text
stringlengths 87
880k
| pmid
stringlengths 1
8
| accession_id
stringlengths 9
10
| license
stringclasses 2
values | last_updated
stringlengths 19
19
| retracted
stringclasses 2
values | citation
stringlengths 22
94
| decoded_as
stringclasses 2
values | journal
stringlengths 3
48
| year
int32 1.95k
2.02k
| doi
stringlengths 3
61
| oa_subset
stringclasses 1
value |
---|---|---|---|---|---|---|---|---|---|---|---|
==== Front
J Neuroengineering RehabilJournal of NeuroEngineering and Rehabilitation1743-0003BioMed Central London 1743-0003-2-141598242510.1186/1743-0003-2-14ResearchPersonal customizing exercise with a wearable measurement and control unit Wang Zhihui [email protected] Tohru [email protected] Naoki [email protected] Graduate School of Science and Technology, Niigata University, 8050 Ikarashi-2nocho, Niigata 950-2181, Japan2005 28 6 2005 2 14 14 7 1 2005 28 6 2005 Copyright © 2005 Wang et al; licensee BioMed Central Ltd.2005Wang et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Recently, wearable technology has been used in various health-related fields to develop advanced monitoring solutions. However, the monitoring function alone cannot meet all the requirements of customizing machine-based exercise on an individual basis by relying on biosignal-based controls. We propose a new wearable unit design equipped with measurement and control functions to support the customization process.
Methods
The wearable unit can measure the heart rate and electromyogram signals during exercise performance and output workload control commands to the exercise machines. The workload is continuously tracked with exercise programs set according to personally customized workload patterns and estimation results from the measured biosignals by a fuzzy control method. Exercise programs are adapted by relying on a computer workstation, which communicates with the wearable unit via wireless connections. A prototype of the wearable unit was tested together with an Internet-based cycle ergometer system to demonstrate that it is possible to customize exercise on an individual basis.
Results
We tested the wearable unit in nine people to assess its suitability to control cycle ergometer exercise. The results confirmed that the unit could successfully control the ergometer workload and continuously support gradual changes in physical activities.
Conclusion
The design of wearable units equipped with measurement and control functions is an important step towards establishing a convenient and continuously supported wellness environment.
wearable unitpersonally customized workload controlinformation technologybiosignalcycle ergometerappropriate exercise level
==== Body
Introduction
In rehabilitation engineering and health promotion, personally customized control of machine-based exercise should be introduced to reflect gradual changes in individual physical work capacity [1]. Biosignal-based workload control systems show great promise as an effective approach to regulate exercise levels [2-4]. Generally, exercise levels are adjusted manually for specific exercise machines, in specific places, typically only by physicians with expertise in sports medicine [5-7]. We have developed an Internet-based cycle ergometer exercise system, aimed at providing a personally customized workload control any time in convenient locations [8,9]. In this system, exercise resources including exercise programs and workload patterns are distributed over the Internet and dynamically integrated on the cycle ergometer. Workload patterns provided by clinicians are computer files defining the time-course of the exercise to meet individual fitness levels and ability. In practical applications, we prepared and set-up measurement equipment, such as computers, amplifiers, and A/D converters, for individual machines. Unlike these conventional systems, significant advances in wearable technology allow us to continuously assess human biometrics more conveniently. Thus, a wearable unit equipped with measurement and control functions can be used on various machines. That is, by setting up one unit, users can perform biosignal-based exercises at a consistent pace, even on a variety of exercise machines. Accordingly, wearable units have the potential to advance the personal customization process, thereby providing a better exercise routine on an individual basis. A lot of attention has been directed to the investigation of health monitoring services, and various types of wearable unit coordinated monitoring function have been studied [10-14]. Still, there are no wearable units suitable for personally customized machine-based exercise. To implement such units, the workload control function must be embedded into the wearable units, and consequently the units can output control signals to the exercise machines to set the appropriate exercise levels.
Because exercise machines used in gyms/health clubs are configured in very different ways, (e.g., some machines have measurement and control functions, while others do not), most users find it very inconvenient to perform exercise in different places. To provide a personal customizing exercise, we need to measure the biosignals and control the workload without any constraints on machines and locations. Therefore, we separated the measurement and control functions from the exercise machines and incorporated these functions into one wearable unit. This allows the personally customized workload control to be implemented at any convenient place. Another disadvantage of traditional exercise machines is that most of them only provide pre-installed exercise programs with limited variations [15]. This is not cost-efficient because upgrading the exercise programs is very complicated and sometimes impossible. In this case, wearable units equipped with measurement and control functions can be used to loosely couple the exercise machines and programs to easily revise and upgrade conventional exercise programs at end users.
We studied biosignal-based workload control, in which the workload can be adjusted using fuzzy inference to continuously adapt the exercise as a function of heart rate and muscle activity [2]. In this paper, we propose a new design of wearable unit for machined-based exercise. To support the personal customization process, we build the measurement and control functions into a single wearable unit. The unit has several different interfaces for measuring multiple biosignals during exercise and then output control commands to exercise machines. To improve convenience, communications between the exercise machines and the wearable unit are by wireless connections. We developed a prototype of this wearable unit for cycle ergometer exercise and used it as part of an Internet-based exercise system. We examined the wearable unit by recruiting nine volunteers over a two-month period. Our results showed that the wearable unit was effective to handle changes in physical activity while controlling the cycle ergometer and was expected to provide continuously supporting appropriate workload patterns for individuals.
Methods
To customize exercise protocols on an individual basis, we need timely updates of workload patterns and continuous workload adjustment, based on the analysis of various biosignals, such as the heart rate (HR) and electromyogram (EMG) signals [1]. Wearable units must offer these measurement and control functions. To enable users to exercise regardless of time and place, the unit must be designed to obtain exercise programs and workload patterns via the Internet and to automatically submit the exercise results.
Wearable Unit Design
Wearable units for machine-based exercise should have interfaces to measure the biosignals. The kind of biosignals required depends on the type of control to be used in exercise programs. We used HR and EMG signals to compute the appropriate exercise levels, according to the idea that gradual changes in physical activity are of interest during an exercise routine. Although exercise programs can be embedded into the wearable unit, they would require a significant amount of the unit's resources, especially if the programs include complicated control methods. Due to the limited processing power and storage capacity available via wearable units, the optimal configuration has wired or wireless communication interfaces to connect to external computers with relatively high performance. If necessary, external computers are utilized for executing exercise programs to provide control parameters. In this case, the wearable unit is a type of middleware, linking the exercise machines to the exercise programs. In addition, like typical designs, the wearable unit needs to have adequate data measurement capacity and transfer speed. Most importantly, the wearable unit should be equipped with an A/D converter and amplifier that operate independently from each exercise machine.
Figure 1 presents our overall design of a wearable unit that meets the requirements of the above design considerations. The low-level control module fixed in the unit is responsible for detecting TCP connections, dealing with temporal biosignal data, and generating control commands according to the specifications of the different exercise machines. Note that the exercise programs can reside either on the wearable unit or on an external computer. The decision about which approach to use depends on the complexity of the exercise programs.
Figure 1 Schematic representation of the design of the wearable unit for machine-based exercise.
Prototype of a Wearable Unit for Cycle Ergometer Exercise
We developed a prototype of the wearable unit to dynamically control the workload during cycle ergometer exercise. It has a Linux (kernel 2.4) operating system, a 66-MHz-CPU, and 2-MB memory capacity. It also has an on-board 12-bit resolution A/D converter, a 60-dB-gain amplifier, a PCMCIA type slot for a wireless LAN card, and an IP address. Additionally, it features 6 channels for biosignal measurements and a sampling frequency of 5 kHz. At the present development stage, infrared wireless communication is used to acquire HR information from, and output workload control commands to, a cycle ergometer.
Our provided exercise program contains a procedure to calculate the appropriate workload by estimating HR and EMG signals, using a set of predefined fuzzy rules and membership functions [2]. The procedure is time-consuming and requires storage space for the measured data (more than 8-MB for each exercise course). The wearable unit cannot work alone to provide the workload because of its low current capacity. Therefore we used external computers to execute the exercise program and compute the workload. Data transmission between the unit and external computers was implemented using TCP socket communication over wired or wireless connections. At the time of workload control, the unit's built-in low-level control module (Fig. 1) created separate threads to communicate with the external computers and cycle ergometer. Hence, the measurement, control, and data transmission processes were performed individually.
Figure 2 shows an acquisition-control sequence diagram of how the wearable unit works with a cycle ergometer and an external computer. Note that at first, the exercise program residing at the external computer opens a TCP connection to the wearable unit. Through this connection, the program acquires and records the HR and EMG signals, measured by the unit. The external computer calculates the workload parameters and sends them to the unit. When receiving the workload parameters, the unit parses them to generate the corresponding workload setting command, and then submits the command to the cycle ergometer. In addition, the exercise program stores all the measured data on the local disk of the external computer for future design of workload patterns. It is worth to emphasize that the exercise program does not reside on cycle ergometers, but rather on external computers. Thus, we can easily upgrade the program without tampering with cycle ergometers.
Figure 2 Acquisition-control sequence diagram for controlling the cycle ergometer through the wearable unit with the help of an external computer.
Applying The Wearable Unit to Internet-Based Exercise Systems
We have developed an Internet-based cycle ergometer exercise system [8,9], which is the backbone of support for the wearable unit, in terms of easy access to various exercise resources at any time from any place. The system provides a central server to process client requests and a history database to store the exercise resources. We have also provided a utility to help clinicians design workload patterns [16]. By coordinating the wearable unit with this system, the practicality and convenience of the personal customization process will improve, because the unit will be able to accommodate various types of cycle ergometers, regardless of whether or not they already have embedded measurement and control functions.
The proposed exercise system (Fig. 3) is composed of a central server and a database server for both the users and physicians with expertise in sports medicine. Clinicians are responsible for designing appropriate workload patterns, based on a review of the database history, and for remotely uploading the patterns. At the user's location, external computers communicate with the central server to download the exercise program and the latest workload pattern designed by clinicians. The downloaded exercise program continuously transmits the workload parameters to the wearable unit via a wireless connection, and then, the unit sets the workload level on the cycle ergometer. The wearable unit gathers HR and EMG and sends this data to the external computer. The exercise program automatically submits all the exercise results to the central server via the Internet after the exercise session is finished.
Figure 3 Layout of Internet-based cycle ergometer exercise system. There is an external computer in the exercise location that communicates with the central server. Clinicians can remotely design and send workload patterns, which will be downloaded by the users at the time of exercise.
Results
We conducted a set of field experiments with the wearable unit over a two-month period in a hypothetical Internet-based environment, using 100-Base-T Ethernet connections, set up in our laboratory. The purpose was to test the system to personally customize workload control while subjects were using a cycle ergometer and physiological data were gathered using the wearable unit. Figure 4 shows an actual exercise session of a subject wearing the unit around his waist. The design utility [16] was installed in advance on a computer operated by a clinician. The experiments were centered on the Microsoft Windows system (Windows 2000). In addition, subjects and clinicians worked in different places.
Figure 4 Photograph of the unit being worn during cycle ergometer exercise.
Seven male and two female young subjects (21.3 ± 1.7 years old) assisted us in carrying out the experiments. They exercised once or twice a week for 30 minutes at a time. The exercise flow was the same as for our previous study on the personal customizing exercise [1]. At first, all subjects took a progressively increasing workload test to evaluate their basic physical work capacity. Then, based on the results of this test, a clinician used the design utility to create customized workload patterns by adjusting the fuzzy rules for each subject. The subjects then downloaded the exercise program and the latest workload pattern from the central server and performed the workload control exercise wearing the unit. The workload control exercises by the subjects and the design of the appropriate workload patterns by the clinician were repeatedly performed after the progressively increasing workload test. It should be noted that we provided a web-based user interface to assist the users in obtaining the exercise programs [17].
Before every exercise session, we downloaded approximately 450-KB of exercise program data as well as 5-KB of workload patterns from the central server to the exercise area. After every session, we uploaded about 8-MB of measured data, including HR and EMG signals, to the central server and stored it in the database.
Figure 5 shows three HR-γARV-MPF scatter graphs, ordered by the exercise date. These represent the changes over a 30-minute time period in a 22-year-old man. A muscular fatigue related index, γARV-MPF, is the correlation coefficient between the averaged rectified value (ARV) and the mean power frequency (MPF) of EMG signals [2], and it became negative as the muscles become fatigued. We also obtained the ratings of perceived exertion (RPE) using Borg's 15-point scale [18] every minute. The RPE is a subjective index widely applied in sports medicine. The exercise levels users found "somewhat hard" are considered efficient based on previous reports. The red squares in each sub-graph represent time slices users found "somewhat hard". There are more samples denoted within the square in (c) (about 30.6%) than there are in (a) (about 10.0%) and (b) (about 18.7%). Therefore, the subject performed more appropriate exercise in Fig. 5 (c). Figure 6 shows the one-to-one time-series graphs for the subject described in Fig. 5. The workload change in (c) was more moderate than it was in (a) and (b). Besides, the maximum workload in (b) and (c) is smaller than in (a). The subject also reported that the workload control pattern shown in Fig. 6 (c), which was designed by reviewing the results of previous exercises, was sufficient to achieve satisfactory exercise. Seven of the nine subjects believed that the workload patterns were challenging at first, but became easier over time. The results of their HRs and EMGs agree with their subjective evaluations. Two male subjects did not obtain satisfying results, but they felt that continuously changing the workload patterns was interesting. The overall results showed that an individualized exercise routine was ensured with the wearable unit in the Internet-based cycle ergometer exercise system.
Figure 5 Change in scatter graph between HR and γARV-MPF for a 22-year-old man during customized exercise session. Exercise (c) is the most effective of the three exercise sessions.
Figure 6 Time-series graphs of different workload patterns for the subject shown in Fig. 5. On the time axis, one frame equals 5 seconds. From top to bottom, workload, heart rate, and γARV-MPF.
Discussion
Wearable Unit for Personally Customized Machine-Based Exercise
Individualized exercise routines are effective for coping with gradual variations in the physical work capacity and for sustaining the motivation to exercise [1]. In machine-based exercise, a practical operation of personal customization is the continuous provision of appropriate workload patterns for users. Thus, when we apply wearable technology to machined-based exercise, the design of the wearable unit must be able to provide the corresponding control function allowing the user to conveniently and easily follow the prescribed workload pattern. However, most wearable unit studies only provide continuous monitoring of various biosignals [10-14], which we believe is insufficient to meet current demands.
We have presented a new wearable unit design equipped with both measurement and control functions for machine-based exercise. The wearable unit gathers measures of the HR and EMG activity and outputs control signals to the exercise machines. Therefore, it is possible to provide appropriate workload control based on individual biosignals. Our results show that a prototype of the wearable unit, combined with an Internet-based exercise system, can achieve personal customization of cycle ergometer exercise. In our experiments, an external computer estimated the appropriate workloads using a biosignal-based fuzzy control method. As a result, the wearable unit formed a link between the user, the exercise machines, and the external computer in which the exercise programs were executed. The wearable unit provided wired and wireless communication interfaces that connected to the external computers. Such designs are very useful if the wearable unit alone cannot perform the computing task in real time. Most importantly, the wearable unit can accommodate various types of cycle ergometers with different specifications, which will greatly improve the convenience of exercising in different places.
The personal customization process has been ensured with the wearable unit. In our experiments, the clinician successfully customized exercise protocols for most of the subjects based on reviewing the subjects' history data. However, two subjects did not perform the anticipated exercises. This had no relationship with the design of the wearable unit, but most likely occurred because our biosignal-based workload control method was not suitable for them. After all, there are great individual differences in terms of functional flexibility and physical work capacity [1]. We require further fundamental studies on providing appropriate exercise levels, based on biosignals. Moreover, cycle ergometer exercise might not be the preferred approach for some subjects. In this case, other types of exercise might be more useful to them.
Information Technology to Support Wearable Units
To continuously support the personally customized workload control without constraints on time and place, the wearable unit must be integrated into an Internet-based support system [1,9,19-21], where the exercise routine or design is provided and the measured data is stored and further processed. By transferring the measured data to a central repository, clinicians can review the exercise history and remotely design appropriate workload patterns at their own convenience. Moreover, complicated computing tasks can be assigned to, and the processed results can be acquired from, external computers over wireless connections.
We showed how a wearable unit could be applied to an Internet-based cycle ergometer exercise system. The wearable unit was able to store small amounts of temporal data, and the completed data was processed in an external computer and then uploaded to the database via the Internet. Additionally, the workload patterns and exercise programs were obtained from a central server. Users could perform the individualized exercise routine at any convenient place. Hence, biosignal-based workload control by a wearable unit and the Internet-based support system is a promising approach for providing appropriate exercise levels that will challenge the user and continuously improve their health.
In fact, if we improve the computing performance of the wearable unit by raising the CPU frequency and the internal memory capacity, the unit will be able to compute exercise levels alone. Accordingly, external computers will become unnecessary for control purpose, thus further improving the convenience of the exercise system. For more flexible designs, a removable storage device, which is now being developed, can be used to increase the storage capacity for exercise data and temporal exercise programs. Such design considerations will be implemented in the next version of the wearable unit.
Range of Application in Health Promotion and Rehabilitation
We described how to apply the wearable unit for an indoor cycle ergometer exercise. The wearable unit could also be effective for outdoor exercises, without requiring any significant changes. We investigated the possibility of using biosignals to control power-assisted bicycles [22]. That study attempted to prevent muscular fatigue during cycling by changing the ratio of rider-generated torque to additional electric-motor-produced torque, based on an evaluation of the measured biosignals. The control process approach is similar to cycle ergometer exercise. Thus, by 1) providing an exercise program that implements the control method, and 2) developing control commands to set the assistance ratio, the wearable unit can also be used to support power-assisted bicycle exercise.
Our wearable unit design for machine-based exercise is suitable for health promotion and rehabilitation. The personal customization process provides an ideal approach and facilitates achievement through the increased motivation of the users, who find convenient not to have to worry about whether or not their exercises are suitable. The workload patterns are remotely designed with the help of clinicians, not by self-assessment of users. Moreover, using Internet-based exercise systems with just one unit, users will be able to perform appropriate exercises on exercise machines that have different specifications. The health promotion and rehabilitation industries are expected to receive favorably control-function-equipped wearable units that can dynamically control the exercise levels, based on measured biosignals.
The wearable unit also reduces the costs of developing and producing exercise machines because the measurement and control functions are separate from the machine. Moreover, loosely coupling exercise machines and exercise programs enables the programs can easily be upgraded without tampering with the hardware, i.e., the exercise machines [23]. The wearable unit helps implement such designs in a more flexible manner, because exercise programs can 1) be installed in the wearable unit to directly control the exercise machines, or 2) reside in an external computer used to communicate with the wearable unit to remotely transfer control signals. Moreover, by taking advantage of the wearable unit, the requirements of exercise machines for the personally customized workload control decrease for practical use, and as a result, the possibility of finding a suitable exercise machine without location constraints would increase.
Conclusion
We embedded measurement and control functions into a single wearable unit to personal customizing machine-based exercise. Moreover, we introduced the Internet technology to support the personal customization process without time and place constraints. A wearable unit capable of outputting control signals provides the appropriate exercise levels, based on exercise programs and measured biosignals. Users wearing this unit can take advantage of various exercise programs using a variety of exercise machines. A prototype of the wearable unit measured heart rate and EMG signals and wirelessly transmitted the control commands. By applying this unit to an Internet-based exercise system, we were able to personally customize cycle ergometer exercise. The design of our wearable unit is a progressive step towards establishing a convenient and continuously supported wellness environment. In the future, we will be able to apply these units to outdoor exercises and rehabilitation.
==== Refs
Kiryu T Sasaki I Shibai K Tanaka K Providing appropriate exercise levels for the elderly IEEE Eng Med Biol Mag 2001 20 116 124 11838243
Kiryu T Takahashi K Ogawa K Multivariate analysis of muscular fatigue during biycle ergometer exercise IEEE Trans Biomed Eng 1997 44 665 672 9254980
Glass SC Knowlton RG Sanjabi PB Sullivan JJ Identifying the integrated electromyographic threshold using different muscles during incremental cycling exercise J Sports Med Phys Fitness 1998 38 47 52 9638032
Mateika J Duffin J The ventilation, lactate and electromyographic thresholds during incremental exercise tests in normoxia, hypoxia and hyperoxia Eur J Applied Physiol 1994 69 110 118
Thompson WR Benardot D Jonas S ACSM fitness book 2003 3 Champaign: Human Kinetics
Williford HN Barfield BR Lazenby RB Olson MS A survey of physicians' attitudes and practices related to exercise promotion Prev Med 1992 21 630 636 1438111
McKenna J Naylor PJ McDowell N Barriers to physical activity promotion by general practitioners and practice nurses Br J Sports Med 1998 32 242 247 9773175
Kiryu T Yamaguchi K Tanaka K Shionoya A Internet based system for adjusting cycle ergometer workload to moderate exercise Proc 21st Annu Int Conf IEEE/EMBS 1999 Atlanta, GA 615
Wang Z Shibai K Kiryu T An Internet-based cycle ergometer by using distributed computing Proc 4th Annu IEEE Conf on ITAB 2003 Birmingham, UK 82 85
Jovanov E Lords AO Raskovic D Cox PG Adhami R Andrasik F Stress monitoring using a distributed wireless intelligent sensor system IEEE Eng Med Biol Mag 2003 22 49 55 12845819
Korhonen I Parkka J Gils MV Health monitoring in the home of the future IEEE Eng Med Biol Mag 2003 22 66 73 12845821
Matsushita S Oba T Otsuki K Toji M Otsuki J Ogawa K A wearable sense of balance monitoring system towards daily health care monitoring Proc 7th IEEE Int Symp Wearable Computers (ISWC) 2003 New York 176 83
Pentland A Healthwear: Medical technology becomes wearable IEEE Computer 2004 37 42 49
Anliker U Ward JA Lukowicz P Tröster G Dolveck F Baer M Keita F Schenker E Catarsi F Coluccini L Belardinelli A Shklarski D Alon M Hirt E Schmid R Vuskovic M AMON: A wearable multiparameter medical monitoring and alert system IEEE Trans Inform Technol Biomed 2004 8 415 427
Stamford BA Choosing and using exercise equipment Phys Sportsmed 1997 25 107 108
Wang Z Kiryu T Development of evaluation utilities for the Internet-based wellness cycle ergometer system Proc IEEE EMBS Asian-Pacific Conf on Biomed Eng 2003 Keihanna, Japan 019265-1.pdf
Wang Z Kiryu T Design of a web-based health promotion system and its practical implementation for cycle ergometer exercise Proc 26th Annu Int Conf IEEE/EMBS 2004 San Francisco, CA 3330 3333
Borg G Ljunggren G Ceci R The increase of perceived exertion, aches and pain in the legs, heart rate and blood lactate during exercise on a bicycle ergometer Eur J Appl Physiol 1985 54 343 349
Siau K Health care informatics IEEE Trans Inform Technol Biomed 2003 7 1 7
Ammenwerth E Gräber S Herrmann G Bürkle T König J Evaluation of health information systems-problems and challenges Int J Med Inf 2003 71 125 135
Blair SN Franklin BA Jakicic JM Kibler B New vision for health promotion within sports medicine Am J Health Promot 2003 18 182 185 14621416
Kiryu T Irishima K Moriya T Mizuno Y Changes in functional activity with prediction during cycling exercise Proc Congress of the International Society of Electrophysiology and Kinesiology 2002 Vienna, Austria 197 198
Wang Z Kiryu T Iwaki M Shibai K An Internet-based cycle ergometer health promotion system for providing personally fitted exercise IEICE Trans Inf Syst
|
15982425
|
PMC1183241
|
CC BY
|
2021-01-04 16:37:40
|
no
|
J Neuroengineering Rehabil. 2005 Jun 28; 2:14
|
utf-8
|
J Neuroeng Rehabil
| 2,005 |
10.1186/1743-0003-2-14
|
oa_comm
|
==== Front
Kinetoplastid Biol DisKinetoplastid Biology and Disease1475-9292BioMed Central London 1475-9292-4-41595524810.1186/1475-9292-4-4Original ResearchApplication of Direct Agglutination Test (DAT) and Fast Agglutination Screening Test (FAST) for sero-diagnosis of visceral leishmaniasis in endemic area of Minas Gerais, Brazil Silva Eduardo S [email protected] Gerard J [email protected] Celia MF [email protected] Reginaldo P [email protected] Raquel S [email protected] Henk DFH [email protected] Universidade do Estado de Minas Gerais, Fundação Educacional de Divinópolis, Divinópolis, MG, Brasil2 KIT (Koninklijk Instituut voor de Tropen / Royal Tropical Institute) KIT Biomedical Research, Meibergdreef 39, 1105 AZ Amsterdam, The Netherlands3 Centro de Pesquisas René Rachou-Fiocruz, Belo Horizonte, MG, Brasil4 Departamento de Bioquímica e Biologia Molecular, IOC/ Fiocruz, Rio de Janeiro, RJ, Brasil2005 14 6 2005 4 4 4 23 12 2004 14 6 2005 Copyright © 2005 Silva et al; licensee BioMed Central Ltd.2005Silva et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The direct agglutination test (DAT) has proved to be a very important sero-diagnostic tool combining high levels of intrinsic validity and ease of performance. Otherwise, fast agglutination screening test (FAST) utilises only one serum dilution making the test very suitable for the screening of large populations.
Results
We have tested FAST and DAT for the detection anti-Leishmania antibodies in serum samples from patients with American visceral (AVL) and cutaneous leishmaniases (ACL) in Minas Gerais State, Brazil. The DAT on serum and blood samples of confirmed AVL patients found all samples positive at a serum dilution of ≥ 1:800. This dilution was subsequently used as cut off value in the study. The blood and serum samples of these confirmed patients could also be clearly read in FAST using a 1:100 dilution with the same high sensitivity. DAT and FAST were not able to detect significant amounts of antibodies in samples from ACL patients and are not suitable for the diagnosis of this manifestation of the disease.
Conclusion
We suggest that both DAT and FAST are very practical diagnostic tools for the sero-diagnosis of AVL under rural conditions as both serological tests do not require sophisticated equipment, a cold chain and are very simple to perform.
==== Body
Background
American visceral leishmaniasis (AVL) a protozoan disease caused by Leishmania chagasi parasites, constitutes a major health problem in Brazil. In the last few years the number of human cases of AVL in the metropolitan region of Belo Horizonte (MRBH), state of Minas Gerais, Brazil has increased, indicating an elevation in the transmission rate of the disease [1]. Dogs and fox are considered to be the main reservoirs host for this parasite [2,3]. The diagnosis of AVL is based on clinical-epidemiological characteristics, by microscopical demonstration of the parasite in biopsies of aspirates, indirectly by serological tests and culturing or molecular methods like the polymerase chain reaction (PCR) [4-6]. Several techniques can be used for the sero-diagnosis of AVL. The indirect fluorescence technique (IFAT) was the technique of choice until 1974 [7,8]. Since then counter-current immuno-electrophoresis and enzyme-linked immunosorbent assay (ELISA) have been found to be powerful tools for the sero-diagnosis of leishmaniasis [9]. In addition, several other serological tests have been developed. The direct agglutination test (DAT) has proved to be a very important sero-diagnostic tool combining high levels of intrinsic validity and ease of performance [10-12]. The test uses whole, stained promastigotes either as a suspension or in a freeze-dried form [10,11,13,14]. By using the freeze-dried antigen, logistic problems, such as the need of a cold chain for storage of antigen, are avoided, making the DAT very suitable for use under field conditions.
Although the direct agglutination test (DAT) for the sero-diagnosis of visceral leishmaniasis has a high sensitivity and specificity [6,13,14], it still has some limitations like the relative long incubation time (18 h) and the need for serial dilutions of blood or serum. In order to circumvent these problems Schoone et al. [15] developed a fast agglutination-screening test (FAST) for the rapid detection of anti-Leishmania antibodies in serum samples and in blood collected on filter paper. The FAST utilises only one serum dilution (qualitative result) and requires 3 hours of incubation. This makes the test very suitable for the screening of large populations.
The increasing importance of AVL in man and its high rates of lethality in the Metropolitan Region of Belo Horizonte (MRBH) indicate that a rapid and relatively simple method is needed for the routine diagnosis of this disease. The objective of this study was to evaluate DAT and FAST as potential AVL diagnostic methods using clinical samples from this region in Brazil.
Materials and methods
Serum samples
This study was carried out utilising serum samples from different patients groups from metropolitan region of Belo Horizonte (MRBH), Minas Gerais State and some control samples from other regions (see below). Samples and information of patients (age, sex, symptomatology, clinical, address) were sent to the laboratory through the existing health care system of MRBH, Minas Gerais, Brazil. Written consent was obtained from the patient or their relative for this study.
The following groups of samples were included in the present study:
1. Serum samples of AVL patients parasitologically confirmed (microscopical examination from bone marrow aspirate smears) (n = 16).
2. Serum samples of patients clinically suspected of AVL (fever, spleen enlargement, pallor, weight loss), but not parasitologically confirmed (n = 99).
3. Serum samples of patients presenting with active cutaneous leishmaniasis lesions (n = 85).
4. Serum samples of patients with confirmed Chagas' disease (n = 12)
5. Serum samples of patients (Uganda) with confirmed African trypanosomiasis (n = 5)
6. Serum samples of patients (Kenya) with confirmed P. falciparum malaria (n = 5)
7. Serum samples of confirmed toxoplasmosis patients (The Netherlands) (n = 5)
8. Serum samples of apparently healthy individuals from an AVL endemic region in Brazil (n = 19)
9. Serum samples of healthy blood donors from a non-endemic region (The Netherlands; n = 5)
Serology
The presence of Leishmania antibodies in all the serum samples was determined by FAST and DAT. A sub-set of the samples (groups 1 – 3) were also analysed with IFAT. Antigen for FAST and DAT were prepared as described earlier [13,15]. The FAST was performed according to the protocol described previously [15]. In brief, serum samples were diluted 1:100 in physiological saline (NaCl 0.85%) to which 0.78% β – mercaptoethanol was added in a V-shaped microtitre plate (Greiner, Germany). Next, 20 μl of this 1:100 dilution was transferred to another well of the V-shaped microtitre plate and 20 μl FAST antigen (2 × 108 promastigotes/ml) was added. The plate was carefully shaken, covered with a lid and allowed to incubate for 3 hours at room temperature after which the results were read. Appropriate positive and negative controls were always included on each plate.
The DAT was performed essentially as previously described by [10,13]. In brief, the samples were diluted in physiological saline (0.9% NaCl) containing 0.78% β – mercaptoethanol. Two-fold dilution series of the sera were made in a V-shaped microtitre plate, starting at a dilution of 1:100 (step 1) and going up to a maximum serum dilution of 1:102.400 (step 11). Well 12 was used as a negative control. Fifty μl DAT antigen (concentration of 5 × 107 parasites per ml) was added to each well containing 50 μl diluted serum and the results were read after 18 hours of incubation. The cut-off value of the DAT was set at >1:800.
The IFAT was, due to logistic problems, only performed on a selection of the serum samples using a commercial kit for the diagnosis of human leishmaniasis (Fiocruz/Bio-Manguinhos, Rio de Janeiro, Brazil). IFAT was performed according to the instructions of the manufacturer for detection of antibodies in serum diluted from 1:40 up to 1:640.
Statistical analysis
The sensitivity and specificity of the DAT and FAST in the present study were calculated as follows: Sensitivity = TP/(TP+FN) × 100% and Specificity = TN/(TN+FP) × 100%. Where TN represents true negative, TP true positive, FN false negative and FP false positive. The sensitivity of the two tests, FAST and DAT, was assessed with sera from confirmed AVL patients (n = 16). Sera of healthy controls (n = 24) and sera of patients with confirmed other diseases (n = 112) were used to determine the specificity of DAT and FAST.
The degree of agreement between FAST, DAT and IFAT was determined by calculating Kappa (κ) values with 95% confidence intervals using Epi-info version 6. Kappa values express the agreement beyond change and a κ value of 0.21 – 0.60 represents a fair to moderate agreement, a κ value of 0.60 – 0.80 represents a substantial agreement and a κ>0.80 represents almost perfect agreement beyond change [16]. The calculation of the degree of agreement between DAT and FAST was based on all serum samples, whereas the κ values for DAT -IFAT and FAST – IFAT were only based on the results obtained with the confirmed and suspected AVL serum samples.
Results
The results of the serological analysis with DAT and FAST are presented in Tables 1 and 2, respectively. The sensitivity of the DAT in the present study was calculated to be 100% and its specificity 97.8% (3 false positive results). The FAST had a sensitivity of 100% and a specificity of 92.5% (11 false positive results). The results of the IFAT testing are presented in Table 3. The sensitivity and specificity of the IFAT could not be adequately calculated as a sufficient number of negative samples was not analysed with this test. It should be noted that IFAT tested 1 confirmed AVL patient negative and 7 patients only had a low IFAT titre. In contrast, IFAT found about 50% of all confirmed ACL patients sero-positive, whereas neither DAT nor FAST was able to detect antibodies in these serum samples. The DAT found 77 out of 99 samples of patients suspected (but not parasitologically confirmed) of AVL positive. It was observed that 79/99 patients were FAST positive and 84/99 patients were IFAT positive with serum dilutions varying from 1:40 to 1:640.
Table 1 Direct agglutination test (DAT) results for anti-Leishmania antibodies in various patients groups.
Serial dilution series
Patient group ≤ 800 1600 3200 6400 12800 25600 51200 102.400 Total (n)
1 confirmed AVL - - 1 - 4 1 3 7 16
2 suspected AVL 22 - 3 9 8 15 5 37 99
3 confirmed CL 84 - 1 - - - - - 85
4 confirmed Chagas' disease 10 2 - - - - - - 12
5 confirmed trypanosomiasis 5 - - - - - - - 5
6 confirmed malaria 5 - - - - - - - 5
7 confirmed toxoplasmosis 5 - - - - - - - 5
8 healthy endemic controls 19 - - - - - - - 19
9 healthy non-endemic controls 5 - - - - - - - 5
Table 2 Fast agglutination screening test (FAST) results for anti-Leishmania antibodies in various patients groups.
Patient group FAST negative Fast Positive Total (n)
1 – Confirmed AVL 0 16 16
2 – Suspected AVL 20 79 99
3 – Confirmed CL 82 3 85
4 – Confirmed Chagas' disease 6 6 12
5 – Confirmed trypanosomiasis 4 1 5
6 – Confirmed malaria 5 0 5
7 – Confirmed toxoplasmosis 5 0 5
8 – Healthy endemic controls 18 1 19
9 – Healthy non-endemic controls 5 0 5
Table 3 Indirect Immunofluoresence Test (IFAT) results for anti-Leishmania antibodies in serum of sub-set of the different patients groups (NR = no reaction).
Patient group NR 1:40 1:80 1:160 1:320 1:640 Positive/Total (n)
1 – Confirmed AVL 1 1 6 - 6 2 15/16
2 – Suspected AVL 15 7 20 20 23 14 84/99
3 – Confirmed ACL 37 3 23 12 9 1 48/85
A high degree of agreement (96%) was observed between FAST and DAT (Table 4a). The agreement beyond change (κ value) was 0.92. In addition, substantial agreement was observed between DAT and IFAT (93%; Table Table 4b) or FAST and IFAT (94%; Table Table 4c), with κ values of 0.75 and 0.80, respectively.
Table 4 Comparison between DAT, FAST and IFAT using serum samples from suspected visceral leishmaniasis patients from the Belo Horizonte Metropolitan Region.
A. Comparison between DAT and FAST
FAST + FAST - Total
DAT + 96 0 96
DAT - 10 145 155
Total 106 145 251
Comparison between DAT and IFAT
IFAT + IFAT - Total
DAT + 92 1 93
DAT - 7 15 22
Total 99 16 115
C. Comparison between FAST and IFAT
IFAT + IFAT - Total
FAST + 94 1 95
FAST - 5 15 20
Total 99 16 115
Discussion
In view of the public health importance of AVL and the inherent difficulties of conventional diagnosis techniques, we evaluated in the present study the performance of the sero-diagnostic tests, DAT and FAST. Both test displayed a very high sensitivity and specificity corroborating with previous studies [10,13-15]. It is noted that the sensitivity of DAT and FAST observed in the present study was determined using a relatively low number of confirmed patients, and therefore 100% sensitivity is not claimed.
The antigen on which DAT and FAST are based is a strain of L. donovani, whereas human and canine visceral leishmaniasis in Brazil is caused by Leishmania chagasi, both species belonging to the L. donovani complex. Apparently the use of an heterologous antigen did not affect the performance of both tests for the detection of anti-Leishmania antibodies in Brazilian AVL patients. The DAT found all parasitologically confirmed AVL patients positive at a serum dilution of ≥ 1:800, which was subsequently used as the cut off dilution in the present study. This serum dilution is comparable to the cut off dilutions found in several other studies [13,14]. The FAST also found all confirmed cases positive, which should be the case as this test is intended as a screening test that should not miss any AVL patient. This result is even better than a previous evaluation of the FAST in which some cases were missed [15].
We have also compared the performance of IFAT with DAT and FAST on serum samples of confirmed AVL cases and suspects. The IFAT missed one confirmed patient that had a very high DAT serum dilution. On the other hand, IFAT found slightly more suspected ALV patients positive (84/99) than DAT (77/99) or FAST (79/99) did. However, there was in general a very good agreement between the performance of the three tests with regard to the seo-diagnosis of AVL. In contrast, both FAST and DAT found only a very limited number of ACL cases sero-positive. IFAT found approximately 50% of these cases positive, albeit with generally very low serum titres. DAT and FAST based on L. donovani antigen are not suitable for the sero-diagnosis of ACL [13].
Conclusion
As final remarks, we can conclude that DAT and FAST are very suitable tools for the sero-diagnosis of AVL. Both tests are easy to interpret, as well as being specific and sensitive. The DAT is very practical under field or rural conditions, as no specialised equipment is required nor a cold chain is necessary for the storage of antigen. In addition, the FAST requires only one serum dilution and the results can be read within 3 hours. The FAST can be used to screen large populations, for example in situations such as epidemics where large number of suspects are seen at the clinic or cases where immediate treatment is necessary.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
Silva ES performed the practical work as part of his PhD study, partly in Brazil and partly in the Netherlands, wrote the concept of the paper. Schoone GJ research technician performed part of the sample and data analysis, produced DAT and FAST antigen and invented the FAST test. Gonijo CMF took part in the study design, supervised practical work in Brazil and assisted in writing the concept of the paper. Pacheco RS and Brazil RP former supervisors of Silva ES took part in study design. Schallig HDFH supervised the practical work in the Netherlands, data analysis, final preparation of manuscript, corresponding author
Acknowledgements
This study was supported by a grant from the Hubrecht-Janssen Foundation (Amsterdam, The Netherlands). This investigation received financial support from the UNDP/World Bank/WHO Special Programme for Research and Training in Tropical Diseases (TDR), grant 10245 from Director's Initiative Fund.
==== Refs
Silva ES Gontijo CMF Pacheco RS Fiuza VO Brazil RP Visceral leishmaniasis in the Metropolitan Region of Belo Horizonte, State of Minas Gerais, Brazil Mem Inst Oswaldo Cruz 2001 96 285 291 11313633
Deane LM Deane MP Visceral leishmaniasis in Brazil: geographical distribution and transmission Rev Inst Med Trop São Paulo 1962 4 198 212
Silva ES Pirmez C Gontijo CMF Fernandes O Brazil RP Visceral leishmaniasis in the crab-eating fox (Cerdocyon thous) in south-east Brazil The Vet Rec 2000 147 421 422 11072988
Mathis A Deplazes P PCR, in vitro cultivation for detection of Leishmania spp. In diagnostic samples from human and dogs J Clin Microbiol 1995 33 1145 1149 7615719
Silva ES Pacheco RS Gontijo CMF Carvalho IR Brazil RP Visceral leishmaniasis caused by Leishmania (Viannia) braziliensis in a patient infected with human immunodeficiency virus Rev do Inst Med Trop São Paulo 2002 44 145 149
Schallig HDFH Oskam L Molecular biological applications in the diagnosis and control of leishmaniasis and parasite identification Trop Med Intern Health 2002 7 641 651 10.1046/j.1365-3156.2002.00911.x
Camargo ME Rebonato C Cross-reactivity in immunofluorescense for Trypanosoma and Leishmania antibodies Am J Trop Med Hyg 1969 18 500 505 4978596
Zuckerman A Current status of the immunology of blood and tissue Protozoa. I. Leishmania Exp Parasitol 1975 38 370 400 1107055 10.1016/0014-4894(75)90123-X
Mukerji K Roy S Mukhopadhyay P Gupta PK Ghosh DK Evaluation of different subcellular fractions of Leishmania donovani for immunodiagnosis of visceral leishmaniasis Indian J Exp Biol 1984 22 120 122 6440858
Harith AE Kolk AHJ Leuwenburg J Muigai R Huifgai E Jelsma T Kager A Improvement of a direct agglutination test for field studies of visceral leishmaniasis J Clin Microbiol 1988 26 1321 1325 3410946
Zijlstra EE Osman OF Hofland HWC Oskan L Ghalib HW El-Hassan AM Kager PA Meredith SEO The direct agglutination test for diagnosis of visceral leishmaniasis under field conditions in Sudan: comparison of aqueous and freeze-dried antigens Trans R Soc Trop Med Hyg 1997 91 671 673 9509176 10.1016/S0035-9203(97)90518-6
Boelaert M El Safi S Jacquet D de Muynck A van der Stuyft P Le Ray D Operational validation of the Direct Agglutination Test for diagnosis of visceral leishmaniasis Am J Trop Med Hyg 1999 60 129 134 9988336
Meredith SEO Kroon NCM Sondorp E Seaman J Goris MGA van Ingen CW Oosting H Schoone GJ Terpstra WJ Oskam L Leish Kit, a stable direct agglutination test based on freeze-dried antigen for the serodiagnosis of visceral leishmaniasis J Clin Microbiol 1995 33 1742 1745 7665640
Oskam L Nieuwenhuys JL Hailu A Evaluation of the direct agglutination test (DAT) using freeze dried antigen for the detection of anti-Leishmania antibodies in stored sera from various patients groups in Ethiopia Trans R Soc Trop Med Hyg 1999 93 275 277 10492758 10.1016/S0035-9203(99)90021-4
Schoone GJ Hailu A Kroon CCM Nieuwenhuys JL Schallig HDFH Oskam L A fast agglutination test (FAST) for the detection of anti-Leishmania antibodies Trans R Soc Trop Med Hyg 2001 95 400 401 11579883 10.1016/S0035-9203(01)90196-8
Altman DG Practical Statistics for Medical Research 2001 Chapman & Hall, London, U.K.
|
15955248
|
PMC1183242
|
CC BY
|
2021-01-04 16:38:32
|
no
|
Kinetoplastid Biol Dis. 2005 Jun 14; 4:4
|
utf-8
|
Kinetoplastid Biol Dis
| 2,005 |
10.1186/1475-9292-4-4
|
oa_comm
|
==== Front
Microb Cell FactMicrobial Cell Factories1475-2859BioMed Central London 1475-2859-4-201598969110.1186/1475-2859-4-20ReviewComparative modelling of protein structure and its impact on microbial cell factories Centeno Nuria B [email protected] Joan [email protected] Baldomero [email protected] Structural Bioinformatics Laboratory, Research Group on Biomedical Informatics (GRIB), IMIM/UPF. c/ Dr. Aiguader 80. 08003 Barcelona, Spain2005 30 6 2005 4 20 20 3 5 2005 30 6 2005 Copyright © 2005 Centeno et al; licensee BioMed Central Ltd.2005Centeno et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Comparative modeling is becoming an increasingly helpful technique in microbial cell factories as the knowledge of the three-dimensional structure of a protein would be an invaluable aid to solve problems on protein production. For this reason, an introduction to comparative modeling is presented, with special emphasis on the basic concepts, opportunities and challenges of protein structure prediction. This review is intended to serve as a guide for the biologist who has no special expertise and who is not involved in the determination of protein structure. Selected applications of comparative modeling in microbial cell factories are outlined, and the role of microbial cell factories in the structural genomics initiative is discussed.
==== Body
Review
Introduction
On the last two decades the development of recombinant DNA techniques has extended the use of microbial organisms to produce target proteins. The enteric bacterium Escherichia coli is one of the most extensively used prokaryotic organisms for genetic manipulations and for industrial production of proteins of therapeutic or commercial interest [1,2]. However, bacterial organisms often fail to produce target proteins due to problems related with protein misfolding and protein glycosilation. Yeast and fungal protein expression systems are used for the industrial production of relevant enzymes in such cases [3].
There are two main interests in the industrial production of proteins: i) Redefining the optimal properties of the target protein and ii) Avoiding problems of high-scale production. Knowledge of three-dimensional structure of the proteins may be helpful to redesign a modified protein. Computational prediction methods play an essential role to provide us with structural information of a sequence whose structure has not been experimentally determined. Homology based or comparative modeling [4] is the most detailed and accurate of all current protein structure prediction techniques [5]. Its aim is to build a three-dimensional model for a protein of unknown structure on the basis of sequence similarity to proteins of known structure [6]. Comparative modeling relies on the fact that structure is more conserved than sequence during evolution. Therefore, similar sequences exhibit nearly identical structures, and even distantly related sequences share the same fold [7,8]. Comparative modeling critically depends on the knowledge of three-dimensional structure of homologous proteins. The progress of structural genomics initiatives [9] allow to model a large amount of protein sequences. Besides, the number of unique structural folds that proteins adopt is limited {Zhang, 1997; #81; Liu, 2004 #48}. Consequently, it is likely that at least one example of most structural folds will be known, making comparative modeling applicable to most protein sequences. In term, an essential step of structural genomics is production of target proteins. Microbial cell factories play a key role in this context.
This review is intended to give a primer addressed to scientists of disciplines related to microbial cell factories who has no expertise in comparative modeling. Our goal is to provide the seeding background to understand concepts, opportunities and challenges of comparative modeling. We will describe each step in the comparative modeling process, discuss the most common errors and how to solve them, as well as outlining the applications of comparative modeling in the field of microbial cell factories.
We will emphasize the simplest and most reliable methodologies to follow up along with their range of application with a reduced number of useful programs and web servers. Many other authors have also written excellent reviews on the comparative modeling field [6,10-14].
Steps in comparative modeling
All current comparative modeling methods consist of four sequential steps: template selection, target-template alignment, model building and model evaluation. Essentially, this is an iterative procedure until a satisfactory model is obtained (Figure 1). In this process a variety of programs and web servers can be used (Table 1). Additionally, protein modeling meta-servers are emerging. They automatically implement the full process in a multi-step protocol, using simultaneously different methods [15].
Table 1 Useful servers and programs for protein comparive modeling.
PROGRAM Server/Web adress Reference
Template Selection
PSI-BLAST [20]
HMMER (HMM search) [35]
TOPITS [17]
FUGUE [26]
Threader [27]
3D-PSSM [28]
PFAM [25]
PHYLIP [31]
Target-Template alignment
CLUSTALW [34]
HMMER (HMM align) [35]
STAMP [36]
CE [37]
DSSP [38]
Model Building
COMPOSER [39]
SwissModel [41]
3D-JIGSAW [44]
MODELLER [46]
Loop Modeling
MODLOOP [50]
ARCHDB [51]
Sloop [52]
Sidechain Modeling
WHAT IF [55]
SCWRL [56]
Evaluation of the model
PROCHECK [67]
PROSA II [70]
Biotech
Refinement
GROMOS [74]
CHARMM [75]
AMBER [76]
Figure 1 Flowchart of methods used for comparative modeling. Scheme of the methods used for comparative modeling, comprising template(s) selection, template-target alignment, model (backbone and loops) building, sidechain modeling, model evaluation, and model refinement steps. Programs and servers referring to these steps are listed in table 1.
Template selection
The starting point in comparative modelling is to identify protein structures related to the target sequence and then to select those that will be used as templates. Such templates may be found by sequence comparison methods or by sequence-structure methods also known as threading methods. Sequence comparison methods can be safely used above a certain threshold in terms of sequence identity (i.e. percentage of identical paired residues in an alignment). It has been shown that above that threshold -which is strongly dependent on sequence length, sequence homology implies structural identity [16]. Even though, below that threshold structural likeness is still possible. Some protein pairs sharing very little sequence similarity may have become similar by convergent or divergent evolution. The alignment of these proteins pairs, which define the so called "midnight zone" in sequence alignment [17], is usually addressed with threading methods. Finally, there is a range in terms of sequence identity amid the safe zone and the "midnight zone" in which the relationship between structural homology and the phylogeny is unclear: the "twilight zone" [18]. Within this range, which is usually defined between 20 and 35% sequence identity [19], additional caution must be taken on the sequence alignment.
PSI-BLAST [20], an iterative sequence comparison method, is probably the most widely used program to detect remote similarities. In more difficult scenarios, where sequence homology is not so evident, templates can be found by searching in sequence space using intermediate sequence search (ISS) methods [21-23].
Sequence comparisons can also be made through Hidden Markov Models (HMMs) [24] as implemented, for instance, in HMMER. HMMs profiles of protein domain families are available in Pfam database. These profiles can be used to automatically identify protein domain(s) within the target even if it shares weak sequence similarity with templates [25].
Threading methods have been developed to find more distant relationships. For this reason, they are the most promising choice in the absence of homologues to the target sequence. Threading methods involve performing sensitive sequence searches and characterizing sequence compatibility with the structural environments of putative templates. Features analysed by this kind of methods include secondary structure and solvent accessibility predictions as well as functional annotation. Most used methods of this kind are TOPITS [17], FUGUE [26], Threader [27] and 3D-PSSM [28]. Recent examples of the combined use of these servers and further modeling [29] prove its use. Other methods are being developed based on the analysis of protein-protein interactions to search for remote similarities [30].
Once a list of related proteins with known structure has been obtained, it is necessary to select those templates that are appropriate for the given modelling problem. The feasibility of a template can be assessed by means of its expectation value, E-value [20], which is one of the parameters in the searches outputs. As a general rule, the lower the E-value, the better the template is.
Besides, several other factors should be considered when selecting a template:
1) Quality of the experimental template structure. Because errors in templates will be passed onto the models, the better templates are the most accurate structures available. Accuracy of the templates can be assessed by the resolution and the R-factor for a crystallographic structure, or by the number of distance restraints per residue in the case of NMR structure.
2) Environment likeness. Experimental factors of interest for the target (i.e. the presence of a ligand in the structure, pH, and solvent features ...) should be found as similar as possible in the chosen templates.
3) Phylogenetic similarity. It is helpful to build a multiple alignment and a phylogenetic tree [31] of the target and templates, in order to select templates from the subfamily that is closest to the target sequence. The phylogenetic tree can be constructed by means of PHYLIP set of programs [31].
Depending on the purpose of the model, some of the factors listed above will be more important than others. For instance, resolution of the template is probably the most important factor if the reason for building the model is to design mutants of a binding site, since an accurate geometrical description is needed.
It is important to emphasize that it is not mandatory to select only one template. Actually, methods using multiple templates seems to perform better than those based on a single template [14,32], especially if the main modes extracted from them are taken into account [33]. Finally, it is noteworthy to be aware that, implicitly, choosing templates means the recognition of the target's overall fold.
Target-template alignment
Once templates have been selected, an optimal alignment between the target sequence and templates is needed to further construct a three dimensional model of the target.
From easiest to more complex, some strategies for aligning target and templates are:
1) Obtaining a multiple alignment of the templates and the target using CLUSTALW [34].
2) Aligning the query sequence to a HMM profile of the templates family built from a Pfam alignment [25] using HMMER [35].
3) Aligning the query sequence to a HMM profile of the templates built from a structural alignment using HMMER [35]. Structural alignments required for this strategy can be obtained using STAMP [36] or CE [37]; an automatic web server is available for the later one.
In our experience this third strategy, in which not only the sequence similarity but also the structural information inherent in the templates guide the alignment, make it more trustworthy.
The obtained alignments must be critically evaluated in terms of the number, length, and position of the gaps opened. Some of them can be manually refined, taking into account the secondary structure of the templates and their accessible surface, both of them calculated with DSSP program [38], in order to avoid gaps which are opened within secondary structural elements. This will be important if the alignment strategy is based only on sequence similarity. In any case, at this step, if necessary, the selection of templates may be revisited, either to search a new template to overcome a gap in a particular region or to remove redundant or inadequate templates.
Model building
Comparative model building generates an all-atom model of the sequence based on its alignment to one or more templates. It includes either sequential or simultaneous modelling of the core of the protein, loops, and side-chains.
The original comparative approach, which is still widely used, is modelling by rigid-body assembly [4]. This method constructs the model from a few regions which are obtained from dissecting related structures. In order to assemble the dissected parts, a framework is calculated by averaging of Cα atoms of structurally conserved regions in template structures. Structurally variable regions are modeled by choosing from a database of all known proteins those regions that better fit the anchor conserved regions. COMPOSER [39] is one of the programs that use this methodology in a semiautomated procedure. SwissModel is a commonly used automated web server also based in this approach [40-42].
Modeling by segment matching is another approach which relies on the approximate positions of conserved atoms in templates [27,43]. This is accomplished by breaking the target into a set of short segments, and searching in a database for matching fragments which are fitted onto an initial framework of the target structure. Database searching is based on sequence similarity, conformational similarity and compatibility with the target structure. 3D-JIGSAW is one of the successful programs that uses this approach [44]
Another approach is modeling by satisfaction of the spatial restrains obtained from the alignment [45]. Probably the most used program based on this approach is MODELLER [46]. First, this automated procedure derives many distance and dihedral angle restraints on the target sequence from its alignment with template three-dimensional structures. Next, this homology-derived restraints and energy terms ensuring proper stereochemistry are combined into a function. Finally, the model is obtained by optimizing this function in such a way that the model violates the input restraints as little as possible. Several slightly different models in agreement with the restraints can be calculated.
Any of the three methods above described produce models of similar accuracy if they are optimally applied. In the difficult cases, modeling by satisfaction of spatial restraints is perhaps the most accurate technique, since it can use many different types of information about the target sequence. In this way, available experimental data can be added as new restraints, making the model more reliable.
Loop modeling
Along with alignment, loop modeling is probably the most difficult step in comparative modelling process. Errors in loops are the dominant problem in comparative modelling when target and template share above 35% sequence identity. This is a very active area of research and it is not practical to consider all available methods (details of some of them can be found in [47-49]). In this review we will present the state-of-the art of methods that can be easily used. Furthermore, it must be pointed out that although existing methods can provide reasonably accurate models of short loop regions; modeling of long loops is still an unsolved problem [11]. Loop modeling methods can be classified in two approaches: ab initio methods and database searching or knowledge-based methods.
Ab initio loop prediction is based on a conformational search guided by a scoring or energy function -the later describing the physico-chemical properties of a protein and its environment. There are many such methods, making use of different protein representations, energy functions and optimisation procedures [11]. Among them, there is an option to use implemented in MODELLER or in a web server MODLOOP [50].
The database approach to loop prediction begins by finding segments of main chain that fit the two stems of a loop. The stems are defined as the main chain atoms that precede and follow the loop but are not part of it. The search is performed through a database of many known protein structures, not only homologues of the modeled protein. Usually, many different hits are obtained and possibly sorted according different criteria (geometric or sequence similarity). The selected segments are then superposed and annealed onto the stem regions. These initial crude models are often refined by optimisation of some energy function. Databases searching approach is more accurate and efficient if it is precedent by an structural classification of the loops present in the database. Web-servers based on structural classification [51,52] are available (see table 1). When database searching is used, it must be keep on mind that the bigger the length of the loop, the lesser the number of putative solutions that will be in the database. At the present, this fact makes this approach specially useful for loops up to 7 residues long [53].
Finally, it must be remarked that prediction of a loop conformation is hindered by two main factors: i) the exponential increase of number of possible conformations as the length of the loop grows, and ii) the conformation of a loop is influenced by the core stem regions that span the loop as well as by the structure of rest of the protein that encircles the loop. These two factors make loop modeling one of the most difficult tasks of the comparative modeling process.
Sidechain modeling
Similarly to what happens in loop modeling, sidechain conformation can be predicted from knowledge-based approaches or taking into account steric or energetic considerations [54].
Knowledge-based approaches, which are the most widely used, employ libraries of common rotamers extracted from high resolution X-ray structures. Rotamers are tried successively and scored with a variety of energy functions [13]. This approach is implemented in most automatic homology modeling procedures. Among the available software to do so it is worth mentioning: the CORALL module of WHATIF [55] and the SCWRL program [56]. Several works have probed the biological relevance of side-chain modeling, as they may imply behavioural changes in protein-protein interactions and dimerization [57-59].
Errors in comparative models
Structural models obtained by homology will have regions that resemble the true structure and regions that do not. That is, all models contain a certain amount of errors, which are more frequent as sequence identity decreases. Any stage of the comparative modeling process has its own source of errors; accordingly, they can be divided in five categories [6]:
1) Incorrect templates. This is a problem when templates share less than 25% sequence identity with the target.
2) Missalignments errors. Accuracy of the alignments is still the key limitation on the quality and usefulness of the models, being the optimal placement of gaps its limiting factor [60]. If the target and the templates have over 40% sequence identity, the alignment is almost always correct. As percentage of identity decreases, regions of local low sequence similarity appear, and alignment errors are more feasible to occur. Alignment errors increase rapidly below 30% sequence identity and become the major source of errors in this kind of models [11,14]. Target-template alignment is probably the most crucial step in comparative modelling, since any errors at this step are usually impossible to correct later [47]. Therefore, it is indeed important to devote efforts to attain the most precise alignment.
3) Structural distortions in correctly aligned regions. As sequence identity decreases, it is possible that a segment correctly aligned adopts different local structure than the target, without disruption of the overall fold. It is convenient to use multiple templates whenever they are available to overcome this problem [61].
4) Errors in regions without a template. Insertions are the most challenging regions to model, because there is not a equivalent region in the template. The complexity of the problem increases with the length of the segment. Database searching [62] or energy-based methods [63] can be applied to predict the conformation of the insertion. If there are alignment errors at stem residues or at the other environment residues, insertion modeling is not likely to result in an accurate model [49]. Therefore, the most accurate environment surrounding the insertion, the better results are obtained.
5) Errors in sidechain packing. As sequence identity decreases below 30%, there is a rapid decrease in the conservation of sidechain packing. That is, rotamers of identical residues are not conserved because the overall surroundings are changed. In addition, it must be pointed out that the correct prediction of sidechain conformation is hampered by the coupling between mainchain and sidechains and by the continuous nature of the distribution of dihedral angles [54]. This kind of error can be critical if affecting residues implicated in protein function. As we will see later, a refinement of the structure by energy minimization or molecular dynamics can sometimes surmount this problem [64].
Summarising, consequences of the errors are more serious if they are made in the initial steps of the comparative modeling process: if the selection of the template is wrong, the model based on it will be wrong; if the alignment is incorrect, local features of the model will be incorrect. Remaining errors are mainly due to incorrect description of the environment of a particular region of the structure.
Evaluation of the models
The quality of the obtained model establish the limits of the information than can be safely extracted from it. Although all structural models obtained by enclose mistakes, they become less of a problem when it is possible to detect them. Once an error is identified, it is possible to discriminate whether it affects key structural or functional regions. Accordingly, strategies to surmount errors should be taken in consideration. Therefore, an essential step in the comparative modeling process is the detection of wrongly modelled regions.
There are two different approaches to estimate errors in a structure: 1) checking the consistency of the model with experimental data of the target protein, and 2) evaluating stereochemistry and other spatial features of the model by means of methods based on statistics derived from experimentally determined protein structures.
On the first approach experimental data is used to certainly determine if particular regions of the protein are correctly modeled. Biochemical data of the most important residues regarding protein overall structure and function can be used to validate the model [65,66]. That is, they should be in close proximity in 3D space and in the correct orientation to perform their role. A consistent modeling of such residues does not ensure a good prediction; conversely, inconsistency is a important reason for concern.
One essential requisite for a model is to have a good stereochemistry. Programs used to check the stereochemistry are based in the analysis of datasets of experimentally determined protein structures. With this respect, the most widely used program is PROCHECK [67], which provide an assessment of the overall quality of the structure and highlight regions that may need further investigation.
Besides stereochemistry, there are other spatial features in the proteins, that could be used as indicators of errors in the models: packing, creation of a hydrophobic core, residue and atomic solvent accessibilities, spatial distribution of charged groups, distribution of atom-atom distances and main-chain hydrogen bonding structures [47]. This kind of information is exploited in another group of programs based on the use of energetic profiles introduced by statistical criteria [68,69]. PROSAII [70] is probably the most widely used program of this category. Although there is a concern about the theoretical validity of the energy profiles for detecting local error in models [6], this approach have been successfully applied [71,72].
It is important to note here that it is highly recommended to analyse the experimentally determined structure of templates with PROCHECK and PROSA II programs. This should allow to discriminate between errors coming from the model and errors already present in the templates.
As a final step, energy minimization and/or molecular dynamics simulations [73] of the model can be done to minimize errors detected with PROCHECK and PROSA II. The most common used programs for this purpose are GROMOS [74], CHARMM [75] and AMBER [76], which explore and evaluate the multiple possible conformations of the protein.
Performing this step is still a controversial issue [77], because the description of the physico-chemical properties of the protein and its environment is not accurate enough [11]. Even though, new evidences are suggesting that long molecular dynamics simulations with explicit solvent could overcome errors in comparative modeling [64]. Over more, strategies focusing on the appropriate sampling of biologically relevant conformations of the protein have been proved to be useful refining the model. This can be achieved by restraining the movement of specific aminoacids [78] or to particular directions in the space [33].
Comparative modeling applications in the field of microbial cell factories
On structure-function relationships
Besides other general applications of protein comparative modeling [6], there are two of them which can be of particular interest in microbial cell factories:
1) Proposing residues for site-directed mutagenesis experiments in target proteins to assess its biological function. There are many examples of how comparative modeling has been used to propose mutants, dealing with different structural features of the protein, such as electrostatic charge and surface shape [79], loop flexibility and residue accessibility [80], the protein binding or enzyme active site [81] or an enzyme alosteric site [82], among others. It is not prudent to apply comparative modeling for this purpose if the target and templates do not share at least around 30% sequence identity, since the required degree of resolution of the model will be not enough to describe the affected structural features on the target protein [6].
2) Detecting an functional important regions of a protein. The knowledge achieved in the process must allow to design proteins with altered or improved functionality. The location of a binding site can be identified by localizing clusters of charged residues [83,84] or using data of deleterious mutations [82] Biological important regions tend to be predicted better than other parts of the model [14], because amino acids in the active and binding sites are often more conserved than other structural features in a protein [85]. In addition, activity is mostly based on the physicochemical properties of residues and its spatial orientation [86]. Consequently, the degree of sequence similarity shared by the target and the templates is less restrictive for this particular application, and thus homology modeling can be applied in a wide range of scenarios, including when sequence similarity drops below 30%.
On solving protein production related problems
There are other topics in protein production processes by means of cell factories in which structural-related features play a major role. One example is protein aggregation leading to bacterial inclusion bodies, which constitute a major bottleneck in protein production [87]. Recently, it has been shown that aggregation depends on specific interactions between solvent-exposed hydrophobic stretches which adopt the form of β-sheet structures [88]. This structural knowledge provides some insight on how to solve this problem: such interactions should be specifically disrupted to avoid aggregation of β-sheets. However, full understanding of this phenomena requires also comprehending the structural details on how two or more proteins interacts. This constitutes a challenging problem known as protein-protein docking prediction [89,90]. Recent works suggest that comparative modeling can be still helpful in combination with other experimental techniques to adress this problem [91,92].
On the meeting point of comparative modeling and cell microbial factories: structural genomics
A major necessity of medium- to high-scale protein production has recently arose with the development of initiatives on structural genomics [93,94]. These initiatives, which pursue to elucidate the tree-dimensional structures of all proteins [95,96], demand optimized and further robotized protein expression systems [87]. This aim will be achieved by a focused, large-scale determination of protein structures by X-ray crystallography and NMR spectroscopy, combined efficiently with accurate protein structure modeling techniques [6].
Structural genomics, as a first step, involves ensuring that each family of proteins is represented by a known structure, avoiding unworthy efforts that will result in redundant structural information. It must be pointed out that, nowadays, there are still families of proteins which must be excluded for this kind of large-scale studies. These problematic cases include integral membrane proteins, highly disulfide-bridge proteins and large complexes [87]. All projects employ exhaustively computational methods for target selection and family exclusion [97]. For the rest of proteins, three-dimensional models can be inferred from the previously resolved family representatives. As a result, a huge amount of structural data will be available, which in turn can serve as starting point for a rational protein production design.
A complete success of the structural genomics initiative critically depends on the advances in protein production technologies. This includes new approaches in expression of targets that show challenges on protein folding [87] and also in the development of automated or semi-automated methods, robust and inexpensive for protein purification [96].
Conclusion
We have attempted to establish the capabilities and limitations of current methods of comparative modeling, as well as a general strategy to follow up in a practical case, that hopefully could serve as a guide for biologist in this field. This methods are becoming important as tools for scientists working in microbial cell factories. We have shown in this review few examples where the use of comparative modeling have been used in this area.
Comparative modeling can be safely used when target and templates share at least 30% sequence identity. Below this threshold, modeling becomes a difficult task even for experts. In any case, models must be critically evaluated to be sure that they are correct enough, devoting most of efforts to the region involved in function.
Many challenging aspects of comparative modeling are active areas of research. The state-of-the-art of the protein structure prediction strategies and methodologies is tested every two years in the CASP (Critically Assessment of techniques for protein Structure Prediction) meeting. A carefully reading of the proceedings of the meeting is probably the best way to update the progress made by the field. See supplement 6 of volume 53 of Proteins for the last report available [98].
As a final advice, it is a good policy make use of different strategies to build the model and compare them. This is always pertinent but specially as sequence identity decreases. Consistency between different models does not ensure a good prediction; however, inconsistency is a meaningful cause of concern.
With the help of structural genomics, the structure of at least one member of the most globular folds will be determined in the next years, making comparative modeling more easy. However, this is not already true for membrane proteins, which constitute a more difficult scenario [99], and more improvements in both structure determination and modeling techniques are needed.
Finally, we do believe that comparative modeling should play key role in the microbial cell factories. It will help biologists to choose which are the most interesting mutant proteins to produce, to design new proteins with a desired function, or to modify a protein to avoid production-related problems.
List of abbreviations
Target: protein to be modelled. Templates: set of proteins, homologous to the target, for which three-dimensional structure is known. Model: inferred three-dimensional structure of the target. NMR: Nuclear Magnetic Resonance. HMM: Hidden Markov Model. Structural alignment: sequence alignment based on structural similarities. Cα : Alpha-carbon; carbon atom joining the carboxyl group and the amino group in an amino acid. Restraint: as referred to in this paper, a restraint is a reduction of the conformational space of a protein on account of a prior knowledge. Main chain: sequence of atoms within a protein formed by the carboxyl group, the alpha-carbon and the amino group of each of its amino acids. Side-Chain: atoms of an amino acid not belonging to the main chain. Stem: structured boundary of a loop. Rotamer: a particular conformation of the side-chain of an amino acid regarding the position of its main chain
Authors' contributions
NBC reviewed the comparative modeling sections and updated its methods. JP reviewed the protein expression systems and the applications of comparative modeling to the microbial cell factories field. BO coordinated the design and redaction of the manuscript. All authors read and approved the final manuscript.
Acknowledgements
Authors acknowledge Fundación Ramón Areces and MCyT grant BIO2002-0369.
==== Refs
Baneyx F Recombinant protein expression in Escherichia coli Curr Opin Biotechnol 1999 10 411 421 10508629 10.1016/S0958-1669(99)00003-8
Baneyx F Mujacic M Recombinant protein folding and misfolding in Escherichia coli Nat Biotechnol 2004 22 1399 1408 15529165 10.1038/nbt1029
Gerngross TU Advances in the production of human therapeutic proteins in yeasts and filamentous fungi Nat Biotechnol 2004 22 1409 1414 15529166 10.1038/nbt1028
Blundell TL Sibanda BL Sternberg MJ Thornton JM Knowledge-based prediction of protein structures and the design of novel molecules Nature 1987 326 347 352 3550471 10.1038/326347a0
Sanchez R Pieper U Melo F Eswar N Marti-Renom MA Madhusudhan MS Mirkovic N Sali A Protein structure modeling for structural genomics Nat Struct Biol 2000 7 986 990 11104007 10.1038/80776
Marti-Renom MA Stuart AC Fiser A Sanchez R Melo F Sali A Comparative protein structure modeling of genes and genomes Annu Rev Biophys Biomol Struct 2000 29 291 325 10940251 10.1146/annurev.biophys.29.1.291
Chothia C Lesk AM The relation between the divergence of sequence and structure in proteins Embo J 1986 5 823 826 3709526
Lesk AM Chothia C How different amino acid sequences determine similar protein structures: the structure and evolutionary dynamics of the globins J Mol Biol 1980 136 225 270 7373651 10.1016/0022-2836(80)90373-3
McPherson A Protein crystallization in the structural genomics era J Struct Funct Genomics 2004 5 1 2 15263837
Baker D Sali A Protein structure prediction and structural genomics Science 2001 294 93 96 11588250 10.1126/science.1065659
Fiser A Feig M Brooks CL 3rdSali A Evolution and physics in comparative protein structure modeling Acc Chem Res 2002 35 413 421 12069626 10.1021/ar010061h
Edwards YJ Cottage A Bioinformatics methods to predict protein structure and function. A practical approach Mol Biotechnol 2003 23 139 166 12632698 10.1385/MB:23:2:139
Krieger E Nabuurs SB Vriend G Homology modeling Methods Biochem Anal 2003 44 509 523 12647402
Kretsinger RH Ison RE Hovmoller S Prediction of protein structure Methods Enzymol 2004 383 1 27 15063644
Kosinski J Cymerman IA Feder M Kurowski MA Sasin JM Bujnicki JM A "FRankenstein's monster" approach to comparative modeling: merging the finest fragments of Fold-Recognition models and iterative model refinement aided by 3D structure evaluation Proteins 2003 53 369 379 14579325 10.1002/prot.10545
Rost B Twilight zone of protein sequence alignments Protein Eng 1999 12 85 94 10195279 10.1093/protein/12.2.85
Rost B Schneider R Sander C Protein fold recognition by prediction-based threading J Mol Biol 1997 270 471 480 9237912 10.1006/jmbi.1997.1101
Doolittle RF Of URFs and ORFs: a primer on how to analyze derived amino acid sequences 1986 Mill Valley, CA, USA: University Science Books
Vogt G Etzold T Argos P An assessment of amino acid exchange matrices in aligning protein sequences: the twilight zone revisited J Mol Biol 1995 249 816 831 7602593 10.1006/jmbi.1995.0340
Altschul SF Madden TL Schaffer AA Zhang J Zhang Z Miller W Lipman DJ Gapped BLAST and PSI-BLAST: a new generation of protein database search programs Nucleic Acids Res 1997 25 3389 3402 9254694 10.1093/nar/25.17.3389
Park J Teichmann SA Hubbard T Chothia C Intermediate sequences increase the detection of homology between sequences J Mol Biol 1997 273 349 354 9367767 10.1006/jmbi.1997.1288
Li W Pio F Pawlowski K Godzik A Saturated BLAST: an automated multiple intermediate sequence search used to detect distant homology Bioinformatics 2000 16 1105 1110 11159329 10.1093/bioinformatics/16.12.1105
John B Sali A Detection of homologous proteins by an intermediate sequence search Protein Sci 2004 13 54 62 14691221 10.1110/ps.03335004
Krogh A Brown M Mian IS Sjolander K Haussler D Hidden Markov models in computational biology. Applications to protein modeling J Mol Biol 1994 235 1501 1531 8107089 10.1006/jmbi.1994.1104
Bateman A Coin L Durbin R Finn RD Hollich V Griffiths-Jones S Khanna A Marshall M Moxon S Sonnhammer EL The Pfam protein families database Nucleic Acids Res 2004 D138 141 14681378 10.1093/nar/gkh121
Shi J Blundell TL Mizuguchi K FUGUE: sequence-structure homology recognition using environment-specific substitution tables and structure-dependent gap penalties J Mol Biol 2001 310 243 257 11419950 10.1006/jmbi.2001.4762
Jones DT Taylor WR Thornton JM A new approach to protein fold recognition Nature 1992 358 86 89 1614539 10.1038/358086a0
Kelley LA MacCallum RM Sternberg MJ Enhanced genome annotation using structural profiles in the program 3D-PSSM J Mol Biol 2000 299 499 520 10860755 10.1006/jmbi.2000.3741
Garriga D Diez J Oliva B Modeling the helicase domain of Brome mosaic virus 1a replicase J Mol Model (Online) 2004 10 5 6
Espadaler J Aragues R Eswar N Marti-Renom MA Querol E Aviles FX Sali A Oliva B Detecting remotely related proteins by their interactions and sequence similarity Proc Natl Acad Sci U S A 2005 102 7151 7156 15883372 10.1073/pnas.0500831102
Felsenstein Confidence-limits on phylogeneis – an approach using the bootstrap Evolution 1985 39 783 791
Tramontano A Leplae R Morea V Analysis and assessment of comparative modeling predictions in CASP4 Proteins 2001 22 38 11835479 10.1002/prot.10015
Qian B Ortiz AR Baker D Improvement of comparative model accuracy by free-energy optimization along principal components of natural structural variation Proc Natl Acad Sci U S A 2004 101 15346 15351 15492216 10.1073/pnas.0404703101
Jeanmougin F Thompson JD Gouy M Higgins DG Gibson TJ Multiple sequence alignment with Clustal X Trends Biochem Sci 1998 23 403 405 9810230 10.1016/S0968-0004(98)01285-7
Eddy SR Profile hidden Markov models Bioinformatics 1998 14 755 763 9918945 10.1093/bioinformatics/14.9.755
Russell RB Barton GJ Multiple protein sequence alignment from tertiary structure comparison: assignment of global and residue confidence levels Proteins 1992 14 309 323 1409577 10.1002/prot.340140216
Shindyalov IN Bourne PE Protein structure alignment by incremental combinatorial extension (CE) of the optimal path Protein Eng 1998 11 739 747 9796821 10.1093/protein/11.9.739
Kabsch W Sander C Dictionary of protein secondary structure: pattern recognition of hydrogen-bonded and geometrical features Biopolymers 1983 22 2577 2637 6667333 10.1002/bip.360221211
Sutcliffe MJ Haneef I Carney D Blundell TL Knowledge based modelling of homologous proteins, Part I: Three-dimensional frameworks derived from the simultaneous superposition of multiple structures Protein Eng 1987 1 377 384 3508286
Guex N Peitsch MC SWISS-MODEL and the Swiss-PdbViewer: an environment for comparative protein modeling Electrophoresis 1997 18 2714 2723 9504803 10.1002/elps.1150181505
Schwede T Kopp J Guex N Peitsch MC SWISS-MODEL: An automated protein homology-modeling server Nucleic Acids Res 2003 31 3381 3385 12824332 10.1093/nar/gkg520
Kopp J Schwede T The SWISS-MODEL Repository of annotated three-dimensional protein structure homology models Nucleic Acids Res 2004 D230 234 14681401 10.1093/nar/gkh008
Unger R Harel D Wherland S Sussman JL A 3D building blocks approach to analyzing and predicting structure of proteins Proteins 1989 5 355 373 2798411 10.1002/prot.340050410
Bates PA Kelley LA MacCallum RM Sternberg MJ Enhancement of protein modeling by human intervention in applying the automatic programs 3D-JIGSAW and 3D-PSSM Proteins 2001 39 46 11835480 10.1002/prot.1168
Havel TF Snow ME A new method for building protein conformations from sequence alignments with homologues of known structure J Mol Biol 1991 217 1 7 1988672 10.1016/0022-2836(91)90603-4
Sali A Blundell TL Comparative protein modelling by satisfaction of spatial restraints J Mol Biol 1993 234 779 815 8254673 10.1006/jmbi.1993.1626
Sali A Modeling mutations and homologous proteins Curr Opin Biotechnol 1995 6 437 451 7579655 10.1016/0958-1669(95)80074-3
Sanchez R Sali A Advances in comparative protein-structure modelling Curr Opin Struct Biol 1997 7 206 214 9094331 10.1016/S0959-440X(97)80027-9
Fiser A Do RK Sali A Modeling of loops in protein structures Protein Sci 2000 9 1753 1773 11045621
Fiser A Sali A ModLoop: automated modeling of loops in protein structures Bioinformatics 2003 19 2500 2501 14668246 10.1093/bioinformatics/btg362
Espadaler J Fernandez-Fuentes N Hermoso A Querol E Aviles FX Sternberg MJ Oliva B ArchDB: automated protein loop classification as a tool for structural genomics Nucleic Acids Res 2004 D185 188 14681390 10.1093/nar/gkh002
Burke DF Deane CM Blundell TL Browsing the SLoop database of structurally classified loops connecting elements of protein secondary structure Bioinformatics 2000 16 513 519 10980148 10.1093/bioinformatics/16.6.513
Fernandez-Fuentes N Querol E Aviles FX Sternberg MJ Oliva B Prediction of the conformation and geometry of loops in globular proteins. Testing ArchDB, a structural classification of loops Proteins 2005
Vasquez M Modeling side-chain conformation Curr Opin Struct Biol 1996 6 217 221 8728654 10.1016/S0959-440X(96)80077-7
Vriend G WHAT IF: a molecular modeling and drug design program J Mol Graph 1990 8 52 56 2268628 10.1016/0263-7855(90)80070-V
Canutescu AA Shelenkov AA Dunbrack RL Jr A graph-theory algorithm for rapid protein side-chain prediction Protein Sci 2003 12 2001 2014 12930999 10.1110/ps.03154503
Repiso A Oliva B Vives Corrons JL Carreras J Climent F Glucose phosphate isomerase deficiency: enzymatic and familial characterization of Arg346His mutation Biochim Biophys Acta 2005 1740 467 471 15949716
de Atauri P Repiso A Oliva B Lluis Vives-Corrons J Climent F Carreras J Characterization of the first described mutation of human red blood cell phosphoglycerate mutase Biochim Biophys Acta 2005 1740 403 410 15949708
Andres AM Soldevila M Navarro A Kidd KK Oliva B Bertranpetit J Positive selection in MAOA gene is human exclusive: determination of the putative amino acid change selected in the human lineage Hum Genet 2004 115 377 386 15349769
Moult J Fidelis K Zemla A Hubbard T Critical assessment of methods of protein structure prediction (CASP): round IV Proteins 2001 2 7 11835476 10.1002/prot.10054
Venclovas C Comparative modeling of CASP4 target proteins: combining results of sequence search with three-dimensional structure assessment Proteins 2001 47 54 11835481 10.1002/prot.10008
van Vlijmen HW Karplus M PDB-based protein loop prediction: parameters for selection and methods for optimization J Mol Biol 1997 267 975 1001 9135125 10.1006/jmbi.1996.0857
Mehler EL Periole X Hassan SA Weinstein H Key issues in the computational simulation of GPCR function: representation of loop domains J Comput Aided Mol Des 2002 16 841 853 12825797 10.1023/A:1023845015343
Fan H Mark AE Refinement of homology-based protein structures by molecular dynamics simulation techniques Protein Sci 2004 13 211 220 14691236 10.1110/ps.03381404
Carrieri A Centeno NB Rodrigo J Sanz F Carotti A Theoretical evidence of a salt bridge disruption as the initiating process for the alpha1d-adrenergic receptor activation: a molecular dynamics and docking study Proteins 2001 43 382 394 11340655 10.1002/prot.1051
Gutierrez-de-Teran H Centeno NB Pastor M Sanz F Novel approaches for modeling of the A1 adenosine receptor and its agonist binding site Proteins 2004 54 705 715 14997566 10.1002/prot.10617
Laskowski RA MacArthur MW Moss DB Thornton JM PROCHECK: a program to check the stereochemical quality of protein structures J Appl Crystallogr 1993 26 283 291 10.1107/S0021889892009944
Sippl MJ Calculation of conformational ensembles from potentials of mean force. An approach to the knowledge-based prediction of local structures in globular proteins J Mol Biol 1990 213 859 883 2359125
Luthy R Bowie JU Eisenberg D Assessment of protein models with three-dimensional profiles Nature 1992 356 83 85 1538787 10.1038/356083a0
Sippl MJ Recognition of errors in three-dimensional structures of proteins Proteins 1993 17 355 362 8108378 10.1002/prot.340170404
Zheng QC Li ZS Xiao JF Sun M Zhang Y Sun CC Homology modeling and PAPS ligand (cofactor) binding study of bovine phenol sulfotransferase J Mol Model (Online) 2005
Aloy P Mas JM Marti-Renom MA Querol E Aviles FX Oliva B Refinement of modelled structures by knowledge-based energy profiles and secondary structure prediction: application to the human procarboxypeptidase A2 J Comput Aided Mol Des 2000 14 83 92 10702927 10.1023/A:1008197831529
Hansson T Oostenbrink C van Gunsteren W Molecular dynamics simulations Curr Opin Struct Biol 2002 12 190 196 11959496 10.1016/S0959-440X(02)00308-1
Scott WRP Hunenberger PH Mark AE Billeter SR Fennen J Torda AE Huber T Kruger P van Gunsteren WF The GROMOS biomiolecular simulation program package J Phys Chem A 1999 103 3596 3607 10.1021/jp984217f
Brooks BR Bruccoleri RE Olafson BD States DJ Swaminathan S Karplus M CHARMM – A program for macromolecular energy, minimization, and dynamics calculations J Comput Chem 1983 16 513 519
Pearlman DA Case DA Caldwell JW Ross WS Cheatham TE Debolt S Ferguson D Seib el G Collman P AMBER, a package of computer-programs for applying molecular mechanics,<normal-mode analysis, molecular-dynamics and free-energy calculations to simulate the structural and energetic properties of molecules Comput Phys Commun 1995 91 1 41 10.1016/0010-4655(95)00041-D
Tramontano A Morea V Assessment of homology-based predictions in CASP5 Proteins 2003 53 352 368 14579324 10.1002/prot.10543
Flohil JA Vriend G Berendsen HJ Completion and refinement of 3-D homology models with restricted molecular dynamics: application to targets 47, 58, and 111 in the CASP modeling competition and posterior analysis Proteins 2002 48 593 604 12211026 10.1002/prot.10105
Iverson GM Reddel S Victoria EJ Cockerill KA Wang YX Marti-Renom MA Sali A Marquis DM Krilis SA Linnik MD Use of single point mutations in domain I of beta 2-glycoprotein I to determine fine antigenic specificity of antiphospholipid autoantibodies J Immunol 2002 169 7097 7103 12471146
Feliu JX Benito A Oliva B Aviles FX Villaverde A Conformational flexibility in a highly mobile protein loop of foot-and-mouth disease virus: distinct structural requirements for integrin and antibody binding J Mol Biol 1998 283 331 338 9769208 10.1006/jmbi.1998.2104
Sathyanarayanan PV Siems WF Jones JP Poovaiah BW Calcium-stimulated autophosphorylation site of plant chimeric calcium/calmodulin-dependent protein kinase J Biol Chem 2001 276 32940 32947 11399751 10.1074/jbc.M009648200
Gloyn AL Odili S Zelent D Buettger C Castleden HA Steele AM Stride A Shiota C Magnuson MA Lorini R Insights into the structure and regulation of glucokinase from a novel mutation (V62M), which causes maturity-onset diabetes of the young J Biol Chem 2005 280 14105 14113 15677479 10.1074/jbc.M413146200
Matsui E Abe J Yokoyama H Matsui I Aromatic residues located close to the active center are essential for the catalytic reaction of flap endonuclease-1 from hyperthermophilic archaeon Pyrococcus horikoshii J Biol Chem 2004 279 16687 16696 14742430 10.1074/jbc.M313695200
Ishino T Pasut G Scibek J Chaiken I Kinetic interaction analysis of human interleukin 5 receptor alpha mutants reveals a unique binding topology and charge distribution for cytokine recognition J Biol Chem 2004 279 9547 9556 14662768 10.1074/jbc.M309327200
Valdar WS Thornton JM Conservation helps to identify biologically relevant crystal contacts J Mol Biol 2001 313 399 416 11800565 10.1006/jmbi.2001.5034
Villa-Freixa J Bonet J Khan AK Johnston M Evaluating the relationship between residue stability and enzyme active site preorganization in protein regulated enzymes Theor Chem Acc 2005
Yokoyama S Protein expression systems for structural genomics and proteomics Curr Opin Chem Biol 2003 7 39 43 12547425 10.1016/S1367-5931(02)00019-4
Carrio M Gonzalez-Montalban N Vera A Villaverde A Ventura S Amylod-like properties of bacterial inclusion bodies J Mol Biol 2005 347 1025 1037 15784261 10.1016/j.jmb.2005.02.030
Wodak SJ Mendez R Prediction of protein-protein interactions: the CAPRI experiment, its evaluation and implications Curr Opin Struct Biol 2004 14 242 249 15093840 10.1016/j.sbi.2004.02.003
Schneidman-Duhovny D Nussinov R Wolfson HJ Predicting molecular interactions in silico: II. Protein-protein and protein-drug docking Curr Med Chem 2004 11 91 107 14754428 10.2174/0929867043456223
Seri M Savino M Bordo D Cusano R Rocca B Meloni I Di Bari F Koivisto PA Bolognesi M Ghiggeri GM Epstein syndrome: another renal disorder with mutations in the nonmuscle myosin heavy chain 9 gene Hum Genet 2002 110 182 186 11935325 10.1007/s00439-001-0659-1
Almstedt K Lundqvist M Carlsson J Karlsson M Persson B Jonsson BH Carlsson U Hammarstrom P Unfolding a folding disease: folding, misfolding and aggregation of the marble brain syndrome-associated mutant H107Y of human carbonic anhydrase II J Mol Biol 2004 342 619 633 15327960 10.1016/j.jmb.2004.07.024
Kim SH Shining a light on structural genomics Nat Struct Biol 1998 5 643 645 9699614 10.1038/1334
Goldsmith-Fischman S Honig B Structural genomics: computational methods for structure analysis Protein Sci 2003 12 1813 1821 12930981 10.1110/ps.0242903
Montelione GT Anderson S Structural genomics: keystone for a Human Proteome Project Nat Struct Biol 1999 6 11 12 9886282 10.1038/4878
Edwards AM Arrowsmith CH Christendat D Dharamsi A Friesen JD Greenblatt JF Vedadi M Protein production: feeding the crystallographers and NMR spectroscopists Nat Struct Biol 2000 7 970 972 11104003 10.1038/80751
Brenner SE Target selection for structural genomics Nat Struct Biol 2000 7 967 969 11104002 10.1038/80747
CASP5. Proceedings of the 5th Meeting on the Critical Assessment of Techniques for Protein Structure Prediction. 1–5 December Asilomar, California, USA Proteins 2002 53 333 595
Becker OM Shacham S Marantz Y Noiman S Modeling the 3D structure of GPCRs: advances and application to drug discovery Curr Opin Drug Discov Devel 2003 6 353 361 12833668
|
15989691
|
PMC1183243
|
CC BY
|
2021-01-04 16:24:35
|
no
|
Microb Cell Fact. 2005 Jun 30; 4:20
|
utf-8
|
Microb Cell Fact
| 2,005 |
10.1186/1475-2859-4-20
|
oa_comm
|
==== Front
Microb Cell FactMicrobial Cell Factories1475-2859BioMed Central London 1475-2859-4-201598969110.1186/1475-2859-4-20ReviewComparative modelling of protein structure and its impact on microbial cell factories Centeno Nuria B [email protected] Joan [email protected] Baldomero [email protected] Structural Bioinformatics Laboratory, Research Group on Biomedical Informatics (GRIB), IMIM/UPF. c/ Dr. Aiguader 80. 08003 Barcelona, Spain2005 30 6 2005 4 20 20 3 5 2005 30 6 2005 Copyright © 2005 Centeno et al; licensee BioMed Central Ltd.2005Centeno et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Comparative modeling is becoming an increasingly helpful technique in microbial cell factories as the knowledge of the three-dimensional structure of a protein would be an invaluable aid to solve problems on protein production. For this reason, an introduction to comparative modeling is presented, with special emphasis on the basic concepts, opportunities and challenges of protein structure prediction. This review is intended to serve as a guide for the biologist who has no special expertise and who is not involved in the determination of protein structure. Selected applications of comparative modeling in microbial cell factories are outlined, and the role of microbial cell factories in the structural genomics initiative is discussed.
==== Body
Review
Introduction
On the last two decades the development of recombinant DNA techniques has extended the use of microbial organisms to produce target proteins. The enteric bacterium Escherichia coli is one of the most extensively used prokaryotic organisms for genetic manipulations and for industrial production of proteins of therapeutic or commercial interest [1,2]. However, bacterial organisms often fail to produce target proteins due to problems related with protein misfolding and protein glycosilation. Yeast and fungal protein expression systems are used for the industrial production of relevant enzymes in such cases [3].
There are two main interests in the industrial production of proteins: i) Redefining the optimal properties of the target protein and ii) Avoiding problems of high-scale production. Knowledge of three-dimensional structure of the proteins may be helpful to redesign a modified protein. Computational prediction methods play an essential role to provide us with structural information of a sequence whose structure has not been experimentally determined. Homology based or comparative modeling [4] is the most detailed and accurate of all current protein structure prediction techniques [5]. Its aim is to build a three-dimensional model for a protein of unknown structure on the basis of sequence similarity to proteins of known structure [6]. Comparative modeling relies on the fact that structure is more conserved than sequence during evolution. Therefore, similar sequences exhibit nearly identical structures, and even distantly related sequences share the same fold [7,8]. Comparative modeling critically depends on the knowledge of three-dimensional structure of homologous proteins. The progress of structural genomics initiatives [9] allow to model a large amount of protein sequences. Besides, the number of unique structural folds that proteins adopt is limited {Zhang, 1997; #81; Liu, 2004 #48}. Consequently, it is likely that at least one example of most structural folds will be known, making comparative modeling applicable to most protein sequences. In term, an essential step of structural genomics is production of target proteins. Microbial cell factories play a key role in this context.
This review is intended to give a primer addressed to scientists of disciplines related to microbial cell factories who has no expertise in comparative modeling. Our goal is to provide the seeding background to understand concepts, opportunities and challenges of comparative modeling. We will describe each step in the comparative modeling process, discuss the most common errors and how to solve them, as well as outlining the applications of comparative modeling in the field of microbial cell factories.
We will emphasize the simplest and most reliable methodologies to follow up along with their range of application with a reduced number of useful programs and web servers. Many other authors have also written excellent reviews on the comparative modeling field [6,10-14].
Steps in comparative modeling
All current comparative modeling methods consist of four sequential steps: template selection, target-template alignment, model building and model evaluation. Essentially, this is an iterative procedure until a satisfactory model is obtained (Figure 1). In this process a variety of programs and web servers can be used (Table 1). Additionally, protein modeling meta-servers are emerging. They automatically implement the full process in a multi-step protocol, using simultaneously different methods [15].
Table 1 Useful servers and programs for protein comparive modeling.
PROGRAM Server/Web adress Reference
Template Selection
PSI-BLAST [20]
HMMER (HMM search) [35]
TOPITS [17]
FUGUE [26]
Threader [27]
3D-PSSM [28]
PFAM [25]
PHYLIP [31]
Target-Template alignment
CLUSTALW [34]
HMMER (HMM align) [35]
STAMP [36]
CE [37]
DSSP [38]
Model Building
COMPOSER [39]
SwissModel [41]
3D-JIGSAW [44]
MODELLER [46]
Loop Modeling
MODLOOP [50]
ARCHDB [51]
Sloop [52]
Sidechain Modeling
WHAT IF [55]
SCWRL [56]
Evaluation of the model
PROCHECK [67]
PROSA II [70]
Biotech
Refinement
GROMOS [74]
CHARMM [75]
AMBER [76]
Figure 1 Flowchart of methods used for comparative modeling. Scheme of the methods used for comparative modeling, comprising template(s) selection, template-target alignment, model (backbone and loops) building, sidechain modeling, model evaluation, and model refinement steps. Programs and servers referring to these steps are listed in table 1.
Template selection
The starting point in comparative modelling is to identify protein structures related to the target sequence and then to select those that will be used as templates. Such templates may be found by sequence comparison methods or by sequence-structure methods also known as threading methods. Sequence comparison methods can be safely used above a certain threshold in terms of sequence identity (i.e. percentage of identical paired residues in an alignment). It has been shown that above that threshold -which is strongly dependent on sequence length, sequence homology implies structural identity [16]. Even though, below that threshold structural likeness is still possible. Some protein pairs sharing very little sequence similarity may have become similar by convergent or divergent evolution. The alignment of these proteins pairs, which define the so called "midnight zone" in sequence alignment [17], is usually addressed with threading methods. Finally, there is a range in terms of sequence identity amid the safe zone and the "midnight zone" in which the relationship between structural homology and the phylogeny is unclear: the "twilight zone" [18]. Within this range, which is usually defined between 20 and 35% sequence identity [19], additional caution must be taken on the sequence alignment.
PSI-BLAST [20], an iterative sequence comparison method, is probably the most widely used program to detect remote similarities. In more difficult scenarios, where sequence homology is not so evident, templates can be found by searching in sequence space using intermediate sequence search (ISS) methods [21-23].
Sequence comparisons can also be made through Hidden Markov Models (HMMs) [24] as implemented, for instance, in HMMER. HMMs profiles of protein domain families are available in Pfam database. These profiles can be used to automatically identify protein domain(s) within the target even if it shares weak sequence similarity with templates [25].
Threading methods have been developed to find more distant relationships. For this reason, they are the most promising choice in the absence of homologues to the target sequence. Threading methods involve performing sensitive sequence searches and characterizing sequence compatibility with the structural environments of putative templates. Features analysed by this kind of methods include secondary structure and solvent accessibility predictions as well as functional annotation. Most used methods of this kind are TOPITS [17], FUGUE [26], Threader [27] and 3D-PSSM [28]. Recent examples of the combined use of these servers and further modeling [29] prove its use. Other methods are being developed based on the analysis of protein-protein interactions to search for remote similarities [30].
Once a list of related proteins with known structure has been obtained, it is necessary to select those templates that are appropriate for the given modelling problem. The feasibility of a template can be assessed by means of its expectation value, E-value [20], which is one of the parameters in the searches outputs. As a general rule, the lower the E-value, the better the template is.
Besides, several other factors should be considered when selecting a template:
1) Quality of the experimental template structure. Because errors in templates will be passed onto the models, the better templates are the most accurate structures available. Accuracy of the templates can be assessed by the resolution and the R-factor for a crystallographic structure, or by the number of distance restraints per residue in the case of NMR structure.
2) Environment likeness. Experimental factors of interest for the target (i.e. the presence of a ligand in the structure, pH, and solvent features ...) should be found as similar as possible in the chosen templates.
3) Phylogenetic similarity. It is helpful to build a multiple alignment and a phylogenetic tree [31] of the target and templates, in order to select templates from the subfamily that is closest to the target sequence. The phylogenetic tree can be constructed by means of PHYLIP set of programs [31].
Depending on the purpose of the model, some of the factors listed above will be more important than others. For instance, resolution of the template is probably the most important factor if the reason for building the model is to design mutants of a binding site, since an accurate geometrical description is needed.
It is important to emphasize that it is not mandatory to select only one template. Actually, methods using multiple templates seems to perform better than those based on a single template [14,32], especially if the main modes extracted from them are taken into account [33]. Finally, it is noteworthy to be aware that, implicitly, choosing templates means the recognition of the target's overall fold.
Target-template alignment
Once templates have been selected, an optimal alignment between the target sequence and templates is needed to further construct a three dimensional model of the target.
From easiest to more complex, some strategies for aligning target and templates are:
1) Obtaining a multiple alignment of the templates and the target using CLUSTALW [34].
2) Aligning the query sequence to a HMM profile of the templates family built from a Pfam alignment [25] using HMMER [35].
3) Aligning the query sequence to a HMM profile of the templates built from a structural alignment using HMMER [35]. Structural alignments required for this strategy can be obtained using STAMP [36] or CE [37]; an automatic web server is available for the later one.
In our experience this third strategy, in which not only the sequence similarity but also the structural information inherent in the templates guide the alignment, make it more trustworthy.
The obtained alignments must be critically evaluated in terms of the number, length, and position of the gaps opened. Some of them can be manually refined, taking into account the secondary structure of the templates and their accessible surface, both of them calculated with DSSP program [38], in order to avoid gaps which are opened within secondary structural elements. This will be important if the alignment strategy is based only on sequence similarity. In any case, at this step, if necessary, the selection of templates may be revisited, either to search a new template to overcome a gap in a particular region or to remove redundant or inadequate templates.
Model building
Comparative model building generates an all-atom model of the sequence based on its alignment to one or more templates. It includes either sequential or simultaneous modelling of the core of the protein, loops, and side-chains.
The original comparative approach, which is still widely used, is modelling by rigid-body assembly [4]. This method constructs the model from a few regions which are obtained from dissecting related structures. In order to assemble the dissected parts, a framework is calculated by averaging of Cα atoms of structurally conserved regions in template structures. Structurally variable regions are modeled by choosing from a database of all known proteins those regions that better fit the anchor conserved regions. COMPOSER [39] is one of the programs that use this methodology in a semiautomated procedure. SwissModel is a commonly used automated web server also based in this approach [40-42].
Modeling by segment matching is another approach which relies on the approximate positions of conserved atoms in templates [27,43]. This is accomplished by breaking the target into a set of short segments, and searching in a database for matching fragments which are fitted onto an initial framework of the target structure. Database searching is based on sequence similarity, conformational similarity and compatibility with the target structure. 3D-JIGSAW is one of the successful programs that uses this approach [44]
Another approach is modeling by satisfaction of the spatial restrains obtained from the alignment [45]. Probably the most used program based on this approach is MODELLER [46]. First, this automated procedure derives many distance and dihedral angle restraints on the target sequence from its alignment with template three-dimensional structures. Next, this homology-derived restraints and energy terms ensuring proper stereochemistry are combined into a function. Finally, the model is obtained by optimizing this function in such a way that the model violates the input restraints as little as possible. Several slightly different models in agreement with the restraints can be calculated.
Any of the three methods above described produce models of similar accuracy if they are optimally applied. In the difficult cases, modeling by satisfaction of spatial restraints is perhaps the most accurate technique, since it can use many different types of information about the target sequence. In this way, available experimental data can be added as new restraints, making the model more reliable.
Loop modeling
Along with alignment, loop modeling is probably the most difficult step in comparative modelling process. Errors in loops are the dominant problem in comparative modelling when target and template share above 35% sequence identity. This is a very active area of research and it is not practical to consider all available methods (details of some of them can be found in [47-49]). In this review we will present the state-of-the art of methods that can be easily used. Furthermore, it must be pointed out that although existing methods can provide reasonably accurate models of short loop regions; modeling of long loops is still an unsolved problem [11]. Loop modeling methods can be classified in two approaches: ab initio methods and database searching or knowledge-based methods.
Ab initio loop prediction is based on a conformational search guided by a scoring or energy function -the later describing the physico-chemical properties of a protein and its environment. There are many such methods, making use of different protein representations, energy functions and optimisation procedures [11]. Among them, there is an option to use implemented in MODELLER or in a web server MODLOOP [50].
The database approach to loop prediction begins by finding segments of main chain that fit the two stems of a loop. The stems are defined as the main chain atoms that precede and follow the loop but are not part of it. The search is performed through a database of many known protein structures, not only homologues of the modeled protein. Usually, many different hits are obtained and possibly sorted according different criteria (geometric or sequence similarity). The selected segments are then superposed and annealed onto the stem regions. These initial crude models are often refined by optimisation of some energy function. Databases searching approach is more accurate and efficient if it is precedent by an structural classification of the loops present in the database. Web-servers based on structural classification [51,52] are available (see table 1). When database searching is used, it must be keep on mind that the bigger the length of the loop, the lesser the number of putative solutions that will be in the database. At the present, this fact makes this approach specially useful for loops up to 7 residues long [53].
Finally, it must be remarked that prediction of a loop conformation is hindered by two main factors: i) the exponential increase of number of possible conformations as the length of the loop grows, and ii) the conformation of a loop is influenced by the core stem regions that span the loop as well as by the structure of rest of the protein that encircles the loop. These two factors make loop modeling one of the most difficult tasks of the comparative modeling process.
Sidechain modeling
Similarly to what happens in loop modeling, sidechain conformation can be predicted from knowledge-based approaches or taking into account steric or energetic considerations [54].
Knowledge-based approaches, which are the most widely used, employ libraries of common rotamers extracted from high resolution X-ray structures. Rotamers are tried successively and scored with a variety of energy functions [13]. This approach is implemented in most automatic homology modeling procedures. Among the available software to do so it is worth mentioning: the CORALL module of WHATIF [55] and the SCWRL program [56]. Several works have probed the biological relevance of side-chain modeling, as they may imply behavioural changes in protein-protein interactions and dimerization [57-59].
Errors in comparative models
Structural models obtained by homology will have regions that resemble the true structure and regions that do not. That is, all models contain a certain amount of errors, which are more frequent as sequence identity decreases. Any stage of the comparative modeling process has its own source of errors; accordingly, they can be divided in five categories [6]:
1) Incorrect templates. This is a problem when templates share less than 25% sequence identity with the target.
2) Missalignments errors. Accuracy of the alignments is still the key limitation on the quality and usefulness of the models, being the optimal placement of gaps its limiting factor [60]. If the target and the templates have over 40% sequence identity, the alignment is almost always correct. As percentage of identity decreases, regions of local low sequence similarity appear, and alignment errors are more feasible to occur. Alignment errors increase rapidly below 30% sequence identity and become the major source of errors in this kind of models [11,14]. Target-template alignment is probably the most crucial step in comparative modelling, since any errors at this step are usually impossible to correct later [47]. Therefore, it is indeed important to devote efforts to attain the most precise alignment.
3) Structural distortions in correctly aligned regions. As sequence identity decreases, it is possible that a segment correctly aligned adopts different local structure than the target, without disruption of the overall fold. It is convenient to use multiple templates whenever they are available to overcome this problem [61].
4) Errors in regions without a template. Insertions are the most challenging regions to model, because there is not a equivalent region in the template. The complexity of the problem increases with the length of the segment. Database searching [62] or energy-based methods [63] can be applied to predict the conformation of the insertion. If there are alignment errors at stem residues or at the other environment residues, insertion modeling is not likely to result in an accurate model [49]. Therefore, the most accurate environment surrounding the insertion, the better results are obtained.
5) Errors in sidechain packing. As sequence identity decreases below 30%, there is a rapid decrease in the conservation of sidechain packing. That is, rotamers of identical residues are not conserved because the overall surroundings are changed. In addition, it must be pointed out that the correct prediction of sidechain conformation is hampered by the coupling between mainchain and sidechains and by the continuous nature of the distribution of dihedral angles [54]. This kind of error can be critical if affecting residues implicated in protein function. As we will see later, a refinement of the structure by energy minimization or molecular dynamics can sometimes surmount this problem [64].
Summarising, consequences of the errors are more serious if they are made in the initial steps of the comparative modeling process: if the selection of the template is wrong, the model based on it will be wrong; if the alignment is incorrect, local features of the model will be incorrect. Remaining errors are mainly due to incorrect description of the environment of a particular region of the structure.
Evaluation of the models
The quality of the obtained model establish the limits of the information than can be safely extracted from it. Although all structural models obtained by enclose mistakes, they become less of a problem when it is possible to detect them. Once an error is identified, it is possible to discriminate whether it affects key structural or functional regions. Accordingly, strategies to surmount errors should be taken in consideration. Therefore, an essential step in the comparative modeling process is the detection of wrongly modelled regions.
There are two different approaches to estimate errors in a structure: 1) checking the consistency of the model with experimental data of the target protein, and 2) evaluating stereochemistry and other spatial features of the model by means of methods based on statistics derived from experimentally determined protein structures.
On the first approach experimental data is used to certainly determine if particular regions of the protein are correctly modeled. Biochemical data of the most important residues regarding protein overall structure and function can be used to validate the model [65,66]. That is, they should be in close proximity in 3D space and in the correct orientation to perform their role. A consistent modeling of such residues does not ensure a good prediction; conversely, inconsistency is a important reason for concern.
One essential requisite for a model is to have a good stereochemistry. Programs used to check the stereochemistry are based in the analysis of datasets of experimentally determined protein structures. With this respect, the most widely used program is PROCHECK [67], which provide an assessment of the overall quality of the structure and highlight regions that may need further investigation.
Besides stereochemistry, there are other spatial features in the proteins, that could be used as indicators of errors in the models: packing, creation of a hydrophobic core, residue and atomic solvent accessibilities, spatial distribution of charged groups, distribution of atom-atom distances and main-chain hydrogen bonding structures [47]. This kind of information is exploited in another group of programs based on the use of energetic profiles introduced by statistical criteria [68,69]. PROSAII [70] is probably the most widely used program of this category. Although there is a concern about the theoretical validity of the energy profiles for detecting local error in models [6], this approach have been successfully applied [71,72].
It is important to note here that it is highly recommended to analyse the experimentally determined structure of templates with PROCHECK and PROSA II programs. This should allow to discriminate between errors coming from the model and errors already present in the templates.
As a final step, energy minimization and/or molecular dynamics simulations [73] of the model can be done to minimize errors detected with PROCHECK and PROSA II. The most common used programs for this purpose are GROMOS [74], CHARMM [75] and AMBER [76], which explore and evaluate the multiple possible conformations of the protein.
Performing this step is still a controversial issue [77], because the description of the physico-chemical properties of the protein and its environment is not accurate enough [11]. Even though, new evidences are suggesting that long molecular dynamics simulations with explicit solvent could overcome errors in comparative modeling [64]. Over more, strategies focusing on the appropriate sampling of biologically relevant conformations of the protein have been proved to be useful refining the model. This can be achieved by restraining the movement of specific aminoacids [78] or to particular directions in the space [33].
Comparative modeling applications in the field of microbial cell factories
On structure-function relationships
Besides other general applications of protein comparative modeling [6], there are two of them which can be of particular interest in microbial cell factories:
1) Proposing residues for site-directed mutagenesis experiments in target proteins to assess its biological function. There are many examples of how comparative modeling has been used to propose mutants, dealing with different structural features of the protein, such as electrostatic charge and surface shape [79], loop flexibility and residue accessibility [80], the protein binding or enzyme active site [81] or an enzyme alosteric site [82], among others. It is not prudent to apply comparative modeling for this purpose if the target and templates do not share at least around 30% sequence identity, since the required degree of resolution of the model will be not enough to describe the affected structural features on the target protein [6].
2) Detecting an functional important regions of a protein. The knowledge achieved in the process must allow to design proteins with altered or improved functionality. The location of a binding site can be identified by localizing clusters of charged residues [83,84] or using data of deleterious mutations [82] Biological important regions tend to be predicted better than other parts of the model [14], because amino acids in the active and binding sites are often more conserved than other structural features in a protein [85]. In addition, activity is mostly based on the physicochemical properties of residues and its spatial orientation [86]. Consequently, the degree of sequence similarity shared by the target and the templates is less restrictive for this particular application, and thus homology modeling can be applied in a wide range of scenarios, including when sequence similarity drops below 30%.
On solving protein production related problems
There are other topics in protein production processes by means of cell factories in which structural-related features play a major role. One example is protein aggregation leading to bacterial inclusion bodies, which constitute a major bottleneck in protein production [87]. Recently, it has been shown that aggregation depends on specific interactions between solvent-exposed hydrophobic stretches which adopt the form of β-sheet structures [88]. This structural knowledge provides some insight on how to solve this problem: such interactions should be specifically disrupted to avoid aggregation of β-sheets. However, full understanding of this phenomena requires also comprehending the structural details on how two or more proteins interacts. This constitutes a challenging problem known as protein-protein docking prediction [89,90]. Recent works suggest that comparative modeling can be still helpful in combination with other experimental techniques to adress this problem [91,92].
On the meeting point of comparative modeling and cell microbial factories: structural genomics
A major necessity of medium- to high-scale protein production has recently arose with the development of initiatives on structural genomics [93,94]. These initiatives, which pursue to elucidate the tree-dimensional structures of all proteins [95,96], demand optimized and further robotized protein expression systems [87]. This aim will be achieved by a focused, large-scale determination of protein structures by X-ray crystallography and NMR spectroscopy, combined efficiently with accurate protein structure modeling techniques [6].
Structural genomics, as a first step, involves ensuring that each family of proteins is represented by a known structure, avoiding unworthy efforts that will result in redundant structural information. It must be pointed out that, nowadays, there are still families of proteins which must be excluded for this kind of large-scale studies. These problematic cases include integral membrane proteins, highly disulfide-bridge proteins and large complexes [87]. All projects employ exhaustively computational methods for target selection and family exclusion [97]. For the rest of proteins, three-dimensional models can be inferred from the previously resolved family representatives. As a result, a huge amount of structural data will be available, which in turn can serve as starting point for a rational protein production design.
A complete success of the structural genomics initiative critically depends on the advances in protein production technologies. This includes new approaches in expression of targets that show challenges on protein folding [87] and also in the development of automated or semi-automated methods, robust and inexpensive for protein purification [96].
Conclusion
We have attempted to establish the capabilities and limitations of current methods of comparative modeling, as well as a general strategy to follow up in a practical case, that hopefully could serve as a guide for biologist in this field. This methods are becoming important as tools for scientists working in microbial cell factories. We have shown in this review few examples where the use of comparative modeling have been used in this area.
Comparative modeling can be safely used when target and templates share at least 30% sequence identity. Below this threshold, modeling becomes a difficult task even for experts. In any case, models must be critically evaluated to be sure that they are correct enough, devoting most of efforts to the region involved in function.
Many challenging aspects of comparative modeling are active areas of research. The state-of-the-art of the protein structure prediction strategies and methodologies is tested every two years in the CASP (Critically Assessment of techniques for protein Structure Prediction) meeting. A carefully reading of the proceedings of the meeting is probably the best way to update the progress made by the field. See supplement 6 of volume 53 of Proteins for the last report available [98].
As a final advice, it is a good policy make use of different strategies to build the model and compare them. This is always pertinent but specially as sequence identity decreases. Consistency between different models does not ensure a good prediction; however, inconsistency is a meaningful cause of concern.
With the help of structural genomics, the structure of at least one member of the most globular folds will be determined in the next years, making comparative modeling more easy. However, this is not already true for membrane proteins, which constitute a more difficult scenario [99], and more improvements in both structure determination and modeling techniques are needed.
Finally, we do believe that comparative modeling should play key role in the microbial cell factories. It will help biologists to choose which are the most interesting mutant proteins to produce, to design new proteins with a desired function, or to modify a protein to avoid production-related problems.
List of abbreviations
Target: protein to be modelled. Templates: set of proteins, homologous to the target, for which three-dimensional structure is known. Model: inferred three-dimensional structure of the target. NMR: Nuclear Magnetic Resonance. HMM: Hidden Markov Model. Structural alignment: sequence alignment based on structural similarities. Cα : Alpha-carbon; carbon atom joining the carboxyl group and the amino group in an amino acid. Restraint: as referred to in this paper, a restraint is a reduction of the conformational space of a protein on account of a prior knowledge. Main chain: sequence of atoms within a protein formed by the carboxyl group, the alpha-carbon and the amino group of each of its amino acids. Side-Chain: atoms of an amino acid not belonging to the main chain. Stem: structured boundary of a loop. Rotamer: a particular conformation of the side-chain of an amino acid regarding the position of its main chain
Authors' contributions
NBC reviewed the comparative modeling sections and updated its methods. JP reviewed the protein expression systems and the applications of comparative modeling to the microbial cell factories field. BO coordinated the design and redaction of the manuscript. All authors read and approved the final manuscript.
Acknowledgements
Authors acknowledge Fundación Ramón Areces and MCyT grant BIO2002-0369.
==== Refs
Baneyx F Recombinant protein expression in Escherichia coli Curr Opin Biotechnol 1999 10 411 421 10508629 10.1016/S0958-1669(99)00003-8
Baneyx F Mujacic M Recombinant protein folding and misfolding in Escherichia coli Nat Biotechnol 2004 22 1399 1408 15529165 10.1038/nbt1029
Gerngross TU Advances in the production of human therapeutic proteins in yeasts and filamentous fungi Nat Biotechnol 2004 22 1409 1414 15529166 10.1038/nbt1028
Blundell TL Sibanda BL Sternberg MJ Thornton JM Knowledge-based prediction of protein structures and the design of novel molecules Nature 1987 326 347 352 3550471 10.1038/326347a0
Sanchez R Pieper U Melo F Eswar N Marti-Renom MA Madhusudhan MS Mirkovic N Sali A Protein structure modeling for structural genomics Nat Struct Biol 2000 7 986 990 11104007 10.1038/80776
Marti-Renom MA Stuart AC Fiser A Sanchez R Melo F Sali A Comparative protein structure modeling of genes and genomes Annu Rev Biophys Biomol Struct 2000 29 291 325 10940251 10.1146/annurev.biophys.29.1.291
Chothia C Lesk AM The relation between the divergence of sequence and structure in proteins Embo J 1986 5 823 826 3709526
Lesk AM Chothia C How different amino acid sequences determine similar protein structures: the structure and evolutionary dynamics of the globins J Mol Biol 1980 136 225 270 7373651 10.1016/0022-2836(80)90373-3
McPherson A Protein crystallization in the structural genomics era J Struct Funct Genomics 2004 5 1 2 15263837
Baker D Sali A Protein structure prediction and structural genomics Science 2001 294 93 96 11588250 10.1126/science.1065659
Fiser A Feig M Brooks CL 3rdSali A Evolution and physics in comparative protein structure modeling Acc Chem Res 2002 35 413 421 12069626 10.1021/ar010061h
Edwards YJ Cottage A Bioinformatics methods to predict protein structure and function. A practical approach Mol Biotechnol 2003 23 139 166 12632698 10.1385/MB:23:2:139
Krieger E Nabuurs SB Vriend G Homology modeling Methods Biochem Anal 2003 44 509 523 12647402
Kretsinger RH Ison RE Hovmoller S Prediction of protein structure Methods Enzymol 2004 383 1 27 15063644
Kosinski J Cymerman IA Feder M Kurowski MA Sasin JM Bujnicki JM A "FRankenstein's monster" approach to comparative modeling: merging the finest fragments of Fold-Recognition models and iterative model refinement aided by 3D structure evaluation Proteins 2003 53 369 379 14579325 10.1002/prot.10545
Rost B Twilight zone of protein sequence alignments Protein Eng 1999 12 85 94 10195279 10.1093/protein/12.2.85
Rost B Schneider R Sander C Protein fold recognition by prediction-based threading J Mol Biol 1997 270 471 480 9237912 10.1006/jmbi.1997.1101
Doolittle RF Of URFs and ORFs: a primer on how to analyze derived amino acid sequences 1986 Mill Valley, CA, USA: University Science Books
Vogt G Etzold T Argos P An assessment of amino acid exchange matrices in aligning protein sequences: the twilight zone revisited J Mol Biol 1995 249 816 831 7602593 10.1006/jmbi.1995.0340
Altschul SF Madden TL Schaffer AA Zhang J Zhang Z Miller W Lipman DJ Gapped BLAST and PSI-BLAST: a new generation of protein database search programs Nucleic Acids Res 1997 25 3389 3402 9254694 10.1093/nar/25.17.3389
Park J Teichmann SA Hubbard T Chothia C Intermediate sequences increase the detection of homology between sequences J Mol Biol 1997 273 349 354 9367767 10.1006/jmbi.1997.1288
Li W Pio F Pawlowski K Godzik A Saturated BLAST: an automated multiple intermediate sequence search used to detect distant homology Bioinformatics 2000 16 1105 1110 11159329 10.1093/bioinformatics/16.12.1105
John B Sali A Detection of homologous proteins by an intermediate sequence search Protein Sci 2004 13 54 62 14691221 10.1110/ps.03335004
Krogh A Brown M Mian IS Sjolander K Haussler D Hidden Markov models in computational biology. Applications to protein modeling J Mol Biol 1994 235 1501 1531 8107089 10.1006/jmbi.1994.1104
Bateman A Coin L Durbin R Finn RD Hollich V Griffiths-Jones S Khanna A Marshall M Moxon S Sonnhammer EL The Pfam protein families database Nucleic Acids Res 2004 D138 141 14681378 10.1093/nar/gkh121
Shi J Blundell TL Mizuguchi K FUGUE: sequence-structure homology recognition using environment-specific substitution tables and structure-dependent gap penalties J Mol Biol 2001 310 243 257 11419950 10.1006/jmbi.2001.4762
Jones DT Taylor WR Thornton JM A new approach to protein fold recognition Nature 1992 358 86 89 1614539 10.1038/358086a0
Kelley LA MacCallum RM Sternberg MJ Enhanced genome annotation using structural profiles in the program 3D-PSSM J Mol Biol 2000 299 499 520 10860755 10.1006/jmbi.2000.3741
Garriga D Diez J Oliva B Modeling the helicase domain of Brome mosaic virus 1a replicase J Mol Model (Online) 2004 10 5 6
Espadaler J Aragues R Eswar N Marti-Renom MA Querol E Aviles FX Sali A Oliva B Detecting remotely related proteins by their interactions and sequence similarity Proc Natl Acad Sci U S A 2005 102 7151 7156 15883372 10.1073/pnas.0500831102
Felsenstein Confidence-limits on phylogeneis – an approach using the bootstrap Evolution 1985 39 783 791
Tramontano A Leplae R Morea V Analysis and assessment of comparative modeling predictions in CASP4 Proteins 2001 22 38 11835479 10.1002/prot.10015
Qian B Ortiz AR Baker D Improvement of comparative model accuracy by free-energy optimization along principal components of natural structural variation Proc Natl Acad Sci U S A 2004 101 15346 15351 15492216 10.1073/pnas.0404703101
Jeanmougin F Thompson JD Gouy M Higgins DG Gibson TJ Multiple sequence alignment with Clustal X Trends Biochem Sci 1998 23 403 405 9810230 10.1016/S0968-0004(98)01285-7
Eddy SR Profile hidden Markov models Bioinformatics 1998 14 755 763 9918945 10.1093/bioinformatics/14.9.755
Russell RB Barton GJ Multiple protein sequence alignment from tertiary structure comparison: assignment of global and residue confidence levels Proteins 1992 14 309 323 1409577 10.1002/prot.340140216
Shindyalov IN Bourne PE Protein structure alignment by incremental combinatorial extension (CE) of the optimal path Protein Eng 1998 11 739 747 9796821 10.1093/protein/11.9.739
Kabsch W Sander C Dictionary of protein secondary structure: pattern recognition of hydrogen-bonded and geometrical features Biopolymers 1983 22 2577 2637 6667333 10.1002/bip.360221211
Sutcliffe MJ Haneef I Carney D Blundell TL Knowledge based modelling of homologous proteins, Part I: Three-dimensional frameworks derived from the simultaneous superposition of multiple structures Protein Eng 1987 1 377 384 3508286
Guex N Peitsch MC SWISS-MODEL and the Swiss-PdbViewer: an environment for comparative protein modeling Electrophoresis 1997 18 2714 2723 9504803 10.1002/elps.1150181505
Schwede T Kopp J Guex N Peitsch MC SWISS-MODEL: An automated protein homology-modeling server Nucleic Acids Res 2003 31 3381 3385 12824332 10.1093/nar/gkg520
Kopp J Schwede T The SWISS-MODEL Repository of annotated three-dimensional protein structure homology models Nucleic Acids Res 2004 D230 234 14681401 10.1093/nar/gkh008
Unger R Harel D Wherland S Sussman JL A 3D building blocks approach to analyzing and predicting structure of proteins Proteins 1989 5 355 373 2798411 10.1002/prot.340050410
Bates PA Kelley LA MacCallum RM Sternberg MJ Enhancement of protein modeling by human intervention in applying the automatic programs 3D-JIGSAW and 3D-PSSM Proteins 2001 39 46 11835480 10.1002/prot.1168
Havel TF Snow ME A new method for building protein conformations from sequence alignments with homologues of known structure J Mol Biol 1991 217 1 7 1988672 10.1016/0022-2836(91)90603-4
Sali A Blundell TL Comparative protein modelling by satisfaction of spatial restraints J Mol Biol 1993 234 779 815 8254673 10.1006/jmbi.1993.1626
Sali A Modeling mutations and homologous proteins Curr Opin Biotechnol 1995 6 437 451 7579655 10.1016/0958-1669(95)80074-3
Sanchez R Sali A Advances in comparative protein-structure modelling Curr Opin Struct Biol 1997 7 206 214 9094331 10.1016/S0959-440X(97)80027-9
Fiser A Do RK Sali A Modeling of loops in protein structures Protein Sci 2000 9 1753 1773 11045621
Fiser A Sali A ModLoop: automated modeling of loops in protein structures Bioinformatics 2003 19 2500 2501 14668246 10.1093/bioinformatics/btg362
Espadaler J Fernandez-Fuentes N Hermoso A Querol E Aviles FX Sternberg MJ Oliva B ArchDB: automated protein loop classification as a tool for structural genomics Nucleic Acids Res 2004 D185 188 14681390 10.1093/nar/gkh002
Burke DF Deane CM Blundell TL Browsing the SLoop database of structurally classified loops connecting elements of protein secondary structure Bioinformatics 2000 16 513 519 10980148 10.1093/bioinformatics/16.6.513
Fernandez-Fuentes N Querol E Aviles FX Sternberg MJ Oliva B Prediction of the conformation and geometry of loops in globular proteins. Testing ArchDB, a structural classification of loops Proteins 2005
Vasquez M Modeling side-chain conformation Curr Opin Struct Biol 1996 6 217 221 8728654 10.1016/S0959-440X(96)80077-7
Vriend G WHAT IF: a molecular modeling and drug design program J Mol Graph 1990 8 52 56 2268628 10.1016/0263-7855(90)80070-V
Canutescu AA Shelenkov AA Dunbrack RL Jr A graph-theory algorithm for rapid protein side-chain prediction Protein Sci 2003 12 2001 2014 12930999 10.1110/ps.03154503
Repiso A Oliva B Vives Corrons JL Carreras J Climent F Glucose phosphate isomerase deficiency: enzymatic and familial characterization of Arg346His mutation Biochim Biophys Acta 2005 1740 467 471 15949716
de Atauri P Repiso A Oliva B Lluis Vives-Corrons J Climent F Carreras J Characterization of the first described mutation of human red blood cell phosphoglycerate mutase Biochim Biophys Acta 2005 1740 403 410 15949708
Andres AM Soldevila M Navarro A Kidd KK Oliva B Bertranpetit J Positive selection in MAOA gene is human exclusive: determination of the putative amino acid change selected in the human lineage Hum Genet 2004 115 377 386 15349769
Moult J Fidelis K Zemla A Hubbard T Critical assessment of methods of protein structure prediction (CASP): round IV Proteins 2001 2 7 11835476 10.1002/prot.10054
Venclovas C Comparative modeling of CASP4 target proteins: combining results of sequence search with three-dimensional structure assessment Proteins 2001 47 54 11835481 10.1002/prot.10008
van Vlijmen HW Karplus M PDB-based protein loop prediction: parameters for selection and methods for optimization J Mol Biol 1997 267 975 1001 9135125 10.1006/jmbi.1996.0857
Mehler EL Periole X Hassan SA Weinstein H Key issues in the computational simulation of GPCR function: representation of loop domains J Comput Aided Mol Des 2002 16 841 853 12825797 10.1023/A:1023845015343
Fan H Mark AE Refinement of homology-based protein structures by molecular dynamics simulation techniques Protein Sci 2004 13 211 220 14691236 10.1110/ps.03381404
Carrieri A Centeno NB Rodrigo J Sanz F Carotti A Theoretical evidence of a salt bridge disruption as the initiating process for the alpha1d-adrenergic receptor activation: a molecular dynamics and docking study Proteins 2001 43 382 394 11340655 10.1002/prot.1051
Gutierrez-de-Teran H Centeno NB Pastor M Sanz F Novel approaches for modeling of the A1 adenosine receptor and its agonist binding site Proteins 2004 54 705 715 14997566 10.1002/prot.10617
Laskowski RA MacArthur MW Moss DB Thornton JM PROCHECK: a program to check the stereochemical quality of protein structures J Appl Crystallogr 1993 26 283 291 10.1107/S0021889892009944
Sippl MJ Calculation of conformational ensembles from potentials of mean force. An approach to the knowledge-based prediction of local structures in globular proteins J Mol Biol 1990 213 859 883 2359125
Luthy R Bowie JU Eisenberg D Assessment of protein models with three-dimensional profiles Nature 1992 356 83 85 1538787 10.1038/356083a0
Sippl MJ Recognition of errors in three-dimensional structures of proteins Proteins 1993 17 355 362 8108378 10.1002/prot.340170404
Zheng QC Li ZS Xiao JF Sun M Zhang Y Sun CC Homology modeling and PAPS ligand (cofactor) binding study of bovine phenol sulfotransferase J Mol Model (Online) 2005
Aloy P Mas JM Marti-Renom MA Querol E Aviles FX Oliva B Refinement of modelled structures by knowledge-based energy profiles and secondary structure prediction: application to the human procarboxypeptidase A2 J Comput Aided Mol Des 2000 14 83 92 10702927 10.1023/A:1008197831529
Hansson T Oostenbrink C van Gunsteren W Molecular dynamics simulations Curr Opin Struct Biol 2002 12 190 196 11959496 10.1016/S0959-440X(02)00308-1
Scott WRP Hunenberger PH Mark AE Billeter SR Fennen J Torda AE Huber T Kruger P van Gunsteren WF The GROMOS biomiolecular simulation program package J Phys Chem A 1999 103 3596 3607 10.1021/jp984217f
Brooks BR Bruccoleri RE Olafson BD States DJ Swaminathan S Karplus M CHARMM – A program for macromolecular energy, minimization, and dynamics calculations J Comput Chem 1983 16 513 519
Pearlman DA Case DA Caldwell JW Ross WS Cheatham TE Debolt S Ferguson D Seib el G Collman P AMBER, a package of computer-programs for applying molecular mechanics,<normal-mode analysis, molecular-dynamics and free-energy calculations to simulate the structural and energetic properties of molecules Comput Phys Commun 1995 91 1 41 10.1016/0010-4655(95)00041-D
Tramontano A Morea V Assessment of homology-based predictions in CASP5 Proteins 2003 53 352 368 14579324 10.1002/prot.10543
Flohil JA Vriend G Berendsen HJ Completion and refinement of 3-D homology models with restricted molecular dynamics: application to targets 47, 58, and 111 in the CASP modeling competition and posterior analysis Proteins 2002 48 593 604 12211026 10.1002/prot.10105
Iverson GM Reddel S Victoria EJ Cockerill KA Wang YX Marti-Renom MA Sali A Marquis DM Krilis SA Linnik MD Use of single point mutations in domain I of beta 2-glycoprotein I to determine fine antigenic specificity of antiphospholipid autoantibodies J Immunol 2002 169 7097 7103 12471146
Feliu JX Benito A Oliva B Aviles FX Villaverde A Conformational flexibility in a highly mobile protein loop of foot-and-mouth disease virus: distinct structural requirements for integrin and antibody binding J Mol Biol 1998 283 331 338 9769208 10.1006/jmbi.1998.2104
Sathyanarayanan PV Siems WF Jones JP Poovaiah BW Calcium-stimulated autophosphorylation site of plant chimeric calcium/calmodulin-dependent protein kinase J Biol Chem 2001 276 32940 32947 11399751 10.1074/jbc.M009648200
Gloyn AL Odili S Zelent D Buettger C Castleden HA Steele AM Stride A Shiota C Magnuson MA Lorini R Insights into the structure and regulation of glucokinase from a novel mutation (V62M), which causes maturity-onset diabetes of the young J Biol Chem 2005 280 14105 14113 15677479 10.1074/jbc.M413146200
Matsui E Abe J Yokoyama H Matsui I Aromatic residues located close to the active center are essential for the catalytic reaction of flap endonuclease-1 from hyperthermophilic archaeon Pyrococcus horikoshii J Biol Chem 2004 279 16687 16696 14742430 10.1074/jbc.M313695200
Ishino T Pasut G Scibek J Chaiken I Kinetic interaction analysis of human interleukin 5 receptor alpha mutants reveals a unique binding topology and charge distribution for cytokine recognition J Biol Chem 2004 279 9547 9556 14662768 10.1074/jbc.M309327200
Valdar WS Thornton JM Conservation helps to identify biologically relevant crystal contacts J Mol Biol 2001 313 399 416 11800565 10.1006/jmbi.2001.5034
Villa-Freixa J Bonet J Khan AK Johnston M Evaluating the relationship between residue stability and enzyme active site preorganization in protein regulated enzymes Theor Chem Acc 2005
Yokoyama S Protein expression systems for structural genomics and proteomics Curr Opin Chem Biol 2003 7 39 43 12547425 10.1016/S1367-5931(02)00019-4
Carrio M Gonzalez-Montalban N Vera A Villaverde A Ventura S Amylod-like properties of bacterial inclusion bodies J Mol Biol 2005 347 1025 1037 15784261 10.1016/j.jmb.2005.02.030
Wodak SJ Mendez R Prediction of protein-protein interactions: the CAPRI experiment, its evaluation and implications Curr Opin Struct Biol 2004 14 242 249 15093840 10.1016/j.sbi.2004.02.003
Schneidman-Duhovny D Nussinov R Wolfson HJ Predicting molecular interactions in silico: II. Protein-protein and protein-drug docking Curr Med Chem 2004 11 91 107 14754428 10.2174/0929867043456223
Seri M Savino M Bordo D Cusano R Rocca B Meloni I Di Bari F Koivisto PA Bolognesi M Ghiggeri GM Epstein syndrome: another renal disorder with mutations in the nonmuscle myosin heavy chain 9 gene Hum Genet 2002 110 182 186 11935325 10.1007/s00439-001-0659-1
Almstedt K Lundqvist M Carlsson J Karlsson M Persson B Jonsson BH Carlsson U Hammarstrom P Unfolding a folding disease: folding, misfolding and aggregation of the marble brain syndrome-associated mutant H107Y of human carbonic anhydrase II J Mol Biol 2004 342 619 633 15327960 10.1016/j.jmb.2004.07.024
Kim SH Shining a light on structural genomics Nat Struct Biol 1998 5 643 645 9699614 10.1038/1334
Goldsmith-Fischman S Honig B Structural genomics: computational methods for structure analysis Protein Sci 2003 12 1813 1821 12930981 10.1110/ps.0242903
Montelione GT Anderson S Structural genomics: keystone for a Human Proteome Project Nat Struct Biol 1999 6 11 12 9886282 10.1038/4878
Edwards AM Arrowsmith CH Christendat D Dharamsi A Friesen JD Greenblatt JF Vedadi M Protein production: feeding the crystallographers and NMR spectroscopists Nat Struct Biol 2000 7 970 972 11104003 10.1038/80751
Brenner SE Target selection for structural genomics Nat Struct Biol 2000 7 967 969 11104002 10.1038/80747
CASP5. Proceedings of the 5th Meeting on the Critical Assessment of Techniques for Protein Structure Prediction. 1–5 December Asilomar, California, USA Proteins 2002 53 333 595
Becker OM Shacham S Marantz Y Noiman S Modeling the 3D structure of GPCRs: advances and application to drug discovery Curr Opin Drug Discov Devel 2003 6 353 361 12833668
|
15989689
|
PMC1183244
|
CC BY
|
2021-01-04 16:40:08
|
no
|
Mol Pain. 2005 Jun 30; 1:21
|
latin-1
|
Mol Pain
| 2,005 |
10.1186/1744-8069-1-21
|
oa_comm
|
==== Front
Nutr JNutrition Journal1475-2891BioMed Central London 1475-2891-4-211596975910.1186/1475-2891-4-21ResearchCalcium-fortified beverage supplementation on body composition in postmenopausal women Haub Mark D [email protected] Tammy R [email protected] Chad M [email protected] Valentina M [email protected] Enas K [email protected] Carol Ann [email protected] Department of Human Nutrition, Kansas State University, Manhattan, KS, USA2 Galichia Center on Aging, Kansas State University, Manhattan, KS, USA2005 21 6 2005 4 21 21 16 2 2005 21 6 2005 Copyright © 2005 Haub et al; licensee BioMed Central Ltd.2005Haub et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
We investigated the effects of a calcium-fortified beverage supplemented over 12 months on body composition in postmenopausal women (n = 37, age = 48–75 y).
Methods
Body composition (total-body percent fat, %FatTB; abdominal percent fat, %FatAB) was measured with dual energy x-ray absorptiometry. After baseline assessments, subjects were randomly assigned to a free-living control group (CTL) or the supplement group (1,125 mg Ca++/d, CAL). Dietary intake was assessed with 3-day diet records taken at baseline and 12 months (POST). Physical activity was measured using the Yale Physical Activity Survey.
Results
At 12 months, the dietary calcium to protein ratio in the CAL group (32.3 ± 15.6 mg/g) was greater than the CTL group (15.2 ± 7.5 mg/g). There were no differences from baseline to POST between groups for changes in body weight (CAL = 0.1 ± 3.0 kg; CTL = 0.0 ± 2.9 kg), %FatTB (CAL = 0.0 ± 2.4%; CTL = 0.5 ± 5.4%), %FatAB (CAL = -0.4 ± 8.7%; CTL = 0.6 ± 8.7%), or fat mass (CAL = 1.3 ± 2.6 kg; CTL = 1.3 ± 2.7 kg).
Conclusion
These results indicate that increasing the calcium to protein ratio over two-fold by consuming a calcium-fortified beverage for 12 months did not decrease body weight, body fat, or abdominal fat composition in postmenopausal women.
agingolderascorbic acidjuiceobesity.
==== Body
Background
According to results and concepts from Zemel et al. [1] and epidemiologic studies [2-4], dietary calcium plays a role in regulating body composition. The epidemiological data demonstrate an inverse association between calcium intake and body weight and body fat mass [2,3,5]. Davies et al. [2] derived a prediction equation using the ratio of intakes of calcium (mg/day) and protein (g/day) (calcium:protein ratio) to estimate annual changes in body weight. The benefit of the calcium:protein ratio is that it tends to correct inaccuracies of self-reported dietary intakes of the individual nutrients [2]. Zemel et al. [1] reported a significant decrease in body fat in African-American males with Type 2 diabetes when they supplemented their diets with 600 mg of calcium from dairy products. In a different randomized clinical trial, Zemel [6] illustrated that those with a higher calcium intake compared to those consuming a calcium deficient (400–500 mg calcium/day) diet had less body fat. While these studies are valuable in demonstrating associations and the effects of deficient diets, there is limited data illustrating the effect of calcium supplementation on body composition in those on self-selected free-living diets. A retrospective study by Shapses et al. [7] observed that calcium supplementation over six months did not increase weight loss or fat loss in postmenopausal women consuming a hypoenergetic diet. Likewise, Barr et al. [8] observed that milk-supplementation (three servings/day) did not elicit weight loss or improve metabolic risk factors in older adults. Given the discrepancy between studies, the effect of calcium intake on body weight and body composition remains inconclusive.
Most studies have imposed calcium supplementation with either pills or dairy products; however, some consumers choose to supplement their calcium via calcium-fortified beverages. In support of this supplementation option, one of the more bioavailable forms of calcium, calcium citrate malate (CCM) [9], is available in commercially available calcium-fortified fruit juice. Previous research indicates that consumption of fruit juice fortified with CCM results in significantly greater calcium absorption than milk or calcium carbonate [9,10]. Thus, CCM may have a better likelihood of eliciting a metabolic effect. Limited, if any, data exist demonstrating the efficacy of calcium-fortified juices at attenuating fat accretion. Therefore, the purpose of this investigation was to determine the effectiveness of increasing the calcium:protein ratio using a beverage fortified with calcium (1,125 mg Ca++/d of CCM) at decreasing body weight and body fat in free-living postmenopausal women.
Methods
Study design and supplementation
The study was approved by the Kansas State University Institutional Review Board, and written informed consent was obtained from each subject. Subjects were recruited through newspaper advertisements, word of mouth, and flyers posted on the Kansas State University campus and in the surrounding community. Respondents were screened for eligibility during a telephone interview. Women were eligible for the study if they were between 45 and 75 years of age and postmenopausal. Menopause was defined as one year from the date of the last menstrual period, or the date of operation for surgical menopause. Exclusion criteria included current diagnosis of chronic disease; smoking; fracture within past year; beginning or ending hormone replacement therapy within six months; and taking any of the following medications known to affect bone metabolism: bisphosphonates, thiazides, corticosteroids, calcitonin, or tamoxifen.
Forty-one healthy postmenopausal women were enrolled, and 37 completed the study (age = 60 ± 7 yr; wt = 69.5 ± 9.9 kg; BMI = 26.1 ± 3.7; years post-menopause = 12.5 ± 9.6 yr). The subjects were randomly assigned, with stratification according to body composition, to either a supplement group (CAL, n = 17) or a free-living comparison group (CTL; n = 20). Subjects in the CAL group consumed 591 ml/day of a beverage containing milk (7%) and fruit juice (15%) (Cal-C™, Nutrijoy, Inc., Manhattan, KS). They were asked to consume half the beverage in the morning the remainder later in the day. Control subjects were asked to continue their usual lifestyle (diet and daily activities). Four subjects in the CAL group withdrew before completion of the study, two due to unwillingness to continue daily supplementation and two due to personal reasons not related to the study. Subjects (n = 2) who travelled during the year for an extended period (two to three weeks) were given the option to carry the beverages with them or take two calcium citrate pills (600 mg calcium per pill) per day in place of the beverage. Compliance was 97%, as measured by weekly consumption reports, which subjects turned in when they picked up their drinks each week. Reasons for missing included: gastrointestinal discomfort, forgetfulness, flu, family emergency, and surgery.
Measurements
At baseline and after the intervention (POST), the subjects reported to the laboratory for height, weight, and dual-energy x-ray absorptiometry (DXA) measurements; to turn in 3-day diet records; and to complete the Yale Physical Activity Survey (YPAS) [11]. Total body composition (body fat percentage, %BFTB; fat mass, FM; and non-bone fat-free mass, FFM) was measured via DXA (v5.6, GE Lunar Corp., Milwaukee, WI). Abdominal percentage fat (%BFAb) was determined by using a customized region of the abdomen (L2 to the iliac crest) as previously described [12]. The same technician analyzed baseline and POST assessments and was blinded to treatment. All subjects were given oral and written instructions for the completion of the diet records at baseline and 12 months. The 3-d diet records (2 weekdays and 1 weekend day) were analyzed using commercially available software (Nutritionist Pro™, v 2.0, First DatBank, Inc., San Bruno, CA). One CTL group subject did not complete a baseline diet record. Physical activity was estimated by using the YPAS [11], which yields an index score that has been associated with VO2max in the elderly and is related to energy expenditure measured via doubly labelled water [13].
Statistical analyses
A general linear model analysis of variance with repeated measures (group by time) was used to determine interaction and main effects, and independent and paired t-tests were used to compare differences at each testing period. Significance was set at p < 0.05. All analyses were run using SPSS for Windows software (version 11.5; SPSS Inc., Chicago, IL). Post hoc Pearson product moment correlations were used to determine the relationship between nutrient intake, nutrient ratios, and changes in body composition. Statistical software (Axum, version 7.0, MathSoft, Cambridge, MA) was used to calculate statistical power. All values are presented as mean ± SD.
Results
Body composition
There were no differences between groups at baseline for any of the body-composition variables (Table 1). There was a main effect of time for body fat. At POST in the CTL group, the non-bone fat-free mass was decreased compared to baseline.
Table 1 Whole-body Body composition measures of participants at baseline and 12 months, with (CAL) and without (CTL) calcium supplementation.
Baseline POST ρ
Body Weight (kg)
CAL 69.6 ± 11.5 70.2 ± 12.6 0.5 ± 3.3
CTL 69.3 ± 8.7 69.3 ± 8.3 0.0 ± 2.9
Whole-body fat (%)
CAL 43.0 ± 7.3 43.1 ± 6.7 0.0 ± 2.3
CTL 43.1 ± 5.5 43.1 ± 5.4 0.1 ± 2.4
Abdominal fat (%)
CAL 45.3 ± 9.7 44.7 ± 8.7 -0.6 ± 3.4
CTL 44.0 ± 8.1 44.2 ± 8.3 0.2 ± 3.7
Fat Mass (kg)#
CAL 29.3 ± 9.2 30.9 ± 9.9 1.6 ± 2.8
CTL 29.0 ± 7.1 30.1 ± 6.7 1.3 ± 2.7
Fat-free Mass (kg)
CAL 39.1 ± 2.9 39.2 ± 2.7 0.1 ± 1.3
CTL 40.0 ± 3.7 39.6 ± 4.0* -0.4 ± 0.7
Bone Mineral Content (kg)
CAL 2.5 ± 0.3 2.5 ± 0.4 0.0 ± 0.1
CTL 2.4 ± 0.3 2.4 ± 0.3 0.0 ± 0.1
Mean ± SD; Abdominal fat = % fat from custom region of interest of the abdomen; 1475-2891-4-21-i1.gif"/> = absolute difference between POST and baseline; #= main effect for time. * = within group difference from baseline to post.
Dietary intake and physical activity
Total energy, carbohydrate, and calcium intakes, and the calcium:protein ratio (mg/g), increased significantly compared with baseline values in the CAL group (Table 2). At POST, dietary calcium and the calcium:protein ratio were greater in the CAL group than in the CTL group. There was also a difference in the delta values between groups for carbohydrate and calcium intakes and in the calcium:protein ratio. The CTL group did not experience any significant changes in dietary calcium intake over time. There was no difference between or within groups for reported physical activity, as measured by the YPAS index, at baseline (CAL = 40 ± 18; CTL = 51 ± 25) or POST (CAL = 40 ± 14; CTL = 49 ± 23).
Table 2 Dietary intake of total energy, macronutrients and calcium values as reported using 3-day diet records.
Baseline POST ρ1475-2891-4-21-i1.gif"/>
Energy Intake (MJ/d)
CAL 6.88 ± 1.28 7.89 ± 1.77† 1.0 ± 1.8
CTL 7.96 ± 2.50 7.96 ± 2.16 -0.0 ± 1.7
Carbohydrate (g/d)
CAL 219 ± 40 270 ± 65†† 51 ± 48*
CTL 239 ± 60 243 ± 69 1 ± 63
Protein (g/kg/d)
CAL 0.97 ± 0.2 1.09 ± 0.5 0.1 ± 0.3
CTL 1.10 ± 0.3 1.07 ± 0.4 -0.0 ± 0.5
Carbohydrate:Protein Ratio (g/g/d)
CAL 3.4 ± 0.8 3.9 ± 0.8 1.2 ± 0.4
CTL 3.3 ± 0.9 3.5 ± 1.2 1.1 ± 0.4
Fat (g/d)
CAL 59 ± 20 62 ± 16 3 ± 18
CTL 72 ± 39 72 ± 34 0 ± 26
Calcium (mg/d)‡#
CAL 919 ± 435 2201 ± 763** † 1281 ± 824**
CTL 1244 ± 734 1153 ± 647 -90 ± 567
Calcium:Protein Ratio (mg:g/d)‡#
CAL 14.6 ± 8.2 32.3 ± 15.6** † † 17.7 ± 15.8**
CTL 16.8 ± 8.7 15.2 ± 7.5 -1.0 ± 7.5
Calcium:Energy (mg:MJ/d)‡#
CAL 134 ± 67 286 ± 105** † † 152 ± 108**
CTL 162 ± 93 144 ± 93 -18 ± 78
Mean ± SD; 1475-2891-4-21-i1.gif"/> = absolute difference between POST and baseline; 1475-2891-4-21-i2.gif"/> = fold change from baseline to POST; ‡ = main effect for time; #= main effect for diet; † p < 0.05, ††p < 0.001 different from zero months within group; * p < 0.05, ** p < 0.01, difference between CAL and CTL groups.
Discussion
The primary finding of this study was that calcium supplementation (1,125 mg/d) did not decrease body weight or fat in the CAL group. The present results do not support the notion that increasing calcium intake leads to decreased body weight in apparently healthy older women.
The results from the present 12 month study differ from those reported by Zemel et al. [1]. The calcium supplement (600 mg Ca++/d) in the Zemel et al. study [1] was an unsweetened dairy product (yogurt), whereas the calcium from this study was suspended in fruit juice. Also, the men in the study by Zemel et al. [1] were obese and diagnosed with Type 2 diabetes; and, they consumed less than 500 mg Ca++/d, which is indicative of an unhealthy diet [14]. The women in the present study were apparently healthy, most were overweight or obese, and consumed almost twice the amount of calcium per day than the men in Zemel's study [1].
An explanation for a lack of change in body composition in the present study might be that levels of serum calcium and 1,25-dihydroxyvitamin D (1,25-(OH)2 D) did not change with the intervention as would likely occur with calcium deficient individuals. This is critical, as the transgenic animal model suggests that changes in serum calcium and 1,25-(OH)2 D are what drive the upregulation of lipolysis and inhibit lipogenesis in adipocytes via alterations in intracellular calcium concentrations and fatty acid synthesis mRNA expression [15]. In the dairy supplementation study by Gunther et al. [16] no changes were observed for 1,25-(OH)2 D or parathyroid hormone following increased calcium intake from dairy sources (1,000 – 1,400 mg Ca++/d).
Another explanation might be that the low carbohydrate:protein ratio of the dairy product played a role in the effectiveness of the dairy products to decrease body fat by 4.9 kg in 12 months. Layman et al. [17] have demonstrated that the carbohydrate:protein ratio plays an important role in weight loss, with a lower ratio being preferred. The macronutrient composition of the present calcium supplement may be important since the carbohydrate:protein ratio of milk is ~1.3 (12 g carbohydrate and 9 g of protein), whereas the ratio of the supplement used in this study was ~25 (25 g carbohydrate and <1 g protein per serving). However, this theory is not supported by results from Gunther et al. [16], which illustrate that one year of calcium supplementation, via dairy products, did not change body composition in young women. Thus, in healthy younger women, the use of a low carbohydrate:protein ratio means of supplementing calcium was not efficacious.
In support of the present data, it was noted in a review by Barr [18] that only one out of 17 studies reviewed observed decreased body weight during calcium supplementation. Moreover, in a short-term clinical trial by Barr et al. [8], healthy older adults were randomly assigned to control group or a group that consumed three servings of milk (skim or 1% fat) per day for 12 weeks. They reported a significant interaction effect (treatment by time) with the milk-consuming group gaining 0.6 kg more weight than the control group.
What was somewhat surprising was the amount of calcium habitually consumed by the women in the present study at baseline (1,112 ± 630 mg calcium/d), compared with the reported intakes of older women from other studies [19,20]; however, an upper limit or threshold of calcium intake has not been suggested relative to effects on body composition. This higher-than-expected calcium intake might reflect the socioeconomic and/or educational background [21] of the women in the present study, a factor we did not assess.
Also, even with a two-fold increase in the calcium:protein ratio there were no changes in any measure of body composition. Based on the prediction equation by Davies et al. [2], the CAL group should have lost 0.49 kg of body weight and the control group should have gained 0.14 kg. This expected decrease in body weight may have been prevented by the fact the CAL group reported a significant increase in energy intake at POST compared with baseline. However, it is difficult to know whether the 3-day diet records and the YPAS accurately reflect the energy intake and expenditure patterns over the 12 months of this study, or if the calcium-fortified beverage attenuated the expected decrease in body fat via the high carbohydrate:protein ratio of the drink. Regardless, these results illustrate that merely increasing the calcium:protein ratio alone is not enough to change body weight or body fat as predicted or suggested by previous epidemiological and animal studies.
The strengths and novelty of this randomized clinical study were: 1) the efficacious nature of the design; 2) the use of a non-dairy calcium-fortified beverage; 3) the use of the calcium:protein ratio as a dependent variable relative to changes in body composition; and 4) this was one of the longer calcium supplementation studies specifically reporting changes in abdominal adiposity. As stated by Barr [18], there is limited published data from randomized trials that have investigated the effect of calcium supplementation on body composition. Contrary to the epidemiological and animal studies, this study clearly demonstrates that a dramatic change in the calcium:protein does not ubiquitously decrease body weight, fat weight, or percent body fat over a 12 month period in apparently healthy free-living women as predicted. That said, given the reported increase in energy intake over three days, the increased calcium intake may have attenuated a potential increase in body weight and/or body fat. Future studies utilizing controlled diets to maintain consistent macronutrient intakes are needed to insure minimal changes in other nutrients occur and to establish whether effects observed in animal studies occur in healthy humans.
Conclusion
In conclusion, 12 months of supplementation with a calcium-fortified beverage seems to have no effect on body composition in free-living postmenopausal women who already consume higher-than-expected levels of calcium. Thus, significantly increasing the calcium:protein ratio alone did not decrease body weight or body fat in these free-living postmenopausal women.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
TRS and MDH worked together to write the first draft of the manuscript. TRS was responsible for subject recruitment and baseline data collection. CMC, EKA, and MDH assisted with baseline data collection and performed data collection and analysis at the end of the study. VMR and CAH helped develop the research design and edited the manuscript.
Acknowledgements
This project was supported by a grant from the Kansas Health and Nutrition Fund provided by the Kansas Attorney General's Office. The calcium supplements were provided by NutriJoy®, Inc (Manhattan, KS). Manuscript editing was supported by the Kansas Agriculture Experiement Station (Contribution no. 04-337-J). We would like to thank Elizabeth Adams for her assistance with data collection and our subjects for their willingness to participate and commitment to complete this project.
==== Refs
Zemel MB Shi H Greer B Dirienzo D Zemel PC Regulation of adiposity by dietary calcium Faseb J 2000 14 1132 1138 10834935
Davies KM Heaney RP Recker RR Lappe JM Barger-Lux MJ Rafferty K Hinders S Calcium intake and body weight J Clin Endocrinol Metab 2000 85 4635 4638 11134120 10.1210/jc.85.12.4635
Heaney RP Davies KM Barger-Lux MJ Calcium and weight: clinical studies J Am Coll Nutr 2002 21 152S 155S 11999544
Heaney RP Normalizing calcium intake: projected population effects for body weight J Nutr 2003 133 268S 270S 12514306
Parikh SJ Yanovski JA Calcium intake and adiposity Am J Clin Nutr 2003 77 281 287 12540383
Zemel MB Thompson W Milstead A Morris K Campbell P Calcium and dairy acceleration of weight and fat loss during energy restriction in obese adults Obes Res 2004 12 582 590 15090625
Shapses SA Heshka S Heymsfield SB Effect of calcium supplementation on weight and fat loss in women J Clin Endocrinol Metab 2004 89 632 637 14764774 10.1210/jc.2002-021136
Barr SI McCarron DA Heaney RP Dawson-Hughes B Berga SL Stern JS Oparil S Effects of increased consumption of fluid milk on energy and nutrient intake, body weight, and cardiovascular risk factors in healthy older adults J Am Diet Assoc 2000 100 810 817 10916520 10.1016/S0002-8223(00)00236-4
Smith KT Heaney RP Flora L Hinders SM Calcium absorption from a new calcium delivery system (CCM) Calcif Tissue Int 1987 41 351 352 3124946
Miller JZ Smith DL Flora L Slemenda C Jiang XY Johnston CCJ Calcium absorption from calcium carbonate and a new form of calcium (CCM) in healthy male and female adolescents Am J Clin Nutr 1988 48 1291 1294 3189218
Dipietro L Caspersen CJ Ostfeld AM Nadel ER A survey for assessing physical activity among older adults Med Sci Sports Exerc 1993 25 628 642 8492692
Park YW Heymsfield SB Gallagher D Are dual-energy X-ray absorptiometry regional estimates associated with visceral adipose tissue mass? Int J Obes Relat Metab Disord 2002 26 978 983 12080453 10.1038/sj.ijo.0801982
Starling RD Matthews DE Ades PA Poehlman ET Assessment of physical activity in older individuals: a doubly labeled water study J Appl Physiol 1999 86 2090 2096 10368377
Barger-Lux MJ Heaney RP Packard PT Lappe JM Recker RR Nutritional correlates of low calcium intake Clin Appl Nutr 1992 2 39 44
Zemel MB Regulation of adiposity and obesity risk by dietary calcium: mechanisms and implications J Am Coll Nutr 2002 21 146S 151S 11999543
Gunther CW Legowski PA Lyle RM McCabe GP Eagan MS Peacock M Teegarden D Dairy products do not lead to alterations in body weight or fat mass in young women in a 1-y intervention Am J Clin Nutr 2005 81 751 756 15817848
Layman DK Boileau RA Erickson DJ Painter JE Shiue H Sather C Christou DD A reduced ratio of dietary carbohydrate to protein improves body composition and blood lipid profiles during weight loss in adult women J Nutr 2003 133 411 417 12566476
Barr SI Increased dairy product or calcium intake: is body weight or composition affected in humans? J Nutr 2003 133 245S 248S 12514301
Devine A Dick IM Heal SJ Criddle RA Prince RL A 4-year follow-up study of the effects of calcium supplementation on bone density in elderly postmenopausal women Osteoporos Int 1997 7 23 28 9102058 10.1007/BF01623455
Storm D Eslin R Porter ES Musgrave K Vereault D Patton C Kessenich C Mohan S Chen T Holick MF Rosen CJ Calcium supplementation prevents seasonal bone loss and changes in biochemical markers of bone turnover in elderly New England women: a randomized placebo-controlled trial J Clin Endocrinol Metab 1998 83 3817 3825 9814452 10.1210/jc.83.11.3817
Pfister AK Wul JTJ Saville PD Factors determining calcium intake in elderly women of Appalachia South Med J 2001 94 1006 1012 11702812
|
15969759
|
PMC1183245
|
CC BY
|
2021-01-04 16:39:31
|
no
|
Nutr J. 2005 Jun 21; 4:21
|
utf-8
|
Nutr J
| 2,005 |
10.1186/1475-2891-4-21
|
oa_comm
|
==== Front
Popul Health MetrPopulation Health Metrics1478-7954BioMed Central London 1478-7954-3-81604277710.1186/1478-7954-3-8ResearchPsychometric properties of the CDC Symptom Inventory for assessment of Chronic Fatigue Syndrome Wagner Dieter [email protected] Rosane [email protected] Christine [email protected] James F [email protected] Elizabeth R [email protected] William C [email protected] Division of Viral and Rickettsial Diseases, National Center for Infectious Diseases, Centers for Disease Control and Prevention, 1600 Clifton Rd, Mail Stop A15, Atlanta, GA, USA2 St Michaels Hospital, Inner City Health Research Unit, Toronto, Canada3 Department of Psychiatry and Behavioral Sciences, Emory University School of Medicine, Atlanta, GA, USA2005 22 7 2005 3 8 8 14 1 2005 22 7 2005 Copyright © 2005 Wagner et al; licensee BioMed Central Ltd.2005Wagner et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Objectives
Validated or standardized self-report questionnaires used in research studies and clinical evaluation of chronic fatigue syndrome (CFS) generally focus on the assessment of fatigue. There are relatively few published questionnaires that evaluate case defining and other accompanying symptoms in CFS. This paper introduces the self-report CDC CFS Symptom Inventory and analyzes its psychometric properties.
Methods
One hundred sixty-four subjects (with CFS, other fatiguing illnesses and non fatigued controls) identified from the general population of Wichita, Kansas were enrolled. Evaluation included a physical examination, a standardized psychiatric interview, three previously validated self-report questionnaires measuring fatigue and illness impact (Medical Outcomes Survey Short-Form-36 [MOS SF-36], Multidimensional Fatigue Inventory [MFI], Chalder Fatigue Scale), and the CDC CFS Symptom Inventory. Based on theoretical assumptions and statistical analyses, we developed several different Symptom Inventory scores and evaluated them on their ability to differentiate between participants with CFS and non-fatigued controls.
Results
The Symptom Inventory had good internal consistency and excellent convergent validity. A Total score (all symptoms), Case Definition score (CFS case defining symptoms) and Short Form score (6 symptoms with minimal correlation) differentiated CFS cases from controls. Furthermore, both the Case Definition and Short Form scores distinguished people with CFS from fatigued subjects who did not meet criteria for CFS.
Conclusion
The Symptom Inventory appears to be a reliable and valid instrument to assess symptoms that accompany CFS. It is a positive addition to existing instruments measuring fatigue because it allows other dimensions of the illness to be assessed. Further research is needed to confirm and replicate the current findings in a normative population.
==== Body
1. Background
Chronic fatigue syndrome (CFS) is an incapacitating illness defined by disabling chronic fatigue and characteristic accompanying symptoms [1]. CFS has no confirmatory physical signs or characteristic laboratory abnormalities and its etiology and pathophysiology remain unknown [2]. Many of the problems searching for markers of CFS reflect ambiguities of definition. Recently, an International CFS Study Group identified areas of vagueness in the case definition and proposed revisions in its utilization [3]. The Group recommended that validated standardized instruments be used to evaluate the functional impairment, fatigue, and accompanying symptoms associated with CFS. They recommended that future research utilize the Medical Outcomes Survey Short-Form-36 (SF-36) [4] to assess functional impairment and that the Checklist Individual Strength (CIS) [5] or Multidimensional Fatigue Inventory (MFI) [6] be used to measure general dimensions of fatigue. The Group was unaware of standardized and validated instruments that assessed the CFS symptom complex and suggested that investigators consider using a symptom checklist developed by the US Centers for Disease Control and Prevention (CDC) for use in population-based surveys. The CDC Symptom Inventory assesses the full range of CFS associated symptoms, has been used in several population-based studies (i.e., comparative data are available), and is publicly available. However, the Symptom Inventory has not been formally validated. The objective of the present study was to validate the Symptom Inventory as a questionnaire for evaluation of CFS associated symptoms.
Methods
Study design/subjects
This study adhered to human experimentation guidelines of the U.S. Department of Health and Human Services and the Helsinki Declaration. The CDC Human Subjects committee approved study protocols. All participants were volunteers who gave informed consent.
This study enrolled subjects who had previously participated in the 1997 through 2000 Wichita CFS Surveillance Study [7,8]. In brief, the Surveillance Study used a random-digit-dialing telephone survey to screen 56,146 adult residents (18 to 69 years of age) of Wichita, Kansas. A surveillance cohort of 3,528 adults who reported fatigue of at least 1-month duration and 3,634 non-fatigued persons completed a detailed telephone interview and eligible subjects were clinically evaluated to assess CFS. A subset of the cohort was followed at 12-, 24-and 36-months with a telephone interview and clinical evaluation. Fatigued participants in the present study were a subset of the 659 fatigued adults identified during surveillance who were classified as CFS by 1994 research case definition criteria [1] or unexplained chronic fatigue not meeting criteria for CFS (ISF). Non-fatigued controls were randomly selected from the cohort who participated in all telephone interviews at baseline, 12-, 24-, and 36-month follow-up periods, who had never reported fatigue of at least 1-month duration, and who had never been identified with medical or psychiatric conditions exclusionary for CFS.
Clinical Measurements
People who agreed to participate were admitted to a Wichita hospital research unit for 2-days. Upon arrival, they provided a standardized past medical history, a review of current medications; they then underwent a brief standardized physical examination and standardized psychiatric evaluation of Axis I disorders (DIS) [9] that exclude classification of CFS (melancholic major depression, bipolar disorder, psychosis, substance abuse, eating disorders) [1,3]. We also collected blood and urine for routine analysis, including: a complete blood count with differential, C-reactive protein, alanine aminotransferase, albumin, alkaline phosphatase, aspartate aminotransferase, total bilirubin, calcium, carbon dioxide, chloride, creatinine, glucose, potassium, total protein, sodium, urea nitrogen BUN, pregnancy test, TSH, free T-4, and urinalysis. In addition they completed the CDC Symptom Inventory and 3 well standardized and validated self-administered questionnaires (described below). Subjects were classified based on results of laboratory, physical and psychologic examination, fatigue and symptoms at the time of the study as: CFS (currently meeting the 1994 CFS research case definition) [1]; ISF (currently unexplained fatigue but not meeting CFS criteria); remission (prior CFS or ISF but currently not fatigued); never fatigued (non-fatigued controls); or, excluded (missing data, laboratory or psychological abnormalities).
The Medical Outcomes Survey Short-Form (SF-36) [4] (QualityMetric Incorporated, Lincoln, Rhode Island) assesses function and well-being in 8 areas: 1) limitations in physical activities because of health problems; 2) limitations in social activities because of physical or emotional problems; 3) limitations in usual role activities because of physical health problems; 4) bodily pain; 5) general mental health; 6) limitations in usual role activities because of emotional problems; 7) vitality (energy and fatigue); and 8) general health perceptions. Scores in each area reflect ability to function (higher values being better).
The Multidimensional Fatigue Inventory (MFI) [6] is a 20-item self-report instrument that measures 5 dimensions of fatigue; General Fatigue, Physical Fatigue, Mental Fatigue, Reduced Motivation and Reduced Activity. The score in each dimension reflects severity of fatigue (higher values being worse). The MFI has been primarily used to measure fatigue in cancer patients [10].
The Chalder Fatigue Scale (Chalder) [11] includes 10 questions measuring physical and mental fatigue. We added 2 questions regarding muscle pain: "Do your muscles hurt at rest?" and "Do your muscles hurt after exercise?".
CDC CFS Symptom Inventory
The Symptom Inventory collects information about the presence, frequency, and intensity of 19 fatigue and illness-related symptoms during the month preceding the interview; these include all 8 CFS-defining symptoms (post-exertional fatigue, unrefreshing sleep, problems remembering or concentrating, muscle aches and pains, joint pain, sore throat, tender lymph nodes and swollen glands, and headaches). It also catalogues diarrhea, fever, chills, sleeping problems, nausea, stomach or abdominal pain, sinus or nasal problems, shortness of breath, sensitivity to light, and depression. Perceived frequency of each symptom was rated on a four-point scale (1 = a little of the time, 2 = some of the time, 3 = most of time, 4 = all of the time), and severity or intensity of symptoms was measured on a three-point scale (1 = mild, 2 = moderate, 3 = severe).
Symptom Inventory Scoring
To summarize the degree of distress associated with each symptom, individual symptom scores were calculated by multiplying the frequency score by the intensity score. We transformed the intensity scores into equidistant scores before multiplication (i.e., 0 = symptom not reported 1 = mild, 2.5 = moderate, 4 = severe) resulting in range 0–16 for each symptom. We calculated a Total score for each person by summing the 19 individual symptom scores (possible range from 0 to 304). We also defined a Case Definition score as the sum of the 8 individual CFS case-definition symptom scores and an Other Symptoms score by considering only the 11 non-CFS symptoms.
Short Form of the CDC Symptom Inventory
We also explored the possibility of deriving a shorter version of the Symptom Inventory that would be a reliable and economic screening instrument. We created a Short Form of the Symptom Inventory by consecutively eliminating those symptoms whose scores had a corrected item-total score correlation < 0.60. The Short Form retained 6 symptoms: unusual fatigue after exertion, unrefreshing sleep, muscle aches, sleeping problems, problems with memory, and problems with concentration.
Statistical Analyses
To evaluate the internal consistency of the MFI and the Symptom inventory, we performed reliability analyses based on the model of averaging the inter-item correlation. Pearson's correlation coefficients between the CDC Symptom Inventory, MFI, Chalder Fatigue Scale and SF-36 were determined to evaluate convergent validity. We assessed construct validity by using one-way analyses of variance and Bonferroni post-hoc group comparisons to compare the CDC Symptom Inventory scores across the fatigue groups. We compared the Short and Total Forms with respect to psychometric properties (i.e., internal consistency and validity). The practicability of both scores was compared with three more intuitive scores derived from the Symptom Inventory; sum of all 19 individual frequency scores (Frequency Score), sum of all 19 individual intensity scores (Intensity Score), and the number of reported symptoms.
Results
Two hundred twenty-seven people participated in the 2-day clinical evaluation and participation rates (64 to 78%) were similar among the categories (p = .26). Five of the 227 were excluded because of incomplete psychiatric interviews, 29 because of exclusionary medical conditions, 3 because of psychiatric conditions, and 26 because of current major depression disorder with melancholic features, resulting in 164 subjects for analysis. Twenty-four subjects with prior fatigue (7 CFS and 17 ISF) were classified as in remission. There were no differences in age, body mass index, or sex between the classification groups (Table 1).
Table 1 Characteristics by subject classification (N = 164). BMI is body mass index, CFS includes subjects with chronic fatigue syndrome, ISF includes those with unexplained chronic fatigue not meeting criteria for CFS, and NF are never fatigued controls
N Women (%) Age (Mean ± SD) BMI (Mean ± SD)
Classification
CFS 52 44 (84.6%) 49.9 ± 7.9 28.4 ± 5.1
ISF 40 29 (72.5%) 49.7 ± 9.3 28.7 ± 4.7
Remission 24 17 (70.8%) 51.2 ± 9.1 28.8 ± 5.1
Never Fatigued 48 41 (85.4%) 50.3 ± 8.5 28.9 ± 5.1
Reliability analyses
Reliability analyses revealed good internal consistency for the reduced motivation subscale of the MFI and excellent internal consistency for the other four subscales. Cronbach's alpha coefficients were 0.89 for general fatigue, 0.82 for physical fatigue, 0.90 for reduced activity, 0.77 for reduced motivation, and 0.92 for mental fatigue. These findings are similar to those of Smets and colleagues [6].
The Symptom Inventory Total score also reflected excellent internal consistency, with a Cronbach's alpha coefficient of 0.88: Cronbach's alpha was 0.87 for the Symptom Inventory Short-Form. Table 2 shows the corrected item-total correlations (product terms) of all symptoms for the Total score and Short Form. In addition, reliability analyses revealed a Cronbach's alpha of 0.82 for the Case Definition score and 0.74 for the Other Symptoms score. Table 3 shows the descriptive data for the Total, Case Definition, Short Form, and the Other Symptoms scores.
Table 2 Corrected item to total correlations for the Symptom Inventory Total Score and the Symptom Inventory Short-Form Score
Symptom Corrected item to total correlations
Total score Short-form score
Sore throat .43
Tender nodes .48
Diarrhea .37
Unusual fatigue after exertion .69 .64
Muscle aches .70 .64
Joint pain .54
Feverishness .28
Chills .52
Unrefreshing sleep .77 .79
Sleeping problems .65 .70
Headaches .43
Memory problems .62 .66
Concentration .59 .67
Nausea .40
Stomach pain .32
Sinus problems .51
Shortness of breath .41
Sensitivity to light .41
Depression .52
Table 3 Descriptive data of the CDC Symptom Inventory Scores
CDC Symptom Inventory Scores Mean SD Min Max
Total 36.22 33.87 0 153.50
Short-form 19.51 20.01 0 96
CDC Case definition 21.21 21.54 0 102
Other symptoms 15.02 14.11 0 62
Validity
Convergent validity of Symptom Inventory
The Total, Case Definition and Short Form scores all had good convergent validity as determined by correlations with the MFI, Chalder Fatigue Scale, and SF-36 subscales (Table 4). As expected, high scores on the Symptom Inventory correlated with high fatigue scores and low levels of function. In other words, subjects who scored high on questionnaires assessing fatigue and low on those assessing functioning and well being, generally had high Symptom Inventory scores.
Table 4 Pearson's correlation matrix of CDC Symptom Inventory scores and MFI, Chalder Fatigue Scale, and SF-36 subscales (N = 164)
Total score Short-form Case definition score
Questionnaires r P r P r P
MFI
General fatigue .64 < .001 .67 < .001 .63 < .001
Physical fatigue .60 < .001 .62 < .001 .62 < .001
Reduced activity .60 < .001 .62 < .001 .57 < .001
Reduced motivation .53 < .001 .53 < .001 .50 < .001
Mental fatigue .54 < .001 .56 < .001 .54 < .001
Chalder Fatigue Scale .74 < .001 .76 < .001 .75 < .001
SF-36
Physical functioning -.58 < .001 -.56 < .001 -.60 < .001
Role-physical -.64 < .001 -.56 < .001 -.62 < .001
Bodily pain -.67 < .001 -.56 < .001 -.68 < .001
General health -.59 < .001 -.59 < .001 -.60 < .001
Vitality -.68 < .001 -.69 < .001 -.67 < .001
Social functioning -.66 < .001 -.62 < .001 -.63 < .001
Role-emotional -.39 < .001 -.40 < .001 -.37 < .001
Mental health -.46 < .001 -.48 < .001 -. 41 < .001
CDC Symptom Inventory Scores
Total .94 < .001 .97 < .001
Short-form .94 < .001 .95 < .001
Construct validity
The extent to which Symptom Inventory scores discriminate between subgroups classified as to fatigue status (e.g., CFS versus not fatigued or CFS versus ISF, CFS versus remission) is one measure of the Inventory's practicability for assessing fatiguing illnesses. All Bonferroni post-hoc comparisons between never fatigued controls and those classified as CFS or ISF showed significant mean differences related to Symptom Inventory scores (Figures 1 and 2). Also, those classified as in remission had significantly lower symptom impact than those with CFS or ISF. The Total score and the Case Definition score distinguished between subjects classified CFS or ISF (Bonferroni post-hoc test; p < .05), while the Symptom Inventory Short Form revealed only a trend towards higher symptom impact for the CFS. Subjects classified as CFS and ISF were similar with respect to the Frequency score, the Intensity score, and number of symptoms.
Figure 1 Mean scores for Symptom Inventory by subject classification. All post-hoc comparisons between never fatigued and CFS or ISF were significant (at least: p < 0.05). All post-hoc comparisons between Remission and CFS or ISF were significant (at least: p < 0.05).
Figure 2 Mean differences for the different Symptom Inventory Product Term Scores between subjects whose overall fatiguing illness history status was classified as classified as CFS, ISF, Remission and never fatigued controls. All post-hoc comparisons between never fatigued and CFS or ISF were significant (at least: p < 0.05). All post-hoc comparisons between Remission and CFS or ISF were significant (at least: p < 0.05 – not valid for other symptoms: Ever ISF vs. Remission).
Discussion
This study showed that the CDC CFS Symptom Inventory is a reliable and valid instrument for assessing symptoms associated with CFS. Many studies conducted in tertiary care settings have used ad hoc (non-validated) questionnaires to assess the frequency or intensity of CFS-defining symptoms. In addition to lack of validation, one unanswered question from such studies concerns what is worse – a severe symptom that occurs sporadically or a clinically minimal symptom occurring every day. The Symptom Inventory obviates this problem because it is scored as a product-term of intensity and frequency and better represents the variance of each symptom.
Both the Total and the Case Definition scores effectively assessed initial study classification as CFS, ISF, and never fatigued. They also showed excellent psychometric properties, including good internal consistency and validity. One might question the possibility of circular logic – defining an illness by its symptoms then assessing psychometric properties of a scale that measures the same symptoms. Unfortunately, as yet, CFS has no confirmatory physical signs or characteristic laboratory abnormalities [2]; so, in lieu of a 'gold standard', it is defined by disabling chronic fatigue and characteristic accompanying symptoms [1,3]. For initial classification in this study, we defined CFS by literally applying criteria of the 1994 case definition [1]. We classified subjects as CFS who stated that they had been fatigued for at least 6-months; that the fatigue severely affected their occupational, educational, social, or recreational activities; who endorsed the presence of at least 4 CFS defining symptoms; and, who had no exclusionary medical or psychiatric conditions. Subjects who had medically/psychiatrically unexplained chronic fatigue not fulfilling all these criteria were considered ISF. This operational classification notwithstanding, CFS represents a multi-faceted illness. Fatigue is a complex construct and we employed 2 standardized and validated instruments (the MFI and Chalder Fatigue Scale) to evaluate its various dimensions (e.g., physical fatigue, mental fatigue). Similarly, the impairment associated with CFS is not unidimensional; so we utilized the SF-36 to quantify the 8 major dimensions of function and wellbeing. Finally, CFS includes a characteristic symptom complex and we used the Symptom Inventory to evaluate the intensity and frequency of accompanying symptoms. The Total, Case Definition, and Short Form scores all had good convergent validity as determined by correlations with the 3 multidimensional measures of fatigue and impairment. The Symptom Inventory scores also discriminated between subgroups originally classified as CFS, ISF or not fatigued and this is a measure of the instrument's practicability for assessing fatiguing illness.
As noted above, there is no objective test to unequivocally diagnose CFS so it is premature to evaluate sensitivity or specificity of the various Symptom Inventory scores. Rather, our objective was to evaluate the Inventory's psychometric properties as baseline for its use in future studies. Studies of the clinical characteristics of CFS and other unexplained fatiguing illnesses should utilize the Symptom Inventory in conjunction with other instruments that assess different dimensions and consequences of fatigue (e.g., the MFI, SF-36) [3,10]. Beside the Symptom Inventory Total and the Case Definition scores, the Symptom Inventory Short Form score, consisting of only 6 symptoms (unusual fatigue after exertion, unrefreshing sleep, muscle aches, sleeping problems, memory and concentration problems), appears to be an economic and precise screening measurement for the current status of fatiguing illnesses.
The Symptom Inventory includes 10 symptoms that are not used to define CFS. Although not considered in the case definition, these symptoms are commonly reported by chronically ill people and have proven useful for stratification during analysis of descriptive and case control studies. In addition, the International CFS Study Group recommended additional research to further develop the CFS case definition and such research must assess a comprehensive range of symptoms. Finally, analytic studies of CFS must consider somatization disorders (both as confounders and comorbid conditions); the full range of symptoms in the Inventory is needed to categorize such disorders.
At least one important limitation must be considered. Although study subjects were recruited from the community and do not reflect the strong biases inherent of clinic patient populations, they did not represent the general population; rather they comprised a sample of people with and without unexplained fatiguing illnesses. Thus, the excellent psychometric properties of the Symptom Inventory cannot be generalized to the general population. To further validate and evaluate the Symptom Inventory, additional testing in a larger population-based sample not stratified by fatigue is required. There is also a need to determine the test-retest-reliability and stability of the Symptom Inventory. The present study provides preliminary results to encourage researchers to administer the Symptom Inventory along with other standardized questionnaires measuring fatigue and functional impairment in studies of CFS and other fatiguing illnesses.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
DW conceived of the scale to summarize symptom impact, had primary responsibility for statistical analysis and wrote the manuscript; RN collaborated in study design, in data analysis and writing the manuscript; CH was instrumental in the conception and design of the study, participated in fieldwork, collaborated in analysis and interpretation of the data, and writing the manuscript; JFJ collaborated in the clinical study and in preparation of the manuscript; ERU was instrumental in the conception and design of the study, collaborated in interpretation of the data, and writing the manuscript; WCR conceived of the study, served as principal investigator throughout its execution and collaborated in writing the manuscript.
==== Refs
Fukuda K Straus SE Hickie I Sharpe MC Dobbins JG Komaroff A The chronic fatigue syndrome; a comprehensive approach to its definition and study Ann Int Med 1994 121 953 959 7978722
Afari N Buchwald D Chronic fatigue syndrome: a review Am J Psychiatry 2003 60 221 236 12562565 10.1176/appi.ajp.160.2.221
Reeves WC Lloyd A Vernon SD Klimas N Jason LA Bleijenberg G Evengard B White PD Nisenbaum R Unger ER Identification of ambiguities in the 1994 chronic fatigue syndrome research case definition and recommendations for resolution BMC Health Services Research 2003 3 25 14702202 10.1186/1472-6963-3-25
Ware JE Sherbourne CD The MOS 36-item Short Form health survey (SF-36): conceptual framework and item selection Med Care 1992 30 473 483 1593914
Bultmann U de Vries M Beurskens AJ Bleijenberg G Vercoulen JH Kant IJ Measurement of prolonged fatigue in the working population: determination of a cut-off point for the Checklist Individual Strength J Occup Health Psychol 2000 5 411 416 11051524 10.1037//1076-8998.5.4.411
Smets EM Garssen BJ Bonke B DeHaes JC The multidimensional fatigue inventory (MFI) psychometric qualities of an instrument to assess fatigue J Psychosom Res 1995 39 315 325 7636775 10.1016/0022-3999(94)00125-O
Reyes M Nisenbaum R Hoaglin DC Emmons C Stewart G Randall B Unger ER Stewart J Abbey S Jones J Gantz N Minden S Reeves WC Prevalence and incidence of chronic fatigue syndrome in Wichita, Kansas Arch Int Med 2003 163 1530 1536 12860574 10.1001/archinte.163.13.1530
Nisenbaum R Jones JF Unger ER Reyes M reeves WC Clinical course of chronic fatigue syndrome in Wichita, Kansas BMC Hlth Quality Life Outcomes 2003 1 49 10.1186/1477-7525-1-49
Robbins L Cottler L Bucholz K Compton W Diagnostic Interview Schedule for DSM-IV (DIS-IV) 1995 St. Louis, MO: Washington University
Dittner AJ Wessely SC Brown RG The assessment of fatigue – A practical guide for clinicians and researchers J Psychosom Res 2004 56 157 170 15016573 10.1016/S0022-3999(03)00371-4
Chalder T Berelowitz G Pawlikowska T Watts L Wessely S Wright D Development of a fatigue scale J Psychosom Res 1993 37 147 153 8463991 10.1016/0022-3999(93)90081-P
|
16042777
|
PMC1183246
|
CC BY
|
2021-01-04 16:37:42
|
no
|
Popul Health Metr. 2005 Jul 22; 3:8
|
utf-8
|
Popul Health Metr
| 2,005 |
10.1186/1478-7954-3-8
|
oa_comm
|
==== Front
Proteome SciProteome Science1477-5956BioMed Central London 1477-5956-3-71596323110.1186/1477-5956-3-7MethodologyCognate peptide-receptor ligand mapping by directed phage display Stratmann Thomas [email protected] Angray S [email protected] Department of Molecular Biology, The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, CA 92037, USA2 Universidad de Barcelona, Departamento de Fisiologia, Diagonal 645, 3°, 08028 Barcelona, Spain2005 17 6 2005 3 7 7 21 10 2004 17 6 2005 Copyright © 2005 Stratmann and Kang; licensee BioMed Central Ltd.2005Stratmann and Kang; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
A rapid phage display method for the elucidation of cognate peptide specific ligand for receptors is described. The approach may be readily integrated into the interface of genomic and proteomic studies to identify biologically relevant ligands.
Methods
A gene fragment library from influenza coat protein haemagglutinin (HA) gene was constructed by treating HA cDNA with DNAse I to create 50 – 100 bp fragments. These fragments were cloned into plasmid pORFES IV and in-frame inserts were selected. These in-frame fragment inserts were subsequently cloned into a filamentous phage display vector JC-M13-88 for surface display as fusions to a synthetic copy of gene VIII. Two well characterized antibodies, mAb 12CA5 and pAb 07431, directed against distinct known regions of HA were used to pan the library.
Results
Two linear epitopes, HA peptide 112 – 126 and 162–173, recognized by mAb 12CA5 and pAb 07431, respectively, were identified as the cognate epitopes.
Conclusion
This approach is a useful alternative to conventional methods such as screening of overlapping synthetic peptide libraries or gene fragment expression libraries when searching for precise peptide protein interactions, and may be applied to functional proteomics.
==== Body
Background
Conventional approaches to elucidate peptide protein interactions include the screening of synthetic sequentially overlapping peptides [1,2], peptide fragments created by enzymatic digestion [3] or expressed gene fragments created by recombinant DNA technologies [4]. For more than a decade, phage display technology has been applied to elucidate protein-protein interactions. Random peptide libraries have been useful to predict epitope sequence mimics of unknown ligands. The epitopes for antibodies have been predicted from consensus sequences derived from aligning motifs with the original protein sequence of interest [5,6]. When the original ligand is not known, it is possible to predict potential binding consensus sequences. Castaño and colleagues predicted a binding motif for murine CD1 using a 22 amino acid random library [7]. In some instances, the ligand and receptor are known and the precise interactions between the two can be determined empirically as outlined above. A semi-empirical approach involves displaying randomly generated fragments of the target genes on fd phage [8]. The method has been used to elucidate epitopes for mAbs directed against the bluetongue virus outer capsid protein [9], human plasminogen-activator inhibitor 1 [10], large subunit of Drosophila RNA polymerase II, human p53, and the human cytokeratin 19 [11].
In this study, we report the epitope mapping of a monoclonal and polyclonal antibody directed against the influenza coat protein HA by a two-step approach. First, the target cDNA is digested randomly with DNAse I and cloned into a pre-selection plasmid Open Reading Frame Expression and Secretion (pORFES IV), described in US Patent 6,586,236 [12], between the ompA leader sequence and the β-lactamase gene. The leader sequence directs peptide-β-lactamase fusion to the periplasm of E. coli.
Fragments, which do not restore the open reading frame of the β-lactamase, as well as non-secreting inserts, are eliminated by selectively propagating in the presence of carbenicillin. Secondly, the selected inserts are transferred to a phage surface display vector JC-M13-88 [13,14] for display on p8 (encoded by a synthetic copy of gene VIII). The method allows the construction of libraries from complex systems (mix of cDNAs) or possibly genomic DNA. The approach demonstrated here allows rapid and precise mapping of epitopes recognized by antibodies using cDNA encoding the target protein to construct an epitope display library.
Results
Construction of a random haemagglutinin fragment epitope library as a β-lactamase fusion protein
An epitope fragment library from the influenza HA gene (influenza X-31 [H3N2], [15]) was generated by randomly digesting the HA DNA with DNAse I. The conditions were adjusted in order to create predominantly 50 – 100 bp fragments that were subsequently cloned into the modified pORFES vector [12]. The "empty" pORFES IV vector contains an ompA leader followed by the β-lactamase gene, which is out of frame as shown in Figure 1. Insertion of DNA fragments at the NaeI site that restore the reading frame between the ompA leader and the β-lactamase coding sequence, permits secretion and allows for positive selection using β-lactam antibiotics. E. coli XLOLR electrocompetent cells were transfected with the ligation products. After 1-hour recovery at 37°C, aliquots of the cells were plated on LB agar plates containing either chloramphenicol alone or in combination with carbenicillin. Since cell growth in the presence of carbenicillin requires restoration of the β-lactamase reading frame, colonies growing on chloramphenicol alone served as a representation of the total number of transformation events regardless of the presence or nature of the insert (i.e., containing either the self-ligated vector without an insert, or the vector ligated with an insert but not necessarily in-frame with β-lactamase). As expected, approximately 1/3 of the transformation events resulted in restoring the correct reading frame. The size of the primary HA library in pORFES was determined by plating on agar plates containing carbenicillin and chloramphenicol.
Figure 1 Insert modification of pORFES to pORFES IV to permit cloningof blunt-ended fragments. The original pORFES [12, 14] vector was designed to accept inserts as NheI-HindIII fragments. To create pORFES IV, the NheI – HindIII stuffer was replaced with aninsert to include a NaeI site, an additional bp and a (Gly)4 spacer. The ligation of inserts restoring the correct frame of the β-lactamase confers carbenicillin resistance.
Approximately 1.7 × 106 independent clones were generated. Since the insert could have one of two possible orientations in three reading frames, and since some of these fragments may have included stop codons, theoretically less than 1/6 of all clones should encode cognate HA peptides. Based on this assumption, the library size for cognate HA peptides was estimated to contain approximately 2.8 × 105 independent clones.
Cloning of the haemagglutinin fragment library into JC-M13-88
Amplification of the HA fragment library in pORFES IV permitted the positive selection of open reading frames and provided ample material for directional cloning into the display vector JC-M13-88. The unique restriction sites, XbaI-HindIII, facilitated directional transfer of the library from pORFES IV into JC-M13-88. These XbaI-HindIII fragments contain the ribosome-binding site, the ompA leader sequence followed by the random insert of the library. Although the DNA digestion yielded mainly 50 – 100 bp fragments, larger fragments were also found in the pORFES vector. In order to obtain only peptides of a length of about 15 to 30 amino acids, the XbaI-HindIII fragments were resolved on a TBE gel and DNA fragments with the appropriate size eluted. Cloning the libraries into JC-M13-88 afforded 6 × 106 independent clones for the HA phage library. This figure is ten-fold higher than the original pORFES IV HA library size. This over-representation ensures that most of the library members are transferred to the display system.
Panning of the haemagglutinin fragment library against mAb 12CA5
The well-characterized mAb 12CA5 that binds to the HA sequence YPYDVPDYAS [16] was used to screen the initial HA library prior to panning in order to determine the initial frequency of the cognate peptide fragment. Approximately 4100 plaques of the naïve (i.e. prior to panning) library were analyzed by probing filter lifts with mAb 12CA5. Initially, only 0.4% plaques were immunoreactive. This number increased with each round of selection. The library was subjected to 3 rounds of panning against mAb 12CA5. The second round of panning resulted in a 12,000-fold increase of eluted phage (2300 fold over background). Approximately 18% of the plaques stained strongly with mAb 12CA5. This number increased to 42% after the third round of selection (Figures 2, 3, 4). Phage from 6 stained plaques were isolated and the insert DNA sequence determined. The results are shown in Table 1 (see additional file 1). Half the clones were identical whereas the other 3 clones differed from each other. All 6 clones contained the YPYDVPDYAS motif.
Figure 2 Panning of the HA fragment phage library with mAb 12CA5. The HA fragment phage display library was selected against the anti-haemagglutinin mAb 12CA5 or wells coated with BSA (no antibody). The eluted phage was quantified after each round of panning by plaque assay.
Figure 3 Panning of the HA fragment phage library with mAb 12CA5. The enrichment factor was calculated as the ratio of eluted phage from wells coatedwith antibody or BSA only, respectively, for each round of panning
Figure 4 Panning of the HA fragment phage library with mAb 12CA5. An aliquots of eluted phage after each round of panning were analyzed by filter lifts for binding to mAb 12CA5. The filters were blocked with BSA, incubated with mAb 12CA5 and subsequently detected with goat anti-mouse kappa Ab coupled to alkaline phosphatase.
Panning of the haemagglutinin fragment library against pAb 07431
Subsequently, the library was evaluated using a rabbit polyclonal IgG raised against the peptide CKRGPDSGFFSRLNWLYKSG. When 2400 plaques of the initial library were screened by filter lift for immunoreactive clones, a higher background staining was observed relative to filter lifts stained with mAb 12CA5. However, one plaque was found to stain strongly with the pAb 07431. During selection, the number of eluted phage increased with each round. After the third round, 91% of plaques bound to pAb 07431 as determined by filter lifts (Figures 5, 6, 7). A sequence analysis of 16 clones revealed only 3 different sequences as shown in Table 2 (see additional file 2). All of the selected sequences contained the GFFSRLNWLTKS motif that is part of the peptide used for the immunizations. The immunizing peptide was derived from influenza strain X47 (HA1, [17]) and differed by a single substitution of T to Y from the gene used for this study (derived from influenza X-31 [H3N2]) as shown in Table 2. Despite this point variation the approach permitted recovery of the epitope.
Figure 5 Panning of the HA fragment phage library with pAb 07431. The HA fragment phage display library was selected against the anti-haemagglutinin pAb 07431 or wells coated with BSA only (no antibody). The Eluted phage was quantified after each round of panning by plaque assay.
Figure 6 Panning of the HA fragment phage library with pAb 07431. The enrichment factor was calculated as the ratio of eluted phage from wells coated with pAb 07431 or BSA only, respectively.
Figure 7 Panning of the HA fragment phage library with pAb 07431. An aliquots of eluted phage after each round of panning were analyzed by filter lifts for binding to pAb 07431. The filters were blocked with BSA, incubated with pAb 07431 and subsequently detected with goat anti-rabbit IgG Ab coupled to alkaline phosphatase.
Interestingly, though, clones TSS 399 and 401–403 encoded additional sequences resulting from frame shifts and a multiple insertion event. A BLAST search of the nucleotide sequence encoding the TSS 399 peptide on phage (5' – TCA TGT GGG CCT GCC AGA GAG GCA ACA TTA GGT GCA ACA TTT GCA TTT GAG TGT GGT AGC GGT TTT TTC AGT AGA CTG AAC TGG TTG ACC AAA TCA-3') in GenBank (access code V01103), revealed that nucleotide1–54 (shown in bold) aligned with nucleotides 1681–1734 of HA (incorrect frame). As expected, a search with the corresponding encoded 18-mer peptide (SCGPAREATLG ATFAFEC) failed to align on the HA polypeptide. Nucleotides 57–96 encoding the cognate epitope aligned with HA nucleotides 509–548 (in frame).
Discussion
Here we describe an approach for the mapping of linear peptide epitopes of fragmented genes by phage display. The procedure differs notably from other approaches of phage display of targeted gene fragments, as functional inserts are preselected prior to phage display. The vector pORFES IV (plasmid Open Reading Frame Expression and Secretion, [12]) permits the positive selection of fragments that give rise to open reading frames and allow expression of β-lactamase. The vector is designed such that ompA leader sequence cleavage occurs immediately prior to the insert. The advantage of such positioning of the cloning and cleavage site is that the insert encodes the N-terminus of the displayed polypeptide, not the vector. This may be important if the ligand requires a free N-terminus for receptor binding. By selecting bacteria in the presence of carbenicillin we are able to rescue plasmids with functional inserts and transfer fragments of desired size into the phage display vector. Conventional phage display libraries contain a mixture of functional and nonfunctional inserts along with phage without inserts. The propagation of phage without inserts or with non-functional inserts is favored, which affects phage library composition, reducing the functional complexity of the library. This qualitative step ensures a greater population of the initial phage libraries encode and display peptide. Approximately 87% of all primary transformation events were eliminated using this pre-selection step, implying that only 13% of the transformation events were functional. This figure is in accordance with the predicted number of potential open reading frames generated by this approach. This also highlights the degree of redundancy in screening conventional libraries for expressed sequences. The preselection step does not preclude the occurrence and accumulation of deletions or frame shifts during phage propagation. However, since the initial starting library is highly functional (i.e., it contains only phage displaying peptides), these acquired deletions or frame shifts should not significantly impact on the subsequent use of the library.
We validated the semi-empirical approach to targeted phage display by constructing a random epitope library from the gene encoding influenza virus protein HA. The DNA was digested and purified by agarose gel electrophoresis to create fragments encoding peptides of 15 to 35 amino acids. Affinity selection was carried out using a mAb and a pAb, which had been created by immunizing a mouse and a rabbit, respectively, with synthetic peptides representing short HA sequences. Thus, at the on-set of these experiments, the epitopes recognized by these antibodies were known. By subjecting the library to three rounds of selection against either of the antibodies, we were able to retrieve both sequences from the library. Panning against mAb 12CA5 revealed a 13 amino acid long consensus sequence, only 3 amino acids longer than the reported epitope for this Ab. When the library was panned against pAb 07431, a consensus sequence of 12 amino acids of length was found, representing the HA peptide 162–173 which is contained by the peptide used for immunization (HA 157–178).
Since we retrieved two distinct epitopes from this random epitope library, it is reasonable to assume that a large representation of open reading frames had been generated and displayed. In addition to selecting the pAb 07431 epitope, clone TSS 399 also encoded additional sequences that also produced an in-frame peptide. Simple BLAST analysis of the nucleotide and peptide sequences allowed us to determine the source of the additional sequence. The BLAST searches revealed that two inserts in tandem had been cloned into the vector, the first out of frame and the second insert inframe, resulting in a chimeric peptide containing the cognate sequence.
Random fragment epitope phage display libraries in combination with affinity selection are very useful tools to find cognate ligands to monoclonal or polyclonal antibodies, at least for linear epitopes. The method described here utilizes a preselection step to enrich for sequences with open reading frames that are readily secreted as fusion proteins with β-lactamase, a prerequisite for subsequent phage display. This ensures that a greater proportion of the phage library not only encodes a polypeptide, but that it also has the potential to display it. This approach to building a phage display peptide library has two key features; first N-terminus of the displayed peptide is encoded by the insert, and secondly that the displayed peptides are approximately 15–35 amino acids in length. This is critical if selecting for peptides binding to MHC class I or II molecules. The approach has been successfully applied for the selection of peptide epitopes derived from ovalbumin and glutamic acid decarboxylase specific for MHC class II molecules [18,19]. This technology might very well be successfully extended to more complex systems in which several cDNAs encoding up-regulated tumor associated genes are processed for epitope screening against antibodies, MHC class I or II molecules [18], or in other protein-protein interactions.
Conclusion
With the speed of library generation from either genomic and cDNA outpacing the emerging field of proteomics, a backlog in correlating gene to function and mapping protein-protein interactions is creating a bottleneck. Here, we hint at a possible route to assist in elucidating the interacting biologically relevant motifs directly from DNA. The example described utilized two antibodies (proteins) and an antigen encoding nucleic acid (DNA) to develop a proof of concept. Additionally, this approach has also been applied to elucidation of GAD peptides binding to MHC class II I-Ag7 molecules [18]. Once a key protein has been identified, it may be possible to use genetic information and selection to identify cognate ligands.
The use of pORFES IV to eliminate "junk" sequences enhances the quality of the displayed library, resulting in enhanced panning efficiency. This approach has already been applied to improving antibody fragment libraries [14], and may also be applied to creating higher quality random peptide display libraries.
Methods
Monoclonal and polyclonal antibodies
The mAb 12CA5 specific for a HA peptide (YPYDVPDYAS)[16] and the rabbit polyclonal serum 07431 against the peptide (CKRGPDSGFFSRLNWLYKSG) derived from Influenza X-47 HA1 were kindly provided by Dr. R. A. Lerner (The Scripps Research Institute). The rabbit IgG fraction was purified from the sera using protein A-sepharose (Pierce, Rochford, IL, USA).
Plasmids and bacterial strains
The plasmid Open Reading Frame Expression and Secretion IV is based on(pORFES) [12,14] with an insert modification shown in Figure 1. The phage display vector JC-M13-88 has been described elsewhere [12-14]. The host strains Escherichia coli XLOLR and XL1-Blue were purchased from Stratagene (La Jolla, CA, USA). Plasmid pCMU encoding HA cDNA (derived from influenza X-31 [H3N2], [15]) was a kind gift from Dr. G. Aichinger (Hammersmith Hospital, London, UK). All enzymes were purchased from Boehringer-Mannheim (Indianapolis, IN, USA).
Haemagglutinin gene fragmentation process
The plasmid pCMU (100 μg) was digested with SmaI/XbaI to release the HA coding insert. The gel-purified inserts (~10 μg DNA) was digested with 70 ng of DNAse in 50 mM Tris-HCl, pH 7.2, containing 40 mM MnCl2 (130 μl) for 4 minute sat 37°C and the reaction was terminated by adding EDTA to a final concentration of 70 mM. After precipitation, the fragments were treated with 0.001 units mung bean nuclease per μg DNA for 5 minutes at 37°C. The resulting mixtures of blunt ended fragments were ligated between the ompA leader sequence and the β-lactamase gene into pORFES IV that had been digested with NaeI and treated with calf intestine phosphatase. The ligation products were transfected into non-suppressing E. coli XLOLR via electroporation and propagated overnight in Super Broth containing 100 μg/ml carbenicillin at 37°C. The inserts that restore the reading frame and that are readily translocated into the periplasm along with the fused β-lactamase confer β-lactam antibiotic resistance. The transfection efficiency was monitored by plating aliquots on agar plates containing 25 μg/ml of chloramphenicol with or without 100 μg/ml carbenicillin. The pORFES IV library DNA was recovered from the overnight culture and inserts released by digestion with XbaI and HindIII. The fragments of 160–210 bp were resolved on a 3% agarose TBE gel.
Preparation of M13 HA fragment display library
The 160–210 bp XbaI/HindIII fragments encode a ribosome binding site, the ompA leader and the HA derived inserts. These were directionally cloned into the phage vector JC-M13-88, between a lac promoter and a synthetic copy of gene VIII. The ligation products were transfected into XL1-Blue E. coli cells via electroporation and the transfection efficiency monitored by plaque assay. The phage were propagated for 16 h at 37°C in Super Broth containing 1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) and isolated by PEG precipitation twice following standard methods [20]. The phage pellets were resuspended in phosphate buffered saline (PBS) and stored at 4°C.
Panning
The JC-M13-88 library containing the HA gene fragments was panned against mAb 12CA5 and polyclonal rabbit IgG 07431. Multi-well maixisorp strips (Nalge Nunc International, Naperville, IL USA) were coated with 100 μl of a 8 μg/ml antibody solution in PBS or with 5 mg/ml BSA in PBS (negative control) over night at 4°C. The wells were washed 3 times with PBS containing 0.1% Tween (PBST), blocked with 0.5% BSA in PBS for 1 h at 37°C and washed as before. Approximately 1010 plaque forming units (pfu) were added in 100 μl of PBS containing 0.25% BSA and 0.05% Tween. The panning was carried out in duplicate. After incubation for 90 minutes at 37°C, the wells were washed 10 times with PBST, the phage eluted with 100 μl of 100 mM glycine, pH 2.2, at ambient temperature for 10 minutes and the solution neutralized by adding 10 μl of 1 M Tris solution (pH not adjusted). An aliquot of the solution was retained to assess the phage output and the rest was used to infect E. coli XL1-Blue cells for propagation as described above.
Filter lifts
To determine the relative abundance of individual binding clones, filter lifts were carried out on the library prior to panning, and subsequently after each round of panning. Phage plaques were overlaid with nitrocellulose filters for 10 minutes at 37°C. The filters were subsequently removed and blocked with 0.5% BSA in PBS for 1 h at RT and washed 3 times with PBST. The mAb 12CA5 and pAb 07431 were adjusted to 2 μg/ml PBST and the filters incubated with the Ab solution for 1 h at RT. The filters were washed as before and incubated with a goat anti-mouse antibody (Southern Biotechnology Associates, Inc., Birmingham, AL USA), or a goat anti-rabbit IgG F(ab)2 fragment (United States Biochemical Corp., Cleveland, OH, USA) conjugated to alkaline phosphatase. Both secondary antibodies were diluted in PBST(1 μg/ml final concentration) and the filters incubated for 1 h at RT. After washing as before, the filters were stained with 5-bromo-4-chloro-3-indolylphosphate (BCIP) and 4-nitro blue tetrazolium chloride (NBT).
Analysis of individual clones
Individual phage plaques were picked and grown overnight after infection with E. coli XL1-Blue cells [20]. The double stranded replicative form (RF) DNA template was sequenced.
Competing interests
The author(s) declare that they have no competing interests.
Disclaimer
The views and opinions expressed herein are those of the authors and do not purport to reflect those of the Universidad de Barcelona.
Authors' contributions
TS was the graduate researcher on this project. ASK was the PI.All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
Table 1 – Alignment of haemagglutinin amino acids 112–133 with 6 peptide sequences displayed on JC-M13-88 after panning against mAb 12CA5. Six clones of the HA phage display library were analyzed after three rounds of panning with mAb 12CA5. All clones showed a positive reaction in a filter lift using mAb 12CA5. Three clones were identical, and all the clones contained the consensus sequence YPYDVPDYAS against which the mAb is directed (in red bold letters).
Click here for file
Additional File 2
Table 2 – Alignment of haemagglutinin amino acids 157–178 with 15 peptide sequences displayed on JC-M13-88 after panning against the polyclonal rabbit IgG 07431. 15 clones of the HA phage display library were analyzed after three rounds of panning with pAb 07431 raised against the haemagglutinin derived peptide CKRGPDSGFFSRCNWLYKSG. All clones showed a positive reaction in a filter lift using pAb 07431. Several clones were identical and 3 different sequences were identified. All the clones contained the consensus sequence GFFSRLNWLTKS (in blue bold letters).
Click here for file
Acknowledgements
We thank Rener Xu (Fudan University, China) for complete sequence determination of pORFES, Lucia Morkes, for technical assistance and Luc Teyton for helpful discussions. ASK is a recipient of an Investigators Award from the Cancer Research Institute/Partridge Foundation and this work was supported by McKnight Foundation. This article has been assigned manuscript number 11514-MB from The Scripps Research Institute.
==== Refs
Delvig AA Rosenqvist E Oftung F Robinson JH T-Cell epitope mapping the PorB protein of serogroup B Neisseria meningitidis in B10 congenic strains of mice Clin Immunol Immunopathol 1997 85 134 42 9344695 10.1006/clin.1997.4437
Patel SD Cope AP Congia M Chen TT Kim E Fugger L Wherrett D Sonderstrup-McDevitt G Identification of immunodominant T cell epitopes of human glutamic acid decarboxylase 65 by using HLA-DR(alpha1*0101,beta1*0401) transgenic mice Proc Natl Acad Sci U S A 1997 94 8082 7 9223318 10.1073/pnas.94.15.8082
Aoubala M Douchet I Bezzine S Hirn M Verger R De Caro A Immunological techniques for the characterization of digestive lipases Methods Enzymol 1997 286 126 49 9309649
Kelly CG Booth V Kendal H Slaney JM Curtis MA Lehner T The relationship between colonization and haemagglutination inhibiting and B cell epitopes of Porphyromonas gingivalis Clin Exp Immunol 1997 110 285 91 9367414
Bonnycastle LL Mehroke JS Rashed M Gong X Scott JK Probing the basis of antibody reactivity with a panel of constrained peptide libraries displayed by filamentous phage J Mol Biol 1996 258 747 62 8637007 10.1006/jmbi.1996.0284
Scott JK Smith GP Searching for peptide ligands with an epitope library Science 1990 249 386 90 1696028
Castano AR Tangri S Miller JE Holcombe HR Jackson MR Huse WD Kronenberg M Peterson PA Peptide binding and presentation by mouse CD1 Science 1995 269 223 6 7542403
Smith GP Filamentous fusion phage: novel expression vectors that display cloned antigens on the virion surface Science 1985 228 1315 7 4001944
Wang LF Du Plessis DH White JR Hyatt AD Eaton BT Use of a gene-targeted phage display random epitope library to map an antigenic determinant on the bluetongue virus outer capsid protein VP5 J Immunol Methods 1995 178 1 12 7530266 10.1016/0022-1759(94)00235-O
van Zonneveld AJ van den Berg BM van Meijer M Pannekoek H Identification of functional interaction sites on proteins using bacteriophage-displayed random epitope libraries Gene 1995 167 49 52 8566810 10.1016/0378-1119(95)00614-1
Petersen G Song D Hugle-Dorr B Oldenburg I Bautz EK Mapping of linear epitopes recognized by monoclonal antibodies with gene-fragment phage display libraries Mol Gen Genet 1995 249 425 31 8552047 10.1007/BF00287104
Kang AS Modulation of polypeptide display on modified filamentous phage use 2003 US Patent 6586236
Chappel JA He M Kang AS Modulation of antibody display on M13 filamentous phage J Immunol Methods 1998 221 25 34 9894895 10.1016/S0022-1759(98)00094-5
Chappel JA Rogers WO Hoffman SL Kang AS Molecular dissection of the human antibody response to the structural repeat epitope of Plasmodium falciparum sporozoite from a protected donor Malar J 2004 3 28 15283866 10.1186/1475-2875-3-28
Gething MJ Bye J Skehel J Waterfield M Cloning and DNA sequence of double-stranded copies of haemagglutinin genes from H2 and H3 strains elucidates antigenic shift and drift in human influenza virus Nature 1980 287 301 6 7421990 10.1038/287301a0
Field J Nikawa J Broek D MacDonald B Rodgers L Wilson IA Lerner RA Wigler M Purification of a RAS-responsive adenylyl cyclase complex from Saccharomyces cerevisiae by use of an epitope addition method Mol Cell Biol 1988 8 2159 65 2455217
Jou WM Verhoeyen M Devos R Saman E Fang R Huylebroeck D Fiers W Threlfall G Barber C Carey N Emtage S Complete structure of the hemagglutinin gene from the human influenza A/Victoria/3/75 (H3N2) strain as determined from cloned DNA Cell 1980 19 683 96 6153930
Corper AL Stratmann T Apostolopoulos V Scott CA Garcia KC Kang AS Wilson IA Teyton L A structural framework for deciphering the link between I-Ag7 and autoimmune diabetes Science 2000 288 505 11 10775108 10.1126/science.288.5465.505
Stratmann T Apostolopoulos V Mallet-Designe V Corper AL Scott CA Wilson IA Kang AS Teyton L The I-Ag7 MHC class II molecule linked to murine diabetes is a promiscuous peptide binder J Immunol 2000 165 3214 25 10975837
Sambrook J Fritsch E Maniatis T Molecular Cloning: A laboratory manual 1989 New York: Cold Spring Harbor Laboratory Press
|
15963231
|
PMC1183247
|
CC BY
|
2021-01-04 16:37:15
|
no
|
Proteome Sci. 2005 Jun 17; 3:7
|
utf-8
|
Proteome Sci
| 2,005 |
10.1186/1477-5956-3-7
|
oa_comm
|
==== Front
RetrovirologyRetrovirology1742-4690BioMed Central London 1742-4690-2-421599241010.1186/1742-4690-2-42ResearchInhibition of Tat activity by the HEXIM1 protein Fraldi Alessandro [email protected] Francesca [email protected] Giuliana [email protected] Annemieke A [email protected] Barbara [email protected] Olivier [email protected] Luigi [email protected] Department of Structural and Functional Biology, University of Naples 'Federico II', Naples, Italy2 UMR 8541 CNRS, Ecole Normale Supérieure, Laboratoire de Régulation de l'Expression Génétique, Paris, France3 Telethon Institute of Genetics and Medicine (TIGEM) Naples, Italy2005 2 7 2005 2 42 42 29 6 2005 2 7 2005 Copyright © 2005 Fraldi et al; licensee BioMed Central Ltd.2005Fraldi et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The positive transcription elongation factor b (P-TEFb) composed by CDK9/CyclinT1 subunits is a dedicated co-factor of HIV transcriptional transactivator Tat protein. Transcription driven by the long terminal repeat (LTR) of HIV involves formation of a quaternary complex between P-TEFb, Tat and the TAR element. This recruitment is necessary to enhance the processivity of RNA Pol II from the HIV-1 5' LTR promoter. The activity of P-TEFb is regulated in vivo and in vitro by the HEXIM1/7SK snRNA ribonucleic-protein complex.
Results
Here we report that Tat transactivation is effectively inhibited by co-expression of HEXIM1 or its paralog HEXIM2. HEXIM1 expression specifically represses transcription mediated by the direct activation of P-TEFb through artificial recruitment of GAL4-CycT1. Using appropriate HEXIM1 mutants we determined that effective Tat-inhibition entails the 7SK snRNA basic recognition motif as well as the C-terminus region required for interaction with cyclin T1. Enhanced expression of HEXIM1 protein modestly affects P-TEFb activity, suggesting that HEXIM1-mediated repression of Tat activity is not due to a global inhibition of cellular transcription.
Conclusion
These results point to a pivotal role of P-TEFb for Tat's optimal transcription activity and suggest that cellular proteins that regulate P-TEFb activity might exert profound effects on Tat function in vivo.
==== Body
Background
The positive transcription elongation factor b (P-TEFb) composed by CDK9/CyclinT1, has emerged as a significant co-factor of the HIV Tat protein. P-TEFb complex has been shown to associate with and phosphorylate the carboxyl-terminal domain (CTD) of RNA pol II, thereby enhancing elongation of transcription [1-3]. Tat protein binds an uracil containing bulge within the stem-loop secondary structure of the Tat-activated region (TAR-RNA) in HIV-1 transcripts [4-6]. Tat functions as an elongation factor and stabilizes the synthesis of full-length viral mRNAs by preventing premature termination by the TAR-RNA stem-loop. Physical and functional interactions between Tat and P-TEFb have been well documented [7,8]. Tat binds to P-TEFb by direct interaction with the human cyclinT1, and the critical residues required for interaction have been delineated [9,10]. The current model for recruitment of P-TEFb to the LTR, predicts the formation of the Tat-P-TEFb complex, which efficiently binds TAR, allowing CDK9 to phosphorylate the CTD of RNAPII, thereby, enhances processivity of the polymerase to produce full-length mRNAs [3,7-10].
Like other CDKs, the P-TEFb activity is regulated by a dedicated inhibitor. Two different P-TEFb complexes exist in vivo [11,12]. The active complex is composed of two subunits, the CDK9 and its regulatory partners cyclinT1 or T2. In addition, a larger inactive complex has been identified, which comprises of four subunits, CDK9, cyclinT1 or T2, the abundant small nuclear RNA 7SK and the HEXIM1 protein [13-17]. It has been recently shown that HEXIM1 has the inherent ability to associate with cyclin T1 and binding of 7SK snRNA turns the HEXIM1 into a P-TEFb inhibitor [15-17]. The relative presence of core and inactive P-TEFb complexes changes rapidly in vivo [11,12]. Several stress-inducing agents trigger dissociation of the inactive P-TEFb complex and subsequent accumulation of kinase active P-TEFb [11]. Thus, the 7SK-HEXIM1 ribonucleic complex represents a new type of CDK inhibitor that contributes to regulation of gene transcription. A further level of complexity of this system comes from the recent identification of HEXIM2, a HEXIM1 paralog, which regulates P-TEFb similarly as HEXIM1 through association with 7SK RNA [18,19].
It has been showed that Tat binds exclusively to the active P-TEFb complex [13]. Thus the presence of HEXIM1/7SK snRNA in P-TEFb complexes prevents Tat binding. Since the association between 7SK RNA/HEXIM1 and P-TEFb appears to compete with binding of Tat to cyclinT1, we have speculated that the TAR RNA/Tat system may compete with the cellular 7SK snRNA/HEXIM1 system in the recruitment of the active P-TEFb complex [13]. Accordingly, it has been shown that over-expression of HEXIM1 represses Tat function [14,17].
We show here that HEXIM1, or its paralog HEXIM2, inhibits Tat trans-activation of HIV-LTR driven gene expression, and more importantly, we demonstrated the role of the 7SK snRNA recognition motif as well as the binding to cyclin T1 as crucial elements for efficient Tat inhibition.
Results
Tat activity is inhibited by HEXIM1
Tat activity involves direct interaction with CDK9/CyclinT1 (P-TEFb) complex. However, two major P-TEFb-containing complexes exits in human cells [11,12]. One is active and restricted to CDK9 and cyclin T, the other is inactive and it contains HEXIM1 or 2 and 7SK snRNA in addition to P-TEFb [15,17]. We have previously shown that Tat interacts only with the active P-TEFb complex [13]. Because the two complexes are in rapid exchange, we sought to determine the functional consequences of the over-expression of HEXIM1 and 7SK snRNA on HIV-1 LTR-driven gene transcription. To this end we performed transient transfections in human 293 cells using the HIV-LTR-Luc reporter in the presence of increasing amounts of Flag-taggeted HEXIM1 and 7SK snRNA, respectively. Dose-dependent expression of F:HEXIM1 was monitored by immunoblotting with anti-HEXIM1 antibody (Fig. 1 panel A). As presented in Fig. 1B, we found that basal transcription from the LTR sequences was unaffected by the presence of F:HEXIM1 or 7SK RNA. In contrast, Tat-mediated transactivation of the HIV-1 LTR was inhibited by the over-expression of F:HEXIM1 in a dose-dependent manner. Ectopic expression of 7SK RNA did not significantly affected HIV-LTR-Luc expression either alone or in combination with F:HEXIM1. Thus, it appears that HEXIM1 is able to repress Tat-mediated activation. To further substantiate the inhibitory function of HEXIM1 we sought to extend our analysis using the murine CHO cells. Tat protein is a potent activator of HIV-1 LTR transcription in primate cells but only poorly functional in rodent cells [6,7]. However, Tat-mediated activation can be rescued by enforced expression of human cyclin T1 [6,7]. As presented in Fig. 1C we found that, while hCycT1 rescued Tat function, ectopic expression of HEXIM1 effectively inhibits Tat activity. Most importantly, Tat enhancement mediated by hCycT1 was effectively abrogated by co-expression of HEXIM1 in a dose-dependent manner. Finally, like in human cells, ectopic expression of 7SK snRNA did not have any significant effect on Tat activity.
Figure 1 Overexpression of HEXIM1 protein represses Tat transactivation. Panel A, Increasing amounts (10, 100 and 500 ng) of Flag-taggeted HEXIM1 were transfected into 293, cellular extracts were prepared at 48 hr after transfection and the relative levels on endogenous and exogenous HEXIM1 proteins were visualized by immunoblotting with anti-HEXIM1 antibody. Panel B, the HIV-Luc reporter (50 ng) was transfected into 293 cells in the presence of pSV-tat (50 ng) along with increasing (10, 100 and 500 ng) amounts of F:HEXIM1 and 7SK RNA (10, 100 and 500 ng), as indicated. Panel C, HEXIM1 decreases the co-operative effect of CycT1 on Tat activation in rodent cells. Chinese hamster ovary cells (CHO) were transfected with the HIV-LTR-Luc reporter (50 ng) in the presence of pSV-Tat (100 ng), lane 1, and together with CMV-hCycT1 (200 ng), lane 2, in the presence of increasing amounts of F:HEXIM1 and 7SK RNA as in panel B. Each histogram bar represents the mean of at least three independent transfections after normalization to Renilla luciferase activity to correct for transfection efficiency with the activity of the reporter without effect set to one. Standard deviations were less than 10%.
The results reported above suggested that ectopic expression of HEXIM1 inhibits Tat activity. A large number of evidences indicate that Tat-transactivation is mainly due to the recruitment of the cellular complex P-TEFb to the LTR, causing phosphorylation of the RNAPII CTD [1,6-10]. Accordingly, we and others have previously showed that artificial recruitment of P-TEFb to the HIV-1 promoter is sufficient to activate the HIV-1 promoter in the absence of Tat [20,21]. We sought to determine the consequences of ectopically expressed F:HEXIM1 on P-TEFb induced transcription in the absence of Tat. We showed that direct recruitment of CyclinT1 to a promoter template by fusion to the GAL4 DNA binding domain, activates transcription from an HIV-1 LTR (G5HIV-Luc) reporter bearing GAL4 sites [20]. Human 293 cells were transfected with the G5HIV-Luc reporter along with GAL4-fusion expression vectors in the presence of F:HEXIM1. As shown in Fig. 2A, we found that GAL4-CycT1 effectively activates transcription from the HIV-1 LTR reporter, and co-expression of F:HEXIM1 resulted in a robust dose-dependent inhibition. The specific effect of HEXIM1 expression was highlighted by the results shown in Fig. 2B. G5HIV-Luc reporter was activated by co-expression of a GAL4-TBP, and such activation was largely unaffected by co-expression of HEXIM1. Thus, it appears that while HEXIM1 represses P-TEFb activity, enforced expression of this protein does not have significant effects on TBP-mediated basal transcription.
Figure 2 HEXIM1 represses GAL4-CycT1-mediated activation. Human 293 cells were transfected with 50 ng of G5-HIV Luc reporter DNA alone (lane 1) or in the presence of GAL4-expression plasmid DNA (200 ng), as indicated. The presence of the cotransfected F:HEXIM1 (10, 100 and 500 ng) is indicated. Each histogram bar represents the mean of three independent transfections after normalization to Renilla luciferase activity. The results are presented as described in figure 1.
Definition of the HEXIM1 regulatory domains involved in repression
To investigate the structural determinants of HEXIM1 protein in repression, the activity of Gal4-CycT1 on G5HIV-Luc was monitored in the presence of co-transfected Flag-tagged deletion mutants of HEXIM1. We found that removal of the C-terminal amino acids affected the inhibition as shown by the HEXIM1 (1–300) and (1–240) mutants (Figure 3 lanes 6–8 and 9–11). In contrast, removal of the 119 N-terminal amino acids of HEXIM1 (120–359) did not abolished inhibition (lanes 12–14). However, further deletion of the N-terminal amino acids (181–359) completely abolished the inhibitory effect (lanes 15–17). Thus, HEXIM1-mediated repression required the presence of the C-terminal domain (300–359aa) as well as a central region between residues 120 and 181. Finally, we found that HEXIM2, which like HEXIM1, associates and inhibits P-TEFb activity, represses Gal4-CycT1 activation in a dose dependent manner (lanes 18–20).
Figure 3 HEXIM1 regulatory domains involved in repression. Human 293 cells were transfected with 50 ng of G5-HIV Luc reporter DNA alone (lane 1) or in the presence of 50 ng of pSV-Tat (lanes 2–20). The presence of increasing amounts (10, 100 and 500 ng) F:HEXIM1 wild-type (lanes 3–5), various deletion mutants (lanes 6–17) and F:HEXIM2 wt(18–20) are indicated, respectively. On the bottom, it is shown the western-blot of whole cells extracts from transfected cells probed with anti-Flag antibody from the indicated co-transfections. The results presented are from a single experiment after normalization to Renilla luciferase activity with the activity of the reporter without effect set to one. This experiment was performed three times with similar results.
We have recently reported that the HEXIM1 C-terminal domain (181–359) is involved in the binding to P-TEFb through direct interaction with the cyclin-box of cyclinT1 [15], and the evolutionarily conserved motif (PYNT aa 202–205) is important for such interaction. The PYND point mutant is impaired in repression and binding either P-TEFb or 7SK RNA in vivo, albeit it retains the ability to bind 7SK in vitro. In addition, we determined that HEXIM1 binds 7SK snRNA directly and the RNA-recognition motif (KHRR) was identified in the central region of the protein (aa 152–155). In fact, the HEXIM1-ILAA mutant fails to interact in vivo and in vitro with 7SK snRNA [15]. To test the importance of these motifs in HEXIM1-mediated repression of Tat activity, HEXIM1 point mutants were co-transfected in 293 cells along with Tat or Gal4-CycT1, respectively. As shown in Figure 4, unlike wild-type HEXIM1, both mutants were unable to repress Tat as well as Gal4-CycT activities, albeit they were expressed at levels comparable to the wild-type protein. Collectively, the results presented in figures 3 and 4 strongly suggest that HEXIM1-mediated inhibition of Tat activity requires interaction with P-TEFb as well as binding to 7SK snRNA.
Figure 4 On top the relevant HEXIM1 functional domains are depicted. Position of the point mutants ILAA and PYND are indicated. G5-HIVLuc reporter (50 ng) was transfected into 293 cells along with Gal4-CycT1 (200 ng) Panel A, or pSV-Tat (50 ng) panel B along with increasing amounts of Flag:HEXIM1 wilt type and mutants (10, 100 and 500 ng) as indicated. Each histogram bar represents the mean of three independent transfections after normalization to Renilla luciferase activity. The results are presented as described in figure 1. Panel C, western-blot with anti-HEXIM1 antibody demonstrated that the HEXIM1 effectors were expressed at comparable levels.
P-TEFb activity in the presence of enhanced expression of HEXIM1
Next we sought to determine whether enhanced expression of HEXIM1 might directly affect the P-TEFb activity. 293 cells were transfected with F:HEXIM1 and cellular extracts from mock and transfected cells were prepared. P-TEFb activity was assayed using as substrate the CTD4 peptide consisting of four consensus repeats of the RNAPII CTD, and time-course kinase assays were performed [15]. Briefly, P-TEFb and its associated factors were affinity purified with anti-CycT1 antibody from mock and F:HEXIM1 transfected cell extracts. Immunoprecipitates were analyzed by immunoblotting for evaluation of CDK9, cyclin T1 and HEXIM1 proteins, respectively. The immunoprecipitates were then treated or not treated with RNase A (Fig. 5). The RNase treatment will degrade the 7SK snRNA thereby relieving the P-TEFb inhibition by HEXIM1/7SK snRNP. In fact, samples treated with RNase showed a robust increase in kinase activity compared those not treated with RNase, indicating that 7SK snRNA had been effectively degraded. We found that the kinase activities of samples that were treated with RNase were quantitatively the same in both mock and F:HEXIM1 transfected extracts indicating equal amounts total of P-TEFb in both samples. A modest, but reproducible reduction of P-TEFb kinase activity (2-fold) was observed in extracts from F-HEXIM1 transfected cells. Altogether, these results demonstrated that over-expression of HEXIM1 resulted in a modest reduction of P-TEFb activity, thus the inhibition of Tat activity is unlikely due to a global reduction of cellular P-TEFb activity.
Figure 5 P-TEFb activity in F:HEXIM1 transfected cells. Human 293 cells were transfected with 100 ng of F:HEXIM1 and cell extracts were prepared from mock and F:HEXIM1 expressing cells at 48 hr after transfection. Cell extracts were immunoprecipitated with anti-cycT1 antisera. The relative amounts of immunopreicipitated cyclinT1, CDK9 and HEXIM1 were quantitated by immunoblotting. Samples were treated or not treated with RNase, as indicated. Kinase assays were performed using a CTD4 peptide and 32P incorporation was quantified in arbitrary units and plotted versus time (min). This experiment was performed four times with similar results. A typical experiment is shown.
To further investigate the mechanism of inhibition of Tat-mediated transcription by HEXIM1, we tested the relative levels of transfected Tat protein in the presence of F:HEXIM1. We found that ectopic expression of HEXIM1 did not affected Tat expression (Figure 6A). Next, we sought to determine whether exogenous expression of HEXIM1 might result in a decrease in Tat-bound CycT1. To this end 293 cells were transfected with pSV-Tat in the presence or absence of F-HEXIM1 using the same transfection conditions used in the Luciferase assays. Cells extracts were immunoprecipitated with CycT1 antibody and the immunoprecipitates were analyzed by immunoblotting for evaluation of Tat, CycT1 and HEXIM1 proteins, respectively. In two different experiments we found a modest, but reproducible decrease in Tat-bound cyclin T1 (Fig. 6B). Thus, it appears that exogenous expression of HEXIM1 results in a decrease of Tat-bound P-TEFb.
Figure 6 Tat-CyclinT1 binding in the presence of HEXIM1. Panel A. 293 cells were transfected with 50 ng of pSV-Tat in the presence or absence of F:HEXIM1 (100 ng) as indicated and at 48 hrs after transfection cell extracts were probe by Western blotting with anti-Tat. For accurate comparison increasing amounts of material (μl) were loaded on the gels. Panel B. 293 cells were transfected as in Panel A, and cell extracts were immunoprecipitated with anti-CycT1. Immunocomplexes were analyzed on Western blots as indicated. I, input, B; bound, FT; flow through. This experiment was performed two times with similar results.
Discussion
Several lines of evidence have suggested that Tat function is largely dependent upon the physical and functional interaction with the cellular transcription factor P-TEFb. The recruitment of P-TEFb to the LTR, involves the formation of the Tat-P-TEFb complex which efficiently binds TAR, allowing CDK9 to phosphorylate the CTD of RNAPII, thereby, enhances processivity of the polymerase to produce full-length mRNAs [6-10]. Two different P-TEFb complexes exist in vivo. The core active P-TEFb comprises two subunits, the catalytic CDK9 and a regulatory partner cyclin T, and a larger inactive P-TEFb complex comprised by CDK9, cyclin T, HEXIM1 protein and the 7SK snRNA [11-17]. The relative presence of core and inactive P-TEFb complexes changes rapidly in vivo [11]. We have previously shown that the presence of HEXIM1/7SK snRNA in P-TEFb complexes prevents Tat binding to P-TEFb [13]. Since the association between 7SK RNA/HEXIM1 and P-TEFb competes with binding of Tat to cyclinT1, we have speculated that the TAR RNA/Tat system may compete with the cellular 7SK snRNA/HEXIM1 system [13]. Accordingly, it has been shown that over-expression of HEXIM1 represses Tat function [14,19] We show here that HEXIM1 inhibits Tat function, while expression of 7SK snRNA does not influence Tat activity. It is pertinent to note that 7SK RNA is an abundant snRNA [23], and it is unlikely that 7SK might be rate-limiting for the assembly of the inactive P-TEFb complex.
We have delineated important structural domains of HEXIM1 required for repression of Tat. First, we found that the C-terminal region is required for inhibition. Previous findings indicated that the C-terminal region of HEXIM1 is involved in binding with cyclinT1 as well as for homo and hetero-dimerization with HEXIM2 [15,18,19]. Second, point mutations in the evolutionarily conserved motif PYNT (aa 202–205) abolished inhibition. It has recently shown a critical role of threonine 205 in P-TEFb binding [15]. Moreover, deletion mutants unable to bind P-TEFb failed to repress Tat (Figure 3). Therefore, it appears that HEXIM1 inhibition is strictly dependent upon the integrity of the protein to interact with P-TEFb. Third, a point mutant in the central part of HEXIM1 (KHRR motif aa 152–155) strongly affects Tat repression. Since this basic motif has been previously shown as the 7SK snRNA recognition motif [15], we conclude that interaction between HEXIM1 and 7SK snRNA is required for Tat repression. Collectively, these findings strongly suggested that HEXIM1-mediated inhibition of Tat required the formation of the P-TEFb/HEXIM1/7SK complex.
We determined that enhanced expression of HEXIM1 resulted in a modest inhibition (2-fold) of P-TEFb activity in vivo. Thus, HEXIM1-mediated inhibition of Tat activity is unlikely due to a global inhibition of P-TEFb activity. Moreover, we found that basal transcription from the LTR sequences was largely unaffected by over-expression of HEXIM1. Finally, ectopic expression of this protein does not have significant effects on TBP-mediated basal transcription. Thus, it appears that P-TEFb is specifically required for Tat-dependent HIV LTR transcription. Our results differ somewhat from those obtained in the Zhou lab who found that exogenous expression of HEXIM1 affects both basal as well as Tat-induced transcription [13]. These apparent discrepancies are possible due to different transfection conditions in which the relative amounts of the over-expressed exogenous proteins are likely different. We found that Tat expression which is under the control of SV40 promoter remains largely unaffected by co-expression of HEXIM. Our findings suggest a dedicated role of P-TEFb in Tat activity. Recent studies point to a specific role of P-TEFb for certain promoters. It has recently found that P-TEFb is recruited to the IL-8 but not to the IkBα promoter [23], and it also represses transcription of regulators such as the nuclear receptor coactivator, PGC-1, in cardiac myocytes [24]. The specific HEXIM-mediated inhibition of Tat activity underlines the pivotal role of P-TEFb in the HIV LTR transcription.
The repression exerted by the HEXIM1 protein is likely the results of a competition between Tat and HEXIM1 in binding the P-TEFb. Since Tat binds only to the active P-TEFb complex, it has been suggested that Tat might trap the active form of P-TEFb as the PTEFb/7SK RNA/HEXIM1 complex appears to undergo continuous formation and disruption in vivo. In this scenario over expression of HEXIM1 may counteract the binding of Tat to P-TEFb, through a competitive association between the ectopic expressed HEXIM1 and P-TEFb. Accordingly, we found that exogenous expression of HEXIM1 results in a small but detectable reduction in Tat-bound- P-TEFb. Our co-immunoprecipitation results are consistent with recent findings showing a mutually exclusive interaction of HEXIM1 and Tat with cyclinT1 using recombinant purified proteins [25]. Because Tat and HEXIM1 interact with the cyclin-box region of cyclinT1, it is plausible if not likely, that the mutually exclusive interaction of these two molecules with cyclinT1 is due to binding to the same domain or to a sterical hindrance. However, these studies have been performed in vitro in the absence of 7SK snRNA.
The results reported here along with previous findings strongly suggest the crucial role of 7SK in the interaction between HEXIM1 and cyclinT1. In fact, HEXIM1 ILAA mutant does not associate with 7SK in vivo and in vitro, and co-immuprecipitation of cyclinT1 and 7SK RNA was markedly reduced with ILAA mutant compared to wild type [15]. Finally, as shown here ILAA mutant failed to repress Tat activity, suggesting an important role of HEXIM1/7SK interaction in Tat inhibition. Thus, association between HEXIM1 and 7SK snRNA appears an important determinant for Tat inhibition. Future in vitro and in vivo interaction studies, in the presence of 7SK snRNA may be instrumental to elucidate the role of 7SK/HEXIM1 complex in Tat activity.
Conclusion
The studies described in this provides further support to the pivotal role of P-TEFb for the optimal transcription Tat activity and highlight the importance of the P-TEFb cellular co-factors HEXIM1/7SK snRNA complex in Tat activity.
Methods
Tissue culture and transfections
Human 293 and rodent CHO cells were grown at 37°C in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (Gibco, Life Technologies). Subconfluent cell cultures were transfected cell cultures were transfected by a liposome method (LipofectAMINE reagent; Life Technologies, Inc.) in 2 cm/dish in multiwells, using 100 ng of reporter DNA and different amounts of activator plasmid DNA as indicated in the text and 20 ng of Renilla luciferase expression plasmid (pRL-CMV, Promega) for normalization of transfections efficiencies. Cells were harvested 48 h after DNA transfections, and cellular extracts were assayed for luciferase activity using Dual-Luciferase Reporter assay (Promega) according to the manufacturer's instructions. The experimental reporter luciferase activity was normalized to transfection efficiency as measured by the activity deriving from pRL-CMV.
Plasmids
The G5HIV-Luc contained the HIV-1 LTR sequences from -83 to +82 of LTR driven the Luc gene with 5 GAL4 DNA-binding sites inserted at -83. The pSV-Tat, GAL4-TBP, GAL4-CycT1, have been described [20]. 7SK snRNA plasmid was kindly provided by S. Murphy [22]. All Flag-taggeted HEXIM1 and HEXIM2 expression vectors were constructed by insertion of the corresponding cDNA regions into the EcoRV site of p3xFlag-CMV10 vector (Clontech). Description of the deletion and point HEXIM1 mutants have been described previously [15]. Full description of the expression vectors used in this work is available upon request.
Western blotting and antibodies
Cells were lysed in ice-chilled buffer A (10 mM HEPES pH 7.9, 1.5 mM MgCl2, 10 mM KCl, 200 mM NaCl, 0.2 mM EDTA), supplemented with 1 mM dithiothreitol, 40 U/ml of RNasin (Promega), protease inhibitor cocktail (P-8340; Sigma), and 0.5 % Nonidet P-40. Lysates were vortexed and incubated for 20 min on ice and clarified by centrifugations. Western blottings were performed using the following antibodies: the rabbit polyclonal anti-HEXIM1 (C4) has been previously described (6); anti-FLAG M2 Monoclonal Antibody (Sigma), goat polyclonal anti-CycT1 (T-18), rabbit polyclonal anti-CDK9 (H-169) from Santa Cruz, anti-Tat (NIH AIDS Research Reagent Program). Binding was visualized by enhanced chemiluminescence (ECL-plus Kit, Amersham Biosciences).
Co-immunoprecipitation and kinase assay
293 cells were transfected with pSV-Tat in the presence or absence of F:HEXIM1 and cell extracts were prepared at 48 hrs after transfection. CycT1 was immunopurified from cell extracts (1 mg) using anti-CycT1 (H-245) (sc-10750, Santa Cruz). Input, immunoprecipited and flow through materials were used in western blottings using anti-cycT1, anti-HEXIM1 and anti-Tat, respectively. For kinase assays 293 cells were transfected with F:HEXIM1 and after 48 hr P-TEFb complex was immunopurified from cell extracts (1 mg) using anti-CycT1 (H-245) (sc-10750, Santa Cruz) as previously described [13,15]. Briefly, whole cell extracts from mock and F:HEXIM1 transfected 293 cells were used in immunoprecipitations together with 40μl of slurry beads (protein G-Sepharose 4 Fast Flow, Amersham Biosciences) pre-adsorbed with anti-CycT1 and the interactions were carried out in buffer A for one hour at 4°C on a wheel. After extensive washes one half of the immunopurified materials was used in western blotting to ensure the presence of equal amounts of CDK9, HEXIM1 and CycT1, respectively. The remaining material was suspended and stirred at room temperature and split in two equal aliquots. One of the aliquot was treated with 10U of RNase A for 15 min at 30°C. Samples treated or not with RNase were stirred at room temperature for three minutes in 65 μl of buffer A containing [γ-32P]ATP (0,1 μCi/μl), 40 mM ATP, 0,1 μg/ml (YSPTSPS)4 peptide CTD4 (6, 8) and RNasin (40 U/ml). Aliquots (20 μl) of the suspension were mixed with SDS-PAGE loading buffer at intervals of three minutes to stop the reaction. The phosphorylated CTD4 substrate was separated on a 15% SDS-PAGE and visualized by radiography. Incorporation of [32P] into CTD peptide was quantified on a Bio-Rad phosphoimager.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
AF carried out the transfection studies and plasmid construction. FV performed studies using the HEXIM1 point mutants. GN carried out the kinase experiments. AAM isolated and constructed the HEXIM2 expression vector. BM and OB participated on discussion of results and drafting the manuscript. LL designed this study and edited the manuscript.
Acknowledgements
We thank S. Murphy for 7SK snRNA plasmid. This work was supported by grants from Istituto Superiore di Sanità Programma Nazionale di Ricerca AIDS and from Italian Association for Cancer Research (AIRC) (L.L.), from Association pour la Recherche sur le Cancer, Agence Nationale de Recherche sur le SIDA (O.B.), and from the Galileo Italy-France exchange program (G.N.).
==== Refs
Price DH P-TEFb, a cyclin-dependent kinase controlling elongation by RNA polymerase II Mol Cell Biol 2000 20 2629 2634 10733565 10.1128/MCB.20.8.2629-2634.2000
Napolitano G Majello B Lania L Role of cyclinT/Cdk9 complex in basal and regulated transcription (Review) Int J Oncol 2002 21 171 177 12063565
Garriga J Grana X Cellular control of gene expression by T-type cyclin/CDK9 complexes Gene 2004 337 15 23 15276198 10.1016/j.gene.2004.05.007
Cullen BR Does HIV-1 Tat induce a change in viral initiation rights? Cell 1993 73 417 420 8490957 10.1016/0092-8674(93)90126-B
Jones KA Peterlin BM Control of RNA initiation and elongation at the HIV-1 promoter Ann Rev Biochem 1994 63 717 743 7979253 10.1146/annurev.bi.63.070194.003441
Brigati C Giacca M Noonan DM Albini A HIV Tat, its TARgets and the control of viral gene expression FEMS Microbiol Lett 2003 220 57 65 12644228 10.1016/S0378-1097(03)00067-3
Mancebo HS Lee G Flygare J Tomassini J Luu P Zhu Y Peng J Blau C Hazuda D Price D Flores O P-TEFb kinase is required for HIV Tat transcriptional activation in vivo and in vitro Genes Dev 1997 11 2633 2644 9334326
Zhu Y Pe'ery T Peng J Ramanathan Y Marshall N Marshall T Amendt B Mathews MB Price DH Transcription elongation factor P-TEFb is required for HIV-1 Tat transactivation in vitro Genes Dev 1997 11 2622 2632 9334325
Bieniasz PD Grdina TA Bogerd HP Cullen BR Recruitment of a protein complex containing Tat and cyclin T1 to TAR governs the species specifity of HIV-1 Tat EMBO J 1998 17 7056 7065 9843510 10.1093/emboj/17.23.7056
Garber ME Wei P KewalRamani VN Mayall TP Herrmann CH Rice AP Littman DR Jones KA The interaction beween HIV-1 Tat and human cyclin T1 requires zinc and a critical cysteine residue that is not conserved in the murine CycT1 protein Genes Dev 1998 12 3512 3527 9832504
Nguyen VT Kiss T Michels AA Bensaude O 7SK small nuclear RNA binds to and inhibits the activity of CDK9/cyclin T complexes Nature 2001 414 322 325 11713533 10.1038/35104581
Yang Z Zhu Q Luo K Zhou Q The 7SK small nuclear RNA inhibits the CDK9/cyclin T1 kinase to control transcription Nature 2001 414 317 322 11713532 10.1038/35104575
Michels AA Nguyen VT Fraldi A MAQ1 and 7SK RNA interact with CDK9/cyclin T complexes in a transcription-dependent manner Mol Cell Biol 2003 23 4859 4869 12832472 10.1128/MCB.23.14.4859-4869.2003
Yik JH Chen R Nishimura R Jennings JL Link AJ Zhou Q Inhibition of P-TEFb (CDK9/Cyclin T) kinase and RNA polymerase II transcription by the coordinated actions of HEXIM1 and 7SK snRNA Mol Cell 2003 12 971 982 14580347 10.1016/S1097-2765(03)00388-5
Michels AA Fraldi A Li Q Binding of the 7SK snRNA turns the HEXIM1 protein into a P-TEFb (CDK9/cyclin T) inhibitor Embo J 2004 23 2608 2619 15201869 10.1038/sj.emboj.7600275
Chen R Yang Z Zhou Q Phosphorylated positive transcription elongation factor b (P-TEFb) is tagged for inhibition through association with 7SK snRNA J Biol Chem 2004 279 4153 4160 14627702 10.1074/jbc.M310044200
Yik JH Chen R Pezda AC Samford CS Zhou Q A human immunodeficiency virus type 1 Tat-like arginine-rich RNA-binding domain is essential for HEXIM1 to inhibit RNA polymerase II transcription through 7SK snRNA-mediated inactivation of P-TEFb Mol Cell Biol 2004 24 5094 5105 15169877 10.1128/MCB.24.12.5094-5105.2004
Byers SA Price JP Cooper JJ Li Q Price DH HEXIM2, A HEXIM1 related protein, regulates P-TEFb through association with 7SK J Biol Chem 2005 280 16360 16367 15713662 10.1074/jbc.M500424200
Yik JH Chen R Pezda AC Zhou Q Compensatory contributions of HEXIM1 and HEXIM2 in maintaining the balance of active and inactive P-TEFb complexes for control of transcription J Biol Chem 2005 280 16368 16376 15713661 10.1074/jbc.M500912200
Majello B Napolitano G Giordano A Lania L Transcriptional regulation by targeted recruitment of cyclin-dependent CDK9 kinase in vivo Oncogene 1999 18 4598 4605 10467404 10.1038/sj.onc.1202822
Taube R Lin X Irwin D Fujinaga K Peterlin BM Interaction between P-TEFb and the C-terminal domain of RNA polymerase II activates transcriptional elongation from sites upstream or downstream of target genes Mol Cell Biol 2002 22 321 331 11739744 10.1128/MCB.22.1.321-331.2002
Murphy S Yoon JB Gerster T Roeder RG Oct-1 and Oct-2 potentiate functional interactions of a transcription factor with the proximal sequence element of small nuclear RNA genes Mol Cell Biol 1992 12 3247 3261 1535687
Luecke HF Yamamoto KR The glucocorticoid receptor blocks P-TEFb recruitment by NFkappaB to effect promoter-specific transcriptional repression Genes Dev 2005 19 1116 1127 15879558 10.1101/gad.1297105
Sano M Wang SC Shirai M Activation of cardiac Cdk9 represses PGC-1 and confers a predisposition to heart failure Embo J 2004 23 3559 3569 15297879 10.1038/sj.emboj.7600351
Schulte A Czudnochowski N Barboric M Identification of a cyclin T-binding domain in Hexim1 and biochemical analysis of its binding competition with HIV-1 Tat J Biol Chem 2005 280 24968 24077 15855166 10.1074/jbc.M501431200
|
15992410
|
PMC1183248
|
CC BY
|
2021-01-04 16:36:39
|
no
|
Retrovirology. 2005 Jul 2; 2:42
|
utf-8
|
Retrovirology
| 2,005 |
10.1186/1742-4690-2-42
|
oa_comm
|
==== Front
Respir ResRespiratory Research1465-99211465-993XBioMed Central London 1465-9921-6-851604864710.1186/1465-9921-6-85ResearchRole of contractile prostaglandins and Rho-kinase in growth factor-induced airway smooth muscle contraction Schaafsma Dedmer [email protected] Reinoud [email protected] I Sophie T [email protected] Herman [email protected] Johan [email protected] S Adriaan [email protected] Department of Molecular Pharmacology, University of Groningen, Antonius Deusinglaan 1, 9713 AV Groningen, The Netherlands2005 27 7 2005 6 1 85 85 13 5 2005 27 7 2005 Copyright © 2005 Schaafsma et al; licensee BioMed Central Ltd.2005Schaafsma et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
In addition to their proliferative and differentiating effects, several growth factors are capable of inducing a sustained airway smooth muscle (ASM) contraction. These contractile effects were previously found to be dependent on Rho-kinase and have also been associated with the production of eicosanoids. However, the precise mechanisms underlying growth factor-induced contraction are still unknown. In this study we investigated the role of contractile prostaglandins and Rho-kinase in growth factor-induced ASM contraction.
Methods
Growth factor-induced contractions of guinea pig open-ring tracheal preparations were studied by isometric tension measurements. The contribution of Rho-kinase, mitogen-activated protein kinase (MAPK) and cyclooxygenase (COX) to these reponses was established, using the inhibitors Y-27632 (1 μM), U-0126 (3 μM) and indomethacin (3 μM), respectively. The Rho-kinase dependency of contractions induced by exogenously applied prostaglandin F2α (PGF2α) and prostaglandin E2 (PGE2) was also studied. In addition, the effects of the selective FP-receptor antagonist AL-8810 (10 μM) and the selective EP1-antagonist AH-6809 (10 μM) on growth factor-induced contractions were investigated, both in intact and epithelium-denuded preparations. Growth factor-induced PGF2α-and PGE2-release in the absence and presence of Y-27632, U-0126 and indomethacin, was assessed by an ELISA-assay.
Results
Epidermal growth factor (EGF)-and platelet-derived growth factor (PDGF)-induced contractions of guinea pig tracheal smooth muscle preparations were dependent on Rho-kinase, MAPK and COX. Interestingly, growth factor-induced PGF2α-and PGE2-release from tracheal rings was significantly reduced by U-0126 and indomethacin, but not by Y-27632. Also, PGF2α-and PGE2-induced ASM contractions were largely dependent on Rho-kinase, in contrast to other contractile agonists like histamine. The FP-receptor antagonist AL-8810 (10 μM) significantly reduced (approximately 50 %) and the EP1-antagonist AH-6809 (10 μM) abrogated growth factor-induced contractions, similarly in intact and epithelium-denuded preparations.
Conclusion
The results indicate that growth factors induce ASM contraction through contractile prostaglandins – not derived from the epithelium – which in turn rely on Rho-kinase for their contractile effects.
==== Body
Background
Growth factors have been reported to be involved in proliferation and differentiation of smooth muscle cells from a variety of tissues, including vasculature and airways [1,2]. In addition, several growth factors have been shown to induce contraction of vascular smooth muscle [3,4]. The mechanisms by which growth factors induce contraction have only been partly elucidated. Recent evidence has indicated that growth factor-receptors, such as the insulin-like growth factor-1 (IGF-1)-receptor, can activate the Rho/Rho-kinase pathway directly [5] and may be involved in smooth muscle contraction via Rho-kinase [6]. Smooth muscle contraction is mainly regulated by the phosphorylation level of the 20 kDa regulatory myosin light chain (MLC) [7]. MLC phosphorylation can be initiated by an increase in intracellular Ca2+-concentration ([Ca2+]i) followed by the Ca2+-calmodulin-dependent activation of myosin light chain kinase (MLCK). The extent of MLC phosphorylation is determined by the ratio of MLCK (MLC-phosphorylation) to myosin light chain phosphatase (MLCP)(MLC-dephosphorylation) activities [8]. Activated Rho-kinase mainly exerts its effect through inhibition of MLCP, resulting in an enhanced MLC phosphorylation and thus an increased level of contraction at a fixed [Ca2+]i (Ca2+-sensitization) [6,9].
In bovine airway smooth muscle, it has been demonstrated that prolonged incubation with growth factors modulates the phenotypic state of the muscle [10,11]. They have also been described to exert acute contractile effects on guinea pig tracheal smooth muscle [12,13]. Recently, we showed that growth factors are also capable of inducing human bronchial smooth muscle contraction. Thus, angiotensin II as well as IGF-1 induced a sustained contraction, which was completely dependent on Rho-kinase [14].
These observations may be of pathophysiological and pharmacotherapeutical interest, as expression levels both of growth factors (EGF)[15] and of receptors of growth factors (EGF[15], PDGF[15,16]) have been found elevated in asthmatic airways. Also, increased levels of PDGF have been found in exhaled breath condensate of asthmatic children with severe airflow limitation [17]. Moreover, previous studies showed an augmented role of Rho-kinase in acetylcholine induced bronchial smooth muscle contraction after repeated allergen challenge in rats [18,19]. Furthermore, we have recently demonstrated that the process of active allergic sensitization by itself, without subsequent allergen exposure, is sufficient to induce an enhanced role of Rho-kinase in guinea pig airway smooth muscle contraction ex vivo and airway resistance in vivo [20]. Therefore, a better understanding of the mechanisms by which growth factors induce a Rho-kinase dependent contraction is of pathophysiological and pharmacotherapeutical interest.
Epidermal growth factor (EGF) causes contraction of guinea pig tracheal smooth muscle via arachidonic acid metabolism in which presumably a tyrosine kinase and phospholipase A2 are involved [12,13]. It is well documented that receptor tyrosine kinases can activate mitogen-activated protein kinase (MAPK)/extracellular signal-regulated kinase (ERK)-kinase (MEK)[21-23]. Activation of MAPK by MEK may result in the activation of cytosolic phospholipase A2 (cPLA2) [24] and subsequent production of arachidonic acid and prostaglandins. Several studies have demonstrated that contractile prostaglandins are dependent on Rho-kinase [20,25]. Altogether, it can be hypothesized that growth factor-induced contraction is mediated via the MEK-dependent, cPLA2-mediated production of prostaglandins and subsequent activation of Rho-kinase. Therefore, we investigated the effects of inhibition of Rho-kinase, MEK and cyclooxygenase (COX) on growth factor-induced prostaglandin-production and contraction, using guinea pig tracheal smooth muscle preparations. In addition, we investigated the effects of selective prostaglandin receptor antagonists on growth factor-induced contraction.
Methods
Animals
Outbred specified pathogen-free male Dunkin Hartley guinea pigs (Harlan, Heathfield, U.K.), weighing 500–700 g, were used in this study. All protocols described in this study were approved by the University of Groningen Committee for Animal Experimentation.
Isometric tension measurements
After experimental concussion and rapid exsanguination the trachea was removed and transferred to Krebs-Henseleit (KH) buffer solution (composition in mM: NaCl 117.5, KCl 5.6, MgSO4 1.18, CaCl2 2.5, NaH2PO4 1.28, NaHCO3 25.00 and D-glucose 5.55; pregassed with 95% O2 and 5% CO2; pH 7.4) at 37°C. The trachea was carefully prepared free of serosa and connective tissue. In some cases, the airway epithelium was carefully removed by moving a 15-cm woollen thread up and down the trachea twice. Epithelium denudation was confirmed by histological examination after fixating cryostat sections (5 μm) in acetone and staining with hematoxylin eosin. Single open-ring tracheal preparations were prepared and mounted for isometric recording, using Grass FT-03 transducers, in 20 ml water-jacketed organ baths (37°C) containing KH solution. During a 90 min equilibration period, with washouts every 30 min, resting tension was gradually adjusted to 0.5 g. Subsequently, the preparations were precontracted with 20 and 40 mM KCl. Following two wash-outs, maximal relaxation was established by the addition of 0.1 μM isoprenaline and tension was re-adjusted to 0.5 g, immediately followed by two changes of fresh KH-buffer. After another equilibration period of 30 min EGF (0.1, 1, 3, 10 or 30 ng/ml) or PDGF (0.1, 1, 3, 10 or 30 ng/ml) was applied or cumulative concentration response curves (CRCs) were constructed to stepwise increasing concentrations of histamine (1 nM – 100 μM), PGE2 (1 nM – 3μM) or PGF2α (1 nM – 10 μM). When maximal agonist-induced contraction was obtained, the tracheal rings were washed several times and maximal relaxation was established using isoprenaline. When used, the inhibitors of Rho-kinase (Y-27632, 1 μM), MAPK-ERK-kinase (MEK) (U-0126, 3 μM) or COX (indomethacin, 3 μM) were applied to the organ bath 30 min before agonist addition. This was also the case for the FP-receptor-and EP1-receptor-antagonists AL-8810 and AH-6809 (10 μM both, applied individually to separate preparations), respectively.
Measurement of prostaglandin F2α and prostaglandin E2 production
Guinea pig tracheal rings were incubated using a 24-wells plate at 37°C. Each well contained 1 ml KH-buffer and 7 tracheal rings. Twenty-one rings were isolated from every trachea, so three conditions per preparation could be tested. Following a 30 min pre-incubation period, 100 μl of the medium was taken as the first sample. Subsequently, PDGF (10 ng/ml) was applied. To determine the time dependency of prostaglandin (PG)-production, samples were collected at 5, 10, 15, 20 and 30 min after PDGF-addition. Sampling was performed under a 95 % O2 / 5 % CO2 atmosphere. PGF2α-and PGE2-production was determined using an ELISA-assay according to the manufacturer's protocol (R&D Systems, U.K.).
Data analysis
All data represent means ± s.e. mean from n separate experiments. Statistical significance of differences was evaluated using either a one way analysis of variance (ANOVA) followed by a Bonferroni post-test or by a paired or unpaired two-tailed Student's t-test when appropriate, and significance was accepted when P<0.05.
Chemicals
Platelet-derived growth factor AB (PDGF-AB, human recombinant) was from Bachem (Bubendorf, Switzerland) and epidermal growth factor (human recombinant), indomethacin, histamine dihydrochloride and (-)-isoprenaline hydrochloride were obtained from Sigma Chemical Co. (St. Louis, MO, U.S.A.). PGF2α was obtained from Pharmacia and Upjohn (Puurs, Belgium) and PGE2 was from BIOMOL (U.S.A). 1,4-diamino-2,3-dicyano-1,4-bis [2-aminophenylthio]butadiene (U-0126), (+)-(R)-trans-4-(1-aminoethyl)-N-(4-pyridyl) cyclohexane carboxamide (Y-27632) and 6-isopropoxy-9-xanthone-2-carboxylic acid (AH-6809) were obtained from Tocris Cookson Ltd. (Bristol, U.K.). 9α, 15R-dihydroxy-1 1β-fluoro-15-(2,3-dihydro-1H-inden-2-yl)-16, 17, 18, 19, 20-pentanor-prosta-5Z, 13E-dien-1-oic acid (AL-8810) was obtained from Cayman Chemical (U.S.A). All other chemicals were of analytical grade.
Results
To investigate the contractile effects of EGF and PDGF on guinea pig tracheal smooth muscle, CRCs of the growth factors were constructed (Fig. 1). Both EGF and PDGF were capable of inducing concentration-dependent contractions, with a potency (EC50) of 6.7 ± 2.3 ng/ml for EGF and 6.4 ± 2.8 ng/ml for PDGF. As shown in Fig. 2, both growth factors induced a slowly developing sustained contraction, which was prevented almost completely in the presence of either Y-27632 (1 μM), U-0126 (3 μM) or indomethacin (3 μM). Also, basal myogenic tone (expressed with respect to maximal relaxation established with isoprenaline) was abolished by these inhibitors (Fig. 2).
Figure 1 EGF (A)-and PDGF (B)-induced contraction of guinea pig open-ring tracheal smooth muscle preparations. Responses shown are corrected for basal myogenic tone, which amounted to 0.21 ± 0.06 g on average (0 % growth factor effect). Maximal effects were reached at concentrations of 30 ng/ml and amounted 0.31 ± 0.04 g (EGF, 100 % effect) and 0.24 ± 0.06 g (PDGF, 100 % effect), corresponding to 19.8 ± 2.8 % and 15.3 ± 4.1 %, respectively, of maximal histamine-induced contraction. Data represent means ± s.e. mean of seven (EGF) or four (PDGF) experiments, each performed in duplicate.
Figure 2 Effects of Y-27632 (1 μM), U-0126 (3 μM) and indomethacin (3 μM) on (A) EGF (10 ng/ml)-and (B) PDGF (10 ng/ml)-induced guinea pig trachealsmooth muscle contraction. Data represent means ± s.e. mean of five (EGF) or six (PDGF) experiments, each performed in duplicate.
Since both MEK-(U-0126) and COX-inhibition (indomethacin) prevented growth factor-induced contraction, we envisaged that growth factor-induced prostaglandin production would be responsible for the observed contractions.
Stimulation of tracheal smooth muscle preparations with PDGF for 30 min greatly enhanced the release of prostaglandin E2 (PGE2) by 255 ± 78 % (from 963 ± 245 to 2762 ± 138 pg/ml; p < 0.01 at t = 30 min; Fig. 3A) and prostaglandin F2α (PGF2α) by 182 ± 38 % (from 1093 ± 204 to 2929 ± 570 pg/ml; p < 0.05 at t = 30 min; Fig. 3B). As shown in Fig. 4, both the release of PGE2 and PGF2α were significantly reduced in the presence of U-0126 (3 μM) and indomethacin (3 μM). In contrast to growth factor-induced contraction, no significant effect of treatment with Y-27632 (1 μM) was found on PGE2 (p = 0.23) or PGF2α (p = 0.08) release. These findings would suggest that prostaglandins produced in response to growth factor stimulation are capable of inducing a Rho-kinase-dependent contraction. Application of PGE2 caused ASM contraction in concentrations up to 0.03 μM (pEC50 = 8.22 ± 0.07, Emax = 58.3 ± 11.2 %), but caused relaxation in higher concentrations (Fig. 5A). Indeed, Rho-kinase inhibition resulted in a decreased potency (pEC50 = 7.9 ± 0.2; p < 0.05) and maximal contraction (Emax = 11.7 ± 3.5 %; p < 0.05) of PGE2-induced contraction. PGF2α-induced contractions (pEC50 = 6.8 ± 0.2; Emax = 71.9 ± 8.2 %) were dependent on Rho-kinase as well, as indicated by the significantly decreased potency (pEC50 = 6.2 ± 0.2 ; p < 0.05) and maximal contraction (Emax = 41.8 ± 9.3 %; p < 0.05) after treatment with Y-27632 (Fig. 5B).
Figure 3 Growth factor-induced PGE2 (A) and PGF2α (B) release from guinea pig tracheal smooth muscle preparations. Basal release amounted to 963 ± 245 pg/ml (PGE2) and 1093 ± 204 pg/ml (PGF2α). Data represent means ± s.e.mean of five experiments.
Figure 4 Effects of U-0126 (3 μM), indomethacin (3 μM) and Y-27632 (1 μM) on growth factor-induced PGE2 (A) and PGF2α (B) release. Data represent means ± s.e.mean of six (PGE2) and five (PGF2α) experiments. *p < 0.05, **p < 0.01 compared to PDGF.
Figure 5 Effects of Rho-kinase inhibition on prostaglandin-induced contraction. PGE2 (A)-and PGF2α (B)-induced contraction in the absence and presence of Y-27632 (1 μM) of guinea pig open-ring tracheal smooth muscle preparations. Data represent means ± s.e.mean of four (PGE2) and seven (PGF2α) experiments, each performed in duplicate.
To establish the functional contribution of the contractile PGE2-sensitive EP1-receptor and the PGF2α-sensitive FP-receptor to growth factor-induced contraction, the selective EP1-receptor antagonist AH-6809 (10 μM) and the selective FP-receptor antagonist AL-8810 (10 μM) were used. Both EGF-and PDGF-induced contractions were significantly reduced after treatment with AL-8810 (46,7 ± 13.0 % and 52.7 ± 13.2 % inhibition, respectively; p < 0.01 both), whereas contractions were almost abolished after treatment with AH-6809 (95.1 ± 3.1 % and 94.4 ± 4.7 % inhibition, respectively; p < 0.001 both)(Fig. 6A,B). To determine whether the epithelium was the source of the prostaglandins involved in growth factor-induced contraction, the effects of AL-8810 and AH-6809 on epithelium-denuded tracheal preparations were studied. Complete denudation was achieved as illustrated in Fig. 7. In these preparations, PDGF induced a slightly higher contraction compared to that in intact preparations, however the difference was not significant. Similar to intact preparations, PDGF-induced contraction was significantly reduced by both AL-8810 (48.8 ± 7.1 % inhibition; p < 0.05; Fig. 6C) and AH-6809 (92.1 ± 3.0 % inhibition; p < 0.01); Fig. 6C). Moreover, the inhibition in denuded preparations was very similar to that in intact preparations, both for AL-8810 and AH-6809, indicating that FP-and EP1-receptor stimulation involved in growth factor-induced contraction occurs independently of epithelium.
Figure 6 EGF (10 ng/ml, A)-and PDGF (10 ng/ml, B,C)-induced contraction of intact (A,B) and epithelium-denuded (C) guinea pig open-ring tracheal smooth muscle preparations in the absence or presence of AL-8810 (10 μM) or AH-6809 (10 μM). Data represent means ± s.e. mean of five (A,B) and three (C) experiments, each performed in duplicate. *p < 0.05, **p < 0.01 and ***p < 0.001 compared to control.
Figure 7 Representative photomicrograph of an intact (A) and epithelium-denuded (B) tracheal preparation. The photographs were taken at 100 × magnification.
Discussion
In this study we demonstrate that the growth factors EGF and PDGF induce contractions of guinea pig tracheal smooth muscle in a concentration dependent fashion. The concentration-effect range of EGF and PDGF (0.1 – 30 ng/ml) represents a pharmacological range very similar to other effects, such as mitogenesis of airway smooth muscle [26,10]. Since contractile effects of EGF have previously been associated with the production of eicosanoids [13] and contractions induced by of IGF-1 and angiotensin II appeared to be dependent on Rho-kinase [13,14], we analyzed whether contractions induced by submaximal concentrations of growth factors are dependent not only on Rho-kinase, but also on COX and MEK. This might be characteristic for growth factor-induced contraction, since potency and maximal contraction of histamine were shown to be independent of Rho-kinase, COX [20] and MEK (Schaafsma et al, unpublished observations) in guinea pig tracheal smooth muscle. Similarly, muscarinic receptor mediated contractions are only partially Rho-kinase-dependent [27,28], further illustrating the agonist-dependent role of Rho-kinase mediated calcium sensitization.
The role of Rho-kinase in growth factor-mediated effects could depend on the duration of growth factor stimulation. For instance, phenotypic modulation, as a consequence of 8 days stimulation with growth factors, or growth factor-induced proliferation of bovine tracheal smooth muscle, has been shown to be independent of Rho-kinase. However, in accordance with the effects of Rho-kinase inhibition on growth factor-induced contraction of human isolated bronchus [14], we demonstrate that Y-27632 fully inhibits growth factor-induced contraction of guinea pig tracheal smooth muscle. This indicates that growth factor-induced acute (smooth muscle contraction) and long term (e.g. modulation of smooth muscle phenotype) effects in airway smooth muscle may be differentially dependent on Rho-kinase.
Since MEK and COX inhibition almost abrogated growth factor-induced contraction, it can be suggested that growth factor-induced contraction relies on the production of prostaglandins. In several studies, it has been demonstrated that cytosolic phospholipase A2 (PLA2) can be activated in response to growth factors in a MAPK-dependent fashion, which results in subsequent arachidonic acid production [29-31]. In addition, contractile activity of EGF in guinea pig tracheal smooth muscle has been reported to be inhibited by indomethacin and by the phospholipase A2 inhibitor mepacrine [12]. As indicated by our results, PGF2α and PGE2 are being produced in response to PDGF-stimulation in a time-dependent fashion, similar to that of growth factor-induced contraction. Both prostaglandins are contractile agonists for airway smooth muscle [20,32,33]. Contractions induced by (exogenous) PGF2α and PGE2 were found to be largely dependent on Rho-kinase activity, which corresponds to observations in vascular smooth muscle [25,34], indicating that Rho-kinase plays an essential role in PGF2α-and PGE2-induced contractions. Interestingly, Rho-kinase inhibition had a more pronounced effect on PGE2-than on PGF2α-induced contractions. This can be explained, however, by realizing that the EP2-receptor mediated relaxation [35], as seen with the higher PGE2-concentrations, is suppressing the contractile phase more effectively when its Rho-kinase-dependent component is being inhibited.
In addition to direct contractile effects on guinea pig airway smooth muscle, PGF2α has been shown to augment cholinergic responsiveness of bovine airway smooth muscle [36], indicating an important role for PGF2α in regulating airway smooth muscle tone. PGF2α has been described to exert its contractile effects on smooth muscle through the FP-receptor [37,38]. Also, PGF2α-induced Ca2+-mobilization in vascular smooth muscle cells was dose-dependently inhibited by the selective FP-receptor antagonist AL-8810 [39]. In our study, a selective and effective concentration of AL-8810 [40,39] reduced EGF-and PDGF-induced contractions, indicating that PGF2α contributes to growth factor-induced contraction through the FP-receptor.
Smooth muscle contractions induced by PGE2 are predominantly mediated through activation of the EP1-receptor [41,32]. In guinea pig airway smooth muscle it has been previously found that PGE2-induced contractions could be dose-dependently inhibited by the EP1-receptor antagonist SC-19220 without modulating the relaxant activity (Van Amsterdam, 1991). Also, like PGF2α, PGE2 enhances cholinergic airway responsiveness of bovine airway smooth muscle [36]. In the present study we found that growth factor-induced contraction of guinea pig tracheal smooth muscle is essentially dependent on EP1-receptor stimulation, since the selective EP1-receptor antagonist AH-6809 [36] abrogated growth factor-induced contractions. Interestingly, these contractions were partially inhibited by FP-receptor blockade as well. From these observations, it may be hypothesized that PGF2α-mediated contractions partially rely on EP1-receptor stimulation (possibly by releasing small amounts of PGE2, selectively activating EP1-receptors) and that synergistic contractile effects of concomitant EP1-and FP-receptor stimulation occur.
Several growth factors, including EGF and PDGF, have been implicated in airway inflammation as they can be released from inflammatory cells, such as macrophages and eosinophils. Moreover, they can be derived from extravasated plasma, epithelial cells and the airway smooth muscle itself [2,42]. Growth factors are involved in tissue repair processes, therefore growth factor-induced contraction could protect damaged areas in the airways from the environment during these processes. In the pathophysiology of asthma, the repair process is usually not restricted to a single segment of the airways and growth factors may then contribute to airflow obstruction. Inhibition of such contractions might therefore be relevant under such pathophysiological conditions.
Conclusion
Our overall results indicate that EGF and PDGF induce airway smooth muscle contraction through contractile prostaglandins. These prostaglandins are presumably produced by the consecutive actions of MEK, cytosolic PLA2 and COX and in turn are dependent on Rho-kinase for their contractile effects (Fig. 8). Since growth factor-induced contractions were inhibited by antagonists of contractile prostaglandin receptors both in intact and epithelium-denuded preparations, it can be concluded that the prostaglandins involved in growth factor-induced contraction are not primarily derived from the epithelium. Since both growth factors and increased Rho-kinase activity are associated with pathophysiological conditions and growth factor-induced contraction is fully Rho-kinase dependent, inhibition of Rho-kinase might be of therapeutical interest in the treatment of inflammatory (airway) diseases.
Figure 8 Putative mechanism of growth factor-induced airway smooth muscle contraction. Growth factors, like EGF and PDGF, bind to their receptors with intrinsic tyrosine kinase activity (RTK) and activate MAPK, which may result in increased levels of arachidonic acid (AA) via cytosolic phospholipase A2 (cPLA2) activation. As a consequence of cyclooxygenase (COX)-mediated conversion of AA, prostaglandins (PGs) are produced. These (contractile) prostaglandins, like PGF2α and PGE2, may in turn couple to their receptors and induce an airway smooth muscle contraction which is largely dependent on Rho-kinase. U-0126, indomethacin (indo) and Y-27632 are inhibitors of MAPK, COX and Rho-kinase, respectively.
Abbreviations
AA, arachidonic acid; AHR, airway hyperresponsiveness; ASM, airway smooth muscle; COX, cyclooxygenase; cPLA2, cytosolic phospholipase A2; CRC, cumulative concentration response curve; EGF, epidermal growth factor; EP1-receptor, prostaglandin E2-receptor type 1; FP-receptor, prostaglandin F2α-receptor; IGF-1, insulin-like growth factor-1; Indo, indomethacin; KH, Krebs-Henseleit; MAPK, mitogen-activated protein kinase; MEK, mitogen-activated protein kinase/extracellular signal-regulated kinase-kinase (MEK); pEC50, -log10 of the concentration causing 50 % of the effect; PDGF, platelet-derived growth factor; PG, prostaglandin; PGE2, prostaglandin E2; PGF2α, prostaglandin F2α; RTK, receptors with intrinsic tyrosine kinase activity
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
DS designed and coordinated the study, performed a major part of the experiments, performed the statistical analysis and drafted the manuscript. RG participated in the design of the study, assisted in performing part of the experiments and contributed to the preparation of the manuscript. ISTB substantially assisted in performing the experiments. HM participated in the design of the study and the interpretation of the results. JZ participated in the design of the study, interpretation of results and final revision of the manuscript. SAN supervised the study, participated in its design and in the preparation of the manuscript. All authors read and approved the final manuscript.
Acknowledgements
We thank the Netherlands Asthma Foundation for financial support (grant 01.83).
==== Refs
Bayes-Genis A Conover CA Schwartz RS The insulin-like growth factor axis: A review of atherosclerosis and restenosis Circ Res 2000 86 125 130 10666406
Hirst SJ Airway smooth muscle as a target in asthma Clin Exp Allergy 2000 30 Suppl 1 54 59 10849477 10.1046/j.1365-2222.2000.00099.x
Sauro MD Thomas B Tyrphostin attenuates platelet-derived growth factor-induced contraction in aortic smooth muscle through inhibition of protein tyrosine kinase(s) J Pharmacol Exp Ther 1993 267 1119 1125 8263773
Berk BC Alexander RW Brock TA Gimbrone MAJ Webb RC Vasoconstriction: a new activity for platelet-derived growth factor Science 1986 232 87 90 3485309
Taya S Inagaki N Sengiku H Makino H Iwamatsu A Urakawa I Nagao K Kataoka S Kaibuchi K Direct interaction of insulin-like growth factor-1 receptor with leukemia-associated RhoGEF J Cell Biol 2001 155 809 820 11724822 10.1083/jcb.200106139
Fukata Y Amano M Kaibuchi K Rho-Rho-kinase pathway in smooth muscle contraction and cytoskeletal reorganization of non-muscle cells Trends Pharmacol Sci 2001 22 32 39 11165670 10.1016/S0165-6147(00)01596-0
Pfitzer G Invited review: regulation of myosin phosphorylation in smooth muscle J Appl Physiol 2001 91 497 503 11408468
SOMLYO ANDREWP SOMLYO AVRILV Ca2+ Sensitivity of Smooth Muscle and Nonmuscle Myosin II: Modulated by G Proteins, Kinases, and Myosin Phosphatase Physiol Rev 2003 83 1325 1358 14506307
Wettschureck N Offermanns S Rho/Rho-kinase mediated signaling in physiology and pathophysiology J Mol Med 2002 80 629 638 12395147 10.1007/s00109-002-0370-2
Gosens R Meurs H Bromhaar MM McKay S Nelemans SA Zaagsma J Functional characterization of serum- and growth factor-induced phenotypic changes in intact bovine tracheal smooth muscle Br J Pharmacol 2002 137 459 466 12359627 10.1038/sj.bjp.0704889
Gosens R Nelemans SA Hiemstra M Grootte Bromhaar MM Meurs H Zaagsma J Insulin induces a hypercontractile airway smooth muscle phenotype Eur J Pharmacol 2003 481 125 131 14637184 10.1016/j.ejphar.2003.08.081
Patel P Itoh H Lederis K Hollenberg MD Contraction of guinea pig trachea by epidermal growth factor--urogastrone Can J Physiol Pharmacol 1988 66 1308 1312 3149204
Nasuhara Y Munakata M Sato A Amishima M Homma Y Kawakami Y Mechanisms of epidermal growth factor-induced contraction of guinea pig airways Eur J Pharmacol 1996 296 161 168 8838452 10.1016/0014-2999(95)00692-3
Gosens R Schaafsma D Grootte Bromhaar MM Vrugt B Zaagsma J Meurs H Nelemans SA Growth factor-induced contraction of human bronchial smooth muscle is Rho-kinase-dependent Eur J Pharmacol 2004 494 73 76 15194453 10.1016/j.ejphar.2004.04.035
Amishima M Munakata M Nasuhara Y Sato A Takahashi T Homma Y Kawakami Y Expression of epidermal growth factor and epidermal growth factor receptor immunoreactivity in the asthmatic human airway Am J Respir Crit Care Med 1998 157 1907 1912 9620926
Lewis CC Chu HW Westcott JY Tucker A Langmack EL Sutherland ER Kraft M Airway fibroblasts exhibit a synthetic phenotype in severe asthma J Allergy Clin Immunol 2005 115 534 540 15753901 10.1016/j.jaci.2004.11.051
Leung TF Wong GW Ko FW Li CY Yung E Lam CW Fok TF Analysis of Growth Factors and Inflammatory Cytokines in Exhaled Breath Condensate from Asthmatic Children Int Arch Allergy Immunol 2005 137 66 72 15832052 10.1159/000085106
Chiba Y Takada Y Miyamoto S MitsuiSaito M Karaki H Misawa M Augmented acetylcholine-induced, Rho-mediated Ca2+ sensitization of bronchial smooth muscle contraction in antigen-induced airway hyperresponsive rats Br J Pharmacol 1999 127 597 600 10401547 10.1038/sj.bjp.0702585
Chiba Y Sakai H Suenaga H Kamata K Misawa M Enhanced Ca2+ sensitization of the bronchial smooth muscle contraction in antigen-induced airway hyperresponsive rats Res Commun Mol Pathol Pharmacol 1999 106 77 85 11127810
Schaafsma D Gosens R Bos IS Meurs H Zaagsma J Nelemans SA Allergic sensitization enhances the contribution of Rho-kinase to airway smooth muscle contraction Br J Pharmacol 2004 143 477 484 15381630 10.1038/sj.bjp.0705903
Zwick E Hackel PO Prenzel N Ullrich A The EGF receptor as central transducer of heterologous signalling systems Trends Pharmacol Sci 1999 20 408 412 10577253 10.1016/S0165-6147(99)01373-5
Lopez-Ilasaca M Signaling from G-protein-coupled receptors to mitogen-activated protein (MAP)-kinase cascades Biochem Pharmacol 1998 56 269 277 9744561 10.1016/S0006-2952(98)00059-8
Fischer OM Streit S Hart S Ullrich A Beyond Herceptin and Gleevec Curr Opin Chem Biol 2003 7 490 495 12941424 10.1016/S1367-5931(03)00082-6
Lin LL Wartmann M Lin AY Knopf JL Seth A Davis RJ cPLA2 is phosphorylated and activated by MAP kinase Cell 1993 72 269 278 8381049 10.1016/0092-8674(93)90666-E
Ito K Shimomura E Iwanaga T Shiraishi M Shindo K Nakamura J Nagumo H Seto M Sasaki Y Takuwa Y Essential role of rho kinase in the Ca2+ sensitization of prostaglandin F(2alpha)-induced contraction of rabbit aortae J Physiol 2003 546 823 836 12563007 10.1113/jphysiol.2002.030775
Kelleher MD Abe MK Chao TS Jain M Green JM Solway J Rosner MR Hershenson MB Role of MAP kinase activation in bovine tracheal smooth muscle mitogenesis Am J Physiol 1995 268 L894 L901 7611431
Gosens R Schaafsma D Meurs H Zaagsma J Nelemans SA Role of Rho-kinase in maintaining airway smooth muscle contractile phenotype Eur J Pharmacol 2004 483 71 78 14709328 10.1016/j.ejphar.2003.10.027
Janssen LJ Wattie J Lu-Chao H Tazzeo T Muscarinic excitation-contraction coupling mechanisms in tracheal and bronchial smooth muscles J Appl Physiol 2001 91 1142 1151 11509509
Margolis BL Holub BJ Troyer DA Skorecki KL Epidermal growth factor stimulates phospholipase A2 in vasopressin-treated rat glomerular mesangial cells Biochem J 1988 256 469 474 3146974
Bornfeldt KE Campbell JS Koyama H Argast GM Leslie CC Raines EW Krebs EG Ross R The mitogen-activated protein kinase pathway can mediate growth inhibition and proliferation in smooth muscle cells. Dependence on the availability of downstream targets J Clin Invest 1997 100 875 885 9259587
Boulven I Palmier B Robin P Vacher M Harbon S Leiber D Platelet-derived growth factor stimulates phospholipase C-gamma 1, extracellular signal-regulated kinase, and arachidonic acid release in rat myometrial cells: contribution to cyclic 3',5'-adenosine monophosphate production and effect on cell proliferation Biol Reprod 2001 65 496 506 11466218
Ndukwu IM White SR Leff AR Mitchell RW EP1 receptor blockade attenuates both spontaneous tone and PGE2-elicited contraction in guinea pig trachealis Am J Physiol 1997 273 L626 L633 9316498
Van Amsterdam RG Beta-adrenoceptor responsiveness in non-allergic and allergic airways - An in-vitro approach 1991 68 69
Shum WW Le GY Jones RL Gurney AM Sasaki Y Involvement of Rho-kinase in contraction of guinea-pig aorta induced by prostanoid EP3 receptor agonists Br J Pharmacol 2003 139 1449 1461 12922932 10.1038/sj.bjp.0705393
Tilley SL Hartney JM Erikson CJ Jania C Nguyen M Stock J McNeisch J Valancius C Panettieri RAJ Penn RB Koller BH Receptors and pathways mediating the effects of prostaglandin E2 on airway tone Am J Physiol Lung Cell Mol Physiol 2003 284 L599 L606 12618422
Catalli A Janssen LJ Augmentation of bovine airway smooth muscle responsiveness to carbachol, KCl, and histamine by the isoprostane 8-iso-PGE2 Am J Physiol Lung Cell Mol Physiol 2004 287 L1035 L1041 15257985 10.1152/ajplung.00138.2004
Coleman RA Smith WL Narumiya S International Union of Pharmacology classification of prostanoid receptors: properties, distribution, and structure of the receptors and their subtypes Pharmacol Rev 1994 46 205 229 7938166
Funk CD Prostaglandins and leukotrienes: advances in eicosanoid biology Science 2001 294 1871 1875 11729303 10.1126/science.294.5548.1871
Kelly CR Williams GW Sharif NA Real-time intracellular Ca2+ mobilization by travoprost acid, bimatoprost, unoprostone, and other analogs via endogenous mouse, rat, and cloned human FP prostaglandin receptors J Pharmacol Exp Ther 2003 304 238 245 12490597 10.1124/jpet.102.042556
Griffin BW Klimko P Crider JY Sharif NA AL-8810: a novel prostaglandin F2 alpha analog with selective antagonist effects at the prostaglandin F2 alpha (FP) receptor J Pharmacol Exp Ther 1999 290 1278 1284 10454504
Sametz W Hennerbichler S Glaser S Wintersteiger R Juan H Characterization of prostanoid receptors mediating actions of the isoprostanes, 8-iso-PGE(2) and 8-iso-PGF(2alpha), in some isolated smooth muscle preparations Br J Pharmacol 2000 130 1903 1910 10952681 10.1038/sj.bjp.0703522
McKay S Sharma HS Autocrine regulation of asthmatic airway inflammation: role of airway smooth muscle Respir Res 2002 3 11 11806846 10.1186/rr160
|
16048647
|
PMC1183249
|
CC BY
|
2021-01-04 16:23:27
|
no
|
Respir Res. 2005 Jul 27; 6(1):85
|
utf-8
|
Respir Res
| 2,005 |
10.1186/1465-9921-6-85
|
oa_comm
|
==== Front
Thromb JThrombosis Journal1477-9560BioMed Central London 1477-9560-3-71602273210.1186/1477-9560-3-7Book ReviewReview of "Thrombosis in Clinical Practice" Altman Raul [email protected] 15 7 2005 3 7 7 Thrombosis in Clinical Practice .
Taylor & Francis . Andrew D Blann, Gregory YH Lip, Alexander GG Turpie . 2005 . pp. ISBN 1-84214-163-5 . 1 7 2005 15 7 2005 Copyright © 2005 Blann et al; licensee BioMed Central Ltd.2005Blann et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
==== Body
As mentioned by the editors, the expectant readers of Thrombosis in Clinical Practice are physicians, general practitioners, nurse practitioners, and healthcare workers. With this in mind I browsed Thrombosis in Clinical Practice.
Thrombosis is a fast moving area and is not easy to give updated information of its pathophysiology and treatment in about 300 pages as it was successfully done in this book.
This multiauthored text combines basic concepts of mechanisms of thrombosis formation as well as therapeutic approaches.
It provides short overviews on risk factors, therapeutics and a systematic discussion of arterial and venous thrombosis management with final recommendations according to the authors experience. It successfully delivers concise information on all aspects of a given type of thrombosis for physicians and nurses who care for patients with thrombosis related diseases.
The first five chapters review the mechanism of thrombus formation (although importance of inflammation was omitted in Chapter 1), the role of the hemostatic system in arterial and venous thrombosis, characteristics of what the authors call "commonly used anticoagulant and antiplatelet drugs" (which include aspirin, heparin, and warfarin starting with an interesting historical perspective of each of them).
Without doubt the risk of bleeding is the most critical aspect of patients under antithrombotic treatment; "why it happens and what to do" are questions answered in Chapter 4.
Factors which determine thrombophilia are discussed in Chapter 5. Practical recommendations for screening patients with potential or already developed thrombotic process are noteworthy.
The following Chapters 6 to 12 deal with treatment of arterial and venous thrombosis and include management of atrial fibrillation where the known, well established therapies are discussed. In cardiac valves, Chapter 7, a comprehensive indication for the used of oral anticoagulant with or without aspirin and the management of pregnant patients was explained. Since some not so recent data have been included in the beginning of this chapter authors missed to mention that in patients with cardiac valves prostheses, the first combined therapy (oral anticoagulant plus dipyridamole) for thrombotic preventing treatment was published by Sullivan et al. (N Engl J Med. 1971;284:1391-4), and belongs to our group the first report on the use of aspirin plus oral anticoagulant (Altman et al. J.Thorac.Cardiov. Surg. 72: 127,1976).
Chapters 8 to 11 review the benefit and risk of antithrombotic therapy in coronary artery disease, thrombosis prevention after coronary interventions and, an updated discussion on risk factors, pathophysiology, diagnosis and antithrombotic therapy in peripheral arterial disease and ischemic stroke.
Chapter 12 deals with an important and still controversial topic: "Management of venous thromboembolism during pregnancy". The authors discuss the conflictive positions in the diagnostic as well as in the treatment of deep venous thrombosis in pregnant women. The antithrombotic treatment in women with recurrent miscarriage and thrombophilia which is conflictive and still unresolved is not discussed in this chapter.
Chapter 13 was dedicated to thrombophilia. I wonder why it was included at the end of the book and not among the first 5 Chapters. I also wonder why Ian Jennings omits to discuss whether the decrease of fibrinolysis is (or not) a potential thrombotic risk factor.
Importantly, a chapter on thrombosis in children was incorporated in this book and finally, a chapter on new drugs for anticoagulant therapy, which could be called "the after coumadin/heparin era", was also included with updated information on the pentasacharides fondaparinux and idraparinux, on drugs affecting factor VII activity, direct Factor Xa and direct thrombin inhibitors. Therapeutic characteristics of antiplatelet drugs is provide in this Chapter. This part of the article shows the feeble beneficial effect of clopidogrel plus aspirin, in reducing the combining risk of ischemic stroke, MI or vascular death compared with aspirin in several artery diseases and likely to be beneficial in patients underwent percutaneous coronary intervention. The antithrombotic capacity of glicoprotein IIb-IIIa (integrin α2b/β3) inhibitors, the potential value of von Willebrand factor antibody AJW 200 and thrombolytic therapy are also discussed.
This textbook is particularly interesting for general practitioners, medical students and trainees in haematology. It will be useful for physicians in several disciplines.
In short: It fulfils its stated purpose.
|
16022732
|
PMC1183250
|
CC BY
|
2021-01-04 16:36:24
|
no
|
Thromb J. 2005 Jul 15; 3:7
|
utf-8
|
Thromb J
| 2,005 |
10.1186/1477-9560-3-7
|
oa_comm
|
==== Front
Virol JVirology Journal1743-422XBioMed Central London 1743-422X-2-461591670610.1186/1743-422X-2-46ResearchRespiratory syncytial virus-induced acute and chronic airway disease is independent of genetic background: An experimental murine model Chávez-Bueno Susana [email protected]ías Asunción [email protected]ómez Ana M [email protected] Kurt D [email protected]íos Ana M [email protected] Mónica [email protected] Octavio [email protected] Hasan S [email protected] Division of Pediatric Infectious Diseases, Department of Pediatrics, The University of Texas Southwestern Medical Center at Dallas and Children's Medical Center Dallas, Dallas, Texas, USA2 Department of Pathology, The University of Texas Southwestern Medical Center at Dallas and Children's Medical Center Dallas, Dallas, Texas, USA2005 25 5 2005 2 46 46 26 4 2005 25 5 2005 Copyright © 2005 Chávez-Bueno et al; licensee BioMed Central Ltd.2005Chávez-Bueno et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Respiratory syncytial virus (RSV) is the leading respiratory viral pathogen in young children worldwide. RSV disease is associated with acute airway obstruction (AO), long-term airway hyperresponsiveness (AHR), and chronic lung inflammation. Using two different mouse strains, this study was designed to determine whether RSV disease patterns are host-dependent. C57BL/6 and BALB/c mice were inoculated with RSV and followed for 77 days. RSV loads were measured by plaque assay and polymerase chain reaction (PCR) in bronchoalveolar lavage (BAL) and whole lung samples; cytokines were measured in BAL samples. Lung inflammation was evaluated with a histopathologic score (HPS), and AO and AHR were determined by plethysmography.
Results
Viral load dynamics, histopathologic score (HPS), cytokine concentrations, AO and long-term AHR were similar in both strains of RSV-infected mice, although RSV-infected C57BL/6 mice developed significantly greater AO compared with RSV-infected BALB/c mice on day 5. PCR detected RSV RNA in BAL samples of RSV infected mice until day 42, and in whole lung samples through day 77. BAL concentrations of cytokines TNF-α, IFN-γ, and chemokines MIG, RANTES and MIP-1α were significantly elevated in both strains of RSV-infected mice compared with their respective controls. Viral load measured by PCR significantly correlated with disease severity on days 14 and 21.
Conclusion
RSV-induced acute and chronic airway disease is independent of genetic background.
Viral pneumoniamouse modelairway hyperresponsivenessPCRcytokines
==== Body
Background
Human respiratory syncytial virus (RSV) is classified in the genus Pneumovirus, subfamily Pneumovirinae, family Paramixoviridae; and is a major cause of lower respiratory tract infection (LRTI) in young children and the elderly [1]. RSV LRTI is associated with increased risk of long-term recurrent wheezing [2-5], however, the pathogenesis of this relationship is not well understood. RSV LRTI elicits a host response including the release of inflammatory mediators and recruitment of different cell populations. The genetic variability of the host response might partially explain the different susceptibilities of individual patients to the acute and long-term effects of RSV infection, as suggested by the higher rates of RSV hospitalization among Native American and Alaskan Native children compared with other groups [6].
Animal models facilitate the study of RSV-induced acute and long-term disease in a more controlled manner. Our laboratory has previously established a mouse model of RSV-induced acute and long-term airway disease [7]. The present studies were designed to characterize the influence of mouse genetic background and the dynamics of viral replication on the chronic manifestations of RSV infection. The BALB/c mouse strain is one of the most commonly used for RSV experimental models, however, C57BL/6 mice frequently provide background for transgenic strains of mice. Therefore characterizing and establishing a comprehensive model of acute and long-term RSV disease in C57BL/6 is essential to further understanding the pathogenesis of RSV disease.
Results
1. RSV alone induces airway obstruction (AO) and airway hyperresponsiveness (AHR) in both C57BL/6 and BALB/c mice
RSV infection alone, without allergic pre-sensitization induced AO in both strains of mice as demonstrated by significantly increased enhanced pause (Penh) values compared with uninfected controls. Baseline Penh values increased transiently on day 1 after RSV inoculation in both strains, decreased by day 2, but continued to be significantly greater than in controls. Airway obstruction increased again and peaked on day 5, when C57BL/6 RSV-infected mice showed significantly higher Penh values than RSV-infected BALB/c mice (p < 0.001) (Figure 1). AO decreased thereafter during the first two weeks after RSV inoculation but remained significantly greater than the respective controls in both strains, for 21 days in BALB/c and 28 days in C57BL/6 mice (Figure 1 inset). RSV infection also induced AHR in both strains as evidenced by a greater difference between pre- and post-methacholine Penh values (delta Penh) compared with controls. Significantly increased AHR was persistently present for 42 days post-inoculation in BALB/c mice, while C57BL/6 mice showed significantly increased AHR for up to 28 days post-inoculation (Figure 2).
Figure 1 Effect of RSV on airway obstruction (AO) in two mouse strains. BALB/c (△) and C57BL/6 (●) mice were inoculated with sterile 10% EMEM (control) and were compared with RSV A2 infected BALB/c (□) and C57BL/6 (◆) mice to evaluate differences in airway obstruction (AO), by measuring Penh via whole-body plethysmography. Penh values are presented as means ± SEM. Comparisons were made by t-test when data normally distributed, or by Mann-Whitney Rank sum test when data were not normally distributed.
Figure 2 Airway Hyperresponsiveness (AHR) in BALB/c and C57BL/6 Mice. Data presented as Delta Penh values which is the difference between pre-and post methacholine Penh for each group of mice, in sham inoculated () and RSV inoculated () BALB/c mice, and sham inoculated () and RSV inoculated () C57BL/6 mice from days 14 to 77. Values represent the mean SEM from 10–30 mice per group. Data shown are the result of four separate experiments. p < 0.05, comparison by t-test when data normally distributed, or by Mann-Whitney Rank sum test when data were not normally distributed.
2. C57BL/6 and BALB/c mice demonstrate acute and persistent inflammatory changes after RSV infection
RSV-inoculated C57BL/6 and BALB/c mice, compared with their controls, showed greater histopathologic scores (HPS) which peaked on day 5 after RSV inoculation (Figures 3, 4A–D). Although the acute inflammatory changes observed in both strains gradually declined, RSV-infected mice had significantly greater HPS than the sham-inoculated controls for up to 77 days post-inoculation (Figures 3, 4E and 4F).
Figure 3 Comparison of acute and long-term histopathologic scores after RSV inoculation. BALB/c (△) and C57BL/6 (●) mice were inoculated with sterile 10% EMEM (control) and were compared with RSV A2 infected BALB/c (□) and C57BL/6 (◆) mice. Serial formalin fixed lung samples were obtained between day 0 (+2 hours) and day 77 after inoculation. HPS scores are represented as means ± SEM. p < .05, by t-test when data normally distributed, or by Mann-Whitney Rank sum test when data were not normally distributed.
Figure 4 RSV induced histopathology. Lung specimens were harvested on days 5 and 77 from groups of C57BL/6 and BALB/c mice inoculated with sterile medium (control) or RSV. Sections from control C57BL/6 and BALB/c mice above (4A and 4B, respectively), show rare, scattered, small lymphocytic infiltrates on day 5 after inoculation with medium, similar to sections of control mice harvested 77 days after (not shown). Acute and chronic inflammatory infiltrates, surrounding airways and vessels are demonstrated in RSV-inoculated mice of both strains, on days 5 and 77 after inoculation (4C to F).
3. RSV infection induces similar cytokine production in the respiratory tract of C57BL/6 and BALB/c mice
BAL concentrations of TNF-α, IFN-γ, MIG, RANTES, and MIP-1α followed similar dynamics in both strains of mice during the acute phase of the infection (Figure 5A–E). Overall, there was a trend for greater BAL cytokine concentrations of IFN-γ, TNF-α, RANTES and MIP-1α in RSV-infected BALB/c mice compared with C57BL/6 mice. (Figure 5A–E). No significant differences were observed in BAL concentrations of IL-4 and IL-10 between controls and infected mice of both strains at any time point evaluated (data not shown).
Figure 5 Cytokine and chemokine concentrations in bronchoalveolar lavage (BAL) samples after RSV inoculation. BAL samples were obtained from BALB/c (△) and C57BL/6 (●) mice inoculated with sterile 10% EMEM (control), and RSV A2 infected BALB/c (□) and C57BL/6 (◆) mice, to measure concentrations of pro-inflammatory cytokines (A) IFN-γ and (B)TNF-α; and the chemokines (C) RANTES, (D) MIP-1α, and (E) MIG. Values presented in means ± SEM pg/ml. p < .05, by t-test when data were normally distributed, or by Mann-Whitney Rank sum test when data were not normally distributed.
4. RSV load dynamics
4a. RSV loads measured by plaque assay in BAL samples follow similar dynamics in both C57BL/6 and BALB/c mice
On day 1 after RSV inoculation, RSV loads in BAL samples from both C57BL/6 and BALB/c mice were significantly greater than in controls by plaque assay (Figure 6). Compared with day 1, plaque assay RSV loads peaked on days 3–5 after inoculation in both strains representing active viral replication (p = 0.002 for day 1 vs days 3–5 in BALB/c mice; ANOVA), and were significantly greater in BALB/c than in C57BL/6 mice (Figure 7). BAL RSV loads declined below the limit of detection by day 7 and remained undetectable through day 77 post-inoculation.
Figure 6 RSV loads in BAL samples measured by the plaque assay method. Groups of 4–16 BALB/c (△) and C57BL/6 (●) mice per group per time point were inoculated intranasally with sterile 10% EMEM (control), and were compared with RSV A2 infected BALB/c (□) and C57BL/6 (◆) mice. Viral load was determined by HEp-2 plaque assay in BAL samples. Data are presented as mean ± SEM Log10 PFU/ml of BAL. p < .05, by t-test when data normally distributed, or by Mann-Whitney Rank sum test when data were not normally distributed.
Figure 7 RSV loads in RSV infected BALB/c mice BAL and lung supernatant samples measured by PCR vs. plaque assay. RSV loads measured by PCR in BAL samples (▽) remain positive up to 42 days after inoculation, while viral loads measured by plaque assay (□) become negative on day 7 post-inoculation (upper panel). Viral load measured in lung supernatants by the plaque assay also become undetectable by day 7 after inoculation, whereas RSV loads measured by PCR in lung supernatants remain detectable throughout all the time points evaluated (lower panel). All pair-wise multiple comparisons made by One-Way ANOVA. † p < 0.01 between D1 and D5 and *p < 0.05 comparing D1 with D3 and D5.
4b. Real Time PCR (RLT-PCR) demonstrates RSV RNA after the virus is no longer detectable by plaque assay
To further characterize the dynamics of RSV infection, we used RLT-PCR in parallel with plaque assays, to measure RSV loads in both BAL samples and lung homogenate supernatants. These experiments were initially conducted in BALB/c mice. RSV loads measured by RLT-PCR and plaque assay in both BAL and lung supernatant samples peaked on days 3 to 5 after inoculation. Similar to plaque assay, RSV load by RLT-PCR also demonstrated a significant increase in viral copies between day 1 and days 3–5, likely demonstrating active replication (Figure 7). In contrast to RSV loads measured by plaque assay, which became undetectable by day 7 after inoculation (Figure 7, dashed-line plots), RSV loads measured by RLT-PCR remained positive for 42 days in BAL samples and throughout 77 days in lung supernatants (Figure 7, solid line plots).
Additional experiments in both mouse strains demonstrated persistence of RSV RNA in lung supernatants for 77 days after inoculation (Figure 8) . Similar to the previous findings using plaque assays, RSV loads were greater in BALB/c than in C57BL/6 mice. Control mice of both strains had undetectable RSV load by RLT-PCR.
Figure 8 Comparison of RSV loads measured by PCR in lung supernatants of BALB/c and C57BL/6 mice. Groups of 2–12 BALB/c (△) and C57BL/6 (●) mice per group per time point were inoculated intranasally with sterile 10% EMEM (control), and were compared with RSV A2 infected Balb/c (□) and C57Bl/6 (◆) mice. Viral load was determined by PCR to detect RSV N gene. Data are presented as mean ± SEM Log10 PFU/ml of BAL. p < .05, by t-test when data normally distributed, or by Mann-Whitney Rank sum test when data were not normally distributed.
5. Correlations among disease severity markers, inflammatory indices, and viral load dynamics
Correlations were determined during both the acute phase of the disease, day 5, and during the progression to the chronic phase on days 14, 21 and 77 after inoculation.
During the acute phase in both mouse strains, airway obstruction (AO) peaked on day 5 and strongly correlated with histopathologic scores (HPS), BAL concentrations of RANTES, IFN-γ, MIP-1α and MIG, and RSV loads measured by both plaque assay in BAL samples and RLT-PCR in lung supernatants (Additional file 1). In BALB/c mice, AO also correlated with BAL TNF-α concentrations, RSV loads measured by plaque assay in BAL samples significantly correlated with those measured in lung supernatants by RLT-PCR in both C57BL/6 and BALB/c mice.
In addition, there were significant correlations between airway hyperresponsiveness (AHR) and HPS on day 14 in both BALB/c (r = 0.99, P = 0.0005, n = 4), and C57BL/6 mice (r = 0.85, P = 0.01, n = 7). AHR and HPS also correlated significantly on day 77 in C57BL/6 mice (r = 0.95, P = 0.012, n = 5).
In BALB/c mice alone, RSV loads measured by RLT-PCR correlated with AHR on day 14 (r = 0.69, P = 0.01, n = 12). RLT-PCR also correlated with AO on day 21 after RSV inoculation (r = 0. 71, P = 0.01, n = 12).
Discussion
Intranasal inoculation of RSV induced acute and chronic effects on the respiratory dynamics and inflammatory response in C57BL/6 mice, similar to those previously reported in BALB/c mice [7]. The acute manifestations of RSV disease in C57BL/6 mice appeared similar but there were several differences compared with BALB/c mice. Acute airway obstruction was more severe and more prolonged in C57BL/6 than in BALB/c mice. BALB/c mice, however, demonstrated more severe histopathologic changes, greater viral loads, and BAL concentrations of most of the cytokines measured.
It is known that RSV infection causes variable degree of disease severity in different mouse strains [8] but there was no information on whether the different genetic backgrounds influence the long term outcome of RSV infection including chronic inflammation and persistent AO and AHR. The varied responses of these two mouse strains to different pathogens have been attributed in part to differences in their T helper (Th) lymphocyte response. Studies in mice infected with Leishmania showed that a Th2 skewed response caused more severe disease in BALB/c mice than in C57BL/6 mice that developed a Th1-dominant response [9]. In some localized bacterial infections, however, the severity of disease was milder in BALB/c mice and was described as Th1 skewed [10]. Indistinct Th responses were found in both mouse strains when infected with certain viruses [11]. Hence, the differences in the immune response might not only depend on the genetic background of the host, but also on the site and the specific pathogen involved in the infectious process. Several groups have demonstrated AHR in C57BL/6 mice during the acute stages of RSV infection. [22-24]. However, so far, our study is the first to shed light on the dynamics of inflammatory infiltrates, airway dysfunction (AO and AHR), and inflammatory cytokines during the long-term phase of RSV-induced disease, up to 77 days after primary infection in both C57BL/6 and BALB/c mice.
Researchers found that during RSV re-challenge following sensitization with different RSV proteins, mice developed a variable Th cytokine response [12-15] which was not exclusively dependent on the strain haplotype [16]. Our experiments, conducted in mice with primary RSV infection, demonstrated that IFN-γ, a Th1 cytokine, was elevated after RSV infection in both strains. IFN-γ was increased in the respiratory tract of children with acute RSV infection [17,18] and correlated with disease severity [19]. In mice, IFN-γ also correlates with the development of RSV disease [20]. Our studies demonstrate that RSV infection induces IFN-γ in both strains of mice and the magnitude of the response is not consistent with the traditional notion that BALB/c are Th-2 weighted and C57BL/6 are Th1-weighted. BALB/c mice showed greater IFN-γ response than C57BL/6 mice as previously shown by others [21]. In our experiments IFN-γ demonstrated the greatest correlation with disease severity (Additional file 1).
Like IFN-γ, the dynamics of TNF-α, RANTES, MIP-1α were comparable in both mouse strains. TNF-α has been linked to RSV disease severity [25]. In our model, BAL concentrations of TNF-α peaked shortly after RSV infection and correlated with markers of disease severity in BALB/c mice (Additional file 1), and were significantly greater in BALB/c than in C57BL/6 mice (Figure 5).
The CC chemokines, RANTES and MIP-1α, and the CXC chemokine MIG were also elevated in mice of both strains shortly after RSV inoculation. The biphasic production of RANTES, MIP-1α and MIG after RSV infection, with the second peak coinciding with the peak of viral replication and histological inflammation, has been previously described in BALB/c mice [7,26] but not in C57BL/6 mice. Although MIP-1α has been detected in C57BL/6 mice [27], no previous studies demonstrated the production of RANTES in the respiratory tract of these mice after RSV infection. These two chemokines correlate with acute disease severity both in humans and in mice. [19,28-30]. Increased production of MIG after RSV infection has been documented in vitro [31] and by us in BALB/c mice[7]. Others have linked MIG to persistent airway inflammation due to continuous chemotaxis of mononuclear cells [32]. In the present study we demonstrate significant correlation between BAL MIG concentrations and AO, and viral load dynamics in both strains of mice. Taken together, these results indicate that RSV induces severe acute pulmonary disease in both strains of mice despite different genetic backgrounds. Although there are quantitative differences during the acute phase of the disease, there were no significant distinctions in the pattern of cytokine responses, lung inflammation or clinical manifestations of disease, and both strains developed long-term airway disease defined by chronic inflammatory infiltrates and abnormal AHR. These findings provide further experimental support to the link between RSV and chronic airway disease in humans. We believe that this is the first description of RSV-induced long-term pulmonary disease in C57BL/6 mice.
The exact mechanism by which different inflammatory mediators participate in the pathogenesis of the acute RSV-induced disease process is not completely understood, and their potential role in determining the chronic consequences of RSV infection, such as AO, AHR and chronic inflammation is even less characterized. Studies in murine models and in humans have not demonstrated yet a direct relation between the acute inflammatory response to RSV and its chronic consequences in the respiratory tract pathology. Our studies demonstrated prolonged AO and persistent AHR in both mouse strains, long after any of the cytokines, potentially responsible for persistent inflammation that we measured, became undetectable in BAL samples. It is possible that the use of more sensitive methods, perhaps in specific compartments, will contribute to clarify the role of these molecules in the chronic respiratory function abnormalities induced by RSV.
Until recently, it was suggested that initial RSV infection, although of short duration, was sufficient to elicit an exaggerated inflammatory response which would evolve into chronic inflammation and long-term airway abnormalities. An alternative, but not contradictory explanation for the development of chronic airway disease could be related to long-term antigenic stimulation. In this hypothesis, the presence of low level viral infection could represent a persistent stimulus responsible for the chronic inflammatory changes and the lasting pulmonary function abnormalities. The application of the RLT-PCR assay to this model demonstrated the persistence of RSV RNA in the respiratory tract for 11 weeks after inoculation in both strains of mice and was also associated with long-term pulmonary disease.
RSV can persist and replicate in vitro [33-35]. The persistence of RSV genomic RNA sequences and viral antigens has been described in guinea pigs for several weeks after acute primary infection [36-38]. Bovine RSV infection in calves resulted in persistent detection of RSV N gene for several weeks [39]. Miller et al. reported detection of G protein RNA by RLT-PCR in lungs of BALB/c mice up to day 12 after inoculation. [26].
It appears unlikely for a RNA virus to persist without at least low-level replication in the host. This is supported by the studies by Schwarze et al., who found RSV genomic and messenger RNA in mouse lung homogenates for 100 days after inoculation, and were able to recover low titer virus in culture after treatment with anti-CD4 and anti-CD8 antibodies [40].
Clinical studies have demonstrated that adults with stable COPD without acute exacerbation had RSV RNA detected by PCR in the upper respiratory tract [41]. The persistence of RSV RNA in children has not been extensively studied. Post-mortem specimens of children who died of sudden infant death syndrome showed positive results by PCR targeting the RSV N gene in 27% of cases and in 18% of the controls. Since some children died during summer the authors suggested the possibility of viral persistence [42]. The persistence of RSV in the respiratory tract of children and its potential contribution to the development and maintenance of AHR is a critical question that needs to be addressed.
In summary, we provide evidence that RSV-induced acute and chronic airway disease is reproducible in two genetically distinct mouse strains. The long-term airway disease coincided with persistent detection of RSV genome in the respiratory tract. It is yet unclear whether the persistence of RSV RNA plays a role in the development of chronic airway disease. Further studies are needed to characterize the potential mechanisms that would allow viral persistence and its possible contribution to the pathogenesis of RSV-induced long-term pulmonary disease.
Materials and methods
Mice
Female, 7–8-week old pathogen-free C57BL/6 and BALB/c mice (Charles River Laboratories, Wilmington, MA) were maintained in filter top cages, and routinely monitored for other pathogens [7]. After inoculation, all mice were kept under the same conditions and were provided identical care. Sentinel mice housed in the mouse storage room are routinely used for health surveillance. Sentinel mice had no detectable antibodies against mouse hepatitis virus, Sendai virus, pneumonia virus of mice, reo-3 virus, mouse encephalitis virus (GD-7), mouse rotavirus (EDIM), minute virus of mice, and Mycoplasma pulmonis; screening for pinworm and mites was also negative. This study was approved by the Institutional Animal Care and Use Committee at the University of Texas Southwestern Medical Center at Dallas.
Virus
RSV A-2 strain was maintained in our laboratory as described [7]. Results of plaque assays were reported in Log10 PFU/mL, with 1.7 Log10 PFU/mL being the lowest limit of detection.
Inoculation
Mice were anesthetized using inhaled methoxyfluorane and intranasally inoculated with 107 PFU of RSV in 100 μl of 10% Eagle's minimal essential medium (EMEM) [6]. Uninfected control animals were sham-inoculated with 100 μl of sterile 10% EMEM. Animals were allowed 30 seconds to aspirate the inoculum while held upright until fully recovered from the anesthesia.
Plethysmography
Unrestrained, whole-body plethysmography (Buxco Electronics, Inc. Sharon, CT) was used to measure the Enhanced Pause (Penh) to evaluate airway obstruction (AO) and airway hyperresponsiveness (AHR), as previously described [7,43]. Briefly, mice were allowed to acclimate to the chamber, and then plethysmograph readings were recorded to establish baseline Penh values to determine AO. Next the mice were exposed to aerosolized methacholine (Sigma; 50 mg/ml) for 4 minutes; after exposure, plethysmograph readings were recorded again to determine AHR. Groups of infected and control mice were always evaluated in parallel at all time points during the entire study. Preliminary studies showed that normal C57BL/6 mice have greater baseline Penh values than normal BALB/c mice; therefore, groups of infected and control mice of each strain were evaluated in parallel during the entire study. Plethysmography was performed in the groups of mice to be sacrificed for sample collection at each time point, and in an additional group of mice that were followed throughout each experiment. A total of 4 independent experiments including 10–30 mice per group per time point were conducted.
Sample collection
Mice were anesthetized with an intraperitoneal injection of 75 mg/kg of ketamine and 5 mg/kg of acepromazine before euthanasia by exsanguination. On average, 4–12 mice were sacrificed per group per time point. Bronchoalveolar lavage (BAL) specimens were obtained as described previously [7,43] to measure cytokine concentrations and RSV loads. Histologic evaluation was performed in whole-lung specimens fixed with a 10% buffered formalin solution. Whole lung samples were also collected from another group of mice for determination of viral loads by plaque assay and by real-time polymerase chain reaction (RLT-PCR). Specimens were obtained on days 0 (within 2 hours of inoculation), 1, 3, 5, 8, 9, 12, 14, 21, 28, 42, and 77 post-inoculation.
Histopathology
Formalin fixed lungs were paraffin embedded, sectioned and stained with hematoxylin and eosin. Histopathologic score (HPS) was determined by a pathologist, unaware of the infection status of the animals. Each section was graded on the basis of a cumulative score from 5 categories: (1) peribronchiolar and bronchial infiltrates, (2) bronchiolar and bronchial luminal exudates, (3) perivascular infiltrates, (4) the number of monocytes, and (5) parenchymal pneumonia. This HPS system assigns values from 0 to 21. This scoring system has been previously validated in RSV infection [7,43].
BAL cytokines
Concentrations of TNF-α, IFN-γ, IL-4, IL-10, MIG, RANTES, and MIP-1α were measured in BAL specimens by ELISA (R&D Systems, Minneapolis, Minn.) for up to 14 days after inoculation. The lower limit of detection for each of these assays were: 120 pg/mL for TNF-α, 50 pg/mL for IFN-γ, 40 pg/mL for IL4, 16 pg/mL for IL-10, 16 pg/ mL for MIG, 200 pg/mL for RANTES and 60 pg/mL for MIP-1α. For statistical analysis, samples with optical density readings below the limit of the standard curve of the assay were assigned a value half that of the detection level.
RSV loads by plaque assay and Real-time Polymerase Chain Reaction (RLT-PCR)
Plaque assay tissue culture was used to measure viral loads in BAL specimens and lung homogenate supernatants as previously described [7]. Lung supernatant samples were prepared by placing whole lung specimens in 1 mL of 10% EMEM, on ice. Lungs were homogenized using sterile probes and a rotor tissue homogenizer (Omni International Inc., Marietta, GA). Lung supernatants were separated by centrifugation at 2000 g for 10 minutes at 4°C and 100 μL were used immediately for determination of viral loads by plaque assay; the remaining supernatant was frozen at -80°C for viral load measurements by RLT-PCR. In a subset of experiments, lung samples used for RLT-PCR were placed in 1.5 mL of RNA stabilization reagent (RNA later, Qiagen Inc., Valencia, CA) immediately after sample collection. Lung supernatants were then obtained by homogenization and centrifugation. The RLT-PCR viral load values measured using this method were similar to those obtained when no RNA later was used; therefore, all results of lung homogenate supernatants are presented together. One sample per mouse was evaluated as a single specimen. Quantitative RLT-PCR was used targeting the conserved region of the RSV N-gene. Forward (5'-AGA TCA ACT TCT GTC ATC CAG CAA) and reverse (5'-TTC TGC ACA TCA TAA TTA GGA GTA TCA AT) primers amplified an 85-bp region containing the 25-mer FAM-labeled probe (5'-CAC CAT CCA ACG GAG CAC AGG AGA T), as previously described [44]. RSV RNA was extracted from 1 mL of lung supernatant samples using ion-exchange mini-columns (Qiagen RNeasy Mini Kit, Valencia, CA, USA) and cDNA was prepared by reverse transcription using 2.5 uM random hexamers for 10 min at 22°C, 30 min at 42°C and 5 min at 95°C. Real-time PCR was performed using a Perkin-Elmer /Applied Biosystems 7700 sequence detector (Foster City, CA) using 10 μl cDNA/ in a total volume of 50 μl master mix with the following run conditions: 1 cycle for 2 min at 50°C and 10 min at 95°C each, followed by 50 cycles for 15 seconds at 95°C and 60 seconds at 60°C. RSV A known concentrations were used to derive a standard curve. Standards and negative controls were run together with each PCR assay. The lower limit of detection of the assay was 10 viral copies/mL.
Statistical methods
T-test or Mann-Whitney Rank Sum test were used according to data distribution for comparisons between groups and the Spearman Rank Order test was used for correlations, as all the data taken together were not normally distributed, using the SigmaStat® (SPSS, Inc. Chicago, Illinois) software package.
Abreviations used
PCR polymerase chain reaction
TNF tumor necrosis factor
IL interleukin
IFN interferon
MIP macrophage inflammatory protein
RANTES regulated on activation, normal T-cell expressed and secreted
MIG monokine induced by interferon-gamma
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
SCB study design, RSV infection, data analyses; AM BAL, formalin fixation, plethysmography; AMG histopathology and special staining; KDO real-time PCR; AMR and MFA plethysmography and BAL; OR data interpretation; HSJ project design, experimental analyses and interpretation.
Supplementary Material
Additional File 1
Significant Correlations on Day 5 after RSV Inoculation in C57BL/6 and BALB/c mice. r = correlation coefficient; p = p-value, n = number of samples analyzed. Spearman's rank order analysis was used. Statistically significant correlations (p < 0.05) listed in bold font. N/A, not applicable, samples from different animals; * TNF-α concentrations undetectable on day 5.
Click here for file
Acknowledgements
S.C.B. is supported in part by a Pediatric Infectious Disease Society Fellowship Award sponsored by GlaxoSmithKline Pharmaceuticals.
A.M. was supported in part by the Pediatric Fellowship Award in Viral Respiratory Infectious Diseases supported by MedImmune Inc., at the Pediatric Academic Societies Annual Meeting.
H.S.J. is supported in part by grants from the Children's Clinical Research Advisory Committee and the Children's Medical Center of Dallas Foundation and American Lung Association.
Approved by the Institutional Animal Care and Use Committee at the University of Texas Southwestern Medical Center at Dallas.
==== Refs
Shay DK Holman RC Newman RD Liu LL Stout JW Anderson LJ Bronchiolitis-associated hospitalizations among US children, 1980-1996 Jama 1999 282 1440 1446 10535434 10.1001/jama.282.15.1440
Martinez FD Wright AL Taussig LM Holberg CJ Halonen M Morgan WJ Asthma and wheezing in the first six years of life. The Group Health Medical Associates N Engl J Med 1995 332 133 138 7800004 10.1056/NEJM199501193320301
Stein RT Sherrill D Morgan WJ Holberg CJ Halonen M Taussig LM Wright AL Martinez FD Respiratory syncytial virus in early life and risk of wheeze and allergy by age 13 years Lancet 1999 354 541 545 10470697 10.1016/S0140-6736(98)10321-5
Sigurs N Bjarnason R Sigurbergsson F Kjellman B Respiratory syncytial virus bronchiolitis in infancy is an important risk factor for asthma and allergy at age 7 Am J Respir Crit Care Med 2000 161 1501 1507 10806145
Sigurs N Gustafsson PM Bjarnason R Lundberg F Schmidt S Sigurbergsson F Kjellman B Severe respiratory syncytial virus bronchiolitis in infancy and asthma and allergy at age 13 Am J Respir Crit Care Med 2005 171 137 141 15516534 10.1164/rccm.200406-730OC
Karron RA Singleton RJ Bulkow L Parkinson A Kruse D DeSmet I Indorf C Petersen KM Leombruno D Hurlburt D Santosham M Harrison LH Severe respiratory syncytial virus disease in Alaska native children. RSV Alaska Study Group J Infect Dis 1999 180 41 49 10353859 10.1086/314841
Jafri HS Chavez-Bueno S Mejias A Gomez AM Rios AM Nassi SS Yusuf M Kapur P Hardy RD Hatfield J Rogers BB Krisher K Ramilo O Respiratory syncytial virus induces pneumonia, cytokine response, airway obstruction, and chronic inflammatory infiltrates associated with long-term airway hyperresponsiveness in mice J Infect Dis 2004 189 1856 1865 15122522 10.1086/386372
Byrd LG Prince GA Animal models of respiratory syncytial virus infection Clin Infect Dis 1997 25 1363 1368 9431379
Sacks D Noben-Trauth N The immunology of susceptibility and resistance to Leishmania major in mice Nat Rev Immunol 2002 2 845 858 12415308 10.1038/nri933
Hazlett LD Pathogenic mechanisms of P. aeruginosa keratitis: a review of the role of T cells, Langerhans cells, PMN, and cytokines DNA Cell Biol 2002 21 383 390 12167240 10.1089/10445490260099665
Ramakrishna C Ravi V Desai A Subbakrishna DK Shankar SK Chandramuki A T helper responses to Japanese encephalitis virus infection are dependent on the route of inoculation and the strain of mouse used J Gen Virol 2003 84 1559 1567 12771426 10.1099/vir.0.18676-0
Bangham CR Cannon MJ Karzon DT Askonas BA Cytotoxic T-cell response to respiratory syncytial virus in mice J Virol 1985 56 55 59 2411951
Alwan WH Record FM Openshaw PJ Phenotypic and functional characterization of T cell lines specific for individual respiratory syncytial virus proteins J Immunol 1993 150 5211 5218 8515055
Hancock GE Tebbey PW Scheuer CA Pryharski KS Heers KM LaPierre NA Immune responses to the nonglycosylated ectodomain of respiratory syncytial virus attachment glycoprotein mediate pulmonary eosinophilia in inbred strains of mice with different MHC haplotypes J Med Virol 2003 70 301 308 12696122 10.1002/jmv.10395
Hussell T Georgiou A Sparer TE Matthews S Pala P Openshaw PJ Host genetic determinants of vaccine-induced eosinophilia during respiratory syncytial virus infection J Immunol 1998 161 6215 6222 9834108
Srikiatkhachorn A Chang W Braciale TJ Induction of Th-1 and Th-2 responses by respiratory syncytial virus attachment glycoprotein is epitope and major histocompatibility complex independent J Virol 1999 73 6590 6597 10400756
Garofalo RP Hintz KH Hill V Ogra PL Welliver RCS Production of interferon gamma in respiratory syncytial virus infection of humans is not associated with interleukins 12 and 18 J Med Virol 2004 73 289 294 15122806 10.1002/jmv.20089
van Benten IJ van Drunen CM Koopman LP KleinJan A van Middelkoop BC de Waal L Osterhaus AD Neijens HJ Fokkens WJ RSV-induced bronchiolitis but not upper respiratory tract infection is accompanied by an increased nasal IL-18 response J Med Virol 2003 71 290 297 12938205 10.1002/jmv.10482
Garofalo RP Patti J Hintz KA Hill V Ogra PL Welliver RC Macrophage inflammatory protein-1alpha (not T helper type 2 cytokines) is associated with severe forms of respiratory syncytial virus bronchiolitis J Infect Dis 2001 184 393 399 11471095 10.1086/322788
van Schaik SM Obot N Enhorning G Hintz K Gross K Hancock GE Stack AM Welliver RC Role of interferon gamma in the pathogenesis of primary respiratory syncytial virus infection in BALB/c mice J Med Virol 2000 62 257 266 11002257 10.1002/1096-9071(200010)62:2<257::AID-JMV19>3.0.CO;2-M
Johnson TR Hong S Van Kaer L Koezuka Y Graham BS NK T cells contribute to expansion of CD8(+) T cells and amplification of antiviral immune responses to respiratory syncytial virus J Virol 2002 76 4294 4303 11932395 10.1128/JVI.76.9.4294-4303.2002
Tekkanat KK Maassab H Berlin AA Lincoln PM Evanoff HL Kaplan MH Lukacs NW Role of interleukin-12 and stat-4 in the regulation of airway inflammation and hyperreactivity in respiratory syncytial virus infection Am J Pathol 2001 159 631 638 11485921
Makela MJ Kanehiro A Dakhama A Borish L Joetham A Tripp R Anderson L Gelfand EW The failure of interleukin-10-deficient mice to develop airway hyperresponsiveness is overcome by respiratory syncytial virus infection in allergen-sensitized/challenged mice Am J Respir Crit Care Med 2002 165 824 831 11897651
Schwarze J Cieslewicz G Hamelmann E Joetham A Shultz LD Lamers MC Gelfand EW IL-5 and eosinophils are essential for the development of airway hyperresponsiveness following acute respiratory syncytial virus infection J Immunol 1999 162 2997 3004 10072551
Rutigliano JA Graham BS Prolonged production of TNF-alpha exacerbates illness during respiratory syncytial virus infection J Immunol 2004 173 3408 3417 15322205
Miller AL Bowlin TL Lukacs NW Respiratory syncytial virus-induced chemokine production: linking viral replication to chemokine production in vitro and in vivo J Infect Dis 2004 189 1419 1430 15073679 10.1086/382958
Domachowske JB Bonville CA Gao JL Murphy PM Easton AJ Rosenberg HF MIP-1alpha is produced but it does not control pulmonary inflammation in response to respiratory syncytial virus infection in mice Cell Immunol 2000 206 1 6 11161432 10.1006/cimm.2000.1730
Noah TL Becker S Chemokines in nasal secretions of normal adults experimentally infected with respiratory syncytial virus Clin Immunol 2000 97 43 49 10998316 10.1006/clim.2000.4914
Noah TL Ivins SS Murphy P Kazachkova I Moats-Staats B Henderson FW Chemokines and inflammation in the nasal passages of infants with respiratory syncytial virus bronchiolitis Clin Immunol 2002 104 86 95 12139952 10.1006/clim.2002.5248
Sheeran P Jafri H Carubelli C Saavedra J Johnson C Krisher K Sanchez PJ Ramilo O Elevated cytokine concentrations in the nasopharyngeal and tracheal secretions of children with respiratory syncytial virus disease Pediatr Infect Dis J 1999 18 115 122 10048682 10.1097/00006454-199902000-00007
Bitko V Garmon NE Cao T Estrada B Oakes JE Lausch RN Barik S Activation of cytokines and NF-kappa B in corneal epithelial cells infected by respiratory syncytial virus: potential relevance in ocular inflammation and respiratory infection BMC Microbiol 2004 4 28 15256003 10.1186/1471-2180-4-28
Kelsen SG Aksoy MO Yang Y Shahabuddin S Litvin J Safadi F Rogers TJ The Chemokine Receptor, CXCR3, and its Splice Variants are Expressed in Human Airway Epithelial Cells Am J Physiol Lung Cell Mol Physiol 2004
Pringle C R Shirodaria PV Cash P Chiswell DJ Malloy P Initiation and maintenance of persistent infection by respiratory syncytial virus J Virol 1978 28 199 211 702647
Panuska JR Cirino NM Midulla F Despot JE McFadden ERJ Huang YT Productive infection of isolated human alveolar macrophages by respiratory syncytial virus J Clin Invest 1990 86 113 119 2365811
Guerrero-Plata A Ortega E Gomez B Persistence of respiratory syncytial virus in macrophages alters phagocytosis and pro-inflammatory cytokine production Viral Immunol 2001 14 19 30 11270594 10.1089/08828240151061347
Hegele RG Robinson PJ Gonzalez S Hogg JC Production of acute bronchiolitis in guinea-pigs by human respiratory syncytial virus Eur Respir J 1993 6 1324 1331 8287949
Hegele RG Hayashi S Bramley AM Hogg JC Persistence of respiratory syncytial virus genome and protein after acute bronchiolitis in guinea pigs Chest 1994 105 1848 1854 8205887
Streckert HJ Philippou S Riedel F Detection of respiratory syncytial virus (RSV) antigen in the lungs of guinea pigs 6 weeks after experimental infection and despite of the production of neutralizing antibodies Arch Virol 1996 141 401 410 8645083 10.1007/BF01718305
Valarcher JF Bourhy H Lavenu A Bourges-Abella N Roth M Andreoletti O Ave P Schelcher F Persistent infection of B lymphocytes by bovine respiratory syncytial virus Virology 2001 291 55 67 11878876 10.1006/viro.2001.1083
Schwarze J O'Donnell DR Rohwedder A Openshaw PJ Latency and persistence of respiratory syncytial virus despite T cell immunity Am J Respir Crit Care Med 2004 169 801 805 14742302 10.1164/rccm.200308-1203OC
Rohde G Wiethege A Borg I Kauth M Bauer TT Gillissen A Bufe A Schultze-Werninghaus G Respiratory viruses in exacerbations of chronic obstructive pulmonary disease requiring hospitalisation: a case-control study Thorax 2003 58 37 42 12511718 10.1136/thorax.58.1.37
Cubie HA Duncan LA Marshall LA Smith NM Detection of respiratory syncytial virus nucleic acid in archival postmortem tissue from infants Pediatr Pathol Lab Med 1997 17 927 938 9353832 10.1080/107710497174372
Mejias A Chavez-Bueno S Rios AM Saavedra-Lozano J Fonseca Aten M Hatfield J Kapur P Gomez AM Jafri HS Ramilo O Anti-respiratory syncytial virus (RSV) neutralizing antibody decreases lung inflammation, airway obstruction, and airway hyperresponsiveness in a murine RSV model Antimicrob Agents Chemother 2004 48 1811 1822 15105140 10.1128/AAC.48.5.1811-1822.2004
van Woensel JB Lutter R Biezeveld MH Dekker T Nijhuis M van Aalderen WM Kuijpers TW Effect of dexamethasone on tracheal viral load and interleukin-8 tracheal concentration in children with respiratory syncytial virus infection Pediatr Infect Dis J 2003 22 721 726 12913774
|
15916706
|
PMC1183251
|
CC BY
|
2021-01-04 16:39:00
|
no
|
Virol J. 2005 May 25; 2:46
|
utf-8
|
Virol J
| 2,005 |
10.1186/1743-422X-2-46
|
oa_comm
|
==== Front
Virol JVirology Journal1743-422XBioMed Central London 1743-422X-2-461591670610.1186/1743-422X-2-46ResearchRespiratory syncytial virus-induced acute and chronic airway disease is independent of genetic background: An experimental murine model Chávez-Bueno Susana [email protected]ías Asunción [email protected]ómez Ana M [email protected] Kurt D [email protected]íos Ana M [email protected] Mónica [email protected] Octavio [email protected] Hasan S [email protected] Division of Pediatric Infectious Diseases, Department of Pediatrics, The University of Texas Southwestern Medical Center at Dallas and Children's Medical Center Dallas, Dallas, Texas, USA2 Department of Pathology, The University of Texas Southwestern Medical Center at Dallas and Children's Medical Center Dallas, Dallas, Texas, USA2005 25 5 2005 2 46 46 26 4 2005 25 5 2005 Copyright © 2005 Chávez-Bueno et al; licensee BioMed Central Ltd.2005Chávez-Bueno et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Respiratory syncytial virus (RSV) is the leading respiratory viral pathogen in young children worldwide. RSV disease is associated with acute airway obstruction (AO), long-term airway hyperresponsiveness (AHR), and chronic lung inflammation. Using two different mouse strains, this study was designed to determine whether RSV disease patterns are host-dependent. C57BL/6 and BALB/c mice were inoculated with RSV and followed for 77 days. RSV loads were measured by plaque assay and polymerase chain reaction (PCR) in bronchoalveolar lavage (BAL) and whole lung samples; cytokines were measured in BAL samples. Lung inflammation was evaluated with a histopathologic score (HPS), and AO and AHR were determined by plethysmography.
Results
Viral load dynamics, histopathologic score (HPS), cytokine concentrations, AO and long-term AHR were similar in both strains of RSV-infected mice, although RSV-infected C57BL/6 mice developed significantly greater AO compared with RSV-infected BALB/c mice on day 5. PCR detected RSV RNA in BAL samples of RSV infected mice until day 42, and in whole lung samples through day 77. BAL concentrations of cytokines TNF-α, IFN-γ, and chemokines MIG, RANTES and MIP-1α were significantly elevated in both strains of RSV-infected mice compared with their respective controls. Viral load measured by PCR significantly correlated with disease severity on days 14 and 21.
Conclusion
RSV-induced acute and chronic airway disease is independent of genetic background.
Viral pneumoniamouse modelairway hyperresponsivenessPCRcytokines
==== Body
Background
Human respiratory syncytial virus (RSV) is classified in the genus Pneumovirus, subfamily Pneumovirinae, family Paramixoviridae; and is a major cause of lower respiratory tract infection (LRTI) in young children and the elderly [1]. RSV LRTI is associated with increased risk of long-term recurrent wheezing [2-5], however, the pathogenesis of this relationship is not well understood. RSV LRTI elicits a host response including the release of inflammatory mediators and recruitment of different cell populations. The genetic variability of the host response might partially explain the different susceptibilities of individual patients to the acute and long-term effects of RSV infection, as suggested by the higher rates of RSV hospitalization among Native American and Alaskan Native children compared with other groups [6].
Animal models facilitate the study of RSV-induced acute and long-term disease in a more controlled manner. Our laboratory has previously established a mouse model of RSV-induced acute and long-term airway disease [7]. The present studies were designed to characterize the influence of mouse genetic background and the dynamics of viral replication on the chronic manifestations of RSV infection. The BALB/c mouse strain is one of the most commonly used for RSV experimental models, however, C57BL/6 mice frequently provide background for transgenic strains of mice. Therefore characterizing and establishing a comprehensive model of acute and long-term RSV disease in C57BL/6 is essential to further understanding the pathogenesis of RSV disease.
Results
1. RSV alone induces airway obstruction (AO) and airway hyperresponsiveness (AHR) in both C57BL/6 and BALB/c mice
RSV infection alone, without allergic pre-sensitization induced AO in both strains of mice as demonstrated by significantly increased enhanced pause (Penh) values compared with uninfected controls. Baseline Penh values increased transiently on day 1 after RSV inoculation in both strains, decreased by day 2, but continued to be significantly greater than in controls. Airway obstruction increased again and peaked on day 5, when C57BL/6 RSV-infected mice showed significantly higher Penh values than RSV-infected BALB/c mice (p < 0.001) (Figure 1). AO decreased thereafter during the first two weeks after RSV inoculation but remained significantly greater than the respective controls in both strains, for 21 days in BALB/c and 28 days in C57BL/6 mice (Figure 1 inset). RSV infection also induced AHR in both strains as evidenced by a greater difference between pre- and post-methacholine Penh values (delta Penh) compared with controls. Significantly increased AHR was persistently present for 42 days post-inoculation in BALB/c mice, while C57BL/6 mice showed significantly increased AHR for up to 28 days post-inoculation (Figure 2).
Figure 1 Effect of RSV on airway obstruction (AO) in two mouse strains. BALB/c (△) and C57BL/6 (●) mice were inoculated with sterile 10% EMEM (control) and were compared with RSV A2 infected BALB/c (□) and C57BL/6 (◆) mice to evaluate differences in airway obstruction (AO), by measuring Penh via whole-body plethysmography. Penh values are presented as means ± SEM. Comparisons were made by t-test when data normally distributed, or by Mann-Whitney Rank sum test when data were not normally distributed.
Figure 2 Airway Hyperresponsiveness (AHR) in BALB/c and C57BL/6 Mice. Data presented as Delta Penh values which is the difference between pre-and post methacholine Penh for each group of mice, in sham inoculated () and RSV inoculated () BALB/c mice, and sham inoculated () and RSV inoculated () C57BL/6 mice from days 14 to 77. Values represent the mean SEM from 10–30 mice per group. Data shown are the result of four separate experiments. p < 0.05, comparison by t-test when data normally distributed, or by Mann-Whitney Rank sum test when data were not normally distributed.
2. C57BL/6 and BALB/c mice demonstrate acute and persistent inflammatory changes after RSV infection
RSV-inoculated C57BL/6 and BALB/c mice, compared with their controls, showed greater histopathologic scores (HPS) which peaked on day 5 after RSV inoculation (Figures 3, 4A–D). Although the acute inflammatory changes observed in both strains gradually declined, RSV-infected mice had significantly greater HPS than the sham-inoculated controls for up to 77 days post-inoculation (Figures 3, 4E and 4F).
Figure 3 Comparison of acute and long-term histopathologic scores after RSV inoculation. BALB/c (△) and C57BL/6 (●) mice were inoculated with sterile 10% EMEM (control) and were compared with RSV A2 infected BALB/c (□) and C57BL/6 (◆) mice. Serial formalin fixed lung samples were obtained between day 0 (+2 hours) and day 77 after inoculation. HPS scores are represented as means ± SEM. p < .05, by t-test when data normally distributed, or by Mann-Whitney Rank sum test when data were not normally distributed.
Figure 4 RSV induced histopathology. Lung specimens were harvested on days 5 and 77 from groups of C57BL/6 and BALB/c mice inoculated with sterile medium (control) or RSV. Sections from control C57BL/6 and BALB/c mice above (4A and 4B, respectively), show rare, scattered, small lymphocytic infiltrates on day 5 after inoculation with medium, similar to sections of control mice harvested 77 days after (not shown). Acute and chronic inflammatory infiltrates, surrounding airways and vessels are demonstrated in RSV-inoculated mice of both strains, on days 5 and 77 after inoculation (4C to F).
3. RSV infection induces similar cytokine production in the respiratory tract of C57BL/6 and BALB/c mice
BAL concentrations of TNF-α, IFN-γ, MIG, RANTES, and MIP-1α followed similar dynamics in both strains of mice during the acute phase of the infection (Figure 5A–E). Overall, there was a trend for greater BAL cytokine concentrations of IFN-γ, TNF-α, RANTES and MIP-1α in RSV-infected BALB/c mice compared with C57BL/6 mice. (Figure 5A–E). No significant differences were observed in BAL concentrations of IL-4 and IL-10 between controls and infected mice of both strains at any time point evaluated (data not shown).
Figure 5 Cytokine and chemokine concentrations in bronchoalveolar lavage (BAL) samples after RSV inoculation. BAL samples were obtained from BALB/c (△) and C57BL/6 (●) mice inoculated with sterile 10% EMEM (control), and RSV A2 infected BALB/c (□) and C57BL/6 (◆) mice, to measure concentrations of pro-inflammatory cytokines (A) IFN-γ and (B)TNF-α; and the chemokines (C) RANTES, (D) MIP-1α, and (E) MIG. Values presented in means ± SEM pg/ml. p < .05, by t-test when data were normally distributed, or by Mann-Whitney Rank sum test when data were not normally distributed.
4. RSV load dynamics
4a. RSV loads measured by plaque assay in BAL samples follow similar dynamics in both C57BL/6 and BALB/c mice
On day 1 after RSV inoculation, RSV loads in BAL samples from both C57BL/6 and BALB/c mice were significantly greater than in controls by plaque assay (Figure 6). Compared with day 1, plaque assay RSV loads peaked on days 3–5 after inoculation in both strains representing active viral replication (p = 0.002 for day 1 vs days 3–5 in BALB/c mice; ANOVA), and were significantly greater in BALB/c than in C57BL/6 mice (Figure 7). BAL RSV loads declined below the limit of detection by day 7 and remained undetectable through day 77 post-inoculation.
Figure 6 RSV loads in BAL samples measured by the plaque assay method. Groups of 4–16 BALB/c (△) and C57BL/6 (●) mice per group per time point were inoculated intranasally with sterile 10% EMEM (control), and were compared with RSV A2 infected BALB/c (□) and C57BL/6 (◆) mice. Viral load was determined by HEp-2 plaque assay in BAL samples. Data are presented as mean ± SEM Log10 PFU/ml of BAL. p < .05, by t-test when data normally distributed, or by Mann-Whitney Rank sum test when data were not normally distributed.
Figure 7 RSV loads in RSV infected BALB/c mice BAL and lung supernatant samples measured by PCR vs. plaque assay. RSV loads measured by PCR in BAL samples (▽) remain positive up to 42 days after inoculation, while viral loads measured by plaque assay (□) become negative on day 7 post-inoculation (upper panel). Viral load measured in lung supernatants by the plaque assay also become undetectable by day 7 after inoculation, whereas RSV loads measured by PCR in lung supernatants remain detectable throughout all the time points evaluated (lower panel). All pair-wise multiple comparisons made by One-Way ANOVA. † p < 0.01 between D1 and D5 and *p < 0.05 comparing D1 with D3 and D5.
4b. Real Time PCR (RLT-PCR) demonstrates RSV RNA after the virus is no longer detectable by plaque assay
To further characterize the dynamics of RSV infection, we used RLT-PCR in parallel with plaque assays, to measure RSV loads in both BAL samples and lung homogenate supernatants. These experiments were initially conducted in BALB/c mice. RSV loads measured by RLT-PCR and plaque assay in both BAL and lung supernatant samples peaked on days 3 to 5 after inoculation. Similar to plaque assay, RSV load by RLT-PCR also demonstrated a significant increase in viral copies between day 1 and days 3–5, likely demonstrating active replication (Figure 7). In contrast to RSV loads measured by plaque assay, which became undetectable by day 7 after inoculation (Figure 7, dashed-line plots), RSV loads measured by RLT-PCR remained positive for 42 days in BAL samples and throughout 77 days in lung supernatants (Figure 7, solid line plots).
Additional experiments in both mouse strains demonstrated persistence of RSV RNA in lung supernatants for 77 days after inoculation (Figure 8) . Similar to the previous findings using plaque assays, RSV loads were greater in BALB/c than in C57BL/6 mice. Control mice of both strains had undetectable RSV load by RLT-PCR.
Figure 8 Comparison of RSV loads measured by PCR in lung supernatants of BALB/c and C57BL/6 mice. Groups of 2–12 BALB/c (△) and C57BL/6 (●) mice per group per time point were inoculated intranasally with sterile 10% EMEM (control), and were compared with RSV A2 infected Balb/c (□) and C57Bl/6 (◆) mice. Viral load was determined by PCR to detect RSV N gene. Data are presented as mean ± SEM Log10 PFU/ml of BAL. p < .05, by t-test when data normally distributed, or by Mann-Whitney Rank sum test when data were not normally distributed.
5. Correlations among disease severity markers, inflammatory indices, and viral load dynamics
Correlations were determined during both the acute phase of the disease, day 5, and during the progression to the chronic phase on days 14, 21 and 77 after inoculation.
During the acute phase in both mouse strains, airway obstruction (AO) peaked on day 5 and strongly correlated with histopathologic scores (HPS), BAL concentrations of RANTES, IFN-γ, MIP-1α and MIG, and RSV loads measured by both plaque assay in BAL samples and RLT-PCR in lung supernatants (Additional file 1). In BALB/c mice, AO also correlated with BAL TNF-α concentrations, RSV loads measured by plaque assay in BAL samples significantly correlated with those measured in lung supernatants by RLT-PCR in both C57BL/6 and BALB/c mice.
In addition, there were significant correlations between airway hyperresponsiveness (AHR) and HPS on day 14 in both BALB/c (r = 0.99, P = 0.0005, n = 4), and C57BL/6 mice (r = 0.85, P = 0.01, n = 7). AHR and HPS also correlated significantly on day 77 in C57BL/6 mice (r = 0.95, P = 0.012, n = 5).
In BALB/c mice alone, RSV loads measured by RLT-PCR correlated with AHR on day 14 (r = 0.69, P = 0.01, n = 12). RLT-PCR also correlated with AO on day 21 after RSV inoculation (r = 0. 71, P = 0.01, n = 12).
Discussion
Intranasal inoculation of RSV induced acute and chronic effects on the respiratory dynamics and inflammatory response in C57BL/6 mice, similar to those previously reported in BALB/c mice [7]. The acute manifestations of RSV disease in C57BL/6 mice appeared similar but there were several differences compared with BALB/c mice. Acute airway obstruction was more severe and more prolonged in C57BL/6 than in BALB/c mice. BALB/c mice, however, demonstrated more severe histopathologic changes, greater viral loads, and BAL concentrations of most of the cytokines measured.
It is known that RSV infection causes variable degree of disease severity in different mouse strains [8] but there was no information on whether the different genetic backgrounds influence the long term outcome of RSV infection including chronic inflammation and persistent AO and AHR. The varied responses of these two mouse strains to different pathogens have been attributed in part to differences in their T helper (Th) lymphocyte response. Studies in mice infected with Leishmania showed that a Th2 skewed response caused more severe disease in BALB/c mice than in C57BL/6 mice that developed a Th1-dominant response [9]. In some localized bacterial infections, however, the severity of disease was milder in BALB/c mice and was described as Th1 skewed [10]. Indistinct Th responses were found in both mouse strains when infected with certain viruses [11]. Hence, the differences in the immune response might not only depend on the genetic background of the host, but also on the site and the specific pathogen involved in the infectious process. Several groups have demonstrated AHR in C57BL/6 mice during the acute stages of RSV infection. [22-24]. However, so far, our study is the first to shed light on the dynamics of inflammatory infiltrates, airway dysfunction (AO and AHR), and inflammatory cytokines during the long-term phase of RSV-induced disease, up to 77 days after primary infection in both C57BL/6 and BALB/c mice.
Researchers found that during RSV re-challenge following sensitization with different RSV proteins, mice developed a variable Th cytokine response [12-15] which was not exclusively dependent on the strain haplotype [16]. Our experiments, conducted in mice with primary RSV infection, demonstrated that IFN-γ, a Th1 cytokine, was elevated after RSV infection in both strains. IFN-γ was increased in the respiratory tract of children with acute RSV infection [17,18] and correlated with disease severity [19]. In mice, IFN-γ also correlates with the development of RSV disease [20]. Our studies demonstrate that RSV infection induces IFN-γ in both strains of mice and the magnitude of the response is not consistent with the traditional notion that BALB/c are Th-2 weighted and C57BL/6 are Th1-weighted. BALB/c mice showed greater IFN-γ response than C57BL/6 mice as previously shown by others [21]. In our experiments IFN-γ demonstrated the greatest correlation with disease severity (Additional file 1).
Like IFN-γ, the dynamics of TNF-α, RANTES, MIP-1α were comparable in both mouse strains. TNF-α has been linked to RSV disease severity [25]. In our model, BAL concentrations of TNF-α peaked shortly after RSV infection and correlated with markers of disease severity in BALB/c mice (Additional file 1), and were significantly greater in BALB/c than in C57BL/6 mice (Figure 5).
The CC chemokines, RANTES and MIP-1α, and the CXC chemokine MIG were also elevated in mice of both strains shortly after RSV inoculation. The biphasic production of RANTES, MIP-1α and MIG after RSV infection, with the second peak coinciding with the peak of viral replication and histological inflammation, has been previously described in BALB/c mice [7,26] but not in C57BL/6 mice. Although MIP-1α has been detected in C57BL/6 mice [27], no previous studies demonstrated the production of RANTES in the respiratory tract of these mice after RSV infection. These two chemokines correlate with acute disease severity both in humans and in mice. [19,28-30]. Increased production of MIG after RSV infection has been documented in vitro [31] and by us in BALB/c mice[7]. Others have linked MIG to persistent airway inflammation due to continuous chemotaxis of mononuclear cells [32]. In the present study we demonstrate significant correlation between BAL MIG concentrations and AO, and viral load dynamics in both strains of mice. Taken together, these results indicate that RSV induces severe acute pulmonary disease in both strains of mice despite different genetic backgrounds. Although there are quantitative differences during the acute phase of the disease, there were no significant distinctions in the pattern of cytokine responses, lung inflammation or clinical manifestations of disease, and both strains developed long-term airway disease defined by chronic inflammatory infiltrates and abnormal AHR. These findings provide further experimental support to the link between RSV and chronic airway disease in humans. We believe that this is the first description of RSV-induced long-term pulmonary disease in C57BL/6 mice.
The exact mechanism by which different inflammatory mediators participate in the pathogenesis of the acute RSV-induced disease process is not completely understood, and their potential role in determining the chronic consequences of RSV infection, such as AO, AHR and chronic inflammation is even less characterized. Studies in murine models and in humans have not demonstrated yet a direct relation between the acute inflammatory response to RSV and its chronic consequences in the respiratory tract pathology. Our studies demonstrated prolonged AO and persistent AHR in both mouse strains, long after any of the cytokines, potentially responsible for persistent inflammation that we measured, became undetectable in BAL samples. It is possible that the use of more sensitive methods, perhaps in specific compartments, will contribute to clarify the role of these molecules in the chronic respiratory function abnormalities induced by RSV.
Until recently, it was suggested that initial RSV infection, although of short duration, was sufficient to elicit an exaggerated inflammatory response which would evolve into chronic inflammation and long-term airway abnormalities. An alternative, but not contradictory explanation for the development of chronic airway disease could be related to long-term antigenic stimulation. In this hypothesis, the presence of low level viral infection could represent a persistent stimulus responsible for the chronic inflammatory changes and the lasting pulmonary function abnormalities. The application of the RLT-PCR assay to this model demonstrated the persistence of RSV RNA in the respiratory tract for 11 weeks after inoculation in both strains of mice and was also associated with long-term pulmonary disease.
RSV can persist and replicate in vitro [33-35]. The persistence of RSV genomic RNA sequences and viral antigens has been described in guinea pigs for several weeks after acute primary infection [36-38]. Bovine RSV infection in calves resulted in persistent detection of RSV N gene for several weeks [39]. Miller et al. reported detection of G protein RNA by RLT-PCR in lungs of BALB/c mice up to day 12 after inoculation. [26].
It appears unlikely for a RNA virus to persist without at least low-level replication in the host. This is supported by the studies by Schwarze et al., who found RSV genomic and messenger RNA in mouse lung homogenates for 100 days after inoculation, and were able to recover low titer virus in culture after treatment with anti-CD4 and anti-CD8 antibodies [40].
Clinical studies have demonstrated that adults with stable COPD without acute exacerbation had RSV RNA detected by PCR in the upper respiratory tract [41]. The persistence of RSV RNA in children has not been extensively studied. Post-mortem specimens of children who died of sudden infant death syndrome showed positive results by PCR targeting the RSV N gene in 27% of cases and in 18% of the controls. Since some children died during summer the authors suggested the possibility of viral persistence [42]. The persistence of RSV in the respiratory tract of children and its potential contribution to the development and maintenance of AHR is a critical question that needs to be addressed.
In summary, we provide evidence that RSV-induced acute and chronic airway disease is reproducible in two genetically distinct mouse strains. The long-term airway disease coincided with persistent detection of RSV genome in the respiratory tract. It is yet unclear whether the persistence of RSV RNA plays a role in the development of chronic airway disease. Further studies are needed to characterize the potential mechanisms that would allow viral persistence and its possible contribution to the pathogenesis of RSV-induced long-term pulmonary disease.
Materials and methods
Mice
Female, 7–8-week old pathogen-free C57BL/6 and BALB/c mice (Charles River Laboratories, Wilmington, MA) were maintained in filter top cages, and routinely monitored for other pathogens [7]. After inoculation, all mice were kept under the same conditions and were provided identical care. Sentinel mice housed in the mouse storage room are routinely used for health surveillance. Sentinel mice had no detectable antibodies against mouse hepatitis virus, Sendai virus, pneumonia virus of mice, reo-3 virus, mouse encephalitis virus (GD-7), mouse rotavirus (EDIM), minute virus of mice, and Mycoplasma pulmonis; screening for pinworm and mites was also negative. This study was approved by the Institutional Animal Care and Use Committee at the University of Texas Southwestern Medical Center at Dallas.
Virus
RSV A-2 strain was maintained in our laboratory as described [7]. Results of plaque assays were reported in Log10 PFU/mL, with 1.7 Log10 PFU/mL being the lowest limit of detection.
Inoculation
Mice were anesthetized using inhaled methoxyfluorane and intranasally inoculated with 107 PFU of RSV in 100 μl of 10% Eagle's minimal essential medium (EMEM) [6]. Uninfected control animals were sham-inoculated with 100 μl of sterile 10% EMEM. Animals were allowed 30 seconds to aspirate the inoculum while held upright until fully recovered from the anesthesia.
Plethysmography
Unrestrained, whole-body plethysmography (Buxco Electronics, Inc. Sharon, CT) was used to measure the Enhanced Pause (Penh) to evaluate airway obstruction (AO) and airway hyperresponsiveness (AHR), as previously described [7,43]. Briefly, mice were allowed to acclimate to the chamber, and then plethysmograph readings were recorded to establish baseline Penh values to determine AO. Next the mice were exposed to aerosolized methacholine (Sigma; 50 mg/ml) for 4 minutes; after exposure, plethysmograph readings were recorded again to determine AHR. Groups of infected and control mice were always evaluated in parallel at all time points during the entire study. Preliminary studies showed that normal C57BL/6 mice have greater baseline Penh values than normal BALB/c mice; therefore, groups of infected and control mice of each strain were evaluated in parallel during the entire study. Plethysmography was performed in the groups of mice to be sacrificed for sample collection at each time point, and in an additional group of mice that were followed throughout each experiment. A total of 4 independent experiments including 10–30 mice per group per time point were conducted.
Sample collection
Mice were anesthetized with an intraperitoneal injection of 75 mg/kg of ketamine and 5 mg/kg of acepromazine before euthanasia by exsanguination. On average, 4–12 mice were sacrificed per group per time point. Bronchoalveolar lavage (BAL) specimens were obtained as described previously [7,43] to measure cytokine concentrations and RSV loads. Histologic evaluation was performed in whole-lung specimens fixed with a 10% buffered formalin solution. Whole lung samples were also collected from another group of mice for determination of viral loads by plaque assay and by real-time polymerase chain reaction (RLT-PCR). Specimens were obtained on days 0 (within 2 hours of inoculation), 1, 3, 5, 8, 9, 12, 14, 21, 28, 42, and 77 post-inoculation.
Histopathology
Formalin fixed lungs were paraffin embedded, sectioned and stained with hematoxylin and eosin. Histopathologic score (HPS) was determined by a pathologist, unaware of the infection status of the animals. Each section was graded on the basis of a cumulative score from 5 categories: (1) peribronchiolar and bronchial infiltrates, (2) bronchiolar and bronchial luminal exudates, (3) perivascular infiltrates, (4) the number of monocytes, and (5) parenchymal pneumonia. This HPS system assigns values from 0 to 21. This scoring system has been previously validated in RSV infection [7,43].
BAL cytokines
Concentrations of TNF-α, IFN-γ, IL-4, IL-10, MIG, RANTES, and MIP-1α were measured in BAL specimens by ELISA (R&D Systems, Minneapolis, Minn.) for up to 14 days after inoculation. The lower limit of detection for each of these assays were: 120 pg/mL for TNF-α, 50 pg/mL for IFN-γ, 40 pg/mL for IL4, 16 pg/mL for IL-10, 16 pg/ mL for MIG, 200 pg/mL for RANTES and 60 pg/mL for MIP-1α. For statistical analysis, samples with optical density readings below the limit of the standard curve of the assay were assigned a value half that of the detection level.
RSV loads by plaque assay and Real-time Polymerase Chain Reaction (RLT-PCR)
Plaque assay tissue culture was used to measure viral loads in BAL specimens and lung homogenate supernatants as previously described [7]. Lung supernatant samples were prepared by placing whole lung specimens in 1 mL of 10% EMEM, on ice. Lungs were homogenized using sterile probes and a rotor tissue homogenizer (Omni International Inc., Marietta, GA). Lung supernatants were separated by centrifugation at 2000 g for 10 minutes at 4°C and 100 μL were used immediately for determination of viral loads by plaque assay; the remaining supernatant was frozen at -80°C for viral load measurements by RLT-PCR. In a subset of experiments, lung samples used for RLT-PCR were placed in 1.5 mL of RNA stabilization reagent (RNA later, Qiagen Inc., Valencia, CA) immediately after sample collection. Lung supernatants were then obtained by homogenization and centrifugation. The RLT-PCR viral load values measured using this method were similar to those obtained when no RNA later was used; therefore, all results of lung homogenate supernatants are presented together. One sample per mouse was evaluated as a single specimen. Quantitative RLT-PCR was used targeting the conserved region of the RSV N-gene. Forward (5'-AGA TCA ACT TCT GTC ATC CAG CAA) and reverse (5'-TTC TGC ACA TCA TAA TTA GGA GTA TCA AT) primers amplified an 85-bp region containing the 25-mer FAM-labeled probe (5'-CAC CAT CCA ACG GAG CAC AGG AGA T), as previously described [44]. RSV RNA was extracted from 1 mL of lung supernatant samples using ion-exchange mini-columns (Qiagen RNeasy Mini Kit, Valencia, CA, USA) and cDNA was prepared by reverse transcription using 2.5 uM random hexamers for 10 min at 22°C, 30 min at 42°C and 5 min at 95°C. Real-time PCR was performed using a Perkin-Elmer /Applied Biosystems 7700 sequence detector (Foster City, CA) using 10 μl cDNA/ in a total volume of 50 μl master mix with the following run conditions: 1 cycle for 2 min at 50°C and 10 min at 95°C each, followed by 50 cycles for 15 seconds at 95°C and 60 seconds at 60°C. RSV A known concentrations were used to derive a standard curve. Standards and negative controls were run together with each PCR assay. The lower limit of detection of the assay was 10 viral copies/mL.
Statistical methods
T-test or Mann-Whitney Rank Sum test were used according to data distribution for comparisons between groups and the Spearman Rank Order test was used for correlations, as all the data taken together were not normally distributed, using the SigmaStat® (SPSS, Inc. Chicago, Illinois) software package.
Abreviations used
PCR polymerase chain reaction
TNF tumor necrosis factor
IL interleukin
IFN interferon
MIP macrophage inflammatory protein
RANTES regulated on activation, normal T-cell expressed and secreted
MIG monokine induced by interferon-gamma
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
SCB study design, RSV infection, data analyses; AM BAL, formalin fixation, plethysmography; AMG histopathology and special staining; KDO real-time PCR; AMR and MFA plethysmography and BAL; OR data interpretation; HSJ project design, experimental analyses and interpretation.
Supplementary Material
Additional File 1
Significant Correlations on Day 5 after RSV Inoculation in C57BL/6 and BALB/c mice. r = correlation coefficient; p = p-value, n = number of samples analyzed. Spearman's rank order analysis was used. Statistically significant correlations (p < 0.05) listed in bold font. N/A, not applicable, samples from different animals; * TNF-α concentrations undetectable on day 5.
Click here for file
Acknowledgements
S.C.B. is supported in part by a Pediatric Infectious Disease Society Fellowship Award sponsored by GlaxoSmithKline Pharmaceuticals.
A.M. was supported in part by the Pediatric Fellowship Award in Viral Respiratory Infectious Diseases supported by MedImmune Inc., at the Pediatric Academic Societies Annual Meeting.
H.S.J. is supported in part by grants from the Children's Clinical Research Advisory Committee and the Children's Medical Center of Dallas Foundation and American Lung Association.
Approved by the Institutional Animal Care and Use Committee at the University of Texas Southwestern Medical Center at Dallas.
==== Refs
Shay DK Holman RC Newman RD Liu LL Stout JW Anderson LJ Bronchiolitis-associated hospitalizations among US children, 1980-1996 Jama 1999 282 1440 1446 10535434 10.1001/jama.282.15.1440
Martinez FD Wright AL Taussig LM Holberg CJ Halonen M Morgan WJ Asthma and wheezing in the first six years of life. The Group Health Medical Associates N Engl J Med 1995 332 133 138 7800004 10.1056/NEJM199501193320301
Stein RT Sherrill D Morgan WJ Holberg CJ Halonen M Taussig LM Wright AL Martinez FD Respiratory syncytial virus in early life and risk of wheeze and allergy by age 13 years Lancet 1999 354 541 545 10470697 10.1016/S0140-6736(98)10321-5
Sigurs N Bjarnason R Sigurbergsson F Kjellman B Respiratory syncytial virus bronchiolitis in infancy is an important risk factor for asthma and allergy at age 7 Am J Respir Crit Care Med 2000 161 1501 1507 10806145
Sigurs N Gustafsson PM Bjarnason R Lundberg F Schmidt S Sigurbergsson F Kjellman B Severe respiratory syncytial virus bronchiolitis in infancy and asthma and allergy at age 13 Am J Respir Crit Care Med 2005 171 137 141 15516534 10.1164/rccm.200406-730OC
Karron RA Singleton RJ Bulkow L Parkinson A Kruse D DeSmet I Indorf C Petersen KM Leombruno D Hurlburt D Santosham M Harrison LH Severe respiratory syncytial virus disease in Alaska native children. RSV Alaska Study Group J Infect Dis 1999 180 41 49 10353859 10.1086/314841
Jafri HS Chavez-Bueno S Mejias A Gomez AM Rios AM Nassi SS Yusuf M Kapur P Hardy RD Hatfield J Rogers BB Krisher K Ramilo O Respiratory syncytial virus induces pneumonia, cytokine response, airway obstruction, and chronic inflammatory infiltrates associated with long-term airway hyperresponsiveness in mice J Infect Dis 2004 189 1856 1865 15122522 10.1086/386372
Byrd LG Prince GA Animal models of respiratory syncytial virus infection Clin Infect Dis 1997 25 1363 1368 9431379
Sacks D Noben-Trauth N The immunology of susceptibility and resistance to Leishmania major in mice Nat Rev Immunol 2002 2 845 858 12415308 10.1038/nri933
Hazlett LD Pathogenic mechanisms of P. aeruginosa keratitis: a review of the role of T cells, Langerhans cells, PMN, and cytokines DNA Cell Biol 2002 21 383 390 12167240 10.1089/10445490260099665
Ramakrishna C Ravi V Desai A Subbakrishna DK Shankar SK Chandramuki A T helper responses to Japanese encephalitis virus infection are dependent on the route of inoculation and the strain of mouse used J Gen Virol 2003 84 1559 1567 12771426 10.1099/vir.0.18676-0
Bangham CR Cannon MJ Karzon DT Askonas BA Cytotoxic T-cell response to respiratory syncytial virus in mice J Virol 1985 56 55 59 2411951
Alwan WH Record FM Openshaw PJ Phenotypic and functional characterization of T cell lines specific for individual respiratory syncytial virus proteins J Immunol 1993 150 5211 5218 8515055
Hancock GE Tebbey PW Scheuer CA Pryharski KS Heers KM LaPierre NA Immune responses to the nonglycosylated ectodomain of respiratory syncytial virus attachment glycoprotein mediate pulmonary eosinophilia in inbred strains of mice with different MHC haplotypes J Med Virol 2003 70 301 308 12696122 10.1002/jmv.10395
Hussell T Georgiou A Sparer TE Matthews S Pala P Openshaw PJ Host genetic determinants of vaccine-induced eosinophilia during respiratory syncytial virus infection J Immunol 1998 161 6215 6222 9834108
Srikiatkhachorn A Chang W Braciale TJ Induction of Th-1 and Th-2 responses by respiratory syncytial virus attachment glycoprotein is epitope and major histocompatibility complex independent J Virol 1999 73 6590 6597 10400756
Garofalo RP Hintz KH Hill V Ogra PL Welliver RCS Production of interferon gamma in respiratory syncytial virus infection of humans is not associated with interleukins 12 and 18 J Med Virol 2004 73 289 294 15122806 10.1002/jmv.20089
van Benten IJ van Drunen CM Koopman LP KleinJan A van Middelkoop BC de Waal L Osterhaus AD Neijens HJ Fokkens WJ RSV-induced bronchiolitis but not upper respiratory tract infection is accompanied by an increased nasal IL-18 response J Med Virol 2003 71 290 297 12938205 10.1002/jmv.10482
Garofalo RP Patti J Hintz KA Hill V Ogra PL Welliver RC Macrophage inflammatory protein-1alpha (not T helper type 2 cytokines) is associated with severe forms of respiratory syncytial virus bronchiolitis J Infect Dis 2001 184 393 399 11471095 10.1086/322788
van Schaik SM Obot N Enhorning G Hintz K Gross K Hancock GE Stack AM Welliver RC Role of interferon gamma in the pathogenesis of primary respiratory syncytial virus infection in BALB/c mice J Med Virol 2000 62 257 266 11002257 10.1002/1096-9071(200010)62:2<257::AID-JMV19>3.0.CO;2-M
Johnson TR Hong S Van Kaer L Koezuka Y Graham BS NK T cells contribute to expansion of CD8(+) T cells and amplification of antiviral immune responses to respiratory syncytial virus J Virol 2002 76 4294 4303 11932395 10.1128/JVI.76.9.4294-4303.2002
Tekkanat KK Maassab H Berlin AA Lincoln PM Evanoff HL Kaplan MH Lukacs NW Role of interleukin-12 and stat-4 in the regulation of airway inflammation and hyperreactivity in respiratory syncytial virus infection Am J Pathol 2001 159 631 638 11485921
Makela MJ Kanehiro A Dakhama A Borish L Joetham A Tripp R Anderson L Gelfand EW The failure of interleukin-10-deficient mice to develop airway hyperresponsiveness is overcome by respiratory syncytial virus infection in allergen-sensitized/challenged mice Am J Respir Crit Care Med 2002 165 824 831 11897651
Schwarze J Cieslewicz G Hamelmann E Joetham A Shultz LD Lamers MC Gelfand EW IL-5 and eosinophils are essential for the development of airway hyperresponsiveness following acute respiratory syncytial virus infection J Immunol 1999 162 2997 3004 10072551
Rutigliano JA Graham BS Prolonged production of TNF-alpha exacerbates illness during respiratory syncytial virus infection J Immunol 2004 173 3408 3417 15322205
Miller AL Bowlin TL Lukacs NW Respiratory syncytial virus-induced chemokine production: linking viral replication to chemokine production in vitro and in vivo J Infect Dis 2004 189 1419 1430 15073679 10.1086/382958
Domachowske JB Bonville CA Gao JL Murphy PM Easton AJ Rosenberg HF MIP-1alpha is produced but it does not control pulmonary inflammation in response to respiratory syncytial virus infection in mice Cell Immunol 2000 206 1 6 11161432 10.1006/cimm.2000.1730
Noah TL Becker S Chemokines in nasal secretions of normal adults experimentally infected with respiratory syncytial virus Clin Immunol 2000 97 43 49 10998316 10.1006/clim.2000.4914
Noah TL Ivins SS Murphy P Kazachkova I Moats-Staats B Henderson FW Chemokines and inflammation in the nasal passages of infants with respiratory syncytial virus bronchiolitis Clin Immunol 2002 104 86 95 12139952 10.1006/clim.2002.5248
Sheeran P Jafri H Carubelli C Saavedra J Johnson C Krisher K Sanchez PJ Ramilo O Elevated cytokine concentrations in the nasopharyngeal and tracheal secretions of children with respiratory syncytial virus disease Pediatr Infect Dis J 1999 18 115 122 10048682 10.1097/00006454-199902000-00007
Bitko V Garmon NE Cao T Estrada B Oakes JE Lausch RN Barik S Activation of cytokines and NF-kappa B in corneal epithelial cells infected by respiratory syncytial virus: potential relevance in ocular inflammation and respiratory infection BMC Microbiol 2004 4 28 15256003 10.1186/1471-2180-4-28
Kelsen SG Aksoy MO Yang Y Shahabuddin S Litvin J Safadi F Rogers TJ The Chemokine Receptor, CXCR3, and its Splice Variants are Expressed in Human Airway Epithelial Cells Am J Physiol Lung Cell Mol Physiol 2004
Pringle C R Shirodaria PV Cash P Chiswell DJ Malloy P Initiation and maintenance of persistent infection by respiratory syncytial virus J Virol 1978 28 199 211 702647
Panuska JR Cirino NM Midulla F Despot JE McFadden ERJ Huang YT Productive infection of isolated human alveolar macrophages by respiratory syncytial virus J Clin Invest 1990 86 113 119 2365811
Guerrero-Plata A Ortega E Gomez B Persistence of respiratory syncytial virus in macrophages alters phagocytosis and pro-inflammatory cytokine production Viral Immunol 2001 14 19 30 11270594 10.1089/08828240151061347
Hegele RG Robinson PJ Gonzalez S Hogg JC Production of acute bronchiolitis in guinea-pigs by human respiratory syncytial virus Eur Respir J 1993 6 1324 1331 8287949
Hegele RG Hayashi S Bramley AM Hogg JC Persistence of respiratory syncytial virus genome and protein after acute bronchiolitis in guinea pigs Chest 1994 105 1848 1854 8205887
Streckert HJ Philippou S Riedel F Detection of respiratory syncytial virus (RSV) antigen in the lungs of guinea pigs 6 weeks after experimental infection and despite of the production of neutralizing antibodies Arch Virol 1996 141 401 410 8645083 10.1007/BF01718305
Valarcher JF Bourhy H Lavenu A Bourges-Abella N Roth M Andreoletti O Ave P Schelcher F Persistent infection of B lymphocytes by bovine respiratory syncytial virus Virology 2001 291 55 67 11878876 10.1006/viro.2001.1083
Schwarze J O'Donnell DR Rohwedder A Openshaw PJ Latency and persistence of respiratory syncytial virus despite T cell immunity Am J Respir Crit Care Med 2004 169 801 805 14742302 10.1164/rccm.200308-1203OC
Rohde G Wiethege A Borg I Kauth M Bauer TT Gillissen A Bufe A Schultze-Werninghaus G Respiratory viruses in exacerbations of chronic obstructive pulmonary disease requiring hospitalisation: a case-control study Thorax 2003 58 37 42 12511718 10.1136/thorax.58.1.37
Cubie HA Duncan LA Marshall LA Smith NM Detection of respiratory syncytial virus nucleic acid in archival postmortem tissue from infants Pediatr Pathol Lab Med 1997 17 927 938 9353832 10.1080/107710497174372
Mejias A Chavez-Bueno S Rios AM Saavedra-Lozano J Fonseca Aten M Hatfield J Kapur P Gomez AM Jafri HS Ramilo O Anti-respiratory syncytial virus (RSV) neutralizing antibody decreases lung inflammation, airway obstruction, and airway hyperresponsiveness in a murine RSV model Antimicrob Agents Chemother 2004 48 1811 1822 15105140 10.1128/AAC.48.5.1811-1822.2004
van Woensel JB Lutter R Biezeveld MH Dekker T Nijhuis M van Aalderen WM Kuijpers TW Effect of dexamethasone on tracheal viral load and interleukin-8 tracheal concentration in children with respiratory syncytial virus infection Pediatr Infect Dis J 2003 22 721 726 12913774
|
15963240
|
PMC1183252
|
CC BY
|
2021-01-04 16:39:00
|
no
|
Virol J. 2005 Jun 20; 2:51
|
latin-1
|
Virol J
| 2,005 |
10.1186/1743-422X-2-51
|
oa_comm
|
==== Front
World J Surg OncolWorld Journal of Surgical Oncology1477-7819BioMed Central London 1477-7819-3-381596975310.1186/1477-7819-3-38ResearchE-cadherin and β-catenin expression in early stage cervical carcinoma: a tissue microarray study of 147 cases Fadare Oluwole [email protected] Harini [email protected] Jun [email protected] Denise [email protected] Peter E [email protected] Wenxin [email protected] Department of Pathology, Yale University School of Medicine, New Haven, CT, USA2 Department of Obstetrics, Gynecology and Reproductive Sciences, Yale University School of Medicine, New Haven, CT, USA3 Department of Preventive Medicine, Keck School of Medicine, University of Southern California, Los Angeles, CA, USA2005 21 6 2005 3 38 38 12 3 2005 21 6 2005 Copyright © 2005 Fadare et al; licensee BioMed Central Ltd.2005Fadare et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The disruption of intercellular adhesions is an important component of the acquisition of invasive properties in epithelial malignancies. Alterations in the cell-cell adhesion complex, E-Cadherin/β-Catenin, have been implicated in the oncogenesis of carcinomas arising from various anatomic sites and have been correlated with adverse clinico-pathologic parameters. In this study, the authors investigated the immunohistochemical expression of E-Cadherin and β-Catenin in a cohort of early stage cervical cancers to determine its prognostic significance and to investigate differences between the three major histological subtypes.
Patients and methods
A tissue microarray of 147 cases of FIGO stage 1A and 1B cervical carcinomas [96 squamous cell carcinomas (SCC), 35 adenocarcinomas (AC), 12 adenosquamous carcinomas (ASQ), 4 miscellaneous types] was constructed from our archived surgical pathology files and stained with monoclonal antibodies to E-Cadherin and β-Catenin. Cases were scored by multiplying the intensity of staining (1 to 3 scale) by the percentage of cells stained (0–100%) for a potential maximum score of 300. For both markers, "preserved" expression was defined as bright membranous staining with a score of 200 or above. "Impaired" expression included any of the following: negative staining, a score less than 200, or exclusively cytoplasmic or nuclear delocalization.
Results
Impaired expression of β-Catenin was found in 85.7%, 66.7%, & 58.3% of AC, SCC & ASQ respectively. Impaired expression of E-Cadherin was found in 94.3%, 86.5% & 100% of cases of AC, SCC, & ASQ respectively. The differences between the histologic subtypes were not significant. For the whole cohort, a comparsion of cases showing impaired versus preserved of E-Cadherin and β-Catenin expression showed no significant differences with respect to recurrence free survival, overall survival, patient age, histologic grade, and frequency of lymphovascular invasion or lymph node involvement. There was no correlation between the status of both markers for all three histological subtypes (overall spearman correlation co-efficient r = 0.12, p = 0.14)
Conclusion
Impairment of E-Cadherin and β-Catenin expression is very frequent in early stage cervical cancers, and alterations in the E-Cadherin/β-Catenin cell adhesion complex are therefore likely involved in the pathogenesis of cervical carcinomas even at their earliest stages. None of the three major histological subtypes of cervical carcinoma (SCC, ADCA, ADSQ) is significantly more likely than the others to show impairment in E-Cadherin and β-Catenin expression. Overall, the expression of both markers does not significantly correlate with clinico-pathological parameters of prognostic significance.
==== Body
Background
The resounding success of the routine papanicolaou smear in reducing the incidence and mortality of cervical cancer has been chronicled extensively [1-6]. However, epidemiological data from 2004 indicates that 10,520 new cases of invasive cervical cancer are still diagnosed annually in the United States, with an associated mortality rate of approximately 37% [7]. For those patients with early stage (FIGO stage I and II) cervical cancers treated primarily with surgical therapy, the decision to administer adjuvant therapy is dependent on a variety of clinical and histopathological parameters. The latter includes the presence or absence of lymphovascular, or deep stromal invasion, large tumor size, tumor involvement of resection margins, involvement of regional lymph nodes etc [8]. Although histopathological parameters classify these patients into broad groups – each comprised of patients with similar risk-profiles for recurrence, the groupings are inherently non-specific [9]. Therefore, it is anticipated that some patients with cervical cancer receive unnecessary adjuvant therapy, with their attendant toxicities. It would thus be of interest to identify biomarkers which can be evaluated on the primary tumor and may further help refine the subset of patients at the most significant risk for recurrence and thus in need of intensive adjuvant management.
The disruption of intercellular adhesions is an important component of the acquisition of invasive properties in epithelial malignancies. Alterations in the cell-cell adhesion complex, E-Cadherin/β-Catenin, have been implicated in the oncogenesis of carcinomas arising from various anatomic sites and have been correlated with adverse clinico-pathological parameters [10-22]. Epithelial Cadherin (E-Cadherin) is a 120 kDa transmembrane glycoprotein which is involved in both homotypic and heterotypic Ca2+-dependent cellular adhesions [23-26]. Inactivation of E-Cadherin may occur through mutations, methylations or deletions of the E-cadherin gene, suppression of the E-Cadherin gene promoter, or posttranslational modification of the protein leading to cytoplasmic delocalization [27-30]. The strength of E-Cadherin-mediated intercellular adhesion is significantly increased by interactions between the cytoplasmic tail of E-cadherin and the cytoskeletal network [30]. This interaction is mediated through the cytoplasmic proteins β-Catenin, α-Catenin and γ-Catenin [23-26]. β-Catenin is a 92 kDa protein that, in addition to its cell-adhesion properties, alsoplays a role as a transcriptional co-activator in the Wnt signaling pathway; the latter is involved in cellular development, differentiation and oncogenesis [32,33]. Deregulation of the Wnt pathway may occur through an activating mutation of the β-Catenin gene, leading to accumulated levels of β-Catenin in the cytoplasm and nucleus, and culminating in the altered transcription of a variety of critical genes [26].
The cervix, in which a dysplasia-to-carcinoma sequence is well-established, offers a useful medium to comparatively study the expression of proteins involved in cell-to-cell adhesion in dysplastic and invasive epithelium. Previous studies have shown that the expression of E-Cadherin and β-Catenin, as evaluated immunohistochemically, is inversely proportional to the histologic grade in squamous intraepithelial lesions (SIL) of the cervix: expression of both markers is generally maintained in low-grade lesions and is lost in high grade lesions [34-36]. In the present study, the expression of E-Cadherin and β-Catenin was evaluated on a cohort of already invasive, albeit early stage cervical carcinomas. Our objectives are 1) to determine the frequency of loss of expression of E-Cadherin and β-Catenin in early stage cervical carcinomas 2) to determine whether the expression of either of these biomarkers is of prognostic significance 3) to specifically investigate whether there are any differences between the three major histological subtypes of cervical carcinoma with respect to expression of E-Cadherin and β-Catenin.
Patients and methods
Case selections and microarray construction
Following approval from our institutional review board, a tissue microarray (TMA) [37] of 147 cases of international federation of gynecology and obstetrics (FIGO) stage 1a and 1b cervical carcinomas was constructed. Technical details on microarray construction at the Yale tissue microarray facility are outlined elsewhere [38,39]. The cases were retrieved from the archived surgical pathology files at Yale New Haven Hospital and consisted of carcinomas diagnosed between 1987 and 2001. These cases were not consecutive but were selected on the basis of availability of at least one evaluable tumor-representative hematoxylin and eosin-stained slide and a paraffin block. On each case, slides were reviewed to select a representative tumor spot. The corresponding spot on the associated paraffin block was then cored and placed on a tissue microarrayer (Beecher Instruments, Silver Spring, MD, USA). The microarray was constructed with a 2-fold redundancy (2 spots for each patient), which has been previously shown by Camp et al [40] to accurately represent standard tissue sections in at least 95% of cases. The finalized arrays were then cut into 5 μm-thick sections and mounted on glass slides using an adhesive tape-transfer system (Instrumedics Inc., Hackensack, NJ, USA) with ultraviolet cross-linking.
Normal controls
To evaluate the staining patterns of E-Cadherin and β-Catenin in the normal cervix, 5 such cases were retrieved from our database. These 5 cases were hysterectomies performed for adnexal masses in which the final diagnoses were non-neoplastic (endometriosis [n = 4], tubo-ovarian abscess [n = 1]), Sections from the cervix in these cases were then stained with monoclonal antibodies to E-Cadherin and β-Catenin.
Immunohistochemistry
The tissue microarray slides were stained with antibodies to E-Cadherin (Clone NCH-38, dilution 1:100, DakoCytomation Corporation, Carpinteria, CA, USA) and β-Catenin (Clone 17C2, dilution 1:100, NovoCastra Laboratories Ltd., Tyne, UK) using a DAKO autostainer based on the standard avidin-biotin complex method. Tissue microarray slides were deparaffinized with xylene and graded alcohols then rehydrated with distilled water. Endogenous peroxidase activity was blocked by placing the slides in 0.5% hydrogen peroxidase/methanol for 10 minutes followed by a tap water rinse. Background staining was reduced by incubating slides in 0.3% bovine serum albumin/Tris-buffered saline. Antigen retrieval entailed placing the slides in a pressure cooker with an antigen unmasking solution (0.01M citrate buffer, pH 6.0) for 1 minute. Slides were subsequently incubated with the primary (4°C overnight), then biotinylated secondary antibodies and streptavidin-biotin peroxidase. 0.05% 3'3' diaminobenzidine (DAB) was used as chromogen, followed by counterstaining with hematoxylin.
Scoring of immunohistochemical staining
Since both E-Cadherin and β-Catenin are normally expressed in a bright membranous fashion in the cervical epithelium, aberrations on this basic theme were deemed abnormal for the purposes of this study. Thus, cytoplasmic staining, nuclear staining and no staining were considered "impaired" expression. Cases were scored by multiplying the intensity of staining (1 to 3 scale, see figure 1) by the percentage of cells stained (0–100%) for a potential maximum score of 300. For both markers, "preserved" expression was defined as bright membranous staining with a score of 200 or above. "Impaired" expression included any of the following: negative staining, a score less than 200, or exclusively cytoplasmic and/or nuclear delocalization. Only the core showing the highest score was utilized in each case. In 144 of 147 cases, the scores did not differ between the 2 cores on each case by more than 50 points. None of the 3 cases with such a discrepancy (i.e. differences of more than 100 points between the 2 cores) affected the "impaired" versus "preserved" expression distinction, because all 3 cases showed impaired expression in both cores. The microarrays were scored by 2 independent observers (OF and DH). Discrepancies were resolved at a consensus session. Results were then correlated with a variety of pathologic parameters, including tumor size, frequency of lymphovascular invasion, lymph node involvement and histological grade.
Figure 1 Representative cores illustrating the criteria used for scoring staining intensity (β-Catenin is shown; E-Cadherin was similar, although with an overall lower intensity).
Statistical analysis
For comparison of cases showing impaired versus preserved expression with respect to various clinico-pathological parameters, two-sided likelihood ratio chi-square test or Fisher's exact tests were used with p < 0.05 as the "statistically significant" cut-off point. Survival analysis was carried out by Kaplan-Meier method and Wilcoxon test was used to compare survival. Spearman correlation coefficients were calculated to examine the correlations between the two markers. All analyses were performed using SAS version 9.0 (SAS Institute, NC).
Validation
To externally validate the staining patterns observed on the microarray, 14 cases representing approximately 10% of the array were selected in a semi-random fashion using a randomization function (Microsoft Excel®, Microsoft Inc, Redmond, WA, USA) on a list of the entire cohort. The goal was to obtain a representative mix of cases with negative and positive stainings (4 groups, each consisting of 3 cases with a score of 0, 1–99, 100–199, 200–300; 2 additional cases for the last group). Thus, each case selected from the list by the randomization function was included in the validation list consecutively until the preset quantitative requirement for that group was met. On each of these 14 cases, a full representative tissue section was then stained with E-Cadherin and β-Catenin and scored using the same system used with the microarray.
Results
Normal epithelium
The staining patterns of E-Cadherin and β-Catenin were identical. However, β-Catenin staining was generally of greater intensity than E-Cadherin staining. In the endocervices of all 5 cases, both markers brightly decorated the epithelium in a circumferentially membranous fashion (basolateral and glycocalyceal). No cytoplasmic or nuclear staining was noted (Figures 2a and 2b). In the ectocervix, staining was also membranous but was limited to the lower two-thirds in 4 cases and the lower half in 1 case (Figure 2c). Squamous metaplastic epithelium was present in 3 cases, all of which showed transmural bright membranous staining (Figure 2d). In all 5 cases, the basal and parabasal cells displayed the greatest intensity of staining (3/3). There was no stromal staining.
Figure 2 Expression of E-Cadherin in the normal endocervix (2a, hematoxylin and eosin, original magnification ×100; 2b: hematoxylin and eosin, original magnification ×200), ectocervix (2c: hematoxylin and eosin, original magnification ×100) and squamous metaplastic epithelium at the transformation zone (2d) (hematoxylin and eosin, original magnification ×200).
Tissue microarray
The 147 cases consisted of 96 squamous cell carcinomas (SCC), 35 adenocarcinomas (ADCA), 12 adenosquamous carcinomas (ADSQ) and a heterogeneous group of 4 cases hereafter referred to as "others". The latter group included 1 small cell carcinoma, 1 pure large cell neuroendocrine carcinoma, 1 squamotransitional carcinoma and 1 adenosquamous carcinoma with morphologically-evident neuroendocrine differentiation. 4 cases were deemed unsatisfactory during scoring by virtue of poor tissue preservation on the microarray slides. The clinicopathological characteristics of the cohort are outlined in table 1.
Table 1 Summary of Clinical and Pathological characteristics of Cohort
Clinicopathologic Parameter n
Age (years) Median 40
Mean 42
Range 20–77
FIGO stage 1A1 14
1A2 15
1B1 109
1B2 9
Lymphovascular invasion Present 62
Absent 85
Histological grade 1 33
2 54
3 60
Lymph node involvement Present 18
Absent 129
Tumor Size (cm) 0.1–0.5 44
0.6–1.0 28
1.1–2.0 33
2.1–3.0 21
>3.1 21
Follow-up (years) Median 4.75
Mean 5.84
Range 0.12–17.43
FIGO: International Federation of Gynecology and Obstetrics
β-Catenin
Impaired expression of β-Catenin was present in 107 (75%) of the 143 cases and in the majority of all the histological subtypes. ADCA had the highest percentage of impaired beta-catenin expression (85.7%, 30/35). Compared to ADCA, a smaller proportion of cases of SCC showed impaired beta-catenin expression (66.7%, 64/96). However, the difference was not significant (p = 0.06). ADSQ showed an even lower proportion of cases with impaired beta-catenin expression (58.3%, 7/12). Compared to ADCA, the difference was not significant (p = 0.06). The difference between SCC and ADSQ was also not significant (p = 0.47). Impairment of expression was most commonly in the form of absent staining (79/107); 19 cases did not meet out quantitative criteria for positivity while 5 cases showed exclusively cytoplasmic delocalization. There were no cases of exclusively nuclear staining (Table 2). The mean age for the patients with preserved and impaired β-Catenin expression was 41.5 years (± 10.6) (n = 39) and 41.6 years (± 11.5) (n = 104) respectively. There was no significant difference (p = 0.99) between these 2 groups. Similarly, there were no statistically significant differences with respect to recurrence free survival (Figure 3), overall survival (Figure 4), frequency of lymphovascular invasion, histologic grade, and frequency of lymph node involvement (Table 3) when cases with impaired expression of β-Catenin were compared with those in which β-Catenin was preserved.
Table 2 Number of patients expressing E-Cadherin and β-Catenin among the various histologic subtypes
Histologic subtype of cervical carcinoma Preserved¶ Impaired U
Reduced Staining ▼ Exclu-Sively Cyto-Plasmic Stain Ing Negative Staining Nuclear Staining
β-catenin expression
Squamous cell Carcinoma (n = 96) 29 10 2 52 0 3
Adenocarcinoma (n = 35) 5 7 1 22 0 0
Adenosquamous (n = 12) 5 1 2 4 0 0
Others (n = 4) 1 1 0 1 0 1
Total (n = 147) 40 19 5 79 0 4
E-cadherin expression
Squamous cell Carcinoma (n = 96) 11 16 2 65 0 2
Adenocarcinoma (n = 35) 1 5 3 25 0 1
Adenosquamous (n = 12) 0 3 2 7 0 0
Others (n = 4) 1 0 1 2 0 0
Total (n = 147) 13 24 8 98 0 4
U: unsatisfactory
¶ membranous staining with 200 score or higher
▼less than 200 score
Figure 3 Kaplan-Meier recurrence-free survival curves comparing patients whose cervical carcinomas showed preserved expression of E-Cadherin and β-Catenin to those with impaired expression (▲ indicates censored observations).
Figure 4 Kaplan-Meier overall survival curves comparing patients whose cervical carcinomas showed preserved expression of E-Cadherin and β-Catenin to those with impaired expression (▲ indicates censored observations).
Table 3 Lack of association between expression of E-Cadherin or β-Catenin and lymphovascular invasion, frequency of lymph node involvement or histologic grade
Preserved¶ Impaired
Reduced Staining ▼ Exclusively Cytoplasmic Staining Negative Staining Nuclear Staining p value (impaired versus preserved expression)
1) Lymphovascular invasion
β-Catenin expression
Absent (n = 83) 20 7 2 54 0 0.4
Present (n = 60) 20 12 3 25 0
E-Cadherin expression
Absent (n = 83) 6 11 3 63 0 0.4
Present (n = 60) 7 13 5 35 0
2) Histological grade*
β-Catenin expression
1 (n = 31) 8 3 1 19 0 >0.05 for all comparisons
2 (n = 53) 20 7 1 25 0
3 (n = 59) 12 9 3 35 0
E-Cadherin expression
1 (n = 31) 1 5 2 23 0 >0.05 for all comparisons
2 (n = 53) 4 13 3 33 0
3 (n = 59) 8 6 3 42 0
3) Lymph node Involvement
β-Catenin expression
Absent (n = 125) 33 15 5 72 0 0.3
Present (n = 18) 7 4 0 7 0
E-Cadherin expression
Absent (n = 125) 9 21 7 88 0 0.06
Present (n = 18) 4 3 1 10 0
¶ membranous staining with 200 score or higher
▼less than 200 score
E-Cadherin
Of the 143 cases, 134 (94%) showed impaired expression of E-Cadherin. Notably, impaired expression of E-Cadherin was present in 100% of the 12 ADSQ; however, high frequencies of impaired expression were also present in ADCA and SCC and the differences between the histologic subtypes were not statistically significant. Impairment of expression was most commonly in the form of absent staining (99/134); 24 cases did not meet out quantitative criteria for positivity while 8 cases showed exclusively cytoplasmic delocalization. No case showed nuclear staining (Table 2). The mean age for the patients with preserved and impaired E-Cadherin expression was 36.5 years (± 9.1) (n = 12) and 42.4 years (± 11.7) (n = 131) respectively. Thus, patients with impaired E-Cadherin expression appeared to be older than patients with preserved E-Cadherin expression. However, the difference was not statistically significant (p = 0.09). As with β-Catenin, there were no statistically significant differences with respect to recurrence free survival (Figure 3), overall survival (Figure 4), frequency of lymphovascular invasion, lymph node involvement and histologic grade (Table 3) when cases with impaired expression of E-Cadherin were compared with those in which E-Cadherin was preserved.
β-Catenin and E-Cadherin correlation analysis
Only six (4%) of 143 cases showed preservation of both E-Cadherin and β-Catenin expression and 93 (65%) of 143 cases showed impairment in the expression of both markers, 7 cases showed impairment in β-Catenin and preservation of E-Cadherin whereas the reverse was observed in 34 cases. Overall, the expressions of E-Cadherin and β-Catenin were not significantly correlated (spearman correlation coefficient r = 0.12, p = 0.14). Similar findings were noted when each of the 3 major histological subtypes was subjected to the same analysis (table 4)
Table 4 Lack of correlation between E-Cadherin and β-Catenin expression among the 3 major histologic subtypes of invasive carcinoma
Histological Subtype E-Cadherin expression Correlation
Preserved Impaired
β-Catenin expression Squamous cell carcinoma Preserved 5 24
Impaired 6 56 r = 0.11 p = 0.3
Adenocarcinoma Preserved 0 5
Impaired 1 28 r = -0.07 p = 0.7
Adenosquamous carcinoma Preserved 0 5
Impaired 0 7 Not applicable
Validation
In 13 of 14 cases, the scores obtained on standard histologic sections were within 50 points of the scores obtained on the corresponding tissue microarray spot (average difference: 30). In the 14th case, the tissue microarray spot was scored 0 whereas the corresponding histologic section was scored 100. In none of the 14 cases did a score discrepancy affect the "preserved" versus "impaired" distinction.
Discussion
In this study, we demonstrate that 1) loss of expression of both E-Cadherin and β-Catenin are frequent events in early stage cervical carcinomas of all three major histological subtypes, and both proteins are thus likely participants in the pathogenesis of cervical carcinoma, 2) Loss of expression of both proteins is of no prognostic significance with respect to the following parameters: recurrence free survival, overall survival, frequency of lymphovascular invasion, histologic grade, and frequency of lymph node involvement. Our findings are in concordance with the recent study of Van de Putte et al [41], in which cases of stage 1B SCC were analyzed. In that study, loss of expression of E-Cadherin and β-Catenin in greater than 50% of tumor cells was found in 90% and 72% of cases respectively. However, the authors found no correlation between the expression of E-Cadherin and β-Catenin (and other catenins) and prognosis [41]. In contrast, Jeffers et al [42] found strong expression of E-Cadherin in all 20 cases of invasive cervical carcinoma that were studied. However, those 20 cases were an admixture of all stages and the prognostic significance of E-Cadherin was not specifically investigated since all cases were positive. Other studies have also showed a high frequency of loss of E-Cadherin expression in cervical cancers. Applying the criteria of the present study, impaired expression of E-cadherin was found in 89% of invasive SCCs in one study [34]. In the study of Sun et al [43], 60 cases of invasive cervical carcinoma of all stages were investigated for E-cadherin expression. Loss of expression of E-Cadherin was found in 53.3% of cases. Additionally, the authors found a correlation between abnormal E-Cadherin expression and clinical stage and, in contrast to the present study, lymph node involvement and histologic grade. Similarly, among all stages of cervical adenocarcinoma, abnormal expression of β-Catenin was significantly associated with advanced pathologic stage and thus disease-free survival in another study [44]. In the latter two studies, the noted discrepancies with the current one may be related to qualitative technical differences (antibody concentration, antigen retrieval methods etc) or differences in interpretation (i.e. criteria for positivity). None of the above studies have specifically compared the 3 major histologic subtypes in early stage cervical carcinoma. However, the majority of studies show that loss of expression of E-Cadherin and β-Catenin is frequent in cervical carcinomas. Although their impact on prognosis of this expression is less certain, our study finds that the expression of both E-Cadherin and β-Catenin lacks prognostic significance at least in early stage cervical carcinoma.
An orderly, membranous expression of E-Cadherin and β-Catenin is found in the normal cervix. The loss of expression of both proteins in a high proportion of high-grade squamous intraepithelial lesions [34-36] suggests that dysregulation of this pathway is an early event in cervical carcinogenesis. We investigated cytoplasmic and nuclear staining as a manifestation of impaired expression of both E-Cadherin and β-Catenin [26]. In the typical normal cell β-Catenin's complex with E-Cadherin and the cytoskeletal network is inversely proportional to the association of β-Catenin with the adenomatous polyposis coli protein, a large multifunctional cytosolic protein. Thus, normally, only small portions of β-Catenin are found in the cytoplasm since the association of β-Catenin with APC eventuates in its lytic degradation. Cytoplasmic accumulation often leads to nuclear accumulation, where β-Catenin may interact with a variety of proteins, culminating in transcriptional activation of a variety of critical genes [23-26]. Experimental studies have shown that tyrosine phosphorylation of β-Catenin by oncogenic products or growth factor receptors may cause dissociation of the E-Cadherin-associated adhesion complex from the cytoskeleton resulting in cellular (non-membranous) redistribution and disassociation of adherens junctions, a well-known feature of epithelial malignancies [45,46]. Exclusively cytoplasmic localization of β-Catenin and E-Cadherin, which was not present in any of the normal cervix epithelia, was seen in 3.4% and 5.4% of our carcinoma cases respectively. A progressive increase in the proportion of cases showing this finding was found in low grade SIL, high grade SIL and invasive carcinomas in one study [34]. However, the logical extension of that finding, a correlation with histological grade in invasive carcinomas, was not found in this study. Our study is also in concordance with that of Shinohara et al [47], in which no relationship was found between the frequency of cytoplasmic/nuclear localization of β-Catenin and histologic grade (no cases of nuclear localization β-Catenin was found in our study).
The possibility that our findings may have been significantly affected by our methods cannot be entirely excluded. Tissue microarray technology as a high throughput modality has gained wide acceptance in pathology investigations and is used routinely [48-52]. Our TMA was constructed with a two-fold redundancy to minimize sampling errors [40]. Furthermore, in our validation set which constituted approximately 10% of the TMA, no significant discrepancies were seen. Nonetheless, error introduced by tumor heterogeneity remains a possibility. Most notably, the minimization of errors from the redundant (2-fold) construction of our TMA presumes a similar degree of tumor heterogeneity between breast [40] and cervical cancer, which may be untrue.
The importance of defining thresholds in any investigations reporting the loss of the immunohistochemical expression of a biomarker cannot be overemphasized. As previously noted, a threshold of 200 was used in this study because both E-Cadherin and β-Catenin are normally expressed in a membranous fashion in most of the cervical epithelium (scores 270–300 for the ectocervix and 300 for the endocervix). However, we performed separate statistical analyses when the threshold was lowered, i.e. when "impaired" expression included any of the following: negative staining, a score less than 99, or exclusively cytoplasmic and/or nuclear delocalization. The results on all previous analyses remained the same (p > 0.05), with 2 exceptions: 1) Overall, the expressions of E-Cadherin and β-Catenin became significantly correlated (spearman correlation coefficient r = 0.33, p = 0.0001). 2) 73% of LVI-negative carcinomas showed impaired β-Catenin expression, compared to 47% of LVI-positive carcinomas (p = 0.03). The significance of the latter finding is unclear, and it can be anticipated that changes in the threshold in either direction (lowering or raising) would result in at least one parameter attaining statistical significance with each change. In our opinion, for a marker that is normally expressed, a high threshold should be used for "loss" of expression, and 200/300 seems reasonable.
Conclusion
In the present report we demonstrated that impairment of E-Cadherin and β-Catenin expression is very frequent in early stage cervical cancers. None of the three major histological subtypes of cervical carcinoma (SCC, ADCA, ADSQ) is significantly more likely than the others to show impairment in E-Cadherin and β-Catenin expression. Alterations in the E-Cadherin/β-Catenin cell adhesion complex are therefore likely involved in the pathogenesis of cervical carcinomas even at their earliest stages. However, the expression of both markers does not significantly correlate with clinicopathologic parameters of prognostic significance.
Competing Interests
The author(s) declare that they have no competing interests.
Authors' Contributions
OF wrote the initial version of the manuscript
HR collected clinical data and summarized clinicopathologic information
JW performed statistical analysis and helped summarize clinicopathologic information
DH and OF scored the tissue microarray
WZ and PES conceived of, supervised, and sponsored the project. Both revised the manuscript.
All authors have read and approved the final manuscript
Acknowledgements
This study was presented in part at the 94th annual meeting of the United States and Canadian Academy of Pathology, Feb 26th – March 3rd, 2005, San Antonio, TX, USA
==== Refs
Koss LG The Papanicolaou test for cervical cancer detection: a triumph and tragedy JAMA 1989 261 737 743 2642983 10.1001/jama.261.5.737
van de Graaf Y Vooijs GP Zielhuis GA Cervical screening revisited Acta Cytol 1990 34 366 372 2111625
Devesa SS Young JL JrBrinton LA Fraumeni JF Jr Recent trends in cervix uteri cancer Cancer 1989 64 2184 2190 2804908
Wingo PA Cardinez CJ Landis SH Greenlee RT Ries LA Anderson RN Thun MJ Long-term trends in cancer mortality in the United States, 1930–1998 Cancer 2003 97 3133 3275 12784323 10.1002/cncr.11380
Adami HO Ponten J Sparen P Bergstrom R Gustafsson L Friberg LG Survival trend after invasive cervical cancer diagnosis in Sweden before and after cytologic screening, 1960–1984 Cancer 1994 73 140 147 8275416
Benedet JL Anderson GH Matistic JP A comprehensive program for cervical cancer detection and management Am J Obstet Gynecol 1992 166 1254 1259 1566781
Jemal A Tiwari RC Murray T Ghafoor A Samuels A Ward E Feuer EJ Thun MJ Cancer Statistics, 2004 CA Cancer J Clin 2004 54 8 29 14974761
Schilder JM Stehman FB Stage Ia-IIa cancer of the cervix Cancer J 2003 9 395 403 14690315
Sedlis A Bundy BN Rotman MZ Lentz SS Muderspach LI Zaino RJ A randomized trial of pelvic radiation therapy versus no further therapy in selected patients with stage IB carcinoma of the cervix after radical hysterectomy and pelvic lymphadenectomy: A Gynecologic Oncology Group Study Gynecol Oncol 1999 73 177 183 10329031 10.1006/gyno.1999.5387
Shimazui T Schalken JA Giroldi LA Jansen CF Akaza H Koiso K Debruyne FM Bringuier PP Prognostic value of cadherin-associated molecules (α-, β-, and γ-catenins and p120cas) in bladder tumors Cancer Res 1996 56 4154 4158 8797585
Krishnadath KK Tilanus HW van Blankenstein M Hop WC Kremers ED Dinjens WN Bosman FT Reduced expression of the cadherin-catenin complex in esophageal adenocarcinoma correlates with poor prognosis J Pathol 1997 182 331 338 9349237 10.1002/(SICI)1096-9896(199707)182:3<331::AID-PATH860>3.0.CO;2-D
van der Wurff AA Vermeulen SJ van der Linden EP Mareel MM Bosman FT Arends JW Patterns of α-and β-catenin and E-cadherin expression incolorectal adenomas and carcinomas J Pathol 1997 182 325 330 9349236 10.1002/(SICI)1096-9896(199707)182:3<325::AID-PATH865>3.0.CO;2-Y
De Leeuw WJ Berx G Vos CB Peterse JL Van de Vijver MJ Litvinov S Van Roy F Cornelisse CJ Cleton-Jansen AM Simultaneous loss of E-cadherin and catenins in invasive lobular breast cancer and lobular carcinoma in situ J Pathol 1997 183 404 411 9496256 10.1002/(SICI)1096-9896(199712)183:4<404::AID-PATH1148>3.0.CO;2-9
Shimazui T Bringuier PP van Berkel H Ruijter E Akaza H Debruyne FM Oosterwijk E Schalken JA Decreased expression of alpha-catenin is associated with poor prognosis of patients with localized renal cell carcinoma Int J Cancer 1997 74 523 528 9355975 10.1002/(SICI)1097-0215(19971021)74:5<523::AID-IJC8>3.0.CO;2-5
Zheng Z Pan J Chu B Wong YC Cheung AL Tsao SW Down regulation and abnormal expression of E-cadherin and β-catenin in nasopharyngeal carcinoma: Close association with advanced disease stage and lymph node metastasis Hum Pathol 1999 30 458 466 10208469 10.1016/S0046-8177(99)90123-5
Perl AK Wilgenbus P Dahl U Semb H Christofori G A causal role for E-cadherin in the transition from adenoma to carcinoma Nature 1998 392 190 191 9515965 10.1038/32433
Jawhari AU Noda M Farthing MJ Pignatelli M Abnormal expression and function of the E-cadherin-catenin complex in gastric carcinoma cell lines Br J Cancer 1999 80 322 330 10408833 10.1038/sj.bjc.6690358
Ross JS Figge HL Bui HX del Rosario AD Fisher HA Nazeer T Jennings TA Ingle R Kim DN E-cadherin expression in prostatic carcinoma biopsies: Correlation with tumor grade, DNA content, pathologic stage, and clinical outcome Mod Pathol 1994 7 835 841 7530850
Umbas R Isaacs WB Bringuier PP Schaafsma HE Karthaus HF Oosterhof GO Debruyne FM Schalken JA Decreased E-cadherin expression is associated with poor prognosis in patients with prostate cancer Cancer Res 1994 54 3929 3933 7518346
Faleiro-Rodrigues C Macedo-Pinto I Pereira D Lopes CS Loss of β-Catenin is associated with poor survival in ovarian carcinomas Int J Gynecol Pathol 2004 23 337 346 15381903 10.1097/01.pgp.0000139711.22158.14
Li Z Ren Y Lin SX Liang YJ Liang HZ Association of E-cadherin and beta-catenin with metastasis in nasopharyngeal carcinoma Chin Med J 2004 117 1232 1239 15361301
Chang HJ Jee CD Kim WH Mutation and altered expression of beta-catenin during gallbladder carcinogenesis Am J Surg Pathol 2002 26 758 766 12023580 10.1097/00000478-200206000-00009
Conacci-Sorrell M Zhurinsky J Ben-Ze'ev The cadherin-catenin adhesion system in signaling and cancer J Clin Invest 2002 109 987 991 11956233 10.1172/JCI200215429
Gooding JM Yap KL Ikura M The cadherin-catenin complex as a focal point of cell adhesion and signaling: new insights from three-dimensional structures BioEssays 2004 26 497 511 15112230 10.1002/bies.20033
Nelson WJ Nusse R Convergence of Wnt, β-Catenin, and Cadherin pathways Science 2004 303 1483 1487 15001769 10.1126/science.1094291
Hajra KM Fearon ER Cadherin and Catenin alterations in human cancer Genes, Chromosomes Cancer 2002 34 255 268 12007186 10.1002/gcc.10083
Berx G Cleton-Jansen AM Nollet F de Leeuw WJ van de Vijver M Cornelisse C van Roy F E-cadherin is a tumour/invasion suppressor gene mutated in human lobular breast cancers EMBO J 1995 14 6107 6115 8557030
Vos CB Cleton-Jansen AM Berx G de Leeuw WJ ter Haar NT van Roy F Cornelisse CJ Peterse JL van de Vijver MJ E-cadherin inactivation in lobular carcinoma in situ of the breast: an early event in tumorigenesis Br J Cancer 1997 76 1131 1133 9365159
Nawrocki B Polette M Van Hengel J Tournier JM Van Roy F Birembault P Cytoplasmic redistribution of E-cadherin-catenin adhesion complex is associated with down-regulated tyrosine phosphorylation of E-cadherin in human bronchopulmonary carcinomas Am J Pathol 1998 153 1521 1530 9811344
Hirohashi S Molecular aspects of adhesion-epigenetic mechanisms for inactivation of the E-Cadherin-mediated cell adhesion system in cancers Verh Dtsch Ges Pathol 2000 84 28 32 11217445
Yap AS Brieher Wm Pruschy M Grumbiner WM Lateral clustering of the adhesive ectodomain: a fundamental determinant of Cadherin function Curr Biol 1997 7 308 315 9133345 10.1016/S0960-9822(06)00154-0
Miller JR Hocking AM Brown JD Moon RT Mechanism and function of signal transduction by the Wnt/beta-Catenin and Wnt/Ca2+ pathways Oncogene 1999 18 7860 7872 10630639 10.1038/sj.onc.1203245
Peifer M Polakis P Wnt signaling in oncogenesis and embryogenesis-a look outside the nucleus Science 2000 287 1606 1609 10733430 10.1126/science.287.5458.1606
Faleiro-Rodrigues C Lopes C E-cadherin, CD44 and CD44v6 in squamous intraepithelial lesions and invasive carcinomas of the uterine cervix: an immunohistochemical study Pathobiology 2004 71 329 336 15627844 10.1159/000081729
de Boer CJ van Dorst E van Krieken H Jansen-van Rhijn CM Warnaar SO Fleuren GJ Litvinov SV Changing roles of cadherins and catenins during progression of squamous intraepithelial lesions in the uterine cervix Am J Pathol 1999 155 505 515 10433943
Yang JZ Zhang XH Wu WX Yan X Liu YL Wang JL Wang FR Expression of EP-CAM, beta-catenin in the carcinogenesis of squamous cell carcinoma of uterine cervix Zhonghua Zhong Liu Za Zhi 2003 25 372 375 12921570
Kononen J Bubendorf L Kallioniemi A Barlund M Schraml P Leighton S Torhorst J Mihatsch MJ Sauter G Kallioniemi OP Tissue microarrays for high-throughput molecular profiling of tumor specimens Nat Med 1998 4 844 847 9662379 10.1038/nm0798-844
Charette L Rimm DL Yale tissue microarray construction protocols. Version 1.0, 1/2001
Rimm DL Camp RL Charette LA Olsen DA Provost E Amplification of tissue by construction of tissue microarrays Exp Mol Pathol 2001 70 255 264 11418004 10.1006/exmp.2001.2363
Camp RL Charette LA Rimm DL Validation of tissue microarray technology in breast carcinoma Lab Invest 2000 80 1943 1949 11140706
Van de Putte G Kristensen GB Baekelandt M Lie AK Holm R E-cadherin and catenins in early squamous cervical carcinoma Gynecol Oncol 2004 94 521 527 15297198 10.1016/j.ygyno.2004.05.046
Jeffers MD Paxton J Bolger B Richmond JA Kennedy JH McNicol AM E-cadherin and integrin cell adhesion molecule expression in invasive and in situ carcinoma of the cervix Gynecol Oncol 1997 64 481 486 9062155 10.1006/gyno.1996.4578
Sun H Liu X Li M E-cadherin expression and its clinical significance in cervical cancer Zhonghua Zhong Liu Za Zhi 2000 22 496 498 11235574
Imura J Ichikawa K Takeda J Fujimori T Beta-catenin expression as a prognostic indicator in cervical adenocarcinoma Int J Mol Med 2001 8 353 358 11562771
Behrens J Vakaet L Friis R Winterhager E Van Roy F Mareel MM Birchmeier W Loss of epithelial differentiation and gain of invasiveness correlates with tyrosine phosphorylation of the E-cadherin/beta-catenin complex in cells transformed with a temperature-sensitive v-SRC gene J Cell Biol 1993 120 757 766 8425900 10.1083/jcb.120.3.757
Aberle H Schwartz H Kemler R Cadherin-catenin complex: protein interactions and their implications for cadherin function J Cell Biochem 1996 61 514 523 8806074 10.1002/(SICI)1097-4644(19960616)61:4<514::AID-JCB4>3.0.CO;2-R
Shinohara A Yokoyama Y Wan X Takahashi Y Mori Y Takami T Shimokawa K Tamaya T Cytoplasmic/nuclear expression without mutation of exon 3 of the beta-catenin gene is frequent in the development of the neoplasm of the uterine cervix Gynecol Oncol 2001 82 450 455 11520139 10.1006/gyno.2001.6298
Henshall S Tissue microarrays J Mammary Gland Biol Neoplasia 2003 8 347 358 14973378 10.1023/B:JOMG.0000010034.43145.86
Hsu FD Nielsen TO Alkushi A Dupuis B Huntsman D Liu CL van de Rijn M Gilks CB Tissue microarrays are an effective quality assurance tool for diagnostic immunohistochemistry Mod Pathol 2002 15 1374 1380 12481020 10.1097/01.MP.0000039571.02827.CE
Rimm DL Camp RL Charette LA Costa J Olsen DA Reiss M Tissue microarray: a new technology for amplification of tissue resources Cancer J 2001 7 24 31 11269645
van de Rijn M Gilks CB Applications of microarrays to histopathology Histopathology 2004 44 97 108 14764053 10.1111/j.1365-2559.2004.01766.x
DiVito KA Charette LA Rimm DL Camp RL Long-term preservation of antigenicity on tissue microarrays Lab Invest 2004 84 1071 1078 15195116 10.1038/labinvest.3700131
|
15969753
|
PMC1183253
|
CC BY
|
2021-01-04 16:39:04
|
no
|
World J Surg Oncol. 2005 Jun 21; 3:38
|
utf-8
|
World J Surg Oncol
| 2,005 |
10.1186/1477-7819-3-38
|
oa_comm
|
==== Front
CytojournalCytoJournal1742-6413BioMed Central London 1742-6413-2-111598518510.1186/1742-6413-2-11MethodologyDiagnostic utility of p16 immunocytochemistry for Trichomonas in urine cytology Pantanowitz Liron [email protected] Q Jackie [email protected] Robert A [email protected] Christopher N [email protected] Department of Pathology, Baystate Medical Center, Tufts University School of Medicine, Springfield, USA2005 29 6 2005 2 11 11 2 4 2005 29 6 2005 Copyright © 2005 Pantanowitz et al; licensee BioMed Central Ltd.2005Pantanowitz et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
We present a case in which p16 immunocytochemistry helped establish the diagnosis of Trichomonas in urine from a male patient. Based on this finding, we recommend p16 immunocytochemistry as a diagnostic tool for unexpected patients or specimen types in which potential trichomonads are identified following routine cytologic evaluation.
==== Body
Article
We received 50 ml of voided urine from a 84-year-old diabetic male who presented with hematuria. Routine urine cultures yielded no growth after 24 hours. A Papanicolaou-stained ThinPrep™ slide was prepared which revealed benign urothelial cells, neutrophils including "cannonballs" (i.e. neutrophils aggregated around epithelial cells), red blood cells and numerous Trichomonas organisms (Figure 1). The diagnosis of Trichomonas was based upon the presence of a discernible nucleus and cytoplasmic granules that were identified in several of the trichomonads. A visible nucleus and well-defined cytoplasmic granules at 40x magnification are specified as important morphological features required for a confident diagnosis of Trichomonas in liquid-based Pap tests [1]. Although we did not identify flagella in the trichomonads in our case, the finding of flagella, while helpful, is not always required to make a diagnosis of Trichomonas [1]. In fact, the morphologic identification of Trichomonas on liquid-based Pap tests has been shown to be highly accurate [2]. Nevertheless, exfoliated cells including microorganisms in urine are often degenerated, which makes the identification of Trichomonas in these specimens by morphology alone difficult. Therefore, confirmatory testing may be needed. However, traditional methods to detect Trichomonas including culture and wet-mount microscopy, as well as molecular studies, may not always be readily available, particularly on fixed samples received in liquid-based vials for cytologic evaluation.
Figure 1 Group of Trichomonas organisms present in urine (ThinPrep™, Papanicolaou stain).
In order to differentiate parasites from degenerated urothelial cells in our case we prepared additional ThinPrep™ slides from the residual specimen to perform immunocytochemistry for cytokeratin (using a cocktail of high- and low-molecular weight keratins) and p16 (using primary purified mouse anti-human p16 antibody, clone G175-405, supplied by BD Biosciences Pharmingen, San Diego, CA, USA). We included p16 based upon recent published data by our group indicating that Trichomonas vaginalis organisms in cervicovaginal specimens are immunoreactive for p16 [3]. In all ten of these cervicovaginal specimens T. vaginalis were p16 positive and demonstrated strong cytoplasmic staining. Immunocytochemical staining for p16, a proven biomarker for high grade dysplasia associated with Human Papillomavirus (HPV) infection, has been previously applied successfully to cervicovaginal cytology specimens [4]. However, p16 immunoreactivity is not specific for HPV-infected epithelium, as immunoreactivity has previously been documented with inflammatory cells, multinucleated giant cells, bacteria and mucus in cervicovaginal specimens [4-6]. It is unclear if p16 staining of Trichomonas organisms reflects specific immunoreactivity (to unknown epitopes) or may be non-specific. In our case, urothelial and squamous cells were strongly immunoreactive for cytokeratin (Figure 2) but were negative for p16, whereas trichomonads demonstrated strong p16 immunoreactivity (Figure 3 and Figure 4) and failed to react with cytokeratin. Appropriate controls were included in this study (data not shown). Following the diagnosis of Trichomonas, our patient was treated with a course of metronidazole.
Figure 2 Cytokeratin immunocytochemistry. A single urothelial cell demonstrates strong cytokeratin immunoreactivity whereas surrounding trichomonads are negative. Degenerated inflammatory cells are also present in this field.
Figure 3 p16 immunocytochemistry. A group of trichomonads demonstrate strong p16 immunoreactivity whereas an adjacent degenerated urothelial cell and squamous cell are negative.
Figure 4 A single trichomonad, adjacent to an unstained exfoliated squamous epithelial cell, is shown to be p16 immunoreactive.
Trichomonas infection in male patients may be asymptomatic or associated with non-gonococcal urethritis, prostatitis, and urethral strictures. Protozoa in men may be harbored in the uncircumcised prepuce, urethra, seminal vesicles, prostate gland or bladder. As shown in the present case, it is often difficult to establish the diagnosis of Trichomonas in men, particularly when the diagnosis is based solely on the cytological evaluation of poorly preserved organisms in a urine specimen. The diagnosis may be especially problematic when the patient denies sexual contact, or when a sexually transmitted disease in a particular patient, such as an infant [7] or the elderly, seems implausible. Adding to the problem is the fact that when in urine, trichomonads usually assume variable shapes [8]. Parasite variation in size and shape may be further exaggerated in air-dried urine smears [9]. In male patients, the organisms also tend to be somewhat smaller than their counterpart in female patients [10]. In addition to the present patient, our laboratory has diagnosed Trichomonas in Papanicolaou-stained urine specimens in eight men of mean age 59 years (range 47–82 years), over a 15 year period. This finding is in keeping with published data suggesting that higher organism loads occur in older men [11]. We do not know if any of these individuals also had diabetes.
In conclusion, as illustrated in the case presented, we recommend the use of p16 immunocytochemistry to help establish the diagnosis of Trichomonas in unexpected patients or specimen types in which potential trichomonads are identified following routine cytologic evaluation.
==== Refs
Aslan DL McKeon DM Stelow EB Gulbahce HE Kjeldahl K Pambuccian SE The diagnosis of trichomonas vaginalis in liquid-based Pap tests: Morphological characteristics Diagn Cytopathol 2005 32 253 259 15830365 10.1002/dc.20231
Aslan DL Gulbahce HE Stelow EB Setty S Brown CA McGlennen RC Pambuccian SE The diagnosis of Trichomonas vaginalis in liquid-based Pap tests: Correlation with PCR Diagn Cytopathol 2005 32 341 344 15880709 10.1002/dc.20256
Pantanowitz L Florence RR Goulart RA Otis CN Trichomonas vaginalis p16 immunoreactivity in cervicovaginal pap tests: a diagnostic pitfall Modern Pathology 2005 18 73A 74A 10.1038/modpathol.3800441
Bibbo M Klump WJ DeCecco J Kovatich AJ Procedure for immunocytochemical detection of P16INK4A antigen in thin-layer, liquid-based specimens Acta Cytol 2002 46 25 29 11843554
Saqi A Pasha TL McGrath CM Yu GH Zhang P Gupta P Overexpression of p16INK4A in liquid-based specimens (SurePath™) as marker of cervical dysplasia and neoplasia Diagn Cytopathol 2002 27 365 370 12451568 10.1002/dc.10205
Sahebali S Depuydt CE Segers K Moeneclaey LM Vereecken AJ Van Marck E Bogers JJ P16INK4a as an adjunct marker in liquid-based cervical cytology Int J Cancer 2004 108 871 876 14712490 10.1002/ijc.11589
Schares T Machtinger S D'Harlingue AE Maloney JR Trichomonas vaginalis urinary tract infection in an infant Pediatr Infect Dis 1982 1 340 341 6984181
Malyszko E Januszko T Detection of Trichomonas vaginalis in men [Article in Polish] Pol Tyg Lek 1991 46 997 999 1669192
Loo CK Gune S Trichomonas vaginalis in urine cytology Acta Cytol 2000 44 484 485 10834019
Summers JL Ford ML The Papanicolaou smear as a diagnostic tool in male trichomoniasis J Urol 1972 107 840 842 4537135
Wendel KA Erbelding EJ Gaydos CA Rompalo AM Use of urine polymerase chain reaction to define the prevalence and clinical presentation of Trichomonas vaginalis in men attending an STD clinic Sex Transm Infect 2003 79 151 153 12690140 10.1136/sti.79.2.151
|
15985185
|
PMC1183263
|
CC BY
|
2021-01-04 16:36:22
|
no
|
Cytojournal. 2005 Jun 29; 2:11
|
utf-8
|
Cytojournal
| 2,005 |
10.1186/1742-6413-2-11
|
oa_comm
|
==== Front
CytojournalCytoJournal1742-6413BioMed Central London 1742-6413-2-91596702310.1186/1742-6413-2-9Case ReportFNAB cytology of extra-cranial metastasis of glioblastoma multiforme may resemble a lung primary: A diagnostic pitfall Chivukula Mamatha [email protected] HE [email protected] Julie A [email protected] Hendrikus G [email protected] Grant [email protected] Vinod [email protected] Departments of Pathology, Medical College of Wisconsin, Milwaukee, WI, USA2 Pulmonary and Critical Care Medicine, Medical College of Wisconsin, Milwaukee, WI, USA3 Neurology and Neurosurgery; Medical College of Wisconsin, Milwaukee, WI, USA2005 20 6 2005 2 9 9 8 3 2005 20 6 2005 Copyright © 2005 Chivukula et al; licensee BioMed Central Ltd.2005Chivukula et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
As extra-cranial metastasis of glioblastoma multiforme (GBM) is rare, it may create a diagnostic dilemma especially during interpretation of fine needle aspiration biopsy (FNAB) cytology.
Case presentation
We present transbronchial FNAB findings in a 62-year-old smoker with lung mass clinically suspicious for a lung primary. The smears of transbronchial FNAB showed groups of cells with ill-defined cell margins and cytological features overlapping with poorly differentiated non-small cell carcinoma. The tumor cells demonstrated lack of immunoreactivity for cytokeratin, thyroid transcription factor-1, and usual neuroendocrine markers, synaptophysin and chromogranin in formalin-fixed cellblock sections. However, they were immunoreactive for the other neuroendocrine immunomarker, CD56, suggesting neural nature of the cells. Further scrutiny of clinical details revealed a history of GBM, 13 months status-post surgical excision with radiation therapy and systemic chemotherapy. The tumor recurred 7 months earlier and was debulked surgically and with intra-cranial chemotherapy. Additional evaluation of tumor cells for glial fibrillary acidic protein (GFAP) immunoreactivity with clinical details resulted in final interpretation of metastatic GBM.
Conclusion
Lack of clinical history and immunophenotyping may lead to a diagnostic pitfall with possible misinterpretation of metastatic GBM as poorly differentiated non-small cell carcinoma of lung in a smoker.
Glioblastoma multiformeFine needle aspiration biopsy cytologyFNAlung tumor
==== Body
Background
Glioblastoma multiforme (GBM) represents extreme anaplasia in astrocytic tumors. Similar to other CNS tumors, the extra-cranial metastasis of GBM is rare and is usually observed after the tumor has infiltrated the dural veins, the cranium, the extra-cranial soft tissue, or more frequently after the tumor debulking surgery for a recurrent tumor [4].
A literature search revealed only a few case reports of extra-cranial metastasis of primary GBM to various organs such as spleen [5], skin [8], heart [7], bone [6], cervical lymph nodes [9,11] and lung [10,12-14]. In all these cases the metastases occurred after the resection of primary intra-cranial tumor with an average time interval of 10 months. Only one case of a spontaneous metastasis of primary GBM to lungs is reported [13]. The diagnosis in most cases was based on biopsy and immunophenotyping with glial fibrillary acidic protein (GFAP).
Since the metastasis of GBM is rare, cytopathological interpretation for its differentiation from other tumors may be challenging. Correct interpretation for origin of the tumor is important in rendering a proper clinical management. We report a case in which optimum clinical history and immunohistochemical evaluation played a crucial role in preventing a diagnostic pitfall of misinterpreting FNAB cytology of lung mass in a smoker.
Case presentation
A 62-year-old male presented with shortness of breath for one week. He had a smoking history of 10 pack-years [1 pack per day, per year, for 10 years]. The X-ray chest revealed a large pleural effusion with suggestion of a mass in the right upper lobe of lung (RUL) with massive hilar lymphadenopathy. The left lung parenchyma was normal and the pulmonary vasculature was not congested. A computerized tomography scan of the chest confirmed a mass in RUL.
Clinically, the tumor was considered as lung primary and pulmonary consult was requested. Bronchoscopy of RUL revealed an irregular friable mucosa. A transbronchial FNAB of the RUL mass was performed with other evaluations.
The patient had previous history of GBM in the right occipital area 13 months ago for which he underwent initial surgical excision with radiation therapy and systemic chemotherapy. The tumor recurred after 6 months for which he underwent extensive debulking and placement of gliadel wafers (BNCU-impregnated wafers).
Methods
Cytomorphological features on adequacy evaluation of transbronchial FNAB smears of the RUL mass were consistent with poorly differentiated non-small cell tumor. Additional material for cellblock was recommended for elective immunophenotyping of tumor cells.
Other evaluations included an endobronchial biopsy from RUL, bronchoalveolar lavage of anterior segment of RUL for cytology, and thoracocentesis of right pleural effusion for cytology.
Results
Cytopathological findings of transbronchial FNAB of the mass
The smears of the fine needle aspirates were relatively cellular with cohesive groups of tumor cells showing mild pleomorphism. The cells had cyanophilic scant cytoplasm with indistinct cell borders (Figure-1). The hyperchromatic nuclei showed powdery to clear chromatin with small but distinct nucleoli. Initial cytological differential diagnosis included poorly differentiated non-small cell carcinoma, paraganglioma, lymphoma, melanoma, and non-neoplastic lesion such as granuloma with addition of metastatic GBM after considering the history of surgical intervention for recurrent GBM.
Figure 1 Cohesive groups of tumor cells showed mild pleomorphism. The cells had cyanophilic scant cytoplasm with indistinct cell borders with morphological features overlapping with poorly differentiated carcinoma.
As the smears were relatively cellular with significant proportion of tumor cell groups showing hyperchromasia and pleomorphism, granuloma could be excluded even with cytomorphological evaluation. However, a few groups without significant pleomorphism and without admixture with large tumor cells demonstrated overlapping morphological features with granuloma (Figure-2). In these groups, the cells showed ill-defined cytoplasm with slightly elongated, overlapping, oval nuclei. Occasional lymphocytes mixed with these cell groups were present, but multinucleated giant cells were absent.
Figure 2 A few groups (a through d) did not show pleomorphism but showed occasional lymphocyte (arrowheads in e & f) admixed with the tumor cells demonstrating ill-defined cell borders superficially resembling a granuloma. However, pleomorphic cells (arrows in e & f) were not uncommon in this relatively cellular specimen.
Other malignant lesions in the differential diagnosis, except lymphoma, were relatively difficult to exclude by morphology alone. The cohesive tumor cells with clear to dusty chromatin and significant pleomorphism did not favor lymphoma. The nucleoli were relatively less prominent and did not favor melanoma and large cell carcinoma of lung. Paragaglioma could not be excluded morphologically except lack of Zellballen pattern (difficult to be evaluated cytologically) in the cellblock sections.
Immunophenotyping of tumor cells was performed on formalin-fixed paraffin-embedded cellblock sections (Table 1).
Table 1 Immnophenotype of tumor cells.
S.No. Immunomarker Antibody (Clone, source, dilution) Antigen retrieval (Pre-treatment of sections) Immunoreactivity pattern of tumor cells
1 Cytokeratin Dako, 34BE12, 1:300 for 30 mts Proteinase,-K for 4 mts Non-immunoreactive
2 Cytokeratin 7 Dako, OVTL12/30, 1:100 for 30 mts PH 6.0 at 98 C for 35 mts Non-immunoreactive
3 Cytokeratin 20 Dako, Ks20.8, 1:300 for 30 mts Proteinase, -K for 4 mts Non-immunoreactive
4 TTF-1 Dako, 8G7G3/1, 1:150 for 20 mt Heat DAKO, PH 6.0 for 35 mts Non-immunoreactive
5 LCA Dako, PD7/26/16 and 2B11, 1:200 for 30 mts Proteinase, -K for 4 mts Non-immunoreactive
6 MART-1 SIGNET, M2-7C10 1:600 for 30 mts Heat, PH6.0 for 35 mts, 1:500 for 35 mts Non-immunoreactive
7 Chromogranin Novocastra, LK2H10, 1:600 for 30 mts HeatPH 6.0 for 1:600 for 35 mts Non-immunoreactive
8 Synaptophysin Dako, 1:100 for 30 mts HeatPH 6.0 for 35 mts, 1:200 for 30 mts Non-immunoreactive
9 CD56 Novocastra, 186, 1:100 for 45 mts Heat DAKO, PH 6.0 for 30 mts, cool 20 mts Immunoreactive
The tumor cells were non-immunoreactive for cytokeratin 7, cytokeratin 20, cocktail of cytokeratin AE1/AE3 – Cam 5.2, thyroid transcription factor (TTF-1), leucocyte common antigen (LCA, CD45), and MART-1. They were also non-immunoreactive for neuroendocrine immunomarkers, chromogranin and synaptophysin. Non-immunoreactivity for cytokeratin (Figure-3b) ruled out poorly differentiated carcinoma including lung primary, LCA non-immunoreactivity (Figure-3c) ruled out lymphoma and granuloma, nonimmunoreactivity for chromogranin and synaptophysin ruled out paraganglioma, and MART-1 non-immunoreactivity (Figure-3d) ruled out melanoma.
Figure 3 Representative immunoprofile of tumor cells on cellblock sections (see also table 1) a. H&E; b. Cytokeratin (non-immunoreactive); c. LCA (non-immunoreactive); d. MART-1 (non-immunoreactive); e. CD56 (immunoreactive); f. GFAP (immunoreactive).
However, the tumor cells were immunoreactive for CD56 (Figure-3e), consistent with neural differentiation. Based on the past clinical history of recent surgical intervention for recurrent GBM, demonstration of immunoreactivity for GFAP (Figure-3f) confirmed the diagnosis of metastatic GBM. Previously resected GBM demonstrated comparable morphology and immunoprofile (Figure-4)
Figure 4 Histomorphology and GFAP immunoreactivity of previously resected GBM.
Endobronchial biopsy
Endobronchial biopsy was negative for tumor.
Cytopathology of other material
The bronchoalveolar lavage and pleural fluid cytology were negative for malignant cells.
Discussion
GBM is by far the most common high-grade glioma representing anaplastic spectrum of fibrillary and diffuse cytoplasmic astrocytoma [2]. It shows marked cellularity, pleomorphism, many mitotic figures, necrosis, and vascular proliferation [2]. The tumor presents usually in the cerebral hemispheres in adults and in brain stem in children. Despite the external-beam radiotherapy and surgery, the median survival for GBM is relatively short (approximately 12 months) [2].
A very few reports describing the cytological features of GBM are found in the literature [9,11,19]. Of the CNS tumors, glioblastoma multiforme is a relatively easy to recognize cytologically as malignant. A review article described cytopathology of various CNS tumors including 15 cases of GBM [19].
However when GBM is sampled from the extra-cranial sites such as lung or lymph node, a differential diagnosis from other tumors may be challenging. Cytologically the smears of GBM show large, spindle to oval, hyperchromatic and pleomorphic nuclei. A fibrillary or necrotic background is also noticed in some of these tumor cells of GBM may be recognizable as astrocytic [1]. But due to the marked pleomorphism of cohesive groups of tumor cells with frequent mitoses, the recognition of GBM from other poorly differentiated tumors may be challenging.
The cell borders of the cohesive tumor cells in the present case were predominantly indistinct with differential diagnosis of poorly differentiated carcinoma, melanoma, paraganglioma, sarcoma, and even lymphoma may be challenging. In the smears with scant cellularity with absence of significant pleomorphism, some groups may resemble granuloma. The smears showing cytoplasmic type of astrocytic tumor cells with well-defined cytoplasm with distinct cell margins may particularly resemble poorly differentiated carcinoma, lymphoma, and melanoma [1].
The tumor cells of GBM are immunoreactive for GFAP and Vimentin [2]. Vimentin immunoreactivity is non-diagnostic due to significant overlap with other lesions in the differential diagnosis of poorly differentiated tumors. Immunoreactivity of tumor cells for GFAP confirms glial differentiation. Generally the astrocytic type of cells shows immunoreactivity for GFAP, but small-undifferentiated cells are often weakly immunoreactive or non-immunoreactive [2]. It is unusual for the gliomas to be positive for epithelial markers such as cytokeratin, thus differentiating them from carcinomas. Although rare, 4% of GBMs may be immunoreactive for cytokeratin, but this is usually weak and focal [2].
Gliosarcoma is other rare subtype of GBM, which shows a characteristic biphasic appearance, consisting of GBM component admixed with sarcomatous elements [17,18]. Extra-cranial metastases of gliosarcoma with sarcomatoid component have been reported in 17 cases (mostly imprint smears with one case as fine needle aspiration biopsy cytology; 7 cases in lung, 6 cases in liver, 2 cases in lymph nodes, 1 in adrenal gland, and 1 in vertebral body) [18].
Although rare, extra-cranial metastasis of GBM may be observed after craniotomy and trephination procedures for initial GBM or more frequently after tumor debulking surgery for a recurrent tumor. A case of metastatic seeding along the needle biopsy tract of a GBM has been reported [16]. Such metastases are attributed to the infiltration of tumor cells into surrounding tissue and extra cranial blood vessels.
Presentation of extracranial metastasis of GBM as a single mass in lung may suggest a differential diagnosis of a poorly differentiated carcinoma [9]. Small cell carcinoma cytologically shows cohesive groups of hyperchromatic, usually small cells, with scant cytoplasm without significant pleomorphism. Rare case reports of a metastatic glioblastoma to cervical lymph node resembling small cell carcinoma have been described in literature [9,11]. In the present case, the small cell carcinoma was not a significant differential diagnosis, as it demonstrated pleomorphism with dusty chromatin and small but visible nucleoli with possibility of poorly differentiated non-small cell carcinoma. Mitoses and apoptotic cells are frequent in small cell carcinoma with many fragile nuclei disrupted as strings of DNA material in the background. The nuclei may be round to oval with 'salt and pepper' chromatin without perceptible nucleoli. Nuclear molding with zig-saw puzzle-like accommodation of nuclear shapes with adjacent nuclei may be present due to scant cytoplasm and delicate nuclear membranes. The tumor cells in this case were non-immunoreactive for cytokeratins and neuroendocrine markers such as chromogranin and synaptophysin, further ruling out the possibility of poorly differentiated carcinoma of lung, both small cell and non-small cell type.
The collection of epitheliod histiocytes in a granuloma show elongated oval carrot shaped nuclei, which usually are not hyperchromatic. The indistinct cell borders impart syncitial appearance to the group. These epitheliod cells are usually admixed with small lymphocytes and may be associated with multinucleated giant cells. In the present case, a few groups of cells did not show pleomorphism but showed occasional lymphocyte admixed with the tumor cells demonstrating ill-defined cell borders superficially resembling a granuloma (Figure-2). In the present case, the granuloma could be ruled out even morphologically, as the smears were relatively cellular with significant proportion of tumor cells showing hyperchromatic and pleomorphic nuclei. In addition, the cells were non-immunoreactive for LCA, which is usually immunoexpressed in epitheloid cells of granuloma.
A paraganglioma shows tumor cells with ill-defined cell borders and nuclei with small nearly visible red nucleoli. The tumor cells may show pleomorphism. The cytoplasm is wispy and finely granular with frayed cell margins. Presence of bare nuclei is frequent [3]. Zellballen arrangement may be observed in cellblock sections. Although, some of the cytomorphological features overlapped, the tumor cells in this case were non-immunoreactive for neuroendocrine markers such as chromogranin and synaptophysin in cellblock sections without Zellballen arrangement excluding paraganglioma.
Individual morphological features of melanoma including prominence of nucleoli, nuclear pseudoinclusions, cytoplasmic melanin pigment, and frequent binucleation with 'demon-eye' nuclei were absent in the present tumor. As melanoma is a great mimicker with diverse morphological patterns, it was ruled out with non-immunoreactivity of these tumor cells for melanoma marker such as, MART-1.
The cells in high-grade large cell lymphomas may show heterogenous population of medium to large pleomorphic non-cohesive cells with clumped nuclear chromatin. The cell borders of singly scattered lymphoma cells are usually distinct [3]. The tumor cells in the present case showed powdery to clear chromatin with cohesive groups of cells without distinct cell borders. They were non-immunoreactive for LCA further ruling out lymphoma.
Conclusion
Due to the rarity of event, in the absence of proper clinical correlation, extracranial metastases of GBM may be misdiagnosed. In the present case, although the cytomorphological features resembled poorly differentiated non-small cell carcinoma of lung, they were not convincing enough to suggest a specific diagnosis. Further evaluation, including clinical history of previously resected GBM with recurrence and immunophenotyping with appropriate immuno panel, avoided the pitfall leading to proper interpretation of the lung mass in this case as metastatic GBM.
List of abbreviations
BNCU, Bi chloroethylnitrosourea, (Chemotherapy drug); CNS, central nervous system; FNAB, Fine needle aspiration biopsy; GBM, Glioblastoma multiforme; GFAP, glial fibrillary acidic protein; LCA, leucocyte common antigen; RUL, right upper lobe of lung.
Authors' contributions
1. (MC) Cytology fellow of the case, collected all the data, participated in cytological evaluation and drafting of manuscript.
2. (HD) Pulmonary fellow of the case and reviewed the manuscript.
3. (JB) Pulmonologist of the case and reviewed the manuscript.
4. (HK) Neuro-oncologist of the case and reviewed the manuscript.
5. (GS) Neuro-surgeon of the case and reviewed the manuscript.
6. (VS) Cytologist of the case, conceptual organization and manuscript writing.
Consent
This case report is published after consent by the spouse of the deceased patient.
Acknowledgements
CytoJournal thanks Yener Erozan, M.D. ([email protected]), Professor, Dept. of Pathology, The Johns Hopkins Hospital, 406 Pathology Building 2: 600 North Wolfe Street, Baltimore, MD21287-6940 for organizing and completing the peer-review process for this manuscript as its Academic Editor."
==== Refs
Aspiration cytology, the art and science of Cytopathology, RichardM Demay
Atlas of Tumor pathology, Tumors of the Central nervous system, P. Burger, B. Scheithauer, AFIP, third series, AFIP fascicle 10
Atkinson Barbara Atlas of Diagnostic cytopathology 2 Saunders
Terheggen HG Muller W Extra cerebrospinal metastasis in glioblastom a. Case Report and review of literature Eur J Pediatrics 124 155 64 1977 jan 26 10.1007/BF00477550
Tamiya YT Meguro T Ichikawa T Sato Y Date I Nakashima H Ohomoto T Glioblastoma with metastasis to spleen – case report, NeurologicaMedico-Chirurgica 2003 43 452 6
Sadik AR Port R Garfinkel B Bravo J Extracranial metastasis of cerebral Glioblastoma multiforme: Case report, Neurosurgery 1984 15 549 51
Hata N Katsuta T Inoue T Arikawa K Yano T Takeshita M Iwaki T Extra-cranial metastasis of glioblastoma to lung and heart with a histological resemblence to small cell carcinoma of the lung: an autopsy case: No Shinkei Geka-neurological surgery 2001 29 438 8
Figuera P Lupton JR Remington T Olding M Jones RV Sekhar LN Sulica VI Cutaneous metastasis from an intracranial glioblastoma multiforme J Am Acad Dermatol 2002 46 297 300 11807444 10.1067/mjd.2002.104966
Campora RG Salverri CO Ramirez FV Villadiego MS Davidson HG Metastatic Glioblastoma multiforme in cervical lymph nodes Report of a case with diagnosis by fine needle aspiration, Acta cytologica 1993 7 938 42
Granjon O Lange F Aurtheir FJ Lebragy F Pulmonary metastases of glioblastoma, revue des maladies respiratoires 1995 12 489 91
Al-Rikabi AC Al-Sohaibani MO Jamjoom A Al-rayess MM Metastatic deposits of a high-grade malignant glioma in cervical lymphnode diagnosed by fine-needle aspiration cytology-case report and literature review Cytopathology 1997 8 421 7 9439895
Laraqui L Amarti A Zouaidia F Maher M Kettani F Saidi A Pulmonary metastasis from a gliblastoma A case report (Article in French), Rev Pneumol Clin, 2001 Jin 2001 57 225 8
Chung YH Wong SL Huang HY Endobronchial metastasis of glioblastoma multiforme diagnosed by fiber-optic bronchoscopic biopsy, Journal of Formosan medical Association 1999 98 133 5
Greif J Horovitz M Marmor S Pleuropulmonary metastasis from an intracranial glioblastoma Lung cancer 1998 20 135 7 9711532 10.1016/S0169-5002(98)00029-4
Vural G Hagmar B Walaas L Extra-cranial metastasis of glioblastoma multiforme diagnosed by fine-needle aspiration: a report of 2 cases and a review of literature, Diagnostic cytopathology 1996 15 60 5
Steinmetz P Barnett GH Chidel MA Suh JH Metastatic seeding of the stereotactic biopsy tract in glioblastoma multiforme: a case report and review of the literature J neuro onco 2001 55 167 71 10.1023/A:1013873431159
Parwani A Berman D Burger P Ali S Gliosarcoma: Cytopathologic characteristics on Fine-needle aspiration (FNA) and Intraoperative Touch imprint Diagnostic Cytopathology 2004 30 77 81 14755755 10.1002/dc.10368
Yokoyama H Ono H Mori K kishikawa M kihara M Extracranial metastasis of Glioblastoma with sarcomatous component, Surgical Neurology 1985 24 641 5
Liwnicz B Henderson K Masukawa T Smith R Needle aspiration cytology of intracranial lesions, Acta cytologica 1982 26 779 86
|
15967023
|
PMC1183264
|
CC BY
|
2021-01-04 16:36:22
|
no
|
Cytojournal. 2005 Jun 20; 2:9
|
utf-8
|
Cytojournal
| 2,005 |
10.1186/1742-6413-2-9
|
oa_comm
|
==== Front
Saline Syst
Saline Syst
Saline Systems
1746-1448
BioMed Central
1746-1448-1-4
16176591
10.1186/1746-1448-1-4
Research
Effect of salinity stress on the life history variables of Branchipus schaefferi Fisher, 1834 (Crustacea: Anostraca)
Sarma SSS [email protected]
Beladjal Lynda [email protected]
Nandini S [email protected]
Cerón-Martínez Gerardo [email protected]
Tavera-Briseño Karina [email protected]
1 Laboratory of Aquatic Zoology, UMF, Division of Research and Postgraduate Studies, National Autonomous University of Mexico, Campus Iztacala, Av. de los Barrios S/N, Los Reyes, AP 314, CP 54090, Tlalnepantla, State of Mexico, Mexico
2 Laboratory of Animal Ecology, Ghent University, Ledeganckstraat 35, B-9000 Gent, Belgium
3 UIICSE, Division of Research and Postgraduate Studies, National Autonomous University of Mexico, Campus Iztacala, Av. de los Barrios S/N, Los Reyes, AP 314, CP 54090, Tlalnepantla, State of Mexico, Mexico
2005
4 7 2005
1 44
26 4 2005
4 7 2005
Copyright © 2005 Sarma et al; licensee BioMed Central Ltd.
2005
Sarma et al; licensee BioMed Central Ltd.
https://creativecommons.org/licenses/by/2.0/ This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Freshwater anostracans inhabit ephemeral water bodies in which as the water level decreases due to evaporation the salt concentration increases. Thus, for most anostracans salinity becomes the major stress factor.
Results
We tested five concentrations of NaCl (0 to 8 g/l) on the life table demography of Branchipus schaefferi fed Chlorella (alga). Age-specific survivorship curves of male and female B. schaefferi showed nearly a similar pattern in that increased salt concentration resulted in decreased survivorship. The age-specific reproduction (mx) of females showed several peaks of cyst production at 0 and 1 g/l salinity while in treatments containing salt at 4 or 8 g/l, there were fewer peaks. Average lifespan, life expectancy at birth, gross and net reproductive rates, generation time and the rate of population increase were all significantly influenced by the salt concentration in the medium. The highest value of net reproductive rate (970 cysts/female) was in treatments containing 0 g/l of salt, while the lowest was 13 cysts/female at 8 g/l. The rate of population increase (r) varied from 0.52 to 0.32 per day depending on the salt concentration in the medium.
Conclusion
The low survival and offspring production of B. schaefferi at higher salinity levels suggests that this species is unlikely to colonize inland saline water bodies. Therefore, the temporary ponds in which it is found, proper conservative measures must be taken to protect this species.
==== Body
pmcBackground
Freshwater anostracans usually inhabit ephemeral waterbodies, which periodically dry, especially during summer months. This condition forces them to adapt to a) mostly one population cycle, b) produce cysts that resist desiccation and c) survive under changing ionic composition of the ambient medium [1]. The ionic composition of freshwater bodies is controlled by many factors. Among abiotic factors, temperature, through evaporation, plays an important role. Usually as the water level decreases due to evaporation, salt concentration increases. Therefore, both flora and fauna of such waterbodies must show some degree of salt tolerance in order to survive [2]. Knowledge on the pond salinity effects on the osmoregulation and conformation of anostracans is much limited. For example, Branchinecta gigas Lynch, 1937 and B. mackini Dexter, 1956 from saline lakes in Central Washington revealed that both these species have the capacity of hyperosmoregulation during low saline conditions and osmoconformation during high salinity conditions [3]. On the other hand, Branchinella compacta Linder, 1941 occurs in low saline conditions and has less tolerance to high salinity [4]. Thus, it remains unknown the tolerance capacities of many anostracan species.
Freshwater anostracans can survive during the drying period, yet their tolerance to increased salt levels is not well documented [1]. Mere survival under a given natural stress is not adequate for the continuation of a species; the reproductive output too is important. Thus, the reproductive strategies of freshwater anostracans reveal the vicissitudes of their habitat [5]. For most freshwater organisms including crustaceans, salinity is a major stress factor [6]. Increase in salinity leads to reduced survival, reduced reproductive output or both. For a given freshwater species, at salt concentrations below the median tolerance limits, reproduction is more drastically affected than survival or other variables such as swimming speed or the feeding rate [7].
The life history characteristics of crustaceans, in general, are best understood using life table demographic studies [8]. The variables more sensitive to stress are a) average lifespan, b) life expectancy at hatching, c) gross reproductive rate and d) net reproductive rates, e) generation time and f) the rate of population increase. However, not all variables are consistently sensitive to the same stress. Usually, the rate of population increase is thought to be more sensitive than the rest of the life history variables since it integrates both the mortality and natality [9]. Similar trends may be found in freshwater anostracans to salinity, but quantitative data are lacking for many genera except Artemia [10].
The aim of the present work was to evaluate the effect of different concentrations of sodium chloride on the life table demography of the freshwater anostracan Branchipus schaefferi Fisher, 1834.
Results
Survivorship curves
Age-specific survivorship curves (Fig. 1) of female (A) and male (B) B. schaefferi showed a nearly similar pattern in relation to salt concentration (from 0 g to 8 g/l). At the highest salt concentration (8 g/l), females continued to live without mortality for the first week. The age-specific reproduction (mx) (Fig. 2) of females showed several peaks of cyst production at 0 and 1 g/l salinity while in treatments containing 4 or 8 g/l, there were fewer peaks.
Figure 1 Age-specific surviorship (lx) curves of B. schaefferi subjected to different concentrations of NaCl. Column A: female; Column B: male. Values represent mean ± standard error based on 3 cohorts.
Figure 2 Age-specific fecundity (mx) curves of female B. schaefferi subjected to different concentrations of NaCl. Values represent mean ± standard error based on 3 cohorts.
Life history variables
All the selected life history variables of B. schaefferi were significantly affected at even the lowest salinity increase (Table 1). Life expectancy at birth of males in 0 g/l salinity was slightly higher than females. However, in treatments containing a certain concentration of NaCl, this was nearly similar for both sexes. Also, regardless of sex, increased concentration of salt in the medium decreased the life expectancy. At the highest salt concentration, the life expectancy was nearly 1/6th of the controls (0 g/l). Gross and net reproductive rates also decreased with increasing salt level in the medium. The highest net reproductive rate (970 cysts/female) was in treatments containing 0 g/l salt, while the lowest (13 cysts/female) at 8 g/l.
Table 1 Selected life history variables of Branchipus schaefferi exposed to different concentrations of salt. For a given variable, treatments carrying same alphabet are not statistically significant (p > 0.05, Tukey's test)
Salt conc. (g/l) Life Expectancy at birth (days) Gross Reproductive rate (cysts / female) Net reproductive rate (cysts / female) Rate of pop. increase (r)
Female
0 49.5 ± 1. 9a 990.8 ± 47.0a 969.7 ± 37.4a 0.515 ± 0.004a
1 24.8 ± 0.8b 412.3 ± 14.6b 395.0 ± 20.9b 0.542 ± 0.002b
2 14.2 ± 0.2c 63.5 ± 6.9c 62.0 ± 8.1c 0.480 ± 0.003c
4 12.2 ± 0.3c,d 84.0 ± 24.1c,d 84.0 ± 24.1c 0.490 ± 0.02c
8 8.5 ± 0.1d 13.0 ± 1.8d 13.0 ± 1.8c 0.318 ± 0.01c
Male
0 53.8 ± 1.7a
1 26.8 ± 1.4b
2 18.5 ± 1.5c
4 14.5 ± 0.2d
8 8.3 ± 0.2e
The rate of population increase (r) varied from 0.52 to 0.32 depending on the salt concentration in the medium. As in the case of other survivorship variables, r decreased with an increase in NaCl in the medium. Generation time varied from 8 to 25 days, depending on the salt concentration. There was a positive relation between average lifespan and the generation time of B. schaefferi (Fig. 3).
Figure 3 Relation between average lifespan and generation time in B. schaefferi under different concentration of NaCl. Plotted are the replicated data for each treatment.
Statistically, all the tested variables (survivorship variables: average lifespan, life expectancy at birth, reproductive variables: gross and net reproductive rates, generation time and the rate of population increase) were significantly influenced by the salt concentration in the medium (p < 0.001, one-way ANOVA). Tukey's tests revealed that all the above variables at 0 g/l salinity were significantly different (p < 0.05) from the treatments containing some quantity of NaCl. However, reproductive variables at 2, 4 and 8 g/l were not significantly different (p > 0.05).
Discussion
Anostracans are particularly well suited for studying the impact of salt concentrations on the life history characteristics because they live in aquatic environments with a continually changing ionic composition [11]. Consequently, within its lifespan, an individual experiences different salinities, which in turn may affect the survival and reproductive performance [5].
Many species of freshwater anostracans feed on green algae [1]. Branchipus schaefferi was earlier cultured on the green agla Scenedesmus [12], which suggests that the use of algal diet was adequate for both survival and reproduction. This was evident in our study too, where in the absence of salt stress, B. schaefferi was able to survive and reproduce cysts.
Temporary waterbodies are characterized by opportunitistic species including anostracans. High reproductive output, short lifespan and wide diet breadth are some of the characteristics of crustacean species inhabiting temporary waters [2,5]. Many species of freshwater anostracans have a lifespan varying from about two weeks to several months [1]. In the present study, regardless of salt concentration, the maximum lifespan varied from 10 to 55 days. Compared to a previous study on the same species [13], the maximum lifespan in this study was slightly shorter (55 vs 77 days). However, as documented earlier, both males and females had nearly the same lifespan [13]. Depending on the environmental conditions, the age-specific life survivorship curves of anostracans may be type I (low initial mortality), type II (mortality rate independent of the age) or type III (heavy initial mortality) [14]. Usually under stressful conditions, the age-specific survivorship curves tend to be type III [15]. Increase in salt concentration in this study was a stress for B. schaefferi and therefore, with increasing NaCl level in the medium, there was steep fall in the survivorship in the initial age groups. Salt levels as low as 1 g/l caused about 80% mortality within 30 days for both males and females. At 8 g/l, both males and females lived for about a week without mortality and thereafter the survival rapidly declined.
Variable (saw tooth-like) cyst production is characteristic of crustaceans including anostracans that live more than 2 months [14,16], where offspring production is pulsed, i.e., after 5–10 day intervals neonate production peaks and during the interim period fewer eggs are produced. In B. schaefferi too, in treatments containing no salt or 1 g/l level, the offspring production was pulsed. With increased salinity in the medium fewer offspring were produced suggesting that salt levels beyond 2 g/l are highly stressful for B. schaefferi. The maximum cysts per brood was about 70. Reproduction was extremely low at 8 g/l. The number of cysts per brood for Streptocephalus may be as high as 900; however under inappropriate conditions this number may be as low as 1 cyst/brood [17]. In our study, the peak cyst production per brood was <20 at the highest salinity (8 g/l). When cultured on a diet of Scenedesmus at a density of 1 × 106 cells/ml, the maximum number of cysts per brood of B. schaefferi was 192 [13]. In the present study, this was much lower. This was probably due to the food density used. Based on dry weight [18], the quantity of algal diet used by Beladjal et al. [13] was nearly twice that used in this study. There is abundant evidence that increase in algal density increases the cyst production in anostracan species [1]. Generation time and rate of population increase observed for B. schaefferi are in broad agreement with those reported for freshwater anostracans [14].
Interrelationships exist among different life history variables of organisms. For example, body size and clutch size relation in crustaceans is generally positive and linearly or curvilinearly related [19]. Similarly generation time and lifespan are linearly related for many species of zooplankton such as rotifers [20], cladocerans [19], a few anostracans (e.g., Streptocephalus mackini Moore, 1966) [14] and now also in B. schaefferi.
The low tolerance capacity of B. schaefferi to NaCl as observed in the present study is also confirmed from field observations. For example, Maier et al. [11] have reported the occurrence of B. schaefferi in man-made freshwater bodies in Germany. However, with an increase in conductivity (higher than 300 μS/cm which is equivalent to <0.5 g/l), B. schaefferi was almost eliminated. There are also some differences with reference to the ionic composition of water in naturally drying ponds and NaCl used here [6]. However, species tolerant to other salts are also tolerant to NaCl or vice versa [21]. For example, Branchinecta sandiegonensis Fugate, 1993 and Streptocephalus woottoni Eng, Belk and Eriksen, 1990 which are generally found in dilute coastal vernal pools, are strong hyperregulators when external Na+ levels are below 60 mmol-1 or under conditions of alkalinities up to 0.8–1.0 g l-1 [22]. Among anostracans, Artemia spp. are both hypo-and hyperosmotic regulators. For example, Artemia franciscana Kellog, 1906 showed little variations in haemolymph ion concentrations when the external salinity (as NaCl) was < 0.3 g l-1 to supersaturated levels. However, this capacity has been reduced when exposed to low pH conditions [23]. Therefore it appears unlikely that B. schaefferi in natural ponds tolerates salinity levels higher than 8 g/l, especially under low pH conditions.
Conclusion
B. schaefferi which inhabits temporary freshwater ponds tolerated only low NaCl levels. When raised on Chlorella at a density of 1 × 106 cells/ml under 0 g /l of NaCl concentration, both males and females had similar lifespan. However, when the salinity of the medium was increased from 0 to 8 g/l, both survival and reproduction were decreased. NaCl as low as 1 g/l caused negative influence on the life expectancy at birth, gross and net reproductive rates as well as the rate of population increase. Thus the low tolerance to salinity of this species suggests that it is unlikely to colonize inland saline waterbodies. Even in those freshwater ponds where it is found, anthropogenic factors leading to elevated salinity may eventually dislodge this species from its natural habitat [24]. In Germany B. schaefferi is considered as one of the most endangered crustacean species. Therefore, the temporary ponds in which it is found, proper conservative measures must be taken to protect this species.
Methods
Culture of test species
The original parental stock of Branchipus schaefferi was obtained from Boughzoul, Algeria and mass cultured using Scenedesmus. Cyst collection was done according to methods described previously [13]. Cysts were hatched following Ali et al. [17]. From the naupliar stage until the termination of the experiments, we used the single-celled alga Chlorella vulgaris. C. vulgaris (Strain CL-V-3, CICESE, Ensenada, Baja California, Mexico) was mass cultured using Bold's basal medium [25]. Based on a preliminary study, we selected 1.0 × 106 cells/ ml of Chlorella as the food density. Analytical grade sodium chloride was used for preparing different salinity levels. For cyst hatching, maintaining B. schaefferi in cultures or for experiments we used reconstituted moderately hard water (the EPA medium) [26] as the medium, which was prepared by dissolving 0.9 g of NaHCO3, 0.6 g of CaSO4, 0.6 g of MgSO4 and 0.04 g of KCl per litre of distilled water.
Experimental design
Based on a preliminary study, five nominal concentrations of NaCl (0, 1, 2, 4 and 8 g/l) chosen. From a stock solution of 32 g/l, the desired concentrations were prepared through serial dilution using EPA medium. The effect of dilution on the algal density while preparing different salinity levels were considered and accordingly adjustments were done so as to obtain the final algal density of 1.0 × 106 cells /ml. The general test conditions were: fluorescent illumination (2000 Lux) continuous and diffused; temperature 23 ± 1°C, pH: 7.0–7.5, medium renewed completely after every 24 h.
Fifty nauplii were introduced into each of the 5 salt concentrations with an algal density of 1.0 × 106 cells/ml present in transparent test jars of 500 ml and the medium was renewed daily. After 10 days, by which time it was possible to distinguish the sex [13], the juveniles were used for conducting life table demographic studies. Into each of the 15 (5 salt concentrations X 3 replicates) test jars containing 100 ml medium we introduced one male and one female B. schaefferi (juveniles). Following initiation of the experiments, we counted the number of cysts from each test jar and the medium was renewed with appropriate salt level and algal density daily. If a male partner in a given test jar died then it was replaced by another one being grown simultaneously under similar conditions from the stock. Similarly if a female died then it was replaced by another female of similar test conditions, although the cyst count was not considered further [13]. The experiments were discontinued when every individual of the original pair died in each replicate. Based on the data collected, we derived age-specific survivorship (lx) and fecundity (mx) curves. The following formulae were used for obtaining life history variables [15]:
lx = Proportion of survivorship per day
mx = Proportion of offspring produced per female per day
where, Tx = number of individuals per day
nx = number of living individuals at the initiation and the age × (days)
Rate of population increase, Euler-Lotka equation (solved iteratively and using jackknife method [27]):
where r = rate of population increase per day, w = age at maturity (days)
Differences in the data on life history variables obtained under different salt concentrations were statistically evaluated using analysis of variance (one-way ANOVA) and Tukey's tests [28].
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
SSSS: Idea, design, some part of data collection and interpretation and write up; participated sufficiently to be an author of this manuscript.
LB: Culture, maintenance of organisms, some part of data analysis and interpretation; participated sufficiently to be an author of this manuscript.
SN: Some parts of data collection, analysis, interpretation and write up; participated sufficiently to be an author of this manuscript.
GCM: Some parts of data collection, interpretation and write up; participated sufficiently to be an author of this manuscript.
KTB: Some parts of data collection, interpretation and write up; participated sufficiently to be an author of this manuscript.
Acknowledgements
Two anonymous reviewers have greatly improved our manuscript. This study was supported by a project from the Mexican National Council for Science and Technology (CONACyT-41786). Thanks are also due to Marcela Carmarillo Ortiz (Head Librarian, UNAM Campus Iztacala) for bibliographic search.
==== Refs
Dumont H Negrea S Introduction to the Class Branchiopoda. Guides to the Identification of the Microinvertebrates of the Continental Waters of the World 2002 Backhuys Publishers, The Netherlands
Williams DD The ecology of temporary waters 1987 Croom Helm. London/Sidney, Timber Press, Oregon
Broch ES Osmoregulatory patterns of adaptation to inland astatic waters by two species of fairy shrimps, Branchinecta gigas Lynch and Branchinecta mackini Dexter Journal of Crustacean Biology 1988 8 383 391
Geddes MC Seasonal fauna of some ephemeral saline waters in western Victoria with particular reference to Parartemia zietziana Sayce (Crustacea: Anostraca) Aust J Mar Freshwater Res 1976 27 1 22
Dodson SI Frey DG Thorp JH, Covich AP Cladocera and other branchiopoda Ecology and classification of North American freshwater invertebrates 2001 Academic Press, London 850 914
Williams WD Salinity as a determinant of the structure of biological communities in salt lakes Hydrobiologia 1998 381 191 201 10.1023/A:1003287826503
Peredo-Alvarez VM Sarma SSS Nandini S Combined effect of concentrations of algal food (Chlorella vulgaris) and salt (sodium chloride) on the population growth of Brachionus calyciflorus and Brachionus patulus (Rotifera) Rev Biol Trop 2003 51 399 408 15162733
Lynch M The evolution of cladoceran life histories Quarterly Reviews of Biology 1980 55 23 42 10.1086/411614
Forbes VE Calow P Is the per capita rate of increase a good measure of population-level effects in ecotoxicology? Environ Toxicol Chem 1999 18 1544 1556 10.1897/1551-5028(1999)018<1544:ITPCRO>2.3.CO;2
Baxevanis AD El-Bermawi N Abatzopoulos TJ Sorgeloos P Salinity effects on maturation, reproductive and life span characteristics of four Egyptian Artemia populations (International Study on Artemia. LXVIII) Hydrobiologia 2004 513 87 100 10.1023/B:hydr.0000018174.72317.cf
Maier G Hoessler J Tessenow U Succession of physical and chemical conditions and of crustacean communities in some small, man-made, water bodies Int Rev Hydrobiol 1998 83 405 418
Beladjal L Peiren N Dierckens KR Mertens J Feeding strategy of two sympatric anostracan species (Crustacea) Hydrobiologia 1997 359 207 212 10.1023/A:1003177828859
Beladjal L Peiren N Vendekerckhove TTM Mertens J Different life histories of the co-occurring fairy shrimps Branchipus schaefferi and Streptocephalus torvicornis J Crust Biol 2003 23 300 307
Anaya-Soto A Sarma SSS Nandini S Longevity of the freshwater anostracan Streptocephalus mackini (Crustacea: Anostraca) in relation to food (Chlorella vulgaris) concentration Freshwater Biol 2003 48 432 439 10.1046/j.1365-2427.2003.01015.x
Krebs CJ Ecology; the experimental analysis of distribution and abundance 1985 Harper & Row, New York
Nandini S Sarma SSS ifetable demography of four cladoceran species in relation to algal food (Chlorella vulgaris L) density Hydrobiologia 2000 435 117 126 10.1023/A:1004021124098
Ali AJ Sarma SSS Dumont HJ Cyst production in the fairy shrimp, Streptocephalus proboscideus (Anostraca) in relation to algal and loricated rotifer diet Crustaceana 1999 72 517 530 10.1163/156854099503564
Nandini S Sarma SSS Population growth of some genera of cladocerans (Cladocera) in relation to algal food (Chlorella vulgaris) levels Hydrobiologia 2003 491 211 219 10.1023/A:1024410314313
Sarma SSS Nandini S Gulati RD Life history strategies of cladocerans: comparisons of tropical and temperate taxa Hydrobiologia 2005 542 315 333 10.1007/s10750-004-3247-2
King CE Dingle H, Hegmann JP The evolution of lifespan Evolution and genetics of life histories 1982 Springer Verlag, New York 121 128
Sarma SSS Elguea-Sánchez B Nandini S Effect of salinity on competition between the rotifers Brachionus rotundiformis Tschugunoff and Hexarthra jenkinae (De Beauchamp) (Rotifera) Hydrobiologia 2002 474 183 188 10.1023/A:1016535821741
Gonzalez RJ Drazen J Hathaway S Bauer B Simovich M Physiological correlates of water chemistry requirements in fairy shrimps (Anostraca) from southern California J Crust Biol 1996 16 315 322
Doyle JE McMahon BR Effects of acid exposure in the brine shrimp Artemia franciscana during development in seawater Comparative Biochemistry and Physiology 1995 112A 123 129
Maier G The status of large branchiopods (Anostraca: Notostraca, Conchostraca) in Germany Limnologica 1998 28 223 228
Borowitzka MA Borowitzka LJ Micro-algal biotechnology 1988 Cambridge University Press, United Kingdom
Weber CI Methods for measuring the acute toxicity of effluents and receiving waters to freshwater and marine organisms 1993 United States Environmental Protection Agency, Cincinnati, Ohio
Meyer JS Ingersoll CG McDonald LL Boyce MS Estimating uncertainty in population growth rates: Jackknife vs bootstrap techniques Ecology 1986 67 1156 1166
Zar JH Biostatistical analysis 2003 Pearson Education Pvt. Limited, Singapore
|
16176591
|
PMC1183265
|
CC BY
|
2022-09-08 23:19:43
|
no
|
Saline Syst. 2005 Jul 4; 1:4
|
utf-8
|
Saline Syst
| 2,005 |
10.1186/1746-1448-1-4
|
oa_comm
|
==== Front
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv12_04_0013
From the Editor in Chief
Health Monitoring and Life on the Mississippi
Wilcox Lynne S. MD, MPH Editor in Chief
4 2004
15 3 2004
1 2 A012004
==== Body
Designing health monitoring systems is a complex task. This issue of Preventing Chronic Disease includes a report and commentary on measuring the burden of diabetes at the individual level in minority populations (1, 2) and a report on measuring heart disease and stroke indicators at the policy level (3).
To inspire stalwart professionals to design such systems, I turn to an individual recognized for his insightful commentary — Mark Twain, also known as Samuel Clemens. Twain had a keen eye for the idiosyncrasies of human behavior, and his nonfiction works suggest he was adept at amateur qualitative research. Though he was a man of letters rather than a scientist, he clearly appreciated the issues involved in gathering quality information:
There is something fascinating about science. One gets such wholesome returns of conjecture out of such a trifling investment of fact (4).
The balance of conjecture and fact is a source of ongoing tension in public health: collecting data is time-consuming and costly, but operating health programs based on conjecture is risky.
Although Healthy People 2010 (5) emphasizes the elimination of health disparities, the nation lacks an accurate way of measuring the burden of diabetes in minority populations. Surveillance systems, for example, may treat Spanish-speaking populations as a homogeneous Hispanic group while individuals within the group may have originated from different Spanish-speaking countries (1). Without an understanding of the diversity of these cultures, health programs may lack the context to serve these populations effectively. A report in this issue of Preventing Chronic Disease, prepared by an expert panel at the Centers for Disease Control and Prevention (CDC), recommends extending the capacity of existing surveys to obtain better measurements of minority populations instead of developing new surveys, in light of the high costs of taking the latter route (1).
Community-level policy and environmental indicators related to stroke and heart disease prevention present a different problem. Researchers in 2 states, Alabama and South Carolina, examined data sources for 31 pilot indicators and found that, while data sources for most indicators are available in the school setting and are available for indicators of tobacco policies across all settings examined, data sources are least available in the health care and work site settings(3). This report calls for combining current data sources with new surveillance efforts. These efforts are also likely to be costly, but will focus on the policies of states, work sites, and health care organizations, rather than on data from individual respondents.
There are three kinds of lies: lies, damned lies and statistics (6).
The game of "lying with statistics" creates confusion for both public health professionals and the people who use their data. Even when diligent public health professionals strive to explain the nuances of surveillance findings, numbers are often misconstrued by well-meaning policy makers.
The CDC expert panel on using survey data for diabetes surveillance among minority populations points out that most diabetes surveys lack sufficient sample size to provide statistics for the burden of diabetes among smaller minority populations, especially at state or local levels. These populations thus remain hidden and the causes of health disparities remain obscure. The panel recommends enhancing community-level health surveys to gain more details on populations within each community.
The study on policy indicators for heart disease and stroke identifies another challenge in surveillance statistics: indicators may lack the sensitivity or specificity needed to assess a policy's effects on health (3). The study's investigators define an insensitive indicator as one that states the simple presence or absence of a policy without indicating the extent to which the policy addresses an issue. Indicators with poor specificity are ambiguous or lack definition of key terms. The report found that both characteristics were frequent problems among the indicators examined. The authors state that improving sensitivity and specificity of indicators is critical to measuring improvements in population health.
I was gratified to be able to answer promptly, and I did. I said I didn't know (7).
There are times when humility is the best response. The expert panel and policy indicator research group labored long and hard to establish recommendations. Nevertheless, the recommendations are still only conjecture, and until they are implemented and validated, we must fall back on Twain's prompt response as a young riverboat pilot.
When we have no answers, we can contemplate quotes from a master. Indeed, these quotes may carry you through any number of trying circumstances in public health. Twain was a pragmatic man. He would understand if you quote him frequently, and occasionally without proper reference.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Wilcox LS. Health monitoring and Life on the Mississippi. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/04_0013.htm.
==== Refs
1 Burrows NR Lojo J Engelgau MM Geiss LS Using survey data for diabetes surveillance among minority populations: a report of the Centers for Disease Control and Prevention's expert panel meeting. Prev Chronic Dis [serial online] 2004 Apr [2004 Mar 15] Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0018.htm 2004 4 2 A02
2 Desai J State-based diabetes surveillance among minority populations Prev Chronic Dis [serial online] 2004 Apr [2004 Mar 15] Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0030.htm 2004 4 2 A03
3 Pluto DM Phillips MM Matson-Koffman D Shepard DM Raczynski JM Brownstein JN Policy and environmental indicators for heart disease and stroke prevention: data sources in two states Prev Chronic Dis [serial online] 2004 Apr [2004 Mar 15] Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0019.htm 2004 4 21 78 93
4 Twain M Life on the Mississippi 2001 11 7 New York (NY) Signet Publishing 106
5 U.S. Department of Health and Human Services Healthy People 2010: understanding and improving health Available from: URL: http://www.healthypeople.gov/publications/ 2000 2nd 76
6 Neider C Twain M The autobiography of Mark Twain 1959 New York (NY) HarperCollins, Publishers, Inc. 195
7 Twain M Life on the Mississippi 2001 11 7 New York (NY) Signet Publishing 36
|
15663877
|
PMC1183493
|
CC BY
|
2022-01-24 23:39:57
|
no
|
Prev Chronic Dis. 2004 Mar 15; 1(2):A01
|
utf-8
|
Prev Chronic Dis
| 2,004 |
nan
|
oa_comm
|
==== Front
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv12_03_0018
Editorial
Using Survey Data for Diabetes Surveillance Among Minority Populations: A Report of the Centers for Disease Control and Prevention’s Expert Panel Meeting
Burrows Nilka Ríos MT, MPH 4770 Buford Hwy, Mail Stop K-10, Atlanta, GA 30341 770-488-1057 [email protected]
Lojo José MPH
Engelgau Michael M. MD, MS
Geiss Linda S. MA
Division of Diabetes Translation, National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention
4 2004
15 3 2004
1 2 A022004
Introduction
Data on diabetes morbidity and mortality and the quality of care among U.S. minority populations are necessary to assess progress toward eliminating racial/ethnic disparities and to design and implement effective interventions. This paper summarizes the discussions and recommendations of an expert panel to address the use of survey data for diabetes surveillance among minority populations.
Methods
The Centers for Disease Control and Prevention's Division of Diabetes Translation convened an expert panel of persons with survey experience and awareness of the problems in conducting health-related surveys among minority populations. Panel members were asked to 1) determine ways to enhance the ability of existing survey systems to address diabetes surveillance among minority populations; 2) identify survey systems that could be used to address surveillance needs; and 3) determine whether new minority-specific survey systems need to be developed.
Results
Panel members concluded that, although no existing survey system is completely adequate for diabetes surveillance among minority populations, new systems should not be developed. They recommended 1) investigating the use of community-based surveys; 2) exploring the ability of national surveys to increase sample sizes and produce state-level estimates; and 3) encouraging government agencies and public health programs to coordinate and integrate diabetes-related survey data and share analytic methodology.
Conclusion
No existing survey is suitable for conducting minority-specific diabetes surveillance. Modifying and expanding existing surveys to establish a diabetes surveillance system of sentinel minority populations would be more feasible than developing a new one. Interagency coordination and collaboration will be critical in this effort.
==== Body
Introduction
An estimated 17 million people in this country have diabetes, and of these, nearly 6 million are not aware of their condition (1). According to the American Diabetes Association, diabetes costs nearly $100 billion in direct medical care costs and indirect costs such as lost productivity (2). Appropriate preventive care practices could prevent or delay a large proportion of the costly and disabling consequences of diabetes (3).
Compared with the general population, certain racial/ethnic communities, as well as older Americans and economically disadvantaged Americans, are disproportionately affected by diabetes and are at increased risk for some diabetes-related complications (4). Potential reasons for these groups being disproportionately affected include a greater prevalence of risk factors and comorbid conditions, inadequate access to medical care, and suboptimal diabetes-related preventive care.
The Division of Diabetes Translation (DDT) at the Centers for Disease Control and Prevention (CDC) uses many data sources to conduct public health surveillance of diabetes and to estimate the burden of diabetes at the national and state levels. These data sources include surveys such as the Behavioral Risk Factor Surveillance System (BRFSS) and the National Health Interview Survey (NHIS), administrative databases such as the U.S. Renal Data System, databases of the Health Care Financing Administration, and vital statistics data. However, several problems are associated with the use of these data in diabetes surveillance among minority populations: 1) surveys cannot reach all minority populations of interest; 2) administrative data sometimes do not include information on race and ethnicity; 3) small sample sizes of minority populations do not allow for accurate estimates of the diabetes burden at the state or community level; and 4) minority populations often are erroneously treated as homogeneous (for example, Mexican Americans, Cuban Americans, and other distinct ethnic groups are usually considered to be part of a homogenous group called "Hispanics").
In 1998, the Department of Health and Human Services' Initiative to Eliminate Racial and Ethnic Disparities in Health established as a national priority the elimination of racial and ethnic disparities in health outcomes among U.S. residents. Data for monitoring progress toward this objective, however, are lacking for many minority populations. Obtaining high-quality surveillance data on diabetes-related morbidity and mortality and quality of diabetes care among minority populations is thus important in identifying disparities and monitoring progress toward reducing these disparities. To discuss the use of survey data for the surveillance of diabetes among minority populations, DDT convened an expert panel meeting.
Methods
DDT organized and conducted the August 5–6, 2002, expert panel meeting. The panel consisted of 18 persons from agencies or organizations outside of DDT, most of whom had survey experience and were aware of the challenges in conducting surveys of health-related information among minority populations. Panel members represented 4 federal agencies, 4 university-affiliated institutions, 2 private nonprofit organizations, and one state health department (Table 1). Panel members were asked to help the CDC to 1) determine ways to enhance the ability of existing survey systems to address diabetes surveillance among minority populations; 2) identify survey systems that could be used, but currently were not being used, to address surveillance needs; and 3) determine whether new minority-specific survey systems need to be developed.
Results
The expert panel members identified numerous national, state, and community-based survey systems and discussed the systems' abilities to address the health information needs of minority populations (Table 2). Most of the identified surveys were cross-sectional in design and national in scope, administered either face-to-face or by telephone, and sponsored or supported by various federal agencies. The expert panel members noted several issues and problems associated with using national survey data for disease and risk factor surveillance. These included sample sizes for minority populations being too small, sampling being limited to the larger minority populations, discrete minority groups being treated as homogeneous rather than diverse populations, and national data being inadequate for estimating state and community problems.
Moreover, the panel questioned whether a national sample of specific minority populations could produce meaningful results given the diversity within these populations. State and community surveys share some of these same problems, but they have the advantage of access to local data, which have more immediate relevancy in planning and evaluating community-based interventions to improve public health. Although local surveys may have the greatest potential for targeting subgroups of minorities, data on such subgroups may not be generalizable to the larger minority populations.
The panel members recognized that the list of surveys discussed as potential sources of data was probably not complete and recommended several strategies to identify additional survey systems that could be useful. These strategies included 1) a summary review of existing longitudinal studies and other relevant data; 2) a search of surveys sponsored by federal or public agencies other than the Department of Health and Human Services; and 3) a query of state health programs to establish an inventory of special surveys.
Recognizing that diabetes is an important growing public health problem (5), the panel stressed that the inclusion of diabetes-related data in existing survey systems is critical. The panel highlighted the need to coordinate the diabetes surveillance efforts of survey systems by using uniform diabetes-related questions, sharing analytical techniques (such as pooling data), and promoting the standardization of measurement and analysis practices.
The panel also discussed the need to examine survey content to ensure that surveys are capable of producing the thorough data necessary to design effective public health interventions. In addition to producing data on minority racial and ethnic groups, surveys also should produce data on other disadvantaged populations as measured by socioeconomic status, social capital, community resources, education, and access to or denial of health insurance, because these factors generally underlie many racial and ethnic disparities.
Finally, the panel discussed the need to establish new minority-specific survey systems and concluded that developing and maintaining new survey systems would be too costly and time-consuming. Instead, the panel recommended expanding upon and enhancing existing survey systems at the state and local levels. Specifically, the panel recommended investigating the use of community-based surveys, such as those in the Racial and Ethnic Approaches to Community Health 2010 project, state- and local-specific surveys of the BRFSS, and the State and Local Area Integrated Telephone Survey. The panel also suggested modifying existing surveys by, for example, increasing sample size, adding supplementary content, sampling additional minority groups, and developing the capacity of national surveys such as NHIS to collect state and community-level data.
Discussion
Data on diabetes-related morbidity and mortality and quality of diabetes care among different U.S. minority populations are necessary to 1) assess progress toward eliminating racial/ethnic disparities in the health burden of diabetes, and 2) design and implement effective interventions for minority groups that are disproportionately affected by diabetes.
No existing survey is suitable for conducting minority-specific diabetes surveillance. Modifying and expanding existing surveys to establish a diabetes surveillance system of sentinel minority populations, however, would be more feasible than developing a new one. Interagency coordination and collaboration will be critical in establishing such a system.
Figures and Tables
Table 1 Members of the Centers for Disease Control and Prevention's Expert Panel on Using Survey Data for Diabetes Surveillance Among Minority Populations, 2002
Centers for Disease Control and Prevention Other Federal Agencies University-affiliated Institution Private Nonprofit Organization State Health Department
Lawrence Barker, PhD
Stephen J. Blumberg, PhD
Nilka Ríos Burrows, MPH
Michael M. Engelgau, MD, MS
Clark Denny, PhD
Linda S. Geiss, MA
H. Wayne Giles, MD
Howard Goldberg, PhD
José Lojo, MPH
Ali H. Mokdad, PhD
Kathryn S. Porter, MD, MS
Edward F. Tierney, MPH
Elizabeth Zell, MStat Karen Beauregard, MHA, Agency for Healthcare Research and Quality Charlotte Steeh, PhD, Georgia State University, School of Policy Studies Heather DH Mann, MA, National Indian Council on Aging Robert W. Indian, MS, Ohio Department of Health
Gerald S. Adler, MPhil, Centers for Medicare And Medicaid Services Carol-Ann Emmons, PhD University of Chicago, National Opinion Research Center Robert Valdez, PhD, Rand Institute
Adrienne Oneto, MA, U.S. Census Bureau David Weir, PhD, University of Michigan, Institute for Social Research
Elisa T. Lee, PhD, University of Oklahoma, Health Sciences Center
Table 2 Description of National, State and Community-based Survey Systems Relevant to Addressing Health Information Needs of Minority Populations, Centers forDisease Control and Prevention, 2002
National Surveys Agencya Target Population Survey Design Survey Mode Frequency
American Community Survey U.S. Census Bureau Civilian, non-institutionalized household population Cross-sectional Mail, phone, person Annual beginning in 2003
Health and Retirement Study NIA Civilian, non-institutionalized household population aged >50 years Panel Person Every 2 years
Medicare Current Beneficiary Survey (MCBS) CMS Medicare beneficiaries aged >64 years Panel Person (computer-assisted) 3 times/year for 4 years
Medical Expenditure Panel Survey (MEPS) AHRQ Civilian, non-institutionalized household population and nursing home residents Panel Person (computer-assisted) Several times/year for 2 years
National Survey Family Growth CDC Civilian, non-institutionalized population aged 15-44 years Cross-sectional Person Periodic
National Health Interview Survey (NHIS) CDC Civilian, non-institutionalized household population Cross-sectional Person Annually
National Health and Nutrition Examination Survey (NHANES) CDC Civilian, non-institutionalized household population Cross-sectional Person Periodic
Consumer Assessment of Health Plans Survey (CAHPS) AHRQ Persons currently enrolled in health plans Cross-sectional Phone (computer-assisted) To be determined
State and Community-based Surveys Agency Target Population Survey Design Survey Mode Frequency
State and Local Area Integrated Telephone Survey (SLAITS) CDC Subgroups of civilian, non-institutionalized household population (e.g., low-income households) Cross-sectional Phone Annual
Behavioral Risk Factor Surveillance System (BRFSS) CDC State-based, civilian, non-institutionalized household population aged >17 years Cross-sectional Phone Annual
Youth Risk Behavior Survey CDC Students in grades 9-12 Cross-sectional Person Every 2 years
BRFSS special surveys State Civilian, non-institutionalized population aged >17 years Cross-sectional As specified by state As specified by state
Racial/Ethnic Approaches to Community Health (REACH 2010) CDC 21 minority communities per the REACH 2010 grantees Cross-sectional Phone Annually for 4 years
a AHRQ: Agency for Healthcare Research and Quality; CDC: Centers for Disease Control and Prevention; CMS: Centers for Medicare and Medicaid Services; NIA: National Institute on Aging.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Burrows NR, Lojo J, Engelgau MM, Geiss LS. Using survey data for diabetes surveillance among minority populations: a report of the Centers for Disease Control and Prevention's expert panel meeting. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0018.htm.
==== Refs
1 Centers for Disease Control and Prevention. National diabetes fact sheet: general information and national estimates on diabetes in the United States, 2003 U.S. Department of Health and Human Services, Centers for Disease Control and Prevention Atlanta (GA) 2003 Available from: URL: http://www.cdc.gov/diabetes/pubs/pdf/ndfs_2003.pdf
2 American Diabetes Association Economic consequences of diabetes mellitus in the U.S. in 1997 Diabetes Care 1998 21 2 296 309 9539999
3 Narayan KM Gregg EW Fagot-Campagna A Engelgau MM Vinicor F Diabetes — a common, growing, serious, costly, and potentially preventable public health problem Diabetes Res Clin Pract 2000 10 50 Suppl 2 S77 S84 11024588
4 Cowie CC Eberhardt MS Manson JE Stampfer MJ Rimm EB Speizer FE Sociodemographic characteristics of persons with diabetes Diabetes in America. 2nd ed U.S. Department of Health and Human Services, Public Health Service, National Institutes of Health Washington (DC) 1995 85 116 Harris MI Cowie CC Stern MP Boyko EJ Reiber GE Bennett PH
5 Mokdad AH Ford ES Bowman BA Nelson DE Engelgau MM Vinicor F Diabetes trends in the U.S.: 1990-1998 Diabetes Care 2000 9 23 9 1278 1283 10977060
|
15663878
|
PMC1183494
|
CC BY
|
2022-01-27 23:15:21
|
no
|
Prev Chronic Dis. 2004 Mar 15; 1(2):A02
|
utf-8
|
Prev Chronic Dis
| 2,004 |
nan
|
oa_comm
|
==== Front
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv12_04_0006
Essay
The Burden of Chronic Disease: The Future is Prevention
Introduction to Dr. James Marks' presentation, “The Burden of Chronic Disease and the Future of Public Health”
Hardy George E. Jr. MD, MPH Executive Director Association of State and Territorial Health Officials
1275 K Street NW, Suite 800, Washington, DC 20005-4006 Association of State and Territorial Health Officials 202-371-9090 [email protected]
4 2004
15 3 2004
1 2 A042004
==== Body
Chronic diseases impose an enormous financial and societal burden on the United States. According to the Centers for Disease Control and Prevention (CDC), chronic diseases today account for 70% of the deaths of all Americans and 75% of this country’s annual health care costs. Unless we take steps now to deal effectively with chronic diseases, our nation is headed for a serious financial and quality-of-life crisis. Among the contributing factors to this crisis are the aging of our population; increases in obesity, particularly among adolescents; and the tragedy of tobacco addiction.
No one speaks with more passion, conviction, and vision about the need to address this pending crisis than Dr. James Marks, director of the CDC’s National Center for Chronic Disease Prevention and Health Promotion. As he demonstrates so clearly in his presentation, "The Burden of Chronic Disease and the Future of Public Health," public health prevention programs can, with real societal and political will, substantially reduce or even prevent the burden of many major chronic disease conditions.
His presentation makes a strong case for moving from a palliative medical model to a prevention-based approach. He argues most persuasively that preventing chronic diseases can provide Americans with a better quality of life, reduce unnecessary medical costs and lost productivity, and strengthen our national economy.
Funding for research and medical advances alone will alleviate neither the cost nor the suffering of individuals faced with a chronic disease. Without concomitant investments in public health prevention and control efforts to implement the prevention research results we already have, rising health care costs and preventable deaths will continue. Americans deserve a national commitment to translating critically important research findings into practical public health solutions.
Despite limited resources, state public health agencies are working to create healthy opportunities for all Americans by partnering with public, private, and voluntary organizations to ensure a strategic application of known prevention successes to chronic disease prevention programming and resources. Public health professionals are also working with Medicaid agencies to develop clinical assistance programs and to explore policy changes that could benefit individuals and communities by more effectively targeting scarce prevention resources. Imagine what could be done if a real national commitment were made to provide the resources necessary to truly address the challenges and opportunities of an aging population and the risk factors attendant to chronic diseases.
Because of the increasing burden of chronic diseases, the United States faces a potential financial and health care crisis of unparalleled proportion. We must not lose this opportunity to do whatever we can to reduce the costly and unnecessary burden of chronic disease that will continue to fuel that crisis. As health professionals, it is incumbent upon us to join Dr. Marks in ensuring that our nation’s political leaders and the citizens they represent better understand the profound burden of chronic disease and the positive efforts that need to be taken now to reduce that burden. Our nation deserves no less.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Hardy GE Jr. The burden of chronic disease: the future is prevention. Introduction to Dr. James Marks' presentation, The Burden of Chronic Disease and the Future of Public Health. Prev Chronic Dis [serial online] 2004 April [date cited]. Available from URL: http://www.cdc.gov/pcd/issues/2004/apr/04_0006.htm.
|
15663880
|
PMC1183496
|
CC BY
|
2022-01-31 23:16:18
|
no
|
Prev Chronic Dis. 2004 Mar 15; 1(2):A04
|
utf-8
|
Prev Chronic Dis
| 2,004 |
nan
|
oa_comm
|
==== Front
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv12_03_0019
Original Research
PEER REVIEWEDPolicy and Environmental Indicators for Heart Disease and Stroke Prevention: Data Sources in Two States
Pluto Delores M. PhD Prevention Research Center, Arnold School of Public Health
730 Devine St, University of South Carolina, Columbia, SC 29208 [email protected]
803-576-5994
Phillips Martha M. PhD, MPH, MBA Department of Epidemiology, School of Public Health, University of Alabama at Birmingham (presently with the Dept. of Psychiatry and Behavioral Sciences, Centers for Mental Healthcare Research, University of Arkansas for Medical Sciences)
Matson-Koffman Dyann DrPH, MPH Cardiovascular Health Branch, Division of Adult and Community Health, National Center for Chronic Disease Prevention and Health Promotion, CDC
Shepard Dennis M. MAT Prevention Research Center, Arnold School of Public Health, University of South Carolina
Raczynski James M. PhD Center for Health Promotion and Department of Health Behavior, School of Public Health, University of Alabama at Birmingham (presently with the College of Public Health, University of Arkansas for Medical Sciences)
Brownstein J. Nell PhD Cardiovascular Health Branch, Division of Adult and Community Health, National Center for Chronic Disease Prevention and Health Promotion, CDC
4 2004
15 3 2004
1 2 A052004
Introduction
Investigators in South Carolina and Alabama assessed the availability of data for measuring 31 policy and environmental indicators for heart disease and stroke prevention. The indicators were intended to determine policy and environmental support for adopting heart disease and stroke prevention guidelines and selected risk factors in 4 settings: community, school, work site, and health care.
Methods
Research teams used literature searches and key informant interviews to explore the availability of data sources for each indicator. Investigators documented the following 5 qualities for each data source identified: 1) the degree to which the data fit the indicator; 2) the frequency and regularity with which data were collected; 3) the consistency of data collected across time; 4) the costs (time, money, personnel) associated with data collection or access; and 5) the accessibility of data.
Results
Among the 31 indicators, 11 (35%) have readily available data sources and 4 (13%) have sources that could provide partial measurement. Data sources are available for most indicators in the school setting and for tobacco control policies in all settings.
Conclusion
Data sources for measuring policy and environmental indicators for heart disease and stroke prevention are limited in availability. Effort and resources are required to develop and implement mechanisms for collecting state and local data on policy and environmental indicators in different settings. The level of work needed to expand data sources is comparable to the extensive work already completed in the school setting and for tobacco control.
==== Body
Introduction
Beginning in 1998, the Centers for Disease Control and Prevention (CDC) received federal funding to support state heart disease and stroke prevention programs. The purpose of these state programs is to develop comprehensive programs emphasizing community-based policy and environmental strategies to reduce risk factors related to heart disease and stroke, such as physical inactivity, poor nutrition, tobacco use, and hypertension. The CDC recommends that assessment and policy development be included within the 10 core public health services to support individual and community health efforts. To monitor their progress on developing community-based policy and environmental strategies, state programs require intermediate evaluation measures of policy and environmental factors. Community-level indicators have been used to measure such intermediate policy and environmental outcomes for other community-based disease prevention programs (1,2). For example, community-level indicators for tobacco use include the existence and quality of clean air laws and the presence of cigarette vending machines in restaurants.
The Cardiovascular Health Branch of the CDC, in collaboration with other units within the National Center for Chronic Disease Prevention and Health Promotion, used literature searches, expert recommendations, and a Delphi process to identify policy and environmental indicators associated with physical activity, nutrition, tobacco control, and national heart disease and stroke prevention guidelines. A draft list of 31 pilot policy and environmental indicators was developed with the intention of revising the list upon feedback from this study. The indicators were selected, in part, because they were thought to be feasible for consistent measurement across 50 states. For example, one indicator can be used to track the number of states that have policies requiring daily physical education for grades K–12. The indicators were categorized by community, school, work site, or health care setting (3).
Because literature on community-level indicators was limited, little was known about the availability of data sources for use by state heart disease and stroke prevention programs. Hence, the Cardiovascular Health Branch staff asked the Alabama and South Carolina heart disease and stroke prevention program directors to assess the availability of data sources for the 31 pilot indicators in those 2 states and to provide their perspectives on the feasibility of using these indicators. These 2 states were selected because of their proximity to the CDC in Atlanta for technical assistance and because each state program has a close relationship with its Prevention Research Center. Each state program collaborated with its Prevention Research Center (the Center for Health Promotion at the University of Alabama at Birmingham and the Prevention Research Center at the University of South Carolina) to carry out the assessment. This paper summarizes the findings and provides recommendations for collecting data and refining community-level indicators for the surveillance of heart disease and stroke prevention.
Methods
Between October 2000 and October 2001, research teams at the South Carolina and Alabama Prevention Research Centers worked in tandem to identify and examine possible data sources and to assess sensitivity and specificity for each indicator. To identify possible data sources, the research teams completed a systematic search within each of 4 settings: community, school, work site, and health care. They identified individuals in state departments of health and education, other state agencies, and private organizations who might have access to or be aware of relevant data sources (Table 1).
Individuals were identified using a snowball technique that began with people or organizations known to research team members as well as contacts identified from Web sites. As individuals were identified, a team member contacted them by telephone. A conversational interview was used to ask respondents if they collected any data related to a given indicator, and if so, they were asked to provide details about the data source. If the agency or organization did not collect relevant data, the research team requested names of other potential informants or sources of data. These new informants were contacted and the process was repeated until all identified individuals or agencies were contacted.
Additionally, the research teams completed literature and on-line searches using keywords from each indicator (e.g., sidewalks, mixed-use, bicycle) to identify additional data sources and possible contacts. Once data sources were identified, the research teams reviewed each data source, taking note of the degree to which the data fit the indicator; the frequency and regularity with which data were collected; the consistency of the data collected across time; the costs (time, money, personnel) associated with data collection and/or data access; and the accessibility of data.
In addition to evaluating the data sources, the research teams made a general assessment of the sensitivity and specificity of each indicator. Sensitivity refers to the extent to which an indicator allows for documentation of incremental change. Indicators were flagged as lacking sensitivity if they referred only to the presence or absence of a policy rather than the extent to which a policy addressed an issue. Indicators were also flagged as lacking sensitivity if they measured change at an inappropriate level (i.e., if an indicator asked about state policy when policy is set at the local level). Specificity refers to the extent to which an indicator precisely and accurately describes an environmental feature or policy being measured. Indicators were flagged as lacking specificity if they were ambiguous or failed to define key terms.
During this project, research teams participated in regular conference calls with personnel from the CDC's Cardiovascular Health Branch and the state program managers in Alabama and South Carolina to review progress, clarify issues, and share protocols and information. Although each research team completed tasks independently and had a different contractual relationship with its state program, efforts were made to ensure that working protocols (including evaluation criteria and reporting formats) were consistent.
Results
Among the 31 pilot indicators, 11 (35%) had readily available data sources and 4 (13%) had data sources that could provide at least partial measurement. Data sources were available for most indicators in the school setting and for indicators related to tobacco policies across all settings. Data sources were least available in the work site and health care settings. Most data sources identified were maintained by a national agency or organization (e.g., CDC, U.S. Department of Agriculture [USDA], National Transportation Enhancements Clearinghouse). State agencies often report data to these national data sources. Neither research team found a data source unique to its state.
The list of indicators was in draft form at the time of this assessment; thus, many pilot indicators were found to lack specificity. Ten (37%) indicators were flagged as lacking specificity because of ambiguous or imprecise definitions. In addition, 9 (29%) indicators were flagged as lacking sensitivity because they considered only the presence or absence of state legislation, not the quality or degree to which recommendations were included in the legislation. More detailed results are presented about the data sources found in each of the 4 settings.
Community setting
Two of the 8 pilot indicators in the community setting — clean indoor air laws and smoking in the home — have readily available data sources (Table 2).
The legislative database in the State Tobacco Activities Tracking and Evaluation (STATE) system summarizes state tobacco legislation, including smoke-free indoor air ordinances for restaurants, day care centers, and public places (4,5). The Office on Smoking and Health at the CDC maintains the database, based on a quarterly search of the LexisNexis legal database (4,5). The database can be used to monitor the presence or absence of state policies and the content of those policies (e.g., restrictions, penalties, enforcement). The legislative database, however, does not capture municipal ordinances that might be enacted in the absence of state policies. Beginning in 1998, the optional Tobacco Indicators module of the annual Behavioral Risk Factor Surveillance System (BRFSS) asked respondents if anyone smoked anywhere in their homes. In 2001, this was changed to ask if smoking was allowed in their homes (6). The Tobacco Indicators module was used by 25 states in 2002.
Data sources also are available that partially measure 2 other community indicators: highway funding of transportation alternatives and the number of farmers' markets. The National Transportation Enhancements Clearinghouse maintains a database of transportation enhancements funds allocated and spent by each state under the Transportation Equity Act for the 21st Century (TEA-21). This searchable, on-line database is updated annually (7). Funds for transportation alternatives under TEA-21, however, do not represent the entire state budget for transportation alternatives, and the database does not include the total amount of the state transportation budget. The research teams found no additional data sources that provide relevant details on highway spending at the state or local level.
The USDA maintains a list of farmers' markets searchable on-line by state (8). The database depends on reports from individual state departments of agriculture. Because the definition of a farmers' market varies by state, the data might be inconsistent or incomplete across states. For example, at the time of this study, the South Carolina listing included only 3 state-run, year-round farmers' markets. The list was recently updated to include smaller local markets that operate on a seasonal basis.
Although regional milk production figures are available, no state data were found on milk production or sales. The research teams also noted that this indicator is not a measure of environment or policy but a community-level indicator of purchasing behavior.
School setting
Ten pilot indicators for heart disease and stroke prevention were identified in the school setting (Table 3). Seven indicators that refer to state policies on physical education requirements, student physical education assessments, food availability, certifications for food service staff and physical and health education teachers, and health education curriculum have readily available data sources.
All 7 of these indicators can be assessed using data from the School Health Policies and Programs Study (SHPPS), which is conducted every 6 years. The study surveys all state departments of education and a nationally representative sample of districts and schools (11). The state survey includes questions related to each of the 7 school indicators. These indicators assume that such policies are enacted at the state level; however, in states like South Carolina and Alabama, school policies are under the authority of school districts or the schools themselves.
The School Health Education Profile (SHEP) collects data that provide partial measurement of school health councils and tobacco-free schools. SHEP is a survey completed every 2 years by a sample of school principals and lead health educators in public schools containing classrooms at the sixth-grade level or higher (12). Because no similar data source is available for elementary schools, SHEP can only partially measure these indicators. In addition, the survey does not currently include questions that lead to the assessment of all components of the tobacco-free school policies recommended by the CDC.
Work site setting
Only one of the 8 pilot work site indicators — clean indoor air laws for work sites — has a readily available data source (Table 4). Neither research team found any data sources for other work site indicators.
The STATE system contains information that measures state clean air laws that apply to work sites (4,5). This indicator is subject to the same sensitivity concerns previously noted for other clean indoor air laws — it notes only the presence or absence of state policies. The BRFSS optional Tobacco Indicators module collects information from individuals about their work site tobacco policies, but it does not measure state indoor air laws. Data on work site policies collected by the optional module would provide an estimate of the percentage of employed adults protected by a work site smoking policy.
Questions from the National Worksite Health Promotion Survey could be used to assess on-site physical activity programs and nutrition or weight management programs (13). This survey collects and provides national data for Healthy People 2010 (14). The sample is too small, however, to draw conclusions by state. Other measurement tools assess policies and environmental characteristics related to heart disease and stroke prevention within work sites, including Heart Check (15) and the Checklist of Health Promotion Environments at Worksites (16). However, these instruments are not commonly used across the country and are not designed to be used as surveillance tools.
Health care setting
Among the 5 pilot indicators identified in the health care setting, only one has a readily available data source: smoking cessation advice delivered by health care professionals (Table 5). The proportion of smokers who received advice to quit smoking in the past year has been included in the optional Tobacco Indicators module of the BRFSS since 2000.
Discussion
In Alabama and South Carolina, the school setting has data to measure — at least partially — all but one of the pilot indicators for heart disease and stroke prevention. The community, work site, and health care settings have data sources for fewer than half of the indicators.
Improving data collection
Given the overall lack of data in most settings assessed in this study, consideration should be given to designing and implementing new data collection processes. Vehicles for new data collection efforts are likely to be surveillance efforts now supported by the CDC (e.g., BRFSS, Youth Risk Behavior Surveillance System, SHPPS, SHEP). The SHPPS and SHEP are designed to collect policy data and are updated regularly to include more complete information. For example, SHEP 2002 included questions related to 2 school indicators: the percent of schools that provide health education instruction that includes the physical education topics listed in CDC's School Health Index and the proportion of schools that have adopted tobacco-free policies that meet CDC recommendations (20,21). Although the BRFSS is an individual-level surveillance tool, the optional Tobacco Indicators module already allows states to collect data to measure 2 indicators indirectly (smoking in the home and receiving advice to quit). Because this module is optional, the data are not available in all states. The availability and variability of relevant data across states can have important implications for achieving consistency within a national surveillance system for heart disease and stroke prevention. This study, however, did not explore a sufficient number of states to determine the extent of this variability.
Systems similar to the legislative database of the STATE system could be developed to monitor other state policies. In fact, in late 2003, the CDC Division of Nutrition and Physical Activity launched an on-line searchable database containing bill information related to physical activity and nutrition from all 50 states (22). Few existing national surveillance efforts, however, gather information from local governments, work sites, and health insurers. Important issues of cost — in terms of time, personnel, financial resources, and participant burden — must be considered when developing new data collection efforts or revising existing systems.
Although the research teams made extensive efforts to consult with a wide range of organizations, other data sources might exist. The research teams restricted their exploration to data that are collected either nationally or within their states. While this project did not complete an exhaustive review of data sources in other states, it did identify some noteworthy examples, such as New York's Heart Check (15). Additional surveys developed by other states (e.g., Montana, North Carolina) can be found on the Cardiovascular Health Council of the Chronic Disease Directors Web site: http://www.chronicdisease.org/cvh_council/Key%20Elements/ State%20Survey/CVH_state_survey.htm. The mechanisms illustrated at this site can serve as models for other states.
An additional challenge of data collection is assessing the impact of policy and environmental changes on behavior and health. Policy and environmental indicators provide only one part of the equation. For example, assessing the impact of school policies on children's behavior presents challenges in obtaining informed consent from the children, school administration, and/or parents.
Refining indicators
To be useful to state programs, indicators for heart disease and stroke prevention examined in this study need to be refined to improve specificity and sensitivity. Including clear definitions would improve the specificity of the indicator and the accuracy and consistency of data collected. Sensitivity for many indicators could be enhanced by establishing criteria for evaluating policies and laws beyond consideration of their presence or absence at the state level. Some data sources like STATE and SHPPS already collect detailed information that could be used to evaluate the content and quality of policies in addition to tracking their presence or absence.
While it may be sufficient to look at states' policies for national surveillance, state programs might need additional surveillance data that show progress in meeting prevention goals within their own states. In some cases, particularly within school and community settings, it might be more relevant — albeit more costly — to assess the percentage of local jurisdictions (counties, municipalities, school districts) that implement a given policy.
The health care indicators provide the greatest challenge for surveillance. As worded, the indicators look at the percentage of insurers that provide a specific type of coverage. Knowing this information might not reflect the percentage of the population covered by those companies. For example, South Carolina currently has only 5 health maintenance organizations, which cover less than 10% of the state's population (23). Even if data indicated that all of these organizations followed the recommended guidelines, the data would not include 90% of the South Carolinians who might or might not have coverage under some other type of health care plan. In addition, insurance companies tend to negotiate with individual employers about the content of health insurance plans rather than having standard plans. Nationally, employers provide coverage for 58% of the population (23). If employer surveys are developed for other work site indicators, these surveys could include questions about health insurance provided by the employers.
The results of this investigation support the need for more attention, resources, and research to provide a consistent, documentable system for measuring indicators for heart disease and stroke prevention. It also will be important to improve the sensitivity and specificity of each indicator and to evaluate how each indicator corresponds to risk factors and health outcomes. These recommendations are consistent with the new Public Health Action Plan to Prevent Heart Disease and Stroke, which recommends enhancing data sources and systems to monitor key indicators for heart disease and stroke and "to systematically evaluate policy and program interventions" (24). Currently, the CDC is funding other projects to refine and validate these and other potential indicators for heart disease and stroke. With the evolving importance of policy and environmental factors influencing primary and secondary prevention efforts in public health, it is vital that a system be developed that will provide national, state, and possibly local data on indicators for heart disease and stroke. During the next decade, these indicators could provide valuable measurements to determine how environmental and policy changes are affecting heart disease and stroke prevention in this nation.
This journal article was supported by grant numbers U50/CCU416128 and U50/CCU416100 from the Centers for Disease Control and Prevention. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the Centers for Disease Control and Prevention.
Figures and Tables
Table 1 Examples of Agencies and Organizations Contacted for Information on Data Sources for Heart Disease and Stroke Prevention, South Carolina and Alabama, 2001
Setting Agency or Organization
Community Federal and state departments of transportation
State and local departments of parks and recreation
Federal and state departments of agriculture
National Transportation Enhancements Clearinghouse
Associations of mayors
State and national dairy associations
CDC Office on Smoking and Health
CDC Behavioral Surveillance Branch
School State departments of education
CDC Division of Adolescent and School Health
Work site Better Business Bureau
Local work site wellness associations
CDC Division of Adult and Community Health
Health care State insurance commissioners
Major third-party insurers (e.g., Blue Cross/Blue Shield, health maintenance organizations)
Table 2 Pilot Indicators and Data Sources for Heart Disease and Stroke Prevention, Community Setting, South Carolina and Alabama, 2001
Indicator Data Sources and Comments
1. Percent of highway funds devoted to transportation alternatives (e.g., bicycle lanes linked to public transportation, mass transit systems, facilities and roadway changes; supports such as parking hubs and bicycle racks).a 1. National Transportation Enhancements Clearinghouse (http://www.enhancements.org). Includes only data on funding spent under the federal Transportation Enhancements Program.c
2. Percent of counties or municipalities with policies requiring sidewalks in all new and redeveloped residential and mixed-use communities. 2. No data source found.
3. Percent of counties or municipalities with policies that promote recreation facilities (e.g., bikeways, parks, fields, gyms, pools, tennis courts, and playgrounds) in new and redeveloped residential and mixed-use communities. 3. No data source found.
4. State policies and percent of counties or municipalities with policies and strategic plans to promote bicycle use for transportation purposes. 4. No data source found.
5. Percent of low-fat milk sales in the state (1= or less). 5. No data source found. Regional milk production data are available but do not reflect state sales.
6. Number of farmers' markets per capita in the state.a 6. U.S. Dept. of Agriculture Farmers' Market database (http://www.ams.usda.gov/ farmersmarkets/). Incomplete due to inconsistent reporting and definition of farmers' markets across states.c
7. State with laws on smoke-free indoor air that prohibit smoking or limit it to separately ventilated areas in restaurants, day care centers, and other public places.b 7. State Tobacco Activities Tracking and Evaluation (STATE) System (http://www2a.cdc.gov/ nccdphp/osh/state/).d
8. Proportion of smokers who report that smoking is not allowed anywhere inside their homes. 8. Behavioral Risk Factor Surveillance System (BRFSS), optional Tobacco Indicators module (http://www.cdc.gov/brfss).d
a 2 indicators (25%) lack specificity (ambiguous, lack precision).
b 1 indicator (12%) lacks sensitivity (unable to measure incremental change, measured at inappropriate level).
c 2 indicators (25%) have data sources that partially measure indicator.
d 2 indicators (25%) have adequate data sources.
Table 3 Pilot Indicators and Data Sources for Heart Disease and Stroke Prevention, School Setting, South Carolina and Alabama, 2001
Indicator Data Sources and Comments
1. State policies that require daily physical education or its equivalent in minutes per week, for all students in K–12, with no substitution of other courses or activities for physical education.a 1. School Health Policy and Programs Study (SHPPS) (www.cdc.gov/ nccdphp/dash/shpps).c
2. State policies that require schools to assess students on the knowledge and skills specified by the state's physical education standards, frameworks, or guidelines.a 2. SHPPS.c
3. State policies requiring that the foods and beverages available at schools outside of school meal programs reinforce the principles of the Dietary Guidelines for Americans (9).a 3. SHPPS.c
4. State policies that require newly hired school food service managers to have a nutrition-related baccalaureate or graduate degree and certification/credentialing in food service from either the state or the American School Food Service Association.a 4. SHPPS.c
5. State policies that require all newly hired staff who teach physical education to be certified, licensed, or endorsed by the state to teach physical education.a 5. SHPPS.c
6. State policies that require all newly hired staff who teach health education to be certified, licensed, or endorsed by the state to teach health education.a 6. SHPPS.c
7. States policies that require schools to assess students on the knowledge and skills specified by the state's health education standards, frameworks, or guidelines.a 7. SHPPS.c
8. Percent of schools that provide health education instruction that includes the physical education, nutrition, and tobacco use prevention topics listed in School Health Index (10). 8. No data source found. Questions from School Health Index could be useful for surveillance, if survey mechanism is developed.
9. Proportion of schools with School Health Councils.b 9. School Health Education Profile (SHEP) (http://www.cdc.gov/nccdphp/dash/profiles). SHEP is completed by sample of principals and lead health educators in schools having at least one of the grades 6–12. No data source available for elementary schools.d
10. Proportion of schools that have adopted tobacco-free school policies that meet CDC recommendations.b 10. SHEP. See 9 above. SHEP does not include questions to thoroughly assess if tobacco policies meet recommendations.d
a 7 indicators (70%) lack sensitivity (unable to measure incremental change, measured at inappropriate level).
b 2 indicators (20%) lack specificity (ambiguous, lack precision).
c 7 indicators (70%) have adequate data sources.
d 2 indicators (20%) have data source that could partially measure indicator
Table 4 Pilot Indicators and Data Sources for Heart Disease and Stroke Prevention, Work Site Setting, South Carolina and Alabama, 2001
Indicator Data Sources and Comments
1. Percent of work sites that have policies supporting the engagement of all employees in physical activity during work time (e.g., flexible scheduling, relaxed dress codes). 1. No data source found.
2. Percent of work sites that provide showers and changing facilities to support physically active employees. 2. No data source found.
3. Percent of work sites that provide and promote on-going, on-site employee physical activity programs (e.g., walking, stretching, aerobics) during the previous 24 months. 3. No data source found. National Worksite Health Promotion Survey measures this indicator at the national level, but the sample is too small for state analysis.
4. Percent of work sites with vending machines and/or snack bars that offer heart-healthy food and beverage choices, including water or flavored water, 1% or less milk products, 100% juice products, fruits, vegetables, and products labeled low or reduced calorie, low or reduced sodium, and those labeled 3 grams or less of fat per serving. 4. No data source found.
5. Percent of work sites with cafeterias that offer heart-healthy food and beverage choices including water or flavored water, 1% or less milk products, 100% juice products, fruits, vegetables, and products labeled low or reduced calorie, low or reduced sodium, and those labeled 3 grams or less of fat per serving. 5. No data source found.
6. Percent of work sites that offer nutrition or weight management classes or counseling.a 6. No data source found. National Worksite Health Promotion Survey measures this indicator at the national level, but the sample is too small for state analysis.
7. States with laws on smoke-free indoor air that prohibit smoking or limit it to separately ventilated areas in government and private work sites.b 7. State Tobacco Activities Tracking and Evaluation System (STATE) (http://www2a.cdc.gov/nccdphp/osh/state/").c
8. Proportion of work sites (segmented by number of employees) that cover smoking cessation programs.a 8. No data source found.
a Two indicators (25%) lack specificity (ambiguous, lack precision).
b One indicator (12%) lacks sensitivity (unable to measure incremental change, measured at inappropriate level).
c One indicator (12%) has adequate data source.
Table 5 Pilot Indicators and Data Sources for Heart Disease and Stroke Prevention, Health Care Setting, South Carolina and Alabama, 2001
Indicator Data Sources and Comments
1. Percent of managed care organizations that adopt a policy to incorporate nationally accredited guidelines (e.g., the AHA Guide to Primary Prevention of Cardiovascular Diseases (17)) as part of their standard care package.a 1. No data source found.
2. Percent of managed care organizations that adopt a policy to incorporate nationally accredited guidelines (e.g., the AHA Guide to Comprehensive Risk Reduction for Patients with Coronary and other Vascular Disease (18)) as part of their standard care package.a 2. No data source found.
3. Percent of managed care organizations (e.g., health maintenance organizations, independent provider organizations, and preferred provider organizations) that have policies or guidelines to routinely provide or reimburse for assessments and counseling for physical activity, medical nutrition therapy, and tobacco cessation to plan members as part of their standard care package, according to the Guide to Clinical Preventive Services (19).a 3. No data source found.
4. Percent of health insurance plans that have policies or guidelines to routinely provide or reimburse for assessments and counseling for physical activity, medical nutrition therapy, and tobacco cessation to plan members as a covered benefit, according to the Guide to Clinical Preventive Services (19).a 4. No data source found.
5. Proportion of current and recent smokers who received advice to quit smoking from a health professional. 5. Behavioral Risk Factor Surveillance System (BRFSS), optional Tobacco Indicators module (http://www.cdc.gov/brfss).b
a 4 indicators (80%) lack specificity (ambiguous, lack precision).
b 1 indicator (10%) has adequate data source.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Pluto DM, Phillips MM, Matson-Koffman D, Shepard DM, Raczynski JM, Brownstein JN. Policy and environmental indicators for heart disease and stroke prevention: data sources in two states. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0019.htm.
==== Refs
1 Cheadle A Wagner E Koepsell T Kristal A Patrick D Environmental indicators: a tool for evaluating community-based health-promotion programs Am J Prev Med 1992 Nov–Dec 8 6 345 350 1482574
2 Glanz K Lankenau B Foerster S Temple S Mullis R Schmid T Environmental and policy approaches to cardiovascular disease prevention through nutrition: opportunities for state and local action Health Educ Q 1995 11 22 4 512 527 8550374
3 Cheadle A Sterling TD Schmid TL Fawcett SB Promising community-level indicators for evaluating cardiovascular health-promotion programs Health Educ Res 2000 2 15 1 109 116 10788197
4 Fishman JA Allison H Knowles SB Fishburn BA Woollery TA Marx WT State laws on tobacco control — United States, 1998 MMWR Surveill Summ 1999 6 25 48 3 21 40
5 State Tobacco Activities Tracking & Evaluation (STATE) System [Internet] U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Promotion and Health Prevention, Office on Smoking and Health Atlanta (GA) 2002 cited 2004 January 29 Available from: URL:http://www2.cdc.gov/nccdphp/osh/state
6 Centers for Disease Control and Prevention. Behavioral Risk Factor Surveillance System survey questionnaire U.S. Department of Health and Human Services, Centers for Disease Control and Prevention Atlanta (GA) 2002
7 National Transportation Enhancements Clearinghouse. TE Projects [database on the Internet]. Rails-to-Trails Conservancy Washington (DC) 2002 Available from: URL:http://www.enhancements.org
8 AMS farmers markets [database on the Internet] U.S. Department of Agriculture Washington (DC) 2002 Available from: URL:http://www.ams.usda.gov/farmersmarkets/index.htm
9 Nutrition and your health: dietary guidelines for Americans Home and Garden Bulletin U.S. Department of Agriculture USDA Washington (DC) 2000 232 5th Available from: URL:http://www.health.gov/dietaryguidelines/dga2000/ document/frontcover.htm
10 School Health Index for physical activity, healthy eating, and a tobacco-free lifestyle [Internet] Atlanta (GA) U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Division of Adolescent and School Health cited 2004 Jan 29 Available from: URL:http://www.cdc.gov/nccdphp/dash/shi
11 SHPPS: School Health Policies and Programs Study [Internet] Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Division of Adolescent and School Health Atlanta (GA) 2002 cited 2004 Jan 29 Available from: URL:http://www.cdc.gov/nccdphp/dash/shpps
12 Grunbaum JA Kann L Williams BI Kinchen SA Collins JL Baumler ER Surveillance for characteristics of health education among secondary schools — school health education profiles, 1998 MMWR Surveill Summ 2000 8 18 49 8 iv 41
13 1999 National Worksite Health Promotion Survey Association for Worksite Health Promotion Association for Worksite Health Promotion Northbrook (IL) 1999
14 Tracking healthy people 2010 U.S. Department of Health and Human Servicesi U.S. Government Printing Office Washington (DC) 2000 996
15 Golaszewski T Fisher B Heart check: the development and evolution of an organizational heart health assessment Am J Health Promot 2002 Nov–Dec 17 2 132 153 12471865
16 Oldenburg B Sallis JF Harris D Owen N Checklist of Health Promotion Environments at Worksites (CHEW): development and measurement characteristics Am J Health Promot 2002 May–Jun 16 5 288 299 12053440
17 Pearson TA Blair SN Daniels SR Eckel RH Fair JM Fortmann SP AHA guidelines for primary prevention of cardiovascular disease and stroke: 2002 update Circulation 2002 7 16 106 3 388 391 12119259
18 Smith SC Blair SN Bonow RO Brass LM Cerqueira MD Dracup K AHA/ACC Guidelines for preventing heart attack and death in patients with atherosclerotic cardiovascular disease: 2001 update Circulation 2001 9 25 104 13 1577 1579 11571256
19 U.S. Preventive Services Task Force Guide to clinical preventive services, 3rd ed, periodic updates. Vol 2: Chemoprevention and counseling 3rd 2003 Rockville (MD) U.S. Department of Health and Human Services, Agency for Healthcare Research and Quality
20 2002 school health education profile: school principal questionnaire Centers for Disease Control and Prevention U.S. Government Printing Office Washington (DC) 2002
21 2002 school health education profile: lead health educator questionnaire Centers for Disease Control and Prevention U.S. Government Printing Office Washington (DC) 2002
22 State legislative information [database on the Internet]. Available from: URL: http://apps.nccd.cdc.gov/DNPALeg/ Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Division of Nutrition and Physical Activity Atlanta (GA) 2003 cited 2004 Jan 29
23 State health facts online [database onthe Internet] Henry J. Kaiser Family Foundation Menlo Park (CA) 2001 Available from: URL:http://www.statehealthfacts.kff.org
24 A public health action plan to prevent heart disease and stroke: executive summary and overview U.S. Department of Health and Human Services, Centers for Disease Control and Prevention Atlanta (GA) 2003
|
15663881
|
PMC1183497
|
CC BY
|
2022-01-26 23:15:43
|
no
|
Prev Chronic Dis. 2004 Mar 15; 1(2):A05
|
utf-8
|
Prev Chronic Dis
| 2,004 |
nan
|
oa_comm
|
==== Front
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv12_03_0017
Original Research
PEER REVIEWEDSocial and Cultural Barriers to Diabetes Prevention in Oklahoma American Indian Women
Keim Kathryn S. PhD, RD, LD Department of Nutritional Sciences, Oklahoma State University
301 Human Environmental Services, Department of Nutritional Sciences, Oklahoma State University, Stillwater, OK 74078. Oklahoma State University 405-744-8293 [email protected]
Taylor Christopher MS, RD Department of Nutritional Services, Oklahoma State University
Sparrer Alicia MS, RD, LD Harris Methodist Fort Worth Hospital
Van Delinder Jean PhD Department of Sociology, Oklahoma State University
Parker Stephany PhD Cooperative Extension Service, Oklahoma State University
4 2004
15 3 2004
1 2 A062004
Introduction
The prevalence of diabetes is disproportionately higher among minority populations, especially American Indians. Prevention or delay of diabetes in this population would improve quality of life and reduce health care costs. Identifying cultural definitions of health and diabetes is critically important to developing effective diabetes prevention programs.
Methods
In-home qualitative interviews were conducted with 79 American Indian women from 3 tribal clinics in northeast Oklahoma to identify a cultural definition of health and diabetes. Grounded theory was used to analyze verbatim transcripts.
Results
The women interviewed defined health in terms of physical functionality and absence of disease, with family members and friends serving as treatment promoters. Conversely, the women considered their overall health to be a personal issue addressed individually without burdening others. The women presented a fatalistic view of diabetes, regarding the disease as an inevitable event that destroys health and ultimately results in death.
Conclusion
Further understanding of the perceptions of health in at-risk populations will aid in developing diabetes prevention programs.
==== Body
Introduction
The American Indian people and culture have sustained serious hardships throughout the last 2 centuries; their greatest struggle, however, may be impending. The rate of diabetes is disproportionately higher in minority populations, especially the American Indian population (1-4). Indian Health Service (IHS) national outpatient data indicate that the age-adjusted prevalence rate of diabetes among American Indians is an estimated 88.7 per 1000 for individuals older than 15 years (5). In the Oklahoma City area, the largest of IHS areas, the age-adjusted prevalence rate of diabetes is 60 per 1000 individuals (3), indicating that American Indians are 2.43 times more likely to have diabetes than the general population at 39 per 1000 individuals (6). Furthermore, national data indicate age-adjusted prevalence rates are greater for American Indian women (12.0%) compared to American Indian men (9.7%) (3). Lee et al observed in an Oklahoma sample that 38% of American Indian men had diabetes compared with 42% of American Indian women (7).
Diabetes is a multifaceted disease that is reaching epidemic proportions in the American Indian community (1). If diabetes could be prevented or delayed in this population, the benefits in quality of life and health care cost savings would be considerable. Rhoades et al estimated 882 years of productive life lost due to diabetes mellitus over a 3-year period among American Indians receiving health care services from IHS (8). Diabetes results in compromises to longevity and quality of life and in economic disadvantages. Health care costs for treatment of non-American Indian patients with diabetes in 1994 were 2.4 times greater than non-American Indian controls, with long-term complications accounting for 38% of the costs (9). Through a Monte Carlo study based on American patients with diabetes, intensive blood sugar control was estimated to produce a 3% reduction in health care costs over 30 years (10). Additionally, Oklahoma Behavioral Risk Factor Surveillance System data demonstrated a significantly greater number of days of disability and poor physical health for patients with diabetes compared to control subjects without diabetes (11). These data have obvious ramifications for workplace productivity. Success at delaying or preventing the onset of diabetes will reduce the costs of diabetes treatment and prolong an individual's potential to be a contributing member of the economy.
A greater understanding of American Indian perceptions of health and diabetes is paramount to the success of diabetes prevention programs among these populations (12-15). Perceptions of the inevitability of diabetes within the reservation environment have been reported (16-18). Perceptions of health among American Indian elders in an urban setting have also been presented (19). Data is lacking on the relationship of diabetes to health and the social environment as well as the perception of the feasibility of diabetes prevention. This study used in-depth qualitative interviews to ascertain a cultural definition of health and diabetes from American Indian women residing outside a reservation setting. The information learned will be used to plan culturally appropriate nutrition education and health promotion programs aimed at preventing or delaying the onset of diabetes among American Indians in Oklahoma.
Methods
The data contained herein represent a portion of a larger study that involved a series of 3 sessions with each study participant. The first session included demonstration of informed consent, completion of a demographic questionnaire and a rank-order assessment of life concerns, and training for a 4-day weighed-food record collection. During the second interview, the participants responded verbally to questions from the Cultural Structure of Health and Diabetes questioning guide (Table 1) and completed a free food sort of previously determined most commonly consumed foods. The food sort allowed participants to group foods based on their own classifications. The final session included a weight valuation interview to identify the cultural perceptions of body image and a trichotomous food sort of the most commonly consumed foods. In this sort, participants sorted food into groups based on their perceptions of health value and fat and sugar content. The research protocol was reviewed and approved by the Institutional Review Board at Oklahoma State University and the executive counsels for the cooperating tribes.
Interviewer training
Five female American Indian interviewers were hired to conduct the in-depth interviews. Each interviewer completed a one-day course on subject recruitment, interview structure, data collection techniques, and response recording. The training consisted of equipment usage, essential techniques of qualitative interviewing (listening and directive questioning skills, for example) and the logistics of the qualitative interviews. The interviewers were compensated $100 for training and $125 for each participant who completed 3 interviews.
Participants
Women of at least one quarter American Indian blood, between ages 18 and 65 years, who were not pregnant or lactating, were eligible for the study. A diagnosis of chronic disease, including diabetes, did not prevent inclusion; however, women diagnosed with chronic diseases that have an impact on appetite (including women receiving cancer treatment) were excluded from the study. Women were recruited proportionately from tribal health clinics in northeast Oklahoma using a non-probablility sampling design. To increase participation rates, women who successfully completed the interview process received $125.
Key informants and the American Indian interviewers at each of the clinics recruited potential subjects for the research study. Articles were published in tribal newsletters and newspapers to promote the study. Women interested in participating were referred to one of the interviewers to receive more information, determine eligibility, and schedule interviews. Additional subjects were recruited from tribal diabetes education programs and the 3 tribal general health clinics.
Data collection and analysis
This study reports results of the interviews during the second session, in which participants responded verbally to questions from the Cultural Structure of Health and Diabetes questioning guide (Table 1). Questions from previous research (20) were modified to identify cultural perceptions of health and diabetes. Questions focused on areas of interest that were consistent with the objectives of the study, such as perceived causes, treatments, and efficacy of diabetes prevention behaviors. Key informants within each clinic reviewed the questions for cultural sensitivity prior to their administration. Results based on each interviewer's session with the first participant served as a pilot; responses were analyzed as the data became available and appropriate changes were made to the questioning guide.
Two researchers analyzed the verbatim transcripts from the audiotapes during data collection. Grounded theory guided analysis of the transcripts (21). An initial list of code words was derived from recurring themes in the transcripts (Table 2). Then, key concepts or recurring themes derived from the qualitative interviews were integrated into the questioning guide using the method of constant comparisons. The transcripts were reviewed throughout the interviewing process. Code word definitions were drafted to encompass the meaning of text segments. When new themes recurred in the transcripts, they were either assigned a new code word or a subcategory of an existing code word. Furthermore, the questioning guide was modified to capture more detail about the emerging themes. Text segments were coded with the corresponding code words using Ethnograph (version 5.04, Qualis Research Associates, Denver, CO). Following open coding, axial coding was used to identify subcategories with code words (21). The final step, selective coding, provided the means to assess the relationship among constructs and to assess how concepts were related to their constructs to establish an overall phenomenon.
Results
Eighty-one American Indian women completed the qualitative interviews. Two transcripts were not available because of technical failure of the recording devices, resulting in 79 usable interviews. Demographic characteristics of the sample are provided in Table 3. The mean age of the women was 43 ± 11 years while mean degree of American Indian blood was 65%. Though the sample was collected from 3 tribal health clinics, 16 different tribal affiliations were reported, making analysis by tribe impractical. Of the 79 women, 26 (33%) reported a previous clinical diagnosis of diabetes. Approximately 70% reported education beyond high school; however, 72% indicated an annual household income of less than $25,000.
Twenty-nine unique code words were developed during the open coding of the transcripts (Table 2). Text segments coded for each code word were then analyzed to establish subcategories and relations among code words. The results of the analysis for code words associated with health and diabetes are presented below.
Cultural definition of health
The American Indian women who took part in this study defined health predominantly in terms of lifestyle behaviors. Individuals performing positive behaviors — such as consuming a "healthy" diet, exercising, and not smoking — were considered healthier than those who did not. Being overweight was also considered to reflect negatively on health status.
Health was also defined in terms of the presence or absence of disease. For example, when individuals were asked to define their current health, they sometimes mentioned the presence or absence of several chronic conditions, including arthritis, diabetes, heart disease, and cancer. In the absence of a chronic disease, individuals considered themselves to be healthy. Even if clinically diagnosed with disease, individuals did not perceive diminished health until there was a physical feeling of illness. Until an individual perceived a feeling of illness, they considered their health to be satisfactory. One woman said, "I haven't been throwing-up sick in years, but a little cold here and there." This was especially true of diabetes, as the women interviewed did not consider the disease to be severe until it was manifested through long-term complications.
Another indicator of health status was defined through physical functionality. The women considered poor health to be an impairment of one's ability to perform daily tasks: "Oh, my current health. I feel like I'm pretty healthy. I can still lift up things and get around." The women viewed being healthy as having the capacity and energy to perform daily tasks and other activities. However, certain accommodations were made for age. Furthermore, the women expected health to decline with age; many defined their health status according to expectations for their current age. One woman described feeling "not too good about my health and myself. It seems like I've been more tired. But I guess that's just this age."
Cultural definition of diabetes
Diabetes was defined most commonly in terms of long-term complications, which were often tied to fear and concern. The most frequently noted complication was amputation, expressed by one woman as "becoming a member of the stub club." Some women were confused about diabetes and its symptomology and long-term complications. Many women were unclear about long-term complications; some women said that dialysis and blindness were symptoms of diabetes. Confusion about hyperglycemia and hypoglycemia — and which one indicated diabetes — also existed. The women expressed the belief that hypoglycemia is an early symptom of diabetes that later converts to hyperglycemia.
Similarly, others expressed a fear of diabetes, calling it a "scary disease." Diabetes was portrayed as devastating. As one participant said, "It ruins your health, and ultimately it will kill you." Furthermore, diabetes was considered a malicious disease. One woman stated: "Diabetes is scary. It's a scary process. It's demeaning. I think it is a very, very cruel breakdown of your system." The perception existed that a body being "out of balance" causes diabetes, and an error in the inner workings of the body results in a blood sugar imbalance.
"Fatalism" (16) toward diabetes and its complications was a strong theme among the women. One woman said, "I knew it was going to happen, but when it did happen, it was a surprise to me. And I felt like I was doomed." The women interviewed expressed the concern that being of American Indian descent leads to a belief in increased susceptibility to diabetes as well as a belief in the inevitability of getting diabetes. Furthermore, the women feared having diabetes for an extended period of time without being diagnosed. The American Indian social network also fostered apprehension about diabetes, as most of the interview participants knew someone with the disease.
Another prominent concern among the American Indian women was the possibility of their own or a family member's diagnosis of diabetes. Interestingly, the women were more concerned about their children being diagnosed with diabetes than about their own possible diagnosis. Their statements about children being at risk reflected an overall concern for children developing diabetes. The women were also concerned about other family members, including siblings, spouses, and parents.
The women expressed the idea that after an individual is diagnosed with diabetes, his or her lifestyle behaviors must change. Diabetes was perceived to require thorough, demanding care. Appropriate care involved eating right, taking medicine, and doing "what the doctor tells you to do." The women regarded diabetes care and behavior change as solely the responsibility of the individual.
Diabetes prevention
When asked if it was possible to prevent diabetes, many of the women responded in terms of personal behaviors that may prevent or help delay the onset of diabetes. These responses centered on changing behaviors that cause diabetes, such as eating a poor diet and not exercising. To explore those responses, we asked further questions about when potential preventative behaviors should begin, and a portion of the respondents indicated the need to reach young children. Other participants with a more fatalistic view of diabetes suggested that diabetes was inevitable in individuals with a strong family history of the disease.
Barriers to diabetes prevention and treatment
Some interview participants indicated that frequent visits to their health care professionals represented an appropriate method of diabetes prevention. Furthermore, the women perceived diabetes screening as a method of diabetes prevention in the absence of changing lifestyle factors. Issues of denial and avoidance of diagnosis were also strong, providing an additional challenge to diabetes prevention and treatment. Despite efforts to increase public awareness and opportunities for diabetes screening, women still avoided screening. Because an individual was considered to be in good health in the absence of physically feeling ill or the clinical diagnosis of a chronic disease, avoiding a visit to a health care professional (thus avoiding a screening) freed the individual from diagnosis and evaded the need for self-care — despite a personal suspicion of having the disease. One woman mentioned "[t]here might be a tendency for people to suspect it but not want to have it confirmed maybe." In such situations, care for diabetes is delayed and the likelihood of long-term complications increases.
Furthermore, individuals did not express personal concern about diabetes until they were themselves facing diagnosis. If a positive diagnosis was made, those women expressed a strong sense of denial. One participant mentioned a family member who was "in denial, and won't go to the doctor, and then it gets worse, and then they'll go after it starts getting too bad." Individuals often postponed care until they perceived a physical ailment, likely indicative of long-term complications.
Supporting social structure
The women mentioned many sources of social support and information. They cited health care professionals as only one of many sources of information about health and diabetes. Community and family members served as considerable sources of both information and misinformation. Misconceptions ranged from the idea that individuals with diabetes are forced into strict dietary modifications with a complete absence of sugar to the idea that diabetes can be "gotten rid of, if you take care [of yourself]." The women obtained much information about diabetes prevention, symptoms, and treatment from discussions with — or observation of the treatments received by — immediate or extended family and friends. Shared knowledge within these circles does not reflect the current state of diabetes care, but defers to an older pedagogy of diabetes care.
In addition to serving as sources of information, families were portrayed as mediators of health self-care. Many self-care concerns are rooted in the women's family caregiver roles, especially as gatekeepers of healthy meals. Their roles are challenged by having to make personal lifestyle behavior changes. For example, American Indian women are responsible for providing meals that satisfy the entire family. If their health requires dietary changes, they find it unacceptable to put their needs above the wants or needs of the family unit, greatly reducing the likelihood of behavior modification.
When asked how the family could aid in diabetes prevention efforts, familial and parental support was most commonly reported. Family discussions about health and diabetes as well as family attendance of educational sessions were indicated as methods of family involvement in diabetes prevention. However, one woman indicated that when she suspected she might have diabetes, her family discouraged screening because they thought it unlikely she would be diagnosed with the condition. This demonstrates both the positive and negative social environment affecting diabetes prevention and treatment.
Discussion
The qualitative method used in our study demonstrates an attempt to obtain a cultural definition of health and diabetes from American Indian women. The pervasiveness of diabetes was readily apparent: most participants had at least one family member or friend with diabetes. Although one third of the participants had diabetes, the responses received from those without diabetes mirrored the responses of those with diabetes. An analysis of responses stratified by diagnosis of diabetes would provide little additional information.
Many factors — historical, political, sociocultural, and geographical — impact health perceptions among American Indians (19). Challenges abound in trying to define health as American Indians perceive it, especially through the lens of Western medicine. A gap exists between the discernment of a biologically defined chronic disease and the more culturally relevant presence of physical symptoms; this gap presents a strong barrier to accurate assessment of personal health status (18,22,23). In one study of Diné (Navajo) families with asthmatic children, asthma is perceived by the families as a series of individual episodic reactions requiring attention instead of an underlying physiological chronic inflammatory condition (23); the findings of the Diné study agree with our findings. Hatton reported that elderly American Indians define their health in terms of the absence or presence of various chronic diseases (19). These results also concur with perceptions found in our sample. Also in the Hatton report, the capacity of individuals to perform activities of daily living and take care of themselves was regarded as an important aspect of personal health assessment (19); this capacity was deemed important by our study participants as well.
Of particular interest was the mutual dependency of the cultural definitions of health and diabetes. The women in our study held the belief that being unhealthy was discernable by physically feeling ill. Interviews with older American Indians residing in urban areas considered themselves to be healthy in the absence of any outward, perceivable sign of illness (19). The disjointed impression among our respondents that long-term complications are symptoms of disease (instead of the consequences of poor diabetes control) may be explained by the perception that one is not unhealthy unless a perceptible feeling of illness is present. When our study participants faced a clinical diagnosis of diabetes, they delayed self-care until long-term complications — accompanied by a decrease in physical function — became evident. To these women, long-term complications serve as the only tangible evidence of illness. It is this strong reliance on physical symptomology that provides a great obstacle to diabetes prevention, screening, and care.
A strong sense of inevitability pervaded the many ideas surrounding the pursuit of health and prevention of diabetes among our American Indian sample. Many, but not all, participants believed that diabetes is inevitable and ultimately leads to death, especially for individuals who have strong family histories of the disease. Previous research with the Gila River Indian Community describes these feelings of inevitability as "fatalistic" attitudes that moderate the perception of diabetes prevention and may serve as additional barriers to adopting prevention behaviors (16). Kozak reported an overall sense of surrender to diabetes, which was viewed as an inevitable, uncontrollable disease that resulted in death (16). Additionally, Judkins reported "highly fatalistic attitudes and verbalizations" about diabetes among the Seneca, accompanied by a feeling of powerlessness against the disease (17). It has been theorized that fatalism has developed as a social coping mechanism to deal with the severity of the diabetes epidemic and the resulting compromised quality of life (16). Compensatory mechanisms built into cultural personality to deal with environmental and personal stress may precipitate denial or avoidance behaviors (17). A sense of inevitability may ultimately result in a decreased propensity to take necessary steps for disease prevention, which is often misconstrued by the administrators of Western medicine as non-compliance (16,18).
Additional barriers were evident in the prevention and treatment of diabetes in these American Indian women. Family dynamics play a critical role in health care in American Indians (22). With a shift from traditional economic strategies to mainstream business practices, traditional American Indian families are shifting toward more Western nuclear families, which has an impact on family dynamics (22). Additionally, family resistance to alterations in dietary habits serves as an additional barrier to diabetes prevention and care in American Indian women. To achieve successful behavior change, nutrition education and diabetes prevention programming must involve the family unit. To what extent will family obligations or positive social support structures within Native American communities allow self-care behaviors? How receptive are American Indian individuals to external support, and what is their capacity to overcome barriers for health promotion? These questions — as yet unanswered — require more research.
In addition to conflicts between healthy lifestyle behaviors and family obligations, avoidance of diabetes screening serves as an additional barrier to diabetes treatment. The women expressed an inclination to avoid screening even if they harbored suspicions of having the disease. Many diagnoses were reported while women were seeking medical treatment for unrelated reasons. If a clinical diagnosis is made, denial is likely, especially when no physical symptoms are apparent. Similarly, Huttlinger et al reported a case of a Diné woman who was taken to the doctor for a routine check-up against her will and subsequently diagnosed with diabetes (18). Though vehemently claiming she felt fine, she had to undergo amputation because of the serious progression of her uncontrolled diabetes. This case demonstrates the family's role in encouraging women to seek care and how the lack of the physical signs of disease can hinder treatment.
There are several limitations of the current study. First, American Indian women were hired from local American Indian communities to conduct the interviews, regardless of previous experience in qualitative interviewing techniques. Despite training in such techniques, they varied in the amount they probed interview participants on topics important to the research team. To address this concern, transcript reviewers analyzed interview tapes soon after retrieval and provided feedback to interviewers as additional training and guidance. Second, because transcript reviewers functioned as the research instrument, the lens through which reviewers read the transcripts provided bias. To address this concern, transcripts were read by the 2 researchers independently from each other and then discussed until consensus was reached on coding. We achieved an inter-rater reliability of more than 90%. Furthermore, the lack of responses related to traditional healing practices and the role of spirituality may have been due to the recruitment of participants through tribal health clinics and is not likely representative of all American Indian cultures. Though snowball sampling aided in recruitment, participants recruited from the health clinics may have been more likely to seek medical treatment through the health clinics than through traditional healing practices. Finally, the sample was derived through non-probability methods. Though these methods may decrease the generalizability of the findings, they are often needed to identify individuals from an at-risk population (19).
Despite these limitations, the congruency of the data to other reports of perceptions of diabetes among other American Indian groups provides support for our findings (19). Though results similar to ours have been reported, they were derived from reservation-living American Indian groups; we have identified perceptions of health and diabetes among a sample population outside the reservation setting. Our findings indicate a more comprehensive approach to the underlying issues in health promotion and diabetes prevention than previous reports. Previous reports did not address the interrelationships of perceptions of health nor did they discuss issues of diabetes prevention. We have attempted to address some of these issues; however, each new issue presents new unanswered questions, indicating a need for further investigation of the cultural definitions of health and diabetes.
Efforts to identify disparities in health perceptions and worldviews are essential for developing nutrition education interventions that precipitate behavior change (24-26). Previous research and multi-site programs, including Awakening the Spirit: Pathways to Diabetes Prevention & Control (American Diabetes Association) and the National Diabetes Prevention Program (National Institute of Diabetes, Digestive and Kidney Diseases), have demonstrated improved diabetes prevention and treatment by targeting specific lifestyle behaviors within the context of American Indian communities (27-29). American Indian communities vary widely in tribal affiliation and location; future researchers must identify the characteristics of each American Indian population studied to ensure they meet the community's specific needs (30). The importance of solid formative data on a population is paramount, especially considering that a large portion of research is conducted on reservations. Furthermore, the extent to which the perceptions held by American Indian women in Northeast Oklahoma are congruent with other American Indians within and outside of Oklahoma needs to be examined to assist in designing effective education programs.
Funding for this research was provided by the Okalahoma Center for the Advancement of Science and Technology and the Dean's Incentive Fund at Oklahoma State University. We thank the tribal health clinics and interviewers for their support and guidance in the research process.
Figures and Tables
Table 1 Cultural Structure of Health and Diabetes: Questioning Guide for Interviewing Oklahoma American Indian Women about Cultural Perceptions of Health and Diabetes, 2003
1. Describe your current health to me.
2. Describe how you feel about your health.
3. What, if any, health concerns do you have?
4. What are the major health concerns of other Indian women you know?
5. What do you think is the leading cause of death for Indian women in the United States?
6. What comes to mind when I mention diabetes?
7. Let's discuss diabetes a bit.
What do you think causes a person to get diabetes?
Why do you think these things (mentioned above) cause diabetes?
If eating right, describe how people should eat.
What keeps people from eating right?
If exercise, what should they do and how often?
What keeps people from getting exercise?
How did you find the information that you just told me?
8. What do you think happens to a woman once she develops diabetes?
9. Can you think of anyone who is at risk for developing diabetes? (Is he or she Indian?)
10. How can a person tell if he or she has diabetes? How do they feel?
11. Tell me about anything that you know of that might keep a woman from developing diabetes.
Why do you think these things (mentioned above) prevent diabetes?
Where did you find this information?
If read, where? Books, magazines (which ones)?
If heard, where? From whom?
12. What may prevent a woman from doing the things that may prevent diabetes?
13. What treatments are there for diabetes that you know about? If diet, describe the diet.
14. Who are you concerned about developing diabetes?
What are the reasons that you are concerned about this person(s)?
15. What can parents or family do to help prevent this person/child from developing diabetes?
16. What can the tribe or community do to help prevent this person/child from developing diabetes?
17. How do you feel about diabetes?
18. What is your greatest fear about diabetes?
19. What control do you think a person has over diabetes?
20. Can you prevent diabetes?
When can a person begin to do these things to prevent diabetes?
21. How would you describe a traditional (Indian) diet (the old way of eating)?
What would you think of shifting the diet back toward the old ways Indians used to eat?
Do you think eating a more traditional diet would help Indians prevent diabetes?
22. Is there anything else would you like to tell me about diabetes?
Table 2 Parent Code Words Used in Initial Coding of Verbatim Transcripts of Interviews of Oklahoma American Indian Women About Cultural Perceptions of Health and Diabetes, 2003
Code Word Definition
AT RISK Characteristics of those at-risk for developing diabetes
AVOIDANCE Avoiding screening or seeking treatment for ailment for fear of diabetes diagnosis
AWARENESS Lack of or increased awareness of diabetes
BAR DM Barriers in controlling diabetes
BAR EXER Perceived barriers preventing exercise
BAR FOOD Perceived barrier to eating a healthy diet
CONTROL Methods used to control diabetes
DEF DIET Cultural definition of a healthy diet
DEF DM Cultural definition of diabetes
DEF EXER Cultural definition of exercise
DEF HEALTH Cultural definition of health
DENIAL Denial experienced post-diagnosis of diabetes
DM CAUSES Perceived causes of diabetes
DM CONCERN Concerns and fear about diabetes
DM DIET Perceived diabetic diet and dietary changes required by diabetes diagnosis
DM LT COMP Perceived long-term complications of diabetes
DM PREVENT Methods to prevent or delay onset of diabetes
DM SYMPTOM Perceived symptoms of diabetes onset
DM TREAT Perceived treatments for diabetes
FAMILY Role of family in diabetes prevention and treatment
HEALTHCARE Issues in quality and continuity of health care
MEN HEALTH Perceptions of men and health
NA DEATH Perceived leading causes of death for American Indian women
NA WOMEN Perceived health issues of other American Indian women
PERSONAL Personal health concerns
SOCIAL Social aspects of diabetes care and prevention
SOURCE Sources of health and nutrition knowledge
TRAD DIET Cultural definition of a "traditional diet"
TRIBE Role of tribe/community in diabetes prevention and treatment
Table 3 Demographic Characteristics of Oklahoma American Indian Women Interviewed About Cultural Perceptions of Health and Diabetes, 2003
Characteristica
Age (years) 43.4 ± 11.4
Degree of Indian blood (%) 64.6 ± 0.3
Body mass indexb(kg/m2) 32.1 ± 6.9
N %
Marital statusc
Married 30 38.5
Not married 48 61.5
Education
High school or less 24 7.6
Some college 44 55.7
College degree 11 13.9
Employment
Employed 52 65.8
Not employed 27 34.2
Annual household incomec
< $15,000 36 46.2
$15,000-$24,000 20 25.6
>;$25,000 22 27.2
BMI categories (kg/m2)b,c
Healthy (18.5-24.9) 9 12.2
Overweight (25.0-29.9) 21 28.4
Obese (>30) 44 59.5
Diagnosed Diabetesc
Yes 26 32.9
No 50 63.3
a Presented as mean ± SD.
b Based on self-reported height and weight.
c Does not sum to n = 79 due to missing data.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Taylor C, Keim KS, Sparrer A, Van Delinder J, Parker S. Social and cultural barriers to diabetes prevention in Oklahoma American Indian women. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0017.htm.
==== Refs
1 Gohdes D Kaufman S Valway S Diabetes in American Indians. An overview Diabetes Care 1993 16 1 239 243 8422782
2 Jackson MY Nutrition in American Indian health: past, present, and future J Am Diet Assoc 1986 11 86 11 1561 1565 3534063
3 Shalala DE Trujillo MH Hartz GJ Paisano EL Trends in Indian health, 1998-1999 U.S. Department of Health and Human Services Rockville (MD) 2000 33
4 Smith CJ Nelson RG Hardy SA Manahan EM Bennett PH Knowler WC Survey of the diet of Pima Indians using quantitative food frequency assessment and 24-hour recal. Diabetic Renal Disease Studyl J Am Diet Assoc 1996 8 96 8 778 784 8683009
5 Prevalence of diagnosed diabetes among American Indians/Alaskan Natives — United States, 1996. MMWR Morb Mortal Wkly Rep 2000 10 30 47 42 901 904
6 Valway S Freeman W Kaufman S Welty T Helgerson SD Gohdes D Prevalence of diagnosed diabetes among American Indians and Alaska Natives, 1987. Estimates from a national outpatient data base Diabetes Care 1993 1 16 2 271 276 8422791
7 Lee ET Howard BV Savage PJ Cowan LD Fabsitz RR Oopik AJ Diabetes and impaired glucose tolerance in three American Indian populations aged 45-74 years. The Strong Heart Study Diabetes Care 1995 May> 18 5 599 610 8585996
8 Rhoades ER Hammond J Welty TK Handler AO Amler RW The Indian burden of illness and future health interventions Public Health Rep 1987 Jul–Aug 102 4 361 368 3112844
9 Clark CM Fradkin JE Hiss RG Lorenz RA Vinicor F Warren-Boulton E Promoting early diagnosis and treatment of type 2 diabetes: the National Diabetes Education Program JAMA 2000 7 19 284 3 363 365 10891969
10 Caro JJ Ward AJ O'Brien JA Lifetime costs of complications resulting from type 2 diabetes in the U.S. J Nutr Elder 2002 3 25 3 476 481
11 Valdmanis V Smith DW Page MR Productivity and economic burden associated with diabetes Am J Public Health 2001 1 91 1 129 130 11189805
12 Designing a diabetes nutrition education program for a Native American community Diabetes Educ 1988 Stegmayer P Lovrien FC Smith M Keller T Gohdes DM 1988 Jan–Feb 14 1 64 66 3335190
13 Pelican S Proulx JM Wilde J Del Vecchio A Dietary guidance workshop helps tribal program cooks make changes J Am Diet Assoc 1995 5 95 5 591 592 7722198
14 Lawn J Lawn P Nutrition education for Native treatment centres Arctic Med Res 1991 Suppl 758 760 1365291
15 Algert SJ Teaching elementary school children about healthy Native American foods J Nutr Educ Behav 2003 Mar–Apr 35 2 105 106 12725717
16 Kozak D Surrendering to diabetes: An embodied response to perceptions of diabetes and death in the Gila River Indian community Omega 1997 35 4 347 359
17 Judkins RA American Indian Medicine and contemporary health problems. IV. Diabetes and perception of diabetes among Seneca Indians N Y State J Med 1978 7 78 8 1320 1323 276674
18 Huttlinger K Krefting L Drevdahl D Tree P Baca E Benally A Doing battle: a metaphorical analysis of diabetes mellitus among Navajo people Am J Occup Ther 1992 8 46 8 706 712 1379778
19 Hatton DC Health perceptions among older urban American Indians West J Nurs Res 1994 16 4 392 403 7941485
20 Understanding Health Risk in Limited-Income Women 2001 Parker SP Oklahoma State University 2002
21 Corbin J Strauss A Grounded theory research: procedures, canons, and evaluative criteria Qual Soc 1990 13 1 3 21
22 Sobralske MC Perceptions of health: Navajo Indians Top Clin Nurs 1985 10 7 3 32 39 3849920
23 Van Sickle D Wright AL Navajo perceptions of asthma and asthma medications: clinical implications Pediatrics 2001 7 108 1 E11 11433090
24 Devine CM Wolfe WS Bisogni CA Frongillo EA Life-course events and experiences: associations with fruit and vegetable consumption in 3 ethnic groups J Am Diet Assoc 1999 3 99 3 303 314
25 Contento IR Michela JL Goldberg CJ Food choice among adolescents: population segmentation by motivations J Nutr Educ Behav 1988 20 6 289 298
26 Murphy A Doner L Doing the best evaluation possibles Charting the course for evaluation: how do we measure the success of nutrition education and promotion in food assistance programs? U.S. Department of Agriculture 1997 28 30
27 Gittelsohn J Evans M Story M Davis SM Metcalfe L Helitzer DL Multisite formative assessment for the Pathways study to prevent obesity in American Indian schoolchildren Am J Clin Nutr 1999 4 69 4 Suppl S767 S772
28 Caballero B Davis S Davis CE Ethelbah B Evans M Lohman T Pathways: a school-based program for the primary prevention of obesity in American Indian childrens Journal of Nutrition Biochemistry 1998 9 535 543
29 Gilliland S Azen AL Perez GE Carter JS Strong in body and spirit: lifestyle intervention for Native American adults with diabetes in New Mexico Diabetes Care 2002 1 25 1 78 83 11772905
30 Pichette EF Berven NL Menz FE LaFromboise TD Effects of cultural identification and disability status on perceived community rehabilitation needs of American Indians The Journal of Rehabilitation 1997 63 4 38 45
|
15663882
|
PMC1183498
|
CC BY
|
2022-01-24 23:45:28
|
no
|
Prev Chronic Dis. 2004 Mar 15; 1(2):A06
|
utf-8
|
Prev Chronic Dis
| 2,004 |
nan
|
oa_comm
|
==== Front
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv12_03_0009
Original Research
PEER REVIEWEDPediatric Hospitalizations for Asthma: Use of a Linked File to Separate Person-level Risk and Readmission
Wallace Jonathan C. MA, MPH Maternal and Child Health Epidemiology Program, New Jersey Department of Health and Senior Services
PO Box 364, Trenton, NJ 08625 [email protected]
609-292-5656
Denk Charles E. PhD Maternal and Child Health Epidemiology Program, New Jersey Department of Health and Senior Services, Trenton, NJ
Kruse Lakota K. MD, MPH Maternal and Child Health Epidemiology Program, New Jersey Department of Health and Senior Services, Trenton, NJ
4 2004
15 3 2004
1 2 A072004
Introduction
Disparities in asthma hospitalization by gender, age, and race/ethnicity are thought to be driven by a combination of 2 factors: disease severity and inadequate health care. Hospitalization data that fail to differentiate between numbers of admissions and numbers of individuals limit the ability to derive accurate conclusions about disparities and risks.
Methods
Hospitalization records for pediatric asthma patients (aged one to 14 years) were extracted from New Jersey Hospital Discharge Files (for the years 1994 through 2000) and then linked by patient identifiers using a probabilistic matching algorithm. The analysis file contained 30,400 hospital admissions for 21,016 children. Hospitalization statistics were decomposed into persons hospitalized and number of hospitalizations. Analysis of readmission within 180 days of discharge used additional records from 2001 to avoid bias due to truncated observation.
Results
Overall, 22.9% of children in our analysis had repeat asthma admissions within the same age interval, accounting for 30.9% of all hospitalizations. Also among all children, 11.7% had at least one readmission within 180 days of a prior discharge. The risk of hospitalization was higher for boys, decreased by age for both genders, was lowest for white children and highest for black children. Readmission rates were higher for black and Hispanic girls than boys in older age groups, but were otherwise relatively uniform by gender and age.
Conclusion
Decomposition of ratios of total hospitalizations to population illuminates components of risk and suggests specific causes of disparity.
==== Body
Introduction
Asthma is one of the most common chronic conditions in the United States and is often cited as the most frequent reason for preventable hospital admissions among children (1-4). The United States Department of Health and Human Services' Healthy People 2010: Objectives for Improving Health established a goal to "reduce asthma morbidity, as measured by a reduction in hospitalizations" (5).
The Asthma and Allergy Foundation of America estimates that there are 142,000 children with asthma in New Jersey (6). Hospital discharge data from 1985 through 2000 indicate that asthma is a major cause of hospitalization for all ages in New Jersey, accounting for approximately 1% of New Jersey's average 1.4 million hospital discharges each year (7). New Jersey mirrors national disparities in asthma hospitalization rates among various population groups by age, gender, and race (1,2).
Few state asthma surveillance systems to date differentiate between the number of individuals hospitalized and the number of admissions, the latter of which can be numerous during an individual's lifetime. The difference between the number of individuals hospitalized and the number of admissions can lead to several forms of bias. First, total hospital admissions overstate the number of individuals affected by severe asthma. Second, the repeat admissions of some individuals may obscure the true sociodemographic distribution of severe disease. Third, routine inference from hospitalizations to individuals implicitly assumes that patterns of readmission reinforce sociodemographic differences in person-level risk or are neutral. Hospitalization for asthma implies more severe disease, less adequate preventive care, or both. Failure to distinguish persons from hospitalizations undermines inferences about the contribution of either of these 2 classes of causation.
The objectives of this study were to use linked hospital asthma admission data for children to accomplish the following: 1) deduplicate records of asthma hospitalizations and assess the scope of readmissions; 2) investigate relative risks of ever being hospitalized; and 3) examine the frequency of hospitalization for each individual admitted and assess the risk of readmission for asthma within 180 days of a prior discharge.
Methods
The study population was defined as all asthma hospitalizations experienced by New Jersey resident children aged one to 14 years from 1994 through 2000. Records from New Jersey Hospital Discharge Files (UB-92) were linked to identify patients with multiple asthma hospitalizations. Asthma hospitalizations were defined using the Centers for Disease Control and Prevention case definition: a primary diagnosis of ICD-9 Code 493 (8). AutoMatch, a probabilistic matching program, was used to link patients with more than one hospitalization in the data subset (9,10). Variables used in the linking process included the following: last name, first name, date of birth, municipality, zip code, race/Hispanic origin, hospital, medical record number, and insurer identification number. In probabilistic matching for deduplication, pools of candidate matches are identified by one or more variables such as last name and birth date. A second set of variables is used to score the similarity of each candidate pair, and the highest scored candidates are linked subject to a minimum cutoff score. The linkage process is iterative, so that variables used to identify candidate matches at one stage were used for confirmatory scoring in other stages and vice versa.
Hospitalization records identified the patient's age, gender, and race/ethnicity. Patient ages were grouped into the following age categories at each hospital admission: one to 4 years, 5 to 9 years, and 10 to 14 years. Children younger than one year were excluded because of issues of uncertainty among physicians surrounding the diagnosis and coding of asthma for children at this age. After age and calendar time exclusions, 32,825 hospitalization records representing 22,990 individuals were available. However, 399 records (1.2%) had missing data for race/ethnicity, and 615 additional persons (3.1%) had inconsistent data across multiple records. In both cases, all records for each person were coded using the following hierarchy:
If any records indicated race as Hispanic, then the person was coded "Hispanic" for all hospitalizations.
If any records indicated race as black, then the person was coded "black" for all hospitalizations.
If no records contradicted race as white, then the person was coded "white" for all hospitalizations.
The final data extract had 30,400 hospitalization records representing 21,016 white, black, and Hispanic persons. Asian and other non-Hispanic individuals were not included in this analysis because of the relatively small number of hospitalizations.
For our analysis, hospitalization ratios are defined as the number of hospitalizations in a population subgroup, including multiple readmissions of the same patient. They were calculated using data on age, gender, and ethnic origin from the 2000 Census as the denominator. Intercensal estimates would not enhance the analysis. Because 7 years of hospitalization records were extracted, census denominators were multiplied by 7.
Person-level hospitalization rates are defined as the number of distinct individuals hospitalized within a population subgroup (including age groups), and used the same 2000 Census denominators (age, gender, and ethnic origin) as hospitalization ratios. Records shared a common person identifier within the linked file. We were thus able to count the number of unique individuals hospitalized within any population subgroup. Individuals who were hospitalized at ages in different categories were counted once per age group. The frequency of admission per child is the ratio of total admissions within an age group to the number of individual children in the age group — essentially, the average number of admissions per individual child.
For a more precise analysis of readmissions, we treated each admission as a prospective opportunity for readmission and checked for a succeeding admission for that individual within 180 days from the date of discharge. We also accessed records from the first half of 2001, so that loss to follow-up would not occur for admissions late in 2000. Age group was classified by the earlier admission in each potential pair, and the entire 180-day readmission window was used even if the child aged out of his or her original age group. Readmission rates are the proportion of admissions in each subpopulation with a succeeding readmission in the time window. Although it departs from the logic of decomposing hospitalization ratios, the analysis of readmission events over a fixed period, constructed to minimize censoring bias, is a stronger approach with fewer inherent limitations.
We judged that the period of 180 days was an appropriate window for clinical follow-up of a chronic condition compared, for example, to much shorter times required for complications from surgical procedures. Also, since aggregate rates of asthma hospitalization have a strong annual cycle, we wanted an interval short enough to avoid confounding with that periodicity. Analysis using other time intervals did not yield substantially different results.
All analysis was performed using SAS Release 8.01 (11). Since the data set captures virtually all persons and hospitalizations for the period examined, statistical inferences based on sampling variability are not applicable. Systematic errors of omission, classification, linkage, and other biases do not follow the same statistical laws as sampling errors. The practical significance of subgroup differences, and whether they exceed expectations of error, must be judged on external rather than statistical criteria.
Results
From 1994 through 2000, 21,016 New Jersey children accounted for 30,400 asthma hospitalizations in New Jersey (Table 1). Overall, 4808 children (22.9% of all children hospitalized for asthma in New Jersey) experienced multiple admissions; their 9384 duplicate admissions accounted for 30.9% of all pediatric asthma admissions. Within this group, 2459 children (11.7% of all children hospitalized for asthma in New Jersey) experienced at least one readmission within 180 days of a prior asthma discharge, totaling 4340 admissions. These latter readmissions, however, accounted for 14.3% of all childhood/pediatric hospitalizations for asthma in New Jersey.
Table 2 presents the decomposition of hospitalization ratios into components by gender, age, and race/ethnicity. To illuminate the most salient points of Table 2, we constructed Figures 1-3.
Figure 1 presents person-level hospitalization rates for all individual children (per thousand population) hospitalized for asthma in New Jersey by age, sex and race/ethnicity. Rates are generally higher for boys than girls. For example, rates for boys are approximately twice the rates for girls among white children aged one to 4 years; the difference varies somewhat by age and race/ethnicity. Rates decline with age, especially between the ages of one to 4 years and 5 to 9 years. For example, rates for white girls drop from 1.8 to 0.7 per 1000 between ages one to 4 years and ages 5 to 9 years, and rates for white boys drop from 3.5 to 1.2 per 1,000 for the same age groups. White children experience the lowest rates, black children the highest, and rates for Hispanic children rates fall in between. Generally, black children have about 4 times the risk of hospitalization compared to white children of comparable age and gender; Hispanic children have about twice the risk compared to their white peers.
Figure 2 presents the frequency of admissions per child who has ever been hospitalized for asthma in New Jersey by age, gender, and race/ethnicity. The qualitative differences in relationships compared to the previous figure are striking; in contrast to person-level hospitalization rates, the frequency of admissions is roughly uniform by age for white children and increases by age for black and Hispanic children. Admission frequencies vary for white girls from 1.21 (ages one to 4 years) to 1.18 (ages 5 to 9 years) to 1.27 (ages 10 to 14 years), but for black girls, admission frequencies increase from 1.35 to 1.44 to 1.72 for the same age groups. Generally, the frequency of admission for boys and girls is equal for white children, although girls have higher averages than boys among older blacks and Hispanics. For example, among white children aged one to 4 years ever hospitalized for asthma, girls experience an average of 1.21 admissions while boys have an average of 1.19; the same comparison for Hispanic children ages 10 to 14 years is 1.70 (girls) vs 1.47 (boys). Finally, compared to the relative advantage of Hispanics over blacks in Figure 1, that trend is eliminated or reversed for frequency of admission, depending on the age group.
Figure 3 presents our analysis of readmission rates more precisely calculated as readmission within 180 days of prior discharge and displays the same general patterns as the admission frequencies in Figure 2 with respect to gender, age and race/ethnicity. Again, readmission rates are roughly uniform by age for white children and increase by age for black and Hispanic girls only. Readmission rates vary for white girls from 11% (ages one to 4 years) to 8% (ages 5 to 9 years) and 10% (ages 10 to 14 years), but for black girls they increase from 16% to 20% to 29% (same age groups). Generally, readmission rates for boys and girls are equal for white children, although girls have higher rates than boys among older black and Hispanic children. For example, among white children ages one to 4 years ever hospitalized for asthma, girls experience a readmission rate of 11% while boys have an average of 10%; the same comparison for Hispanic children ages 10 to 14 years is 21% (girls) vs 17% (boys). Black girls ages 10 to 14 years have a very distinctive spike compared to boys and to Hispanic girls of the same age.
To underscore the comparisons discussed above, Table 3 transforms the measures in Table 2 into relative risks and similar ratios. The first 4 columns contrast younger age groups against the 10-to-14 years age group. These columns most clearly support our generalizations and distinctions about age and race/ethnicity. Hospitalization ratios and person-level rates decline dramatically with age across all race/ethnic groups, with slightly larger declines for boys. Admission frequencies and readmission rates have much weaker and usually opposite effects. It is important to note that wherever age has opposite effects on person-level admission and readmission, the overall hospitalization ratio will be smaller than the person-level admission rate. The last 4 columns of Table 3 contrast black and Hispanic children against whites. In general, both person-level hospitalization and readmission components are substantial and in the same direction, which produces even larger relative differentials in hospitalization ratios. The difference between black and Hispanic children is largely a function of person-level hospitalization — every other ratio in these columns is very similar between the two.
Discussion
Hospitalization ratios in New Jersey have associations with gender, age, and race/ethnicity that are comparable to other recent reports (12-17). Repeat hospitalizations for asthma are a common event for New Jersey children, accounting for almost one third of all admissions for asthma. Much epidemiological evidence and clinical experience suggests that hospitalization for asthma is a mixture of biological disease factors and failures of preventive care. We believe that the significant qualitative differences between person-level hospitalization rates and readmissions provide important leverage to distinguish the influences of the two.
Person-level hospitalization rates give a more precise accounting of how the burden of severe asthma is distributed across individuals (rather than the burden of hospitalization as a distinct event). This is the main implication of the differences between hospitalization ratios and person-level rates in Table 3. On the other hand, readmissions are more than residual events to be eliminated in deduplication. Readmission rates are arguably more driven by issues of disease management, since children in that analysis are more homogeneous — they all have a degree of disease severe enough to require hospitalization. The 2 rates in combination — person-level and readmission — tell the same story as hospitalization ratios, but in a more focused and coherent way.
In general, our findings on readmission rates fit with very simple hypotheses about barriers to appropriate preventive care. Based solely on socioeconomic barriers to access, we would expect comparatively little variation by gender and age (of children, not adults), and we would expect black and Hispanic children to be similar to each other. These expectations are generally met by our data. Person-level rates clearly are more complex, and it would be helpful if we could assess the relative contribution of access and other factors.
For example, numerous studies have documented age and race/ethnic disparities in asthma hospitalization (12-18). Variations in the accessibility and quality of preventive asthma care have been linked to race/ethnicity (19-22). A recent study that explicitly addresses financial access still finds disparities in processes of asthma care such as the use of controller medications (23). Risk factors for asthma morbidity — such as outdoor environment, family smoking, and physical activity — have been well demonstrated to vary by racial and ethnic groups. Less is known about potential biologic factors such as genetic, physiological, pharmacogenomic, and/or environmental-genetic interactions.
In New Jersey, person-level hospitalization rates for black children are 3.2 to 4.3 times higher than for white children (stratified by age and gender, Table 3). This ratio is about twice the disparity in person-level rates experienced by Hispanic children and also twice the disparity in readmission rates for both black and Hispanic children. On the basis of this outsized differential, we observe, as others have, that there could be a biological difference in disease etiology for black children (24,25). Regardless of the cause(s), this disparity demands further investigation.
Numerous studies have documented higher hospitalization ratios for preschool-age children (12-16). In New Jersey, children ages one to 4 years have threefold higher person-level hospitalization rates than preadolescents among girls and more than fourfold higher rates among boys. This differential does not exist at all for readmission rates, contrary to To's findings from Canada (12). This would seem to undermine the hypothesis that preschool-age children present special challenges for home management and lends more support to the distinctive nature of early onset asthma (26,27). On the other hand, there is an anomalous spike in readmission rates for black girls ages 10 to 14 years, which is not mirrored in person-level rates or in black boys or Hispanic girls of the same age. It is more likely that this anomaly is related to asthma management and use of controller medications (28).
Our analysis of linked hospitalizations does not allow us, unfortunately, to explore some issues typical in a cohort analysis. For example, we cannot describe age-specific onset or population prevalence of severe asthma. Since we cannot in most cases identify the first hospitalization for each child, we can only generalize about readmissions by assuming independence among the intervals between hospitalizations.
Matching of annual files does not lend itself, in this case, to the construction and analysis of age cohorts of children. The optimal timeframe for a longitudinal file is limited by the comparability of files over many years and a natural decay in expected matching accuracy. True cohort data, with complete histories from both inpatient and outpatient management, would obviously be much more powerful.
State-level asthma hospitalization surveillance systems like New Jersey's address 2 objectives, both incompletely. Trends in asthma hospitalizations can inform us about prevalence and distribution of the severest forms of the disease. Surveillance of asthma readmissions especially informs us about the effectiveness of asthma care — accessibility of care, management of environmental triggers, and appropriate preventive asthma management. For both objectives, racial and ethnic disparities and age-specific differences are critically important.
This research was supported by the Addressing Asthma from a Public Health Perspective grant from the Centers for Disease Control and Prevention. The analysis and conclusions expressed here are those of the authors.
Figures and Tables
Figure 1 Person-level hospitalization rates for asthma by age, gender, race/ethnicity, New Jersey, 1994–2000. NH indicates non-Hispanic.
Person-level hospitalization rates for asthma by age, gender, race/ethnicity, New Jersey, 1994–2000. NH indicates non-Hispanic.
Figure 2 Average admissions per child hospitalized for asthma by age, gender, and race/ethnicity, New Jersey, 1994–2000. NH indicates non-Hispanic.
Average admissions per child hospitalized for asthma by age, gender, and race/ethnicity, New Jersey, 1994–2000. NH indicates non-Hispanic.
Figure 3 Readmission rates within 180 days of prior asthma discharge by age, gender, and race/ethnicity, New Jersey, 1994–2000. NH indicates non-Hispanic.
Readmission rates within 180 days of prior asthma discharge by age, gender, and race/ethnicity, New Jersey, 1994–2000. NH indicates non-Hispanic.
Table 1 Hospitalization Records for Pediatric Asthma Admissions, New Jersey, 1994–2000
Admissions Children Admissions per Child
n % n %
All pediatric asthma admissions 30,400 100.0 21,016 100.0 1.45
Repeat asthma admissions within same age interval 9,384 30.9 4,808 22.9 2.95
Readmissions within 180 days of previous dischargea 4,340 14.3 2,459 11.7 2.76
a Analysis of readmissions within 180 days of discharge used additional records from 2001 to avoid censoring bias.
Table 2 Pediatric Asthma Hospitalizations and Readmissions by Race/Ethnicity, Age, and Gender, New Jersey, 1994–2000
Race/Ethnicity Age (years) Pediatric Asthma Admissions (n) Children Admitted for Asthma (n) 2000 Census (n) Hospitalization Ratioa Person-level Hospitalization Rateb Mean Hospitalizations per Child Ever Hospitalized for Asthmac Readmission Within 180 Days of Prior Discharged(%)
Female
White 1-4 2006 1655 127,941 2.2 1.8 1.21 11
5-9 1028 871 174,845 0.8 0.7 1.18 8
10-14 921 726 174,529 0.8 0.6 1.27 10
Black 1-4 2353 1746 33,352 10.1 7.5 1.35 16
5-9 1447 1002 47,256 4.4 3.0 1.44 20
10-14 1071 621 46,114 3.3 1.9 1.72 29
Hispanic 1-4 1283 909 38,335 4.8 3.4 1.41 17
5-9 795 526 47,347 2.4 1.6 1.51 21
10-14 565 333 44,174 1.8 1.1 1.70 21
Male
White 1-4 3914 3283 134,718 4.2 3.5 1.19 10
5-9 1782 1570 183,919 1.4 1.2 1.14 7
10-14 1191 953 184,884 0.9 0.7 1.25 11
Black 1-4 4160 3050 34,456 17.2 12.6 1.36 16
5-9 2311 1688 49,335 6.7 4.9 1.37 15
10-14 1385 894 47,466 4.2 2.7 1.55 20
Hispanic 1-4 2339 1691 39,977 8.4 6.0 1.38 17
5-9 1257 886 49,711 3.6 2.5 1.42 17
10-14 592 403 45,770 1.8 1.3 1.47 17
Total 30,400 21,016 1,504,129 N/A N/A N/A N/A
Average N/A N/A N/A N/A 2.0 1.45 15
a Per 1000 population. Hospital ratios are defined as the number of pediatric asthma admissions in a population subgroup, including multiple readmissions of the same patient, divided by the number of individuals in a population subgroup, as determined by the 2000 Census, multiplied by 7.
b Per 1000 population. Person-level hospitalization rates are defined as the number of pediatric asthma individuals hospitalized within a population subgroup, divided by the number of individuals in a population subgroup, as determined by the 2000 Census, multiplied by 7.
c Mean number of hospitalizations per child ever hospitalized for asthma was calculated by dividing the number of pediatric asthma admissions by the number of children admitted for asthma.
d Analysis of readmissions within 180 days of discharge used additional records from 2001 to avoid censoring bias.
Table 3 Relative Risk Ratios, Pediatric Asthma Hospitalizations and Readmissions by Race/Ethnicity, Age, and Gender, New Jersey, 1994-2000
Relative risk ratio by agea
Race/Ethnicity Age (years) Hospitalization Ratiob Person-level Hospitalization Ratec Mean Hospitalizations per Child Ever Hospitalized for Asthmad Readmission Within 180 Days of Prior Dischargee
Female
White 1-4 3.0 3.1 1.0 1.1
5-9 1.1 1.2 0.9 0.8
10-14 -- -- -- --
Black 1-4 3.0 3.9 0.8 0.6
5-9 1.3 1.6 0.8 0.7
10-14 -- -- -- --
Hispanic 1-4 2.6 3.1 0.8 0.8
5-9 1.3 1.5 0.9 1.0
10-14 -- -- -- --
Male
White 1-4 4.5 4.7 1.0 0.9
5-9 1.5 1.7 0.9 0.7
10-14 -- -- -- --
Black 1-4 4.1 4.7 0.9 0.8
5-9 1.6 1.8 0.9 0.8
10-14 -- -- -- --
Hispanic 1-4 4.5 4.8 0.9 1.0
5-9 2.0 2.0 1.0 1.0
10-14 -- -- -- --
Relative risk ratio by race/ethnicityf
Race/Ethnicity Age(years) Hospitalization Ratio Person-level Hospitalization Rate Mean Hospitalizations per Child Ever Hospitalized for Asthma Readmission Within 180 Days of Prior Discharge
Female
White 1-4 -- -- -- --
5-9 -- -- -- --
10-14 -- -- -- --
Black 1-4 4.5 4.0 1.1 1.6
5-9 5.2 4.3 1.2 2.4
10-14 4.4 3.2 1.4 3.0
Hispanic 1-4 2.1 1.8 1.2 1.7
5-9 2.9 2.2 1.3 2.5
10-14 2.4 1.8 1.3 2.2
Male
White 1-4 -- -- -- --
5-9 -- -- -- --
10-14 -- -- -- --
Black 1-4 4.2 3.6 1.1 1.7
5-9 4.8 4.0 1.2 2.1
10-14 4.5 3.7 1.2 1.9
Hispanic 1-4 2.0 1.7 1.2 1.7
5-9 2.6 2.1 1.2 2.4
10-14 2.0 1.7 1.2 1.6
a 1-4 year and 5-9 year age groups are compared to 10-14 year age group.
b Hospital ratios are defined as the number of pediatric asthma admissions in a population subgroup, including multiple readmissions of the same patient, divided by the number of individuals in a population subgroup, as determined by the 2000 Census, multiplied by 7.
c Person-level hospitalization rates are defined as the number of pediatric asthma individuals hospitalized within a population subgroup, divided by the number of individuals in a population subgroup, as determined by the 2000 Census, multiplied by 7.
d Mean number of hospitalizations per child ever hospitalized for asthma was calculated by dividing the number of pediatric asthma admissions by the number of children admitted for asthma.
e Analysis of readmissions within 180 days of discharge used additional records from 2001 to avoid censoring bias.
f Black and Hispanic children are compared to white children.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Wallace JC, Denk CE, Kruse LK. Pediatric hospitalizations for asthma: use of a linked file to separate person-level risk and readmission. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0009.htm.
==== Refs
1 Miller J The effects of race/ethnicity and income on early childhood asthma prevalence and health care use Am J Public Health 2000 3 90 3 428 430 10705865
2 Mannino DM Homa DM Akinbami LJ Moorman JE Gwynn C Redd SC Surveillance for asthma — United States, 1980-1999 MMWR Surveill Summ 2002 3 29 51 1 1 13
3 Fiore BJ Olson JA Hanrahan LP Anderson HA Asthma hospitalizations in Wisconsin: a missed opportunity for prevention WMJ 2000 11 99 8 52 56 11149262
4 Asthma hospitalizations and readmissions among children and young adults — Wisconsin, 1991-19951 MMWR Morb Mortal Wkly Rep 1997 8 8 46 31 726 729 9262074
5 U.S. Department of Health and Human Services Healthy People 2010: With understanding and improving health and objectives for improving health. 2 vols Washington (DC) U.S. Government Printing Office 2000 11
6 Weiss KB Sullivan SD Lyttle CS Trends in the costs of asthma in the United States, 1985-1994 J Allergy Clin Immunol 2000 9 106 3 493 499 10984369
7 2002 1 29 Asthma in New Jersey 286 Trenton (NJ) New Jersey Department of Health and Senior Services, Division of Family Health Services
8 Centers for Disease Control and Prevention. Council of State and Territorial Epidemiologists asthma surveillance definition U.S. Department of Health and Human Services, CDC, National Center for Environmental Health Atlanta (GA) 2001
9 AutoMatch: Generalized Record Linkage System [computer program]. Version 4.2 1998 Kennebunkport (ME) Matchware Technologies Inc
10 Jaro MA Probabilistic linkage of large public health data files Stat Med 1995 3 15 14 (5-7) 491 498 7792443
11 SAS: System for Windows [computer program]. Version 8 1998 Cary (NC) SAS Institute Inc
12 To T Dick P Feldman W Hernandez R A cohort study on childhood asthma admissions and readmissions Pediatrics 1996 8 98 (2 Pt 1) 191 195 8692616
13 Jonasson G Lodrup Carlsen KC Leegaard J Carlsen K-H Mowinckel P Halvorsen KS Trends in hospital admissions for childhood asthma in Oslo, Norway, 1980-95 Allergy 2000 3 55 3 232 239 10753013
14 Korhonen K Reijonen TM Malmstrom K Klaukka T Remes K Korppi M Hospitalization trends for paediatric asthma in eastern Finland: a 10-yr survey Eur Respir J 2002 6 19 6 1035 1039 12108853
15 Engelsvold DH Oymar K Hospital admissions for childhood asthma in Rogaland, Norway, from 1984 to 2000 Acta Paediatr 2003 5 92 5 610 616 12839293
16 Akinbami LJ Schoendorf KC Trends in childhood asthma: prevalence, health care utilization, and mortality Pediatrics 2002 8 110 (2 Pt 1) 315 322 12165584
17 Senthilselvan A Strong in body and spirit: lifestyle intervention for Native American adults with diabetes in New Mexico Thorax 1995 Sep> 50 9 934 936 8539671
18 Aligne CA Auinger P Byrd RS Weitzman M Risk factors for pediatric asthma. Contributions of poverty, race, and urban residence Am J Respir Crit Care Med 2000 9 162 (3 Pt 1) 873 877 10988098
19 Halterman JS Aligne CA Auinger P McBride JT Szilagyi PG Inadequate therapy for asthma among children in the United States Pediatrics 2000 1 105 (1 Pt 3) 272 276 10617735
20 Finkelstein JA Brown RW Schneider LC Weiss ST Quintana JM Goldmann DA Quality of care for preschool children with asthma: the role of social factors and practice setting Pediatrics 1995 95 3 389 394 7862478
21 Homer CJ Szilagyi P Rodewald L Bloom SR Greenspan P Yazdgerdi S Does quality of care affect rates of hospitalization for childhood asthma Pediatrics 1996 7 98 1 18 23 8668406
22 Perrin JM Homer CJ Berwick DM Woolf AD Freeman JL Wennberg JE Variations in rates of hospitalization of children in three urban communities N Engl J Med 1989 5 4 320 18 1183 1187 2710191
23 Lieu TA Lozano P Finkelstein JA Chi FW Jensvold NG Capra AM Racial/ethnic variation in asthma status and management practices among children in managed Medicaid Pediatrics 2002 5 109 5 857 865 11986447
24 Nelson DA Johnson CC Divine DW Strauchman C Joseph CL Owenby DR Ethnic differences in the prevalence of asthma in middle class children Ann Allergy Asthma Immunol 1997 1 78 1 21 26 9012615
25 Nimmagadda SR Evans R Allergy: etiology and epidemiology Pediatr Rev 1999 4 20 4 111 116 10208083
26 Sears MR Greene JM Willan AR Wiecek EM Taylor DR Flannery EM A longitudinal, population-based, cohort study of childhood asthma followed to adulthood N Engl J Med 2003 10 9 349 15 1414 1422 14534334
27 Martinez F Toward asthma prevention — does all that really matters happen before we learn to read? N Engl J Med 2003 10 9 349 15 1473 1475 14534342
28 Lozano P Finkelstein JA Hecht J Shulruff R Weiss KB Asthma medication use and disease burden in children in a primary care population Arch Pediatr Adolesc Med 2003 1 157 1 81 88 12517200
|
15663883
|
PMC1183499
|
CC BY
|
2022-01-31 23:22:34
|
no
|
Prev Chronic Dis. 2004 Mar 15; 1(2):A07
|
utf-8
|
Prev Chronic Dis
| 2,004 |
nan
|
oa_comm
|
==== Front
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv12_04_0002
Special Topics in Public Health
Essay
A Glimpse of Things to Come: Featured Abstracts from the 18th National Conference on Chronic Disease Prevention and Control
Katz David L. MD, MPH Director Yale-Griffin Prevention Research Center
130 Division St, Derby, CT 06418
E-mail: [email protected]
with copy to [email protected]
4 2004
15 3 2004
1 2 A082004
Editor's note
Dr Katz serves as co-chair of the 18th National Conference on Chronic Disease Prevention and Control, held February 18–20, 2004, in Washington, DC.
==== Body
Great ventures are invariably a blend of tradition and innovation. Tradition is a cornerstone of identity, and without it, there is no clearly defined role, no unique character, no reliability. But in a constantly changing world, tradition without innovation is a cumbrous burden. Innovation allows a great venture to keep pace with the currents of change, to remain great and timely.
The annual Chronic Disease Conference is such a venture. Spanning 18 years, the conference has given rise to an array of traditions. It is tradition for the conference to address timely, compelling public health issues. It is tradition for the conference to encompass the breadth and depth of matters relating to preventing chronic disease. It is tradition for content to be shaped by the thoughtful reflection of invited luminaries and the insightful advances conveyed within abstracts submitted for competitive review.
Conference traditions are now to be enhanced by a well-timed innovation: the launch of Preventing Chronic Disease: Public Health Research, Practice, and Policy has created a unique capacity to publish select conference abstracts concurrently with the conference. The efficiencies of electronic publication allow for identifying, selecting, and editing these submissions as the conference itself undergoes final preparation. The 21 most meritorious abstracts from this year's conference appear within this issue of Preventing Chronic Disease. This innovation will itself count among the traditions that define the character, content, and excellence of the Chronic Disease Conference for years to come.
Advantages of this effort are considerable. E-journal readers unable to attend the February conference may nonetheless gain an appreciation for the work of their colleagues. Virtual networking will enhance conference networking, and individuals not accustomed to attending the meeting may be enticed to do so.
Publication is the ultimate "carrot and stick" of academic advancement. Those among us who must be calculating in our use of time now have an added incentive to submit our best work to the Chronic Disease Conference for consideration. The caliber of abstracts the conference planners are privileged to review has always been excellent, and the promise of publication is motivation to ever greater excellence.
With topics ranging from nutrition in schoolchildren to stroke in the elderly, from secondhand smoke to homelessness, the abstracts that follow provide a window to the dynamic work, rigorous science, dedicated effort, and passion for public welfare that so richly furnish the Chronic Disease Conference. They provide, as well, a glimpse of things to come. Preventing Chronic Disease will from now on reward the most laudable of our submissions with online publication. The content of the abstracts directs our efforts and imaginations to the best of innovations in public health practice and to advances in our collective capacity to prevent chronic disease.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Katz DL. A glimpse of things to come: featured abstracts from the 18th National Conference on Chronic Disease Prevention and Control. Prev Chronic Dis [serial online] 2004 April [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/04_0002.htm.
|
0
|
PMC1183500
|
CC BY
|
2022-01-31 23:22:56
|
no
|
Prev Chronic Dis. 2004 Mar 15; 1(2):A08
|
utf-8
|
Prev Chronic Dis
| 2,004 |
nan
|
oa_comm
|
==== Front
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv12_03_0032
Step-by-Step: Making Your Communities Healthier
A Model Community Skin Cancer Prevention Project in Maine
Hayden Christine A. City of Portland, Health and Human Services Department, Public Health Division
389 Congress St, Portland, ME 04101 [email protected]
207-874-8942
4 2004
15 3 2004
1 2 A102004
==== Body
The purpose of our program was to create and test a community skin cancer prevention project for replication throughout the state of Maine. The project was a collaborative effort of the Maine Cancer Consortium, American Cancer Society (ACS), and the City of Portland, Health and Human Services Department, Public Health Division. Portland, Me, served as the pilot site.
The National Cancer Institute (NCI) defines skin cancer as a disease in which abnormal cells divide uncontrollably in the outer layers of the skin (1). The American Cancer Society's Facts and Figures 2001 (the latest year for which these figures are available) estimated that more than 1 million cases of highly curable basal cell or squamous cell cancers would be diagnosed in the United States that year (2). An estimated 9800 U.S. deaths from cancer were projected as well: 7800 from melanoma, the most serious form of skin cancer, and 2000 from other skin cancers. Melanoma was expected to be diagnosed in about 51,400 Americans in 2001. The incidence rate of melanoma has increased about 3% per year on average since 1981. In 2002, NCI announced that researchers showed for the first time that individual risk of melanoma is associated with the intensity of sunlight that a person receives over a lifetime (3).
Target audiences for our program were newborns and their parents, children between 5 and 14 years old and their caregivers, and all people living in the Portland area. Protecting skin from excess sun exposure during childhood and adolescence is important in reducing the risk of all types of skin cancer during adulthood. From our anecdotal evidence, many parents of newborns are unaware that sunscreen is not recommended for babies under 6 months of age, and they need better information about how to protect their newborns from the sun. Teaching children and their caregivers to follow ACS guidelines will help protect their skin for years to come. It will also help children to develop healthy habits they can maintain throughout their lives.
Our skin cancer prevention project, which took place during 2002, mirrored the goals of Healthy Maine 2010 to increase the proportion of people who limit sun exposure and use protection when exposed to sun (4). Our specific objectives were the following:
Increase the proportion of new parents who are aware of the dangers of sun exposure to newborns and the proper ways to protect their babies from sun exposure.
Improve sun protection policies for the 700 youths participating in Portland's Parks and Recreation summer camp program.
Increase community awareness of the dangers of unprotected sun exposure.
To help us design and implement our action plan, in fall 2001 we recruited members for a sun protection team from among our group of partners and from the ACS Cancer Control Advisory Council. The team consisted of one staff person from the City of Portland's Public Health Division, 2 staff members from the ACS, nurses from the 2 participating hospitals, staff from the City of Portland's Parks and Recreation Department, and staff of the Sea Dogs, Portland's minor league baseball team. Not all staff members were involved with all components of the program.
The project was funded through the ACS for $10,000 for a one-year period. One City of Portland staff member was paid for approximately 3 hours of work per week to coordinate the program. In addition, approximately $3500 was spent on supplies, $2000 for advertising and brochures, and $300 for educational sessions.
Our skin cancer prevention project consisted of 3 major components, which are described below.
No Sun for Baby
This hospital-based initiative targeted parents of babies born during May (Skin Cancer Awareness Month) 2002. Of the 3 hospitals in Portland, 2 are known for their maternity or birthing units: Maine Medical Center and Mercy Hospital. We contacted the nurse managers of the maternity unit at Maine Medical Center and the birthing unit at Mercy Hospital to discuss and obtain approval for our plans and were told that a total of between 350 and 400 babies are typically born during May at these 2 hospitals. ACS staff delivered No Sun for Baby kits to each hospital unit. Each kit included a beach pail and shovel with a Slip! Slop! Slap! Wrap! message as well as a baby sun hat and brochure from the Skin Cancer Foundation. Slip! Slop! Slap! Wrap! means "Slip on a shirt, slop on sunscreen with an SPF of 15 or higher, slap on a hat with a wide brim, and wrap on sunglasses" (5). Both hospitals agreed to promote sun safety information by distributing brochures and/or by talking about Skin Cancer Foundation recommendations and/or ACS guidelines in childbirth education classes, Lamaze classes, parenting classes, and discharge packages. An evaluation postcard was also included in the kit for parents to fill out and return to the ACS. We distributed 380 No Sun for Baby kits, and both hospitals have committed to incorporating sun safety messages in childbirth education classes on a continuing basis.
Parks and Recreation Sun Protection Guidelines
In working with the team member from the City of Portland's Parks and Recreation summer program, we learned that there were no sun protection policies or guidelines for the approximately 700 campers aged 5 to 14 years who participate annually. For the 2002 summer camp season, we distributed sun protection guidelines for outdoor recreation or work and provided technical assistance to Parks and Recreation management on adapting the guidelines to meet their needs (6). We conducted educational sessions for 50 Parks and Recreation staff members and gave them Slip! Slop! Slap! Wrap! educational brochures. We also discussed the rationale for having and enforcing guidelines. Enforcing the guidelines would prove to be difficult because some children arrived at camp without shirts, sunscreen, hats, sunglasses, or other sun protection. In the educational sessions with counselors, we stressed the importance of counselors providing positive role models for campers. We encouraged counselors to use the Slip! Slop! Slap! Wrap! model. We also emphasized the importance of creating shady areas for campers. As a result, Parks and Recreation management purchased portable tents for each campsite and created shade structures at the municipal pools.
All parents of campers were provided a copy of the sun safety guidelines as well as a copy of the Slip! Slop! Slap! Wrap! educational brochure. The Parks and Recreation Department received only one phone call during the summer from a parent of a child who had been sunburned while at camp.
Protect the Skin You're In Day
We distributed 500 sunscreen samples and Slip! Slop! Slap! Wrap! educational brochures to approximately 6200 baseball game attendees and 200 staff in collaboration with Portland's minor league baseball team, the Sea Dogs, on June 9, 2002. Slugger, the Sea Dogs mascot, performed the Slip! Slop! Slap! Rap! throughout the game. Game announcers read Slip! Slop! Slap! Wrap! messages (Figure) between innings, and we held drawings for T-shirts donated by the ACS. The Sea Dogs donated space in their program for a half-page Protect the Skin You're In advertisement. The number of programs distributed during summer 2002 was 22,750. In addition, Sea Dogs management agreed to sponsor an annual Protect the Skin You're In Day.
Figure. Public service announcements, Portland Sea Dogs baseball game, community skin cancer prevention project, Portland, Me, June 9, 2002.
Unprotected skin can burn within 12 minutes. Slip on a shirt! Slop on sunscreen! Slap on a hat! Wrap on sunglasses!
About 80% of skin cancers could be prevented by protection of skin from the sun's rays.
Skin cancer is the most common form of cancer — there are more than 1.3 million diagnoses each year.
To protect your skin from the sun's rays, apply sunscreen with SPF 15 or higher every 2 hours.
Sunscreen is not recommended for children under the age of 6 months — they should be shaded from the sun.
Sunlamps and tanning booths are as harmful to your skin as the sun.
It is most important to protect your skin between 10:00 AM and 4:00 PM, when the sun's rays are most intense.
Protect your children from sunburns! Research shows a link between sunburns in children and an increased risk of melanoma and other skin cancers later in life.
All components of the Portland sun protection pilot program could easily be replicated and sustained by other communities. The cost for each component is low. Several steps can decrease costs. For example, hospital volunteers could make sun hats for the No Sun for Baby program. Hospitals could also solicit merchants for donated pails and shovels. Many different kinds of organizations can participate: hospitals, health agencies, municipalities, sports teams, private and public camps, summer camps, YWCAs and YMCAs, Boys & Girls Clubs of America, scouting organizations, and others.
Buy-in from each organization is essential to program success. We often found that individuals within a partner organization who had friends or relatives with sunburns, suspicious moles, or skin cancer itself would often take leadership roles within the organization.
This program offers tremendous news media opportunities for promoting sun protection messages. We contacted local television stations to publicize the No Sun for Baby program. Our staff members also happened to know parents who were expecting babies in May 2002. These prospective parents agreed to be interviewed by the news media, and an interview was aired on Health Beat, a segment of a local television evening news program.
We also exhibited our program at poster sessions at the 17th National Conference on Chronic Disease Prevention and Control in St. Louis, Mo, February 17–19, 2003, and the Centers for Disease Control and Prevention Cancer Conference in Atlanta, Ga, September 15–18, 2003. This gave us the opportunity to share our work with other organizations throughout the country so they could replicate it within their own communities.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Hayden CA. A model community skin cancer prevention project in Maine. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0032.htm.
==== Refs
1 What you need to know about skin cancer [Internet] 2000 9 16 National Institutes of Health (US) Washington (DC) updated 2000 Sep 16; cited 2004 6 Feb Available from: URL: http://www.cancer.gov/cancer_information/cancer_type/skin/
2 American Cancer Society Inc. Cancer facts & figures 2001 2001 American Cancer Society Atlanta (GA) Available from: URL: http://www.cancer.org/downloads/STT/F&F2001.pdf 17 44 17, 44
3 Individuals' risk of melanoma increases with time outdoors, especially in high-sunlight areas [Internet] 2002 7 14 National Cancer Institute Washington (DC) cited 2004 6 Feb Available from: URL: http://www.cancer.gov/newscenter/individualmelanoma
4 Mills DA Healthy Maine 2010: longer and healthier lives 2002 Bureau of Health, Maine Department of Human Services Augusta (ME) Available from: URL: http://www.state.me.us/dhs/boh/healthyme2k/hm2010a.htm 45
5 Slip! Slop! Slap!: ACS helps launch national program to help shed light on skin cancer [Internet] 2004 9 16 American Cancer Society Atlanta (GA) cited 2004 6 Feb Available from: URL: http://www.cancer.org/docroot/NWS/content/ NWS_5_1x_Slip__Slop__Slap_.asp
6 Sun protection policies [Internet]. Prevention, Current Activities Maine Cancer Consortium Topsham (ME) cited 2004 6 Feb Available from: URL: http://www.mainecancerconsortium.org/welcome.html
|
15663885
|
PMC1183501
|
CC BY
|
2022-01-31 23:22:38
|
no
|
Prev Chronic Dis. 2004 Mar 15; 1(2):A10
|
utf-8
|
Prev Chronic Dis
| 2,004 |
nan
|
oa_comm
|
==== Front
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv12_04_0009
Tools and Techniques
Law as a Tool for Preventing Chronic Diseases: Expanding the Spectrum of Effective Public Health Strategies
Part 2
Mensah George A. MD Cardiovascular Health Branch, Division of Adult and Community Health, National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention
4770 Buford Hwy, NE, Mail Stop K-47, Atlanta, GA 30341-3717 [email protected]
Centers for Disease Control and Prevention 770-488-2424
Goodman Richard A. MD, JD, MPH Public Health Law Program, Public Health Practice Program Office
Centers for Disease Control
Zaza Stephanie MD, MPH Office of the Director, NCCDPHP
Centers for Disease Control
Moulton Anthony D. PhD Public Health Law Program, Public Health Practice Program Office
Centers for Disease Control
Kocher Paula L. JD Office of General Counsel
Centers for Disease Control
Dietz William H. MD, PhD Division of Nutrition and Physical Activity, NCCDPHP
Centers for Disease Control
Pechacek Terry F. PhD Office of Smoking and Health, NCCDPHP
Centers for Disease Control
Marks James S. MD, MPH NCCDPHP
Centers for Disease Control
4 2004
15 3 2004
1 2 A112004
==== Body
Introduction
In part one of this 2-part series, we reviewed the important roles that laws have played in public health and provided examples of specific laws and their effectiveness in supporting public health interventions (1). We suggested that conceptual legal frameworks for systematically applying law to preventing and controlling chronic diseases have not been fully recognized and we provided the basic elements of a conceptual legal framework. In part 2 of this series, we first provide an overview of U.S. jurisprudence, describe the legal mechanisms, remedies, and tools for applying law to public health, and summarize the jurisdictional levels at which laws, mechanisms, remedies, and tools operate. We then identify the potential contours for legal frameworks of varying complexity and scope by offering examples of legal frameworks in public health practice. This paper also outlines a plan for increasing the capacity within the Centers for Disease Control and Prevention (CDC) for developing legal frameworks and expanding guidance on using legal tools for preventing and controlling chronic diseases. Finally, we describe resources for building or enhancing the capacity to use law as a tool for preventing diseases, injuries, and disabilities at the local level.
Overview of U.S. jurisprudence and legal methods relevant to public health
The dimensions and elements of a systematic legal framework for preventing chronic diseases and other public health problems can be drawn from examining relevant fields of U.S. jurisprudence, legal theories, and legal methods. These dimensions and elements include the following: 1) basic sources of U.S. law relevant to preventing and controlling public health problems; 2) legal mechanisms, remedies, and tools for applying law to disease prevention and control; and 3) jurisdictional levels at which such laws, mechanisms, and tools might be appropriately applied.
Basic sources of U.S. law relevant to public health
Basic sources of U.S. law include the federal Constitution and state constitutions, federal and state legislative enactments, formally ratified treaties, administrative law promulgated and enforced by agencies to which legislatures have delegated such authorities, and common law (also frequently referred to as case law) articulated by the federal and state judiciary following appellate review. While the term "general welfare" is mentioned twice in the U.S. Constitution — the supreme law of the land — nowhere is the term "public health" mentioned. This absence may possibly reflect the view at the time the Constitution was established that protection of the public’s health was a state responsibility and not a duty to be assigned to the national government (2). In addition, through police powers — reserved to the states by the Tenth Amendment — states retained responsibility for public health (3,4). Despite the absence of the term in the Constitution, several provisions do confer some public health powers on the federal government as well as affect the exercise of police power by the states. For example, one provision (Article I, Section 8) confers on Congress the powers to tax, appropriate monies, and provide for the general welfare of the United States (3). These authorities have enabled Congress to establish agencies with responsibilities in public health within the executive branch, as well as to allocate monies earmarked for public health activities to the states.
In contrast to the U.S. Constitution’s grant of limited, enumerated powers to the federal government, individual state constitutions exist as limits on the sovereign powers of states. While state constitutions vary in their references to public health, many provide for their legislatures to establish state — and sometimes county or local — boards of health. In addition, states may delegate public health responsibilities to such local authorities.
A combination of federal statutes establishes roles and authorities of federal agencies in disease prevention and control activities. The CDC and the Food and Drug Administration (FDA) are 2 such agencies for which Congress has statutorily conferred explicit public health responsibilities and authorities to address public health problems. The responsibilities and powers of these agencies are reflected in provisions of the Public Health Service Act (PHSA) and the Food, Drug, and Cosmetic Act. Sections of Title III of the PHSA encompass a range of powers and duties for disease control and prevention, including conducting scientific research relating to causes, treatment, control, and prevention of diseases (Section 301), and for federal–state cooperation in disease prevention and control (Section 311). Examples of PHSA provisions more targeted to chronic disease issues are Section 317H, which covers surveillance and juvenile diabetes, Sections 399W-Z, which cover programs to improve the health of children, including, for example, grants to promote childhood nutrition and physical activity, and applied research into childhood obesity, and Sections 1501-1510, which cover breast and cervical cancer screening.
State legislatures, by acting under their broad plenary authorities and by expressing their police powers to protect the health and safety of their populations, enact numerous statutes for disease control and prevention. These statutes create public health agencies at state and local levels, articulate express authorities for such agencies to assure the public’s health through regulatory and non-regulatory actions, and may even delegate such powers to lower-level agencies. While the states’ legal authorities for preventing and controlling many infectious diseases have been comprehensively described (5), such information has not been well-characterized for chronic diseases.
In addition to constitutions and statutes, other important sources of law affecting public health include administrative law and common law. Administrative law is created by administrative agencies through rules, regulations, orders, and procedures designed to promote policy goals enacted by legislation (6). Responsibility for implementing and enforcing such regulations may be delegated by legislatures to public health departments and other regulatory agencies which, through the processes of issuing and enforcing regulations, create a body of administrative law. Administrative law requirements may govern a spectrum of public health actions that range from the designation of notifiable diseases reported through public health surveillance and the development of sanitation codes to the enforcement of environmental regulations (7).
The United States has a common law system, encompassing what frequently is referred to as "case law," in which judges interpret constitutions and statutes through written opinions that guide the application of the law. Within this system, the U.S. Supreme Court is the final authority over a hierarchy of courts. The hierarchy extends from municipal and other local courts that hear many public health cases, through the district, appeals, and supreme courts of each state, and finally, to federal district and appellate courts. Within this hierarchy, the opinions of an appeals court are binding on subordinate courts. Although state courts in one state do not bind the courts in other states, state courts often are influenced by courts in other states that have considered similar problems. This common law system allows judges to modify constitutions and statutes to adjust to changing conditions and unanticipated problems. Many cases with important ramifications for public health, such as the U.S. Supreme Court’s landmark decision Jacobson v Massachusetts, illustrate this process of applying constitutional provisions to public health situations that were not anticipated by the Constitution (8). Well-known areas of common law include contract, criminal, real property, and tort law, some of which have been important in public health (9). Tort law, for example, has addressed injuries caused by unsafe conditions. Although historically the body of judge-made tort law was almost entirely a common-law creation of judges, most states have now clarified and limited these judicial decisions by statute and regulation.
Legal mechanisms, remedies, and tools
Under the sources of U.S. law described above, public health departments and other governmental agencies, as well as non-governmental organizations and parties, can opt to employ a variety of legal mechanisms, remedies, and tools for applying law to the prevention of diseases and injuries. Legal mechanisms represent several categories of governmental methods and interventions including not only the powers to tax and spend, but also the direct regulation of individuals (e.g., seatbelt requirements) and of businesses (e.g., licensure, inspections, fines, occupational safety standards) (3,7). Public health agencies also can turn to a broad set of remedies and sanctions to enforce regulations. Such remedies include civil sanctions — fines, suspension or revocation of licensure, and injunctions (also known as court orders) requiring termination of a defined activity required by law — and, in some instances, even criminal sanctions (7). Claims for damages under tort and property theories represent an additional legal tool for states, localities, individuals, or groups addressing public health problems. This tool has been used to protect the public from injury risks associated with products such as motor vehicles and tobacco (10).
Jurisdictional levels
The jurisdictions at which mechanisms, remedies, and tools may be applied to public health problems span the federal, state, and local levels. While correspondence between the levels of enactment and application of laws can be straightforward, the fit and interplay of laws and mechanisms can be complicated for a multitude of public health policies and problems. For some problems, there may be a clear relationship between the source(s) of law and its jurisdictional application. For an ordinance enacted by a county commission to ban smoking in restaurants or entertainment venues, for example, the law will be highly specific to a narrowly defined geographic and political jurisdiction. For other problems, however, the relationship between the source(s) of law and the target public health problem may be extremely complex, involving a combination of federal, state, and local laws, and possibly even invoking the principle of preemption — the legal effect resulting when a superior governmental unit blocks an inferior governmental unit from regulation (7).
Examples of legal frameworks in public health practice
Despite the foundational role of law in framing public health, as well as the important roles laws have played as interventions for public health problems, only a limited number of explicit, conceptual legal frameworks have been developed for preventing and controlling diseases and injuries. An historical example of the role of law in the modern public health movement is the Shattuck Report on sanitary conditions in Boston in 1850 (11). That report concluded with a proposed bill establishing a framework for public health regulation. This approach later was applied to zoning and city planning to improve the public’s health by separating residential housing from industrial areas, creating green spaces, and improving lighting and ventilation in multifamily housing.
A more recent example of the role of law in public health is a model — explicitly labeled as a legal framework — for improving the built environment (12). Other examples can be drawn from analyses of legal authorities related to public health problems such as acute disease and public health emergencies, environmental health, injuries, food-borne illnesses, and tobacco use-related diseases (13).
In outlining a model for modifying the design of the built environment to facilitate healthy behaviors and to create conditions for health, the authors noted that educating people about healthy lifestyles is by itself insufficient and that the built environment must allow for people to engage in healthy behaviors (12). The authors suggested that law be used as a tool to achieve the goals of a modified built environment and they proposed a framework of 5 legal approaches: 1) environmental regulation to reduce toxic emissions; 2) zoning ordinances to designate specific uses for areas; 3) building codes to set standards for structures; 4) taxation to encourage or discourage activities; and 5) spending to provide resources for projects that enhance the built environment (12). These legal approaches reflect not only constitutional principles but also the potential use of laws arising from a variety of federal, state, and local legislative enactments, and from administrative agencies.
While not explicit legal frameworks per se, some legal powers have been outlined as tools for addressing other public health problems. Legal authorities necessary for interventions during public health emergencies draw from a combination of constitutional sources, statutory enactments, and applications of the state’s police power — for example, the powers to seize property, abate nuisances, and implement personal control measures such as quarantine, isolation, and mandatory vaccination (14). To control and prevent food-borne diseases, public health and other government agencies at the federal, state, and local levels rely on a set of laws including federal statutes, such as the Federal Food, Drug, and Cosmetic Act and the Federal Meat Inspection Act, the uses of administrative law by agencies possessing delegated regulatory powers, such as the FDA and the Environmental Protection Agency (EPA), and myriad state and local legislated and delegated authorities (15). Other legal tool constructs exist for environmental health, tobacco control, and injury control (16,17). These constructs employ legislative and administrative laws and also identify a prominent role for litigation.
A final example is an analysis of laws related to potentially modifiable risk factors associated with coronary heart disease (CHD) (18). While this analysis was not advanced explicitly as a legal framework for the control and prevention of CHD, it nonetheless offers ideas and elements for a legal framework for addressing this and other chronic disease problems. The author of this analysis examined selected risk factors for CHD, such as smoking, in relation to laws most directly related to modifiable socio-environmental determinants for the factor (e.g., the Federal Cigarette Labeling and Advertising Act, state and local clean indoor air laws), as well as in relation to laws more remotely related to such determinants (e.g., state product liability laws, public health laws, and consumer fraud laws).
Building the CDC’s public health law capacity in chronic disease prevention and control
Systematic legal frameworks in chronic disease prevention and control, like those described in the previous section, can be used to 1) assure that all potential legal avenues are considered; 2) provide a structure within which legal interventions can be monitored for appropriateness and effectiveness; and 3) assist in ensuring that laws, rules, orders, and regulations developed within these frameworks are implemented and enforced.
Specific legal frameworks could be derived for a number of issues within the arena of chronic disease prevention — prevention of heart disease, stroke, diabetes, asthma, obesity, cancer, or complications of diabetes, for example — and health promotion, such as reducing tobacco use, increasing physical activity, and improving nutrition. Indeed, the CDC’s National Center for Chronic Disease Prevention and Health Promotion (NCCDPHP) in collaboration with the CDC’s Public Health Law Program has launched the development of legal frameworks in 2 of these areas — prevention of cardiovascular disease and obesity — and aims to develop additional frameworks in other high-priority areas.
In addition to these issue-specific frameworks, an overarching legal framework could guide the development of legal tools for the entire range of chronic diseases, health behaviors, and environmental conditions. The framework could borrow effective legal tools from one area and apply them creatively to another. An overarching framework could be developed by incorporating the common elements of issue-specific frameworks and by analyzing cross-cutting issues in chronic disease prevention and health promotion.
To build the capacity of NCCDPHP to provide guidance and technical assistance in the emerging field of public health law, we plan to initiate several projects. First, a public health law work group has formed within NCCDPHP to oversee promising activities. An attorney-analyst and medical epidemiologist will coordinate efforts of a cross-NCCDPHP, multidisciplinary team with representatives from each category-specific division — for example, the Division of Adult and Community Health and the Office on Smoking and Health. Initial plans include a one-day meeting with external (non-CDC) program and public health law experts to develop the next steps in building capacity. Priorities include creating category-specific and overarching legal frameworks for chronic disease prevention and health promotion, hosting seminars on public health law and its current and potential uses in chronic disease, and expanding our ability to guide and collaborate with constituents in using legal tools for chronic disease prevention and health promotion.
Existing resources
A growing body of information resources can assist public health professionals and others interested in building organizational capacity for using law as a tool for chronic disease prevention (17,19-24). The initial activities listed above and others will generate future articles in this and other journals. Readers may review statutes and case law within their own jurisdictions and also consult their legal counsel on laws relating to their program's goals. Finally, public health conferences increasingly are offering educational sessions and programs on laws for preventing chronic diseases. The CDC Public Health Law Program Web site offers information on 2 past conferences (25) and on the upcoming conference, The Public's Health and the Law in the 21st Century, to be held June 14–16, 2004, in Atlanta, Ga (26).
Conclusions
This paper outlined the variety of legal tools, remedies, and mechanisms available to public health practitioners and policy makers for achieving public health goals and also examined law as a tool for expanding strategies for preventing and controlling chronic diseases. We emphasize that the use of law should complement, not supplant, existing strategies based on well-established principles of public health practice (27). Law can bolster existing strategies when used prudently by public health practitioners who have a clear understanding of how it shapes public health infrastructure and can promote program goals.
In addition to giving examples of the effectiveness of laws in public health, we described the broad jursiprudential landscape upon which legal frameworks address chronic diseases across a wide array of programs not limited to officially designated public health agencies. Medicare and Medicaid programs, for example, set policies that determine access for large segments of the U.S. population to screening and secondary prevention activities for a number of chronic disease risk factors and conditions. Similarly, the EPA and its state counterparts determine to a large extent exposure levels to airborne and waterborne toxins and to particulates linked to cancer, asthma, and other chronic diseases. Municipal water systems, many independent of officially designated public health agencies, influence oral health through fluoridation policies. Even federal and state revenue agencies, while established largely for other purposes, affect the public’s health through policies that assign taxable status to preventive medical treatments.
Legal frameworks provide exciting opportunities for expanding the spectrum of effective public health strategies. In collaboration with the CDC’s Public Health Law Program, other legal experts, and our external partners, NCCDPHP will continue to explore the development, dissemination, and use of these legal frameworks for the prevention and control of chronic diseases.
We thank Professor Edward P. Richards of the Louisiana State University Law Center for reviewing this manuscript and offering valuable comments.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Mensah GA, Goodman RA, Zaza S, Moulton AD, Kocher PL, Dietz WH, et al. Law as a tool for preventing chronic diseases: expanding the spectrum of effective public health strategies [Part 2]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/04_0009.htm.
==== Refs
1 Mensah GA Goodman RA Zaza S Moulton AD Kocher PL Dietz WH Law as a tool for preventing chronic diseases: expanding the spectrum of effective public health strategies Prev Chronic Dis [serial online] 2004 1 Available from: URL: http://www.cdc.gov/pcd/issues/2004/jan/03_0033.htm
2 Public health law 3rd 1947 Tobey JA New York(NY) The Commonwealth Fund
3 The law and the public’s health: the foundations 2003 3 22 Gostin LO Koplan JP Grad FP Goodman RA Rothstein MA Hoffman RE Lopez W Matthews GW Law in public health practice New York (NY) Oxford University Press
4 Richards EP The jurisprudence of prevention: the right of societal self-defense against dangerous persons Hast Const L Q 1989 18 329 395
5 Gostin LO Burris S Lazzarini Z The law and the public’s health: a study of infectious disease law in the United States Colum L Rev 1999 99 1 59 128
6 Black’s law dictionary Abridged 6th ed 1991 St. Paul (MN) West Publishing Co
7 Regulating public health: principles and applications of administrative law 2003 23 42 Jacobson PD Hoffman RE Goodman RA Rothstein MA Hoffman RE Lopez W Matthews GW Law in public health practice New York (NY) Oxford University Press
8 Jacobson v Massachusetts 1905 197 U.S. 11
9 Regulating public health: principles and applications of administrative law 2003 63 92 Lazzarini Z Scott S Buehler JW Goodman RA Rothstein MA Hoffman RE Lopez W Matthews GW Law in public health practice New York (NY) Oxford University Press
10 Vernick JS Mair JS Teret SP Sapsin JW Role of litigation in preventing product-related injuries Epidemiol Rev 2003 25 90 98 12923993
11 Shattuck L Report of the sanitary commission of Massachusetts 1850 Available from: URL: http://biotech.law.lsu.edu/cphl/history/books/sr/index.htm Harvard University Press 1948
12 Perdue WC Stone LA Gostin LO The built environment and its relationship to the public’s health: the legal framework Am J Public Health 2003 93 1390 1394 12948949
13 Public health law 1998 1147 1154 14th ed Richards EP Rathbun KS Wallace RB Doebbeling BN Maxcy-Rosenau-Last public health and preventive medicine Stamford (CT) Appleton & Lange
14 Legal authorities for interventions during public health emergencies 2003 195 210 Misrahi JJ Matthews GW Hoffman RE Goodman RA Rothstein MA Hoffman RE Lopez W Matthews GW Law in public health practice New York (NY) Oxford University Press
15 Control of foodborne diseases 2003 285 306 Fain K Sobel J Goodman RA Rothstein MA Hoffman RE Lopez W Matthews GW Law in public health practice New York (NY) Oxford University Press
16 Environmental health and protection 2003 371 398 Locke PA Falk H Kochtitzky CS Bump C Goodman RA Rothstein MA Hoffman RE Lopez W Matthews GW Law in public health practice New York (NY) Oxford University Press
17 2003 371 398 Goodman RA Rothstein MA Hoffman RE Lopez W Matthews GW Law in public health practice New York (NY) Oxford University Press
18 Parmet WE The impact of law on coronary heart disease: some preliminary observations on the relationship of law to "normalized" conditions J Law Med Ethics 2002 30 4 608 620 12561267
19 Gostin LO Public health law in a new century: part III: public health regulation: a systematic evaluation JAMA 2000 283 23 3118 3122 10865307
20 Public health law manual: a handbook on the legal aspects of public health administration and enforcement 1990 New York (NY) American Public Health Association Grad FP 2nd
21 Public health law: power, duty, restraint 2000 Berkeley (CA) University of California Press Gostin LO
22 Public health law and ethics: a reader 2002 Berkeley (CA) University of California Press Gostin LO
23 The legal basis of public health 2003 [course online]. Available from: URL: http://www.phppo.cdc.gov/phtn/legal-basis/info.asp Chicago (IL) School of Public Health of the University of Illinois at Chicago
24 Public health law program [Internet]. 2004 Available from: URL: www.phppo.cdc.gov/od/phlp Centers for Disease Control and Prevention, Public Health Practice Program Office Atlanta (GA)
25 Public health law program [Internet]. 2003 Available from: URL: http://www.phppo.cdc.gov/od/phlp/conference/ ConferenceArchiv.asp Centers for Disease Control and Prevention, Public Health Practice Program Office Atlanta (GA)
26 The public's health and the law in the 21st century: third annual conference on public health law [Internet]. 2004 Available from: URL: http://www.phppo.cdc.gov/od/phlp Centers for Disease Control and Prevention, Public Health Practice Program Office Atlanta (GA)
27 Centers for Disease Control and Prevention. Promising practices in chronic disease prevention and control: a public health framework for action. 2003 Available from: URL: http://www.cdc.gov/nccdphp/promising_practices/index.htm Centers for Disease Control and Prevention, Department of Health and Human Services Atlanta (GA)
|
15663886
|
PMC1183502
|
CC BY
|
2022-01-27 23:22:37
|
no
|
Prev Chronic Dis. 2004 Mar 15; 1(2):A11
|
utf-8
|
Prev Chronic Dis
| 2,004 |
nan
|
oa_comm
|
==== Front
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv12_04_0012
Book Review
Evidence-Based Public Health
MacDonald Goldie PhD Steps to a HealthierUS, National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention, Atlanta, Ga
4 2004
15 3 2004
1 2 A12Editors: Brownson Ross C. , Baker Elizabeth A. , Leet Terry L. and Gillespie Kathleen N.
Evidence-Based Public Health.
New York: Oxford University Press, Inc.2002. 10 ISBN: 0-19-5143760 Price: $39.95 256 pages 2004
==== Body
In a relatively short amount of time, the term “evidence-based public health” has flooded dialogues on program planning, implementation, and evaluation. What is evidence-based public health? Abigail Adams reminded us that “[w]e have too many high sounding words, and too few actions that correspond with them” (1). In Evidence-Based Public Health, Brownson and colleagues provide not only a precise definition of a complex term but also a stepwise framework for decision making toward improved public health practice.
The authors order the text according to a 6-step process for enhancing evidence-based decision making in public health: 1) develop an initial statement of the issue; 2) quantify the issue; 3) search the scientific literature and organize the information; 4) develop and prioritize program options; 5) develop an action plan and implement interventions; and 6) evaluate the program or policy. With every step in the process, the authors provide resources for immediate use, including Wide-ranging OnLine Data for Epidemiologic Research (WONDER), a Centers for Disease Control and Prevention (CDC) program; the Community Health Status Indicators Project; the Annual Review of Public Health; evidence-based information on health care outcomes, quality, cost, use, and access via the Agency for Healthcare Research and Quality (AHRQ); the Guide to Community Preventive Services; the Models that Work Campaign to identify and promote innovative community-based models; the Planned Approach to Community Health (PATCH); PRECEDE-PROCEED; and the CDC Working Group on Evaluation.
Evidence-Based Public Health was prepared for 4 main user groups: public health practitioners, policy makers, researchers, and key stakeholders, including the public. The text should be considered necessary reading in schools of public health; the authors artfully marry science and practice with accessible case examples throughout. The combination of practical steps and supporting resources serves as a foundation for decision making in public health as a tangible product of lessons learned via traditional research, the ongoing translation of diverse sources of evidence, and reflective practice. Without question, the authors demystify “evidence-based public health” and delve into intimately related concepts, including the role and varying quality of “best practices” in public health.
Yet, with an increasing emphasis on more integrated, community-based approaches to chronic disease prevention and health promotion, the authors leave room for others to undertake a much-needed discussion of the back-and-forth relationship between emerging or “promising practices” and how this information might expand on existing evidence to inform public health practice now and in the future. As we strive to identify important leverage points for improving community and health outcomes, how will lessons learned in the front lines of public health practice infiltrate that which constitutes “good” evidence? Do community-based practitioners have the resources necessary to evaluate programs and disseminate key findings? Have we created ample pathways for informing evidence-based decision making in public health?
The book lends itself to a follow-up discussion of community-based participatory research as a possible strategy for enhancing the evidence base relevant to program development to address a wide range of existing and emerging health disparities. Moreover, Evidence-Based Public Health highlights the necessity of continued investment in research syntheses, as well as strategies of dissemination that take into account the real-world challenges faced by practitioners in a climate of uncertain resources and increasing calls for accountability to new and diverse stakeholders. To this end, the authors surely set the stage for rich dialogue on a host of issues critical to advancing chronic disease prevention and health promotion in bold new directions.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: MacDonald G. Evidence-Based Public Health [book review]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/04_0012.htm.
==== Refs
1 John Adams McCullough D New York(NY) Simon and Schuster, Inc. 17
|
0
|
PMC1183503
|
CC BY
|
2022-01-27 23:18:41
|
no
|
Prev Chronic Dis. 2004 Mar 15; 1(2):A12
|
utf-8
|
Prev Chronic Dis
| 2,004 |
nan
|
oa_comm
|
==== Front
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv12_04_0007
Letters
Response to S. Leonard Syme’s Essay
Robinson Robert G. DrPH Senior Science Fellow, Associate Director for Program Development Office on Smoking and Health, Centers for Disease Control and Prevention
4 2004
15 3 2004
1 2 A132004
==== Body
To the Editor:
The recent essay by Dr Leonard Syme contributes constructively to the dialogue on disparities (1). It validates the idea that the traditional focus on the individual and risk factors is limited and underscores the importance of environment and community. The complexity of community, however, is not apparent in the essay, and this oversight adds to deficiencies in interventions. Public health needs a different paradigm for assessment and intervention development.
One barrier mentioned by Dr Syme is arrogance, and to that I would add elitism; both prevent experts from relating and adopting paradigms that use community as the unit of analysis. Both challenge diversity and inclusivity, which are necessary for community partnerships. Also troublesome is a limited definition of competency. Dr Syme illustrates the ineffectiveness of interventions in several studies. Others have outlined limitations in addressing community: McKenzie (2) (on the impact of racism and community), Vena and Weiner (3) (on the social determinants of health and community), and Richards, Kennedy, and Krulewitch (4) (on evaluation models that insufficiently encompass community complexity).
Dr Syme uses environment as a metaphor for community, but environmental change is safe verbiage that disguises the limitations of theory and practice. Environmental change factors are merely risk factors writ large. They are reductionist, failing to build a comprehensive understanding of community and reinforcing traditional analyses, which assess outcomes in terms of etiology or predictive factors. They do not assess relationship to community but impose it. Because risk factors relate to individual well-being, we often incorrectly assume they relate to community outcomes.
Dr Syme also uses social status as a metaphor for community. The construct is simple: draw a circle around an entity and name it community. Indeed, Dr Syme defines as community any group that is targeted: citizens of Richmond, Calif, fifth-graders, bus drivers. Each possesses an ethos and a consciousness, but each also lacks the complexity of community. The most critical mistake in targeting a social stratum is creating the illusion that we are targeting a community. We design an intervention for welfare mothers, for example, and write up our findings as a community intervention. But targeting the poor is not the same thing as targeting the community. Change theory derives from the individual unit of analysis and from constructs that do not reflect the complexity inherent in communities.
Another flaw in Dr Syme's essay is the exclusion of race/ethnicity. This exclusion is compounded by insufficiency of community theory and practice and emphasis on etiology and risk factors. Multivariate analysis suggests variables that are important based on statistical significance. Education and income knock race/ethnicity "out of the box." This exclusion is incorrect. Etiology assumes a core role in developing interventions. This may make sense when the unit of analysis is the individual, but it is unfounded when the target is the community.
Communities defined by race/ethnicity magnify the error. Although poverty is the predictive variable, poor people tend not to live in integrated communities. The social reality of imposed segregation is ignored. Indeed, observations of an area of homelessness in Los Angeles showed that white, black, and Latinos each reside on separate street corners (5).
We must develop interventions at 2 levels: by identifying causal factors and deciding at what depth the intervention is to occur and by relating the causal factors to the target population. What do causal factors mean to the population? What is the best protocol for delivery? Superimposing the community over the multivariate analysis is a paradigm shift from traditional biostatistical training, and we need to explore it.
The challenge for the 21st century is to develop theory and practice that resonate with community and its determinants: history, culture, context, and geography. Community competence, a protocol for intervention development, is one solution (6). It avoids the reductionism inherent in cultural competency, and is enhanced by language, literacy, positive imagery, salient imagery, multiple generations, and diversity.
Progress in public health science and practice throughout the 20th century reflects our understanding of the individual. While progress in environmental health has been obvious, progress within race/ethnic communities is not so evident. Upgrading our sanitation and related regulatory protocols benefited populations defined by geography and work site. African Americans and Native Americans continue to demonstrate disparities. Ethnic communities within Latino and Asian/Pacific Islander aggregations demonstrate similar disparities. Why? Our science and practice fails to assess community trends or develop tailored interventions. The 21st century should be the "century of the community," and the emphasis of efforts to improve theory and practice ought to reflect this paradigm.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Robinson R. Response to S. Leonard Syme's essay [letter to the editor]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/04_0005.htm.
==== Refs
1 Syme L Social determinants of health: the community as an empowered partner Prev Chronic Dis [serial online] Available from: URL: URL:http://www.cdc.gov/pcd/issues/2004/jan/03_0001.htm 2004 1 cited 10 Feb 2004
2 McKenzie K Racism and health: antiracism is an important health issue BMJ 2003 1 11 326 65 66 12521953
3 Vena JE Weiner JM Innovative multidisciplinary research in environmental epidemiology: the challenges and needs Int J Occup Med Environl Health 1999 12 4 353 370
4 Richards L Kennedy PH Krulewitch CJ Wingrove B Katz K Wesley B Achieving success in poor urban minority community-based research: strategies for implementing community-based research within an urban minority population Health Promotion Practice 2002 3 3 410 420
5 LeDuff C Skid row still down on its luck International Herald Tribune 2003 7 15 5
6 Robinson RG Journal of Health Education Practice Forthcoming
|
15663887
|
PMC1183504
|
CC BY
|
2022-01-24 23:35:53
|
no
|
Prev Chronic Dis. 2004 Mar 15; 1(2):A13
|
utf-8
|
Prev Chronic Dis
| 2,004 |
nan
|
oa_comm
|
==== Front
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv12_04_0005
Letters
Response to S. Leonard Syme’s Essay
Roberson Robin National Institutes of Health, Bethesda, Maryland
[email protected]
4 2004
15 3 2004
1 2 A142004
==== Body
To the Editor:
The essay "Social Determinants of Health: The Community as an Empowered Partner" by Dr S. Leonard Syme addresses why, in spite of all the biomedical advances and clinical interventions, most clinical studies are designed simply to assess disease and not to assess disease in people, especially those at the bottom of a social structure.
At the National Institutes of Health (NIH), where I have worked as a chemist for the last 20 years, it is a time of great expectations, tremendous growth, and a newfound enthusiasm under a newly proposed NIH roadmap, which holds the promise of translating basic research discoveries from bench to bedside. However, in my African American west Baltimore neighborhood, where I have resided for more than 20 years, that same air of promise and expectation for healthier lifestyles at the hands of current biomedical research does not exist. In fact, changing the face of biomedical research is not high on the list of most people's daily priorities.
Two Saturdays ago, as I was leaving a drug store in my neighborhood, I was privy to a conversation between two elderly black women, one of whom had either just picked up or dropped off a prescription. The very brief yet insightful exchange, in my opinion, sums up the sentiment many African Americans like myself have about today's health care system and health care providers. One lady commented to the other, "You know, I have a young doctor now and he's prescribed this new medication — just experimenting on me." The other lady responded, "Yes, indeed, that's all they're doing — just experimenting." This conversation exemplifies the lack of trust that still exists among African Americans, old as well as young, with health care and health care research.
To rid our country of the health disparities that still exist, first and foremost we must regain the trust of people most affected by health care disparity. Although these ladies may not have a clue about biomedical research, they realize that in spite of their efforts to establish good health regimes, widely accepted interventions might not resolve their health problems. As suggested by Dr Syme, we will achieve this trust only when researchers realize that people are the most valued resource in biomedical research and are humble enough to admit it.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Roberson R. Response to S. Leonard Syme's essay [letter to the editor]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/04_0007.htm.
|
15688514
|
PMC1183505
|
CC BY
|
2022-01-26 23:18:05
|
no
|
Prev Chronic Dis. 2004 Mar 15; 1(2):A14
|
utf-8
|
Prev Chronic Dis
| 2,004 |
nan
|
oa_comm
|
==== Front
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv12_04_0011
Errata
Erratum, Vol. 1, No. 1
4 2004
15 3 2004
1 2 A152004
==== Body
In the book review of Community-Based Health Research: Issues and Methods, one of the editors, Daniel S. Blumenthal, was incorrectly identified as David S. Blumenthal. The information was corrected on our Web site on Monday, December 15, 2003, and the correct information appears below:
Community-Based Health Research: Issues and Methods Editors: Daniel S. Blumenthal, MD, MPH, Ralph J. DiClemente, PhD New York Springer Publishing Company Publication Date: 27 Oct 2003 240 pages Price: $39.95 (domestic), $44.80 (foreign) ISBN: 0-8261-2025-3
When a book on community-based research opens with comments about Imhotep, Hippocrates, and Aesculapius, it certainly catches the reader's attention. Such is the case with Community-Based Health Research: Issues and Methods, a new book edited by Daniel S. Blumenthal and Ralph J. DiClemente.
The mistake appeared in the introductory information for the book as well as in the first paragraph of the review. The corrected article also appears online at http://www.cdc.gov/pcd/issues/2004/jan/03_0031.htm.
We regret any confusion this error may have caused.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
|
0
|
PMC1183506
|
CC BY
|
2022-01-31 23:31:30
|
no
|
Prev Chronic Dis. 2004 Mar 15; 1(2):A15
|
utf-8
|
Prev Chronic Dis
| 2,004 |
nan
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science 1610389810.1371/journal.pcbi.001000105-PLCB-RA-0002plcb-01-01-07Research ArticleAb Initio Prediction of Transcription Factor Targets Using Structural Knowledge Ab Initio Prediction of Target GenesKaplan Tommy 12Friedman Nir 1*Margalit Hanah 2*1 School of Computer Science and Engineering, The Hebrew University, Jerusalem, Israel
2 Department of Molecular Genetics and Biotechnology, Faculty of Medicine, The Hebrew University, Jerusalem, Israel
Sali Andrej EditorUniversity of California at San Francisco, United States of America*To whom correspondence should be addressed. E-mail: [email protected] (NF), [email protected] (HM)6 2005 24 6 2005 1 1 e110 1 2005 11 2 2005 Copyright: © 2005 Kaplan et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Current approaches for identification and detection of transcription factor binding sites rely on an extensive set of known target genes. Here we describe a novel structure-based approach applicable to transcription factors with no prior binding data. Our approach combines sequence data and structural information to infer context-specific amino acid–nucleotide recognition preferences. These are used to predict binding sites for novel transcription factors from the same structural family. We demonstrate our approach on the Cys2His2 Zinc Finger protein family, and show that the learned DNA-recognition preferences are compatible with experimental results. We use these preferences to perform a genome-wide scan for direct targets of Drosophila melanogaster Cys2His2 transcription factors. By analyzing the predicted targets along with gene annotation and expression data we infer the function and activity of these proteins.
Synopsis
Cells respond to dynamic changes in their environment by invoking various cellular processes, coordinated by a complex regulatory program. A main component of this program is the regulation of transcription, which is mainly accomplished by transcription factors that bind the DNA in the vicinity of genes. To better understand transcriptional regulation, advanced computational approaches are needed for linking between transcription factors and their targets. The authors describe a novel approach by which the binding site of a given transcription factor can be characterized without previous experimental binding data. This approach involves learning a set of context-specific amino acid–nucleotide recognition preferences that, when combined with the sequence and structure of the protein, can predict its specific binding preferences. Applying this approach to the Cys2His2 Zinc Finger protein family demonstrated its genome-wide potential by automatically predicting the direct targets of 29 regulators in the genome of the fruit fly Drosophila melanogaster. At present, with the availability of many genome sequences, there are numerous proteins annotated as transcription factors based on their sequence alone. This approach offers a promising direction for revealing the targets of these factors and for understanding their roles in the cellular network.
Citation:Kaplan T, Friedman N, Margalit H (2005) Ab initio prediction of transcription factor targets using structural knowledge. PLoS Comput Biol 1(1): e1.
==== Body
Introduction
Specific binding of transcription factors to cis-regulatory elements is a crucial component of transcriptional regulation. Previous studies have used both experimental and computational approaches to determine the relationships between transcription factors and their targets. In particular, probabilistic models were employed to characterize the binding preferences of transcription factors, and to identify their putative sites in genomic sequences [1,2]. This approach is useful when binding data are available, but cannot be applied to proteins without extensive experimental binding studies. This difficulty is particularly emphasized in view of the genome projects, where new proteins are classified as DNA-binding according to their sequence, yet there is no information about the genes they regulate.
To address the challenge of profiling the binding sites of novel proteins, we propose a family-wise approach that builds on structural information and on the known binding sites of other proteins from the same family. We use solved protein–DNA complexes [3] to determine the exact architecture of interactions between nucleotides and amino acids at the DNA-binding domain. Although sharing the same structure, different proteins from a structural family have different binding specificities because of the presence of different residues at the DNA-binding positions. To predict their binding site motif, we need to identify the residues at these positions and understand their DNA-binding preferences.
In previous studies, we used the empirical frequencies of amino acid–nucleotide interactions [4,5] in solved complexes (from various protein families) to build a set of “DNA-recognition preferences.” This approach assumed similar DNA-binding preferences of the amino acids for all structural domains and at all binding positions. However, there are clear experimental indications that this assumption is not always valid: a particular amino acid may have different binding preferences depending on its positional context [6–8]. To estimate these context-specific DNA-recognition preferences, we need to determine the appropriate context of each residue, which may depend on its relative position and orientation with respect to the nucleotide. Then, we need to collect statistics about the DNA-binding preferences in this context. This can be achieved from an ensemble of solved protein–DNA complexes from the same family. Unfortunately, sufficient data of this type are currently unavailable.
To overcome this obstacle, we propose to estimate context-specific DNA-recognition preferences from available sequence data using statistical estimation procedures. The input of our method is a set of pairs of transcription factors and their target DNA sequences [2]. We then identify the residues and nucleotides that participate in protein–DNA interaction, and collect statistics about the DNA-binding preferences of residues under different contexts of the binding domain. These are then used to discover the binding site of other transcription factors from the same family, for which no targets are known.
We apply our approach to the Cys2His2 Zinc Finger DNA-binding family. This family is the largest known DNA-binding family in multicellular organisms [9] and has been studied extensively [10]. Members of this family bind DNA targets according to a stringent binding model [11,12], which maps the exact interactions between specific residues in the DNA-binding domain with nucleotides at the DNA site (Figure 1). We use many Zinc Finger proteins together with their native DNA targets (extracted from the TRANSFAC database [2]), and apply an iterative expectation maximization (EM) algorithm [13] to estimate position-specific DNA-recognition preferences (Figure 2). These are used in turn for predicting the DNA binding site motifs of novel proteins in the family (Figure 3), and for performing a genome-wide scan for putative targets.
Figure 1 The Canonical Cys2His2 Zinc Finger DNA Binding Model
Residues at positions 6, 3, 2, and −1 (relative to the beginning of the α-helix) at each finger interact with adjacent nucleotides in the DNA molecule (interactions shown with arrows). (Figure adapted from a figure by Prof. Aaron Klug, with permission.)
Figure 2 Estimating DNA-Recognition Preferences
The DNA-recognition preferences are estimated from unaligned pairs of transcription factors and their DNA targets [2] (above). The EM algorithm [13] is used to simultaneously assess the exact binding positions of each protein–DNA pair (bottom right), and to estimate four sets of position-specific DNA-recognition preferences (bottom left).
Figure 3 Predicting the DNA Binding Site Motifs of Novel Transcription Factors
The protein's DNA-binding domains are identified using the Cys2His2 conserved pattern (top left). The residues at the key positions (6, 3, 2 and −1) of each finger (marked in red in the bottom left panel) are then assigned onto the canonical binding model (bottom right), and the sets of position-specific DNA-recognition preferences (top right panel) are used to construct a probabilistic model of the DNA binding site. For example, the lysine at the sixth position of the third finger faces the first position of the binding site (dotted blue arrow). We predict the nucleotide probabilities at this position using the appropriate recognition preferences (dotted black arrow).
Results
In Silico Reconstruction of DNA-Recognition Preferences
In order to estimate the context-specific DNA-recognition preferences of the Cys2His2 Zinc Finger DNA-binding family we used the canonical binding model learned from the solved protein–DNA complex of Egr-1 [11,12]. According to this model, the binding specificity of each Zinc Finger domain is determined by residues at four key positions (see Figure 1). We aimed to learn a different set of DNA-recognition preferences for each of the four key positions. These sets should express the probability of every amino acid to interact with each nucleotide. Since the number of solved protein–DNA complexes is insufficient to estimate these preferences directly, we resorted to sequence data of proteins and their DNA targets. We extracted 455 protein–DNA pairs from the TRANSFAC 7.3 database [2] (see Materials and Methods). Unfortunately, the exact binding locations of these DNA targets are not pinpointed, and thus we employed statistical tools to infer them (see Figure 2; Materials and Methods). We then used the protein–DNA binding model to identify the interacting residues and nucleotides, and collect statistics on their binding preferences (see Materials and Methods). Based on these we estimated four sets of DNA-recognition preferences (Figure 4; Tables S1 and S2), showing both context-independent preferences (such as the preference of lysine for guanine) and context-dependent ones (e.g., the preference of aspartic acid for cytosine). Table S3 shows the 10%–90% confidence intervals of the estimated probabilities.
Figure 4 Four Sets of Position-Specific DNA-Recognition Preferences in Zinc Fingers
The estimated sets of DNA-recognition preferences for the DNA-binding residues at positions 6, 3, 2, and −1 of the Zinc Finger domain are displayed as sequence logos. At each position, the associated distribution of nucleotides is displayed for each amino acid. The total height of letters represents the information content (in bits) of the position, and the relative height of each letter represents its probability. Color intensity indicates the level of confidence for a given amino acid at a certain position (where paler colors indicate lower confidence due to low occurrences of the amino acid at the specific position in the training data) (see Tables S1 and S2 for full data). Some of the DNA binding preferences are general, regardless of the residue's position within the zinc finger (e.g., lysine's tendency to bind guanine), while others are position-dependent (e.g., the tendency of phenylalanine to bind cytosine only when in position 2).
Learned Recognition Preferences Are Consistent with Experimental Results
We evaluated the four reconstructed sets of DNA-recognition preferences by comparing them with experimental data. First, we compared the derived preferences with qualitative preferences based on phage-display experiments [10] and found the two to be consistent (data not shown). Second, we predicted binding site models for Egr-1 variants for which experimental binding data were available [14], using their sequences and our estimated preferences. These models were used to score the binding of Egr-1 variants to a set of DNA targets that were tested in the experimental study. We found that our predictions were highly correlated with the experimentally measured binding affinities [14] (Table S4).
Next, we evaluated the ability of the estimated recognition preferences to identify binding sites within genomic sequences. We compiled a dataset of binding sites of ten Cys2His2 transcription factors. These involved 43 experimentally verified binding sites within natural genomic promoter sequences with a total length of 14,534 bp (Table S5). Using the recognition preferences, we predicted the binding site models of the ten transcription factors and used them to scan the respective promoter regions for putative binding sites (Figure 5A and 5B; see Materials and Methods). To prevent bias by known sites in our training data, we applied a “leave protein out” cross-validation analysis, and predicted the DNA binding model of a protein using DNA-recognition preferences that were learned from a reduced dataset, from which all its binding sites were removed. Our method marked 30 locations as putative binding sites, out of which 21 matched experimental knowledge (sensitivity of 49% and specificity of 70%, p < 10−48; see Table S6).
Figure 5 Validation of DNA-Recognition Preferences
(A) The predicted binding site model of human Sp1 protein is compared to its known site (matrix V$SP1_Q6 from TRANSFAC [2], based on 108 aligned binding sites). To prevent bias by known Sp1 sites in our training data, the set of DNA-recognition preferences was estimated from the TRANSFAC data after removing all Sp1 sites.
(B) Scanning the 300-bp-long promoter of human dihydrofolate reductase (DHFR) by the predicted Sp1 binding model. The p-value of each potential binding site is shown (y-axis). Four positions achieved a significant p-value (higher than the horizontal red line), out of which three are known Sp1 binding sites [41] (arrows).
(C) A summary of in silico binding experiments for 21 pairs of Zinc Finger transcription factors and their target promoters. Shown is the tradeoff between false positive rate (x-axis) and true positive rate (y-axis) as the significance threshold for putative binding sites is changed. For every threshold point, our set of recognition preferences (EM) achieves better accuracy than the preferences of Mandel-Gutfreund et al. [5] (M-G) and Benos et al. [15] (SAMIE). Interestingly, when the DNA-recognition preferences were estimated from training data expanded to include TRANSFAC's artificial sequences, inferior results were obtained (dotted red line).
(D) Cumulative distribution of Sp1 scores among the sequences of targets/non-targets of unbiased chromatin immunoprecipitation scans of human Chromosomes 21 and 22 [16]. The predicted Sp1 motif appears in 45% of the experimentally bound sequences but in only 5% of the control sequences.
Benos et al. [15] proposed a method (SAMIE) to estimate Cys2His2 Zinc Finger position-specific binding preferences from in vitro SELEX binding experiments. We compared the predictions of the known binding sites within promoter regions provided by our position-specific recognition preferences to those of Benos et al. [15] and of Mandel-Gutfreund et al. [5] (Figure 5C; Table S7). These results suggest that predictions based on our recognition preferences out-perform the predictions based on the other methods.
To further evaluate our predictions, we used the binding locations of Sp1 along human Chromosomes 21 and 22, as mapped by genome-wide chromatin immunoprecipitation [16]. We compiled two datasets of 1-kb-long sequences: one dataset included sequences that exhibited highly significant binding, and the other dataset included sequences that showed no binding at all (to be used as a control; see Materials and Methods). We used the DNA-recognition preferences to predict a binding site model for Sp1, and scanned the genomic sequences with it. We identified Sp1 binding sites in 45% of the experimentally bound sequences, and in only 5% of the control sequences (Figure 5D).
Ab Initio Genome-Wide Prediction of Transcription Factor Binding Sites
In the past few years many genomes were solved, yielding sequences of thousands of putative transcription factors. However, only little is currently known about the binding specificities of these factors and about their target genes. To address this problem, we applied our predictive scheme to the Drosophila melanogaster genome in a fully automated manner. We first scanned the sequences of 16,201 putative gene products and identified 29 canonical Cys2His2 Zinc Finger transcription factors with three or four fingers (see Materials and Methods). We then used their sequences and the estimated DNA-recognition preferences to compile a binding site model for each transcription factor, as in Figure 3 (see Figure S1 and Table S8 for detailed models). Finally, we used these binding site models to scan the upstream promoter regions of 15,665 D. melanogaster genes. Multiple putative direct targets were predicted for each Zinc Finger, as detailed at http://compbio.cs.huji.ac.il/Zinc. The number of putative direct target genes for each transcription factor and the overlap between targets of different factors are shown in Figures S2 and S3. Interestingly, several Zinc Fingers have similar residues at the DNA-binding positions, and are therefore predicted to bind similar sites and to have mutual predicted targets (see Figures S1 and S3). In D. melanogaster, this phenomenon has been reported for at least some transcription factors (e.g., Sp1 and Btd) [17].
To infer the function of the 29 transcription factors, we employed the functional annotations of their predicted target genes (based on the Gene Ontology [GO] terms [18]). The target sets of most transcription factors (21 out of 29) were found to be significantly enriched with at least one GO term (Figure 6A). For some of the transcription factors, the enriched GO terms match prior biological knowledge. For example, the putative targets of Glass were found to be enriched with terms related to photoreceptor cell development, consistent with previous studies that linked the Glass transcription factor with eye photoreceptor development [19]. Similarly, the putative targets of Btd and Sp1 were enriched with developmental terms, such as neurogenesis, development, and organogenesis. Indeed these regulators are known to play essential roles in mechanosensory development [17]. Furthermore, our analysis suggests possible functions for unknown proteins, as well as new annotations for some of the already known regulators (see Figure S4 for complete results).
Figure 6 Inferring the Function and Activity of Zinc Finger Transcription Factors in D. melanogaster
(A) Similar gene annotation enrichment among the putative target sets of 29 transcription factors in D. melanogaster. Blue cells correspond to significant overabundance of a GO term (row) among the predicted targets of a protein (column), using a hyper-geometric test. The binding sites of most factors show enrichment in at least one GO term. For some of the regulators, the enriched GO terms match prior biological knowledge. For example, the putative targets of Glass (gl) were found to be enriched with terms related to photoreceptor cell development (red circle 1). Similarly, the putative targets of Buttonhead (btd) and Sp1 were enriched with developmental terms such as neurogenesis, development, and organogenesis (red circle 2). Closely related GO annotations are not shown; see Figure S4 for full results.
(B) Deducing the activity of the 29 transcription factors using gene expression patterns. Expression data from early (0–12 h) embryogenesis [20] and data from the entire Drosophila life cycle [21] are used to test whether the putative direct targets of a regulator are expressed differently than the rest of the genes in a given experiment. Red cells correspond to significant enrichment of overexpressed targets using a Kolmogorov-Smirnov test, while green cells correspond to enrichment of underexpressed targets. For most of the regulators the analysis resulted in at least one significant embryogenesis experiment, suggesting an active role in early developmental stages (above). Similar results were obtained using the full life cycle gene expression data (below).
We further evaluated the function and activity of the 29 transcription factors based on the mRNA expression profiles of their target genes (Figure 6B). We used expression data from early embryogenesis [20], as well as data from the entire life cycle of D. melanogaster [21]. In each experiment and for each transcription factor, we tested whether its putative targets showed similarity in their expression patterns and differed from the rest of the genes (see Materials and Methods). Such coherent expression supports the suggested relationship between the factor and the genes it is predicted to regulate. Out of the 29 transcription factors we examined, 21 showed such significant associations in at least one embryogenesis experiment, suggesting their active roles throughout early developmental stages (Figure 6B). These transcription factors include many known developmental regulators that are active during embryonic development (e.g., Btd, Sp1, Glass, Odd-skipped, and Stripe) [18,22], as well as other proteins, whose function is currently unknown. Similar results were obtained using the full life cycle gene expression data [21], mapping putative time points at which each regulator is predicted to be active (Figure 6B).
Note that the expression profiles are based on whole embryos, and therefore ignore spatially differential expression patterns. Thus, the correct function of some tissue-specific Zinc Finger proteins may be obscured in these data. Additional insight may be gained by focusing on expression data in homogeneous regions. Specifically, Butler et al. [23] compared gene expression in two homogeneous parts of the Drosophila imaginal wing disc—the body wall and the hinge-wing pouch. In our analysis we used the ratios between the expression levels in the two regions, and examined putative targets for enrichment in one of the regions. We then inferred the regulatory role of a transcription factor (activator or repressor) using its own expression pattern. For example, the putative targets of Stripe show higher expression levels in the body wall than the rest of the genes (enrichment p-value ≤ 0.0002). Stripe itself is enriched more than 9-fold in the body wall, relative to the wing-hinge region. This suggests that Stripe functions mainly in the body-wall region, where it activates its target genes. Indeed, this is consistent with the known role of Stripe as an activator of epidermal muscle attachment genes [24]. Using the same reasoning, we inferred the regulatory roles of four additional D. melanogaster transcription factors within the imaginal wing disc, three of which were previously uncharacterized (Table 1).
Table 1 Analysis of Differential Expression in D. melanogaster Imaginal Wing Disc
Butler et al. [23] measured the gene expression levels at two parts of the imaginal wing disc—the body wall and hinge-wing pouch, and computed the ratios between the two. The regulatory functions of transcription factors are analyzed by comparing their ratio with the ratios of their targets. Activators are expected to have the same directional enrichment as their targets, while repressors are expected to have opposite effects. Each group of targets is assigned a p-value using a two-tailed Kolmogorov-Smirnov test that compares the ratios in the target group to those of the rest of the genes.
aRatio between the transcription factor's mRNA expression levels at the body wall and the wing-hinge pouch.
b
p-Value of targets' enrichment using a Kolmogorov-Smirnov test.
Discussion
In this paper we propose a general framework for predicting the DNA binding site sequence of novel transcription factors from known families. Our framework combines structural information about a specific DNA-binding domain with examples of binding sites for proteins in the family. We apply a statistical estimation algorithm to the canonical Cys2His2 Zinc Finger DNA-binding family, and derive a set of DNA-recognition preferences for each residue at each interacting position in the Zinc Finger DNA-binding domain.
We apply these preferences and predict the binding site models of novel proteins from the same family. Finally, we use the predicted models in genome-wide scans and identify the proteins' putative direct target genes.
Structure-based approaches for prediction of transcription factor binding sites have recently gained much interest [5,8,15,25–29]. Most of the current structural approaches define a binding model based on solved protein–DNA complexes, and attempt to identify DNA subsequences that best fit the amino acids that are determined as interacting with the DNA. Previous studies [4,8] used ensembles of solved protein–DNA complexes (from all DNA-binding domains) to extract general parameters for amino acid–base recognition. Some studies used only the counts of amino acid–nucleotide pairs to derive these parameters [4], whereas others also considered the spatial arrangements [8]. However, for fine grained definition of such potentials, a much larger set of solved protein–DNA complexes is needed than is currently available. An alternative approach to estimate DNA-recognition preferences is to extract them separately for each DNA-binding domain. However, here again, the data of solved complexes are insufficient to allow such derivation.
In a recent study, Benos et al. [15] assigned position-specific DNA-recognition preferences for the Cys2His2 Zinc Finger family. The model they used is similar to ours, with two significant differences. First, they relied on data from in vitro selection assays, such as SELEX and phage display, to train their recognition preferences. Second, their assays screened artificial sequences, both artificial proteins and artificial DNA targets. In contrast, we rely on previously published information of natural binding sites. Our approach does not require specialized experiments, and more importantly, it captures the specificity of natural proteins to DNA sequences. As we showed, our preferences are consistent with independent experimental results [6,7,10] and are superior to such preferences derived by the other computational approaches [5,15]. In addition, previous studies showed that there are discrepancies between SELEX-derived motifs and those derived from natural binding sites [30,31]. Indeed, our method yielded inferior predictions when information on artificial binding sequences was included in our training data. Figure 4C shows that our set of recognition preferences is superior to previous models in identifying genomic binding sites. When comparing the predictions by the various recognition preferences to measured affinities of DNA artificial sequences [14], we report results similar to those of Benos et al. (see Table S4).
Analysis of the Estimated DNA-Recognition Preferences
Analysis of the estimated recognition preferences suggests that the protein–DNA recognition code is not deterministic, but rather spans a range of preferences. Moreover, our analyses show that a residue may have different nucleotide preferences depending on its context. For some amino acids, the qualitative preferences remain the same across various positions, while the quantitative preferences vary (e.g., arginine; see Figure 4). The DNA-binding preferences of other residues change across various positions. For example, histidine at position 3 tends to interact with guanine, while it shows no preference to any nucleotide at all other positions. Another example is the tendency of alanine at position 6 to face guanine. This preference, which was revealed automatically by our analysis, is not consistent with the chemical nature of alanine's side chain nor with general examinations of amino acid–nucleotide interactions [5,8]. We suspect that it is affected by the large number of Sp1 targets in our dataset. This potential interaction was implied before in Sp1 binding sites [32] and may reflect an interaction between the residue at position 2 with the complementary cytosine.
The Protein–DNA Binding Model
In this work, we use a binding model that is based on solved protein–DNA complexes. The model presents a rigid and simplistic representation of the amino acid–base interactions at the Zinc Finger domains. Only some of the Zinc Finger domains (termed “canonical” in this work) use this model for binding, while others maintain more complex interactions. As our results show, by using this model, we manage to recover most of the DNA-binding specificities of amino acids, and use them to predict the binding site models of novel proteins. We believe that this model offers a fair tradeoff between complexity (and number of parameters) and accuracy.
Inter-Position Dependencies in the Binding Site
The Cys2His2 binding model inherently assumes that all positions within the binding site are independent of each other. This assumption is used in most computational approaches that model binding sites. Two recent papers [33,34] discuss this issue in the context of the Cys2His2 Zinc Finger domain. Their analyses of binding affinity measurements suggest that weak dependencies do exist among some positions of the binding sites of Egr-1. Nonetheless, a reasonable approximation of the binding specificities is obtained even when ignoring these dependencies. In another recent study [35], we evaluated probabilistic models that are capable of capturing inter-position dependencies within binding sites. Our results show that dependencies can be found in the binding sites of many proteins from various DNA-binding domains (especially from the helix-turn-helix and the homeo domains). However, our results also suggest that such models of dependencies do not lead to significant improvements in modeling the binding sites of Zinc Finger proteins. Thus, we believe that the Cys2His2 binding model we use here is indeed a reasonable approximation of the actual binding.
Genome-Wide Predictions of Binding Sites and Target Genes
In the current era there is a growing gap between the number of known protein sequences and the number of experimentally verified binding sites. To better understand regulatory mechanisms in newly solved genomes, it is crucial to identify the direct target genes of novel DNA-binding proteins. Our method opens the way for such genome-wide assays. Here we apply it to the Cys2His2 Zinc Finger DNA-binding family. By predicting the binding site models of regulatory proteins, one can classify genes into those that contain significant binding sites at their regulatory promoter regions (hence, putative target genes) and those that do not. As we showed, our approach can scale up to such genome-wide scans and successfully predict the target genes of many novel Zinc Finger proteins in higher eukaryotes. Furthermore, by integrating data from external sources, such as gene expression and GO annotations, it is possible to infer the cellular function and activity of these novel proteins.
Applications to Other DNA-Binding Domains
Theoretically, our approach can be extended to handle other structural families, such as the basic leucine zipper, the homeodomain, and the basic helix-loop-helix, for which enough binding data already exist (1,191, 505, and 201 binding sites per family, respectively). This extension requires that the various proteins in the family show a common DNA binding model, which can be used further for other family members. For such families, our approach should suffice. For other families, where the binding models are more complex and flexible (including other Zinc Finger domains, such as CCCC, CCHC, or even the non-canonical Cys2His2), more advanced models and learning techniques will be needed. In spite of these possible difficulties, we believe that structural approaches, such as the one we show here, open promising directions, leading to successful predictions of binding site models and, following that, to accurate identification of the target genes of novel proteins, even on genome-wide scales. Eventually, such approaches will be utilized to reconstruct larger and larger portions of the transcriptional regulatory networks that control the living cell.
Materials and Methods
Sequences of Zinc Finger proteins and their binding sites.
We trained a profile hidden Markov model [36] on 31 experimentally determined canonical domains [37], and used it to classify the remaining Cys2His2 Zinc Finger domains in TRANSFAC [2] as canonical or non-canonical. From these, we selected proteins with two to four properly spaced canonical fingers. This resulted in 61 canonical Cys2His2 Zinc Finger proteins, and 455 protein–binding site pairs. We used these pairs as our training data in subsequent steps. The total number of fingers in this dataset was 1,320, and the total length of all binding sites was 9,761 bp (average length of 21 bp per site).
Identification of DNA-binding residues.
The interacting residues in each finger are located at positions 6, 3, 2, and −1 relative to the beginning of the α-helix (see Figure 1). We identify these positions using their relative positioning in the Cys2His2 conserved pattern: CX(2–4)CX(11–13)HX(3–5)H. Although, theoretically there can be 204 different combinations of amino acids at the interacting positions, we found only 80 different combinations among the 1,320 fingers in our database. Figures S5 and S6 show the abundance of amino acids at the different DNA-binding positions.
The probabilistic model.
We describe the binding preferences of a protein using a probabilistic model. For a canonical Egr-1-like Zinc Finger protein, we denote by A = {Ai,p : i = {1,…, k}, p ∈ {−1,2,3,6}} the set of interacting residues in the different four positions of the k fingers (ordered from the N- to the C-terminus). Let N
1,…, NL be a target DNA sequence. The conditional probability of an interaction with a DNA subsequence, starting from the jth position in the DNA is
where Pp(N|A) is the conditional probability of nucleotide N given amino acid A at position p. These probabilities are the parameters of the model. For each of the four interacting positions there is a matrix of the conditional probabilities of the four nucleotides given all 20 residues. We call these matrices the DNA-recognition preferences.
The model, as described above, does not account for the interactions by the amino acid in position 2 in each finger. According to the canonical binding model (see Figure 1), the amino acid at position 2 interacts with the nucleotide that is complementary to the nucleotide interacting with position 6 of the previous finger. Thus, when we have a base pair interacting with two amino acids, we replace the term P
6(Nj
+3(i−1)|Ak
+1−i,6) with the term
for i > 1, where α is a weighting coefficient that depends on the number of examples we have seen while estimating the recognition preferences at each position. Moreover, we add the term P
2(Nj
+3(i−1)|Ak
+2−i,2), for i = k + 1, to capture the last nucleotide, which is in interaction with position 2 of the first finger.
Estimating DNA-recognition preferences.
We searched for the DNA-recognition preferences that maximized the likelihood of the DNA sites given the binding proteins. The DNA sequences in our database were reported as containing the binding sites [2], yet the exact binding locations were not pinpointed. Thus, we simultaneously identified the exact binding locations and maximum likelihood recognition preferences using the iterative EM algorithm [13]. Starting with an initial choice of DNA-recognition preferences (possible choices are discussed below), the algorithm proceeds iteratively, by carrying out two steps. In the E-step, the expected posterior probability of binding locations is computed for every protein–DNA pair. This is done using the current sets of preferences. In the M-step, the DNA-recognition preferences are updated to maximize the likelihood of the current binding positions for all protein–DNA pairs based on the distribution of possible binding locations computed in the E-step.
Each iteration of these two steps increases the likelihood of the data until reaching a convergence point [13]. Although the EM algorithm is proved to converge, it does not ensure that the final DNA-recognition preferences are the optimal ones, because of suboptimal local maxima of the likelihood function. This can be overcome by using promising starting points or applying the EM procedure with multiple random starting points (see Figure S7). An additional potential pitfall is over-fitting the recognition preferences of rare residues. To address this problem and ensure that the estimated recognition preferences for rare amino acids are close to uniform distribution (i.e., uninformative), we use a standard method of “pseudo-counts.” We do so by adding a constant (0.7 in the results above) to each amino acid–nucleotide count computed at the end of the E-step. This is equivalent to using a Dirichlet prior on the parameters, and then performing a maximum a posteriori estimation rather than maximum likelihood estimation.
We evaluated the robustness and convergence rate of the EM procedure using a 10-fold cross-validation procedure. In each round, we removed a part of the data, trained on the remaining pairs, and tested the likelihood of the held-out protein–DNA pairs. We used this procedure to test various initialization options. Our evaluation shows that the EM algorithm performs best when initialized with the general recognition preferences of Mandel-Gutfreund et al. [5], converging within a few iterations. Similar results were obtained using random initialization points, although the convergence rate was somewhat slower (see Figure S7). Also, in Figure S8 we demonstrate the correlation between the size of the training dataset and the likelihood of test data.
Predicting the binding sites of novel proteins.
Given the sequence of a novel Cys2His2 Zinc Finger protein, we identified the four key residues at each DNA-binding domain, and utilized the appropriate set of DNA-recognition preferences to construct a probabilistic model of the binding site (see Figure 3).
In silico binding experiments.
We used the predicted binding site models to scan genomic sequences for putative binding sites. We scored each possible binding position using the log of the ratio between the probability assigned to it by the model and the background probability (log-odds score). We then estimated the p-value of these scores and applied a Bonferroni correction to account for multiple tests within the same promoter region [38]. Sites with a significant p-value (≤0.05 after Bonferroni correction) were marked as putative binding sites (see Figure 4B).
Comparison with other computational approaches.
In a similar manner, we generated probabilistic binding site models for these transcription factors using the recognition preferences of Mandel-Gutfreund et al. [5] and SAMIE [15]. We then scanned the corresponding promoter regions using these models.
Ab initio genome-wide prediction of binding sites.
We downloaded genomic sequences of the D. melanogaster from FlyBase [22], release 3–1. These include 2-kb regulatory regions upstream from 15,664 genes, and the sequences of 16,201 putative gene products. We scanned the proteins for canonical Zinc Finger domains using the Cys2His2 conserved pattern and our profile-HMM model (available at http://compbio.cs.huji.ac.il/Zinc). We found 29 proteins with properly spaced three or four fingers (with distances of 28–31 residues between the beginnings of Zinc Finger domains). We then used the learned sets of DNA-recognition preferences to predict probabilistic binding site models for these putative Zinc Finger transcription factors. Finally, we performed in silico binding experiments by scanning each gene's 2-kb upstream region for two significant binding sites (p ≤ 0.05 after Bonferroni correction). The matched genes were marked as putative direct targets of the transcription factor.
Enrichment of GO annotations among the target genes.
FlyBase GO annotations [18,22] were downloaded from the Gene Ontology Consortium (http://www.geneontology.org) in October 2003. The enrichment p-values were calculated by GeneXPress (http://genexpress.stanford.edu), using a hyper-geometric test that compares the abundance of similarly annotated genes among the putative targets to the rest of the genome. We then applied an FDR correction for multiple hypotheses using a false rate of 0.05 [39], and only significant factors/terms are shown.
Inference of activity/function using gene expression data.
We downloaded genome-wide gene expression data from early embryogenesis stages [20] (available from FlyBase; http://www.fruitfly.org/cgi-bin/ex/insitu.pl). The expression level of each gene in each array was transformed to log (base 2) of the ratio of expression to the geometric average of the expression of the gene in all arrays. In addition, we downloaded expression data from along the Drosophila life cycle [21] (available from Stanford Microarray Database; http://genome-www5.stanford.edu). These expression data are represented as log (base 2) of expression compared to a reference sample representing all stages of the life cycle.
For each protein and in each experiment, we used a Kolmogorov-Smirnov test to evaluate whether the expression pattern of the putative direct target genes was different from the expression of the rest of the genome. We then corrected the results for multiple hypotheses using an FDR correction [39] (false rate of 0.05). Similarly, we used differential gene expression data from D. melanogaster imaginal wing disc [23]. For each gene, we computed the ratio of its expression in the body wall to its expression in the hinge-wing pouch, and performed a two-tailed version of the Kolmogorov-Smirnov test to compare these ratios among the putative targets and the rest of the genome.
Supporting Information
Figure S1 Sequence logos of 29 Drosophila Transcription Factors
(617 KB PDF).
Click here for additional data file.
Figure S2 Number of Predicted Direct Targets
(162 KB PDF).
Click here for additional data file.
Figure S3 Percentage of Pairwise Coverage between Targets
(109 KB PDF).
Click here for additional data file.
Figure S4 Results of Complete GO Table
(182 KB PDF).
Click here for additional data file.
Figure S5 Abundance of DNA-Binding Residues in Training Data
(123 KB PDF).
Click here for additional data file.
Figure S6 Abundance of Combinations of DNA-Binding Residues in Training Data
(123 KB PDF).
Click here for additional data file.
Figure S7 Convergence of the EM Algorithm on Held-Out Test Data
(106 KB PDF).
Click here for additional data file.
Figure S8 Likelihood of Held-Out Test Data Given Different Sizes of the Training Datasets
(106 KB PDF).
Click here for additional data file.
Table S1 Four Sets of DNA-Recognition Preferences: Probabilities
(22 KB PDF).
Click here for additional data file.
Table S2 Four Sets of Recognition Preferences: Counts
(20 KB PDF).
Click here for additional data file.
Table S3 Confidence Intervals on Four Sets of DNA-Recognition Preferences
(63 KB PDF).
Click here for additional data file.
Table S4 Correlation with Experimentally Measured Binding Affinities
(514 KB TIF).
Click here for additional data file.
Table S5 21 Protein–DNA Pairs
(2 MB TIF).
Click here for additional data file.
Table S6 Sensitivity and Specificity of Test Set at Different Significance Threshold Values
(328 KB TIF).
Click here for additional data file.
Table S7 Sensitivity and Specificity of Test Set at Different Significance Threshold Values—Other Computational Methods
(440 KB TIF).
Click here for additional data file.
Table S8 Position-Specific Score Matrices of 29 Cys2His2 Transcription Factors from Drosophila melanogaster
(55 KB PDF).
Click here for additional data file.
The authors wish to thank Yael Altuvia, Yoseph Barash, Ernest Fraenkel, Benjamin Gordon, Robert Goldstein, Ruth Hershberg, Dalit May, Lena Nekludova, Aviv Regev, and Eran Segal for helpful discussions. TK is supported by the Yeshaya Horowitz Association through the Center for Complexity Science. NF is supported by the Harry and Abe Sherman Senior Lectureship in Computer Science. This work was supported by grants from the Israeli Ministry of Science and the Israeli Science Foundation. A preliminary version of this manuscript appeared in RECOMB 2005 [40].
Competing interests. The authors have declared that no competing interests exist.
Author contributions. TK, NF, and HM conceived and designed the experiments. TK performed the experiments. TK, NF, and HM analyzed the data and wrote the paper.
Abbreviations
EMexpectation maximization
GOGene Ontology
==== Refs
References
Stormo GD 2000 DNA binding sites: Representation and discovery Bioinformatics 16 16 23 10812473
Wingender E Chen X Fricke E Geffers R Hehl R 2001 The TRANSFAC system on gene expression regulation Nucleic Acids Res 29 281 283 11125113
Luscombe NM Laskowski RA Thornton JM 2001 Amino acid–base interactions: A three-dimensional analysis of protein–DNA interactions at an atomic level Nucleic Acids Res 29 2860 2874 11433033
Mandel-Gutfreund Y Margalit H 1998 Quantitative parameters for amino acid–base interaction: Implications for prediction of protein–DNA binding sites Nucleic Acids Res 26 2306 2312 9580679
Mandel-Gutfreund Y Baron A Margalit H 2001 A structure-based approach for prediction of protein binding sites in gene upstream regions Pac Symp Biocomput 2001 139 150
Choo Y Klug A 1994 Toward a code for the interactions of zinc fingers with DNA: Selection of randomized fingers displayed on phage Proc Natl Acad Sci U S A 91 11163 11167 7972027
Choo Y Klug A 1994 Selection of DNA binding sites for zinc fingers using rationally randomized DNA reveals coded interactions Proc Natl Acad Sci U S A 91 11168 11172 7972028
Kono H Sarai A 1999 Structure-based prediction of DNA target sites by regulatory proteins Proteins 35 114 131 10090291
Tupler R Perini G Green MR 2001 Expressing the human genome Nature 409 832 833 11237001
Wolfe SA Greisman HA Ramm EI Pabo CO 1999 Analysis of zinc fingers optimized via phage display: Evaluating the utility of a recognition code J Mol Biol 285 1917 1934 9925775
Pavletich NP Pabo CO 1991 Zinc finger-DNA recognition: Crystal structure of a Zif268–DNA complex at 2.1 A Science 252 809 817 2028256
Elrod-Erickson M Benson TE Pabo CO 1998 High-resolution structures of variant Zif268–DNA complexes: Implications for understanding zinc finger–DNA recognition Structure 6 451 464 9562555
Dempster AP Laird NM Rubin DB 1977 Maximum likelihood from incomplete data via the EM algorithm J R Stat Soc Ser B 39 1 38
Bulyk ML Huang X Choo Y Church GM 2001 Exploring the DNA-binding specificities of zinc fingers with DNA microarrays Proc Natl Acad Sci U S A 98 7158 7163 11404456
Benos PV Lapedes AS Stormo GD 2002 Probabilistic code for DNA recognition by proteins of the EGR family J Mol Biol 323 701 727 12419259
Cawley S Bekiranov S Ng HH Kapranov P Sekinger EA 2004 Unbiased mapping of transcription factor binding sites along human chromosomes 21 and 22 points to widespread regulation of noncoding RNAs Cell 116 499 509 14980218
Schock F Purnell BA Wimmer EA Jackle H 1999 Common and diverged functions of the Drosophila gene pair D-Sp1 and buttonhead Mech Dev 89 125 132 10559487
Harris MA Clark J Ireland A Lomax J Ashburner M 2004 The Gene Ontology (GO) database and informatics resource Nucleic Acids Res 32 D258 D261 14681407
Moses K Ellis MC Rubin GM 1989 The glass gene encodes a zinc-finger protein required by Drosophila photoreceptor cells Nature 340 531 536 2770860
Tomancak P Beaton A Weiszmann R Kwan E Shu S 2002 Systematic determination of patterns of gene expression during Drosophila embryogenesis Genome Biol 3 RESEARCH0088 12537577
Arbeitman MN Furlong EE Imam F Johnson E Null BH 2002 Gene expression during the life cycle of Drosophila melanogaster
Science 297 2270 2275 12351791
FlyBase Consortium 2003 The FlyBase database of the Drosophila genome projects and community literature Nucleic Acids Res 31 172 175 12519974
Butler MJ Jacobsen TL Cain DM Jarman MG Hubank M 2003 Discovery of genes with highly restricted expression patterns in the Drosophila wing disc using DNA oligonucleotide microarrays Development 130 659 670 12505997
Vorbruggen G Jackle H 1997 Epidermal muscle attachment site-specific target gene expression and interference with myotube guidance in response to ectopic stripe expression in the developing Drosophila epidermis Proc Natl Acad Sci U S A 94 8606 8611 9238024
Suzuki M Gerstein M Yagi N 1994 Stereochemical basis of DNA recognition by Zn fingers Nucleic Acids Res 22 3397 3405 8078776
Steffen NR Murphy SD Tolleri L Hatfield GW Lathrop RH 2002 DNA sequence and structure: Direct and indirect recognition in protein–DNA binding Bioinformatics 18 S22 S30 12169527
Endres RG Schulthess TC Wingree NS 2004 Toward an atomistic model for predicting transcription-factor binding sites Proteins 57 262 268 15340913
Havranek JJ Duarte CM Baker D 2004 A simple physical model for the prediction and design of protein-DNA interactions J Mol Biol 344 59 70 15504402
Paillard G, Deremble C, Lavery R 2004 Looking into DNA recognition: Zinc finger binding specificity Nucleic Acids Res 32 6673 6682 15613596
Robison K McGuire AM Church GM 1998 A comprehensive library of DNA-binding site matrices for 55 proteins applied to the complete Escherichia coli K-12 genome J Mol Biol 284 241 254 9813115
Shultzaberger RK Schneider TD 1999 Using sequence logos and information analysis of Lrp DNA binding sites to investigate discrepancies between natural selection and SELEX Nucleic Acids Res 27 882 887 9889287
Berg JM 1992 Sp1 and the subfamily of zinc finger proteins with guanine-rich binding sites Proc Natl Acad Sci U S A 89 11109 11110 1454785
Benos PV Bulyk ML Stormo GD 2002 Additivity in protein–DNA interactions: How good an approximation is it? Nucleic Acids Res 30 4442 4451 12384591
Bulyk ML Johnson PLF Church GM 2002 Nucleotides of transcription factor binding sites exert interdependent effects on the binding affinities of transcription factors Nucleic Acids Res 30 1255 1261 11861919
Barash Y Elidan G Friedman N Kaplan T 2003 Modeling dependencies in protein–DNA binding sites Vingron M Istrail S Pevzner P Waterman M Proceedings of the Seventh International Conference on Research in Computational Molecular Biology New York ACM Press 28 37
Eddy SR 1998 Profile hidden Markov models Bioinformatics 14 755 763 9918945
Wolfe SA Nekludova L Pabo CO 2000 DNA recognition by Cys2His2 zinc finger proteins Annu Rev Biophys Biomol Struct 29 183 212 10940247
Barash Y Elidan G Kaplan T Friedman N 2005 CIS: compound importance sampling method for protein-DNA binding site p-value estimation Bioinformatics 21 596 600 15454407
Benjamini Y Hochberg Y 1995 Controlling the false discovery rate: A practical and powerful approach to multiple testing J R Stat Soc Ser B 57 289 300
Kaplan T Friedman N Margalit H 2005 Predicting transcription factor binding sites using structural knowledge Miyano S Mesirov JP Kasif S Istrail S Pevzner PA Proceedings of the Ninth International Conference on Research in Computational Molecular Biology: Lecture notes in computer science, Volume 3,500 Berlin Springer-Verlag 522 537
Kriwacki RW Schultz SC Steitz TA Caradonna JP 1992 Sequence-specific recognition of DNA by zinc-finger peptides derived from the transcription factor Sp1 Proc Natl Acad Sci U S A 89 9759 9763 1329106
|
16103898
|
PMC1183507
|
CC BY
|
2021-01-05 09:18:21
|
no
|
PLoS Comput Biol. 2005 Jun 24; 1(1):e1
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010001
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science 1610390310.1371/journal.pcbi.001000205-PLCB-RA-0010R1plcb-01-01-04Research ArticleBioinformatics - Computational BiologyMicrobiologySystems BiologyEubacteriaWhat Makes Ribosome-Mediated Transcriptional Attenuation Sensitive to Amino Acid Limitation? What Makes Transcriptional Attenuation Sensitive?Elf Johan Ehrenberg Måns Department of Cell and Molecular Biology, Uppsala University, Uppsala, SwedenMattick John Stanley EditorUniversity of Queensland, AustraliaE-mail: [email protected] (JE); [email protected] (ME)6 2005 24 6 2005 1 1 e216 1 2005 8 4 2005 Copyright: © 2005 Elf and Ehrenberg.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Ribosome-mediated transcriptional attenuation mechanisms are commonly used to control amino acid biosynthetic operons in bacteria. The mRNA leader of such an operon contains an open reading frame with “regulatory” codons, cognate to the amino acid that is synthesized by the enzymes encoded by the operon. When the amino acid is in short supply, translation of the regulatory codons is slow, which allows transcription to continue into the structural genes of the operon. When amino acid supply is in excess, translation of regulatory codons is rapid, which leads to termination of transcription. We use a discrete master equation approach to formulate a probabilistic model for the positioning of the RNA polymerase and the ribosome in the attenuator leader sequence. The model describes how the current rate of amino acid supply compared to the demand in protein synthesis (signal) determines the expression of the amino acid biosynthetic operon (response). The focus of our analysis is on the sensitivity of operon expression to a change in the amino acid supply. We show that attenuation of transcription can be hyper-sensitive for two main reasons. The first is that its response depends on the outcome of a race between two multi-step mechanisms with synchronized starts: transcription of the leader of the operon, and translation of its regulatory codons. The relative change in the probability that transcription is aborted (attenuated) can therefore be much larger than the relative change in the time it takes for the ribosome to read a regulatory codon. The second is that the general usage frequencies of codons of the type used in attenuation control are small. A small percentage decrease in the rate of supply of the controlled amino acid can therefore lead to a much larger percentage decrease in the rate of reading a regulatory codon. We show that high sensitivity further requires a particular choice of regulatory codon among several synonymous codons for the same amino acid. We demonstrate the importance of a high fraction of regulatory codons in the control region. Finally, our integrated model explains how differences in leader sequence design of the trp and his operons of Escherichia coli and Salmonella typhimurium lead to high basal expression and low sensitivity in the former case, and to large dynamic range and high sensitivity in the latter. The model clarifies how mechanistic and systems biological aspects of the attenuation mechanism contribute to its overall sensitivity. It also explains structural differences between the leader sequences of the trp and his operons in terms of their different functions.
Synopsis
When cells grow and divide, they must continually construct new proteins from the 20 amino acid building blocks according to the instructions of the genetic code. Proteins are made by large macromolecular complexes, ribosomes, where information encoded as base triplets (codons) in messenger RNA sequences, transcribed from the DNA sequences of the genes, is translated into amino acid sequences that determine the functions of all proteins. Rapid growth of cells requires that the supply of each free amino acid is balanced to the demand for it in protein synthesis.
The present work mathematically models a common control mechanism in bacteria, which regulates synthesis of amino acids to eliminate deviations from balanced supply and demand. The mechanism “measures” the speed by which the ribosome translates the codons of a regulated amino acid. When supply is less than demand, translation of these “control” codons is slow, which is sensed by the mechanism and used to increase synthesis of the amino acid.
This paper explains why the mechanism is “hyper-sensitive” to relative changes in supply and demand, and why it is differently designed for control of the enzymes that synthesize the amino acids histidine and tryptophan.
Citation:Elf J, Ehrenberg M (2005) What makes ribosome-mediated transcriptional attenuation sensitive to amino acid limitation? PLoS Comput Biol 1(1): e2.
==== Body
Introduction
Ribosome-mediated attenuation of transcription [1] is commonly used for control of expression of amino acid biosynthetic operons in bacteria [2]. There are several other types of attenuation mechanisms [3], but “attenuation” in this paper specifically refers to its ribosome-mediated variant. By this mechanism, the fate of an initiated round of transcription depends on the outcome of a race between the RNA polymerase (RNAP), transcribing the leader of the regulated operon, and a ribosome, translating the leader transcript. The open reading frame in the leader contains two or more regulatory codons cognate to the amino acid that is synthesized by the enzymes encoded by the mRNA of the operon [1]. If the supply of the amino acid is insufficient to meet the demand from protein synthesis, the ribosome will be slowed down on these codons and transcription will continue into the structural genes. If, in contrast, the amino acid supply is in excess, the ribosome will move quickly over the regulatory codons, which results in the formation of an RNA hairpin that signals termination (attenuation) of transcription. Ribosome-mediated transcriptional attenuation was first found in the trp operon of Escherichia coli [1,4] and the his operon of Salmonella enterica serovar Typhimurium (Salmonella typhimurium) [5,6]. Attenuation mechanisms have since been identified for the leu, thr, ilvGMEDA, ilvBN, and pheA operons of E. coli and S.
typhimurium [2], as well as for biosynthetic operons in many other organisms [7].
Attenuation control mechanisms up-regulate operon expression only in response to a reduced speed of translation of regulatory codons, which has led to the idea that these control mechanisms must reduce the rate of growth of bacteria. The reason is that amino acid production will start to increase only when the rate of peptide elongation and, therefore, the current growth rate have already fallen below their maximal values. This is in contrast to repressor systems controlled by amino acid pools, which can regulate gene expression without reduction of the rate of protein elongation [8]. In their book [9], Ingraham, Maaløe, and Neidhardt describe this as a paradox, and suggest that attenuation must be very sensitive, so that the rate of protein synthesis needs to be slowed down only marginally to activate expression of attenuation-controlled operons. In the present study we show that, indeed, attenuation of transcription can be truly hyper-sensitive in accordance with their expectation.
Our starting point is a consensus scheme for attenuation of transcription (Figure 1), and we model mathematically how gene expression responds to amino acid limitation. We take into account the observation that the RNAP and the ribosome start their race over the leader transcript in synchrony [10–12]. This timing, which we show is essential for the sensitivity of attenuation, was not considered in early models of attenuation [13–15] and was first introduced in a comparative study of repressor and attenuation control of amino acid biosynthetic operons [8]. There are two main sources for hyper-sensitivity of attenuation [8]: one related to the multi-step character of the ribosome and the RNAP movements along the operon leader, and the other to the frequency of the amino-acid-starved codons in the mRNAs on all ribosomes of the cell. The former is a property of the mechanism per se, and the second is the property of the mechanism in the context of a growing cell. Here, we will clarify and refine the model by including how the selective charging of tRNA isoacceptors [16] affects the sensitivity of attenuation. We will also extend the model to include mixed codon usage in the attenuation control region as well as modulation of the basal expression level through ribosomal release at the stop codon. These more detailed aspects of the attenuation mechanism turn out to be necessary to explain the striking difference in design of the trp and his leaders in both E. coli and S. typhimurium.
Figure 1 The Leader-Transcript of the trp Operon in E. coli
Attenuated transcription results in a 141-nucleotide (nt) transcript. Aborted transcription of the paused RNAP results in a 91-nt transcript. The transcript includes an open reading frame of 15 codons, encoding a very short-lived 14-residue leader peptide. The RNAP is released from the pause site when the seventh codon is read [17]. Two of the three codons in the control region (10 and 11) are trp codons. Also, ribosome stalling on the arg-codon (12) prevents I:II-hairpin formation and attenuation (see Figure 6). After reaching the stop codon, the ribosome dissociates in about 1 s. The II:III and III:IV conformations have similar stabilities, and the terminator is formed with 50% probability when the ribosome reaches the stop codon and dissociates from the transcript before region IV is available. This determines the basal read-through level of 10–15%.
A scheme for attenuation control of the trp operon, mainly based on the experimental work by Yanofsky and co-workers [2,17], is shown in Figure 1. The leader sequence contains, starting from the 5′ end, the initiation codon (AUG) for mRNA translation followed by region I, in which there are m regulatory codons for the amino acid that is synthesized by the enzymes encoded by the controlled operon. Region I is followed by region II and then by a strong pause site for the RNAP. Further downstream there are n RNA bases, including leader regions III and IV. Mutually exclusive hairpin structures can be formed by regions II:III or III:IV.
When the RNAP has reached the pause site, it stops and remains there until a ribosome starts melting the hairpin structure formed by regions I and II [10–12]. Then the RNAP resumes transcription in synchrony with the movement of the ribosome. If the ribosome is slow in translating regulatory codons in the control region I, it will remain there when the RNAP finishes transcription of region IV (Figure 1). In this case, the II:III, but not the III:IV, hairpin is formed and the RNAP will continue into the open reading frames of the structural genes. When, in contrast, the amino acid synthetic activity of the enzymes encoded by the operon supersedes demand, the ribosome will move quickly over the regulatory codons in region I, which prevents formation of the anti-terminator loop II:III. In this case, the hairpin III:IV will be formed when the RNAP finishes transcribing region IV, which leads to termination of transcription. If the ribosome reaches the stop codon and dissociates from the leader RNA before hairpin III:IV is formed but after hairpin II:III can be formed, then termination may be aborted in spite of rapid translation of regulatory codons. The probability of this event determines the basal expression of the trp attenuator [18].
Under conditions when initiation of leader peptide synthesis is shut down, the RNAP eventually dissociates spontaneously from its pausing state and continues transcription. In this case, the leader transcript preferentially forms the I:II and the III:IV termination structure [17]. This phenomenon, known as super attenuation, is not an integrated part of our model, but one of its consequences will be discussed.
Results
Mathematical Modeling of Attenuation of Transcription
Molecular details from extensive experimental work will here be used to analyze the sensitivity of ribosome-dependent attenuation of transcription in growing cells of E. coli and S. typhimurium.
Let R(t) be the probability that the ribosome at time t is in the control region of the mRNA leader with its m regulatory codons (Figure 1). At time zero, the RNAP resumes transcription from its pausing state by the approach of a ribosome, so that RNAP and ribosome take off in synchrony from well-defined positions on the leader. Let f(t) be the probability density for the time t at which RNAP leaves the nth base, counted from the pause site (Figure 1). The probability Q, that initiation of transcription of the leader of the operon is continued into its structural genes, is the probability R(t), that the ribosome remains in the control region at any time t when the RNAP moves from the nth base with probability density f(t), i.e.
To simplify, we assume that each one of the m codons in the control region is translated with the first-order rate constant k, which depends on the rate of supply of the controlled amino acid compared to ribosomal demand. Then, the movement of the ribosome is defined by a Poisson Process [19], where the probability Po(i,kt) that the ribosome is at codon i at time t is given by (kt)i
e−kt/i!. Accordingly, the probability R(t) is
The first stochastic treatment of ribosome movement during mRNA translation was introduced by von Heijne et al. almost 30 years ago [20]. They discussed possible effects of mRNA secondary structures on ribosomal step times, and suggested that hairpin formations could slow down the movement of the ribosome. Direct measurements of ribosomal step times in E. coli revealed, however, that the rate of codon translation is only marginally affected by stable secondary structures in the open reading frames of messenger RNAs [21]. This suggests rapid melting of the I:II hairpin in the attenuation leader, motivating our assumption of unhindered ribosome movement during translation of this region.
To simplify further, we also assume that each one of the n bases downstream from the pause site is transcribed with the same first-order rate constant q, so that the movement of the RNAP is also a Poisson Process. Then, the probability density f(t) for the time t when the RNAP leaves base n is given by
This model is apparently similar to Manabe's [13], but there is an important difference. Here, we have accounted for the experimentally identified pause site for the RNAP, which synchronizes the movements of RNAP and ribosome [10–12]. The pause site was not known when Manabe's model was formulated, but is included in later models of attenuation [14,15]. These, however, miss the point, that the RNAP is actively released by the ribosome-dependent melting of the hairpin structure that defines the pause site (Figure 1). This synchronization of translation and transcription [10–12] is in fact a strict requirement for a hyper-sensitive attenuation mechanism [8].
The probability for continued transcription into the structural genes by early ribosome termination at the stop codon in the open reading frame of the leader sequence [18] is described below, and its effects will be discussed when we compare the trp and his attenuator mechanisms.
Partitioning of Hyper-Sensitivity into a Mechanistic and a System-Dependent Factor
The signal s for attenuation of transcription of amino acid biosynthetic operons is the rate of amino acid supply normalized to ribosomal demand [22]. The response is the probability Q in Equation 1, that initiation of transcription of the leader leads to expression of the structural genes of the operon. Steady-state sensitivity is defined as the logarithmic gain or, equivalently, the sensitivity amplification a
Qs [23,24], i.e., the relative change in Q normalized to a relative change in s:
The sensitivity a
Qs can be partitioned into the factors aQk and aks. The first is the relative change in Q caused by a relative change in the rate k of translation of a regulatory codon in the open reading frame of the leader. The second is the relative change in k caused by a relative change in the signal s, i.e.
The first factor aQk depends on the attenuation mechanism per se, and the second factor aks is a system parameter, which relates the rate of reading of a starved codon to the rate of supply of its cognate amino acid in a growing bacterial cell.
Hyper-Sensitivity by Competing Multi-Step Processes
To clarify the origin of hyper-sensitivity relating to aQk in Equation 5, we consider first an attenuation-like mechanism, where there is only one regulatory codon (m = 1), translated with rate constant k′, in the open reading frame of the mRNA leader and only one rate-limiting step (n = 1) with rate constant q′ for the RNAP. Then, Q in Equation 1 is given by the hyperbolic relation [19]
and aQk′ by
aQk′ asymptotically approaches its largest absolute value, one, when k′ increases beyond limit and Q goes to zero.
Next, consider the case when the RNAP transcribes many bases in the leader sequence (n >> 1), while the ribosome translates a single, rate-limiting regulatory codon (m = 1). With q = nq′, the probability Q is
Here, aQk asymptotically approaches its largest absolute value, n, when k′ increases indefinitely and Q goes to zero, as in the previous case. In the limit that n → ∞, Q = e−k′/q′ and aQk' = −k′/q′ (see insert in Figure 2). Accordingly, when a single-step process competes with a multi-step process, the sensitivity amplification can be numerically very large, at the cost of a small Q [25].
Figure 2 The Probability Q of Read-Through of the Attenuation Leader Is Plotted as a Function of k′ = 1/τ Where τ Is the Average Time for Translation of All the m Codons
The translation rates of the individual codons are k = k′ · m. The average time to transcribe the n nucleotides is kept constant at 1/q′ = 1s. The transcription rate for an individual nucleotide is q = q′ · n. For m = 1, n = 1 the response function is given by Equation 6; for m = 1, n = 50 the response function is given by Equation 8. These curves should be compared to the much more sensitive response that is reached with two competing multi-step processes, as in attenuation (e.g., m = 10, n = 50). Insert: The sensitivity amplification aQk is plotted as a function of n, while k is kept fixed, and q is changed to get the read-through probabilities (Q) corresponding to the different curves. The two curve families correspond to the ordinary multi-step mechanism with m = 1 (dashed) and the competing multi-step mechanisms with m = n (solid), respectively. The dashed curves approach ln(Q) as n →∞, whereas the solid curves all go to −∞ for all values of Q.
In realistic attenuation mechanisms, however, both m and n are larger than one, as exemplified by the E. coli case, where n = 50–100 and m = 2–16 [2]. When m increases, the relative uncertainty (standard deviation normalized to mean) of the time the ribosome spends in the control region decreases. Also, when the time that the RNAP leaves base n is well defined, the time during which R(t) is averaged in Equation 1 is short. Accordingly, when R makes a sharp transition from zero to one, the absolute value of the sensitivity amplification can be very high, even for Q values close to one (Figure 2). In the interesting case when m,n → ∞, Q is a step function with infinite sensitivity to changes in k for all values of Q at the point of operation of the control mechanism.
Hyper-Sensitivity in Attenuation due to Low Frequency of Regulatory Codons in Bulk mRNA
To understand the origin of hyper-sensitivity relating to aks in Equation 5, one must take into account that the steady-state rate jaa of amino acid supply from the amino acid biosynthetic enzymes must equal the steady-state rate f[R]v of its consumption [8]. In the case when there is one codon per amino acid, f is its frequency of occurrence on all translating ribosomes in the cell, [R] is the ribosome concentration, and v is the average rate of protein elongation per ribosome. The rate v is the inverse of the average time to translate individual codons on translating ribosomes, weighted by their usage frequencies [16]. To simplify, we take 1/k to be the average time to translate a particular regulatory codon and 1/k
max to be the uniform time to translate all other codons [22], which gives the flow balance relation [8]
The signal s is in Equation 9 is jaa divided by the maximally possible rate, f[R]kmax, of consumption of the controlled amino acid when it is supplied in excess. Accordingly, s is zero when amino acid supply is shut down, and s is one at saturating supply rate. The sensitivity parameter aks in Equation 5 follows by differentiating k in Equation 9 with respect to s and multiplying the derivative by s/k:
Because f is small [0.001–0.1], aks contributes with a factor in the interval 10–1,000 to the overall sensitivity amplification aQs in Equation 5 for values of the signal s close to and below one. This remarkable result can be explained as follows. When one amino acid is rate-limiting for protein synthesis, the total rate of peptide elongation in the cell is determined by its rate of supply [8,22]. When s is close to one and the usage frequency f of the regulatory codon is small, a given reduction in total rate of protein synthesis corresponds to the decreased rate of translation of only a small fraction of the currently translated codons. Hence, a relative decrease in an s value close to one leads to a 1/f times larger relative reduction in the rate of reading of the starved codon. When s decreases, an increasing fraction of ribosomes in the cell will be programmed with the starved codon, so that the current f value increases from a small value toward one, at which point the effect vanishes (Figure 3).
Figure 3 The Probability Q of Read-Through of the Leader Is Plotted as a Function of the Signal s
For the curve with m = 10, k is given by Equation 9, and for the curves with m = 1, k is one-tenth of this value. q′ = 1s
−1 for all curves. The usage frequency of the amino acid in proteins is f = 1/20, and the maximal rate of codon translation at full supply of amino acid is kmax= 20s
−1. Insert: The contributions of the two factors to the sensitivity aQs= aQkaks is illustrated for the case with m = 10 and n = 50. aks approaches 1/f as s → 1. The asymptotic behavior of aQk is described in Figure 2.
Selective Charging of tRNA Isoacceptors and the Sensitivity of Attenuation of Transcription
So far, we have neglected that the genetic code is redundant [26], in that there often are several codons cognate to one amino acid that are translated by several isoaccepting tRNAs [27]. There exists a strong bias among the codons in attenuation leaders [2], and it was suggested early that this bias has evolved to maximize the sensitivity of attenuation mechanisms by the choice of rare codons in attenuation leaders, with the leu operon as the paradigmatic example [2]. This explanation must, however, be an oversimplification, since the bias in other attenuation leaders favors major (e.g., as with the thr operon with eight ACC codons) or intermediate codons [16].
The mystery of biased usage of regulatory codons in attenuation leaders in E. coli was recently resolved by the theory of selective charging of isoaccepting tRNAs during amino acid starvation [16]. This theory, fed with experimental data on total tRNA concentrations and codon usage on translating ribosomes [28,29], was used to identify those regulatory codons in attenuation leaders in E. coli for which the rate of translation is most sensitive to variation in the rate of supply of their cognate amino acid. In each case, the experimentally observed codon bias [2] favors the regulatory codon for which the translation rate is most sensitive to amino acid limitation [16].
Since then, selective charging in Thr, Leu, and Arg tRNA isoacceptor families has been verified experimentally [30] and found to be qualitatively in accordance with theory [16]. The theory has also successfully predicted how ribosomal by-passing [31] responds to inhibition of the seryl-tRNA synthetase, when the ribosomal A site has a codon read by either a Ser3 or a Ser5 isoacceptor [32].
The theory of selective charging assumes that the rate by which a charged tRNA isoacceptor is deacylated is in proportion to the frequency by which its cognate codons occur on translating ribosomes, and that the rate of aminoacylation of an isoacceptor is in proportion to the concentration of its deacylated form. From this follows that an isoacceptor with high total concentration and low codon usage will remain charged with amino acid, while an isoacceptor from the same isoacceptor family with low total concentration and high codon usage will lose charging when the supply of their common amino becomes limiting for protein synthesis [16].
To exemplify, consider a simple case with two different isoacceptors (A and B), where A reads only codon a and B reads only codon b. Codon a occurs in the attenuation leader of the mRNA that encodes the enzymes that synthesize the amino acid cognate to A and B. When codons a and b occur on translating ribosomes with the same frequency, the aminoacylated forms of isoacceptors A and B are consumed by the same rate in protein synthesis. In the steady state, this also means that the deacylated forms of A and B must be charged with amino acid at identical rates. From this follows further, that when the total concentrations of A and B in the cell are different, the concentrations of their aminoacylated forms may differ greatly when their cognate amino acid is rate-limiting for protein synthesis. This is illustrated in Figure 4A, which shows the concentrations of the aminoacylated forms of A and B when the total concentration of A is constant and the total concentration of B is varied at an s value (Equation 4) [22] of 0.95 with equal codon usage frequencies of 2.5% for a and b. When the total concentration of B increases from small values, the concentration of charged B goes up, while the concentration of charged A goes down (Figure 4A). Therefore, under amino acid limitation, the charged level of isoacceptor A depends strongly on the total concentration of isoacceptor B, and this dependence is also reflected in the sensitivity of the attenuation mechanism with a as regulatory codons (insert in Figure 4A). When the total concentration of B is small, the sensitivity of attenuation is very small, but increases sharply toward its asymptote with increasing B concentration. At the asymptote, B isoacceptors are fully charged, and b codons are read with maximal rate. In this limit, the sensitivity follows from Equation 1, with the codon frequency f defined by the usage of a codons, i.e., 2.5% in this case (Figure 4B).
Figure 4 The Choice among Synonymous Codon
(A) The predicted concentration of aminoacylated tRNA isoacceptor A (solid) and B (dashed) is plotted as a function of the total concentration of isoacceptor B for a fixed but limiting supply of the cognate amino acid (s = 0.95). The codon usage of a and b are both equal to 2.5%, and the total concentration of tRNA isoacceptor A is kept constant at 3.33 μM. Insert: The sensitivity in rate of reading a codons by the A isoacceptor in response to a change in amino acid supply is plotted over the same interval as the main figure.
(B) The probability of read-through is plotted as a function of normalized amino acid supply rate. The attenuation control codon a is translated by tRNA isoacceptor A. The same amino acid is also encoded by another codon b, which is translated by tRNA isoacceptor B. The codon usage of a and b are both equal to 2.5%. The different curves correspond to different ratios between the total concentrations of isoacceptor A and B, as indicated in the figure. As m = 10, n = 50, and kmax = 20s
−1, the situation where the concentrations are equal corresponds to the solid curve in Figure 3. Insert: The sensitivity aQs for the three cases. More details about the calculations are given in the Supporting Information and reference [16].
In more realistic scenarios, several isoacceptors may read the same codon, which leads to more complex kinetics [16], but the same principles apply.
The Fraction of Regulatory Codons in the Control Region
In realistic attenuation systems, there is normally a mixture of regulatory and other codons in the control region, defined as the mRNA leader region from the position where the RNAP is released up to the last codon where ribosome stalling promotes anti-terminator conformation. The sensitivity of attenuation mechanisms depends on the fraction of regulatory codons in the control region. To see this, we take into account that regulatory and other codons are translated with different rates by using the master equation [33]
r(i,t) is the probability that the ribosome is at codon i at time t. The enumeration is defined such that the RNAP is released when codon i = 0 is translocated into the A site of the ribosome. Codon m1 is the first, and m2 is the last codon where ribosome stalling causes anti-termination and codon M is the stop codon. ki is the rate by which codon i is read, and kdiss is the rate by which the ribosome dissociates from the stop codon (see next section). The probability that the ribosome is in the part of the control region that promotes anti-terminator conformation at time t is
Equation 12 generalizes Equation 2 for R(t) and should be used in Equation 1 when there is a mix of different codons in the control region or when ribosome stalling in only a part of the control region can promote anti-terminator conformation. When all ki = k and m
1 = 0, m
2 = m − 1, then Equation 12 reduces to Equation 2.
If the fraction of regulatory codons in the control region is small, the sensitivity of the mechanism will be comparatively small; e.g., if m' codons out of m are regulatory, the sensitivity will be smaller than for a mechanism with m' regulatory codons out of m' codons in the control region. Figure 5 shows how the sensitivity of the mechanism changes when the fraction of regulatory codons decreases from 10 out of 10, (#10/#10) to #1 out of #10 (#1/#10), while the rate of transcription and the number of RNAP steps are constant. The translation rate of the non-regulatory codons is taken to be 15s
−1. When the translation rates of regulatory and non-regulatory codons are equal in a control region with m codons, the sensitivity is proportional to the fraction of regulatory codons multiplied by the sensitivity that pertains when all m codons are regulatory. This is illustrated in Figure 5 at the right-most point of the x-axis, where regulatory as well as non-regulatory codons have a translation rate of 15s
−1. When the translation rates of regulatory and non-regulatory codons differ or when there are different codons in the control region, comparison of different cases with respect to sensitivity requires scaling of the transcriptional step rate q. The purpose of the scaling is to equalize the Q-values in the different cases for the particular rate (k) of regulatory codon-translation at which the sensitivity is computed. With this normalization, it is seen that that the sensitivity is higher when there are five regulatory and no other codons, than when there are five regulatory and five non-regulatory codons in the control region. As the rate of translation of regulatory codons decreases, the absolute value of the sensitivity amplification in the mixed codon case increases until the sensitivities converge (insert in Figure 5).
Figure 5 The Sensitivity Amplification aQk Is Plotted as a Function of the Rate of Translation of Regulatory Codons
The different curves correspond to different number of regulatory codons in a control region that has in total ten codons. #i/#j indicates i regulatory codons out of j codons in the control region. The rate of translation for non-regulatory codons is 15s
−1. The RNAP transcribes 50 nt s
−1 (q = 50s
−1) and the total number of RNAP steps is 80 (n = 80). Insert
: The sensitivity amplification aQk is plotted as in the main figure, but now the rate of transcription (q) is scaled such that Q = 0.01 for each point.
The Probability of Anti-Terminator Conformation after Ribosome Release from the Stop Codon
If the ribosome reaches the stop codon and dissociates from the leader RNA before hairpin III:IV is formed but after hairpin II:III can be formed, then termination may be aborted in spite of rapid translation of regulatory codons. The probability Q2 of this event is small for the his attenuator but large for the trp attenuator, where it sets the high basal expression level [18]. Let I be the time integral of the probability that the RNAP is in a region between bases n
1 and n
2 at time t when the ribosome is at the stop codon M with probability r(M,t), from which it dissociates with the first-order rate constant kdiss:
Bases n1 and n2 define the region where the II:III hairpin (Figure 6A) can be formed after dissociation of the ribosome from the stop codon. Accordingly, n1 is the last base in region III that can pair with region II and n2 as the base in the middle of region IV. This is because formation of hairpin II:III requires that region III is available and that region IV has not been completely transcribed. It has been assumed that, when these conditions are fulfilled, formation of either hairpin I:II or hairpin II:III in the trp operon leader will occur with 50% probability, since their stabilities are similar [18]. We will use the same assumption for the his attenuator. Accordingly, the probability Q2 is equal to I/2, and the total probability Qtot that the polymerase will continue into the structural genes of the operon is given by Q+Q2, where Q is defined in Equation 1.
Figure 6 Attenuation Control of the his and trp Operons
(A) The trp and his attenuation leader regions. For the trp mechanism, codon 3 and 4 after the ribosome has released the RNAP at “rel” are trp codons [2,17]. Ribosome stalling on codon 3, 4, or 5 leads to anti-terminator conformation [42]. If the ribosome releases from the stop codon (codon 8) after segment III is available for secondary structure formation (RNAP has transcribed nt 28 + 8) but before more than half of segment IV is available (RNAP has transcribed nt 39 + 8), the probability for anti-terminator formation is 50% [18]. The “+ 8” are the bases that have been transcribed but are unavailable for secondary structure formation [43]. For the his mechanism, codon 2 to 8 after the ribosome has released the RNAP are his codons [2,44]. Ribosome stalling on any of these is assumed to lead to anti-terminator conformation. If the ribosome releases from the stop codon (codon 11) after segment III is available for secondary structure formation (RNAP has transcribed nt 7 + 8) but before more than half of segment IV is available (RNAP has transcribed nt 18 + 8), we assume a 50% chance to get anti-terminator formation. The transcription rate is 50 nt s
−1 [45], the translation rate of codons other than trp or his is 15s
−1 [46], and the rate of ribosome release from the stop codon is 1s
−1.
(B,C) The probability of transcription (y-axis) past the leader for trp and his operons as a function of the rate of translation of respective trp and his codons (x-axis). The read-through probability is a sum of the probabilities for two mutually excusive events: ribosome stalling on the control codons while the RNAP escapes attenuation (dotted) or ribosome release from the stop codon before the termination hairpin is completed (dashed). In the trp case, the second event causes a high basal expression level.
(D) The probability of transcription as a function of normalized supply of his and trp, respectively. The codon usage frequency is 1% for trp and 2% for his [28]. kmax = 15s
−1. Insert: The sensitivity aQs over the interval s = 0.8 to 1.0. The positive sign of the sensitivity in gene expression for an increase in histidine supply in the narrow range of s = 0.995–1.0 is due to reduced probability of ribosome release at the stop codon.
The trp and his Attenuators
The attenuator mechanisms of the trp operon of E. coli and the his operon of S.
typhimurium have been extensively studied (see [2,34] and references therein).
The trp operon is under dual transcriptional control by a Trp-sensitive repressor [35] and an attenuator [36]. During balanced growth, the operon is repressor controlled with a dynamic range of about 70 [37]. During severe Trp starvation, attenuation is turned off, resulting in an additional 10-fold increase in operon expression [37,38]. The dual control of the trp operon requires a high basal read-through of the attenuator at excess supply of Trp. This is achieved at a level of 10%–15%, due mainly to termination of translation and ribosome release from the stop codon, before region IV has become available to form the terminator structure with region III ([18]; Figure 6). Our model predicts a basal level expression of about 8% (Figure 6), when the trp attenuator data (legend in Figure 6) are inserted in Equations 1, 12, and 13. It has, however, been suggested that the basal level also contains a contribution from an intrinsic read-through, not included in our model, of about 3%, as estimated under super attenuation conditions [18]. Taking this additional read-through into account makes our model prediction (8 + 3 = 11%) an even better estimate of the experimental estimate (10%–15%).
The his operon in S. typhimurium
, in contrast, is only attenuation-controlled and the basal level expression at excess supply of His is supposed to be smaller than the basal expression level of the trp attenuator [34]. Our model immediately suggests two different strategies for implementation of low basal read-through in attenuation mechanisms. Their common feature is that the decision to form the transcriptional terminator hairpin occurs well before termination of translation and ribosome release from region II (see Figure 1). This can be achieved either by placing the stop codon far downstream in region II, so that termination of translation is delayed in relation to formation of the transcriptional terminator III:IV; or the same effect can be accomplished by rapid formation of a secondary structure that prevents anti-terminator conformation unless the ribosome is in the control region. The latter option seems to be the design principle of the his operon. Here, an extra hairpin has evolved between the RNAP pause hairpin and the transcriptional terminator (Figure 6A). This extra hairpin forms rapidly when the RNAP resumes transcription after pausing, and the model suggests that this is to prevent read-through of the his attenuator by translation termination and ribosome release. When, in contrast, the extra hairpin structure is removed from the model, the basal level read-through at saturating His supply increases dramatically (Figure 6C). From this, we suggest that the extra hairpin serves to minimize basal read-through of the his attenuator to reduce the metabolic cost of His synthesis when His is supplied externally. The trp operon expression, in contrast, is turned off by the Trp repressor when there is Trp in the medium, and the proper action of the dual control system of the trp operon requires a high basal read-through of the attenuator.
Our models account for experimental observations regarding the trp (Figure 6B) and the his operon (Figure 6C). It is, for instance, clearly seen that the trp attenuator (Figure 6B) requires a much more severe amino acid limitation for full de-repression and has a much a higher basal read-through level than the his attenuator (Figure 6C). We have also estimated the sensitivity (a
Qs; see Equation 4) of gene expression to variation in the amino acid supply signal s for each one of these attenuators (Figure 6D). The his attenuator is much more sensitive due to the higher number and higher fraction of his codons in that control region.
Discussion
We have described two major sources of hyper-sensitivity of ribosome-dependent attenuation of transcription.
The first relates to the odds for selecting the winner of two competing multi-step (Poisson) processes with synchronized starts. One competitor is the RNAP, transcribing the leader of the operon, and the other is the ribosome, translating regulatory codons in that leader. Synchronous starts for transcription and translation in the attenuator are essential for hyper-sensitivity, and synchrony is achieved by the ribosome-dependent melting of the hairpin structure that makes the RNAP pause in the attenuator (see Figure 1). When the number of regulatory codons to be translated as well as the number of bases to be transcribed in the attenuator are large, hyper-sensitivity in the rate of gene expression to variation in the rate of amino acid synthesis emerges and can be combined with a high probability that initiation of transcription of the leader continues into the structural genes of the operon (see Figure 2). The importance of a large number of regulatory codons for sensitivity has been verified experimentally [39]. When, in contrast, there is competition between two single-step processes, the control is hyperbolic and lacks hyper-sensitivity [19]. Although competition between a single- and a multi-step process (exponential control) can lead to hyper-sensitivity, it is only at the cost of a very low probability for an initiation event to result in gene expression (see Figure 2; [19]).
The second source of hyper-sensitivity relates to the frequency (f) of occurrence of regulatory codons in the mRNAs of all translating ribosomes of the cell, just at the onset of amino acid starvation. A fractional decrease of amino acid synthesis is amplified by a factor of 1/f in the resulting fractional increase in the time to translate regulatory codons (see Figure 3). This effect is a direct consequence of the stoichiometric coupling that exists between the rates of deacylation in protein synthesis of all the cell's tRNAs [16,22]. When amino acid limitation becomes increasingly severe, an ever-larger fraction of all ribosomes will expose the starved regulatory codons in their A site, until f becomes one and this sensitivity amplification effect vanishes.
When the regulatory codons in an attenuator are read by a tRNA that belongs to a family of isoaccepting tRNAs, cognate to the same amino acid, then hyper-sensitivity requires that the concentration of the aminoacylated form of the regulatory codon reader is sensitive to amino acid deprivation [16]. In short, the condition is that the concentrations of other members of the isoacceptor family are large in relation to the usage of their codons on all translating ribosomes, as compared to the concentration of the regulatory codon-reading tRNA in relation to the general usage of those codons in translation. When this condition is met, the charged level of the regulatory codon reader is maximally sensitive to amino acid limitation, while the charged levels of the other isoacceptors are insensitive and remain high ([16]; see Figure 4). In this case, the frequency f (Equation 4) refers to how often the regulatory and not all synonymous codons occur on translating ribosomes, which enhances the sensitivity at the onset of starvation. This is in line with results from experiments by Carter et al. [40], in which the three CUA codons, normally used in the S.
typhimurium
leu attenuator, were replaced by CUG codons. This led to reduced sensitivity, with total de-repression of the operon occurring only at severe Leu starvation. Our theory suggests that, for E. coli, CUA is the most suitable leu attenuator control codon, and that CUG is a less optimal but still a reasonable alternative [16]. On the assumption that the charged levels of Leu isoacceptors in E. coli and S.
typhimurium react similarly to Leu limitation, our theory also predicts that if the CUA codons had been replaced by UUG or UUA codons, then the attenuator would have become even less sensitive to Leu limitation and the operon would not have been de-repressed even under severe Leu starvation.
Our modeling of the his and trp attenuators from S.
typhimurium and E. coli, respectively, has reproduced available experimental data on these control circuits with respect to, in particular, regulatory range and basal expression levels. Furthermore, our simulations suggest that attenuation is close to fully relieved when the rate of Trp supply is about 70% of ribosomal demand, i.e., when s = 0.7 (Figure 6D). An s value of 0.7 roughly corresponds to a 30% reduction in growth rate due to Trp starvation [8,16,22], and our result can therefore be compared to experimental data from Yanofsky and Horn [38]. These show a 4-fold, out of a maximally 6-fold, relief of attenuation of the trp operon when the growth rate is reduced by 20% (s = 0.8) due to Trp limitation. These experimental data [38] are, therefore, in good agreement with our predictions (Figure 6D). For s values like these that are significantly less than one, the Trp as well as the Trp-tRNATrp levels are expected to be very low compared to their values at adequate supply of Trp [8,16,22].
Our analysis has also led to a novel suggestion regarding the role of the extra hair pin structure that exists in the his-attenuator of S.
typhimurium. Our model predicts that removal of this hair pin results in a much higher basal level expression from the operon, which suggests that the hair pin serves to lower the probability of anti-terminator formation caused by translation termination and ribosome release from the attenuator. Accordingly, the presence of the extra hair pin leads to an increase of the regulatory range of the his-attenuator and reduces the metabolic cost associated with redundant His-synthesis when there is His in the medium.
Our analysis has, finally, resolved the “Ingraham-Maaløe-Neidhardt paradox” which states that the controlling ability of attenuation mechanisms necessarily leads to reduced growth rate [9], by showing that attenuators can indeed be hyper-sensitive regulators of amino acid synthesis. We also suggest that further improvement of attenuator performance comes from burst-like expression from attenuator-controlled amino acid biosynthetic operons, so that total protein elongation in the cell is marginally limited by amino acid supply only a small fraction of the time. This hypothesis is now addressed experimentally (J. Elf, in preparation).
Supporting Information
Selective Charging
To estimate the effect of selective charging, we must explicitly introduce concentrations of aminoacylated tRNA isoacceptors. From this, the rate of reading a specific codon is given by
where k
cat and K
m are the Michaelis-Menten parameters for peptide elongation. The concentration of cognate aminoacylated tRNA that can read the codon is [aa-tRNAj]. The average time it takes to read the codon is τj = 1/kj. With these definitions, the average rate of protein synthesis, v, is given by [16,41]:
where f
j is the codon usage frequency of codon j.
When one amino acid is rate-limiting for protein synthesis, the tRNAs for the other amino acids are fully charged [22]. Assume, as in the main text, that the limiting amino acid has two cognate codons, ca and cb. Codon ca can be read by tRNA isoacceptor A with concentration [A], and cb can be read by isoacceptor B with concentration [B], respectively. The charged fractions of A and B are α and β, respectively. In this case, Equation 15 becomes
where Km = 1μM and a total concentration of tRNA that can read an individual codon [tRNAtot] = 3.33 μM. Further, we chose k
cat= 26 s
−1 in order to make kmax= 20s
−1 at full charging, which facilitates comparison with Figure 3.
The rates of aminoacylation of the two isoacceptors are assumed to be proportional to the concentrations of their deacylated forms, i.e., to (1 − α)[A] and (1 − β)[B]. If this were not the case, the computations would have to be modified but the same principles would pertain [16]. At the steady state, the rate of aminoacylation equals the rate of consumption of aminoacylation tRNA in protein synthesis, which is the frequency of the codons the isoacceptors reads, fa or fb, respectively, multiplied by the concentration of elongating ribosomes, [R], multiplied by the average rate of peptide elongation, v. This gives the flow balance relations
Equation 17 defines the charged fractions α and β for each amino acid supply level s = v/v
max
, as given by variation in jaa. The rate of translation of the attenuation control codon is given by kj = kcat/(1 + Km/(α[A])), where [A] = 3.33 μM. The curves in Figure 4 are given for [B] = 3.33 μM, [B] = 1.67 μM, and [B] = 6.67 μM, respectively, while fa = fb = 0.025 for all curves.
This work was supported by the Swedish Research Council.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. JE and ME conceived and designed the model. JE performed the numerical computations . JE and ME wrote the paper.
Abbreviations
ntnucleotide
RNAPRNA polymerase
==== Refs
References
Yanofsky C 1981 Attenuation in the control of expression of bacterial operons Nature 289 751 758 7007895
Landick R Yanofsky C 1987 Transcription attenuation Ingraham JL Neidhardt FC Escherichia coli and Salmonella typhimurium: Cellular and molecular biology Washington, D.C. ASM Press 1276 1301
Yanofsky C 2000 Transcription attenuation: Once viewed as a novel regulatory strategy J Bacteriol 182 1 8 10613855
Jackson EN Yanofsky C 1973 The region between the operator and first structural gene of the tryptophan operon of Escherichia coli may have a regulatory function J Mol Biol 76 89 101 4578102
Lewis JA Ames BN 1972 Histidine regulation in Salmonella typhimurium . XI. The percentage of transfer RNA His charged in vivo and its relation to the repression of the histidine operon J Mol Biol 66 131 142 4339187
Kasai T 1974 Regulation of the expression of the histidine operon in Salmonella typhimurium
Nature 249 523 527 4599761
Vitreschak AG Lyubetskaya EV Shirshin MA Gelfand MS Lyubetsky VA 2004 Attenuation regulation of amino acid biosynthetic operons in proteobacteria: Comparative genomics analysis FEMS Microbiol Lett 234 357 370 15135544
Elf J Berg OG Ehrenberg M 2001 Comparison of repressor and transcriptional attenuator systems for control of amino acid biosynthetic operons J Mol Biol 313 941 954 11700051
Ingraham JL Maaloe O Neidhardt FC 1983 Growth of the bacterial cell Sunderland (Massachusetts) Sinauer Associates 435 p.
Landick R Carey J Yanofsky C 1985 Translation activates the paused transcription complex and restores transcription of the trp operon leader region Proc Natl Acad Sci U S A 82 4663 4667 2991886
Winkler ME Yanofsky C 1981 Pausing of RNA polymerase during in vitro transcription of the tryptophan operon leader region Biochemistry 20 3738 3744 6168281
Landick R Carey J Yanofsky C 1987 Detection of transcription-pausing in vivo in the trp operon leader region Proc Natl Acad Sci U S A 84 1507 1511 2436219
Manabe T 1981 Theory of regulation by the attenuation mechanism: Stochastic model for the attenuation of the Escherichia coli tryptophan operon J Theor Biol 91 527 544 6173543
von Heijne G 1982 A theoretical study of the attenuation control mechanism J Theor Biol 97 227 238 6752587
Suzuki H Kunisawa T Otsuka J 1986 Theoretical evaluation of transcriptional pausing effect on the attenuation in trp leader sequence Biophys J 49 425 435 3513858
Elf J Nilsson D Tenson T Ehrenberg M 2003 Selective charging of tRNA isoacceptors explains patterns of codon usage Science 300 1718 1722 12805541
Landick R Turnbough JCL 1992 Transcriptional attenuation McKnight SL Yamamoto KR Transcriptional regulation Woodbury (New York) Cold Spring Harbor Laboratory Press 407 447
Roesser JR Nakamura Y Yanofsky C 1989 Regulation of basal level expression of the tryptophan operon of Escherichia coli
J Biol Chem 264 12284 12288 2663855
Paulsson J Ehrenberg M 2001 Noise in a minimal regulatory network: Plasmid copy number control Q Rev Biophys 34 1 59 11388089
von Heijne G Nilsson L Blomberg C 1977 Translation and messenger RNA secondary structure J Theor Biol 68 321 329 599938
Sorensen M Kurland C Pedersen S 1989 Codon usage determines translation rate in Escherichia coli
J Mol Biol 20 365 377
Elf J Ehrenberg M 2005 Near-critical behavior of aminoacyl-tRNA pools in E. coli at rate limiting supply of amino acids Biophys J 88 132 146 15501947
Goldbeter A Koshland DE Jr 1982 Sensitivity amplification in biochemical systems Q Rev Biophys 15 555 591 6294720
Savageau MA 1976 Biochemical systems analysis: A study of function and design in molecular biology Reading (Massachusetts) Addison-Wesley 379 p.
Ehrenberg M 1996 Hypothesis: Hypersensitive plasmid copy number control for ColE1 Biophys J 70 135 145 8770193
Crick FH 1966 Codon–anticodon pairing: The wobble hypothesis J Mol Biol 19 548 555 5969078
Björk G 1996 Stable RNA modification Neidhardt FC Escherichia coli and Salmonella cellular and molecular biology Washington, D.C. ASM Press 861 886
Dong H Nilsson L Kurland CG 1996 Co-variation of tRNA abundance and codon usage in Escherichia coli at different growth rates J Mol Biol 260 649 663 8709146
Ikemura T 1981 Correlation between the abundance of Escherichia coli transfer RNAs and the occurrence of the respective codons in its protein genes J Mol Biol 146 1 21 6167728
Dittmar K Sørensen M Elf J Ehrenberg M Pan T 2005 Selective charging of tRNA isoacceptors induced by amino acid starvation EMBO Rep 6 151 157 15678157
Gallant J Bonthuis P Lindsley D Cabellon J Gill G 2004 On the role of the starved codon and the takeoff site in ribosome bypassing in Escherichia coli
J Mol Biol 342 713 724 15342232
Lindsley D Bonthuis P Gallant J Tofoleanu T Elf J 2005 Ribosome by-passing at serine codons as a test of the model of selective tRNA charging EMBO Rep 6 147 150 15678161
van Kampen NG 1992 Stochastic processes in physics and chemistry, 2nd ed Amsterdam Elsevier p. 134 165
Winkler ME 1996 Biosynthesis of histidine Neidhardt FC Escherichia coli and Salmonella cellular and molecular biology Washington, D.C. AMS Press 485 505
Yanofsky C Crawford IP 1987 The tryptophan operon Neidhardt FC Escherichia coli and Salmonella typhimurium: Cellular and molecular biology Washington, D.C. ASM Press 1453 1472
Oxender DL Zurawski G Yanofsky C 1979 Attenuation in the Escherichia coli tryptophan operon: Role of RNA secondary structure involving the tryptophan codon region Proc Natl Acad Sci U S A 76 5524 5528 118451
Yanofsky C Kelley RL Horn V 1984 Repression is relieved before attenuation in the trp operon of Escherichia coli as tryptophan starvation becomes increasingly severe J Bacteriol 158 1018 1024 6233264
Yanofsky C Horn V 1994 Role of regulatory features of the trp operon of Escherichia coli in mediating a response to a nutritional shift J Bacteriol 176 6245 6254 7928995
Bartkus JM Tyler B Calvo JM 1991 Transcription attenuation-mediated control of leu operon expression: Influence of the number of Leu control codons J Bacteriol 173 1634 1641 1999384
Carter PW Bartkus JM Calvo JM 1986 Transcription attenuation in Salmonella typhimurium : The significance of rare leucine codons in the leu leader Proc Natl Acad Sci U S A 83 8127 8131 3534884
Ehrenberg M Kurland CG 1984 Costs of accuracy determined by a maximal growth rate constraint Q Rev Biophys 17 45 82 6484121
Zurawski G Elseviers D Stauffer GV Yanofsky C 1978 Translational control of transcription termination at the attenuator of the Escherichia coli tryptophan operon Proc Natl Acad Sci U S A 75 5988 5992 366606
Artsimovitch I Landick R 1998 Interaction of a nascent RNA structure with RNA polymerase is required for hairpin-dependent transcriptional pausing but not for transcript release Genes Dev 12 3110 3122 9765211
Johnston HM Roth JR 1981 DNA sequence changes of mutations altering attenuation control of the histidine operon of Salmonella typhimurium
J Mol Biol 145 735 756 6167727
Gausing K 1972 Efficiency of protein and messenger RNA synthesis in bacteriophage T4-infected cells of Escherichia coli
J Mol Biol 71 529 545 4567464
Bremer H Dennis P 1987 Modulation of chemical composition and other parameters of the cell by growth rate Neidhardt FC Escherichia coli and Salmonella typhimurium: Cellular and molecular biology Washington, D.C. ASM Press 1527 1541
|
16103903
|
PMC1183508
|
CC BY
|
2021-01-05 09:20:35
|
no
|
PLoS Comput Biol. 2005 Jun 24; 1(1):e2
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010002
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science San Francisco, USA 1610390410.1371/journal.pcbi.001000305-PLCB-RA-0020R1plcb-01-01-03Research ArticleBiochemistryBioinformatics - Computational BiologyEvolutionGenetics/Comparative GenomicsGenetics/Functional GenomicsGenetics/EvolutionMolecular Biology - Structural BiologyEukaryotesYeast and FungiSaccharomycesAnimalsPredicting Functional Gene Links from Phylogenetic-Statistical Analyses of Whole Genomes Predicting Functional Gene LinksBarker Daniel Pagel Mark *School of Animal and Microbial Sciences, University of Reading, United KingdomMurray Diana EditorWeill Medical College of Cornell University, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 24 6 2005 1 1 e32 2 2005 13 4 2005 Copyright: © 2005 Barker and Pagel.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.An important element of the developing field of proteomics is to understand protein-protein interactions and other functional links amongst genes. Across-species correlation methods for detecting functional links work on the premise that functionally linked proteins will tend to show a common pattern of presence and absence across a range of genomes. We describe a maximum likelihood statistical model for predicting functional gene linkages. The method detects independent instances of the correlated gain or loss of pairs of proteins on phylogenetic trees, reducing the high rates of false positives observed in conventional across-species methods that do not explicitly incorporate a phylogeny. We show, in a dataset of 10,551 protein pairs, that the phylogenetic method improves by up to 35% on across-species analyses at identifying known functionally linked proteins. The method shows that protein pairs with at least two to three correlated events of gain or loss are almost certainly functionally linked. Contingent evolution, in which one gene's presence or absence depends upon the presence of another, can also be detected phylogenetically, and may identify genes whose functional significance depends upon its interaction with other genes. Incorporating phylogenetic information improves the prediction of functional linkages. The improvement derives from having a lower rate of false positives and from detecting trends that across-species analyses miss. Phylogenetic methods can easily be incorporated into the screening of large-scale bioinformatics datasets to identify sets of protein links and to characterise gene networks.
Synopsis
A typical fully sequenced genome from a bacterial species contains several thousand genes, and those from multicellular animals may contain many thousands of genes. Understanding the function of these genes is one of the key goals of the developing fields of bioinformatics and proteomics, and the results are of interest to life scientists. The authors describe a computational statistical method that can identify pairs of genes whose functions may be linked, in the sense of participating in a common metabolic pathway or from some physical interaction. The method is applied to phylogenetic trees of related organisms and identifies instances in which a pair of genes is either gained or lost together during evolution. They find that genes that have co-evolved like this on two or more occasions during their evolutionary history are almost certainly functionally linked. These methods can be applied in an automated way to large numbers of species for which fully annotated genomes are available to identify candidate sets of functionally linked genes, and to characterize gene networks.
Citation:Barker D, Pagel M (2005) Predicting functional gene links from phylogenetic-statistical analyses of whole genomes. PLoS Comput Biol 1(1): e3.
==== Body
Introduction
Evidence that two or more traits co-evolve across a range of species can be used to test hypotheses about the common selective pressures acting on the traits, and about the functional or adaptive relationship between them. Correlated evolution is increasingly being applied at the genetic level on the premise that genes that are gained and lost together [1–3], or that show similar expression patterns or rates of evolution [4,5], may form a functional linkage. This provides a computational approach that can screen large genomic datasets for functional links [6] and help to identify the functions of uncharacterised genes. Such analyses can also be used to describe metabolic networks [7], and discover gene “modules” or clusters of genes engaged in a common function [8].
Genes and their expression patterns evolve in a phylogenetic context such that functional links of adaptive value tend to be conserved and inherited by descendant species. Among closely related species, shared phylogenetic inheritance can also produce correlated gene profiles for genes that are not linked. Two or more genes might arise independently in a common ancestor and be retained in evolutionary descendants owing to their individual adaptive functions. Figure 1 (numerals in red) shows how this can produce spurious evidence of a functional link when measured across species. By comparison, multiple independent phylogenetic events of the gain/loss of pairs of genes make a compelling statistical case for a functional link (Figure 1, numerals in blue). Phylogenetic methods have uses beyond merely accounting for shared inheritance: they make it possible to investigate ancestral states and to identify the probable temporal ordering of changes in two traits. Knowledge of which of two traits changed first in the evolutionary history the phylogeny describes can be used to test ideas about cause and effect or the dependency of one trait on another [9,10].
Figure 1 Across-Species Correlation Confuses Shared Inheritance with Correlated Evolution but Phylogenetic Method Does Not
The figure shows a hypothetical phylogeny of eight species. Assume all four genes were present in the common ancestor. Only the top (blue) pair provides statistical evidence for correlated evolution. The apparent correlation in the bottom (red) pair arises from shared inheritance of the loss (state “0”) of both genes in the ancestor to the four species on the right of the diagram. Although the two genes were lost at the same time, it may have been for unrelated reasons. By comparison, the correlation in the top pair rests upon four separate events of the correlated loss of both genes. Both genes are retained until near the tips of the tree, at which point both are lost in each of four separate species. It is unlikely that two genes would be simultaneously lost on four independent occasions, unless the two genes were functionally linked. A simple across-species correlation does not discriminate between these two scenarios, whereas one that accounts for phylogeny does. This is an extreme scenario but many others are possible.
Our interest is to evaluate whether incorporating phylogenetic information improves the identification of functional gene links. The need to take account of phylogenetic relationships in comparative studies has long been appreciated in evolutionary biology [11,12] but has received less attention in bioinformatics studies [1,2]. We apply the phylogenetic-statistical method Discrete [9], for assessing correlated evolution among pairs of discrete traits, to data on the presence and absence of pairs of genes. The method identifies independent events of correlated evolution on a phylogeny by comparing the statistical likelihood of the observed data under two alternative scenarios, one in which the two genes are allowed to evolve on the phylogeny independently, and another in which they co-evolve. Trait evolution is modelled as a continuous-time Markov process, and evidence for the model of correlated evolution is assessed by means of the likelihood ratio statistic.
Our dataset consists of a phylogeny of 15 eukaryote species for which complete or nearly complete sequenced genomes are available. There is no limit to the number of species that can be used, but it is important to use fully sequenced and well-annotated genomes to ensure that genes determined to be “absent” are in fact not in the genome. We compare the phylogenetic method's predictions to predictions derived from across-species correlations; the latter have been used in bioinformatics investigations to predict functional gene links [1,2]. We use the Munich Information Center for Protein Sequences (MIPS) [13] database of annotated complexes of yeast proteins as a “known” criterion measure. The MIPS functional links have been determined by low-throughput laboratory procedures and therefore provide a reliable collection of functional links in this species. We find that incorporating phylogenetic information improves predictions by up to 35% over across-species correlations in detecting functional links, and increasingly so for pairs of genes with greater phylogenetic evidence of a functional link. The number of times a pair of genes has been independently gained or lost on the phylogeny is a strong predictor of functional linkage, such that protein pairs with at least two to three correlated events are almost certainly functionally linked.
Results
Phylogenetic Tree
Figure 2 shows the maximum likelihood phylogenetic tree of the 15 species. We also conducted Bayesian Markov chain Monte Carlo [14] phylogenetic analyses [15–21] using the program BayesPhylogenies [21]. The posterior support as derived from the Markov chain Monte Carlo (MCMC) analysis was 100% at all nodes, and our topology for the yeast species agrees with a recent yeast phylogeny [22]. This is not to say that either our tree or that in [22] is the “true” tree (compare to [23]), just that given our data and model of evolution, no other tree was sufficiently likely to be included in the posterior Bayesian sample.
Figure 2 Phylogeny of the 15 Species Showing Two Pairs of Presence/Absence Data for Proteins in MIPS
All nodes of the tree received 100% posterior support in an MCMC analysis (see Results). The protein pairs {CIN4, ORC3} and {L9A, L42B}, marked “1” for presence and “0” for absence, are included to illustrate probable type I (false positive) and type II (false negative) errors by the across-species method in real data (see Results, “False positives”). Probable false positive: The across-species correlation returns a significant (p = 0.0014) correlation between the pair {CIN4, ORC3}. The phylogenetic method regards this as a chance association (p = 0.13) arising from a single event of both genes being gained in the ascomycete yeasts, followed by shared inheritance (as in Figure 1). The pair {L9A, L42B} consists of two functionally linked proteins. These return a significant phylogenetic correlation (p = 0.035) owing to perhaps five correlated losses of both genes (see text). The across-species association is sensitive only to the distribution of the two proteins across the tips, and returns a non-significant result (p = 0.23). This is a probable false negative.
Distribution of Likelihood Ratios
We calculated the likelihood ratio statistic (see Materials and Methods) to test for correlated evolution in 10,551 pairs. We excluded pairs that yielded any evidence of a negative relationship (n = 2,449) on grounds that one gene being present when the other is absent cannot be evidence for a functional link. Figure 3 shows the distribution of the remaining 8,102 likelihood ratios. Larger values of the likelihood ratio (LR) statistic provide stronger evidence for correlated evolution. The blue bar identifies 2,483 likelihood ratios corresponding to pairs of genes in which one or both genes are present in all 15 species and therefore cannot be studied for evidence of correlated evolution. The red bars exclude these pairs, leaving 5,619 LRs for which both genes vary across species. This skewed distribution has a mean of 3.36 ± 2.47, with two values greater than 15.5.
Figure 3 Distribution of 8,102 LRs for MIPS Pairs, Measuring the Strength of Support for the Phylogenetic Correlation
Critical p-value cut-off points are derived from the random pairs data (see Results). The blue bar within the first class represents the 2,483 pairs for which one or both proteins were present in all 15 species (LR ≈ 0). The red bars record the remaining 5,619 LRs for pairs of proteins that both vary across species. Approximately 8% of the results exceed the p ≤ 0.05 level. Two pairs have LRs greater than 15.5 but are not visible on the graph. The excess of LR scores of 4–4.5 and 5.5–6 may arise from misidentified homology in S. kluyveri. This species is identified as having a smaller number of genes than its phylogenetic neighbours. These paired absences will tend to inflate correlations. We left these results in our analyses, as they affect the phylogenetic and across-species analyses equally, and we cannot be sure which absences are real and which are not.
To assign p-value cut-off points to the distribution of varying protein pairs, we simulated the null LR distribution from 9,509 random protein pairs from the MIPS database. These pairs were drawn to have the same across-species distribution as the MIPS pairs, but with the restriction that none of the random pairs came from the same protein complex. The 9,509 pairs produced 6,393 likelihood ratios for pairs of genes that both vary across species, of which 3,722 represent a non-negative relationship. This set of likelihood ratios is well characterized by a gamma probability density distributed as GAM(1.9,1.4). The p-value cut-off points shown on Figure 3 are derived from this null-distribution. Although our expectation is that the random pairs do not represent true links, they are likely to be a conservative control on the MIPS pairs, as some may describe as-yet undiscovered functional links.
Of the MIPS pairs whose patterns both vary across species, 609 (11%) have LRs that exceed the p ≤ 0.05 level. Among these, 185 (3.3% of the total) exceed the p ≤ 0.01 level. There were n = 278 pairs for which both genes are found in all 15 species. If these are assumed to represent functional links, then the total number of across-species functional links remains around 11% = (609 + 278)/8,102; we use 8,102 in the denominator, because we are now including the constant pairs in the calculation. Varying pairs contribute roughly twice as much to this total as do the constant pairs. Even in this hand-picked dataset of known interactions in Saccharomyces cerevisiae, only a comparatively small number generalise across species.
Detection of “Known” Protein Interactions
We wish to know whether the LR method gets better at detecting “known” interactions the more extreme its statistical result. We combined the 8,102 LRs corresponding to all non-negative MIPS relationships, with the 6,838 LRs obtained from the randomly generated pairs in which the relationship is also non-negative. We then assigned the combined data to p-value bins and determined the percentage of the results in each bin that correspond to MIPS pairs. If large values of the likelihood ratio are indicative of functional links, then this percentage should increase as the p-value declines. If, however, random pairs are just as likely to show large p-values, the percentage will not improve. To measure the influence of adopting a phylogenetic perspective, we compared the phylogenetic LR method's performance at identifying true MIPS pairs with the across-species correlation (Fisher exact test, but again excluding pairs with a negative correlation).
Figure 4 compares the two methods' performance, plotting the percentage of the predicted links at or below a given p-value that correspond to annotated functionally linked pairs in the MIPS database. At p ≤ 1.0, both methods declare every pair significant producing a correct rate of 8,102/(8,102 + 6,838) = 54%. As the p-value decreases, the percentage of results that are MIPS pairs increases for both methods. The phylogenetic method correctly classifies a higher percentage of the pairs than the across-species correlation at each p-value level, and eventually includes only MIPS pairs in its predictions. By comparison, the across-species correlation method reaches a plateau beneath 100% correct.
Figure 4 Phylogenetic Method Identifies a Higher Percentage of Functional Links than the Across-Species Correlation
The main graph shows the percentage of the predicted links at or below a given p-value, that correspond to annotated functionally linked pairs in the MIPS database, separately for the two methods. At a p-value of 1.0 or less, both methods declare all of the pairs to be functionally linked, producing a correct percentage of 54% (see Results). Inset: the percentage by which the phylogenetic method improves upon the across-species correlation, where improvement = (percent correct phylogenetic − percent correct across-species)/54.
The inset graph in Figure 4 plots the phylogenetic method's relative improvement over the across-species correlation. The phylogenetic LR shows a pronounced rise at the p ≤ 0.05 level, to 18% improvement, increasing to 35% improvement at more extreme p-values. This is the direct contribution of taking into account shared phylogenetic inheritance. At a p-value of about 0.0006 or less, all of the pairs that the phylogenetic LR method identifies represent known functional links. LRs significant at this level roughly correspond to at least two to three phylogenetically independent pairs of gain/loss events on our phylogenetic tree. This suggests that a consistent pattern of phylogenetic co-evolution almost certainly points to a functional link, and increasingly so as the number of co-evolutionary events increases. The across-species test has no way to discriminate multiple independent events from a single event that is retained and inherited by many species (see Figure 1). This causes it to misclassify many pairs and to fail to improve at identifying functional links even at more extreme p-values.
False Positives
If false positives cause the across-species correlations to classify a lower percentage of the true pairs correctly, this should be apparent from comparing the two methods in the random-pairs data. Figure 5A plots the across-species p-value against the phylogenetic LR p-value for pairs of proteins in the random-pairs dataset. The correlation between the two methods' p-values is r = 0.85 for all pairs. Removing the 3,116 pairs in which at least one gene is found in all 15 species and thus both methods return a p-value of 1.0, the correlation falls to r = 0.73. This means that the methods have only 53% of their variance in common and shows that they respond to different aspects of the data. The diagonal 1:1 line indicates that the across-species p-values tend to fall at or below (more extreme than) the LR p-value; it declares more of the pairs to be functionally linked. The horizontal line through the p = 0.05 level identifies 170 protein pairs in this random pairs data that the across-species correlations declare significant, but the phylogenetic LR method finds not significant. Many of the LRs in this region have large p-values, indicating that there is no evidence of repeated independent events of correlated change. By comparison, the vertical line through p = 0.05 shows that in only 32 cases does the LR method declare a result significant that does not show a significant across-species pattern. Taken together, these results illustrate how the across-species correlation is prone to picking false positives.
Figure 5 Phylogenetic Method Results in Fewer False-Positives than the Across-Species Correlation
The across-species p-value (y-axis) is plotted against the phylogenetic method's p-value (x-axis) for the range of p = 0–0.25; the methods draw similar conclusions for p-values greater than 0.25.
(A) Higher rates of probable false positives for the across-species correlation. The horizontal dashed line defines the region in which the across-species method declares pairs significant (n = 170) but the phylogenetic method finds no evidence for a functional link. The vertical dashed line defines the same region for the phylogenetic method (n = 32).
(B) Same relationship as in (A) but for the MIPS pairs of annotated links. The across-species correlation returns a functional link for n = 278 pairs that the phylogenetic method declares non-significant. Many of these may be false positives arising from chance events (see Results). The phylogenetic method finds n = 186 extra pairs significant. Especially at lower p-values, these are unlikely to be false positives (see Results) (Figure 4).
Figure 5B plots the same comparison, but this time for the MIPS pairs. The across-species correlation again tends to have lower p-values but the trend is less pronounced. The correlation between the two methods is r = 0.86 or 74% shared variance for all patterns, and r = 0.74 or 55% shared variance for pairs in which both genes vary across species. The two methods agree on 423 pairs at the p ≤ 0.05 level. The horizontal dashed line identifies 278 pairs that the across-species correlation declares significant but the LR method finds not significant. Are these false positives? The results in Figure 5A show that the across-species correlation is often misled by shared phylogenetic inheritance. It is tempting to speculate, therefore, that many of these 278 results are false positives, even though they are linked in S. cerevisiae. The vertical dashed line shows that the phylogenetic LR method identifies 186 pairs as significant that the correlation method declares not significant. Figure 4 suggests that LR method's extra 186 MIPS pairs are unlikely to be false positives.
The LR method's improvement over the across-species correlation seems principally to derive from correctly excluding spurious functional links that arise from shared phylogenetic inheritance, but also from correctly identifying some patterns of co-evolution that the across-species correlation misses. The two pairs of proteins shown alongside the phylogeny in Figure 2 illustrate this point. The across-species correlation is significant (p = 0.0014) between the pair {CIN4, ORC3}, whereas the phylogenetic method regards this as a chance association (p = 0.13) arising from a single event of both genes being gained in the ascomycete yeasts, followed by shared inheritance (like the red distribution patterns in Figure 1). In agreement with the phylogenetic approach, the proteins' known functions do not suggest a link. In contrast, the pair {L9A, L42B} consists of two proteins that are functionally linked, as components of the cytoplasmic ribosomal large subunit. The pair returns a significant phylogenetic correlation (p = 0.035). The across-species correlation is sensitive only to the distribution of the two proteins across the tips of the tree and returns a non-significant result (p = 0.23). If we assume that independent gains of the same gene are unlikely, then L9A was present relatively early on in the phylogeny, no later than the common ancestor to the Aspergillus nidulans-S. cerevisiae clade. It was lost in A. nidulans and separately in the Neurospora crassa-Fusarium graminearum group, even though L42B was present. L42B and L9A were lost together on five separate occasions spanning Candida albicans to Saccharomyces mikatae, but both were retained in S. paradoxus and S. cerevisiae. The across-species correlation is not sensitive to these changes, and its result is probably a false negative or type II error.
Contingent Gain or Loss of Genes
Contingent relationships between a pair of genes describe cases in which one gene is more likely to be gained or lost depending upon the state of the other. One example of this might be cases in which two genes are paralogues, and so one of the pair gets lost in each species owing to its redundant function. Other cases might identify instances in which one gene's function depends upon the presence of a second gene, but the second gene performs functions even in the absence of the first. Such contingent linkages may describe and explain many of the large number of cases in which two genes are functionally linked in one species, but they do not exclusively appear together across species. They can be detected by estimating the transition rate parameters of the dependent model (see Materials and Methods) [9] and looking for rates of evolution of one gene being dependent upon the presence or absence of the other.
Three cytoplasmic ribosomal large subunit proteins may provide an example of contingent evolution. Protein L30 is significantly linked to proteins L43A and L43B: both LRs = 9.73, p < 0.007. L43A and L43B are duplicates with identical protein sequence, and L30 is auto-regulatory [24]. The three proteins are present together in nine of the species, and are probably ancestral to the group represented by the phylogeny in Figure 2. Figure 6 represents this scenario on the left side of the diagram as all three proteins present. The remainder of the figure shows a model describing the contingent manner in which these ancestral proteins are lost. Solid arrows indicate most likely events of evolution to other evolutionary states, dashed arrows correspond to events for which no statistical support is found. See Materials and Methods for details of the transition rates qij. L30 is lost (q42 > 0) in two species while leaving L43A and L43B remaining. Once L30 is lost, the other two proteins follow (q21 > 0), yielding four species in which both proteins are absent. In comparison, L43A and L43B are never lost when L30 is present (q43 is not significantly different from zero). This may indicate a dependent relationship amongst these genes such that L43A and L43B acquire their function in the presence of L30.
Figure 6 Detecting Contingent Evolution Between Two Proteins
Protein L30 is significantly linked across species to L43A and L43B (both LRs = 9.73, p < 0.007). The three are present together in nine of the species, and are probably ancestral to the group represented by the phylogeny in Figure 2. The diagram represents the probable ancestral states on the left side. Solid arrows indicate the most likely events of evolution to other evolutionary states, and dashed arrows correspond to events for which no statistical support is found. L30 can be lost (q42 > 0), leaving L43 and L43B remaining, and this happens in two species. Once L30 is lost, the other two proteins follow (q21 > 0), yielding the remaining four species in which both proteins are absent. In comparison, L43A and L43B are never lost in the presence of L30. This suggests a contingent relationship amongst these proteins such that L43A and L43B seem to derive their functions only in the presence of L30.
Discussion
Incorporating phylogenetic information into predictions of functional gene links improved by between 18% and 35% upon predictions derived from across-species correlations, and increasingly so for pairs of genes with greater evidence of correlated evolution on the phylogeny. The phylogeny makes it possible to discriminate across-species patterns that arise by chance through common ancestry from those that indicate multiple independent instances of the correlated gain or loss of a pair of genes. This has implications for methods such as “phylogenetic profiling” [1,2], which, despite its name, does not make use of phylogenetic information when deriving predictions about functional links. In addition to reducing the number of false positives, incorporating phylogenetic information can sometimes recognize a true functional link even when the simple across-species pattern is vague and non-significant.
We find that the pairs of genes that have been gained or lost together on two to three or more occasions are almost certainly functionally linked. To our knowledge, this is the first phylogenetic demonstration that correlated evolutionary events strongly imply functional linkage, and underscores the importance of analysing events of protein evolution on phylogenetic trees. As the number of fully sequenced genomes increases, phylogenetic approaches can be used with increasing sensitivity to detect multiple events of correlated gene evolution, and by inference, pairs of genes with a high probability of being functionally linked.
We studied functional links on only a single phylogenetic tree rather than on a sample of trees, because we wished to compare results to the across-species correlation, which has no way of making use of the phylogenies. But it is straightforward to implement our approach in a Bayesian framework such that functional links are estimated across a sample of trees. Elsewhere we describe how to derive Bayesian posterior probability distributions of the parameters of the continuous-time Markov model of trait evolution, estimated over the posterior probability distribution of phylogenetic trees [25,26]. This accounts for uncertainty about the tree and about the parameters of the model of trait evolution, and can be especially valuable where there are disagreements about the placement of some species or groups of species.
A surprising number of gene pairs that are annotated as functionally linked in yeast do not appear to be linked in other, often closely related, species. Some of these may arise because a gene characterised as “absent” has simply gone unnoticed. We think this is only a small part of the explanation here, as we restricted ourselves to well-annotated, fully sequenced genomes. More likely is that the set of across-species functional links is far smaller than the set of all known links within any given species, and this raises the question of just what an across-species functional link measures. One distinct possibility is that a fundamental set or “backbone” of conserved protein interactions exists, in what might be called the “correlated evolution network.” This set of links is distinctive, in that the pairs of genes tend either to be both present or both absent. If so, their identification should be given a high priority, as they may reveal general organismic “rules of assembly.”
The highly specific nature of functional links also has implications for using model organisms to make predictions about other species, such as humans. Our data suggest that such predictions will often be wrong: Many genes whose functions and links have been identified from in-depth study in a model species may adopt different functions in other species. A phylogenetic method routinely applied to large numbers of species could distinguish the subset of genes whose functions can be reliably assumed to generalise from those that do not. Used in combination with low-throughput single-species studies, a more sophisticated picture may emerge.
In any analyses relying on identification of orthologues across species, multigene families may cause particular headaches. Assuming that the functionally conserved orthologue of a given gene will be under similar selection pressures and therefore have the greatest sequence similarity on average, reciprocal sequence similarity procedures such as we have used (see Materials and Methods) should perform well. Because the possibility of mis-identification can seldom be ruled out with certainty, additional evidence for correct annotation should be sought when a gene is suspected to be part of a larger family. Another approach is more practical: Simply exclude genes from consideration if they appear in multiple copies in a target species [27].
A large number of genes remain uncharacterised. Identifying functional linkages from phylogenetic events of co-evolution with other genes seems a promising way to understand function, and is an approach that can yield insights from currently poorly understood genomes. It is encouraging that we are able to detect functional links with reasonable sensitivity and specificity in a comparatively small number of species. Larger datasets will not only improve the ability to detect correlations; they will also make it possible to link events of correlated evolution to background organismic and ecological variables, and to identify clusters of genes that tend to appear together. Our approach can also be easily modified to use continuously varying data. Such data are increasingly becoming available from sequence similarity searches [3] and micro-array expression studies, and may provide a rich source of information on functional linkages and the nature of mRNA expression evolution [28].
Materials and Methods
The method requires a phylogeny of the organisms to be investigated, plus data on the presence and absence of homologous genes.
Gene-sequence data for phylogenetic tree.
All of our analyses are conducted on a phylogenetic tree of fifteen eukaryote species for which whole-genome data were available in 2003, including the 13 fungal species S. cerevisiae, S. bayanus,
S. mikatae,
S. paradoxus, S. castellii, and S. kluyveri (available from the Saccharomyces Genome Database, at ftp://ftp.yeastgenome.org/yeast/); C. albicans (Stanford Genome Technology Center, at http://www-sequence.stanford.edu/group/candida); S. pombe (Wellcome Trust Sanger Institute, at ftp://ftp.sanger.ac.uk); A. nidulans, F. graminearum, and M. grisea (Broad Institute, at http://www.broad.mit.edu); N. crassa (MIPS, at ftp://ftpmips.gsf.de; also the Broad Institute); and Cryptococcus neoformans (preliminary sequence data obtained through The Institute for Genomic Research, at http://www.tigr.org). In addition, genome data for two animals were used: C. elegans (Wormbase, at http://www.wormbase.org) and D. melanogaster (Berkeley Drosophila Genome Project, at http://www.fruitfly.org). Where two predicted protein sets existed for the same species (i.e., S. bayanus, S. mikatae, and N. crassa), these were combined into a single nonredundant set.
We used gene sequences for EF-1 alpha and EF-2 to infer the phylogenetic tree. We obtained proteins and their corresponding nucleotide sequences for each species, and aligned the data at the protein level using Clustal-X [29] before converting it back to nucleotides with Protal2DNA (http://bioweb.pasteur.fr). Ambiguously aligned codons were detected by eye and removed, yielding 1,425 aligned nucleotide sites.
Phylogenetic inference.
We reconstructed the phylogeny using a general time-reversible model of sequence evolution and allowing for gamma-distributed rate-heterogeneity (GTR+Γ). We found a tree for the 15 species using a heuristic search for the maximum likelihood (ML) tree in PAUP, and we refer to this as the ML tree. For comparison, we sampled the posterior probability distribution of phylogenies using the same model of evolution in a Bayesian MCMC framework (as described in [21] and using the program BayesPhylogenies). Bayesian methods are becoming increasingly popular in phylogenetic studies, providing a statistically rigorous way to describe uncertainty about the true phylogeny. After discarding the first 300,000 trees in the chain as a “burn-in” period, we sampled 500 trees at intervals of 50,000 trees to ensure that successive trees in our sample were independent (autocorrelation of log-likelihoods = 0.00). There was no increase in the mean log-likelihood after burn-in. The consensus tree topology from the posterior distribution was identical to the ML tree, with 100% posterior support at all nodes.
We used the single ML tree in all of our analyses of correlated evolution, rather than calculating correlated evolution across the Bayesian sample [25]. The latter approach is preferable in that it accounts for any effects of phylogenetic uncertainty on our results. In the present case, that uncertainty is limited to variation in branch lengths, because only one tree topology was found in the Bayesian sample. Variation in branch-lengths can influence likelihoods, but our interest is to compare the performance of phylogenetic and across-species methods, and the across-species method has no way of using the extra information in the Bayesian sample of trees.
Gene presence/absence data.
The MIPS database lists 260 known S. cerevisiae protein complexes and the 1,156 proteins that form them. Regarding each protein within a complex as functionally linked to every other different protein in that complex gives 10,551 pairwise links. For each of these S.
cerevisiae proteins, we sought orthologues in the 14 other species. We took a reciprocal best-in-genome global alignment between proteins to indicate “presence” of an orthologue and no such reciprocal hit to indicate “absence.” A heuristic approach was used (conceptually based on [31]), in which up to 20 best local alignments were obtained with BLASTP [32], then scores were re-assigned using Needleman-Wunsch alignment (EMBOSS Needle [33]).
Statistical modelling of correlated gene evolution.
We modelled correlated gene presence/absence on a phylogeny using a continuous-time Markov model [9,34]. The method compares the statistical likelihood of a model in which two binary traits are allowed to evolve independently on the tree, with a model in which the two traits are allowed to evolve in a correlated fashion. Evidence for correlated evolution arises if the dependent or correlated model shows a significantly better fit to the data than the independent model. The method has been applied to molecular evolution studies of prion proteins [35] and co-evolving protein residues [36]; and to estimating ancestral states of artiodactyl ribonucleases [25,37], rates of change in homing endonucleases [38], and the history of lichenization in fungal evolution [20].
We describe the method in some detail below, as our presentation only partially overlaps with that in [9], and elements of our description are pertinent to modelling gene presence/absence data. Here the binary trait is the presence or absence of each gene as observed in the species at the tips of the phylogeny (see, for example, Figure 1). Two binary traits can produce four different pairs of outcomes or states, corresponding to the pairings of presence or absence in two genes. The diagram in Equation 1 links the four states by arrows with parameters that describe the rates of transition between the two states of one of the genes, holding the state of the other constant. If two traits evolve independently of one another, then the rate of change between the presence (“1”) or absence (“0”) of one gene will not depend upon whether the other is present or absent. For example, if the rate of change from state “0” to state “1” in the second gene does not depend upon the state of the first variable, then q12 will be equal to q34. More generally, the model of independent evolution implies that q13
=
q24,
q42 = q31,
q43
=
q21, and q12 = q34 and therefore requires a maximum of four parameters.
The model of correlated evolution does not place any restrictions on the parameters, using a maximum of eight parameters to describe the data. The correlated evolution model will improve on the independent model when the distribution of the traits across the species of the phylogeny implies that some of the pairs of transition rates constrained in the independent model to be equal to each other, in fact differ. Information that pairs of coefficients differ arises not from the number of species that come to inherit a particular set of outcomes, but from the implied number of times the events represented by the rate coefficients have occurred on the tree. This is how the likelihood approach discriminates between the two scenarios of Figure 1.
The method is formally described by a rate matrix Q:
where we use the Q
I,D notation to indicate that the matrix can be configured to either the independent or dependent (correlated) model depending upon whether some pairs of transition rates are constrained to be equivalent. The main diagonal elements are defined as minus the sum of the other rate coefficients in the row of the matrix, such that each row sums to zero. The values of all dual transitions, or cases in which both traits change simultaneously, are set to zero in the matrix in Equation 2. Dual transitions are set to zero because their transition rate parameters measure the probability of both traits changing simultaneously in infinitesimally short interval dt. The probability of both traits changing in the same instant is negligibly small and can be ignored.
The model does allow both traits to change over a longer interval t, however. Thus if the four states in the matrix above are numbered 0, 1, 2, and 3, we can write the probability of a change in the short interval dt as Pij(dt) = qijdt, where i and j can take the values 0 to 3. The probabilities over longer intervals t are found by exponentiating the Q matrix multiplied by the length of the interval:
Combining these probabilities over all branches of the tree yields the likelihood of the data, L. The method of ML finds the values of the rate parameters in Q that make the observed data most probable, given the Q and the phylogenetic tree.
The likelihood is summed over all possible ancestral state reconstructions [9,39], meaning that results do not depend upon any particular inferred evolutionary history of the traits. In the case of gene presence/absence data, this can mean that a gene is allowed to arise or evolve more than once on the tree, something that is probably highly unlikely except in the case of lateral gene transfer. In practice, we expect that most correlated evolution takes the form of coincident losses of genes. However, given that homology is assessed as similarity at the sequence level, small amounts of convergent evolution could make two initially dissimilar sequences become more similar and thereby appear as orthologues. To exclude or reduce the effect of allowing multiple gains among highly diverged species, we fixed the root of the tree at “present” for any gene that was found on both sides of the major bifurcation that the root defines. This causes the model to favour losses for those pairs.
The Discrete method can be used with a single tree and is now implemented in a Bayesian framework in the program BayesDiscrete, to account for uncertainty in both the estimates of the phylogeny and in the parameters of the model of correlated evolution. It is available from M. Pagel and A. Meade at http://www.ams.reading.ac.uk/zoology/pagel.
Hypothesis testing.
When the independent and dependent models are estimated by maximum likelihood, their goodness of fit is compared using the LR statistic: LR = −2 log(H0) − log(H1), where H0 is here the likelihood of the model of independent evolution and H1 is the model of dependent evolution. This test statistic is asymptotically distributed as a χ2 variate with degrees of freedom equal to the difference in the number of parameters of the two models (here four). If LR exceeds the critical value of the χ2 distribution with appropriate degrees of freedom, then the H1 model is judged a better fit to the data.
When the phylogeny contains a small number of species or rates of evolution are low, the LR statistic as defined above is often distributed with fewer than four degrees of freedom [34]. In these cases, the correct null hypothesis distribution can be simulated following [9,34].
Across-species correlation.
We used Fisher's exact test (e.g., [40]) to analyse the 2 × 2 table, recording the number of species with each of the four possible categories of presence/absence of two genes. This test makes no distributional assumptions and returns an exact p-value for the null hypothesis that the two traits are distributed independently.
Supporting Information
Accession Numbers
Swiss-Prot (http://www.ebi.ac.uk/swissprot/) accession numbers for the yeast proteins discussed in this paper are CIN4 (P39110), L9A (P05738), L30 (P14120), L42B (P02405), L43A (P49631), L43B (P49631), and ORC3 (P54790).
We acknowledge the support of the Biotechnology and Biological Sciences Research Council, UK (Grants 19848 and 14980 to MP). Software to implement the methods described here is available from M. Pagel and A. Meade at http://www.ams.reading.ac.uk/zoology/pagel. We thank Jill Harrison and Val Wood for advice.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. DB and MP conceived and designed the experiments, performed the experiments, analysed the data, contributed reagents/materials/analysis tools, and wrote the paper.
Abbreviations
LRlikelihood ratio
MCMCMarkov chain Monte Carlo
MIPSMunich Information Center for Protein Sequences
MLmaximum likelihood
==== Refs
References
Pellegrini M Marcotte EM Thompson MJ Eisenberg D Yeates TO 1999 Assigning protein functions by comparative genome analysis: Protein phylogenetic profiles Proc Natl Acad Sci U S A 96 4285 4288 10200254
Eisenberg D Marcotte EM Xenarios I Yeates TO 2000 Protein function in the post-genomic era Nature 405 823 826 10866208
Date SV Marcotte EM 2003 Discovery of uncharacterized cellular systems by genome-wide analysis of functional linkages Nat Biotechnol 21 1055 1062 12923548
Fraser HB Hirsh AE Wall DP Eisen MB 2004 Coevolution of gene expression among interacting proteins Proc Natl Acad Sci U S A 101 9033 9038 15175431
Pazos F Valencia A 2001 Similarity of phylogenetic trees as indicator of protein-protein interaction Protein Eng 14 609 614 11707606
von Mering C Krause R Snel B Cornell M Oliver SG 2002 Comparative assessment of large-scale data sets of protein-protein interactions Nature 417 399 403 12000970
Jeong H Tombor B Albert R Oltvai ZN Barabàsi A-L 2000 The large-scale organization of metabolic networks Nature 407 651 654 11034217
Ettema T van der Oost J Huynen M 2001 Modularity in the gain and loss of genes: Applications for function prediction Trends Genet 17 485 487 11525815
Pagel M 1994 Detecting correlated evolution on phylogenies: A general method for the comparative analysis of discrete characters Proc R Soc Lond B Biol Sci 255 37 45
Pagel M 1999 Inferring the historical patterns of biological evolution Nature 401 877 884 10553904
Felsenstein J 1985 Phylogenies and the comparative method Am Nat 125 1 15
Harvey P Pagel M 1991 The comparative method in evolutionary biology. Oxford Oxford University Press 239 p.
Mewes HW Frishman D Güldener U Mannhaupt G Mayer K 2002 MIPS: A database for genomes and protein sequences Nucleic Acids Res 30 31 34 11752246
Gilks WR Richardson S Spiegelhalter DJ 1996 Introducing Markov chain Monte Carlo Gilks WR Richardson S Spiegelhalter DJ Markov chain Monte Carlo in practice. London Chapman and Hall 1 19
Rannala B Yang Z 1996 Probability distributions of molecular evolutionary trees: A new method of phylogenetic inference J Mol Evol 43 304 311 8703097
Wilson I Balding D 1998 Genealogical inference from microsatellite data Genetics 150 499 510 9725864
Larget B Simon DL 1999 Markov chain Monte Carlo algorithms for the Bayesian analysis of phylogenetic trees Mol Biol Evol 16 750 759
Mau B Newton M Larget B 1999 Bayesian phylogenetic inference via Markov chain Monte Carlo methods Biometrics 55 1 12 11318142
Huelsenbeck JP Ronquist F Nielsen R Bollback JP 2001 Bayesian inference of phylogeny and its impact on evolutionary biology Science 294 2310 2314 11743192
Lutzoni F Pagel M Reeb V 2001 Major fungal lineages derived from lichen-symbiotic ancestors Nature 411 937 940 11418855
Pagel M Meade A 2004 A phylogenetic mixture model for detecting pattern-heterogeneity in gene sequence or character-state data Syst Biol 53 571 581 15371247
Rokas A Williams BL King N Carroll SB 2003 Genome-scale approaches to resolving incongruence in molecular phylogenies Nature 425 798 804 14574403
Lutzoni F Kauff F Cox CJ McLaughlin D Celio G 2004 Assembling the fungal tree of life: Progress, classification, and evolution of subcellular traits Am J Bot 91 1446 1480 21652303
Mao H White SA Williamson JR 1999 A novel loop-loop recognition motif in the yeast ribosomal protein L30 autoregulatory RNA complex Nat Struct Biol 6 1139 1147 10581556
Pagel M Meade A Barker D 2004 Bayesian estimation of ancestral character states on phylogenies Syst Biol 53 673 684 15545248
Pagel M Meade A 2005 Bayesian estimation of correlated evolution across cultures: A case study of marriage systems and wealth transfer at marriage Mace R Holden CJ Shennan S The evolution of cultural diversity: A phylogenetic approach. London University College London Press 235 256
von Mering C Zdobnov EM Tsoka S Ciccarelli FD Pereira-Leal JB Ouzonis CA Bork P 2003 Genome evolution reveals biochemical networks and functional modules Proc Natl Acad Sci U S A 100 15428 15433 14673105
Khaitovich P Weiss G Lachmann M Hellmann I Enard W 2004 A neutral model of transcriptome evolution PLoS Biol 2 682 689
Thompson JD Gibson TJ Plewniak F Jeanmougin F Higgins DG 1997 The CLUSTAL_X windows interface: Flexible strategies for multiple sequence alignment aided by quality analysis tools Nucleic Acids Res 25 4876 4882 9396791
Swofford DL 2002 PAUP*: Phylogenetic Analysis Using Parsimony (and other methods), version 4.0 Beta 10. Sunderland, Massachusetts Sinauer Associates Available: http://paup.csit.fsu.edu .
Korf I 2003 Serial BLAST searching Bioinformatics 19 1492 1496 12912829
Altschul SF Madden TL Schaffer AA Zhang J Zhang Z 1997 Gapped BLAST and PSI-BLAST: A new generation of protein database search programs Nucleic Acids Res 25 3389 3402 9254694
Rice P Longden I Bleasby A 2000 EMBOSS: The European Molecular Biology Open Software Suite Trends Genet 16 276 277 10827456
Pagel M 1997 Inferring evolutionary processes from phylogenies Zoologica Scripta 26 331 348
Krakauer DC Pagel M Southwood TR Zanotto PM 1996 Phylogenesis of prion protein Nature 380 675 8614459
Pollock DD Taylor WR Goldman N 1999 Coevolving protein residues: Maximum likelihood identification and relationship to structure J Mol Biol 287 187 198 10074416
Schluter D 1995 Uncertainty in ancient phylogenies Nature 377 108 109 7675077
Goddard M Burt A 1999 Recurrent invasion and extinction of a selfish gene Proc Natl Acad Sci U S A 96 13880 13885 10570167
Felsenstein J 1981 Evolutionary trees from DNA sequences: A maximum likelihood approach J Mol Evol 17 368 376 7288891
Zar JH 1996 Biostatistical analysis, 3rd edition. Upper Saddle River, New Jersey Prentice Hall 662 p.
|
16103904
|
PMC1183509
|
CC BY
|
2021-01-05 09:18:21
|
no
|
PLoS Comput Biol. 2005 Jun 24; 1(1):e3
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010003
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science 1610390510.1371/journal.pcbi.001000405-PLCB-ED-0098plcb-01-01-02EditorialBioinformatics - Computational BiologyNone
PLoS Computational Biology: A New Community Journal EditorialBourne Philip E Brenner Steven E Eisen Michael B Philip E. Bourne is founding editor and editor-in-chief (e-mail: [email protected]), Steven E. Brenner is founding editor, and Michael B. Eisen is founding editor of PLoS Computational Biology.
6 2005 24 6 2005 1 1 e4Copyright: © 2005 Bourne, Brenner, and Eisen.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited. Citation:Bourne PE, Brenner SE, Eisen MB (2005) PLoS Computational Biology: A new community journal. PLoS Comput Biol 1(1): e4.
==== Body
Welcome to PLoS Computational Biology, a community journal from the Public Library of Science dedicated to reporting biological advances achieved through computation. The journal is published in partnership with the International Society for Computational Biology (ISCB). The importance of this partnership is described in the accompanying letter from Michael Gribskov, ISCB president.
What motivates us to start a new journal at this time? Computation, driven in part by the influx of large amounts of data at all biological scales, has become a central feature of research and discovery in the life sciences. This work tends to be published either in methods journals that are not read by experimentalists or in one of the numerous journals reporting novel biology, each of which publishes only small amounts of computational research. Hence, the impact of this research is diluted. PLoS Computational Biology provides a home for important biological research driven by computation—a place where computational biologists can find the best work produced by their colleagues, and where the broader biological community can see the myriad ways computation is advancing our understanding of biological systems.
PLoS Computational Biology is governed by one overarching principle: scientific quality. This quality is reflected in the editorial board and the editorial staff. The editorial board members are leaders in their respective scientific areas and have agreed to give their valuable time to support a quality journal in their field. Behind the scenes, through a rigorous presubmission process, three quality reviews for each paper, and an acceptance rate below 20%, the editors and staff already knew in the six months since the journal was launched that we were producing a first-rate product. The scientific content is now here for all of you to see and will continue to build in the months and years to come.
Computational biology thrives on open access to biological data.
The inaugural issue of PLoS Computational Biology includes a Perspective from Sean Eddy on his belief that we are in the “ante”-disciplinary phase of our science—a time ahead of becoming an organized science and not to be confused with the push toward being “inter”-disciplinary, as many would have us become. Perspectives, and, later, Reviews, will be a regular feature of the journal. This is followed by a series of research articles, each including a synopsis distinct from the scientific abstract, written by the authors for nonexperts. Among these articles, Pagel et al. describe a method for determining pairs of functionally related genes and then use that method to study how these pairs are either gained or lost over evolutionary time; Margalit et al. report on the modeling of transcription factor binding sites and discover 29 such binding sites for C2H2 zinc-finger proteins in Drosophila melanogaster; and Ehrenberg et al. model ribosome-mediated transcriptional attenuation and explain the functional differences that result from the structural differences of the leader sequences of the trp and his operons.
The vision we have for PLoS Computational Biology as a community journal is first and foremost to support the dissemination of our science in a way that draws attention to the quality, depth, and scope of our best work. That this is an open-access journal is an integral part of this vision. Open access ensures not only that everything we publish is immediately freely available to anyone, anywhere in the world, but also that the contents of this journal can be redistributed and reused in ways that increase their value. Computational biology thrives on open access to DNA sequences, protein structures, and other types of biological data—it is high time that we apply the same principle to our papers and unleash our creativity to develop new and exciting ways to use the scientific literature.
To realize this vision, we need your help. We need you to submit your best work to the journal and encourage colleagues to do so as well. Beyond that, recognize that we are a unique community in that we use information in novel ways. The journal itself is just another information stream that can be processed and analyzed. We have our own ideas for what that means and how it can change the face of publishing, but we want your input too. These are exciting times for biology and publishing. We hope to capitalize on this excitement, so do not hesitate to tell us what we should be doing by sending e-mail to [email protected].
|
16103905
|
PMC1183510
|
CC BY
|
2021-01-05 09:18:21
|
no
|
PLoS Comput Biol. 2005 Jun 24; 1(1):e4
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010004
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science 1610390610.1371/journal.pcbi.001000505-PLCB-MI-0097plcb-01-01-03Message from ISCBBioinformatics - Computational BiologyNoneAn Open Forum for Computational Biology Message from ISCBGribskov Michael Michael Gribskov is president of the International Society for Computational Biology. E-mail: [email protected]
6 2005 24 6 2005 1 1 e5Copyright: © 2005 Michael Gribskov.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited. Citation:Gribskov M. (2005) A Message from ISCB. PLoS Comput Biol 1(1): e5.
==== Body
It is my pleasure to welcome you to PLoS Computational Biology. The International Society for Computational Biology (ISCB) is proud to be, with the Public Library of Science, the co-sponsor of this journal. We believe that PLoS Computational Biology will rapidly become a leading journal in the area of computational biology and, as an official journal of ISCB, will be an important venue in which ISCB members will publish their findings and learn of the work of others, both ISCB members and nonmembers.
New journals appear at frequent intervals these days, and one may well ask: why a new computational biology journal? PLoS and ISCB feel that there is both need and interest in a journal with a new focus and new approach to computational biology. This feeling grows out of our recognition of the enormous changes that computation has wrought in both science and communication.
In contrast to the situation of only a decade or so ago, almost every area of biology is now affected and enhanced by computational studies. But until the appearance of PLoS Computational Biology, there has been no single publication with a focus on the important contributions of computational studies to the understanding of living systems. PLoS Computational Biology meets this need and provides a forum in which experimentalists and computational scientists can meet, exchange ideas, and develop new solutions to biological questions.
The revolution in communication that has grown out of the Internet and World Wide Web infrastructure provides the other motivating force behind the creation of a new journal. Free availability of protein and nucleic acid sequences, protein structures, and other biological data is critical to practitioners of computational biology—support of open-access journals is ISCB and PLoS launch an open-access journal with a new focus and new approach to computational biology. therefore a logical and natural progression for ISCB. Open access was a key factor in choosing PLoS as a partner. The ISCB believes that open access to scientific research is an important means of more rapidly disseminating scientific advances, and of communicating the findings of ISCB members to scientists around the world. Open access provides distinct advantages to scientific authors, including greater freedom to reuse and republish text, and greater availability of text for electronic data mining. In combination with rapid publication and high editorial standards, these provide compelling reasons for computational biologists to read and publish in PLoS Computational Biology.
In choosing PLoS Computational Biology as an official journal of the society, ISCB was also guided by the quality of PLoS, its board of directors, and senior editorial staff. It has been, and will continue to be, a pleasure collaborating with such an experienced and highly skilled group. PLoS brings great expertise in an experienced editorial staff and in developing the open-access publishing model. The experience and expertise of PLoS ensure that PLoS Computational Biology meets the highest production standards. Scientific leadership from ISCB, including the appointment of Philip E. Bourne, past-president of ISCB, as editor-in-chief will ensure the highest scientific standards. Moreover, the authors of each paper published in PLoS Computational Biology will be able to nominate one of their number to receive a complimentary membership to the ISCB. The Society and the journal thus work together to strengthen our community.
The participation of ISCB members will be critical to the success of PLoS
Computational Biology. As authors, contributors, reviewers, and editorial board members, the members of ISCB will work with PLoS in building a successful journal. Together, ISCB and PLoS make ideal partners, and I am sure that you will join me in eagerly awaiting the first issues of PLoS Computational Biology.
|
16103906
|
PMC1183511
|
CC BY
|
2021-01-05 09:18:21
|
no
|
PLoS Comput Biol. 2005 Jun 24; 1(1):e5
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010005
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science 1610390710.1371/journal.pcbi.001000605-PLCB-PV-0095plcb-01-01-06PerspectivesBioinformatics - Computational BiologyNone“Antedisciplinary” Science PerspectivesEddy Sean R Sean R. Eddy is at Howard Hughes Medical Institute and the Department of Genetics at Washington University School of Medicine in Saint Louis, Missouri, United States of America. E-mail: [email protected]
6 2005 24 6 2005 1 1 e6Copyright: © 2005 Sean R. Eddy.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited. Citation:Eddy SR (2005) “Antedisciplinary” science. PLoS Comput Biol 1(1): e6.
==== Body
“The scale and complexity of today's biomedical research problems demand that scientists move beyond the confines of their individual disciplines and explore new organizational models for team science. Advances in molecular imaging, for example, require collaborations among diverse groups—radiologists, cell biologists, physicists, and computer programmers.” —National Institutes of Health Roadmap Initiative [1]
Reading this made me a little depressed. For starters, the phrase “organizational models for team science” makes me imagine a factory floor of scientists toiling away on their next 100-author paper under the watchful gaze of their National Institutes of Health program officers, like some scene from Terry Gilliam's movie Brazil. It's also depressing to read that the National Institutes of Health thinks that science has become too hard for individual humans to cope with, and that it will take the hive mind of an interdisciplinary “research team of the future” to make progress. But what's most depressing comes from purely selfish reasons: if groundbreaking science really requires assembling teams of people with proper credentials from different disciplines, then I have made some very bad career moves.
I've been a computational biologist for about 15 years now. We're still not quite sure what “computational biology” means, but we seem to agree that it's an interdisciplinary field, requiring skills in computer science, molecular biology, statistics, mathematics, and more. I'm not qualified in any of these fields. I'm certainly not a card-carrying software developer, computer scientist, or mathematician, though I spend most of my time writing software, developing algorithms, and deriving equations. I do have formal training in molecular biology, but that was 15 years ago, and I'm sure my union card has expired. For one thing, they all seem to be using these clever, expensive kits now in my wet lab, whereas I made most of my own buffers (after walking to the lab six miles in the snow, barefoot).
If I thought I was the only person who abandoned disciplinary training to take up a new area of science, after reading about the “research teams of the future,” I might slink away and find something else to do before the future arrives. But I don't think I'm alone. I was recently at a meeting where people started discussing these interdisciplinary “research teams of the future,” and Howard Berg, who had just given a wonderful chalk talk about bacterial chemotaxis, was sitting behind me. I heard him mutter that he wondered how a misfit like him was going to fit into this new world order. Well, he's doomed. He's successfully applied physical, mathematical, and biological approaches to an important problem without enlisting an interdisciplinary team of properly qualified physicists, mathematicians, and biologists. As he recently wrote, perhaps he'll have to start collaborating with himself [2].
I wonder if it's the success of the Human Genome Project that led us to this. The scale of the genome project required “big science” and large teams. The genome project also fueled the explosive growth of the highly successful field of computational biology. Did the ideas of interdisciplinary science and large teams become inappropriately intertwined? Certainly, achieving the goals of the Human Genome Project required engineers, physicists, and computer scientists. It would be silly to argue against large interdisciplinary teams where a mammoth technical goal can be clearly defined. But when I think of new fields in science that have been opened, I don't think of interdisciplinary teams combining existing skills to solve a defined problem—I think of single interdisciplinary people inventing new ways to look at the world.
Focusing on interdisciplinary teams instead of interdisciplinary people reinforces standard disciplinary boundaries rather than breaking them down. An interdisciplinary team is a committee in which members identify themselves as an expert in something else besides the actual scientific problem at hand, and abdicate responsibility for the majority of the work because it's not their field. Expecting a team of disciplinary scientists to develop a new field is like sending a team of monolingual diplomats to the United Nations.
Progress is driven by new scientific questions, which demand new ways of thinking. You want to go where a question takes you, not where your training left you. We may not have a single clarion call to arms like Schrödinger's What is Life? driving physicists into biology right now, as in the beginnings of molecular biology. But we do have powerful new technologies to harness (computational biology), newly revitalized approaches to old problems (systems biology), and new areas altogether (synthetic biology). New disciplines eventually self-organize around new problems and approaches, creating a new shared culture. This shared culture coalesces into the next essential training regimen for the next generation of scientists, and with luck, some of these people will overcome their training to open up more new fields of inquiry. Interdisciplinary science is just the embryonic stage of a new discipline. To value interdisciplinary science for its own sake is to value history over progress—that is, to value people's past training more than their current work.
Don't get me wrong. Certainly experience does affect how problems are approached, and it's synergistic to bring together people with different ideas. It's just a question of emphasis. In a marriage, previous experience affects what is brought to the partnership; but dwelling too much on prior experience causes the commitment to the current project to be called into question. Show me someone working on modeling the yeast cell cycle who still calls himself a physicist, and I'll show you someone with commitment issues.
Consider, for instance, the rise of molecular biology as a discipline. We think of Watson and Crick as molecular biologists, not as an ornithologist and a physicist. The first molecular biologists were a motley crew of misfits and revolutionaries with no particularly relevant training, many of them ex-physicists. These physicists didn't waste much time identifying themselves as physicists any more. They viewed themselves as a new kind of biologist. They burned their bridges. Max Delbrück dropped physics and started studying phage replication because it seemed like the fastest, best way to crack the molecular basis of heredity. It's hard to imagine molecular biology making such dramatic progress if it had involved forming interdisciplinary teams of physicists and biologists. The molecular biologists were viewed as naive infidels. Biochemist Erwin Chargaff sniffed that “molecular biology is the practice of biochemistry without a license” [3].
Molecular biologists even worried about what to call themselves, like we argue over whether we're computational biologists or bioinformaticians. Any revolution needs to find the right slogan to unify under. Francis Crick explained, “I myself was forced to call myself a molecular biologist because when inquiring clergymen asked me what I did, I got tired of explaining that I was a mixture of crystallographer, biophysicist, biochemist, and geneticist, an explanation which in any case they found too hard to grasp” [4].
To encourage the rise of new disciplines as successful as molecular biology, we need to encourage individuals to leave old disciplines behind and forge new fields. New science needs to be judged on its merits, not by the disciplinary credentials of the people doing it—particularly in fast-moving interdisciplinary areas where any formal training may be outdated anyway. If your grant proposal includes statistical analysis, your reviewers shouldn't be acting as enforcers requiring you to have a card-carrying statistician as a collaborator. Maybe in your narrow area, you know how to do the relevant statistics as well as any formally trained statistician. A proposal invoking high-performance computing should not get held up until you enlist collaborating computer scientists, who may not even be interested in your problem. Maybe you know how to use a supercomputer well enough to do what you propose.
Perhaps the whole idea of interdisciplinary science is the wrong way to look at what we want to encourage. What we really mean is “antedisciplinary” science—the science that precedes the organization of new disciplines, the Wild West frontier stage that comes before the law arrives. It's apropos that antedisciplinary sounds like “anti-disciplinary.” People who gravitate to the unexplored frontiers tend to be self-selected as people who don't like disciplines—or discipline, for that matter.
One can't deny that science is getting more complex, because the sheer amount of knowledge is growing. But the history of science is full of ideas that seemed radical, unfathomable, and interdisciplinary at the time, but that now we teach to undergraduates. Every generation, we somehow compress our knowledge just enough to leave room in our brains for one more generation of progress. This is not going to stop. It may take big interdisciplinary teams to achieve certain technical goals as they come tantalizingly within view, but someone also needs to synthesize new knowledge and make it useful to individual human minds, so the next generation will have a taller set of giants' shoulders to stand on. Computer science mythologizes the big teams and great computing engines of Bletchley Park cracking the Enigma code as much as we mythologize the Human Genome Project, but computer science rests more on the lasting visions of unique intellectual adventurers like Alan Turing and John von Neumann. Looking around my desk at the work I'm trying to build on, I do see the human genome paper, but even more, I see the work of individual pioneers who left old disciplines and defined new ones—writing with the coherence, clarity, and glorious idiosyncrasy that can only come from a single mind.
==== Refs
References
National Institutes of Health Roadmap 2005 Research teams of the Future Available: http://nihroadmap.nih.gov/researchteams/index.asp . Bethesda (Maryland) National Institutes of Health Accessed 25 May 2005.
Berg H 2005 Q&A: Howard Berg Curr Biol 15 R189 R190 15825267
Chargaff E 1978 Heraclitean fire: Sketches from a life before nature New York Rockefeller University Press 252 p.
Crick FHC 1965 Recent research in molecular biology: Introduction Br Med Bull 21 183 186 5825405
|
16103907
|
PMC1183512
|
CC BY
|
2021-01-05 09:18:21
|
no
|
PLoS Comput Biol. 2005 Jun 24; 1(1):e6
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010006
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science 1610390810.1371/journal.pcbi.001000705-PLCB-RA-0017R2plcb-01-01-05Research ArticleBioinformatics - Computational BiologyGenetics/GenomicsGenetics/Functional GenomicsMolecular Biology - Structural BiologyEukaryotesSaccharomycesSusceptibility to Superhelically Driven DNA Duplex Destabilization: A Highly Conserved Property of Yeast Replication Origins Superhelically Driven DNA Duplex DestabilizationAk Prashanth *Benham Craig J UC Davis Genome Center, University of California, Davis, California, United States of AmericaBourne Philip EditorUniversity of California at San Diego, United States of America* To whom correspondence should be addressed. E-mail: [email protected] 2005 24 6 2005 1 1 e71 2 2005 10 5 2005 Copyright: © 2005 Ak and Benham.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Strand separation is obligatory for several DNA functions, including replication. However, local DNA properties such as A+T content or thermodynamic stability alone do not determine the susceptibility to this transition in vivo. Rather, superhelical stresses provide long-range coupling among the transition behaviors of all base pairs within a topologically constrained domain. We have developed methods to analyze superhelically induced duplex destabilization (SIDD) in genomic DNA that take into account both this long-range stress-induced coupling and sequence-dependent local thermodynamic stability. Here we apply this approach to examine the SIDD properties of 39 experimentally well-characterized autonomously replicating DNA sequences (ARS elements), which function as replication origins in the yeast Saccharomyces cerevisiae. We find that these ARS elements have a strikingly increased susceptibility to SIDD relative to their surrounding sequences. On average, these ARS elements require 4.78 kcal/mol less free energy to separate than do their immediately surrounding sequences, making them more than 2,000 times easier to open. Statistical analysis shows that the probability of this strong an association between SIDD sites and ARS elements arising by chance is approximately 4 × 10−10. This local enhancement of the propensity to separate to single strands under superhelical stress has obvious implications for origin function. SIDD properties also could be used, in conjunction with other known origin attributes, to identify putative replication origins in yeast, and possibly in other metazoan genomes.
Synopsis
Several DNA functions require the two strands of the DNA duplex to transiently separate. Examples include the initiation of gene expression and of DNA replication. Here the authors examine the strand separation properties of the DNA duplex at autonomously replicating sequences (ARS elements), which are the potential replication origins in yeast.
In vivo, susceptibility to strand separation does not depend only on local DNA properties such as adenine plus thymine content or thermodynamic stability. Rather, stresses imposed on the DNA in vivo couple together the strand-opening behaviors of all base pairs that experience them. The authors use computational methods for analyzing stress-driven strand separation to examine the susceptibility to opening of 39 experimentally well-characterized ARS elements. They show that these ARS elements have strikingly increased susceptibilities to stress-induced separation relative to the surrounding sequences. On average, these ARS elements require 4.78 kcal/mol less free energy to separate than do surrounding sequences, making them more than 2,000 times easier to open. This enhanced susceptibility to stress-driven strand separation has obvious implications for the mechanisms that begin the process of replication. This property is also shared by bacterial and viral replication start points, suggesting that it may be a general attribute of replication origins.
Citation:Ak P, Benham CJ (2005) Susceptibility to superhelically driven DNA duplex destabilization: A highly conserved property of yeast replication origins. PLoS Comput Biol 1(1): e7.
==== Body
Introduction
In eukaryotes, DNA replication is initiated at multiple origins. Potential sites in the genome of the yeast Saccharomyces cerevisiae that may serve this function are referred to as autonomously replicating sequences, or ARS elements [1]. ARS elements are more A+T-rich than the genomic average, and contain regions of low local thermodynamic stability that are thought to be necessary for function [2,3]. However, the duplex unwinding required for replication initiation occurs as an isothermal process within topologically constrained domains of DNA. Under these conditions susceptibility to strand opening is not dependent only on local thermodynamic stability. Instead, superhelical stresses couple together the strand-opening behaviors of all base pairs that experience them. We hypothesize that the superhelical stresses that occur in vivo play a role in regulating the strand opening needed to initiate replication. This suggests that ARS elements should have an increased local susceptibility to superhelically induced duplex destabilization (SIDD). Here we demonstrate that virtually all known ARS elements do indeed show a significant local increase in their predicted SIDD susceptibility. Experiments on four specific ARS-containing regions have shown that each does experience local denaturation when negatively supercoiled [4].
We have calculated the SIDD properties of the entire yeast genome using a previously developed statistical mechanical method that includes both sequence-specific thermal stability and the global coupling induced by superhelical stresses within topological domains. (The algorithms implementing this method have been presented elsewhere [5].) This method computes the destabilization energy G(x) for each base pair (sometimes also called the SIDD energy) under the specified environmental conditions and level of superhelicity. This is the incremental free energy that is needed to guarantee separation of base pair x under these conditions. We note that G(x) is directly related to stability, not to destabilization—the higher the value of G(x), the more energy is needed to force that base pair open, and hence the more stable it is. Base pairs having G(x) near 10 kcal/mol remain essentially as stable under the assumed level of superhelicity as they would be in a relaxed molecule. (The majority of the base pairs act this way; significant superhelical destabilization is limited to a small fraction of the genome, as shown below.) Sites with G(x) near zero are strongly destabilized, and would denature with high probability under these conditions, while partially destabilized sites have intermediate values of G(x).
The SIDD energy G(x) is more informative regarding the extent of destabilization than is the probability p(x) of denaturation, because it also finds positions of partial destabilization. These can be biologically important because partial destabilization, even by only a few kilocalories, can greatly facilitate opening by other processes, such as interactions with regulatory molecules. For example, superhelical destabilization by only 3 kcal/mol (i.e., G(x) changing from 10 kcal/mol in a relaxed molecule to 7 kcal/mol under superhelicity), which is far less than is needed to open the duplex, will still increase the ease of opening by other processes by a factor of 130 (see Materials and Methods.) In this way changes in the level of imposed superhelicity can have strong effects on the rates of occurrence of regulatory events, especially those whose rate-limiting steps involve DNA strand opening.
Although these calculations have no free parameters, comparisons with experiments have shown that their predictions are quantitatively accurate. They determine the locations of opening and the extents of opening, both as functions of imposed superhelicity, at an accuracy comparable to experimental measurements in all sequences on which such experiments have been performed [6,7]. Many sites that these methods had previously calculated would open under stress have subsequently been experimentally shown to separate under these conditions, both in vitro and in vivo [8–11]. This gives confidence in the accuracy of their predictions when applied to other sequences on which experiments have not been performed.
Our approach for analyzing duplex destabilization differs fundamentally from others, such as the THERMODYN and MELTMAP algorithms [12–14], which only consider local thermodynamic stability or A+T content. SIDD does not depend on such local properties alone; rather, transitions in superhelical domains are globally interactive. Because strand separation localizes some of the imposed negative superhelicity as untwisting at the open site, it causes a corresponding relaxation that is felt throughout the topological domain. So denaturation at any site will alter the opening probabilities of every other site in the domain. This global coupling can lead to complex interactive transition behaviors that are not reflected by local thermodynamic stability [9]. An example of the long-range coupling induced by superhelicity is shown in Figure 1.
Figure 1 Comparison of Stress-Induced Duplex Destabilization Calculations with Assessments of Thermodynamic Stability
(A) The SIDD profile showing the G(x) values computed for a 3-kbp region (525701–528700) on Chromosome X of yeast at superhelix density σ = −0.055.
(B) The SIDD profile of a 38-bp deletion mutant of the same region [28], at the same superhelicity. The deletion is at positions 526489–526526, indicated by the red arrow. This deletion causes drastic changes of SIDD properties throughout the region, even 2 kbp away. This is an effect of the global coupling induced by the superhelical stresses.
(C) Thermodynamic stability profiles of the same regions as computed by WEB-THERMODYN [12,13], both before (black) and after (red) the deletion. The only effect of this deletion, whose location is indicated by the red arrow, is to displace the downstream profile by 38 bp. However, as shown in (B), the SIDD profile is profoundly altered throughout the region.
A genome-wide view of destabilization properties offers new perspectives on chromosomal organization and, specifically, on the structural properties of DNA regulatory regions. For yeast these include transcriptional regulatory sites (unpublished data) and the sites regulating the initiation of replication, considered here. SIDD has been implicated in the functioning of replication origins in a variety of organisms. The unique replication origin in E. coli, oriC, is superhelically destabilized, and this destabilization has been implicated in its function [15]. Other work has documented pathological origin activity at SIDD sites created by expansion of the pentameric repeat, causing spinocerebellar ataxia type 10 [7]. A role has been established for SIDD in the function of the Epstein–Barr oriP origin [16]. Here we focus on developing a genome-wide view of yeast replication origins.
Results
SIDD analysis of the complete yeast genome has been performed under the conditions described in the Materials and Methods section. The cumulative distribution of G(x) is shown in Figure 2. One sees that most of this genome is not significantly destabilized; half of the base pairs have G(x) greater than 9.13 kcal/mol. Only 7.23% of the base pairs have G(x) less than 4 kcal/mol under these conditions, while just 3.48 % have G(x) less than 2 kcal/mol, indicative of substantial destabilization. Moreover, the significantly destabilized sites are largely confined to regulatory regions governing either transcription (unpublished data) or replication.
Figure 2 The Cumulative Distribution of Destabilization Levels G(x) for the Entire Yeast Genome
For each value of G on the horizontal axis, this curve plots the number of base pairs (expressed as a percent of the genome) needing that amount of free energy (or less) to strand separate. G = 10 kcal/mol is sufficient to open any base pair in the genome.
The SIDD Properties of ARS Elements
An exhaustive literature search found 39 experimentally well-characterized ARS elements. These are relatively short regions that function as replication origins in a standard in vivo plasmid assay. The site within each element that acts as the replication origin under these circumstances is usually not more precisely identified.
Visual examination of the SIDD profiles of the genomic locations containing these ARS elements shows that the specific ARS sites occur at positions having low G(x) values, and hence are highly susceptible to destabilization by superhelical stress and thereby unusually prone to strand separation. A representative example is presented in Figure 3. (A complete list of these elements and the SIDD profiles of regions containing each element are presented in Table S1 and Protocol S1, respectively.)
Figure 3 ARS Elements Are More Destabilized Than Other Parts of the Genome
(A) The SIDD profile of a region (bp 166936–168740) of Chromosome 6, containing ARS606 (position marked in red).
(B) For contrast, the SIDD profile of a randomly chosen but representative genomic location of equal length (Chromosome 6, bp 21000–22834).
Graphs of all 39 ARS elements are presented in Protocol S1.
To quantify this propensity, we determined the minimum value Gmin of G(x) occurring within each ARS element. For comparison, at each ARS element we also found Gmin in two nearby segments, each the same length as the ARS element, and located symmetrically to either side of it. Here we used comparison regions separated from the ARS element by 250 bp, but equivalent results were found when the comparison regions were chosen to directly abut the ARS elements (data not shown).
The average value of Gmin within these ARS elements was 1.51 kcal/mol, while Gmin within the comparison regions averaged 6.29 kcal/mol. It follows that ARS elements are much more susceptible to SIDD than are neighboring regions. The distributions of Gmin values within the ARS elements and in their comparison regions are shown in Figure 4.
Figure 4 Duplex Destabilization at ARS Elements Compared with Duplex Destabilization at the Surrounding Sequences
This histogram shows the distributions of ARS elements (red), and of comparison regions (black) whose Gmin values fall in the indicated ranges. (Here, as elsewhere, the lower the Gmin value, the more destabilized the region.) The comparison regions were chosen to have the same lengths as the ARS element they flank, and to be positioned 250 bp away from it on either side. There being twice as many comparison regions as ARS elements, these distributions are normalized to show the fraction of sites of each type falling within each interval. Equivalent results were obtained when the comparison regions were chosen to directly abut the ARS elements, so the localization of destabilization at ARS elements is not simply a consequence of their positions within intergenic regions. (ARS elements 302, 303, and 320 on Chromosome III were positioned very close [20 bp separate ARS302 from ARS303, and ARS320 directly abuts ARS303], so for the purpose of these statistical tests these three were regarded as a single site.)
We next compared the distribution of Gmin values within the ARS elements to the genome-wide SIDD distribution shown in Figure 2. Just 2.79% of the base pairs in this genome were destabilized at the level G(x) less than 1.51 kcal/mol, the average Gmin for the ARS elements. This clearly shows that sites that are superhelically destabilized to the extent found at ARS elements are not common.
The Statistical Significance of this Association
We performed a Wilcoxon–Mann–Whitney rank sum test [17] to rigorously assess the statistical significance of this observed difference in destabilization between ARS elements and their comparison regions. The results show that the null hypothesis (that these distributions are the same) must be rejected with very high confidence—the p-value calculated by this test was p = 4.28 × 10−10. A Kolmogorov–Smirnov test [18] of the same distributions, performed for the same purpose, yielded p = 2.91 × 10−10. Together, these two nonparametric tests show that the greater destabilization within ARS elements relative to their flanking regions is statistically highly significant.
Discussion
We have shown that a strong susceptibility to destabilization under stress is a statistically significant attribute of ARS elements, as evaluated in their genomic contexts within the S. cerevisiae genome. The fact that 38 of the 39 analyzed ARS elements are significantly destabilized, and on average are much more destabilized than their neighborhoods, makes this one of the most highly conserved attributes known to occur at ARS elements.
Although the SIDD results reported here are consistent with earlier studies of duplex unwinding elements (DUEs) in ARS elements [4], we note three significant differences. First, unlike the attribute of helical stability used to characterize DUEs, SIDD properties are acutely dependent not just on the ARS element sequence itself, but also on its larger context. So altering nearby sequences can drastically change the SIDD properties of a region. This effect is consistent with the observation that origin activity varies depending on chromosomal location, suggesting the influence of local chromatin structure [1]. Second, in addition to finding unwinding regions, SIDD calculations also identify locations where imposed superhelicity diminishes the energy needed to separate the DNA into single strands. This has an exponential effect on the ease with which other molecules can induce strand separation there (see Materials and Methods). Third, unlike DUEs, the destabilization at ARS element locations is not confined to discrete or specific positions within the element.
The presence of stress-destabilized sites at ARS elements has clear implications for the mechanisms of initiation of DNA replication. Under certain circumstances, the presence of a SIDD site alone has been shown to confer a degree of origin activity on an otherwise inactive region [7]. Other more complex roles also are possible. Observations of SIDD near promoters have shown that protein binding can exert regulatory effects by translocating destabilization from the binding site to other locations [11]. Similar events could occur during origin function. The reported dual roles for B2 elements within the ARS element, as being involved either in duplex unwinding [4] or protein binding [19], could be reconciled if protein binding to a destabilized B2 element were to cause a similar regulatory translocation. If the destabilization were to move to the position where unwinding is required for initiation, this could be the mechanism by which binding activates initiation.
To experimentally investigate the details of the role that SIDD may play in the regulation of specific replication origins, the destabilization properties of a region can be altered without changing its base sequence. This involves inserting at another location a DNA sequence that is also susceptible to some type of superhelical transition [9,20]. Since stresses couple together the transition behaviors of all base pairs that experience them, introducing a new competitive region will change the SIDD propensity of the site of interest. This strategy has been used previously to prove that SIDD is involved in the activation of the ilvPG promoter of E. coli [20].
The complete S. cerevisiae genome has been estimated to contain between 200 and 400 ARS elements [1]. The regions of Chromosomes III, VI, and XIV that have been systematically examined for ARS element sites together constitute 6% of the genome and contain 31 ARS elements [21]. If this density is representative, it would give a slightly higher estimate of approximately 500 ARS elements in this genome. Whichever number is used, it is clear that only a small fraction of the ARS elements in yeast have been located to date.
The statistically highly significant association of SIDD properties with ARS elements reported here suggests that these properties may be useful for finding the precise locations of ARS elements within regions of the yeast genome that are suspected to contain them. Two recent studies identified several such regions on a genome-wide scale. The first study identified DNA segments that showed binding activity for ORC and MCM proteins [22], while the second measured the time of replication across complete chromosomes using density transfer and microarray hybridization [23]. The regions identified by these approaches are roughly 1 kb and 10–20 kb in size, respectively—too large to unambiguously locate ARS elements within them. Since SIDD properties can be calculated with single base pair resolution, predictions of the susceptibility to superhelical destabilization could be used in conjunction with these results to identify potential replication origins throughout the yeast genome. Two illustrative examples are shown in Figure 5. Alternatively, SIDD properties could be used in conjunction with other computational methods (e.g., sequence-based algorithms [24]) of origin prediction to locate potential origins with greater confidence and accuracy.
Figure 5 Localization of ARS Elements
(A) Replication timing profile of Chromosome 3. The two peaks indicated by red stars are predicted with high confidence to contain replication origins. (Data replotted from [23].)
(B and C) SIDD profiles of the two peak regions (from [A]) are plotted to high resolution, along with locations of the known ARS element (red) and the DNA segments within which ORC and MCM proteins were shown to bind [22] (yellow). (B) shows the profile around ARS 310, and (C) shows that of ARS 314.
We have shown that strong susceptibility to destabilization under stress is a highly conserved attribute of ARS elements in S. cerevisiae. Our ongoing research suggests that an enhancement of SIDD propensities might also correlate with replication origin locations in higher eukaryotes.
Materials and Methods
We analyzed the SIDD properties of the complete genome of the S288C strain of S. cerevisiae [25]. We used the method described previously whereby the DNA sequence of each chromosome is partitioned into overlapping windows and each window is analyzed separately [5]. Each window (except perhaps the last) has length N = 5,000 bp, with successive windows offset by 500 bp so each internal base pair appears in ten windows. The final values of the probability p(x) and the destabilization energy G(x) for the base pair at position x are calculated as the weighted averages of their computed values in each of the windows that contain that base pair. A detailed description of this algorithm has been presented elsewhere [5].
In these calculations all conformational and free energy parameters are given their experimentally measured values, so there are no free parameters [8,9]. Here we use values appropriate to a temperature of 37 °C and a [Na+] of 0.01 M, the conditions of the Kowalski nuclease digestion procedure by which superhelical denaturation is most accurately evaluated [26]. We use superhelix density σ = −0.055, a moderate physiological value [27]. These calculations robustly predict the locations where destabilization occurs, although the details of the transition profiles vary somewhat with assumed conditions. In particular, elevated temperature and increased negative superhelicity act synergistically; higher stress is required to achieve a given level of destabilization at lower temperatures, other factors remaining fixed.
This analysis of the complete yeast genome required approximately 12 hours to execute on a 28-node Apple X-Serve cluster, each node containing dual 1 GHz G4 processors. The profile of the entire genome is available on request.
To understand the significance of the destabilization energy, consider a system that can assume multiple states, each with an energy G. Suppose two specific states, which we call 1 and 2, have energies G1 and G2, respectively. At equilibrium the ratio of the number of molecules in each state will vary exponentially with the difference in their energies according to f1/f2 = exp [−(G1 −
G2)/RT], where RT = 0.616 kcal/mol at a temperature T of 37 °C. It follows from this equation that lower energy states are exponentially more highly populated than are higher energy states at equilibrium.
Now, suppose this equilibrium involves the opening of a specific region of DNA by a reversible reaction with another molecule. Let the free energy required for opening this region be G1 in a supercoiled molecule, and G2 in a relaxed molecule, the difference being the destabilization caused by the superhelicity. If this difference is 4.78 kcal/mol (which is the average difference between the ARS elements and their comparison regions), this will favor the open state by f1/f2 = 2,334, so at equilibrium opening will occur more than 2,000 times as often when this region is superhelically destabilized than when it is not, other factors remaining fixed. We note that this amount of destabilization would bring our G(x) from 10 kcal/mol just down to 5.2 kcal/mol, which is less than what would be needed to open the region completely. If strand separation at this site is the rate-limiting step in the initiation of a process, this amount of stress-induced destabilization can have a profound effect on the frequency of initiation.
Supporting Information
Protocol S1 Database of Known ARS Sites and SIDD Profiles of All 39 ARS Elements
(2.3 MB PDF).
Click here for additional data file.
Table S1 Table of Experimentally Well-Characterized ARS Elements and Associated Free Energies of Destabilization
(117 KB DOC).
Click here for additional data file.
The work reported here was supported in part by grants R01-GM68903 and R01-HG01973 from the National Institutes of Health, and grant DBI-0416764 from the National Science Foundation. We thank Oscar Aparicio, Stephen Bell, Carol Newlon, M. K. Raghuraman, Bruce Stillman, and James Theis for helpful discussions, and Miraslava Kaloper for technical assistance.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. PA conceived and designed the experiments, performed the experiments, and analyzed the data. PA and CJB contributed reagents/materials/analysis tools and wrote the paper.
Abbreviations
ARS elementautonomously replicating sequence
DUEduplex unwinding element
SIDDsuperhelically induced duplex destabilization
==== Refs
References
Newlon CS 1996 DNA replication in yeast DePamphilis ML DNA replication in eukaryotic cells Cold Spring Harbor (New York) Cold Spring Harbor Laboratory Press 873 914
Marahrens Y Stillman B 1992 A yeast chromosomal origin of DNA replication defined by multiple functional elements Science 255 817 23 1536007
Huang RY Kowalski D 1993 A DNA unwinding element and an ARS consensus comprise a replication origin within a yeast chromosome EMBO J 12 4521 4531 8223462
Natale D Umek R Kowalski D 1993 Ease of DNA unwinding is a conserved property of yeast replication origins Nucleic Acids Res 21 555 560 8441667
Benham CJ Bi C 2004 The analysis of stress-induced duplex destabilization in long genomic DNA sequences J Comp Biol 11 519 543
Benham CJ 1992 Energetics of the strand separation transition in superhelical DNA J Mol Biol 225 835 847 1602485
Potaman VN Bissler JJ Hashem VI Oussatcheva EA Lu L 2003 Unpaired structures in SCA10 (ATTCT)n.(AGAAT)n repeats J Mol Biol 326 1095 1111 12589756
Benham CJ 1993 Sites of predicted stress-induced DNA duplex destabilization occur preferentially at regulatory loci Proc Natl Acad Sci U S A 90 2999 3003 8385354
Benham CJ 1996 Duplex destabilization in superhelical DNA is predicted to occur at specific transcriptional regulatory regions J Mol Biol 255 425 434 8568887
Fye RM Benham CJ 1999 Exact method for numerically analyzing a model of local denaturation in superhelically stressed DNA Phys Rev E Stat Phys Plasmas Fluids Relat Interdiscip Topics 59 3408 3426
Sheridan SD Benham CJ Hatfield GW 1998 Activation of gene expression by a novel DNA structural transmission mechanism that requires supercoiling-induced DNA duplex destabilization in an upstream activating sequence J Biol Chem 273 21298 21308 9694890
Huang Y Kowalski D 2003 WEB-THERMODYN: Sequence analysis software for profiling DNA helical stability Nucleic Acids Res 31 3819 3821 12824427
Huang Y Kowalski D 2003 WEB-THERMODYN [Web-based computer program] Buffalo (New York) Roswell Park Cancer Institute Department of Cancer Genetics Available: http://wings.buffalo.edu/gsa/dna/dk/WEBTHERMODYN/ . Accessed 19 May 2005.
Lerman LS Silverstein K 1987 Computational simulation of DNA melting and its application to denaturing gradient gel electrophoresis Methods Enzymol 155 482 501 2828875
Kowalski D Eddy MJ 1989 The DNA unwinding element: A novel, cis -acting component that facilitates opening of the Escherichia coli replication origin EMBO J 8 4335 4344 2556269
Polonskaya Z Benham CJ Hearing J 2004 Role for a region of helically unstable DNA within the Epstein-Barr virus latent cycle origin of DNA replication in origin function Virology 328 282 291 15464848
DeGroot MH 1975 Probability and statistics Reading (Massachusetts) Addison-Wesley 607 p.
Chakravarti IM Laha RG Roy J 1967 Handbook of methods of applied statistics, Volume 1 New York John Wiley and Sons
Wilmes GM Bell SP 2002 The B2 element of the Saccharomyces cerevisiae ARS1 origin of replication requires specific sequences to facilitate pre-RC formation Proc Natl Acad Sci U S A 99 101 106 11756674
Sheridan S Benham CJ Hatfield GW 1999 Inhibition of supercoiling-dependent transcriptional activation by a distant B-DNA to Z-DNA transition J Biol Chem 274 8169 8174 10075720
Newlon CS Theis JF 2002 DNA replication joins the revolution: Whole-genome views of DNA replication in budding yeast Bioessays 24 300 304 11948615
Wyrick JJ Aparicio JG Chen T Barnett JD Jennings EG 2001 Genome-wide distribution of ORC and MCM proteins in S. cerevisiae: High-resolution mapping of replication origins Science 294 2357 2360 11743203
Raghuraman MK Winzeler EA Collingwood D Hunt S Wodicka L 2001 Replication dynamics of the yeast genome Science 294 115 121 11588253
Breier AM Chatterji S Cozzarelli NR 2004 Prediction of Saccharomyces cerevisiae replication origins Genome Biol 5 R22 15059255
Stanford University School of Medicine Department of Genetics 2004 Saccharomyces Genome Database [database] Available: http://www.yeastgenome.org/ . Accessed 19 May 2005.
Kowalski D Natale DA Eddy MJ 1988 Stable DNA unwinding, not “breathing,” accounts for single-strand-specific nuclease hypersensitivity of specific A+T-rich sequences Proc Natl Acad Sci U S A 85 9464 9468 2849106
Kouzine F Liu J Sanford S Chung HJ Levens D 2004 The dynamic response of upstream DNA to transcription-generated torsional stress Nat Struct Mol Biol 11 1092 1100 15502847
Zaret KS Sherman F 1982 DNA sequence required for efficient transcription termination in yeast Cell 28 563 573 6280875
|
16103908
|
PMC1183513
|
CC BY
|
2021-01-05 09:18:21
|
no
|
PLoS Comput Biol. 2005 Jun 24; 1(1):e7
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010007
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science 1610390910.1371/journal.pcbi.001000805-PLCB-RA-0024R2plcb-01-01-10Research ArticleBioinformatics - Computational BiologyIn VitroCombinatorial Pattern Discovery Approach for the Folding Trajectory Analysis of a β-Hairpin Pattern Discovery for Protein FoldingParida Laxmi 1Zhou Ruhong 121 Computational Biology Center, IBM Thomas J. Watson Research Center, Yorktown Heights, New York, United States of America
2 Department of Chemistry, Columbia University, New York, New York, United States of America
Lai Luhua EditorPeking University, ChinaE-mail: [email protected] (LP); [email protected] (RZ)6 2005 24 6 2005 1 1 e84 2 2005 18 5 2005 Copyright: © 2005 Parida and Zhou.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.The study of protein folding mechanisms continues to be one of the most challenging problems in computational biology. Currently, the protein folding mechanism is often characterized by calculating the free energy landscape versus various reaction coordinates, such as the fraction of native contacts, the radius of gyration, RMSD from the native structure, and so on. In this paper, we present a combinatorial pattern discovery approach toward understanding the global state changes during the folding process. This is a first step toward an unsupervised (and perhaps eventually automated) approach toward identification of global states. The approach is based on computing biclusters (or patterned clusters)—each cluster is a combination of various reaction coordinates, and its signature pattern facilitates the computation of the Z-score for the cluster. For this discovery process, we present an algorithm of time complexity c∈RO((N + nm) log n), where N is the size of the output patterns and (n × m) is the size of the input with n time frames and m reaction coordinates. To date, this is the best time complexity for this problem. We next apply this to a β-hairpin folding trajectory and demonstrate that this approach extracts crucial information about protein folding intermediate states and mechanism. We make three observations about the approach: (1) The method recovers states previously obtained by visually analyzing free energy surfaces. (2) It also succeeds in extracting meaningful patterns and structures that had been overlooked in previous works, which provides a better understanding of the folding mechanism of the β-hairpin. These new patterns also interconnect various states in existing free energy surfaces versus different reaction coordinates. (3) The approach does not require calculating the free energy values, yet it offers an analysis comparable to, and sometimes better than, the methods that use free energy landscapes, thus validating the choice of reaction coordinates. (An abstract version of this work was presented at the 2005 Asia Pacific Bioinformatics Conference [1].)
Synopsis
The study of protein folding mechanisms continues to be one of the most challenging problems in computational biology. Currently, the protein folding mechanism is often characterized by calculating the free energy landscape versus various reaction coordinates, such as the fraction of native contacts, the radius of gyration, RMSD from the native structure, and so on. In this paper, the authors present a combinatorial pattern discovery approach toward understanding the global state changes during the folding process. This is a first step toward an unsupervised (and perhaps eventually automated) approach toward identification of global states. The authors apply this approach to a β-hairpin folding trajectory and demonstrate that this approach extracts crucial information about protein folding intermediate states and mechanism.
Citation:Parida L, Zhou R (2005) Combinatorial pattern discovery approach for the folding trajectory analysis of a β-hairpin. PLoS Comput Biol 1(1): e8.
==== Body
Introduction
Understanding protein folding is one of the most challenging problems in molecular biology [2–7]. The interest is not only in obtaining the final fold (generally referred to as structure prediction) [8–10] but also in understanding the folding mechanism and folding kinetics involved in the actual folding process. Many native proteins fold into unique globular structures on a very short time scale. The so-called fast folders can fold into the functional structure from random coil in microseconds to milliseconds. Recent advances in experimental techniques that probe proteins at different stages during the folding process have shed light on the nature of the folding kinetics and thermodynamics [11–17]. However, due to experimental limitations, detailed protein folding pathways remain unknown. Computer simulations performed at various levels of complexity, ranging from simple lattice models to all-atom models with explicit solvent, can be used to supplement experiment and fill in some of the gaps in our knowledge about folding mechanisms.
Large-scale simulations about protein folding with realistic all-atom models still remain a great challenge [3–5,7]. Enormous effort is needed for this grand problem; one example is the recent IBM Blue Gene project, which is aimed at building a supercomputer with hundreds-of-teraflop to petaflop computing power to tackle the protein folding problem. Meanwhile, effective analyses of the trajectory data from the protein folding simulations, either by molecular dynamics or Monte Carlo, remains yet another challenge due to the large number of degrees of freedom and the huge amount of trajectory data. [18,19] Currently, the protein folding mechanism is often characterized by calculating the free energy landscape versus the so-called reaction coordinates [3,20,21]. We and others have used various reaction coordinates [3,20,21], such as the fraction of native contacts, the radius of gyration of the entire protein, the root mean square deviation (RMSD) from the native structure, the number of β-strand hydrogen bonds, the number of α-helix turns, the hydrophobic core radius of gyration, and the principal components (PC) from principal component analysis [20,22]. Searching for better reaction coordinates is still of great interest in protein folding mechanism studies. These analyses have provided important information for a better understanding of protein folding. However, it often requires a priori knowledge about the system under study, and the free energy contour maps usually result in too much information reduction due to their limit in dimensionality, which is often as low as two or three. Thus, better or complementary analysis tools are in great demand.
It is also known that the folding process of many proteins takes the amino acid coil through different intermediate states before stabilizing on the final folded state. Therefore, a first step toward understanding the folding process is to identify these states. In this paper, we propose the use of a combinatorial pattern discovery technique to analyze protein folding trajectory data from simulation experiments. A novel aspect of the current algorithm is that it incorporates arbitrary and possibly different distribution functions of the data in each dimension and guarantees complete and accurate solution to the clustering problem. The procedure involves computations of clusters of the data: each cluster has a signature pattern describing all the elements of the cluster. The simplicity of the pattern leads to easy interpretation of and thus better understanding of the underlying processes and facilitates the computation of a Z-score for the cluster. By appropriate redundancy checks, the number of clusters is made manageably small. The results of this method are threefold. Firstly, the method is validated by comparing its results with previously published results with a free energy landscape analysis. Secondly, the method succeeds in extracting meaningful new patterns and structures that had been overlooked before. These new structures provide a better understanding of the folding mechanism of a β-hairpin, which is used as a case study in this paper. These new patterns also interconnect various states in existing free energy contour maps versus different reaction coordinates. This success encourages us to postulate that the automatic discovery will lead to much greater understanding of the folding process. Thirdly, the method validates the choice of reaction coordinates since the pattern discovery analysis based on these reaction coordinates compares well with the previous free energy based approaches.
Results/Discussion
Description of Models
Well-known simulation methods exist to carry out the folding of a protein. However, it is often not sufficient to obtain a succinct understanding of the folding process. The task here is to understand the folding mechanism by recognizing structural patterns or intermediate states that the folding process goes through. For example, the folding of a small protein, a β-hairpin, could be understood at a global level in terms of the states shown in Figure 1. Although we would aim to understand the folding of every protein in this simplistic form, the current state of the art is far from this goal.
Figure 1 A Hypothetical State Diagram of a Folding of Protein
A schema of the folding process for a small protein, a β-hairpin. It starts with an unfolded state, U state, undergoes to a hydrophobic core collapsed H state, and then to a partially folded P state before finally ending at the folded F state.
At each step of the simulation process, a configuration of the solvated protein can be computed. However, the simulation may be carried for nanoseconds to microseconds in units of femtoseconds (10−15), so the number of such intermediate configurations could easily be millions in number. Hence, the task is to identify and capture representative intermediate configurations. Since working in the structure space of the protein is extremely complex, researchers often identify a few key characteristic features of the protein, or often so-called reaction coordinates, and study the trends and variations in these reaction coordinates [21,23].
In this paper, we utilize a four-step process toward understanding the folding of a protein (Figure 2). The first step involves the in silico simulation that gives rise to a large collection of data points, each point being an array of the characteristic features of the folding protein at that time point. For example, the radius of gyration or the number of hydrogen bonds could be such features. In the Results/Discussion section, we study the β-hairpin folding as a show case and describe seven such characteristic features that we have used previously in the study of this particular protein.
Figure 2 The Flowchart of the Process of Understanding a Folding Protein
Step 1 starts with millions of data points obtained from the simulation experiments. Step 2 extracts the recurring patterns, reducing the size of the data to be studied down to thousands. Step 3 further reduces down this to a representative set of a handful states, which are studied in detail in Step 4. The structures are extracted, and a possible state diagram summarizing the path of the folding protein is elucidated.
In the second step, we study these data points to extract the characteristic set of features that we call pattern clusters. Again, in the case of the β-hairpin, the data points are seven-dimensional, corresponding to the characteristic features of the protein at each time interval (see Table 1 for a small portion of the data as an example). In the third step, these patterns are filtered to retain the most significant ones. It is very difficult to model the significant patterns in this domain, so we have combined the second and third steps and use appropriate parameters to filter out possibly insignificant patterns: we use cluster size (in terms of rows) and the Z-scores.
Table 1 A Small Portion of the Raw Data from the REMD Sampling of the β-Hairpin Folding in Explicit Water
The fourth step is to analyze the patterns. This involves extracting the structure of the configuration using the time coordinates and studying the correlation of the different structures. For instance, one could observe that the hydrophobic core is formed before the β-strand hydrogen bonds, or vice versa; and one can interconnect various free energy states in different free energy surfaces by monitoring the high-dimensional (multi-column) patterns. These findings can provide a better understanding of the protein folding mechanism. Further, the time correlation between various patterns or states could be studied. For example, it is extremely useful to know which pattern or state precedes the other and by how much time.
Here, we describe in detail the second and third steps in our approach, as shown in Figure 2. We model the extraction problem as a combinatorial detection problem for at least three specific reasons: (1) The data are obtained from a replica exchange molecular dynamics (REMD) method [24] (more details below). This method is essentially a Monte Carlo method; thus, the time series is not strictly real time due to the random Monte Carlo exchange process. Also, our interest is in finding pattern clusters that are not necessarily correlated in time. (2) This emphasizes that any probabilistic (or non-deterministic) component can be isolated from the algorithm and the problem. Any high-frequency noise can be largely resolved through an introduction of a δ function (see below). (3) The signature pattern of the cluster helps interpret the clusters quite easily. Also, in comparison to the straightforward grouping or clustering algorithms in previous publications [21,25], this provides a complete and efficient (in linear time) method to find the signature patterns. It must be pointed out that this is the critical reason why we chose to use this method, since this enables us to have a tighter control on an acceptable cluster that is also meaningful in terms of the folding process.
A small but important protein system has been selected as an example to demonstrate our approach to understanding the folding process. This small protein is a 16-residue β-hairpin (GEWTYDDATKTFTVTE) from the C-terminus of protein G (residues 41–56 of PDB file 2gb1.pdb). Its folding mechanism and folding free energy states have been studied extensively in previous works [21,23]. The current study will use our new approach to analyzing the existing trajectories from the previous REMD simulations in explicit solvent [21,24]. The REMD method couples molecular dynamics trajectories with a temperature-exchange Monte Carlo process for efficient sampling of the conformational space. In this method, replicas are run in parallel at a sequence of temperatures ranging from the desired temperature to a high temperature at which the replica can easily surmount the energy barriers. From time to time, the configurations of neighboring replicas are exchanged and this exchange is accepted by a Metropolis acceptance criterion that guarantees the detailed balance. Because the high-temperature replica can traverse high-energy barriers, this provides a mechanism for the low-temperature replicas to overcome the quasi-ergodicity they would otherwise encounter in a single-temperature replica.
This β-hairpin has received much attention recently from both experimental and theoretical fronts [11,13,14,18,20,26–30]. The β-sheets and α-helices are the key secondary structures in proteins. It is believed that understanding the folding of these elements will be a foundation for investigating larger and more complex structures. The study of isolated β-sheets has for a long time been limited by the lack of an amenable experimental system. The breakthrough experiments by Serrano [11] and Eaton [13] groups have recently established this β-hairpin as the system of choice to study β-sheets in isolation. These pioneering experiments inspired a number of theoretical works on this system with various models [18,20,21,26,27,31,32]. However, there are still a number of important aspects that remain controversial, such as the relative importance and time-sequential order between the β-strand hydrogen bonds formation and the hydrophobic core formation, and the existence of α-helical intermediates during the folding.
Simulation Parameters
In this study, an all-atom model—The Optimized Potential for Liquid Simulations-All-Atom force field [33] with an explicit solvent model, Simple Point Charge model [34]—is used for the description of the protein solvated in water. A total of 64 replicas of the solvated system consisting of 4,342 atoms is simulated with temperatures spanning from 270 K to 695 K. For each replica, a 3-nanosecond molecular dynamic simulation is run with replica exchanges attempted every 400 femtoseconds. The reader is directed to [21,23] for details of this simulation. For each conformation, seven different reaction coordinates are used (Table 1). There are a total of about 20,000 conformations saved for each replica. Table 1 lists a small portion of the data for the replica at 310 K, which is the biological temperature.
These simulations have revealed a hydrophobic-core-driven folding mechanism from free energy contour map analysis [21]. Since this is a well-studied system and a large amount of data is available, comparisons with other analysis tools, such as the free energy contour map analysis, might be easier and more straightforward. Various reaction coordinates obtained from previous runs serve as the starting point.
Discovery Parameters
Although we developed the framework for a very general δ function, for simplicity, in this section we treat δ(x) to be a constant function. Thus, δ(x) = c for some constant
for each x. The δ functions for each column of Table 1 is given as follows: δ
1(x) = 0.2, δ
2(x) = 0.6, δ
3(x) = 0.35, δ
4(x) = 0.15, δ
5(x) = 5.0, δ
6(x) = 16.5, δ
7(x) = 1.0 for all x. Further, the quorum k is defined to be 2,000. Table 2 lists some representative patterns of size two with these parameters. The time sequences are not shown due to the space constraints. These simple patterns can be directly compared with the previous free energy states in the three-dimensional free energy contour maps. These are three-dimensional plots of free energy versus a pair of reaction coordinates or data columns of Table 1.
Table 2 Simple Patterns of Size Two
These patterns can be easily compared to the three-dimensional free energy landscapes using a pair of corresponding reaction coordinates.
One might often want to study detailed patterns or structures in some predefined subregions such as the structures in the unfolded ensemble. More evidence has shown that the protein structures in unfolded states are not fully extended but often have well-defined structures instead [35]. This can also avoid the problem that important patterns in these less populated areas are being overlooked due to a smaller population than the predefined quorum k. Thus, some less populated free energy states in free energy landscapes can be recovered by reducing the quorum. Hence, another set of parameters have been used, and here we confine our search to data points with
and
5.0 Å (see Table 1 for definitions of these reaction coordinates) with k = 100. Yet another set of parameters have included
and
9.0 Å with k = 50. A subset of the results is shown later. Thus, this approach might be useful for hierarchical pattern searches that gradually zoom into the predefined subsets of data.
Analysis of Results
To obtain a representative structure(s) from a set of configurations ci, the set is partitioned into a minimum number of groups Gj such that for each Gj there exists a representative
, and for each
the structure corresponding to ck is at most 1 Å RMSD from
. Thus, each Gj will be represented by a structure corresponding to
[21,26].
Recovering known free energy states.
Obviously, the first question of importance is: Can we recover the previously found free energy states in the new approach? The “time sequence” of each pattern is then used to extract the corresponding conformations of the protein. Figure 3A shows a representative or most populated structure for the first pattern (
= 4.886 ± 0.2,
= 5.448 ± 0.6 ) in Table 2. This structure mimics the representative structure from the folded state (F state) in the free energy contour map versus
and
very well. Thus this pattern resembles the F state of the free energy contour map. Similarly, the second pattern of Table 2 (
= 2.875 ± 0.2,
= 5.448 ± 0.6) resembles the partially folded state, P state, in the same free energy landscape. The structures for the two patterns are shown in Figure 3. Thus, our approach recovers the most populated states in the free energy landscape analysis.
Figure 3 Representative Structures for Two Patterns
Hydrophobic residues TRP43, TYR45, PHE52, and VAL54 are represented by spacefill, and the rest of the residues are represented by ribbons. (A) Pattern 1 in Table 2 captures the folded state (F state) in free energy contour map analysis [21]. (B) Pattern 2 in Table 2 captures the partially folded state (P state) in the same free energy contour map.
The third and fourth patterns in Table 2 also resemble the F state and P state, respectively, in the same free energy contour map versus
and
. Numerous other patterns have shown similar results, i.e., recovering various previously found free energy states in the free energy contour maps versus different reaction coordinates. It should be noted, though, that many patterns might be redundant, either because the δ() function values given for reaction coordinates are too wide, or because some of the reaction coordinates are highly correlated. For example, the fifth pattern of Table 2 is
= 4.979 ± 0.6, Rg = 8.144 ± 0.35. Clearly, these two reaction coordinates are highly correlated, since
measures the radius of gyration of four key residues out of the total 16 that are measured by Rg. However, for many other cases, it may not be so obvious.
Interconnecting various free energy landscapes.
More complicated patterns with many reaction coordinates are also found in the current approach, which had been previously undetected. In the traditional free energy landscape analysis, typically one or two reaction coordinates are used at each time, since a two- or three-dimensional free energy contour map is usually plotted. It is extremely difficult to visualize high-dimensional free energy landscapes in order to identify the free energy basins or barriers. Table 3 lists some of these complicated patterns with up to six reaction coordinates. Of course, as pointed out earlier, some reaction coordinates might be correlated, so the data in each reaction coordinate may not be totally independent. Nevertheless, it still reveals some interesting new findings. First of all, these patterns can interconnect various free energy states in different free energy landscapes. This might not be so obvious in free energy surfaces themselves. For example, the sixth pattern in Table 3, (Rg = 8.144 ± 0.35, ρ = 0.815 ± 0.15, PC-1 = −5.881 ± 5.0, PC-2 = −33.574 ± 16.5, RMSD = 3.292 ± 1.0), interconnects the following two free energy surfaces, one versus PC-1and PC-2 (Figure 4A), and the other versus ρ and Rg (Figure 4B). The states corresponding to the free energy well (of value ≈ −8 KT) near PC-1 = −5.9, PC-2 = −33.6 in Figure 4A and ρ = 0.82, Rg = 8.1 in Figure 4B are the same free energy state since they consist of the same clusters in the same pattern. In this particular case, obviously they all represent the folded state (F state).
Table 3 Complex Patterns of Size up to Six
Figure 4 Free Energy Landscapes
Free energy landscapes versus (A) the principal components PC-1 and PC-2, and (B) the fraction of native contact ρ and the radius gyration of the peptide Rg at 310 K. The interconnected free energy wells described by the pattern are near −8KT at PC-1 = −5.9, PC-2 = −33.6 in (A) and at ρ = 0.82, Rg = 8.1 in (B) (see text for more details).
Understanding folding mechanism better.
More importantly, the new approach reveals important structures overlooked previously, which might help understand the folding mechanism better. Eaton and coworkers [13,14] proposed a “hydrogen bond zipping” mechanism for this β-hairpin, in which folding initiates at the turn and propagates toward the tails by making β-strand hydrogen bonds one by one, so that the hydrophobic core, from which most of the stabilization derives, forms relatively late during the folding. In our previous study, we proposed a different folding mechanism, in which this β-hairpin undergoes a hydrophobic core collapse first, then makes native β-strand hydrogen bonds to make over the free energy loss due to the loss of H-bonds between the backbone atoms and water. Figure 5A shows a representative structure for the eighth pattern in Table 3, (
= 4.950 ± 0.2, Rg = 8.013 ± 0.35, ρ = 0.848 ± 0.15, PC-1 = −5.881 ± 5.0, PC-2 = −33.574 ± 16.5, RMSD = 3.292 ± 1.0). The structure shows that all five native β-strand H-bonds have been formed, but the hydrophobic core is not completely aligned yet. The loop region also bends toward the hydrophobic core to somewhat offset the non-perfect hydrophobic core. These structures with H-bonds that are formed but with their hydrophobic core not perfectly aligned (RMSDs up to 4 Å) imply that the hairpin can also have a path to form β-strand hydrogen bonds before the core is finalized. The current findings indicate that the final hydrophobic core and β-strand hydrogen bonds might be formed almost simultaneously. This can also be seen from the low free energy barrier in free energy landscapes as discussed before [21]. Interestingly, Thirumalai et al. also found that the lag time between collapse and hydrogen bond formation is very short and the two processes occur nearly simultaneously [32]. It should be pointed out that the turn (loop) formation is critical in this β-hairpin folding mechanism, since the hydrophobic core and β-strand hydrogen bonds need to be packed or formed at right positions. Interestingly, this is also reported by other groups [15–17]. For example, Gai and coworkers studied a related β-hairpin, Trp-zipper hairpin, and found that the rate-limiting event corresponds to the turn formation [15,16]. Moreover, the authors pointed out that a stronger turn-promoting sequence increases the stability of the hairpin primarily by increasing its folding rate, whereas a stronger hydrophobic cluster increases the stability primarily by decreasing its unfolding rate [15,16].
Figure 5 Representative Patterns and Structures
(A) Pattern 6 of Table 3, which represents a new class of structures previously overlooked in free energy landscape analysis.
(B) Pattern 1 of Table 4, which captures the H state (hydrophobic core formed but no β-strand H-bonds) in free energy contour map analysis [21].
(C) Pattern 2 in Table 2 captures the unfolded state (U state) in the same free energy contour map. The hydrophobic residues TRP43, TYR45, PHE52, and VAL54 are represented by spacefill, and the rest are represented by ribbons.
Finally, the patterns of subsets of data in less populated states, such as the unfolded state, are studied in detail by zooming into these regions with a smaller quorum k and a different set of δ(). As mentioned earlier, more evidence has shown that the protein structures in unfolded states are not fully extended, but often have well-defined structures instead [35]. The first pattern in Table 4 (
= 0.0,
= 5.448 ± 0.5) resembles the previous H-state in free energy contour map versus
and
, where the hydrophobic core is largely formed but no native β-strand H-bonds have been made yet. Figure 5B shows a representative structure of this pattern, which mimics the structures from previous H-state very well. Figure 5C shows a representative structure for the sixth pattern in Table 4, (
= 0.0,
= 9.951 ± 0.35, ρ = 0.050 ± 0.15, PC-1 = −21.188 ± 15.0, PC-2 = 36.517 ± 15.0, RMSD = 9.872 ± 0.8). This is the most populated structure of this β-hairpin in unfolded state. Even though not many structural features are found in this structure, it is certainly not fully extended either. Since this is a very small protein with only one secondary structure in the native state, not much has been identified in the unfolded state; for larger and more complicated protein systems, such as lysozymes, more structural features might be expected in the unfolded state as found by recent experiments [35].
Table 4 Clusters with (1) J
1 = 0.0, J
2 ≥ 5.0, k = 50 and (2) J
1 = 0.0, J
2 ≥ 10.0, k = 100
To avoid clutter, the J
1 values are not shown.
Conclusion
In this paper, we have presented a method to enhance our understanding of protein folding mechanisms. At the heart of this method is a combinatorial pattern-discovery algorithm that analyzes multi-dimensional data from the simulation of the protein folding trajectory. The approach is based on pattern computation, each pattern being defined by a cluster of the reaction coordinates. A small but important protein system, a β-hairpin from the C-terminus of protein G, is then used to demonstrate this approach. It is shown that the method not only reproduces the previously found free energy states in free energy contour maps, but also reveals new information overlooked previously in free energy landscape analysis about the intermediate structures and folding mechanism. It is also shown to be useful in making interconnections between various three-dimensional free energy surfaces versus different reaction coordinates and also explains the mechanism behind the folding process. The method also validates the choice of reaction coordinates as the analysis without using free energy values compares well with the ones that use them. The success with β-hairpin is very encouraging, and we are currently exploring the application of this method to other larger protein molecules.
As stated in the Introduction section, it is important to study the time correlation between various patterns or states. For example, it is extremely useful to know which pattern or state precedes the other and by how much time. Of course, this requires real-time trajectory data. The current study uses the previous trajectories of REMD, which is a Monte Carlo method; thus, the time sequence in the data points is not real time. After this method's success with the current data, we believe that we will be able to garner time correlation of the patterns, and we are currently investigating this.
Materials and Methods
We first define the problem at hand and then give a linear time algorithm to solve the problem. The number of clusters can be easily controlled by the use of an appropriate δ() function (see below).
Combinatorial problem description.
In this section, we describe the combinatorial problem. Here, we also make some simple observations that have quite useful and practical implications (such as linear number of δ-clusters and so on). They also indicate to the extent different functions (such as the form of δ()) can be relaxed without sacrificing the general framework presented in this section. A reader may skip the statements and the proofs of these observations without any loss of continuity. Definitions 1 and 2 identify the pattern discovery or the clustering problem used in this paper, and the Results/Discussion section describes an output-sensitive algorithm to discover them.
First, we begin with a general definition of the δ-cluster and δ() function and also present the conditions under which the number of patterns are small.
Definition 1. (δ-cluster, maximal δ cluster) Given δ() : R → R
+,
, 1 ≤ i ≤ n and a quorum k. A δ-cluster is collection of i with
,
such that if
, then
. Further, Vc is maximal if there exists no Vc such that
and Vc is a δ-cluster.
Although using a general δ() function opens the possibility of various pre-processing of the data, it is important to identify a reasonable δ() function. We impose the following condition on δ(), calling it the constrained
δ
function. Given any three data elements with ν
1 <
ν
2 <
ν
3, if (v
3 − δ(v
3)) ≤ (v
1 + δ(v
1)) then (v
2 − δ(v
2)) ≥ (v
3 − δ(v
3)) and (v
2 + δ(v
2)) ≤ (v
1 + δ(v
1)).
This is a reasonable condition on an acceptable δ() function, as can be seen from the consequence of the imposed constraint in Lemma 1. A multitude of continuous functions satisfy this condition, and in the rest of the paper we will assume that δ() function we use also satisfies this condition.
Lemma 1. A δ-cluster on ν
1 <
ν
2 < …
νn
is of the form νi
<
νi
+1,…,νi
+l.
Let V be a δ-cluster with ν
min (νi) as the minimum and ν
max (νi
+l) as the maximum elements. Since ν
max and ν
min are in the δ-cluster, ν
max − δ(ν
max) ≤ ν
min + δ(ν
min). Thus, for any
, by the imposed condition, then νi − δ(νi) ≥ ν
max − δ(ν
max) and νi − δ(νi) ≤ ν
min + δ(ν
min):
Thus, the containment of the intervals is as shown; hence, for each νi , νmin < νi < ν
max,
-cluster.
Lemma 2. The number of maximal δ-clusters is no more than n where δ() is constrained.
By Lemma 1, any δ-cluster is an interval (contiguous elements on the sorted list) on the sorted list of data elements. We will show that any two intervals that correspond to two maximal δ-clusters cannot be such that one is contained in the other. Assume the contrary that one is contained in the other. Clearly, by the definition of maximality, the smaller interval is not maximal, leading to a contradiction. As no interval is contained in the other, it is possible to assign a unique element on the sorted data elements to each interval. Thus, the number of intervals cannot exceed the number of data elements, hence the result.
Corollary 1. If δ(x) = c for some
, then the number of δ-clusters is no more than n.
The bicluster takes into account the different columns or features in the data: the natural definition of such a cluster is given below.
Definition 2. (bicluster, maximal bicluster) Given δ
j() : R → R
+, quorum k and
, 1≤ j≤ m, 1≤ i ≤ n. A bicluster is collection i and j with
such that for each j,
is a δj-cluster. Further, Vc is maximal if there exists no additional i′ or j′ with the corresponding Vc with
such that Vc is a bicluster.
For ease of reference, the bicluster will be also called a pattern cluster since a cluster can be represented by the signature pattern (J
1 = c1, J2 = c2,..., JL = cL), where
, 1 ≤ k ≤ L. These J
1, J2,..., JL represent various reaction coordinates from the protein folding trajectory (shown in Table 1). This representation is more suitable for interpreting the results, as seen in other sections of this paper. The size of the bicluster is L, and k is the number occurrences or quorum of the cluster.
Lemma 3. The following are a consequence of the maximality constraint: (1) If a collection of i is such that
where Vc is a maximal δ-cluster for some j, then there exist no other maximal δ-cluster Vc ≠Vc such that
. (2) If a collection of j is such that
where Vc is a maximal δ-cluster for some j, then there exist no other maximal δ-cluster Vc ≠Vc such that
.
Lemma 4. Given
, 1 ≤ j ≤ m, 1 ≤ i ≤ n. the number of maximal biclusters is no more than n
2
m.
In a maximal bicluster Vc for some j,
is not necessarily maximal. The number of such clusters by Lemma 2 can be no more than n2. By Lemma 3, this can belong to only one maximal bicluster. Thus, there can be no more than n
2
m maximal biclusters, since there are m columns.
The linear time algorithm.
Similar descriptions of bicluster detection appear in [36], in which the authors present only an empirical time bound (linear with output size). G. Alexe and P.L. Hammer also present an incremental polynomial time algorithm with a total running time of O(Nnm
2) (personal communication). N is the number of patterns in the output, and (n × m) is the size of the input. In this section, we present an output-sensitive algorithm that computes all the maximal biclusters. The algorithm has two main steps. In the first step, the maximal δ-clusters are computed, and in the second step, the maximal biclusters are computed using the clusters of the first step.
Step 1: Maximal
δj
-cluster computation. For each j, 1 ≤ j ≤ m, compute the maximal δj -cluster,
. For simplicity, let the number of these be Lj and the clusters be
, 1 ≤ l ≤ Lj and they are computed as described below. We present a simple algorithm that does a linear scan of the sorted entries vij for each fixed j using two pointers i and l: i tracks the start of the cluster, and l tracks the end of the cluster. The end pointer is incremented until it is no longer a cluster satisfying the δ() function, and only then the start pointer is incremented. The pseudocode, Compute-Cluster(), describes the maximal δ-cluster computations, for each j. To avoid clutter, the end-of-input check is not included in the code.
Compute-Cluster()
(1) Sort the νi's to obtain ν
1, ν
2, …, νn
(2) i ← 1, l ← i + 1
(3) If
(4) Then l ← l + 1, go to Step (3)
(5) Else
,
i ← i + 1, go to Step (3)
Next, for each νij, 1 ≤ i ≤ n, 1 ≤ j ≤ m, a set of δ-clusters v′ij is computed as follows:
Step 2: Maximal bicluster computation. The algorithm in this step is based on the set intersection problem described previously [37] in the context of computing redundant motifs from irredundant ones. The algorithm works on v′ij,1 ≤ i ≤ n,1 ≤ j ≤ m, of the last step.
We describe a simple recursive algorithm to solve this problem. This algorithm implicitly constructs a tree in a depth-first manner where (1) each level corresponds to a distinct j, hence the height of the tree is m, and (2) each non-leaf node at level l corresponds to j = (m − l) (the root at level 0 corresponds to (j = m), and has at most (Lj + 1) children, the ℓth child,
, corresponds to the δ-cluster
and the very last child ([Lj + 1]th child) ignores the
δ-clusters. The algorithm is efficient due to the two following factors: (1) use of a data structure (D in the pseudocode below) to store the maximal biclusters, so that searching for an arbitrary one can be done quickly, and (2) terminating the tree traversal appropriately. The data structure suggested for use is a tree so that each query takes log n time. The terminating condition (line [2.4] of the pseudocode) is such that each leaf node corresponds to either the maximal bicluster defined by the δ-clusters (feature values)
where j
1 ≤ j
2 … ≤ jp or its variants of the form
where 1 ≤ q ≤ ρ.
The pseudocode of the recursive routine Generate-Set() shown below, describes the algorithm. Assume a function Add-set (R,C), which inserts R, a subset of integers between one and n, in a tree data structure D, along with the accompanying set C: then a query of the form if a set R exists in D takes O(log n) time. The initial call is Generate-Set ({1,2,…,n},φ,m)).
Generate-Set (R,C,j)
(1) If (j ≤ 0) then exit
(2) For
Let
Let
If
exists in D (as (R″,C″)), add
to C″
Else
Add-set (
) to D
Generate-Set (
(3) Generate-Set (R,C,j − 1)
The maximal biclusters are
, for each computed (R,C) stored in D.
Analysis of the algorithm.
We first show that the algorithm is correct in computing all the maximal biclusters and next show that the algorithm runs in time linear with the size of the output.
Correctness of the Algorithm. We first show that each computed (R,C) is a bicluster. By the construction, for each j,
is a δ-cluster. Thus (R,C) is a bicluster. Next, we have to show that it is maximal. Assume it is not. Then there exists vij such that
is a bicluster. Hence for each j,
is a δ-cluster. Then in the subroutine call Generate-Set (R,C,i′) of the pseudocode , this set must have been created, leading to a contradiction. Hence, the assumption is wrong.
Next, assume there exists
such that
is a bicluster. Hence for each j,
is a δ-cluster. Then in Step 3 of the subroutine call Generate-Set(R,C,i), Vd corresponding to j′ must have been included, leading to a contradiction. Hence, the assumption is wrong. Thus, all the computed sets are maximal biclusters. By similar arguments, it is easy to see that if there is any maximal bicluster defined on the data set, it must one of the computed R's.
Complexity of the Algorithm. Assume the input elements are νij, 1 ≤ i ≤ n, 1 ≤ j ≤ m. Consider the first step of computing the δ-clusters for each j. The sorting of the elements νi, 1 ≤ I ≤ n takes O(n log n) time. The algorithm works by scanning the input from left to right, say i to I + s, where the set {νi , νi+
1 ,…, νi
+s } is a maximal δ-cluster. Then the input is scanned from i + 1, i + s + 1, i + s + 2,… onwards and so on. Thus, each data element is visited no more than twice. Assuming the comparison can be made in constant time, this step of the algorithm takes O(n log n + n) = O(n log n) time for each j.
Next, consider the second step of computing the maximal biclusters. Notice that the search in Step (2.4) of the subroutine Generate-Set can be done in log n time. In the recursive-call tree structure (of the subroutine Generate-Set), each leaf node corresponds to a maximal bicluster. In a tree, the number of internal nodes is bounded by the number of leaf nodes and each leaf node is hit only as many times as the number of features in each pattern, thus assuming the output size is N (the total number of features in all the maximal biclusters) and the second step of the algorithm takes O(N log n) time. Thus, the time taken by the complete algorithm is O((nm + N) log n), where N is the size of the output and nm is the size of the input.
We are grateful to Jiawu Feng, who did the early implementation of the algorithm. We would also like to thank Gustavo Stolovitzky, Ajay Royyuru, Isidore Rigoutsos, Jed Pitera, and William Swope for many helpful discussions.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. RZ conceived and designed the experiments. LP and RZ performed the experiments, analyzed the data, and wrote the paper.
Abbreviations
PCprincipal component
REMDreplica exchange molecular dynamics
RMSDroot mean square deviation
==== Refs
References
Feng J Parida L Zhou R 2005 Protein folding trajectory analysis using patterned clusters [abstract]. Asia Pacific Bioinformatics Conference; 2005 17–21 Jan; Singapore
Fersht AR 1999 Structure and mechanism in protein science New York W.H. Freeman 631 p.
Brooks CL Onuchic JN Wales DJ 2001 Taking a walk on a landscape Science 293 612 11474087
Dobson CM Sali A Karplus M 1998 Protein folding: A perspective from theory and experiment Angew Chem Int Ed Engl 37 868
Brooks CL Gruebele M Onuchic JN Wolynes PG 1998 Chemical physics of protein folding Proc Natl Acad Sci U S A 95 11037 9736683
Saven JG 2003 Connecting statistical and optimized potentials for protein folding via a generalized foldability criterion J Chem Phys 118 6133 6136
Zhou R Huang X Margulius CJ Berne BJ 2004 Hydrophobic collapse in multidomain protein folding Science 305 1605 15361621
Simons KT Bonneau R Ruczinski I Baker D 1999 Structure prediction of casp iii targets using rosetta Proteins 37 171
Xu D Crawford O Locascio P Xu Y 2001 Application of prospect in casp4: Characterizing protein structures with new folds. Proteins S5: 140–148
Zhou R Silverman BD Royyuru A Athma P 2003 Spatial profiling of protein hydrophobicity: Native vs. decoy structures. Proteins52: 561
Blanco FJ Rivas G Serrano L 1994 A short linear peptide that folds in a native stable β-hairpin in aqueous solution Nat Struct Biol 1 584 7634098
Blanco FJ Serrano L 1994 Folding of protein g b1 domain studied by the conformational characterization of fragments comprising its secondary structure elements Eur J Biochem 230 634
Munoz V Thompson PA Hofrichter J Eaton WA 1997 Folding dynamics and mechanism of β-hairpin formation Nature 390 196 9367160
Munoz V Henry ER Hofrichter J Eaton WA 1998 A statistical mechanical model for β-hairpin kinetics Proc Natl Acad Sci U S A 95 5872 9600886
Du D Zhu Y Huang CY Gai F 2004 Understanding the key factors that control the rate of β-hairpin folding Proc Natl Acad Sci U S A 101 15915 15520391
Snow CD Qiu L Du D Gai F Hagen SJ 2004 Trp Zipper folding kinetics by molecular dynamics and temperature-jump spectroscopy Proc Natl Acad Sci U S A 101 4077 15020773
Dyer RB Maness SJ Peterson ES Franzen S Feisinmeyer RM 2004 The mechanism of β-hairpin formation Biochemistry 43 11560 15350142
Zagrovic B Sorin EJ Pande VS 2001 β-hairpin folding simulation in atomistic detail J Mol Biol 313 151 11601853
Yoda T Sugita Y Okamoto Y 2004 Comparisons of force fields for proteins by generalized-ensemble simulations Chem Phys Lett 386 460
Garcia AE Sanbonmatsu KY 2001 Exploring the energy landscape of a β- hairpin in explicit solvent Proteins 42 345 11151006
Zhou R Berne BJ Germain R 2001 The free energy landscape for β-hairpin folding in explicit water Proc Natl Acad Sci 98 14931 11752441
Garcia AE 2001 Large-amplitude nonlinear motions in proteins Phys Rev Lett 42 2696
Zhou R 2001 Free energy landscape of protein folding in water: Explicit vs. implicit solvent Proteins 98 148
Sugita Y Okamoto Y 1999 Replica-exchange molecular dynamics method for protein folding Chem Phys Lett 314 141
Zhou R 2003 Trp-cage: Folding free energy landscape in explicit water Proc Natl Acad Sci U S A 100 13280 14581616
Pande VS Rokhsar DS 1999 Molecular dynamics simulations of unfolding and refolding of a β-hairpin fragment of protein g Proc Natl Acad Sci U S A 96 9062 10430895
Dinner AR Lazaridis T Karplus M 1999 Understanding β-hairpin formation Proc Natl Acad Sci U S A 96 9068 10430896
Roccatano D Amadei A Di Nola A Berendsen HJ 1999 A molecular dynamics study of the 41–56 β-hairpin from b1 domain of protein g Protein Sci 10 2130
Kolinski A Ilkowski B Skolnick J 1999 Dynamics and thermodynamics of β-hairpin assembly: Insights from various simulation techniques Biophys J 77 2942 10585918
Ma B Nussinov R 2000 Molecular dynamics simulations of a β-hairpin fragment of protein g: Balance between side-chain and backbone forces J Mol Bio 296 1091 10686106
Klimov DK Thirumalai D 2000 Mechanism and kinetics of β-hairpin formation Proc Natl Acad Sci U S A 97 2544 10716988
Zhou R Berne BJ 2002 Can a continuum solvent model reproduce the free energy landscape of a β-hairpin folding in water? Proc Natl Acad Sci 99 12777 12242327
Jorgensen WL Maxwell D Tirado-Rives J 1996 Development and testing of the opls all-atom force field on conformational energetics and properties of organic liquids J Am Chem Soc 118 11225
Berendsen HJC Postma JPM van Gunsteren WF Hermans J 1981 Interaction models for water in relation to protein hydration Pullman B Intermolecular forces Dordrecht (Netherlands) Reidel 331 342
Klein-Seetharaman J Oikawa M Grimshaw SB Wirmer J Duchardt E 2002 Long-range interactions within a nonnative protein Science 295 1719 11872841
Lepre J Rice J Tu Y Stolovitzky G 2004
Genes@Work: An efficient algorithm for pattern discovery and multi-variate feature selection in gene expression data Bioinformatics 7 1033 1044
Parida L 2000 Some results on flexible-pattern discovery Combinatorial Pattern Matching (CPM2000) LNCS 1848 33
|
16103909
|
PMC1183514
|
CC BY
|
2021-01-05 09:18:21
|
no
|
PLoS Comput Biol. 2005 Jun 24; 1(1):e8
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010008
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science 1610391010.1371/journal.pcbi.001000905-PLCB-RA-0038R2plcb-01-01-09Research ArticleBioinformatics - Computational BiologyGenetics/Functional GenomicsMolecular Biology - Structural BiologyStructural GenomicsPDUGStructure-FunctionImproving the Precision of the Structure–Function Relationship by Considering Phylogenetic Context Structure-Function Relationship in ContextShakhnovich Boris E Bioinformatics Program, Boston University, Boston, Massachusetts, United States of AmericaBourne Philip EditorUniversity of California at San Diego, United States of AmericaE-mail: [email protected] 2005 24 6 2005 1 1 e928 2 2005 23 5 2005 Copyright: © 2005 Boris E. Shakhnovich.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Understanding the relationship between protein structure and function is one of the foremost challenges in post-genomic biology. Higher conservation of structure could, in principle, allow researchers to extend current limitations of annotation. However, despite significant research in the area, a precise and quantitative relationship between biochemical function and protein structure has been elusive. Attempts to draw an unambiguous link have often been complicated by pleiotropy, variable transcriptional control, and adaptations to genomic context, all of which adversely affect simple definitions of function. In this paper, I report that integrating genomic information can be used to clarify the link between protein structure and function. First, I present a novel measure of functional proximity between protein structures (F-score). Then, using F-score and other entirely automatic methods measuring structure and phylogenetic similarity, I present a three-dimensional landscape describing their inter-relationship. The result is a “well-shaped” landscape that demonstrates the added value of considering genomic context in inferring function from structural homology. A generalization of methodology presented in this paper can be used to improve the precision of annotation of genes in current and newly sequenced genomes.
Synopsis
The author provides a novel perspective on a key problem of structural biology: the structure–function relationship in proteins. While relatedness in protein structure correlates with general description of function, attempts to use this relationship predictively are often complicated by its ambiguous nature. A structure encoded by a family of sequences may be implicated in a set of diverse functions across a variety of organisms. The author outlines an innovative approach that underlines the importance of considering genomic context when using structure-comparison methods for functional prediction.
First, the author defines two distance measures: in genomic space and in function space. Then, the author describes a landscape of functional distance based on both structural and phylogenetic relatedness.
It turns out that this landscape forms a “functional well” where proximity occurs when the structures are similar and occur in the same set of genomes. This result may have implications in future research into functional prediction. With the increasing pace of sequence deposition into databanks, this result suggests a simple way to improve functional prediction via structure homology by complementing existing methods with emerging techniques from comparative genomics.
Citation:Shakhnovich BE (2005) Improving the precision of the structure–function relationship by considering phylogenetic context. PLoS Comput Biol 1(1): e9.
==== Body
Introduction
Since the advent of biological data storage in digital format, researchers have struggled to define quantitative measures of comparison for sequence [1], structure [2], and function [3–5]. While proximity measures for sequence and structure are now well established, the problem of defining functional distance has been particularly daunting. Existing computational methods of describing function using ontologies are not a priori well suited for calculating functional distance [3]. However, using mostly anecdotal evidence, researchers have shown that sequences sharing key structural characteristics often display common function [6].
Nevertheless, quantitatively relating structural homology to function has been complicated by a dearth of functional distance measures and numerous examples of folds performing many unrelated functions. This many-to-many relationship between structure and function has been linked to fundamental biological processes and characteristics such as adaptation, specialization, pleiotropy, or differential regulation [7–9]. Despite these difficulties, understanding the relationship between structure and function is one of the foremost challenges of post-genomic biology [10]. Since protein function often depends on genomic context, defining predominant trends in the coalescent evolution of organisms and proteins may be instrumental in improving our understanding of the structure–function relationship [5].
Results/Discussion
I consider the protein domain universe as the set of all structurally characterized domains [11]. I treat each domain as a structural scaffold encoded by a set of homologous sequences [4]. The power of this approach is its ability to leverage the relative conservation of function inside the structural scaffold [5] to statistically determine the relationship between structure and function. Then, using information about the distribution of the domain universe across the evolutionary tree [12,13], I hope to improve the current level of precision [4] of the structure–function relationship. Thus, for each pair of domains, I start by defining and calculating their structural, functional, and phylogenetic similarity (see Materials and Methods and [5,14–16]).
First, I define a simple but quantitative measure of functional comparison: F-score. F-score is defined as normalized Euclidian distance between GO [17] trees built from annotations of sequences coding for each structural scaffold (see Materials and Methods and [17]). Formally, FA,B = 1/L(Σiε{functions}(pA,i − pB,i)2)1/2
FA,B is the functional distance between domain A and domain B, P[A|B],i is the percentage of sequences that fold into structure A or B that are annotated as function i, and L is a normalization constant that accounts for different depths of annotation on the GO. F-score measures similarity of paths on the GO tree between two sets of homologous sequences. For example, if two domains encode two sets of sequences that follow exactly the same path, F-score will be zero. On the other hand, if the sequences encoding the two domains have no common functional annotations, the F-score will be maximum.
Next, I set out to correlate F-score and structural similarity (Z-score calculated using DALI [2]). I expect a general correlation to hold, since previous research has shown that domains sharing key structural characteristics often perform similar functions [6,8]. Indeed, I observed a robust correlation, on average between Z-score and F-score (Figure 1A). However, the dynamic range of this correlation is small. The difference in F-score between the closest and farthest structures is only 30%. This small dynamic range most likely stems from the ambiguous relationship between structure and function.
Figure 1 The Correlations between Z, F, and P Scores
(A) The correlation between structural comparison Z-score and functional distance F-score. (Pearson's r = 0.96 and slope = 0.007.) Each bin contains at least 200 observations. It is worth noting that the average functional distance (F-score) falls from 0.48 to 0.30, only by a third during two decades of structural similarity [14].
(B) The correspondence between phylogenetic profile distances calculated using mutual information and F-score. Slope of the linear fit is 0.36, with Pearson's r = 0.96. The correlation is averaged, i.e., each data point represents a bin containing 150–200 domains, and the functional distances are averaged inside the bin [14].
(C) The landscape of functional distance with respect to Z and P scores. An average F-score is calculated for each of the 36 bins; each bin contains 100–200 observations. Since F-score is a distance metric, hotter colors represent domains that are farther away and cooler colors represent those that are closer.
From an evolutionary perspective, the environment is often important in defining the precise function of the sequence. Consequently, sequences appearing in the same set of genomes have been shown to perform similar functions [18]. Thus, domains with similar phylogenetic profiles should also display similar F-scores [18]. For a measure of phylogenetic similarity, I used the most commonly used mutual information (see Materials and Methods) between phylogenetic profiles of domains (P-score). Since mutual information is reflective, it is maximum when the two domains appear in the same or exactly opposing subset of genomes, and minimum when the overlap in appearance across the genomes is random. I found that P-score is a slightly better predictor of functional similarity than structural homology, with dynamic range of 50% as measured by F-score (Figure 1B). This implies that genomic context, more than constraints imposed by structure alone, may influence the precise function of the gene.
Finally, quantitative definitions of structure, function, and phylogenetic similarity allowed me to calculate the landscape of F-scores for all pairs of domains with respect to their Z and P scores (Figure 1C). Contrary to naïve expectation of smooth transitions across a small range of F-scores observed for pairwise comparisons in Figure 1A and B, I found that the combination of Z-score and P-score forms a well-shaped functional landscape with a sharp transition in F-score. This suggests that similar structures occurring in different genomes often perform dissimilar functions (see Materials and Methods). Alternatively, genes with similar structures are more likely to perform similar functions if the distribution of their orthologs on the evolutionary tree is also similar. This finding is intuitive, since genes often adapt to the environment through mutation in sequence that alters function but not structure.
The findings presented here suggest that both our understanding of the structure–function relationship and the precision of functional annotation can be greatly improved by considering structural homology in phylogenetic context. I am currently involved in work trying to improve on my naïve measure of functional similarity and assess the robustness of these results to arbitrary cutoff parameters. Furthermore, using these results it may be possible to outline a novel, optimal strategy with respect to functional annotation for the currently ongoing structural genomics projects.
Materials and Methods
Evolution is, at its core, a science of comparison. In order to study evolution, I needed to create a computational framework to represent our current body of knowledge. I chose to approach this problem from a graph-theoretic prospective in which nodes are the basic units of evolution and edges are different comparison measures. Aside from providing a unified framework, evolutionary graphs like these provide a way to organize the diverse glut of experimental data that has become the cornerstone of bioinformatics research. In the case of molecular evolution, given that domains can be functionally independent, can be expressed outside larger protein complexes in genomes, and are often rearranged through alternative splicing, I can define a domain as a good evolutionary basic unit subject to structure–function pressures. Consequently, I chose to work with annotations and comparisons of domains instead of whole proteins.
Structural comparison and building of PDUG.
I employed a Z-score measure of structural proximity as weight for the edges to create a protein domains universe graph (PDUG [19]). Formally, I created a graph where the nodes are the representative set of recurring structural domains identified previously by DALI [20,21], and the edges are the structural comparisons between those domains weighed by their respective Z-score. I used the above graph representation to understand the role that pressure on structure plays in the evolution of protein domains. Using this graph-theoretic paradigm, I could investigate not only the topology of the graph but also the correlation between the structural comparison graph and other dimensions of the same graph based on comparison metrics, such as function and phylogenetic proximity explained in detail below. The names of the domains used in this study are available at http://romi.bu.edu/phylo_context/domain_names.txt. The domain names refer to the DALI nomenclature as described in [22].
After I defined a PDUG, I had to populate it using sequences, so as to correlate the structures and the set of sequences that fold into those structures. I used a non-redundant database of sequences, NRDB [23]. This database straightforwardly uses sequence alignment on all known sequence databases to remove neighbors with more than 90% identity to a representative sequence, analogous to the method described above for structures. In order to map the set of recurrent domains onto sequence space, I used the now canonical BLAST [24] sequence alignment algorithm to find homologous PDUG nodes to all non-redundant sequence representatives obtained from NRDB. For every sequence in NRDB, I found the best matching sequences below 1e-10 threshold. Since structures from DALI are themselves devoid of sequence homologs, at most one structure is found for every non-redundant sequence from NRDB. Since each sequence is annotated with the function that it performs, this yields a mapping not only of non-redundant sequences but also of their respective functions to nodes on PDUG. The distribution of sequences from NRDB that are homologous to DALI structures is given in Figure S1.
Functional domain universe graph.
Since I was interested in the most general description of functionality of protein domains, I defined the function of each domain as the weighted set of functions performed by all the sequences that align to it. Thus, the functionality of the domain is represented by a probabilistic GO [25] tree. This tree is populated by taking all non-redundant sequences matching each PDUG node (as described above) and placing their functional annotations into the canonical GO. I rebuilt the whole GO tree by following all paths that led to root node from the functional annotations mined out of NRDB sequences. I increased the count of a node each time I visited it. Afterwards, all counts were turned into probabilities by normalizing the number of times that I visited each node on every level of the GO tree by the total number of times I visited that level. I ended up with a probabilistic representation of function for each structure at various levels of specificity.
Each node on PDUG now had the representative structure, the set of sequences that fold into that structure, and the set of functions performed by those sequences in the form of a probabilistic, hierarchical GO [25] tree. The benefit of representing functionality in terms of a probabilistic GO tree is that I could now compare functionality of domains by simply comparing their GO trees. If I wanted to understand the “difference” in function between two domains, I needed to take into account all functions that this structure was implicated in. For example, some sequences for a given structure may be involved in creatine phosphorylation, and others can be involved in arginine phosphorylation, as in the case of 1qh4 (Figure S2) [26].
Thus, in order to compare the GO trees, I calculated the Euclidian distance between the nodes on each level of the GO hierarchy by using Equation 1.
Here FA,B is the functional distance between domain A and domain B, pA,i is the percentage of sequences that fold into structure A that are annotated with function i, and the sum is taken over all annotated functions. F-score measures similarity of paths on the GO tree between two sets of homologous sequences. For example, if two domains encode two sets of sequences that follow exactly the same path, the F-score will be zero. On the other hand, if the sequences encoding the two domains have no common functional annotations, the F-score will be maximum. Using the above functional distance measure, I created another dimension of PDUG. In this dimension, the edges are functional comparisons between the domains and are weighed by the F-score.
Phylogenetic distance P-score.
Phylogenetic context (the subset of genomes where the domain is found) can have a profound effect on the function and overall evolution of that domain. Knowing this, I created another dimension of PDUG where each node was annotated with the genomes where it was present. This is done by simply BLASTing [24] the set of non-redundant sequences found in each node in PDUG against all fully sequenced and mapped genomes. This yields a mapping of structural space into genomic space. Thus, each node is annotated with a vector where columns represent the different genomes and the values are zero or one, depending on whether the domain exists in that genome.
The calculation of distance in genome space is non-trivial and is subject to all kinds of qualifications, such as relative distance between genomes on the tree [27,28]. However, I simplified the calculation by employing mutual information as a first-order approximation to distance between every two phylogenetic vectors. The distance between any two nodes in phylogenetic space is then just the mutual information between their vectors, as defined by
where pij is the frequency of occurrence of all four possible combinations of presence or absence in the same genome for nodes i and j, and pi and pj are the marginal probabilities of seeing those domains in all genomes. Mutual information is a reflexive measure, insensitive to correlation or anti-correlation. Thus, mutual information will be maximal if the two phylogenetic vectors are either perfectly correlated or perfectly anti-correlated while the norm of that vector is half the length. This is a useful property for evaluating P-score, since genes that appear in a completely disparate set of genomes have been shown to perform similar functions in a process dubbed “non-orthologous gene displacement” [29]. Using this distance measure, I created the third and final dimension of PDUG where the nodes are the domains with redundant sequences and functional trees, and the edges are weighed by the mutual information measure between the phylogenetic profiles of the nodes.
Looking through the dimensions.
Finally, I correlated all three dimensions of PDUG, by observing the F-score between two nodes with respect to both the structural proximity and the phylogenetic distance (Table S1). The striking observation was that resolution of functional distance increases by almost 100% when considering structural proximity and phylogenetic distance over using any one of these measures alone. The combination of phylogenetic distance and structural similarity differentiates structures with close functional similarity from similar structures without functional similarity, and analogous behavior is observed for sets of domains sharing phylogenetic profiles (data not shown). The protein domains that run contrary to this trend are good candidates for investigating convergent evolution.
Robustness analysis.
To evaluate the robustness of the results reported in Figure 1C, I performed a jackknife analysis to evaluate the standard deviation of each data point on the graph. I sampled 60% of the data 150 times. I then gridded those points, as in Figure 1C, and then I conglomerated the results. The means and standard deviations of F-scores for each pair of Z and P scores can be accessed directly from http://romi.bu.edu/phylo_context/z_p_f_landscape_stat_sig.dat. The difference in the functional similarity score is several standard deviations away from random and is highly significant. Moreover, the overall nature of the results does not depend on the binning, or the way that the jackknife procedure is performed (data not shown).
Supporting Information
Figure S1 The Distribution of Sequences from NRDB That Are Homologous to a Structure
The data are available online from http://romi.bu.edu/phylo_context/count_seqs.out. The structures may be downloaded from the PDB directly and from the ASTRAL compendium using the domain names provided in http://romi.bu.edu/phylo_context/domain_names.txt.
(4.2 MB TIF).
Click here for additional data file.
Figure S2 Example of Uneven Scaffold Annotation on the Functional GO Tree
(1.3 MB TIF).
Click here for additional data file.
Table S1 Use of Phylogenetic Distance for a Particular Structural Similarity Score Differentiates Functionally Related Proteins from Those That Are Not
(449 KB TIF).
Click here for additional data file.
The author would like to acknowledge Eugene Shakhnovich, John Max Harvey, and support from NIH.
Competing interests. The author has declared that no competing interests exist.
Author contributions. BES conceived and designed the experiments, performed the experiments, analyzed the data, contributed reagents/materials/analysis tools, and wrote the paper.
Abbreviations
GOGene Ontology
PDUGprotein domains universe graph
==== Refs
References
Altschul SF Madden TL Schaffer AA Zhang J Zhang Z 1997 Gapped BLAST and PSI-BLAST: A new generation of protein database search programs Nucleic Acids Res 25 3389 3402 9254694
Dietmann S Holm L 2001 Identification of homology in protein structure classification Nat Struct Biol 8 953 957 11685241
Lord PW Stevens RD Brass A Goble CA 2003 Investigating semantic similarity measures across the Gene Ontology: The relationship between sequence and annotation Bioinformatics 19 1275 1283 12835272
Shakhnovich BE Max Harvey J 2004 Quantifying structure-function uncertainty: A graph theoretical exploration into the origins and limitations of protein annotation J Mol Biol 337 933 949 15033362
Shakhnovich BE Dokholyan NV DeLisi C Shakhnovich EI 2003 Functional fingerprints of folds: Evidence for correlated structure-function evolution J Mol Biol 326 1 9 12547186
Orengo CA Bray JE Buchan DW Harrison A Lee D 2002 The CATH protein family database: A resource for structural and functional annotation of genomes Proteomics 2 11 21 11788987
Fraser AG Marcotte EM 2004 A probabilistic view of gene function Nat Genet 36 559 564 15167932
Murzin AG Brenner SE Hubbard T Chothia C 1995 SCOP: A structural classification of proteins database for the investigation of sequences and structures J Mol Biol 247 536 540 7723011
Eisenstein E Gilliland GL Herzberg O Moult J Orban J 2000 Biological function made crystal clear—Annotation of hypothetical proteins via structural genomics Curr Opin Biotechnol 11 25 30 10679350
Montelione GT Anderson S 1999 Structural genomics: Keystone for a Human Proteome Project Nat Struct Biol 6 11 12 9886282
Dokholyan NV Shakhnovich B Shakhnovich EI 2002 Expanding protein universe and its origin from the biological Big Bang Proc Natl Acad Sci U S A 99 14132 14136 12384571
Deeds EJ Hennessey H Shakhnovich EI 2005 Prokaryotic phylogenies inferred from protein structural domains Genome Res 15 393 402 15741510
Yang S Doolittle RF Bourne PE 2005 Phylogeny determined by protein domain content Proc Natl Acad Sci U S A 102 373 378 15630082
Shakhnovich BE Harvey JM Delisi C 2004 ELISA: A unified, multidimensional view of the protein domain universe Genome Inform Ser Workshop Genome Inform 15 213 220
Day R Beck DA Armen RS Daggett V 2003 A consensus view of fold space: Combining SCOP, CATH, and the Dali Domain Dictionary Protein Sci 12 2150 2160 14500873
Dietmann S Park J Notredame C Heger A Lappe M 2001 A fully automatic evolutionary classification of protein folds: Dali Domain Dictionary version 3 Nucleic Acids Res 29 55 57 11125048
Ashburner M Ball CA Blake JA Botstein D Butler H 2000 Gene ontology: Tool for the unification of biology. The Gene Ontology Consortium Nat Genet 25 25 29 10802651
Pellegrini M Marcotte EM Thompson MJ Eisenberg D Yeates TO 1999 Assigning protein functions by comparative genome analysis: Protein phylogenetic profiles Proc Natl Acad Sci U S A 96 4285 4288 10200254
Dokholyan NV Shakhnovich B Shakhnovich EI 2002 Expanding protein universe and its origin from the biological Big Bang Proc Natl Acad Sci U S A 99 14132 14136 12384571
Holm L Sander C 1995 Dali: A network tool for protein structure comparison Trends Biochem Sci 20 478 480 8578593
Holm L Sander C 1997 Dali/FSSP classification of three-dimensional protein folds Nucleic Acids Res 25 231 234 9016542
Dietmann S Park J Notredame C Heger A Lappe M 2001 A fully automatic evolutionary classification of protein folds: Dali Domain Dictionary version 3 Nucleic Acids Res 29 55 57 11125048
Holm L Sander C 1998 Removing near-neighbour redundancy from large protein sequence collections Bioinformatics 14 423 429 9682055
Altschul SF Madden TL Schaffer AA Zhang J Zhang Z 1997 Gapped BLAST and PSI-BLAST: A new generation of protein database search programs Nucleic Acids Res 25 3389 3402 9254694
Ashburner M Ball CA Blake JA Botstein D Butler H 2000 Gene ontology: Tool for the unification of biology. The Gene Ontology Consortium Nat Genet 25 25 29 10802651
Eder M Schlattner U Becker A Wallimann T Kabsch W 1999 Crystal structure of brain-type creatine kinase at 1.41 A resolution Protein Sci 8 2258 2269 10595529
Wu J Kasif S DeLisi C 2003 Identification of functional links between genes using phylogenetic profiles Bioinformatics 19 1524 1530 12912833
Pellegrini M Marcotte EM Thompson MJ Eisenberg D Yeates TO 1999 Assigning protein functions by comparative genome analysis: Protein phylogenetic profiles Proc Natl Acad Sci U S A 96 4285 4288 10200254
Koonin EV Mushegian AR Bork P 1996 Non-orthologous gene displacement Trends Genet 12 334 336 8855656
|
16103910
|
PMC1183515
|
CC BY
|
2021-01-05 09:18:23
|
no
|
PLoS Comput Biol. 2005 Jun 24; 1(1):e9
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010009
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science 1610389910.1371/journal.pcbi.001001005-PLCB-RA-0018R2plcb-01-01-08Research ArticleBioinformatics - Computational BiologyNoneExtraction of Transcript Diversity from Scientific Literature Text Mining for Alternative TranscriptsShah Parantu K 12Jensen Lars J 1Boué Stéphanie 1Bork Peer 12*1 Structural and Computational Biology Program, European Molecular Biology Laboratory, Heidelberg, Germany
2 Max Delbrück Centre for Molecular Medicine, Berlin-Buch, Germany
Bourne Philip EditorUniversity of California at San Diego, United States of America* To whom correspondence should be addressed. E-mail: [email protected] 2005 24 6 2005 1 1 e101 2 2005 21 5 2005 Copyright: © 2005 Shah et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Transcript diversity generated by alternative splicing and associated mechanisms contributes heavily to the functional complexity of biological systems. The numerous examples of the mechanisms and functional implications of these events are scattered throughout the scientific literature. Thus, it is crucial to have a tool that can automatically extract the relevant facts and collect them in a knowledge base that can aid the interpretation of data from high-throughput methods. We have developed and applied a composite text-mining method for extracting information on transcript diversity from the entire MEDLINE database in order to create a database of genes with alternative transcripts. It contains information on tissue specificity, number of isoforms, causative mechanisms, functional implications, and experimental methods used for detection. We have mined this resource to identify 959 instances of tissue-specific splicing. Our results in combination with those from EST-based methods suggest that alternative splicing is the preferred mechanism for generating transcript diversity in the nervous system. We provide new annotations for 1,860 genes with the potential for generating transcript diversity. We assign the MeSH term “alternative splicing” to 1,536 additional abstracts in the MEDLINE database and suggest new MeSH terms for other events. We have successfully extracted information about transcript diversity and semiautomatically generated a database, LSAT, that can provide a quantitative understanding of the mechanisms behind tissue-specific gene expression. LSAT (Literature Support for Alternative Transcripts) is publicly available at http://www.bork.embl.de/LSAT/.
Synopsis
Given the functional complexity of higher eukaryotes, the relatively small number of genes in the human and other mammalian genomes came as a surprise to the scientific community. Later it was discovered that the majority of genes are subject to alternative splicing (“cutting and pasting”) or associated mechanisms that ultimately increase the diversity of transcripts that code for proteins. Studies exploring transcript diversity are currently dominated by high-throughput experiments and computational methods; however, the quality of such data should be assessed against a reliable reference set based on single-gene studies. Unfortunately, the latter type of information is scattered throughout the scientific literature. The authors have thus developed a computational approach for extracting information on alternative transcripts from MEDLINE abstracts and used it to create a database, LSAT. LSAT (Literature Support for Alternative Transcripts) provides information for more than 4,000 genes from about 14,000 abstracts. This database can provide a quantitative understanding of the mechanisms behind tissue-specific gene expression based on single-gene studies, which we show agrees well with EST-based studies (these studies involve tissue-specific splicing detected by the analysis of libraries of expressed sequence tags [ESTs]). These results indicate that mechanisms like alternative splicing, alternative promoters, and alternative polyadenylation work in concert to generate and regulate transcript diversity. More generally, information extraction of complex biological process seems feasible and can also complement large-scale data generation in other areas to assign functions to genes.
Citation:Shah PK, Jensen LJ, Boué S, Bork P (2005) Extraction of transcript diversity from scientific literature. PLoS Comput Biol 1(1): e10.
==== Body
Introduction
Although many model organisms have now been completely sequenced, we are still very far from understanding cellular function from genome sequence. One complicating factor is the expression of multiple alternative mRNA transcripts from a single gene using different mechanisms. Alternative promoters that are active in different tissues or at different developmental stages often regulate the expression of different mRNA isoforms, either directly through different transcription start sites or indirectly by promoter-directed exon inclusion in concert with alternative splicing (AS) [1]. Various AS mechanisms are known: alternative 5′ or 3′ sites can result in exons of different size, exons can be included or skipped, or an entire intron may be retained [2–5]. Alternative polyadenylation (AP), either alone or coupled with AS of 3′ terminal exons, may also generate transcript isoforms that are tissue- or developmental-stage-specific [6].
Generation of multiple alternative transcripts is important for the complexity and evolution of eukaryotic organisms [5,7–9]. In addition, their spatial and temporal expression patterns are believed to be one of the important factors behind the functional specificity of different tissues and organs. Moreover, defects in these processes are associated with various diseases [2]. Thus, developing an exhaustive catalog of alternative transcripts is a crucial task in order to fully understand the complexity of eukaryotes [7].
At present, high-throughput experiments and computational analyses dominate the mapping of the alternative transcript universe [10,11]. However, the quality and the biological meaning of these assignments should be assessed against a highly reliable benchmark set, which can be extracted from single-gene studies published in the scientific literature [3,12,13]. In addition, computational tools to explore the evolutionary conservation of mechanisms that generate transcript diversity (TD) are under development [14], which will also require a trustworthy set for rule learning.
Manual curation of experimentally determined biological events (physical interactions, AS, disease phenotypes, etc.) to generate trustworthy knowledge bases is slow compared to the rapid increase in the body of knowledge represented in the literature. Natural language processing tools thus play an increasingly important role in transferring information from free-form biomedical text to structured databases (see reviews [15–18]). This task can be split in to two steps: (1) a subset of documents describing events or scenarios of interest is identified (information retrieval [IR]), and (2) facts are extracted from these documents and deposited into structured fields (information extraction [IE]).
IR can be performed at the level of full articles, pertinent paragraphs, or sentences. As current IE methods operate at the sentence level, it may be appropriate to perform IR at the same level. Support vector machines have become the method of choice for IR tasks because of their ability to learn patterns and generalize well while handling large sets of input features, a common attribute of the text data [19–21]. Most IE systems use rules written by the domain experts to extract facts about events or scenarios of interest. The performance of most rule-based systems suffers because of the fact that any event or scenario can be written in one of many syntactically correct ways. Thus, an extraction system based only on syntactic patterns would require an exhaustive collection of rules in order to cover all possible patterns. The problem posed by multiple syntactic patterns can be solved by merging multiple syntactic patterns to a single semantic pattern by predicate–argument structures [22–24]. Predicate–argument structures and support vector machines (SVMs) are becoming prevalent in natural language processing and are widely believed to achieve good recall and precision; they were tested here for their applicability to the biomedical literature.
Here we present the benchmark and the results of a new extraction procedure that combines an SVM classifier with rule-based extraction of semantic patterns. The extracted knowledge about TD was stored in a database and subsequently used to quantify the amount of TD in different tissues. We discuss applications of our work for the assignment of MeSH terms (from the National Library of Medicine's Medical Subject Headings thesaurus), providing functional annotations to genes and to the transcript variants generated by computational methods.
Results/Discussion
Overall Strategy and Generation of the Database
To extract information about TD and associated spatiotemporal information scattered throughout MEDLINE, we devised a two-step procedure (Figure 1). In the first step, sentences containing TD information were identified within the papers' abstracts. To do so, and in order to overcome the problem of syntactic patterns, we tan SVM classifier for the sentence classification task by inductive machine learning [25] on an annotated corpus [19–21]. We then processed the entire MEDLINE database and identified sentences describing TD within those abstracts. In the second step, sentences were parsed and the word phrases were assigned different meaningful (semantic) categories (see Materials and Methods).
Figure 1 Creating Specialized Databases for Events of Interest
A database of physiologically occurring AS events can be generated in two steps. Each step may involve machine learning or rule-based methods. The first step involves the identification of sentences from scientific text. These sentences can be parsed in a second step to extract frequently occurring semantic patterns.
Finally, we mapped each abstract with information about alternative transcripts (retrieved by the SVM classifier) to entries in Swiss-Prot [26], RefSeq [27], GenBank [28], and Ensembl [29] databases, when possible. This not only provided the sequence information at genome, transcript, and protein level for the genes described in abstracts but also allowed us to access structural and functional information about these genes stored in various sequence databases. All this information obtained for each MEDLINE entry constitutes an entry in LSAT (Figure S1).
We identified eight different semantic categories describing biologically relevant data in the sentences describing TD, among which are event mechanism, species, tissue specificity, and experimental methods (Table 1; see Materials and Methods). In total we extracted 9,503 instances of event mechanisms from as many abstracts (Table S1) and 5,028 instances of tissues (Table S2) with associated gene names. Overall, the database contains 3,063, 874, and 207 nonredundant instances of AS, differential promoter usage (DP), and AP associated with genes and tissues extracted by entity taggers.
Table 1 Extraction of Semantic Patterns
Performance of the SVM Classifier for Sentence Retrieval
Our SVM classifier retrieved 31,123 putative TD-containing sentences from the MEDLINE database (12,948,515 abstracts). After false positives were removed by manual curation, 20,549 TD-containing sentences in 13,892 abstracts were left, corresponding to a precision of 66%. Details on the training set and SVM training procedure are described in Materials and Methods and Protocol S1.
We determined the recall of the classifier using manually curated AS annotations from MEDLINE and Swiss-Prot for annotations on human, mouse, rat, and Drosophila. All entries from MEDLINE 2004 annotated with the MeSH term “alternative splicing” and describing natural transcript generation (see Materials and Methods) were compared with our results. For each of these four species, we also analyzed our results on MEDLINE entries referred to in Swiss-Prot entries annotated with the keyword “alternative splicing” [26]. The average sensitivity of the classifier was 61% (Table 2; see Materials and Methods). The SVM classifiers thus achieve good recall and precision and can be used for extracting biological events.
Table 2 Recall of the SVM Classifier
Performance of the IE Step
From the sentences retrieved by the SVM classifier, we extracted instances of eight semantic categories (see Materials and Methods) and evaluated the precision and recall by manually inspecting 300 randomly selected sentences for each category (see Table 1). Both precision and recall are highly satisfactory; however, it should be noted that accuracy in finding tag boundaries was not considered. Also, the recall is good for all categories, but not all eight categories are equally represented in the sentences (see Table 1).
Proposing New Annotations in Curated Databases
Annotators at the National Library of Medicine have manually assigned the MeSH term “alternative splicing” to 8,133 abstracts. During the IE step, we identified 1,536 additional abstracts that mention AS but lack the MeSH term “alternative splicing,” corresponding to a 19% increase in annotation. We also identified DP and AP in 874 and 219 abstracts, respectively, for which we propose the new MeSH terms “alternative promoters” and “alternative polyadenylation” (Tables S3–S6).
We also quantified the number of Ensembl genes for which we can propose new annotations for AS (see Materials and Methods). The annotation increase observed was 20%, 52%, and 105% for human, mouse, and rat genomes, respectively (Figure S2). These tentative assignments can supplement the work of curators, and the numbers are likely to reflect the current extent of manual curation for these different genomes. The annotation increase for the human genes was relatively little compared to that for the rat genes because a total of 3,438 genes are already annotated in Swiss-Prot and RefSeq for AS in human, whereas only 342 genes are annotated for AS in rat. Even more annotations could be obtained by manually curating extracted events that could not be automatically mapped to a sequence database entry; we have manually mapped 190 genes exhibiting tissue-specific splicing. The observed increase in the annotation emphasizes the need for automated methods to speed up the process of database curation.
Quantification of the Different Mechanisms That Lead to TD
The majority of vertebrate multi-exon genes undergo AS [10]. Moreover, different promoters may control the transcription of different mRNA isoforms, which may result in directed 5′ exon inclusion/exclusion, and AP signals can control the tissue specificity of alternative 3′ exons. While examples of synergy between these mechanisms are known, the extent of it is currently being explored. We found DP co-mentioned with AS in 14% of abstracts describing genes with differential promoters. A total of 19% of the abstracts providing information about alternative first exon usage also mentioned usage of different promoters. A total of 17% abstracts describing AP also mentioned AS.
The extent to which various mechanisms are utilized for increasing TD may vary across different anatomical systems. To study this, we mapped all vertebrate tissue information to anatomical systems using the MeSH anatomy terms and counted the number of nonredundant events extracted for each mechanism in each system (Figure 2, top panel). AS is utilized equally in most organs except in the nervous system, where AS is significantly overrepresented (Figure 2, bottom panel). Similarly, there is significant overrepresentation of DP in the connective tissues and to a lesser extent in the digestive system and in the genitalia.
Figure 2 Preference for the Utilization of TD-Generating Mechanisms across Anatomical Systems
Nonredundant instances of AS, DP, and AP are plotted against anatomical systems in which expression was found. The color of each square in the top panel signifies the ratio of the number of events detected for the system to the highest number of events within the row. Total number of nonredundant instances for each mechanism is on the left. The bottom panel shows the negative logarithm of p-values (see Materials and Methods for details). The anatomical systems are as follows: A, cardio vascular system; B, cells; C, connective tissues; D, digestive system; E, fetal/embryonic structures; F, endocrine system; G, exocrine glands; H, genitalia; I, immune system; J, integumentary system; K, musculoskeletal system; L, nervous system; M, respiratory system; N, sense regions; O, urinal system.
The information about alternative promoter usage linked with specific gene names and tissues extracted in this study is the largest such collection available, to our knowledge. We expect that it would provide a reliable dataset for development of computational methods for predicting tissue-specific promoter usage.
Tissue-Specific Differences in the Extent of AS
AS has been shown to play an important role in creating functional specialization of tissues and development stages [30,31], but only a small number of instances of tissue-specific splicing are listed in the current AS databases [32,33]. With a large collection of high-quality AS events in hand, tissue-specific differences in AS should become visible. We checked entries in our database containing the field “specificity.” We identified 959 events describing tissue specificity in AS. These represented 675 AS events for pairs of tissues and 284 events where only one tissue was reported. The results contained 400 nonredundant events for 183 human genes. We also mapped a further 190 genes (not included above) from various species to Swiss-Prot identifiers during the manual curation.
To study the extent of tissue-specific AS, we mapped tissues and organs to respective systems as described in the previous section and plotted the results (Figure 3, left panel). The nervous system, genitalia, and immune, digestive, and musculoskeletal systems showed extensive tissue specificity in inter- and intra-systemic AS. These systems also showed expression of unique AS transcripts, with the nervous system showing the highest number of unique transcripts. These tissue-specific patterns of expression extracted from the literature strongly overlap with the 667 tissue-specific AS events derived from analysis of the EST data [33] for 454 human genes across 46 tissues (Figure 3, right panel).
Figure 3 Tissue Specificity in AS
The figure shows the body system distribution of differential/specific splicing. The instances were obtained from literature mining (left panel) and analysis of EST data ([33]; right panel). Each square is colored according to the ratio between the corresponding count and the highest count within the panel. Letter codes for anatomical systems are as in Figure 2. P represents a unique transcript.
The knowledge extracted from the literature confirms EST-based studies [31,33] and earlier experimental studies [34] that showed AS as the preferred mechanism for generating TD across the nervous system. EST-based studies [31] have also suggested that genes in liver (digestive system) and testis (genitalia) show distinct patterns of splicing with alternative exons. Our results indicate that these transcripts may show these different patterns of splicing in combination with different promoter regions. This conclusion seems plausible since AS of first exons is influenced by alternative promoter regions in at least 19% of cases (see above; [35]), and it should be explored further.
Assigning Function to the Transcripts Generated by Computational Analysis
Sometimes experimental biologists speculate about the mechanism responsible for the multiple transcripts observed with a limited number of experiments but the corresponding transcripts are not deposited in GenBank. For example, work by Pisarra et al. [36] on human Dopachrome tautomerase describes two transcripts in melanocytes and melanomas with a “different carboxyl-terminus” generated, concluding that “dopachrome tautomerase can yield different isoforms by alternative poly(A) site usage or by alternative splicing” (Figure 4).
Figure 4 Assignment of Function Using Database Knowledge
This figure shows a database entry that derives very little functional annotation from sequence databases. Text extraction rules were successful in identifying gene name, tissue, and event mechanism for the Dopachrome tautomerase gene. Multiple transcripts of the gene using SPLICE-POA [37] were produced by utilizing alternative 3′ splice sites and polyadenylation signals as speculated in the research article (bottom panel). Pink rectangles denote the exons, black lines describe constitutive splice sites, and blue lines show alternative splice sites. Black arrows show the different proteins generated via AS.
On the other hand, various methods, including those based on aligning EST and other sequence data to genomic regions, are currently in use for detecting AS on a large scale. The function of the isoforms thus generated is largely unknown [37], and these transcripts are poorly annotated in sequence databases.
Using the heaviest bundling algorithm [37] with genomic sequence data from Ensembl [38], and transcript data from UniGene [39] clusters for the gene, we were able to generate two transcript isoforms for Dopachrome tautomerase (Figure 4, bottom) resembling those described by Pisarra et al. [36] and were able to detect an AS event in the 3′ region. Hence, the use of large-scale methods may provide detailed information about underlying events, and text mining would provide functional annotations to the transcript isoforms observed.
Conclusions
We successfully extracted information about the genes that express multiple transcripts and associated spatiotemporal information using state-of-the-art methods in natural language processing and utilized it for function annotations. The information extracted by far exceeds current manual curation efforts and generates reliable results. Our results indicate that mechanisms like AS, DP, and AP work in concert for the generation and regulation of TD. They also suggest that the nervous system preferentially relies on AS over other mechanisms to express the largest set of tissue-specific transcripts. In contrast, genitalia and the digestive system more frequently make use of alternative promoter regions. The knowledge stored in the database about synergy and preference for TD-generating mechanisms across tissues will be integrated to high-throughput data in the future. More generally, IE of complex biological processes seems feasible and can also complement large-scale data generation in other areas to assign function.
Materials and Methods
Training corpus and SVM learning.
A set of 4,240 sentences describing physiological TD and 13,520 negative sentences were selected as a training corpus from article titles and abstracts. Sentences describing mutations, clinical studies involving patients, nucleotide transversions, and splicing mechanisms were considered negative sentences. Sentences describing natural gene expression, gene paralogs, and aberrant transcripts showed word usage similar to those describing TD, making the classification task more challenging. Description of the learning corpus can be found in Protocol S1 and Figure S3.
The text in all the abstracts was split into sentences using the Oak system (S. Sekine, unpublished data; http://nlp.cs.nyu.edu/oak/). All the sentences were tagged with Tree-tagger [40] to give words their part-of-speech tags. Sentences were broken down into constituent words and stemmed to act as features to learn from. Stop words and words with certain part-of-speech tags were removed from the primary features. To add domain knowledge and enrich the features to learn from, frequently occurring word bi-grams and tri-grams were also defined from unprocessed sentences. The feature file was large, containing 23,742 features.
The procedure of inductive learning (see Protocol S1) was applied for the sentence classification task, using the feature vectors described above as input. We compared the performance of naïve Bayes, expectation maximization, maximum entropy, variants of TF-IDF, K-nearest neighbors, and support vector machines for the task [21,41–43]. The SVM with a radial basis function kernel (gamma = 1.5 and C = 100) outperformed other methods and SVM classifiers with linear and sigmoid kernel functions (P. K. Shah and P. Bork, unpublished data).
The classifier was trained to extract only the natural TD from the written text, as contrasted by aberrant transcripts that are caused by genetic changes. For consistency, we removed the 2,767 of the 8,133 MEDLINE entries with the MeSH term “alternative splicing” that also had the MeSH term “mutation,” had no abstract text, or had erroneous assignment of the MeSH term “alternative splicing.”
Definitions of precision and recall.
Precision and recall are used in IR to measure the performance of methods and they are defined as below.
Recall = TP/(TP + FN); Precision = TP/(TP + FP) (1) Where, TP, TN, FP, and FN denote true-positive, true-negative, false-positive, and false-negative elements according to a classification criterion.
Parsing of the sentences using semantic patterns.
An event or a scenario is described in a sentence via the combination of a predicate (normally a verb) and its arguments [22–24,44]. While the same biological relation can be described in many syntactically different ways, only a limited number of semantic categories (e.g., gene name or tissue name) may accompany the predicates (see Protocol S1 for further discussion). Therefore, at this step we can apply rule-based methods without much loss of sensitivity.
We constructed semantic patterns similar to those described in the PASBio database of predicate–argument structure [22]. These patterns match informative parts of sentences, e.g., “gene lacks exon n in tissue.” The Stanford lexical parser was used for parsing the sentences [45,46]. Sentence trees were viewed using the TigerSearch tool for generating extraction rules for taking the semantic patterns from sentences [47]. (See Protocol S1 for examples of rules.)
The success in assigning gene, species, and event mechanisms to abstracts is as follows (Figure S3). A total of 46% of all abstracts were directly mapped to literature entries in sequence databases such as Swiss-Prot, RefSeq, and GenBank. A further 15% of all abstracts were assigned gene names using a gene tagger [48], with the species name extracted from the sentences and/or from the MeSH terms mapped with the synonym list. However, only 54% of all abstracts could be unambiguously assigned to a unique species (see Figure 2, category A in lower right histogram). The rest of the abstracts may have had gene and species information but they could not be assigned to a sequence database. Tissues were tagged using a dictionary made of tissue lists from the Swiss-Prot and RefSeq databases. They were assigned to the relevant anatomical system (top level MeSH anatomy terms) using the MeSH browser. We have submitted these entries for manual curation to EMBL-EBI's Alternative Exon Database [32].
Quantifying the gain in gene annotation.
To quantify the gain in gene annotation, first we mapped sequence information to the MEDLINE identifiers from the SVM classification using literature entries in Swiss-Prot, RefSeq, and GenBank. Second, we mapped sequence-containing entries for human, mouse, and rat genes present in our results and in those databases to Ensembl gene identifiers using the EnsMart system. Then we compared our annotations to those of Swiss-Prot and RefSeq to identify genes that were missed during the manual curation of AS. Special care was taken to avoid annotations that may have arisen because of a single literature entry mapping to multiple database entries. Hence, these annotations were highly significant.
Associating TD-generating mechanisms with organ systems.
The significance of the association of each TD-generating mechanism with each organ system was evaluated using the hypergeometric distribution. We corrected these p-values for multiple testing by calculating the false discovery rate using the Benjamini-Hochberg formula [49]. We found 14 significant associations (out of 45) at a 5% false discovery rate, three of which were also significant at a 1% false discovery rate.
Supporting Information
Figure S1 An Example Database Entry
(1.7 MB TIF).
Click here for additional data file.
Figure S2 Distribution of the Results of the IE Step
(4.6 MB TIF).
Click here for additional data file.
Figure S3 Description of the Training Set
(60 KB PDF).
Click here for additional data file.
Protocol S1 Supplementary Text
(112 KB PDF).
Click here for additional data file.
Table S1 Genes and Associated TD-Generating Mechanism
(423 KB TXT).
Click here for additional data file.
Table S2 Genes and Tissues
(120 KB TXT).
Click here for additional data file.
Table S3 Abstracts Describing AS
(445 KB XLS).
Click here for additional data file.
Table S4 Abstracts Describing Alternative Promoters
(76 KB XLS).
Click here for additional data file.
Table S5 Abstracts Describing Alternative Initiation
(20 KB XLS).
Click here for additional data file.
Table S6 Abstracts Describing AP
(29 KB XLS).
Click here for additional data file.
Authors would like to thank Yi Xing and Dr. Christopher Lee for providing the code for SPLICE-POA and the isoform generation algorithm.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. PKS, LJJ, and PB conceived and designed the experiments. PKS performed the experiments. PKS and SB analyzed the data. PKS contributed reagents/materials/analysis tools. PKS, LJJ, and PB wrote the paper.
Abbreviations
APalternative polyadenylation
ASalternative splicing
DPdifferential promoter usage
IEinformation extraction
IRinformation retrieval
SVMsupport vector machine
TDtranscript diversity
==== Refs
References
Landry JR Mager DL Wilhelm BT 2003 Complex controls: The role of alternative promoters in mammalian genomes Trends Genet 19 640 648 14585616
Garcia-Blanco MA Baraniak AP Lasda EL 2004 Alternative splicing in disease and therapy Nat Biotechnol 22 535 546 15122293
Modrek B Lee C 2002 A genomic view of alternative splicing Nat Genet 30 13 19 11753382
Black DL 2003 Mechanisms of alternative pre-messenger RNA splicing Annu Rev Biochem 72 291 336 12626338
Boue S Letunic I Bork P 2003 Alternative splicing and evolution Bioessays 25 1031 1034 14579243
Edwalds-Gilbert G Veraldi KL Milcarek C 1997 Alternative poly(A) site selection in complex transcription units: Means to an end? Nucleic Acids Res 25 2547 2561 9185563
Graveley BR 2001 Alternative splicing: Increasing diversity in the proteomic world Trends Genet 17 100 107 11173120
Brett D Pospisil H Valcarcel J Reich J Bork P 2002 Alternative splicing and genome complexity Nat Genet 30 29 30 11743582
Lareau LF Green RE Bhatnagar RS Brenner SE 2004 The evolving roles of alternative splicing Curr Opin Struct Biol 14 273 282 15193306
Johnson JM Castle J Garrett-Engele P Kan Z Loerch PM 2003 Genome-wide survey of human alternative pre-mRNA splicing with exon junction microarrays Science 302 2141 2144 14684825
Hu GK Madore SJ Moldover B Jatkoe T Balaban D 2001 Predicting splice variant from DNA chip expression data Genome Res 11 1237 1245 11435406
Modrek B Resch A Grasso C Lee C 2001 Genome-wide detection of alternative splicing in expressed sequences of human genes Nucleic Acids Res 29 2850 2859 11433032
Modrek B Lee CJ 2003 Alternative splicing in the human, mouse and rat genomes is associated with an increased frequency of exon creation and/or loss Nat Genet 34 177 180 12730695
Philipps DL Park JW Graveley BR 2004 A computational and experimental approach toward a priori identification of alternatively spliced exons RNA 10 1838 1844 15525709
Andrade MA Bork P 2000 Automated extraction of information in molecular biology FEBS Lett 476 12 17 10878241
de Bruijn B Martin J 2002 Getting to the (c)ore of knowledge: Mining biomedical literature Int J Med Inform 67 7 18 12460628
Shatkay H Feldman R 2003 Mining the biomedical literature in the genomic era: An overview J Comput Biol 10 821 855 14980013
Hirschman L Park JC Tsujii J Wong L Wu CH 2002 Accomplishments and challenges in literature data mining for biology Bioinformatics 18 1553 1561 12490438
Cristianini N Shawe-Taylor J 2000 An introduction to support vector machines and other kernel-based learning methods Cambridge Cambridge University Press 189 p.
Vapnik VN 2000 The nature of statistical learning theory, 2nd ed New York Springer 314 p.
Joachims T 2001 Learning to classify text using support vector machines: Methods, theory and algorithms Boston Kluwer Academic Publishers 205 p.
Wattarujeekrit T Shah P Collier N 2004 PASBio: Predicate-argument structures for event extraction in molecular biology BMC Bioinformatics 5 155 15494078
Marcus M 1994 The Penn Treebank: A revised corpus design for extracting predicate-argument structure. 1994 ARPA Human Language Technology Workshop; 1994 March; Princeton, New Jersey San Francisco Morgan Kaufmann
Surdeanu M Harabagiu S Williams J Aarseth P 2003 Using predicate-argument structures for information extraction Proceedings of the 41st Annual Meeting of the Association for Computational Linguistics; 2003; Sapporo, Japan. pp. 8 15
Mitchell TM 1997 Machine learning New York McGraw-Hill 414 p.
Bairoch A Apweiler R 2000 The SWISS-PROT protein sequence database and its supplement TrEMBL in 2000 Nucleic Acids Res 28 302 303 10592254
Pruitt KD Maglott DR 2001 RefSeq and LocusLink: NCBI gene-centered resources Nucleic Acids Res 29 137 140 11125071
Benson DA Karsch-Mizrachi I Lipman DJ Ostell J Wheeler DL 2004 GenBank: Update Nucleic Acids Res 32 D23 D26 14681350
Birney E Andrews TD Bevan P Caccamo M Chen Y 2004 An overview of Ensembl Genome Res 14 925 928 15078858
Grabowski PJ Black DL 2001 Alternative RNA splicing in the nervous system Prog Neurobiol 65 289 308 11473790
Yeo G Holste D Kreiman G Burge CB 2004 Variation in alternative splicing across human tissues Genome Biol 5 R74 15461793
Thanaraj TA Stamm S Clark F Riethoven JJ Le Texier V 2004 ASD: The Alternative Splicing Database Nucleic Acids Res 32 D64 D69 14681360
Xu Q Modrek B Lee C 2002 Genome-wide detection of tissue-specific alternative splicing in the human transcriptome Nucleic Acids Res 30 3754 3766 12202761
Mirnics K Pevsner J 2004 Progress in the use of microarray technology to study the neurobiology of disease Nat Neurosci 7 434 439 15114354
Zavolan M Kondo S Schonbach C Adachi J Hume DA 2003 Impact of alternative initiation, splicing, and termination on the diversity of the mRNA transcripts encoded by the mouse transcriptome Genome Res 13 1290 1300 12819126
Pisarra P Lupetti R Palumbo A Napolitano A Prota G 2000 Human melanocytes and melanomas express novel mRNA isoforms of the tyrosinase-related protein-2/DOPAchrome tautomerase gene: Molecular and functional characterization J Invest Dermatol 115 48 56 10886507
Lee C 2003 Generating consensus sequences from partial order multiple sequence alignment graphs Bioinformatics 19 999 1008 12761063
Birney E Andrews D Bevan P Caccamo M Cameron G 2004 Ensembl 2004 Nucleic Acids Res 32 D468 D470 14681459
Wheeler DL Church DM Edgar R Federhen S Helmberg W 2004 Database resources of the National Center for Biotechnology Information: Update Nucleic Acids Res 32 D35 D40 14681353
Schmid H 1994 Probabilistic part-of-speech tagging using decision trees Proceedings of the International Conference on New Methods in Language Processing; 1994 September.
Nigam K Lafferty J McCallum A 1999 Using maximum entropy for text classification IJCAI-99 Workshop on Machine Learning for Information Filtering. pp. 61–67 Available: http://www-ai.cs.uni-dortmund.de/EVENTS/IJCAI99-MLIF/papers (nigam.ps.gz). Accessed 26 May 2005.
Nigam K McCallum A Thrun S Mitchell T 2000 Text classification from labeled and unlabeled documents using EM Mach Learn 39 103 134
McCallum A Nigam K 1998 A comparison of event models for naive Bayes text classification Learning for text categorization: Papers from the AAAI Workshop 1998 July 27; Madison, Wisconsin. Technical Report WS-98–05. Menlo Park (California) AAAI Press
Tateisi Y Ohta T Tsujii J 2004 Annotation of predicate-argument structure on molecular biology text IJCNLP 2004 Workshop on Beyond Shallow Analysis; 2004; Hainan, China.
Klein D Manning CD 2003 Accurate unlexicalized parsing Proceedings of the 41st Annual Meeting of the Association for Computational Linguistics; 2003; Sapporo, Japan.
Klein D , Manning CD 2002 Fast exact inference with a factored model Neural Information Processing Systems Conference; 2002. Available: http://books.nips.cc/papers/files/nips15/CS01.pdf . Accessed 26 May 2005.
Holger V 2002 TIGERin—Grafische Eingabe von Suchenfragen in TIGERSearch [diploma thesis] Stuttgart Universität Stuttgart 81 p.
Mika S Rost B 2004 Protein names precisely peeled off free text Bioinformatics 20 I241 I247 15262805
Reiner A Yekutieli D Benjamini Y 2003 Identifying differentially expressed genes using false discovery rate controlling procedures Bioinformatics 19 368 375 12584122
|
16103899
|
PMC1183516
|
CC BY
|
2021-01-05 09:18:22
|
no
|
PLoS Comput Biol. 2005 Jun 24; 1(1):e10
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010010
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science 1610390010.1371/journal.pcbi.001001105-PLCB-RA-0059R2plcb-01-01-11Research ArticleNeuroscienceMammalsA Phenomenological Theory of Spatially Structured Local Synaptic Connectivity Theoretical Model of Synaptic ConnectivityAmirikian Bagrat Brain Sciences Center, Veterans Affairs Medical Center, Minneapolis, Minnesota, United States of America, and Department of Neuroscience, University of Minnesota, Minneapolis, Minnesota, United States of AmericaFriston Karl J EditorUniversity College London, United KingdomE-mail: [email protected] 2005 24 6 2005 1 1 e1122 3 2005 10 5 2005 Copyright: © 2005 Bagrat Amirikian.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.The structure of local synaptic circuits is the key to understanding cortical function and how neuronal functional modules such as cortical columns are formed. The central problem in deciphering cortical microcircuits is the quantification of synaptic connectivity between neuron pairs. I present a theoretical model that accounts for the axon and dendrite morphologies of pre- and postsynaptic cells and provides the average number of synaptic contacts formed between them as a function of their relative locations in three-dimensional space. An important aspect of the current approach is the representation of a complex structure of an axonal/dendritic arbor as a superposition of basic structures—synaptic clouds. Each cloud has three structural parameters that can be directly estimated from two-dimensional drawings of the underlying arbor. Using empirical data available in literature, I applied this theory to three morphologically different types of cell pairs. I found that, within a wide range of cell separations, the theory is in very good agreement with empirical data on (i) axonal–dendritic contacts of pyramidal cells and (ii) somatic synapses formed by the axons of inhibitory interneurons. Since for many types of neurons plane arborization drawings are available from literature, this theory can provide a practical means for quantitatively deriving local synaptic circuits based on the actual observed densities of specific types of neurons and their morphologies. It can also have significant implications for computational models of cortical networks by making it possible to wire up simulated neural networks in a realistic fashion.
Synopsis
Each neuron communicates signals via synaptic connections simultaneously to several hundreds of neighboring neurons forming a synaptic circuit. Determining the pattern of synaptic connections between local neurons is crucial for understanding a specific cortical function implemented by a synaptic circuit. The connectivity between a pair of neurons is affected by their axonal/dendritic morphologies and relative spatial locations. Although neuroscientists have precise tools to measure neuronal activity caused by the flow of signals between circuit neurons, there are still considerable difficulties in the direct experimental measurement of local synaptic connectivity, which actually determines the underlying activity. This paper presents a theoretical approach to synaptic connectivity accounting for the morphologies of pre- and postsynaptic neurons and providing the average number of synaptic contacts formed between them as a function of their relative locations. An important aspect is the decomposition of the complex structure of an axonal/dendritic arbor into a small number of basic structures. The theory is in very good agreement, within a wide range of cell separations, with empirical data on axonal–dendritic contacts of pyramidal cells and somatic synapses formed by the axons of inhibitory interneurons. The current approach can provide a practical means for quantitatively deriving local synaptic circuits based on the actual observed densities of specific types of neurons and their morphologies.
Citation:Amirikian B (2005) A phenomenological theory of spatially structured local synaptic connectivity. PLoS Comput Biol 1(1): e11.
==== Body
Introduction
Unraveling intrinsic cortical circuitry—the pattern of synaptic connections between neurons within a local region—still is one of the most difficult challenges faced by researchers in neuroscience. The structure of intrinsic circuitry is the key to understanding cortical function and how neuronal functional modules such as cortical columns are formed [1–9]. The central obstacle to progress is the quantification of synaptic connectivity among different types of cells. The connectivity between a pair of pre- and postsynaptic neurons is affected by their axonal and dendritic morphologies, and their relative spatial locations. While neocortical neurons have diverse morphology, the number of basic morphologically distinct types of neurons is only of an order of 101 [10], and it is much smaller than the number of neurons themselves, which is of an order of 1010 for the human cortex [2]. This fact engenders an intuition that the problem of deciphering local cortical circuitry in such a huge network could be theoretically tractable. Indeed, the number of possible morphologically different types of pre- and postsynaptic cell pairs that it might be necessary to consider is relatively small, of an order of 101 × 101 = 102. Furthermore, instead of treating each cell pair individually, synaptic connectivity can be averaged over a whole ensemble of pairs in which all pre- and postsynaptic neurons have their respective underlying morphologies. An orderly relationship found by Sholl [11] describing apparently random branching patterns of dendrites of individual pyramidal and stellate cells inaugurated the idea of a statistical approach to the problem of synaptic connectivity, and motivated subsequent experimental and theoretical studies that have been carried out in this spirit.
To form a synaptic contact it is necessary that the presynaptic axon comes to a close spatial apposition to a certain cellular site (dendrite, soma, axon hillock, etc.) located on the postsynaptic neuron, establishing a physical contact. For the purpose of synaptic connectivity, it is useful to distinguish morphological features of neurons at large and small spatial scales. Large-scale features, such as the characteristic shape and size of a volume occupied by axonal/dendritic ramifications, set the limits of spatial separation between a pair of neurons within which they can potentially establish physical contacts and thus affect how the synaptic connectivity changes as a function of their relative positions.
Small-scale morphological features, such as the length and local curvature of axonal branches, can reveal cellular site specificity of synaptic contacts. The question of whether synaptic connectivity is random or specific has been addressed in several studies [3,6,12–18]. It has been noticed that small-scale features of the axons of most neurons (70%–80%), which are excitatory pyramidal cells [2], have a remarkable property: axon branches usually extend as straight lines for fairly long distances. This, as well as other observations, supports the idea that pyramidal cell axons make contacts in a nonspecific fashion, at random sites on the postsynaptic cell dendrites, when they are encountered by chance along the axon path [15,19].
Morphological properties of the remaining (20%–30%) neocortical cells, which are mostly inhibitory interneurons [2], are more intricate. It has been shown [18] that their axons have significantly smaller branch lengths, and the trajectories are considerably curved. Such geometrical characteristics suggest [18] that inhibitory interneuron axons are attracted to specific targets scattered around, in contrast to pyramidal cell axons, which shoot straight through neuropil for long distances. Indeed, interneurons can be divided into several types on the basis of what specific cellular sites on the postsynaptic neuron their axons are targeting; for example, there are dendrite- and tuft-targeting, dendrite-targeting, proximal dendrite- and soma-targeting, and axon-targeting interneurons [10].
Quantitative studies of synaptic connectivity are usually restricted to the case of nonspecific connections, when geometries of axons and dendrites are assumed to be mutually independent [15,19–23]. They can be divided basically into two categories. Studies in the first category make a number of specific assumptions about the morphologies of pre- and postsynaptic cells and then estimate synaptic connectivity between them [19–21,24]. Although these assumptions are relatively loosely related to the actual morphological structures of cells and to a certain extent are dictated by the convenience of analytical considerations, studies using this kind of approach offer a general framework for the treatment of the problem of synaptic connectivity and provide a broad view of the underlying picture. Studies in the second category, instead of making series of assumptions, use actual three-dimensional (3D) ramification patterns of reconstructed neurons and then estimate synaptic connectivity, supposing that the observed morphologies of presynaptic axons and postsynaptic dendrites are independent [22,23]. Although these kinds of studies are labor intensive, and provide a narrower view, being limited to a specific type of pre- and postsynaptic pair, they are quantitatively much more precise since the morphologies of real neurons are used for the prediction of connectivity.
In this paper I present a theoretical model of synaptic connectivity that strives to bring the strong aspects of each of these two different approaches into a single framework. For example, it takes into consideration large-scale morphological features of pre- and postsynaptic cells. However, it does not require 3D reconstructions of neurons; the necessary structural parameters of the underlying arbors can be directly estimated from plane, two-dimensional (2D) drawings of axons and dendrites. Importantly, such drawings are already available in literature for many different types of neurons. On the other hand, this theory does not explicitly consider small-scale morphological features of arbors but rather, in the spirit of several previous studies [15,19,25], introduces the notion of a synaptic density field. Making reasonable assumptions about the general structure of synaptic fields of axons and dendrites, I derived an expression for the average number of synaptic contacts formed between a pair of cells of given morphological types as a function of their relative locations in 3D space. To evaluate the impact of these assumptions, the theoretical number of contacts was compared with the number of contacts estimated empirically. I found that, within a wide range of cell separations, the theory is in very good agreement with empirical data on (i) axonal–dendritic contacts of pyramidal cells [22,23] and (ii) somatic synapses formed by the axons of inhibitory interneurons [26].
Results
Theoretical Framework
Synaptic connectivity between cortical neurons takes place at short-range (<103 μm) and long-range (>103 μm) scales [2,3]. The former is due to the local—with respect to the cell soma—ramifications of axonal and dendritic processes, whereas the latter is mediated by horizontally running axons and dendrites in cortical layers, and by axons of pyramidal cells leaving the cortex for the white matter and reentering it at distant locations. The present consideration deals only with local intracortical interactions. My approach is inspired by several previous quantitative studies of cortical connectivity [3,15,20,21,23,25].
Synaptic density field.
Suppose that one is able to record the spatial position of each synapse formed by the axon branches of a given presynaptic cell belonging to a particular morphological type μ
A. Let us mark all recorded synaptic sites by dots and place this cell into a 3D coordinate system, so that its soma is at the origin of the system. The presynaptic sites will then form a dispersed cloud of dots distributed in a certain way relative to the origin. Let us now pick up another nearby cell of the same type μ
A, record the spatial positions of all synapses formed by the axon branches of that cell, and then place it into the same coordinate system, again aligning its soma with the origin. Since the distribution of synaptic sites of the second cell is unlikely to be exactly the same as that of the first one, the number of dots would nearly double. However, given that both cells are of the same morphological type and are from the same cortical neighborhood, one would expect that the two dot patterns would be similar. These hypothetical experiments can be repeated many times. As the number of contributing cells increases, the dot clouds will become more and more dense, leading to a formation of a certain structure.
One can describe this structure in terms of the volume density of dots—synaptic sites—averaged over a large number of contributing cells of the same morphological type μ
A. Such ensemble averaging defines the synaptic density field
as the expected density of synapses formed by the axon ramifications of a single cell at spatial location r relative to the cell soma. Specifically, consider an element of volume ΔV with coordinate r. Let
be the number of synaptic sites on the axon branches of cell i (i = 1, 2,…n
A) within the volume ΔV. The contribution of neuron i to the synaptic density field at location r is then given by
. Correspondingly, the sum of individual contributions from the entire ensemble yields the underlying density field
. In the framework of this approach, observed spatial distributions of synapses of different cells, even if belonging to the same morphological type, are random realizations of a certain structure determined by the underlying morphology of axons μ
A. Conversely, synaptic density field
is the average of all possible realizations of the distributions of synaptic sites of individual cells:
where 〈·〉 indicates ensemble averaging.
The consideration above can be carried out, likewise, for synapses formed on the dendrite branches of postsynaptic cells. This would result in a corresponding synaptic density field
determined by the dendrite morphology μ
D.
Synaptic contacts between a pair of cells.
Consider now a pair of cells separated by a displacement vector d pointing from the soma of the presynaptic cell with the axon morphology μ
A to the soma of the postsynaptic cell with the dendrite morphology μ
D. Assume that the locations of all synaptic sites (including synapses formed with other neurons) on the presynaptic cell axon and the postsynaptic cell dendrite are recorded. Let us align the origin of the coordinate system with the soma of the presynaptic cell. Then the postsynaptic cell coordinate is d. One may pick up now another nearby pre- and postsynaptic cell pair composed of the same types of neurons separated by the same displacement d, record spatial positions of all synaptic sites on the presynaptic axon and postsynaptic dendrite, and place the pair into the same coordinate system. In the spirit of the single cell consideration above, this procedure can be repeated many times, giving rise to a large ensemble of contributing cell pairs. In such pairs, all presynaptic cells belong to the morphological type μ
A. Likewise, all postsynaptic cells belong to their own morphological type, μ
D. For a given cell pair i (i = 1, 2,…n), the number of synaptic sites on the axon branches of the presynaptic cell in the element of volume ΔV at location r is given by
, whereas the number of synaptic sites on the dendrite branches of the postsynaptic cell at the same location is given by
. It is assumed that ΔV is small enough, so that within this volume each individual cell may have at most one synaptic site, i.e.,
and
are either zero or one. If cells in the pair form a synaptic contact in ΔV, then the synaptic sites of both cells are present in this volume, i.e.,
, and share a common spatial location. The opposite, however, is not always true. Specifically, if both cells in the pair do have synaptic sites in ΔV, it does not necessarily mean that there is a synapse between them. Indeed, these sites could form synapses with other cells that have axonal or dendritic ramifications in ΔV. Let
be the proportion of pairs that form a synaptic contact in ΔV relative to the total number of pairs in which both cells have synaptic sites in the volume ΔV. The ensemble average of the volume density of synaptic contacts at location r between cell pairs separated by displacement d is then given by
In the framework of this approach,
and
can be considered as random variables, whereas
can be treated as the probability of synaptic contact at location r between pre- and postsynaptic cells given that each of them has a synaptic site in the element of volume ΔV.
I now make two simplifying assumptions. First, I assume that
and
are distributed independently and, therefore, the average of their product is equal to the product of their averages:
This means that, at any given spatial location r, the occurrences of synaptic sites on the axonal and dendritic ramifications of, correspondingly, the pre- and postsynaptic cells are independent from each other. This is true when the interaction (repulsion or attraction) forces between the axons and dendrites are relatively small and could be neglected. Second, I assume that
, the probability of forming a synaptic contact, is determined by the likelihood of spatial proximity of the pre- and postsynaptic branches in ΔV, and that the rate of synapse formation (given that such a close spatial apposition has been established) is constant along the axon and dendrite branches and is independent of the type of morphologies μ
A and μ
D. This can be formally expressed as
where δ is a free parameter representing the characteristic volume to be shared by the axon and dendrite segments in order to form a synapse.
Using equations 3, 4, and 1, the average density of synaptic contacts
given by equation 2 can be then written as
Integrating equation 5 over the entire space one can obtain the average number of synaptic contacts between a pair of cells as a function of the displacement d:
I assume that at any given spatial location, the ensemble distribution of the number of synaptic contacts within an element of volume ΔV, in which
could be considered constant, follows a Poisson distribution with the mean
. Assuming that the distributions at different spatial locations are independent from each other (see, however, [23]), the total number of synaptic contacts, which is the sum of the local contributions, will also be distributed according to Poisson, with the mean
, as given by equation 6. Although observed distributions of the number of synapses per connection could be non-Poisson [27,28], for the purpose of the present consideration a Poisson distribution is a reasonably good approximation, and has been used in previous studies of synaptic connectivity [21,23].
Structure of Synaptic Density Fields
The evaluation of the average number of synaptic contacts
requires knowledge of the synaptic density fields
and
. To make the problem theoretically tractable, I introduce a number of simplifying assumptions about the general structure of
, where T designates either an axonal (T = A) or dendritic (T = D) tree.
Assumption 1.
Synaptic density fields have cylindrical symmetry. The axis of symmetry traverses the cell soma and is oriented vertically, orthogonal to the cortical layers. Thus, the density is isotropic in the horizontal dimension, parallel to the layers. This assumption is motivated by the laminated structure of the cortex [2]. Indeed, although along the vertical dimension, across the layers, cortical physical properties are heterogeneous and may change dramatically (density of cells, type of cells, etc.), in the horizontal dimension, within a local region, the properties are much more homogeneous and isotropic. Correspondingly, while axonal and dendritic arborizations of experimentally reconstructed cells may exhibit certain anisotropy in the horizontal dimension [26], I assume that apparent asymmetries shown by individual cells are averaged out when a large ensemble of cells of the same morphological type is considered, so that the ensemble averaged distribution of synapses is cylindrically symmetrical.
Assumption 2.
The arborization structure of a given morphological type μT can be broken down into a number of basic structures—elementary clouds. The center
of an individual cloud
is located on the axis of symmetry at a certain distance from the cell soma. Accordingly, the synaptic density field representing the arborization μT is a superposition of the density fields
of the constituting clouds:
Likewise, the average total number of synaptic contacts between a pair of cells is the sum of the contributions
from individual axonal–dendritic cloud pairs:
where
(cf. equation 6). This assumption is based on observations that axonal and dendritic morphologies of cortical cells often show distinct branching patterns that are well segregated in space. Figure 1 exemplifies this point. For example, a drawing of the axonal arbor [22] that is typical for layer 2 (L2) pyramidal neurons illustrates that in addition to a dense arborization of collaterals around the cell body, the axons also ramify extensively in deeper layers of the cortex (Figure 1B). It is natural, therefore, to describe the axonal arborization structure of these neurons as a superposition of two distinct clouds.
Figure 1 Decomposition of the Complex Structure of Arbors into Elementary Synaptic Clouds
Synaptic density field of each cloud is illustrated by a set of concentric ellipsoids of different weights. An ellipsoid represents the equal-synaptic-density surface, whereas its weight represents the magnitude of the density. The outer ellipsoid, in addition, encloses the spatial extent of cloud ramifications. Yellow dots depict cell somata. The horizontal ℓ|| and vertical ℓ⊥ dimensions of one of the clouds as well as the displacement r
0 of its center from the soma are shown.
(A) A drawing of the dendritic arbor typical for L3 pyramidal neurons.
(B) A drawing of the axonal arbor typical for L2 pyramidal neurons.
The drawings of arbors are based on data representations in [22] by kind permission of B. Hellwig.
Assumption 3.
Equal synaptic-density surfaces of a given elementary cloud form a continuum of concentric, similar ellipsoids that are aligned at the cloud center. This assumption is motivated by the observation that the contours of the spatial spread of axonal and dendritic clouds often have an ellipsoidal shape (Figure 1). Because of the assumption of cylindrical symmetry, the ellipsoids revolve around the axis of symmetry. The shape of ellipsoids, oblate or prolate, depends on the actual spatial pattern of the ramifications. For example, the shape of the axonal clouds shown in Figure 1B can be approximated by oblate ellipsoids. On the other hand, the oblique apical dendrites of pyramidal cells can be well approximated by prolate ellipsoids, and the basal dendrites by spheres (Figure 1A).
Assumption 4.
Synaptic density falls off exponentially along the longitudinal and transverse axes of the ellipsoids. Specifically, I assume that the elementary cloud density field is given by
where
is the synaptic density at the center of the cloud
;
and
, correspondingly, are the longitudinal (parallel to the cortical layers) and transverse (perpendicular to the layers) components of the vector
originating from the cloud center; and finally,
and
are the longitudinal and transverse space constants characterizing the rate of the density decay in the horizontal and vertical dimensions, respectively.
Such a choice of the density field function is motivated by the work of Sholl [11], who has measured the number of dendritic processes crossing the unit area on a sphere centered at the cell soma. Sholl has shown that for the dendritic systems of the stellate and pyramidal neurons in the striate and motor areas of the cat this number decays exponentially as a function of the sphere radius. Assuming that synapses are distributed uniformly randomly along the cell processes [15,29,30] the exponential law would also be applicable to the density of synapses received by an individual cell. I further postulate that a similar relationship holds for axonal processes.
Given that synaptic density fields are determined by equation 10, one can evaluate from equation 9 the average number of synaptic contacts between a specific pair of axonal
and dendritic
clouds. To simplify the calculations, it is convenient to introduce dimensionless variables:
where
and
are, correspondingly, the longitudinal and transverse components of the displacement vector
connecting the cloud centers. In general, the triple integral (equation 9) cannot be evaluated analytically. It can be reduced, however, to a one-dimensional integral (see Materials and Methods) that is easy to evaluate numerically:
where J
0 (x) is the Bessel function of order zero. In a special case, when
, i.e., the interacting axonal and dendritic clouds have similar shapes (not necessarily spherical), the integral (equation 9) can be evaluated analytically (see Materials and Methods):
where
.
Comparison with Experiments
The central point of the present theory is to reduce the detailed picture of the complex branching patterns of axonal/dendritic processes and the locations of individual synapses along them to a simple but adequate (for the explanation of local synaptic connectivity) representation that is described by a small number of phenomenological parameters. To that end, the synaptic field of a given arbor is represented as a superposition of the synaptic fields of elementary clouds. For each elementary cloud
there are three structural parameters (
and
, characterizing the spatial spread of the cloud in the horizontal and vertical dimensions, respectively, and
, defining the displacement of the cloud center from the cell soma) and one magnitude-scaling parameter (
, describing the synaptic density at the cloud center). The impact of the underlying simplifying assumptions on the capacity of the theoretical model to quantify cortical synaptic connectivity can be assessed by comparing the predictions of the theory with experimental data. Below, using quantitative studies of synaptic connectivity carried out elsewhere, I consider three examples revealing the adequacy of the theory and illustrating how the above parameters can be evaluated from experimental data. Although in these studies only physical, not synaptic, contacts were directly [26] or indirectly [22,23] estimated from experimental data, they all presumed that physical connectivity is a good approximation of synaptic connectivity if the formation of a synaptic contact from an already established physical contact is a random and nonspecific process with a certain fixed rate.
Example 1: Pyramidal neurons in L2 and L3 of rat visual cortex.
Here I exploit Hellwig's work [22] in which axonal and dendritic arborizations of four L2 and four L3 pyramidal neurons of rat visual cortex were 3D reconstructed with the aim of estimating local connectivity from morphologies μ
A and μ
D of pre- and postsynaptic cells. This aim was achieved by counting the number of physical contacts between axonal and dendritic branches in a pair of reconstructed cells positioned at a certain distance from each other. The distance was varied by shifting the postsynaptic cell along an axis parallel to the coronal plane and the cortical surface, beginning from the maximum overlapping position in which a separation
between the cell somata along the shifting axis was zero to a position in which
was 500 μm.
Based on the layer of origin of the cell soma, Hellwig distinguished two different types of axons and two different types of dendrites. I designate them as P2A and P3A for the axons and P2D and P3D for the dendrites of neurons in L2 and L3, respectively (Figure 2A–2D). Thus, given that μ
A = {P2A, P3A} and μ
D = {P2D, P3D}, there can be four different types of pre- and postsynaptic cell pairs, resulting in four different types of axonal–dendritic connections μ
A → μ
D.
Figure 2 Connectivity between Pyramidal Neurons in L2/L3 of Rat Visual Cortex
(A–D) Images representing the average structures of axons (A and B) and dendrites (C and D) of pyramidal neurons originating from L2 (A and C) and L3 (B and D). Yellow dots depict cell somata. Ellipsoids capture the spatial extent of the synaptic clouds identified from these images. The dimensions
and the displacement
μ of each cloud were measured as illustrated in Figure 1B. The images were created using dendritic and axonal arborization drawings based on data representations in [22] by kind permission of B. Hellwig.
(E–H) Average number of contacts between pre- and postsynaptic neurons as a function of the distance between them. The type of axonal–dendritic connection is shown on each plot. Empirical curves [22] are plotted in black. Fitted theoretical curves are plotted in blue and predicted curves are plotted in red. Dots show stochastic variations in the theoretical number of contacts.
For each type of axon and dendrite there were four 3D reconstructions. By averaging over 32 possible combinations of pre- and postsynaptic cell pairs of the same μ
A → μ
D type separated by a given distance (4 axons × 4 dendrites × 2 positions along the shifting axis resulting in the same separation), Hellwig [22] obtained empirical relationships
describing the average number of contacts between pyramidal neurons in L2 and L3 rat visual cortex as a function of horizontal cell separation
(Figure 2E–2H, black curves). To understand how well the present theoretical model can capture these empirical relationships, I adopted the following procedure.
First, I visualized average spatial structures of the reconstructed axons and dendrites. To that end, individual arborization drawings based on data representations in [22] were mirror reflected about the vertical axis. Then, the original and reflected drawings for all four arbors of a given type μT (μT = {P2A, P3A, P2D, P3D}) were laid on top of each other so that the positions of all somata were aligned. The resulting image represented the average spatial structure of the underlying arbor type μT (Figure 2A–2D). Note that this procedure made the average arbor images obtained from a relatively small experimental sample symmetric and implemented the assumption of the theory that synaptic density fields have cylindrical symmetry.
Second, visually inspecting these images, I identified elementary clouds of axons and dendrites, and enclosed them in distinct ellipses capturing the spatial extent of individual cloud ramifications (Figure 2A–2D). The morphologies of both P2D and P3D dendrites were described by two clouds corresponding to basal and apical ramifications represented by a sphere and an oblate ellipsoid, respectively (Figure 2C and 2D). The structures of P2A and P3A axons were described by single clouds corresponding to a dense arborization of collaterals around the cell body and were represented by oblate ellipsoids (Figure 2A and 2B). The clouds formed by the extensive branching of P2A and P3A axons at deeper layers were disregarded in the present consideration. The point is that they are well separated from the P2D and P3D dendrites that ramify in the upper layers and, therefore, their contribution to the connectivity between L2/L3 pyramidal neurons is negligible. Altogether, six distinct clouds were used to describe the two types of neuron morphologies. For each individual cloud
(i = 1 for μ
A and i = 1,2 for μ
D) the horizontal,
, and vertical,
, semi-axes of the corresponding ellipse as well as the position of its center relative to the cell soma,
, were estimated from the drawings in Figure 2A–2D (see also Figure 1B).
Third, these measurements were linked to the parameters of the theory. Specifically, I assumed that the space constants of a given cloud
are proportional to the lengths of the semi-axes of the corresponding ellipse:
and
, where γ is a certain dimensionless constant common to all axonal and dendritic clouds.
defined the displacement of the cloud center from the cell soma along the axis of symmetry. In addition, it was assumed that the synaptic densities at the cloud centers were the same for all four dendritic clouds and were described by a single parameter,
. Similarly, the synaptic densities at both axonal cloud centers were assumed equal and were described by a single parameter,
. Furthermore, since
,
, and δ all enter only as their triple product into equation 12, defining the average number of contacts between a pair of axonal and dendritic clouds, they were merged into one parameter
. The simplifying assumptions above imply that κ is a constant common to all types of axonal–dendritic cloud pairs. Thus, once the dimensions
and
and positions
of individual clouds were estimated and fixed, the number of free parameters of the theory was effectively reduced to just two—γ and κ—that uniformly scaled, respectively, the spatial constants and the local magnitude of the synaptic density fields of all clouds.
Fourth, the parameters γ and κ were estimated by fitting theoretical curves for the average number of contacts
into corresponding empirical relationships
obtained by Hellwig [22].
was calculated using the superposition principle for the synaptic density fields of the underlying clouds (assumption 2) and equations 11 and 12, defining the average number of contacts between individual cloud pairs. Two empirical curves,
and
, describing the intra-layer connections, were used to derive γ and κ. The resulting best fit (least squares) estimates were
and
μm−3. Figure 2E and 2H shows the fitted (blue) and empirical (black) curves.
Finally, predictions of the theory were compared against independent experimental data. Specifically, the two remaining empirical curves,
and
, describing the inter-layer connections, were compared with the corresponding theoretical curves
and
that were calculated using
and
estimated from data on intra-layer connections. Figure 2F and 2G shows the predicted (red) and empirical (black) curves. To facilitate the comparison, theoretical data were also generated in the same representation as original experimental data [22] from which empirical relationships
were obtained. Specifically, at each 1-μm increment of separation
, the number of contacts was drawn from a Poisson distribution with the mean
(see Theoretical Framework) independently 32 times, corresponding to 32 possible combinations of the reconstructed pre- and postsynaptic cell pairs of the same μ
A → μ
D type separated by the distance
in [22], and then averaged and plotted in Figure 2E–2H as a dot.
One can see that overall there is very good agreement between the theory and experiment. Note that a single fixed pair of parameters
and
quantitatively explain a set of four different types of connections between two morphologically distinct types of pyramidal neurons. The theoretical average number of contacts matches fairly well to the experimentally determined ones in all four plots within the entire range of cell separation
explored in [22]. In addition, the stochastic variations in the theoretical number of contacts (dots) are similar to variations seen in the corresponding experimental plots (cf. Figure 8A–8D of [22]).
Example 2: Clutch cells in L4 of cat visual cortex.
In this example I use experimental data obtained by Kisvárday and colleagues [26,31], who studied local connections of clutch (a type of basket) cells in L4 of cat visual cortex. Most synaptic contacts formed by the axons of these types of inhibitory neurons are positioned at the postsynaptic cell soma and proximal dendrites [32]. Budd and Kisvárday [26] carried out a quantitative analysis in which they examined only the somatic connections. Based on an electron microscopy study [31], they assumed that all neuron somata (NS) opposed to boutons of clutch cell axons (CA) are contacted synaptically. Using previously 3D reconstructed axons of two clutch cells and recorded spatial locations of somata contacted by the axonal branches [31], they estimated the number of somatic connections CA → NS made by the individual clutch cell axon as a function of the radial distance R from the cell body. This was done simply by counting the number of contacted somata within a vertical cylindrical shell of a given radius R and a fixed width ΔR, centered at the clutch cell body, and traversing the entire depth of L4. The resulting two radial distributions, one for each cell, had very similar profiles (cf. Figure 3A and 3B of [26]), although the total numbers of counted postsynaptic somata were different (434 and 311).
Figure 3 Somatic Connections of Clutch Cells in L4 of Cat Visual Cortex
(A) Radial distribution of the average number of postsynaptic somata contacted by the axon. Dots with drop-lines show empirical distribution obtained by pooling data from the two cells [26]. Bars show theoretical distribution.
(B) Image representing the average structure of the clutch cell axon. Yellow dot depicts cell soma. The ellipsoid captures the spatial extent of the synaptic cloud identified from the image. The dimensions
and the displacement
of the cloud were measured as illustrated in Figure 1B. The image was created using axonal arborization drawings based on data representations in [31] by kind permission of Z. Kisvárday.
To compare these experimental data with the theory, which provides ensemble averaged quantities, I first pooled data from the two cells and obtained the average observed radial distribution
of somatic connections made by the clutch cell axons (Figure 3A, dots with drop-lines). The corresponding theoretical distribution
was calculated in the following way.
First, utilizing the assumption of cylindrical symmetry, an image representing the average spatial structure of the clutch cell axon was obtained. Specifically, individual drawings based on data representations in [31], depicting the projections of the 3D reconstructed axon on two (nearly frontal and sagittal) planes, were mirror reflected about the vertical axis. The underlying average image was then obtained by overlaying, as in example 1, the original and reflected drawings (Figure 3B).
Second, based on this image, the morphology of the clutch cell axon was described by a single cloud. The horizontal
and vertical
semi-axes of the corresponding enclosing ellipse were estimated from Figure 3B (see also Figure 1B). Since the analysis in Budd and Kisvárday [26] was restricted to L4 somatic connections only, the contribution from descending axonal branches projecting to deeper layers was disregarded.
Third, as in example 1, it was assumed that the axonal field space constants are proportional to the dimensions of the enclosing ellipse:
and
. Axonal–somatic connections were described using the same formalism as in the case of axonal–dendritic connections. Given that the volume of the postsynaptic soma is much smaller than the volume occupied by the clutch cell axonal ramifications, I assumed that the somatic field space constants
and
are small, such that
and
. In this case, the integrand in equation 12 can be expanded in the Taylor's series about
and
; retaining only the free term, I obtained the average number of contacts
made by the clutch cell axon with the cell soma located at
, where
is a dimensionless parameter, and
and
are the synaptic densities at the corresponding cloud centers. Integrating
over the volume of a cylindrical shell of radius R and width ΔR, I obtained the underlying radial distribution of somatic connections:
where K
2(x) is the modified Bessel function of order 2, and g
N is the neuronal density in L4, which was set to 5.4 × 104 mm−3 (cf. [26]). The theoretical distribution is parameterized by ζ and γ. The former scales the overall amplitude of the distribution whereas the latter uniformly scales the spatial constants
and
, which, in turn, shape the radial profile of the distribution.
Finally, the parameters γ and ζ were varied in order to bring the theoretical distribution
into a correspondence with the experimentally obtained distribution
. The values of the adjusted parameters were
and
; the resulting theoretical distribution is shown in Figure 3A (bars). One can see that the theory captures the features of the experimental distribution adequately. Particularly, the profile of
matches very well with the profile of
in the whole range of R. Note also that the value of
derived for the somatic connections of the clutch cell axons is very close to the value of
derived for the connections between the pyramidal neurons in L2/L3.
Example 3: Pyramidal neurons in L5 of rat somatosensory cortex.
This example, illustrated in Figure 4, relies on work of Markram and colleagues [23] in which physical connectivity between pyramidal neurons in L5 of rat somatosensory cortex was estimated based on the 3D reconstructed morphology of 11 axons (P5A) and 14 dendrites (P5D). The key idea of their approach is that statistics of cell arbors could be used to estimate the average number of P5A → P5D axonal–dendritic contacts formed by a pair of neurons as a function of their relative locations. Unlike earlier work of Hellwig [22], considered in example 1, who explicitly averaged the number of physical contacts over pairs of reconstructed axons and dendrites positioned at a given relative distance, Kalisman et al. [23] first averaged the geometry of ramifications over reconstructions of single arbors, separately for the axons and dendrites, to obtain two maps, called the axonal and dendritic templates. Then, using these empirical templates, the connectivity map
—the average number of contacts formed by a pair of neurons as a function of their horizontal,
, and vertical,
, separation—was estimated (Figure 4A).
Figure 4 Connectivity between Pyramidal Neurons in L5 of Rat Somatosensory Cortex
(A and B) Connectivity map showing the average number of contacts formed between the presynaptic cell positioned at the origin and the postsynaptic cell at location
: (A) empirical map adapted from data representations in [23] by kind permission of H. Markram and G. Silberberg; (B) theoretical map.
(C and D) Images representing the average structures of dendrites (C) and axons (D) of pyramidal neurons in L5. Yellow dots depict cell somata. Ellipsoids capture the spatial extent of the synaptic clouds identified from these images. The dimensions
and the displacement
of each cloud were measured as illustrated in Figure 1B. The images were created using dendritic and axonal arborization drawings based on data representations in [23] by kind permission of H. Markram and G. Silberberg.
To compare these results with my theoretical model, I first visualized, as in previous examples, the average spatial structures of the underlying arbors using drawings based on data representations in [23]. The resulting two images are shown in Figure 4C and 4D. The morphology of P5A axons was described by a single cloud corresponding to extensively branching collaterals below the cell soma and was represented by an oblate ellipsoid (Figure 4D). The structure of P5D dendrites was described by two clouds corresponding to basal and oblique apical ramifications represented by a sphere and a prolate ellipsoid, respectively (Figure 4C). The apical tuft dendrites that ramify in the upper cortical L1 and L2 were disregarded because, within the range of the vertical separation
explored in [23], the tuft dendrites nearly do not overlap with the axon collaterals and, therefore, their contribution to the connectivity can be neglected. Thus, altogether, three clouds were used to describe L5 pyramidal neuron morphology. For each cloud
(i = 1 for μ
A = P5A and i = 1,2 for μ
D = P5D) the horizontal,
, and vertical,
, semi-axes of the corresponding ellipse as well as the position of its center relative to the cell soma,
, were estimated from the drawings in Figure 4C and 4D (see also Figure 1B).
The theoretical average number of contacts
was calculated in the same way as explained in example 1. As before, I assumed that the space constants of individual clouds are given by
and
. Since the optimal values of the parameter γ derived in two previous examples were very close (0.207 and 0.219), in this case I simply set γ = 0.215. Thus κ (see example 1) was the only remaining free parameter. The value of κ = 2.10 × 10−5 μm−3 was determined by normalizing the amplitude of the peak in the connectivity map
to 4.8 contacts, in accordance with Kalisman et al. [23]. One can see that the resulting map, shown in Figure 4B, is in good quantitative agreement with the empirical map
, shown in Figure 4A. Particularly, the peak of theoretical connectivity occurred at
μm, when the postsynaptic neuron was located 100 μm below the presynaptic cell soma, as was also observed in the empirical map
. The connectivity at the map origin, which could be considered as the average number of contacts between the axon and dendrite of the same neuron, i.e., the average number of autaptic contacts, was 2.7. This is in good agreement with the 2.3 ± 0.9 autapses per neuron estimated by Lübke et al. [28] from detailed light and electron microscopy study of the same type of neurons, and very close to the 2.8 contacts reported by Kalisman et al. [23] for the empirical map
.
Discussion
In this work I proposed a simple theoretical model of local synaptic connectivity between a pair of cortical neurons that takes into account the morphological structure of axons and dendrites and the relative spatial locations of the pre- and postsynaptic somata. To understand the implications of the underlying simplifying assumptions, the theoretical number of synaptic contacts was compared with the number of contacts estimated empirically in quantitative studies of synaptic connectivity [22,23,26]. In these studies 2D drawings of arbors (necessary for extracting the phenomenological parameters of the theory) and empirically estimated numbers of contacts at various separations between pre- and postsynaptic cells (required for the comparison against the predictions of the theory) were both published. In all three examples considered there was very good agreement between the theory and experiment, within a wide range of pre- and postsynaptic cell separations.
The present approach relies on the assumption that the interactions between axons and dendrites are negligibly small and, therefore, their morphological properties can be treated independently. This is adequate, particularly, for the axons of pyramidal cells that form nonspecific axonal–dendritic contacts (examples 1 and 3). In addition, I demonstrated that the same formalism can be extended to the case of highly specific contacts such as somatic synapses formed by the axons of inhibitory interneurons (example 2), and thus the present approach, unlike the previously suggested method [23], is able to quantitatively predict this type of synaptic connectivity as well. These results suggest that the theoretical framework and the chosen functional form for the synaptic density fields of axons and dendrites effectively capture 3D morphologies of a variety of neurons (GABAergic and pyramidal, from different cortical layers, areas, and organisms) and describe the two types of synaptic connectivity (excitatory axonal–dendritic and inhibitory axonal–somatic) between cell pairs fairly well. It remains to be seen, however, whether this approach will be able to produce satisfactory results for other types of neurons.
An important aspect of the theoretical framework is the “linearization” of the complex structure of an axonal/dendritic arbor of a given morphological type μT, representing it as a linear combination of basic structures—elementary synaptic clouds (see assumption 2). It was demonstrated that by measuring the horizontal
and vertical
spread of individual cloud ramifications observed in the 2D drawings of the underlying arbors, one can estimate the corresponding space constants
and
, assuming a linear isotropic relationship between the physical
and characteristic
sizes of the cloud:
,
. The central result of this paper is that the scaling parameter γ providing the best correspondence between theory and experiment had a nearly constant value regardless of the morphological origin of synaptic clouds considered in the examples. Indeed, γ varied within a narrow range (0.207 < γ < 0.219) and, therefore, appears to be nearly independent of the type of originating arbor (axonal or dendritic), neuron (pyramidal or GABAergic), cortical layer (superficial L2/L3 or deep L5), cortical area (somatosensory or visual), and organism (rat or cat). As a result, it is tempting to think of γ as a kind of “universal” space calibration constant that translates the physical dimensions of any given synaptic cloud into the space constants describing properly the spatial distribution of the underlying cloud synaptic density. Further quantitative studies, however, are necessary to find out whether γ is truly invariant with respect to the whole multitude of diverse neuron morphologies observed in the cortex.
In the present consideration it was assumed that the parameter
, which defines the density at the center of a particular cloud
, is constant for all constituent clouds of the underlying arbor μT:
. In this case, the theoretically predicted average number of contacts between a pair of neurons is affected, in fact, by the product of two such parameters, one for the presynaptic and the other for the postsynaptic arbor. Therefore, regardless the number of pre- and postsynaptic clouds involved, as far as the synaptic connectivity is concerned, a single parameter incorporating this product (κ in examples 1 and 3; ζ in example 2) is what matters. It defines the amplitude scale of synaptic interactions (i.e., the peak number of contacts) between pairs of neurons, in contrast to γ, which defines the spatial scale of such interactions (i.e., how fast the number of contacts decreases as the separation between neurons increases). Unlike γ, which turned out to be nearly constant, the value of κ varied substantially in the examples considered. If one is interested in the absolute number of synaptic contacts, then this parameter should be calibrated for each morphologically distinct type of neuronal pair by comparing the predicted number of synaptic contacts with the experimentally measured one at known cell separations. The main value of the present theoretical framework, however, is in the determination of the relative scaling of the number of synaptic contacts between a pair of cells with their spatial separation, rather than the exact number of contacts.
It is noteworthy that although the number of 3D reconstructed neurons is growing, the existing empirical methods [22,23] for the estimation of synaptic connectivity, which are based on 3D reconstructions, are cumbersome. In contrast, the present theoretical framework, although less accurate than the methods presented in [22,23], provides a straightforward approach for estimating synaptic connectivity by (i) extracting the relevant structural parameters of axons and dendrites from 2D arborization drawings, and then (ii) plugging them into a compact analytical expression providing the number of synaptic contacts as a function of relative cell positions.
The significance of this work, however, goes beyond the derivation of an analytical expression describing synaptic connectivity between morphologically distinct neuronal pairs. For example, the present approach could be used for deciphering the structure of local synaptic circuitry (i.e., the pattern of connections between neurons) in a cortical region of interest. In particular, one could estimate the individual contributions from diverse types of neurons distributed across cortical layers to the net synaptic input received by a neuron of a given type μ
D. The number of synaptic contacts contributed by all presynaptic neurons of a particular type μ
A could be obtained by integrating the theoretical number of contacts formed between the pair μ
A→μ
D over the positions of the presynaptic neurons μ
A in the underlying cortical region. Note that such an approach accounts for both specific morphologies of neurons and spatial distributions of neurons. These are important factors for the quantification of local synaptic circuits because the densities of morphologically distinct types of neurons vary across cortical layers in a specific fashion [2]. Also, vertically oriented anatomical minicolumns, clearly visible in certain cortical areas [2], introduce an additional order in the cortical distribution of neurons. In general, such a local spatial ordering of neurons could further structuralize cortical circuitry and contribute to the formation of functional modules (such as cortical columns) with sharp borders [33]. The theoretical predictions of local connectivity patterns can be compared with independent quantitative experimental studies of cortical synaptic circuits.
Recently, Stepanyants, Hof, and Chklovskii [34] provided an elegant and insightful analysis of information storage capacity associated with local structural plasticity, without major remodeling of dendritic or axonal arbors. The key aspect of their approach is the calculation of the number of potential synapses that a given dendrite can form with axons passing within the spine length from the dendrite. The capacity for altering connectivity patterns through formation and elimination of synapses made by dendritic spines—the information storage capacity—was then characterized in terms of the filling fraction f—the ratio of actual to potential synapses.
The framework of potential synapses could be also used in different contexts, providing insights into different aspects of synaptic connectivity. In the present approach the specificity of synaptic connections is determined by geometrical factors such as the layout of axonal and dendritic branches and the relative spatial positions of pre- and postsynaptic cells. Can the specificity of synaptic connections go beyond the geometry, without major remodeling of dendritic or axonal arbors? This is possible if the number of potential synapses as defined in [34] is greater than the number of actual synapses. The specificity could be achieved by selecting the appropriate subset from the pool of potential synapses. In this interpretation, the filling fraction characterizes the capacity to form specific synaptic connections apart from the geometrical factors considered above. Using estimates of spine length from several brain areas, Stepanyants et al. [34] calculated the filling fraction f and found that the information capacity ranges from three to four bits per synapse of pyramidal neuron. In the context of synaptic specificity this implies that, on average, each presynaptic site can choose its postsynaptic partner roughly from three to four available sites, without major remodeling of axonal or dendritic arbors.
Is this potential for pyramidal neuron local synaptic specificity actually realized in the cortex? Until a short time ago, this was an open question. In a recent paper, Kalisman, Silberberg, and Markram [35], using confocal microscopy and whole-cell recordings from pairs and triplets of thick tufted L5 pyramidal neurons of rat somatosensory cortex, found that axons physically contact neighboring dendrites without any bias. This is consistent with the present theoretical model as well as previous studies [22,23]. The average number of axonal–dendritic touches between synaptically connected pairs of neurons was 6.6 ± 1.5. However, only 1.5 ± 0.3 of those touches were characterized as bouton–spine contacts (putative synapses). Special analysis carried out in [35] strongly suggested that the bouton–spine contacts were indeed synapses. Thus, Markram and colleagues [35] demonstrated that indeed only a small fraction of potential synaptic sites (touches) are realized as actual synapses (bouton–spine contacts). One can think that the conversion of a potential to actual synapse is a random, stochastic process, i.e., a given touch is transformed into a synapse with a certain probability. However, do these conversions of potential synapses occur independently from each other (i.e., uniformly randomly) or in a specifically coordinated fashion (i.e., nonuniformly randomly)? In a recent study, Chklovskii and colleagues [36] probed synaptic connections using quadruple whole-cell recordings from L5 pyramidal neurons in rat visual cortex. Statistical analysis of several hundred such simultaneous recordings revealed that reciprocal synaptic connections as well as several three-neuron connectivity patterns are more common than one would expect in uniformly randomly wired quadruplets. This study, therefore, suggests that presynaptic sites can select their partners from the pool of potential postsynaptic sites in a specific way. Fine-scale specificity has been also reported by Callaway and colleagues [37] in rat visual cortex for connections between adjacent pyramidal neurons in L2/L3 forming a group of selectively interconnected neurons that receive common excitatory input from L4.
Thus, the specificity in synaptic connectivity without major remodeling could occur at least at two levels. While the geometry of axons and dendrites and relative cell positions define the coarse level of specificity, recent work [35–37] suggest that the fine-tuning of synaptic connectivity in local microcircuits could be achieved by selecting an appropriate subset from the pool of potential synapses. The present theoretical framework considers the coarse specificity only. As soon as sufficient data become available, new quantitative models accounting for the fine-tuning of specificity in local cortical circuits should be developed.
In conclusion, the phenomenological approach to local synaptic connectivity described in this paper provides a remarkably simple way for extracting the relevant structural parameters of axons and dendrites from 2D arborization drawings. It was demonstrated that a crude approximation of axonal and dendritic arbors as a superposition of a set of ellipsoids is satisfactory for the purpose of quantitative estimation of synaptic connectivity between specific types of neurons as a function of their relative locations. Since for many types of neurons 2D drawings are available from literature, the present approach could be of principal significance for the practicality of deciphering synaptic microcircuits of a given cortical region based on the actual observed densities of specific types of neurons and their morphologies. It could also have significant implications for computational models of cortical networks by making it possible to wire up simulated neural networks in a realistic fashion.
Materials and Methods
Evaluation of the number of contacts between a pair of elementary clouds.
The aim of this section is to explain how I evaluated the average number of synaptic contacts
formed between the axonal cloud
and the dendritic cloud
separated by the displacement vector Δr connecting their centers. Let us place the origin of a Cartesian coordinate system at the center of the cloud
, so that the z-axis is parallel to the clouds' axes of cylindrical symmetry. Using synaptic density fields given by equation 10, and introducing the dimensionless variables (equation 11), equation 9 defining the average number of synaptic contacts can be rewritten as
where
Note now that the above integral is a convolution of two functions: I(s) = g * h, where
To evaluate I(s), one can first find its Fourier transform
I(s)e
−iksd3
s using the property that the transform of a convolution is the product of the Fourier transforms of the factors of the convolution:
. The transform of g(q) is given by (e.g., [38])
Next, note that
=
. Applying the scaling property of the Fourier transform, one obtains
I(s) can be then found by the inverse transformation:
Exploiting the symmetry present in the problem, let us change to conventional cylindrical coordinates:
The integral (equation 20) can be then rewritten as
where
and
(cf. equation 11). Integrating over ϕ, one obtains
where J
0(x) is the Bessel function of order zero (e.g., [38]). To integrate over z, one can use the method of contour integration (e.g., [39]) provided by the theorem of residues from the theory of functions of a complex variable. The integrand has two poles of order two in the upper-half of complex z-plane. Computing the residues at these poles and choosing the large semicircle in the upper-half plane as the contour of integration, one obtains the final result (equation 12).
Special case of similar shapes.
Consider a case
, in which interacting axonal and dendritic clouds have similar, but not necessarily spherical, shapes (cf. equation 11). Equation 20 then becomes
Using now the higher symmetry of this case, one can change to spherical coordinates:
and, after integrating over ϕ, obtain
where
Making the change of variable
, the double integral (equation 26) can be further simplified and reduced to a one-dimensional integral:
This last integral is evaluated using the same method of contour integration as in the general case considered above. Again, the integrand has two poles of order two in the upper-half of complex k-plane. Computing the residues at these poles and choosing the large semicircle in the upper-half plane as the contour of integration, one obtains equation 13.
I dedicate this paper to the memory of Alexander Lukashin, my teacher, friend, and an inspiring scientist.
I thank Apostolos P. Georgopoulos for his constant support, stimulating conversations, and encouragement. I am grateful to Bernhard Hellwig for his permission to use arborization drawings of L2 and L3 pyramidal neurons of rat visual cortex, Zoltán Kisvárday for his permission to use arborization drawings of clutch cells in L4 of cat visual cortex, and Henry Markram and Gilad Silberberg for their permission to use the connectivity map and arborization drawings of L5 pyramidal neurons of rat somatosensory cortex.
This work was supported by the United States Department of Veterans Affairs and the American Legion Brain Sciences Chair.
Competing interests. The author has declared that no competing interests exist.
Author contributions. BA developed the theoretical model and wrote the paper.
Abbreviations
2Dtwo-dimensional
3Dthree-dimensional
L[number]layer [number]
==== Refs
References
Mountcastle VB 1978 An organizing principle for cerebral function Edelman GM Mountcastle VB The mindful brain Cambridge MIT Press 7 50
Mountcastle VB 1998 Perceptual neuroscience: the cerebral cortex Cambridge Harvard University Press 486 p.
Abeles M 1991 Corticonics: Neural circuits of the cerebral cortex Cambridge Cambridge University Press 280 p.
Douglas RJ Martin KA 2004 Neuronal circuits of the neocortex Annu Rev Neurosci 27 419 451 15217339
Jones EG 2000 Microcolumns in the cerebral cortex Proc Natl Acad Sci U S A 97 5019 5021 10805761
Silberberg G Gupta A Markram H 2002 Stereotypy in neocortical microcircuits Trends Neurosci 25 227 230 11972952
Zhang LI Tan AY Schreiner CE Merzenich MM 2003 Topography and synaptic shaping of direction selectivity in primary auditory cortex Nature 424 201 205 12853959
Poggio T Bizzi E 2004 Generalization in vision and motor control Nature 431 768 774 15483597
Rockland KS Ichinohe N 2004 Some thoughts on cortical minicolumns Exp Brain Res 158 265 277 15365664
Markram H Toledo-Rodriguez M Wang Y Gupta A Silberberg G 2004 Interneurons of the neocortical inhibitory system Nat Rev Neurosci 5 793 807 15378039
Sholl DA 1953 The organization of the visual cortex in the cat J Anat Lond 89 33 46
Szentagothai J 1978 Specificity versus (quasi-) randomness in cortical connectivity Brazier MAB Petsche H Architectonics of the cerebral cortex New York Raven Press 77 97
Szentagothai J 1990 Specificity versus (quasi-) randomness revisited Acta Morphol Hung 38 159 167 2102598
White EL 1989 Cortical circuits: Synaptic organization of the cerebral cortex. Structure, function and theory Boston Birkhauser 223 p.
Braitenberg V Schüz A 1998 Cortex: Statistics and geometry of neuronal connectivity Berlin Springer 249 p.
Kozloski J Hamzei-Sichani F Yuste R 2001 Stereotyped position of local synaptic targets in neocortex Science 293 868 872 11486089
Anderson JC Binzegger T Douglas RJ Martin KA 2002 Chance or design? Some specific considerations concerning synaptic boutons in cat visual cortex J Neurocytol 31 211 229 12815241
Stepanyants A Tamas G Chklovskii DB 2004 Class-specific features of neuronal wiring Neuron 43 251 259 15260960
Braitenberg V 1978 Cortical architectonics: General and real Brazier MAB Petsche H Architectonics of the cerebral cortex New York Raven Press 443 465
Utley AM 1955 The probability of neural connexions Proc R Soc Lond B Biol Sci 144 229 240 13266808
Liley DT Wright JJ 1994 Intracortical connectivity of pyramidal and stellate cells: Estimates of synaptic densities and coupling symmetry Netw Comput Neural Syst 5 175 189
Hellwig B 2000 A quantitative analysis of the local connectivity between pyramidal neurons in layers 2/3 of the rat visual cortex Biol Cybern 82 111 121 10664098
Kalisman N Silberberg G Markram H 2003 Deriving physical connectivity from neuronal morphology Biol Cybern 88 210 218 12647228
Binzegger T Douglas RJ Martin KA 2004 A quantitative map of the circuit of cat primary visual cortex J Neurosci 24 8441 8453 15456817
Krone G Mallot H Palm G Schüz A 1986 Spatiotemporal receptive fields: A dynamical model derived from cortical architectonics Proc R Soc Lond B Biol Sci 226 421 444 2869496
Budd JM Kisvárday ZF 2001 Local lateral connectivity of inhibitory clutch cells in layer 4 of cat visual cortex (area 17) Exp Brain Res 140 245 250 11521157
Markram H Lübke J Frotscher M Roth A Sakmann B 1997 Physiology and anatomy of synaptic connections between thick tufted pyramidal neurones in the developing rat neocortex J Physiol 500 409 440 9147328
Lübke J Markram H Frotscher M Sakmann B 1996 Frequency and dendritic distribution of autapses established by layer 5 pyramidal neurons in the developing rat neocortex: Comparison with synaptic innervation of adjacent neurons of the same class J Neurosci 16 3209 3218 8627359
Hellwig B Schüz A Aertsen A 1994 Synapses on axon collaterals of pyramidal cells are spaced at random intervals: A Golgi study in the mouse cerebral cortex Biol Cybern 71 1 12 7519886
Karube F Kubota Y Kawaguchi Y 2004 Axon branching and synaptic bouton phenotypes in GABAergic nonpyramidal cell subtypes J Neurosci 24 2853 2865 15044524
Kisvárday ZF Martin KA Whitteridge D Somogyi P 1985 Synaptic connections of intracellularly filled clutch cells: A type of small basket cell in the visual cortex of the cat J Comp Neurol 241 111 137 4067011
Kisvárday ZF 1992 GABAergic networks of basket cells in the visual cortex Prog Brain Res 90 385 405 1631306
Ohki K Chung S Ch'ng YH Kara P Reid RC 2005 Functional imaging with cellular resolution reveals precise micro-architecture in visual cortex Nature 433 597 603 15660108
Stepanyants A Hof PR Chklovskii DB 2002 Geometry and structural plasticity of synaptic connectivity Neuron 34 275 288 11970869
Kalisman N Silberberg G Markram H 2005 The neocortical microcircuit as a tabula rasa Proc Natl Acad Sci U S A 102 880 885 15630093
Song S Sjöstrom PJ Reigl M Nelson S Chklovskii DB 2005 Highly nonrandom features of synaptic connectivity in local cortical circuits PLoS Biol 3 e68. DOI: 10.1371/journal.pbio.0030068 15737062
Yoshimura Y Dantzker JL Callaway EM 2005 Excitatory cortical neurons form fine-scale functional networks Nature 433 868 873 15729343
Gradshteyn IS Ryzhik IM 2000 Table of integrals, series, and products San Diego Academic Press 1163 1,163 p.
Mathews J Walker RL 1970 Mathematical methods of physics Menlo Park W. A. Benjamin 501 p.
|
16103900
|
PMC1183517
|
CC BY
|
2021-01-05 09:18:22
|
no
|
PLoS Comput Biol. 2005 Jun 24; 1(1):e11
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010011
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science 1610390110.1371/journal.pcbi.001001205-PLCB-RA-0042R2plcb-01-01-13Research ArticleBioinformatics - Computational BiologyCell BiologyEvolutionGenetics/GenomicsGenetics/Gene ExpressionMicrobiologyStatisticsSystems BiologyEubacteriaThe Genomic Pattern of tDNA Operon Expression in E. coli
Genomic Pattern of tDNA Operon ExpressionArdell David H ¤Kirsebom Leif A Department of Cell and Molecular Biology, Biomedical Center, Uppsala University, Uppsala, SwedenStormo Gary EditorWashington University in St. Louis, St. Louis, Missouri, United States of AmericaE-mail: [email protected] (DHA); [email protected] (LAK)¤ Current address: Linnaeus Centre for Bioinformatics, Uppsala University, Uppsala, Sweden
6 2005 24 6 2005 1 1 e127 3 2005 26 5 2005 Copyright: © 2005 Ardell and Kirsebom.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.In fast-growing microorganisms, a tRNA concentration profile enriched in major isoacceptors selects for the biased usage of cognate codons. This optimizes translational rate for the least mass invested in the translational apparatus. Such translational streamlining is thought to be growth-regulated, but its genetic basis is poorly understood. First, we found in reanalysis of the E. coli tRNA profile that the degree to which it is translationally streamlined is nearly invariant with growth rate. Then, using least squares multiple regression, we partitioned tRNA isoacceptor pools to predicted tDNA operons from the E. coli K12 genome. Co-expression of tDNAs in operons explains the tRNA profile significantly better than tDNA gene dosage alone. Also, operon expression increases significantly with proximity to the origin of replication, oriC, at all growth rates. Genome location explains about 15% of expression variation in a form, at a given growth rate, that is consistent with replication-dependent gene concentration effects. Yet the change in the tRNA profile with growth rate is less than would be expected from such effects. We estimated per-copy expression rates for all tDNA operons that were consistent with independent estimates for rDNA operons. We also found that tDNA operon location, and the location dependence of expression, were significantly different in the leading and lagging strands. The operonic organization and genomic location of tDNA operons are significant factors influencing their expression. Nonrandom patterns of location and strandedness shown by tDNA operons in E. coli suggest that their genomic architecture may be under selection to satisfy physiological demand for tRNA expression at high growth rates.
Synopsis
The concentrations of tRNAs are co-adapted to codon usage frequencies in the transcriptomes of E. coli and other diverse organisms. But how are tRNA concentrations determined? Here, the researchers analyzed the E. coli tRNA concentration profile in its genomic context, using clustering and regression methods to partition tRNA concentration data to tDNA operons that were defined semi-automatically. They found that co-expression in operons explains the tRNA profile much better than tDNA gene dosage alone. Furthermore, they could significantly explain the total expression from tDNA operons by their distance from the genomic origin of replication. Per-copy transcription initiation rates from tDNA operons were also estimated. Although there is some evidence for replication-dependent effects on tDNA operon expression, this cannot explain how constant the tRNA profile is with growth rate. As a consequence, tDNA promoters are predicted to compensate for the location of their operons. Finally, the researchers found pronounced asymmetries between the leading and lagging genomic strands in the locations of tDNA operons, and on the effect of location on their expression. These nonrandom patterns suggest that the genomic location and strandedness of tDNA operons may be under some selection in E. coli to satisfy physiological demand for tRNAs at high growth rates.
Citation:Ardell DH, Kirsebom LA (2005) The genomic pattern of tDNA operon expression in E. coli. PLoS Comput Biol 1(1): e12.
==== Body
Introduction
During balanced growth in rich media, prokaryotic and eukaryotic microorganisms selected to grow efficiently are enriched in “major” isoacceptor tRNAs cognate to “preferred” codons in the transcriptome [1,2]. This is explained as a growth-maximizing strategy: to achieve a high rate of growth, ribosomes must be saturated with ternary complex (tRNA + elongation factor Tu + GTP) at the same time as mass invested in ternary complex must decrease [3]. From this perspective, ribosomal substrates specialize in major isoacceptors to optimize a trade-off between rate and mass of the translational apparatus [3]. We call this phenomenon translational streamlining.
There is a question as to whether the tRNA profile becomes increasingly enriched in major isoacceptors at higher growth rates. Some early studies of tRNA concentrations using Northern blots found this to be the case: major isoacceptors increased more than 4-fold at high growth rates, while most minor isoacceptor concentrations decreased [4,5]. A subsequent highly meticulous study using direct quantitation of radioactively labeled tRNA provides presumably the most precise, accurate, and complete measurements of tRNA concentrations in any organism to date [2]. Its authors find that concentrations of major isoacceptors increase with growth rate but only about 2-fold, from μ = 0.4 to μ = 2.5 doublings/h, less than had been found in the previous studies, while minor isoacceptor concentrations remained approximately the same. They conclude that the data are consistent with the hypothesis of growth-rate-dependent enrichment of the tRNA profile, a hypothesis that we call growth-regulated translational streamlining.
Although codon usage bias in efficiently growing microorganisms such as Escherichia coli has been considered one of the best examples of selection at the molecular level (see e.g., [6]), the factors that determine the cellular tRNA concentrations that co-vary with those codon usage patterns are still largely unknown. The mechanisms underlying growth-dependent modulation of tRNA concentrations have been called a mystery, and speculated to be elaborate. It is a de facto standard in computational studies to use tRNA gene (tDNA) dosage (i.e., copy number in the genome) as a proxy for tRNA concentration [7–9], yet in E. coli, gene dosage explains only about half of the variation in tRNA concentrations [2] at any growth rate. Gene dosage also cannot explain any eventual growth-rate-dependent modulation in the tRNA profile. tDNAs, like other genes, are organized into operons in prokaryotes, and it is natural to ask whether an operon-oriented perspective might afford a better understanding of the forces that determine the tRNA profile.
Furthermore, we wished to investigate whether the genomic organization of tDNA operons plays a role in determining the tRNA profile. Some tDNAs are found in common operons with ribosomal RNA genes (ribosomal DNAs, or rDNAs), and tDNA and rDNA operons have many upstream regulatory features in common [10]. There is a clear effect of genome position on the relative outputs of the seven E. coli rDNA operons: those closer to the origin of replication have relatively higher expression [11]. This is because in bacteria such as E. coli, which can divide faster than the time required for their genome to replicate completely, overlapping rounds of genome replication lead to a higher relative concentration of genes near the origin of replication [12]. The dosage of a gene is to be distinguished from its concentration. Gene
dosage is the number of copies of a gene in a genome, and is static with respect to the physiological state of an organism or the replicative state of its chromosome. Gene (or operon) concentration is the average number per cell volume of a chromosomal region containing a gene or operon copy under specific “balanced” (that is, steady-state exponential) growth conditions. Different copies of the same gene scattered around in the genome will have different concentrations depending on the replicative state of the chromosome. Furthermore, gene concentrations depend on cell volume. Theory exists for calculating relative gene concentrations as a function of genome location, growth rate, and other physiological parameters [13–15]. This theory dictates that operon concentration increases exponentially with proximity to the origin of replication (oriC) at a given growth rate. Experimentally, transposition of certain reporter genes toward the origin of replication increases their total relative expression at a specific growth rate in a manner fully consistent with theory [14,16].
In light of these results, we wanted to ask whether replication-dependent effects of genomic location on operon concentration (position effects) can explain the biased tRNA profile in E. coli. Furthermore, we wanted to see if eventual growth-regulated translational streamlining is also mediated by position effects, if operons expressing major isoacceptors were seen to lie preferentially closer to oriC. We note in passing that, at least in E. coli, gene concentrations alone cannot explain the increased concentration of tRNAs at higher growth rates, since cell volume also increases exponentially with growth rate so that the concentration of oriC is kept approximately constant. This means that the concentration of all genes and operons everywhere else in the genome actually decreases with growth rate [14,15,17]. Therefore, the increasing concentration of tRNAs with growth rate [2] requires a growth-regulated increase in the output of all tDNA operons. We hoped then to describe this growth-dependent increase in the output of tDNA operons and see whether or not it was uniform.
However, other results speak against strong position effects explaining the tRNA profile or its eventual modulation with growth rate. The aforementioned connected problem of the regulation of ribosomal RNA (rRNA) synthesis has itself been the subject of controversy (reviewed in [18–20]). Alternatively to either gene concentration or dosage effects explaining variation, endogenous regulators likely induce feedback on stable RNA synthesis to maintain rRNA concentrations at systemically established levels. Different models have been proposed to explain, for instance, that ribosome concentrations are fairly stable to experimental alteration of rDNA operon dosage (reviewed in [18]). A further indication that operon concentration is not limiting to ribosome synthesis is that the synthesis rate is independent of cell age [21]. The effect of gene concentration on expression rate was specifically shown to be buffered in the case of another feedback-regulated system—namely, tryptophan synthase [14]. In the case of tRNAs, a recent study using microarrays also showed clear roles for processing and degradation on tRNA concentrations [22], suggesting that the idiosyncratic effects of individual tRNA structures and their precursors may have strong roles to play in explaining tRNA concentrations. Thus, it is far from clear that position effects can explain the tRNA profile either across operons within a given growth rate or across growth rates.
In the present work, we set out to re-examine Dong et al.'s data on the E. coli tRNA profile and its growth-rate variation in the genomic context of tDNA operon organization. We were surprised to find only weak evidence for growth-rate-dependent streamlining of the tRNA profile; instead, all tRNAs increase at very similar proportions and the tRNA profile is nearly equally streamlined toward major isoacceptors at all growth rates. Then we successfully mapped true tDNA operons in the E. coli genome using a simple, semi-automated scheme. With these in hand, we used least squares multiple regression and existing models and data for the physical properties of growing E. coli to estimate their total and per-copy expression. We show that this “operon model” explains the tRNA profile much better than gene dosage alone. We show that although a large fraction of the variation in tDNA operon expression must be explained by localized differences in regulatory elements and precursor structure, a significant fraction of variation in the E. coli tRNA profile is explained by the genomic location of tDNA operons. Our per-copy estimates, indicative of promoter strength, were consistent with independent experimental data and predict, surprisingly, that promoters in tDNA operons further away from oriC grow relatively stronger with growth rate. This may compensate for decreasing operon concentrations with growth rate to keep the tRNA profile constant. Finally, we demonstrate a significant asymmetry in the locations, and the effect of location on expression, of tDNA operons in the leading and lagging strands. Because co-expression in operons explains almost all of the variation in tRNA concentrations at any growth rate, and because tRNA concentrations are known to be co-adapted with codon usage, these results imply that the location and strandedness of tDNA operons may be partly influenced by natural selection in the genome of E. coli.
Results/Discussion
The tRNA Concentration Profile in E. coli Is Nearly Equally Streamlined at All Growth Rates
In re-examining the data of Dong et al., we found that the concentration of all tRNA isoacceptors increases with growth rate, and does so with surprising proportionality. Figure 1 shows that linear regressions of tRNA isoacceptor concentrations at μ = 0.4 and μ = 0.7 doublings/h explain a surprisingly high fraction, upwards of 96% of the variation, in tRNA concentrations at 2.5 doublings/h. Both μ = 0.4 and μ = 0.7 are used because there may be some idiosyncratic aspects of the data at the lowest growth rates [23]. Thus, considering that the measurement error in the data of Dong et al. is 10%, the proportional increase of all isoacceptors with growth rate swamps any variation in the increase of individual isoacceptors.
Figure 1 Regression of tRNA Isoacceptor Concentrations at the Highest Measured Growth Rate for E. coli Strain W1485 (A K12 Derivative) against the Same at Lower Growth Rates
tRNA concentrations at the highest growth rate (μ = 2.5 doublings/h) are regressed against the same at (A) μ = 0.4 doublings/h and at (B) μ = 0.7 doublings/h. Concentration data are from [2]. Classification into “major,” “minor,” and “neither” types is from codon usage in ribosomal protein genes and anticodon reading relationships from [2,9]. All isoacceptors increase with growth rate, so that the uniform increase of all isoacceptors swamps variation in increase of individual isoacceptors.
Evidence for growth-regulated translational streamlining in the residual variation is weak. We classified isoacceptors as “major,” “minor,” or “neither” on the basis of whether they were cognate to preferred codons as described in the Materials and Methods section, and compared the distributions of ratio increases in concentrations of isoacceptors in these classes (concentration data and classifications are provided in Dataset S1). Figure 2 shows that the least increasing isoacceptors do fall in the minor class (containing 18 isoacceptors), while the major class (containing nine) shows a slightly greater increase with growth rate. Statistically, by this classification, the mean ratio increase of major isoacceptors is not significantly greater than that for minor isoacceptors. We used one-sided tests, which are liberal for rejecting the null hypothesis of equality of ratios between the major and minor groups. For the increase from 0.7 to 2.5 doublings/h (which shows the strongest difference), a Wilcoxon test finds a borderline difference between the distribution of major and minor isoacceptors (p = 0.06), a Welch's t-test on difference in mean ratios is also borderline significant (p = 0.08), but the bootstrap test on the difference in mean ratios is not significant (p = 0.10). Neither is an analysis of covariance test for the effect of isoacceptor type on concentration at μ = 2.5 controlling for concentration at μ = 0.7 (p = 0.37). p-Values for the increase from 0.4 to 2.5 doublings/h are all much higher, also failing to reject equality of means. Thus, the evidence for preferential enrichment of major isoacceptors with growth rate is not strong.
Figure 2 Frequency Histograms and Density Estimates for the Ratio Increase in Concentration of Different Classes of Isoacceptors After an Increase in Growth Rate
Isoacceptors are grouped into “major,” “minor,” and “neither” classes, and the distributions of concentration ratios are shown for each class after an increase in cellular growth rate (A, light grey) from 0.4 to 2.5 doublings/h and (B, dark grey) from 0.7 to 2.5 doublings/h. White-colored bars correspond to values for Thr1+Thr3 as labeled (see text). While no difference is evident among classes from 0.4 to 2.5 doublings/h, a slight difference is evident from 0.7 to 2.5 doublings/h. This difference is not significant but becomes significant if Thr1+Thr3 and Pro1+Pro3 are removed from analysis, or trimmed means are used to compare groups.
These results should be taken cautiously because they depend on how isoacceptors are classified. For instance, if we move Thr1+3 and Pro1+3 from the major class to the neither class, major isoacceptors do have a significantly higher mean increase with growth rate (Figure 2). Similarly, significance increases if we use trimmed means.
The Thr and Pro tRNAs belong to the only two isoacceptor families where major isoacceptors were not uniquely identified in our classification procedure (see Materials and Methods). Not coincidentally, these tRNAs are among the least abundant in the cell. This points out that it may not be correct to weigh all isoacceptor families equally in this analysis as we have done, because amino acid usage is biased and this bias increases with growth rate [24]. Furthermore, the classification is contingent on correct assignments of codon–anticodon reading pattern rules and preferred codons. Lastly, a more complete analysis could account for uncertainty in the concentration measurements. Nonetheless, it is clear that isoacceptor concentrations increase with growth rate in a much more proportional manner than was previously recognized. We conclude that there is scant evidence of growth-dependent streamlining of isoacceptor concentrations in favor of major isoacceptors.
Although our analysis is inconsistent with a strong effect of growth-regulated translational streamlining, it is not inconsistent with translational streamlining in general. Isoacceptor concentrations are biased in favor of major isoacceptors already at low growth rates. Even the slowest growth rate examined presents a significantly higher concentration of major isoacceptors by the Wilcoxon test (p < 0.01). This may be consistent with selection for translational streamlining at the highest growth rate determining the tRNA profile at all growth rates. In conclusion, growth regulation of the tRNA profile may be inessential to the theory that E. coli achieves a growth advantage through translational streamlining.
Operons Explain tRNA Concentrations Better than Gene Dosage Alone
Like protein-coding genes, tDNAs are co-transcribed in operons. We next set out to ask whether the operonic organization of tDNAs can better explain tRNA expression levels in E. coli at any growth rate better than gene dosage alone. We partitioned 87 tDNAs in the E. coli K12 genome [25] obtained with tRNAscan-SE [26] into 47 clusters. A tDNA or cluster of tDNAs was clustered together if they laid within 300 base pairs (bp) or less of one another (the clustering radius) and fell in the same strand. The clustering of 47 was stable for clustering radii between 200 and 1,000 bp (Figure 3). Although this procedure did split apart three rDNA operons—namely, rrnC, rrnD, and rrnH—known co-transcription relationships [10,27–30] were correctly identified in all other cases (including a previously unnamed operon containing only one Thr-2 tDNA, which we call thrX; for details, see Materials and Methods and [31]). We manually joined the three rDNA operons to produce a final set of 44 operons (Table 1).
Table 1 tDNA Operons in the E. coli K12 Genome
Figure 3 Effect of Clustering Radius on the Number of tDNA Clusters Obtained in the E. coli K12 Genome (Calculated in Steps of 25 bp)
Clustering radius (r) is the maximum distance from one end of a tDNA in bp within which part of another co-linear tDNA must fall to be joined into the same cluster. Vertical dashed line shows the value used in this study (r = 300 bp), which correctly recovered all but three of the 44 experimentally known tDNA operons (indicated by horizontal dashed line). These three, ribosomal operons all, were not correctly recovered until a much higher radius was used but then within only a narrow range (2,400 ≤ r ≤ 2,800 bp) before a false positive was encountered. Thus, the natural proximity of tDNAs within operons made it possible through tDNA coordinates and strandedness alone to recover most of the true operons in E. coli K12.
We used the primer sequences from [2] to map concentration data for 44 tRNA isoacceptors from each of five growth rates (Table 5 in [2]) to these tDNA operons, and then estimated by multiple linear regression by least squares with an intercept term, according to Equation 2 (Materials and Methods), a “standardized concentration” for each operon (the design matrix and corresponding right-hand side are provided in Datasets S2 and S3). We call this the operon model (with intercept) for explaining tRNA concentrations. Solving the operon model requires the placement of additional constraints on the system, as shown in Table 2 and described in Materials and Methods. For comparison to the operon model, we repeated the linear regression in [2] of tRNA concentration on tDNA dosage alone (“gene dosage model,” Equation 4 in Materials and Methods).
Table 2 Constraints Added for Least Squares Estimation of OSCs
Despite the addition of 32 variables, the operon model explains tRNA concentrations statistically significantly better than the gene dosage model at all growth rates (p(F
32,10) < 0.002 for all growth rates, see Table 3). Gene dosage explains only 55–60% of the variation, while the operon model explains 92–94% even after adjusting for the added variables (Table 3). This result suggests that, after controlling for gene dosage differences, inputs of different operons to the same isoacceptor pool, and the tendency for tDNAs to repeat within operons, tDNAs that are co-expressed in operons have significantly similar expression. We conclude that the operon model explains tRNA concentration data significantly better than gene dosage alone.
Table 3 Comparison of Gene and Operon Models for Explaining tRNA Concentrations in E. coli, and Circular Regressions of the Estimated OSCs on Angle (α), Increasing as in min, but with the Origin of Replication at Zero in the E. coli K12 Genome
*p < 0.05, ** p < 0.01, *** p < 0.001
For subsequent work with the operon model, specifically in predicting operon expression, we redid the regression dropping out the intercept term, forcing regression through the origin. This is justified because both gene numbers and concentrations are on ratio scales and concentrations have a natural zero if the number of their encoding genes are zero. That is to say, the intercept term has no clear biological interpretation. On the other hand, for model comparisons and regression statistics just shown, we retained intercepts both to reproduce earlier work and because doing so is recommended statistical practice [32]. Intercept terms were never significant in our regressions. At low growth rates (0.4, 0.7, and 1.07 doublings/h), intercept terms were about 70% of the mean magnitude of coefficient estimates, and could reduce their value by half. At high growth rates, intercept terms were about 5% of the mean magnitude of coefficient estimates and affected coefficient estimates by only about 10%.
In the operon model, there are no polarity effects, so that contributions of gene copies are independent of location within an operon. We found that the improvement of the operon model over the gene dosage model (with or without intercept, data not shown) increases quite a bit if we do not manually join the proximal and distal tRNA genes from the three previously mentioned rDNA operons: without intercept, the probability of fit improvement by chance decreases by two orders of magnitude (p(F
33,10) < 10−5), and the adjusted fraction of variation explained is greater than 98% at all growth rates (design matrix and right-hand side provided in Datasets S4 and S5). This model improvement from not joining the distal tDNAs of the ribosomal operons may be because of degradation, RNA polymerase drop-off, or decoupling through secondary promoters, yielding the most influence on the rDNA operons because they are the longest among our operon set. Indeed, initial analysis of promoters in our tDNA operons indicates that the distal Thr1-tDNA in the rrnD operon may have a secondary promoter while the distal tDNAs in the rrnC and rrnH operons do not [31]. This is borne out by an examination of residuals from the operon model without intercept (Figure 4). However, the residuals do not show a generally consistent trend of overestimated expression from operon distal ends, which would have been consistent with systematic transcriptional drop-off (or other effects of operon polarity) on tRNA concentrations.
Figure 4 Residuals of the Operon Model Regressions Used to Estimate Expression (Without Intercept), Showing Unexplained Variation
Only residuals with absolute values larger than 10−14 are shown. Thus, variation in all but nine of the 44 operons is completely explained. All operons containing tDNAs with these non-zero residuals are arrayed at the bottom, showing true tDNA order from 5′ to 3′. All such tDNAs are in single-copy except for the genes encoding Ala1B in the ribosomal operons, indicated by vertical stacking at the bottom, Arg3, which is repeated three times in the serV operon, and MetM, Gln1, and Gln2, which are each repeated twice in the metT operon (for its true configuration, see Table 1). Residuals are shown in μM units, those of standardized concentrations. Residuals are shown for each tDNA in order of increasing growth rate from 0.4 (leftmost) to 2.5 doublings/h (rightmost). Positive (negative) residuals indicate underestimation (overestimation) by the model.
tRNA Operons Are More Productive the Closer They Lie to oriC
We then re-estimated operon standardized concentrations (OSCs) by repeating the operon model regression dropping the intercept term. OSCs (the
obtained by fitting the model in Equation 5) estimate, for each operon and growth rate, the concentration that a tRNA isoacceptor would have at that growth rate if its gene were contained in single copy in, and only in, that operon. Alternatively, OSCs estimate the hypothetical concentrations of tRNA precursors that would be expressed from each operon in the absence of tRNA precursor processing, all other factors being equal. Estimated OSCs and other data about operons are provided in Dataset S6.
In balanced growth, tRNA concentrations, like the concentrations of all cellular components, are proportional to their rates of synthesis (see e.g., [18]). Assuming that tRNA precursor processing is fast and the degradation of stable RNA is slow, this means that tRNA concentrations in balanced growth are proportional to their rates of transcription. Therefore, under these assumptions, our estimated OSCs at a given growth rate are proportional to the “bulk” rate of transcription from each operon at that growth rate, with different constants of proportionality at different growth rates. In the Materials and Methods section, we show how to calculate these constants of proportionality. Below, we present some statistical analyses in terms of OSCs, but equivalently refer to them in terms of operon bulk expression rates when, and only when, such results are invariant up to a multiplicative constant that is equivalent in statistical analysis.
We explored the spatial variation in the genome of operon expression by plotting OSCs against the genomic location of operons and through the use of circular regressions [32], where 0° was placed at the origin of replication oriC. Including an intercept term in a circular regression of OSC (or, equivalently, bulk expression) on genome location (see Equation 6), we found significant negative dependence of operon bulk expression on distance from oriC at all five growth rates, as indicated by the significant cosine terms in Table 3. In contrast, sine terms of the circular regressions were not significantly different from zero, suggesting symmetry of the expression pattern about oriC.
Figure 5 shows that, despite considerable variation independent of location, average bulk expression of operons (for μ = 2.5 doublings/h in units of initiations per minute per pg of cell culture) increases the closer that operon lies to oriC. Figure 5 also shows a re-estimated circular regression including only the cosine and intercept terms, which we later use to compare models in Table 4.
Table 4 Evidence for Log-Linear Fit of Estimated OSCs (μM) against Genome Location (m)
aAkaike Information Criterion for fit of standardized conc. against cos (α) as defined in Table 3.
bAkaike Information Criterion for fit of log standardized conc. against genome location (m).
cExpectations calculated from Equation 1 and Cμ = (220 /3) − (40/3) μ calculated from [18].
dBCA confidence intervals.
eExcluding outlying data for IleX and IleY.
* p < 0.05
Figure 5 Total Expression (Bulk Output) at μ = 2.5 Doublings/h of tDNA Operons Against Their Location in the E. coli K12 Genome
The angular scale is symbolized by α in the text, with 0° placed at the origin of replication (oriC) shown at the top. Units of the radial axis (expression) are initiations per min per picogram of culture mass. Leading strand operons are indicated in blue and lagging strand operons in gold. The red curve shows a re-estimated circular regression of all the data including only intercept and cosine terms, showing the significant tendency of expression to increase toward oriC, especially for leading strand operons. Values for lagging strand operons asnV, asnU, and asnT are covered but equal by constraint (see Table 2) to the value for asnW.
Proximity to the origin of replication explains 11–17% of the variation in estimated tDNA operon expression in E. coli in a circular regression. However, it is not at all clear whether this implies that expression rate is a cause or an effect of location in the genome, or both. Expression rate as an effect of location could derive either physiologically, by the effect of genome position on gene concentration, or evolutionarily, by some (admittedly highly speculative) genome-position-dependent mutation effect that would tend to make promoters stronger near the origin of replication. Expression rate as an ultimate cause of location could occur through the selection of certain operons to be retained near oriC after they are moved there by translocation, horizontal transfer, or inversion of operons and tDNAs in the genome. This selection could be for higher expression through the position effect on expression or by some hypothetical advantage of operons with strong promoters per se to lie near the origin. In the following, we find evidence of both directions of causality: expression is caused to some degree by genome location, but the location and strandedness of operons has also likely evolved in E. coli to exploit this and other effects to increase satisfaction for tRNA demands of the cell.
tRNA Operon Expression Is Consistent with Replication-Dependent Gene Dosage Effects
To examine the hypothesis that position-dependent effects of replication on gene concentration are causing the genomic pattern of tDNA operon expression in E. coli, we compared our estimated operon expressions to a statistical model based on well-known theory relating gene expression to gene concentration [14,15]. The full derivation of this model is shown in Materials and Methods. This model is as follows:
where Expr(X) is the expression of operon X at growth rate μ and genomic position m (as a fraction of the length of half of a chromosome), kμ is an unknown proportionality constant relating the average expression of a set of genes to their concentrations, ɛ is a stochastic error term, [oriC] is the concentration of the origin of replication, and Cμ is the time required for complete genome replication, considered a constant given the bacterial strain and growth rate μ.
The intercept in this model depends on unknown factors or nuisance parameters such as the concentration of oriC in relation to growth rate (which can vary at low growth rates and is possibly strain-dependent [18,33,34]), the distribution of the stochastic error ɛ, and kμ, which captures the uniform increase in transcription at all tDNA operons as a function of growth rate. However, the slope in this model, and its underlying exponential form, depends on only the well-studied parameter Cμ, and can be directly compared to our expression estimates to evaluate consistency with position effects on gene concentration.
We find general agreement of our estimates with the model in Equation 1. That is to say, we find evidence for exponential fits of operon expression against genome location, decreasing from the origin. First, the Akaike Information Criterion indicates that an exponential model (Equation 9) fits the standardized concentration data much better than the trigonometric model (Equation 6) used in the circular regressions (Table 4). This means that the data are more consistent with an exponential trend than with the trigonometric trend implied in the circular regression, although it does not rule out that some other function would fit the data better. Second, the Box-Cox test suggests log-transformation of the data, also consistent with an exponential trend (Figure S1). Indeed, the 95% confidence intervals of lambda at all growth rates include zero (log-transformation) and exclude one (no transformation of the data). Third, confidence intervals for the slopes in log-linear fits include expected values based on realistic estimates of C at all growth rates (Table 4). Thus the data are broadly consistent with a position effect on gene concentration causing increased expression of operons toward oriC.
On the other hand, for the model in Equation 1 to fit in detail, one would expect the slopes of the log-linear regressions in Table 4 to decrease with increasing growth rates. One expects, in other words, steeper gradients of expression on distance from the origin of replication at higher growth rates. However, our estimated slopes are constant in μ. Indeed, operon expression rates, like raw tRNA isoacceptor concentrations, increase in fairly constant proportions with growth rate. High fractions, albeit lower than with the raw data, of estimated operon expression rates at higher growth rates could be explained by those at lower growth rates, from 82–97%. For instance, when regressed through the origin, the growth-dependent increase factor at 2.5 doublings/h from 0.4 doublings/h is 1.74 ± 0.04 (SE) for tDNA estimated operon expression rates and 1.80 ± 0.04 for raw tRNA concentrations, explaining 97% and 98% of the variation, respectively. This shows that the data are inconsistent with gene concentration as a cause of growth-regulated translational streamlining.
It is remarkable that if E. coli actually were selected for extensive growth-regulated translational streamlining, this could have easily been arranged by holding the relative strength of different tDNA promoters constant with growth rate and passively exploiting the location effect on gene concentration. The fact that the tDNA operon expression profile is fairly constant, despite changes in underlying operon concentrations, implies that the relative strength of different tDNA promoters must change with growth rate. We can ask which growth rate condition better fits the expected slope, which might suggest under which condition expression is most governed by the position effect on operon concentration and least governed by a hypothetical compensating factor working against this effect. Interestingly, the data fit the expected slopes better at higher growth rates (Table 4), suggesting that any hypothetical compensatory effect may be most effective under slow growth conditions. In this case, this hypothetical compensating effect would make, at low growth rates, either origin-proximal promoters stronger, terminus-proximal promoters weaker, or both, than what would be expected based on comparisons with expressions at higher growth rates.
We can also regress out the expected position effect on operon concentration to study trends in estimated per-copy expression rates of each tDNA operon at each growth rate, using the derivation shown in Materials and Methods (Table 5). Our data agree fairly well with independent calculations for rDNA (rrn) operons [35]. According to these calculations, the average expression from all rrn operons combined increases from 8.9 initiations per min per copy at μ = 0.7 to 16.0 at μ = 1.07 and up to 66 initiations per minute per copy at μ = 2.5 doublings/h (Table 5). These estimates are quite similar, if slightly under, those estimated in [35] based on different assumptions, data, and bacterial strains. Based on measurements of total RNA, they calculated the average initiation rate of rrn operons to be about 10 at μ = 0.6, just under 20 at μ = 1.1, about 64 at μ = 2.2, and almost 90 at μ = 2.7. The data in [35] are presented only as figures, not tables, so the values quoted here are approximate. The agreement seems satisfactory. Our estimates for rrn operons may be low because their distal tDNAs are weighted equally (see above), and because of the relative lack of information for the isoacceptors they express (i.e., Ile-tRNAs). Nonetheless, these results suggest that Table 5 shows reasonable predictions for initiation rates from tDNA operons, with one caveat: excess synthesis to compensate for eventual tRNA degradation, or for transcriptional abortion and drop-off, cannot be detected from the combination of data and theory we have used to make these estimates. The estimates in Table 5 should therefore be considered as minimal estimates assuming no losses, or as “net” synthesis rates after such unmeasured losses have taken place.
Table 5 Estimated Average Synthesis Rates Per Operon (Number of Transcripts Initiated per min per Copy) of tRNA Precursors in the E. coli K12 Genomea
aData calculated to two significant digits.
The per-copy estimates of operon expression rates (“promoter velocities”;
in Materials and Methods) have no significant trend against genome location at any growth rate. This argues against strong promoters per se evolving near oriC either by location-dependent mutation effects in situ or by translocation. However, the growth-rate ratio increase in per-copy synthesis rates are very significantly and positively dependent on distance from oriC (Figure 6), both from μ = 0.4 to μ = 2.5 (
) and μ = 0.7 to μ = 2.5 (
). Significance and explained variation increases when we exclude outlying values for the IleX and IleY operons (
and
, respectively). This is the evidence for the aforementioned compensatory effect holding the tRNA profile relatively constant despite the position effect on operon concentration. Since we have not examined any model for the growth regulation of tDNA promoters as a whole, we cannot say whether this compensatory effect acts through greater acceleration of origin-distal promoters or lesser acceleration of origin-proximal promoters at higher growth rates.
Figure 6 Ratio Increase in Per-Copy Synthesis Rates of Operons (Promoter Velocities) as Growth Rate Increases from μ = 0.4 to 2.5 Doublings/h
(A) μ = 0.7 to 2.5 doublings/h. (B) Against fractional distance from the origin of replication oriC with maximum distance set at 1 (m). Outlying values for ileX and ileY are indicated.
To summarize, we have provided strong evidence that the genomic location of tDNA operons plays a significant role in shaping the tRNA profile in E. coli. Even though it is obvious that location-dependent gene concentrations must be accounted for when calculating expression rates from any one operon, our result that such position effects partly explain concentration variation across different tRNAs is unexpected and novel. Yet the results are not fully consistent with the simplest model of operon location determining expression rate. First, location explains only about 15% of expression rate variation (Figure 5), so one may say that intrinsic causes of expression such as promoter velocity are the primary determinant. Second, the tRNA profile is relatively constant at different growth rates while tDNA operon expression is a negative exponential function of the product of growth rate and genomic distance from oriC (Equation 1). This implies the existence of a compensatory effect working against the position effect on expression (Figure 6).
tRNA Operon Locations and Expression Are Different in the Leading and Lagging Strands
tDNA operons in E. coli are different in the leading and lagging strands with respect to both location and the effect of location on expression. We defined the strandedness of an operon by the angular coordinate relative to oriC (α) of its first tDNA and its orientation relative to the K12 genome sequence; if α < 180°, the operon is leading if parallel and lagging if antiparallel, and vice versa otherwise (see Table 1). The leading and lagging strands are quite different with respect to the placement and the expression of tDNA operons (see Figure 4). tRNA genes considered separately lie preferentially on the leading strand (χ
2 = 5.069, df = 1, p = 0.02), but this ignores operonic organization with its constraints of co-orientation and tendency for tDNA repeats within operons. We find that tDNA operons are evenly distributed on the two strands (χ
2 = 1.454, df = 1, p = 0.23). Furthermore, statistically speaking, there is no difference in the mean size (number of tDNAs) of leading and lagging tDNA operons, by either a permutation test (p = 0.35; see Materials and Methods) or the Wilcoxon test (p = 0.52). However, after dividing operons into two sets according to whether they lie closer to oriC or the terminus region, leading strand operons lie significantly closer to the origin and lagging strand operons closer to the termini (Figure 5 and Fisher's exact test, p = 0.010). This observation is partially reaffirmed by circular statistics [32,36]. The Watson two-sample test rejects homogeneity of the spatial distributions of leading and lagging strand tDNA operons (U2 = 0.1949, p < 0.05) and the Rayleigh test rejects circular uniformity of the placement of leading strand operons against an alternative unimodal distribution toward the origin (r
0 = 0.3214, p < 0.01).
However, the distribution of lagging strand operons is not significantly different from uniform by any circular statistical one-sample test that we tried, including Watson's, Kuiper's, and Rayleigh's. Although these tests make different assumptions, they all lead to the same conclusion. Lack of power without including prior knowledge of the biological importance of the origin or termini may partly explain this, as these tests also failed to reject uniformity for leading strand operons. When we supplied an alternative hypothesis that lagging strand operons are oriented toward the termini, the Rayleigh test for lagging strand operons barely failed to reject uniformity (r
0 = 0.2446, p = 0.07). We conclude that leading strand tDNA operons are significantly clustered spatially toward oriC while lagging strand operons are not.
With respect to expression, leading strand operons show an even greater increase in estimated expression toward oriC than all operons taken together (cosine term 2.07 ± 0.75, p = 0.0108, sine term NS), while lagging strand operons alone show no significant relationship of expression on genome location. However, by either analysis of covariance or a Wilcoxon test of residuals, we found no effect of strand—that is, no statistical difference between the two strands—on estimated bulk expression of operons at any growth rate, after accounting for the effect of operon location.
Thus, in E. coli, genome location effects are very strong in the leading strand and statistically insignificant in the lagging strand. In the leading strand, operons are placed nonrandomly with respect to oriC, and there is a strong location effect on expression. However, operons and overall expression are both equally distributed on the two strands. How do we explain these statistical observations? Rocha discusses two hypotheses to explain strand asymmetries in gene placement and gene expression [37]. Both depend on the effects of head-on collisions of DNA polymerase and RNA polymerase during the transcription of lagging strand genes. One hypothesis emphasizes the effect of such collisions on genome replication through stalling of replication forks. The other hypothesis emphasizes the effect of such collisions on transcription through aborted transcripts either failing to meet gene product demand or by their dominant negative effects due to hypothetical toxicity. We speculate that transcriptional output might be diminished after a head-on collision of RNA polymerase with the replication fork not only by abortion of an elongating transcript at the time of collision but also possibly by interference with transcription during resolution of stalled replication forks after a collision.
Systematic studies of protein-coding genes in E. coli and other bacteria have pointed toward gene essentiality rather than gene expression level as a better explanatory variable for predicting strandedness [38]. This favors the interpretation that it is the effect of polymerase collisions on transcription that determines strand asymmetries in gene placement. With tRNA precursors, it seems unlikely to us that incomplete transcripts would have toxic effects, but a detrimental effect from failing to meet the physiological demand of translation seems likely. Constraints of high demand on an operon might select on it being located near the origin with leading strand orientation, while lesser demand might permit an operon to evolve a more random location and orientation through transposition and inversion, with adjustments by the local evolution of promoter strength as the dominating factor setting expression levels in both strands. We note in passing that if there were a toxic effect from the transcriptional abortion of any operon through fork collision with RNA polymerase, it would be amplified by genomic proximity of offending operons to the origin of replication, since the toxicity would be proportional to average gene concentration (the proportion of toxic product to total production from an operon would be independent of location, but the absolute quantity of toxic product would increase with the concentration of the operon). We speculate further that the negative effect of lagging-strandedness on transcriptional productivity would be a constant proportion of output regardless of location, and might therefore be expected to be neutral to location. These speculations demand a quantitative assessment of the data on the mechanisms and kinetics of events during and after such collisions for their further evaluation.
Our results lead us toward the testable conclusion that physiological demand for tRNA can serve as an evolutionary cause of the genomic location and strandedness of tDNA operons. The tRNA profile, which is known to be selected for translational streamlining in covariation with preferred codons, has now been shown to be correlated with the location and strandedness of tDNA operons. This suggests that the genomic architecture of tDNA operons is itself under some degree of natural selection in E. coli.
Stronger conclusions from the present analysis cannot be made before comparative analysis of tDNA operon and promoter sequences is undertaken within and among enterobacterial genomes. This will be the subject of future research.
Materials and Methods
87 tDNAs were identified by tRNAscan-SE [26] from the E. coli K12 MG1655 genome and then matched by regular expressions to reverse complements of oligo probe data from [2]. Only one tDNA was unmatched by any of the oligos: one of the major Leu1 tDNAs with a single mismatch in the variable loop from other copies. This mismatch occurs in the middle of the relevant oligo, so this tDNA was included. We included a Thr2 gene we call thrX at coordinate 296402. This Thr2 gene is annotated in the Genome tRNA Database (http://lowelab.ucsc.edu/GtRNAdb/), but it is not annotated in EcoCyc (http://www.ecocyc.org) nor in the NCBI genome annotation, and there is no evidence for its specific expression. Using tRNAscan-SE, the thrX gene has a covariance score of 42.36, which is almost half that of the other and nearby Thr2 gene thrW at coordinate 262095, indicating that it contains structural irregularities. Closer inspection shows that 1) thrX is a chimera with a 3′ end identical to thrW starting in the anticodon stem 5′ of the loop, 2) that the oligo used against Thr2 in [2] matches the 3′ end of the anticodon stem into the variable loop and would therefore hybridize perfectly to this tRNA were it expressed, 3) this tRNA would probably fold normally if it were expressed, including tertiary contacts, and 4) the thrX gene has a reasonable upstream promoter [31]. We therefore included thrX in the analysis, but removing it from analysis does not change our regression results (design matrix and right-hand side provided in Datasets S7 and S8), because the only other operon containing a Thr2 gene in E. coli K12 is nearby and co-oriented thrW.
All statistical analysis was executed in R [39]. We classified isoacceptors as “major,” “minor,” or “neither” in reference to preferred codons in highly expressed genes at high growth rates. For this codon-based criteria, we analyzed codon usage in 45 ribosomal protein genes from the E. coli K12 genome [25] with codonw (J. Peden, http://www.molbiol.ox.ac.uk/cu/). The top two codons for each amino acid were checked against cognate anticodon reading patterns as according to [2], which is also the source of the correspondence between tRNA isoacceptor numbering and anticodons shown in Table 1. A major isoacceptor was then picked uniquely in all acceptor classes except for two cases (threonine with two tRNAs, Thr1 and Thr3, that both have GGU anticodons matching the preferred codons ACC and ACU, and proline with two tRNAs, Pro1 and Pro3, that both match the preferred codon CGG). Thr1 and Thr3 were added together as were Pro1 and Pro3, and these combined data were assigned to the major class. This assignment of major isoacceptors is identical to that used by Ikemura [40] with the exception of Ser5, which Ikemura did not measure. Two undecidable cases (Ile1+Ile2, which Dong et al. could not distinguish, and Tyr1 and Tyr2, both of which match both Tyr codons) and all tRNAs in single-isoacceptor families were labeled as “neither.” The remainder were assigned to the minor isoacceptor class. The classifications are shown in Dataset S1. The bootstrap test for difference in mean ratios between the major and minor groups was calculated by algorithm 16.2 on p. 224 in [41]).
To predict operons, tDNAs with the same orientation in the genome were clustered automatically using an end-to-end distance (clustering radius) of 300 bp. In known cases of heterogeneous operons (tDNAs mixed with protein-coding genes), tDNAs always come first in the operon [28]. We have found no evidence otherwise in a separate analysis of upstream promoters [31]. The procedure split apart three ribosomal operons that were manually joined as described in the text.
OSCs are derived from concentration data in Table 5 in [2] projected onto the defined operons by least squares multiple regression and are in μM units (see Dataset S5). In order to perform this operon model regression, we had to add ten additional constraints first to bring the design matrix to full rank, and then one additional constraint to enforce nearly equal expression of ribosomal operons (see Table 2). Ten of the 11 constraints amounted to adding assumptions, such as if there are two operons feeding into, and only into, a single isoacceptor pool, that they do so equally and were absolutely necessary to perform the regression. One of these ten involved constraining two of the three ribosomal operons rrnA, rrnD, and rrnH and was necessary to perform the regression, but involved a choice between two alternatives. However, with either of these minimal ten constraints, the two constrained rrn operons were estimated to be expressed at unrealistically low levels and the other at an unrealistically high level, when it is known that all ribosomal operons tend to be expressed at similar fairly high levels [10,11,42] (Design Matrix and right-hand side provided in Datasets S9 and S10). This may have been because there is relatively little data for the isoacceptor pools fed into by ribosomal operons—for instance, Dong et al. were not able to distinguish the Ile1 and Ile2 isoacceptors. Therefore, we added back the additional constraint on these ribosomal operons to enforce their nearly equal expression (see Table 2).
We estimated OSCs
of operon j at growth rate μ by least squares solutions of matrix equations of the form
where S is the number of isoacceptors, P is the number of operons, C is the number of constraints,
is the dosage of tDNA i in operon j,
is the coefficient of constraint i on operon j,
iμ is an intercept term, and
is the concentration of tRNA i at growth rate μ. For ease of reference, we can rewrite Equation 2 as
where D is a S × P matrix, C is a C × P matrix, M is a (S + C) × P matrix,
is a column vector of ones of length N,
is a column vector of
of length P,
is a column vector of
of length S,
is a column vector of zeroes of length C, and
is the concatenation of
and
.
We call a specific matrix M on the left-hand side of Equation 3 a “design matrix.” For what we call the operon model, S = 44, P = 44, and C = 11. All design matrices M and their corresponding
are available in Datasets S2–S5 and S7–S10. For comparison to the operon model, we repeated the linear regression in [2] of tRNA concentration on tDNA dosage alone (gene dosage model), which assumes equal expression of all tDNAs, thereby fitting concentration data using only a single variable xμ and genomic tDNA copy number as a predictor. In our notation, the gene dosage model is simply
where
, and
is a column vector of size S, the ith component of which is gi.
For the purposes of evaluating and comparing the operon and gene dosage models (
and F in Table 3), we included intercept terms in the regressions. The degrees of freedom are 42 for the gene dosage model and 10 for the operon model (44 data points minus 44 operons plus 11 constraints minus 1 intercept term). The adjusted coefficients of determination (
in Table 3) are corrected for degrees of freedom by the definition
.
For the purpose of studying genomic variation in operon expression under the operon model, we regressed the data without an intercept term according to the model
which reasonably implies that these operons are the only sources of tRNAs in the cell, and is also reasonable because tRNA concentrations and gene dosage are on ratio scales with true zeroes.
The circular regression presented in Table 3 uses a model
where
is the standardized concentration of operon j estimated according to the model in Equation 5, αj is angular distance of operon j from oriC in the direction of minutes of genetic distance, βI are regression coefficients, and ɛ is an error term. The regression shown in Figure 5, and the cosine model in Table 4, is based on this model without the sine term, which was insignificant at all growth rates. In the case of Figure 5, the fitted model was scaled by a constant to give predicted bulk operon expressions. The calculation of this constant is described next.
Bulk operon expression (total expression rates) are observed (estimated by fitting the model in Equation 4) or predicted (from a fit to the model in Equation 6) OSCs
multiplied by a growth-rate-dependent factor rμ = (NAVμ/1021
Mμ)(μln2/60), where NA is Avogadro's number, μ is the growth rate in doublings/h, Vμ is average cell volume in μm3, and Mμ is average cell mass in grams as functions of μ. Multiplication by this factor yields units of number of initiations per min per gram of cell culture. Functional relationships for average cell volume (Vμ = 0.4 × 2μ μm3) and average cell mass (Mμ = 1.6 × 10−13 × 2μ g) with growth rate μ were taken from [17]. The first factor in rμ yields a density while the second factor in rμ derives from the relationship between density and synthesis rate during balanced growth [18]. Values in Figure 5 are multiplied by an additional factor of 10−12 to yield values per picogram. Statistical results calculated on data within a growth rate (such as in Tables 3 and 4) are invariant to multiplication by this constant factor. Therefore, some results are discussed and presented equivalently as standardized concentrations or total expression rates.
The average concentration [X] of a gene per concentration [oriC] of the origin of replication oriC at growth rate μ and location m (relative distance from the origin of replication as a fraction of maximal distance, with the length of a half-chromosome set to 1) follows the relationship [14,15]:
where Cμ is the time required for complete genome replication, considered a constant given the bacterial strain and growth rate μ (equivalently, this formula can be presented in terms of the doubling time in min τ, where τ = 60/μ). A derivative stochastic model for the predicted expression Expr(X) of such a gene is therefore
where the unknown proportionality constant kμ relates the estimated average expression of a set of genes to their concentrations as an unknown but common function of growth rate, and ɛ is some stochastic error term. Taking logarithms yields Equation 1 in the text.
The Box-Cox tests the likelihood of different functional families with the data using a single parameter lambda. We used the default range in R to fit lambda, which is from −2 to 2 in 0.1-increments.
Table 4 compares the circular regression model in Equation 6, fitted without a sine term, to the fit of the data to an exponential model of the form in Equation 1, namely:
where mj is the fractional distance from oriC of operon j with maximum 1. The expected slopes in Table 3 are calculated from Equation 1 and the assumption of a linear and strain-independent [33,34] dependence of the genome replication period C on growth rate μ, calibrated from data in [18,33] to be Cμ = (220/3) − (40/3)μ. We also repeated these comparisons excluding outlying estimates for the ileX and IleY operons. That these estimates were outliers could be seen by comparing relative proportional trends of estimates against growth rates (e.g., Table 5 or Figure 6), as well as by their effects on statistical tests (e.g., Table 4). Instability in these estimates came because of the aforementioned lack of data for Ile isoacceptors (Ile1 and Ile2 could not be distinguished by the oligos in Dong et al.'s data) relative to their constraint in the least squares regression.
Per-copy estimates of operon expression rates, which we also call “promoter velocities”
(see text for caveats) and shown in Table 5, are proportional to bulk operon expression rates
through an additional growth-rate-dependent factor
. The first factor in lμ converts grams to spectrophotometric units OD450 from a factor measured in [43]. The second factor is the reciprocal density of oriC per unit OD450 taken to be constant with growth rate at value 10−9 [33]. The third factor is in units of dosage ratio of an operon at location m to oriC given by Equation 7 with Cμ = (220/3) − (40/3)μ as above. Promoter velocities
are therefore in units of initiations per minute per copy. Values in Table 5 are shown at only two significant figures to emphasize their highly approximate nature, owing to the approximately 10% uncertainty in the isoacceptor concentration measurements from which they are derived, and to the rough nature of the assumptions that went into the least squares estimation, and are probably underestimates for reasons discussed in the text. Figure 6 shows ratios
of promoter velocities at a high growth rate μ2 and a lower growth rate μ1.
Circular statistical calculations were calculated in R with the additional CircStats package (S-plus original by Ulric Lund, R port by Claudio Agostinelli, available at http://cran.r-project.org/). To compare the number of tDNAs in leading and lagging strand operons by a permutation test, we calculated the sizes of the operons, where the size sj of the jth operon, 1 ≤ j ≤ P is
. The means and variances were similar among the two groups, with the leading group mean at 2.077 and the lagging group mean at 1.833, and the variances 2.474 and 2.5, respectively. We then carried out a permutation test sampling R = 10,000 permuted assignments of sizes to the leading and lagging strand groups using the standard equal-variance two-sample t-test as a test statistic and report the proportion
, where i is an indicator function equal to one if its argument is true and zero otherwise, t is the value of the test statistic for the observed groups and
is the value of the test statistic for the pth permuted group assignment.
Supporting Information
Dataset S1 Dong et al.'s tRNA Concentration Data with Classification of Isoacceptors
This dataset contains Dong et al's Table 5 with concentration data, and the assignment of isoacceptor types to tRNAs: “major”, “minor,” and “neither.”
Units of concentration are uM. To reproduce the results of Figure 2 and the statistical two-sample tests, values for Pro1 and Pro3 should be added together, as should Thr1 and Thr3.
(3 KB TXT)
Click here for additional data file.
Dataset S2 Design Matrix for Least Squares Regression with 44 Operons and 11 Constraints
This is the “correct” matrix used in the analysis, joining ribosomal operons rrnC, rrnD, and rrnH. Can be input directly into R.
(6 KB TXT).
Click here for additional data file.
Dataset S3 Concentration and Constraint Matrix to be Used with S2
(1 KB TXT).
Click here for additional data file.
Dataset S4 Automated Design Matrix for Least Squares Regression with 47 Operons and 13 Constraints
Corresponds to tDNA clusters found with a clustering radius of 300 bp used in the paper, which splits apart ribosomal operons rrnC, rrnD, and rrnH. Can be input directly into R.
(6 KB TXT).
Click here for additional data file.
Dataset S5 Concentration and Constraint Matrix to be Used with S4
(1 KB TXT).
Click here for additional data file.
Dataset S6 OSCs and Other Correlates
This table gives OSCs made with the “correct” design matrix Dataset S2 called “design_matrix.correct” and estimated by least squares regression through the origin. Additional operon properties are collected here.
(5 KB TXT).
Click here for additional data file.
Dataset S7 Design Matrix for Least Squares Regression with 43 Operons and 10 Constraints
This is the same as the “correct” matrix but excludes the operon thrX discussed in the text. Can be input directly into R.
(7 KB TXT).
Click here for additional data file.
Dataset S8 Concentration and Constraint Matrix to be Used with Dataset S7
(1 KB TXT).
Click here for additional data file.
Dataset S9 Minimal Design Matrix for Least Squares Regression with 44 Operons and 10 Constraints
This imposes minimal possible constraints on expression equality among ribosomal operons rrnC, rrnD, and rrnH. Can be input directly into R.
(5 KB TXT).
Click here for additional data file.
Dataset S10 Concentration and Constraint Matrix to be Used with Dataset S9
(1 KB TXT).
Click here for additional data file.
Figure S1 Log-Likelihood Profile for Box-Cox Test
(4 KB EPS).
Click here for additional data file.
Accession Numbers
The GenBank (http://www.ncbi.nlm.nih.gov/Genbank) accession number for the E. coli K12 MG1655 genome is NC_000913.
We thank Diarmaid Hughes, Kurt Nordström, Otto Berg, Måns Ehrenberg, Charles Kurland, Johan Elf, Hugo de Boer, and anonymous reviewers for helpful comments and discussion, Måns Ehrenberg and Eduardo Rocha for sharing unpublished manuscripts, and acknowledge a grant to LAK from the Wallenberg Consortium North and to DHA and LAK from Vetenskapsrådet, the Swedish Research Council.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. DHA conceived and designed the experiments, performed the experiments, analyzed the data, contributed reagents/materials/analysis tools, and wrote the paper. LAK contributed both materially and intellectually to the development and execution of the project.
Abbreviations
bpbase pair
OSCoperon standardized concentration
rDNAribosomal DNA
rRNAribosomal RNA
==== Refs
References
Ikemura T 1982 Correlation between the abundance of yeast transfer RNAs and the occurrence of the respective codons in protein genes. Differences in synonymous codon choice patterns of yeast and Escherichia coli with reference to the abundance of isoaccepting transfer RNAs J Mol Biol 158 573 597 6750137
Dong HJ Nilsson L Kurland CG 1996 Co-variation of tRNA abundance and codon usage in Escherichia coli at different growth rates J Mol Biol 260 649 663 8709146
Ehrenberg M Kurland CG 1984 Costs of accuracy determined by a maximal growth rate constraint Q Rev Biophys 17 45 82 6484121
Emilsson V Kurland CG 1990 Growth rate dependence of transfer RNA abundance in Escherichia coli
Embo J 9 4359 4366 2265611
Emilsson V Naslund AK Kurland CG 1993 Growth-rate-dependent accumulation of twelve tRNA species in Escherichia coli
J Mol Biol 230 483 491 7681880
Akashi H Eyre-Walker A 1998 Translational selection and molecular evolution Curr Opin Genet Dev 8 688 693 9914211
Percudani R Pavesi A Ottonello S 1997 Transfer RNA gene redundancy and translational selection in Saccharomyces cerevisiae
J Mol Biol 268 322 330 9159473
Kanaya S Yamada Y Kudo Y Ikemura T 1999 Studies of codon usage and tRNA genes of 18 unicellular organisms and quantification of Bacillus subtilis tRNAs: Gene expression level and species-specific diversity of codon usage based on multivariate analysis Gene 238 143 155 10570992
Rocha EP 2004 Codon usage bias from tRNA's point of view: Redundancy, specialization, and efficient decoding for translation optimization Genome Res 14 2279 2286 15479947
Keener J Nomura M 1996 Regulation of ribosome synthesis Neidhardt F Curtiss III R Ingraham JL Lin EC Brooks Low K Escherichia coli and Salmonella : Cellular and molecular biology, 2nd ed Washington, D.C. ASM Press 1417 1431
Condon C Philips J Fu ZY Squires C Squires CL 1992 Comparison of the expression of the seven ribosomal RNA operons in Escherichia coli
Embo J 11 4175 4185 1396599
Cooper S Helmstetter CE 1968 Chromosome replication and the division cycle of Escherichia coli B/r J Mol Biol 31 519 540 4866337
Sueoka N Yoshikawa H 1965 The chromosome of Bacillus subtilis . I. Theory of marker frequency analysis Genetics 52 747 757 4953222
Chandler MG Pritchard RH 1975 The effect of gene concentration and relative gene dosage on gene output in Escherichia coli
Mol Gen Genet 138 127 141 1105148
Bremer H Churchward G 1977 An examination of the Cooper-Helmstetter theory of DNA replication in bacteria and its underlying assumptions J Theor Biol 69 645 654 607026
Schmid MB Roth JR 1987 Gene location affects expression level in Salmonella typhimurium
J Bacteriol 169 2872 2875 3294809
Donachie WD Robinson AC 1987 Cell division: Parameter values and the process Neidhardt F Ingraham JL Low KB Magasanik B Schaechter M Escherichia coli and Salmonella typhimurium , 1st ed Washington, DC American Society for Microbiology 1578 1593
Dennis PP Ehrenberg M Bremer H 2004 Control of rRNA synthesis in Escherichia coli: A systems biology approach Microbiol Mol Biol Rev 68 639 668 15590778
Paul BJ Ross W Gaal T Gourse RL 2004 rRNA transcription in Escherichia coli
Annu Rev Genet 38 749 770 15568992
Gralla JD 2005
Escherichia coli ribosomal RNA transcription: Regulatory roles for ppGpp, NTPs, architectural proteins and a polymerase-binding protein Mol Microbiol 55 973 977 15686546
Dennis PP 1971 Regulation of stable RNA synthesis in Escherichia coli
Nat New Biol 232 43 47 4935732
Dittmar KA Mobley EM Radek AJ Pan T 2004 Exploring the regulation of tRNA distribution on the genomic scale J Mol Biol 337 31 47 15001350
Berg OG Kurland CG 1997 Growth rate-optimised tRNA abundance and codon usage J Mol Biol 270 544 550 9245585
Emilsson V Kurland CG 1990 Growth-rate dependence of global amino-acid-composition Biochim Biophys Acta 1050 248 251 2207150
Blattner FR Plunkett G Bloch CA Perna NT Burland V 1997 The complete genome sequence of Escherichia coli K-12 Science 277 1453 1474 9278503
Lowe TM Eddy SR 1997 tRNAscan-SE: A program for improved detection of transfer RNA genes in genomic sequence Nucleic Acids Res 25 955 964 9023104
Kenri T Imamoto F Kano Y 1994 Three tandemly repeated structural genes encoding tRNA(f1Met) in the metZ operon of Escherichia coli K-12 Gene 138 261 262 8125313
Inokuchi H Yamao F 1995 Structure and expression of prokaryotic tRNA genes Söll D RajBhandary UL tRNA: Structure, biosynthesis and function Washington, D.C. ASM Press 17 30
Champagne N Lapointe J 1998 Influence of FIS on the transcription from closely spaced and non-overlapping divergent promoters for an aminoacyl-tRNA synthetase gene (gltX) and a tRNA operon (valU) in Escherichia coli
Mol Microbiol 27 1141 1156 9570400
Ow MC Kushner SR 2002 Initiation of tRNA maturation by RNase E is essential for cell viability in E. coli
Genes Dev 16 1102 1115 12000793
Pettersson BM Ardell DH Kirsebom LA 2005 The length of the 5′ leader of Escherichia coli tRNA precursors influences bacterial growth J Mol Biol In press.
Zar JH 1999 Biostatistical analysis Saddle River (New Jersey) Prentice-Hall
Bipatnath M Dennis PP Bremer H 1998 Initiation and velocity of chromosome replication in Escherichia coli B/r and K-12 J Bacteriol 180 265 273 9440515
Michelsen O Teixeira de Mattos MJ Jensen PR Hansen FG 2003 Precise determinations of C and D periods by flow cytometry in Escherichia coli K-12 and B/r Microbiology 149 1001 1010 12686642
Zhang X Bremer H 1996 Effects of Fis on ribosome synthesis and activity and on rRNA promoter activities in Escherichia coli
J Mol Biol 259 27 40 8648646
Batschelet E 1981 Circular statistics in biology Sibson R Cohen JE London Academic Press 371 p.
Rocha EP 2004 The replication-related organization of bacterial genomes Microbiology 150 1609 1627 15184548
Rocha EP Danchin A 2003 Essentiality, not expressiveness, drives gene-strand bias in bacteria Nat Genet 34 377 378 12847524
R Development Core Team 2004 R: A language and environment for statistical computing Vienna R Foundation for Statistical Computing
Ikemura T 1985 Codon usage and tRNA content in unicellular and multicellular organisms Mol Biol Evol 2 13 34 3916708
Efron B Tibshirani RJ 1993 An introduction to the bootstrap New York Chapman and Hall
Hirvonen CA Ross W Wozniak CE Marasco E Anthony JR 2001 Contributions of UP elements and the transcription factor FIS to expression from the seven rrn P1 promoters in Escherichia coli
J Bacteriol 183 6305 6314 11591675
Brunschede H Dove TL Bremer H 1977 Establishment of exponential growth after a nutritional shift-up in Escherichia coli B/r: Accumulation of deoxyribonucleic acid, ribonucleic acid, and protein J Bacteriol 129 1020 1033 320174
|
16103901
|
PMC1183518
|
CC BY
|
2021-01-05 09:18:23
|
no
|
PLoS Comput Biol. 2005 Jun 24; 1(1):e12
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010012
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science 1610390210.1371/journal.pcbi.001001305-PLCB-RA-0071R2plcb-01-01-12Research ArticleBioinformatics - Computational BiologyDevelopmentEvolutionSystems BiologyDrosophilaVertebratesmicroRNA Target Predictions across Seven Drosophila Species and Comparison to Mammalian Targets microRNA Targets across Seven Drosophila SpeciesGrün Dominic Wang Yi-Lu Langenberger David Gunsalus Kristin C Rajewsky Nikolaus *Center for Comparative Functional Genomics, Department of Biology, New York University, New York, New York, United States of AmericaEisen Michael EditorLawrence Berkeley National Laboratory, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 24 6 2005 1 1 e137 4 2005 2 6 2005 Copyright: © 2005 Grün et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.microRNAs are small noncoding genes that regulate the protein production of genes by binding to partially complementary sites in the mRNAs of targeted genes. Here, using our algorithm PicTar, we exploit cross-species comparisons to predict, on average, 54 targeted genes per microRNA above noise in Drosophila melanogaster. Analysis of the functional annotation of target genes furthermore suggests specific biological functions for many microRNAs. We also predict combinatorial targets for clustered microRNAs and find that some clustered microRNAs are likely to coordinately regulate target genes. Furthermore, we compare microRNA regulation between insects and vertebrates. We find that the widespread extent of gene regulation by microRNAs is comparable between flies and mammals but that certain microRNAs may function in clade-specific modes of gene regulation. One of these microRNAs (miR-210) is predicted to contribute to the regulation of fly oogenesis. We also list specific regulatory relationships that appear to be conserved between flies and mammals. Our findings provide the most extensive microRNA target predictions in Drosophila to date, suggest specific functional roles for most microRNAs, indicate the existence of coordinate gene regulation executed by clustered microRNAs, and shed light on the evolution of microRNA function across large evolutionary distances. All predictions are freely accessible at our searchable Web site http://pictar.bio.nyu.edu.
Synopsis
MicroRNA genes are a recently discovered large class of small noncoding genes. These genes have been shown to regulate the expression of target genes by binding to partially complementary sites in the mRNAs of the targets. To understand microRNA function it is thus important to identify their targets. Here, the authors use their bioinformatic method, PicTar, and cross-species comparisons of several newly sequenced fly species to predict, genome wide, targets of microRNAs in Drosophila. They find that known fly microRNAs control at least 15% of all genes in D. melanogaster. They also show that genomic clusters of microRNAs are likely to coordinately regulate target genes. Analysis of the functional annotation of target genes furthermore suggests specific biological functions for many microRNAs. All predictions are freely accessible at http://pictar.bio.nyu.edu. Finally, Grün et al. compare the function of microRNAs across flies and mammals. They find that (a) the overall extent of microRNA gene regulation is comparable between both clades, (b) the number of targets for a conserved microRNA in flies correlates with the number of targets in mammals, (c) some conserved microRNAs may function in clade-specific modes of gene regulation, and (d) some specific microRNA–target regulatory relationships may be conserved between both clades.
Citation:Grün D, Wang YL, Langenberger D, Gunsalus KC, Rajewsky N (2005) microRNA target predictions across seven Drosophila species and comparison to mammalian targets. PLoS Comput Biol 1(1): e13.
==== Body
Introduction
Recently, it has been discovered that the genomes of animals contain hundreds of microRNA genes. These small noncoding genes are typically transcribed by RNA polymerase II, processed into hairpins, and exported into the cytoplasm, where they are cleaved by the central enzyme of the RNAi pathway, Dicer, to form single-stranded mature microRNAs [1,2]. In animals, mature microRNAs are thought to bind to partially complementary binding sites in the mRNAs of target genes and, by unknown mechanisms, to regulate their post-transcriptional expression. In all known cases microRNAs repress expression of protein-coding target genes, either by repressing translation while not affecting the mRNA concentration of the target, or potentially by directly inducing a decrease in target mRNA concentrations [3]. To understand the biological function of microRNAs it is therefore important to identify their targets. Since high-throughput experimental methods for microRNA target identification have not been published yet, computational methods that try to identify target sites based on their partial complementarity with microRNAs have become increasingly important [4–13]. In flies, the sensitivity of these methods was sufficient to predict roughly eight targets per microRNA above noise, although the true number of targets has been estimated to be much higher [14]. Cross-species comparisons, which allow for the identification of evolutionarily conserved and thus likely functional target sites, have proven very helpful to boost the sensitivity of microRNA target detection. Recently, three independent studies based on cross-species comparisons of eight vertebrates concluded that in vertebrates, microRNAs are predicted to regulate at least 20%–30% of all genes [8,13,15]. These findings are consistent with experimental results [3].
It has also been widely suggested that microRNAs, similarly to transcription factors, can act in combination (or cooperatively) by binding to the same mRNA in a concentration-dependent manner. Tissue specificity of gene expression could then be in part explained by a “microRNA code” [16] of tissue-specific expression of the trans-acting microRNAs. This idea is supported by experiments [17] and by results from computational approaches that have been used to search for target sites of different microRNAs in the same target mRNA [5,6,13]. In particular, a mammalian gene was predicted and experimentally shown to be coordinately regulated by several co-expressed microRNAs [13].
We used our microRNA-target-finding algorithm, PicTar [13], and cross-species comparisons of seven recently sequenced Drosophila species to predict and analyze microRNA targets in flies. Our underlying model for target site recognition and a comparison of these results to our previous predictions [9] is presented in the Discussion. We also computed predictions for common targets of clustered microRNAs, since recent experiments [18,19] have suggested that microRNA genes that reside in clusters spanning roughly 50 kbp of genomic DNA tend to be co-expressed. To shed light on the specific function of microRNAs, we analyzed the functional annotation for predicted target sets using Gene Ontology (GO) terms [20]. However, to arrive at a more global understanding of microRNA function we then asked whether the extent of microRNA targeting in flies is comparable to targeting in vertebrates, whether certain microRNA–mRNA regulatory relationships are conserved between both clades, and whether individual microRNAs could potentially play a role in clade-specific gene regulation.
Results
Genome-Wide Cross-Species Comparisons of Seven Fly Species Allow High-Specificity and High-Sensitivity microRNA Target Predictions
It has been widely demonstrated that the success of the computational identification of microRNA target sites can be significantly boosted by searching for target sites that are evolutionarily conserved, and therefore likely to be functional. Thus, we set out to make use of the very recent whole-genome sequencing of a number of fly species (Figure 1). The genomic sequence for eight of these species, which include members of the melanogaster, obscura, repleta, and virilis groups, have been already assembled (D. melanogaster, D. simulans, D. yakuba, D. erecta, D. ananassae, D. pseudoobscura, D. virilis, and D.
mojavensis). We discarded the D. simulans assembly since it proved to contain large gaps. The estimated divergence time for these species ranges from a few million years to roughly 40 million years (Figure 1).
Figure 1 Phylogenetic Tree of 12 Drosophila Species
Our datasets include 3′ UTRs for seven of these species: D. melanogaster, D. yakuba, D. erecta, D. ananassae, D. pseudoobscura, D. virilis, and D.
mojavensis. Species underlined in solid red are present in set 1 and set 2. D. erecta (broken red line) is present only in set 2. Source: http://species.flybase.net/.
To identify evolutionarily conserved microRNA target sites in 3′ UTR sequences, it was critical to identify orthologous mRNAs. We experimented with two independently produced sets of genome-wide alignments of the eight species (see Materials and Methods). The first set of alignments (termed set 1), which does not contain sequence for D. erecta, was produced by the UCSC Genome database (http://genome.ucsc.edu/) and is based on pairwise alignments that were subsequently multiply aligned. The second set (termed set 2) came from true genome-wide multiple alignments (C. Dewey, MERCATOR, http://hanuman.math.berkeley.edu/~cdewey/mercator/) [21]. For both sets, we extracted multiple alignments of D. melanogaster 3′ UTRs using the D.
melanogaster FlyBase annotation for 18,892 gene transcripts and obtained 3′ UTR alignments across all eight species for 13,465 transcripts (set 1) and 13,030 transcripts (set 2) (Table 1). We also defined sets of alignments by keeping only the longest 3′ UTR from all transcript variants for the same gene, resulting in approximately 9,800 alignments for each set (termed unique alignments). The coverage of genes is thus roughly comparable between both sets. Additionally we masked repeats in the unique alignments using the UCSC repeat masks for set 1 and using the Tandem Repeat Remover [22] following Rajewsky et al. [23] for set 2. The nucleotide space of the various alignment sets is listed in Table 2 and comprises for each set a total of 2.2–4.1 Mb per species for the repeat-masked unique alignments. Masking repeats thus removed substantial amounts of sequence (up to 22% per species).
Table 1 Statistics of the 3′ UTR Multiple Alignments
Total number of UTR alignments with sequence for all species up to the indicated one, referring to the order D. melanogaster, D. yakuba, D. erecta, D. ananassae, D. pseudoobscura, D. virilis, D. mojavensis.
Table 2 Number of Aligned 3′ UTR Nucleotides
Total number of nucleotides per species in the multiple alignments for set 1 and set 2 (for all genes and for unique genes with both masked and unmasked repeats).
To identify conserved microRNA targets, we used the algorithm PicTar [13]. The key component of PicTar is the notion of a “nucleus” (or “seed”), defined as a stretch of seven bases (starting at the first or second position from the 5′ end of the microRNA), with consecutive perfect Watson–Crick basepairings to the target site. A recent computational and experimental study [14] demonstrated that the presence of such a nucleus is necessary for a substantial fraction of all microRNA target sites in Drosophila. For the remaining sites the nucleus is imperfect and contains mismatches, bulges, or G:U basepairings. Experimental results have suggested that sites with imperfect nuclei seem to be functional only when compensated by additional binding of the 3′ end of the microRNA to the target site [14,17]. Input to PicTar consists of orthologous, aligned 3′ UTR sequences and a search set of one or several microRNAs. PicTar first determines candidate 3′ UTR alignments containing a minimal number of conserved perfect nuclei, termed anchor sites. The minimal number and the degree of conservation of anchor sites are defined by the user. Each candidate UTR is searched separately for sites with perfect and imperfect nuclei. Subsequently, imperfect sites are required to pass a free energy filter. This is currently set to maximally two-thirds of the free energy of the perfectly basepaired microRNA–mRNA duplex and thus removes the vast majority of sites with imperfect nuclei. Sites with a perfect nucleus may optionally be subject to a much milder free energy filtering step (depending on the settings). Finally PicTar computes a score (see Materials and Methods) reflecting the likelihood that a given UTR will be targeted by members of the search set based on a hidden Markov model.
To estimate the extent of microRNA targeting in Drosophila, we used PicTar to count conserved putative target sites with perfect nuclei (anchors). The microRNAs used for these searches consisted of all currently known microRNAs that seemed to be conserved in all species under consideration (see Materials and Methods). To avoid counting target sites more than once, we represented all microRNA “families” that share identical nuclei by just one member of each family. The final set contained 46 microRNAs with unique nuclei conserved in all flies. As in our previous study [13], we recruited cohorts of randomized microRNA sequences to estimate the number of false positives (see Materials and Methods). Specifically, we computed all anchor sites (single conserved nuclei) for set 1 and set 2 with masked and unmasked repeats for real microRNAs, as well as for five sets of randomized cohorts in each case (Figure 2). A measure for the specificity is the signal-to-noise ratio, which is defined as the ratio of the number of anchor sites for real versus randomized microRNAs. In each case, we averaged the result over five cohorts and computed the mean and the standard deviation of the signal-to-noise ratio. We computed specificity and sensitivity, requiring different degrees of evolutionary conservation of anchor sites both with and without free energy filtering (Figure 2). Overall, we observed that using the free energy filter or masking repeats tends to enhance specificity with modest losses in sensitivity. We obtained higher signal-to-noise ratios with set 2, but a higher sensitivity with set 1. We also found that requiring different degrees of evolutionary conservation of anchor sites strongly affects sensitivity and specificity. More precisely, searching for anchor sites conserved between all flies (at various parameter settings) yielded a signal-to-noise ratio of 2.8–3.6 (set 1) and 3.3–4.0 (set 2). The sensitivity was, on average, 25–33 (set 1) and 15–29 (set 2) anchor sites per microRNA above noise. Anchor sites conserved in the melanogaster and obscura groups yielded signal-to-noise ratios of 2.1–2.4 (set 1) and 2.3–2.7 (set 2) with a sensitivity of 47–57 (set 1) and 29–40 (set 2) anchor sites per microRNA above noise (Figure 2).
Figure 2 Signal-to-Noise Ratios of the PicTar Single Target Site Predictions
For both set 1 and set 2 the predicted number of anchor sites for 46 unique microRNAs, conserved in all flies, and corresponding randomized microRNAs (averaged over five cohorts) and the respective signal-to-noise ratio (indicated above the bars) are shown with and without using free energy filtering of anchor sites for UTRs with either masked and unmasked repeats.
(A) Predictions for set 1 with anchor sites conserved in the melanogaster and obscura groups.
(B) Predictions for set 1 with anchor sites conserved in all flies.
(C) Predictions for set 2 with anchor sites conserved in the melanogaster and obscura groups.
(D) Predictions for set 2 with anchor sites conserved in all flies.
Based on these results we defined three settings, termed S1, S2, and S3 (see Materials and Methods) that allowed us to adjust the trade-off between sensitivity and specificity, and to generate predictions of high sensitivity, high specificity, and medium specificity/sensitivity, respectively. For each of the settings S1–S3 we recorded the specificity and the number of targeted transcripts as a function of the PicTar score cutoff, i.e., discarding all predictions with a score lower than a given threshold (Figure 3). We found that high-scoring transcripts tended to have a significantly improved specificity. For example, when using setting S3 the signal-to-noise ratio can be improved by a factor of 1.7 while retaining a sizeable number of predicted transcripts per microRNA. The positive correlation between specificity and PicTar score is consistent with our observation that some non-anchor sites make a contribution to the score. These sites appear to be “scattered”, i.e., are present only in some species or are not found in all species at the same position in the alignment. We experimented with relaxing our anchor site definition to include cases where a perfect nucleus is found in all species under consideration but not necessarily at overlapping positions in the alignments. The signal-to-noise ratio decreased in all settings S1–S3 (for example for S3 from 3.3 to 2.6), with no significant gain in sensitivity. We thus concluded that many scattered sites could be functional but should be scored only when they occur in conjunction with anchor sites, as implemented in the PicTar algorithm.
Figure 3 Sensitivity and Specificity as a Function of PicTar Score
Shown is the average number of predicted targeted genes as a function of a PicTar score cutoff (discarding all target genes with a score below this cutoff) for three different PicTar settings (S1–S3; see Materials and Methods): (A) high-sensitivity setting (S1), (B) high-specificity setting (S2), and (C) medium sensitivity/medium specificity setting (S3). The signal-to-noise ratio also depends on the score cutoff and is indicated above the curve for certain cutoff values. All predictions for all settings can be accessed on the PicTar Web server (not filtered by score cutoffs).
Previous analyses of microRNA targeting in vertebrates [6,8,13,15] and flies [5,14] suggested that a substantial fraction (10%–30%) of all protein-coding genes in both clades are regulated by microRNAs. Using settings S3 (or S2), we found that 15% (13%) of all annotated roughly 10,000 unique melanogaster 3′ UTR transcripts (corresponding to approximately 10,000 genes) have at least one anchor site that is conserved in all seven fly species at a signal-to-noise ratio of about three (four). Thus, with settings S3 or S2, roughly 10% of all transcripts are predicted to be targeted by microRNAs above noise in all flies. To estimate how many genes could be regulated by more than one microRNA, we counted all transcripts with at least two anchor sites. Applying the high-specificity setting S2, we found that searching for multiply targeted transcripts further enhances the specificity to a significant degree (Figure 4). For example, we found seven times as many targeted transcripts with at least two anchor sites for real microRNAs compared to randomized microRNAs. With settings S2 and S3, we predicted that 30% of all targeted transcripts have more than one anchor site. Finally, for our high-sensitivity setting S1 we found that 27% of all transcripts have at least one anchor site at a single-site signal-to-noise ratio of approximately 2.2. Of these, 40% are found to have at least two anchor sites.
Figure 4 Specificity of PicTar Predictions of Genes with Multiple Putative Target Sites
Number of unique genes as a function of the minimal number of anchor sites for 46 unique, conserved microRNAs and for randomized microRNAs (averaged over five cohorts). The ratio of these numbers, reflecting the specificity, is indicated above each bar.
In summary, based on our high-sensitivity setting, we predicted that at least 15% of all D.
melanogaster genes with currently annotated 3′ UTR sequences are regulated by at least one known microRNA, and that at least one-fifth of these Drosophila microRNA targets could be subject to coordinate control by two or more microRNAs from different microRNA families (above noise). We provide ranked PicTar target predictions for all conserved microRNAs, all FlyBase transcripts, and settings S1–S3 at our searchable Web site (http://pictar.bio.nyu.edu). The results, linked to various other public databases, can be queried for genes of interest or microRNAs of interest.
Recovery of Experimentally Validated microRNA Targets in Drosophila
We have previously shown that PicTar has an excellent recovery rate of validated Caenorhabditis elegans microRNA targets [13]. To analyze the recovery of experimentally validated targets in Drosophila, we collected 19 microRNA–target regulatory relationships from the literature [4,12,24]. The overlap with PicTar predictions across settings S1–S3 is summarized in Table 3. The apoptosis gene hid/wrinkled is targeted by the microRNA bantam [24]. For all settings S1–S3, hid is the top-scoring bantam target (PicTar score of 17.3) and has five anchor sites conserved in all flies. Notably, hid targeted by bantam has the second highest PicTar score within all our target predictions. The only gene with a higher score (40.5) is nerfin-1, which contains two anchor sites for miR-286 (or equivalently miR-279) conserved in all flies, and many additional sites for the same microRNA (see Discussion).
Table 3 Recovery of Published Drosophila microRNA Targets with Experimental Support
Experimentally assayed microRNA target sites are listed in the second column, comprising 19 microRNA–gene regulatory relationships with various degrees of experimental support and nine sites that did not show regulatory activity. Columns labeled by S1–S3 refer to the recovery of sites at the corresponding PicTar setting.
The Notch signaling gene hairy was recently predicted [4,9] and validated as a target of miR-7 with a single binding site [4]. PicTar found a miR-7 anchor site conserved in all flies of the melanogaster and obscura groups, whereas the site in D. virilis appears to be slightly shifted upstream. Hence, this target is recovered with setting S1 but not with settings S2 and S3. There is experimental evidence that miR-7 also targets HLHm3 and E(spl)m4, two genes that are located in the E(spl) complex [4]. For HLHm3, PicTar predicts one miR-7 target site conserved in all flies (with all settings). The gene E(spl)m4 did not have an annotated 3′ UTR but was recovered after adding the likely 3′ UTR sequence to our dataset [4]. Another gene of the E(spl) complex, HLHm5, is the highest ranking target gene of miR-7 when searching for targets conserved in all flies (with setting S2; rank 2 with setting S3). Target predictions at a reduced level of conservation (setting S1) also yield HLHm5 as the top-ranking miR-7 target. The Notch gene Bearded is recovered as a target of miR-4 (or miR-79, equivalently). With setting S1 we found three conserved sites in its 3′ UTR. These so called Bearded boxes have been shown to mediate repression of a reporter gene with a Bearded 3′ UTR in vivo [25]. This gene is again very high scoring (15.6) and ranks second in the list of miR-4 target predictions (setting S1). This target is not recovered with the other settings, because the alignments of this gene do not contain sequence for D. mojavensis and D. virilis. The same microRNA is thought to repress bagpipe [14], which ranks second in the list of miR-4 target predictions (S3).
The proapoptotic genes reaper, grim, and sickle are validated targets of the miR-2 family [4]. For sickle we found one conserved site in all flies for miR-2, miR-13, and miR-6, which share the same nucleus. For reaper, we recovered one site for the same microRNAs in the melanogaster and obscura group with setting S1, while the other settings failed to identify this target because of missing sequence for this gene in D. mojavensis.
grim is the only target of this group not recovered by PicTar, because it has only a 6mer nucleus for miR-2.
A recent algorithm for the prediction of microRNA targets did not rely on evolutionary information, but incorporated the 3′ UTR secondary structure to compute putative microRNA targets [12]. Some of the high-scoring predictions could then be supported by luciferase reporter constructs in cell lines. We recovered four targets from this list (miR-7/HLHm5, miR-279/SP555, miR-124/Gli, and miR-310/imd) but failed to locate conserved nuclei for the other six targets (see comments in Table 3). Strikingly, out of nine computationally predicted targets that were experimentally assayed but did not show any repression activity (likely false positives) [12], we only predicted one microRNA–target regulatory relationship (miR-286/boss).
In summary, PicTar recovered 8/9 (89%) of all known targets with experimental in vivo evidence and 4/10 (40%) of targets with other experimental support with setting S1, i.e., requiring conservation of anchor sites only in flies of the melanogaster and obscura groups. Only three of all targets with experimental support were lost when requiring conservation between all fly species and thus were not recovered with settings S2 and S3.
Some Clustered microRNAs Are Likely to Coordinately Regulate Gene Expression
Expression assays have shown that microRNA genes that are located in the same genomic region within 50 kb of each other are often co-expressed [18,19], suggesting the possibility that they may coordinately regulate common target genes. In D. melanogaster, we identified seven clusters within 50-kb regions that contained precursors of at least two conserved microRNAs from different families. To identify common targets of clustered microRNAs in flies, we used PicTar to predict coordinate targets for each of these microRNA clusters (available on the PicTar server). Table 4 gives an overview of all clusters, their location in the Drosophila genome, the abundance of targeted transcripts, and, whenever all microRNA genes of a given cluster are located in an intron of another gene, the identifier of this gene. To evaluate whether clustered miRNAs target the same gene more often than expected by chance, we considered all 1,128 pairwise combinations of all 48 unique conserved microRNAs. While pairs of microRNAs from the same cluster make up only 2.1% of these pairs, 132 genes contained at least one anchor site for each microRNA of these clustered pairs (using setting S1), or 12% of the 1,104 genes that contain at least two different anchor sites for any combination of these 48 microRNAs. Thus, some pairs of microRNAs from clusters are likely to coordinately regulate a significantly higher proportion of genes (12%) than expected (2.1%). Furthermore, the number of target genes predicted for pairs of clustered microRNAs is twice the number expected from randomly drawn sets of 24 pairs among the 48 conserved microRNAs, which is significant by three standard deviations (see Materials and Methods). These findings support the hypothesis of coordinate control executed by clustered microRNAs.
Table 4 Clusters of microRNAs and Their Number of Predicted Target Genes
Clusters of unique microRNAs, conserved in all flies, with precursor sequences, originating from a genomic region of less than 50 kB. The number of unique genes with at least two anchor sites for different microRNAs of a given cluster is indicated. Predictions are computed for both set 1 and set 2, and for anchors conserved in the melanogaster and obscura groups, and in all seven fly species. If clustered microRNA precursors reside in an intron of an annotated FlyBase gene, the identifier is also indicated.
Biological and Molecular Classification of Predicted microRNA Targets
To gain insight into the function of Drosophila microRNAs, we used GeneMerge [26] to analyze the over-representation of specific GO terms [20] in the functional annotation of genes predicted to be targeted by a particular microRNA versus a background gene set (see Materials and Methods). To avoid potentially spurious statistical significances, we chose not to use all genes as the background, but constructed a background set comprising all predicted targets for both real and randomized microRNAs. From the “biological process” ontology, a total of 112 significantly over-represented GO terms were identified; 70% of the gene sets targeted individually by conserved microRNAs and two sets of combinatorial target predictions for microRNA clusters contained at least one over-represented GO term (Figure 5A). For the “molecular function” ontology, a total of 25 significantly over-represented GO categories were obtained among 36% of all individual microRNA target gene sets and one set of microRNA cluster targets (Figure 5B). Consistent with previous estimates [1,2], our data indicate that microRNAs regulate a large variety of genes in many different biological processes. Globally prominent GO terms were morphogenesis, organogenesis, development (including embryonic development, and anterior/posterior and dorsal/ventral axis specification), neurogenesis, signal transduction (including Notch, Torso, Sevenless, and Frizzled signaling), and transcriptional regulation. Our overall overlap with another GO analysis for fly microRNA targets in a recent study was marginal, very likely because of not only the differences in approaches for identifying over-represented GO terms, but also the different nature of target site predictions made by PicTar and the published miRanda algorithm [5].
Figure 5 Significant GO Terms among the Predicted Target Genes of All Single microRNAs and Clusters of Co-Expressed microRNAs
Significantly enriched GO terms for (A) “biological processes” and (B) “molecular function” ontologies. Shown are GO terms with p-values smaller than 0.1, corrected for multiple testing. Hierarchical clustering was performed separately for GO terms and microRNAs (see Materials and Methods).
Our data were consistent with and extended results from a recent study that used GO functional analysis to predict microRNA target genes [4], in which miR-7 was predicted to be active in Notch signaling and miR-277 in valine, leucine, and isoleucine degradation. For miR-277, we recovered all nine predicted targets and found five additional genes (CG3267, CG4389, CG4600, CG6638, and CG8778) at p < 10−7. Targets of miR-7 predicted by PicTar included many Notch pathway genes as well as targets of Notch signaling, including E(spl)m5, Tom, Bob, E(spl)mγ, Bearded, E(spl)m3, and E(spl)m4, most of which were very high scoring (using setting S1). Furthermore, many targets of Notch signaling were also predicted as targets of the Bearded-box microRNAs miR-4 and miR-79 (E(spl)m5, Bearded, E(spl)mγ, and Tom) and of the K-box microRNAs miR-2 and miR-11
(E(spl)m5, E(spl)m2, E(spl)mδ, and E(spl)m3), consistent with previous observations [27]. Other known Notch targets would have been included in PicTar's target lists if their 3′ UTRs were annotated in the current FlyBase release (data not shown). We note that the majority of Notch targets predicted by PicTar would not have been predicted if stringent free energy filtering were applied for predicted microRNA–target duplexes with perfect nuclei.
Comparison of microRNA Targets between Flies and Vertebrates
Previously, we applied PicTar to exhaustively search 3′ UTR alignments of eight vertebrates (human, chimpanzee, mouse, rat, dog, chicken, pufferfish, and zebrafish) for microRNA target sites [13]. To compare the extent of microRNA targeting in flies and vertebrates, we first compared length, repeat content, and conservation of 3′ UTRs between both clades, using our datasets derived from the UCSC database for consistency. We focused on the comparison of 3′ UTRs between D. melanogaster and human since 3′ UTRs from these species were extracted based on annotated transcripts. We found that the length distribution of 3′ UTRs and the distribution of repeats within them are very similar between all mammals and between all flies, respectively, so comparisons between human and D. melanogaster UTRs should reveal essential differences between the two clades. We found a much broader distribution of 3′ UTR lengths in mammals than in flies, yielding on average approximately 900 nucleotides per 3′ UTR for human and approximately 400 nucleotides per 3′ UTR in D. melanogaster (Figure 6), consistent with previous results [28]. Examining the contribution of repeat elements, we found that repeats constitute 11% of all human 3′ UTR sequences compared with 4% in D. melanogaster (Table 5). Interestingly, for short repeats (up to about 50 nucleotides), the length distribution in D. melanogaster and human is similar (Figure 7). For longer elements the distribution in flies continues to decay exponentially with the same slope, whereas the human distribution displays a broad tail with another significant peak centered around approximately 300 nucleotides. To analyze 3′ UTR conservation, we counted all 7mers that appeared to be perfectly conserved in each 3′ UTR multiple alignment and divided these counts by the length of the 3′ UTR sequence. We found that the probability of a nucleotide to reside in a conserved 7mer is comparable between vertebrate alignments (including human, chimp, mouse, rat, dog, and chicken) and alignments covering all fly species in our dataset (0.02 and 0.03, respectively). Similarly, 3′ UTR conservation is comparable between mammals and flies in the melanogaster and obscura groups (0.06 and 0.08, respectively). The contribution of repeat elements to conserved 7mers is substantially different in vertebrates and flies (Table 6). Masking repeats reduced the number of bases in conserved 7mers by about 1% in vertebrates and about 10% in flies. Thus, repeats in 3′ UTRs appear to be much better conserved in flies than in vertebrates and thus may be of functional importance in flies.
Table 5 Repeat Elements in 3′ UTRs of Human and D. melanogaster
Fraction of repeats in the 3′ UTRs of human and D. melanogaster.
Table 6 Conservation of 7mers in 3′ UTRs of Vertebrates and Flies
Fraction of nucleotides residing in 7mers conserved in all flies up to the indicated one (referring to the order D. melanogaster, D. yakuba, D. erecta, D. ananassae, D. pseudoobscura, D. virilis, D. mojavensis) and in vertebrates, with and without the inclusion of repeat elements. Comparison of Table 4 and 5 demonstrates that in vertebrates (flies), repeat elements share less (more) nucleotides than expected with conserved 7mers.
Figure 6 Lengths Distribution of 3′ UTRs in Human and D. melanogaster
Data for set 1 and set 2 on a logarithmic scale. The distribution decays exponentially with increasing length in human much slower than in D. melanogaster. The average 3′ UTR lengths in human and D. melanogaster are approximately 900 and approximately 400 nucleotides, respectively.
Figure 7 Length Distribution of Repeat Elements in 3′ UTRs of Human and D. melanogaster
Data for set 1 on a logarithmic scale. The distribution peaks strongly for both species at a length of 11 nucleotides and decays exponentially for longer repeat elements in D. melanogaster. Up to a length of roughly 50 nucleotides, both distributions are very similar, while for longer elements the distribution for human no longer decays exponentially, but has a broad tail with another significant peak at a length of approximately 300 nucleotides.
The extent of microRNA regulation seems roughly comparable between mammals and flies overall, with several interesting clade-specific differences. In vertebrates, we and others [6,8] found that roughly 30% of all genes may be regulated by microRNAs. This is twice the number we found in flies (15%), but this could be explained by the smaller number of known microRNAs in flies and other reasons (see Discussion). More interestingly, we checked whether individual microRNAs appeared to target similar or significantly different numbers of genes in mammals versus flies, since such differences could be indicative of clade-specific changes in microRNA function. To retain a reasonable sensitivity in target predictions for this analysis, we used human, chimp, mouse, rat, and dog for target predictions in mammals and the melanogaster and obscura groups for predictions in flies. We defined a set of 48 homologous microRNAs in mammals and flies (see Materials and Methods) and computed the average number of microRNA targets in both clades. We then calculated the ratio of predicted targets per microRNA to the average separately for each clade (Table 7). A scatter plot of these ratios (Figure 8) demonstrates a correlation between the numbers of targeted genes for homologous microRNAs in mammals and flies. However, certain microRNAs appear to have a significantly higher number of target genes in either humans (miR-10, miR-133, miR-125, let-7, and miR-285) or flies (miR-184 and miR-210). For example, for let-7 we found 1.64 as many target genes as expected on average in mammals, but only around 50% of the average expected number in flies. It is impossible to determine from this analysis whether microRNAs have acquired more targets in one clade or lost targets in the other, but it is striking that both human homologs of the fly microRNAs miR-184 and miR-210 are expressed at low abundance across many human tissues, while the homologs of miR-10, miR-133, miR-125, let-7, and miR-285 are expressed overall at much higher levels [19]. We stress that the human homologs of miR-10 and miR-133 have average or below average numbers of predicted targets in human. Our data indicate that the above seven microRNAs may function in clade-specific modes of gene regulation.
Table 7 Homologous microRNAs between Mammals and Flies of the melanogaster and obscura Groups and Their Respective Number of Target Genes
The ratio of the number of target genes for a particular microRNA to the number of target genes averaged over all microRNAs is indicated for flies and for vertebrates (termed relative abundances). The ratio of the relative abundances between flies and mammals is plotted in Figure 8.
aIn melanogaster and obscura, in units of the average number of targeted genes per microRNA.
bIn mammals, in units of the average number of targeted genes per microRNA.
Figure 8 Number of Predicted Target Genes for Homologous microRNAs between Mammals and Flies
Scatter plot for relative numbers of targeted genes predicted for homologous microRNAs in mammals and flies. The ratio of the number of predicted target genes of a microRNA and the average number of putative targeted genes per microRNA are plotted in mammals (y-axis) versus flies (x-axis). Conservation in flies included the melanogaster and obscura groups. Outliers (with a ratio of relative numbers of predicted target genes larger than 3.0 or smaller then 0.33) are circled. The microRNA identifiers refer to microRNAs annotated in D. melanogaster.
Finally, we computed which regulatory microRNA–mRNA relationships seemed to be conserved between flies and mammals (see Materials and Methods). From all 8,136 homologous human–D. melanogaster gene pairs in our dataset, 50 unique gene pairs were predicted to be targeted by homologous microRNAs (listed in Table S1). These 50 pairs comprise approximately 60 microRNA–mRNA regulatory relationships. Although these numbers are small, stringent permutation tests indicated that the result was marginally significant (1.7 standard deviations) (see Materials and Methods). Perhaps not surprisingly, almost half of the 50 D. melanogaster genes belong to the GO category “development,” and “histogenesis” is assigned to 13 of these 24 genes. Both results are significant (see Materials and Methods).
Discussion
The Extent of Post-Transcriptional Gene Regulation in Drosophila Mediated by microRNAs
The sequencing of the genomes of several Drosophila species proved to be an invaluable resource for the analysis of microRNA targets in flies. Cross-species comparisons allowed us to arrive at significantly enhanced sensitivity and specificity for microRNA target predictions in comparison with recent approaches. For example, previous studies have predicted on average eight target genes per microRNA (see [14] and references therein), while our data allow us (with high-sensitivity setting S1) to predict 54 target genes per microRNA above noise in D. melanogaster. Requiring conservation in all flies, we still predict on average more than 23 and 30 target genes per microRNA, for settings S2 and S3 respectively, at a strongly enhanced signal-to-noise ratio.
Based on our target predictions, we found that currently known microRNAs are expected to regulate a large fraction of all D. melanogaster genes (15%). This number is almost certainly an underestimate, since (a) the annotation of 3′ UTRs is incomplete, (b) the genomic sequences of several fly species still contain large gaps, and (c) it is expected that many more microRNAs in fly remain to be discovered. Indeed, using an approach analogous to that of a recent comparative study of mammals [15], we analyzed fly 3′ UTRs across all seven species and found strong evidence for the existence of a substantial number of yet undiscovered fly microRNA genes (N. Rajewsky, unpublished data).
The number of targets per microRNA we predicted is consistent with recent estimates of the true number of microRNA targets by Brennecke et al. [14]. In that study, the authors analyzed the statistical significance of conserved 8mer nuclei and conserved 7mer nuclei and concluded that the vast majority of computationally detectable target sites possessed at least one conserved 7mer nucleus. Our method is similar to this approach, but differs in the larger number of species included in our conservation analysis. Requiring similar levels of sequence conservation yields roughly comparable numbers of target genes per microRNA for both methods. In a number of cases in our dataset, gaps in the assemblies artificially decrease the number of predicted targets. On the other hand, using all seven Drosophila species allowed us to almost double the signal-to-noise ratio. In the future, further completion of the assemblies of the Drosophila genomes will almost certainly boost the number of PicTar predictions.
Comparison to Our Previous Algorithm
Previously, we had published an algorithm for microRNA target identification and used it to predict microRNA targets within a set of central developmental genes involved in the body patterning of Drosophila [9]. In our model for target site recognition, we had introduced the notion of the nucleus as a stretch of perfect Watson–Crick basepairings between the microRNA and the target site and had shown that the nucleus (a) is typically 6–8 bases long, (b) is the central component of the specificity of target recognition, and (c) may serve as a nucleation site to allow a rapid zip up of the nucleus region of the microRNA–mRNA duplex [9]. This model for target site recognition explicitly proposed an explanation for the physical basis of target site recognition that combined kinetic and thermodynamic components. A recent experimental paper supports this idea [29]. We had also observed that the position of the nucleus within the microRNA is oftentimes conserved and at the 5′ end, indicating that the same cis-regulatory motif may be used to coordinate the action of a microRNA across different genes. We compared our previously predicted microRNA–mRNA regulatory relationships to our current PicTar predictions. We found that out of all cases where genes were present in both datasets, 11 out of 30 previous predicted sites were precisely recovered by PicTar. A number of the predictions are not recovered by PicTar because our previous algorithm did not restrict the nucleus to the 5′ end of the microRNA.
Future PicTar Improvements
The highest scoring gene from all single microRNA target site predictions was nerfin-1, with two anchor sites for miR-286 conserved in all flies and many additional, non-aligned sites present in all flies. Errors or ambiguities in the alignment can oftentimes explain the presence of these “scattered” sites. Additionally, compensatory mutations could lead to non-aligned and yet functionally conserved target sites in a 3′ UTR. At present, PicTar scores these scattered sites in the same way as it scores conserved sites, as long as both of them occur in the same UTR. Future refinements of the algorithm should explore (a) explicit evolutionary models for the evolution of 3′ UTR sequences and microRNA target sites, (b) improved probabilistic scoring for sites with imperfect nuclei [14], (c) the incorporation of secondary structure information [12], (d) incorporation of mRNA expression levels (e.g., from microarray experiments), and (e) expression levels of microRNAs.
Our data indicated that some clustered microRNAs are likely to coordinately regulate target genes. In addition, it has been shown that clustered microRNAs are likely to be co-expressed. Using multiple co-expressed microRNAs to coordinately regulate target genes could be an efficient way to increase the specificity of target gene regulation, and may also enhance the robustness of target gene expression levels against fluctuations in individual microRNA concentrations. We note that our data only suggest that clustered microRNAs are more likely to coordinately regulate target genes by coordinate binding to their 3′ UTRs than non-clustered microRNAs. Many microRNAs that reside in clusters also seem to target genes without additional binding sites for microRNAs in the same cluster. Conversely, there appear to be many possibilities for microRNAs from different clusters to coordinately bind the same target genes.
The Evolution of microRNA Function across Large Evolutionary Distances
microRNAs offer the exciting possibility to study the evolution of trans-acting regulatory genes together with the evolution of their cis-regulatory target sites using computational methods. In this study, we have only touched upon this problem by comparing the estimated number of targeted genes per microRNA in one clade to the predicted number of targets for the homologous microRNA in another clade, which, by our definition of homology, is likely to bind to the same cis-regulatory sites. We caution that our definition of homology would also refer to microRNAs that may have evolved independently in one or both clades. However, our comparison yielded a nontrivial correlation between the numbers of targeted genes per microRNA in flies and vertebrates, indicating that the relative number of microRNA targets per microRNA tends to be conserved over very large evolutionary distances. In contrast, only a relatively modest number of specific microRNA–mRNA regulatory relationships seemed to be conserved between both clades. This scenario hints at conservation of global “network” features of gene regulation mediated by microRNAs while implicating microRNAs in an extensive rewiring of post-transcriptional gene regulation during organismal evolution.
It was striking that some microRNAs (including let-7) that are likely to have a large number of target genes in vertebrates seem to have a strongly reduced relative number of targets in flies, and vice versa. We singled out three microRNAs (miR-184, miR-304, and miR-210) with a drastically enhanced relative number of targets in flies compared to vertebrates. Our GO term analysis for microRNA targets revealed that one of them (miR-210) had over 70 predicted target genes, which as a group were significantly enriched (p < 0.03 after correcting for multiple testing) for 11 genes with the GO annotation “female gamete generation” (see Figure 5A). These 11 predicted miR-210 targets are cut, egghead, germ cell-less, gurken, lozenge, par-1, Ras oncogene at 85D, rhomboid-4, RNA-binding protein 9, singed, and slalom. Most of these genes are evolutionarily conserved and have a known role in Drosophila oogenesis, either in development and patterning of the oocyte or in differentiation of the somatic follicle cells that surround the developing egg chamber, and seven of the 11 are implicated in developmentally critical signaling pathways involving receptor tyrosine kinases, Notch, wingless, or hedgehog (see Protocol S1). Development of a mature Drosophila oocyte involves an elaborate sequence of events that must be precisely orchestrated in time. A surprising number of the genes in the above list play roles in important events that must take place within a specific window of time during oogenesis, many of which involve signaling between the germline and soma. Thus, an important emergent theme of miRNA regulation may revolve around the widespread need for precise control of spatiotemporally restricted events during development. In addition, oogenesis in Drosophila occurs through a very different developmental program than in vertebrates. It is thus intriguing that a single microRNA has potentially evolved to include a wide array of target genes that are important for this developmentally divergent process. However, many of these potential targets are not restricted to oogenesis but also function at other times and places, including the eye, nervous system, and epithelia, and a number of other predicted miR-210 targets also function in these tissues (e.g., arrowhead, cacophony, trio, Sema-1b, makorin, Van Gogh, Syntaxin 17, G-oα47A, RhoGAP92B, cul-2, Apc, and Scm). Thus, this microRNA may play more complex pleiotropic roles in developmental networks. We conclude that some microRNAs could be candidates for genes that mediate clade-specific differences in gene expression, and could play an important role in shaping the diversity of life.
Materials and Methods
3′ UTR alignments.
We used two sets of 3′ UTR alignments for flies. Set 1 was created on the basis of alignments, retrieved from the UCSC Genome Browser database at http://www.genome.ucsc.edu [30], by assembling aligned contigs of six fly species. The following assemblies were used to construct the multiz alignments [31]: D. melanogaster Apr. 2004 (dm2), D. yakuba Apr. 2004 (droYak1), D. ananassae Jul. 2004 (droAna1), D. pseudoobscura Aug. 2003 (dp2), D. virilis Jul. 2004 (droVir1), D. mojavensis Aug. 2004 (droMoj1), Anopheles gambiae Feb. 2003 (anoGam1), and Apis mellifera Jul. 2004 (apiMel1). The detailed amount of nucleotides and aligned sequence for all flies are shown in Tables 1 and 2. The 3′ UTR alignments of set 2 were extracted from genome-wide multiple alignments generated by the Pachter group at UC Berkeley (http://hanuman.math.berkeley.edu/genomes/drosophila.html) [21] using the following assemblies: D. melanogaster Apr. 2004 (dm2), D. ananassae Jul. 2004 (droAna1), D. yakuba Apr. 2004 (droYak1), D. erecta Oct. 2004, D. pseudoobscura Aug. 2003 (dp1), D. virilis Jul. 2004 (droVir1), D. mojavensis Dec. 2004. For both datasets we used FlyBase release 4.1 to extract 3′ UTRs in D. melanogaster.
microRNA sequences.
We downloaded all D. melanogaster microRNA precursors and mature microRNAs from the microRNA registry at Rfam [32] (release 5.0, http://www.sanger.ac.uk/Software/Rfam/mirna/index.shtml). For each microRNA, we checked for conservation of the precursor sequence in all fly species, using multiple alignments retrieved from the UCSC Genome database. We required the first 8mer of the mature microRNA to be perfectly conserved, but applied a less stringent conservation constraint, a percentage identity of 75%, to the precursor sequence. From the 79 mature D. melanogaster microRNAs, we found 69 to be conserved in all flies and 73 to be conserved in the melanogaster and obscura groups. Statistics were generated with a subset of 46 microRNAs with unique nuclei, i.e., each nucleus is specific for only one microRNA in this list. Lists of these microRNAs are provided as Tables S2–S4.
Randomized microRNAs.
Randomized microRNAs [13] were produced by extracting 8mers with the same genome-wide abundance (± 15%) in all D. melanogaster 3′ UTRs of the first and the second 7mer nucleus compared to the respective 7mers of the corresponding real microRNA. The 3′ end of the real microRNA was attached to this 8mer. We produced five cohorts of unique randomized microRNAs each for set 1 and set 2, in either case both with masked and unmasked repeats.
Different settings for PicTar predictions.
Comparing anchor site predictions based on the two different alignment sets (see Figure 2), we found that using alignment set 1 yielded an overall higher sensitivity, while target predictions based on set 2 had a higher specificity. A major determinant of sensitivity and specificity is the required level of conservation of anchor sites. According to these findings, we defined three PicTar settings (termed S1, S2, and S3) to cover the observed ranges of sensitivity and specificity. Masking repeats and applying free energy filtering of anchor sites served to fine-tune the trade-off between sensitivity and specificity for each setting. The high-sensitivity setting (S1) had repeat-masked UCSC alignments (set 1) as input sequences, required conservation of anchor sites only between species of the melanogaster and obscura groups, and applied no free energy filtering of perfect nuclei. Setting S2, providing high-specificity predictions, used alignments of set 2 with unmasked repeats as input sequences and required conservation of anchors in all flies and free energy filtering of perfect nuclei. The medium sensitivity/medium specificity setting S3 was equal to setting S1, but used conservation of anchors in all flies.
Phylogenetic PicTar score.
Given an alignment of a 3′ UTR for all flies, PicTar computes a likelihood score for the UTR of each species separately. The final score of the whole alignment is a weighted average of the single species scores, with weights reflecting the phylogenetic grouping of the species. More precisely, the score of all flies in the melanogaster subgroup was averaged and the resulting score was further averaged with the score for D. ananassae and D. pseudoobscura, yielding a score for the melanogaster and obscura groups. The scores for D. mojavensis and D. virilis, which have similar evolutionary distances to the melanogaster group, were averaged. This outgroup score and the score of the melanogaster and obscura groups were averaged to obtain the final PicTar score for all flies.
Homologous microRNAs between vertebrates and flies.
According to a recent study, the nucleus of a given microRNA is presumably sufficient to achieve repression of a gene [14]. We thus applied a relaxed definition of homology. Whenever the first or second 7mer of a microRNA in Drosophila was also present as one of the nuclei in a human microRNA, these two microRNAs were assumed to be homologs. Comparing all microRNAs conserved in the melanogaster and obscura groups with all microRNAs conserved in mammals, we obtained 48 pairs of homologous microRNAs between mammals and flies.
Target numbers for random microRNA pairs.
To assess the significance of targeting by 24 pairs of microRNAs extracted from clustered microRNA genes, we used 1,000 sets of 24 pairs of microRNAs drawn randomly from the set of all possible 1,128 distinct pairs (using all 48 unique microRNAs conserved in the melanogaster and obscura groups). For conservation of anchor sites in the melanogaster and obscura groups, on average 18 (± 2) out of 24 random pairs had at least one target gene, compared to 22 of the co-expressed pairs. We obtained on average 70 (± 21) unique target genes per random set, compared to 132 unique targets of the clustered pairs with a high Z-value (Z = 3). When requiring conservation between all flies, the results were more significant: 19 out of 24 clustered pairs targeted 50 unique genes, while on average 11 (± 2) out of 24 randomly drawn doublets were predicted to target approximately 23 (± 8) unique genes (Z = 3.5).
Homologous genes in vertebrates and flies.
Homologous genes between D. melanogaster and human were extracted from HomoloGene [33] (ftp://ftp.ncbi.nih.gov/pub/HomoloGene/current/) with annotations of 14 March 2005. This list contained 19,685 human genes and 7,983 fly genes. Keeping only pairs of homologous genes for which we were able to assign a FlyBase CG number and a RefSeq gene identifier [34], our reduced list contained 4,623 pairs of homologous genes. We extracted an additional list of homologous human–D. melanogaster transcripts from the Ensembl Genomebrowser (http://www.ensembl.org/). After merging both lists, we obtained a final list containing 8,136 pairs of homologous transcripts.
Shuffling test for homology relationships.
To asses the significance of the number of conserved microRNA–target relations of homologous target genes and microRNAs between vertebrates and flies, we shuffled homology relations in vertebrates and flies in the following way: All nonhomologous genes and microRNAs were discarded from our table of microRNA–target gene assignments. All microRNAs of a given family with equal 7mers at the 5′ end were represented by one specific member of this family. Similarly, we discarded multiple transcript variants, keeping only the longest variant for each gene. We constructed a list with assignments of each microRNA to all its target genes. Shuffling was performed by permuting the microRNA entries of this list, thereby assigning a new set of target genes to each microRNA. We counted the number of homology relationships for these permuted microRNA–target assignments and averaged the results over 1,000 runs. We obtained on average 45 (± 9) homology relationships for the shuffled lists, while we counted 60 real homology relationships, when using only unique lists of genes and microRNAs. The described shuffling strategy models a situation of nonconserved microRNA–target relations, but keeps the number of microRNAs targeting a particular gene constant.
GO term analysis.
To evaluate the PicTar target predictions for all single microRNAs, we searched for significantly overrepresented GO terms [20] of all target genes for each microRNA separately using the GeneMerge software [26]. GeneMerge computes the significance of occurrences of particular GO terms for a set of genes compared to a background gene set. To use an extensive background gene set that captures features of genes targeted by microRNAs as best possible, we lumped together all genes predicted to be targeted by all microRNAs (setting S1) or genes that were hit by the five cohorts of randomized microRNAs. Finally, p-values were conservatively corrected for multiple testing as provided by GeneMerge and recorded below a cutoff of 0.1. We performed the analysis separately for all GO terms in the “biological processes” ontology, and the most specific “biological processes” GO term for each gene, as well as for all GO terms in the “molecular function” ontology. These three classes of GO terms are provided by GeneMerge. Results from the first two analyses were merged into one output file, keeping the lower p-value for GO terms that were present twice. To visualize the results, we used two-way hierarchical clustering based on the linear correlation coefficient of the negative logarithm of the p-value [35]. To compute p-values for the overrepresentation of GO terms for genes that are (a) conserved between D. melanogaster and human, and (b) predicted to be targeted by homologous microRNAs in flies and mammals, we used a background gene set that was obtained by intersecting the background gene set described above with the set of all D. melanogaster genes with homologs in human.
Supporting Information
Protocol S1 Detailed Discussion of Predicted miR-210 Targets
(170 KB DOC).
Click here for additional data file.
Table S1 Homologous Genes between Flies and Mammals, Targeted by Homologous microRNAs
(71 KB XLS).
Click here for additional data file.
Table S2 Mature microRNAs Conserved in All Flies of Our Dataset
(18 KB XLS).
Click here for additional data file.
Table S3 Mature microRNAs Conserved in the melanogaster and obscura Groups
(20 KB XLS).
Click here for additional data file.
Table S4 Set of Unique, Conserved Mature microRNAs Used to Compute Signal-to-Noise Ratios
(18 KB XLS).
Click here for additional data file.
We acknowledge the Agencourt Bioscience Corporation (http://www.agencourt.com/) for the Drosophila ananassae, Drosophila erecta, Drosophila mojavensis, Drosophila virilis sequence data, the Genome Sequencing Center, WUSTL School of Medicine (http://genome.wustl.edu/) for the Drosophila yakuba sequence data, and the Human Genome Sequencing Center at the Baylor College of Medicine (http://www.hgsc.bcm.tmc.edu/) for the Drosophila pseudoobscura sequence data. We are indebted to Colin Dewey, Nicolas Bray, and Lior Pachter for providing us with the seven-way multiple alignment. We thank Jim Kent and Angie Hinrichs for help with the UCSC Genome Browser database. We also thank Thadeous Kacmarczyk for excellent administration of our computers and Nicholas Socci for help with graphing the clustering results. We thank S. Cohen for discussions. DG acknowledges a German Academic Exchange Service (DAAD). This research was supported in part by the Howard Hughes Medical Institute grant through the Undergraduate Biological Sciences Education Program to New York University.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. DG and NR conceived and designed the experiments. DG and NR performed the experiments. DG, DL, and NR analyzed the data. KCG analyzed and discussed miR-210 targets. YLW and NR contributed reagents/materials/analysis tools. DG, KCG, and NR wrote the paper.
Abbreviation
GOGene Ontology
==== Refs
References
Ambros V 2004 The functions of animal microRNAs Nature 431 350 355 15372042
Bartel DP 2004 MicroRNAs: Genomics, biogenesis, mechanism, and function Cell 116 281 297 14744438
Lim LP Lau NC Garrett-Engele P Grimson A Schelter JM 2005 Microarray analysis shows that some microRNAs downregulate large numbers of target mRNAs Nature 433 769 773 15685193
Stark A Brennecke J Russell RB Cohen SM 2003 Identification of Drosophila microRNA targets PLoS Biol 1 e13. DOI: 10.1371/journal.pbio.0000060 12975658
Enright AJ John B Gaul U Tuschl T Sander C 2003 MicroRNA targets in Drosophila
Genome Biol 5 R1 14709173
John B Enright AJ Aravin A Tuschl T Sander C 2004 Human microRNA targets PLoS Biol 2 e363. DOI: 10.1371/journal.pbio.0020363 15502875
Lewis BP Shih IH Jones-Rhoades MW Bartel DP Burge CB 2003 Prediction of mammalian microRNA targets Cell 115 787 798 14697198
Lewis BP Burge CB Bartel DP 2005 Conserved seed pairing, often flanked by adenosines, indicates that thousands of human genes are microRNA targets Cell 120 15 20 15652477
Rajewsky N Socci ND 2004 Computational identification of microRNA targets Dev Biol 267 529 535 15013811
Rehmsmeier M Steffen P Hochsmann M Giegerich R 2004 Fast and effective prediction of microRNA/target duplexes RNA 10 1507 1517 15383676
Kiriakidou M Nelson PT Kouranov A Fitziev P Bouyioukos C 2004 A combined computational-experimental approach predicts human microRNA targets Genes Dev 18 1165 1178 15131085
Robins H Li Y Padgett RW 2005 Incorporating structure to predict microRNA targets Proc Natl Acad Sci U S A 102 4006 4009 15738385
Krek A Grün D Poy MN Wolf R Rosenberg L 2005 Combinatorial microRNA target predictions Nat Genet 37 495 500 15806104
Brennecke J Stark A Russell RB Cohen SM 2005 Principles of microRNA-target recognition PLoS Biol 3 e85. DOI: 10.1371/journal.pbio.0030085 15723116
Xie X Kulbokas EJ Golub TR Mootha V Lindblad-Toh K 2005 Systematic discovery of regulatory motifs in human promoters and 3′ UTRs by comparison of several mammals Nature 434 338 345 15735639
Hobert O 2004 Common logic of transcription factor and microRNA action Trends Biochem Sci 29 462 468 15337119
Doench JG Sharp PA 2004 Specificity of microRNA target selection in translational repression Genes Dev 18 504 511 15014042
Sempere LF Sokol NS Dubrovsky EB Berger EM Ambros V 2003 Temporal regulation of microRNA expression in Drosophila
melanogaster mediated by hormonal signals and broad-Complex gene activity Dev Biol 259 9 18 12812784
Baskerville S Bartel DP 2005 Microarray profiling of microRNAs reveals frequent coexpression with neighboring miRNAs and host genes RNA 11 241 247 15701730
The Gene Ontology Consortium 2000 Gene Ontology: Tool for the unification of biology Nature Genet 25 25 29 10802651
Bray N Pachter L 2004 MAVID: Constrained ancestral alignment of multiple sequences Genome Res 14 693 699 15060012
Benson G 1999 Tandem repeats finder: A program to analyze DNA sequences Nucleic Acids Res 27 573 580 9862982
Rajewsky N Vergassola M Gaul U Siggia ED 2002 Computational detection of genomic cis -regulatory modules applied to body patterning in the early Drosophila embryo BMC Bioinformatics 3 30 12398796
Brennecke J Hipfner DR Stark A Russell RB Cohen SM 2003
bantam encodes a developmentally regulated microRNA that controls cell proliferation and regulates the pro-apoptotic gene hid in Drosophila
Cell 113 25 36 12679032
Lai EC Posakony JW 1997 The Bearded box, a novel 3′ UTR sequence motif, mediates negative posttranscriptional regulation of Bearded and Enhancer of split complex gene expression Development 124 4847 4856 9428421
Castilllo-Davis CI Hartl DL 2003 GeneMerge—Post-genomic analysis, data mining, and hypothesis testing Bioinformatics 19 891 892 12724301
Lai EC 2003 Micro RNAs are complementary to 3′ UTR sequence motifs that mediate negative post-transcriptional regulation Nat Genet 30 363 364
Mignone F Gissi C Liuni S Pesole G 2002 Untranslated regions of mRNAs Genome Biol 3 REVIEWS0004 11897027
Ma JB Yuan YR Meister G Pei Y Tuschl T 2005 Structural basis for 5′-end-specific recognition of guide RNA by the A. fulgidus Piwi protein Nature 434 666 670 15800629
Karolchik D, Baertsch R, Diekhans M, Furey TS, Hinrichs, et al 2003 The UCSC Genome Browser Database Nucleic Acids Res 31 51 54 12519945
Blanchette M Kent WJ Riemer C Elnitski L Smit AFA 2004 Aligning multiple genomic sequences with the threaded blockset aligner Genome Res 14 708 715 15060014
Griffiths-Jones S 2004 The microRNA Registry Nucleic Acids Res 32 D109 D111 14681370
Wheeler DL, Barret T, Benson DA, Bryant SH, Canese K, et al 2005 Database resources of the National Center for Biotechnology Information Nucleic Acids Res 33 D39 D45 15608222
Pruitt KD Tatusova T Maglott DR 2005). NCBI Reference Sequence (RefSeq): A curated non-redundant sequence database of genomes, transcripts, and proteins Nucleic Acids Res 33 D501 D504 15608248
Herrero J Al-Shahrour F Díaz-Uriarte R Mateos Á Vaquerizas JM (2003 GEPAS, a web-based resource for microarray gene expression data analysis Nucleic Acids Res 31 3461 3467 12824345
|
16103902
|
PMC1183519
|
CC BY
|
2021-01-05 09:18:22
|
no
|
PLoS Comput Biol. 2005 Jun 24; 1(1):e13
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010013
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 10.1371/journal.pgen.001000105-PLGE-RA-0002R1plge-01-01-05Research ArticleBioinformatics - Computational BiologyMicrobiologyGenetics/GenomicsGenetics/Gene DiscoveryYeast and FungiA Human-Curated Annotation of the Candida albicans Genome Annotation of the Candida albicans Genome Braun Burkhard R 1van het Hoog Marco 2d'Enfert Christophe 3Martchenko Mikhail 2Dungan Jan 4Kuo Alan 4Inglis Diane O 1Uhl M. Andrew 1Hogues Hervé 2Berriman Matthew 5Lorenz Michael 6Levitin Anastasia 2Oberholzer Ursula 2Bachewich Catherine 2Harcus Doreen 2Marcil Anne 2Dignard Daniel 2Iouk Tatiana 2Zito Rosa 2Frangeul Lionel 7Tekaia Fredj 8Rutherford Kim 5Wang Edwin 2Munro Carol A 9Bates Steve 9Gow Neil A 9Hoyer Lois L 10Köhler Gerwald 4Morschhäuser Joachim 11Newport George 4Znaidi Sadri 12Raymond Martine 12Turcotte Bernard 13Sherlock Gavin 14Costanzo Maria 14Ihmels Jan 15Berman Judith 16Sanglard Dominique 17Agabian Nina 4Mitchell Aaron P 18Johnson Alexander D 1Whiteway Malcolm 2Nantel André 2*1 Department of Microbiology and Immunology, University of California, San Francisco, California, United States of America
2 Biotechnology Research Institute, National Research Council Canada, Montreal, Quebec, Canada
3 Unité Postulante Biologie et Pathogénicité Fongiques, INRA USC 2019, Institut Pasteur, Paris, France
4 Department of Stomatology, University of California, San Francisco, California, United States of America
5 The Sanger Centre, Cambridge, United Kingdom
6 Department of Microbiology and Molecular Genetics, Utah-Houston Medical School, Houston, Texas, United States of America
7 Plate-Forme Intégration et Analyse Génomique, Institut Pasteur, Paris, France
8 Unité de Génétique Moléculaire des Levures, Institut Pasteur, Paris, France
9 School of Medical Sciences, University of Aberdeen, Institute of Medical Sciences, Foresterhill, Aberdeen, United Kingdom
10 Department of Veterinary Pathobiology, University of Illinois at Urbana-Champaign, Urbana, Illinois, United States of America
11 Institut für Molekulare Infektionsbiologie, Universität Wurzburg, Wurzburg, Germany
12 Institut de Recherches Cliniques de Montreal, Montreal, Quebec, Canada
13 Department of Medicine, Royal Victoria Hospital, McGill University, Montreal, Quebec, Canada
14 Department of Genetics, Stanford University School of Medicine, Palo Alto, California, United States of America
15 Department of Molecular Genetics, Weizmann Institute of Science, Rehovot, Israel
16 Department of Genetics, Cell Biology, and Development, University of Minnesota, Minneapolis, Minnesota, United States of America
17 Institute of Microbiology, University Hospital Lausanne, Lausanne, Switzerland
18 Department of Microbiology and Institute of Cancer Research, Columbia University, New York, New York, United States of America
Snyder Michael EditorYale University, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 17 6 2005 1 1 e110 1 2005 14 3 2005 Copyright: © 2005 Braun et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Recent sequencing and assembly of the genome for the fungal pathogen Candida albicans used simple automated procedures for the identification of putative genes. We have reviewed the entire assembly, both by hand and with additional bioinformatic resources, to accurately map and describe 6,354 genes and to identify 246 genes whose original database entries contained sequencing errors (or possibly mutations) that affect their reading frame. Comparison with other fungal genomes permitted the identification of numerous fungus-specific genes that might be targeted for antifungal therapy. We also observed that, compared to other fungi, the protein-coding sequences in the C. albicans genome are especially rich in short sequence repeats. Finally, our improved annotation permitted a detailed analysis of several multigene families, and comparative genomic studies showed that C. albicans has a far greater catabolic range, encoding respiratory Complex 1, several novel oxidoreductases and ketone body degrading enzymes, malonyl-CoA and enoyl-CoA carriers, several novel amino acid degrading enzymes, a variety of secreted catabolic lipases and proteases, and numerous transporters to assimilate the resulting nutrients. The results of these efforts will ensure that the Candida research community has uniform and comprehensive genomic information for medical research as well as for future diagnostic and therapeutic applications.
Synopsis
Candida albicans is a commonly encountered fungal pathogen usually responsible for superficial infections (thrush and vaginitis). However, an estimated 30% of severe fungal infections, most due to Candida, result in death. Those who are most at risk include individuals taking immune-suppressive drugs following organ transplantation, people with HIV infection, premature infants, and cancer patients undergoing chemotherapy. Current therapies for this pathogen are made more difficult by the significant secondary effects of anti-fungal drugs that target proteins that are also found in the human host.
Recent sequencing and assembly of the genome for the fungal pathogen C. albicans used simple automated procedures for the identification of putative genes. Here, we report a detailed annotation of the 6,354 genes that are present in the genome sequence of this organism, essentially writing the dictionary of the C. albicans genome.
Comparison with other fungal genomes permitted the identification of numerous fungus-specific genes that are absent from the human genome and whose products might be targeted for antifungal therapy. The results of these efforts will thus ensure that the Candida research community has uniform and comprehensive genomic information for medical research, for the development of functional genomic tools as well as for future diagnostic and therapeutic applications.
Citation:Braun BR, van het Hoog M, d'Enfert C, Martchenko M, Dungan J, et al. (2005) A human-curated annotation of the Candida albicans genome. PLoS Genet 1(1): e1.
==== Body
Introduction
Candida albicans is a commonly encountered fungal pathogen responsible for infections generally classed as either superficial (thrush and vaginitis) or systemic (such as life-threatening blood-borne candidiasis) [1,2]. Its life cycle has fascinating aspects that have generated great excitement over the last decade, with an influx of workers and new molecular techniques brought to bear on long-standing problems [3]. Topics of particular interest are the organism's capacity to shift into several different phenotypic states, some with distinct roles in infection, and its recently discovered capacity to mate, providing at least part of a sexual cycle, although population genetic studies indicate that it is still largely a clonal diploid population. Other special adaptations for infection include a battery of externally displayed proteins and secreted digestive enzymes; complex interactions with the host immune system normally keep C. albicans at bay as a minor part of the mucosal flora [1,4,5].
Here, we report a detailed annotation of the genome sequence of this organism, bringing the previously available raw sequence to a new level of stability and usability. The genome of C. albicans has previously been shotgun sequenced to a level of 10.9-fold coverage [6]. However the assembly of this sequence faced special difficulties because the organism is diploid but with little or no gene exchange in the wild. Thus homologous chromosomes show substantial divergence, and many genes are present as two distinctive alleles. This required that the assembly process be aware of the diploid status and be prepared to segregate reads into two alleles for any section of the genome. At the same time, the genome is rich in recently diverged gene families that are easily confused with alleles. This task was further complicated by the absence of a complete physical map of the C. albicans genome. Nevertheless, this arduous assembly process resulted in a dataset (assembly 19, with 266 primary contigs over eight chromosomes) that has already yielded a number of significant advances including the production of DNA microarrays [7], libraries of systematic gene knockouts [8], large-scale transposon mutagenesis [9], and the ability of many individual researchers to identify novel genes using bioinformatic tools [10]. Unfortunately, due to the mostly computational methods used in its development, the current genome assembly still contains a significant number of predicted genes that are fragmented, overlapping, or otherwise erroneous. As a consequence, different groups have been using different methods for the identification and classification of C. albicans genes, which has hindered communication and complicated comparisons between large-scale datasets.
Following the publication of these early functional genomics studies, it was realized that the needs of the C. albicans research community would be better served by a unified gene nomenclature. The results of this community-based effort were initially based on the version 19 computational assembly and preliminary annotation produced independently by various research groups. We used visual inspection of 11,615 putative coding sequences and various bioinformatic tools to refine the quality and description of each open reading frame (ORF).
In all, we provide unique identifiers, coordinates, names, and descriptions for 6,354 genes. With the exception of certain large gene families, we have not annotated the portion of the assembly 19 DNA that was set aside as secondary alleles, instead concentrating on the primary sequence that forms one haploid genome equivalent. Investigation of the identity and relative divergence of all alleles will be an important further project for the C. albicans genome, as will finishing and linking the small number of gaps that remain in the primary sequence. In addition, we describe a variety of gene families and we discuss insights into virulence. Finally, we use comparative genomics to point out a variety of additional insights that are illuminated by the high-quality annotation provided here. This project serves as a model for community-based annotation that could be applied by other research communities that wish to improve on automated sequencing pipeline output that may be available for their organisms of interest.
Results/Discussion
The Annotation Process
Compilation of Candida annotation data.
As detailed in Materials and Methods, we used assembly version 19 of the C. albicans genome [6] to identify 11,615 putative ORFs. These included genes encoding proteins greater than 150 aa as well as genes encoding smaller proteins of 50–149 aa that have a coding function greater than 0.5 as determined with a GeneMark matrix [11]. These ORFs were then compared to the set of 7,680 C. albicans ORFs defined by the Stanford Genome Technology Center (SGTC), thus permitting their classification using the same systematic identifiers of the format orf19.n [6]. The 3,936 novel ORFs without an orf19.n counterpart were assigned a new reference number of the format orf19.n.i where orf19.n is the five-prime closest (contig-wise) ORF defined by the SGTC and i is an integer that varies between one and the number of novel ORFs found in the orf19.n to orf19.(n + 1) interval. To simplify correlation with previously published data that use the orf6.n or earlier nomenclatures, we have produced a Web-accessible translation tool (http://candida.bri.nrc.ca).
Positional information for each ORF was merged with data from a variety of different sources, including the SGTC (http://www-sequence.stanford.edu/group/candida/index.html), CandidaDB (http://genolist.pasteur.fr/CandidaDB; [12]), the Agabian laboratory (http://agabian.ucsf.edu/canoDB/anno.php), and the Johnson/Fink laboratories [13], whose annotation data had been updated with a Magpie annotation [7]. This large dataset was then reformatted into EMBL-style files, thus allowing for input in the Artemis annotation software [14]. Volunteer annotators accessed a custom-made database to reserve and download EMBL files containing sequence and annotation data for each of the 266 DNA sequence contigs. To help in validating various and sometimes conflicting sources of information, translated protein sequences from putative C. albicans ORFs were compared to putative protein sequences extracted from five fungal genomes—Saccharomyces cerevisiae [15,16], Schizosaccharomyces pombe [17], Neurospora crassa [18], Aspergillus nidulans (Aspergillus nidulans Database, http://www-genome.wi.mit.edu/annotation/fungi/aspergillus/), and Magnaporthe grisea (Magnaporthe grisea Database, http://www-genome.wi.mit.edu/annotation/fungi/magnaporthe/)—as well as to the genomes of five other eukaryotes—Arabidopsis thaliana [19], Drosophila melanogaster [20], Caenorhadbitis elegans [21], Mus musculus [22], and Homo sapiens [23]—and the GenBank non-redundant (NR) protein database. Comparisons against the translated C. albicans genome were also performed to help identify overlapping genes and putative gene families.
To help interpret such a large number of sequence comparisons, we organized sequence similarity data in a Web-accessible database using a novel visualization concept whereby we used a colorimetric display to indicate BLAST similarity, which was easy and rapid to scan visually (Figure 1). The annotators could thus rapidly determine which genes are potentially unique to C. albicans (e.g., orf19.4741 and orf19.4786), those that are members of gene families (e.g., orf19.4736 and orf19.4779), genes that only have homologs in fungal genomes (e.g., orf19.4756 and orf19.4778), or those with homologs in all eukaryotic genomes (e.g., orf19.4732 and orf19.4784). Finally, a strong hit against the complete NR database, but not in the other genomes (orf19.4772 and orf19.4800), allowed us to identify C. albicans genes that had already been described and submitted to the sequence databases prior to the publication of assembly orf19. Clicking on the relevant boxes opened an additional window containing the precompiled sequence alignments, thus permitting the validation of interesting observations. These visualization tools and the results of sequence comparisons are available at http://candida.bri.nrc.ca/candida/index.cfm?page=blast.
Figure 1 Visualization of Protein Sequence Similarities
Sample from a Web page used by annotators of the C. albicans genome to visualize the significance of the best hit from whole-proteome BLASTP searches. Each putative ORF was compared to the NR database, the Candida ORF list itself (Ca19; showing results from the four top hits), and amino acid sequences from the proteomes of S. cerevisiae (Sac), S. pombe (S.p), M. grisea (Mag), N. crassa (Neu), H. sapiens (H.S), M. musculus (M.m), D. melanogaster (Dro), C. elegans (C.e), and A. thaliana (A.t). The BLASTP e-value from the top hit was converted to a color scale as indicated. Examples of C. albicans genes with interesting similarity patterns are indicated.
The coordinates and annotations for all 11,615 putative ORFs were thus verified, corrected, and (if necessary) rewritten by the annotators. We removed ORFs smaller than 300 bp with no significant sequence similarity to other genes, either within the C. albicans genome or in the sequence databases. In cases where two ORFs overlapped by more than 50%, the smallest gene was removed unless it showed even a slight sequence similarity to another gene in the sequence databases. In other cases, we encountered two, or more, contiguous ORFs that obviously were part of the same gene. These interruptions were usually due to unidentified introns or presumed sequencing errors. In these cases, we decided to merge the relevant gene fragments into a single entry. A total of 5,262 ORF entries were thus removed from the database, or merged with neighboring ORFs, leaving 6,354 confirmed genes. Sequence and/or annotation data can be obtained in Dataset S1 or at http://candida.bri.nrc.ca.
A nomenclature for C. albicans genes.
Following consultations with the C. albicans research community during the fifth and sixth American Society for Microbiology Conferences on Candida and Candidiasis, it was agreed that C. albicans gene names should follow the format established for S. cerevisiae [24]. Gene names consist of three letters (the gene symbol) followed by an integer (e.g., ADE12); the gene symbol should be an acronym for, or relate to, the gene function, gene product, or mutant phenotype. It is preferable that a given gene symbol have only one meaning, so that all genes using that symbol are related in some way, for instance, by sharing a function, participating in a shared pathway, or belonging to the same gene family. In addition, gene symbols that are used in S. cerevisiae gene names should retain the same meaning when used for C. albicans genes. The prefix ‘Ca' has sometimes been used on gene names to denote that a gene is derived from C. albicans; however, while the use of prefixes adds clarity to discussions of genes from different species that share a name (e.g., comparing CaURA3 to ScURA3), the prefix is not considered part of the gene name proper. Finally, allele designations and deletion symbols should come after the gene name (ICG1–8 and icg1Δ for example). For more details on genetic nomenclature, see the Candida Genome Database (CGD; [25]) Web page on this topic (http://www.candidagenome. org/Nomenclature.html).
Wherever possible, genes that are orthologous between C. albicans and S. cerevisiae should share the same name. We have provided 3,409 suggested names (in the SuggGene field of the EMBL files) for many C. albicans ORFs based on their orthology to S. cerevisiae genes; these are not yet considered the standard C. albicans gene names, but rather provide guidance for investigators wishing to name these genes. CGD assigns standard names to C. albicans genes for which there are published data (the PubGene field). The annotation contains 355 such entries. Generally, CGD considers the first published name in the correct format to be the standard name; common usage and uniqueness are also considered. All names that have been used for a gene are collected in CGD, regardless of their format, so that information from the literature can be traced to the correct gene. In the current annotation, additional published gene names have been placed in the Synonym field.
Public access to the data.
The complete annotation dataset, results of BLAST sequence similarity searches, and the identification of conserved protein domains can be obtained from our Web page (http://candida.bri.nrc.ca/). Furthermore, CGD (http://www.candidagenome.org), funded by the National Institute for Dental and Craniofacial Research of the National Institutes of Health, will curate the scientific literature and provide tools for accessing and analyzing the C. albicans genome sequence. In addition, CGD will act as a central repository for gene names and modifications, as approved by the C. albicans research community at the American Society for Microbiology Candida and Candidiasis meeting in Austin, Texas, in March 2004. CGD itself will not name C. albicans genes, but instead will act as a clearinghouse for the standard gene names and aliases, as the Saccharomyces Genome Database (SGD) does for the S. cerevisiae community. CGD hopes that researchers will follow CGD's gene nomenclature guidelines (see above) and keep CGD informed of any new gene names. Prior to publication, researchers may reserve a gene name, which will then become the standard name upon publication. Finally, the CandidaDB database (http://genolist.pasteur.fr/CandidaDB) [12], which has provided an annotation of the C. albicans genome sequence since January 2001, will be updated to take into account the complete annotation dataset and will continue to provide tools for accessing and analyzing the C. albicans genome sequence complementary to those available at the CGD and the Biotechnology Research Institute.
Content and General Statistics
As detailed in Tables 1 and 2, we identified 6,354 genes in version 19 of the C. albicans genome assembly. This number is certain to change slightly with time as more data come to light. For instance, 80 of these genes are probably duplicates, having almost identical counterparts near the extremities of sequence contigs. Novel genes may also lie in unsequenced/unassembled gaps between the DNA sequence contigs. We identified 246 genes containing mutations or sequencing errors that result in a frameshift, or the insertion of a stop codon, that will have to be confirmed through resequencing. In the meantime, these elements have been joined as a single ORF entry and tagged with the entry “sequencing error?” inside their Note field. We have also identified 190 genes truncated at the ends of contigs, only 35 of which have an identical counterpart on a potentially overlapping contig. New information will be continuously integrated into the community data as it is submitted.
Table 1 Features of Completed Fungal Genomes
aNumber of base pairs in genome divided by number of genes.
bNumber and proportion of proteins with no significant similarity to known proteins.
nd, not determined.
Table 2 Statistics of the C. albicans Annotation
aExcluding “unknown.”
The mean protein coding length of 1,439 bp (480 aa) is almost identical to what has been observed in S. cerevisiae and S. pombe, while the gene density stands at one gene per 2,342 bp. Short descriptions for all gene products were provided by annotators, usually based on sequence similarity. A total of 1,218 (19.2%) genes encode unique proteins with no significant homologs in the sequence databases, a percentage almost identical to that observed in the current version of the S. cerevisiae annotation [16]. An additional 819 (12.9%) gene products exhibited significant similarities to other proteins of unknown function. Furthermore, we have provided Enzyme Commission (EC) numbers and Gene Ontology (GO) terms for 1,334 and 3,586 gene products, respectively.
Intron analysis.
There are 215 ORFs containing at least one intron, four of which have two introns, one gene (encoding the Hxt4p transporter) has three, and the SIN3 gene has four. A total of 43 (20.2%) of these genes encode ribosomal proteins, 63 (29.6%) encode products with enzymatic activity, and 26 (12.2%) encode trans-membrane proteins involved in small molecule transport. We measured the relative position of introns in their host ORFs and observed that a significant proportion of them are located in the 5′ end of ORFs, with 32% of introns being located within the first 10% of the coding sequences. A survey of the distribution of introns in 18 eukaryotic genomes, including S. cerevisiae and H. sapiens, also indicated a similar bias in intron-poor genomes. It has been argued that this 5′ bias is an indication that introns are particularly difficult to remove by cDNA recombination, because of the high activity of these genes and paucity of full-length cDNA, and that this finding lends some support to the idea that introns are being lost more frequently than they are being gained in these lineages [26], although a more recent study of four fungal genomes suggests the presence of additional mechanisms [27].
We surveyed the intron phase distribution and found that C. albicans has 50.5%, 20.4%, and 29.1% of phase zero, one, and two introns, respectively. A similar result was observed in fungal, plant, and animal genomes [27,28], suggesting that a similar intron phase distribution may be present in ancient introns and that the intron loss has no preference selection on intron phases. Seventy out of 215 intron-containing ORFs have reciprocal best matches with S. cerevisiae genes that also contain introns. Among these 70 ORFs, 25 introns (35.7%) share the same position and the same phase. This suggests that these commonly positioned introns descended from a common ancestor, as suggested previously [29].
Analysis of protein domains.
Table 3 shows the most abundant protein domains that were identified in the C. albicans proteome. As a comparison, we also performed this analysis on the same ten eukaryotic proteomes that were used in the BLASTP sequence comparisons. Compared to the S. pombe and S. cerevisiae proteomes, the C. albicans proteome shows a slight increase in the abundance of leucine-rich repeats (IPR001611), some zinc finger transcription factors (IPR001138), esterases/lipases (IPR000379), and trans-membrane transporters for polyamines (IPR002293) and for amino acids (IPR004841). If the analysis is expanded to the other fungal proteomes, only the increased abundance in leucine-rich repeats appears to be unique to C. albicans.
Table 3 Number, Abundance Ranking, and Proportion of Gene Products Containing the Indicated Interpro Protein Domain in C. albicans and Other Eukaryotes
Numbers represent how many gene products have the given domain. Ordered ranking of each domain is given in parentheses. Percentages represent the proportion of gene products that contain at least one of the domains.
DOI: 10.1371/journal.pgen.0010001.t003
Genome-Based Identification of Antifungal Targets
One of the main arguments supporting large-scale sequencing projects for fungal pathogens is the hope of finding novel antifungal targets, particularly those that are absent from the genome of their host. Table 4 shows a list of 228 C. albicans genes that have a very strong sequence homolog (based on a top hit BLASTP expect value (e-value) < 1e−45) in all five fungal genomes but no significant sequence similarity (best BLASTP e-value > 1e−10) to genes in the genomes of either humans or mice. For example, this list includes FKS1, which encodes a 1,3-beta-glucan synthase that is the target for the cell wall agents called echinocandins [30]. The list includes 46 gene products that are assumed to be located on the plasma membrane, 71 that are predicted to be involved in the transport of small molecules, and 21 that appear to be involved, directly or indirectly, with cell wall synthesis. Furthermore, 41 gene products have been associated with an EC number, indicating an enzymatic activity, with phospholipases being the most abundant. The roles and sites of action of these gene products suggest that they would be both accessible and theoretically amenable to inhibition by small molecules.
Table 4 Genes from C. albicans with a Strong Homolog in the S. cerevisiae, S. pombe, A. niger, M.
grisea, and N. crassa genomes but Absent from the H. sapiens and M. musculus Genomes
Table 4 Continued
Table 4 Continued
Table 4 Continued
Short Tandem Repeats
Short tandem repeats (STRs), also called short sequence repeats or microsatellite DNA, play an important role in evolution and have been used to characterize population variability. Although they can arise through DNA polymerase slippage and unequal recombination, whole-genome analysis has suggested that additional mechanisms for the control of STR production/correction remain to be identified [31–33]. Jones et al. [6] scanned the C. albicans genome for STRs of unit sizes between two and five and identified 1,940 trinucleotide repeats in their ORF sequences. To confirm that this high STR frequency is indeed a hallmark of the C. albicans genome, we used a statistical approach to measure repeat frequencies in four completed fungal genomes with an emphasis on STRs that affect protein sequences. We used randomized genome sequences to calculate the probability that each potential STR (including mutations that may arise following the amplification event) is nonrandom, and used only those with greater than 95% probability.
As can be seen in Datasets S2–S5 and Table 5, the STR frequencies in C. albicans and N. crassa are significantly greater than the frequencies observed in S. cerevisiae and S. pombe. Repeats that occur inside coding sequences are further characterized in Table 5. As would be expected, repeats with a modulo of three are more common in coding sequences, although we note that species with the greatest STR frequency have the smallest proportion of repeats that would break a reading frame. While coding sequence STRs in C. albicans and the other fungi most commonly encode for repeats of glutamine, asparagine, glutamic acid, and aspartic acid, we note that some of the repeats that are prevalent in C. albicans genes are distinct. Repeats of the ACT (threonine) and TCA (serine) codons are known to be especially rare in most taxa [31,33]. Correlating STR distribution with Gene Ontology annotations shows that a significant proportion of the C. albicans genes whose products are classified as DNA-binding proteins or cytoskeletal elements also contain STRs. Several gene products have been shown to play a role in the generation/correction of novel STRs in eukaryotes [34]. A comparison of the aa sequences of Rad51p, Rad52p, Mre1p, Hpr5p, and Pob3p from C. albicans, S. cerevisiae, S. pombe, and N. crassa did not reveal any significant correlation that could be associated with changes in the STR distribution. The high proportion of STRs in C. albicans genes argues that this organism would make a better model than S. cerevisiae for studying the creation and elongation of these elements that cause a variety of neuromuscular pathologies in humans. Our observations further indicate that future studies on STR frequency in eukaryotic genomes should include a broader spectrum of fungal genomes. The S. cerevisiae genome has been used as the fungal representative in comparative studies published to date [31–33].
Table 5 Frequency and Characteristics of Short Tandem Repeats in the Coding Sequences of Fungal Genomes
aSTRs with a less than 5% chance of being random
Table 5 Continued
Identification of Spurious Genes
Some of the 6,354 predicted ORFs are likely to be spurious. We used data from S. cerevisiae to model an approach that combines gene length, gene homology, and gene expression data to search for spurious gene candidates. Theoretically, genes with no sequence similarity and with expression profiles that do not correlate with other known genes are much more likely to be spurious. In an earlier study, spurious genes in S. cerevisiae were identified by sequence comparison between four closely related yeast species [16]. Most did not have orthologs with other eukaryotes, were of short length, and had expression profiles that were not significantly correlated with those of other genes in the genome (Figure 2A and 2B). Combining both the criteria of sequence homology and expression correlation produced a list of S. cerevisiae candidate genes that was highly enriched for ORFs that were considered to be spurious based on the separate sequence comparison between the closely related species. We repeated this homology/expression/length analysis on genes of the C. albicans genome. C. albicans genes with an ortholog in other eukaryotes are assumed to be real and were excluded as candidates (510 of 513 S. cerevisiae genes ruled spurious by the reading frame conservation test [16] had no ortholog in C. albicans). In the above analysis, approximately 1,000 gene expression experiments were analyzed for S. cerevisiae [35], while approximately 200 currently available experiments were analyzed for C. albicans (see Materials and Methods). Table S1 includes a ranked list of the 349 C. albicans genes that are the most likely to be spurious.
Figure 2 Identification of Spurious Genes
Assessing criteria that identify candidate spurious genes in S. cerevisiae, using a reference set of known spurious genes [16].
(A) For every gene in S. cerevisiae, the average Pearson correlation coefficient with all other genes was calculated. Shown are histograms of the correlations associated with genes characterized as spurious in the reading frame conservation test ([16]; red) and all genes in the genome (black).
(B) The distribution of gene lengths is shown for genes characterized as spurious (red) and for all genes of the genome (black).
(C) Assessing the likelihood of being spurious as a function of gene length and correlation score. Shown is the proportion of spurious genes out of all genes whose length and correlation score fall into each of the intervals. The proportion is color-coded according to the color bar shown. S. cerevisiae genes with an ortholog in C. albicans were excluded from the analysis.
Multigene Families
Many putative and demonstrated virulence factors of C. albicans are members of large multigene families. Well-known examples of such families encode secreted aspartyl proteinases [36,37]), agglutinins [38], secreted lipases [39], high-affinity iron transporters [40], and ferric reductases [41]. Members of each of these families are differentially expressed as a function of the yeast–hyphae transition, phenotypic switching, or timing during experimental infection. Also, each of these families is large relative to the corresponding homolog or family of homologs in S. cerevisiae, leading to the concept that expansion of many C. albicans gene families may be an adaptation to a commensal lifestyle and may be, in part, responsible for C. albicans's unusual ability to occupy a variety of host niches.
The sequencing of the genome provides an opportunity to survey the global occurrence and extent of multigene families as a first step in assessing their contribution to colonization and disease. We devised a purely computational method to define a comprehensive list of multigene families using NCBI-BLAST and custom Perl scripts. Each translated ORF in the annotated ORF set was compared to every other ORF in the set; if an ORF pair's BLAST alignment had an expectation value less than 1e−30 and a length greater than 60% of the length of the longer of the two ORFs, then the two ORFs were considered to be members of the same family. A transitive closure rule was applied to ensure that each ORF had membership in one and only one family. In all, 23% of the ORFs were members of families, a percentage comparable to that seen in other eukaryotes [18]. The approach yielded 451 families, with an average of 3.27 members each; 13 of the families have ten or more members, while the largest family has 39 members, consisting of proteins with possible leucine-rich repeat domains.
A striking difference between C. albicans and S. cerevisiae is the manner in which they acquire nutrients from the environment. In addition to the well-described secreted aspartyl proteinases, lipases, and high-affinity iron transporters, C. albicans possesses expanded families of acid sphingomyelinases (with four genes per haploid genome), phospholipases B (six genes), oligopeptide transporters (seven genes), and amino acid permeases (23–24 genes). Another striking difference is the emphasis by C. albicans on respiratory catabolism, as reflected in expanded families of peroxisomal enzymes. These include families of acyl-CoA oxidases (three genes), 3-ketoacyl-CoA thiolases (four genes), acyl-CoA thioesterases (three or four genes), fatty acid–CoA synthases (five genes), and glutathione peroxidases (four genes).
Additional families that may pertain to colonization or pathogenesis include those encoding the estrogen-binding protein OYE1 (seven genes), the fluconazole-resistance transporter FLU1 (13 genes), and the vacuolar protein PEP3/VPS16 (four genes), whose Aspergillus homolog is required for nuclear migration and polarized growth.
The ATP-binding cassette transporter superfamily.
The ATP-binding cassette (ABC) protein superfamily represents one of the largest protein families known to date among available genome sequences. These proteins share similar molecular architecture with the presence of at least one conserved ABC domain and the presence of membrane-spanning segments (transmembrane segments [TMSs]). The ABC domain typically contains Walker A and Walker B motifs and an ABC signature motif. The ABC domain and TMSs can be arranged in a duplicated forward (TMS6-ABC)2 or reverse (ABC-TMS6)2 topology, however “half size” ABC proteins also exist. As indicated in Table 6, the C. albicans genome contains at least 27 genes with ABC domains that include these topologies. These genes have been categorized, according to a classification established in S. cerevisiae, into six subfamilies (the MDR, PDR, MRP/CFTR, ALD, YEF3, and RLI subfamilies) [42]. The MDR, PDR, MRP/CFTR, and ALD subfamilies likely all encode transporter proteins, while the other subfamilies, YEF3 and RLI, generally lack TMSs and are considered as non-transporter ABC proteins. The C. albicans ABC proteins fall neatly into the categories developed for S. cerevisiae, and they are also present in approximately the same numbers (with the exception of the MRP/CFTR subfamily; see below). The predicted topology of each protein detailed in Table 6 is also largely comparable between the two yeast species. Among the 27 ABC proteins so far identified in C. albicans, the functions of only nine have been previously characterized. The largest group of known ABC transporters belongs to the CDR gene family, among which are CDR1 and CDR2, two genes upregulated in azole-resistant clinical isolates that function in multidrug resistance [43–45]. CDR3 and CDR4 have been shown to function as phospholipid flippases and their expression is controlled by the white–opaque switching system [46,47]. Four MRP/CFTR-like transporters are present in C. albicans, and among them three show the NH2-terminal extension with additional transmembrane segments that is typical for many MRP-like transporters (see Table 6). For unknown reasons, homologs of additional members of this family, such as the S. cerevisiae genes ScYBT1, ScNFT1, and ScVMR1, are lacking in C. albicans [42,48]. Interestingly, the vacuolar MRP-like transporter encoded by MLT1 has been implicated in virulence [49]. Since MRP/CFTR transporters are often involved in detoxification of heavy metals or xenobiotics, the presence or absence of discontinuous alleles of some ABC transporter genes (e.g., orf19.6383) may indicate strain differences in ABC transporter function and resulting susceptibility to environmental stresses. Most of the ABC transporter genes listed in Table 6 were given names through their closest homologs in S. cerevisiae; however, the functional assignments of these genes awaits further investigation.
Table 6 Genes Encoding Members of the ABC Transporter Family
aSubfamily nomenclature as proposed by Bauer et al. [42].
bPublished names are underlined.
The ALS family.
The ALS genes encode large cell-surface glycoproteins that function in host–pathogen interactions [50,51]. The ALS genes are composed of three domains: a 5′ domain that is approximately 1,300 bp in length and relatively conserved in sequence across the family, a central domain composed entirely of tandemly repeated copies of a 108-bp sequence, and a 3′ domain of variable length and sequence that encodes a serine/threonine-rich portion of the protein [50]. Efforts to characterize the ALS genes started independently of the C. albicans genome project [38,52–54]) and were aided greatly by information that emerged as the genome sequencing effort progressed [55–57]. Table 7 lists the current ORFs that correspond to genes in the ALS family. The ALS family includes eight different genes [55], each with an extensive degree of allelic variability, sometimes within a given strain (Table 7) or across the wider population of C. albicans isolates [58–60]. Because of sequence assembly difficulties, mainly attributable to the length and repetitive nature of sequences within the ALS central domain, only three of ALS ORFs in this project are in agreement with ALS gene sequences derived independently of the genome project and reported in the literature (Table 7). The annotation effort described here did not edit the underlying assembly 19 sequence. However, gap sequencing that is presently being carried out and the production of a final genome assembly will correct these errors. Published ALS gene sequences can be found on the CGD Web site.
Table 7 Assembly 19 ORFs That Correspond to ALS Genes
Assembly of the C. albicans genome sequence revealed the contiguous positions of ALS5,
ALS1, and ALS9 on Chromosome 6, which was verified by independent studies [57]. Additional testing revealed that, in SC5314, the large alleles of ALS5,
ALS1, and ALS9 occupy the same chromosome while the small alleles of each gene are found on the homologous chromosome [57]. But allelic variability and arrangement on homologous chromosomes will vary for each C. albicans strain. Allelic variation can be extreme for ALS genes, and is most commonly associated with the tandem repeat domain, although it is also present within other domains of the coding region [56,57,59]. Presenting the sequence of a single ALS allele, as done in these annotation data, loses the sense of allelic diversity that can have a significant effect on evaluation of ALS protein function. For example, testing the two ALS3 alleles from strain SC5314 in a common adhesion assay format showed that the allele with more tandem repeat copies produced a protein with greater adhesive capability than the smaller allele [60]. Table 7 notes GenBank entries for ALS alleles from strain SC5314 that aid understanding of allelic diversity for the various ALS genes.
The MEP family.
Members of the MEP gene family encode ammonium permeases and, along with the OPT family described below, feature prominently in our list of fungal-specific genes. They thus represent potentially interesting targets for the development of antifungal drugs. Experimental evidence suggests that MEP1 and MEP2 encode the only specific ammonium permeases in C. albicans, since Δmep1 Δmep2 double mutants exhibited no detectable ammonium uptake and were unable to grow at ammonium concentrations below 5 mM [61], a phenotype that is similar to that of S. cerevisiae mutants deleted for all three ammonium permeases [62,63]. The third C. albicans gene, represented by orf19.4446, encodes a protein with much lower similarity to the other ammonium permeases of C. albicans and S. cerevisiae (approximately 44% to all proteins) but might encode an ammonium permease that is not expressed under the growth conditions used in these assays.
In addition to its role in ammonium transport, Mep2p also controls nitrogen-starvation-induced filamentous growth of C. albicans. Mutants in which only the MEP2 gene was deleted grew as well as the wild-type strain at low ammonium concentrations but failed to filament under these conditions. This role of MEP2 in filamentous growth of C. albicans at low ammonium concentrations is similar to the function of its counterpart ScMEP2 in pseudohyphal growth of S. cerevisiae under limiting ammonium conditions [63]. However, in contrast to the latter, MEP2 seems to have a much broader role in filamentous growth of C. albicans since Δmep2 mutants also had a filamentation defect when amino acids or urea instead of ammonium served as the limiting nitrogen source (J. Morschhäuser, personal communication).
The OPT family.
Oligopeptide transporters represent another group of fungal-specific surface proteins that transport peptides of four or five amino acids in length into the cell and together with the di- and tripetide transporters allow growth when peptides are the only available nitrogen source. This is presumably the position of C. albicans cells when they have invaded host tissues and are secreting their battery of peptidases and other catabolic enzymes. The founding member of the oligopeptide transporter gene family was OPT1 from C. albicans [64]. Analysis of the C. albicans genome sequence as well as cloning of the corresponding genes demonstrated that C. albicans in fact possesses a large gene family encoding putative oligopeptide transporters. The OPT genes were annotated according to their decreasing similarity to OPT1. The OPT2,
OPT3, and OPT4 genes are highly similar to each other. The similarity of the remaining members of the family then drops considerably, but we have detected genes now named OPT6,
OPT7, and OPT8. Deletion of the OPT1 alleles in the C. albicans wild-type strain SC5314 resulted in increased resistance of the mutants to a toxic tetrapeptide, providing experimental evidence that Opt1p indeed functions as an oligopeptide transporter in C. albicans [65]. Preliminary observations indicate that at least the OPT2 to OPT5 genes also encode functional oligopeptide transporters (O. Reuß and J. Morschhäuser, unpublished data).
Zinc cluster transcription factors.
Proteins of the zinc finger superfamily represent one of the largest classes of DNA-binding proteins in eukaryotes. Several different classes of zinc finger domains exist that differ in the arrangement of their zinc-binding residues [66]. One of these domains, which appears to be restricted to fungi, consists of the Zn(II)2Cys6 binuclear cluster motif in which six cysteines coordinate two zinc atoms [67,68]. S. cerevisiae possesses 54 zinc cluster factors defined by the presence of the zinc cluster signature motif CX2CX6CX5–16CX2CX6–8C, which is generally located at the N-terminus of the protein. These proteins function as transcriptional regulators involved in various cellular processes including primary and secondary metabolism (e.g., Gal4p, Ppr1p, Hap1p, Cha4p, Leu3p, Lys14p, and Cat8p), pleiotropic drug resistance (e.g., Pdr1p, Pdr3p, and Yrr1p), and meiosis (Ume6p) [68,69]. Quite often, they bind as homo- or heterodimers to two CGG triplets organized as direct, indirect, or inverted repeats and separated by sequences of variable length [68,70]. A large proportion of these factors (50%) also contain a middle homology region (Fungal_trans in the Pfam Protein Families Database) located in the central portion of the protein that has been proposed to participate in DNA binding and to assist in DNA target discrimination [67].
Analysis of the C. albicans proteome using a combination of sequence analyses tools (SMART, Pfam, and PHI-BLAST) allowed us to identify 77 binuclear cluster proteins. These factors are characterized by the presence of the zinc cluster signature motif CX2CX6CX5–24CX2CX6–9C generally located at the N-terminus of the protein (72 out of 77) and with a spacing between cysteines 3–4 and 5–6 slightly different from the S. cerevisiae motif. As observed in S. cerevisiae, a large proportion of the C. albicans factors also contain a middle homology region (29 out of 77). To our knowledge, only six of the C. albicans zinc cluster genes have been characterized in detail, including SUC1, involved in sucrose utilization [71], FCR1, implicated in pleiotropic drug resistance [72], CWT1, required for cell wall integrity [73], and CZF1,
FGR17, and FGR27, involved in filamentous growth [9,74]. The functions of many uncharacterized C. albicans zinc cluster factors (approximately 20%) can be inferred from the fact that they display high levels of sequence similarity (top BLASTP e-value ≤ 1e−20) with the products of S. cerevisiae genes with a known function. In the case of GAL4, however, the C. albicans homologous ORF identified (orf19.5338) encodes a significantly smaller protein (261 aa) than S. cerevisiae Gal4p (881 aa), lacking the C-terminal two-thirds of the protein that contains one of two transcriptional activating domains, and must therefore have a somewhat different function. Approximately half of the C. albicans zinc cluster genes do not appear to have homologs in S. cerevisiae (using a BLAST cutoff of < 1e−20) and are therefore likely to participate in processes specific to C. albicans. Finally, it is noteworthy that many of the zinc cluster factors known to be involved in pleiotropic drug resistance in S. cerevisiae, such as Pdr1p, Pdr3p, Yrr1p, Yrm1p, Rds1p, and Rdr1p, do not appear to possess close structural homologs in C. albicans. Since pleiotropic drug resistance is frequently observed in C. albicans, it is likely that this organism possesses functional homologs of these genes or other novel processes that remain to be identified.
Lipid and Amino Acid Metabolism
Some of the C. albicans ORFs that do not have clear homologs in S. cerevisiae but do have homologs in other fungi, bacteria, and/or vertebrates encode catabolic enzymes, oxidoreductases, and proteins involved in environmental sensing pathways. The list of genes that C. albicans does not share with S. cerevisiae is skewed towards enzymes involved in the catabolism of fatty acids and ketone bodies in the peroxisome. There are also numerous oxidoreductases, some of which may be involved in activating hydrophobic organic compounds as a prelude to their oxidative degradation. This metabolic arrangement may reflect, in part, the state of the common ancestor with S. cerevisiae, as also reflected in Yarrowia lipolytica, C. antartica, C. rugosa, C. tropicalis, C. maltosa, and C. deformans, which are model organisms in the study of lipases and alkane oxidation for industrial purposes. It is worth mentioning, however, that the genus Candida arose originally to identify fungi that were unclassifiable, asexual, and ascomycetous—properties that appear to correlate with parasitism and the presence of catabolic gene families, such as lipases and alkane-assimilating cytochrome P-450 enzymes. Beta-oxidation in fungi is predominantly peroxisomal, and the number of enzymes participating in the process is greater in C. albicans than in S. cerevisiae. C. albicans also encodes a related ethanolamine kinase (orf19.6912), a malonyl-CoA acyl carrier protein acyltransferase (MCT1), and an enoyl-CoA hydratase (orf19.6830) not found in S. cerevisiae. Further supplying substrates for oxidation are several enzymes encoded by C. albicans that participate in the degradation of asparagine (asparaginase; orf19.3791), cysteine (cysteine dioxygenase [CDG1] and cysteine sulfinate decarboxylase [orf19.5393]), valine (3-hydroxyisobutyrate dehydrogenase [orf19.5565]), and arginine (orf19.3498). Other catabolic enzymes come as a surprise in that they may relate to the scavenging of unsuspected carbon sources. C. albicans encodes three D-amino acid oxidases (IFG3, DAO1, and DAO2) whose substrates might be derived from bacterial cell walls, various oxidoreductases whose substrates are likely to be aromatic and aliphatic compounds not used by the host, a pathway consistent with omega oxidation of fatty acids (which would convert alkanes into alpha-omega diols, fatty acids, and dicarboxylic acids), and a benzene desulfurase (orf19.3901).
Acetyl-CoA generated in the peroxisome is transferred to the mitochondrion, where the most notable difference from S. cerevisiae is the presence of a respiratory Complex I, which can now largely be reconstructed based on sequence similarity to components found in other organisms. The importance of Complex I in the biology of C. albicans is inferred from the observation that deletion of one of its subunits results in a defect in filamentation [75] and the observation that subunit 49 is essential for vegetative growth [8]. An additional difference is the presence of two alternative oxidases that may be involved in protection against oxidative stress [76]. Thus, it is not yet clear whether the omnivorous catabolic capacity of C. albicans reflects its heritage and role as a fungal saprophyte aiding organic decomposition, or whether these capacities have been elaborated and tuned in response to the specific problem of consuming mammalian host cells.
Phospholipases.
Depending on the site of attack, phospholipases are classified as phospholipase A, B, C, or D. Phospholipase A enzymes hydrolyze the 1-acyl ester (PLA1) or the 2-acyl ester (PLA2) of phospholipids. In fungi, phospholipase B enzymes hydrolyze both acyl groups and often also have lysophospholipase activity, removing the remaining acyl moiety on lysophospholipids [77]. Phospholipase C and phospholipase D enzymes are phosphodiesterases that cleave the glycerophosphate bond and remove the base group of phospholipids, respectively. While a major role of phospholipase function is membrane homeostasis, additional functions comprise nutrient digestion and generation of signaling molecules. Some phospholipases are toxins or components of venoms. Bacterial phospholipases have been shown to be involved in pathogenesis by promoting hemolysis, cytolysis, and tissue destruction, as well as interfering with host signal transduction [78].
As indicated in Table 8, the largest and best-characterized group of phospholipases in C. albicans is the five-member phospholipase B gene family. A related gene family is present in S. cerevisiae, albeit with three members, again reflecting the general increase in gene numbers for enzymes involved in lipid metabolism in C. albicans. All PLB proteins harbor NH2-terminal signal peptides for secretion; Plb3p, Plb4p, and Plb5p additionally contain hydrophobic COOH termini with putative GPI anchor attachment sites for localization to the plasma membrane or further processing for tethering to the cell wall [79,80]. To date, PLB1 and PLB2 are the best-characterized members of the gene family [81–84]. Inactivation of PLB1 [82,83] and PLB5 (S. Theiss, G. Ishdorj, M. Kretschman, C. Y. Lan, T. Nichterlein, et al., unpublished data) reduced virulence in animal models.
Table 8 Phospholipases in C. albicans
aPublished names are underlined.
Putative PLC and PLD phosphodiesterases are also represented in the C. albicans genome. Orf19.6629 is a likely homolog to S. cerevisiae
ScISC1, which encodes a PLC with neutral sphingomyelinase activity. Besides the recently published PLC1 gene [85], two almost identical genes encode phosphatidylinositol phospholipase C proteins (PI-PLC). The latter lack homologs in S. cerevisiae, but are similar to bacterial PI-PLCs. PLD1 was shown to be involved in the morphological transition from yeast to hyphae and required for full virulence in animal models [86]. Interestingly, the PLD1 gene product and another phospholipase-like protein (encoded by orf19.4151) show significant sequence similarity to S. cerevisiae proteins that are involved in meiosis and sporulation (ScSpo1p, ScSpo14p, and ScSpo22p). As already shown for PLD1, the ScSPO1 homolog, the functional roles of these proteins are likely to differ from their counterparts in S. cerevisiae since C. albicans has not been shown to undergo meiosis [10].
Another intriguing group of phospholipase genes in C. albicans are patatin-like phospholipases encoded by orf19.1504, orf19.5426, and orf19.6396. These proteins might account for phospholipase A activities in C. albicans that could be involved in intracellular storage or mobilization of lipids.
Sphingolipid metabolism.
C. albicans also displays differences from S. cerevisiae with respect to sphingolipid metabolism. Pathways leading to and from fungal-type sphingomyelins have been studied extensively in S. cerevisiae, where by-products mediate many important structural and signaling functions that affect cell proliferation, the definition of cell membrane domains and polarity, apoptosis, and stress responses [87,88]. Many of the associated enzymes are essential and are targets of fungal toxins, and thus are candidates for anti-fungal drug development [88]. C. albicans shares the same fundamental pathways in sphingolipid biosynthesis/degradation plus four additional enzymes. Two of these, a glucosyl transferase (CGT1) and a delta-4 sphingolipid desaturase (DES1), have been previously studied. The presence of glycosyl ceramides in C. albicans has been known for some time [89,90], and the gene responsible for their synthesis has been cloned and expressed in Pichia [91]. The molecules play a common role in differentiation in dimorphic fungi [92]. Homologs of the delta-4 sphingolipid desaturase enzyme include the mouse, human, and Drosophila degenerative spermatocyte proteins, which play a role in meiosis [93]; its function in C. albicans may relate to membrane structure, or the production of signaling molecules, as is the case in plants. An interesting component of sphingolipid metabolism in C. albicans is a sphingomyelin transfer protein (Het1p) similar to the Podospora anserina HET-C2 protein. The P. anserina protein is involved in self/non-self discrimination, a fungal version of the vertebrate major histocompatibility locus [94,95]. It is possible that the protein is involved in regulating the sphingomyelin composition of C. albicans membranes, a factor that may relate to acquisition of resistance to amphotericin B and azoles [96]. Finally, C. albicans encodes four acid sphingomyelinases, two of which may be secreted, that have not been studied in fungi. Based on the actions of metazoan secreted acid sphingomyelinases, these enzymes may be involved in regulation of membrane raft formation and generation of ceramide, a second messenger that is known to regulate apoptosis in higher eukaryotes. Secreted sphingomyelinases of pathogenic bacteria, which are enzymatically similar but structurally unrelated to those of C. albicans, have been shown to lyse phagosomal membranes [97], facilitate entry into both phagocytic and nonphagocytic cells [98,99], act as hemolysins that abet piracy of iron from the host [100,101], and induce host cell apoptosis [102,103].
Signal Transduction
Differences in signal transduction and regulatory pathways between C. albicans and S. cerevisiae are numerous. Many of these C. albicans–specific genes encode proteins that are responsive to changes in the environment. They may thus be responsive to colonization of a new anatomical site (e.g., passage through the stomach), fluctuations in the availability of nutrients, or the appearance of host inflammatory reactions. Gene products falling into this category include (1) a homolog (TIP120) of a TBP-interacting protein in humans and rats, which acts as global regulator of class I, II, and III genes in response to abrupt changes in ambient conditions [104], (2) a relative (orf 19.1798) of tuberin, a negative regulator of cell growth in response to low cellular energy levels in mammals [105], (3) a conserved group of stomatin-like proteins (orf 19.7296 and SLP2) that may play a role in mechanoreception, (4) a family of pirin homologs that obviously arose from a recent duplication event (PRN1, PRN2, PRN3, and PRN4)—these are nuclear factors whose homologs interact with the human oncogene Bcl-3 product and with an A. thaliana G protein alpha-subunit involved in regulating seed germination and early seedling development [106])—and (5) a rhomboid protein (orf 19.5234), probably located on the plasma membrane, whose homologs in eukaryotes and bacteria mediate the proteolytic release of signaling peptides from a larger precursor [107]. In addition to differences traceable to novel genes, other pathways that share components have doubtless been altered in their role and regulation, such as the mating pathway [108].
Two of the most important enzyme families that are involved in signal transduction pathways are the kinases and small GTPases. The C. albicans annotation identifies 96 protein kinases, most of which have strong orthologs in S. cerevisiae. The C. albicans genome contains two genes encoding GTPases of the heterotrimeric G protein alpha-subunit family—GPA1 and GPA2. In addition, it contains 29 small GTPases of the p21 superfamily. These include a single Ras protein (Ras1p), various members of the Rho and Rab families, the Ran1 homolog Gsp1p, and several members of the ADP ribosylation subfamily. Most of these proteins have clear S. cerevisiae orthologs. However, S. cerevisiae does not have a Rac homolog, while orf19.6237 appears to encode a C. albicans Rac protein and has thus been named RAC1. As well, orf19.5902 appears to be distantly related to Ras but lacks any strong equivalent in any organism, and has been designated Rlp1p, for Ras-like protein, while orf19.2975 is a YPT/RAB family member that has been named RAB7 because it has no clear S. cerevisiae YPT ortholog.
Conclusions
We have coordinated a community-wide effort to manually confirm, edit, and annotate 6,354 genes from assembly 19 of the C. albicans genome. This annotation includes 214 intron-containing genes, 246 genes with either missense mutations or sequencing errors, and 190 truncated genes that terminate at the ends of the sequence contigs. C. albicans genes were found to be exceptionally rich in short sequence repeats, especially compared to the genomes of S. pombe and S. cerevisiae. Correlation with transcriptional profiling data was used to identify potentially spurious genes. This improved dataset allowed the identification fungal-specific genes and permitted a detailed analysis of several large multigene families. Comparative genomic studies indicate that C. albicans is much more versatile in its production of secreted lipid- and amino-acid-degrading enzymes and in its ability to import the resulting nutrients.
Materials and Methods
Identification of C. albicans ORFs and merging of preliminary annotations.
Nucleotide sequence data for assembly 19 were retrieved from the SGTC Web site (http://www-sequence.stanford.edu//group/candida/). Assembly 19 is composed of a haploid supercontig set (contigs 19–831 to 19–10262), here referred to as the haploid set, and a allelic supercontig set (contigs 19–20001 to 19–20161), here referred to as the allelic set [6].
The CAAT-Box software package [109] was used to identify annotation-relevant ORFs in assembly orf19. A set including ORFs longer than 300 codons and a set with all intergenic regions obtained after subtraction of ORFs larger than 80 codons were created. These sets were used to build a GeneMark matrix [11] that was subsequently used to evaluate the coding probability of all ORFs in assembly 19. ORFs longer than 150 codons, and ORFs longer than 40 codons and with a GeneMark coding function greater than 0.5 [11] over their whole length, were selected and assigned a reference number of the format IPFn.i where IPF stands for individual protein file, n is an integer specific to the IPF, and i corresponds to the number of times the IPF has been modified between assembly 5, 6, and 19 of the C. albicans genome sequence. In total, 11,025 and 9,089 IPFs were selected in the haploid and allelic sets, respectively. IPFs shorter than 150 codons in the haploid set were further inspected for (1) overlaps with larger IPFs on a different frame and (2) homology to proteins in the NR database of non-redundant proteins from GenBank. IPFs that overlapped with a larger IPF or did not show a significant homolog (BLASTP e-value < 1e−3) [110] were designated FALSORF. Of the 11,025 IPFs identified in the haploid set, 3,505 were FALSORFs.
All IPFs identified in the haploid set were compared through reciprocal BLASTP to the set of 7,680 C. albicans ORFs defined at the SGTC that uses the systematic designation orf19.n. BLASTP results were parsed using Readblast [111]. IPFs without an orf19.n counterpart were assigned a new reference number of the format orf19.n.i, where orf19.n is the closest upstream (using SGTC contig coordinates) ORF defined by the SGTC and i is an integer that varies between 1 and the number of IPFs located between orf19.n and orf19.(n + 1). For instance, if three ORFs were found between orf19.1234 and orf19.1235, these would be referred to as orf19.1234.1, orf19.1234.2, and orf19.1234.3. Taken together, 11,616 orf19 ORFs were identified in the haploid set, of which 3,936 were not present in the SGTC orf19 set.
A similar procedure was applied to the allelic set of sequences, and a total of 9,552 orf19 ORFs were identified, of which 3,012 were not present in the SGTC orf19 set. The haploid and allelic sets of orf19 ORFs were compared by reciprocal BLASTP in order to define allelic and unique sequences in the allelic set (see below).
The 11,615 orf19 ORFs identified in the haploid set were compared by reciprocal BLASTP to the 9,168 ORFs identified by the SGTC using assembly 6 of the C. albicans genome sequence (designated orf6.n). A similar reciprocal comparison was run using the set of 6,165 C. albicans proteins available in the CandidaDB database that have been defined by applying a procedure similar to that outlined above on assembly 6 and through a manual curation aiming to reach a non-redundant protein set (http://genolist.pasteur.fr/CandidaDB; [12]). Furthermore, orf19 ORFs were reciprocally compared to the S. cerevisiae proteome using data available at the SGD [112]. All data were parsed using Readblast [111], and a matrix was generated that correlated ORFs from each dataset.
We used the genome annotation tool Artemis [14], which provides very detailed annotation capability, visually mapping desired features onto the target sequence. In preparation, we loaded our heterogeneous data into the required EMBL-style files: (1) the orf19 reference (field: Assembly_ID); (2) the orf6 reference (field: old_Assembly_ID); (3) the CandidaDB entry number (field: db_xref); (4) the entry number for the Comprehensive Yeast Genome Database, which provides a detailed analysis of protein features of all entries available in CandidaDB [113] (field: db_xref); (5) the GenBank entry number for C. albicans proteins previously characterized and annotated (field: db_xref); (6) the IPF reference (field: db_xref); (7) annotation data available from CandidaDB (proposed gene name and proposed function; field: Annotator); (8) annotation data of the Agabian's laboratory based on the orf6 protein set (http://agabian.ucsf.edu/canoDB/anno.php) (field: Annotator); (9) annotation data of the Fink's and Johnson's laboratories based on the orf6 protein set (unpublished) (field: note_AJ); (10) annotation comments available from CandidaDB (field: note_GF); (11) Pfam matches [114] obtained using the orf19 protein set (field: pfam_match); (12) Clusters of Orthologous Groups matches [115] obtained from the Comprehensive Yeast Genome Database using the CandidaDB protein set (field: COGs_MIPS); (13) EC number matches obtained from the Comprehensive Yeast Genome Database using the CandidaDB protein set (field: EC_number_MIPS); (14) S. cerevisiae closest protein including e-value, putative orthology, protein function, gene name, and alternate gene names, genome reference number obtained from SGD (field: Note); (15) GO annotation of the S. cerevisiae protein obtained from SGD (field: GO); (16) the reference of the allelic orf19 in the allelic set of sequences including supercontig number and location (field: Allele); and (17) chromosome assignment data available from the Magee's and Whiteway's laboratories (http://206.167.190.233/candida/index.cfm?page=CaChrom) (field: Chromosome).
Furthermore all ORFs were assigned a color code in order to facilitate annotation using the annotation tool Artemis [14]. All orf19 ORFs corresponding to an IPF classified as FALSORF and orf19 ORFs identified at the SGTC and not found among IPFs were color-coded in grey. orf19 ORFs with an unambiguous allele (90% identical amino acids over the whole length of the longest ORF) were color-coded in red (>150 codons) or pink (<150 codons). orf19 ORFs with a questionable allele (90% identical amino acids over the whole length of the shortest ORF) were color-coded in green (>150 codons) or pale green (<150 codons). orf19 ORFs without a clear allele (less than 90% identical amino acids over the whole length of the shortest ORF or no reciprocal match) were color-coded in blue (>150 codons) or light blue (<150 codons).
From this “master” file of sequences and their corresponding preliminary annotation, groups of contigs were selected and saved as partially annotated subsequences that were reserved and retrieved by members of the annotation consortium, and once fully annotated, were returned to a central Web site. A version of Artemis was distributed to the consortium that included a modified “options” file [116] allowing project-specific qualifiers to be used, and also featuring the C. albicans–specific translation Table 12 [117].
Whole genome BLAST searches and visualization of sequence homologies.
ORF sequences were translated to proteins using the translation table for C. albicans [117], and compared using the BLASTP algorithm [118] with the NR database, the C. albicans proteome itself, the putative proteomes of five fungi (S. cerevisiae, S. pombe, N. crassa,
A. nidulans, and M. grisea), and the proteomes of five other eukaryotes (A. thaliana, D. melanogaster, C. elegans, M. musculus, and H. sapiens). Sequence data were obtained from the EMBL-EBI Integr8 Browser (http://www.ebi.ac.uk/integr8/EBI-Integr8-HomePage. do) and the Broad Institute for Genome Research (http://www.broad. mit.edu/annotation/). For the most similar proteins by BLAST, negative exponents of the e-values were parsed from the output files, collected in a relational database, and visualized as a color range associated with each reference protein (see Figure 1). For this purpose, a Web-accessible visualization tool was constructed using Macromedia Cold Fusion and an Apache2 Web server. These results can be consulted at For this purpose, a Web-accessible visualization tool was constructed using Maromedia Cold Fusion and an Apache2 Web server. These results can be consulted at http://candida.bri.nrc.ca/candida/index.cfm?page =blast.
Bioinformatic identification of putative introns.
Intron locations were predicted by constructing regular expressions based on consensus data drawn from several known introns in C. albicans (SOD1, EFB1, CMD1, CSK1, and others) as well as the extensive knowledge of the splice site consensus in S. cerevisiae [119] and matching of those expressions to the genomic DNA. Two regular expressions were used: /(.{15,}?TG[AT]A[CT]G)/ and /(.{700}) ([ATC]TACTAAC.{4,24}?[ATC][CTA]AG)(.{90})/. These incorporate several conserved aspects of mRNA splice sites including (1) the splice donor site of G(T/C)A(T/A)GT, where the initial guanine is the first residue within the intron, (2) a gap of between 15 and 700 nucleotides between the donor site and the branch point, (3) the branch point consensus sequence (A/T/C)TACTAAC, (4) a gap of between four and 24 nucleotides between the branch point and the splice acceptor site, (5) the splice acceptor site (T/A/C)(T/A/C)AG, where the final guanine is the final residue within the intron. This procedure provided landmarks that annotators could use to decide whether a gene contained introns, a decision also based on the position of nearby ORFs, on the ability of the intron/exon assembly to extend the size of the reading frame, and, most important, on the ability of the putative intron/exon assembly to improve similarity to sequences in other species. It predicted 1,297 introns in the C. albicans genome, many of which were not near ORFs or were otherwise incorrect. Nevertheless, this procedure was useful in providing markers for confirmation by the annotators.
Identification and classification of STRs.
A STR was defined as a short nucleotide element (1–20 nt) repeated a number of times with a periodicity P, a length L, and a tolerated mutation window W. Since allelic indels are rare in C. albicans, we considered only mutations in STR sequences and not insertion or deletion from the consensus periodic pattern. W indicates the minimun distance between two mutations within a STR. For example, the STR “CTACAACAACAGCAAC” has P = 3, L = 16, and W ≤ 10. When two STRs overlap they are merged together defining a single STR domain. Since short STRs can randomly occur, depending on the G/C content of the genome, we established a significant threshold length L
MIN for every periodicity P and mutation window W to represent the STR length for which the number of STR domains found in a randomized genome is less than 5% of the number found in the real genome. We call STR95 the set of all STR domains that are longer than the threshold value. In the STR95 set, every STR domain has a less than 5% chance of being a random event. With this method of setting the minimal STR length, no significant STR can be found in any randomized genome or in genomes that do not contain STRs in sufficient number to beat the odds 20 to one. More information and data can be found in Dataset S6.
Identification of spurious genes.
For the calculation of gene expression correlation, we used a set of approximately 1,000 S. cerevisiae microarray experiments [127], and 216 genome-wide C. albicans expression profiles [7,13,108,120–126]. The iterative signature algorithm [35,127] was applied to the C. albicans expression data as described. We analyzed ORFs present on the arrays for which an orf19 number could be determined, including some that were subsequently removed from the final set of annotated genes. Pairwise Pearson correlation coefficients were calculated for each ORF with respect to all other ORFs across all of the experiments contained in the dataset. Random subsets of the S. cerevisiae data were generated by randomly selecting 200 experiments from the complete set of approximately 1,000 profiles. ORFs whose correlation coefficient exceeded the threshold value of 3σ with at least one other ORF in the dataset were recorded and excluded from the list of spurious gene candidates. The standard deviation σ of the background correlation of random gene pairs was measured to be σ = 0.16 for S. cerevisiae and σ = 0.21 for C. albicans. Similarly, genes possessing an ortholog in S. cerevisiae were excluded from the list of C. albicans candidates, and vice versa. All remaining ORFs were subsequently ordered by their length, and the 50 shortest ORFs were excluded (many of the shortest 50 genes correspond to real genes in S. cerevisiae).
Supporting Information
Dataset S1 Coordinates and All of the Annotation Fields for the 6,354 Confirmed C. albicans Genes, Based on the Version 19 Genome Assembly
Please note that Microsoft Excel may convert some of the gene names to dates and fail to import some of the largest fields.
(2 MB TXT)
Click here for additional data file.
Dataset S2 Sequence and Position of All Statistically Significant STRs in C. albicans Coding Sequences
(291 KB TXT)
Click here for additional data file.
Dataset S3 Sequence and Position of All Statistically Significant STRs in S. pombe Coding Sequences
(11 KB TXT)
Click here for additional data file.
Dataset S4 Sequence and Position of All Statistically Significant STRs in S. cerevisiae Coding Sequences
(66 KB TXT)
Click here for additional data file.
Dataset S5 Sequence and Position of All Statistically Significant STRs in N. crassa Coding Sequences
(488 KB TXT)
Click here for additional data file.
Dataset S6 Detailed Description of Our STR Identification Algorithm
(4 KB TXT)
Click here for additional data file.
Table S1 List of Potentially Spurious Genes
(34 KB XLS)
Click here for additional data file.
Accession Numbers
The GenBank (http://www.ncbi.nlm.nih.gov/Genbank/) accession numbers for ORFs discussed in this paper are ALS3–1 (AY223552), ALS3–2 (AY223551), ALS4–2 (AF272027), ALS5–1 (AY227440), ALS5–2 (AY227439), strain SC5314 sequence corresponding to ALS6 (AY225310), strain SC5314 ALS9–1 (AY269423), and strain SC5314 ALS9–2 (AY269422).
This project, and many others that applied functional genomic techniques to the study of C. albicans, would not have been possible without the public release of genomic sequence data by Ted Jones, Stewart Scherer, and their colleagues at the SGTC. We also wish to thank Keith James for Perl-scripting, Susumu Goto for corrections to the EC numbers, and numerous members of the Candida research community for their encouragements. We thank the Magnaporthe and Aspergillus Sequencing Projects of North Carolina State University and the Broad Institute of the Massachusetts Institute of Technology and Harvard (http://www.broad.mit.edu) for access to preliminary genome sequence data. Many of the annotators were supported by National Institutes of Health grant # RO1 AI49187 (to ADJ), European Comission grants # QLK2–2000–00795 and MCRTN-CT-2003–504148 (to CD), National Institute for Dental and Craniofacial Research grant # P01 DE07946 (to NA), and Canadian Institutes of Health Research grant # MOP-42516 (to MW). Finally, we gratefully acknowledge direct support for the community effort from the Pharmacia Corporation, the Burroughs Wellcome Fund, the Wellcome Trust, and the National Research Council Canada's Genome and Health Initiative. This is National Research Council Canada publication #46227.
Competing interests. BRB is an employee of both Incyte and University of California at San Francisco (UCSF). Incyte played no role in this work, and the resources used were all from UCSF.
Author contributions. BRB, CD, JD, DOI, MB, ML, APM, ADJ, MW, and AN conceived and designed the experiments. BRB, MM, JD, DOI, MAV, AL, VO, CB, DH, AM, DD, TL, RZ, MW, and AN performed the experiments. BRB, MV, CD, AK, HH, MB, CAM, SB, NAG, LLH, GK, JM, GN, SZ, MR, BT, JHI, JB, DS, NA, MW, and AN analyzed the data. BRB, MV, CD, JD, AK, DOI, MAV, HH, MB, ML, LF, FT, KR, EW, GS, MC, NA, APM, ADJ, and AN contributed reagents/materials/analysis tools. BRB, CD, AK, MB, EW, NAG, LLH, GK, JM, GN, SZ, MR, BT, GS, JB, DS, ADJ, MW, and AN wrote the paper.
Abbreviations
ABCATP-binding cassette
CGD
Candida Genome Database
e-valueexpect value
ECEnzyme Commission
GOGene Ontology
IPFindividual protein file
NRnon-redundant
ORFopen reading frame
SGD
Saccharomyces Genome Database
SGTCStanford Genome Technology Center
STRshort tandem repeat
TMStransmembrane segment
==== Refs
References
Odds FC 1994 Pathogenesis of Candida infections J Am Acad Dermatol 31 S2 S5 8077502
Fradin C Hube B 2003 Tissue infection and site-specific gene expression in Candida
albicans
Adv Appl Microbiol 53 271 290 14696322
Berman J Sudbery PE 2002
Candida
albicans: A molecular revolution built on lessons from budding yeast Nat Rev Genet 3 918 930 12459722
Fidel PL Jr 2002 Distinct protective host defenses against oral and vaginal candidiasis Med Mycol 40 359 375 12230215
Naglik JR Challacombe SJ Hube B 2003
Candida albicans secreted aspartyl proteinases in virulence and pathogenesis Microbiol Mol Biol Rev 67 400 428 12966142
Jones T Federspiel NA Chibana H Dungan J Kalman S 2004 The diploid genome sequence of Candida
albicans
Proc Natl Acad Sci U S A 101 7329 7334 15123810
Nantel A Dignard D Bachewich C Harcus D Marcil A 2002 Transcription profiling of Candida
albicans cells undergoing the yeast to hyphal transition Mol Biol Cell 13 3452 3465 12388749
Roemer T Jiang B Davison J Ketela T Veillette K 2003 Large-scale essential gene identification in Candida
albicans and applications to antifungal drug discovery Mol Microbiol 50 167 181 14507372
Uhl MA Biery M Craig N Johnson AD 2003 Haploinsufficiency-based large-scale forward genetic analysis of filamentous growth in the diploid human fungal pathogen C. albicans
EMBO J 22 2668 2678 12773383
Tzung KW Williams RM Scherer S Federspiel N Jones T 2001 Genomic evidence for a complete sexual cycle in Candida
albicans
Proc Natl Acad Sci U S A 98 3249 3253 11248064
Lukashin AV Borodovsky M 1998 GeneMark.hmm: New solutions for gene finding Nucleic Acids Res 26 1107 1115 9461475
d'Enfert C, Goyard S, Rodriguez-Arnaveilhe S, Frangeul L, Jones L, et al. 2005 CandidaDB: A genome database for Candida
albicans pathogenomics Nucleic Acids Res 33 D353 D357 15608215
Bennett RJ Uhl MA Miller MG Johnson AD 2003 Identification and characterization of a Candida
albicans mating pheromone Mol Cell Biol 23 8189 8201 14585977
Rutherford K Parkhill J Crook J Horsnell T Rice P 2000 Artemis: Sequence visualization and annotation Bioinformatics 16 944 945 11120685
Goffeau A Barrell BG Bussey H Davis RW Dujon B 1996 Life with 6000 genes Science 274 546 563 547 8849441
Kellis M Patterson N Endrizzi M Birren B Lander ES 2003 Sequencing and comparison of yeast species to identify genes and regulatory elements Nature 423 241 254 12748633
Wood V Gwilliam R Rajandream MA Lyne M Lyne R 2002 The genome sequence of Schizosaccharomyces
pombe
Nature 415 871 880 11859360
Galagan JE Calvo SE Borkovich KA Selker EU Read ND 2003 The genome sequence of the filamentous fungus Neurospora
crassa
Nature 422 859 868 12712197
Arabidopsis Genome Initiative 2000 Analysis of the genome sequence of the flowering plant Arabidopsis thaliana
Nature 408 796 815 11130711
Adams MD Celniker SE Holt RA Evans CA Gocayne JD 2000 The genome sequence of Drosophila melanogaster
Science 287 2185 2195 10731132
C. elegans Sequencing Consortium (1998) Genome sequence of the nematode C. elegans: A platform for investigating biology Science 282 2012 2018 9851916
Waterston RH Lindblad-Toh K Birney E Rogers J Abril JF 2002 Initial sequencing and comparative analysis of the mouse genome Nature 420 520 562 12466850
Lander ES Linton LM Birren B Nusbaum C Zody MC 2001 Initial sequencing and analysis of the human genome Nature 409 860 921 11237011
Cherry JM 1995 Genetic nomenclature guide. Saccharomyces
cerevisiae
Trends Genet 1995 11 12
Arnaud MB Constanzo MC Skrzypek MS Binkley G Lane C 2005 The Candida Genome Database (CGD), a community resource for Candida albicans gene and protein information Nucleic Acid Res 33 D358 D363 15608216
Mourier T Jeffares DC 2003 Eukaryotic intron loss Science 300 1393 12775832
Nielsen CB Friedman B Birren B Burge CB Galagan JE 2004 Patterns of intron gain and loss in fungi PLoS Biol 2 e422. DOI: 10.1371/journal. pbio.0020422 15562318
Fedorov A Merican AF Gilbert W 2002 Large-scale comparison of intron positions among animal, plant, and fungal genes Proc Natl Acad Sci U S A 99 16128 16133 12444254
Kersanach R Brinkmann H Liaud MF Zhang DX Martin W 1994 Five identical intron positions in ancient duplicated genes of eubacterial origin Nature 367 387 389 8114942
Boucher HW Groll AH Chiou CC Walsh TJ 2004 Newer systemic antifungal agents: Pharmacokinetics, safety and efficacy Drugs 64 1997 2020 15341494
Toth G Gaspari Z Jurkat J 2000 Microsatellites in different eukaryotic genomes: Survey and analysis Genome Res 10 967 981 10899146
Katti MV Ranjekar PK Gupta VS 2001 Differential distribution of simple sequence repeats in eukaryotic genome sequences Mol Biol Evol 18 1161 1167 11420357
Astolfi P Bellizzi D Sgaramella V 2003 Frequency and coverage of trinucleotide repeats in eukaryotes Gene 317 117 125 14604799
Nag DK Suri M Stenson EK 2004 Both CAG repeats and inverted DNA repeats stimulate spontaneous unequal sister-chromatid exchange in Saccharomyces
cerevisiae
Nucleic Acids Res 32 5677 5684 15494455
Ihmels J Friedlander G Bergmann S Sarig O Ziv Y 2002 Revealing modular organization in the yeast transcriptional network Nat Genet 31 370 377 12134151
Magee BB Hube B Wright RJ Sullivan PJ Magee PT 1993 The genes encoding the secreted aspartyl proteinases of Candida albicans constitute a family with at least three members Infect Immun 61 3240 3243 8335356
White TC Miyasaki SH Agabian N 1993 Three distinct secreted aspartyl proteinases in Candida
albicans
J Bacteriol 175 6126 6133 8407785
Hoyer LL Payne TL Bell M Myers AM Scherer S 1998
Candida
albicans ALS3 and insights into the nature of the ALS gene family Curr Genet 33 451 459 9644209
Hube B Stehr F Bossenz M Mazur A Kretschmar M 2000 Secreted lipases of Candida
albicans: Cloning, characterisation and expression analysis of a new gene family with at least ten members Arch Microbiol 174 362 374 11131027
Ramanan N Wang Y 2000 A high-affinity iron permease essential for Candida
albicans virulence Science 288 1062 1064 10807578
Lan CY Rodarte G Murillo LA Jones T Davis RW 2004 Regulatory networks affected by iron availability in Candida
albicans
Mol Microbiol 53 1451 1469 15387822
Bauer BE Wolfger H Kuchler K 1999 Inventory and function of yeast ABC proteins: About sex, stress, pleiotropic drug and heavy metal resistance Biochim Biophys Acta 1461 217 236 10581358
Prasad R De Wergifosse P Goffeau A Balzi E 1995 Molecular cloning and characterization of a novel gene of Candida
albicans , CDR1, conferring multiple resistance to drugs and antifungals Curr Genet 27 320 329 7614555
Sanglard D Kuchler K Ischer F Pagani JL Monod M 1995 Mechanisms of resistance to azole antifungal agents in Candida
albicans isolates from AIDS patients involve specific multidrug transporters Antimicrob Agents Chemother 39 2378 2386 8585712
Sanglard D Ischer F Monod M Bille J 1997 Cloning of Candida
albicans genes conferring resistance to azole antifungal agents: Characterization of CDR2, a new multidrug ABC-transporter gene Microbiology 143 405 416 9043118
Lan CY Newport G Murillo LA Jones T Scherer S 2002 Metabolic specialization associated with phenotypic switching in Candida
albicans
Proc Natl Acad Sci U S A 99 14907 14912 12397174
Smriti Krishnamurthy S Dixit BL Gupta CM Milewski S 2002 ABC transporters Cdr1p, Cdr2p and Cdr3p of a human pathogen Candida
albicans are general phospholipid translocators Yeast 19 303 318 11870854
Mason DL Mallampalli MP Huyer G Michaelis S 2003 A region within a lumenal loop of Saccharomyces
cerevisiae Ycf1p directs proteolytic processing and substrate specificity Eukaryot Cell 2 588 598 12796304
Theiss S Kretschmar M Nichterlein T Hof H Agabian N 2002 Functional analysis of a vacuolar ABC transporter in wild-type Candida
albicans reveals its involvement in virulence Mol Microbiol 43 571 584 11929516
Hoyer LL 2001 The ALS gene family of Candida
albicans
Trends Microbiol 9 176 180 11286882
Sheppard DC Yeaman MR Welch WH Phan QT Fu Y 2004 Functional and structural diversity in the Als protein family of Candida albicans
J Biol Chem 279 30480 30489 15128742
Hoyer LL Scherer S Shatzman AR Livi GP 1995
Candida
albicans ALS1: Domains related to a Saccharomyces
cerevisiae sexual agglutinin separated by a repeating motif Mol Microbiol 15 39 54 7752895
Hoyer LL Payne TL Hecht JE 1998 Identification of Candida
albicans ALS2 and ALS4 and localization of Als proteins to the fungal cell surface J Bacteriol 180 5334 5343 9765564
Gaur NK Klotz SA 1997 Expression, cloning, and characterization of a Candida
albicans gene, ALA1, that confers adherence properties upon Saccharomyces
cerevisiae for extracellular matrix proteins Infect Immun 65 5289 5294 9393828
Hoyer LL Hecht JE 2000 The ALS6 and ALS7 genes of Candida
albicans
Yeast 16 847 855 10861907
Hoyer LL Hecht JE 2001 The ALS5 gene of Candida albicans and analysis of the Als5p N-terminal domain Yeast 18 49 60 11124701
Zhao X Pujol C Soll DR Hoyer LL 2003 Allelic variation in the contiguous loci encoding Candida albicans ALS5, ALS1 and ALS9 Microbiology 149 2947 2960 14523127
Lott TJ Holloway BP Logan DA Fundyga R Arnold J 1999 Towards understanding the evolution of the human commensal yeast Candida albicans
Microbiology 145 1137 1143 10376829
Zhang N Harrex AL Holland BR Fenton LE Cannon RD 2003 Sixty alleles of the ALS7 open reading frame in Candida albicans: ALS7 is a hypermutable contingency locus Genome Res 13 2005 2017 12952872
Oh SH Cheng G Nuessen JA Jajko R Yeater KM 2004 Functional specificity of Candida albicans Als3p proteins and clade specificity of ALS3 alleles discriminated by the number of copies of the tandem repeat sequence in the central domain Microbiology 151 673 681
Biswas K Morschhäuser J 2005 The Mep2p ammonium permease controls nitrogen starvation-induced filamentous growth in Candida albicans
Mol Microbiol 56 649 669 15819622
Marini AM Soussi-Boudekou S Vissers S Andre B 1997 A family of ammonium transporters in Saccharomyces
cerevisiae
Mol Cell Biol 17 4282 4293 9234685
Lorenz MC Heitman J 1998 The MEP2 ammonium permease regulates pseudohyphal differentiation in Saccharomyces cerevisiae
EMBO J 17 1236 1247 9482721
Lubkowitz MA Hauser L Breslav M Naider F Becker JM 1997 An oligopeptide transport gene from Candida
albicans
Microbiology 143 387 396 9043116
Reuß O Vik Å Kolter R Morschhäuser J 2004 The SAT1 flipper, an optimized tool for gene disruption in Candida
albicans
Gene 341 119 127 15474295
Matthews JM Sunde M 2002 Zinc fingers—Folds for many occasions Life 54 351 355 12665246
Schjerling P Holmberg S 1996 Comparative amino acid sequence analysis of the C6 zinc cluster family of transcriptional regulators Nucleic Acids Res 24 4599 4607 8967907
Todd RB Andrianopoulos A 1997 Evolution of a fungal regulatory gene family: The Zn(II)2Cys6 binuclear cluster DNA binding motif Fungal Genet Biol 21 388 405 9290251
Akache B Wu K Turcotte B 2001 Phenotypic analysis of genes encoding yeast zinc cluster proteins Nucleic Acids Res 29 2181 2190 11353088
Hellauer K Rochon MH Turcotte B 1996 A novel DNA binding motif for yeast zinc cluster proteins: The Leu3p and Pdr3p transcriptional activators recognize everted repeats Mol Cell Biol 16 6096 6102 8887639
Kelly R Kwon-Chung KJ 1992 A zinc finger protein from Candida
albicans is involved in sucrose utilization J Bacteriol 174 222 232 1729210
Talibi D Raymond M 1999 Isolation of a putative Candida
albicans transcriptional regulator involved in pleiotropic drug resistance by functional complementation of a pdr1 pdr3 mutation in Saccharomyces
cerevisiae
J Bacteriol 181 231 240 9864335
Moreno I Pedreno Y Maicas S Sentandreu R Herrero E 2003 Characterization of a Candida
albicans gene encoding a putative transcriptional factor required for cell wall integrity FEMS Microbiol Lett 226 159 167 13129622
Brown DH Jr Giusani AD Chen X Kumamoto CA 1999 Filamentous growth of Candida albicans in response to physical environmental cues and its regulation by the unique CZF1 gene Mol Microbiol 34 651 662 10564506
McDonough JA Bhattacherjee V Sadlon T Hostetter MK 2002 Involvement of Candida
albicans NADH dehydrogenase complex I in filamentation Fungal Genet Biol 36 117 127 12081465
Huh WK Kang SO 2001 Characterization of the gene family encoding alternative oxidase from Candida
albicans
Biochem J 356 595 604 11368790
Ghannoum MA 2000 Potential role of phospholipases in virulence and fungal pathogenesis Clin Microbiol Rev 13 122 143 10627494
Schmiel DH Miller VL 1999 Bacterial phospholipases and pathogenesis Microbes Infect 1 1103 1112 10572314
Eisenhaber B Schneider G Wildpaner M Eisenhaber F 2004 A sensitive predictor for potential GPI lipid modification sites in fungal protein sequences and its application to genome-wide studies for Aspergillus
nidulans , Candida
albicans , Neurospora
crassa , Saccharomyces
cerevisiae and Schizosaccharomyces
pombe
J Mol Biol 337 243 253 15003443
Lee SA Wormsley S Kamoun S Lee AF Joiner K 2003 An analysis of the Candida
albicans genome database for soluble secreted proteins using computer-based prediction algorithms Yeast 20 595 610 12734798
Hoover CI Jantapour MJ Newport G Agabian N Fisher SJ 1998 Cloning and regulated expression of the Candida albicans phospholipase B (PLB1) gene FEMS Microbiol Lett 167 163 169 9809417
Leidich SD Ibrahim AS Fu Y Koul A Jessup C 1998 Cloning and disruption of caPLB1, a phospholipase B gene involved in the pathogenicity of Candida
albicans
J Biol Chem 273 26078 26086 9748287
Mukherjee PK Seshan KR Leidich SD Chandra J Cole GT 2001 Reintroduction of the PLB1 gene into Candida
albicans restores virulence in vivo Microbiology 147 2585 2597 11535799
Sugiyama Y Nakashima S Mirbod F Kanoh H Kitajima Y 1999 Molecular cloning of a second phospholipase B gene, caPLB2 from Candida
albicans
Med Mycol 37 61 67 10200936
Bennett DE McCreary CE Coleman DC 1998 Genetic characterization of a phospholipase C gene from Candida albicans: Presence of homologous sequences in Candida species other than Candida albicans
Microbiology 144 55 72 9467900
Hube B Hess D Baker CA Schaller M Schafer W 2001 The role and relevance of phospholipase D1 during growth and dimorphism of Candida
albicans
Microbiology 147 879 889 11283284
Dickson RC Lester RL 2002 Sphingolipid functions in Saccharomyces
cerevisiae
Biochim Biophys Acta 1583 13 25 12069845
Obeid LM Okamoto Y Mao C 2002 Yeast sphingolipids: Metabolism and biology Biochim Biophys Acta 1585 163 171 12531550
Matsubara T Hayashi A Banno Y Morita T Nozawa Y 1987 Cerebroside of the dimorphic human pathogen, Candida
albicans
Chem Phys Lipids 43 1 12 3555875
Ghannoum MA Janini G Khamis L Radwan SS 1986 Dimorphism-associated variations in the lipid composition of Candida
albicans
J Gen Microbiol 132 2367 2375 3540201
Leipelt M Warnecke D Zahringer U Ott C Muller F 2001 Glucosylceramide synthases, a gene family responsible for the biosynthesis of glucosphingolipids in animals, plants, and fungi J Biol Chem 276 33621 33629 11443131
Barreto-Bergter E Pinto MR Rodrigues ML 2004 Structure and biological functions of fungal cerebrosides An Acad Bras Cienc 76 67 84 15048196
Ternes P Franke S Zahringer U Sperling P Heinz E 2002 Identification and characterization of a sphingolipid delta 4-desaturase family J Biol Chem 277 25512 25518 11937514
Saupe S Descamps C Turcq B Begueret J 1994 Inactivation of the Podospora
anserina vegetative incompatibility locus het-c, whose product resembles a glycolipid transfer protein, drastically impairs ascospore production Proc Natl Acad Sci U S A 91 5927 5931 8016091
Mattjus P Turcq B Pike HM Molotkovsky JG Brown RE 2003 Glycolipid intermembrane transfer is accelerated by HET-C2, a filamentous fungus gene product involved in the cell-cell incompatibility response Biochemistry 42 535 542 12525182
Mukhopadhyay K Prasad T Saini P Pucadyil TJ Chattopadhyay A 2004 Membrane sphingolipid-ergosterol interactions are important determinants of multidrug resistance in Candida
albicans
Antimicrob Agents Chemother 48 1778 1787 15105135
Gonzalez-Zorn B Dominguez-Bernal G Suarez M Ripio MT Vega Y 1999 The smcL gene of Listeria
ivanovii encodes a sphingomyelinase C that mediates bacterial escape from the phagocytic vacuole Mol Microbiol 33 510 523 10417642
Grassme H Gulbins E Brenner B Ferlinz K Sandhoff K 1997 Acidic sphingomyelinase mediates entry of N. gonorrhoeae into nonphagocytic cells Cell 91 605 615 9393854
Hauck CR Grassme H Bock J Jendrossek V Ferlinz K 2000 Acid sphingomyelinase is involved in CEACAM receptor-mediated phagocytosis of Neisseria
gonorrhoeae
FEBS Lett 478 260 266 10930579
Projan SJ Kornblum J Kreiswirth B Moghazeh SL Eisner W 1989 Nucleotide sequence: The beta-hemolysin gene of Staphylococcus
aureus
Nucleic Acids Res 17 3305 2726469
Marshall MJ Bohach GA Boehm DF 2000 Characterization of Staphylococcus
aureus beta-toxin induced leukotoxicity J Nat Toxins 9 125 138 10868340
Esen M Schreiner B Jendrossek V Lang F Fassbender K 2001 Mechanisms of Staphylococcus
aureus induced apoptosis of human endothelial cells Apoptosis 6 431 439 11595832
Tseng HJ Chan CC Chan EC 2004 Sphingomyelinase of Helicobacter
pylori -induced cytotoxicity in AGS gastric epithelial cells via activation of JNK kinase Biochem Biophys Res Commun 314 513 518 14733936
Makino Y Yogosawa S Kayukawa K Coin F Egly JM 1999 TATA-binding protein-interacting protein 120, TIP120, stimulates three classes of eukaryotic transcription via a unique mechanism Mol Cell Biol 19 7951 7960 10567521
Inoki K Zhu T Guan KL 2003 TSC2 mediates cellular energy response to control cell growth and survival Cell 115 577 590 14651849
Lapik YR Kaufman LS 2003 The Arabidopsis cupin domain protein AtPirin1 interacts with the G protein alpha-subunit GPA1 and regulates seed germination and early seedling development Plant Cell 15 1578 1590 12837948
Gallio M Sturgill G Rather P Kylsten P 2002 A conserved mechanism for extracellular signaling in eukaryotes and prokaryotes Proc Natl Acad Sci U S A 99 12208 12213 12221285
Tsong AE Miller MG Raisner RM Johnson AD 2003 Evolution of a combinatorial transcriptional circuit: A case study in yeasts Cell 115 389 399 14622594
Frangeul L Glaser P Rusniok C Buchrieser C Duchaud E 2004 CAAT-Box, contigs-assembly and annotation tool-box for genome sequencing projects Bioinformatics 20 790 797 14752000
Altschul SF Gish W Miller W Myers EW Lipman D 1990 Basic local alignment search tool J Mol Biol 215 403 410 2231712
Tekaia F Blandin G Malpertuy A Llorente B Durrens P 2000 Genomic exploration of the hemiascomycetous yeasts: 3. Methods and strategies used for sequence analysis and annotation FEBS Lett 487 17 30 11152878
Christie KR Weng S Balakrishnan R Costanzo MC Dolinski K 2004 Saccharomyces Genome Database (SGD) provides tools to identify and analyze sequences from Saccharomyces
cerevisiae and related sequences from other organisms Nucleic Acids Res 32 D311 D314 14681421
Mewes HW Amid C Arnold R Frishman D Guldener U 2004 MIPS: Analysis and annotation of proteins from whole genomes Nucleic Acids Res 32 D41 D44 14681354
Bateman A Coin L Durbin R Finn RD Hollich V 2004 The Pfam protein families database Nucleic Acids Res 32 D138 D141 14681378
Wheeler DL Church DM Edgar R Federhen S Helmberg W 2004 Database resources of the National Center for Biotechnology Information: Update Nucleic Acids Res 32 D35 D40 14681353
Berriman M Rutherford K 2003 Viewing and annotating sequence data with Artemis Brief Bioinform 4 124 132 12846394
Ohama T Suzuki T Mori M Osawa S Ueda T 1993 Non-universal decoding of the leucine codon CUG in several Candida species Nucleic Acids Res 21 4039 4045 8371978
Altschul SF 1997 Evaluating the statistical significance of multiple distinct local alignments Suhai S Theoretical and computational methods in genome research New York Plenum 1 14
Cheng SC 1994 Formation of the yeast splicing complex A1 and association of the splicing factor PRP19 with the pre-mRNA are independent of the 3′ region of the intron Nucleic Acids Res 22 1548 1554 8202353
Garcia-Sanchez S Aubert S Iraqui I Janbon G Ghigo JM 2004
Candida
albicans biofilms: A developmental state associated with specific and stable gene expression patterns Eukaryot Cell 3 536 545 15075282
Bensen ES Martin SJ Li M Berman J Davis DA 2004 Transcriptional profiling in C. albicans reveals new adaptive responses to extracellular pH and functions for Rim101p Mol Microbiol 54 1335 1351 15554973
Cowen LE Nantel A Tessier D Whiteway M Thomas DY 2002 Population genomics of drug resistance in experimental populations of Candida
albicans
Proc Natl Acad Sci U S A 99 9284 9289 12089321
Enjalbert B Nantel A Whiteway M 2003 Stress induced gene expression in Candida
albicans: Absence of a general stress response Mol Biol Cell 14 1460 1467 12686601
Lee CM Nantel A Jiang L Whiteway M Shen SH 2004 The serine/threonine protein phosphatase SIT4 modulates yeast-to-hypha morphogenesis and virulence in Candida
albicans
Mol Microbiol 51 691 709 14731272
Karababa M Coste AT Rognon B Bille J Sanglard D 2004 Comparison of gene expression profiles of Candida
albicans azole-resistant clinical isolates and laboratory strains exposed to drugs inducing multidrug transporters Antimicrob Agents Chemother 48 3064 3079 15273122
Rogers PD Barker KS 2003 Genome-wide expression profile analysis reveals coordinately regulated genes associated with stepwise acquisition of azole resistance in Candida
albicans clinical isolates Antimicrob Agents Chemother 47 1220 1227 12654650
Ihmels J Bergmann S Barkai N 2004 Defining transcription modules using large-scale gene expression data Bioinformatics 20 1993 2003 15044247
Dujon B Sherman D Fischer G Durrens P Casaregola I 2004 Genome evolution in yeast Nature 430 35 44 15229592
Loftus BJ Fung E Roncaglia P Rowley D Amedeo P 2005 The genome of the basidiomycetous yeast and human pathogen Cryptococcus Neoformans
Science 307 1321 1324 15653466
Balan I Alarco AM Raymond M 1997 The Candida albicans CDR3 gene codes for an opaque-phase ABC transporter J Bacteriol 179 7210 7218 9393682
Franz R Michel S Morschhauser J 1998 A fourth gene from the Candida
albicans CDR family of ABC transporters Gene 220 91 98 9767132
Raymond M Dignard D Alarco AM Mainville N Magee BB 1998 A Ste6p/P-glycoprotein homolog from the asexual yeast Candida
albicans transports the a-factor mating pheromone in Saccharomyces
cerevisiae
Mol Microbiol 27 587 598 9489670
Ogawa A Hashida-Okado T Endo M Yoshioka H Tsuruo T 1998 Role of ABC transporters in aureobasidin A resistance Antimicrob Agents Chemother 42 755 761 9559778
Di Domenico BJ Lupisella J Sandbaken M Chakraburtty K 1992 Isolation and sequence analysis of the gene encoding translation elongation factor 3 from Candida
albicans
Yeast 8 337 352 1626427
Sturtevant J Cihlar R Calderone R 1998 Disruption studies of a Candida
albicans gene, ELF1: A member of the ATP-binding cassette family Microbiology 144 2311 2321 9720054
Andaluz E Coque JJ Cueva R Larriba G 2001 Sequencing of a 4.3 kbp region of chromosome 2 of Candida albicans reveals the presence of homologs of SHE9 from Saccharomyces cerevisiae and of bacterial phosphatidylinositol-phospholipase C Yeast 18 711 721 11378898
Kanoh H Nakashima S Zhao Y Sugiyama Y Kitajima Y 1998 Molecular cloning of a gene encoding phospholipase D from the pathogenic and dimorphic fungus, Candida
albicans
Biochim Biophys Acta 9 359 364
|
16103911
|
PMC1183520
|
CC BY
|
2021-01-05 08:00:24
|
no
|
PLoS Genet. 2005 Jul 17; 1(1):e1
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010001
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 10.1371/journal.pgen.001000205-PLGE-RA-0004R1plge-01-01-01Research ArticleNeuroscienceGenetics/Gene DiscoveryCaenorhabditisUsing Microarrays to Facilitate Positional Cloning: Identification of Tomosyn as an Inhibitor of Neurosecretion Microarray-Assisted CloningDybbs Michael 12Ngai John 2Kaplan Joshua M 1*1 Department of Molecular Biology, Massachusetts General Hospital, Boston, Massachusetts, United States of America
2 Department of Molecular and Cell Biology, Functional Genomics Laboratory, Helen Wills Neuroscience Institute, University of California, Berkeley, California, United States of America
Barsh Gregory S EditorStanford University School of Medicine, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 25 7 2005 1 1 e213 1 2005 1 2 2005 Copyright: © 2005 Dybbs et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Forward genetic screens have been used as a powerful strategy to dissect complex biological pathways in many model systems. A significant limitation of this approach has been the time-consuming and costly process of positional cloning and molecular characterization of the mutations isolated in these screens. Here, the authors describe a strategy using microarray hybridizations to facilitate positional cloning. This method relies on the fact that premature stop codons (i.e., nonsense mutations) constitute a frequent class of mutations isolated in screens and that nonsense mutant messenger RNAs are efficiently degraded by the conserved nonsense-mediated decay pathway. They validate this strategy by identifying two previously uncharacterized mutations: (1) tom-1, a mutation found in a forward genetic screen for enhanced acetylcholine secretion in Caenorhabditis elegans, and (2) an apparently spontaneous mutation in the hif-1 transcription factor gene. They further demonstrate the broad applicability of this strategy using other known mutants in C. elegans,
Arabidopsis, and mouse. Characterization of tom-1 mutants suggests that TOM-1, the C. elegans ortholog of mammalian tomosyn, functions as an endogenous inhibitor of neurotransmitter secretion. These results also suggest that microarray hybridizations have the potential to significantly reduce the time and effort required for positional cloning.
Synopsis
Genetic screens are commonly used to figure out which genes are involved in a biological process. The first step in a genetic screen is to isolate mutant animals that are defective in the process being studied. The next step is to find which of the thousands of genes has the mutation that causes the observed defect. Positional cloning, the tried-and-true method for locating mutations, is slow and expensive. The authors propose using microarray hybridizations to speed the process. Their approach relies on the fact that a large fraction of the mutations found in screens are the results of premature stop codons, a particularly severe type of mutation. In cells, messages containing premature stop codons are rapidly destroyed by a protective pathway, called nonsense-mediated decay, thus making them directly detectable by microarray hybridization.
The authors apply this strategy retrospectively to known mutants in Caenorhabditis elegans, Arabidopsis, and mouse. They identify two uncharacterized mutations in C. elegans, including one, tom-1, found in a forward genetic screen for enhancers of neurotransmission. Interestingly, their characterization of tom-1 mutants suggests that the highly conserved protein tomosyn inhibits neurotransmission in neurons. This study shows that microarray hybridizations will help reduce the time and effort required for positional cloning.
Citation:Dybbs M, Ngai J, Kaplan JM (2005) Using microarrays to facilitate positional cloning: Identification of tomosyn as an inhibitor of neurosecretion. PLoS Genet 1(1): e2.
==== Body
Introduction
Forward genetic screens have been traditionally utilized in model systems (e.g., Caenorhabditis elegans, Drosophila, yeast, and Arabidopsis). More recently, large-scale screens have been undertaken in vertebrate systems such as zebrafish [1,2] and mouse [3–5]. Mutations isolated in genetic screens are typically identified by positional cloning. The difficulty posed by positional cloning is determined by the size of the genome, the recombination rate, and the difficulty of assessing the mutant phenotype. For example, the mouse genome comprises 3,600 centimorgans (cM) and 3 × 109 base pairs. The ultimate goal of a typical positional cloning project is to analyze a sufficient number of recombinants to map the mutation to a small genetic interval (typically approximately 0.1 cM). Once a mutation has been precisely mapped, gene identification is typically achieved by a variety of strategies: direct sequencing of the region (100 kb in the mouse), candidate gene testing, or screening for informative alleles (e.g., microdeletions). The difficulty of a particular positional cloning can be compounded by the nature of the mutant phenotype. This problem is particularly acute for behavioral mutants, which often have phenotypes that must be scored in multiple trials, or in populations of animals. Together, these issues conspire to make traditional positional cloning a significant and costly bottleneck.
To circumvent these difficulties, several new technologies have been developed to isolate mutations by reverse genetics. Reverse genetic strategies include use of insertional mutagens [6−10], PCR screens for randomly induced deletions [11], homologous gene targeting [12,13], and physical or genetic detection of point mutations in sequenced genes [14,15]. While reverse genetic strategies circumvent the positional cloning bottleneck, these approaches also have limitations. Mutations isolated by reverse genetics often lack obvious phenotypic defects (e.g., because they are in functionally redundant genes). Phenotypic differences observed in mutants isolated by reverse genetics can be confounded by other mutations in the genetic background, particularly since animals are typically heavily mutagenized in these strategies. For these reasons, it would be useful to develop methods that would allow more rapid characterization of mutations isolated in forward genetic screens.
We wondered whether microarray expression data could facilitate the identification of mutations responsible for behavioral defects isolated in forward genetic screens. It is well established that nonsense mutations result in the degradation of the mutant messenger RNA (mRNA) via the nonsense-mediated decay (NMD) pathway. A surveillance mechanism common to all eukaryotes, NMD serves as a quality control system to destroy faulty mRNAs whose translation would lead to an inappropriately truncated protein [16−18]. NMD protects cells by eliminating inactive or potentially deleterious dominant negative proteins that are the result of somatic mutation, transcriptional mistakes, or splicing errors.
It has been proposed that NMD could be used as a basis to identify nonsense mutations in cell lines [19,20]. In principle, a nonsense mutation in mutant animals could be identified using microarray hybridizations to find transcripts with decreased abundance. In practice, microarray data alone are unlikely to be sufficient to identify nonsense mutations. In addition to the expected statistical noise associated with microarray experiments, there are likely to be transcriptional changes in other genes that are caused by the mutation being studied. The most powerful cloning approach would thus be one that uses microarray data together with traditional mapping information. Here, we present evidence supporting the feasibility and general utility of this strategy.
Results
To test the feasibility of using microarrays to facilitate positional cloning, we will address four questions. (1) How frequently are nonsense alleles recovered in forward genetic screens? (2) Are microarray hybridizations sensitive enough to detect the decreased abundance of a nonsense mutant transcript? (3) Can microarray hybridizations be used to identify an uncloned behavioral mutant in C. elegans? (4) Is this microarray-based strategy applicable to other model organisms?
Nonsense Alleles Represent a Large Fraction of C. elegans Mutations
The utility of microarrays in cloning depends on the frequency with which nonsense alleles are recovered in phenotypic screens. Since 15 of the 61 amino acid–encoding codons are mutable to stop codons by a single base-pair substitution, nonsense alleles are likely to represent a large fraction of all alleles recovered after random mutagenesis with agents that increase the rate of nucleotide misincorporation. To assess the prevalence of nonsense alleles isolated following random mutagenesis, we compiled a list of sequenced C. elegans mutant alleles by downloading information from WormBase and conducting targeted literature searches (Figure 1; Table S1). We focused on misincorporation mutations because these represent the most common lesion caused by alkylating agents such as ethyl methane sulfonate or N-ethyl-N-nitrosourea. In total, we examined 943 single-nucleotide-substitution loss-of-function alleles in 246 genes, which were published by 99 laboratories. We then classified these alleles as putative NMD targets (nonsense) or non-NMD targets (missense). We found that 41% of the alleles were putative NMD targets. Interestingly, this figure is comparable to estimates that one-third of human genetic diseases are the result of nonsense mutations [21,22].
Figure 1 Fraction of C. elegans Alleles That Are Nonsense Mutations
Molecular information about alleles was obtained from WormBase and literature searches. Graph includes 117 genes for which molecular characterization of three or more alleles was available (770 alleles total). Of these 117 alleles, 22 (19%) have no known nonsense mutations. Many of these no-nonsense alleles are in genes that are required for viability.
We calculated the percentage of nonsense alleles recovered for each of the 117 genes in our dataset with three or more characterized alleles (Figure 1). For 20% of these genes, no nonsense alleles have been identified, even in cases where ten alleles are known. Many of these genes are known to cause a lethal phenotype as a null mutation (e.g., acy-1, unc-17, and let-502). Since many screens demand homozygous viable phenotypes, it is not surprising that nonsense alleles were rarely recovered in these genes. For another 20% of these genes, all known alleles are nonsense mutations. These might comprise genes for which mutant phenotypes are expressed only when gene function is completely eliminated. For all other genes, there appears to be a broad distribution in the fraction of nonsense alleles recovered, with a mode occurring at 40% nonsense alleles. Thus, while a high frequency of nonsense alleles seems to be a general feature common to many screens in C. elegans, this is likely to vary considerably for different gene classes.
Proof of Principle: mec-3 and unc-43 CaMKII Mutations Are Detectable by Microarray
Are microarray hybridizations sensitive enough to detect changes in mutant transcript abundance due to a nonsense lesion above the global variation in gene expression between mutant and control strains? Some potential sources of variance in gene expression include random fluctuations in gene expression [23,24], uncontrolled differences between the mutant and control populations (e.g., differences in developmental stage or physiological status), and differences in genetic backgrounds [25,26]. Perhaps the most important potential limitation is changes in gene expression that are a secondary consequence of a mutation. This could be particularly problematic for mutations in genes encoding transcription factors or other components of signal transduction cascades, the loss of which would be expected to alter the expression of many downstream genes.
To address some of these concerns, we examined the large collection of microarray experiments used to build a whole-genome expression profile for C. elegans [27]. Most of these experiments, which were done with printed microarrays, were designed to identify gene-expression profiles associated with various developmental programs or specific tissues. However, one set of experiments analyzed changes in gene expression in mutants lacking the MEC-3 transcription factor (see Materials and Methods) [28]. The mec-3(e1338) allele corresponds to a W69Stop mutation, and homozygous animals carrying this mutation are touch-insensitive [29,30].
Using this dataset, we classified genes as differentially expressed in mec-3(e1338) based on two criteria: average fold-change in expression level and statistical significance using a Student's t-test. We constructed a volcano plot with the log2(fold-change) on the x-axis and negative log10(p-value) on the y-axis [31]. This provides a useful way to visualize differentially expressed genes—those whose expression level is down (negative on the x-axis) and that show high statistical significance (large on the y-axis). Seventy genes were identified as having significantly reduced expression in mec-3(e1338), using fold-change greater than −1.0 (log2 scale) and p < 0.01 as thresholds for decreased expression (Figure 2). Had e1338 been an uncharacterized mutation that we were attempting to clone, the next step would be to narrow the candidate list of 70 genes using mapping data. Fifteen of these genes are on Chromosome 4, which contains mec-3 and approximately 2,900 other genes. Of these, only three differentially expressed genes fall within a two-map-unit interval spanning mec-3 and approximately 100 other genes. Thus, even in the case of a transcription factor, microarray hybridizations are sufficiently sensitive to detect changes in the mutant mRNA abundance despite broader changes in gene expression. In the case of MEC-3, it is likely that the reduced abundance of e1338 mRNA is due both to NMD and to positive autoregulation of mec-3 transcription [32]. Therefore, triangulating between rough mapping information and microarray data would have facilitated the rapid cloning of mec-3.
Figure 2 Analysis of mRNA Abundance in mec-3(e1338) Animals
Fold-change (x-axis) is plotted against the statistical significance (y-axis) for each probeset. Fold-changes are shown on log2 scale. p-Values are shown on a negative log10 scale. The symbol × indicates genes with reduced expression in mec-3(e1338) animals (fold-change < −1, p < 0.01). Light blue circles indicate genes with reduced expression that are also on Chromosome 4. Dark blue circles indicate genes with reduced expression that are within 1 cM to the left or right of mec-3. The open red circle indicates the mec-3 gene.
To further address the sensitivity of microarray-assisted cloning, we analyzed changes in gene expression observed in KP3365 unc-43(n1186) mutants. The unc-43 gene encodes type II calcium- and calmodulin-dependent protein kinase (CaMKII), which is broadly expressed in the worm nervous system as well as in muscles and in the intestine [33]. This provides another demanding test case because CaMKII plays a pivotal role in calcium-mediated signaling in neurons, and unc-43 mutations are known to cause changes in the expression of other genes [34]. The n1186 allele corresponds to a Q67Stop mutation, and homozygous animals carrying this mutation have relatively subtle behavioral defects [33].
We hybridized total RNA isolated from wild-type and KP3365 unc-43(n1186) CaMKII mutant animals to the Affymetrix C. elegans GeneChip (Dataset S1). Using fold-changes greater than 0.5 (log2 scale) and p < 0.01 as thresholds, we found 20 probesets with decreased expression in KP3365 unc-43(n1186) CaMKII mutants as compared to wild-type controls (Figure 3). Eight of these probesets correspond to sequences on Chromosome 4, which contains unc-43 CaMKII and approximately 2,900 other genes. Strikingly, seven of these eight probesets correspond to the unc-43 gene. These seven probesets correspond to nonoverlapping regions of the coding sequence, as well as the 5′ and 3′ untranslated region of unc-43 (Figure S1). Only two of these probesets (193459_s_at and 193463_s_at) were annotated as corresponding to the unc-43 CaMKII mRNA transcript according to the annotation of the C. elegans GeneChip provided by Affymetrix (downloadable at http://www.affymetrix.com). During our examination of the eight candidate probesets that showed decreased expression and were on Chromosome 4, we discovered the five additional probesets that corresponded to unc-43 CaMKII. These additional probesets provided a serendipitous blind control, since we were not aware of their existence until they appeared on our candidate list from the hybridization. This redundancy in probes is a result of overlap in the various databases used to design the GeneChip and inaccuracies in gene predications at the time of design (December 2000). Approximately 2,400 (13%) of the genes on the C. elegans GeneChip are represented by multiple probes. While it is unusual to have seven probesets on a microarray corresponding to a single gene, this illustrates the robustness of our ability to measure the changes in expression of unc-43 CaMKII. This demonstrates that microarray hybridizations can be used to measure the decrease in unc-43 transcript in the unc-43(n1186) mutant. These results also suggest that microarray hybridizations utilizing printed arrays and those utilizing Affymetrix chips are both sufficiently sensitive to detect changes in the abundance of mutant mRNAs, even in the relatively stringent cases of transcription factors and signal transduction components.
Figure 3 Analysis of mRNA Abundance in the KP3365 unc-43(n1186) CaMKII Strain
Expression data are illustrated as described in Figure 2. The symbol × indicates probesets with reduced expression in KP3365 animals (fold-change < −0.5, p < 0.01). Filled blue circles indicate probesets with reduced expression that are also on Chromosome 4. Open red circles indicate probesets corresponding to the unc-43 CaMKII gene. The black square indicates the probeset corresponding to the hif-1 gene.
Identification of a hif-1 Polymorphism in the KP3365 Strain
One potential limitation of our strategy is that mutant strains may contain multiple mutations, some of which do not contribute to the mutant phenotype. This will be particularly true in heavily mutagenized strains, and in cases where the mutants have not been extensively backcrossed with wild-type strains. Therefore, we examined the KP3365 unc-43(n1186) CaMKII hybridization data for other genes with significantly reduced expression. Interestingly, the gene with the largest decrease in expression in KP3365 unc-43(n1186) animals was not unc-43; rather, it was hif-1 (Figure 3), which encodes the worm ortholog of hypoxia-inducible factor 1α, a transcription factor that mediates transcriptional responses to oxygen deprivation [35]. This is the only probe corresponding to hif-1 on the C. elegans GeneChip. There are two likely explanations for the observed decrease in hif-1 transcript levels: either hif-1 expression is regulated by unc-43 CaMKII, or the KP3365 strain contains a loss-of-function polymorphism in the hif-1 gene. Sequencing genomic DNA from KP3365 animals revealed two mutations in the last exon of hif-1 (nu469) (Figure 4A). Since neither of these mutations results in a premature stop, why is transcript level decreased? To address this issue, we sequenced hif-1 cDNA made from wild-type and KP3365 animals. This revealed that the nu469 mutations cause an aberrant splicing of the hif-1 mRNA, removing 135 base pairs from the last exon (Figure 4A). We confirmed this change in hif-1 splicing by RT-PCR (Figure 4B). The aberrantly spliced hif-1(nu469) transcript is likely to have reduced stability.
Figure 4 Characterization of the Splicing Defect in hif-1(nu469)
(A) Diagram of hif-1 gene structure and the two mutations in hif-1(nu469), a previously uncharacterized lesion in the background of the KP3365 strain. This lesion consists of two closely linked mutations: (1) C→A at position 2315 of the coding sequence, resulting in a P771Q mutation, and (2) an insertion of TTATCA after position 2373. Sequencing cDNA from KP3365 revealed that these two mutations result in the inappropriate splicing of the hif-1 transcript (indicated by a dashed line), removing 135 base pairs of the last exon. Location of primers for PCR are indicated by half-arrows above exons 9 and 11.
(B) RT-PCR confirmed the altered splicing of the hif-1(nu469) mRNA.
In aerobic conditions, the HIF-1 protein is constitutively degraded by the von Hippel–Lindau ubiquitin ligase [36–38]; consequently, the hif-1(nu469) mutation would presumably be phenotypically silent in normal growth conditions. The hif-1(nu469) mutation was not present in several other strains containing the unc-43(n1186) allele, suggesting that this mutation occurred spontaneously during culturing in our laboratory (data not shown). In summary, KP3365 animals carry a previously uncharacterized polymorphism in hif-1, which we identified based solely on our microarray hybridization results. Identifying such polymorphisms may allow researchers to explain unexpected aspects of mutant phenotypes of particular strains.
Using Microarrays to Identify a Mutation in Tomosyn, an Inhibitor of Neurotransmitter Secretion
To further address whether microarray hybridizations can be used to identify uncharacterized mutations, we analyzed a behavioral mutant that was isolated in a forward genetic screen for inhibitors of neurotransmitter secretion. Neurotransmission serves as the primary mode of communication between cells in the nervous system. Neurotransmitters such as acetylcholine (ACh) are secreted by presynaptic nerve cells, and activate receptors on postsynaptic cells. Behavioral and pharmacological screens in C. elegans have proven to be a powerful approach to identifying molecules involved in synaptic transmission and nervous system function [39–42]. The cholinesterase inhibitor aldicarb is widely used as a means to monitor ACh secretion at the C. elegans neuromuscular junction [41,43–46]. In the presence of aldicarb, ACh accumulates in the synaptic cleft, causing the body wall muscles to become hypercontracted and animals to become paralyzed. Mutations that increase ACh secretion cause hypersensitivity to the paralytic effects of aldicarb [46–48]. To identify negative regulators of ACh secretion, we used hypersensitivity to aldicarb as the basis for a forward genetic screen. One of the strongest mutations recovered in our screen was nu468 (filled squares in Figure 5).
Figure 5 Aldicarb Sensitivity and Rescue of KP3293 nu468
Levels of ACh secretion were assayed by following the time course of paralysis of animals on 1 mM aldicarb. Filled circle, wild-type; filled square, nu468, mutant recovered in our screen; filled triangle, ok285 deletion allele of tom-1; open circle, nu468 with a transgene expressing tom-1 cDNA under the unc-17 promoter. Data shown are averages from seven trials, except for ok285, which is an average of five trials. Error bars represent standard error.
We meiotically mapped nu468 to Chromosome 1, which contains approximately 2,700 genes. We then hybridized RNA from KP3293 nu468 animals to the C. elegans GeneChip, comparing the hybridizations to wild-type hybridizations as previously described (Dataset S1). Six probesets showed significantly decreased expression in KP3293 animals (fold-change < −0.5, p < 0.01) (Figure 6), of which two corresponded to genes on Chromosome 1. Sequencing DNA from the mutant revealed a nonsense mutation in one of these genes, tom-1, the C. elegans ortholog of mammalian tomosyn (Figures 7A and S2). This lesion, a predicted NMD target, is consistent with the decreased transcript levels that we observed by the microarray hybridization. This is the only probe corresponding to tom-1 on the C. elegans GeneChip.
Figure 6 Positional Cloning of tom-1(nu468)
Expression data are illustrated as described in Figure 2. The symbol × indicates probesets with reduced expression in KP3293 nu468 (fold-change < −0.5, p < 0.01). Filled blue circles indicate probesets with reduced expression in KP3293 nu468 that are also on Chromosome 1. The open red circle indicates the probeset corresponding to tom-1. Sequencing of the tom-1 gene in KP3293 nu468 revealed a W212Stop mutation in the tom-1 gene (see Figure 7A).
Figure 7 Expression of TOM-1, the C. elegans Ortholog of Tomosyn
(A) Schematic of worm tomosyn indicating the location of the premature stop found in nu468 and deletion in ok285.
(B–D) Expression pattern of tom-1 characterized with 4.2 kb of sequence upstream of the start codon driving expression of green fluorescent protein. Expression is seen in ventral cord motor neurons, with cell bodies indicated by arrowheads (B) and a number of neurons in the head (C) and the tail (D). Scale bars = 10 μm.
We performed several experiments to confirm that the tom-1(nu468) mutation caused the aldicarb hypersensitivity observed in the KP3293 strain. First, we tested a second tom-1 allele, ok285, which was generated by the C. elegans Gene Knockout Consortium (http://celeganskoconsortium.omrf.org). This allele, tom-1(ok285), encodes a mutant protein lacking 202 residues in a highly conserved region, and homozygous tom-1(ok285) mutants exhibited aldicarb hypersensitivity similar to that observed in KP3293 tom-1(nu468) mutants (triangles in Figure 5). Second, the aldicarb hypersensitivity of KP3293 tom-1(nu468) animals was rescued by a transgene driving expression of a tom-1 cDNA in cholinergic neurons (open circles in Figure 5). Third, we found that a 4.2-kb tom-1 promoter fragment was sufficient to drive expression of green fluorescent protein in the ventral cord motor neurons (Figure 7B), consistent with the idea that tomosyn acts as an inhibitor of ACh secretion in motor neurons. Tomosyn also showed expression in several neurons in head (Figure 7C) and tail ganglia (Figure 7D).
Tomosyn is an approximately 1,100–amino acid protein with two functional domains: (1) the C-terminal coiled-coil domain, which shares homology with synaptobrevin and has been shown to bind to syntaxin and SNAP-25, and (2) the approximately 600-residue WD40-rich N-terminal region, which shows strong homology to the fly tumor suppressor protein Lethal giant larvae (Figure 7A) [49,50]. Previous studies have shown that the synaptobrevin-like coiled-coil domain of rat tomosyn binds syntaxin and SNAP-25, forming a SNARE-like complex that occludes synaptobrevin [51]. This suggests a mechanism whereby tomosyn competitively inhibits secretion by preventing SNARE complex formation. Supporting this hypothesis, overexpression of tomosyn in neuroendocrine cells results in a decrease in exocytosis in response to depolarization [49–52]. While these overexpression studies show that tomosyn can function to inhibit dense core vesicle release, they do not address the endogenous function of tomosyn. Our results provide the first in vivo evidence suggesting that endogenously expressed tomosyn inhibits neurotransmitter secretion in neurons.
Generalizability of Microarray-Assisted Cloning
Since NMD functions in all eukaryotes [16,18], we wondered whether our strategy could be applied to other model systems. To address this, we conducted a retrospective analysis of microarray data from mutants in other organisms. We searched the public microarray databases for experiments in which researchers had analyzed mutants in other organisms. Specifically, we looked for hybridizations where mutant RNA had been compared to wild-type RNA and where the mutation was the result of a premature stop codon (and thus a predicted NMD target). For practical reasons, we also required that the mutant gene be represented and detectable on the microarray. Surprisingly, we found that only two experiments met these criteria. The first was a study of pmr4
(powdery mildew resistant 4), a cell-wall biosynthesis gene in Arabidopsis that confers pathogen resistance when mutated. The lesion used in the microarray studies was a premature stop codon in the second exon (PMR4 dataset) [53]. The second was a study of the mdx mouse, an animal model of Duchenne muscular dystrophy, with a premature stop codon in exon 23 of dystrophin (MDX dataset) [54,55]. In both of these studies, the authors knew the nature of the mutation and were attempting to find genes whose expression changed in the mutant background.
For these two examples, we asked retrospectively whether hybridization data would have aided identification of the mutant genes. To do this, we reanalyzed the PMR4 and MDX data as described for the C. elegans mutants and constructed volcano plots (Figure 8). Each of these datasets had a different distribution on the fold-change and significance axes because of differences in the sample preparation, labeling efficiency, type of GeneChip, and number of hybridizations. We therefore adjusted our significance and ratio thresholds for differential expression (see Materials and Methods). In the PMR4 hybridizations, 17 genes showed significantly decreased expression (fold-change < −1.0, p < 0.01). Of these, two genes were on Chromosome 4, the most significant of which was PMR4 (Figure 8A). In the case of the MDX data, 22 genes showed significantly decreased expression (fold-change < −1.5, p < 0.0001). Of these, only the dystrophin gene was on X (Figure 8B). In each case, there was only one probe on the GeneChip corresponding to the mutant gene. For both of these examples, the combination of microarray data and chromosomal mapping quickly reduced the number of candidates to one or two genes.
Figure 8 Analysis of Nonsense Mutants in Arabidopsis and Mouse
(A) PMR4 mutant; (B) MDX mutant. Expression data are illustrated as described in Figure 2. The symbol × indicates probesets with reduced expression in the nonsense mutant (fold-change < −1.0, p < 0.01) for PMR4 and (fold-change < −1.5, p < 0.0001) for MDX. Numbers of genes with significantly reduced expression are indicated for both mutants. Filled blue circles indicate probesets with reduced expression that are on same chromosome as the mutant gene. The open red circle indicates the probeset corresponding to the mutant gene.
Discussion
We present evidence demonstrating the utility of microarray hybridizations in facilitating the rapid identification of mutations isolated in forward genetic screens. Several results suggest that this technique will be widely applicable. This strategy was successful in identification of C. elegans, mouse, and Arabidopsis mutations. Mutations were successfully identified in both transcription factors and signal transduction components, which are likely to be the most challenging cases. Mutant genes were successfully detected using data obtained with both printed arrays and Affymetrix chips. And finally, we were able to identify two previously uncharacterized C. elegans mutations with this approach.
Will this strategy work for genes that regulate the expression of many other genes? We provide examples for successful identification of three genes that directly affect transcription—two transcription factors (mec-3 and hif-1) and a protein kinase that regulates neuronal gene expression (unc-43). Although 70 genes were differentially expressed in mec-3 mutants, only three differentially expressed genes mapped within a 2-cM interval containing mec-3 and 100 other genes. Therefore, microarray hybridizations would have facilitated identification of mec-3.
The success rate for this strategy depends on three factors: (1) the fraction of genes that are detectable by microarray, (2) the frequency of nonsense alleles recovered in screens, and (3) the efficiency with which nonsense mutated mRNAs are degraded by NMD. In our hybridizations using mRNA prepared from whole worms, 80% of the genes on the array showed detectable expression. In cases where a mutation affects a particular cell type or tissue, the likelihood of detecting a particular transcript can be increased using RNA isolated from that tissue or cell type [28,56].
What fraction of newly isolated mutations will be nonsense alleles (see Figure 1)? Our analysis suggests that for 20% of C. elegans genes, nonsense mutations are rarely recovered. For the remaining 80% of C. elegans genes, 45% of alleles recovered were nonsense alleles. Given these ratios, the rate of successful gene identification by microarray hybridization could be increased by analyzing multiple alleles of the same gene and by selecting genes for which null alleles are the most frequently isolated class of alleles recovered in screens. Furthermore, microarray-assisted cloning will likely be useful for other categories of alleles that decrease transcript abundance, e.g., the spontaneous hif-1(nu469) mutation. These include mutations altering pre-mRNA splicing, mutations in promoters, frameshift mutations, and mutations yielding transcripts that lack termination codons [57–59].
What fraction of nonsense alleles are efficiently targeted by the NMD machinery? In each of the six examples we present above, this was the case, but how often do nonsense transcripts evade degradation by the NMD machinery? Rules governing NMD recognition of mutant mRNAs have been described in yeast, C. elegans, and mammals [16–18,59–63]. The NMD machinery distinguishes premature stop codons from natural stops using the exon-junction complexes that are deposited at exon–exon boundaries by the spliceosome. Stops that are greater than 50–55 base pairs upstream of the last exon-junction complex are recognized by the NMD machinery as premature and are efficiently targeted for destruction [61,64]. Prior studies have shown that 100% (n = 23) of C. elegans nonsense mutations were susceptible to NMD surveillance (measured either by mRNA abundance or by suppression of mutant phenotypes by NMD pathway mutations) [17]. Of these, six mutations (26%) were judged to be only partially targeted by NMD. Based on these examples and those we describe here, we estimate that 75%–100% of nonsense alleles in C. elegans would show a detectable decrease in mRNA levels. Considering all three of these factors (gene detection by microarray, nonsense allele frequency, and NMD efficiency), we expect microarray-assisted cloning to be successful in 25%–30% of positional clonings (assuming only one allele is hybridized per gene).
The principal costs of positional cloning are those incurred in isolating, phenotyping, and genotyping a sufficient number of recombinants (i.e., informative meioses) to map a mutation to a small genetic interval. A typical positional cloning requires 2,000–10,000 informative meioses. Our results suggest that microarray hybridizations can significantly reduce the number of meioses required for positional clonings. In five of six cases, microarray data in conjunction with chromosomal linkage data were sufficient for gene identification. Therefore, while we expect that this strategy will be useful in many genetic systems, microarray-assisted cloning promises to provide the greatest value in organisms such as mouse and zebrafish, where long generation times and large genomes make meiotic mapping more time consuming and costly. Furthermore, microarray-assisted cloning may be particularly useful in cases where mutant phenotypes are more difficult to assess, such as behavioral mutants, or incompletely penetrant or complex (i.e., multigenic) traits [65,66]. Given the effort and challenges involved in meiotic mapping and the relative ease and speed of microarray hybridizations, we believe that this microarray-based strategy provides significant benefit, even though it will be successful in only a subset of cases.
Can microarrays be used to aid the cloning of human disease genes? One-third of human disease genes are predicted to be caused by nonsense lesions or mutations that decrease transcript abundance [21,22]. Furthermore, nonsense mutant transcripts encoded by disease genes such as BRCA1 and hepatocyte nuclear factor 1α have been shown to be effectively degraded by NMD [67,68]. Given the enormous time and expense involved in mapping genes in humans, the strategy described here could provide a valuable addition to the toolbox of human geneticists.
Materials and Methods
Allele analysis.
Information about 930 recessive single base-pair substitution alleles was downloaded from WormBase (http://www.wormbase.com), Release WS123 (see Table S1). Information about 82 additional alleles was obtained through literature searches. Based on their molecular description, 943 alleles were classified as either NMD targets (nonsense) or non-NMD targets (missense). Excluded from the analysis were 69 alleles that could not be definitively classified. These alleles included those with incomplete molecular descriptions and those with lesions such as splice site mutations that could not be classified without further characterization.
RNA sample preparation.
Animals analyzed in microarray experiments were first synchronized by hypochlorite treatment and arrested at the first larval stage by incubation for 22 h in M9 [27,69]. Animals were then grown at 20 °C on 15-cm NG HB101 plates until the fourth larval stage (approximately 46 h). Animals were washed, harvested in M9, and then flash-frozen in liquid nitrogen and stored at −80 °C. Total RNA was prepared by Trizol extraction (Invitrogen, Carlsbad, California, United States).
Microarray target preparation and hybridization.
Targets were prepared and hybridized at the Harvard Medical School Biopolymer Facility. Starting with 10 μg of total RNA, first-strand cDNA was synthesized as described in the Affymetrix (Santa Clara, California, United States) expression technical manual. Briefly, 10 μg of RNA was added to 1 μl of 50 μM T7 primer (HPLC purified) (Integrated DNA Technologies, Coralville, Iowa, United States) in a volume of 9 μl. Then 1 μl of each Poly A spike control (5 nM) was added to the RNA, and T7 was added as an internal control. Poly A spikes were created from Poly A–tailed genes from Bacillus subtilis cloned into Stratagene (La Jolla, California, United States) pBluescript as an XhoI-to-NotI insert 5′–3′, respectively, and commercially available through ATCC (Manassas, Virginia, United States) (see Affymetrix technical expression manual). The RNA, T7, and Poly A spike controls were heated to 70 °C for 10 min and then placed on ice for 5 min. The RNA, T7, and Poly A mix was then heated to 42 °C. Then 4 μl of 5× first-strand buffer (Invitrogen), 2 μl of 0.1 M DTT (Invitrogen), 1 μl of 10 mM dNTP (Invitrogen), and 1 μl of Superscript II, RNase H− was added to the RNA and incubated at 42 °C for 1 h. Double-strand DNA was created via a replacement reaction under the following conditions. To the 20-μl first-strand reaction was added 91 μl of water, 30 μl of second-strand buffer (Invitrogen), 3 μl of 10 mM dNTP (Invitrogen), 1μl of Escherichia coli DNA ligase (Invitrogen), 1 μl of RNase H (Invitrogen), and 4 μl of E. coli DNA polymerase (Invitrogen). This 130-μl second-strand mix was added to the first-strand reaction and incubated at 16 °C for 2 h, then 2 μl of T4 DNA polymerase was added for 5 min at 16 °C, then the reaction was phenol-chloroform-extracted using 150 μl of phenol chloroform isoamyl alcohol (pH 7) (Ambion, Austin, Texas, United States), and the organic and aqueous phases were separated using a 1.5-ml phase lock heavy gel (Brinkmann Eppendorf, Westbury, New York, United States). The 150-μl aqueous layer was removed and precipitated in 375 μl of 100% ethanol and 15 μl of 3 M sodium acetate (Sigma, St. Louis, Missouri, United States). The cDNA pellet was isolated using an Eppendorf (Hamburg, Germany) 5415C centrifuge at room temperature for 20 min. Ethanol was aspirated and the pellet washed in 75% ethanol, centrifuged for 10 min, and aspirated. The cDNA pellet was rehydrated using 22 μl of nuclease-free water (Ambion) and used with the BioArray HighYield RNA Transcript Labeling Kit T7 (Enzo Life Sciences, Farmingdale, New York, United States). The resulting biotinylated cRNA probes were purified using RNeasy columns (Qiagen, Valencia, California, United States) and quantitated using A260 with an Agilent (Palo Alto, California, United States) 8453 spectrophotometer. Then 15 μg of labeled probe was fragmented with 5× fragmentation buffer (see Affymetrix technical manual) and combined with hybridization controls (Affymetrix), herring sperm DNA (Promega, Madison, Wisconsin, United States), and BSA (Invitrogen) to create 300 μl of hybridization mix. Of this, 200 μl was added to the Affymetrix C. elegans GeneChip. Hybridization was done in a GeneChip Hybridization Oven 320 for 16 h at 45 °C, processed on an Affymetrix Fluidics Station 400 using double amplification staining (see Affymetrix technical manual), and washed using fluidics protocol EukGE-WS2v4. The GeneChips were then scanned on a Hewlett-Packard (Palo Alto, California, United States) GeneArray Scanner.
Public datasets.
Descriptions of all available hybridizations in the C. elegans whole-genome expression profiles were downloaded from the Stanford Microarray Database (http://genome-www5.stanford.edu). These were then searched to find direct mutant versus wild-type comparisons. The 368 hybridizations that were publicly accessible included a large number of developmental time courses, aging experiments, and heat-shock and tissue-specific expression profiles (see Figure 2 in [27] for more detail). The only direct mutant versus wild-type comparison was the mec-3(e1338) analysis, which consisted of six hybridizations. For these experiments, normalized log expression ratios were downloaded from the Stanford Microarray Database (ExptSetNo = 1461). Affymetrix expression values for mouse and Arabidopsis datasets were downloaded from NCBI's Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo).
Microarray data analysis.
For Affymetrix data, probesets were first filtered to eliminate those that showed no detectable signal. A threshold of 32 was used for the C. elegans and Arabidopsis data. A threshold of 256 was used for the mdx data because these data showed significantly higher signals than the other datasets. This is most likely because the RNA for these experiments was prepared from a single tissue (mouse skeletal muscle), as opposed to the C. elegans and Arabidopsis RNA, which was derived from the whole organism. For printed arrays, only spots that showed detectable signal (mean signals greater than 1.5 standard deviations above background) were included in the analysis.
Probes were classified as differentially expressed based on two criteria: fold-change and statistical significance. For Affymetrix data, fold-change was calculated as the average expression in the mutant divided by average expression in wild-type. For printed arrays (mec-3), fold-change was calculated by averaging the expression ratio in each of the mutant–versus–wild-type replicate hybridizations. This ratio provides a measure of the magnitude of expression difference between mutant and wild-type samples. To assess the statistical significance of expression differences, we compared the replicate expression values in the mutant hybridizations to replicate expression values in the wild-type hybridizations using a Student's t-test, and calculated a p-value. We then constructed a volcano plot with the log2(fold-change) on the x-axis and negative log10(p-value) on the y-axis [31]. Cutoffs for differential expression were based on shape and distribution of individual volcano plots.
Raw image files were converted to probeset data (.cel files) in Microarray Suite (MAS 5.0). The nine probeset data files were normalized together and expression values were determined using the Robust Multi-chip Average (RMA) method as implemented in RMA Express (http://stat-www.berkeley.edu/~bolstad/RMAExpress/RMAExpress.html). Subsequent analysis was done using the R statistical computing package (http://www.r-project.org) and the Bioconductor libraries (http://www.bioconductor.org). Graphs were produced in Igor Pro 4.0 (WaveMetrics, Lake Oswego, Oregon, United States). Probeset annotations were downloaded from the Affymetrix Web site (http://www.affymetrix.com).
Molecular characterization of hif-1(nu469).
Using RNA prepared from KP3365 and wild-type animals, first-strand cDNA was synthesized using a primer specific for the 3′ UTR of hif-1 (Invitrogen). The hif-1 gene was then amplified by PCR from this cDNA and sequenced.
Isolation and mapping of the tom-1 mutation.
The tom-1(nu468) allele was isolated in an ethyl methane sulfonate screen for mutants that displayed hypersensitivity to aldicarb. F2 progeny of mutagenized animal were transferred to agar plates containing 0.5 mM aldicarb (Chem Service, West Chester, Pennsylvania, United States). After 1 h, a time point at which all wild-type worms were still moving, paralyzed animals were transferred to separate plates and rescreened for aldicarb sensitivity in subsequent generations. nu468 was determined to be recessive and was mapped to Chromosome 1 using conventional meiotic mapping.
Analysis of aldicarb sensitivity.
Aldicarb sensitivity was assessed essentially as described [45,46]. Briefly, for each experiment, 20 to 25 animals were transferred to agar plates containing 1 mM aldicarb (Chem Service). Paralysis was assessed every 10 min by prodding each animal with a platinum wire. Data from independent trials were averaged and used to calculate standard error. All experiments were conducted blind with respect to the genotype of the animals.
Molecular characterization of tom-1(ok285).
Genomic sequence from VC223 animals, available on WormBase, shows a deletion of 1,580 nucleotides that removes all of exons 11–13 and part of exon 10. To characterize the effect of this lesion on the tom-1 gene product, we purified RNA from VC223 animals and amplified the mutated tom-1 mRNA by RT-PCR (Invitrogen). Sequencing revealed that the genomic deletion results in an in-frame lesion in the mRNA, removing 606 nucleotides of coding sequence, and adding 23 nucleotides of intronic sequence and a 490-base-pair alternative exon from isoform C of tom-1 that is located just downstream of the deletion.
Rescue of tom-1(nu468).
tom-1(nu468) was rescued using the full-length cDNA of the major splice form of tom-1 (M01A10.2a) under the promoter of unc-17 synaptic vesicle ACh transporter. A transgenic strain was isolated by microinjecting the rescuing plasmid at 100 ng/μl using pttx-3::dsRed as a marker into nu468. During characterization and rescue of tom-1, we discovered that the start site and first exon were incorrectly predicted and described in WormBase. We identified the correct start site and initial two exons of tom-1 by performing RT-PCR using a primer complementary to the trans-splice acceptor (SL1). This corrected version of tom-1 shows much better alignment to the N-terminus of mammalian homologs (see Figure S2).
Supporting Information
Dataset S1 Microarray Expression Data from Wild-Type, unc-43(n1186), and tom-1(nu468) Animals
(2 MB TDS)
Click here for additional data file.
Figure S1 Probeset Alignments to unc-43 CaMKII Isoform H (K11E8.1h)
Target sequences were downloaded (http://www.affymetrix.com) and aligned to unc-43 CaMKII isoform H. Since a majority of genes on the C. elegans GeneChip are represented by one probeset, unc-43 represents an atypica1 case. To explain this, it is useful to consider the history and design of the C. elegans GeneChip (see http://www.affymetrix.com/support/technical/datasheets/celegans_drosophila_datasheet.pdf). The targets for this GeneChip were designed by Affymetrix based on more than 18,800 Sanger Center predicted transcripts from the December 2000 genome sequence, as well as 2,300 3' EST clusters and 300 GenBank mRNAs (Release 121). Despite efforts to eliminate redundancy, there is not a strict one-to-one correspondence between the current set of genes and the probesets on the GeneChip. Our analysis indicates that 13% of the genes on the GeneChip are represented by more than one probe. Even so, having unc-43 CaMKII represented by seven probesets is an unusual situation. However, only two of these probesets (193459_s_at and 193463_s_at) are annotated as corresponding to the unc-43 CaMKII mRNA transcript according to annotations of the C. elegans GeneChip provided by Affymetrix in the March 28, 2003, update (downloadable at http://www.affymetrix.com). During our examination of the eight candidate probesets that showed decreased expression and were on Chromosome 4, we discovered five additional probesets that corresponded to unc-43 CaMKII. Four of these probesets (172058_x_at, 175820_s_at, 175821_s_at, and 175824_s_at) were based on GenBank sequences, and one (187759_s_at) was based on a predicted open reading frame, Y43C5B.1, that was part of the genome as of December 2000 but has since been shown to correspond to the 5′ UTR of unc-43 CaMKII. These additional probesets provided a serendipitous blind control, since we were not aware of their existence until they appeared on our candidate list from the hybridization. There is also an additional (eighth) probeset (173423_at) described in the Affymetrix annotation as corresponding to unc-43 CaMKII. However, based on the current gene model, 173423_at aligns to the intron between exons 11 and 12. As would be expected, this probeset shows no detectable expression and thus was not considered in our analysis of the unc-43 CaMKII mutant.
(33 KB PDF)
Click here for additional data file.
Figure S2 Alignment of TOM-1 to Human and Rat Tomosyn
Sequences of human and mouse tomosyn were downloaded from Ensembl (http://www.ensembl.org). The multiple sequence alignment was performed using T-Coffee (http://www.ch.embnet.org/software/TCoffee.html). Alignment output was produced using GeneDoc (http://www.psc.edu/biomed/genedoc).
(104 KB PDF)
Click here for additional data file.
Table S1 List of Missense and Nonsense Alleles
(51 KB PDF)
Click here for additional data file.
Accession Numbers
The National Center for Biotechnology Information Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo) accession number for the data generated by the authors and discussed in this publication is GSE2210. The GEO accession numbers for datasets downloaded by the authors and discussed in this paper are mdx mouse (GDS236), mdx mutant (GDS236), PMR4 Arabidopsis (GDS417), and PMR4 mutant (GDS417).
The GenBank (http://www.ncbi.nlm.nih.gov/Genbank) accession numbers for genes and gene products discussed in this paper are tom-1(ok285) (AY912103) and tomosyn (AY912102).
The Ensembl (http://www.ensembl.org/) accession numbers for genes and gene products discussed in this paper are human tomosyn (ENSP00000179882) and mouse tomosyn (ENSRNOP00000018806).
We thank K. Vranizan for statistical and programming advice; T. Rector for Affymetrix sample preparation and hybridizations; J. Rine, D. Altshuler, V. Mootha, S. McCarroll, and C. Murphy for helpful discussions; T. Speed for statistical input; P. Juo, L. Drier, J. Dittman, and I. Ruvinsky for reading this manuscript; other members of the Kaplan and Ruvkun labs for suggestions; K. Scott for providing a fantastic foster lab for MD in Berkeley; and Gregoire for sustenance. Some of the strains described here were provided by the C. elegans Genetic Stock Center. This work was supported by a grant from the National Institutes of Health to JK (GM54728). MD was supported by a Howard Hughes Medical Institute predoctoral fellowship.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. MD, JN, and JMK conceived and designed the experiments. MD performed the experiments and analyzed the data. MD and JMK wrote the paper.
Abbreviations
AChacetylcholine
CaMKIItype II calcium- and calmodulin-dependent protein kinase
cMcentimorgan
mRNAmessenger RNA
NMDnonsense-mediated decay
==== Refs
References
Haffter P Granato M Brand M Mullins MC Hammerschmidt M 1996 The identification of genes with unique and essential functions in the development of the zebrafish, Danio rerio
Development 123 1 36 9007226
Driever W Solnica-Krezel L Schier AF Neuhauss SC Malicki J 1996 A genetic screen for mutations affecting embryogenesis in zebrafish Development 123 37 46 9007227
Zarbalis K May SR Shen Y Ekker M Rubenstein JL 2004 A focused and efficient genetic screening strategy in the mouse: Identification of mutations that disrupt cortical development PLoS Biol 2 e219. DOI: 10.1371/journal.pbio.0020219 15314648
O'Brien TP Frankel WN 2004 Moving forward with chemical mutagenesis in the mouse J Physiol 554 13 21 14678486
Nadeau JH Balling R Barsh G Beier D Brown SD 2001 Sequence interpretation. Functional annotation of mouse genome sequences Science 291 1251 1255 11233449
Bessereau JL Wright A Williams DC Schuske K Davis MW 2001 Mobilization of a Drosophila transposon in the Caenorhabditis elegans germ line Nature 413 70 74 11544527
Cooley L Kelley R Spradling A 1988 Insertional mutagenesis of the Drosophila genome with single P elements Science 239 1121 1128 2830671
Gaiano N Amsterdam A Kawakami K Allende M Becker T 1996 Insertional mutagenesis and rapid cloning of essential genes in zebrafish Nature 383 829 832 8893009
Breindl M Harbers K Jaenisch R 1984 Retrovirus-induced lethal mutation in collagen I gene of mice is associated with an altered chromatin structure Cell 38 9 16 6088079
Korswagen HC Durbin RM Smits MT Plasterk RH 1996 Transposon Tc1-derived, sequence-tagged sites in Caenorhabditis elegans as markers for gene mapping Proc Natl Acad Sci U S A 93 14680 14685 8962114
Jansen G Hazendonk E Thijssen KL Plasterk RH 1997 Reverse genetics by chemical mutagenesis in Caenorhabditis elegans
Nat Genet 17 119 121 9288111
Rong YS Golic KG 2000 Gene targeting by homologous recombination in Drosophila
Science 288 2013 2018 10856208
Bradley A Zheng B Liu P 1998 Thirteen years of manipulating the mouse genome: A personal history Int J Dev Biol 42 943 950 9853825
Stemple DL 2004 TILLING—A high-throughput harvest for functional genomics Nat Rev Genet 5 145 150 14726927
McCallum CM Comai L Greene EA Henikoff S 2000 Targeting induced local lesions IN genomes (TILLING) for plant functional genomics Plant Physiol 123 439 442 10859174
Hentze MW Kulozik AE 1999 A perfect message: RNA surveillance and nonsense-mediated decay Cell 96 307 310 10025395
Mango SE 2001 Stop making nonSense: The C. elegans smg genes Trends Genet 17 646 653 11672865
Wilusz CJ Wang W Peltz SW 2001 Curbing the nonsense: The activation and regulation of mRNA surveillance Genes Dev 15 2781 2785 11691829
Huusko P Ponciano-Jackson D Wolf M Kiefer JA Azorsa DO 2004 Nonsense-mediated decay microarray analysis identifies mutations of EPHB2 in human prostate cancer Nat Genet 36 979 983 15300251
Noensie EN Dietz HC 2001 A strategy for disease gene identification through nonsense-mediated mRNA decay inhibition Nat Biotechnol 19 434 439 11329012
Frischmeyer PA Dietz HC 1999 Nonsense-mediated mRNA decay in health and disease Hum Mol Genet 8 1893 1900 10469842
Mendell JT Dietz HC 2001 When the message goes awry: Disease-producing mutations that influence mRNA content and performance Cell 107 411 414 11719181
Raser JM O'Shea EK 2004 Control of stochasticity in eukaryotic gene expression Science 304 1811 1814 15166317
Fraser HB Hirsh AE Giaever G Kumm J Eisen MB 2004 Noise minimization in eukaryotic gene expression PLoS Biol 2 e137. DOI: 10.1371/journal.pbio.0020137 15124029
Brem RB Yvert G Clinton R Kruglyak L 2002 Genetic dissection of transcriptional regulation in budding yeast Science 296 752 755 11923494
Schadt EE Monks SA Drake TA Lusis AJ Che N 2003 Genetics of gene expression surveyed in maize, mouse and man Nature 422 297 302 12646919
Kim SK Lund J Kiraly M Duke K Jiang M 2001 A gene expression map for Caenorhabditis elegans
Science 293 2087 2092 11557892
Zhang Y Ma C Delohery T Nasipak B Foat BC 2002 Identification of genes expressed in C. elegans touch receptor neurons Nature 418 331 335 12124626
Way JC Chalfie M 1988
mec-3, a homeobox-containing gene that specifies differentiation of the touch receptor neurons in C. elegans
Cell 54 5 16 2898300
Xue D Tu Y Chalfie M 1993 Cooperative interactions between the Caenorhabditis elegans homeoproteins UNC-86 and MEC-3 Science 261 1324 1328 8103239
Cui X Churchill GA 2003 Statistical tests for differential expression in cDNA microarray experiments Genome Biol 4 210 12702200
Way JC Chalfie M 1989 The mec-3 gene of Caenorhabditis elegans requires its own product for maintained expression and is expressed in three neuronal cell types Genes Dev 3 1823 1833 2576011
Reiner DJ Newton EM Tian H Thomas JH 1999 Diverse behavioural defects caused by mutations in Caenorhabditis elegans
unc-43 CaM kinase II Nature 402 199 203 10647014
Sagasti A Hisamoto N Hyodo J Tanaka-Hino M Matsumoto K 2001 The CaMKII UNC-43 activates the MAPKKK NSY-1 to execute a lateral signaling decision required for asymmetric olfactory neuron fates Cell 105 221 232 11336672
Maxwell PH Wiesener MS Chang GW Clifford SC Vaux EC 1999 The tumour suppressor protein VHL targets hypoxia-inducible factors for oxygen-dependent proteolysis Nature 399 271 275 10353251
Epstein AC Gleadle JM McNeill LA Hewitson KS O'Rourke J 2001
C. elegans EGL-9 and mammalian homologs define a family of dioxygenases that regulate HIF by prolyl hydroxylation Cell 107 43 54 11595184
Jaakkola P Mole DR Tian YM Wilson MI Gielbert J 2001 Targeting of HIF-alpha to the von Hippel-Lindau ubiquitylation complex by O2 -regulated prolyl hydroxylation Science 292 468 472 11292861
Ivan M Kondo K Yang H Kim W Valiando J 2001 HIFalpha targeted for VHL-mediated destruction by proline hydroxylation: Implications for O2 sensing Science 292 464 468 11292862
Trent C Tsuing N Horvitz HR 1983 Egg-laying defective mutants of the nematode Caenorhabditis elegans
Genetics 104 619 647 11813735
Avery L 1993 The genetics of feeding in Caenorhabditis elegans
Genetics 133 897 917 8462849
Miller KG Alfonso A Nguyen M Crowell JA Johnson CD 1996 A genetic selection for Caenorhabditis elegans synaptic transmission mutants Proc Natl Acad Sci U S A 93 12593 12598 8901627
Nonet ML 1999 Studying mutants that affect neurotransmitter release in C. elegans
Bellen H Neurotransmitter release New York Oxford University Press 265 303
Nguyen M Alfonso A Johnson CD Rand JB 1995
Caenorhabditis elegans mutants resistant to inhibitors of acetylcholinesterase Genetics 140 527 535 7498734
Nonet ML Grundahl K Meyer BJ Rand JB 1993 Synaptic function is impaired but not eliminated in C. elegans mutants lacking synaptotagmin Cell 73 1291 1305 8391930
Lackner MR Nurrish SJ Kaplan JM 1999 Facilitation of synaptic transmission by EGL-30 Gqalpha and EGL-8 PLCbeta: DAG binding to UNC-13 is required to stimulate acetylcholine release Neuron 24 335 346 10571228
Nurrish S Segalat L Kaplan JM 1999 Serotonin inhibition of synaptic transmission: Galpha(0) decreases the abundance of UNC-13 at release sites Neuron 24 231 242 10677040
van der Linden AM Simmer F Cuppen E Plasterk RH 2001 The G-protein beta-subunit GPB-2 in Caenorhabditis elegans regulates the G(o)alpha-G(q)alpha signaling network through interactions with the regulator of G-protein signaling proteins EGL-10 and EAT-16 Genetics 158 221 235 11333232
Wang ZW Saifee O Nonet ML Salkoff L 2001 SLO-1 potassium channels control quantal content of neurotransmitter release at the C. elegans neuromuscular junction Neuron 32 867 881 11738032
Masuda ES Huang BC Fisher JM Luo Y Scheller RH 1998 Tomosyn binds t-SNARE proteins via a VAMP-like coiled coil Neuron 21 479 480 9768835
Fujita Y Shirataki H Sakisaka T Asakura T Ohya T 1998 Tomosyn: A syntaxin-1-binding protein that forms a novel complex in the neurotransmitter release process Neuron 20 905 915 9620695
Hatsuzawa K Lang T Fasshauer D Bruns D Jahn R 2003 The R-SNARE motif of tomosyn forms SNARE core complexes with syntaxin 1 and SNAP-25 and down-regulates exocytosis J Biol Chem 278 31159 31166 12782620
Yizhar O Matti U Melamed R Hagalili Y Bruns D 2004 Tomosyn inhibits priming of large dense-core vesicles in a calcium-dependent manner Proc Natl Acad Sci U S A 101 2578 2583 14983051
Nishimura MT Stein M Hou BH Vogel JP Edwards H 2003 Loss of a callose synthase results in salicylic acid-dependent disease resistance Science 301 969 972 12920300
Sicinski P Geng Y Ryder-Cook AS Barnard EA Darlison MG 1989 The molecular basis of muscular dystrophy in the mdx mouse: A point mutation Science 244 1578 1580 2662404
Tseng BS Zhao P Pattison JS Gordon SE Granchelli JA 2002 Regenerated mdx mouse skeletal muscle shows differential mRNA expression J Appl Physiol 93 537 545 12133862
Roy PJ Stuart JM Lund J Kim SK 2002 Chromosomal clustering of muscle-expressed genes in Caenorhabditis elegans
Nature 418 975 979 12214599
Frischmeyer PA van Hoof A O'Donnell K Guerrerio AL Parker R 2002 An mRNA surveillance mechanism that eliminates transcripts lacking termination codons Science 295 2258 2261 11910109
van Hoof A Frischmeyer PA Dietz HC Parker R 2002 Exosome-mediated recognition and degradation of mRNAs lacking a termination codon Science 295 2262 2264 11910110
Wickens M Anderson P Jackson RJ 1997 Life and death in the cytoplasm: Messages from the 3′ end Curr Opin Genet Dev 7 220 232 9115434
Lykke-Andersen J 2001 mRNA quality control: Marking the message for life or death Curr Biol 11 R88 R91 11231165
Singh G Lykke-Andersen J 2003 New insights into the formation of active nonsense-mediated decay complexes Trends Biochem Sci 28 464 466 13678954
Wagner E Lykke-Andersen J 2002 mRNA surveillance: The perfect persist J Cell Sci 115 3033 3038 12118059
Culbertson MR 1999 RNA surveillance. Unforeseen consequences for gene expression, inherited genetic disorders and cancer Trends Genet 15 74 80 10098411
Nagy E Maquat LE 1998 A rule for termination-codon position within intron-containing genes: When nonsense affects RNA abundance Trends Biochem Sci 23 198 199 9644970
Glazier AM Nadeau JH Aitman TJ 2002 Finding genes that underlie complex traits Science 298 2345 2349 12493905
Korstanje R Paigen B 2002 From QTL to gene: The harvest begins Nat Genet 31 235 236 12089518
Harries LW Hattersley AT Ellard S 2004 Messenger RNA transcripts of the hepatocyte nuclear factor-1alpha gene containing premature termination codons are subject to nonsense-mediated decay Diabetes 53 500 504 14747304
Perrin-Vidoz L Sinilnikova OM Stoppa-Lyonnet D Lenoir GM Mazoyer S 2002 The nonsense-mediated mRNA decay pathway triggers degradation of most BRCA1 mRNAs bearing premature termination codons Hum Mol Genet 11 2805 2814 12393792
Murphy CT McCarroll SA Bargmann CI Fraser A Kamath RS 2003 Genes that act downstream of DAF-16 to influence the lifespan of Caenorhabditis elegans
Nature 424 277 283 12845331
|
16103915
|
PMC1183521
|
CC BY
|
2021-01-05 08:00:24
|
no
|
PLoS Genet. 2005 Jul 25; 1(1):e2
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010002
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 10.1371/journal.pgen.001000305-PLGE-RA-0026R1plge-01-01-03Research ArticleAnimal BehaviorNeuroscienceGenetics/Gene DiscoveryCatPseudogenization of a Sweet-Receptor Gene Accounts for Cats' Indifference toward Sugar Sweet Taste in the Cat: Use It or Lose ItLi Xia 1Li Weihua 1Wang Hong 1Cao Jie 1Maehashi Kenji 1¤Huang Liquan 1Bachmanov Alexander A 1Reed Danielle R 1Legrand-Defretin Véronique 2Beauchamp Gary K 13Brand Joseph G 145*1 Monell Chemical Senses Center, Philadelphia, Pennsylvania, United States of America
2 The WALTHAM Centre for Pet Nutrition, Melton Mowbray, Leicestershire, United Kingdom
3 Department of Psychology, School of Arts and Sciences and Department of Anatomy, School of Veterinary Medicine, University of Pennsylvania, Philadelphia, Pennsylvania, United States of America
4 Department of Biochemistry, School of Dental Medicine, University of Pennsylvania, Philadelphia, Pennsylvania, United States of America
5 Veterans Affairs Medical Center, Philadelphia, Pennsylvania, United States of America
Flint Jonathan EditorUniversity of Oxford, United Kingdom*To whom correspondence should be addressed. E-mail: [email protected]¤ Current address: Department of Fermentation Science, Tokyo University of Agriculture, Tokyo, Japan
7 2005 25 7 2005 1 1 e316 2 2005 26 3 2005 Copyright: © 2005 Li et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Although domestic cats (Felis silvestris catus) possess an otherwise functional sense of taste, they, unlike most mammals, do not prefer and may be unable to detect the sweetness of sugars. One possible explanation for this behavior is that cats lack the sensory system to taste sugars and therefore are indifferent to them. Drawing on work in mice, demonstrating that alleles of sweet-receptor genes predict low sugar intake, we examined the possibility that genes involved in the initial transduction of sweet perception might account for the indifference to sweet-tasting foods by cats. We characterized the sweet-receptor genes of domestic cats as well as those of other members of the Felidae family of obligate carnivores, tiger and cheetah. Because the mammalian sweet-taste receptor is formed by the dimerization of two proteins (T1R2 and T1R3; gene symbols Tas1r2 and Tas1r3), we identified and sequenced both genes in the cat by screening a feline genomic BAC library and by performing PCR with degenerate primers on cat genomic DNA. Gene expression was assessed by RT-PCR of taste tissue, in situ hybridization, and immunohistochemistry. The cat Tas1r3 gene shows high sequence similarity with functional Tas1r3 genes of other species. Message from Tas1r3 was detected by RT-PCR of taste tissue. In situ hybridization and immunohistochemical studies demonstrate that Tas1r3 is expressed, as expected, in taste buds. However, the cat Tas1r2 gene shows a 247-base pair microdeletion in exon 3 and stop codons in exons 4 and 6. There was no evidence of detectable mRNA from cat Tas1r2 by RT-PCR or in situ hybridization, and no evidence of protein expression by immunohistochemistry. Tas1r2 in tiger and cheetah and in six healthy adult domestic cats all show the similar deletion and stop codons. We conclude that cat Tas1r3 is an apparently functional and expressed receptor but that cat Tas1r2 is an unexpressed pseudogene. A functional sweet-taste receptor heteromer cannot form, and thus the cat lacks the receptor likely necessary for detection of sweet stimuli. This molecular change was very likely an important event in the evolution of the cat's carnivorous behavior.
Synopsis
Although sweet sugars are ubiquitous in human foods, they are seldom added to cat food, and owners usually do not feed sweets to their cats. This is because, in contrast to most other mammals, both domestic cats and their wild cousins, the big cats, do not show a preference for and, most likely, cannot detect sweet-tasting compounds. Other than this sweet blindness, the cat's sense of taste is normal. The molecular mechanism for this unique behavior towards sweets was not known, until now. Sweet compounds, including sugars and artificial sweeteners, are recognized by a special taste bud receptor composed of the products of two genes. The authors found that in cats, one of these genes is not functional and is not expressed. (It is called a pseudogene.) Because the sweet receptor cannot be formed, the cat cannot taste sweet stimuli. During the evolution of the cats' strictly carnivorous behavior, selection to maintain a functional receptor was apparently relaxed. This research provides a molecular explanation for the common observation that the cat lives in a different sensory world than the cat owner.
Citation:Li X, Li W, Wang H, Cao J, Maehashi K, et al. (2005) Pseudogenization of a sweet-receptor gene accounts for cats' indifference toward sugar. PLoS Genet 1(1): e3.
==== Body
Introduction
The domestic cat (Felis silvestris catus), of the family Felidae in the order Carnivora, is an obligate carnivore. Its sense of taste is distinguished by a lack of attraction to, or indifference toward, compounds that taste sweet to humans, such as sweet carbohydrates (sugars) and high-intensity sweeteners [1–3]. This behavior toward sweet stimuli is in marked contrast to the avidity for sweets shown by most omnivores and herbivores and even some other carnivores such as the dog [4]. The indifference that cats display toward sweet-tasting compounds contrasts with their otherwise normal taste behavior toward stimuli of other taste modalities. For example, they show preference for selected amino acids [5] and generally avoid stimuli that to humans taste either bitter or very sour [1,5]. Congruent with these behavioral responses to taste stimuli, recordings from cat taste nerve fibers, and from units of the geniculate ganglion innervating taste cells, demonstrate responses to salty, sour, and bitter stimuli as well as to amino acids and nucleotides, but do not show neural responses to sucrose and several other sugars [5–12]. The sense of taste in the cat, in general, is therefore similar to that of other mammals, with the exception of an inability to taste sweet stimuli.
The molecular basis for this sweet blindness in cats is not known. Because the taste blindness appears specific to this single modality, we postulated that the defect in the cat (and likely in other obligate carnivores of Felidae) lies at the receptor step subtending the sweet-taste modality. The possible defects at the molecular level that might cause this sweet blindness could range from a single to a few amino acid substitutions, such as is found between sweet “taster” and “nontaster” strains of mice [13,14], to more radical mechanisms, such as an unexpressed pseudogene.
To distinguish among these possibilities, we identified the DNA sequence and examined the structures of the two known genes, Tas1r2 and Tas1r3, that in other mammals encode the sweet-taste receptor heteromer, T1R2/T1R3. We compared these with the sequence and structure of the same genes in dog, human, mouse, and rat—all species that display a functional sweet-taste modality. We also sought to detect the expression of the two cat genes at both the RNA and protein levels. Our results lead us to conclude that Tas1r3 is expressed in cat taste buds and very likely is functional, whereas cat Tas1r2 is an unexpressed pseudogene. The immediate repercussion of this unexpressed gene is that the heteromer normally acting as a sweet-taste receptor in most other mammals likely does not form in the cat.
Results
We identified DNA sequences of Tas1r3 and Tas1r2 from the domestic cat by screening a feline BAC library and using a PCR strategy on cat genomic DNA with degenerate primers. The feline sequences were compared with those of other species, and gene structures were determined. The expression of these two receptors was then evaluated by in situ hybridization and immunohistochemistry.
Molecular Cloning of Cat Tas1r3 and Tas1r2: Sequence and Gene Structure
By sequencing positive BAC clones retrieved from a feline genomic BAC library (Felis silvestris catus; BACPAC Resources, Oakland, California, United States), we obtained more than 3 kb of genomic sequences containing the open reading frame for cat Tas1r3, and approximately 10 kb of genomic sequences containing the open reading frame for cat Tas1r2. Because exons 1 and 2 of Tas1r2 were not found in the positive BAC clones, we employed a PCR strategy using degenerate primers to amplify these regions from cat genomic DNA (Novagen, San Diego, California, United States) (See Materials and Methods). We aligned the cDNA sequences and the deduced amino acid sequences from cat Tas1r3 and Tas1r2 with their dog, human, mouse, and rat orthologs (Figure 1). (We obtained the sequences of domestic dog genes, Tas1r3 and Tas1r2, by screening a dog genomic library using the same overgo probes and methods as for the feline genomic BAC library and by taking advantage of the limited data available at that time from the public dog genome database at http://www.ncbi.nlm.nih.gov/genome/guide/dog/).
Figure 1 Alignment of Deduced Amino Acid Sequences of T1R3 and T1R2 from Five Species
This figure shows the alignment of the deduced sequences of the taste receptor proteins, T1R3 and T1R2, from domestic cat, domestic dog, human, mouse, and rat. Amino acids that are identical among species are shaded in black; conservative amino acid substitutions are shaded in gray. The cat T1R3 sequence shows high similarity with that of human and rodents, with especially high similarity with that of dog. The predicted cat T1R2 sequence is truncated at amino acid 355 due to a premature stop codon at bp 57–59 in exon 4, which results from a 247-bp deletion in exon 3. The underlined amino acids from 316 to 355 of the cat T1R2 result from the frame shift brought by the 247-bp deletion in exon 3. Note that the deduced amino acid sequence of dog T1R2 predicts an apparently normal protein showing high similarity with that of rat, mouse, and human.
Table 1 presents the percent similarity of the Tas1r3 and Tas1r2 genes at both the cDNA and the protein levels between all possible pairs of five species: cat, dog, human, mouse, and rat. The cat Tas1r3 gene shows high similarity at the cDNA level with that of dog (87%), human (79%), rat (75%), and mouse (74%) (Table 1). The cat Tas1r3 gene predicts a protein of 865 amino acids (Figure 1) showing 85% similarity with deduced protein of dog, and 73%, 72%, and 72% with that of human, mouse, and rat, respectively (Table 1). Initially we predicted the exon–intron boundaries of cat Tas1r3 by comparison with the known boundaries of human TAS1R3. To confirm these exon–intron boundaries for cat Tas1r3, we performed both RT-PCR on cDNA from cat taste bud–containing circumvallate and fungiform papillae, and PCR on cat genomic DNA using intron-spanning primers. By comparing the cDNA sequence with the genomic sequence, we confirmed the boundaries predicted from human TAS1R3 (Figure 2A). Both the cat Tas1r3 and human TAS1R3 genes are composed of six similarly sized exons and five introns (Figure 2A).
Table 1 Similarity of Sweet Receptors between Species
T1R2 and T1R3 are the protein names, and Tas1r2 and Tas1r3 are the corresponding gene names. Columns for T1R2 and T1R3 show percent similarity between predicted amino acid sequences; columns for Tas1r2 and Tas1r3 show percent similarity between cDNA sequences.
Figure 2 Gene Structures of Cat Tas1r3, Human TAS1R3, and Cat Tas1r2, Human TAS1R2
The exons are shown in black (size in bp of each exon is in parentheses). Boundaries of gene sequences used to produce probes for in situ hybridization studies (Figure 3) are shown by the horizontal lines labeled “P1” and “P2” under the sketch of the cat Tas1r3 and cat Tas1r2. Boundaries of sequence used to generate peptide antigens for immunohistochemical studies (Figure 4) are shown by the horizontal lines labeled “A” under the sketch of the cat Tas1r3 and cat Tas1r2. The locations referred to in the vertical explanation text above the asterisks and the spade symbol indicate the position in bp within each exon. Intron sizes shown in the figure are not proportionally scaled on both (A) and (B) because of the large size of Tas1r2 introns. Under each human exon is the percent similarity between each human exon and its cat counterpart at the nucleotide level (Figure 2B). The exons for cat Tas1r2 refer to parts corresponding to human exons. The spade symbol (♠) indicates the position of microdeletion in exon 3 of cat Tas1r2. Asterisks (*) indicate the stop codon positions in exon 4 and 6 of cat Tas1r2. Note that nucleotide numbers of the exon 3 in human TAS1R2 and cat Tas1r2 are not identical.
Figure 3 RNA Expression of Cat Tas1r2 and Tas1r3 from Circumvallate Papillae
Digoxigenin-labeled sense and antisense cRNA probes corresponding to exons 3 and 6 of cat Tas1r2 and Tas1r3 were synthesized using DIG RNA labeling kit (Roche Applied Science, Indianapolis, Indiana, United States) (See Figure 2 for the locations of in situ probes, and Table 3 for identity of primers.) Hybridizations were carried as described [39]. Panel (A) shows result of antisense probes for Tas1r3, whereas panel (B) shows the result of the sense probes for Tas1r3. Panel (C) shows results of the antisense probes for Tas1r2 whereas panel (D) shows results of the sense probes. Scale bar, shown only in panel (A), = 60 μm for (A), (B), (C), and (D).
Figure 4 Protein Expression of Cat T1R2 and T1R3
Cat T1R3 expression is detected in taste buds of circumvallate papilla (CV) (A) and a fungiform papilla (Fun) just anterior to the intermolar eminence (B) by labeling with anti-mouse T1R3 antibody. Cat T1R2 expression is not detectable in either circumvallate (C) or fungiform (D) using an anti-cat T1R2 antibody. Control studies demonstrated that the anti-cat T1R2 antibody labeled a subset of taste bud cells in rat circumvallate (data not shown). Scale bar, shown only in panel (A) and (B), = 60 μm for (A) and = 45 μm for (B). Scale for panel (C) is the same as that of panel (A); scale for panel (D) is the same as that of panel (B).
We identified the exon–intron boundaries of cat Tas1r2 by comparison with known human boundaries (Figure 2B). Examining the sequence of cat Tas1r2, we discovered a microdeletion of 247 base pairs (bp) within exon 3. This deletion is responsible for a frame shift that results in a premature stop codon at bp 57–59 of exon 4 (Figure 2B). Assuming, for the moment, that a protein is translated from cat Tas1r2, then, because of the deletion and premature stop codon, the gene sequence predicts a peptide of 355 amino acids, the first 315 of which show high similarity with their rat, mouse, human, and dog counterparts (see Figure 1). Because of the frame shift introduced by the 247-bp deletion, the remaining deduced 40 amino acids show no similarity with their rat, mouse, human, or dog counterparts (underlined sequence of cat T1R2; Figure 1). The predicted similarity of this hypothetical 355–amino acid protein was compared with its truncated counterparts from dog, human, mouse, and rat. It ranges from 55% to 69% (Table 1). In contrast, the percent similarity of the full-length T1R2 protein within pairs of other species is between 91% (mouse–rat) and 69% (mouse–human).
By aligning cat Tas1r2 DNA sequences of exons 4, 5, and 6 with their human counterparts, we found four additional stop codons: one in exon 4 due to a deletion at bp 123, and three in exon 6 due to a substitution at bp 95 and a deletion at bp 247 (Figure 2B). The multiple stop codons indicate that the cat Tas1r2 is a pseudogene.
In an attempt to confirm the cat Tas1r2 exon–intron boundaries, we performed RT-PCR on cDNA from cat circumvallate and fungiform taste papillae. Despite using numerous (> 70) primers corresponding to deduced message from the Tas1r2 gene, we were unable to detect it.
RNA and Protein Expression
Having detected message from cat Tas1r3, but not from cat Tas1r2, by RT-PCR, we used the more tissue-specific approaches of in situ hybridization and immunohistochemistry to refine the search for cat Tas1r2 gene expression, using the cat Tas1r3 gene for comparison. Probes for in situ hybridization were constructed from the gene sequences corresponding to the lines marked “P” in Figures 2A and 2B. (See Materials and Methods for details.) Figure 3 shows that message from Tas1r3 is expressed in taste buds of cat circumvallate papillae whereas Tas1r2 expression is not detectable by in situ hybridization. Antisense probes for Tas1r3 result in positive labeling (Figure 3A); the arrows indicate three of the many labeled taste buds. The control sense probes show no labeling (Figure 3B). In contrast, antisense probes for cat Tas1r2 show no detectable labeling (Figure 3C) as is the case for the sense control (Figure 3D).
To detect the presence of taste receptor proteins from Tas1r2 and from Tas1r3, we exposed 10-μm sections of cat circumvallate and fungiform papillae to polyclonal antibodies developed against deduced amino acid peptide antigens marked by the line labeled “A” in Figure 2A and 2B. T1R3-like immunoreactivity was present in the taste buds of every circumvallate (10) and fungiform (4) papilla used in this study (Figure 4A and 4B) whereas immunoreactivity to T1R2 was not detected in these same tissues (Figure 4C and 4D). (Each circumvallate papilla of the cat contains approximately 400 taste buds, whereas the large fungiform papillae used in this study, located in the area of the eminence, contain from 1 to about 15 taste buds each.) The antibody to cat T1R2 did, however, label a subset of taste buds in rat circumvallate papillae (results not shown).
Confirmation of Tas1r2 Sequence in Six Individual Cats, Tiger, and Cheetah
Because the feline BAC genomic library was constructed from a single individual cat, we confirmed the sequence of Tas1r2 in six additional, unrelated, healthy adult domestic cats. Genomic DNA was obtained by cheek swabs from five of the six cats and through a blood sample from the remaining cat, amplified by PCR using primers that flanked the deletion and stop codons of the known cat Tas1r2, and sequenced. In addition, to test whether other species of Felidae display similar sequence anomalies in their Tas1r2 gene, we performed PCR on genomic DNA of one tiger (Therion International, Saratoga Springs, New York, United States) and one cheetah (a gift from the San Diego Zoo). We found that Tas1r2 in all six cats, the tiger, and the cheetah show the identical 247-bp deletion in exon 3, and all have stop codons at the same positions in exon 4 (Table 2). In exon 6, we found evidence for two alleles at position 93–95 in domestic cat, wherein two cats show the stop codon, TGA (homozygotes TGA/TGA); one cat shows TGR (heterozygote TGA/TGG); and three of the domestic cats, the one tiger, and the single cheetah show TGG (homozygotes TGG/TGG) (Table 2). The second exon 6 stop codon is also common to all three species (TGA for domestic cat, TAG for tiger and cheetah). Although the third stop codon of exon 6 at bp 697–699 was found in all six domestic cats, the corresponding region in tiger and cheetah could not be amplified by PCR.
Table 2
Tas1r2 Stop Codons in Species of Felidae
Stop codons are shown in bold.
aLocation (bp) refers to the position within each exon (Figure 2B).
bTwo cats are homozygotes TGA/TGA, one cat is heterozygote TGA/TGG, and three cats are homozygotes TGG/TGG.
UN, unknown, region could not be amplified by PCR.
Collectively, these data indicate that cat Tas1r3 is an expressed and likely functional receptor, whereas cat Tas1r2 is an unexpressed pseudogene.
Discussion
The taste receptors for sweetness and for umami (an amino acid–taste modality) are members of the T1R family of taste receptors [15,16,17]. These are Class C, family 3, G protein–coupled receptors (GPCR). The three known members of the T1R family are T1R1, T1R2, and T1R3 (for review, see [18]). In rodents and primates the primary sweet-taste receptor is composed of a dimer of two closely related GPCRs, T1R2 and T1R3 [14,15,16,17].
For this study, we made the working assumption that the Felidae T1R family shows specificity similar to that known from rodents and primates. Because the umami receptor is composed of the heteromer, T1R1/T1R3, and because cats can taste amino acids, it would appear likely that both of these proteins should be functional. The sweet-taste receptor is composed of the heteromer T1R2/T1R3. Because the cat cannot taste sweet stimuli, the most likely assumption is that the cat T1R2 is non-functional.
Molecular Features of Cat Tas1r3
By comparison with other known T1R3 proteins and other proteins of Class C, family 3, the sequence and gene structure of cat Tas1r3 predict a functional receptor of 865 amino acids (see Figure 1). Cat Tas1r3 is assumed to be located on cat Chromosome C1, syntenic with human 1p36, where human TAS1R3 is located [19,20]. As with other Tas1r3 genes, the cat Tas1r3 is composed of six exons, each approximately the same size as those of human (see Figure 2A). The sequence of cat Tas1r3 predicts a seven-transmembrane GPCR with extended N-terminal domain (first transmembrane region spanning amino acids 572–595), features common to other T1R3 receptors. Important Class C, family 3, structural motifs can also be located in cat T1R3 including the xPKxY motif at amino acids 814–818, and the FHSCCY motif at amino acids 517–522. Additionally, although most members of Class C, family 3, GPCRs show a highly conserved arginine residue at the extreme 3′ end of transmembrane segment 3, an exception is found with human, mouse, and rat T1R3, which substitute glutamic acid (E) for arginine (R) [21]. This substitution is also found in cat T1R3 at amino acid 660 (see Figure 1; the deduced dog T1R3 substitutes glutamine (Q) for arginine at the end of TM3).
Available evidence indicates that the products of cat Tas1r3 are expressed in taste buds. RT-PCR readily detected the message from Tas1r3 in lingual taste bud–containing tissues (results not shown). In situ hybridization studies confirmed the presence of message and localized it to taste buds (see Figure 3A). Polyclonal antibodies developed against T1R3 labeled taste buds in both cat circumvallate (Figure 4A) and fungiform papillae (Figure 4B). While only a few cells showed evidence of T1R3-like immunoreactivity, nearly every taste bud was labeled by both in situ hybridization and immunohistochemistry.
These commonalities in gene structure and sequence, together with evidence that the cat Tas1r3 gene is expressed, are consistent with the assumption that cat Tas1r3 codes for a functional receptor.
Molecular Features of Cat Tas1r2
Cat Tas1r2, on the other hand, while retaining structure similar with that of the human TAS1R2 gene (see Figure 2B), is an unexpressed pseudogene. The likely important molecular event that resulted in cat Tas1r2 becoming a pseudogene is the 247-bp deletion in exon 3. This deletion results in a frame shift that brings about a premature stop codon in exon 4 (Figure 2B). An additional stop codon can be found in exon 4, with three more in exon 6 (Figure 2B). This apparent accumulation of mutations suggests that there is no pressure from natural selection on the cat Tas1r2 gene. To determine if this gene is expressed, we performed studies to detect message from cat Tas1r2 by RT-PCR of taste bud–containing lingual papillae and by in situ hybridization. For RT-PCR, numerous (>70) primers were constructed based on sequences from exons 1–6. For in situ hybridization, probes were designed from exon 3 and from exon 6 (Figure 2B;
Table 3). Both techniques failed to detect message from cat Tas1r2 (see Figure 3C and 3D). Consistent with these attempts to detect message from cat Tas1r2, immunohistochemistry using an antibody developed from a deduced amino acid sequence spanning exons 2 and 3 revealed no labeling of taste buds in circumvallate or fungiform papillae (Figure 4C and 4D).
These results suggest that the cat Tas1r2 pseudogene is not transcribed, or if it is transcribed, it rapidly degrades, perhaps through a nonsense-mediated mRNA decay pathway [22].
Table 3 Primers for In Situ Probes
Tm, melting (annealing) temperature.
Tas1r2 in Felidae
The generality of the pseudogene nature of cat Tas1r2 was confirmed by sequencing the deletion and stop codon areas from six individual healthy adult cats. All showed the deletion and similar stop codons with some polymorphism (see Table 2). To assess the generality of the pseudogene nature of Tas1r2 in Felidae, we sequenced the stop codon areas and the area including the exon 3 microdeletion from genomic DNA of tiger and cheetah. These too displayed microdeletion and stop codons at the same location as the domestic cat. These observations, suggesting that in at least three species of Felidae Tas1r2 is not expressed, are consistent with behavioral evidence showing that, not only domestic cats, but also tigers and cheetahs do not prefer sweetened water over plain water [1].
According to morphological and molecular evidence, the available phylogeny of the order Carnivora consists of two groups, the Feliformia (cats, mongooses, civets, and hyenas) and the Caniformia (wolves, bears, raccoons, mustelids, and pinnipeds) [23,24]. It is difficult to determine when the alteration of Tas1r2 occurred and whether it preceded or followed the cat ancestor's change in diet to exclude plants. Clearly, because dogs have a human-like T1R2 structure (see Figure 1) and an avidity for sweet carbohydrates [25], the changes in the cat Tas1r2 must have occurred after the divergence of the Feliformia and the Caniformia.
Genes Affecting Taste Behavior
Taste receptors are shaped by and reflect a species' food choices. The genes encoding taste receptors often show a good deal of variation both among species and among individuals. These variations, both subtle and obvious, can have a variety of effects on taste sensitivity and preference behavior. A textbook example of this effect is the individual variation seen in sensitivity to the bitter compound, phenylthiocarbamide (PTC). A gene of the human TAS2R family of bitter taste receptors, TAS2R38, associated with this individual variation, shows three coding single-nucleotide polymorphisms giving rise to five haplotypes worldwide, accounting for the 55% to 85% of the variance in PTC sensitivity [26]. Further, in Drosophila, the behavioral and electrophysiological responses to trehalose are diminished in two mutants that carry deletions in the trehalose recognition gene, Gr5a [27]. In the mouse, variation in preference for sweet-tasting stimuli maps to the gene for T1R3, located within the Sac locus [28,29]. This gene is allelic in mice, and several reports identify a missense mutation (I60T) as being the most likely mutation accounting for the phenotypic differences [13,14,16,30–33]. However, the same alleles are not involved in strain-dependent sweet-taste preference in rats [34].
In addition to the modulation of behavior that can be caused by point mutations, more profound behavioral changes can result from the abolishment of gene function through, for example, the generation of pseudogenes. An example of this effect in mammalian chemoreception lies within the large repertoire of olfactory receptor genes. More than 60% of the human olfactory receptor genes are pseudogenes [35], whereas, only 20% are classified as such in mouse [35,36]. Strikingly, the accumulation of these olfactory pseudogenes in primates reportedly occurred concomitant with the acquisition of trichromatic color vision, perhaps reflecting the overarching behavioral changes that such an acquisition engendered [37]. Similar generation of bitter-taste receptor pseudogenes, accompanied by a large number of coding region single-nucleotide polymorphism, can account for the broad diversity displayed by the bitter-taste receptor family. This diversity may possibly play an important role in both species-specific and individually manifested taste preference [38].
In the extreme case, where a species fails to respond to stimuli representative of an entire modality, such as the cat with sweet taste, the development of a unique food preference behavior, based on the remaining taste receptors, might be anticipated. Because, with the exception of the sweetness modality, the taste system of the cat is organized much like that of most other mammals, discovering the molecular basis for the cats' lack of response to sweet-tasting compounds gives us a window on the development of strict carnivorous behavior in Felidae.
Conclusion
It is known that Felidae do not detect sweetness of carbohydrates yet can taste amino acids. Our results indicate that the gene encoding one member of the sweet-taste receptor heteromer is an unexpressed pseudogene. Given this observation, we suggest that the most parsimonious explanation for the inability of Felidae to respond to sweeteners is the lack of a functional T1R2 protein.
Materials and Methods
Animal tissue.
We obtained cat taste tissue from healthy young-adult animals euthanized for reasons unrelated to this study. Animals were cared for under protocols 033400 and 057600 approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania to Dr. Mark Haskins of the School of Veterinary Medicine, University of Pennsylvania.
Preparation of overgo probes.
Overgo probes are comprised of two 22mers with a complementary eight-base overlap. They can be designed by a computer program (http://genome.wustl.edu/tools/?overgo=1) and are readily synthesized. To identify cat Tas1r2 and Tas1r3, overgo probes were designed by aligning conserved coding regions of Tas1r2 and Tas1r3 sequences from different species. The single-stranded overhangs (14 bases) were filled in with 33P-labeled dATP and dCTP, and the overgo probes were used in hybridization procedures with the BAC libraries.
Screening a feline genomic BAC library.
Tas1r2 and Tas1r3 overgo probes were radioactively labeled by the random hexa-nucleotide method, and hybridization and washing of membranes were as described [29]. We identified 47 positive BAC clones for cat Tas1r2 and cat Tas1r3, and sequenced all of the positive BAC ends. By aligning BAC ends sequences with human syntenic regions (human TAS1R2 and TAS1R3 are located on chromosome 1p36), we picked BAC clones positive for cat Tas1r2 and Tas1r3 for shotgun library preparation.
Production of shotgun libraries for BACs containing cat Tas1r2 and Tas1r3.
We prepared BAC DNAs from positive clones by using a Qiagen Large Construct Kit (Valencia, California, United States). The BAC DNAs were digested using Sau3A I and the digested BAC DNA fragments subcloned into pGEM-3Z (Promega) vector. After transformants were arrayed to a nylon membrane, two separate hybridizations were performed by using pooled Tas1r2 and Tas1r3 overgo probes. By sequencing positive clones from the shotgun libraries and by using a chromosome walking strategy, we obtained the full coding region of the cat Tas1r3 and exon 3 to exon 6 of cat Tas1r2.
Identification of exon 1 and exon 2 of the cat Tas1r2 by PCR strategy.
Because exon 1 and exon 2 of the cat Tas1r2 were not present in the positive BACs selected above, we designed degenerate primers based on Tas1r2 alignments from different species (human, rodents, and dog) and performed PCR using cat genomic DNA as a template. The PCR products were sequenced. The feline BAC library was then re-screened using PCR products as probes, and new positive BAC clones were retrieved. Using a chromosome walking strategy, we obtained the complete sequence of exon 1 and exon 2 of cat Tas1r2 from these newly retrieved BAC clones.
RT-PCR.
To examine the RNA expression and to determine the intron–exon boundaries of the cat Tas1r2 and Tas1r3 genes, we extracted total RNA using TRIZOl Reagent (Life Technologies Inc., Rockville, Maryland, United States) from cat taste bud–containing tissues, followed by reverse transcription (Superscript reverse transcriptase, Life Technologies). The cDNA samples were amplified using AmpliTaq DNA Polymerase with GeneAmp (Perkin Elmer Corporation, Branchburg, New Jersey, United States) and intron-spanning primers selected to distinguish genomic and cDNA. Single bands of expected sizes were excised from the gel, purified, and sequenced.
In situ hybridization.
The probes corresponding to exons 3 and 6 of cat Tas1r2 and Tas1r3 were amplified by PCR using the primers described in Table 3. Digoxigenin-labeled cRNA probes were synthesized using a DIG RNA labeling kit (Roche). Taste bud–containing vallate tongue tissue was obtained as above. Fresh frozen sections (14 μm/section) of cat circumvallate papillae were attached to clean SuperFrost/Plus slides and prepared for in situ hybridization [39]. High-stringency hybridizations were carried out at 70 °C overnight in 50% formamide, 5X SSC, 5X Denhardt's, 250 μg/ml yeast RNA, and 500 μg/ml sperm DNA using the mixed cRNA probes. Sections were washed at 72 °C with 0.2X SSC three times. Signals were detected using alkaline phosphatase–conjugated antibodies to digoxigenin and standard chromogenic substrates and observed with a Nikon SA Microphot Microscope. Control hybridizations were performed with sense probes.
Immunohistochemistry.
Polyclonal anti-cat T1R2 rabbit antisera directed against an N-terminal peptide of cat T1R2 (exons 2 and 3; see Figure 1) were generated by Zymed Laboratories, Inc. (South San Francisco, California, United States). Generation of antisera directed against N-terminal peptide of mouse T1R3 has been described previously [33]. Lingual tissue blocks containing cat circumvallate and fungiform papillae were fixed in 4% paraformaldehyde for 2–6 h, then processed [40]. The antibodies were incubated with the sections (10 μm/section) for 60 h at 4 °C. After washing, the sections were incubated with the secondary antibody (Cy3-conjugated goat anti-rabbit IgG; The Jackson Laboratory, Bar Harbor, Maine, United States) and observed with a Leica TCS SP2 Spectral Confocal Microscope (Leica Microsystems Inc., Mannheim, Germany). Single-channel fluorescence images (average projection of 20–25, 0.3-μm optical sections) were processed with Adobe Photoshop software and overlaid on their respective difference interference contrast images.
Examination of stop codons in six individual cats and other species within Feliformia.
To confirm that Tas1r2 is a pseudogene in other cats, we obtained genomic DNA from cheek swabs or blood of six unrelated healthy adult cats. We sequenced the areas around the microdeletion and the stop codons by PCR using primers that flanked these areas of interest. In addition, to test whether other species of Felidae have a functional Tas1r2 gene, we performed PCR on genomic DNA of one tiger (Therion International, Saratoga Springs, New York, United States) and one cheetah (a gift from the San Diego Zoo) using the same primers above. All the PCR products are purified and sequenced.
Supporting Information
Accession Numbers
The GenBank (http://www.ncbi.nlm.nih.gov/Genbank/) accession numbers for the genomes discussed in this paper are cat Tas1r3 (AY819786), cat Tas1r2 (AY819787), dog Tas1r2 (AY916758), dog Tas1r3 (AY916759), human TAS1R3 (BK000152), and human TAS1R2 (NM_152232).
We thank Mr. Douglas L. Bayley, Ms. Kirsten J. Mascioli, Ms. Linda Wysocki, and Mr. Minliang Zhou for excellent technical assistance. We also thank Dr. Chenyan Wu for helping to design overgo probes and for many stimulating discussions. We gratefully acknowledge Dr. Taufiqul Huque for helping with early preliminary experiments, and Dr. Paul A.S. Breslin for suggesting an experiment and for critically reading the manuscript. We thank Lynn S. Hall and Mary Chatterton for providing the sample of blood or cheek swabs from cats. We acknowledge the laboratory of Dr. Mark Haskins of the School of Veterinary Medicine, University of Pennsylvania, for the procurement of animal tissue and the dedicated assistance of Patty O'Donnell and Karyn Cullen of that laboratory (NIH grants 02512 and DK25759 to Dr. Haskins).
This work was supported by a grant from The WALTHAM Centre for Pet Nutrition (to XL and JGB), and by National Institutes of Health grants R01DC00882 (GKB), R03DC05154 (LH), training grant T32DC00014 (to Monell Center, Dr. C. Wysocki, PI), by a grant from the United States Department of Veterans Affairs (JGB), and a grant from the National Science Foundation (DBJ-0216310). This project is funded, in part, under a grant with the Pennsylvania Department of Health. The Department specifically disclaims responsibility for any analyses, interpretations, or conclusions.
The funding agencies had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests. HW, JC, KM, and LH declare that they have no competing interests of a financial, professional, or personal nature. VLD is an employee of the Masterfoods division of Mars. GKB is on an advisory board to the WALTHAM Centre. Patents describing the uses of the feline receptors have been filed, and name as inventors: XL, WL, JGB, DRR, and AAB.
Author contributions. XL, LH, DRR, VLD, GKB, and JGB conceived and designed the experiments. XL, WL, HW, JC, KM, and JGB performed the experiments. XL, WL, JC, AAB, and DRR analyzed the data. XL, HW, LH, AAB, and JGB contributed reagents/materials/analysis tools. XL, AAB, DRR, GKB, and JGB wrote the paper.
Abbreviations
bpbase pair
GPCRG protein–coupled receptors
==== Refs
References
Beauchamp GK Maller O Rogers JG 1977 Flavor preferences in cats (Felis catus and Panthera sp.) J Comp Physiol Psychol 91 1118 1127
Carpenter JA 1956 Species differences in taste preferences J Comp Physiol Psychol 49 139 144 13319525
Bartoshuk LM Jacobs HL Nichols TL Hoff LA Ryckman JJ 1975 Taste rejection of nonnutritive sweeteners in cats J Comp Physiol Psychol 89 971 975 1184803
Bradshaw JW 1991 Sensory and experiential factors in the design of foods for domestic dogs and cats Proc Nutr Soc 50 99 106 1881936
White TD Boudreau JC 1975 Taste preferences of the cat for neurophysiologically active compounds Physiol Psychol 3 405 410
Boudreau JC Bradley BE Bierer PR Kruger S Tsuchitani C 1971 Single unit recordings from the geniculate ganglion of the facial nerve of the cat Exp Brain Res 13 461 488 5137297
Boudreau J Alev N 1973 Classification of chemoresponsive tongue units of the cat geniculated ganglion Brain Res 17 157 175
Boudreau JC 1977 Chemical stimulus determinants of cat neural taste responses to meats J Am Oil Chem Soc 54 464 466 915172
Dinger B Fidone SJ Stensaas FJ 1984 Gustatory trophic action of arterial chemosensory neurones in the cat J Physiol 356 49 64 6084060
Robinson PP 1988 The characteristics and regional distribution of afferent fibres in the chorda tympani of the cat J Physiol 406 345 357 3254415
Boudreau JC White TD 1978 Flavor chemistry of carnivore taste system Bullard RW Flavor chemistry of animal foods: A symposium sponsored by the Division of Agricultural and Food Chemistry at the 174th meeting of the American Chemical Society, Chicago, Ill, August 29, 1977 Washington, DC American Chemical Society 102 128
Beidler LM Fishman IY Hardiman CW 1955 Species differences in taste responses Am J Physiol 181 235 239 14376602
Bachmanov AA Li X Reed DR Ohmen JD Li S 2001 Positional cloning of the mouse saccharin preference (Sac) locus Chem Senses 26 925 933 11555487
Max M Shanker YG Huang L Rong M Liu Z 2001 Tas1r3, encoding a new candidate taste receptor, is allelic to the sweet responsiveness locus Sac Nat Genet 28 58 63 11326277
Nelson G Chandrashekar J Hoon MA Feng L Zhao G 2002 An amino-acid taste receptor Nature 416 199 202 11894099
Nelson G Hoon MA Chandrashekar J Zhang Y Ryba NJ 2001 Mammalian sweet taste receptors Cell 106 381 390 11509186
Li X Staszewski L Xu H Durick K Zoller M 2002 Human receptors for sweet and umami taste Proc Natl Acad Sci U S A 99 4692 4696 11917125
Montmayeur JP Matsunami H 2002 Receptors for bitter and sweet taste Curr Opin Neurobiol 12 366 371 12139982
Liao J Schultz PG 2003 Three sweet receptor genes are clustered in human chromosome 1 Mamm Genome 14 291 301 12856281
Murphy WJ Sun S Chen Z Yuhki N Hirschmann D 2000 A radiation hybrid map of the cat genome: Implications for comparative mapping Genome Res 10 691 702 10810092
Pin JP Galvez T Prezeau L 2003 Evolution, structure, and activation mechanism of family 3/C G-protein-coupled receptors Pharmacol Ther 98 325 354 12782243
Rajavel KS Neufeld EF 2001 Nonsense-mediated decay of human HEXA mRNA Mol Cell Biol 21 5512 5519 11463833
Flynn JJ Nedbal MA 1998 Phylogeny of the Carnivora (Mammalia): congruence vs incompatibility among multiple data sets Mol Phylogenet Evol 9 414 426 9667990
Mattern MY McLennan DA 2000 Phylogeny and speciation of Felids Cladistics 16 232 253
Grace J Russek M 1968 The influence of previous experience on the taste behavior of dogs toward sucrose and saccharin Physiol Behav 4 553 558
Kim U Jorgenson E Coon H Leppert M Risch N 2003 Positional cloning of the human quantitative trait locus underlying taste sensitivity to phenylthiocarbamide Science 299 1221 1225 12595690
Dahanukar A Foster K van der Goes van Naters WM Carlson JR 2001 A Gr receptor is required for response to the sugar trehalose in taste neurons of Drosophila Nat Neurosci 4 1182 1186 11704765
Li X Inoue M Reed DR Huque T Puchalski RB 2001 High-resolution genetic mapping of the saccharin preference locus (Sac) and the putative sweet taste receptor (T1R1) gene (Gpr70) to mouse distal Chromosome 4 Mamm Genome 12 13 16 11178737
Li X Bachmanov AA Li S Chen Z Tordoff MG 2002 Genetic, physical, and comparative map of the subtelomeric region of mouse Chromosome 4 Mamm Genome 13 5 19 11773963
Kitagawa M Kusakabe Y Miura H Ninomiya Y Hino A 2001 Molecular genetic identification of a candidate receptor gene for sweet taste Biochem Biophys Res Commun 283 236 242 11322794
Montmayeur JP Liberles SD Matsunami H Buck LB 2001 A candidate taste receptor gene near a sweet taste locus Nat Neurosci 4 492 498 11319557
Sainz E Korley JN Battey JF Sullivan SL 2001 Identification of a novel member of the T1R family of putative taste receptors J Neurochem 77 896 903 11331418
Reed DR Li S Li X Huang L Tordoff MG 2004 Polymorphisms in the taste receptor gene (Tas1r3) region are associated with saccharin preference in 30 mouse strains J Neurosci 24 938 946 14749438
Lu K McDaniel A Tordoff M Li X Beauchamp G 2005 No relationship between sequence variation in protein coding regions of the Tas1r3 gene and saccharin preference in rats Chem Senses 30 231 240 15741599
Gilad Y Man O Paabo S Lancet D 2003 Human specific loss of olfactory receptor genes Proc Natl Acad Sci U S A 100 3324 3327 12612342
Young JM Friedman C Williams EM Ross JA Tonnes-Priddy L 2002 Different evolutionary processes shaped the mouse and human olfactory receptor gene families Hum Mol Genet 11 535 546 11875048
Gilad Y Wiebe V Przeworski M Lancet D Paabo S 2004 Loss of olfactory receptor genes coincides with the acquisition of full trichromatic vision in primates PLoS Biol 2 E5 14737185
Parry CM Erkner A le Coutre J 2004 Divergence of T2R chemosensory receptor families in humans, bonobos, and chimpanzees Proc Natl Acad Sci U S A 101 14830 14834 15466715
Schaeren-Wiemers N Gerfin-Moser A 1993 A single protocol to detect transcripts of various types and expression levels in neural tissue and cultured cells: In situ hybridization using digoxigenin-labelled cRNA probes Histochemistry 431–440
Grosvenor W Kaulin Y Spielman AI Bayley DL Kalinoski DL 2004 Biochemical enrichment and biophysical characterization of a taste receptor for L-arginine from the catfish, Ictalurus puntatus BMC Neurosci 5 25 15282034
|
16103917
|
PMC1183522
|
CC BY
|
2021-01-05 08:00:24
|
no
|
PLoS Genet. 2005 Jul 25; 1(1):e3
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010003
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 10.1371/journal.pgen.001000405-PLGE-RA-0005R2plge-01-01-02Research ArticleCell BiologyNeuroscienceOphthalmologyGenetics/Disease ModelsGenetics/Complex TraitsEukaryotesAnimalsVertebratesMammalsMus (Mouse)Susceptibility to Neurodegeneration in a Glaucoma Is Modified by Bax Gene Dosage Bax in Glaucoma Susceptibility Libby Richard T 1Li Yan 2Savinova Olga V 13Barter Joseph 1Smith Richard S 13Nickells Robert W 3John Simon W.M 134*1 The Jackson Laboratory, Bar Harbor, Maine, United States of America
2 Department of Ophthalmology and Visual Sciences, University of Wisconsin, Madison, Wisconsin, United States of America
3 The Howard Hughes Medical Institute, Bar Harbor, Maine, United States of America
4 Department of Ophthalmology, Tufts University School of Medicine, Boston, Massachusetts, United States of America
Valle David EditorJohns Hopkins Institute, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 25 7 2005 1 1 e416 1 2005 6 4 2005 Copyright: © 2005 Libby et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.In glaucoma, harmful intraocular pressure often contributes to retinal ganglion cell death. It is not clear, however, if intraocular pressure directly insults the retinal ganglion cell axon, the soma, or both. The pathways that mediate pressure-induced retinal ganglion cell death are poorly defined, and no molecules are known to be required. DBA/2J mice deficient in the proapoptotic molecule BCL2-associated X protein (BAX) were used to investigate the roles of BAX-mediated cell death pathways in glaucoma. Both Bax
+/− and Bax
−/− mice were protected from retinal ganglion cell death. In contrast, axonal degeneration was not prevented in either Bax
+/− or Bax
−/− mice. While BAX deficiency did not prevent axonal degeneration, it did slow axonal loss. Additionally, we compared the effects of BAX deficiency on the glaucoma to its effects on retinal ganglion cell death due to two insults that are proposed to participate in glaucoma. As in the glaucoma, BAX deficiency protected retinal ganglion cells after axon injury by optic nerve crush. However, it did not protect retinal ganglion cells from N-methyl-D-aspartate (NMDA)-induced excitotoxicity. BAX is required for retinal ganglion cell death in an inherited glaucoma; however, it is not required for retinal ganglion cell axon degeneration. This indicates that distinct somal and axonal degeneration pathways are active in this glaucoma. Finally, our data support a role for optic nerve injury but not for NMDA receptor-mediated excitotoxicity in this glaucoma. These findings indicate a need to understand axon-specific degeneration pathways in glaucoma, and they suggest that distinct somal and axonal degeneration pathways may need to be targeted to save vision.
Synopsis
Glaucoma is a group of diseases whose unifying characteristic is death of nerve cells (retinal ganglion cells) that connect the eye to the brain. Glaucoma is often associated with a harmfully high pressure inside the eye (intraocular pressure) contributing to nerve cell death. Various treatments are used to lower eye pressure, but currently no commonly used treatments directly protect the nerve cells. DBA/2J mice develop elevated eye pressure with age, and this pressure kills retinal nerve cells. The authors use this mouse model to investigate how these nerve cells die in glaucoma. They show that there are distinct degeneration pathways activated in different parts of the retinal nerve cells. They found that the biochemical pathway in the nerve cell body, which resides in the retina, requires a molecule called BAX (BCL2-associated X protein). In contrast, pathways in the part of the cell (axon) that connects the cell body to the brain do not require BAX. Because degeneration pathways in the cell body and of the axon also may be molecularly different in human glaucoma, it will be important to consider them all when designing therapies. Their data also suggest that the BAX gene is a candidate to modulate glaucoma susceptibility.
Citation:Libby RT, Li Y, Savinova OV, Barter J, Smith RS, et al. (2005) Susceptibility to neurodegeneration in a glaucoma is modified by Bax gene dosage. PLoS Genet 1(1): e4.
==== Body
Introduction
Glaucoma is a common blinding disease affecting approximately 70 million people worldwide [1]. Glaucoma is often associated with elevated intraocular pressure (IOP). IOP elevation and glaucoma are typically spontaneous, progressive, idiopathic processes and are most common in the elderly [2]. Although IOP-lowering treatments slow the development and progression of glaucoma in many patients [3,4], it is not always possible to reduce IOP to a “safe” level [5]. Vision loss in glaucoma is the result of retinal ganglion cell (RGC) death with accompanying optic nerve atrophy, so glaucoma is a neuropathy. IOP elevation is not detected in a significant subset of glaucomas [6,7]. Thus, the unifying characteristic of glaucoma is RGC death. While there are several hypotheses as to why elevated IOP kills RGCs, both the precise biochemical cascades that are triggered within RGCs and the nature of the proximal insult(s) that trigger these cascades remain superficially defined [8]. No treatments that directly protect the neurons are in routine clinical use.
The complex nature of glaucoma makes studies of its pathogenesis difficult [9]. Consequently, no specific molecules have been shown to be essential for RGC death in glaucoma. Standard glaucoma-relevant models include direct RGC trauma, direct optic nerve trauma, and suddenly induced IOP elevation [10–18]. Although these induced models have provided valuable information, the relevance of specific damaging mechanisms may differ significantly between spontaneous and experimentally induced glaucomas. Thus, studies using inherited glaucoma models are also necessary.
Apoptosis is known to contribute to RGC death following experimentally induced insults including axotomy and IOP elevation (e.g., [19,20]), and there is also some evidence that apoptosis is involved in human glaucoma [21,22]. A number of molecules that are known to affect apoptosis are reported to be important regulators of RGC death after various induced insults. These include X-linked inhibitor of apoptosis protein (XIAP) [23–25], p38 [26], several caspases [27–30], the B-cell lymphoma/leukemia 2 (BCL2) family of apoptotic regulators [20,31–34], and members of the c-Jun N-terminal kinase (JNK) [35,36] and tumor necrosis factor (TNF) [35,37] signaling pathways. One of these molecules, BCL2-associated X protein (BAX; a proapoptotic member of the BCL2 family), has a major role in mitochondrial-mediated apoptosis in different neuronal cell types [38,39]. In mice, BAX deficiency increases the number of RGCs in the adult retina by 220% by allowing more RGCs to survive during development [39]. Genetic or induced BAX deficiency is also known to prevent RGC apoptosis after optic nerve crush and axotomy [13,18,40]. Thus, BAX-mediated apoptosis is clearly an important mechanism of stress-induced RGC death. Whether or not this pathway has a role in IOP-induced RGC death in either experimentally induced or inherited glaucomas is not known.
Understanding the pathophysiologic mechanisms of RGC death in glaucoma and the genetic susceptibility factors contributing to this process is important for the development of effective and individualized treatments. Here, we use the genetically uniform DBA/2J mouse model of glaucoma [41–43] to assess the importance of mitochondrially mediated apoptosis in an inherited glaucoma. Importantly, we show that in this model of inherited glaucoma there are distinct RGC death and axonal degeneration pathways. The RGC death pathway is BAX dependent and, therefore, apoptotic. The axonal degeneration pathway is BAX independent. Finally, our data suggest that reducing BAX levels in the retina may retard the rate of vision loss in glaucoma.
Results
Apoptosis Is Physiologically Relevant for RGC Death in an Inherited Glaucoma
To determine if RGC apoptosis has a significant role in an inherited glaucoma, we assessed DNA fragmentation, chromatin condensation, and cellular ultrastructure in glaucomatous DBA/2J retinas. We identified hallmarks of apoptosis including the presence of TUNEL-positive cells that peaked between 10 and 13 mo of age (the period when the majority of RGC death occurs in this model) (Figure 1). These results confirm earlier suggestive studies that apoptotic pathways are important mediators of RGC death in spontaneous glaucoma [19–22].
Figure 1 Dying RGCs Have Characteristic Features of Apoptosis
(A–C) A double-labeling assay that identifies fragmented DNA using fluorescently labeled dUTP (A) and detects chromatin condensation by binding of the dye YOYO-1 (B) was used to assess the presence of these hallmarks of apoptosis in glaucomatous DBA/2J eyes at 10–11 mo of age (a time when many RGCs die). A cell in the retinal ganglion cell layer (GCL, arrowhead) has both of these features of apoptosis as indicated by double labeling (C). INL, inner nuclear layer.
(D–F) Electron microscopy provided further evidence for apoptosis. (D) An example of a healthy RGC. (E) Chromatin condensation (a hallmark of apoptosis) along the inner surface of the nuclear envelope in a ganglion cell (arrows). The internal limiting membrane of the retina is indicated by arrowheads. (F) An apoptotic body in the ganglion cell layer (arrows) containing a nuclear fragment with prominent condensed chromatin (asterisk) and other cell remnants.
(G) A TUNEL assay (see Materials and Methods) was used to assess the prevalence of cell death at different ages. TUNEL labeling was not detected at 7 mo (an age prior to glaucomatous cell death) and peaked at 10–13 mo, when most RGCs die. No TUNEL-positive cells were detected in nonglaucomatous, age-matched control mice. These results support an important role of apoptosis in RGC death in spontaneous glaucoma. Scale bar, 1 μm.
Homozygous but Not Heterozygous BAX Deficiency Alters RGC Number
To test the role of BAX in glaucomatous RGC death, we extensively backcrossed a previously characterized null allele of Bax
(Baxtm1Sjk) [44] onto the inbred DBA/2J background. In mammals, approximately twice as many RGCs are produced during retinal development than survive into adulthood [45–47]. As expected from previous studies of retinal development on a different genetic background [39], complete BAX deficiency increased the number of RGC-layer somata in adult DBA/2J mice by 220% (average cell number per 40× field ± standard error of the mean [SEM], number of retinas analyzed: Bax
+/+, 199 ± 5.6, n = 7; Bax
−/−, 437 ± 15.3, n = 8). In agreement with this, Bax
−/− mice had 217% more RGC axons than Bax
+/+ mice (Bax
+/+, 50,504 ± 1,988, n = 8; Bax
−/−, 108,907 ± 10,322, n = 4; p < 0.001). Reflecting the increased number of RGC axons and the proportional increase in glial cell types [48], the cross sectional area of Bax
−/− optic nerves was significantly increased (average ± SEM, number of optic nerves measured: Bax
+/+, 0.157 ± 0.005 mm2, n = 13; Bax
−/−, 0.278 ± 0.008 mm2, n = 16; p < 0.001). In heterozygous Bax
+/− mice, RGC number (average per 40× field ± SEM, 212 ± 14.0, n =5) and the optic nerve area (0.171 ± 0.007 mm2, n = 13) was not different from Bax
+/+ mice (p = 0.352 and p = 0.107, respectively). Thus, heterozygous levels of BAX are sufficient for death of the normal numbers of RGCs during retinal development.
BAX Ablation Preserves RGC Numbers but Does Not Prevent RGC Axonal Degeneration in Glaucoma
To determine the role of BAX in glaucomatous RGC death, we assessed the effects of BAX deficiency on RGC survival and on RGC axonal degeneration (see Materials and Methods). Our results show that BAX is not required for RGC axon degeneration. Bax
−/− mice developed severe optic nerve damage, including essentially complete loss of axons (Figure 2). In contrast, our experiments show that BAX is required for RGC death in glaucoma (Figure 3). Despite severe axonal degeneration, the numbers of cell bodies in the RGC layer of Bax
−/− mice were normal. As a stringent test of this observation, we counted RGC-layer cell bodies in the retinas of mice with severe (≥ 95% loss) axon degeneration. The number of RGC cell bodies was normal in Bax
−/− mice with more than 95% axon loss (Figure 3). Importantly, Bax
+/− mice were also protected against glaucomatous RGC death. Bax
+/− mice with an axon loss of 95% or more also had substantially increased survival of RGC cell bodies as compared to Bax
+/+ controls (Figure 3E). Thus, RGC death and axonal degeneration are clearly distinguished in these experiments.
Figure 2 BAX Is Not Required for Glaucomatous Optic Nerve Degeneration
To assess the effects of Bax deficiency on optic nerve degeneration, we analyzed PPD-stained optic nerve cross sections from Bax
+/+ and Bax
−/− mice (n > 49 for each genotype; see Materials and Methods).
(A and B) Before the DBA/2J glaucoma damages RGCs, the optic nerves of both Bax
+/+ (A) and Bax
−/− mice (B) had a normal organization. The axons appeared healthy with a clear axoplasm and darkly stained myelin sheath.
(C and D) BAX deficiency did not prevent glaucomatous optic nerve damage. Severe degeneration involving extensive to complete axon loss and scarring occurred in both Bax
+/+ (C) and Bax
−/− mice (D). The majority of mice of both genotypes had this severe degree of damage by 12 mo of age. These experiments show that BAX is not required for glaucomatous axon degeneration. Scale bar, 50 μm.
Figure 3
Bax Deficiency Prevents Glaucomatous RGC Death
To determine the effects of BAX deficiency on RGC death in glaucoma, we analyzed RGC layer cells at stages with and without glaucomatous optic nerve damage (see Materials and Methods). All shown images are from a similar region of the superior, peripheral retina.
(A and B) In both Bax+/+ (A) and Bax
−/− (B) mice without glaucomatous optic nerve damage, the retinas appear healthy. The retinas of both genotypes are similar except that Bax
−/− mice have extra RGCs (since BAX is important in normal developmental RGC death [39]).
(C and D) In contrast, an obvious difference was evident between the retinas of Bax
+/+ and Bax
−/− mice that had all suffered severe glaucomatous damage with 95% or more axon degeneration. As expected, for Bax
+/+ retinas (C) from eyes with 95% or more optic nerve axon loss, there was a noticeable decrease in RGC layer cells (compare [C] to [A]). In contrast, retinas from Bax
−/− mice with correspondingly damaged optic nerves (D) had suffered no obvious loss of RGC layer cells (compare [D] to [B]). This suggests that BAX is required for RGC death in DBA/2J glaucoma. As is well established for both RGCs and other neurons, Bax
−/− RGCs that survive without axons have a shrunken morphology [38,82]. This is clearly evident in the Bax
−/− glaucomatous mice (D).
(E) RGC layer cell counts for eyes with 95% or more axon degeneration confirmed that BAX is necessary for RGC death in this glaucoma. To allow comparison between genotypes, the percent of surviving cells is shown (% soma in mice with 95% or more axon loss compared to mice of the same genotype without glaucomatous damage). At 12 mo of age, Bax
+/+ mice had 61.4% ± 3.8% of their RGC layer cells remaining, Bax
−/− mice had no appreciable cell loss (101% ± 5.3%). The RGC layer cells of Bax
+/− mice were also protected (89.2% ± 5.7%). The p values comparing differences in cell counts between nonglaucomatous and very severely glaucomatous (≥ 95% axon loss) eyes of the same genotype were: Bax
+/+, p < 0.001; Bax
+/−, p = 0.207; Bax
−/−, p = 0.426. No cell loss was seen in Bax
−/− mice even out to 18 mo (94.8% ± 4.4% cells surviving, p = 0.524 compared to nonglaucomatous Bax
−/− mice). These findings show that BAX gene dosage has an important effect on the susceptibility of RGCs to glaucomatous death. Scale bar, 50 μm.
Other Proapoptotic Molecules Do Not Compensate for BAX Deficiency
In some neuronal cell types, BAX deficiency delays but does not prevent apoptosis [49]. This is because other proapoptotic molecules (e.g., another BCL2 family member, BAK) mediate cell death in the BAX-deficient neurons [50]. To test this possibility in the DBA/2J model, we aged Bax
−/− mice to18 mo. As expected in a complex age-related disease, the severity of glaucomatous damage varies between individual DBA/2J eyes at any age. Nevertheless, by 12 mo of age, the majority of eyes have severe optic nerve damage (see below). Therefore, 18 mo of age is 6 mo after the majority of eyes have severe axon loss. Despite this extensive axonal degeneration, there was no obvious reduction in RGC numbers in any of the 18-mo-old Bax
−/− eyes (Figure 3E). This result indicates that other molecules do not substitute for BAX and that BAX is essential for RGC apoptosis in DBA/2J inherited glaucoma.
Homozygous BAX Deficiency Alters IOP
DBA/2J mice develop a form of pigmentary glaucoma that is secondary to a progressive iris disease. Iris pigment and cell debris enter the ocular drainage structures, resulting in subsequent IOP elevation [41,42]. The increase in IOP induces RGC death. Manipulations that alleviate the iris disease and prevent IOP elevation also prevent RGC death in this strain [41]. To assess the effects of BAX deficiency, the clinical phenotypes and IOP profiles of mice of each Bax genotype were carefully examined at multiple ages.
Periodic assessment of the progression of iris abnormalities by slit-lamp examinations (approximately every 2 mo between 3 and 12 mo of age) revealed no differences between mice of each Bax genotype. Histologic analysis confirmed these observations (unpublished data). Thus, BAX-mediated processes are not necessary for the progression of the iris disease.
Although iris damage was similar in mice of all three genotypes, Bax genotype did have an effect on IOP. The peak period of IOP elevation in DBA/2J mice is from 9 to 12 mo of age, with the IOP distribution clearly shifting upward between 8 and 9 mo. We monitored IOP at key ages (9, 10.5, and 12 mo). Surprisingly, Bax
−/− mice tended to have lower IOP than either Bax
+/− or Bax
+/+ mice at 9 mo, and the difference was statistically significant at 10.5 mo of age (Figure 4). In contrast, the IOPs of Bax
+/− and Bax
+/+ mice were not different. Because of the lower IOP in Bax
−/− mice, we analyzed IOP in 4-mo-old mice of each genotype and found no differences (average ± SEM, number of eyes examined: Bax
+/+, 13.54 ± 0.45 mm Hg, n = 22; Bax
+/−, 13.77 ± 0.33, n = 22; Bax
−/− 13.46 ± 0.34, n = 20; p > 0.5). This result indicates that BAX deficiency does not alter baseline IOP but does have an effect as the IOP increases to glaucomatous levels in older mice.
Figure 4 Complete Bax Deficiency Has a Protective Effect against IOP Elevation
To assess the possibility that Bax deficiency may delay axon degeneration by lessening the glaucomatous insult to which RGCs are exposed, we analyzed IOP at key ages of IOP elevation. (A) The average IOP for each genotype (± SEM) and (B) the actual IOP values recorded. Bax deficiency did not prevent IOP elevation. At both 9 mo and 10.5 mo, the average IOP of both Bax
+/− and Bax
−/− mice was significantly elevated compared to preglaucomatous DBA/2J mice (p < 0.001). The width of the horizontal line (black in [A], gray in [B]) represents the mean IOP ± SEM of a group of wild-type preglaucomatous DBA/2J mice that were 3 mo old. The degree of IOP elevation, however, was altered in Bax
−/− mice. In both 9- and 10.5-mo-old Bax
−/− mice, the average IOP was less than that of Bax
+/+ and Bax
+/− mice. This reduction in IOP elevation was significant at 10.5 mo (p < 0.01). The IOP of Bax
+/− mice did not differ from wild type at either 9 or 10.5 mo (p > 0.82). By 12 mo, there was no difference in IOP between mice of any genotype (p > 0.25).
The lower IOP insult in Bax
−/− mice does not account for the survival of their RGCs. This conclusion is supported by the normal RGC numbers remaining in Bax
+/− mice with indistinguishable IOP from Bax
+/+ mice. Previous studies have shown that BAX deficiency allows RGC survival following axotomy or optic nerve crush [13]. By contrast, even when neuroprotective treatments are administered, only a small number RGCs survive in the short term (4–6 wk) in Bax
+/+ mice exposed to severe axon trauma [51,52]. Thus, there is no reasonable explanation for the finding of prolonged survival of RGCs that have no axons other than that BAX is a necessary RGC-intrinsic molecule for apoptosis in this glaucoma model.
Bax Deficiency Delays Axon Degeneration
Although we have shown that axon degeneration is not dependent upon BAX, our results clearly identify BAX as an endogenous susceptibility factor for both RGC death and axonal degeneration in DBA/2J glaucoma. As discussed above, complete or partial BAX deficiency had a profound rescuing effect on RGC cell bodies. Importantly, decreasing functional Bax gene dosage also decreased susceptibility to glaucoma by delaying the progression of axon damage (Figure 5). At 10.5 mo of age, the majority of Bax
+/+ mice had moderate or severe optic nerve damage (see Materials and Methods), with only 20% being mildly affected. In contrast, 53% of Bax
−/− and 44% of Bax
+/− mice were only mildly affected at 10.5 mo of age. At 12.0 mo of age, the distribution of optic nerve damage was indistinguishable among mice of the three Bax genotypes (Figure 5). Since mice of each Bax genotype were littermates that were housed in the same cages throughout aging, these results provide compelling evidence that decreasing BAX levels delays optic nerve damage.
Figure 5 Glaucomatous Optic Nerve Degeneration Is Delayed in Bax-Deficient Mice
BAX deficiency did not prevent glaucomatous axon degeneration. Nevertheless, the beneficial effect on RGC survival raised the possibility that it may have a protective effect on the axon and delay optic nerve damage. To assess this, three investigators masked to genotype determined the severity of optic nerve damage for mice of each Bax genotype at each age (n = 49–71 per genotype at each age; see Materials and Methods). At 10.5 mo, both Bax
+/− and Bax
−/− mice had significantly less optic nerve damage than Bax+/+ mice. Mild damage was evident in 53% of Bax
−/− and 44% of Bax
+/− optic nerves, compared to only 20% of Bax
+/+ (p < 0.001 for both Bax
+/− and Bax
−/− compared to Bax
+/+; Chi2 test). With disease progression to 12 mo of age, the distribution of optic nerve damage became indistinguishable among mice of different Bax genotypes (p > 0.10). Quantitative assessment of a random subset of nerves assigned each damage level (more than eight of each) demonstrates that the number of axons that remain in optic nerves having each damage level are clearly different (see Materials and Methods).
The delay of optic nerve damage in Bax
+/− mice (note: Bax
+/− mice had similar IOP insults to Bax
+/+ mice) suggests that partially decreasing BAX levels in RGCs protects RGC axons. However, since complete BAX deficiency limited IOP elevation, a further protective effect of BAX deficiency by lowering IOP is also possible and may explain the trend toward greater axonal protection in Bax
−/− mice. Thus, it is possible that either low-expressing or low-activity alleles of BAX may affect glaucoma susceptibility both by limiting and/or delaying IOP elevation and by directly protecting RGCs from damaging effects of harmfully high IOP.
An Integrated Approach Supports a Role of Direct Optic Nerve Injury in Glaucoma
Comparing the specific pathways active in glaucomatous RGC death to the pathways induced by acute, experimental manipulations can provide information about the initial insult(s) to RGCs in glaucoma. N-methyl-D-aspartate (NMDA) receptor-mediated excitotoxic injury and direct axon injury are two insults that have been proposed to kill RGCs in glaucoma. Acute experimental procedures can be used to mimic these insults. Intraocular NMDA injection is used to mimic excitotoxic RGC insult, and controlled optic nerve crush is used to mimic a direct axon insult [53,54]. To assess the likely roles of these insults in a spontaneous glaucoma, we subjected preglaucomatous DBA/2J mice of differing Bax genotypes to these procedures. This allowed direct comparison of RGC death induced by these distinct excitotoxic and axonal insults to the naturally progressing glaucoma (Figure 6) in a single genetic context. Bax genotype had absolutely no effect on RGC death initiated by intraocular injection of the excitotoxin NMDA. In contrast, the RGCs of both Bax
+/− and Bax
−/− mice were profoundly protected against optic nerve crush. Since RGC death in the DBA/2J glaucoma is also BAX dependent, these data support a role for axon injury, but not for excitotoxicity (at least through the NMDA receptor) in this glaucomatous RGC death.
Figure 6 Mechanical Axon Insult, but Not Excitotoxicity, Induces BAX-Dependent RGC Death
To help distinguish between the likely roles of mechanical axon insult and excitotoxicity in cell death induction in spontaneous glaucoma, we subjected preglaucomatous DBA/2J mice of each Bax genotype to either controlled optic nerve crush or NMDA-mediated excitotoxicity. For controlled crush and NMDA, the percent RGC survival in the manipulated eye compared to the contralateral control eye is shown. For ease of comparison, the data for glaucomatous damage are the same as shown in Figure 3. In contrast to the spontaneous glaucoma, NMDA-mediated RGC death is not dependent on BAX, as evident by the complete lack of protection from death in Bax
−/− mice. As for the spontaneous glaucoma, RGC death induced by controlled optic nerve crush was completely dependent on BAX and prevented in both Bax
+/− and Bax
−/− mice. Overall, the effects of BAX in the face of spontaneous glaucoma and controlled crush were remarkably similar.
Discussion
BAX-Mediated Apoptosis Is Important in an Inherited Glaucoma
Our findings provide important new information about RGC injury and death in glaucoma. BAX deficiency completely prevents RGC death in DBA/2J mice. These results conclusively demonstrate that apoptosis plays a pivotal role in this inherited model of glaucoma. BAX is the first molecule shown to be completely necessary for RGC death in any glaucoma. Considering the protection we demonstrate in this mouse model, it is worth assessing BAX pathways as important targets for new treatments in human glaucoma.
Distinct Pathways Mediate RGC Death and Axonal Degeneration in Glaucoma
Intrinsic axonal degeneration pathways have recently been identified [55,56]. The molecular components of these pathways appear to be distinct from those active in classical somal apoptosis [57,58]. Thus, the different compartments of a neuron can degenerate by different molecular processes. In glaucoma, it is not clear whether the same or different degeneration pathway(s) are activated in the cell body and axon. Our study demonstrates that BAX is required for RGC death but not for RGC axonal degeneration in DBA/2J glaucoma. This indicates that the axonal degeneration pathway is distinct from apoptosis in this inherited glaucoma. Our findings clearly demonstrate that axon degeneration is not a consequence of RGC death, since severe axon degeneration occurred in Bax
−/− mice without RGC death. It is not yet clear whether the RGC apoptosis and axonal degeneration pathways have some common features or are completely distinct. However, for the design of therapeutic strategies for human glaucoma, our studies suggest that both apoptotic and axonal degeneration pathways should be considered.
Alternative Glaucoma Hypotheses
The initial RGC compartments that are insulted in glaucoma, as well as the nature of the damaging insults that induce degeneration, are not completely clear. In the excitotoxic hypothesis of glaucoma, elevated IOP leads to elevated intraocular glutamate levels [59]. The elevated glutamate levels are proposed to cause excessive stimulation of glutamate receptors (NMDA type), leading to increased intracellular calcium levels and RGC death. A different glaucoma hypothesis involves direct optic nerve injury. In this hypothesis, high pressure places stress on the optic nerve as the nerve exits the eye through the lamina cribrosa [60]. Important studies report that the first damage to RGCs is evident in the axon segment near the lamina cribrosa in the optic nerve head [61,62], so it was suggested that this is the first site of IOP-induced insult (see Quigley [60]). Although it definitively shows local axonal dysfunction, the occurrence of initial damage in this region does not conclusively indicate that this is the first or only site of neuronal insult. Because of optic nerve head architecture and the stress at the lamina cribrosa, it is conceivable that the axon segment at the lamina cribrosa may take substantial resources to maintain, especially when IOP is elevated. Somal stress may decrease available resources for axon maintenance and repair. Therefore, somal stress or damage may contribute to the abnormalities observed in the optic nerve head. As a group, Bax
+/− mice had an indistinguishable IOP insult compared to Bax
+/+ mice, but their RGCs did not undergo pressure-induced cell death. Importantly, RGC axonal degeneration was delayed in these Bax
+/− mice. Therefore, our data imply that shielding the RGC cell bodies has a protective effect against axon degeneration.
Direct Optic Nerve Damage Resembles Glaucoma
To provide insight to the nature and location of the damaging insults that occur in glaucoma, we compared the effects of BAX deficiency on RGC death in inherited glaucoma to RGC death induced by either direct optic nerve injury or excitotoxicity (all in the genetically uniform DBA/2J strain). Intraocular NMDA injection was used to model excitotoxic RGC death, and controlled optic nerve crush was used to mimic direct optic nerve damage [53,54]. Unlike the DBA/2J glaucoma, our experiments show that the excitotoxic insult does not require BAX to induce RGC death. Although these experiments cannot rule out the possibility of an intrinsic excitotoxic mechanism, these results do not support a role of NMDA receptor-mediated excitotoxicity as a primary cause of glaucomatous RGC death. Similar to the DBA/2J glaucoma, RGC death following optic nerve crush requires BAX, and both Bax
+/− and Bax
−/− mice are profoundly protected. Along with our demonstration of an axon intrinsic degeneration pathway, these results further support the hypothesis [60] that direct optic nerve and axon injury is an important pathogenic component leading to RGC death in glaucoma.
Bax Can Modulate Neuronal Susceptibility in Glaucoma
Individual patients have different levels of susceptibility to glaucomatous RGC death [2,63]. Our experiments clearly identify Bax as an important modulator of neuronal susceptibility in DBA/2J glaucoma. BAX deficiency prevented RGC death and delayed optic nerve degeneration in both Bax
+/− and Bax
−/− mice. These results suggest that the use of BAX inhibitors could potentially be used to delay glaucomatous vision loss. In situations where BAX is important, pharmacologically suppressing BAX activity may significantly slow the progression of glaucoma. Since RGCs were maintained for an extended period after axon degeneration in Bax
−/− mice, treatments that inhibit BAX pathways may allow long-term preservation of RGC cell bodies. Such treatments may allow the RGCs of patients to be stored in their own retinas until future treatment strategies are developed that can stimulate axonal growth and restore vision.
Complete Bax Deficiency Limits IOP Elevation
In addition to implicating BAX as a target for direct neuroprotective treatments, the lower IOP of Bax
−/− mice suggests that BAX inhibition may delay or limit IOP elevation. These results suggest that apoptotic death of cells affecting aqueous humor drainage contributes to IOP elevation, at least in secondary glaucomas where the drainage structures are insulted by pigment and cell debris. In a previous study assessing neuroprotection by an apoptosis inhibitor in a rat model of glaucoma, the treated rats had lower IOP than the other group [25]. Although not a conclusion of this rat study, the IOP data support a role for apoptosis in IOP elevation. In humans, cell death has been speculated to contribute to common forms of glaucoma (due to loss of drainage structure cells in old individuals and at late stages of glaucoma [64,65]). However, a primary role for ocular drainage pathway cell death during IOP elevation is not clearly established. Importantly, a recent study convincingly demonstrated endoplasmic reticulum stress and subsequent cell death in primary cultures of drainage pathway cells expressing human glaucoma mutations [66]. Together with our finding that complete BAX deficiency delays IOP elevation in a glaucoma setting, these results strongly support further investigation of apoptotic pathways and effects of antiapoptotic drugs on IOP in human glaucoma.
BAX Is a Candidate Human Glaucoma Susceptibility Gene
The profound protection against RGC death and the delay in axon degeneration in Bax
+/− mice together suggest BAX as a candidate human glaucoma susceptibility gene. It is important to note that we considered the possibility that a closely linked gene that was transferred from the 129/SV strain (in which the Bax mutation was generated) hitchhiked into the DBA/2J background along with Bax and explains the protection in heterozygotes. We conclude that this possibility is remote on the basis of the following observations. First, the RGCs of wild-type mice of the parental 129/SV strain are not protected from optic nerve crush. Bax heterozygosity protected the animals from both optic nerve crush and glaucoma in our experiments. This strongly implies that the parental strain does not have a modifier gene that would account for the protection we observed. Second, almost all RGCs were saved in the Bax
+/− mice despite complete axon degeneration. To our knowledge, only two genes have been documented that can save the cell when the axon is destroyed. Substantial overexpression of Bcl2 (a BAX antagonist) can do this, as can Bax deficiency. Thus, it is very unlikely that there is a similarly potent gene in the congenic interval, and scanning the flanking chromosome identifies no obvious candidates.
Complete BAX deficiency has developmental consequences [44] and is unlikely to be common in the human population. However, human BAX alleles that quantitatively affect the level of BAX are identified, and are reported to affect the development and progression of some but not other diseases [67–72]. Other factors that control BAX expression could also be important. Lower levels of BAX are associated with a worse prognosis for some types of cancer [73]. Our findings in Bax
+/− mice support the hypothesis that quantitative variation in the level of BAX gene product may alter the prognosis of glaucomatous damage in individuals with high IOP. Although further studies are needed to assess this possibility, quantitative variation of BAX activity among human patients may have a substantial effect on susceptibility and disease progression. It is possible that lower-activity alleles may result in slower or less severe damage, whereas high-activity alleles may be detrimental. Characterization of BAX alleles may have important predictive value for disease progression.
Materials and Methods
Animals and husbandry.
Mice were housed in a 14 h light to 10 h dark cycle under previously described conditions [74]. The Jackson Laboratory (Bar Harbor, Maine, United States) pathogen surveillance program regularly screened for pathogens. All experiments were conducted in accordance with the Association for Research in Vision and Ophthalmology's statement on the use of animals in ophthalmic research and were approved by our institutional animal care and use committees. Both male and female mice were used. For each age group and genotype, approximately equal numbers of males and females were used. A Bax null allele (Baxtm1Sjk [44]; herein referred to as Bax
−) was backcrossed from B6.129X1-Bax
tm1SjK (obtained from The Jackson Laboratory) onto DBA/2J for more than 12 generations to generate the congenic strain D2.129X1(B6)-Bax
tm1Sjk
/Sj. Congenic DBA/2J Bax
+/− mice were intercrossed to produce Bax
+/+
,
Bax
+/−, and Bax
−/− littermates. All three genotypes were housed together and analyzed simultaneously. DBA/2J mice were from our colony (Sj) that was initiated with mice purchased from The Jackson Laboratory. DBA/1J mice were obtained from The Jackson Laboratory.
Cell death related assays.
Eyes from DBA/2J or control DBA/1J mice were fixed in 4% paraformaldehyde in 0.1M phosphate buffer (pH 7.2) for 3 h, transferred to 0.4% paraformaldehyde in 0.1 M phosphate buffer for 48 h, and infiltrated with paraffin. Eyes from two 10- to 11-mo-old DBA/2J mice and two control mice were sectioned at 5 μm thickness and subjected to a modified double labeling protocol that involved in situ end-labeling (equivalent to a TUNEL assay) of fragmented DNA (using BODIPY fluorophores; Molecular Probes, Eugene, Oregon, United States) and detection of condensed chromatin (with the dimeric cyanine dye YOYO-1; Molecular Probes) as published [75]. Samples were analyzed with a confocal microscope. Conventional TUNEL assays were performed as previously reported [13] and conducted on the following numbers of DBA/2J mice of each age group: 7 mo (six), 8–9 mo (ten), 10–11 mo (16), 12–13 mo (nine), and 15–18 mo (eight). Five 10- to 12-mo-old control DBA/1J mice and more than 15 control mice of mixed genetic background ranging from 10 to 14 mo old were also analyzed. Counts of TUNEL positive cells were done as previously reported [53]. Briefly, the number of TUNEL-positive RGC layer cells was counted for eight to 12 sagittal sections from each eye, and average values for each age group are reported. Eyes used for electron microscopy and histology were processed as previously described [76], except that tissue blocks were oriented for en face retinal sectioning through the ganglion cell layer.
Clinical examination and intraocular pressure measurement.
DBA/2J mice develop a pigmentary form of glaucoma that follows a characteristic easily detectable clinical course. DBA/2J mice (all genotypes) used in the spontaneous glaucoma experiments were assessed with a slit lamp to ensure that the Bax mutation did not alter the course of the disease. Slit-lamp examination and evaluation criteria (including pigment dispersion and transillumination) were previously described [41,42]. Examination of at least 40 mice of each genotype at 6 and 9 mo of age and at the time of harvest (10.5 or 12 mo) was performed. Additionally, smaller groups of mice (12–20 of each genotype) were analyzed at other ages between 3 and 12 mo of age. IOP was recorded [77,78] for mice of each genotype. The number of mice of each genotype successfully assessed at each age were as follows. For 4 mo, Bax
+/+
n = 22, Bax
+/−
n = 22, Bax
−/−
n = 20; for 9 mo, Bax
+/+
n = 21, Bax
+/−
n = 25, Bax
−/−
n = 18; for 10.5 mo, Bax
+/+
n = 50, Bax
+/−
n = 54, Bax
−/−
n = 52; and for 12 mo, Bax
+/+
n = 42, Bax
+/−
n = 42, Bax
−/−
n = 37. Student's t-tests were used for statistical comparisons.
Optic nerve damage.
Optic nerves were dissected, processed, embedded in plastic, sectioned and stained with paraphenylenediamine (PPD) as previously described [76], except that the staining time was increased to 35 min and Embed 812 medium was used. PPD stains all myelin sheaths, but differentially stains the axoplasm of sick or dying axons darkly. Counts of normal-appearing axons were performed using established nonbiased counting methods. Prior to beginning axon counts, the optic nerve was outlined at 100× magnification, and its cross-sectional area was automatically calculated. Magnification of the same nerve section was increased to 1,000×, and a total of 20 fields at 1,000× were electronically collected. The fields were spaced in a regular fashion across the entire nerve, taking care to avoid field overlap so that the same area was not counted twice. The 20 collected pictures were stacked on the computer screen so that only the final picture was visible to the operator. For nerves with a large number of axons (mildly and moderately affected nerves), a rectangular box that contained a minimum of 200 axons was then drawn on the twentieth image. For nerves with severe axon loss, a larger box was drawn so that a significant proportion of the nerve could be counted. The software program then “cut” a rectangle centred at the same location in all 20 images. Since the operator could only see the top image, this removed the possibility of unconscious operator bias and made the selection of axons to be counted random. Axons were counted manually and marked using the computer. The program tracked the total area counted and the total axon count for all 20 images. The total counted area averaged 12.1%, 14.2%, and 20.5% of the total nerve area for mildly, moderately, and severly affected nerves, respectively. The final count was calculated and expressed as number of axons per optic nerve. With this approach, the nerves with 95% or more axon loss were selected for RGC counts by comparing the remaining axon number to the average for unaffected nerves of the same genotype.
Because of the large number of mice (approximately 50–70 mice of each genotype at each age), an optic nerve rating scale was used for the glaucoma progression study (see Figure 5). The indicated damage levels are readily distinguishable upon inspection of the nerve without counting. Nevertheless, axon counts were performed on at least eight randomly selected nerves of each damage grade to provide quantitative information about these distinct stages of disease (see below). Two investigators (masked to genotype, age, and the damage level assigned by the other investigator) assigned a damage level to each nerve. The two investigators assigned the same grade more than 90% of the time (321 out of 355 nerves). For the nerves on which the initial two investigators differed, a third (masked) investigator was utilized. The third investigator's grade always agreed with one of the initial grades, and the most common assigned grade was used. The number of nerves of each genotype assessed at each age were as follows. For 10.5 mo, Bax
+/+
n = 49, Bax
+/−
n = 62, Bax
−/−
n = 58; for 12 mo, Bax
+/+
n = 71, Bax
+/−
n = 50, Bax
−/−
n = 65.
The damage levels and typical numbers of normal axons present at each stage (determined through axon counts by an investigator masked to damage grade) follow. The representative axon counts were determined for randomly selected nerves of each grade using the counting procedure described above. In mildly affected nerves, there was very mild or no damage, with healthy axons having a clear axoplasm and intact myelin sheath (average number of axons ± SEM: 50,504 ± 1,988; n = 8). In moderately affected nerves, darkly stained, degenerating axons were readily detectable, but the vast majority of axons appeared completely normal (average number of axons ± SEM: 31,410 ± 2,199; n = 8 [79]). In severely affected nerves, there was extensive axon damage throughout the optic nerve with obvious axon loss (average number of axons ± SEM: 7,970 ± 2,150; n = 17). The axon number was significantly different between optic nerves of each damage level (p < 0.001 for all comparisons, t-tests).
Ganglion cell death.
Eyes were fixed and retinas were flat-mounted and Nissl-stained with cresyl violet using a modification of the technique reported by Stone [80]. Retinal ganglion cells make up approximately 40%–60% of the neurons in the ganglion cell layer of the mouse retina, and all RGC subtypes cannot be reliably distinguished from the other resident neuron in the ganglion cell layer (the displaced amacrine cell) based on cellular morphology [53,81]. This is especially true during disease, when morphology and marker expression can change dramatically. Consequently, cell loss was measured as a function of the change in total cell number compared to control eyes (strain and genotype matched nonglaucomatous eyes for the spontaneous glaucoma experiments and the contralateral nonmanipulated eye for the controlled crush and excitotoxic experiments). RGC density varies greatly with respect to retinal location. Therefore, two 40× fields were counted in each retinal quadrant and care was taken to ensure that the fields were the same distance from the periphery. For each individual eye, the eight counts for each retina were averaged. To assess RGC survival in the spontaneous glaucoma, retinas from eyes with very severely affected nerves that had fewer than 5% surviving axons were compared to retinas from unaffected eyes without glaucomatous nerve damage. RGC number was counted in approximately eight severely affected eyes and eight unaffected eyes of each genotype, except for unaffected control Bax
+/− mice (five eyes) and 18 mo unaffected Bax
−/− mice (four eyes).
NMDA injections and controlled optic nerve crush.
These experiments were performed as described previously [53]. For NMDA injections, 2 μl of an 80 mM solution of NMDA in balanced saline solution was injected intravitreally into one eye of each mouse using a glass micropipet. After 4 d the eyes were harvested and cells counted as described above. Data were collected from ten Bax
+/+ and eight Bax
−/− mice. For optic nerve crush, the nerve of one eye was exposed and clamped approximately 0.5 mm from the globe with self-closing jeweler's forceps for 4 s. Eyes were harvested 21 d after surgery and cells counted. Data were collected from nine Bax
+/+, nine Bax
+/−, and seven Bax
−/− mice. In each paradigm, cell loss was measured relative to the cell number present in the control eye of each mouse examined.
Supporting Information
Accession Numbers
The GenBank (http://www.ncbi.nlm.nih.gov/) accession number for Bax is 12028.
We thank G. Cox, R. Burgess, and Edward Leiter for critical reading of the manuscript, and Amy Snow, Larry Wilson, Adriana Zabaleta, and Mihai Cosma for technical assistance with the experiments. Scientific support services at The Jackson Laboratory are subsidized by a core grant from the National Cancer Institute (CA34196). This work was supported in part by R29EY12223 (RWN) and F32EY014515 (RTL). SWMJ is an Investigator of The Howard Hughes Medical Institute.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. RTL, RWN, and SWMJ conceived and designed the experiments. RTL, YL, OVS, JB, RSS, RWN, and SWMJ performed the experiments. RTL, YL, OVS, JB, RSS, RWN, and SWMJ analyzed the data. RTL and SWMJ wrote the paper.
Abbreviations
BAXBCL2-associated X protein
IOPintraocular pressure
NMDAN-methyl-D-aspartate
RGCretinal ganglion cell
SEMstandard error of the mean
==== Refs
References
Quigley HA 1996 Number of people with glaucoma worldwide Br J Ophthalmol 80 389 393 8695555
Leske MC 1983 The epidemiology of open-angle glaucoma: A review Am J Epidemiol 118 166 191 6349332
Kass MA Heuer DK Higginbotham EJ Johnson CA Keltner JL 2002 The ocular hypertension treatment study: A randomized trial determines that topical ocular hypotensive medication delays or prevents the onset of primary open-angle glaucoma Arch Ophthalmol 120 701 713 ; discussion 829 830 12049574
The Advanced Glaucoma Intervention Study investigators 2000 The Advanced Glaucoma Intervention Study: 7. The relationship between control of intraocular pressure and visual field deterioration. The AGIS Investigators Am J Ophthalmol 130 429 440 11024415
The Advanced Glaucoma Intervention Study investigators 2002 The Advanced Glaucoma Intervention Study: 11. Risk factors for failure of trabeculectomy and argon laser trabeculoplasty Am J Ophthalmol 134 481 498 12383805
Klein BE Klein R Sponsel WE Franke T Cantor LB 1992 Prevalence of glaucoma. The Beaver Dam eye study Ophthalmology 99 1499 1504 1454314
Kamal D Hitchings R 1998 Normal tension glaucoma—A practical approach Br J Ophthalmol 82 835 840 9924383
Osborne NN Wood JP Chidlow G Bae JH Melena J 1999 Ganglion cell death in glaucoma: What do we really know? Br J Ophthalmol 83 980 986 10413706
John SW Anderson MG Smith RS 1999 Mouse genetics: A tool to help unlock the mechanisms of glaucoma J Glaucoma 8 400 412 10604301
Grozdanic SD Betts DM Sakaguchi DS Allbaugh RA Kwon YH 2003 Laser-induced mouse model of chronic ocular hypertension Invest Ophthalmol Vis Sci 44 4337 4346 14507878
Mabuchi F Aihara M Mackey MR Lindsey JD Weinreb RN 2003 Optic nerve damage in experimental mouse ocular hypertension Invest Ophthalmol Vis Sci 44 4321 4330 14507876
Ueda J Sawaguchi S Hanyu T Yaoeda K Fukuchi T 1998 Experimental glaucoma model in the rat induced by laser trabecular photocoagulation after an intracameral injection of India ink Jpn J Ophthalmol 42 337 344 9822959
Li Y Schlamp CL Poulsen KP Nickells RW 2000 Bax-dependent and independent pathways of retinal ganglion cell death induced by different damaging stimuli Exp Eye Res 71 209 213 10930325
Johnson EC Morrison JC Farrell S Deppmeier L Moore CG 1996 The effect of chronically elevated intraocular pressure on the rat optic nerve head extracellular matrix Exp Eye Res 62 663 674 8983948
Aihara M Lindsey JD Weinreb RN 2003 Experimental mouse ocular hypertension: Establishment of the model Invest Ophthalmol Vis Sci 44 4314 4320 14507875
Levkovitch-Verbin H Quigley HA Martin KR Valenta D Baumrind LA 2002 Translimbal laser photocoagulation to the trabecular meshwork as a model of glaucoma in rats Invest Ophthalmol Vis Sci 43 402 410 11818384
Garcia-Valenzuela E Shareef S Walsh J Sharma SC 1995 Programmed cell death of retinal ganglion cells during experimental glaucoma Exp Eye Res 61 33 44 7556468
Isenmann S Engel S Gillardon F Bahr M 1999 Bax antisense oligonucleotides reduce axotomy-induced retinal ganglion cell death in vivo by reduction of Bax protein expression Cell Death Differ 6 673 682 10453078
Quigley HA Nickells RW Kerrigan LA Pease ME Thibault DJ 1995 Retinal ganglion cell death in experimental glaucoma and after axotomy occurs by apoptosis Invest Ophthalmol Vis Sci 36 774 786 7706025
Nickells RW 1999 Apoptosis of retinal ganglion cells in glaucoma: An update of the molecular pathways involved in cell death Surv Ophthalmol 43 Supp 1 S151 S161 10416758
Tatton NA Tezel G Insolia SA Nandor SA Edward PD 2001 In situ detection of apoptosis in normal pressure glaucoma. A preliminary examination Surv Ophthalmol 45 Suppl 3 S268 S272 ; discussion S273 S266 11377447
Kerrigan LA Zack DJ Quigley HA Smith SD Pease ME 1997 TUNEL-positive ganglion cells in human primary open-angle glaucoma Arch Ophthalmol 115 1031 1035 9258226
Kugler S Straten G Kreppel F Isenmann S Liston P 2000 The X-linked inhibitor of apoptosis (XIAP) prevents cell death in axotomized CNS neurons in vivo Cell Death Differ 7 815 824 11042676
Straten G Schmeer C Kretz A Gerhardt E Kugler S 2002 Potential synergistic protection of retinal ganglion cells from axotomy-induced apoptosis by adenoviral administration of glial cell line-derived neurotrophic factor and X-chromosome-linked inhibitor of apoptosis Neurobiol Dis 11 123 133 12460552
McKinnon SJ Lehman DM Tahzib NG Ransom NL Reitsamer HA 2002 Baculoviral IAP repeat-containing-4 protects optic nerve axons in a rat glaucoma model Mol Ther 5 780 787 12027563
Kikuchi M Tenneti L Lipton SA 2000 Role of p38 mitogen-activated protein kinase in axotomy-induced apoptosis of rat retinal ganglion cells J Neurosci 20 5037 5044 10864961
Kermer P Klocker N Labes M Thomsen S Srinivasan A 1999 Activation of caspase-3 in axotomized rat retinal ganglion cells in vivo FEBS Lett 453 361 364 10405176
Kermer P Ankerhold R Klocker N Krajewski S Reed JC 2000 Caspase-9: Involvement in secondary death of axotomized rat retinal ganglion cells in vivo Brain Res Mol Brain Res 85 144 150 11146116
McKinnon SJ Lehman DM Kerrigan-Baumrind LA Merges CA Pease ME 2002 Caspase activation and amyloid precursor protein cleavage in rat ocular hypertension Invest Ophthalmol Vis Sci 43 1077 1087 11923249
Hanninen VA Pantcheva MB Freeman EE Poulin NR Grosskreutz CL 2002 Activation of caspase 9 in a rat model of experimental glaucoma Curr Eye Res 25 389 395 12789547
Tatton WG Chalmers-Redman RM Tatton NA 2001 Apoptosis and anti-apoptosis signalling in glaucomatous retinopathy Eur J Ophthalmol 11 Suppl 2 S12 S22 11592526
Wakabayashi T Kosaka J Hommura S 2002 Up-regulation of Hrk, a regulator of cell death, in retinal ganglion cells of axotomized rat retina Neurosci Lett 318 77 80 11796190
Napankangas U Lindqvist N Lindholm D Hallbook F 2003 Rat retinal ganglion cells upregulate the pro-apoptotic BH3-only protein Bim after optic nerve transection Brain Res Mol Brain Res 120 30 37 14667574
Bonfanti L Strettoi E Chierzi S Cenni MC Liu XH 1996 Protection of retinal ganglion cells from natural and axotomy-induced cell death in neonatal transgenic mice overexpressing bcl-2 J Neurosci 16 4186 4194 8753880
Tezel G Yang X Yang J Wax MB 2004 Role of tumor necrosis factor receptor-1 in the death of retinal ganglion cells following optic nerve crush injury in mice Brain Res 996 202 212 14697498
Yoshida K Behrens A Le-Niculescu H Wagner EF Harada T 2002 Amino-terminal phosphorylation of c-Jun regulates apoptosis in the retinal ganglion cells by optic nerve transection Invest Ophthalmol Vis Sci 43 1631 1635 11980884
Tezel G Yang X 2004 Caspase-independent component of retinal ganglion cell death, in vitro Invest Ophthalmol Vis Sci 45 4049 4059 15505055
Deckwerth TL Elliott JL Knudson CM Johnson EM Jr Snider WD 1996 BAX is required for neuronal death after trophic factor deprivation and during development Neuron 17 401 411 8816704
Mosinger Ogilvie J Deckwerth TL Knudson CM Korsmeyer SJ 1998 Suppression of developmental retinal cell death but not of photoreceptor degeneration in Bax-deficient mice Invest Ophthalmol Vis Sci 39 1713 1720 9699561
Qin Q Patil K Sharma SC 2004 The role of Bax-inhibiting peptide in retinal ganglion cell apoptosis after optic nerve transection Neurosci Lett 372 17 21 15531080
Anderson MG Smith RS Hawes NL Zabaleta A Chang B 2002 Mutations in genes encoding melanosomal proteins cause pigmentary glaucoma in DBA/2J mice Nat Genet 30 81 85 11743578
John SW Smith RS Savinova OV Hawes NL Chang B 1998 Essential iris atrophy, pigment dispersion, and glaucoma in DBA/2J mice Invest Ophthalmol Vis Sci 39 951 962 9579474
Chang B Smith RS Hawes NL Anderson MG Zabaleta A 1999 Interacting loci cause severe iris atrophy and glaucoma in DBA/2J mice Nat Genet 21 405 409 10192392
Knudson CM Tung KS Tourtellotte WG Brown GA Korsmeyer SJ 1995 Bax-deficient mice with lymphoid hyperplasia and male germ cell death Science 270 96 99 7569956
Strom RC Williams RW 1998 Cell production and cell death in the generation of variation in neuron number J Neurosci 18 9948 9953 9822750
Potts RA Dreher B Bennett MR 1982 The loss of ganglion cells in the developing retina of the rat Brain Res 255 481 486 7066701
Young RW 1984 Cell death during differentiation of the retina in the mouse J Comp Neurol 229 362 373 6501608
Burne JF Staple JK Raff MC 1996 Glial cells are increased proportionally in transgenic optic nerves with increased numbers of axons J Neurosci 16 2064 2073 8604051
Doughty ML De Jager PL Korsmeyer SJ Heintz N 2000 Neurodegeneration in Lurcher mice occurs via multiple cell death pathways J Neurosci 20 3687 3694 10804210
White FA Keller-Peck CR Knudson CM Korsmeyer SJ Snider WD 1998 Widespread elimination of naturally occurring neuronal death in Bax-deficient mice J Neurosci 18 1428 1439 9454852
Cheng L Sapieha P Kittlerova P Hauswirth WW Di Polo A 2002 TrkB gene transfer protects retinal ganglion cells from axotomy-induced death in vivo J Neurosci 22 3977 3986 12019317
Mo X Yokoyama A Oshitari T Negishi H Dezawa M 2002 Rescue of axotomized retinal ganglion cells by BDNF gene electroporation in adult rats Invest Ophthalmol Vis Sci 43 2401 2405 12091443
Li Y Schlamp CL Nickells RW 1999 Experimental induction of retinal ganglion cell death in adult mice Invest Ophthalmol Vis Sci 40 1004 1008 10102300
Levin LA 2001 Animal and culture models of glaucoma for studying neuroprotection Eur J Ophthalmol 11 Suppl 2 S23 S29 11592527
Raff MC Whitmore AV Finn JT 2002 Axonal self-destruction and neurodegeneration Science 296 868 871 11988563
Coleman MP Perry VH 2002 Axon pathology in neurological disease: A neglected therapeutic target Trends Neurosci 25 532 537 12220882
Finn JT Weil M Archer F Siman R Srinivasan A 2000 Evidence that Wallerian degeneration and localized axon degeneration induced by local neurotrophin deprivation do not involve caspases J Neurosci 20 1333 1341 10662823
Whitmore AV Lindsten T Raff MC Thompson CB 2003 The proapoptotic proteins Bax and Bak are not involved in Wallerian degeneration Cell Death Differ 10 260 261 12700655
Vorwerk CK Gorla MS Dreyer EB 1999 An experimental basis for implicating excitotoxicity in glaucomatous optic neuropathy Surv Ophthalmol 43 Suppl 1 S142 S150 10416757
Quigley HA 1999 Neuronal death in glaucoma Prog Retin Eye Res 18 39 57 9920498
Quigley HA Hohman RM Addicks EM Massof RW Green WR 1983 Morphologic changes in the lamina cribrosa correlated with neural loss in open-angle glaucoma Am J Ophthalmol 95 673 691 6846459
Quigley HA Addicks EM Green WR Maumenee AE 1981 Optic nerve damage in human glaucoma. II. The site of injury and susceptibility to damage Arch Ophthalmol 99 635 649 6164357
Libby RT Gould DG Anderson MG Smith RS John SWM 2005 Complex genetics of glaucoma susceptibility Annu Rev Genomics Hum Genet In press.
Alvarado J Murphy C Polansky J Juster R 1981 Age-related changes in trabecular meshwork cellularity Invest Ophthalmol Vis Sci 21 714 727 7298275
McMenamin PG Lee WR Aitken DA 1986 Age-related changes in the human outflow apparatus Ophthalmology 93 194 209 3951826
Liu Y Vollrath D 2004 Reversal of mutant myocilin non-secretion and cell killing: Implications for glaucoma Hum Mol Genet 13 1193 1204 15069026
Moshynska O Moshynskyy I Misra V Saxena A 2005 G125A single-nucleotide polymorphism in the human BAX promoter affects gene expression Oncogene 24 2042 2049 15688029
Starczynski J Pepper C Pratt G Hooper L Thomas A 2003 The P2X7 receptor gene polymorphism 1513 A→C has no effect on clinical prognostic markers, in vitro sensitivity to fludarabine, Bcl-2 family protein expression or survival in B-cell chronic lymphocytic leukaemia Br J Haematol 123 66 71 14510944
Moshynska O Sankaran K Saxena A 2003 Molecular detection of the G(-248)A BAX promoter nucleotide change in B cell chronic lymphocytic leukaemia Mol Pathol 56 205 209 12890741
Zeng SM Yankowitz J Widness JA Strauss RG 2003 Sequence-based polymorphisms in members of the apoptosis Bcl-2 gene family and their association with hematocrit level J Gend Specif Med 6 36 42 14714449
Kuhlmann T Glas M zum Bruch C Mueller W Weber A 2002 Investigation of bax, bcl-2, bcl-x and p53 gene polymorphisms in multiple sclerosis J Neuroimmunol 129 154 160 12161031
Saxena A Moshynska O Sankaran K Viswanathan S Sheridan DP 2002 Association of a novel single nucleotide polymorphism, G(-248)A, in the 5′-UTR of BAX gene in chronic lymphocytic leukemia with disease progression and treatment resistance Cancer Lett 187 199 205 12359369
Sturm I Kohne CH Wolff G Petrowsky H Hillebrand T 1999 Analysis of the p53/BAX pathway in colorectal cancer: Low BAX is a negative prognostic factor in patients with resected liver metastases J Clin Oncol 17 1364 1374 10334520
Smith RS Zabaleta A Kume T Savinova OV Kidson SH 2000 Haploinsufficiency of the transcription factors FOXC1 and FOXC2 results in aberrant ocular development Hum Mol Genet 9 1021 1032 10767326
Tatton NA Maclean-Fraser A Tatton WG Perl DP Olanow CW 1998 A fluorescent double-labeling method to detect and confirm apoptotic nuclei in Parkinson's disease Ann Neurol 44 S142 S148 9749586
Smith RS Zabaleta A John SW Bechtold LS Ikeda S 2002 General and specific histopathology Smith RS Systemic evaluation of the mouse eye New York CRC Press 265 297
Savinova OV Sugiyama F Martin JE Tomarev SI Paigen BJ 2001 Intraocular pressure in genetically distinct mice: an update and strain survey BMC Genet 2 12 11532192
John SWM Hagaman JR MacTaggart TE Peng L Smithes O 1997 Intraocular pressure in inbred mouse strains Invest Ophthalmol Vis Sci 38 249 253 9008647
Anderson MG Libby RT Gould DB Smith RS John SWM 2005 High-dose radiation with bone marrow transfer prevents neurodegeneration in an inherited glaucoma Proc Natl Acad Sci U S A 102 4566 4571 15758074
Stone J 1981 The wholemount handbook Sydney Maitland Publishing 3 23
Jeon CJ Strettoi E Masland RH 1998 The major cell populations of the mouse retina J Neurosci 18 8936 8946 9786999
Sun W Oppenheim RW 2003 Response of motoneurons to neonatal sciatic nerve axotomy in Bax-knockout mice Mol Cell Neurosci 24 875 886 14697655
|
16103918
|
PMC1183523
|
CC BY
|
2021-01-05 08:00:57
|
no
|
PLoS Genet. 2005 Jul 25; 1(1):e4
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010004
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 10.1371/journal.pgen.001000505-PLGE-RA-0029R1plge-01-01-04Research ArticleEvolutionGenetics/Population GeneticsEukaryotesYeast and FungiSaccharomycesEvidence for Domesticated and Wild Populations of Saccharomyces cerevisiae
Yeast DomesticationFay Justin C *Benavides Joseph A Department of Genetics, Washington University School of Medicine, St. Louis, Missouri, United States of AmericaHaber James E EditorBrandeis University, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 25 7 2005 1 1 e523 2 2005 8 4 2005 Copyright: © 2005 Fay and Benavides.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
Saccharomyces cerevisiae is predominantly found in association with human activities, particularly the production of alcoholic beverages. S. paradoxus, the closest known relative of S. cerevisiae, is commonly found on exudates and bark of deciduous trees and in associated soils. This has lead to the idea that S. cerevisiae is a domesticated species, specialized for the fermentation of alcoholic beverages, and isolates of S. cerevisiae from other sources simply represent migrants from these fermentations. We have surveyed DNA sequence diversity at five loci in 81 strains of S. cerevisiae that were isolated from a variety of human and natural fermentations as well as sources unrelated to alcoholic beverage production, such as tree exudates and immunocompromised patients. Diversity within vineyard strains and within saké strains is low, consistent with their status as domesticated stocks. The oldest lineages and the majority of variation are found in strains from sources unrelated to wine production. We propose a model whereby two specialized breeds of S. cerevisiae have been created, one for the production of grape wine and one for the production of saké wine. We estimate that these two breeds have remained isolated from one another for thousands of years, consistent with the earliest archeological evidence for winemaking. We conclude that although there are clearly strains of S. cerevisiae specialized for the production of alcoholic beverages, these have been derived from natural populations unassociated with alcoholic beverage production, rather than the opposite.
Synopsis
The budding yeast, Saccharomyces cerevisiae, has been used to make bread, beer, and wine for thousands of years. To investigate the evolutionary history of this species, the authors examined DNA sequence variation from a large collection of yeast strains isolated from a variety of sources, including saké wine, grape wine, clinical samples, tree exudates, and fruit. The DNA sequence diversity among these strains shows that both saké and grape wine strains form two distinct groups that have remained isolated for a substantial period of time. The data suggest that S. cerevisiae consists of both “wild” and “domesticated” populations and that at least two independent domestication events lead to extant grape wine and saké wine strains.
Citation:Fay JC, Benavides JA (2005) Evidence for domesticated and wild populations of Saccharomyces cerevisiae. PLoS Genet 1(1): e5.
==== Body
Introduction
Sensu strictu species of the genus Saccharomyces, as their scientific name implies, are yeast specialized for growth on sugar. In comparison to other yeasts, Saccharomyces favor aerobic fermentation over respiration in the presence of high concentrations of sugar [1]. Fermentation results in the production of ethanol and a competitive advantage, as these yeasts are tolerant to high concentrations of ethanol [2]. One of these species, S. cerevisiae, has served as one of the best model systems for understanding the eukaryotic cell and has served as the dominant species for the production of beer, bread, and wine [3]. However, it is worth noting that strains of S. bayanus are sometimes used for wine production and strains of S. pastorianus, hybrids between S. cerevisiae and S. bayanus, are used to brew lagers [4].
Since the discovery of yeast as the cause of fermentation [5], numerous strains of S. cerevisiae have been isolated, the majority of which have been found associated with the production of alcoholic beverages [6–9]. In many instances, the strains are clearly specialized for use in the lab [10] and the production of wine [11], beer [12], and bread [13]. This has lead to the common view that S. cerevisiae is a domesticated species that has continuously evolved in association with the production of alcoholic beverages [3,6,14]. Under this model, the occasional strains of S. cerevisiae found in nature are thought to be migrants from human-associated fermentations.
The first use of S. cerevisiae is likely to have been for the production of wine, rather then bread or beer [3,15]. S. cerevisiae has been associated with winemaking since 3150 BC, based on extraction of DNA from ancient wine containers [16], and the earliest evidence for winemaking is to 7000 BC from the molecular analysis of pottery jars found in China [17]. The idea that S. cerevisiae was first used to produce wine rather than beer or bread is further supported by the fact that the production of wine requires no inoculum of yeast [7]. In addition, strains associated with whisky, ale, and bakeries show amplified fragment length polymorphism (AFLP) profiles similar to various wine strains [18].
To examine the relationship between vineyard and non-vineyard strains of S. cerevisiae and to understand their evolutionary origin, we have surveyed DNA sequence variation in 81 strains isolated from geographically and ecologically diverse sources (Table 1). These include 60 strains associated with human fermentations, predominantly from vineyards, and 19 strains not associated with human fermentations, predominantly from immunocompromised patients and tree exudates.
Table 1 Strains Studied and Their Source
NA, not available; seg., segregant.
Table 1 Continued
Results/ Discussion
DNA sequence variation was examined in 81 yeast strains at five unlinked loci (see Materials and Methods). A total of 184 polymorphic sites were found. Figure 1 shows all of the variable sites along with a neighbor-joining tree constructed from these sites. There are two immediately striking features of the data. First, there are high levels of linkage disequilibrium between sites found in unlinked genes. This linkage disequilibrium cannot be explained by a lack of recombination because the four gamete test [19] shows evidence of recombination both within and between loci. The high level of linkage disequilibrium is most likely caused by population subdivision and suggests that the data from these five genes provide a genomic view of population differentiation among these strains. Second, there are significant levels of population differentiation based on the source from which the samples were isolated (see Materials and Methods). A number of strains are worth noting. Y9 is very closely related to the saké strains and was obtained from Indonesian ragi, or yeast cake, which like saké is made by fermenting koji, a mixture of rice and the mold Aspergillus oryzae [20]. Y3 and Y12 were isolated from African palm wine, made from fermenting sap of the oil palm, Elaeis guineensis. Y5 was isolated from African bili wine.
Figure 1 A Neighbor-Joining Tree Shows Differentiation among Yeast Strains Isolated from Different Sources
The tree was constructed from polymorphic sites found at five unlinked loci and was rooted using S. paradoxus. Strains are colored according to the substrates from which they were isolated. The right side shows color-coded polymorphism data with minor alleles shown in black, major alleles shown in white, missing data shown in light gray, and heterozygous sites shown in orange.
If strains of S. cerevisiae that are not associated with human fermentations have escaped their manmade environments, their progenitors should be closely related to strains isolated from human fermentations. Two aspects of the data indicate this is not the case. First, the oldest lineages at the root of the tree, that are most similar to S. paradoxus, were isolated from tree exudates in North America and Africa, or from immunocompromised patients. Although one of the clinical samples is most closely related to vineyard strains, the majority of clinical isolates are not closely related to strains obtained from human-associated fermentations. Second, strains from grape wine and saké wine production contain significantly less variation, as measured by the average number of pairwise differences between strains [21], than is found in natural and clinical isolates, which contain just as much variation as is found in the total sample (Table 2). However, diversity in strains associated with human fermentations other than grape and saké wine production is not reduced compared to the clinical and natural isolates. The four strains associated with fermentations, three of which were isolated from traditional African wines, show the greatest diversity and represent some of the oldest lineages. This raises the possibility that S. cerevisiae was domesticated in Africa and that most vineyard and saké strains were derived from a domesticated African strain. If so, one would expect clinical and natural isolates to be more closely related to strains isolated from vineyards, which have a cosmopolitan distribution compared to strains from traditional African wine. Clinical and natural isolates, however, show no obvious relationship to strains associated with manmade fermentations.
Table 2 Diversity among Strains
aOnly strains without missing data are used.
bπ is the average number of pairwise differences between strains, per basepair. The standard deviation is shown in parentheses.
Although the genealogical relationships among strains of S. cerevisiae show that the species as a whole is not domesticated, the data do support the hypothesis that some strains are domesticated. Based on the low levels of diversity within vineyard and saké strains and the clear separation of these two groups, we propose two domestication events, one for yeast used to produce grape wine and one for yeast used to produce rice wine. When might these events have occurred? Domestication would have occurred after the divergence between the vineyard and saké strains but before differentiation among the vineyard and among the saké strains. These two time points can be roughly estimated by the average number of differences per synonymous site between the saké and vineyard strains, 1.28 × 10−2, and the average number of differences among the vineyard, 2.92 × 10−3, and among the saké strains, 4.06 × 10−3, respectively (see Materials and Methods). Assuming a point mutation rate of 1.84 × 10−10 per base pair (bp) per generation and 2,920 generations per year, the estimate for the divergence time between the two groups is approximately 11,900 years ago, and within the vineyard group and saké group is approximately 2,700 and approximately 3,800 years ago, respectively (see Materials and Methods). These dates could easily be an order of magnitude older if the number of generations per year is one tenth that obtained assuming an exponential growth rate. Interestingly, the time period is consistent with the earliest archeological evidence for winemaking, approximately 9,000 years ago [17]. It should be noted that proof that these strains are domesticated requires evidence that they have acquired characteristics advantageous to humans through human activity, whether intentional or not. The alternative hypothesis to domestication is that initial fermentations selected those natural isolates most amenable to alcoholic beverage production and that these initial isolates have been used by humans ever since.
The source population for both the saké and grape wine strains is not clear, but is likely similar to the source population for the clinical strains. Insects, particularly fruit flies, present one possibility [22,23]. Numerous strains of S. cerevisiae and S. paradoxus have been isolated from oak tree exudates in North America [24], and tree exudates are often visited by insects [22]. Three of these oak tree isolates were included in our study and are among the most diverse of the strains (Figure 1). Given that S. paradoxus is most often found in association with tree exudates from both Europe [25,26] and North America [24], strains of S. cerevisiae isolated from tree exudates may be truly “wild” yeast. Whether the yeast isolated from African palm wine is domesticated remains an open question, although it is worth noting that African palm wine is made by collecting sap tapped from oil palm trees and fermentation occurs naturally without the addition of yeast.
Materials and Methods
Strains were obtained from a number of individuals and stock centers. B1–B6 were obtained from B. Dunn; I14 from J. Fay; CDB and PR from Red Star, Berkeley, California, United States; K1–K15 from N. Goto-Yamamoto and the NODAI culture collection; M1–M34 from R. Mortimer; SB from Whole Foods, Berkeley, California, United States; UC1–UC10 from the University of California, Davis stock center; Y1–Y12 from C. Kurtzman and the ARS culture collection; YJM145–YJM1129 from J. McCusker; and YPS163–YPS1009 were from the collection of P. Sniegowski.
Five genes, CCA1, CYT1, MLS1,
PDR10, and ZDS2, and their promoters were sequenced in 81 strains (see Table 1). These genes were randomly chosen from all divergently transcribed intergenic sequences upstream of functionally annotated genes with clear orthologs in S. paradoxus. The sequenced regions include 3,671 bp of coding sequence and 3,561 bp of noncoding sequence. For each gene, both strands of purified PCR products were sequenced using Big Dye (Perkin Elmer, Boston, Massachusetts, United States) termination reactions. Sequence variation was identified using phred, phrap, and consed [27]. For construction of the neighbor-joining tree, a single allele was used from strains with heterozygous sites. The allele was randomly chosen from the two haplotypes inferred by PHASE [28].
Sequence data were analyzed using DNASP [29]. Population subdivision was tested by a permutations test according to the source categories from which each strain was obtained (Table 1). The average time since divergence of two strains was obtained by k = 2μt, where k is the substitution rate, μ is the mutation rate per bp and t is the time in generations. The mutation rate has been estimated at CAN1 and SUP3 at 2.25 × 10−10 per base pair per generation [30]. Given that 82% of spontaneous mutations are single base substitutions [31], we estimate the point mutation rate is 1.84 × 10−10 per bp per generation. S. cerevisiae can reproduce in 90 min, or 16 generations per day. However, even under optimal laboratory conditions the number of generations over a 24-h period is typically much less. To obtain divergence time in years rather than generations, we assumed S. cerevisiae can go through a maximum of eight generations per day or 2,920 generations per year.
Supporting Information
Accession Numbers
The sequences of the genes CCA1, CYT1, MLS1, PDR10, and ZDS2 that are discussed in this paper have been deposited into GenBank (http://www.ncbi.nlm.nih.gov/Genbank/) as accession numbers AY942206–AY942556.
We thank two anonymous reviewers, P. Sniegowski and members of the Fay lab for comments and suggestions. We also thank B. Dunn, N. Goto-Yamamoto, R. Mortimer, C. Kurtzman, J. McCusker, and P. Sniegowski for contributing yeast strains and Heidi Kuehne for the collection of strains associated with oak exudates. Without their help this study would not have been possible.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. JCF conceived and designed the experiments. JAB performed the experiments. JCF analyzed the data and wrote the paper.
Abbreviations
bpbase pair
==== Refs
References
Otterstedt K Larsson C Bill RM Stahlberg A Boles E 2004 Switching the mode of metabolism in the yeast Saccharomyces cerevisiae
EMBO Rep 5 532 537 15071495
Sipiczk M Romano P Lipani G Miklos I Antunovics Z 2001 Analysis of yeasts derived from natural fermentation in a Tokaj winery Antonie Van Leeuwenhoek 79 97 105 11392490
Mortimer RK 2000 Evolution and variation of the yeast (Saccharomyces) genome Genome Res 10 403 409 10779481
Nguyen HV Gaillardin C 2005 Evolutionary relationships between the former species Saccharomyces uvarum and the hybrids Saccharomyces bayanus and Saccharomyces pastorianus; reinstatement of Saccharomyces uvarum (Beijerinck) as a distinct species FEMS Yeast Res 5 471 483 15691752
Pasteur L 1866 Études sur le vin Paris (France) Imprimeurs Imperials 266 p.
Martini A 1993 Origin and domestication of the wine yeast Saccharomyces cerevisiae
J Wine Res 4 165 176
Mortimer R Polsinelli M 1999 On the origins of wine yeast Res Microbiol 150 199 204 10229949
Naumova ES Bulat SA Mironenko NV Naumov GI 2003 Differentiation of six sibling species in the Saccharomyces sensu stricto complex by multilocus enzyme electrophoresis and UP-PCR analysis Antonie Van Leeuwenhoek 83 155 166 12785309
Teresa F-EM Barrio E Querol A 2003 Analysis of the genetic variability in the species of the Saccharomyces sensu stricto complex Yeast 20 1213 1226 14587104
Mortimer RK Johnston JR 1986 Genealogy of principal strains of the yeast genetic stock center Genetics 113 35 43 3519363
Kunkee RE Bisson LF 1993 Wine-making yeasts Rose AH Harrison JS The yeasts. Volume 5, Yeast technology New York Academic Press 69 126
Hammond JRM 1993 Brewers yeast Rose AH Harrison JS The yeasts. Volume 5, Yeast technology New York Academic Press 7 67
Rose AH Vijayalakshmi G 1993 Baker's yeasts. In: Rose AH, Harrison JS, editors. The yeasts. Volume 5, Yeast technology New York Academic Press 357 397
Naumov GI 1996 Genetic identification of biological species in the Saccharomyces sensu stricto complex J Ind Appl Microbiol 17 295 302
McGovern PE 2003 Ancient wine: The search for the origins of viniculture Princeton (New Jersey) Princeton University Press 335 p.
Cavalieri D McGovern PE Hartl DL Mortimer R Polsinelli M 2003 Evidence for S. cerevisiae fermentation in ancient wine J Mol Evol 57 (Suppl 1) S226 S232 15008419
McGovern PE Zhang J Tang J Zhang Z Hall GR 2004 Fermented beverages of pre- and proto-historic China Proc Natl Acad Sci U S A 101 17593 17598 15590771
Azumi M Goto-Yamamoto N 2001 AFLP analysis of type strains and laboratory and industrial strains of Saccharomyces sensu stricto and its application to phenetic clustering Yeast 18 1145 1154 11536336
Hudson RR Kaplan NL 1985 Statistical properties of the number of recombination events in the history of a sample of DNA sequences Genetics 111 147 164 4029609
Kodama K 1993 Saké-brewing yeasts Rose AH Harrison JS The yeasts. Volume 5, Yeast technology New York Academic Press 129 168
Nei M 1987 Molecular evolutionary genetics New York Columbia University Press 512 p.
Phaff HJ Starmer WT 1987 Yeasts associated with plants, insects and soil Rose AH Harrison JS The yeasts. Volume 1, Biology of yeasts New York Academic Press 123 180
Naumov GI Naumova ES Sniegowski PD 1998
Saccharomyces paradoxus and Saccharomyces cerevisiae are associated with exudates of North American oaks Can J Microbiol 44 1045 1050 10029999
Sniegowski PD Dombrowski PG Fingerman E 2002
Saccharomyces cerevisiae and Saccharomyces paradoxus coexist in a natural woodland site in North America and display different levels of reproductive isolation from European conspecifics FEM Yeast Res 1 299 306
Johnson LJ Koufopanou V Goddard MR Hetherington R Schafer SM 2004 Population genetics of the wild yeast Saccharomyces paradoxus
Genetics 166 43 52 15020405
Naumov GI Naumova ES Sniegowski PD 1997 Differentiation of European and Far East Asian populations of Saccharomyces paradoxus by allozyme analysis Int J Syst Bacteriol 47 341 344 9103619
Ewing B Green P 1998 Base-calling of automated sequencer traces using phred. II. Error probabilities Genome Res 8 186 194 9521922
Stephens M Smith NJ Donnelly P 2001 A new statistical method for haplotype reconstruction from population data Am J Hum Genet 68 978 989 11254454
Rozas J Rozas R 1999 DnaSP version 3: An integrated program for molecular population genetics and molecular evolution analysis Bioinformatics 15 174 175 10089204
Drake JW 1991 A constant rate of spontaneous mutation in DNA-based microbes Proc Natl Acad Sci U S A 88 7160 7164 1831267
Kang XL Yadao F Gietz RD Kunz BA 1992 Elimination of the yeast RAD6 ubiquitin conjugase enhances base-pair transitions and G.C----T.A transversions as well as transposition of the Ty element: Implications for the control of spontaneous mutation Genetics 130 285 294 1311695
|
16103919
|
PMC1183524
|
CC BY
|
2021-01-05 08:00:24
|
no
|
PLoS Genet. 2005 Jul 25; 1(1):e5
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010005
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 10.1371/journal.pgen.001000605-PLGE-RA-0008R2plge-01-01-07Research ArticleDevelopmentEvolutionPlant ScienceEukaryotesPlantsArabidopsisDiversity of Flowering Responses in Wild Arabidopsis thaliana Strains Flowering Time Variation in A. thalianaLempe Janne 1Balasubramanian Sureshkumar 1Sureshkumar Sridevi 1Singh Anandita 1Schmid Markus 1Weigel Detlef 12*1 Department of Molecular Biology, Max Planck Institute for Developmental Biology, Tübingen, Germany
2 Plant Biology Laboratory, Salk Institute for Biological Studies, La Jolla, California, United States of America
Doebley John EditorUniversity of Wisconsin, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 25 7 2005 1 1 e620 1 2005 21 4 2005 Copyright: © 2005 Lempe et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Although multiple environmental cues regulate the transition to flowering in Arabidopsis thaliana, previous studies have suggested that wild A. thaliana accessions fall primarily into two classes, distinguished by their requirement for vernalization (extended winter-like temperatures), which enables rapid flowering under long days. Much of the difference in vernalization response is apparently due to variation at two epistatically acting loci, FRI and FLC. We present the response of over 150 wild accessions to three different environmental variables. In long days, FLC is among those genes whose expression is most highly correlated with flowering. In short days, FRI and FLC are less important, although their contribution is still significant. In addition, there is considerable variation not only in vernalization response, but also in the response to differences in day length or ambient growth temperature. The identification of accessions that flower relatively early or late in specific environments suggests that many of the flowering-time pathways identified by mutagenesis, such as those that respond to day length, contribute to flowering-time variation in the wild. In contrast to differences in vernalization requirement, which are mainly mediated by FRI and FLC, it seems that variation in these other pathways is due to allelic effects at several different loci.
Synopsis
Flowering is a quintessential adaptive trait in plants: Its correct timing ensures, for example, that plants do not produce seeds when they will not find favorable conditions for dispersal or germination. Befitting its importance, flowering is affected by several different environmental variables. The authors have compared the flowering times of over 150 Arabidopsis thaliana wild strains in response to three environmental factors: ambient growth temperature, day length, and vernalization (extended winter-like temperatures). Genetic and molecular analyses confirmed the important role of the previously identified FRI and FLC genes in flowering time. Genome-wide expression studies showed that FLC is among the genes whose expression is most highly correlated with flowering. Their studies, however, also revealed that the impact of FRI and FLC depends not only on vernalization treatment, which leads to repression of FRI and FLC activity, but also on day length. Within the groups of relatively early- and late-flowering strains, they find several unique responses, suggesting that many of the signaling pathways identified in mutant studies of laboratory strains are also being used to generate flowering diversity in the wild.
Citation:Lempe J, Balasubramanian S, Sureshkumar S, Singh A, Schmid M, et al. (2005) Diversity of flowering responses in wild Arabidopsis thaliana strains. PLoS Genet 1(1): e6.
==== Body
Introduction
Arabidopsis thaliana, a facultative long-day plant, occurs throughout the northern hemisphere, and wild accessions show extensive variation in several traits including flowering time [1]. The major environmental factors that control flowering time are light quantity (day length) and quality, ambient growth temperature, and vernalization. In general, long days, high ambient temperature, and vernalization promote flowering, and many genes with positive and negative roles in mediating these effects have been identified, primarily through the analysis of laboratory-induced mutations.
Analysis of wild accessions has shown that there are winter annuals, in which flowering is strongly delayed unless plants are vernalized, as well as rapidly cycling strains [2,3]. A major factor that prevents A. thaliana from flowering rapidly without vernalization is FRI (FRIGIDA) [4]. FRI delays flowering by maintaining high expression levels of FLC (FLOWERING LOCUS C), which encodes a MADS domain protein that represses flowering [5–7]. The widely used laboratory strains Landsberg erecta (Ler) and Columbia (Col-0) lack FRI activity because of deletions at the FRI locus. These two deletions are found in many other rapid-cycling accessions, indicating that FRI is a major determinant of natural variation in flowering time [4]. In addition, less commonly occurring deletions and point mutations in FRI have been reported [8–10]. Apart from loss of FRI function, attenuation of FLC activity provides an alternative mechanism to achieve rapid-cycling behavior in wild strains [8,11].
Natural variation in the flowering responses to other environmental variables has not been studied in as much detail as the vernalization response has been analyzed. A naturally occurring deletion in FLM, which causes early flowering in both long and short days [12], has so far only been found in accessions from Niederzenz [13]. Similarly, the EDI (EARLY DAY-LENGTH INSENSITIVE) allele of the photoreceptor CRY2 (CRYPTOCHROME 2) causes early flowering under short days, but appears to be restricted to accessions from the Cape Verde Islands [14]. It has been proposed that other CRY2 alleles affect flowering in a wide range of accessions, but experimental data supporting this suggestion are still lacking [15].
Despite the immense interest in exploiting natural variation to identify A. thaliana genes controlling flowering time, large studies that simultaneously compare the responses of wild accessions to different environments have been lacking. In three previous reports, summary statistics for short and long days, along with flowering times for about 70 strains in common garden experiments, were published [10,15,16]. In these 70 strains, FRI and FLC significantly delayed flowering only in long days. Here, we present detailed data on the flowering responses of over 150 wild strains along with several mutants under four different conditions. We report substantial variation in pathways other than those affected by vernalization, including the pathway that mediates responses to ambient temperature, which so far has not been studied in a range of natural accessions.
Results/Discussion
Extensive Variation in Flowering Responses of A. thaliana Accessions
We analyzed flowering time in a random set of accessions of single-seed descent, reflecting much of the geographic diversity represented in the European Arabidopsis Stock Centre (http://www.arabidopsis.info) at the time (Dataset S1; Table S1; Figure S1). We measured several traits, including days to flowering (DTF) and total leaf number (TLN), which were highly correlated under different conditions (r
2 = 0.6 to 0.9), with environment-dependent variation in growth rate found in only a few accessions. Heritability of TLN was higher than that of DTF (Table S2). We studied the effects of vernalization, ambient growth temperature, and photoperiod, using four different environments: 23 °C long days (23LD), 16 °C long days (16LD), 16 °C long days after 5 wk vernalization (16LDV), and 23 °C short days (23SD). DTF and TLN data from 155 wild strains, along with the common laboratory strains Col-0 and Ler and 32 flowering time mutants in Col-0 or Ler backgrounds, passed our quality controls in at least two conditions. A complete dataset across all environments was obtained from 177 strains.
In general, the distribution of flowering times reflected the action of genetic pathways known from the analysis of laboratory strains (Figure 1). In these, both long days and elevated temperatures accelerate flowering, and the wild accessions flowered on average earlier in 23LD compared to 23SD and 16LD. Similarly, a large tail of late-flowering accessions seen in 16LD largely disappeared upon vernalization. In contrast to the prominent peak of early-flowering accessions observed in 23LD (Figure 1A), flowering time in 23SD was more evenly distributed (Figure 1D).
Figure 1 Distribution of Flowering Times among Wild Accessions
(A–D) Distribution of flowering times expressed as TLN in23LD, 16LD, 16LDV, and 23SD. White and black bars represent accessions with and without Col- or Ler-type deletions in FRI, respectively. Gray bars indicate strains subsequently identified to have defects in FRI or FLC.
(E) Mean flowering times of putatively FRI/FLC functional (black bars) and nonfunctional (white bars) accessions in four environments.
Having found substantial variation in the flowering time of accessions, we asked how similar the accessions were in their responses to environmental cues. The degree of genetic correlation between different environments indicated that the largest differences were in the response to vernalization, and the smallest in the response to ambient temperature (Table 1). Genotype-by-environment interactions accounted for 27% of flowering-time variation in response to vernalization (16LDV vs. 16LD); 9% in response to photoperiod (23LD vs. 23SD); and 3% in response to ambient temperature (23LD vs. 16LD), all of which were significant at p < 0.0001. The variation in sensitivity to different environmental factors was obvious when flowering times were regressed on the environmental mean (Figure 2) [17]. Accessions also differed extensively in other traits related to flowering, such as juvenile, adult, and cauline leaf number, indicating further differences in developmental physiology (see Table S1). Taken together, these results demonstrate that there is extensive variation in the responses of accessions to several environmental cues.
Table 1 Genetic Correlation across Environments
95% confidence intervals in parentheses.
Figure 2 Environmental Responses Plotted against TLN
Environmental responses were calculated from regression of TLN onto environmental means. (A) 23LD vs. 16LD; (B) 16LD vs. 16LDV; (C) 23LD vs. 23 SD; 95% of all values are between the gray dotted lines. The means are not centered because the responses are not normally distributed.
Co-variation between environmental factors and a particular trait can be evidence for selection. Environmental factors such as light and temperature, in turn, are strongly correlated with latitude, and latitudinal clines for different phenotypes are not uncommon [18]. Previously, a strong correlation between flowering time and latitude was demonstrated for accessions that were sown in fall and overwintered in a common garden in Rhode Island. Among strains with apparently functional FRI/FLC (see below), this correlation was particularly high (r
2 = 0.38, n = 21) [10].
In our set of accessions, we found the strongest correlation between latitude and flowering (TLN) in 16LDV (Figure 3A; r
2 = 0.18, p < 0.0001), suggesting that the latitude correlation in the aforementioned common garden experiment is dependent on vernalization during natural winter conditions. This correlation is even higher in the group of accessions that was shared between our study and that of Stinchcombe and colleagues [10], namely r
2 = 0.38, n = 9. Thus, the differences in the strength of the latitudinal cline that we and Stinchcombe and colleagues [10] observed might be due to different sampling biases.
Figure 3 A Latitudinal Cline in Flowering after Vernalization
Correlation of latitude with (A) flowering time, (B) and vernalization sensitivity (expressed as the regression coefficient of TLN on 16LD and 16LDV grand means). Both accessions with putatively functional alleles (black dots) and nonfunctional alleles (grey dots) at FRI/FLC are shown. Regression line with only FRI/FLC functional accessions (black line) and with all accessions (grey line) is shown separately. FRI/FLC functional accessions are indicated by black dots, FRI/FLC nonfunctional accessions by hollow circles. The correlation of flowering in FRI/FLC functional accessions with latitude (black line; r
2 = 0.18) is higher than of all accessions (grey line; r
2 = 0.13).
Consistent with flowering time being correlated with the magnitude of the vernalization response, there is also a significant latitudinal cline in the vernalization sensitivity of accessions (Figure 3B; r2 = 0.12, p < 0.0001). Because there was very little correlation between latitude and longitude for the origin of strains (r
2 = 0.035), we did not include longitude as a covariant in our analyses. Although population structure may be a confounding factor in evaluating latitudinal clines, this concern is somewhat mitigated by the relatively small amount of population structure in A. thaliana [19,20].
The Contribution of FLC Activity to Flowering-Time Variation
In both laboratory-induced mutant strains and natural accessions, there is broad correlation of FLC expression levels, as determined by RNA blots, and flowering time, such that plants flower late when FLC levels are high and early when FLC levels are low [5,6,8,11,21]. However, a recent study of 17 accessions revealed also considerable variation in FLC expression among late-flowering accessions [22]. To evaluate more precisely the relationship between absolute FLC levels and quantitative variation in flowering time, we analyzed FLC expression by qRT-PCR in 149 accessions grown at 23LD (Figure 4A). Across all strains, variation in FLC levels accounted for 42% and 40% of TLN and DTF variation (r
2), respectively (p < 0.0001). As expected, this is largely due to FLC mediating the vernalization response, which can be calculated as the relative reduction in TLN or DTF when 16LDV and 16LD are compared. The correlation of FLC expression level with vernalization response using TLN and DTF is 47% and 52%, respectively.
Figure 4 Correlation of FLC Expression with Flowering Time
(A) Correlation of FLC RNA levels and DTF in 149 accessions, as measured by qRT-PCR. Hollow circles indicate accessions with lesions in FRI.
(B) Correlation of expression levels of 68 flowering regulators with flowering time. Arrow indicates FLC.
(C) Correlation of expression estimates (expressed as log2) of 22,500 genes represented on Affymetrix ATH1 array with flowering times across 34 accessions. Arrow indicates FLC, and line indicates distribution from 1,000 permutations.
In other members of the Brassicaceae family, loci controlling flowering time co-localize with FLC orthologs [23], but the relationship between expression of different FLC paralogs and flowering time appears to be more complex than in A. thaliana [24,25]. Furthermore, FLC expression in Brassica oleracea var. capitata shows differential responses to vernalization, depending on the tissue examined [26].
To determine whether the substantial correlation between FLC expression levels and flowering time across 149 A. thaliana accessions was unusual compared to other genes known to control flowering, or even to the remainder of the transcriptome in general, we examined Affymetrix ATH1 array data for 34 accessions, which had been generated as part of the AtGenExpress project (Table S3). The correlation of FLC with TLN and DTF at 23LD that we obtained using array analyses was 37% and 42%, respectively, which is quite similar to the 42% and 40% estimated from qRT-PCR analyses. We found that, among a set of 68 floral regulators, FLC was the one most highly associated with either TLN or DTF (Figure 4B; Table S4). In addition, when examining other conditions, we found none of the other floral regulators were as highly correlated with flowering time as FLC was for 23LD.
When we considered all genes represented on the ATH1 array, we found FLC to be among the ten most highly correlated genes at 23LD (both positively and negatively correlated) (Figure 4C). This was true regardless of whether expression was regressed linearly or logarithmically onto flowering time, and whether Pearson or Spearman rank correlation was calculated. With respect to DTF, FLC levels were considerably more correlated with this trait than expression levels of any other gene. We determined significance by permutation analysis, and found that FLC was the only gene significantly correlated with DTF at a p < 0.05 level under both Pearson and Spearman rank correlation. The prominence of FLC, whose levels are highly correlated with vernalization response, may reflect the fact that this response is the most important variable affecting flowering time in A. thaliana accessions. We note, however, that the analysis was biased, because we analyzed young seedlings, a stage in which FLC levels are particularly high.
There was a general association of low FLC levels with the Col- and Ler-type deletions in FRI and early flowering, but this relationship was not absolute. Several exceptions to this rule are explained by other nonfunctional FRI alleles or variation in FLC itself, as shown by a combination of genetic and molecular analyses (Table 2; Figure 5; Figure S2). Several of the early accessions have missense mutations in FRI that are different from those described before [8,9], whereas one nonfunctional FRI allele has extensive polymorphisms (Figure 5A). Because the effects of FRI on flowering time are entirely dependent on FLC [7], we were also curious whether there are any molecular effects of FRI in the absence of FLC. To this end, we compared the global expression profile at the shoot apex in response to floral induction by long days between FRI flc-3 and fri-Col flc-3 plants, similar to a design used before [27]. We detected no significant differences between the two genotypes (data not shown), confirming that FRI function absolutely requires FLC activity.
Table 2 Accessions with Compromised FRI or FLC Activity
aThe first 19 accessions (all but the last) were analyzed because of early flowering.
bThe new FLC allele in this strain was found during FLC RT-PCR analyses of all strains.
n/d, no data.
Figure 5 Newly Identified Alleles of FRI and FLC
(A) Diagrams of FRI alleles. Exons are represented by boxes, introns by lines. Black lines indicate nonsynonymous changes, gray lines synonymous changes as well as polymorphisms in the non-coding region compared to the reference sequence of H51 [4]. Insertions and deletions (indels) are designated by triangles, numbers indicate size in base pairs (bp). Premature stop codons are caused by a 1-bp insertion in Ang-1, single bp polymorphisms in BG1 and Ri-0, or a deletion and inversion (gray box) in Pog-0. Note extensive polymorphisms in Jl-1. BG4, BG6, BG9, and HR5 were only partially sequenced.
(B) Changes in FLC transcripts in three accessions. Dotted lines indicate exons that are missing in part or completely. These deletions cause frame shifts and thus altered amino acid sequences (grey boxes). Premature stop codons are indicated. The reasons for the aberrant transcript processing in Ll-2 are not known. In Cen-0, an alternative splice acceptor site in the last exon is used, leading to a deletion of exon sequences and a frame shift, whereas in Cal-0 an alternative splice acceptor site in the last intron is used, which adds additional sequences and also causes a frame shift of sequences of the last exon.
(C) Large insertions in the first intron of FLC in several accessions.
Although null alleles of FRI are easily found in natural accessions [4,9], only transposon insertions with reduced expression level of a wild-type transcript have been described for FLC [6,8,11,16], which has led to the proposal that null alleles of FLC have reduced fitness in the wild [3]. However, we discovered three natural FLC alleles with severely affected protein function (Figure 5B). At least one of the very early accessions, Ll-2, likely carries a null allele, because transgenic overexpression of the Ll-2 FLC cDNA, which lacks exons 2 to 6, in an flc-3 background has no effect on flowering (not shown). These findings demonstrate that FLC alleles with severely compromised protein function can be recovered from wild populations. We also found several new transposon insertions in the first intron of FLC (Figure 5C).
Variation in Responses to Photoperiod and Ambient Temperature
In total, our collection of 155 strains included 67 strains that carry either the common Col- or Ler-type lesions in FRI, and at least 23 accessions with alternative FRI/FLC lesions. A two-way analysis of variance (ANOVA) model showed that FRI and FLC can account for 63% of variation in 23LD TLN. In contrast, FRI and FLC accounted for only 23% variation in 23SD, consistent with this pathway being most important when other floral inductive stimuli are strong, as in long days. Nevertheless, our finding that FRI and FLC significantly affect flowering in short days in a large set of wild strains is at variance with previous surveys that did not find a significant effect of FRI [10,16]. Twenty-four wild strains are in common between the previous studies and our study. A comparison of flowering times indicates limited correlation in these 24 strains (r
2 = 0.55 for long days, 0.35 for short days). Consistent with the differences in observed flowering time, FRI and FLC are significant factors in our long day and short day conditions in this set of 24 strains, indicating that these discrepancies are not due to sampling bias, but different experimental conditions. Our results are consistent with earlier studies using FRI/FLC introgressed laboratory strains, which also displayed late flowering in short days [28,29]. In addition, we did not limit our assessment of FRI and FLC activity to genotyping of major, previously known alleles.
It has been suggested that a disadvantage of growth chamber experiments compared to common garden experiments is an increased contribution of random environmental variation, which may obscure ecologically relevant responses [10,30]. We believe that we have been able to limit random environmental variation with very accurate growth facilities along with a completely randomized experimental design (see Materials and Methods).
Consistent with FRI and FLC mediating most of the vernalization response, there is no significant difference between groups with and without FRI/FLC activity in 16LDV, suggesting that our vernalization treatment is largely saturating (see Figure 1E). However, although vernalization strongly reduces late flowering, it does not abolish variation in flowering time, which still varies more than 4-fold (see Figure 1C).
One of our long-term goals is to identify genes and pathways that contribute to vernalization-independent variation in flowering time of natural accessions. To identify candidate accessions with interesting variation in flowering responses, we applied hierarchical clustering, which has been used before to group accessions according to light sensitivity during seedling development [31] (Figures 6A–6C, S3–S5). Potentially interesting accessions were also identified as those having large residuals when comparing contrasting conditions, such as 23LD versus 16LD, 23LD versus 23SD, and 16LD versus 16LDV. These specific comparisons yielded similar results to hierarchical clustering (Figure 6D–6G).
Figure 6 Variation in Responses to Environmental Cues Other than Vernalization
(A–C) Hierarchical clustering identifies groups with different flowering behaviors (see Figure S3 for entire cluster diagram). Green and red indicate earlier and later flowering than the mean, respectively. Gray indicates missing data. Conditions are, from left to right: 16LD, 23LD, 16LDV, 23SD; with six columns each (DTF, juvenile rosette leaf number, adult rosette leaf number, rosette leaf number, cauline leaf number, TLN).
(A) Accessions that cluster with photoperiodic mutants.
(B) Accessions that cluster with mutants of the autonomous pathway.
(C) Accessions that flower late in 16LD.
(D–G) Comparison of specific accessions with laboratory strains Col-0 and Ler.
(D) Accessions that are temperature-insensitive and not delayed in 16LD (black) compared to 23LD (white bars).
(E) Accessions that do not respond to vernalization and flower similarly in 16LD (black) and 16LDV (white).
(F) Accessions that flower late in long days (23LD, black), but not short days (23SD, white). For comparison, the photoperiodic co-1 mutant is included.
(G) Accessions that flower early in short days (23SD, white) compared to long days (23LD, black).
To associate accessions with variation in specific genetic pathways, the cluster analysis included a large number of known flowering-time mutants. The first major distinction was between accessions that flowered either relatively early or late in long days, with the first group comprising accessions that carry lesions at FRI or FLC. The second cluster consisted of late-flowering, vernalization-responsive accessions along with strains that behaved similarly to very late-flowering photoperiodic mutants such as constans (co) or gigantea (gi) [32,33]. Because all the mutants we examined had been isolated from backgrounds that lacked FRI function, we independently clustered rapid-cycling accessions with laboratory-induced mutants (see Figure S4), which resulted in several major groups, two of which were similar to photoperiodic mutants in that there was relatively little difference in flowering under long and short days. One was relatively late in long days (Figure 6F), whereas the other was relatively early in short days (Figure 6G). Another group behaved similarly to autonomous pathway mutants and included accessions that were rather late in 16LD and 23LD, but not in 23SD (Figure 6B).
We looked specifically for variation among putatively FRI/FLC functional accessions, by clustering them together with some of the accessions that we had identified as lacking FRI/FLC activity. This resulted in three major groups, which differ in their vernalization sensitivity (Figure S5). Several accessions from lower latitudes were found in a class that becomes particularly early after vernalization (Figure S5). This observation, although consistent both with our observation of a latitudinal cline in vernalization sensitivity (Figure 5B) and with earlier findings [10], does not imply that FRI promotes flowering after vernalization, but rather may reflect that other mechanisms, which otherwise delay flowering, are dispensable in this cluster of accessions.
As a first step toward identifying the genetic mechanisms underlying the newly identified flowering behaviors, we examined F2 populations from crosses of Ei-6 and Fr-2, two strains that flower relatively early in 23SD, to the reference strain Col-0. In both cases, the earliness appeared to be recessive compared to Col-0 (Figure 7A). Consistent with the recessive nature of earliness, neither strain carries the naturally occurring, dominantly acting EDI allele of CRY2, which can cause early flowering in short days accession [14]. In addition, when we crossed Ei-6 and Fr-2 to each other, the F1 hybrids were later than either parent, indicating that earliness in short days has a different genetic basis in these strains (Figure 7A). The F2 progeny from a cross of Fr-2 to Col-0 included a distinct early class, suggesting that early flowering is due to variation at a major locus (Figure 7B). In contrast, in the Ei-6 × Col-0 cross, continuous variation can be seen, indicating the involvement of multiple loci (Figure 7C).
Figure 7 Genetic Analysis of Two Accessions that Flower Early in 23SD
(A) Flowering time of accessions and F1 progeny.
(B) A distinct class of early-flowering plants in the Fr 2 × Col-0 F2 population.
(C) Continuous segregation of flowering times in Ei-6 × Col-0 F2 population.
Conclusions
We provide an extensive dataset on flowering-time associated traits in a large number of A. thaliana accessions across multiple environments, which forms a valuable resource for both experimental and population genetic studies. We have tested the response to three major environmental variables affecting flowering time: photoperiod, vernalization, as well as ambient growth temperature, whose effects had not been studied before in natural strains. Although our data confirm that FRI and FLC are indeed major factors regulating the rapid-cycling or winter-annual habit, loss of FRI and FLC function cannot explain all of the variation observed. Some of the remaining variation may be due to more subtle differences in the activity of the vernalization pathway.
Although there is less variation in the ambient temperature response compared to vernalization and photoperiod, we could identify accessions that responded more strongly or more weakly to temperature than the majority of accessions (see Figure 6C and 6D). The identification of accessions that flower relatively early or late in specific environments suggests that many of the flowering-time pathways identified by mutagenesis, such as those that respond to photoperiod or temperature, participate in generating flowering-time variation in the wild. In contrast to variation in vernalization requirement, which appears to be largely due to FRI and FLC, it seems that variation in these other pathways is due to allelic effects at several different loci, a conclusion that is supported by a growing number of other studies [13,14,22]. The hypothesis of a complex genetic basis of non-vernalization variation is further supported by the finding that, in contrast to vernalization response, none of the known flowering-time regulators are particularly highly correlated with photoperiod or temperature response. It will be interesting to learn which genes control variation in these other pathways.
Materials and Methods
Stocks.
Accessions were obtained from the Nottingham Arabidopsis Stock Centre (http://www.arabidopsis.info). Strains with different combinations of FRI and FLC alleles have been described [6,7,28].
Growth conditions.
Plants were grown in controlled growth rooms with a temperature variability of about plus or minus 0.1 °C under a 1:1 mixture of Cool White and Gro-Lux Wide Spectrum fluorescent lights, with a fluence rate of 125 to 175 μmol m−2 s−1. All light bulbs were of the same age. Long days had 16 h of light, short days had 8 h. Maximal humidity was 65%. Light, temperature, and humidity were continuously monitored online and logged data were stored in a Structured Query Language (SQL) database.
Plant culture.
Seeds were stratified at 4 °C for 7 d (to minimize variation due to differences in stratification requirements) in 0.1% agarose, then planted on soil. Before vernalization, seed germination was induced at 16 °C for 24 h. Plants were cultivated for 5 wk at 4 °C in a vernalization room with 8 h light of about 50 μmol m−2 s−1, before transfer to normal growth rooms.
Experimental design.
Twelve seeds per genotype were sown in a completely randomized design in individually numbered positions of multiple-well flats. If more than one seed had been deposited at a certain position, only one plant was allowed to remain after 1 wk. Flats were watered on alternating days and rotated every time to minimize the effects of positional light or stagnant water. Only genotypes with at least four germinated plants in a given environment were included in further analysis. Typically, eight to ten plants per genotype and condition were analyzed. For the first set of experiments (which included mostly the vernalization-requiring accessions), seeds from the stock center were used. For the second set of experiments, progeny derived by single-seed descent were used. We analyzed the efficiency of randomization by observing the total average across shelves. There was no statistically significant variation across the shelves.
Scoring of traits.
The following traits were scored: DTF, TLN, juvenile leaf number (JLN), adult leaf number (ALN), cauline leaf number (CLN), and rosette leaf number (RLN). Flowering was scored on a daily basis for first macroscopic appearance of the inflorescence. DTF was calculated from the date of sowing. For vernalized plants, the vernalization period was subtracted. In our conditions, plants did not produce leaves during vernalization. Individual plants that had flowered were removed. Flats remained in their specific conditions throughout the entire experiment. Leaves were counted under the microscope, and juvenility was determined by the presence of trichomes on the abaxial side.
Quality control.
As a first step of quality control, we analyzed variation between identical genotypes that had been included in both sets of experiments. No statistically significant variation could be detected. Therefore, we pooled the two datasets and treated them as a single experiment for further analysis. Some accessions displayed unusually large variation, suggestive of segregation or seed contamination. We therefore removed those genotypes from the analysis.
Statistical analyses.
Data were analyzed using JMP (version 5.1, SAS Institute, Cary, North Carolina, United States), the “base” package of R (http://www.R-project.org) [34], or Excel (Microsoft, Redmond, Washington, United States). The total variance was partitioned into between-line variance and the residuals with a one-way ANOVA using phenotype as a response and accession as a single factor of random effect. The variance component was estimated using a restricted maximum likelihood method (REML). Broad-sense heritability (H
2) was calculated as between-line variance (V
G) divided by total variance. The coefficient of genetic variation (CV
G) was calculated as (100 × )/mean for each trait. Genetic correlations (r
GE) were calculated from covariance and variance components estimated through REML, using those genotypes for which data across all environments were available. The significance of each genetic correlation was determined using a t-test after Z transformation of the correlation coefficient [35]. Genotype-by-environment interactions were tested using a two-way ANOVA with strain and conditions as classifying factors. Hierarchical cluster analysis was performed with Cluster 3.0 [36]. The data were normalized to the overall mean, followed by log transformation. Noncentric, average linkage (UPGMA) clustering was then performed using Pearson correlation. To evaluate environmental sensitivity, TLN was regressed onto the environmental mean, [log(TLNenvironmentA) − log(TLNenvironmentB)]/[log(mean TLNenvironmentA) − log(mean TLNenvironmentB)] [17].
DNA analyses.
DNA was extracted using the CTAB method [27] with minor modifications. For sequence analysis of FRI and FLC, 3.3 kb and 4.2 kb genomic fragments, respectively, were amplified using ExTaq polymerase (TaKaRa Biomedical, Shiga, Japan). Pooled products from four independent PCR reactions were directly sequenced on both strands. See Table S5 for oligonucleotide primers used.
Expression studies.
The aerial parts of plants grown on soil in 23LD for 12 d were collected at dusk and flash frozen in liquid nitrogen. RNA was extracted using Trizol, and 1 μg of total RNA was reverse transcribed using a Reverse Transcription kit (Invitrogen, Carlsbad, California, United States). See Table S5 for a list of oligonucleotide primers. PCR was performed in the presence of SYBR Green (Invitrogen), and amplification was monitored in real time with the Opticon Continuous Fluorescence Detection System (MJ Research, Reno, Nevada, United States). Two biological replicates, each with two technical replicates, were analyzed. Threshold cycles (cT) were based on a reaction reaching a specific fluorescence intensity in the log-linear phase of the amplification curve. ΔcT was calculated as the difference in cT between FLC and UBQ10. PCR efficiency was assumed to be the same and relative transcript abundance compared to Col-0 was calculated as 2−ΔΔcT.
Microarray data were generated as part of the AtGenExpress project (http://arabidopsis.org/info/expression/ATGenExpress.jsp) (see Table S3). RNA isolated from the aerial parts of 4-d-old seedlings was hybridized to Affymetrix ATH1 arrays as described [27], and normalized expression estimates were obtained using gcRMA (bioconductor.org), a modification of the robust multi-array analysis (RMA) algorithm [37]. Microarray data have been deposited with the ArrayExpress database (http://www.ebi.ac.uk/arrayexpress; experiment E-TABM-18 and E-TABM-19). In addition, normalized microarray data are available from our website (http://www.weigelworld.org/resources/microarray).
Supporting Information
Dataset S1 Means of Flowering Time Traits in Accessions and Mutant Strains
Data file of flowering-time-related traits in accessions and mutants. Also available as a PDF file (Table S1).
(38 KB CSV)
Click here for additional data file.
Figure S1 Geographic Distribution of Wild Strains Used in the Survey
Open circles indicate known FRI/FLC defects.
(47 KB PDF)
Click here for additional data file.
Figure S2 Genetic Analysis of Rapid-Cycling Accessions
Flowering times of F1 progeny from crosses of accessions that do not carry Ler- or Col-type deletions in FRI to fri-Col flc-3 (white), FRI-Sf2 flc-3 (gray), fri-Col FLC (black), and FRI-Sf2 FLC (striped).
(A) F1 progeny in which FRI causes late flowering.
(B) F1 progeny in which FLC causes late flowering.
(C) F1 progeny in which only simultaneous introduction of FRI and FLC causes late flowering.
(78 KB PDF)
Click here for additional data file.
Figure S3 Hierarchical Cluster Analysis of All Genotypes
Columns for each condition are DTF, JLN, ALN, RLN, CLN, and TLN. Hierarchical clustering identifies groups with different flowering behaviors. Green indicates earlier flowering than the mean, red later flowering than the mean, and gray indicates missing data.
(110 KB PDF)
Click here for additional data file.
Figure S4 Hierarchical Clustering of Accessions Lacking Functional FRI or FLC
(A) Accessions that are similar to Col.
(B) Accessions that are similar to Ler, with a reduced adult phase at 16LD.
(C) Accessions that are severely delayed in 16LD compared to 23LD.
(D) Accessions that cluster together with mutants of the photoperiodic pathway. This group includes several accessions that flower early in 23SD.
(E) Accessions that cluster with mutants of the autonomous pathway.
(950 KB PDF)
Click here for additional data file.
Figure S5 Hierarchical Clustering of Accessions with Functional FRI/FLC Pathway
(Top) Accessions with moderate vernalization response.
(Middle) Accessions with extreme vernalization response. Several accessions in this cluster originate from lower latitudes (given in parentheses).
(Bottom) Accessions that are delayed after vernalization, possibly indicating FRI-independent variation. Note that BG6, HR5, and Wil-1, which lack functional FRI, cluster with this group.
(594 KB PDF)
Click here for additional data file.
Table S1 Means of Flowering Time Traits in Accessions and Mutant Strains
A, B, C, etc., denote multiple stocks and repeated assays. Latitude information was obtained from Geographic Names Information System (US Geological Survey).
ALN, adult rosette leaf number; CLN, cauline leaf number; DTF, days to flowering; JLN, juvenile rosette leaf number; TLN, total leaf number.
The data are also available as a CSV file (Dataset S1).
(62 KB PDF)
Click here for additional data file.
Table S2 Summary Statistics Based on Individual Plants
Asterisk indicates ± 95% confidence intervals (2 × SEM).
(31 KB PDF)
Click here for additional data file.
Table S3 Key to ATH1 Array Data
(33 KB PDF)
Click here for additional data file.
Table S4 Pearson Correlation of Known Flowering Regulators with DTF and TLN in 23LD
(33 KB PDF)
Click here for additional data file.
Table S5 Oligonucleotide Primers Used
Uses: 1/3, amplification of 3.3-kb FRI genomic region; 1–13, FRI sequence analysis; 12/13, PCR genotyping of Col-type deletion in FRI; 3/4, PCR genotyping of Ler-type deletion in FRI; 6/7, FRI qRT-PCR; 30/31 and 28/34, amplification of 4.2-kb FLC genomic region (divided into two fragments); 14–35, FLC sequence analysis; 34/35, FLC cDNA amplification; 36–38, PCR genotyping of Ler-type insertion in FLC; 39/40, UBQ10 qRT-PCR; 41/42, FLC qRT-PCR.
(36 KB PDF)
Click here for additional data file.
Accession Numbers
GenBank (http://www.ncbi.nlm.nih.gov/Genbank/) accession numbers of the new sequences discussed in this paper are AY964090–AY964098 and AY0528–AY970557.
We thank the Nottingham Arabidopsis Stock Centre and Rick Amasino for seed stocks; Josip Perkovic for technical assistance; Jan Lohmann for help with microarray experiments; Norman Warthmann for help with programming in R; and John Stinchcombe for providing the flowering data referred to in his Proceedings of the National Academy of Science of the United States (2004) paper [10]. We thank Kirsten Bomblies, Justin O. Borevitz, Joanne Chory, Richard Clark, Vava Grbic, Yasushi Kobayashi, Julin Maloof, Norman Warthmann, and Jonathan Werner for discussion and critical reading of the manuscript; and two anonymous reviewers for insightful comments. We acknowledge the use of microarray data produced in the context of the AtGenExpress project, which is coordinated by Lutz Nover (Frankfurt), Thomas Altmann (Potsdam), and D.W., and supported by funds from the Deutsche Forschungsgemeinschaft (DFG) and the Max Planck Society. Our work was also supported by a European Molecular Biology Organization (EMBO) Long-Term Fellowship to S.B., a National Institutes of Health (NIH) grant (GM62932) to D.W., and by the Max Planck Society, of which D.W. is a director.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. SB and DW conceived and designed the experiments. JL, SB, SS, AS, and MS performed the experiments. JL, SB, and DW analyzed the data and wrote the paper.
Abbreviations
Col-0Columbia
DTFdays to flowering
TLN; total leaf number
LerLandsberg erecta
16LD16 °C long days
16LDV16 °C long days after 5 wk vernalization
23LD23 °C long days
23SD23 °C short days
==== Refs
References
Alonso-Blanco C Koornneef M 2000 Naturally occurring variation in Arabidopsis : An underexploited resource for plant genetics Trends Plant Sci 5 22 29 10637658
Sung S Amasino RM 2004 Vernalization and epigenetics: How plants remember winter Curr Opin Plant Biol 7 4 10 14732435
Koornneef M Alonso-Blanco C Vreugdenhil D 2004 Naturally occurring genetic variation in Arabidopsis thaliana
Annu Rev Plant Physiol Plant Mol Biol 55 141 172
Johanson U West J Lister C Michaels S Amasino R 2000 Molecular analysis of FRIGIDA , a major determinant of natural variation in Arabidopsis flowering time Science 290 344 347 11030654
Sheldon CC Burn JE Perez PP Metzger J Edwards JA 1999 The FLF MADS box gene: a repressor of flowering in Arabidopsis regulated by vernalization and methylation Plant Cell 11 445 458 10072403
Michaels SD Amasino RM 1999
FLOWERING LOCUS C encodes a novel MADS domain protein that acts as a repressor of flowering Plant Cell 11 949 956 10330478
Michaels SD Amasino RM 2001 Loss of FLOWERING LOCUS C activity eliminates the late-flowering phenotype of FRIGIDA and autonomous pathway mutations but not responsiveness to vernalization Plant Cell 13 935 941 11283346
Gazzani S Gendall AR Lister C Dean C 2003 Analysis of the molecular basis of flowering time variation in Arabidopsis accessions Plant Physiol 132 1107 1114 12805638
Le Corre V Roux F Reboud X 2002 DNA polymorphism at the FRIGIDA gene in Arabidopsis thaliana : Extensive nonsynonymous variation is consistent with local selection for flowering time Mol Biol Evol 19 1261 1271 12140238
Stinchcombe JR, Weinig C, Ungerer M, Olsen KM, Mays Cet al. 2004 A latitudinal cline in flowering time in Arabidopsis thaliana modulated by the flowering time gene FRIGIDA
Proc Natl Acad Sci U S A 101 4712 4717 15070783
Michaels SD He Y Scortecci KC Amasino RM 2003 Attenuation of FLOWERING LOCUS C activity as a mechanism for the evolution of summer-annual flowering behavior in Arabidopsis
Proc Natl Acad Sci U S A 100 10102 10107 12904584
Scortecci KC Michaels SD Amasino RM 2001 Identification of a MADS-box gene, FLOWERING LOCUS M , that represses flowering Plant J 26 229 236 11389763
Werner JD Borevitz JO Warthmann N Trainer GT Ecker JR 2005 Quantitative trait locus mapping and DNA array hybridization identify an FLM deletion as a cause for natural flowering-time variation Proc Natl Acad Sci U S A 102 2460 2465 15695584
El-Assal SE-D Alonso-Blanco C Peeters AJ Raz V Koornneef M 2001 A QTL for flowering time in Arabidopsis reveals a novel allele of CRY2
Nat Genet 29 435 440 11726930
Olsen KM Halldorsdottir SS Stinchcombe JR Weinig C Schmitt J 2004 Linkage disequilibrium mapping of Arabidopsis CRY2 flowering time alleles Genetics 167 1361 1369 15280248
Caicedo AL Stinchcombe JR Olsen KM Schmitt J Purugganan MD 2004 Epistatic interaction between Arabidopsis
FRI and FLC flowering time genes generates a latitudinal cline in a life history trait Proc Natl Acad Sci U S A 101 15670 15675 15505218
Falconer DS Mackay TFC 1996 Introduction to quantitative genetics, 4th ed Harlow (United Kingdom) Addison Wesley Longman 463 p.
Endler JA 1977 Geographic variation, speciation, and the clines Princeton (New Jersey) Princeton University Press 262 p.
Sharbel TF Haubold B Mitchell-Olds T 2000 Genetic isolation by distance in Arabidopsis thaliana : Biogeography and postglacial colonization of Europe Mol Ecol 9 2109 2118 11123622
Nordborg M Borevitz JO Bergelson J Berry CC Chory J 2002 The extent of linkage disequilibrium in Arabidopsis thaliana
Nat Genet 30 190 193 11780140
Sheldon CC Rouse DT Finnegan EJ Peacock WJ Dennis ES 2000 The molecular basis of vernalization: the central role of FLOWERING LOCUS C (FLC)
Proc Natl Acad Sci U S A 97 3753 3758 10716723
Werner JD Borevitz JO Uhlenhaut NH Ecker JR Chory J 2005
FRIGIDA –independent variation in flowering time of natural A. thaliana accessions. Genetics E-pub ahead of print DOI: 10.1534/genetics.104.036889
Kole C Quijada P Michaels SD Amasino RM Osborn TC 2001 Evidence for homology of flowering-time genes VFR2 from Brassica rapa and FLC from Arabidopsis thaliana Theor Appl Genet 425–430
Tadege M Sheldon CC Helliwell CA Stoutjesdijk P Dennis ES 2001 Control of flowering time by FLC orthologues in Brassica napus
Plant J 28 545 553 11849594
Pires JC Zhao J Schranz ME Leon EJ Quijada PA 2004 Flowering time divergence and genomic rearrangements in resynthesized Brassica polyploids (Brassicaceae) Biol J Linnean Soc 82 675 688
Lin SI Wang JG Poon SY Su CL Wang SS 2005 Differential regulation of FLOWERING LOCUS C expression by vernalization in cabbage and Arabidopsis Plant Physiol 137 1037 1048 15734903
Schmid M Uhlenhaut NH Godard F Demar M Bressan R 2003 Dissection of floral induction pathways using global expression analysis Development 130 6001 6012 14573523
Lee I Bleecker A Amasino R 1993 Analysis of naturally occurring late flowering in Arabidopsis thaliana
Mol Gen Genet 237 171 176 8455554
Lee I Amasino RM 1995 Effect of vernalization, photoperiod, and light quality on the flowering phenotype of Arabidopsis plants containing the FRIGIDA gene Plant Physiol 108 157 162 12228459
Endler JA
1986
Natural selection in the wild
In: Monographs in Population Biology
21 Levin SA, Horn HS, editors.
Princeton (New Jersey)
Princeton University Press
354
p.
Maloof JN Borevitz JO Dabi T Lutes J Nehring RB 2001 Natural variation in light sensitivity of Arabidopsis
Nat Genet 29 441 446 11726931
Mouradov A Cremer F Coupland G 2002 Control of flowering time: Interacting pathways as a basis for diversity Plant Cell 14 S111 130 12045273
Simpson GG Dean C 2002
Arabidopsis , the Rosetta stone of flowering time? Science 296 285 289 11951029
Ihaka R Gentleman R 1996 R: A language for data analysis and graphics J Comput Graph Stat 5 299 314
Sokal RR Rohlf FJ 1981 Biometry, 2nd ed New York WH Freeman and Company 859 p.
Eisen MB Spellman PT Brown PO Botstein D 1998 Cluster analysis and display of genome-wide expression patterns Proc Natl Acad Sci U S A 95 14863 14868 9843981
Irizarry RA Bolstad BM Collin F Cope LM Hobbs B 2003 Summaries of Affymetrix GeneChip probe level data Nucleic Acids Res 31 e15 12582260
|
16103920
|
PMC1183525
|
CC BY
|
2021-01-05 08:00:57
|
no
|
PLoS Genet. 2005 Jul 25; 1(1):e6
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010006
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 10.1371/journal.pgen.001000705-PLGE-RA-0018R2plge-01-01-08Research ArticleCell BiologyNeuroscienceGenetics/Genetics of DiseaseGenetics/Disease ModelsMus (Mouse)Cell-Autonomous Death of Cerebellar Purkinje Neurons with Autophagy in Niemann-Pick Type C Disease Cell-Autonomous Neurodegeneration in NPCKo Dennis C Milenkovic Ljiljana Beier Steven M Manuel Hermogenes Buchanan JoAnn Scott Matthew P *Departments of Developmental Biology, Genetics, and Bioengineering, Howard Hughes Medical Institute, Stanford University School of Medicine, Stanford, California, United States of AmericaOrr Harry EditorUniversity of Minnesota, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 25 7 2005 1 1 e77 2 2005 17 5 2005 Copyright: © 2005 Ko et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Niemann-Pick type C is a neurodegenerative lysosomal storage disorder caused by mutations in either of two genes, npc1 and npc2. Cells lacking Npc1, which is a transmembrane protein related to the Hedgehog receptor Patched, or Npc2, which is a secreted cholesterol-binding protein, have aberrant organelle trafficking and accumulate large quantities of cholesterol and other lipids. Though the Npc proteins are produced by all cells, cerebellar Purkinje neurons are especially sensitive to loss of Npc function. Since Niemann-Pick type C disease involves circulating molecules such as sterols and steroids and a robust inflammatory response within the brain parenchyma, it is crucial to determine whether external factors affect the survival of Purkinje cells (PCs). We investigated the basis of neurodegeneration in chimeric mice that have functional npc1 in only some cells. Death of mutant npc1 cells was not prevented by neighboring wild-type cells, and wild-type PCs were not poisoned by surrounding mutant npc1 cells. PCs undergoing cell-autonomous degeneration have features consistent with autophagic cell death. Chimeric mice exhibited a remarkable delay and reduction of wasting and ataxia despite their substantial amount of mutant tissue and dying cells, revealing a robust mechanism that partially compensates for massive PC death.
Synopsis
Niemann-Pick disease type C is a deadly neurodegenerative disease that is most often due to mutations in a gene called npc1. As a consequence of intracellular lipid trafficking defects, patients with Niemann-Pick type C, and mice with the same disease, lose an important class of cerebellar neurons called Purkinje cells (PCs). Npc1 (the protein coded by npc1) might be needed in other cell types to produce substances that nourish PCs or within the PCs themselves. To see which is true, the researchers constructed genetically mosaic mice in which some cells have mutant Npc1 and some have normal Npc1 function. In the cerebella of these mosaic mice, PCs lacking Npc1 continued to die even while surrounded by normal cells, while normal PCs appeared unaffected by their partially mutant surroundings. From these findings, the researchers concluded that the neurodegeneration is due to a problem within PCs and not due to a lack of supporting factors provided by other cells or an extrinsic toxic or inflammatory insult. Npc1 probably functions within PCs to allow critical transport processes necessary for cell survival. The researchers also found that the degenerating PCs undergo a complex process called autophagy in which the cells sense a lack of key nutrients and start to break down their own structures to feed themselves. By identifying exactly which cells require Npc1 function, the researchers set the stage for investigating the exact molecular roles of Npc1 protein in the cells where it is most needed.
Citation:Ko DC, Milenkovic L, Beier SM, Manuel H, Buchanan J, et al. (2005) Cell-autonomous death of cerebellar Purkinje neurons with autophagy in Niemann-Pick type C disease. PLoS Genet 1(1): e7.
==== Body
Introduction
Niemann-Pick type C (NPC) is a devastating autosomal recessive neurodegenerative disorder characterized by the accumulation of cholesterol and other lipids in viscera and the central nervous system. The clinical presentation typically includes progressive ataxia, dystonia, and dementia, first presenting in early childhood and ultimately leading to death in the early teens [1]. NPC is one of over 40 known lysosomal storage disorders (reviewed in [2,3]), which collectively have an incidence of one in 8,000 live births [4]. The specific molecules that accumulate in each disease vary, but most of the disorders share the feature of prominent neurological symptoms. Though great progress has been made in characterizing the biochemical and genetic defects in these diseases, the pathways that lead from these defects to cell and tissue dysfunction are inadequately understood.
The decline in neurological function in NPC patients is caused by mutations in either the npc1 or npc2 gene [5–7]. Npc1 is a multiple-membrane-spanning late endosomal/lysosomal protein that can bind cholesterol [8] and acts as a transmembrane pump when expressed in bacteria [9]. The Npc1 protein has a sequence related to the “sterol sensing domain” of known regulators of cholesterol metabolism (SCAP and HMG-CoA reductase) and Patched, a receptor for the secreted developmental signaling protein Hedgehog. The exact role of Npc1 in regulating lipid homeostasis and in maintaining neurological function is far from clear. Even less is known about Npc2, the small cholesterol-binding protein [10–12] responsible for a minority (<5%) of NPC cases.
Phenotypes similar to human NPC are seen in two mouse strains, C57BLKS/J spm [13] and BALB/c npc1nih [14], both of which harbor spontaneous mutations in npc1 [6]. The most striking and well-documented histological change in npc1 mice is the progressive loss of cerebellar Purkinje cells (PCs) [15,16]. Stereotactic cell counting confirmed that the PC is the type that exhibits the greatest percentage loss in npc1 mice (<10% remaining at 10 wk) but surprisingly revealed that glia in the corpus callosum have a greater early reduction in number (48% for glia versus 13% for PCs, at 3 wk) [17]. This early decline in glia and the detection of Npc1 protein in astrocytic processes [18] led to the idea that the primary function of the Npc1 protein may be in glia and that loss of PCs and other neurons may be a secondary consequence of glial dysfunction [17,19]. However, Npc1 mRNA is abundantly present in neurons as well as glia throughout the brain [20]. Thus, it is unclear which cells in the brain require Npc1 protein and why PCs degenerate in mice lacking Npc1.
Four broad categories of possible explanations for the PC degeneration have emerged (Figure 1). (1) Npc1 is required within PCs and loss of the protein results in degeneration as a consequence of the toxic accumulation of lipids. High concentrations of unesterified cholesterol can cause cell death by altering membrane rigidity, forming intracellular crystals that may interfere with organelle function, or triggering apoptotic signaling events [21]. (2) Npc1 is required within PCs and loss of the protein causes a deficiency at specific subcellular locations. In this model, the accumulation of cholesterol and other lipids is not harmful per se, but the diminished capacity to transport these or other molecules to the correct places within the cell leads to degeneration. (3) Npc1 is required within glia or another cell type within the cerebellum, or in a distant organ such as a secretory gland, and loss of the protein results in degeneration of PCs through a deficiency in a secreted molecule. This could be a trophic factor or possibly even cholesterol itself, as studies have shown that glia-derived cholesterol is required for synaptogenesis [22]. (4) Loss of Npc1 function results in the release of toxic/pro-inflammatory compounds into the local environment or systemic circulation, causing PC degeneration.
Figure 1 Models of PC Degeneration in npc1 Mice
Four basic mechanisms are depicted. (1) The accumulation of cholesterol, sphingolipids, or other molecules within PCs lacking Npc1 could be toxic. (2) Loss of Npc1 function could cause a block in trafficking that leads to a localized subcellular deficiency in lipids and/or proteins. (3) Other cells, most likely glia, could produce a secreted factor, such as apoE-cholesterol or a trophic factor required for PC survival, whose export is reduced or blocked without Npc1 function. (4) Any cell type in the npc1 mouse could produce toxic metabolites, such as released lysosomal hydrolases or beta amyloid, that could kill surrounding cells or particularly susceptible cell types regardless of genotype.
In this study, we employed chimeric mice to help resolve the fundamental mystery of why neurons degenerate in NPC disease. This powerful approach can distinguish intrinsic and extrinsic causes of cell death but has only rarely been used to study human neurodegenerative disease [23,24]. The results indicate that PC loss in npc1
−/− mutants is a cell-autonomous process, i.e., Npc1 function is critical within PCs. Ultrastructural and biochemical analyses support the idea that this cell-autonomous loss is due to the activation of a particular genetic program, autophagy, that leads to cell death.
Results
Pattern of Cerebellar Degeneration in npc1 Mice
As a first step it was important to observe and measure progressive PC loss. PC loss in the cerebellar vermis progressed in an anterior (lobule I) to posterior (lobule X) sweep, as previously reported [15,16] (Figure 2A and 2B). There was a normal number and density of PCs in cerebellar sections from 30-d-old npc1−/− mice. At 50 d, most PCs in lobules I–III were no longer present. By 70 d, only lobule X PCs remained intact.
Figure 2 Features of PC Loss in npc1 Mice
(A) Sections from the cerebellar vermis of a 70-d-old npc1−/− mouse were stained with anti-Calbindin (green) to visualize PCs and 7AAD (red) for nuclei. Although there are scant PCs remaining in lobule II, the PCs in lobule X have not decreased.
(B) Quantification of progressive anterior-to-posterior PC loss. PC densities from 30-d npc1
+/+ and 30-, 50-, and 70-d npc1
−/− mice were quantified from five sections from two mice each. Error bars show standard deviation.
(C) Bergmann glia visualized in 70-d npc1
+/+ and npc1
−/− mice using anti-S100β (green) have normal morphology. Note that the glial cell bodies near the PC layer and the radial processes extending into the molecular layer are intact in the npc1
−/− mouse despite the loss of PCs.
(D) Oligodendrocyte cell bodies stained with anti-CC1 (green) have a similar distribution and morphology in the cerebellar white matter of 70-d npc1−/− and wild-type mice. TOTO-3 (red) was used as a nuclear counterstain. Bar = 5 μm.
(E) Axonal spheroids in npc1
−/− mice. Spheroids (arrows), seen as Calbindin-positive (red) swellings surrounded by myelin (anti-myelin basic protein; green), are numerous in the npc1−/− cerebellum (50-d-old, lobule X shown) but are not seen in the wild-type control. Bar = 5 μm.
(F) Sections from wild-type and npc1 mice were stained with filipin to visualize free cholesterol. PCs from npc1 mice show an accumulation of intracellular cholesterol at 30 d regardless of their lobular location. Images are oriented such that the PCs (arrowheads) are in the center of the field, with the molecular layer above and the inner granule cell layer below.
We next investigated whether Bergmann glia are affected in npc1−/− mice. Bergmann glia appose the PC soma and extend processes into the molecular layer that sheaths the PC dendritic trees (reviewed in [25]). Loss or dysfunction of these glia could precede and precipitate the PC loss. In contrast to the degeneration and loss of PCs, Bergmann glia stained with anti-S100β maintain their density and their characteristic radial morphology, even at 70 d, in all lobules (Figure 2C). Oligodendrocytes, the glia that myelinate axons, were normal in distribution and cell body morphology (Figure 2D). Staining of myelin revealed axonal spheroids as previously reported (Figure 2E) [15,26]. Granule cells, the most abundant neurons in the cerebellum, were also grossly normal based on anti-GABAα6 staining (data not shown). This indicates that loss of PCs in npc1 mice is not due secondarily to a loss or other obvious morphological derangement of these other cell types. However, these observations do not rule out the possibility that a more subtle dysfunction in these cells contributes to the degeneration of PCs in npc1mice.
The anterior-to-posterior gradient of PC loss provides the opportunity to determine how cholesterol accumulation changes in relation to the PCs. If intracellular cholesterol accumulation is toxic to PCs, a correlation is expected between the amount of cholesterol and the rate or severity of degeneration. Instead, we find that at 30 d, intracellular cholesterol accumulation is noted even in lobule X PCs (which do not degenerate during the normal npc1 mouse life span). We did not find any consistent difference in filipin staining of cholesterol among the different cerebellar lobules in npc1 mice (Figure 2F). Therefore, lysosomal cholesterol accumulation per se is not sufficient to explain the pattern of PC death.
PC Degeneration in npc1 Mice Is Cell-Autonomous
To determine whether PC loss is due to dysfunction within the PCs, alterations in the surrounding glia and microenvironment, or a combination of both, we generated mice containing a mixture of npc1 mutant and wild-type cells. Morulas derived from mating npc1 heterozygotes (BALB/c npcnih/+) and from matings of wild-type mice homozygous for a ubiquitously expressed green fluorescent protein (GFP) gene (FVB.Cg-Tg(GFPU)5Nagy/J; [27]) were aggregated to form chimeric embryos. A quarter of the embryos from npc1/+ parents should be homozygous for the npc1 mutation, but these embryos could not be distinguished at the time of morula isolation, so the embryos derived from aggregation had to be analyzed later to determine which were the useful chimeras. Chimeras were also created in which npc1 mutant cells were marked with GFP (genetic background 75% FVB/25% BALB/c; see Materials and Methods for crosses), ensuring that the results were independent of the genetic background. Of 33 chimeric mice generated, seven were of the desired genotype, as determined using PCR analysis (two npc1
−/− BALB/c↔GFP; two npc1
−/− 75% FVB/25% BALB/c↔CD-1; three npcl
−/− 75%FVB/25% BALB/c↔C57BL6/J) (Figure 3A). These seven mice had wild-type contributions ranging from 11% to 61% based on skin fluorescence, granule cell counts, and lobule X PC counts. Wild-type contributions estimated from skin fluorescence always corresponded well to the value determined from cell counts; the numbers differed by 4%–22% (see Table 1). The reliability of skin fluorescence and granule cell counts as estimates of overall chimerism percentage was further validated by examination of two npc1 heterozygous chimeras that demonstrated similar chimerism percentage whether determined by skin, granule cell counts, or, most importantly, PC counts (Table 1). For each chimeric mouse, all PCs in five sagittal sections of the cerebellar vermis were counted and each cell's genotype was determined based on GFP fluorescence. Characteristics of the mice are summarized in Table 1.
Figure 3 Cell-Autonomous Degeneration of PCs in Chimeric Mice
(A) Genotyping of chimeric mice. Upper gel: PCR to amplify a fragment of the npc1 gene from npc1
−/−↔GFP, npc1
+/−↔GFP, and npc1
+/−mice results in npc1− (1,067 bp) and npc1+ (947 bp) bands. Lower gel: The three genotypes can be distinguished by ApaLI digestion of the PCR products. The npc1− allele gives rise to 937- and 130-bp bands. The wild-type allele varies depending on the source: the BALB/c allele is digested to 477-, 340-, and 130-bp bands while the allele from the GFP mouse is digested to 817- and 130-bp bands. Note that the 477- and 340-bp bands are absent from the homozygous mutant chimera.
(B) Quantification of npc1
−/− PC density in chimeric mice. For each mouse, the actual density observed (the average of the mean densities counted in each lobule) is compared to the number expected if no PC loss occurs (calculated from the density at 30 d in npc1
−/− mice and the chimerism percentage). Five sections separated by at least 100 μm were analyzed for each mouse. For the three mice sacrificed at 70 d, only lobules I–IX were included in the analysis because of the lack of degeneration in lobule X in npc1
−/− mice at this age. A value for the number of npc1
−/− PCs expected if loss does occur (calculated from the density at 70 d in npc1
−/− mice and the chimerism percentage) is also included for these three mice. No rescue of npc1
−/− PCs was observed for any of the chimeric mice (p < 0.0001 for each mouse comparing observed versus expected if no loss occurs).
(C) Quantification of npc1−/− PC density by lobule in C1.4 at 70 d. Observed and expected densities are shown for npc1
−/− PCs in each lobule. No clear rescue of npc1
−/− PCs was observed in any lobule (p > 0.05 for all lobules comparing observed versus expected if loss occurs, except II and III, where p-value is between 0.01 and 0.05; p < 0.001 for all lobules comparing observed versus expected if no loss occurs, except lobule X, where p > 0.05).
(D and E) Images of lobules II and X taken from C1.4 at 70 d. GFP is green and Calbindin staining is red. In lobule II, the number of PCs is clearly reduced and of the five remaining PCs, four are wild-type (GFP-positive). In lobule X (where no degeneration occurs during the normal npc1
−/− life span), PC density is normal and the majority of PCs (23 of 32; 72%) are mutant, as expected based on npc1
−/− contribution to this mouse (67%).
(F) Quantification of wild-type PC density in chimeric mice. Observed and expected densities are shown for wild-type PCs. For mice with more than 15% wild-type contribution, the density of wild-type PCs is similar to the number expected if these cells have not degenerated (p > 0.05 for each mouse comparing observed versus expected if no loss occurs, except C5.2, where p = 0.02, and C5.6, where p = 0.0005).
(G) Quantification of wild-type PC density by lobules in C1.4 at 70 d. No loss of wild-type PCs was observed in any of the lobules (p > 0.05 for all lobules comparing observed versus expected if no loss occurs; p < 0.001 for all lobules comparing observed versus expected if loss occurs, except for lobule X, where p > 0.05).
Table 1 Genotype and Percentage Chimerism for the Mice Used in This Study
aPercentage wild-type based on PCs was estimated from calculating the density of PCs in five sections at least 100 μm apart. Standard deviation is in parentheses.
bPercentage wild-type based on granule cells was estimated from counting of anti-NeuN-stained granule cells in three 65 × 65 μm regions from two sections at least 100 μm apart.
cSevere ataxia is defined as a balance beam score of two or less. Homozygous mutant ages are the average of three male and three female mice.
dThis small percentage of non-GFP PCs is likely an artifact from the thin sectioning; if a small fraction of a PC is in the section, it may not be scored GFP-positive. Examination of the same PC in the adjacent section shows that all PCs are in fact GFP-positive.
eOnly lobule X PCs were used in estimating the wild-type contribution in mutant chimeric mice sacrificed at 70 d, prior to degeneration of this lobule.
fND indicates not determined, as this mouse was sacrificed prior to developing severe ataxia.
gNM indicates not meaningful, as lobule X PCs are only a useful measure of wild-type contribution when mice are sacrificed by 70 d.
CD1, CD-1 strain (Charles River Laboratories); C57, C57BL/6J strain (Jackson Laboratory).
The data from all seven mice demonstrated that PC degeneration in npc1−/− mice is primarily a cell-autonomous process. PC counts revealed that the presence of a milieu that was more than 60% wild-type was insufficient to prevent npc1 PC loss. In the two oldest mice analyzed (190 d), chimeras 4.4 and 4.8 (C4.4 and C4.8), nearly all npc1 PCs had degenerated, including those in lobule X. In these mice, the wild-type contributions were estimated to be 54% and 61%. Mutant npc1 PCs were observed at densities of 0.5 and 0.6 PCs/mm in the mice, strikingly lower than the expected densities of 12 and 10 PCs/mm if no degeneration had occurred (Figure 3B). Expected PC densities, assuming no degeneration, were calculated by multiplying the observed density in 30 d npc1−/− mice by the percentage of chimerism. Similar measurements were made in all seven chimeras. Two mice sacrificed at a younger age (120 d; C2.1 and C5.8) had lost all but a few npc1 PCs, which were primarily in lobule X. The results demonstrate that surrounding wild-type cells are unable to prevent npc1 PC degeneration. These four mice also demonstrate that lobule X PCs will eventually degenerate; their loss is delayed rather than completely prevented.
Three chimeric mice were analyzed at 70 d, at the end of the normal life span of npc1
−/− mice. One of these three mice had a substantial wild-type contribution (33%; C1.4), but no significant prevention or delay of npc1 PC degeneration was observed (Figure 3C). The vast majority of mutant PCs (except those in lobule X) degenerated despite the presence of a significant fraction of wild-type cells (Figure 3D and 3E).
In contrast to the mutant PCs, wild-type PCs did not degenerate in most of the chimeric mice. The observed wild-type PC densities agreed well with the expected densities based on the wild-type contribution in each mouse (Figures 3F and 3G). The lone exceptions occurred in the two mice with 11% and 13% wild-type contribution. These mice were expected to have a very low number of wild-type PCs, but there were even fewer than expected. No significant decrease in wild-type PC density was observed in the other five chimeric mice, including C2.1, which was expected to have only a marginally greater wild-type contribution (15%). Although fluctuation in how small numbers of wild-type cells are distributed and consequent variability among analyzed sections provide a likely explanation, another interpretation is that when very few wild-type cells are present, the entire cerebellum begins to break down during the late stages of cerebellar degeneration. Toxins released by dying cells could contribute to PC loss but are clearly not the main cause of neurodegeneration.
Other Histologic Features of npc1 Chimeric Mice
Examination of the relationships of other cell types to the PCs in the chimeras provides further evidence for the cell-autonomous nature of PC degeneration. Microglia have been observed in the cerebellum of npc1−/− mice, but their role in PC degeneration is unclear [28]. Microglia are the phagocytic cells of the central nervous system, and their presence is indicative of an inflammatory process. The appearance of microglia within the cerebellum follows the anterior–posterior pattern of PC death. In wild-type mice or npc1 mutant mice at 30 d, almost no microglia are observed (data not shown). In 50-d-old npc1 mutant mice, very few microglia are present in lobule X (Figure 4A). The other cerebellar lobules have numerous microglia in the white matter tract and granule cell layer prior to the loss of PCs (Figure 4B). After more extensive PC loss (lobules I–IV at 50 d and I–IX at 70 d), microglia also appear in the molecular and PC layers (Figure 4C). Thus, the arrival of microglia coincides with PC loss. However, the infiltration of numerous microglia is not sufficient to trigger PC degeneration. Microglia were present in all the chimeric mice (though to a lesser extent in mice sacrificed at a later age) and were in close proximity to wild-type PCs (Figure 4D, white arrows). Despite this close association, there was no decrease in wild-type PC density in the chimeras except the two noted above. Furthermore, if nonspecific microglial killing of PCs were the main mechanism of PC loss, then wild-type and mutant PCs should demonstrate comparable decline percentages. This is not the case in any of the chimeras.
Figure 4 Other Histological Characteristics in the npc1 Mutant and Chimeric Mice
(A–C) Microglia characteristics. Sections from 50-d-old npc1 mutants were stained with anti-Calbindin (green) to visualize PCs and anti-F4/80 (red) to mark microglia. (A) Very few microglia are present in lobule X, where no PC degeneration has yet occurred. (B) Lobule VIII demonstrates an infiltration of microglia in the granule cell layer and white matter tract. (C) In lobule II, nearly all PCs are gone and numerous microglia are present throughout the cerebellum. The dashed line indicates the edge of the granule cell layer.
(D) Microglia infiltration is not sufficient to induce PC degeneration. C1.4 demonstrates numerous microglia marked by anti-F4/80 (red) in all layers of the cerebellum, including immediately adjacent to wild-type (GFP-positive) PCs (arrows). Despite this close approximation, no loss of wild-type PCs was detected by cell counting. The GFP transgene is apparently poorly expressed in microglia, as all microglia appear GFP-negative regardless of their genotype.
(E) Mutant PCs are lost even when surrounded by wild-type Bergmann glia. A stretch of the PC layer in C4.8 is shown where no npc1−/− (GFP-positive) PCs remain despite the presence of numerous wild-type (GFP-negative) glia (arrowheads) marked by anti-S100β (red). The space occupied by two wild-type PCs is indicated (arrows) with two npc1−/− (GFP-positive) glia directly underneath.
(F) Wild-type PCs do not degenerate even when surrounded by mutant Bergmann glia. Three wild-type PCs (GFP-positive; arrows) in C1.4 have not degenerated despite only mutant glia (arrowheads) being in the immediate vicinity.
(G) Oligodendrocytes in chimeric mice are a mixture of wild-type and mutant cells. Oligodendrocyte cell bodies are stained with anti-CC1 (red). Wild-type (GFP-positive; arrows) and npc1−/− (GFP-negative; arrowheads) oligodendrocytes are interspersed in C1.4.
Bar = 50 μm for (A), (D), and (G); 10 μm for (E) and (F).
Mutant PC loss occurred even with wild-type Bergmann glia in the immediate vicinity. In the chimeras, PCs were strikingly absent in regions containing abundant wild-type glia (Figure 4E, arrowheads). Conversely, wild-type PCs (Figure 4F, white arrows) survived even when surrounded by mutant glia (Figure 4F, arrowheads). Oligodendrocytes of both genotypes were scattered in the cerebellar white matter of the chimeric mice (Figure 4G). The loss of mutant PCs is unlikely to be due to the lack of a secreted factor from supporting cells. If Npc1 were necessary for the efficient secretion of this hypothetical survival factor, then surrounding wild-type glia should have been able to provide the molecule to mutant PCs. The ongoing degeneration of npc1 mutant PCs surrounded by a high percentage of wild-type cells indicates that cell-autonomous mechanisms play the primary role in PC death.
Prevention of Ataxia and Weight Loss in Chimeric Mice
Despite substantial loss of npc1 PCs, some chimeric mice displayed minimal ataxia and no weight loss. Ataxia was assessed through measuring agility on a balance beam [29] and by quantifying step size in a gait assay [30]. At 190 d, more than twice the normal life span of npc1 mice, the two chimeras with more than 50% wild-type cells (C4.4 and C4.8) displayed no signs of wasting and only minimal ataxia (Figure 5). Even more impressive was a mouse with only 15% wild-type cells (C2.1) that had significantly reduced symptoms. The genetic background of the mutant cells appears to influence the rate of disease progression. Delayed symptoms were not noted in C5.2 and C5.6, which had only slightly less wild-type contribution than C2.1. The npc1 mutant cells in the C5.2 and C5.6 mice were derived from a 75% FVB/25% BALB/c↔GFP+/+ background, and mutant mice of this genetic background demonstrated a faster progression of disease compared to the mutation in the BALB/c background of C2.1 (average life span of 67 ± 5 d versus 81 ± 8 d). It is remarkable that some 40% of the PCs can die without leading to severe ataxia or weight loss, and that even much smaller numbers of wild-type cells (15%) provide temporary protection against wasting, ataxia, and death.
Figure 5 Organismal Phenotypes in Chimeric Mice
(A) Chimeric mice with more than 50% wild-type contribution do not exhibit wasting. Weekly weights for each of the mice were recorded from 4 wk until the mice were sacrificed. For wild-type and npc1−/− homozygous mice, the curve shown is the mean weight from three to four mice. A progressive decrease in weight of npc1 mice begins to show around 7 wk. The chimeric mice display a delay in, or even absence of, wasting depending on the amount of wild-type contribution.
(B) Prevention of ataxia in chimeric mice. Mice were assessed for ataxia weekly during two 3-min trials on a balance beam (see Materials and Methods).
(C) Gait of chimeric mice. Front and back paws of mice were dipped in red or green paint, and mice walked across a box lined with paper. The npc1
−/− mouse displays shorter stride length and a smearing of the footprints as the paws are not as well lifted between steps. Chimeric mice with high amounts of wild-type contribution (C4.8; C4.4 not shown) exhibited a gait indistinguishable from wild-type.
(D) Quantification of stride length. For wild-type and mutant controls, three mice were assessed. For chimeric mice, gait was measured prior to sacrifice at the ages noted. Error bars indicate the standard deviation. Comparison with the GFP mice demonstrates that npc1+/+, C1.4, C4.4, and C4.8 do not have significantly different stride lengths (p > 0.05) while npc1−/−, C5.2, C5.6, C2.1, and C5.8 have significantly shorter stride lengths (p < 0.001)
Activation of Autophagy in Degenerating PCs
The evidence from chimeric mice highlighted the importance of possible cell-autonomous mechanisms of PC degeneration. Necrosis, apoptosis, and autophagy of PCs all occur in mouse models of neurodegeneration (reviewed in [31]). The cell death mechanism at work in NPC is unknown, although a very recent report demonstrated TUNEL staining and increased caspase-8 levels in npc1
−/− cerebellum, supporting a role for apoptosis in neurodegeneration in NPC [32]. We addressed this question by ultrastructural comparison of wild-type and npc1 PCs using electron microscopy (EM). No apoptotic bodies were identified by EM, nor did we identify any TUNEL-positive PCs using immunofluorescence labeling of sections from 48-d-old npc1−/− mice (data not shown).
In contrast, numerous structures with features consistent with autophagic vacuoles (AVs) were identified in npc1 PCs by EM. Previous EM studies of npc1 PCs [13,16] noted the appearance of “lamellar inclusion bodies.” Such structures are consistent with the activation of autophagy, although to our knowledge there have been no reports of autophagy in NPC. AVs can be classified as autophagosomes or autolysosomes based on morphological criteria [33,34]. Autophagosomes are bound by two or more membranes and contain seemingly unaltered cytoplasmic components. The autophagic and endocytic pathways converge following fusion of autophagosomes with late endosomes and subsequently with lysosomes [35]. These fusion events result in autolysosomes, organelles enclosed by a single limiting membrane with degrading membranes and cellular components within. Numerous AVs are found in PCs from 48-d-old npc1 homozygous mice (Figure 6A–6E). Such structures are rarely seen in 48-d-old wild-type mice (Figure 6F–6H). Since NPC cells accumulate lipids within organelles containing late endocytic and lysosomal markers, the structures we identified as AVs could be aberrant late endosomes or lysosomes, with no autophagy involved. However, we can clearly identify cytoplasmic contents (free ribosomes, endoplasmic reticulum [ER] cisternae, and mitochondria) within these structures, a diagnostic for AVs. Quantification of AVs in wild-type and npc1 PCs demonstrates that these structures occupy 15 times more area in npc1 PCs (Figure 6I). Rarely, the cell body of an npc1 PC filled up almost completely with autolysosomes (Figure 6E). Degenerating PCs in npc1 mice clearly have morphological features consistent with autophagic cell death.
Figure 6 Increased Autophagy in npc1 PCs
EM of npc1−/− and wild-type PCs from anterior vermis of 48-d-old mice.
(A) Low-power field of a typical npc1 PC.
(B) Magnification of the red box in (A) showing accumulated multivesicular and multilamellar organelles.
(C) Magnification of the blue box in (A) showing a large AV (arrowheads) containing ribosomes and membranes.
(D) Another AV (arrowhead) with ER cisternae within the lumen.
(E) An npc1 PC probably in the late stages of degeneration, with many AVs, some with apparently degrading mitochondria (arrowheads).
(F) Low-power field of two wild-type PCs.
(G and H) Magnification of the red (G) and blue (H) boxes in (F) showing no AVs and very few multivesicular organelles (arrow).
(I) Quantification of area occupied by AVs and other multivesicular/multilamellar organelles expressed as a percentage of total cytoplasmic area for seven wild-type and seven npc1−/− PCs; p = 0.0009 for comparing AV areas and p = 0.0003 for comparing other multivesicular/multilamellar areas, indicating both types of organelles occupy significantly more area in npc1−/− PCs.
Bar = 5 μm for (A) and (F); for all others, bar = 1 μm.
Increased modification of an autophagosomal marker protein provides further evidence for the activation of autophagy in npc1 cerebellum. During autophagy induced by starvation or rapamycin, the light chain 3 (LC3) protein undergoes lipid modification by a ubiquitin-like conjugating system [36,37]. This lipid-modified form of the protein (LC3-II) associates with autophagosome membranes, and levels of LC3-II correlate with the amount of autophagosome formation [38]. In adult cerebellar cortex, LC3 protein is reported to be present primarily in PCs, based on immunofluorescence studies [39]. The amount of LC3-II protein in npc1
−/− cerebellum increases dramatically with age (Figure 7A). When 22-d-old mice are compared, LC3-II levels are nearly equivalent. In wild-type mice, the level stays the same or may slightly decrease in 50- and 67-d-old mice. In the npc1−/− mice, there was a 7- and 11-fold increase in LC3-II compared to age-matched controls of these respective ages. Thus, morphological and biochemical criteria show that autophagy is activated in degenerating npc1−/− cerebella.
Figure 7 Levels of the Autophagosomal Marker LC3-II Are Increased in Degenerating npc1 Cerebellum
(A) Immunoblots of 70 μg of protein resolved by 15% SDS-PAGE show increased levels of LC3-II in 50- and 67-d-old npc1 cerebellar extracts. Two mice of each age and genotype were analyzed, and relative band intensities were quantified. p-Values derived from comparing wild-type and npc1−/− cerebella at each age are 0.3, 0.01, and 0.001 for 22-, 50-, and 67-d-old mice, respectively.
(B) Immunoblots of CHO cell extracts resolved by 15% SDS-PAGE do not show increased levels of LC3-II in CT60 cells compared to wild-type CHO-KI cells or CT60 cells stably transfected with NPC1-YFP. No difference is noted with 10 μM U18666A, a drug that mimics npc1 loss of function. Rapamycin treatment (1 μM) of CHO-KI cells demonstrates robust activation of autophagy with increased LC3-II levels.
To determine whether loss of npc1 is sufficient to induce autophagy in all cell types, we examined CT60 cells, a cell line harboring a loss-of-function mutation in npc1 [40]. No elevation of LC3-II was detected, though LC3-II accumulation can clearly be induced in Chinese hamster ovary (CHO) cells with rapamycin treatment (Figure 7B). The lack of autophagy activation in an npc1 cell line and in cerebella from 22-d-old npc1−/− mice suggests that autophagy is activated only in npc1−/− cells actively undergoing degeneration.
Discussion
Neurodegeneration is among the most devastating and feared types of human disease. No treatments exist that stop neurodegeneration and very few even provide transient symptomatic relief [41]. Alzheimer disease, the most common cause of dementia, affects roughly 12 million people worldwide [42]. Even for Alzheimer disease the pathophysiology is unclear. It is still controversial whether extracellular or intracellular beta-amyloid deposits are responsible for the bulk of neurotoxicity [43,44], and the importance of inflammation to progression of the disease is unknown [45,46]. For most neurodegenerative diseases, how much neuronal loss can be attributed to cell-autonomous versus non-cell-autonomous factors is unknown. We examined this crucial question in the neurodegenerative disorder NPC.
People born with NPC disease undergo a progressive decline in neurological function that can include ataxia, tremor, dystonia, dementia, and seizures. As the prominent ataxia would suggest, cerebellar involvement, including PC degeneration, has been well documented [47–49]. The mouse model of NPC faithfully recapitulates many of the human symptoms, and PC loss serves as an easily quantifiable measure of neurodegeneration. The involvement of circulating molecules such as steroids and sterols, and the ubiquitous expression of both NPC genes, make determining when and where the NPC genes are required a key first step in understanding neurodegeneration. A previous study highlighted the importance of events within the central nervous system, as expression of Npc1 driven by the prion promoter could prevent neurodegeneration even without complete rescue of the liver phenotype [30]. In this study we have found that the most prominent neurodegenerative event, the death of the PCs, is due to lost Npc1 function within those cells and to a distinctive programmed cellular response, autophagy.
Cell-Autonomous PC Death in NPC Disease
We have demonstrated that the loss of PCs in the mouse model of NPC, which has a mutant npc1 gene, is primarily a cell-autonomous defect. This finding places constraints on possible mechanisms of PC degeneration in NPC disease, and allows future investigations to focus on the cell type that matters. As is described in the introduction, several lines of evidence hinted that neurodegeneration in NPC disease could be a non-cell-autonomous process. However, the inability of a wild-type milieu, including surrounding wild-type glia, to prevent npc1−/− PC loss indicates that the degeneration of PCs is not due to the loss of a secreted factor such as glia-derived sterols, neurotrophins, or Npc2 protein. Consistent with this, two very recent studies failed to detect a difference in levels of secretion of cholesterol from wild-type versus npc1−/− glia [50,51].
In addition to surrounding glial cells, the health of PCs could be affected by their axonal targets in the cerebellar deep nuclei. Expression of GFP is very low in neurons of the cerebellar deep nuclei in the GFP mice (data not shown), so we were unable to assess the genotype of these cells in the chimeras. During the migration of PCs, their axons do not remain attached to their future post-synaptic targets [52]. If PC death in npc1 mice was caused by loss of a target-derived factor, then the absence of a genotypic match between target neurons and PCs would mean that wild-type and npc1 PCs should be equally affected. This is not what we observe. In addition, completely depriving PCs of their axonal targets by axotomy does not cause PC degeneration even up to a year following axotomy [53,54]. Thus, derangement of cerebellar deep nuclei is unlikely to be causing PC degeneration in NPC.
Secreted or released toxic molecules are also not responsible for PC death. Microglia produce pro-inflammatory and cytotoxic compounds including cytokines, proteases, and free radicals [55–57] and are present in higher-than-normal numbers in the NPC mutant cerebellum [28]. A robust inflammatory response may play a role in the progressive neurodegeneration seen in NPC but is not the primary factor in causing neuronal loss: in the chimeric mice most wild-type PCs do not degenerate even while in the midst of numerous microglia. Our experiments do not exclude a more cell-specific role for microglia in PC death. Microglia could help trigger cell death by specific molecular interactions with susceptible PCs, as is proposed to occur during PC death normally seen at postnatal day 3 [58]. A microglia-based mechanism would require a cell-autonomous component—PCs expressing the proper membrane receptors—and a non-cell-autonomous component—microglia to carry out the directed engulfment of only those cells.
Internal Factors in Neurodegeneration
PCs may degenerate either (1) because the buildup of sterols and other lipids, or other unknown substances, poisons the PCs or (2) because trafficking defects resulting from Npc1 loss affect cell survival. Although excess cholesterol can have toxic effects [21], intracellular cholesterol levels within PCs do not correlate with the loss of these cells: lobule X PCs accumulate cholesterol and do not degenerate during the 70 d that normally constitute the npc1 mouse life span. More subtle differences among lobules, in sterol mass or subcellular localization of cholesterol, are unlikely to be detected with filipin staining. In any case, blocking lipid accumulation is insufficient to slow neuronal loss, again arguing against the idea that cell death occurs because of accumulated lipid. Crossing the npc1 mouse to a GalNAcT-knockout mouse unable to synthesize complex gangliosides decreases levels of sphingolipids and cholesterol without any effect on neurodegeneration [59]. Crossing the npc1 mouse to a low density lipoprotein receptor knockout also does not slow neurodegeneration [60]. Although both of these experiments have their weaknesses—possible buildup of less complex sphingolipids in the GalNAcT experiment and possible redundancy in lipoprotein receptors in the low density lipoprotein receptor knockout—the findings at least suggest that intracellular toxicity due to lipid accumulation is not the primary mechanism of PC death.
Defects in lipid trafficking are likely to have particularly detrimental effects on synthetic processes that rely on lipids as substrates. A recent study demonstrates that neurosteroid synthesis is reduced in npc1 mice and administration of allopregnanolone increases life span and neuronal survival [61]. PC loss involving interference with production of neurosteroids acting in a paracrine or endocrine manner is untenable in light of our findings. An autocrine mechanism of action, with PCs making substances that act in the producing cell, is still plausible and needs to be explored further. PCs synthesize several different neurosteroids and express progesterone receptors [62]. PCs are unique in having an extensive smooth ER, the “hypolemmal cisternae,” that is “so well developed in PCs that it may be considered a specific characteristic…since it extends under the plasmalemma of the entire cell body, the dendritic tree, and the axon” [63]. Palay and Chan-Palay [63] also note that mitochondria are often found against the inner surface of the hypolemmal cisternae. Impaired cholesterol trafficking to mitochondria and the smooth ER may lead to a deficiency at these sites, lowering the production of neurosteroids.
Autophagy in NPC PC Death
Loss of npc1 function within PCs leads to increased autophagy and cell death. Apoptosis may also be involved in the death of neurons from NPC disease; both processes are likely at work in the degenerating NPC brain. In contrast to apoptosis, which has been demonstrated in many instances of neurodegeneration (reviewed in [31]), activated autophagy has been demonstrated in only a few cases, but notably including death of PCs in Lurcher mice [64]. The role of autophagy in neurodegenerative disease extends to Huntington disease and Parkinson disease [65–67]. These neurodegenerative diseases differ from NPC in that they are disorders of protein aggregation. Formation of insoluble aggregates of Huntingtin or alpha-synuclein is an early event in disease progression [68–71]. Autophagy may be activated by cells in order to dispose of the aggregates. Deliberately enhancing degradation of protein aggregates by triggering autophagy in animal models of Huntington disease can ameliorate the affects of the disease [67]. No protein aggregation has been reported in NPC, but an analogous mechanism could be at work: during the first several weeks of life, cells may activate autophagy to degrade the abnormal lipid-laden lysosomes and allow the survival of PCs. If this is true, then cells with lipid-laden lysosomes should have high autophagic activity. This prediction is clearly contradicted by our data: activation of autophagy is seen only in older, degenerating PCs, and npc1 cells not undergoing cell death do not exhibit increased levels of the autophagic marker LC3-II.
Why then is autophagy induced in degenerating PCs? Autophagy is a normal, cellular process that is regulated by nutritional status and by hormones. Loss of npc1 leads to trapping of lipids within aberrant membrane compartments, and this may induce a “lipid-starvation response” analogous to the well-characterized autophagic response to amino acid deprivation. In fact, there is already evidence that such a response occurs in npc1 homozygotes: despite the seeming overabundance of cholesterol in npc1−/− cells, the SREBP signaling pathway is activated. This results in increased cholesterol synthesis and low density lipoprotein receptor activity [72]. Thus, NPC cells act as though they are “cholesterol starved.” NPC cells therefore may have three ways to adjust to a shortage of sterols and other lipids: increased lipid synthesis, increased lipid uptake, and autophagy. Microarray analysis of SREBP targets in liver [73] did not reveal increased transcription of autophagy genes, but the targets may differ between tissues, or regulation may occur at the post-transcriptional level. In NPC cells, the autophagic response would be futile since transport systems are damaged. Instead of rescuing the cell, autophagy would hasten cell death by further sequestering and reducing available lipids.
Destructive autophagy in NPC mutant PCs may also, or instead, be stimulated hormonally. In mammals, autophagy is inhibited by insulin [74], and in insects, steroids regulate autophagy of neurons by a cell-autonomous mechanism [75,76]. In Manduca (tobacco hawkmoth) motor neurons and in salivary glands and fat body of Drosophila, the steroid ecdysone triggers autophagy at specific developmental time points. We would infer a different regulatory relationship in mammalian NPC. Neurosteroids might inhibit autophagy in PCs and when their synthesis is severely decreased, as in NPC [61], autophagic cell death might ensue. Neurotrophins may also be involved in regulating PC autophagy. Cultured PCs undergo autophagic cell death following neurotrophin withdrawal [77]. These findings are relevant to NPC because neurons from NPC mice have decreased responsiveness to neurotrophins [78].
Levels of Cerebellum Involvement in Chimeric npc1−/− Mice
The chimeric mice also provide information about the progression of the disease. Homozygous npc1 mice at 70 d are unable to maintain posture on the balance beam for even a few seconds and weigh only half as much as controls. Remarkably, some chimeras had only minimal ataxia and no wasting despite the loss of approximately 40% of PCs. The minimal ataxia phenotype, manifested as occasional falls off the balance beam while turning, stopped progressing by approximately 140 d. By this age, nearly all mutant PCs, including those in lobule X, had died. Presumably ataxia worsens until all mutant PCs die, at which point the phenotype stabilizes.
With so many PCs lost, other regions of the brain, other cerebellar cells, or the remaining PCs may compensate. Morphological changes that have been observed in the remaining npc1 neurons, such as occurrence of ectopic dendritic spines and meganeurites [26,79] and increased width of the dendritic tree [80], may be the physical manifestations of the remaining PCs taking on functions of their deceased counterparts. Compensation may also occur in other regions of the brain. In the pcd and Lurcher mouse models, compensatory mechanisms in the efferent targets of the PCs, the cerebellar deep nuclei, have been postulated to be responsible for the mild ataxia that occurs despite severe PC loss [81]. Lurcher chimeric mice, in which PCs undergo cell-autonomous loss, have no behavioral abnormalities even when they have few wild-type cells [82]. Recent studies using this model of PC degeneration have defined a minimum number of PCs required for normal motor control between 1,000–7,000, or roughly 1%–5% of the normal number of PCs [83,84]. Thus, the loss of PCs may be significantly ameliorated by still-to-be-identified mechanisms, and slowing loss of PCs by any mechanism may allow increased compensation. Some of the phenotypes of the npc1−/− mouse are likely to be due to loss or derangement of other cells, such as certain populations of neurons in the thalamus [85] or dorsal root ganglion [86]. There appears to be a critical threshold of wild-type contribution between 30% and 60% required for near-complete rescue of both ataxia and weight loss. The delayed progression seen with low levels of wild-type cells (approximately 15%) is quite encouraging with respect to the potential for therapies that preserve even some of the cells.
Several questions regarding the loss of PCs in NPC have been answered in this study. Of four possible mechanisms of PC loss (see Figure 1), the chimeric mice demonstrate that the two strictly non-cell-autonomous models are invalid. Of the two cell-autonomous models, internal toxicity versus subcellular deficiency, we favor the latter based on the observed defects in organelle trafficking in npc1 cultured cells [87,88], the lack of correlation between cholesterol levels and PC loss, the decreased levels of steroidogenesis in npc1 mice [61], and the previous transgenic studies designed to decrease cholesterol or sphingolipid accumulation [59,60]. We envision that in response to the proposed intracellular deficiency, cells compensate by increasing endogenous synthesis of cholesterol, by endocytosis, and/or by autophagy. In cells with a high lipid demand and insufficient supply, autophagy increases and cell death ensues. It may be that a similar series of events—defective trafficking leading to subcellular deficiency and autophagic cell death—occurs in many other lysosomal storage diseases. To our knowledge, Danon disease, caused by deficiency in the lysosomal membrane protein LAMP-2, is the only other lysosomal storage disease that has a demonstrated increase in AVs [89]. The pattern of degeneration erupting in each disease would be determined by the cell-specific response to being deprived of a particular molecule or class of molecules. PCs appear to be highly sensitive to insufficient levels of sterols or sterol metabolites (and perhaps also to sphingolipid deficiency, since PC loss is prominent in Niemann-Pick type A as well [90,91]).
Our results are particularly interesting in light of the two other human neurodegenerative diseases that have been analyzed with chimeric mouse models. In amyotrophic lateral sclerosis, surrounding wild-type non-neuronal cells can prevent the degeneration of SOD1 mutant motor neurons, and the genotype of the motor neurons themselves appears to have no bearing on their probability of survival [23]. In the mnd mouse, a model for neuronal ceroid lipofuscinosis, proteolipid storage is cell-autonomous, but retinal degeneration equivalent to the mnd controls was seen in all chimeric mice [24]. The fact that three neurodegenerative diseases examined with this experimental approach exhibit three different modes of degeneration (non-cell-autonomous rescue for amyotrophic lateral sclerosis, cell-autonomous storage but non-cell-autonomous degeneration for neuronal ceroid lipofuscinosis, and cell-autonomous autophagic death for NPC) demonstrates the remarkable diversity in neuronal degeneration mechanisms and highlights how much still needs to be learned about the biology of neurodegeneration.
Materials and Methods
Materials.
BALB/c npc1nih, FVB.Cg-Tg(GFPU)5Nagy/J, and C57BL/6J mice were obtained from Jackson Laboratory (Bar Harbor, Maine, United States). CD-1 mice were from Charles River Laboratories (Wilmington, Massachusetts, United States). Mouse anti-Calbindin D-28K (1:250 dilution used), mouse anti-S100β (1:250) ascites, rapamycin, and filipin were obtained from Sigma (St. Louis, Missouri, United States). Rabbit anti-GABAα6 (1:250), rabbit anti-NeuN (1:100), and rat anti-myelin basic protein (1:500) were obtained from Chemicon International (Temecula, California, United States). Rat anti-F4/80 (1:100) was obtained from Serotec (Raleigh, North Carolina, United States). Mouse anti-CC1 (1:100) was from EMD Biosciences (San Diego, California, United States). Rabbit anti-MAP LC3 (1:1,000) was the kind gift of Marlene Rabinovitch and Lihua Ying (Stanford University, Palo Alto, California, United States). Nuclear dyes 7AAD and TOTO-3 were obtained from Molecular Probes (Eugene, Oregon, United States).
Generation of npc1 chimeric mice.
Aggregation chimeras were made according to established procedures [92]. Mice made up of npc1 cells and GFP-expressing wild-type cells were constructed using BALB/c npc1nih and FVB.Cg-Tg(GFPU)5Nagy/J embryos. In order to make reciprocal chimeras, npc1 heterozygous mice were mated to FVB.Cg-Tg(GFPU)5Nagy/J mice. Offspring heterozygous for npc1 and GFP were backcrossed to FVB.Cg-Tg(GFPU)5Nagy/J mice to obtain mice heterozygous for npc1 and homozygous for GFP. Sibling matings were used to generate GFP-expressing npc1−/− embryos. Wild-type embryos were obtained from CD-1 and C57BL/6J strains. Week-old pups were examined with a blue LED and blue-filtering glasses to determine the ratio of mutant to wild-type cells based on skin fluorescence. The percentage of mutant cells in chimeras made with the C57BL/6J strain was also evaluated by measuring relative areas of coat color; the results were consistent with the percentage of GFP fluorescent cells.
Tail preparation DNA was used to genotype mice using PCR primers for npc1 [6]. To distinguish between the wild-type npc1 alleles in the BALB/c, FVB.Cg-Tg(GFPU)5Nagy/J, CD-1, and C57BL/6J strains, we took advantage of a polymorphism that could be detected with an ApaLI restriction digest. Digestion of the BALB/c, CD-1, or C57BL/6J wild-type PCR product results in 477-, 340-, and 130-bp fragments; the FVB.Cg-Tg(GFPU)5Nagy/J wild-type PCR product digest results in 817- and 130-bp fragments; the mutant npc1 PCR product digest results in 937- and 130-bp fragments. Therefore, for npc1−/− ↔FVB.Cg-Tg(GFPU)5Nagy/J mice, we expect 937-, 817-, and 130-bp bands. For npc1−/−;GFP +/+↔CD-1 or C57BL/6J mice, we expect 937-, 477-, 340-, and 130-bp bands.
Mouse assays.
Mice were weighed weekly on a gram scale. For balance beam measurements, mice were placed perpendicularly on a stainless steel bar (2 cm in diameter, 120 cm in length) wrapped in laboratory tape as described [29]. Alternating 10-cm segments were marked by varying tape color. The level of ataxia was assessed based on time (180 s maximum), number of sections crossed, and qualitatively on a five-point ataxia scale: five, no ataxia; four, inability to turn around on the bar; three, difficulty walking to the end of the bar without falling off; two, the mouse can only cling to the bar and is unable to correct itself from its initial perpendicular orientation; and one, postural instability as the mouse quickly falls off the bar even when placed along the long axis.
For gait assays, front paws were dipped in red paint and back paws in green paint, and mice were placed on Whatman paper at one end of a 14 × 44 cm box. The distance between the back edge of each same-side paw print was used to determine stride length.
Cerebellar sectioning and staining.
Mice were anesthetized with avertin and perfused with 4% PFA. Brains were post-fixed for 24 h in 4% PFA, transferred to 25% sucrose overnight, embedded and frozen in OCT, and cut into 12-μm sections with a cryostat.
For immunostaining, sections were rehydrated in PBS and blocked/permeabilized with 10% normal goat serum and 0.2% Triton X-100 in PBS. Incubation with primary antibody was carried out overnight at 4 °C. Following three washes with PBS, sections were incubated for 2–4 h with the appropriate fluorescently conjugated secondary antibody. Sections were washed again and mounted with Fluormount-G (Southern Biotechnology Associates, Birmingham, Alabama, United States). Filipin staining was carried out with 50 μg/ml filipin following treatment with 0.02% saponin as described previously [79].
PC counts.
All PCs in five sections at least 100 μm apart were counted for each mouse. The length of the PC layer was measured using MetaMorph software (Universal Imaging, Downington, Pennsylvania, United States), and the density was determined by dividing the number of PCs by this length. Expected densities if no degeneration has occurred were calculated by multiplying the percentage of chimerism by the density of PCs observed in npc1 mice at 30 d. PC densities in the three chimeric mice sacrificed at 70 d were also compared to expected densities with degeneration (calculated from the percentage of chimerism and the density of PCs observed in npc1 mice at 70 d). p-Values were calculated with the unpaired t-test or one-way ANOVA with Tukey's multiple comparison test using GraphPad Prism (GraphPad Software, San Diego, California, United States).
LC3 immunoblotting.
Cell extracts were obtained by solubilization with 2% IGEPAL-CA630, 0.2% SDS, and mini complete protease inhibitor tablet (Roche Applied Science, Indianapolis, Indiana, United States) in PBS, followed by centrifugation at 10,000 RPM for 10 min. Extracts were kept on ice and resolved with 15% SDS-PAGE. Following transfer to PVDF membrane (Bio-Rad, Hercules, California, United States), blocking was accomplished with 5% milk in PBS plus 0.1% Triton X-100. Primary incubation with anti-LC3 (rabbit, 1:1,000 in 1% milk PBST) was carried out overnight at 4 °C. Following three washes, the membrane was incubated with anti-rabbit-HRP (1:10,000), washed three times, and developed with WestPico reagent (Pierce Biotechnology, Rockford, Illinois, United States).
Electron microscopy.
Mice were anesthetized using avertin and were perfused with PBS followed by a solution of 2.5% glutaraldehyde and 4% paraformaldehyde in 0.1 M cacodylate buffer (pH 7.4) for 15 min using gravity flow. For microwave processing, a Pelco laboratory microwave equipped with variable wattage, Cold Spot water recirculator, and vacuum chamber was used (Ted Pella, Mountain Lakes, California, United States). Brains were removed, immersed in the same fixative, and microwaved at 100 W for 1 min on, 1 min off, 1 min on, with the Cold Spot set at 15 °C. A vacuum chamber was applied and the brains microwaved at 450 W for 20 s on, 20 s off, 20 s on, three times. After an additional 15 min at room temperature, the cerebellum was dissected away and sliced into anterior and posterior samples. Post-fixation was carried out using 2% osmium tetroxide containing 0.8% potassium ferricyanide and 5% sucrose in 0.1 M cacodylate buffer (pH 7.4) using the same microwave parameters as for primary fixation and followed by an additional 45 min at room temperature. After rinsing in distilled water, the tissue was dehydrated in an ascending alcohol (50%–95%) series at 350 W for 45 s, then switched to 100% ethanol under vacuum for 1 min on, 1 min off, 1 min on, three times. The tissue was microwaved in 100% acetone for 1 min on, 1 min off, 1 min on, at power level 3 using the vacuum chamber. The tissue was then infiltrated in Epon-Araldite-acetone mixture in the microwave for 4 min on, 4 min off, 4 min on, at 350 W for each resin mixture (1:2, 1:1, 2:1) under vacuum. For the final infiltration, the samples were placed in 100% resin and microwaved at 350 W under vacuum for 10 min on, 10 min off, 10 min on. The samples were embedded in fresh Epon-Araldite and hardened at 65 °C for 1–2 d. Semi-thin sections (1–2 μm) were cut using a Histoknife (Diatome, Fort Washington, Pennsylvania, United States) and stained with toludine blue. Thin sections (70 nm) were cut and mounted on formvar-coated multi-slot grids. Sections were post-stained with 5% aqueous uranyl acetate and lead citrate and examined at 80 kV in a JEOL (Tokyo, Japan) 1230 electron microscope. Digital images were captured with a Gatan (Pleasanton, California, United States) 967 slow-scan, cooled CCD camera. Images were analyzed with ImageJ Software (National Institutes of Health, Bethesda, Maryland, United States).
Supporting Information
Accession Numbers
The OMIM (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=OMIM) accession number for NPC is 257220. The NCBI Entrez (http://www.ncbi.nih.gov/entrez/query.fcgi?db=gene) accession numbers for the genes and gene products discussed in this paper are LC3 (GeneID 84557), npc1 (GeneID 4864), and npc2 (GeneID 10577).
We thank the Ara Parseghian Medical Research Foundation for their support of this research. We thank Marlene Rabinovitch and Lihua Ying for the rabbit polyclonal LC3 antibody. We thank Nat Heintz, Zhenyu Yue, and William Jackson for helpful discussions about autophagy. We thank Xun Huang and Emily Ray for comments on the manuscript. DCK was supported by a Medical Scientist Training Grant. MPS is an Investigator of the Howard Hughes Medical Institute.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. DCK, LM, SMB, and MPS conceived and designed the experiments. DCK, LM, HM, and JB performed the experiments. DCK analyzed the data. DCK, LM, SMB, HM, and JB contributed reagents/materials/analysis tools. DCK, LM, and MPS wrote the paper.
Abbreviations
AVautophagic vacuole
C[number]chimera [number]
CHOChinese hamster ovary
EMelectron microscopy
ERendoplasmic reticulum
GFPgreen fluorescent protein
LC3light chain 3
NPCNiemann-Pick type C
PCPurkinje cell
==== Refs
References
Patterson MC Vanier MT Suzuki K Morris JA Carstea ED 2001 Niemann-Pick disease type C: A lipid trafficking disorder Scriver CR Beaudet AL Sly WS Valle D The metabolic and molecular bases of inherited disease New York McGraw Hill 36110 33633
Gieselmann V 1995 Lysosomal storage diseases Biochim Biophys Acta 1270 103 136 7727535
Futerman AH van Meer G 2004 The cell biology of lysosomal storage disorders Nat Rev Mol Cell Biol 5 554 565 15232573
Meikle PJ Hopwood JJ Clague AE Carey WF 1999 Prevalence of lysosomal storage disorders JAMA 281 249 254 9918480
Carstea ED Morris JA Coleman KG Loftus SK Zhang D 1997 Niemann-Pick C1 disease gene: Homology to mediators of cholesterol homeostasis Science 277 228 231 9211849
Loftus SK Morris JA Carstea ED Gu JZ Cummings C 1997 Murine model of Niemann-Pick C disease: Mutation in a cholesterol homeostasis gene Science 277 232 235 9211850
Naureckiene S Sleat DE Lackland H Fensom A Vanier MT 2000 Identification of HE1 as the second gene of Niemann-Pick C disease Science 290 2298 2301 11125141
Ohgami N Ko DC Thomas M Scott MP Chang CC 2004 Binding between the Niemann-Pick C1 protein and a photoactivatable cholesterol analog requires a functional sterol-sensing domain Proc Natl Acad Sci U S A 101 12473 2478 15314240
Davies JP Chen FW Ioannou YA 2000 Transmembrane molecular pump activity of Niemann-Pick C1 protein Science 290 2295 2298 11125140
Okamura N Kiuchi S Tamba M Kashima T Hiramoto S 1999 A porcine homolog of the major secretory protein of human epididymis, HE1, specifically binds cholesterol Biochim Biophys Acta 1438 377 387 10366780
Friedland N Liou HL Lobel P Stock AM 2003 Structure of a cholesterol-binding protein deficient in Niemann-Pick type C2 disease Proc Natl Acad Sci U S A 100 2512 2517 12591954
Ko DC Binkley J Sidow A Scott MP 2003 The integrity of a cholesterol-binding pocket in Niemann-Pick C2 protein is necessary to control lysosome cholesterol levels Proc Natl Acad Sci U S A 100 2518 2525 12591949
Miyawaki S Mitsuoka S Sakiyama T Kitagawa T 1982 Sphingomyelinosis, a new mutation in the mouse: A model of Niemann-Pick disease in humans J Hered 73 257 263 7202025
Pentchev PG Gal AE Booth AD Omodeo-Sale F Fouks J 1980 A lysosomal storage disorder in mice characterized by a dual deficiency of sphingomyelinase and glucocerebrosidase Biochim Biophys Acta 619 669 679 6257302
Higashi Y Murayama S Pentchev PG Suzuki K 1993 Cerebellar degeneration in the Niemann-Pick type C mouse Acta Neuropathol (Berl) 85 175 184 8382896
Tanaka J Nakamura H Miyawaki S 1988 Cerebellar involvement in murine sphingomyelinosis: A new model of Niemann-Pick disease J Neuropathol Exp Neurol 47 291 300 3130465
German DC Quintero EM Liang CL Ng B Punia S 2001 Selective neurodegeneration, without neurofibrillary tangles, in a mouse model of Niemann-Pick C disease J Comp Neurol 433 415 425 11298365
Patel SC Suresh S Kumar U Hu CY Cooney A 1999 Localization of Niemann-Pick C1 protein in astrocytes: Implications for neuronal degeneration in Niemann- Pick type C disease Proc Natl Acad Sci U S A 96 1657 1662 9990080
Ong WY Kumar U Switzer RC Sidhu A Suresh G 2001 Neurodegeneration in Niemann-Pick type C disease mice Exp Brain Res 141 218 231 11713633
Prasad A Fischer WA Maue RA Henderson LP 2000 Regional and developmental expression of the Npc1 mRNA in the mouse brain J Neurochem 75 1250 1257 10936208
Tabas I 2002 Consequences of cellular cholesterol accumulation: Basic concepts and physiological implications J Clin Invest 110 905 911 12370266
Mauch DH Nagler K Schumacher S Goritz C Muller EC 2001 CNS synaptogenesis promoted by glia-derived cholesterol Science 294 1354 1357 11701931
Clement AM Nguyen MD Roberts EA Garcia ML Boillee S 2003 Wild-type nonneuronal cells extend survival of SOD1 mutant motor neurons in ALS mice Science 302 113 117 14526083
Lipman RD Donohue LR Hoppe P Bronson RT 1996 Evidence that lysosomal storage of proteolipids is a cell autonomous process in the motor neuron degeneration (mnd) mouse, a model of neuronal ceroid lipofuscinosis Neurosci Lett 219 111 114 8971792
Yamada K Watanabe M 2002 Cytodifferentiation of Bergmann glia and its relationship with Purkinje cells Anat Sci Int 77 94 108 12418089
March PA Thrall MA Brown DE Mitchell TW Lowenthal AC 1997 GABAergic neuroaxonal dystrophy and other cytopathological alterations in feline Niemann-Pick disease type C Acta Neuropathol (Berl) 94 164 172 9255392
Hadjantonakis AK Gertsenstein M Ikawa M Okabe M Nagy A 1998 Generating green fluorescent mice by germline transmission of green fluorescent ES cells Mech Dev 76 79 90 9867352
German DC Liang CL Song T Yazdani U Xie C 2002 Neurodegeneration in the Niemann-Pick C mouse: Glial involvement Neuroscience 109 437 450 11823057
Voikar V Rauvala H Ikonen E 2002 Cognitive deficit and development of motor impairment in a mouse model of Niemann-Pick type C disease Behav Brain Res 132 1 10 11853852
Loftus SK Erickson RP Walkley SU Bryant MA Incao A 2002 Rescue of neurodegeneration in Niemann-Pick C mice by a prion-promoter-driven Npc1 cDNA transgene Hum Mol Genet 11 3107 3114 12417532
Yuan J Lipinski M Degterev A 2003 Diversity in the mechanisms of neuronal cell death Neuron 40 401 413 14556717
Wu YP Mizukami H Matsuda J Saito Y Proia RL 2005 Apoptosis accompanied by up-regulation of TNF-alpha death pathway genes in the brain of Niemann-Pick type C disease Mol Genet Metab 84 9 17 15639190
Dunn WA Jr 1990 Studies on the mechanisms of autophagy: Maturation of the autophagic vacuole J Cell Biol 110 1935 1945 2161853
Dunn WA Jr 1990 Studies on the mechanisms of autophagy: Formation of the autophagic vacuole J Cell Biol 110 1923 1933 2351689
Liou W Geuze HJ Geelen MJ Slot JW 1997 The autophagic and endocytic pathways converge at the nascent autophagic vacuoles J Cell Biol 136 61 70 9008703
Tanida I Tanida-Miyake E Komatsu M Ueno T Kominami E 2002 Human Apg3p/Aut1p homologue is an authentic E2 enzyme for multiple substrates, GATE-16, GABARAP, and MAP-LC3, and facilitates the conjugation of hApg12p to hApg5p J Biol Chem 277 13739 13744 11825910
Tanida I Sou YS Ezaki J Minematsu-Ikeguchi N Ueno T 2004 HsAtg4B/HsApg4B/autophagin-1 cleaves the carboxyl termini of three human Atg8 homologues and delipidates microtubule-associated protein light chain 3- and GABAA receptor-associated protein-phospholipid conjugates J Biol Chem 279 36268 36276 15187094
Kabeya Y Mizushima N Ueno T Yamamoto A Kirisako T 2000 LC3, a mammalian homologue of yeast Apg8p, is localized in autophagosome membranes after processing EMBO J 19 5720 5728 11060023
Mann SS Hammarback JA 1996 Gene localization and developmental expression of light chain 3: A common subunit of microtubule-associated protein 1A(MAP1A) and MAP1B J Neurosci Res 43 535 544 8833088
Cadigan KM Spillane DM Chang TY 1990 Isolation and characterization of Chinese hamster ovary cell mutants defective in intracellular low density lipoprotein-cholesterol trafficking J Cell Biol 110 295 308 2404988
Lansbury PT Jr 2004 Back to the future: The “old-fashioned” way to new medications for neurodegeneration Nat Med 10 S51 S57 15298008
Citron M 2004 Strategies for disease modification in Alzheimer's disease Nat Rev Neurosci 5 677 685 15322526
Meyer-Luehmann M Stalder M Herzig MC Kaeser SA Kohler E 2003 Extracellular amyloid formation and associated pathology in neural grafts Nat Neurosci 6 370 377 12598899
Zhang Y McLaughlin R Goodyer C LeBlanc A 2002 Selective cytotoxicity of intracellular amyloid beta peptide1–42 through p53 and Bax in cultured primary human neurons J Cell Biol 156 519 529 11815632
Lim GP Yang F Chu T Chen P Beech W 2000 Ibuprofen suppresses plaque pathology and inflammation in a mouse model for Alzheimer's disease J Neurosci 20 5709 5714 10908610
Weggen S Eriksen JL Das P Sagi SA Wang R 2001 A subset of NSAIDs lower amyloidogenic Abeta42 independently of cyclooxygenase activity Nature 414 212 216 11700559
Anzil AP Blinzinger K Mehraein P Dozic S 1973 Niemann-Pick disease type C: Case report with ultrastructural findings Neuropadiatrie 4 207 225 4740443
Harzer K Schlote W Peiffer J Benz HU Anzil AP 1978 Neurovisceral lipidosis compatible with Niemann-Pick disease type C: Morphological and biochemical studies of a late infantile case and enzyme and lipid assays in a prenatal case of the same family Acta Neuropathol (Berl) 43 97 104 209660
Patterson MC 2003 A riddle wrapped in a mystery: Understanding Niemann-Pick disease, type C Neurologist 9 301 310 14629784
Mutka AL Lusa S Linder MD Jokitalo E Kopra O 2004 Secretion of sterols and the npc2 protein from primary astrocytes J Biol Chem 279 48654 48662 15355983
Karten B Hayashi H Francis GA Campenot RB Vance DE 2004 Generation and function of astroglial lipoproteins from Niemann-Pick type C1-deficient mice Biochem J 387 779 788
Sotelo C 2004 Cellular and genetic regulation of the development of the cerebellar system Prog Neurobiol 72 295 339 15157725
Dusart I Sotelo C 1994 Lack of Purkinje cell loss in adult rat cerebellum following protracted axotomy: Degenerative changes and regenerative attempts of the severed axons J Comp Neurol 347 211 232 7814665
Morel MP Dusart I Sotelo C 2002 Sprouting of adult Purkinje cell axons in lesioned mouse cerebellum: “Non-permissive” versus “permissive” environment J Neurocytol 31 633 647 14501204
Thery C Chamak B Mallat M 1991 Cytotoxic effect of brain macrophages on developing Eur J Neurosci 3 1155 1164 12106245
Combs CK Karlo JC Kao SC Landreth GE 2001 Beta-amyloid stimulation of microglia and monocytes results in TNFalpha-dependent expression of inducible nitric oxide synthase and neuronal apoptosis J Neurosci 21 1179 1188 11160388
Flavin MP Zhao G Ho LT 2000 Microglial tissue plasminogen activator (tPA) triggers neuronal apoptosis in vitro Glia 29 347 354 10652444
Marin-Teva JL Dusart I Colin C Gervais A van Rooijen N 2004 Microglia promote the death of developing Purkinje cells Neuron 41 535 547 14980203
Liu Y Wu YP Wada R Neufeld EB Mullin KA 2000 Alleviation of neuronal ganglioside storage does not improve the clinical course of the Niemann-Pick C disease mouse Hum Mol Genet 9 1087 1092 10767333
German DC Quintero EM Liang C Xie C Dietschy JM 2001 Degeneration of neurons and glia in the Niemann-Pick C mouse is unrelated to the low-density lipoprotein receptor Neuroscience 105 999 1005 11530237
Griffin LD Gong W Verot L Mellon SH 2004 Niemann-Pick type C disease involves disrupted neurosteroidogenesis and responds to allopregnanolone Nat Med 10 704 711 15208706
Tsutsui K Sakamoto H Ukena K 2003 A novel aspect of the cerebellum: Biosynthesis of neurosteroids in the Purkinje cell Cerebellum 2 215 222 14509571
Palay SL Chan-Palay V 1974 Cerebellar cortex: Cytology and organization New York Springer 348 p.
Yue Z Horton A Bravin M DeJager PL Selimi F 2002 A novel protein complex linking the delta 2 glutamate receptor and autophagy: Implications for neurodegeneration in lurcher mice Neuron 35 921 933 12372286
Cuervo AM Stefanis L Fredenburg R Lansbury PT Sulzer D 2004 Impaired degradation of mutant alpha-synuclein by chaperone-mediated autophagy Science 305 1292 1295 15333840
Ravikumar B Duden R Rubinsztein DC 2002 Aggregate-prone proteins with polyglutamine and polyalanine expansions are degraded by autophagy Hum Mol Genet 11 1107 1117 11978769
Ravikumar B Vacher C Berger Z Davies JE Luo S 2004 Inhibition of mTOR induces autophagy and reduces toxicity of polyglutamine expansions in fly and mouse models of Huntington disease Nat Genet 36 585 595 15146184
Davies SW Turmaine M Cozens BA DiFiglia M Sharp AH 1997 Formation of neuronal intranuclear inclusions underlies the neurological dysfunction in mice transgenic for the HD mutation Cell 90 537 548 9267033
Meriin AB Zhang X He X Newnam GP Chernoff YO 2002 Huntington toxicity in yeast model depends on polyglutamine aggregation mediated by a prion-like protein Rnq1 J Cell Biol 157 997 1004 12058016
Lee MK Stirling W Xu Y Xu X Qui D 2002 Human alpha-synuclein-harboring familial Parkinson's disease-linked Ala-53 –> Thr mutation causes neurodegenerative disease with alpha-synuclein aggregation in transgenic mice Proc Natl Acad Sci U S A 99 8968 8973 12084935
Giasson BI Duda JE Quinn SM Zhang B Trojanowski JQ 2002 Neuronal alpha-synucleinopathy with severe movement disorder in mice expressing A53T human alpha-synuclein Neuron 34 521 533 12062037
Liscum L Faust JR 1987 Low density lipoprotein (LDL)-mediated suppression of cholesterol synthesis and LDL uptake is defective in Niemann-Pick type C fibroblasts J Biol Chem 262 17002 17008 3680287
Horton JD Shah NA Warrington JA Anderson NN Park SW 2003 Combined analysis of oligonucleotide microarray data from transgenic and knockout mice identifies direct SREBP target genes Proc Natl Acad Sci U S A 100 12027 12032 14512514
Pfeifer U 1978 Inhibition by insulin of the formation of autophagic vacuoles in rat liver. A morphometric approach to the kinetics of intracellular degradation by autophagy J Cell Biol 78 152 167 670291
Weeks JC 2003 Thinking globally, acting locally: Steroid hormone regulation of the dendritic architecture, synaptic connectivity and death of an individual neuron Prog Neurobiol 70 421 442 14511700
Kinch G Hoffman KL Rodrigues EM Zee MC Weeks JC 2003 Steroid-triggered programmed cell death of a motoneuron is autophagic and involves structural changes in mitochondria J Comp Neurol 457 384 403 12561078
Florez-McClure ML, Linseman DA, Chu CT, Barker PA, Bouchard RJ, et al. 2004 The p75 neurotrophin receptor can induce autophagy and death of cerebellar Purkinje neurons J Neurosci 24 4498 4509 15140920
Henderson LP Lin L Prasad A Paul CA Chang TY 2000 Embryonic striatal neurons from Niemann-Pick type C mice exhibit defects in cholesterol metabolism and neurotrophin responsiveness J Biol Chem 275 20179 20187 10770933
Zervas M Dobrenis K Walkley SU 2001 Neurons in Niemann-Pick disease type C accumulate gangliosides as well as unesterified cholesterol and undergo dendritic and axonal alterations J Neuropathol Exp Neurol 60 49 64 11202175
Sarna JR Larouche M Marzban H Sillitoe RV Rancourt DE 2003 Patterned Purkinje cell degeneration in mouse models of Niemann-Pick type C disease J Comp Neurol 456 279 291 12528192
Grusser-Cornehls U Baurle J 2001 Mutant mice as a model for cerebellar ataxia Prog Neurobiol 63 489 540 11164620
Wetts R Herrup K 1982 Interaction of granule, Purkinje and inferior olivary neurons in lurcher chimaeric mice. I. Qualitative studies J Embryol Exp Morphol 68 87 98 7108427
Martin LA Goldowitz D Mittleman G 2003 The cerebellum and spatial ability: Dissection of motor and cognitive components with a mouse model system Eur J Neurosci 18 2002 2010 14622233
Martin LA Escher T Goldowitz D Mittleman G 2004 A relationship between cerebellar Purkinje cells and spatial working memory demonstrated in a lurcher/chimera mouse model system Genes Brain Behav 3 158 166 15140011
Yamada A Saji M Ukita Y Shinoda Y Taniguchi M 2001 Progressive neuronal loss in the ventral posterior lateral and medial nuclei of thalamus in Niemann-Pick disease type C mouse brain Brain Dev 23 288 297 11504598
Ohara S Ukita Y Ninomiya H Ohno K 2004 Axonal dystrophy of dorsal root ganglion sensory neurons in a mouse model of Niemann-Pick disease type C Exp Neurol 187 289 298 15144855
Ko DC Gordon MD Jin JY Scott MP 2001 Dynamic movements of organelles containing Niemann-Pick C1 protein: NPC1 involvement in late endocytic events Mol Biol Cell 12 601 614 11251074
Zhang M Dwyer NK Love DC Cooney A Comly M 2001 Cessation of rapid late endosomal tubulovesicular trafficking in Niemann-Pick type C1 disease Proc Natl Acad Sci U S A 98 4466 4471 11296289
Tanaka Y Guhde G Suter A Eskelinen EL Hartmann D 2000 Accumulation of autophagic vacuoles and cardiomyopathy in LAMP-2-deficient mice Nature 406 902 906 10972293
Otterbach B Stoffel W 1995 Acid sphingomyelinase-deficient mice mimic the neurovisceral form of human lysosomal storage disease (Niemann-Pick disease) Cell 81 1053 1061 7600574
Horinouchi K Erlich S Perl DP Ferlinz K Bisgaier CL 1995 Acid sphingomyelinase deficient mice: A model of types A and B Niemann-Pick disease Nat Genet 10 288 293 7670466
Joyner AL 1993 Gene targeting: A practical approach Oxford IRL Press at Oxford University Press 234 p.
|
16103921
|
PMC1183526
|
CC BY
|
2021-01-05 08:00:57
|
no
|
PLoS Genet. 2005 Jul 25; 1(1):e7
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010007
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 10.1371/journal.pgen.001000805-PLGE-RA-0034R3plge-01-01-11Research ArticleBioinformatics - Computational BiologyCancer BiologyCell BiologyDiabetes - Endocrinology - MetabolismMolecular Biology - Structural BiologyPathologyGenetics/GenomicsGenetics/Genetics of DiseaseGenetics/Disease ModelsGenetics/Gene ExpressionHomo (human)In VitroA HIF1α Regulatory Loop Links Hypoxia and Mitochondrial Signals in Pheochromocytomas Hypoxia/Energy Link in PheochromocytomaDahia Patricia L. M 1*¤Ross Ken N 2Wright Matthew E 1Hayashida César Y 3Santagata Sandro 4Barontini Marta 5Kung Andrew L 6Sanso Gabriela 5Powers James F 7Tischler Arthur S 7Hodin Richard 8Heitritter Shannon 4Moore Francis Jr.4Dluhy Robert 4Sosa Julie Ann 9Ocal I. Tolgay 9Benn Diana E 10Marsh Deborah J 10Robinson Bruce G 10Schneider Katherine 11Garber Judy 11Arum Seth M 12Korbonits Márta 13Grossman Ashley 13Pigny Pascal 14Toledo Sérgio P. A 3Nosé Vania 4Li Cheng 15Stiles Charles D 11 Department of Cancer Biology, Dana-Farber Cancer Institute, Harvard Medical School, Boston, Massachusetts, United States of America
2 Broad Institute, Massachusetts Institute of Technology, Cambridge, Massachusetts, United States of America
3 University of São Paulo School of Medicine, São Paulo, Brazil
4 Brigham and Women's Hospital, Boston, Massachusetts, United States of America
5 Center of Endocrine Investigations, Hospital de Niños R. Gutierrez, Buenos Aires, Argentina
6 Department of Pediatric Oncology, Dana-Farber Cancer Institute, Harvard Medical School, Boston, Massachusetts, United States of America
7 Tufts–New England Medical Center, Boston, Massachusetts, United States of America
8 Massachusetts General Hospital, Boston, Massachusetts, United States of America
9 Yale University, New Haven, Connecticut, United States of America
10 Royal North Shore Hospital and Kolling Institute of Medical Research, University of Sydney, Australia
11 Division of Population Sciences, Dana-Farber Cancer Institute, Harvard Medical School, Boston, Massachusetts, United States of America
12 Boston Medical Center, Boston, Massachusetts, United States of America
13 St. Bartholomew's Hospital, London, United Kingdom
14 Regional University Hospital, Lille, France
15 Department of Biostatistical Science, Dana-Farber Cancer Institute, Harvard Medical School, Boston, Massachusetts, United States of America
Frankel Wayne N EditorThe Jackson Laboratory, United States of America*To whom correspondence should be addressed. E-mail: [email protected]¤Current address: Department of Medicine and Cellular and Structural Biology, University of Texas Health Science Center, San Antonio, Texas, United States of America
7 2005 25 7 2005 1 1 e828 2 2005 9 5 2005 Copyright: © 2005 Dahia et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Pheochromocytomas are neural crest–derived tumors that arise from inherited or sporadic mutations in at least six independent genes. The proteins encoded by these multiple genes regulate distinct functions. We show here a functional link between tumors with VHL mutations and those with disruption of the genes encoding for succinate dehydrogenase (SDH) subunits B (SDHB) and D (SDHD). A transcription profile of reduced oxidoreductase is detected in all three of these tumor types, together with an angiogenesis/hypoxia profile typical of VHL dysfunction. The oxidoreductase defect, not previously detected in VHL-null tumors, is explained by suppression of the SDHB protein, a component of mitochondrial complex II. The decrease in SDHB is also noted in tumors with SDHD mutations. Gain-of-function and loss-of-function analyses show that the link between hypoxia signals (via VHL) and mitochondrial signals (via SDH) is mediated by HIF1α. These findings explain the shared features of pheochromocytomas with VHL and SDH mutations and suggest an additional mechanism for increased HIF1α activity in tumors.
Synopsis
Pheochromocytomas (also known as paragangliomas) are highly vascular tumors that arise from mutations in a diverse and apparently unrelated group of tumor suppressor genes and oncogenes. The authors show here that three of the genes that cause hereditary pheochromocytomas have a common function. Specifically, these genes, VHL,
SDHB, and SDHD, encode proteins that regulate a transcription factor known as hypoxia-inducible factor 1 subunit α (HIF1α), which helps cells adapt to hypoxia (low oxygen levels). VHL is named after its role in von Hippel-Lindau disease (VHL), an inherited disorder that predisposes individuals to pheochromocytomas and other tumors. Previous studies showed that when cells lack VHL, HIF1α is not degraded, resulting in a signal that resembles hypoxia. The authors found that loss of two genes that cause two distinct pheochromocytoma syndromes (the genes SDHB and SDHD, which encode the subunits B and D of succinate dehydrogenase, a component enzyme of the energy and respiratory system in mitochondria) also triggers a HIF1α response. The researchers further discovered that high H1F1α levels can suppress SDHB. This suggests a regulatory loop that further enhances the “hypoxia” profile of tumors. This finding provides a rational explanation for the shared features of these distinct syndromes and may be relevant for other cancers with a prominent hypoxic pattern.
Citation:Dahia PLM, Ross KN, Wright ME, Hayashida CY, Santagata S, et al. (2005) A HIF1α regulatory loop links hypoxia and mitochondrial signals in pheochromocytomas. PLoS Genet 1(1): e8.
==== Body
Introduction
Adrenal and extra-adrenal pheochromocytomas (also known as paragangliomas) are catecholamine-secreting tumors derived from chromaffin cells of neural crest origin [1]. Pheochromocytomas can arise as a result of mutations in the following disease-associated genes: RET in multiple endocrine neoplasia type 2 (MEN2); VHL in von Hippel-Lindau disease (VHL); NF1 in neurofibromatosis type 1 (NF1); and succinate dehydrogenase (SDH)
subunits B, C, or D in familial paraganglioma syndromes type 4 (PGL4), type 3 (PGL3), and type 1 (PGL1), respectively [2]. The various pheochromocytoma susceptibility genes modulate a variety of signaling pathways that are superficially unrelated to one another. However, the uniform phenotype of the tumors that arise from these distinct genetic lesions suggests the presence of underlying biochemical links.
The VHL tumor suppressor is a key mediator of the hypoxia response. It targets the hypoxia-inducible factor 1 subunit α (HIF1α) for ubiquitin-mediated degradation under normal oxygen conditions [3]. HIF1α has been shown to be critical for the oncogenic effects resulting from VHL mutations in specific cellular contexts [4,5]. Two other genes related to familial paraganglioma, SDHB and SDHD, encode subunits of SDH, the enzyme that composes mitochondrial complex II [6,7]. This enzyme is both a component of the Krebs cycle, by oxidizing succinate to fumarate, and of the mitochondrial respiratory chain, by transferring electrons to the ubiquinone pool [8]. Familial paragangliomas associated with SDHB and SDHD mutations resemble the carotid body growths that occur as a result of chronic hypoxia exposure in individuals living at high altitudes [6]. These clinical observations and the finding of increased expression of HIF targets in tumors with SDH mutations [9,10] have suggested the possibility that the VHL and SDH syndromes intersect at the molecular level.
As an entry-level screen for interacting signals, we generated global expression signatures of 76 hereditary and sporadic primary pheochromocytomas and paragangliomas. We show here that pheochromocytomas with VHL and SDHB or SDHD mutations form a tight cluster with a clear hypoxia and reduced oxidoreductase signature. This observation led to the identification of suppressed SDHB protein in tumors with VHL mutation and to the genetic demonstration that this effect is HIF-dependent. Our findings link pheochromocytomas with mutations in distinct genes—VHL,
SDHB, and SDHD—and suggest that mitochondrial complex II inhibition contributes to development of pheochromocytomas with VHL mutation.
Results
Expression Profiling Links Pheochromocytomas with VHL and SDHB or SDHD Mutations
Unsupervised hierarchical cluster analysis of a cohort of 76 sporadic and hereditary pheochromocytomas (Dataset S1) identified two dominant expression clusters (Figure 1; Dataset S2). Cluster 1 comprised all VHL and SDH tumors. Cluster 2 contained all MEN2 and NF1 pheochromocytomas. The remaining unknown familial tumors and 37 sporadic samples were partitioned into one or the other of the two major clusters. Cluster 1 contained 12 of the 13 extra-adrenal tumors. However, the bipartite distribution of tumor sets is not a simple reflection of anatomical location of the tumor because more than half of Cluster 1 tumors are adrenal in origin (Figure 1).
Figure 1 Unsupervised Analysis of Pheochromocytomas Links Tumors with VHL and SDHB or SDHD Mutations
Unsupervised hierarchical clustering identifies two major clusters in pheochromocytomas: Cluster 1 contains VHL (V), SDHD (D), and SDHB (B) tumors; Cluster 2 contains MEN2 (M) and NF1 (N) pheochromocytomas. Multiple tumors from seven independent unclassified families with recurrent pheochromocytoma (numbered 1–7) and also sporadic tumors (S) are distributed between the two clusters. Letters or numbers on the first row indicate the various tumor classes, as described above. The second row identifies tumor location as adrenal (A) or extra-adrenal (E). Mutations were later detected in samples marked with an asterisk, guided by cluster distribution (see text and Table 1 for details).
Table 1 Mutations Identified in Previously Unclassified Hereditary and Sporadic Pheochromocytomas and Their Association with Expression Cluster Distribution
aN/A indicates no familial history; these are considered “sporadic” cases.
bGermline DNA unavailable.
cThis mutation results in exon 5 skipping, leading to premature truncation of the predicted protein at 248 amino acids. Samples 92, 93, and 152 are from the same family.
To validate the expression clusters, we initially confirmed the expression difference between the two clusters by quantitative real-time PCR or Western blot analysis of genes identified by the unsupervised analysis (Dataset S3). Next, we sequenced all familial samples and also 20 of the sporadic tumors for mutations in known pheochromocytoma-associated genes. We detected novel VHL,
SDHB, and RET mutations in six samples derived from four independent families (Table 1). In all cases, the mutations resided in component genes predicted by the cluster distribution.
All VHL tumors, including the newly identified mutated samples in Cluster 1, as well as the MEN2 tumors of Cluster 2, were used to generate two-class predictors (Dataset S4). Likewise, gene predictors were also created by comparing MEN2 and another component of Cluster 1, SDH tumors (including both SDHB and SDHD mutants). An extensive overlap was seen between the genes that discriminate MEN2 from SDH tumors and those that distinguish MEN2 from VHL tumors (Figure 2; Dataset S4). Over 92% of the entire sample cohort was correctly assigned to one of the two classes, in agreement with the unsupervised clustering distribution (Dataset S4). These results outline the molecular similarities between subcomponent tumors of Cluster 1 (VHL and SDH) and validate the two expression clusters detected by the unsupervised analysis (see Figure 1).
Figure 2 Similarity between VHL and SDH Tumors from Cluster 1 by Supervised Learning Methods
Supervised analysis reveals an extensive overlap between genes that discriminate MEN2 from VHL (left) and MEN2 from SDH tumors (right). Samples are shown in columns and genes are represented in rows. Expression levels are normalized for each gene, where the mean is zero. Red indicates high-level expression and blue, low-level expression. The color scale at the bottom indicates relative expression and standard deviations from the mean. Some representative genes are displayed in a color-coded manner according to their functional classes (green, kinase receptor signaling and adrenergic metabolism; pink, oxidative response; blue, hypoxia-responsive/angiogenesis genes). Within each of these functional classes the order of gene appearance in the heat map has been maintained for each class comparison. Complete gene lists are available as Dataset S4.
Suppressed SDHB Expression Is a Common Feature of VHL and SDH Tumors
Cluster 1 tumors are associated with a set of biological programs that differs markedly from Cluster 2 pheochromocytomas (Table 2; Figure 2; Dataset S4). These tumors display a rich signature of angiogenesis, hypoxia, extracellular matrix elements, and coordinated suppression of oxidoreductase enzymes. The first three components have been associated with mutations in the VHL gene, whose product is a known regulator of the activity of the HIF pathway [11]. Of note, pheochromocytoma was the sole manifestation in one of the VHL families in this series (tumors 155 and 156; see Figure 1). These features are suggestive of VHL type 2C, which has been deemed HIF-independent [12,13]. These tumors clustered with the remaining VHL samples in the expression profiling analysis, suggesting that transcription similarities between these cases and the remaining VHL tumors outnumber differences that they may bear. However, only long-term follow-up, not available in this kindred, can precisely define these tumors as VHL type 2C. The similar transcription profile of pheochromocytomas with mutations in SDH subunits indicates that the mechanism by which these tumors develop also involves the hypoxia-sensing pathway. Owing to the limited amount of tumor material, we were not able to quantitate HIF1α and HIF2α protein levels in our primary samples.
Table 2 Gene-Set Enrichment Analysis (GSEA) of Pathways Significantly Represented in Three Genetic Classes of Pheochromocytoma
aNES is calculated as described in Materials and Methods; negative sign indicates inverse correlation of the indicated pathway in the specific genetic class of tumors.
bNOM p-value is the significance of pathway enrichment as described in Materials and Methods; pathways are ordered by the NOM p-value.
CCR3, chemokine (C-C motif) receptor 3; CNS, central nervous system; IGF1, insuline-like growth factor; MAP kinase, mitogen-activated protein kinase; NGF, nerve growth factor.
Another element of the Cluster 1 signature—a synchronized suppression of mitochondrial functions—was noted by predominantly reduced expression of components of the oxidative response and Krebs cycle in both the unsupervised and supervised analyses of these tumors. This profile generated multiple significant scores by the pathway-enrichment analysis method, gene set enrichment analysis (GSEA) (Table 2), which measures the degree of inverse correlation of oxidative pathways of a rank-ordered gene list derived from the pairwise comparisons. Depressed oxidoreductase function has not been previously linked to a HIF-mediated signature. Mitochondrial complex II is a component of the electron transport chain, and mutations of SDHB or SDHD genes that abrogate the oxidoreductase function of complex II can cause pheochromocytomas [9,14,15]. Because of the role of SDH subunits as tumor suppressors, we reasoned that the oxidoreductase signature observed in pheochromocytomas from Cluster 1 (Table 2) might indicate that complex II disruption could contribute to other tumors besides those with SDH mutations. This prompted us to examine the link between pheochromocytomas with VHL and SDH mutations in our series by first determining the protein expression of the catalytic unit of complex II, SDHB. We found that the expression of SDHB is reduced in all tumors with SDH (both SDHB and SDHD) mutations in this cohort (Figure 3A), indicating that low SDHB expression functions as a surrogate for disruption of complex II. Importantly, suppressed SDHB levels were also found in the majority of tumors with VHL mutations and sporadic pheochromocytomas from Cluster 1 tested by immunoblots (Figure 3B). To confirm this finding, we performed immunostaining of SDHB in pheochromocytomas or paragangliomas representative of the various genetic syndromes using available paraffin-embedded material. SDHB immunostaining was highly concordant with the immunoblot findings (Figure 3C), suggesting that SDHB downregulation may be a feature of a broader group of pheochromocytomas. Although SDHB mRNA was collectively lower in Cluster 1 tumors than in Cluster 2 pheochromocytomas (Dataset S5), levels of the SDHB protein did not exactly parallel mRNA abundance in individual tumor samples, suggesting that a transcription defect cannot entirely account for the differences in SDHB expression observed at the protein level.
Figure 3 Low Expression of SDHB Is a General Feature of Cluster 1 Tumors
(A) Expression of SDHB protein in pheochromocytomas with SDHB or SDHD mutations. Western blot analysis of SDHB of whole cell lysates from primary tumors was performed as described in Methods. Lane 1 is normal adrenal medulla used as control and lanes 2–6 are tumors 140, 158, 136, 58, and 220, respectively, from Figure 1 and Table 1. β-actin was used as a loading control.
(B) SDHB expression segregates with cluster membership. Cluster 2 tumors, comprising MEN2, NF1, and other sporadic tumors, are shown in lanes 2–4 (tumors 105, 91, and 196, respectively, from Figure 1). Cluster 1 contains tumors with VHL and SDHB mutations and a subset of sporadic samples (lanes 5–7 are tumors 16, 85, 101, and 152, respectively, from Figure 1). Lane 1 is normal adrenal medulla. β-actin was used as a loading control.
(C) Immunostaining of SDHB protein in pheochromocytomas or paragangliomas with various genetic backgrounds. A MEN2-related pheochromocytoma is shown on the top row, followed by tumors with mutations in NF1, SDHB, SDHD, and VHL genes. Corresponding hematoxylin/eosin staining is shown on the left.
In contrast, Cluster 2 tumors exhibited a distinct set of biological programs, including genes that mediate translation initiation, protein synthesis, and kinase signaling (Table 2; Dataset S4). The two prototype genes of Cluster 2 (RET and NF1) are linked by their common ability to activate the RAS/RAF/MAP kinase signaling cascade [16,17]. Activated RAS signaling has been shown by expression profiling to be associated with increased translation events [18]. Thus, the anabolic functions of activated RAS may constitute the biochemical mechanism that underlies the assignment of MEN2 and NF1 tumors to Cluster 2. Increased expression of genes defining a neural/neuroendocrine profile and adrenergic metabolism were also prominent features of this cluster (Table 2; Dataset S4).
HIF1α Contributes to SDHB Regulation
Because of the critical role of VHL in controlling availability of HIF in normoxic conditions, we next investigated whether downregulation of SDHB was HIF-dependent. In two cell line models, HEK293 and mouse pheochromocytoma cell line (MPC) 9/3L, exposure to the hypoxia-mimetic agent cobalt chloride reduced SDHB protein expression (Figure 4A). Further, transient expression of a mutant, nondegradable form of HIF1α, HIF1αP402A/P564A, was able to downregulate SDHB (Figure 4B). In contrast, suppression of HIF1α in neural crest–derived A2058 melanoma cell lines stably expressing HIF1α short hairpin RNA (shRNA) prevented the reduction of SDHB after exposure to cobalt chloride (Figure 4C), in contrast to cells expressing a control shRNA sequence, supporting a central role of HIF1α in regulating SDHB levels. This finding implicates HIF1α as a mediator of the clustering between VHL- and SDH-mutated primary pheochromocytomas identified in our expression profiling studies. Similar to what we found in primary pheochromocytomas, no decrease in SDHB mRNA was detected when these cell lines were treated with hypoxia-mimetic drugs (data not shown), suggesting that a posttranscriptional phenomenon is related to these findings.
Figure 4 HIF1α Attenuates SDHB Levels
(A) HIF1α expression was induced by treatment of mouse pheochromocytoma MPC 9/3L cells with 150 μM cobalt chloride for the indicated times. SDHB expression decreased in treated cells. Glut1 indicates increased activity of HIF1α, and β-actin was used as a loading control.
(B) Transient expression in HEK293 cells of a HIF1α double mutant PA (P402A/P564A) that is resistant to VHL-mediated degradation reduced expression of SDHB.
(C) A2058 cell lines stably expressing HIF1α shRNA do not show change in SDHB after cobalt chloride exposure, while SDHB is downregulated in control GFP shRNA cells treated with cobalt chloride.
(D) Proposed model of HIF1α and SDHB interregulation. HIF1α downregulates SDHB, which leads to complex II dysfunction. High succinate levels resulting from loss of complex II, in turn, inhibit prolyl hydroxylase (PHD) activity [19]. Non-hydroxylated HIF1α is resistant to VHL-mediated targeting for degradation and can therefore activate downstream genes, such as angiogenic factors. “E3 complex” indicates the E3 ubiquitin ligase complex for which VHL is the substrate recognition factor.
Discussion
Transcription profiling of a large series of primary pheochromocytomas reveals that tumors with VHL and SDH mutations are closely linked. The hypoxia-angiogenesis signature identified by our analysis of primary tumors with SDHB or SDHD mutations confirms and extends recent observations on the role of SDH proteins in cultured cell lines. Selak et al. showed that disruption of the mitochondrial complex II results in increased HIF1α activity and that this upregulation is channeled through inhibition of prolyl hydroxylase function [19]. This hydroxylation step is essential for VHL-dependent HIF1α degradation [20,21]. Our data show that mitochondrial complex II mutations lead to upregulation of HIF1α targets in human tumor tissue and indicate an additional level of interplay between the SDHB and HIF1α proteins, i.e., a reciprocal effect of HIF1α in modulating components of the mitochondrial complex II. Our findings favor the existence of an autoregulatory loop whereby HIF1α contributes to attenuation of SDHB levels, resulting in complex II inhibition (Figure 4D). High levels of succinate resulting from loss of complex II function can in turn block HIF1α degradation through inhibition of prolyl hydroxylases. However, while succinate accumulation, but not oxidative stress, was considered the oncogenic trigger by Selak et al. [19], increased levels of reactive oxygen species have been reported in animal models of SDHC dysfunction [22–24]. The latter results are consistent with the oxidoreductase defect of our primary tumor samples. The precise mechanisms for the interaction between HIF1α and SDHB still remain to be identified, but our data suggest that a posttranscriptional response is likely to be involved. Of note, and in keeping with our current data, HIF2α /EPAS1-null mice were reported to have increased SDH activity in muscle [25].
One provocative possibility suggested by these findings is that the tumorigenic effects of VHL mutations in chromaffin tissue might involve dysfunction of mitochondrial complex II. Hence, we propose that in VHL-derived tumors two complementary mechanisms play a role in stabilizing HIF1α: the loss of VHL-dependent targeting of HIF1α for proteasome-mediated degradation, and a second mechanism that is dependent on low levels of HIF1α hydroxylation resulting from complex II dysfunction. The effects of HIF1α in our model were less marked than those observed with hypoxia-mimetic agents, which inhibit prolyl hydroxylases. This suggests that additional factors, besides HIF1α, might be involved in SDHB suppression. As such, it will be relevant to determine how SDHB and mitochondrial complex II are regulated in VHL type 2C variants that have been proposed to impart distinct, HIF-independent signaling outcomes [12,13].
A relationship between SDH function and oxygen regulation has been suspected based on previous identification of increased expression of angiogenic factors in cases of SDH-mutant pheochromocytomas [9]. Also, clinical similarities besides pheochromocytoma have been noted in families with germline mutations of VHL and SDHB [26]. This is in agreement with our transcription results and biochemical data indicating that HIF1α is involved in this association. The bipartite transcription clustering of pheochromocytomas has thus provided an explanation for the link between two genetic subtypes of pheochromocytomas. We also showed that this distribution has high predictive value, as determined by the identification of previously undetected mutations in tumor samples segregating with the appropriate cluster. The successful distinction of tumors from Cluster 1 and Cluster 2 by SDHB immunostaining in our pilot series suggests that this may be developed into a new screening method to classify pheochromocytomas in one of two major categories that reflect the underlying genetic defect. Of interest, in a recent study, immunohistochemistry of head and neck paragangliomas with SDHB and SDHD mutations revealed similar suppression of SDHB, which was accompanied by morphologically abnormal mitochondria [27]. This is in line with our results of catecholamine-secreting tumors and suggests that SDHB downregulation is a general marker of complex II dysfunction. This study also describes a number of sporadic head and neck paragangliomas with low SDHB staining; these tumors might correspond with Cluster 1 pheochromocytomas for which no detectable mutation was identified and that also appear to arise from disruption of related pathways. It remains to be tested whether the predominant hypoxic-angiogenic profile of pheochromocytomas with VHL and SDH mutations will render these tumors targets for antiangiogenic therapies. This will be particularly relevant for SDHB-mutant pheochromocytomas which have been suggested to be more prone to malignancy [10].
Materials and Methods
Tumor specimens.
Tumor samples were obtained from patients with catecholamine-secreting pheochromocytomas and thoracic or abdominal paragangliomas according to institutionally approved protocols. Fragments were obtained from the core of the tumor and contained more than 70% tumor cells. Samples with a clear adjacent cortical component were macrodissected. Specimens were snap-frozen at time of surgical resection and stored at −70 °C or in liquid nitrogen until processed.
Diagnosis of pheochromocytoma and/or paraganglioma was confirmed by histology in every case. Heredity status was defined by the presence of clinical features associated with well-known familial syndromes (medullary thyroid carcinoma, hyperparathyroidism, hemangioblastomas of retina and/or central nervous system, renal cell carcinoma, neurofibromas, café-au-lait spots, and head and neck paragangliomas) or diagnosis of pheochromocytoma and/or paraganglioma in at least one first-degree relative. In all, 76 catecholamine-secreting pheochromocytomas or paragangliomas representing well-characterized hereditary variants cited above, familial tumors of undetermined genetic cause, and sporadic tumors were included in this study (see Dataset S1). Of these, 15 samples belonged to seven different families presenting with bilateral tumors and/or familial history of recurrent pheochromocytoma and/or paraganglioma. No additional clinical features associated with hereditary pheochromocytoma were identified in these individuals. Of the seventy-six tumors, 13 were located outside the adrenal gland (one was mediastinal, one retrocardiac, and the remaining 11 were in periaortic or perirenal locations). No head or neck paragangliomas or tumors with SDHC mutations were included in this series.
RNA isolation and microarray preparation.
Total RNA was extracted from each frozen tumor specimen, and biotinylated cRNAs were generated using Trizol (Invitrogen, Carlsbad, California, United States) according to the manufacturer's instructions. Eighty-four tumor samples (including six replicates) were hybridized overnight to U133A oligonucleotide microarrays (Affymetrix, Santa Clara, California, United States), which included approximately 22,000 probe sets. In four duplicate cases two aliquots of RNA were separately used for target preparation and subsequent analysis, and in two cases, the same source cRNA was used in two independent hybridizations. Arrays were subsequently developed with phycoerythrin-conjugated streptavidin (SAPE) and biotinylated anti-streptavidin, and scanned to obtain quantitative gene expression levels. The raw gene expression values were scaled to account for differences in global chip intensity using MAS software (Affymetrix). Four scans (two duplicates and two unique tumors) were excluded because of poor quality. In total, 76 unique tumor samples were used for the analysis.
Normalization and model-based expression analysis.
All arrays were normalized using dChip software v1.3 [28]. Model-based expression index was obtained by PM/MM difference algorithm in dChip. Gene filtering and hierarchical clustering analysis were also performed in dChip [29]. To deal with variable degrees of quality in sample hybridization and resulting control parameters, such as absolute (%P) call and 5′/3′ ratios of housekeeping genes (GAPDH and actin), we subclassified the samples into three categories: superior, good, and satisfactory quality. For each gene, samples belonging to each category were separately computed for coefficient of variation (standard deviation/mean) for gene filtering and standardized (to achieve mean = 0 and SD = 1) for gene and sample clustering. Duplicated samples were combined prior to standardization.
Unsupervised clustering analysis.
Samples were clustered using an unsupervised hierarchical clustering method to delineate groups with biological distinction. Filtering parameters were set to define the genes that showed the highest variation among the sample set. A selection was made of 508 genes using the following parameters: 0.6 < average SD/mean across categories < 10, and having a mean intensity value of >100 units in at least 20% of samples.
The reliability of the clusters was verified by a resampling method based on the standard errors for expression values [30]. This subsampling procedure was repeated 100 times to refine the cluster parameters. The most stable clusters resulting from multiple iterations were defined and used for subsequent analysis.
Supervised analysis.
Once the main clusters were defined by the method above, tumors that represented individual genetic classes were used for supervised analysis: MEN2 versus VHL pheochromocytomas, or MEN2 versus combined SDHB and SDHD (SDH) samples. The strength of gene expression differences between each pair of classes (defined as the “training set”) was assessed using two supervised machine learning algorithms, k nearest neighbor and weighted-voting, as previously described [31,32]. Duplicate samples were also included in this analysis. Data were preprocessed by applying thresholds of ten minimum and 16,000 maximum genes, which then were filtered by requiring a 5-fold minimum variation and 50 minimum absolute difference. Models were evaluated using leave-one-out cross-validation. The differential genes were reselected after each sample withdrawal. Probes (features) to be used in the models were selected by ranking the genes according to the signal-to-noise metric [31]. After the number of probes was selected to find the minimum error rate, a model was trained using data for the pair of tumor classes and tested on data for the remaining samples (defined as the “test set”). Prediction performance on the test samples was used to confirm the similarity of sample types.
Pathway analysis.
To gather insights into the function of genes associated with the clusters defined by the comparisons described above, we used a statistical method designed to test for the enrichment of groups of genes in data generated from expression studies, GSEA [33]. GSEA considers predefined gene sets representing pathways of interest and determines whether the members of these sets are overrepresented in a list of genes that has been ordered by their correlation with a specific phenotype or class distinction.
The output of GSEA is a normalized enrichment score (NES) that represents a measure of the degree of enrichment of the gene set at the top (highly correlated) or bottom (anti-correlated) of the ordered gene list. The NES is used to produce a p-value that measures the significance of that score. This is obtained by permutation testing, which involves shuffling the class template associated with the data to determine how often an observed NES occurs by chance. p-Values are adjusted to account for multiple hypothesis testing [34]. The gene sets used for analysis in the current study were obtained from Gene Ontology (www.geneontology.org), GenMAPP (www.genmapp.org), and Biocarta (www.biocarta.com); manually curated proteome datasets were also used [34–36].
DNA sequencing.
Direct sequencing of familial samples and 20 sporadic tumors was performed from tumor and germline DNA, whenever available, using PCR products that encompass exons and intron–exon boundaries of RET (exons 10, 11, and 13–16), VHL (exons 1–3), SDHD (exons 1–4), and SDHB (exons 1–8), as previously described [7,37,38]. SDHC was also sequenced in familial tumors with no detectable mutation [7].
Western blots and transfections.
Whole cell lysates from tumors and normal adrenal medullas were prepared as previously described [39], and 50 μg was run on 12% SDS gels, transferred to PVDF membranes and hybridized with antibodies against SDHB and SDHA (Molecular Probes, Eugene, Oregon, United States), RET (Immuno-Biological Laboratories, Gunma, Japan), or β-actin (Sigma, St. Louis, Missouri, United States), according to the manufacturers' instructions. Filters were developed with a chemiluminescence assay (Pierce Biotechnology, Rockford, Illinois, United States) and images captured using the VersaDoc Imaging system (Bio-Rad, Hercules, California, United States).
The MPC 9/3L cell lines derived from NF1± mice were cultured as described [40]. HEK293 cells were cultured in DMEM and 10% fetal bovine serum supplemented with 100 U/ml penicillin and 100 μg/ml streptomycin. HIF1α expression was induced in MPC 9/3L or HEK293 cells by treatment with 150 μM cobalt chloride, which blocks prolyl hydroxylation of HIF1α and its binding to VHL [41], for the indicated times (see Figure 4A). A HIF1α double mutant (P402A/P564A) that is resistant to VHL-mediated proteasome degradation [42] was generated by site-directed mutagenesis (Quick Change, Stratagene, La Jolla, California, United States) and cloned into p3X-FLAG vector (Sigma). HEK293 cells were transfected with the HIF1α P402A/P564A double mutant or an empty vector using Lipofectamine 2000, as recommended by the manufacturer (Invitrogen). Transfected cells were harvested at 48 h and assayed by Western blotting, as above. Membranes were probed with the following antibodies: SDHB, as described above, HIF1α (BD Biosciences, San Jose, California, United States), Glut1, used as a surrogate for HIF1α activity (Alpha Diagnostic, San Antonio, Texas, United States), and FLAG (Sigma). β-actin was used as a loading control, as above. The lentiviral shRNA expression vector FSIPPW was used, as previously described [43]. The shRNA expression construct targeting HIF1α (FSIPPW-HIF) is directed against the sequence 5′-AACTAACTGGACACAGTGTGTTT-3′, which is conserved in mouse, rat, and human HIF1α. FSIPPW-eGFP, packaging of lentiviruses, and infection of cell lines were performed as previously described [43]. A2058 melanoma cells stably expressing HIF1α shRNA (FSIPPW-HIF) or control pEGFP shRNA (FSIPPW-EGFP) were cultured in DMEM, 10% FBS, and 2 μg/ml puromycin. Infected cell lines were selected with 2 μg/ml puromycin (Sigma). Cells were exposed to 150 μM cobalt chloride for 24 or 48 h, and lysates obtained as above.
Quantitative Real-Time PCR.
Quantitative real-time PCR was performed in cDNA from 20 tumors (ten from Cluster 1 and ten from Cluster 2) from the cohort above using the iCycler iQ Real-Time PCR Detection System (Bio-Rad). SYBR green fluorescence (Bio-Rad) was used for quantification, according to the manufacturer's instructions. Primer sequences and PCR conditions are available upon request.
Immunohistochemistry.
Immunohistochemical analysis was performed on 4-μm-thin sections of formalin-fixed tissue obtained from the archives of Brigham and Women's Hospital and from consultation. Clinical data and additional follow-up information were provided by the referring clinician and/or pathologists. Slides were processed according to standard protocol using primary antibodies SDHB (1:1,000 dilution, Molecular Probes) and the Envision Plus Detection System (Dako, Carpinteria, California, United States) for antigen–antibody detection. Heart muscle and normal adrenal tissue were included as positive controls. Negative controls (no primary antibody) were also maintained throughout. Immunoreactivity was graded semi-quantitatively using the following scale: 0, no staining; 1+, <5% of tumor cells reactive; 2+, 5%–25% of tumor cells reactive; 3+, 25%–50% of tumor cells reactive; 4+, >50% of tumor cells reactive (weak intensity); 5+, >50% of tumor cells reactive (moderate intensity); and 6+, >50% of tumor cells reactive (strong intensity).
Supporting Information
Dataset S1 Summary of Clinical Data and Classification of Pheochromocytomas by Genetic Groups
(13 KB PDF)
Click here for additional data file.
Dataset S2 Filtered Gene List from the Unsupervised Analysis
(137 KB PDF)
Click here for additional data file.
Dataset S3 Results of Real-Time PCR and Western Blot Analyses of Differentially Expressed Genes
(27 KB PDF)
Click here for additional data file.
Dataset S4 Cross-Validation and Gene Set Analysis Results (Gene Lists and Heat Map Images) of Two-Class Comparisons by Supervised Learning Methods
(345 KB PDF)
Click here for additional data file.
Dataset S5 Expression of Mitochondrial Complex II Subunits in Pheochromocytomas by Cluster Distribution
(12 KB PDF)
Click here for additional data file.
Accession Numbers
The data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/) and are accessible through GEO Series accession number GSE2841.
We thank Ricardo Aguiar for advice and critical review of the manuscript; Todd Golub for input and suggestions; Scott Pomeroy, John Alberta, and William Kaelin, Jr., for comments; Margaret Shipp and Graeme Eisenhofer for access to prepublication data; Christian Colin for technical assistance; and Gail Adler for access to tumor samples. We also thank the Massachusetts General Hospital Tumor Bank, Brain and Tissue Bank for Developmental Disorders at the University of Maryland, and the Microarray Core Facility at the Dana-Farber Cancer Institute. PLMD is recipient of a Claudia Adams Barr Investigator Award and a Pan-Mass Agencourt Sequencing Grant. This work was supported in part by the Charles A. Dana Foundation Project in Neuro-Oncology (CDS), NIH-PO1 HD24926–12 (CDS), RO1 CA48017 (AST), and RO1 NS37685 (AST).
Competing interests. The authors have declared that no competing interests exist.
Author contributions. PLMD and CDS conceived and designed the experiments. PLMD, KNR, MEW, SS, VN, and CL performed the experiments. PLMD, KNR, SS, VN, and CL analyzed the data. PLMD, CYH, MB, ALK, GS, JFP, AST, RH, SH, FM, RD, JAS, ITO, DEB, DJM, BGR, KS, JG, SMA, MK, AG, PP, and SPAT contributed reagents/materials/analysis tools. PLMD and CDS wrote the paper.
Abbreviations
GSEAgene set enrichment analysis
HIFhypoxia-inducible factor
HIF1αhypoxia-inducible factor 1 subunit α
MEN2multiple endocrine neoplasia type 2
MPC 9/3Lmouse pheochromocytoma cell line 9/3L
NESnormalized enrichment score
NF1neurofibromatosis type 1
PGL[number]paraganglioma syndrome type [number]
SDHsuccinate dehydrogenase
shRNAshort hairpin RNA
VHLvon Hippel-Lindau disease
==== Refs
References
Manger WM Gifford RW 2002 Pheochromocytoma J Clin Hypertens (Greenwich) 4 62 72 11821644
Bryant J Farmer J Kessler LJ Townsend RR Nathanson KL 2003 Pheochromocytoma: The expanding genetic differential diagnosis J Natl Cancer Inst 95 1196 1204 12928344
Kaelin WG Jr 2003 The von Hippel-Lindau gene, kidney cancer, and oxygen sensing J Am Soc Nephrol 14 2703 2711 14569079
Kondo K Kim WY Lechpammer M Kaelin WG Jr 2003 Inhibition of HIF2α is sufficient to suppress pVHL-defective tumor growth PLoS Biol 1 e83. DOI: 10.1371/journal.pbio.0020289 . 14691554
Kondo K Klco J Nakamura E Lechpammer M Kaelin WG Jr 2002 Inhibition of HIF is necessary for tumor suppression by the von Hippel-Lindau protein Cancer Cell 1 237 246 12086860
Baysal BE Ferrell RE Willett-Brozick JE Lawrence EC Myssiorek D 2000 Mutations in SDHD, a mitochondrial complex II gene, in hereditary paraganglioma Science 287 848 851 10657297
Astuti D Latif F Dallol A Dahia PL Douglas F 2001 Gene mutations in the succinate dehydrogenase subunit SDHB cause susceptibility to familial pheochromocytoma and to familial paraganglioma Am J Hum Genet 69 49 54 11404820
Ackrell BA 2000 Progress in understanding structure-function relationships in respiratory chain complex II FEBS Lett 466 1 5 10648801
Gimenez-Roqueplo AP Favier J Rustin P Mourad JJ Plouin PF 2001 The R22X mutation of the SDHD gene in hereditary paraganglioma abolishes the enzymatic activity of complex II in the mitochondrial respiratory chain and activates the hypoxia pathway Am J Hum Genet 69 1186 1197 11605159
Gimenez-Roqueplo AP Favier J Rustin P Rieubland C Crespin M 2003 Mutations in the SDHB gene are associated with extra-adrenal and/or malignant phaeochromocytomas Cancer Res 63 5615 5621 14500403
Kaelin WG Jr 2002 Molecular basis of the VHL hereditary cancer syndrome Nat Rev Cancer 2 673 682 12209156
Hoffman MA Ohh M Yang H Klco JM Ivan M 2001 von Hippel-Lindau protein mutants linked to type 2C VHL disease preserve the ability to downregulate HIF Hum Mol Genet 10 1019 1027 11331612
Clifford SC Cockman ME Smallwood AC Mole DR Woodward ER 2001 Contrasting effects on HIF-1α regulation by disease-causing pVHL mutations correlate with patterns of tumourigenesis in von Hippel-Lindau disease Hum Mol Genet 10 1029 1038 11331613
Gimenez-Roqueplo AP Favier J Rustin P Rieubland C Kerlan V 2002 Functional consequences of a SDHB gene mutation in an apparently sporadic pheochromocytoma J Clin Endocrinol Metab 87 4771 4774 12364472
Yankovskaya V Horsefield R Tornroth S Luna-Chavez C Miyoshi H 2003 Architecture of succinate dehydrogenase and reactive oxygen species generation Science 299 700 704 12560550
Ballester R Marchuk D Boguski M Saulino A Letcher R 1990 The NF1 locus encodes a protein functionally related to mammalian GAP and yeast IRA proteins Cell 63 851 859 2121371
Califano D Rizzo C D'Alessio A Colucci-D'Amato GL Cali G 2000 Signaling through Ras is essential for ret oncogene-induced cell differentiation in PC12 cells J Biol Chem 275 19297 19305 10748077
Rajasekhar VK Viale A Socci ND Wiedmann M Hu X 2003 Oncogenic Ras and Akt signaling contribute to glioblastoma formation by differential recruitment of existing mRNAs to polysomes Mol Cell 12 889 901 14580340
Selak MA Armour SM Mackenzie ED Boulahbel H Watson DG 2005 Succinate links TCA cycle dysfunction to oncogenesis by inhibiting HIFα prolyl hydroxylase Cancer Cell 7 77 85 15652751
Ivan M Kondo K Yang H Kim W Valiando J 2001 HIFα targeted for VHL-mediated destruction by proline hydroxylation: Implications for O2 sensing Science 292 464 468 11292862
Jaakkola P Mole DR Tian YM Wilson MI Gielbert J 2001 Targeting of HIFα to the von Hippel-Lindau ubiquitylation complex by O2-regulated prolyl hydroxylation Science 292 468 472 11292861
Ishii T Yasuda K Akatsuka A Hino O Hartman PS 2005 A mutation in the SDHC gene of complex II increases oxidative stress, resulting in apoptosis and tumorigenesis Cancer Res 65 203 209 15665296
Guo J Lemire BD 2003 The ubiquinone-binding site of the Saccharomyces cerevisiae succinate-ubiquinone oxidoreductase is a source of superoxide J Biol Chem 278 47629 47635 13129931
Ishii N Fujii M Hartman PS Tsuda M Yasuda K 1998 A mutation in succinate dehydrogenase cytochrome b causes oxidative stress and ageing in nematodes Nature 394 694 697 9716135
Scortegagna M Ding K Oktay Y Gaur A Thurmond F 2003 Multiple organ pathology, metabolic abnormalities and impaired homeostasis of reactive oxygen species in Epas1−/− mice Nat Genet 35 331 340 14608355
Vanharanta S Buchta M McWhinney SR Virta SK Peczkowska M 2004 Early-onset renal cell carcinoma as a novel extraparaganglial component of SDHB-associated heritable paraganglioma Am J Hum Genet 74 153 159 14685938
Douwes Dekker PB Hogendoorn PC Kuipers-Dijkshoorn N Prins FA van Duinen SG 2003 SDHD mutations in head and neck paragangliomas result in destabilization of complex II in the mitochondrial respiratory chain with loss of enzymatic activity and abnormal mitochondrial morphology J Pathol 201 480 486 14595761
Schadt EE Li C Ellis B Wong WH 2001 Feature extraction and normalization algorithms for high-density oligonucleotide gene expression array data J Cell Biochem Suppl S37 120 125
Eisen MB Spellman PT Brown PO Botstein D 1998 Cluster analysis and display of genome-wide expression patterns Proc Natl Acad Sci U S A 95 14863 14868 9843981
Li C Wong WH 2001 Model-based analysis of oligonucleotide arrays: Expression index computation and outlier detection Proc Natl Acad Sci U S A 98 31 36 11134512
Golub TR Slonim DK Tamayo P Huard C Gaasenbeek M 1999 Molecular classification of cancer: Class discovery and class prediction by gene expression monitoring Science 286 531 537 10521349
Armstrong SA Staunton JE Silverman LB Pieters R den Boer ML 2002 MLL translocations specify a distinct gene expression profile that distinguishes a unique leukemia Nat Genet 30 41 47 11731795
Mootha VK Lindgren CM Eriksson KF Subramanian A Sihag S 2003 PGC-1α-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes Nat Genet 34 267 273 12808457
Sweet-Cordero A Mukherjee S Subramanian A You H Roix JJ 2005 An oncogenic KRAS2 expression signature identified by cross-species gene-expression analysis Nat Genet 37 48 55 15608639
Mootha VK Bunkenborg J Olsen JV Hjerrild M Wisniewski JR 2003 Integrated analysis of protein composition, tissue diversity, and gene regulation in mouse mitochondria Cell 115 629 640 14651853
Semenza G 2002 Signal transduction to hypoxia-inducible factor 1 Biochem Pharmacol 64 993 998 12213597
Eng C Crossey PA Mulligan LM Healey CS Houghton C 1995 Mutations in the RET proto-oncogene and the von Hippel-Lindau disease tumour suppressor gene in sporadic and syndromic phaeochromocytomas J Med Genet 32 934 937 8825918
Aguiar RC Cox G Pomeroy SL Dahia PL 2001 Analysis of the SDHD gene, the susceptibility gene for familial paraganglioma syndrome (PGL1), in pheochromocytomas J Clin Endocrinol Metab 86 2890 2894 11397905
Dahia PLM Aguiar RCT Alberta J Kum JB Caron S 1999 PTEN is inversely correlated with the cell survival factor Akt/PKB and is inactivated via multiple mechanisms in haematological malignancies Hum Mol Genet 8 185 193 9931326
Powers JF Evinger MJ Tsokas P Bedri S Alroy J 2000 Pheochromocytoma cell lines from heterozygous neurofibromatosis knockout mice Cell Tissue Res 302 309 320 11151443
Yuan Y Hilliard G Ferguson T Millhorn DE 2003 Cobalt inhibits the interaction between hypoxia-inducible factorα and von Hippel-Lindau protein by direct binding to hypoxia-inducible factorα J Biol Chem 278 15911 15916 12606543
Masson N Willam C Maxwell PH Pugh CW Ratcliffe PJ 2001 Independent function of two destruction domains in hypoxia-inducible factorα chains activated by prolyl hydroxylation EMBO J 20 5197 5206 11566883
Kanellopoulou C Muljo SA Kung AL Ganesan S Drapkin R 2005 Dicer-deficient mouse embryonic stem cells are defective in differentiation and centromeric silencing Genes Dev 19 489 501 15713842
|
16103922
|
PMC1183527
|
CC BY
|
2021-01-05 08:00:24
|
no
|
PLoS Genet. 2005 Jul 25; 1(1):e8
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010008
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 10.1371/journal.pgen.001000905-PLGE-RA-0024R3plge-01-01-09Research ArticleBioinformatics - Computational BiologyCell BiologyEvolutionGenetics/Comparative GenomicsGenetics/Chromosome BiologyDrosophilaHeterochromatinHp1EvolutionOogenesisPositive Selection Drives the Evolution of rhino, a Member of the Heterochromatin Protein 1 Family in Drosophila
Positive Selection of rhinoVermaak Danielle 1Henikoff Steven 12Malik Harmit S 1*1 Fred Hutchinson Cancer Research Center, Seattle, Washington, United States of America
2 Howard Hughes Medical Institute, Basic Sciences Division, Fred Hutchinson Cancer Research Center, Seattle, Washington, United States of America
Clark Andy G EditorCornell University, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 25 7 2005 1 1 e97 2 2005 13 5 2005 Copyright: © 2005 Vermaak et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Heterochromatin comprises a significant component of many eukaryotic genomes. In comparison to euchromatin, heterochromatin is gene poor, transposon rich, and late replicating. It serves many important biological roles, from gene silencing to accurate chromosome segregation, yet little is known about the evolutionary constraints that shape heterochromatin. A complementary approach to the traditional one of directly studying heterochromatic DNA sequence is to study the evolution of proteins that bind and define heterochromatin. One of the best markers for heterochromatin is the heterochromatin protein 1 (HP1), which is an essential, nonhistone chromosomal protein. Here we investigate the molecular evolution of five HP1 paralogs present in Drosophila melanogaster. Three of these paralogs have ubiquitous expression patterns in adult Drosophila tissues, whereas HP1D/rhino and HP1E are expressed predominantly in ovaries and testes respectively. The HP1 paralogs also have distinct localization preferences in Drosophila cells. Thus, Rhino localizes to the heterochromatic compartment in Drosophila tissue culture cells, but in a pattern distinct from HP1A and lysine-9 dimethylated H3. Using molecular evolution and population genetic analyses, we find that rhino has been subject to positive selection in all three domains of the protein: the N-terminal chromo domain, the C-terminal chromo-shadow domain, and the hinge region that connects these two modules. Maximum likelihood analysis of rhino sequences from 20 species of Drosophila reveals that a small number of residues of the chromo and shadow domains have been subject to repeated positive selection. The rapid and positive selection of rhino is highly unusual for a gene encoding a chromosomal protein and suggests that rhino is involved in a genetic conflict that affects the germline, belying the notion that heterochromatin is simply a passive recipient of “junk DNA” in eukaryotic genomes.
Synopsis
Eukaryotic genomes are organized into good and bad neighborhoods. In fruit fly genomes, most genes are found in euchromatin—good neighborhoods that tend to be amenable to gene expression and deficient in selfish mobile elements. Conversely, heterochromatic regions are deficient in genes but chock full of mobile genetic elements, both dead and alive. Cells expend considerable effort to maintain this organization, to prevent bad neighborhoods from exerting their negative influence on the rest of the genome. At the forefront of this organization are the HP1 proteins, which are involved in the compaction and silencing of heterochromatic sequences. First discovered in Drosophila, HP1 proteins have been subsequently found in virtually all fungi, plants, and animals.
Most HP1 proteins evolve under stringent evolutionary pressures, suggesting that they lack any discriminatory power in their action. However, a recent paper by Vermaak finds that one of the five HP1 encoding genes in Drosophila genomes, rhino, bucks the trend and evolves rapidly. rhino is predominantly expressed in ovaries, which is where many mobile elements are also active. Their results suggest that rhino has been constantly evolving to police a particularly dynamic, novel compartment in heterochromatin with exquisite specificity. Thus, instead of a genomic wasteyard that genes shun and where transposons go to die, heterochromatin now appears to have been shaped by a constant struggle for evolutionary dominance.
Citation:Vermaak D, Henikoff S, Malik HS (2005) Positive selection drives the evolution of rhino, a member of the heterochromatin protein 1 family in Drosophila. PLoS Genet 1(1): e9.
==== Body
Introduction
Repetitive DNA sequences can constitute large parts of many genomes (approximately 30% in human and fly genomes) and are involved in fundamental cellular processes [1–3]. For example, centromeres in higher eukaryotes consist of large, repetitive regions required for accurate chromosome segregation during each cell division [4]. Heterochromatin flanks the centromere and is also essential for segregation [5–7]. It is composed largely of repetitive DNA and transposable elements and their relics, but can contain genes important for fertility and viability [8,9]. Transcriptionally silent heterochromatin can influence the expression of not only mobile elements embedded in heterochromatin, but also euchromatic genes [6,10–12]. Given the importance of heterochromatin, it is not surprising that perturbation of heterochromatic proteins is associated with cancer and other diseases [13,14].
The study of repetitive heterochromatic DNA lags far behind that of euchromatic regions because heterochromatin is hard to sequence and manipulate experimentally. Even when DNA sequence is available, the underlying evolutionary forces that shape patterns of rapidly changing repetitive sequences and chromosomal architecture are hard to discern. A complementary approach is to study the evolution of protein components that associate with repetitive DNA instead of studying the DNA directly. These protein components have been well studied, especially in Drosophila genomes [15–18]. Using a similar strategy, the discovery of positive selection acting on the proteins that bind centromeric DNA has led to the centromere-drive hypothesis that may account for the sequence complexity of centromeres [19–21].
Here, we examine the evolutionary pressures that shape proteins that bind heterochromatic DNA. Heterochromatin protein 1 (HP1) is a ubiquitous component of heterochromatin that is the best available surrogate to study heterochromatin complexity. HP1 was first identified in flies [18,22] and is present in most eukaryotes where it is required for maintenance of most aspects of the heterochromatic state [6,10,11,23]. HP1 consists of a N-terminal chromo domain, a hinge region, and a C-terminal chromo shadow (or simply “shadow”) domain that structurally resembles the chromo domain and mediates homodimerization [16,22,24–26]. The chromo domain binds to histone H3 tails methylated at lysine 9 (H3K9me), a covalent modification associated with heterochromatin maintenance and transcriptional silencing [10,11,27,28] and can directly influence the targeting of HP1 in vivo [29].
Multiple HP1-like genes, which may have different functions, can be found in the same genome. In vertebrates, for example, there are at least three HP1-like genes (HP1α, HP1β, and HP1γ) that each encode proteins with distinct localization patterns, despite being about 65% identical [22,30–33]. Drosophila melanogaster contains five genes with HP1-like domain organization. We undertook a molecular evolutionary study of these HP1 paralogs in Drosophila, aiming to use them as a surrogate for studying heterochromatic DNA evolution. HP1A (or Su[Var]205) was the first of these to be identified. This HP1A gene encodes the prototypic HP1 protein required for heterochromatin maintenance [18,34]. The functions of the other four HP1 proteins are unknown. However, HP1B and HP1C differ from HP1A in their chromatin localization [35], suggesting that their function is not redundant with HP1A. The fourth HP1-like protein, HP1D/Rhino (hereafter referred to as “Rhino”), was discovered in a screen for female sterile mutants [36] whereas we identified the fifth, HP1E, using bioinformatic criteria in this study.
rhino mutants display a variety of late-stage eggshell defects, among them the fused dorsal appendages for which the gene was named [36]. Careful characterization of mutant egg chambers revealed several defects [36]. First, nurse cells failed to undergo a higher-order chromatin structure reorganization from a “five-blob” state to a dispersed state at stage 5. Second, although transcript levels of several patterning genes were unaffected, transcripts of key patterning genes such as gurken and oskar were mislocalized. Furthermore, Gurken protein synthesis was delayed in early egg chambers and germaria, and Gurken protein showed aberrant accumulation in later egg chambers [36]. Unlike other HP1 proteins, Rhino is expressed predominantly during oogenesis [36]. Its unusual expression pattern suggested that the evolutionary constraints on rhino might more accurately reflect pressures on heterochromatin in the female germline, relatively free from constraints imposed during somatic expression.
In this report, we show that tagged Rhino protein localizes to distinct foci within the heterochromatic domain of tissue culture cells. Remarkably, we find that all three domains of Rhino show strong evidence of recurrent positive selection. Such positive selection implies that rhino is involved in a heritable and recurrent genetic conflict, raising the intriguing possibility that heterochromatin itself might represent a paleontological record of this genetic conflict.
Results
HP1 Paralogs in Drosophila Genomes
D. melanogaster contains five HP1-like genes, defined as such because they all encode an N-terminal chromo domain and a C-terminal shadow domain (Figure 1A). Four of these paralogs have been identified in previous analyses [36,37], whereas HP1E is newly identified in this report. These paralogs show differences in their conservation across Drosophila species. HP1A, HP1B, and HP1C are highly conserved, even between D. melanogaster and the more distantly related D. pseudoobscura (Figure 1A). In contrast, rhino differs significantly in size and amino acid sequence between D. melanogaster and D. simulans. In addition, the HP1E gene appears to have degenerated in the D. pseudoobscura genome, whereas D. pseudoobscura possesses HP1F, a novel HP1 that the D. melanogaster genome lacks altogether.
Figure 1 HP1 Paralogs in Drosophila
(A) Proteins encoded by D. melanogaster
HP1s and selected orthologs (obtained by PCR from syntenic locations) are drawn to scale (indicated at bottom) with a dark rectangle resembling the N-terminal chromo domain and a lighter rectangle the C-terminal chromo shadow domain. The HP1E open reading frame is no longer preserved in D. pseudoobscura, and D. melanogaster does not contain HP1F. The hinge regions and N- and C-terminal extensions cannot be aligned between different HP1 types, for example HP1A versus HP1B. HP1D/Rhino contains a very long hinge region that is poorly conserved between species.
(B) A neighbor-joining phylogenetic tree based on an alignment of selected HP1 chromo and (C) shadow domains. The monophyletic vertebrate HP1 paralogs are shown for comparison. rhino evolution is clearly distinct from vertebrate or other Drosophila
HP1s. HP1 orthologs between D. melanogaster, D. erecta, and D pseudoobscura are shown connected by bold branches (HP1E is not conserved in D. pseudoobscura). The divergence times for D. melanogaster–D. erecta and D. melanogaster–D. pseudoobscura are approximately 9 and 25 million years respectively, whereas those for mouse–human are approximately 80 million years. Clearly, the rhino chromo and shadow domains are far more divergent between these Drosophila species than the chromo domains of HP1A, -B, and -C.
Among the HP1 paralogs, the HP1D/rhino gene appears to be particularly rapidly evolving. In phylogenetic analyses, both the rhino chromo and shadow domains appear to have evolved far more rapidly (Figure 1B–C) than their counterparts in other HP1s in Drosophila (compare branch lengths between D. melanogaster, D. erecta, and D. pseudoobscura orthologs, which have bold branches). HP1E also appears to evolve rapidly in its chromo domain but is not preserved in D. pseudoobscura. Thus, rhino appears unique among the HP1-like genes in being well conserved yet evolving rapidly. Because rhino is evolving so rapidly, orthologs are not likely to be unambiguously identified in other organisms.
rhino Is Expressed Predominantly in Ovaries
Previous Northern blot analysis had detected a 1.6 kb rhino mRNA in female flies, early embryos, and ovary, but not in male flies and rhino mutants [36]. In situ analysis showed that the rhino transcript was present both within the germline and somatic cells of the ovary [36]. However, an abundant and much larger band on the Northern blot did not show the same restricted expression pattern. This band was also present in RNA made from rhi2 mutant flies suggesting that it did not contain rhino transcript. In order to further delineate the expression pattern of this unusual HP1 gene, we used RT-PCR to assess the presence of rhino mRNA in male or female flies and in different tissues, because it provides a more sensitive assay that complements the previous Northern analysis (Figure 2).
Figure 2 RT-PCR Analyses of the Various HP1 Paralogs
(A) The rhino gene from D. melanogaster is drawn to scale. Exons are boxed (grey fill indicates coding sequence) and lines indicate introns. The position of a P[lacZ, ry+] (PZ) element in the rhi2 mutant is shown (triangle; not to scale). Dmid1f and Dmid2b RT-PCR primers span the first rhino intron. RT-PCR was carried out on roughly equivalent amounts of RNA using a primer set for rhino or actin-42A (primer sequences in Table S2). Control reactions contained no RNA or D. melanogaster genomic DNA.
(B) The rhino gene from D. bipectinata is schematized and primers used for RT-PCR indicated. RT-PCR analysis shows that rhino is specifically expressed in ovaries. D. bipectinata separated from D. melanogaster approximately 13 million years ago.
(C) RT-PCR reactions carried out for the other HP1 paralogs in D. melanogaster. HP1A, -B, and -C are ubiquitously expressed in adult tissues whereas HP1E expression appears to be predominantly restricted to the male testes.
We confirmed the predominant expression of rhino in D. melanogaster ovaries, although low levels of transcript could also be detected in testis, head, and faintly in carcass, likely below detection limits for Northern analysis (Figure 2A). Endogenous rhino transcript was also present in S2 tissue culture cells that were used for our localization studies. Furthermore, the absence of any rhino transcript from rhi2 mutant flies by RT-PCR confirms that the large cross-reacting band seen on previous Northern analysis [36] does not contain rhino transcript. We have extended this finding to show that the predominant expression of rhino is restricted to ovaries in another distantly related species, D. bipectinata (Figure 2B). In contrast, we found that HP1A, -B, and -C genes were abundantly expressed in all gross adult tissues that we examined (Figure 2C). Interestingly, HP1E showed an expression pattern restricted predominantly to the male testis, suggesting that two of the five HP1 paralogs in D. melanogaster are each devoted predominantly to testes and ovaries respectively. This may highlight the fact that chromatin structure is likely to be inherently different in somatic versus germline cells, that may have spurred this specialization.
Rhino Localization in D. melanogaster Cells
The localization of protein products of three HP1 genes have been tested so far in Drosophila tissue culture cells. Only HP1A was found to localize predominantly to heterochromatin, whereas HP1C localized to euchromatin and HP1B to both euchromatin and heterochromatin [35]. Therefore, we decided to first study the localization pattern of Rhino to determine whether it localized to heterochromatin. Drosophila S2 interphase cells have a DAPI-dense staining area that helps demarcate cytological boundaries of heterochromatin, although it is worth noting that DAPI does not stain all heterochromatic DNA, owing to sequence-dependent DNA-binding preference [38]. H3K4me is an excellent cytological marker for euchromatin, whereas H3K9me marks heterochromatin [10,11]. The localization patterns of green fluorescent protein (GFP) fused to HP1A, HP1B, or HP1C and expressed in tissue culture cells were previously shown to be faithful representations of the localization of the endogenous proteins by antibody staining [35]. We therefore expressed rhino as a C-terminal GFP fusion protein in Drosophila S2 cells, followed by immunostaining with antibodies to HP1A, HP1B, HP1C, or specific modifications of histone H3 (Figure 3) for comparison of localization patterns.
Figure 3 Rhino-GFP Localizes to Distinct Foci in the Heterochromatic Domain
A C-terminal GFP fusion protein of rhino was transiently expressed in Drosophila tissue culture cells (green in merge). Nuclei were stained with DAPI that stains DNA (blue in merge) and antibodies (red in merge) to HP1A, HP1B, HP1C, H3K9me, H3K4me, or fibrillarin (a nucleolar protein). H3K4me stains euchromatin whereas HP1A, H3K9me, Rhino-GFP, and bright DAPI staining all fall within heterochromatin. Rhino-GFP does not overlap with any of the antibody staining patterns, but appears to localize adjacent to HP1A and H3K9me within the heterochromatic domain.
The localization pattern of Rhino-GFP differed from that of HP1A, -B, and -C in interphase tissue culture cells (Figure 3 and Figure S1). Rhino-GFP formed distinct foci that occupied a limited area in the nucleus. These Rhino-GFP foci were located in the heterochromatic compartment as defined by the absence of H3K4me staining. Strikingly, Rhino-GFP also did not directly overlap with common markers of heterochromatin, HP1A or H3K9me, rather appearing interspersed with, or surrounding these signals. Thus, unlike HP1A, we expect that the Rhino chromo domain does not bind H3K9me. The Rhino-GFP localization pattern was not an artifact of GFP-tagging because it was also observed with a Rhino protein that was N-terminally tagged with a biotinylated peptide (Figure S1). We conclude that among HP1 paralogs, Rhino-GFP has a unique localization pattern within the heterochromatic domain in tissue culture cells. Its localization pattern in oocytes is currently unknown.
Molecular Evolution of rhino: Positive Selection of the Hinge and Chromo Shadow Domains
The indication that rhino may be a rapidly evolving HP1 (see Figure 1), its predominant expression in ovaries (see Figure 2), and its interesting cytological localization pattern (Figure 3), led us to investigate its evolutionary history in further detail. Uncovering evolutionary constraints under which different HP1 genes evolve can provide insight into the evolutionary forces that shape heterochromatin. To study the molecular evolution of HP1 proteins in Drosophila, we obtained DNA sequence for HP1 orthologs in the closely related D. simulans species (diverged from D. melanogaster about 2.5 million years ago) by PCR.
Rapid evolution of HP1s may be attributed to relaxed constraint, allowing sequence changes to accumulate, especially if different gene copies are functionally redundant. Alternatively, amino acid replacement changes may confer a selective advantage, in which case they would be expected to accumulate at a rate higher than expected under neutral evolution (positive selection). To evaluate whether any of the HP1s are undergoing such positive selection between the closely related D. melanogaster and D. simulans species, we performed a 100-bp sliding window analysis of the number of replacement changes per site (dN) compared to the number of synonymous changes per site (dS) (Figure 4). HP1B, HP1C and HP1E had dN < dS in all windows, consistent with purifying selection, as expected for structural proteins evolving under strict constraints. To our surprise, we found dN > dS for several windows in the rhino gene corresponding to the hinge region of the encoded protein (Figure 4E). We used Monte Carlo simulations in the K-estimator program to show that three of these windows were statistically significant (dN > dS, p < 0.02, indicated by asterisks), consistent with positive selection of rhino between D. melanogaster and D. simulans. We also found two windows with dN > dS for the HP1A-encoded hinge with borderline significance (p-value approximately 0.05), but further detailed analysis including a population study of D. melanogaster and D. simulans, and dN/dS comparisons among several other pairs of closely related Drosophila species (D. Vermaak, H. S. Malik, unpublished data) led to the conclusion that there was no positive selection of HP1A. Thus, no other HP1 homolog other than rhino showed any evidence of positive selection, suggesting that HP1D/rhino is again unique in this respect, not just among Drosophila
HP1 paralogs, but also among all HP1 genes identified so far.
Figure 4 Comparison of D. melanogaster and D. simulans HP1s
(A–E) Different D. melanogaster and D. simulans HP1 coding DNA sequences were aligned (indels and unalignable sequences were removed) and dN (black line) and dS (grey line) values were calculated using K-estimator [75] with a sliding window of 100 bases and a 35-bp step size. The domain structure of each HP1 is shown schematically and to scale beneath each plot, with the dark rectangle representing the chromo domain and the grey rectangle the chromo shadow domain. For HP1A, dN exceeds dS in the hinge region, but dS is very low in these windows. In contrast, for rhino, dN is higher throughout and exceeds dS in several windows corresponding to the hinge region (dN/dS values are also plotted for rhino). Windows in which statistically significant values for positive selection were obtained (dN/dS > 1, p < 0.02), are indicated by asterisks and map to the hinge region.
Sliding window dN/dS analyses suggest that rhino is subject to positive selection. To follow up on this initial observation, we undertook a more detailed study in D. melanogaster and D. simulans. We used PCR to obtain rhino sequence from 17 strains of D. melanogaster and 11 strains of D. simulans. DNA sequence changes were categorized as replacement (R) or synonymous (S) (Table S1). Changes were further classified as either fixed between species (f) or polymorphic within species (p) (Table S1). Under a neutral evolutionary model, the ratio of replacement to synonymous changes that have been fixed between species (Rf:Sf) is expected to be roughly the same as the ratio for polymorphic changes (Rp:Sp) (McDonald-Kreitman test) [39]. We did not find a significant deviation from neutrality when the entire rhino sequence was considered (Table 1, entire coding region, p = 0.13). However, a sliding window analysis clearly showed that the observed fixed replacement changes far exceeded those expected under neutral evolution in the C terminal part of the protein (Figure 5). Indeed, the shadow domain had a highly significant deviation from neutrality (p < 0.01), suggesting that this domain has been subject to strong positive selection (Table 1, shadow). We used parsimony to assign each DNA sequence change within the shadow domain to either the melanogaster or simulans lineage by polarizing the changes to outgroup species D. teissieri and D. yakuba (Table S1, changes in the melanogaster lineage [m] or simulans lineage [s]). We concluded that the shadow domain has been subject to positive selection in the D. simulans lineage (Table 1, shadow D. simulans only, p < 0.05), but there were not enough polymorphisms to reach a similar conclusion for the D. melanogaster lineage despite a strong Rf:Sf ratio. Although the complete hinge alone does not reject neutrality, separating the long hinge domain into N- and C-terminal segments suggests that the C-terminal region of the hinge, abutting the shadow domain, has been subject to positive selection (p < 0.01). We could not determine whether the positive selection in the hinge was lineage specific because of ambiguity in the alignment with outgroup species. Despite this strong signal for positive selection, we were unable to detect evidence of recent adaptive “sweeps” using Fu and Li [40] or Tajima [41] tests, suggesting that any such sweeps were not recent enough to result in standing single polymorphisms. Thus, both the hinge and shadow domains of the protein encoded by rhino show strong evidence for relatively old episodes of positive selection between the D. melanogaster and D. simulans lineages.
Table 1 McDonald-Kreitman Test of rhino in D. melanogaster and D. simulans
aG-value (with Williams correction) for test of 2 × 2 independence.
b
p-Value calculated using Fisher's exact test because the G-value calculation does not permit zeros.
cWe arbitrarily define the N- and C-terminal hinge boundaries at position 750 in the rhino alignment based on Figure 5.
Figure 5 Population Genetics of HP1D/rhino between D. melanogaster and D. simulans
Replacement changes that have been fixed between D. melanogaster (17 strains) and D. simulans (11 strains) (Rf obs [observed], open bars) were calculated with a 300-nucleotide sliding window, 25-nucleotide step size. The number of expected replacement changes for each window (Rf exp; solid bars) were calculated from the neutral expectation of the McDonald-Kreitman test (Rf:Sf ≈ Rp:Sp). Rf obs exceeds Rf exp in the C-terminal part of rhino (the C-terminal part of the hinge and the shadow domain as shown beneath), consistent with positive selection (also see Table 1). The chromo and chromo shadow domains are represented by dark and light rectangles, respectively.
rhino Evolution in Other Drosophila Species
Is the positive selection of rhino limited to the melanogaster species group? To address this question, we identified D. pseudoobscura
rhino by synteny with D. melanogaster;
rhino is contained within an intron of another gene in both species. We used RT-PCR to confirm the predicted splice sites for rhino from the obscura species group. D. pseudoobscura
rhino is very different in length (317 vs. 418 encoded amino acids) and sequence from D. melanogaster
rhino (see Figure 1). In fact, the hinge region of the Rhino protein is changing so rapidly that it is unrecognizable in a BLAST comparison between D. melanogaster and D. pseudoobscura (e-value > 1,000). To trace the evolution of rhino beyond the melanogaster species group, we obtained rhino sequence from intervening species between D. melanogaster and D. pseudoobscura. Despite the evolutionary distance of D. melanogaster from D. pseudoobscura, we could identify noncoding conserved sequences both upstream and downstream of the rhino gene, allowing us to design primers to amplify rhino from 12 additional Drosophila species, shown schematically in Figure 6A.
Figure 6 Positive Selection of rhino in 25 Million Years of Drosophila Evolution
(A) rhino was PCR amplified and sequenced from the indicated Drosophila species. Predicted protein sequences are drawn to scale with amino acid length shown on the right. The chromo and chromo shadow domains are relatively conserved and are indicated by the large dark and light rectangles, respectively. The hinge regions are rapidly evolving. They differ dramatically in size and sequence and cannot be aligned between different species groups (indicated on the left) and sometimes not even within the same species group, for example the D. bipectinata versus D. ananassae hinge. Within the melanogaster species group, D. melanogaster rhino appears to have undergone large deletions up to 50 codons in the hinge region compared with its closest relative D. simulans (indicated by slanted lines). These deletions are adjacent to the adaptively evolving hinge region identified between D. melanogaster and D. simulans. A 58 amino acid duplication present in the ananassae species group is indicated by grey arrows. Thin black rectangles indicate runs of serine ranging between 70% and 100% serine.
(B) dN and dS calculations for rhino from alignments of two pairs of closely related species from the melanogaster and takahashii species groups show multiple windows in which dN exceeds dS, indicative of positive selection.
We find that rhino is evolving at an unprecedented rate for an HP1. The hinge regions cannot be unambiguously aligned between different species groups or in some cases not even within the same species group. We did not detect any significant similarity of the Rhino hinge regions to other proteins or motifs, yet all the hinge regions share certain sequence features, most noticeably long runs of serines as well as proline- and glutamine-rich sequences (Figure 6A). In some instances, we found clear evidence of positive selection (dN/dS > 1) for alignable segments of hinge regions of closely related pairs, D. yakuba versus D. teissieri, D. erecta versus D. orena, D. lutescens versus D prostipennis, D. bipectinata versus D. parabipectinata, and D. pseudoobscura versus D. miranda (representative examples shown in Figure 6B). Our ability to detect instances of dN/dS > 1 within the hinge region for multiple species pairs within a small sampling of Drosophila species suggests that positive selection of the hinge is a common feature in rhino evolution.
Positive Selection of rhino Chromo Domain
Phylogenetic analyses (see Figure 1B and 1C) suggested that not just the hinge region, but also the chromo and shadow domains of Rhino are diverging more rapidly than similar domains of other HP1s. For the hinge and shadow domains, we have already presented evidence that this rapid evolution is not due to lack of selective constraint, but rather due to positive selection. However, we were unable to detect positive selection within the chromo domain using dN/dS or McDonald-Kreitman tests, nor were we able to detect significant evidence of an adaptive sweep using standard tests (Fu and Li [40], Hudson-Kreitman-Aguade [42], Tajima's D [41]). We reasoned that it may be hard to detect positive selection of the chromo domain because the majority of its codons are likely to be functionally constrained and therefore under purifying selection. However, such purifying selection may be masking positive selection of a small number of codons within the chromo domain. We therefore used a codon by codon maximum likelihood test, PAML [43], to ask if we could detect any codons that have been under repeated, strong positive selection.
We used a DNA sequence alignment of the rhino gene corresponding to the encoded chromo domain from different Drosophila species. The corresponding amino acid sequence alignment is shown in Figure 7A. We note that a tree based upon this amino acid alignment is in agreement with the accepted Drosophila phylogeny [44], suggesting that we are considering strict orthologs. Remarkably, models that allow codons to evolve under positive selection (M8 and M2) fit the data significantly better than associated models that do not permit positive selection (M7 and M1) (p < 0.001 in all cases, Table 2). Just a few codons account for this positive selection. In particular, three codons repeatedly show highly significant posterior probabilities (1E, 9L, and 25S in Table 2; arrows Figure 7A). The Rhino chromo domain structure is likely to be similar to that of known HP1 chromo domains, so we show the likely positions of the three adaptively evolving amino acids of the Rhino chromo domain on the known structure of Drosophila HP1A chromo domain bound to H3K9me peptide (Figure 7A; [45]). Position 1 is in close proximity to the groove that binds the methylated peptide, suggesting that this amino acid may be driven to adapt to a constantly changing substrate of Rhino. We cannot rule out that positions 9 and 25 may also be adapting to a substrate binding in the same position, because they could influence the overall conformation and thus binding specificity of the chromo domain. However, positions 9 and 25 are expected to be solvent accessible on the opposite side of the Rhino chromo domain and may represent an additional, potentially novel, interaction surface.
Figure 7 Positive Selection of the Chromo and Chromo Shadow Domains of rhino
(A) An amino acid alignment of the chromo domain of different Drosophila species is shown with the distantly related HP1A and human HP1α chromo domains for comparison. The neighbor-joining tree based on this alignment (shown on the left) recapitulates known Drosophila phylogeny. Amino acids of the HP1A chromo domain that are involved in binding to H3K9me are color coded: Blue amino acids form an aromatic cage that recognize K9me, and pink amino acids form a complementary surface for recognition of the H3 peptide [45]. The corresponding DNA sequence alignment was used in a PAML analysis. Three codons that have been under repeated and strong positive selection are indicated by arrows. The corresponding positions (red) are indicated on the known structure of the Drosophila
HP1A chromo domain (light blue) bound to H3K9me (purple) [45].
(B) Amino acid alignment of representative chromo shadow domains of rhino orthologs from Drosophila. The neighbor-joining tree based on this alignment also recapitulates Drosophila phylogeny. Amino acids of mouse HP1β known to be involved in dimerization are shown in pink and those required for the shadow fold in blue [46]. We use arrows to indicate codons identified as being under positive selection by our PAML analysis. Corresponding positions of the mouse HP1β chromo shadow domain are indicated (red) on one of the shadow domains (light blue) of the dimer [46]. These positions are shown in yellow on the other shadow domain (green).
Table 2 PAML Analyses of rhino Chromo and Shadow Domains in Drosophila
Codon positions are as defined in D. melanogaster. λ refers to log likelihood. Branch length, S, refers to the number of nucleotide substitutions per codon, and can vary from one model to another (range shown). Analyses shown were carried out using the F61 model of codon frequencies, but similar results were obtained with the F3×4 model.
We have already shown that the shadow domain is under strong positive selection between D. melanogaster and D. simulans. To find out if some codons of the shadow domain have also been under continuous positive selection, we carried out a PAML analysis. A tree based on the shadow domain amino acid alignment also recapitulates Drosophila phylogeny (Figure 7B). We found significant evidence for positive selection, with most of the signal coming from just three codons (Table 2; Positions S31, I33, and I59 in Figure 7B). On the structure of a mouse HP1β shadow domain dimer [46], positions 31 and 33 are on the same side of the shadow dimer and should be available for protein–protein interactions. The vertebrate HP1 chromo shadow domain dimerization site is known to bind to many proteins through their PxVxL motif [47–49]. It is unclear if these interactions are conserved in Rhino, but rapid evolution in this region (including position 59 identified in our PAML analysis) certainly has the potential to easily influence protein–protein interactions.
PAML analyses like these are very useful to highlight codons that have been repeatedly subject to positive selection [43,50]; however, they do run a risk of false positives. This is somewhat ameliorated in our dataset because the tree lengths are of moderate value (Table 2). Similar tree lengths have been shown by simulations to have a significantly lower risk of false positives [51]. Nonetheless, the true test for the significance of these positively selected residues will come from functional assays on Rhino function and localization.
Discussion
In this paper, we have undertaken an evolutionary study of HP1-like proteins, with the ultimate aim of discerning the selective pressures that act on heterochromatin. We have found that Rhino, the only HP1 paralog that is expressed predominantly in ovaries, encodes a protein that has a unique localization pattern in S2 cells. Although it is excluded from the euchromatic compartment, the Rhino protein does not overlap with HP1A or H3K9me. This immediately suggests that H3K9me or HP1A does not mark all Drosophila heterochromatin, and that Rhino has a uniquely different specificity for a previously unappreciated compartment in heterochromatin.
It has not been easy to discern the molecular function of rhino from mutant phenotypes in eggshell defects. Possibilities range from a role for Rhino in gross chromatin structural changes, to transcriptional or translational regulation and even microtubule organization in the oocyte [36]. Despite our current lack of knowledge about the molecular function of rhino, the fact that mutations are female sterile, point to its importance to proper oogenesis [36]. HP1A, which is far better understood, is an essential gene. Such chromosomal proteins serving crucial functions are expected to be under strong evolutionary constraints and purifying selection. Although this is true for four of the five D. melanogaster HP1s including HP1A (see Figure 4), we find that all three domains of rhino have evolved under positive selection using multiple criteria, including dN/dS ratios, McDonald-Kreitman, and PAML analyses [39,43]. What could be driving this positive selection of such an important HP1 protein?
Can co-evolutionary pressures explain the positive selection acting on rhino? For instance, rhino might be continually “catching up” to mutations in interacting proteins required for its function. We believe this is unlikely, because mutations that compromise a required interaction are likely to be culled out of the population by purifying selection, long before a chance compensatory mutation in rhino can occur. A second possibility is that the positive selection of rhino may be driven by changes in the regulation of key genes between two species. Although we cannot formally rule out such a scenario, it appears unlikely to explain the relatively constant positive selection that we have seen for approximately 25 million years of Drosophila evolution.
Positive selection need not involve rhino's “normal” function, whatever that may be, but rather underlie a second and unrelated “defense” function of rhino. In such a scenario the positive selection on rhino would be driven by a recurrent intracellular conflict that yields a selection advantage to the “winner.” Genes encoding proteins involved in direct host–parasite interactions are often subject to positive selection. In this case, changes that are beneficial for the parasite (to evade interactions for instance) will be followed by selection favoring changes in the host proteins (that restore interactions). Thus, two antagonistic entities locked in genetic conflict face repeated episodes of positive selection, only to arrive at the same quasi-steady state, a scenario formalized as the “Red Queen” hypothesis [52]. rhino may be subject to the same kind of genetic conflict that occurs intracellularly. It is especially intriguing that the only HP1 we have found to be subject to positive selection is expressed predominantly in ovaries ([36] and Figure 2), where such a competitive advantage has directly heritable consequences. We consider two models of genetic conflict to explain rhino's positive selection.
Under the first model, rhino participates in suppressing “selfish” behavior of centromeres, which can compete to maximize their transmission advantage in female meiosis, where only one of four meiotic products is destined to become the egg [21]. We have previously proposed that this kind of drive can have deleterious consequences for male meiosis and is likely to be suppressed either by centromeric proteins altering their DNA-binding specificity [19,20] or by heterochromatin proteins evolving to limit centromere boundaries, and thereby limiting “strength” [4,21,53]. Similar selective pressures have been previously proposed to result in deleterious mutations in the nod chromokinesin in D. melanogaster [54]. rhino may represent another repressor of the drive by directly or indirectly influencing centromere strength.
A second model is that positive selection on rhino is a direct result of genetic conflict between rhino and mobile genetic elements. Although we have no evidence to support this hypothesis, it is attractive for several reasons. Transposable elements can evolve rapidly and differ significantly between Drosophila species, including D. melanogaster and D. simulans [55,56]. Rhino-GFP localizes to the heterochromatic region of the nucleus (see Figure 3), which is highly enriched in transposable elements [57]. Finally, genome-bound transposable elements can only increase their genomic copy number by transposing in the germline, increasing selective pressures on host proteins that act as suppressors of germline transposition. Rhino may either interact with the integration machinery of transposons to direct their integration into transcriptionally silent heterochromatin, or it may directly bind and transcriptionally repress transposable elements that are newly introduced into heterochromatin. Some transposable elements are known to be major in vivo targets of HP1A, apparently involving the RNAi (RNA interference) pathway [1,11,58–62]. Similarly, rhino may be under continual selection to directly bind transposable elements.
Whatever is driving the positive selection of rhino, mutations in any of Rhino's three domains appear to be selected to give rhino the upper hand in the current round of competition. The chromo and related shadow domains are very versatile interaction domains that can influence binding to DNA, RNA, and proteins [63]. The hinge domain can also strongly influence localization of HP1-like proteins [64,65]. Future experiments will address the functional role of the three amino acids under recurrent positive selection in the chromo and shadow domains (Figure 7) and help to distinguish between our models of what drives the positive selection of rhino. These experiments promise to reveal insights into the organization of a substantial portion of Drosophila genomes. It is probably not a coincidence that we have found positive selection only in an HP1-family member that is expressed predominantly in ovaries. Indeed, a restricted expression pattern may have allowed detection of a previously unremarked conflict that shapes at least a fraction of Drosophila heterochromatin, via the positive selection of rhino. Such a signal may have been masked for other HP1s due to their constrained roles in other tissues.
Our results complement previous findings that other proteins that bind heterochromatin appear to be among the most rapidly evolving proteins in an unbiased screen in Drosophila [67–68], although this does not appear to be the result of positive selection [69]. Polymorphisms in heterochromatin-binding proteins can have direct effects on non-disjunction frequencies [54,70,71]. Similarly, although HP1A, -B, and -C appear to be conserved and evolving under purifying selection, HP1 evolution (in both sequence and gene copy number; see Figure 1) in general appears quite rapid for a chromosomal protein with a highly conserved function in most eukaryotes. Thus, rapid changes in the genomic landscape may underlie rapid diversification of genes encoding HP1s and chromosomal proteins in general.
Materials and Methods
Sequences from Drosophila and databases and RT-PCR.
Drosophila species and strains (Table S1) were obtained from the Drosophila stock center (currently in Tucson, Arizona) and genomic DNA was prepared by standard methods [19]. The rhino locus was amplified using PCR Supermix High Fidelity (Invitrogen, Carlsbad, California, United States) with the primers indicated in Table S2. PCR products were either sequenced directly or following Topo-TA cloning (Invitrogen). RNA was prepared from whole male or female flies or different tissues (head, ovary, testis, or carcass) using a kit (Qiagen RNeasy; Qiagen, Valencia, California, United States) and cleared of genomic DNA by DNase I digestion (Ambion DNA-free; Ambion, Austin, Texas, United States). RNA concentrations were measured from various tissues, and the same amount of total RNA was used as template in the RT-PCR analysis. RT-PCR (Invitrogen) to evaluate the presence of rhino mRNA was carried out using Dmid1f and Dmid2b primers (Table S2) that span the rhino intron, along with actin-42A primers [72] as a control. For D. bipectinata, primers dv15 and dv230 that span the rhino intron were used. RT-PCR and sequencing was carried out to confirm the predicted splice-site positions for rhino from D. simulans (strain 2), D. bipectinata, and D. miranda. Splice sites for rhino from other species were predicted using Berkeley Drosophila Genome Project Splice site predictor (http://www.fruitfly.org/seq_tools/splice.html). All sequences have been deposited in Genbank (accession numbers AY944308–AY944358, Table S2).
Sequence analysis.
Sequences were assembled using DNA Strider [73]. Clustal_X [74] was used to obtain pairwise or multiple alignments and to generate formatted files for further analysis. Pairwise sequence alignments used for dN/dS analysis were hand edited, using the amino acid sequence as a guide to place indels. For instance, there is an 80 amino acid length difference between the D. melanogaster and D. simulans hinge regions. These regions cannot be compared in tests for positive selection. Pairwise dN and dS comparisons and confidence values were calculated using the K-estimator software [75,77]. Sliding window size was arbitrarily chosen as 100 bases with 35 base steps for all pairwise dN/dS comparisons. Confidence interval estimates were calculated using Monte Carlo simulations, taking into account (1) dN and dS values, (2) the number of codons, (3) transition: transversion ratio, and (4) GC content and amino acid composition. Thus, K-estimator [75] at least takes into account most of the confounding variables that are known to give false positives in terms of dN/dS. We also present a dN/dS analysis using the reconstructed hypothetical ancestors to all the D. melanogaster and D. simulans
rhino sequences (Figure S2).
The DNASP software package [77] was used to perform several tests for positive selection using genomic sequence of rhino from 17 strains of D. melanogaster and 11 strains of D. simulans. The Fu and Li [40], Tajima's D [41], and Hudson-Kreitman-Aguade [42] tests were carried out on the complete sequence, including the intron, whereas the McDonald-Kreitman test [39] was carried out on coding regions only (1,209 total positions with indels removed). Fixed replacement changes in the chromo and chromo shadow domains were polarized using D. yakuba and D. teissieri sequences as outgroups, but we could not unambiguously polarize all changes in the hinge region. The expected fixed replacement changes (Rfexpected) shown in Figure 4B were calculated from the ratio Rfexpected = Sfobserved(Rpobserved/Spobserved) according to the neutral expectation in the McDonald-Kreitman test, where R = replacement, S = synonymous, f = fixed between population, p = polymorphic within the population (similar to the previously proposed “Neutrality Index” [78]). A sliding window of 300 nucleotides with step size of 25 was used for presentation purposes.
Neighbor-joining phylogenetic trees were constructed using the PAUP software, version 4.0b10 [79] and appropriate Clustal_X multiple alignments of either the chromo or chromo shadow domains. A total of 1,000 replicates were carried out for bootstrapping. Maximum likelihood analysis was performed with the PAML software package [43] in separate analyses for multiple alignments of the chromo domain and the shadow domains (the rapid evolution of the hinge in both size and sequence precluded its comparison in such a multiple alignment). Codons that were repeatedly subject to positive selection were identified using N sites models (M1, M7) that do not permit positive selection compared to models (M2, M8) that permit sites to evolve under positive selection. The strength of positive selection was calculated by comparing twice the log likelihood difference (M2 vs. M1, M8 vs. M7) in a chi-square test with two degrees of freedom. Codons that were identified as having evolved under positive selection with high posterior probabilities (p > 0.95) were highlighted on a three-dimensional structure of the respective domains and visualized using the Cn3D software (version 4.0) [80].
Plasmid constructs.
A plasmid for expressing rhino as a C-terminal GFP fusion protein under control of the hsp70 heat shock promoter (HSRhiGFP) was constructed as follows: rhino coding sequence flanked by XbaI and NotI restriction enzyme sites was amplified by RT-PCR from D. melanogaster (Canton S) using primers KcRhiF and KcRhiB (Table S1). The PCR product was digested and cloned into a modified heat shock expression plasmid [81] that had been digested with XbaI and EagI and phosphatase treated to yield the rhino open reading frame followed by a six amino acid linker and GFP. Correct cloning was verified by sequencing. An N-terminal fusion protein of a biotin recognition peptide (MAGGLNDIFEAQKIEWHEDTGGS) to rhino (BLRPRhi) was constructed as follows: Primers dv99 and dv100 were used to amplify rhino coding sequence with flanking NotI and BamHI restriction enzyme sites from the HSRhiGFP plasmid. The PCR fragment was TA cloned and the sequence verified before digestion of the TA clone and subcloning of the gel-isolated fragment into a BLRP expression vector with a metallotheionine promoter [82,83]. A plasmid (pBirA) expressing the Escherichia coli biotin ligase enzyme (BirA) from a metallotheionine promoter was a gift from Takehito Furuyama.
Cell culture, transfection, and immunostaining.
S2 cells (Invitrogen, D-mel2) were maintained in serum-free insect media (Invitrogen) supplemented with 90 ml/l of 200 mM L-Glutamine (Sigma, St. Louis, Missouri, United States). Twenty micrograms of the HSRhiGFP plasmid was transfected as previously described [81]. Cells were heat shocked for 1 h on the next day and allowed to recover for 2 h before immunostaining [84]. In the case of the BLRPrhino construct, 10 μg of plasmid DNA were co-transfected with 10 μg of pBirA plasmid that contains the biotin ligase under control of a metallotheionine promoter. After overnight incubation, cells were induced for 3 h with 500 μM CuSO4, added directly to the media, followed by immunostaining. HP1A, HP1B, and HP1C antibodies have been previously described [35]. Antibodies to H3K9me or H3K4me were purchased from Upstate Biotech (Waltham, Massachusetts, United States). Monoclonal mouse anti-Fibrillarin antibody was purchased from Encor Biotechnology Inc (Alachua, Florida, United States). All antibodies, including the secondary Texas-red fluorescently labeled goat anti-rabbit or anti-mouse antibodies (Amersham, Piscataway, New Jersey, United States), were used at a dilution of 1/200, with the exception of the anti-fibrillarin antibody that was used at 1/500. Images of nuclei were obtained and de-convolved using the Deltavision software (Applied Precision, Issaquah, Washington, United States).
Supporting Information
Figure S1 Rhino-GFP Localization in Drosophila S2 Cells
These additional images of Rhino-GFP show a localization pattern that is distinct from HP1A, H3K4me, and H3K9me. In addition, an N-terminal biotinylated-tagged Rhino protein shows the same localization pattern as that of the C-terminal GFP-tagged Rhino protein.
(5.2 MB PDF)
Click here for additional data file.
Figure S2 A Sliding Window dN/ dS Analysis
Only those changes that were found to have been fixed differences between D. melanogaster and D. simulans were used. All intraspecific polymorphisms were eliminated for this analysis. Compared to Figure 4, the signal for positive selection now appears concentrated exclusively in the C-terminal region of rhino.
(203 KB PDF)
Click here for additional data file.
Table S1 All Polymorphisms within the Coding Region of the rhino Gene in D. melanogaster and D. simulans Are Shown
Changes are highlighted as being either fixed (f) between species or polymorphic (p) within species, as replacement (R) or synonymous (s) changes. Fixed changes were polarized using an outgroup species to changes along either the D. melanogaster (m) or D. simulans (s) lineages. Many changes could not be unambiguously polarized.
(34 KB DOC)
Click here for additional data file.
Table S2 List of Primers Used and Accession Numbers of Sequences Obtained in This Study
(36 KB XLS)
Click here for additional data file.
Accession Numbers
The Flybase (http://flybase.bio.indiana.edu) accession numbers of the genes discussed in this paper are rhino (CG10683) and HP1E (CG8120). New sequences obtained during the course of this study have been deposited in Genbank under the accession numbers AY944308–AY944358. The Molecular Modeling Database (MMDB; http://www.ncbi.nlm.nih.gov/Structure/MMDB/mmdb.shtml) accession numbers of the proteins discussed in this paper are H3K9me (19011, PDB 1KNE) and HP1β shadow domain dimer (13286, PDB 1DZ1).
We thank the Drosophila stock center (Tucson, Arizona) for various Drosophila species stocks and Judy O'Brien for maintenance of Drosophila stocks. We are grateful to Kami Ahmad and Celeste Berg for useful discussions throughout this project, and Jiro Yasuhara, Barbara Wakimoto, Sara Sawyer, and Julie Kerns for comments on the manuscript. We also gratefully acknowledge Terri Bryson for help with maintenance of Drosophila tissue culture cells, and Takehito Furuyama for the BLRP and pBirA plasmids. This work was supported by a Damon Runyon Cancer Postdoctoral Fellowship (DV), by the Howard Hughes Medical Institute (SH), startup funds from the Fred Hutchinson Cancer Research Center, and a Scholar Award from the Sidney Kimmel Foundation (HSM). HSM is an Alfred P. Sloan Fellow in Computational and Evolutionary Molecular Biology.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. DV and HSM conceived and designed the experiments. DV performed the experiments. DV and HSM analyzed the data. DV and SH contributed reagents/materials/analysis tools. DV, SH, and HSM wrote the paper.
Abbreviations
dNnumber of replacement changes per site
dSnumber of synonymous changes per site
ffixed between species
GFPgreen fluorescent protein
HP1heterochromatin protein 1
H3K4mehistone H3 dimethylated at lysine 4
H3K9mehistone H3 dimethylated at lysine 9
ppolymorphic within species
Rreplacement
Ssynonymous
==== Refs
References
Hodgetts R 2004 Eukaryotic gene regulation by targeted chromatin re-modeling at dispersed, middle-repetitive sequence elements Curr Opin Genet Dev 14 680 685 15531164
Kazazian HH Jr 2004 Mobile elements: Drivers of genome evolution Science 303 1626 1632 15016989
Deininger PL Moran JV Batzer MA Kazazian HH Jr 2003 Mobile elements and mammalian genome evolution Curr Opin Genet Dev 13 651 658 14638329
Malik HS Henikoff S 2002 Conflict begets complexity: The evolution of centromeres Curr Opin Genet Dev 12 711 718 12433586
Choo KH 2001 Domain organization at the centromere and neocentromere Dev Cell 1 165 177 11702777
Dillon N 2004 Heterochromatin structure and function Biol Cell 96 631 637 15519697
Bernard P Maure JF Partridge JF Genier S Javerzat JP 2001 Requirement of heterochromatin for cohesion at centromeres Science 294 2539 2542 11598266
Wakimoto BT Hearn MG 1990 The effects of chromosome rearrangements on the expression of heterochromatic genes in chromosome 2L of Drosophila
melanogaster
Genetics 125 141 154 2111264
Hoskins RA Smith CD Carlson JW Carvalho AB Halpern A 2002 Heterochromatic sequences in a Drosophila whole-genome shotgun assembly Genome Biol 3 RESEARCH0085 12537574
Richards EJ Elgin SC 2002 Epigenetic codes for heterochromatin formation and silencing: Rounding up the usual suspects Cell 108 489 500 11909520
Grewal SI Moazed D 2003 Heterochromatin and epigenetic control of gene expression Science 301 798 802 12907790
Weiler KS Wakimoto BT 1995 Heterochromatin and gene expression in Drosophila
Annu Rev Genet 29 577 605 8825487
Kirschmann DA Lininger RA Gardner LM Seftor EA Odero VA 2000 Down-regulation of HP1Hsalpha expression is associated with the metastatic phenotype in breast cancer Cancer Res 60 3359 3363 10910038
Luciani JJ Depetris D Missirian C Mignon-Ravix C Metzler-Guillemain C 2004 Subcellular distribution of HP1 proteins is altered in ICF syndrome Eur J Hum Genet 13 41 51
Li Y Danzer JR Alvarez P Belmont AS Wallrath LL 2003 Effects of tethering HP1 to euchromatic regions of the Drosophila genome Development 130 1817 1824 12642487
Li Y Kirschmann DA Wallrath LL 2002 Does heterochromatin protein 1 always follow code? Proc Natl Acad Sci U S A 99 Suppl 4 16462 16469 12151603
Pal-Bhadra M Leibovitch BA Gandhi SG Rao M Bhadra U 2004 Heterochromatic silencing and HP1 localization in Drosophila are dependent on the RNAi machinery Science 303 669 672 14752161
Eissenberg JC Morris GD Reuter G Hartnett T 1992 The heterochromatin-associated protein HP-1 is an essential protein in Drosophila with dosage-dependent effects on position-effect variegation Genetics 131 345 352 1644277
Malik HS Henikoff S 2001 Adaptive evolution of Cid, a centromere-specific histone in Drosophila
Genetics 157 1293 1298 11238413
Talbert PB Bryson TD Henikoff S 2004 Adaptive evolution of centromere proteins in plants and animals J Biol 3 18 15345035
Henikoff S Ahmad K Malik HS 2001 The centromere paradox: Stable inheritance with rapidly evolving DNA Science 293 1098 1102 11498581
Eissenberg JC Elgin SC 2000 The HP1 protein family: getting a grip on chromatin Curr Opin Genet Dev 10 204 210 10753776
Cowell IG Aucott R Mahadevaiah SK Burgoyne PS Huskisson N 2002 Heterochromatin, HP1 and methylation at lysine 9 of histone H3 in animals Chromosoma 111 22 36 12068920
Aasland R Stewart AF 1995 The chromo shadow domain, a second chromo domain in heterochromatin-binding protein 1, HP1 Nucleic Acids Res 23 3168 3173 7667093
Singh PB Miller JR Pearce J Kothary R Burton RD 1991 A sequence motif found in a Drosophila heterochromatin protein is conserved in animals and plants Nucleic Acids Res 19 789 794 1708124
Paro R Hogness DS 1991 The Polycomb protein shares a homologous domain with a heterochromatin-associated protein of Drosophila
Proc Natl Acad Sci U S A 88 263 267 1898775
Lachner M O'Carroll D Rea S Mechtler K Jenuwein T 2001 Methylation of histone H3 lysine 9 creates a binding site for HP1 proteins Nature 410 116 120 11242053
Bannister AJ Zegerman P Partridge JF Miska EA Thomas JO 2001 Selective recognition of methylated lysine 9 on histone H3 by the HP1 chromo domain Nature 410 120 124 11242054
Platero JS Hartnett T Eissenberg JC 1995 Functional analysis of the chromo domain of HP1 EMBO J 14 3977 3986 7664737
Ma J Hwang KK Worman HJ Courvalin JC Eissenberg JC 2001 Expression and functional analysis of three isoforms of human heterochromatin-associated protein HP1 in Drosophila
Chromosoma 109 536 544 11305786
Hayakawa T Haraguchi T Masumoto H Hiraoka Y 2003 Cell cycle behavior of human HP1 subtypes: Distinct molecular domains of HP1 are required for their centromeric localization during interphase and metaphase J Cell Sci 116 3327 3338 12840071
Minc E Allory Y Courvalin JC Buendia B 2001 Immunolocalization of HP1 proteins in metaphasic mammalian chromosomes Methods Cell Sci 23 171 174 11741155
Minc E Courvalin JC Buendia B 2000 HP1gamma associates with euchromatin and heterochromatin in mammalian nuclei and chromosomes Cytogenet Cell Genet 90 279 284 11124534
James TC Elgin SC 1986 Identification of a nonhistone chromosomal protein associated with heterochromatin in Drosophila melanogaster and its gene Mol Cell Biol 6 3862 3872 3099166
Smothers JF Henikoff S 2001 The hinge and chromo shadow domain impart distinct targeting of HP1-like proteins Mol Cell Biol 21 2555 2569 11259603
Volpe AM Horowitz H Grafer CM Jackson SM Berg CA 2001
Drosophila rhino encodes a female-specific chromo-domain protein that affects chromosome structure and egg polarity Genetics 159 1117 1134 11729157
Adams MD Celniker SE Holt RA Evans CA Gocayne JD 2000 The genome sequence of Drosophila melanogaster Science 287 2185 2195 10731132
Mohan S Yathindra N 1994 A study of the interaction of DAPI with DNA containing AT and non-AT sequences—molecular specificity of minor groove binding drugs J Biomol Struct Dyn 11 849 867 8204219
McDonald JH Kreitman M 1991 Adaptive protein evolution at the Adh locus in Drosophila
Nature 351 652 654 1904993
Fu YX Li WH 1993 Statistical tests of neutrality of mutations Genetics 133 693 709 8454210
Tajima F 1989 Statistical method for testing the neutral mutation hypothesis by DNA polymorphism Genetics 123 585 595 2513255
Hudson RR Kreitman M Aguade M 1987 A test of neutral molecular evolution based on nucleotide data Genetics 116 153 159 3110004
Yang Z 1997 PAML: A program package for phylogenetic analysis by maximum likelihood Comput Appl Biosci 13 555 556 9367129
Malik HS Vermaak D Henikoff S 2002 Recurrent evolution of DNA-binding motifs in the Drosophila centromeric histone Proc Natl Acad Sci U S A 99 1449 1454 11805302
Jacobs SA Khorasanizadeh S 2002 Structure of HP1 chromodomain bound to a lysine 9-methylated histone H3 tail Science 295 2080 2083 11859155
Brasher SV Smith BO Fogh RH Nietlispach D Thiru A 2000 The structure of mouse HP1 suggests a unique mode of single peptide recognition by the shadow chromo domain dimer EMBO J 19 1587 1597 10747027
Cowieson NP Partridge JF Allshire RC McLaughlin PJ 2000 Dimerisation of a chromo shadow domain and distinctions from the chromodomain as revealed by structural analysis Curr Biol 10 517 525 10801440
Thiru A Nietlispach D Mott HR Okuwaki M Lyon D 2004 Structural basis of HP1/PXVXL motif peptide interactions and HP1 localisation to heterochromatin EMBO J 23 489 499 14765118
Smothers JF Henikoff S 2000 The HP1 chromo shadow domain binds a consensus peptide pentamer Curr Biol 10 27 30 10660299
Wong WS Yang Z Goldman N Nielsen R 2004 Accuracy and power of statistical methods for detecting adaptive evolution in protein coding sequences and for identifying positively selected sites Genetics 168 1041 1051 15514074
Anisimova M Bielawski JP Yang Z 2002 Accuracy and power of Bayes prediction of amino acid sites under positive selection Mol Biol Evol 19 950 958 12032251
Van Valen L 1973 A new evolutionary law Evolutionary Theory 1 1 30
Sullivan BA 2002 Centromere round-up at the heterochromatin corral Trends Biotechnol 20 89 92 11841851
Zwick ME Salstrom JL Langley CH 1999 Genetic variation in rates of nondisjunction: Association of two naturally occurring polymorphisms in the chromokinesin nod with increased rates of nondisjunction in Drosophila
melanogaster
Genetics 152 1605 1614 10430586
Daniels SB Peterson KR Strausbaugh LD Kidwell MG Chovnick A 1990 Evidence for horizontal transmission of the P transposable element between Drosophila species Genetics 124 339 355 2155157
Kaminker JS Bergman CM Kronmiller B Carlson J Svirskas R 2002 The transposable elements of the Drosophila
melanogaster euchromatin: A genomics perspective Genome Biol 3 RESEARCH0084 12537573
Moshkin YM Belyakin SN Rubtsov NB Kokoza EB Alekseyenko AA 2002 Microdissection and sequence analysis of pericentric heterochromatin from the Drosophila melanogaster mutant Suppressor of Underreplication Chromosoma 111 114 125 12111334
Greil F van der Kraan I Delrow J Smothers JF de Wit E 2003 Distinct HP1 and Su(var)3–9 complexes bind to sets of developmentally coexpressed genes depending on chromosomal location Genes Dev 17 2825 2838 14630943
Sun LV Chen L Greil F Negre N Li TR 2003 Protein-DNA interaction mapping using genomic tiling path microarrays in Drosophila
Proc Natl Acad Sci U S A 100 9428 9433 12876199
Sun FL Haynes K Simpson CL Lee SD Collins L 2004 cis-Acting determinants of heterochromatin formation on Drosophila melanogaster chromosome four Mol Cell Biol 24 8210 8220 15340080
van Steensel B Henikoff S 2000 Identification of in vivo DNA targets of chromatin proteins using tethered dam methyltransferase Nat Biotechnol 18 424 428 10748524
Matzke MA Matzke AJ 2004 Planting the seeds of a new paradigm PLoS Biol 2 E133 15138502
Eissenberg JC 2001 Molecular biology of the chromo domain: an ancient chromatin module comes of age Gene 275 19 29 11574148
Badugu R Yoo Y Singh PB Kellum R 2005 Mutations in the heterochromatin protein 1 (HP1) hinge domain affect HP1 protein interactions and chromosomal distribution Chromosoma 113 370 384 15592864
Meehan RR Kao CF Pennings S 2003 HP1 binding to native chromatin in vitro is determined by the hinge region and not by the chromodomain EMBO J 22 3164 3174 12805230
Schmid KJ Tautz D 1997 A screen for fast evolving genes from Drosophila
Proc Natl Acad Sci U S A 94 9746 9750 9275195
Shareef MM Badugu R Kellum R 2003 HP1/ORC complex and heterochromatin assembly Genetica 117 127 134 12723692
Cenci G Siriaco G Raffa GD Kellum R Gatti M 2003 The Drosophila HOAP protein is required for telomere capping Nat Cell Biol 5 82 84 12510197
Schmid KJ Nigro L Aquadro CF Tautz D 1999 Large number of replacement polymorphisms in rapidly evolving genes of Drosophila . Implications for genome-wide surveys of DNA polymorphism Genetics 153 1717 1729 10581279
Afshar K Scholey J Hawley RS 1995 Identification of the chromosome localization domain of the Drosophila nod kinesin-like protein J Cell Biol 131 833 843 7490288
Zhang P Knowles BA Goldstein LS Hawley RS 1990 A kinesin-like protein required for distributive chromosome segregation in Drosophila
Cell 62 1053 1062 2144792
Tobin SL Cook PJ Burn TC 1990 Transcripts of individual Drosophila actin genes are differentially distributed during embryogenesis Dev Genet 11 15 26 1694472
Douglas SE 1994 DNA Strider. A Macintosh program for handling protein and nucleic acid sequences Methods Mol Biol 25 181 194 8004164
Thompson JD Gibson TJ Plewniak F Jeanmougin F Higgins DG 1997 The CLUSTAL_X windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools Nucleic Acids Res 25 4876 4882 9396791
Comeron JM 1999 K-Estimator: Calculation of the number of nucleotide substitutions per site and the confidence intervals Bioinformatics 15 763 764 10498777
Sawyer SL Emerman M Malik HS 2004 Ancient adaptive evolution of the primate antiviral DNA-editing enzyme APOBEC3G PLoS Biol 2 E275 15269786
Rozas J Sanchez-DelBarrio JC Messeguer X Rozas R 2003 DnaSP, DNA polymorphism analyses by the coalescent and other methods Bioinformatics 19 2496 2497 14668244
Rand DM Kann LM 1996 Excess amino acid polymorphism in mitochondrial DNA: Contrasts among genes from Drosophila, mice, and humans Mol Biol Evol 13 735 748 8754210
Swofford DL 2001 PAUP*: Phylogenetic analysis using parsimony (*and other methods) Version 4: Sunderland (Massachusetts) Sinauer Associates
Wang Y Geer LY Chappey C Kans JA Bryant SH 2000 Cn3D: Sequence and structure views for Entrez Trends Biochem Sci 25 300 302 10838572
Vermaak D Hayden HS Henikoff S 2002 Centromere targeting element within the histone fold domain of Cid Mol Cell Biol 22 7553 7561 12370302
de Boer E Rodriguez P Bonte E Krijgsveld J Katsantoni E 2003 Efficient biotinylation and single-step purification of tagged transcription factors in mammalian cells and transgenic mice Proc Natl Acad Sci U S A 100 7480 7485 12802011
Beckett D Kovaleva E Schatz PJ 1999 A minimal peptide substrate in biotin holoenzyme synthetase-catalyzed biotinylation Protein Sci 8 921 929 10211839
Henikoff S Ahmad K Platero JS van Steensel B 2000 Heterochromatic deposition of centromeric histone H3-like proteins Proc Natl Acad Sci U S A 97 716 721 10639145
|
16103923
|
PMC1183528
|
CC BY
|
2021-01-05 08:00:24
|
no
|
PLoS Genet. 2005 Jul 25; 1(1):e9
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010009
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 10.1371/journal.pgen.001001005-PLGE-RA-0067R2plge-01-01-06Research ArticleDevelopmentPediatricsGenetics/Genetics of DiseaseGenetics/Disease ModelsMammalsMus (Mouse)Homo (Human)
Fog2 Is Required for Normal Diaphragm and Lung Development in Mice and Humans Fog2 Required for Diaphragm and Lung DevelopmentAckerman Kate G 12*Herron Bruce J 3Vargas Sara O 4Huang Hailu 2Tevosian Sergei G 5Kochilas Lazaros 6Rao Cherie 2Pober Barbara R 7Babiuk Randal P 8Epstein Jonathan A 9Greer John J 8Beier David R 21 Division of Emergency Medicine, Department of Medicine, Children's Hospital and Harvard Medical School, Boston, Massachusetts, United States of America
2 Divison of Genetics, Brigham and Women's Hospital and Harvard Medical School, Boston, Massachusetts, United States of America
3 Genomics Institute, Wadsworth Center, Troy, New York, United States of America
4 Department of Pathology, Children's Hospital and Harvard Medical School, Boston, Massachusetts, United States of America
5 Department of Genetics, Dartmouth Medical School, Hanover, New Hampshire, United States of America
6 Department of Pediatrics, Brown Medical School, Providence, Rhode Island, United States of America
7 Department of Surgery, Children's Hospital and Harvard Medical School, Boston, Massachusetts, United States of America
8 Department of Physiology, University of Alberta, Edmonton, Alberta, Canada
9 Cardiovascular Division, Department of Medicine and Department of Cell and Molecular Biology, University of Pennsylvania Medical School, Philadelphia, Pennsylvania, United States of America
Frankel Wayne EditorJackson Laboratory, United States of America* To whom correspondence should be addressed. E-mail: [email protected] 2005 17 6 2005 1 1 e101 4 2005 20 5 2005 Copyright: © 2005 Ackerman et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Congenital diaphragmatic hernia and other congenital diaphragmatic defects are associated with significant mortality and morbidity in neonates; however, the molecular basis of these developmental anomalies is unknown. In an analysis of E18.5 embryos derived from mice treated with N-ethyl-N-nitrosourea, we identified a mutation that causes pulmonary hypoplasia and abnormal diaphragmatic development. Fog2 (Zfpm2) maps within the recombinant interval carrying the N-ethyl-N-nitrosourea-induced mutation, and DNA sequencing of Fog2 identified a mutation in a splice donor site that generates an abnormal transcript encoding a truncated protein. Human autopsy cases with diaphragmatic defect and pulmonary hypoplasia were evaluated for mutations in FOG2. Sequence analysis revealed a de novo mutation resulting in a premature stop codon in a child who died on the first day of life secondary to severe bilateral pulmonary hypoplasia and an abnormally muscularized diaphragm. Using a phenotype-driven approach, we have established that Fog2 is required for normal diaphragm and lung development, a role that has not been previously appreciated. FOG2 is the first gene implicated in the pathogenesis of nonsyndromic human congenital diaphragmatic defects, and its necessity for pulmonary development validates the hypothesis that neonates with congenital diaphragmatic hernia may also have primary pulmonary developmental abnormalities.
Synopsis
Birth defects involving the diaphragm are as common and as serious as genetic disorders such as cystic fibrosis, yet the underlying causes of these defects are unknown. Most babies born with diaphragmatic defects have very small lungs, and many die in the newborn period with severe breathing difficulties. It is unknown whether the small lungs occur because these children have a diaphragmatic defect, or whether some patients might have a genetic abnormality that affects the development of both the lung and the diaphragm simultaneously.
In a screen of fetal mice carrying chemically induced genetic mutations, the authors found that a mutation in the gene Fog2
(Friend of gata 2), causes abnormal diaphragm development and small lungs. The lungs have a primary developmental abnormality that includes the specific loss of one lung lobe. Based on this result, the authors studied children affected with diaphragmatic abnormalities, and identified one human baby with a serious mutation in the human gene FOG2 who died at five hours of life with severe breathing difficulties, a diaphragmatic defect, and small lungs.
The authors have established that Fog2 is necessary for both diaphragm and lung development in mice and in humans. This is the first known cause of a nonsyndromic congenital diaphragmatic defect and establishes that some patients may have a primary developmental abnormality of both the lung and the diaphragm.
Citation:Ackerman KG, Herron BJ, Vargas SO, Huang H, Tevosian SG, et al. (2005) Fog2 is required for normal diaphragm and lung development in mice and humans. PLoS Genet 1(1): e10.
==== Body
Introduction
Congenital diaphragmatic defects are a spectrum of relatively common birth defects. The Bochdalek or posterolateral hernias (often referred to as congenital diaphragmatic hernia [CDH]) occur in 1/3,000 live births [1], and although these are the most common type of diaphragmatic defect presenting at birth, diaphragmatic aplasia and diaphragmatic muscularization defects (eventrations) may have a similar clinical presentation.
Making specific anatomic distinctions among these types of defects can be difficult without direct gross (intraoperative or postmortem) examination. Pulmonary hypoplasia associated with these diaphragmatic defects causes severe mortality and morbidity. The pathogenesis and developmental relationship between diaphragmatic defects and pulmonary hypoplasia is not understood. Although advances in the medical management of pulmonary hypoplasia may have decreased the mortality associated with CDH patients who survive to receive care at high-volume centers [2,3], the population-based mortality has been reported to be as great as 62%, and there are a large number of deaths prior to birth or to transfer to a tertiary care facility [4]. As these patients commonly present with severe respiratory failure at birth, therapy has been centered around developing better methods to provide ventilatory support while not producing further lung injury. Extracorporeal membrane oxygenation (ECMO) is used in some centers to provide an extended period of cardiopulmonary bypass [5,6], while other centers have success using other ventilatory support techniques [7]. The morbidity in those who survive is high, and many patients survive with chronic respiratory insufficiency, cognitive and neuromotor deficits, and hearing loss as a result of necessary intensive interventions and associated structural and irreversible developmental abnormalities [8–11].
To date, there have been no specific mutations found to be associated with nonsyndromic diaphragmatic defects and pulmonary hypoplasia in humans. The heritability of these defects is unclear, as the high morbidity and mortality limit the collection of multigenerational families for analysis. The genetic etiologies are likely to be complex and probably arise from different mutations in various parts of the molecular developmental pathways required for diaphragmatic development. Indeed, there are numerous reports implicating different chromosomal abnormalities in the pathogenesis of CDH [12,13]. Given the difficulty of studying lethal developmental abnormalities in humans, it is of great potential utility to develop animal models of human birth defects, as the specific genetic abnormalities found in animal models can then be investigated in human populations.
We screened mice treated with the chemical mutagen N-ethyl-N-nitrosourea (ENU) for lines with developmental defects that present in the perinatal period [14]. From this screen, we identified a line of mice carrying a recessive mutation that results in primary pulmonary hypoplasia and abnormal diaphragmatic and cardiac development. Positional cloning analysis identified Fog2 (Zfpm2) as a likely candidate, and DNA sequencing revealed a mutation in a splice donor site that generates an abnormal transcript encoding a truncated protein. This result suggested that we examine the orthologous gene in humans with similar developmental defects, and we report the finding of a de novo nonsense mutation in FOG2 in a patient who died at birth with a diaphragmatic defect and severe pulmonary hypoplasia. This is the first reported mutation associated with these abnormalities in a human. We present additional data that provide direct evidence that pulmonary hypoplasia may be a primary component of this spectrum of disorders.
Results
Identification of the little lung Mutation in Fog2
We screened third-generation progeny of ENU-mutagenized mice at embryonic day 18.5 (E18.5) for abnormal developmental phenotypes [14]. One line of mice was found to have multiple embryos in independent litters that displayed pulmonary hypoplasia and a thin diaphragm. The mutation, which we called little lung
(lil), was mapped to Chromosome 15 by utilizing a strategy of interval haplotype analysis (data not shown) [15]. For high-resolution mapping, F2 progeny from two crosses were analyzed. In 450 progeny of an intercross of F1 (A/J × FVB/N) lil/+ mice, the interval containing the mutation was defined by 19 recombinants between D15Mit220 and D15Mit154. Because of the lack of informative markers within this interval, an additional 39 F2 progeny from an A/J × C57BL/6J cross were tested. The identification of two recombinants established D15Mit85 as the proximal and D15Mit5 as the distal flanking markers.
The lil phenotype was identified at E18.5 to have bilateral pulmonary hypoplasia and an abnormal diaphragm. Pulmonary hypoplasia was apparent in all mutant mice that survived to birth. In a comparison between wild-type and mutant mice found dead on day one of life, body weights were not different; however, the lung weights were significantly lower in the mutant mice. The average mutant lung weight was 9.6 ± 2.5 mg while the average wild-type lung weight was 26.4 ± 4.6 mg (p < 0.001). All lungs from mutant mice were lacking an accessory lobe on the right side and had underdevelopment of the anterior right middle lobe (Figure 1A). Diaphragms from mutant lil mice were intact, but muscularization was absent in the dorsal regions. Myotubules were present in a limited and abnormal distribution. More specifically, myotubules normally radiate in a mediolateral fashion to meet the lateral body walls, with a normal paucity of muscle fibers in the central tendonous region. In the mutant diaphragm, myotubules radiated in a dorsal–ventral orientation, and muscle tissue did not meet the entire surface of the body walls (Figure 1B).
Figure 1 Abnormal Pulmonary and Diaphragmatic Development in the lil Mouse
(A) The mutant hypoplastic lung (right) lacks the development of the accessory lobe and the anterior portion of the right middle lobe (marked with arrows on the control sample on the left).
(B) Whole diaphragms show a lack of normal muscularization in the posterolateral regions and the peripheral regions of the mutant diaphragm (control on left and lil diaphragm on the right).
The number of lil mutant mice that survived to birth was less than 5% of total progeny, rather than the 25% expected for a recessive mutation. We evaluated litters at different embryonic time points to determine whether the reduced number was due to intra-uterine demise. Litters were collected at E12.5, 13.5, 15.5, 17.5, and 18.5, and embryos were genotyped and evaluated for evidence of intra-uterine demise, including growth retardation, pallor, and tissue friability. lil embryos had a progressively higher rate of demise between E13.5 and E15.5. At E12.5, 20 out of 87 (23%) were homozygous for the mutation, and all of these appeared viable. At E13.5, 13 of 49 (27%) were mutant and two had died. At E15.5, 22 out of 91 (24%) were mutant and the majority of mutant embryos (17 out of the 22) had died. By E18.5, nine of 29 embryos (31%) were homozygous for the mutation, of which only one embryo was viable. Diaphragmatic muscularization was abnormal in all mutant mice examined (n > 25). Pulmonary hypoplasia was observed in 100% of mutants evaluated for that phenotype between E11.5 and birth (n > 50).
Examination of cardiac morphology showed that hearts from E15.5 lil mutant mice had a variety of developmental defects, including enlarged and abnormally developed endocardial cushions, a double-outlet right ventricle, and a complete atrioventricular canal. The myocardium was also poorly developed, with thinning of the outer compact layer and decreased vascularity. The cohort of mutant mice that survived to birth also had cardiac malformations including atrioventricular-canal-type ventricular septal defects, ostium primum atrial septal defects, and enlarged atria (data not shown). All mutants specifically evaluated for a cardiac phenotype (n = 10) had abnormal cardiac development.
Examination of the 3-Mb region between D15Mit85 and D15Mit5 in DNA sequence databases revealed three predicted genes and four known genes, including Fog2. Targeted mutations of Fog2 have cardiac defects strikingly similar to those we identified in lil mutant mice, including atrioventricular canal defects, thinned myocardium, and absent coronary vasculature [16]. RT-PCR amplification of the proximal portion of Fog2 revealed longer transcripts in the lil mutant than in the wild-type embryos (Figure 2A). Sequencing of the mutant transcript revealed a point mutation (from thymine to cytosine) 2 bp after position 301, which is in the splice donor site at the end of the third exon. This mutation causes a splicing defect that results in the insertion of 85 bp of intronic sequence into the mutant transcript, and introduces a stop codon that generates a severely truncated protein product (Figure 2B). Heterozygous lil mutant mice were crossed with a Fog2
+/− (null allele) mutant generated by gene targeting [16]. Doubly heterozygous mice had an embryonic lethal phenotype; this failure to complement proves that lil is a mutation in Fog2. The variable phenotype of lil mice (relative to that found for the Fog2 null mutant) is likely due to the generation of a low level of normal transcript despite the presence of the splice site mutation.
Figure 2 The ENU-Induced lil Mutation in Fog2 is a Splice Site Mutation
(A) RT-PCR revealed a lengthened transcript in mice with the lil phenotype: the first band is a lil mouse, and the second band is a control mouse.
(B) Sequencing Fog2 revealed a splice site mutation that causes the insertion of 85 bp of intronic sequence in the mutant mouse. This results in a premature stop codon prior to zinc finger transcription.
Lung and Diaphragm Development in the Fog2 Mutant Mouse
Experiments were conducted to evaluate the role of Fog2 in pulmonary and diaphragmatic development. The pulmonary phenotype is characterized by diffuse hypoplasia and specific loss of the accessory lobe and a portion of the right middle lobe.
It is well established that abnormalities in either diaphragmatic development or fetal breathing can result in a secondary pulmonary hypoplasia, although loss of normal structure has never been documented in this context [17,18]. Fog2 is expressed diffusely in the pulmonary mesenchyme during the period of branching morphogenesis, while later expression is restricted to the smooth muscles of airways and pulmonary vessels (Figure 3). This, and the observation that lungs appeared small on transverse sections evaluated prior to diaphragmatic muscularization or function, suggests that the pulmonary hypoplasia occurs independently of the diaphragmatic defect. To test this hypothesis, lungs were dissected from Fog2
−/− mice and littermate controls before the onset of fetal diaphragmatic motion. Fog2
−/− mice were used for this experiment, as we wanted to avoid potential phenotypic variance from the lil hypomorphic mutation. Lungs dissected at E12 from Fog2
−/− embryos were smaller in size and lacked the development of an accessory lobe. In 11 viable Fog2−
/− lung culture explants, there was never development of an accessory lobe, and the weights of mutant lungs cultured for 5 d were significantly lower than those of littermate controls (Figure 4). These data demonstrate that the pulmonary hypoplasia in Fog2 mutant mice is a primary defect.
Figure 3
Fog2 Is Expressed in the Developing Lung and Diaphragm
Fog2 is expressed in the diffuse pulmonary mesenchyme at E13.5 (A) (arrow shows mesenchyme) and is restricted to the bronchial and vascular smooth muscle (sm) at E16.5 (B). Fog2 is expressed diffusely in the developing diaphragm (Dia) both prior to (E11.5) (C) and after muscularization (E13.5) (D).
Figure 4
Fog2 Is Necessary for Primary Lung Development
Fog2 null lungs removed prior to diaphragmatic muscularization and grown in vitro show no accessory lobe development. Accessory lobe is labeled with black arrow in control littermate lungs.
Branching in the unaffected lobes appeared to be delayed by 6–12 h, but all mutants developed an intricate branching pattern in the unaffected lobes after culture for 5 d.
Because the Fog2 phenotype is striking for specific lobar loss, the spatial pattern of Fog2 expression was evaluated in normal embryos during the period of early lobar establishment to determine whether Fog2 expression is specifically different at these lobar buds. Expression was evaluated in mice carrying a lacZ gene incorporated into the Fog2 locus (S. Tevosian, unpublished data). In nine mice examined at E11.5, all lungs showed a specific enhancement of Fog2 expression in the mesenchyme surrounding the accessory bud and the right middle lobe bud, which are the lobes that do not develop normally in Fog2 mutant mice (Figure 5). By E12.5, expression was diffuse in the pulmonary mesenchyme, as was seen previously with in situ hybridization on tissue sections.
Figure 5
Fog2 Expression in Embryonic Non-Mutant Lungs
In embryonic non-mutant lungs, Fog2 is most highly expressed at the tips of the accessory (single arrows) and right middle lobes (double arrows) at E11.25 (A) and E11.75 (B), while expression is diffuse by E12.75 (C).
Diaphragms from Fog2 lil mice show an intact membrane with a defect in muscular patterning (see Figure 1B). The membranous portion of the diaphragm is populated by a migratory population of muscle precursor cells, much like the limbs [19,20]. Mice with defects in genes known to be important for the control of this process have intact but amuscular diaphragms [17,21,22]. Hepatocyte growth factor/Scatter factor
(HGF) is one potential candidate responsible for the guidance of muscle precursors to the membranous diaphragm. It has been shown that HGF is expressed along this anatomic pathway [23], and mice with absence of the HGF receptor c-Met fail to form muscularized pleuroperitoneal folds (PPFs), and thus have amuscular diaphragms [24,25]. Fog2 is expressed diffusely in the early amuscular diaphragm at E11.5 as well as in the later muscularized diaphragm (see Figure 3C and 3D). Pax3 and MyoD, transcription factors required for appropriate migration and determination of myogenic precursors, were detected in the PPFs of Fog2 lil mice (data not shown). However, in situ expression analysis demonstrated that the expression of HGF in the region where this structure meets the membranous diaphragm was markedly reduced in Fog2 mutant mice (Figure 6). We hypothesize that Fog2 is required (either directly or indirectly) for the induction of HGF in the developing diaphragm, and dysregulation of HGF patterning along the path of muscle precursor cell migration between the PPF and the diaphragm accounts for the abnormal phenotype in these mice.
Figure 6
HGF Patterning Is Abnormal in Fog2 Null Mice
In situ hybridization of HGF in E12.5 wild-type (A) and Fog2
−/− (B) embryos demonstrates decreased expression in the region where the PPF meets the membranous diaphragm. Li, liver; Lu, lung.
FOG2 Mutation in a Patient with Diaphragm and Lung Abnormalities
FOG2 sequence analysis was performed on autopsy tissue from 30 of 32 deceased children with an anatomic diagnosis of diaphragmatic defect evaluated at the Children's Hospital in Boston, Massachusetts, between 1993 and 2003. Autopsy reports were reviewed to determine the specific diagnoses. Of these 30 cases, 17 (57%) had Bochdalek CDH, two (7%) had diaphragmatic agenesis, seven (23%) had diaphragmatic eventration/muscularization defects (without Bochdalek CDH), and four (13%) had Bochdalek hernia of one hemidiaphragm and eventration of the other. Pulmonary hypoplasia was assessed using lung/body weight ratio and radial alveolar counts [26]. The material available for review included written reports and histologic slides in all cases and gross kodachromes in a subset of cases.
One child carried a highly significant FOG2 sequence change. The patient was a full-term 3,500-g baby girl who developed severe respiratory failure at birth and died after 5 h of resuscitation. Antemortem radiographs showed opacified lung fields and possible bowel in the chest. The patient's clinical diagnosis was CDH.
Review of autopsy material revealed severe bilateral pulmonary hypoplasia (combined lung weight, 11.1 g; expected for body length/weight, 46.8 ± 26.2 g; [27]), most markedly involving the left lung. The lung/body weight ratio was 0.0037 (expected > 0.010) [28]. There were a reduced number of bronchial generations, and the radial alveolar count was 2–3 (expected = 5) [29]. There were incomplete lung fissures bilaterally. A deep posterior diaphragmatic eventration was present on the left side. Additionally, two muscularized ligamentous bands resembling diaphragmatic remnants encircled the left lobe of the liver, creating an abnormal fissured liver contour. Away from the eventration, the diaphragm appeared well muscularized, measuring 0.3 cm in thickness. A complete autopsy revealed no other malformations; the heart was determined to be grossly normal and was donated for valve harvest.
Sequence analysis revealed a cytosine to thymine heterozygous change in exon 4 that changes the 112th amino acid from arginine to a stop codon. This mutation produces a severely truncated peptide that does not contain zinc finger domains (Figure 7). This base change was not present in the analysis of DNA from 400 normal adults. To assess the likelihood that the mutation was causal for the developmental phenotype, we examined both parents. Paternity was confirmed, and sequence analysis revealed that the neither parent carried this mutation, proving that the patient had a de novo mutation in FOG2 (Figure 7).
Figure 7
FOG2 Mutation in a Patient with Diaphragm and Lung Abnormalities
Sequencing revealed a de novo heterozygote nonsense mutation in a patient who died at birth with severe pulmonary hypoplasia and a posterior deep diaphragmatic eventration. She was clinically diagnosed with CDH. This nonsense mutation occurs prior to the functional zinc finger domains.
Discussion
Congenital diaphragmatic defects are a heterogeneous group of disorders of unknown etiology. The defects that present in the pre- or perinatal period include Bochdalek hernia, diaphragmatic aplasia, and various degrees of muscularization defects or eventrations. Different types of defects occur in the same patients or in siblings, suggesting these represent variable expression of the same underlying pathogenesis [30,31]. Clinical differentiation between these defects may be very difficult, as the residual membranous diaphragm of a muscularization defect is thin and may not be easily visible on prenatal ultrasound or postnatal chest radiographs [32]. Although diaphragmatic muscularization defects were historically considered to be predictive of a good outcome, there have been inadequate population-based studies that include fetal or neonatal cases and autopsy diagnoses to make this conclusion definitive. In fact, the series of patients we report here and the published literature indicate that an eventration defect may be associated with displacement of abdominal contents and also with severe pulmonary hypoplasia and respiratory insufficiency [33,34].
Numerous chromosome abnormalities have been found in association with congenital diaphragm abnormalities [12,35]. Human FOG2 maps to Chromosome 8q23.1, and, importantly, several patients with diaphragm defects and rearrangements involving this locus have been reported. Specifically, there are three unrelated CDH patients with cytogenetically balanced translocations at or near the FOG2 locus [36,37]. Additionally, two patients with deletions apparently encompassing the FOG2 locus have died from multiple congenital anomalies including CDH [38–40]. Inactivation of this gene due to chromosomal rearrangement or deletion would result in a heterozygous null mutation similar to that found in the patient we report.
Because the FOG2 mutation we report is de novo and the phenotypes of the pulmonary and diaphragmatic defects are similar between mouse and human, we suggest that this mutation in FOG2 is the first reported cause of a human developmental diaphragmatic and pulmonary defect. In contrast to the affected child, mice heterozygous for a null mutation of Fog2 appear normal. However, there is ample precedent for the observation that haploinsufficiency of a gene with developmental functions is much less well tolerated in humans than mice [41].
It is unclear how the Fog2 diaphragmatic defect relates to the more common Bochdalek CDH, as the pathogenic mechanisms for both are largely unknown. Muscle precursors destined to populate the diaphragm migrate from the lateral dermomyotome of cervical somites. Prior to migration onto the diaphragm, they populate the PPF, a wedge-shaped tissue that tapers medially from the lateral body wall to the esophageal mesentery and fuses ventrally with the septum transversum [42]. Muscle precursors reach the PPF by E11, where they proliferate, differentiate, and then migrate toward the dorsolateral costal, sternal costal, and crural regions of the developing diaphragm. Thus, a defect in PPF formation subsequently results in the abnormal formation of the diaphragm [43]. We have shown that the Fog2 mutant does have an abnormal pattern of HGF expression in the region through which muscle precursor cells migrate onto the developing diaphragm. This finding may account for the abnormally patterned muscle that develops in the Fog2 mutant diaphragm. Although Pax3 and MyoD expression is detected in the PPF, a detailed analysis of transcription factors responsible for muscle precursor cell migration and differentiation will need to be completed both in the PPF and along the pathway of muscle precursor cell migration between the PPF and the membranous diaphragm. Fog2 can interact with any of the Gata factors, Gatas 1–6, as well as other transcription factors such as CoupTFII [44,45]. It is known that a Fog2–Gata4 interaction is critical for normal cardiac and gonadal development, but interacting factors in the lung and diaphragm have not yet been determined.
The severity of pulmonary hypoplasia in the patient we report was out of proportion to that of the diaphragm defect. Pulmonary hypoplasia is associated with abnormal diaphragmatic anatomy or function, and is known to occur as a secondary developmental defect in models of diaphragmatic dysfunction such as complete amuscularization [17] or phrenic nerve disruption [46]. It occurs in a surgical model of CDH in which a hernia is physically created in an in utero lamb [47,48]. However, the possibility that primary pulmonary developmental abnormalities occur with, rather than secondary to, diaphragmatic defects has been suggested by others based on a teratogenic model of CDH [49–51] and has long been suspected by clinicians who care for these patients. In addition, the high incidence of lobar abnormalities associated with CDH [52] supports the possibility that this disorder can be associated with a primary developmental pulmonary abnormality.
Our analysis of mice carrying mutations of Fog2 proves that there is a primary defect in lung development that results in specific loss of the accessory lobe and partial loss of the right middle lobe. The specific lobar defects prompted us to evaluate Fog2 expression at the time of early lobar budding. While Fog2 expression is diffuse in the pulmonary mesenchyme after lobar structure is well established (E12.5), it is more focally expressed in the mesenchyme surrounding the right middle lobe and accessory buds as these lobes form. This matches the phenotype of right middle lobe and accessory lobe loss, and suggests that Fog2 has a specific patterning role in establishment of these lobes. It is less clear whether loss of Fog2 results in a global branching defect, as Fog2 lungs appear to have a slight developmental delay, which could result from many causes. Cultured Fog2 lungs do develop an intricate branching pattern in the unaffected lobes that appears similar in the pattern to wild-type lungs after 5 d in culture.
In this report, we show that a mutation of Fog2 in the mouse causes the phenotype of abnormal diaphragmatic muscularization and primary pulmonary hypoplasia. We furthermore demonstrate that a mutation in this gene is associated with a lethal defect in lung and diaphragm development in a child. It is notable that, despite extensive analysis of Fog2 biology and the generation of a Fog2 knock-out mouse, its role in diaphragm and lung development was previously not recognized. It is only as a consequence of phenotype-driven analyses such as those we are pursuing that one has the opportunity to assay all of the potential molecular derangements that may result in human disease.
Materials and Methods
These investigations were conducted with approval of the institutional review board for Children's Hospital, Boston, and Brigham and Women's Hospital, Boston. Animal use was approved by the Center for Animal Resources and Comparative Medicine (Harvard Medical School).
Genetic mapping of the mouse mutation lil.
The lil mutation was identified as described in results. Wild-type FVB/N and C57BL/6J mice used for genetic crosses were obtained from the Jackson Laboratory (Bar Harbor, Maine, United States). Mice carrying a null mutation of Fog2 generated by gene targeting [16] were the generous gift of Dr. Stuart Orkin.
Developmental analysis of mice.
Timed pregnancies were set up for collection of E11.5–E17.5 embryos. Embryos were fixed, dehydrated, and embedded in paraffin prior to sectioning. In older embryos, a median sternotomy was performed under microscopic guidance, and diaphragm, lungs, and heart were examined. The lungs and tracheobronchial tree were removed and weighed. Whole diaphragms were isolated from fixed thoracic tissue from E15.5 and E17.5 embryos. For lung explant culture, lungs were dissected from fresh embryos at E11.5 and E12.5 and placed on porous 24-mm (0.4-μ) polyester membranes floated in wells containing 2 ml of Dulbecco's modified Eagle's medium, nutrient mixture F-12 (11039–021, Gibco, San Diego, California, United States), supplemented with 10% fetal bovine serum, 0.3 mg/ml L-glutamine, 100 units/ml penicillin, 100 mcg/ml streptomycin, and 0.25 mcg/ml amphotericin B. Lung explants were cultured at 37 °C in 95% air/5% CO2 for up to 5 d. They were photographed daily with a dissecting microscope (MZ12.5, Leica, Wetzlar, Germany) equipped with a Leica DC500 digital camera.
Transgenic mice carrying the lacZ gene driven by the Fog2 promoter have been developed by S. Tevosian. In these animals, the lacZ gene is incorporated (“knocked-in”) into the Fog2 locus to allow β-galactosidase expression as a fusion protein in frame with the first 235 amino acids of the FOG2 protein. The Fog2-lacZ module is followed by an ires-eGFP cassette. This creates a null allele of Fog2 gene. The Fog2-LacZ-eGFP construct was linearized with KspI and electroporated into the CJ7 ES cells. The correctly targeted clone was selected by the Southern blot analysis and injected into C57BL/6J blastocysts. Fog2-lacZ-eGFP animals were maintained on the mixed C57BL/6J/129 background. lacZ Expression in whole dissected embryonic lungs was analyzed by staining for β-galactosidase activity with X-gal after fixation for 30 min.
RT-PCR and sequence analysis in the mouse.
RNA was extracted by standard techniques from thoracic embryonic tissue. RT-PCR was performed using six primer sets designed to cover the Fog2 gene. RT-PCR was repeated with radiolabeled primers to amplify an abnormally spliced region of the gene (Table S1), and the product was run on a denaturing sequencing gel according to standard techniques. The RT-PCR product was cloned into pCR2.1 vector using TOPO TA Cloning Kit (Invitrogen, Carlsbad, California, United States) and sequenced using gene-specific primers. Sequence analysis was done using Sequencher 4.1 (Gene Codes, Ann Arbor, Michigan, United States).
In situ hybridization.
After dehydration and embedding in paraffin wax, 10-μ sections were subjected to radioactive in situ hybridization as described [53]. Probes labeled with 35S were prepared by run-off transcription of linearized plasmid templates and hybridized to tissue sections. Nuclei were counterstained with Hoescht 33258, and signal was imaged using fluorescent and darkfield microscopy.
Human DNA extraction and sequence analysis.
DNA was isolated from paraffin blocks by phenol-chloroform extraction [54,55], and from frozen tissues by standard techniques. Primers were designed to amplify FOG2 coding exons plus 50 bp of flanking upstream and downstream sequence. PCR amplification and sequencing were performed by standard methods. Primer sequences used are listed in Table S2. Sequence analysis was done with Sequencher 4.1 (Gene Codes).
DNA from the parents of one autopsy patient was extracted from fresh blood samples. A second set of blood samples was sent to an outside CLIA-certified laboratory for DNA extraction, PCR, sequencing, and analysis. Paternity testing was performed by the outside laboratory using a standard panel of markers. SNP genotyping was done using Harvard Partners Center for Genetics and Genomics genotyping core facility (Cambridge, Massachusetts, United States).
Supporting Information
Table S1 Primers for RT-PCR (Mouse): Amplification of Abnormal Transcript in Fog2 Mutant (lil) Mice
(26 KB DOC)
Click here for additional data file.
Table S2 Human Primers for Amplification of FOG2 Coding Sequence from Genomic DNA
(46 KB DOC)
Click here for additional data file.
This manuscript is dedicated to Baby Lucy and her parents for the contribution that they have made to helping us to understand developmental diaphragmatic and lung defects. This work was funded by National Institute of Child Health and Human Development (NIHCD) grant HD36404 (DRB) and the Hearst Foundation (KGA). KGA also received salary support from NICHD grant T32HD040129–02. We would like to thank Stuart Orkin, M. D., for the generous gift of his Fog2
−/− mice. We would also like to thank Raju Kucherlapati, Ph. D., and Birgit Funke, Ph. D., for providing control DNA samples.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. KGA, JJG, and DRB conceived and designed the experiments. KGA, BJH, HH, SGT, LK, CR, RPB, and JAE performed the experiments. KGA, BJH, HH, LK, CR, JAE, JJG, and DRB analyzed the data. KGA, SOV, SGT, BRP, JAE, JJG, and DRB contributed reagents/materials/analysis tools. KGA and DRB wrote the paper.
Abbreviations
CDHcongenital diaphragmatic hernia
E[number]embryonic day [number]
ENUN-ethyl-N-nitrosourea
HGF
Hepatocyte growth factor/Scatter factor
lil
little lung
PPFpleuroperitoneal fold
==== Refs
References
Langham MR Jr Kays DW Ledbetter DJ Frentzen B Sanford LL 1996 Congenital diaphragmatic hernia. Epidemiology and outcome Clin Perinatol 23 671 688 8982563
Downard CD Jaksic T Garza JJ Dzakovic A Nemes L 2003 Analysis of an improved survival rate for congenital diaphragmatic hernia J Pediatr Surg 38 729 732 12720181
Javid PJ Jaksic T Skarsgard ED Lee S 2004 Survival rate in congenital diaphragmatic hernia: The experience of the Canadian Neonatal Network J Pediatr Surg 39 657 660 15136994
Stege G Fenton A Jaffray B 2003 Nihilism in the 1990s: The true mortality of congenital diaphragmatic hernia Pediatrics 112 532 535 12949279
Harrington KP Goldman AP 2005 The role of extracorporeal membrane oxygenation in congenital diaphragmatic hernia Semin Pediatr Surg 14 72 76 15770591
Bouchard S Johnson MP Flake AW Howell LJ Myers LB 2002 The EXIT procedure: Experience and outcome in 31 cases J Pediatr Surg 37 418 426 11877660
Cacciari A Ruggeri G Mordenti M Ceccarelli PL Baccarini E 2001 High-frequency oscillatory ventilation versus conventional mechanical ventilation in congenital diaphragmatic hernia Eur J Pediatr Surg 11 3 7 11370980
Stefanutti G Filippone M Tommasoni N Midrio P Zucchetta P 2004 Cardiopulmonary anatomy and function in long-term survivors of mild to moderate congenital diaphragmatic hernia J Pediatr Surg 39 526 531 15065021
Ahmad A Gangitano E Odell RM Doran R Durand M 1999 Survival, intracranial lesions, and neurodevelopmental outcome in infants with congenital diaphragmatic hernia treated with extracorporeal membrane oxygenation J Perinatol 19 436 440 10685274
Rasheed A Tindall S Cueny DL Klein MD Delaney-Black V 2001 Neurodevelopmental outcome after congenital diaphragmatic hernia: Extracorporeal membrane oxygenation before and after surgery J Pediatr Surg 36 539 544 11283873
Trachsel D Selvadurai H Bohn D Langer JC Coates AL 2005 Long-term pulmonary morbidity in survivors of congenital diaphragmatic hernia Pediatr Pulmonol 39 433 439 15704183
Lurie IW 2003 Where to look for the genes related to diaphragmatic hernia? Genet Couns 14 75 93 12725592
Enns GM Cox VA Goldstein RB Gibbs DL Harrison MR 1998 Congenital diaphragmatic defects and associated syndromes, malformations, and chromosome anomalies: A retrospective study of 60 patients and literature review Am J Med Genet 79 215 225 9788565
Herron BJ Lu W Rao C Liu S Peters H 2002 Efficient generation and mapping of recessive developmental mutations using ENU mutagenesis Nat Genet 30 185 189 11818962
Neuhaus IM Beier DR 1998 Efficient localization of mutations by interval haplotype analysis Mamm Genome 9 150 154 9457677
Tevosian SG Deconinck AE Tanaka M Schinke M Litovsky SH 2000 FOG-2, a cofactor for GATA transcription factors, is essential for heart morphogenesis and development of coronary vessels from epicardium Cell 101 729 739 10892744
Tseng BS Cavin ST Booth FW Olson EN Marin MC 2000 Pulmonary hypoplasia in the myogenin null mouse embryo Am J Respir Cell Mol Biol 22 304 315 10696067
Harding R 1997 Fetal pulmonary development: The role of respiratory movements Equine Vet J Suppl 24 32 39
Birchmeier C Brohmann H 2000 Genes that control the development of migrating muscle precursor cells Curr Opin Cell Biol 12 725 730 11063939
Greer JJ Allan DW Martin-Caraballo M Lemke RP 1999 An overview of phrenic nerve and diaphragm muscle development in the perinatal rat J Appl Physiol 86 779 786 10066685
Babiuk RP Greer JJ 2002 Diaphragm defects occur in a CDH hernia model independently of myogenesis and lung formation Am J Physiol Lung Cell Mol Physiol 283 L1310 L1314 12388344
Li J Liu KC Jin F Lu MM Epstein JA 1999 Transgenic rescue of congenital heart disease and spina bifida in Splotch mice Development 126 2495 2503 10226008
Dietrich S Abou-Rebyeh F Brohmann H Bladt F Sonnenberg-Riethmacher E 1999 The role of SF/HGF and c-Met in the development of skeletal muscle Development 126 1621 1629 10079225
Bladt F Riethmacher D Isenmann S Aguzzi A Birchmeier C 1995 Essential role for the c-met receptor in the migration of myogenic precursor cells into the limb bud Nature 376 768 771 7651534
Yang XM Vogan K Gross P Park M 1996 Expression of the met receptor tyrosine kinase in muscle progenitor cells in somites and limbs is absent in Splotch mice Development 122 2163 2171 8681797
Emery JL Mithal A 1960 The number of alveoli in the terminal respiratory unit of man during late intrauterine life and childhood Arch Dis Child 35 544 547 13726619
Wigglesworth JS Singer DB 1998 Textbook of fetal and perinatal pathology, 2nd ed Malden (Massachusetts) Blackwell Science 1205 p.
Page DV Stocker JT 1982 Anomalies associated with pulmonary hypoplasia Am Rev Respir Dis 125 216 221 7065526
Cooney TP Thurlbeck WM 1982 The radial alveolar count method of Emery and Mithal: A reappraisal 2—Intrauterine and early postnatal lung growth Thorax 37 580 583 7179186
Thomas MP Stern LM Morris LL 1976 Bilateral congenital diaphragmatic defects in two siblings J Pediatr Surg 11 465 467 957074
Rodgers BM Hawks P 1986 Bilateral congenital eventration of the diaphragms: Successful surgical management J Pediatr Surg 21 858 864 3783371
Yang JI 2003 Left diaphragmatic eventration diagnosed as congenital diaphragmatic hernia by prenatal sonography J Clin Ultrasound 31 214 217 12692831
Rais-Bahrami K Gilbert JC Hartman GE Chandra RS Short BL 1996 Right diaphragmatic eventration simulating a congenital diaphragmatic hernia Am J Perinatol 13 241 243 8724727
Elberg JJ Brok KE Pedersen SA Kock KE 1989 Congenital bilateral eventration of the diaphragm in a pair of male twins J Pediatr Surg 24 1140 1141 2809986
Tibboel D Gaag AV 1996 Etiologic and genetic factors in congenital diaphragmatic hernia Clin Perinatol 23 689 699 8982564
Temple IK Barber JC James RS Burge D 1994 Diaphragmatic herniae and translocations involving 8q22 in two patients J Med Genet 31 735 737 7815446
Howe DT Kilby MD Sirry H Barker GM Roberts E 1996 Structural chromosome anomalies in congenital diaphragmatic hernia Prenat Diagn 16 1003 1009 8953633
Cappellini A Sala E Colombo D Villa N Mariani S 1996 Monosomy 8q and features of Fryns' syndrome [abstract] Eur J Hum Genet 4 29
Wilson WG Wyandt HE Shah H 1983 Interstitial deletion of 8q. Occurrence in a patient with multiple exostoses and unusual facies Am J Dis Child 137 444 448 6601906
Harnsberger J Carey JC Morgan A 1982 Interstitial deletion of the long arm of the 8 chromosome and the Langer-Gidion syndrome [abstract]. Proceedings of the 1982 Birth Defects Conference; 1982 June; Birmingham, Alabama
Roessler E Belloni E Gaudenz K Jay P Berta P 1996 Mutations in the human Sonic Hedgehog gene cause holoprosencephaly Nat Genet 14 357 360 8896572
Babiuk RP Zhang W Clugston R Allan DW Greer JJ 2003 Embryological origins and development of the rat diaphragm J Comp Neurol 455 477 487 12508321
Greer JJ Cote D Allan DW Zhang W Babiuk RP 2000 Structure of the primordial diaphragm and defects associated with nitrofen-induced CDH J Appl Physiol 89 2123 2129 11090558
Cantor AB Orkin SH 2005 Coregulation of GATA factors by the Friend of GATA (FOG) family of multitype zinc finger proteins Semin Cell Dev Biol 16 117 128 15659346
Huggins GS Bacani CJ Boltax J Aikawa R Leiden JM 2001 Friend of GATA 2 physically interacts with chicken ovalbumin upstream promoter-TF2 (COUP-TF2) and COUP-TF3 and represses COUP-TF2-dependent activation of the atrial natriuretic factor promoter J Biol Chem 276 28029 28036 11382775
Fewell JE Lee CC Kitterman JA 1981 Effects of phrenic nerve section on the respiratory system of fetal lambs J Appl Physiol 51 293 297 6894915
de Lorimier AA Tierney DF Parker HR 1967 Hypoplastic lungs in fetal lambs with surgically produced congenital diaphragmatic hernia Surgery 62 12 17
Wilcox DT Irish MS Holm BA Glick PL 1996 Animal models in congenital diaphragmatic hernia Clin Perinatol 23 813 822 8982573
Keijzer R Liu J Deimling J Tibboel D Post M 2000 Dual-hit hypothesis explains pulmonary hypoplasia in the nitrofen model of congenital diaphragmatic hernia Am J Pathol 156 1299 1306 10751355
Guilbert TW Gebb SA Shannon JM 2000 Lung hypoplasia in the nitrofen model of congenital diaphragmatic hernia occurs early in development Am J Physiol Lung Cell Mol Physiol 279 L1159 L1171 11076806
Jesudason EC Connell MG Fernig DG Lloyd DA Losty PD 2000 Early lung malformations in congenital diaphragmatic hernia J Pediatr Surg 35 124 127 10646789
Nose K Kamata S Sawai T Tazuke Y Usui N 2000 Airway anomalies in patients with congenital diaphragmatic hernia J Pediatr Surg 35 1562 1565 11083423
Wawersik S Epstein JA 2000 Gene expression analysis by in situ hybridization. Radioactive probes Methods Mol Biol 137 87 96 10948528
Mies C 1994 Molecular biological analysis of paraffin-embedded tissues Hum Pathol 25 555 560 7516909
Fox EA 2000 Preparation of DNA from fixed, paraffin-embedded tissue. In: Dracopoli NC, Haines JL, Korf BR, Morton CC, Seidman CE, et al., editors. Current protocols in human genetics. New York: John Wiley and Sons. pp. A.3I.1–A.3I.5
|
16103912
|
PMC1183529
|
CC BY
|
2021-01-05 08:00:24
|
no
|
PLoS Genet. 2005 Jul 17; 1(1):e10
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010010
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 1741133210.1371/journal.pgen.001001405-PLGE-I-0130plge-01-01-13InterviewThe Whole Side of It—An Interview with Neil Risch InterviewsGitschier Jane Jane Gitschier is in the Department of Medicine and Pediatrics and the Howard Hughes Medical Institute, University of California, San Francisco, California, United States of America. E-mail: [email protected]
7 2005 25 7 2005 1 1 e14Copyright: © 2005 Jane Gitschier.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Citation:Gitschier J (2005) The whole side of it—An interview with Neil Risch. PLoS Genet 1(1): e14.
==== Body
Neil Risch is the Lamond Distinguished Professor in Human Genetics and Director of the Center for Human Genetics at the University of California, San Francisco, California, United States. He has held faculty appointments at Columbia, Yale, and Stanford Universities.
Trying to track down Neil Risch is the stuff of legend. E-mails to him can bounce back with the comment line “overwhelmed by email.” Phone calls lead to voicemail, and faxes to a tepid response from his assistant. It's not that he's reclusive or off windsurfing, he's simply a whirlwind of genetics ideas and activity.
Indeed, when the University of California at San Francisco was looking for a director of the new Center for Human Genetics, Neil's name quickly came to the top of the heap. With wide-ranging experience and interests, he was described by one of the field's founding fathers as “the statistical geneticist of our time.” It didn't hurt that he is a mensch.
I managed to trap Neil in his bright new office on the ninth floor of the west tower off Parnassus Avenue. Still spartan, with only a computer, a phone, and a chair, the office's view spanned Golden Gate Park, the Marin headlands, and the Pacific Ocean. The vista was interrupted only by the jarring copper-clad tower of the new museum under construction in Golden Gate Park. It was a brilliant blue, warm afternoon, and I looked forward to spending some one-on-one time with this man, with his infectious laugh and his intellectual stamina.
Jane Gitschier: Let's start with the broad view. What really interests you?
Neil Risch: My passion, really, is the interplay between population genetics and clinical applications—to see the whole side of it.
When I was in graduate school, I came out of math. Three weeks into my first course in human genetics [as part of a new biomathematics graduate program at the University of California at Los Angeles], I knew that it was what I wanted to do. One, I loved the subject matter. Two, I loved the quantitative aspect of it. Three, I loved the intuitive aspect of it. It was almost like I could predict the next lecture. It was a perfect fit for me. My other passion was population genetics, but there weren't many career opportunities in that field back then. So always there was this latent passion for population genetics without the opportunity to act on it.
Neil Risch
The thing that has been exciting for me is that I saw ten years ago where the field was going. Having the sequence of the human genome provides the opportunity to look at the variation in that sequence as well, which leads to the marriage of several areas, but particularly human population genetics with disease studies—genetic epidemiology.
But the fields have never been so intimately related as they are now, and I am just thrilled. I get to marry the two things I love to do. In the old days, the NIH [National Institutes of Health] would never fund a study in population genetics, but now it does because you need to understand human population history and genetics to undertake all these studies of human genetic variation underlying disease susceptibility.
Gitschier: So many people want to collaborate with you. How do you choose what projects to become involved with? Did you initiate most of the projects you work on, or do people come to you?
Risch: In my Stern Award address [presented to the American Society of Human Genetics in 2004], I talked about the population genetics analogy of selection and drift. Some things are planned and have a natural scheme to them, and some things are just random. Some things I have become interested in and I have initiated, and other things have come to me and I've participated in them either because it was appropriate, given the setting where I was working, or I was interested in the project and wanted to have a collegial relationship.
There has been quite a range of the projects I have gotten involved in, especially on the clinical side. I think it's been good for me that I haven't focused on one particular area, like cancer or psychiatry. Some people could say, “He's like a dilettante,” but I think it has given me a broad perspective on the clinical application and the commonalities in terms of the issues involved across diseases, and some of the unique aspects. I value the fact that I've been able to be involved in a lot of different things. I feel incredibly fortunate to have become established in a field so that I've had a lot of opportunities in terms of different people approaching me for collaborations. It's been great.
Gitschier: You are an Ashkenazi Jew and belong to a conservative temple here in San Francisco. You also are well known for your work on diseases, such as torsion dystonia, that affect this particular population. Do you think you gravitate to these problems because of your ethnicity?
Risch: Early on, I was interested in population genetics and I knew about this debate about the presence of the lysosomal storage diseases in the Jews. Why did the Jews have all these diseases?
Gitschier: I just assumed it was selective advantage. What was the debate?
Risch: Between selection and drift—as it always is.
This actually started with the dystonia work. At the time, there was a raging question about the mode of inheritance. They knew it [severe early-onset idiopathic torsion dystonia] was more prevalent in Ashkenazi Jews. Some people thought it was dominant, but there was a major paper that said it was recessive just like all other major Ashkenazi diseases.
And I was interested. It was a nice statistical problem, and it was in the Jewish population, which I was interested in both for scientific and historical reasons, and because of my own identity.
So we did a study. Susan Bressman, my colleague at Columbia, systematically went out and clinically examined all the first- and second-degree relatives of all the early-onset Jewish cases—parents, siblings, children, nieces and nephews, half-sibs, uncles and aunts, maybe grandparents. We analyzed the data, and the rates of dystonia in the relatives were the same in all the first-degree relatives; the siblings did not have a higher risk. About 15% for everybody, across the board. We concluded that it was autosomal dominant with low penetrance [30%].
We did a formal segregation analysis. We could clearly, overwhelmingly reject a recessive model. And at the time I thought, “Well, this is fascinating!” because we had a dominant disease with a founder effect in the Jewish population. And we even suggested that this would be valuable for gene mapping—by linkage disequilibrium [LD] analysis. And after mapping the disease, we found strong LD right away. So this was very clear evidence that we were dealing with a relatively recent founder mutation.
I know this is a long story, but it was through that project that I got interested in Jewish population genetics.
Gitschier: It seems as though every time I open the science section of the New York Times, you are featured in it. These articles, at least lately, focus on your adherence to an often politically incorrect idea, such as the genetic basis for race or the way NIH should spend its money on diseases of addiction. Do you deliberately choose controversy?
Risch: I think historically I have avoided it. Perhaps this is what job security offers you—the opportunity to get involved in potentially more controversial questions. And I think I've decided that playing it safe is not the way to go. I just don't believe that anymore. These are big important subjects and I just don't think they should be avoided.
Gitschier: Let's talk about the former, the genetic basis of race. As you know, I went to a session for the press at the ASHG [American Society for Human Genetics] meeting in Toronto, and the first words out of the mouth of the first speaker were “Genome variation research does not support the existence of human races.”
Risch: What is your definition of races? If you define it a certain way, maybe that's a valid statement. There is obviously still disagreement.
Gitschier: But how can there still be disagreement?
Risch: Scientists always disagree! A lot of the problem is terminology. I'm not even sure what race means, people use it in many different ways.
In our own studies, to avoid coming up with our own definition of race, we tend to use the definition others have employed, for example, the US census definition of race. There is also the concept of the major geographical structuring that exists in human populations—continental divisions—which has led to genetic differentiation. But if you expect absolute precision in any of these definitions, you can undermine any definitional system. Any category you come up with is going to be imperfect, but that doesn't preclude you from using it or the fact that it has utility.
We talk about the prejudicial aspect of this. If you demand that kind of accuracy, then one could make the same arguments about sex and age!
You'll like this. In a recent study, when we looked at the correlation between genetic structure [based on microsatellite markers] versus self-description, we found 99.9% concordance between the two. We actually had a higher discordance rate between self-reported sex and markers on the X chromosome! So you could argue that sex is also a problematic category. And there are differences between sex and gender; self-identification may not be correlated with biology perfectly. And there is sexism. And you can talk about age the same way. A person's chronological age does not correspond perfectly with his biological age for a variety of reasons, both inherited and non-inherited. Perhaps just using someone's actual birth year is not a very good way of measuring age. Does that mean we should throw it out? No. Also, there is ageism—prejudice related to age in our society. A lot of these arguments, which have a political or social aspect to them, can be made about all categories, not just the race/ethnicity one.
Gitschier: I have heard you say, “Don't politicize the human genome.”
Risch: I have a strong problem with the way politicians use this information. [Former President] Clinton, for example, when the first draft of the human genome sequence came out, made a statement about how all people in the world, in terms of their genetic makeup, are 99.9% the same. His intent—to reduce conflict among peoples—is noble. People on the left, anthropologists and sociologists, do the same thing. They use the 99.9% figure as an argument for social equality. But the truth is that people do differ by that remaining 0.1% and that people do cluster according to their ancestry. The problem is that others could use that information to create division.
Gitschier: Do you ever feel that the press misrepresents you?
Risch: Don't we all feel that to some extent? They always take the simple side, which leads to misinterpretation. So there are risks in talking to the press, especially on controversial subjects.
Gitschier: You have a brother who is an academic, and at one point you and he were both on the faculty at Yale. Is that genetic? Tell us about the environment that you grew up in that ultimately led to producing two academicians. I don't know what your brother's field is.
Risch: So interesting. You tell me, is this “nature” or “nurture”?
My brother, my only sibling [Harvey Risch], is 20 months older than I. Mathematics was his skill set also. He went to Cal Tech, starting a year before me—both of us math majors. Then he decided to go to medical school, so he did additional coursework, and we graduated simultaneously. He went to UCSD to medical school, and had to do a thesis—so he came to UCLA in my department and lived with me. Then he applied to do a PhD in biomathematics in my department [but went to the University of Chicago instead]. His first paper and my first paper appeared in the same issue of Annals of Human Genetics, and we didn't even know it. Then Harvey decided to do epidemiology as a post-doc, while I was learning epidemiology at Columbia. Our grades, SAT scores, GRE scores—everything pretty much the same.
Gitschier: Sounds like the premise for a simple quantitative analysis. Do you think you are both hardwired to do mathematical problems? Or did your family just sit around at dinner doing math problems?
Risch: Not at all. My mom [Sonia Risch] was very artistic and intuitive. Great artist, writer, actress—and brilliant at all of it. My father was a clinical psychiatrist. But if you look at my family history, on my mother's side there are a lot of MDs and on my father's side there is a lot of math. So if you want to make a genetic hypothesis here, my brother is the confluence of both, maybe, and maybe me, too. My brother and I would talk about stuff a lot. My mother would say, “Oh, they're talking Fortran.”
“My passion, really, is the interplay between population genetics and clinical applications—to see the whole side of it.”
Gitschier: You have been working with the epidemiologists at Kaiser Permanente [a health-care provider] in Oakland for the past seven years. I understand you spend every Wednesday there. How did this collaboration come about, and what is it you are trying to accomplish? Other than numbers, is there something the Kaiser resource can do that say, Iceland, couldn't?
Risch: Can it ever!
Gitschier: Why don't you tell me what's so cool about it?
Risch: When I went to Stanford, at the back of my mind was this issue about Kaiser. This comes from my epidemiology background. This is the advantage of being multidisciplinary. If I could push anything, it's the value of seeing the links between various disciplines and marrying them. One, Kaiser's membership is a cohort—you don't have to construct it from scratch. And it's followed over time, for many people over 20 years. Two, they have computerized databases where every contact a patient has with the health-care system, every inpatient, outpatient, and pharmacy visit, every visit with doctors outside the Kaiser system gets recorded, every X ray, every lab test, basic biochemistries—all computerized. Three, it is the health-care provider for one-third of the Bay Area and a very good representation of Bay Area population. It is missing only the very high and very low end of the socioeconomic ladder. All major ethnic groups are represented. To me, it's the most wonderful laboratory for doing population genetic and genetic epidemiologic research.
As you can tell, I'm a very strong believer of inclusion of a variety of people of varying racial/ethnic backgrounds in research. There is everything to be gained from doing so—not just politically, but scientifically. Also, ethically it's the right thing to do. And I'm concerned in this whole discussion that people may be scared off from the genetics research and that's the battle.
One big issue that I think will go a long way towards addressing this problem is to do everything we can to recruit more minority scientists to human genetics. For the research to have credibility in minority communities, there must be representation from those communities among scientists. And I want to be involved in that process also.
Gitschier: In Iceland, you've got all the ancestry data, and so you can do traditional linkage analysis, but in Kaiser, you're going to do association studies.
Risch: That's right. Because of the way the technology is moving, this is a tremendous resource for doing that.
Gitschier: This is a very long-term study. There must be an element of “Oh, I see this gelling” and it just can't go fast enough.
Risch: You're right, it's been a long process. There are complexities and financial issues.
This is really the best opportunity we have in the United States to do something along these lines, and it's been a little frustrating that it's been difficult to get the support and funding for it, but I'm patient. Because when you believe in something and know it's right, you have the patience to see it though.
|
17411332
|
PMC1183530
|
CC BY
|
2021-01-05 07:59:36
|
no
|
PLoS Genet. 2005 Jul 25; 1(1):e14
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010014
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 10.1371/journal.pgen.001001705-PLGE-RA-0032R3plge-01-01-10Research ArticleSystems BiologyGenetics/Genome ProjectsEukaryotesCaenorhabditisNew Genes Tied to Endocrine, Metabolic, and Dietary Regulation of Lifespan from a Caenorhabditis elegans Genomic RNAi Screen RNAi Aging Screen in C. elegansHansen Malene Hsu Ao-Lin ¤aDillin Andrew ¤bKenyon Cynthia *Department of Biochemistry and Biophysics, University of California, San Francisco, California, United States of AmericaKim Stuart EditorStanford University School of Medicine, United States of America*To whom correspondence should be addressed. E-mail: [email protected]¤a Current address: Department of Internal Medicine and Geriatrics Center, University of Michigan, Ann Arbor, Michigan, United States of America
¤b Current address: Molecular and Cell Biology Laboratory, The Salk Institute, La Jolla, California, United States of America
7 2005 25 7 2005 1 1 e1724 2 2005 9 6 2005 Copyright: © 2005 Hansen et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Most of our knowledge about the regulation of aging comes from mutants originally isolated for other phenotypes. To ask whether our current view of aging has been affected by selection bias, and to deepen our understanding of known longevity pathways, we screened a genomic Caenorhabditis elegans RNAi library for clones that extend lifespan. We identified 23 new longevity genes affecting signal transduction, the stress response, gene expression, and metabolism and assigned these genes to specific longevity pathways. Our most important findings are (i) that dietary restriction extends C. elegans' lifespan by down-regulating expression of key genes, including a gene required for methylation of many macromolecules, (ii) that integrin signaling is likely to play a general, evolutionarily conserved role in lifespan regulation, and (iii) that specific lipophilic hormones may influence lifespan in a DAF-16/FOXO-dependent fashion. Surprisingly, of the new genes that have conserved sequence domains, only one could not be associated with a known longevity pathway. Thus, our current view of the genetics of aging has probably not been distorted substantially by selection bias.
Synopsis
Lifespan in C. elegans is influenced by several genetic pathways and processes; a great deal of the information about this regulation of aging comes from genetic mutants originally identified because of other phenotypes. Therefore, to ask whether the current view of the genetics of aging has been significantly affected by selection bias, and to deepen the understanding of known longevity pathways, Hansen et al. screened a genome-wide RNAi library for bacterial clones that extend lifespan when fed to the nematode Caenorhabditis elegans.
The investigators identified 23 new longevity genes affecting signal transduction, the stress response, gene expression, and metabolism and assigned these genes to specific longevity pathways. Their most important findings were (i) that dietary restriction extended C. elegans' lifespan by down-regulating expression of key genes, including a gene required for methylation of many macromolecules, (ii) that integrin signaling is likely to play a general, evolutionarily conserved role in lifespan regulation, and (iii) that specific lipophilic hormones may influence lifespan through the conserved insulin/IGF-1 signaling pathway.
Surprisingly, the authors found that of the new genes that have conserved sequence domains, only one could not be associated with a known longevity pathway. Thus, the current view of the genetics of aging has probably not been distorted substantially by selection bias. The authors expect the further study of these genes to provide valuable information about the mechanisms of aging, not only in C. elegans but also in higher organisms.
Citation:Hansen M, Hsu AL, Dillin A, Kenyon C (2005) New genes tied to endocrine, metabolic, and dietary regulation of lifespan from a Caenorhabditis elegans genomic RNAi screen. PLoS Genet 1(1): e17.
==== Body
Introduction
A number of mutations and environmental conditions can extend the lifespan of Caenorhabditis elegans. Although the underlying mechanisms are not fully understood, these perturbations appear to affect at least three distinct regulatory systems. The first is the insulin/IGF-1/FOXO system [1,56]. Inhibiting insulin/IGF-1 signaling can double the lifespan of the animal [54]. This lifespan extension requires the FOXO transcription factor DAF-16 [54], which, in turn, influences the expression of a diverse set of downstream antioxidant, metabolic, chaperone, antimicrobial, and novel genes that act in a cumulative way to influence lifespan [2,3]. The process of autophagy, which is increased in these long-lived mutants, may also make an important contribution [4]. In addition to DAF-16, the heat-shock transcription factor HSF-1 [5–7] as well as AMP kinase [8] are required for the lifespan extension of insulin/IGF-1 pathway mutants. The insulin/IGF-1/FOXO pathway acts exclusively during adulthood to influence aging [9] and appears to be regulated by sensory cues [10,11]. The reproductive system also affects insulin/IGF-1 signaling [12]. Killing the germline increases lifespan, and this lifespan increase requires the presence of the somatic gonad [12]. Germline ablation is thought to extend lifespan by activating a steroid hormone pathway and DAF-16/FOXO, and the somatic gonad appears to exert a counterbalancing influence on lifespan by affecting the activity of the insulin/IGF-1 pathway [12,13]. Finally, other regulators, such as the sir-2 deacetylase and Jun kinase, can also increase lifespan in a daf-16-dependent fashion [14–17].
Dietary restriction (DR) (i.e., caloric restriction) also extends the lifespan of C. elegans. How this occurs is not clear. This lifespan extension does not require DAF-16 activity, suggesting that it is distinct from the insulin/IGF-1 signaling pathway [18,19] and the sir-2 histone deacetylase [14]. So far, only one gene, the ubiquinone biosynthetic gene clk-1, has been implicated in the response to DR in C. elegans [18].
Third, inhibition of mitochondrial respiration or ATP synthesis increases lifespan [20–22]. Like the lifespan extension produced by DR, this lifespan extension is daf-16-independent. However, DR extends lifespan when it is initiated during adulthood, whereas respiratory-chain inhibition does not [21]. Thus, these two types of perturbations may act in different ways to increase lifespan.
In an effort to learn whether additional pathways might influence aging in C. elegans, and to identify additional genes in known pathways, we screened for enhanced longevity using a genomic RNA interference (RNAi) bacterial-feeding library that covers ~87% of the C. elegans open reading frames (total of 16,757) [23,24]. In our screen, we did not expect to identify all of the genes whose normal functions shorten lifespan. In previous Chromosome I RNAi screens, both the Ruvkun lab and our lab identified a number of mitochondrial RNAi clones that increased lifespan, but we identified only one gene in common [21,22] (legend in Table 1). Likewise, the Plasterk group found a high level of false negatives in each of their two RNAi screens [25]. In addition, we would not expect to find functionally redundant genes, genes with essential roles in other biological processes, or neuronal genes (since neurons are refractory to RNAi [24,26,27]). Nevertheless, since most biological pathways involve many genes, we hoped to identify genes that function in most or all of the pathways that influence lifespan.
Table 1 Mean Lifespan Extensions Observed in First Retests of fer-15(b26); fem-1(hc17) and of N2 Worms Grown on Long-Lived Candidate RNAi Clones
RNAi clones identified in the screen were first retested on fer-15; fem-1, and when found to increase lifespan significantly (p < 0.02), were then tested on N2. Genes in top section of table produced daf-16-dependent lifespan extension when inhibited. Genes in bottom section of table were specifically annotated with a mitochondrial function. N2 lifespan data on cyc-1, cco-1, nuo-2, and atp-3 were published previously [21]. The genes cco-1 and cchl-1 were previously identified in the Ruvkun lab's Chromosome I/II screen [22]. (All seven of the mitochondrial clones and the clone for the metabolic gene F57B10.3 that were identified in the Ruvkun lab's Chromosome I/II but not our screen extended mean lifespan when we assayed them directly; though, interestingly, most did not extend maximum lifespan notably.)
fer-15; fem-1 but not N2 animals showed significant lifespan extensions when grown on library clones for inx-8, zig-6, gei-9, and ril-3/F26F2.1 (see Table S1). The lifespan trials were generally carried out at 20 °C, with some exceptions in which the experiments were performed at 25 °C (see Table S1 for details). fer-15; fem-1 mean lifespan when grown on control RNAi bacteria was 19.1 ± 1.3 d (total number of experiments, n = 10), N2 mean lifespan on control RNAi was 20.0 ± 1.8 d (n = 5), both at 20 °C. Protein ID refers to primary accession numbers from UniProt.
p-values were calculated as pair-wise comparisons relative to control (no RNAi insert) using the Log-rank (Mantel-Cox) method. Please note that each RNAi clone was tested on N2 one or more times, and also on fer-15; fem-1 one or more times and found to extend lifespan each time (this table shows one such trial, see Table S1 for complete dataset).
a Lifespan extensions of animals grown on RNAi clone compared to control vector-only bacteria (no RNAi insert).
b Genes were named in this study.
Chr, chromosome position of gene; ddl, daf-16-dependent longevity; drr, dietary restriction response; ril, RNAi-induced longevity.
Results/Discussion
RNAi Screen for Clones That Increase Longevity
To identify longevity genes, we cultured animals on RNAi bacteria from the time of hatching and then looked for plates containing live individuals at a time when age-matched controls were all dead (see Materials and Methods). We estimate that we screened 70% of the open reading frames of C. elegans (Figure S1). Of the 94 candidate clones retested, 29 caused a highly significant lifespan extension, ranging from ~10–90%. (Five of these clones were published previously in Chromosome I/II screens [21,22], Table 1). These clones were tested a third time, and found to produce similar lifespan extensions (Tables 1 and S1).
Surprisingly, we recovered only one known longevity gene, the insulin/IGF-1-like receptor gene daf-2. To better understand this finding, we tested the RNAi clones of several known longevity genes (Table S2). Of the ten known genes we considered, only five were represented in our RNAi library. RNAi clones for three of these genes, daf-2,
age-1, and akt-1 (but not clk-1 or glp-1), extended lifespan significantly (Table S2), indicating that we had a significant level of false negatives in our screen. (False negatives can arise for many reasons, including small sample size at the time of scoring, or censoring because of plate contamination or progeny production; see Materials and Methods. In addition, clones that extend only mean lifespan but not maximal lifespan would qualify as false negatives in our screen.) Together, these findings suggest that although we identified many new longevity genes in our screen, more remain to be found.
Some of the RNAi clones we recovered had relatively small effects on lifespan. Unexpectedly, one of these was the daf-2 library clone (Figure S2), which repeatedly lengthened lifespan ~20% (see Tables 1 and S1). In contrast, a daf-2 RNAi clone we constructed using the same vector but a different daf-2 insert doubled lifespan (data not shown, Figure S2). Thus for daf-2, one could say that the RNAi library contained a “weak allele.” Likewise, for the other genes we analyzed, it is possible that longer lifespans could be produced by further reductions in gene activity.
The 23 new genes affected a wide variety of processes, including signal transduction, gene expression/nucleic-acid metabolism, the stress response, glucose and amino-acid metabolism, mitochondrial function, and regulation of vesicle trafficking (Table 1). We assigned these genes names based on our functional analysis and DNA sequence information (Table 1).
New Components of the Insulin/IGF-1 and Reproductive Pathways
The DAF-16/FOXO transcription factor is required for mutations in the insulin/IGF-1 pathway, or germline ablation, to extend lifespan [1]. We found that seven of the new RNAi clones (ttr-1/K03H1.6, maoc-1/E04F6.3, gpi-1/Y87G2A.8, sinh-1/Y57A10A.20, ddl-1/F59E12.10, ddl-2/Y48E1B.1, and ddl-3/Y54G11A.8), as well as the daf-2/Y55D5A.5 clone, failed, in multiple trials, to extend the lifespan of daf-16(mu86) null mutants (Table 2). Such a failure to extend lifespan could conceivably be due to sickness, but the animals appeared healthy. (This was the case for all the other animals we examined, unless stated otherwise.) Therefore, the simplest explanation is that the activity of daf-16 is specifically required to extend the lifespan of animals treated with these RNAi clones. In principle, these “daf-16-dependent” genes could function in the insulin/IGF-1 pathway, the reproductive pathway, both pathways, or other DAF-16-dependent pathways.
Table 2 Genetic Epistasis Analysis of RNAi Clones Whose Effects Require DAF-16
Mean lifespan extensions observed in the first lifespan performed with each mutant. p-values were calculated as pair-wise comparisons relative to the control (no RNAi insert) of that experiment using the Log-rank (Mantel-Cox) method. Significant lifespan extensions (p < 0.02) are in bold. Lifespan experiments in which no extension was observed were repeated at least once (except for daf-2); each lifespan trial in this table represents one of the repeats (see Table S1 for complete dataset). The mutants we tested were daf-16(mu86),
daf-2(e1370), glp-1(e2141), and daf-12(rh61rh411), and lifespan analysis was generally carried out at 20 °C, with some exceptions, in which the experiments were performed at 25 °C (see Table S1 for details).
a Lifespan extension of animals grown on RNAi clone compared to control bacteria (no RNAi insert).
New genes in the insulin/IGF-1 pathway are particularly interesting because their counterparts in humans could potentially play a role not only in longevity regulation but also in insulin/IGF-1-related metabolic diseases such as diabetes and cancer. To investigate whether the genes identified by these RNAi clones might be part of the insulin/IGF-1 pathway, we carried out two experiments. First, while not a conclusive test (since daf-2 mutants are not null), we were curious to know whether the RNAi clones might be unable to further extend the lifespan of daf-2(e1370) mutants, as is the case with mutations in the downstream gene age-1/PI 3-kinase [28]. We found that this was the case for all but one of these clones (see below; Table 2). In contrast, all but one of the daf-16-independent clones that we examined increased the lifespan of e1370 mutants (see below; Table 3). We also asked whether these RNAi clones might affect a different process regulated by the insulin/IGF-1 pathway: dauer formation. Wild-type juveniles enter dauer, a pre-pubescent growth-arrested state, in response to food limitation. Strong daf-2 mutants become dauers even in the presence of food, in a daf-16-dependent fashion. To assay dauer formation, we asked whether our daf-16-dependent RNAi clones enhanced the weak dauer-constitutive (Daf-c) phenotype of daf-2(e1370) mutants cultured at 22.5 °C. We found that four daf-16-dependent RNAi clones (ttr-1,
gpi-1,
sinh-1, and ddl-2) enhanced dauer formation to a significant extent, whereas two clones (maoc-1 and ddl-1) did not (Figure 1). (daf-2(e1370) worms grown on ddl-3 RNAi produced almost no progeny. Therefore, this daf-16-dependent RNAi clone was not assayed.) Because ttr-1,
gpi-1,
sinh-1, and ddl-2 RNAi clones all enhanced dauer formation and their function was daf-16-dependent, these genes are likely to be part of the daf-2 pathway.
Table 3 Genetic Epistasis Analysis of RNAi Clones Whose Effects Do Not Require DAF-16
Mean lifespan extensions observed in the first experiments carried out with each mutant strain. p-values were calculated as pair-wise comparisons relative to the vector-only control (no RNAi insert) using the Log-rank (Mantel-Cox) method. Significant lifespan extensions (p < 0.02) are in bold. RNAi of atp-4 gave a slightly higher p-value, which was still considered significant.
Lifespan experiments in which no extension was observed were repeated at least once (except for daf-2; see Table S1 for complete dataset). Lifespan data on cyc-1, cco-1, nuo-2, and atp-3 were published previously [21]. The genes cchl-1 and cco-1 were previously identified in the Ruvkun lab's Chromosome I/II screen [22]. The mutants we tested were daf-16(mu86),
daf-2(e1370), and eat-2(ad1116), and lifespan analysis was generally carried out at 20 °C, with some exceptions in which the experiments were performed at 25 °C (see Table S1 for details).
a Lifespan extension of animals grown on RNAi clone compared to control bacteria (no RNAi insert).
b
eat-2 animals grown on either pat-4 or pat-6 RNAi appeared unhealthy (see Table S1 and text).
Figure 1 The Dauer-Constitutive Phenotype of daf-2(e1370) Is Enhanced by Many RNAi Clones That Extend Lifespan in a daf-16-Dependent Fashion
Relative dauer formation of daf-2(e1370) animals grown at 22.5 °C on RNAi clones versus vector control is shown, average of two to three experiments. 30–50% of the animals on vector control become dauers at 22.5 °C. Total number of dauers/total number of animals observed is noted on top of bars. Error bars: ± SEM. ‘*', previously characterized RNAi clones [9] served as negative (daf-16, RNAi insert consists of first 1.2 kb cDNA) and positive (daf-2, RNAi insert consists of first 2.2 kb cDNA; see also Figure S2) controls for the dauer experiment. daf-2(e1370) worms grown on ddl-3 RNAi gave rise to almost no progeny; therefore, this daf-16-dependent RNAi clone was not assayed.
To test whether our daf-16-dependent RNAi clones were part of the germline pathway, we asked whether they could further extend the lifespans of glp-1(e2141ts) animals (which lack a germline when raised at high temperature [29]), and also mutants defective in the putative steroid receptor daf-12 [12]. The germline pathway is distinct from the daf-2 pathway in that it does not appear to affect dauer formation (see data in Materials and Methods). Consistent with this, our daf-2 clone further extended the lifespan of glp-1 mutants (see Table 2). As described below, five of our daf-16-dependent RNAi clones (ttr-1, sinh-1, ddl-1, ddl-2, and ddl-3) failed to extend the lifespan of glp-1 mutants. In principle, these clones could simply prevent the development of the germline. However, this seems unlikely, because none of these RNAi clones reduced brood size (data not shown). Instead, we favor the interpretation that germline ablation extends lifespan, at least in part, by inhibiting the activities of these five genes. These and other daf-16-dependent genes will now be described in more detail.
One of the most interesting daf-16-dependent clones contains a transthyretin-family domain and the corresponding protein is thus predicted to be a member of the transthyretin-related protein family (ttr-1, for transthyretin-related protein). In vertebrates, transthyretin is one of three specific carrier proteins involved in the transport of both thyroid hormones and retinol. These carrier proteins might also play a role in regulating the uptake of the free circulating hormones in various tissues [30,31]. C. elegans TTR-1 is a member of a distinct but related protein family with a completely conserved predicted hydrophobic pocket [32], suggesting that it could function as a carrier of a lipophilic hormone(s) that influences lifespan. We found that ttr-1 RNAi enhanced dauer formation (Figure 1) and failed to further extend the lifespan of daf-2(e1370) mutants (see Table 2), suggesting that it (and the putative lipophilic hormone[s] it binds) acts in the insulin/IGF-1 pathway either upstream of or in parallel to daf-16.
The ttr-1 RNAi clone did not further extend the lifespan of glp-1 mutants (see Table 2), suggesting that it also functions in the germline pathway to influence lifespan. The lifespan extension of germline-ablated animals requires the putative steroid hormone receptor DAF-12 [12], and we found that ttr-1 was unable to extend the lifespan of DAF-12 mutants. Thus, one plausible model is that TTR-1 binds to and limits the availability of a ligand that activates DAF-12.
Another daf-16-dependent RNAi clone, maoc-1, encodes a protein containing a MaoC-like dehydratase domain, which is also found in enzymes such as human type IV estradiol 17-β-dehydrogenase and in the fatty acid synthase β subunit. Therefore, maoc-1 could act in the synthesis of a lipophilic hormone that influences lifespan in a daf-16-dependent fashion. This clone did not further extend the lifespan of daf-2(e1370) mutants (see Table 2), suggesting that the gene acts in the insulin/IGF-1 pathway; however, it failed to enhance the Daf-c phenotype (Figure 1). Furthermore, the clone enhanced the longevity of glp-1 (germline-defective) mutants (see Table 2), arguing that it does not extend lifespan by affecting germline signaling. The clone also increased the lifespan of daf-12-null mutants, so it is unlikely to synthesize a DAF-12 ligand. Since this clone acts differently from ttr-1 to affect lifespan, these findings raise the possibility that multiple lipophilic hormones influence lifespan in a daf-16-dependent fashion.
The third daf-16-dependent gene, gpi-1, encodes a glucose-6-phosphate isomerase/neuroleukin homolog. This RNAi clone enhanced dauer formation (Figure 1) and extended the lifespan of glp-1 and daf-12 mutants (see Table 2). Thus, it probably functions in the insulin/IGF-1 pathway, but not in the reproductive pathway to affect lifespan. In mammals, glucose-6-phosphate isomerase functions in glycolysis and also as a secreted neuronal growth factor. We favor the possibility that glycolysis influences aging, because inhibition of another glycolysis gene, phosphoglycerate mutase, also extends lifespan [22]. One possibility is that this RNAi clone extends lifespan by reducing ATP levels, which in turn inhibits the release of insulin-like DAF-2 ligands—analogous to the effect of ATP on insulin release in mammals [33].
The fourth daf-16-dependent gene, sinh-1, encodes a homolog of Schizosaccharomyces pombe
SIN1, stress-activated MAP kinase interacting protein 1 [34], which functions in the response to DNA damage. In S.
pombe, SIN1 inactivation increases sensitivity to some environmental stresses, such as high temperature and osmotic stress, but not to oxidative stress [34]. Unexpectedly, we found that sinh-1 RNAi significantly increased both thermotolerance (mean lifespan of wild-type worms at 35 °C: control 16.0 h; sinh-1(RNAi) 20.6 h; p < 0.0001) and the resistance to oxidative stress (mean lifespan of wild-type worms treated with paraquat (free-radical generator): control 4.4 h; sinh-1(RNAi) 7.1 h; p < 0.0001). Thus, in C. elegans (and perhaps other multicellular organisms as well), this gene appears to prevent, rather than promote, stress resistance.
sinh-1 RNAi enhanced dauer formation (Figure 1) and failed to further extend the longevity of daf-2(e1370) mutants (see Table 2), suggesting that it acts in the insulin/IGF-1 pathway. sinh-1 RNAi also failed to extend the longevity of glp-1 mutants, suggesting that sinh-1 functions in the germline pathway (see Table 2). Interestingly, sinh-1 RNAi did extend the lifespan of daf-12/steroid-receptor mutants (see Table 2). The regulatory relationship between DAF-16 and DAF-12 is not known. Thus, one possible explanation is that in response to germline ablation, DAF-12 activates DAF-16/FOXO by inhibiting the activity of SINH-1. Alternatively, germline cells might regulate the activities of DAF-16/FOXO and DAF-12 in parallel, and sinh-1 might act only in the pathway that regulates DAF-16/FOXO.
The protein encoded by the fifth daf-16-dependent RNAi clone, ddl-1 (for daf-16-dependent longevity), has been reported to interact with heat-shock factor binding protein (HSB-1) [35], a negative regulator of C. elegans heat-shock transcription factor (HSF-1) activity [36]. In C. elegans, HSF-1 promotes longevity [5–7] and is required along with daf-16/FOXO for the increased lifespan of daf-2 receptor mutants [6,7]. Thus, wild-type ddl-1 might inhibit longevity by reducing the activity of HSF-1. ddl-1 RNAi did not further extend the lifespan of daf-2(e1370) mutants (see Table 2), consistent with a role in the insulin/IGF-1 pathway, but it did not enhance dauer formation (Figure 1).
HSF-1 is also required for the longevity of animals lacking a germline (hsf-1(RNAi) mean lifespan, 10.4 d at 20 °C, n=53; glp-1(e2141); hsf-1(RNAi) mean lifespan 11.0 d, n=60; p = 0.25). ddl-1 RNAi did not extend the lifespan of glp-1 mutants, suggesting that it, too, functions in the germline pathway. Interestingly, like sinh-1 RNAi, ddl-1 RNAi was able to extend the lifespans of daf-12 mutants (see Table 2). This finding suggests that ddl-1 and sinh-1, which was also implicated in the stress response, act in the same process to affect longevity.
The sixth daf-16-dependent clone, ddl-2, encodes a protein that has been reported to interact with DDL-1 [35]. This RNAi clone enhanced dauer formation (Figure 1) and failed to increase the longevity of daf-2(e1370) mutants (see Table 2), consistent with a role in the insulin/IGF-1 pathway. Like ddl-1 RNAi, ddl-2 RNAi failed to increase the lifespan of germline-deficient animals (see Table 2), suggesting that it, too, functions in the germline pathway. However, unlike ddl-1 RNAi, in several experiments, this clone failed to extend the lifespan of daf-12 mutants (see Table 2). Assuming that DDL-1 and DDL-2 do interact, this suggests that they function in a complex fashion to influence lifespan.
The final daf-16-dependent clone, (ddl-3), is predicted to have a tetratrico-peptide-repeat (TPR) protein interaction motif. This clone failed to increase the longevity of daf-2 mutants (see Table 2), suggesting that ddl-3 acts in the insulin/IGF-1 pathway. It also failed to further extend the lifespan of glp-1/germline-defective and daf-12/steroid-receptor mutants (see Table 2), suggesting that DDL-3 also acts in the germline pathway, possibly as an adaptor protein in a signaling cascade that regulates DAF-16 and DAF-12.
Mitochondrial RNAi Clones
Given the large number and large effects of Chromosome I/II respiratory-chain RNAi clones [21,22], we were not surprised to find that 12 of our RNAi clones encoded components of the mitochondrial respiratory chain. (Of these, four Chromosome I clones [nuo-2/T10E9.7, cyc-1/C54G4.8, cco-1/F26E4.9, and atp-3/F27C1.7] and one Chromosome II clone [cchl-1/T06D8.6] were published previously [21,22]. The seven new clones were nuo-3/Y57G11C.12, nuo-4/K04G7.4, nuo-5/Y45G12B.1, cco-2/Y37D8A.14, atp-4/T05H4.12, atp-5/C06H2.1, and asb-2/F02E8.1 [see Table 1].) Respiratory-chain RNAi clones are thought to affect a pathway that is independent of the insulin/IGF-1 pathway [21,22]. Consistent with this, we found that all of the new mitochondrial RNAi clones were able to extend the lifespan of daf-2 mutants, and that their activities were daf-16 independent (Table 3).
RNAi of respiratory-chain components decreases body size and slows movement and eating behavior (pumping) [21]. Interestingly, while lifespan and size/pumping rate are inversely correlated for most of the animals in this class, there were several exceptions (see Table S3). For example, we found that animals subjected to nuo-5 (NADH-ubiquinone oxidoreductase) or cchl-1 (cytochrome C
heme-lyase) RNAi had almost normal body size (and, for nuo-5, also normal pumping rate) but substantial lifespan extensions (about 30%; see Table 1). Thus, the longevity of animals with reduced respiration is unlikely to be causally connected to their small size or reduced rate of pumping.
Genes That May Be Involved in the Longevity Response to Dietary Restriction
Nine additional RNAi clones also increased the lifespan of daf-16 mutants (Table 3). DR extends the lifespan of C. elegans in a daf-16-independent manner [18,19]. We therefore asked whether any of these clones might be involved in the response to DR. To do this, we asked whether they might fail to further extend the long lifespan of the eat-2(ad1116) mutant, which is a genetic model for DR [18,37]. We found that this was the case for four (sams-1/C49F5.1, rab-10/T23H2.5, drr-1/F45H10.4, and drr-2/T12D8.2) of the nine genes (Table 3). This was interesting because to date, the lifespan of all but one long-lived mutant examined, clk-1, can be extended further by mutations in eat-2 [18].
One of these genes, sams-1, encodes S-adenosyl methionine synthetase, a protein that functions as a universal methyl group donor in many biochemical reactions. Inhibition of this enzyme can affect methylation of histones, DNA, RNA, proteins, phospholipids, and other small molecules [38]. To further investigate the possibility that C. elegans sams-1 plays a role in the response to DR, we asked whether this RNAi clone, like DR [39] (and D. Crawford and C. Kenyon, unpublished data) reduced brood size and delayed reproduction. We found that brood size was reduced and reproductive timing was slightly delayed (Figure 2). In addition, like DR animals, sams-1 RNAi, worms were slender (see Table S1). The rate of pumping (eating) was not affected (data not shown), suggesting that this RNAi clone exerts its effects via changes in metabolism rather than changes in appetite. Together, these findings suggest that DR may extend lifespan, at least in part, by inhibiting the activity of sams-1. Consistent with this idea, we found that sams-1 mRNA levels were reduced 3-fold in eat-2 mutants (Figure 3). We therefore propose the model that DR may initiate a longevity response, at least in part, by triggering a regulated decrease in sams-1 mRNA levels and, consequently, cellular S-adenosyl methionine levels. Reducing dietary methionine levels is known to increase the lifespan of mice [40] and rats by 40–45% [41–43]; however, whether methionine limitation and general DR activate the same longevity mechanisms is not known. Our findings suggest that this might be the case.
Figure 2
sams-1,
rab-10, and drr-1 RNAi Affect Reproduction
Progeny profile of N2 animals grown on RNAi clones for (A) sams-1, (B) rab-10, (C) drr-1, and (D) drr-2 (note that drr-2 RNAi did not affect reproduction). Number of progeny per worm at each time interval is shown. Error bars: ± SEM. (E) Total brood size of N2 worms grown on RNAi clones for either sams-1,
rab-10, drr-1, or drr-2. The number of progeny produced by each worm was calculated from the progeny profile data in (A)–(D) and averaged. The p-values were calculated relative to control of the experiment as Student's t-test.
Figure 3 mRNA Levels of sams-1,
rab-10, drr-1, and drr-2 Are Reduced in eat-2(ad1116) Mutants
Relative mRNA levels of sams-1,
rab-10, drr-1, and drr-2 in eat-2(ad1116) compared to N2 were measured by quantitative PCR, and average of four different sample sets are shown. The relative mRNA levels were normalized against the act-1 (beta-actin) level in each sample. The RNAi clone for gei-9 is shown as a control; this clone does not cause significant lifespan extension when fed to N2 or eat-2 worms (Table S1, and data not shown). Error bars: ± SEM.
Another RNAi clone that failed to further extend the lifespan of eat-2 mutants was rab-10, which encodes a Rab-like GTPase similar to those that regulate vesicle transport. Like sams-1 RNAi, this clone did not affect pumping, but did reduce and delay reproduction (see Figure 2) and produced a slender appearance (see Table S1). The expression of this gene, too, was down-regulated in response to DR (2-fold; Figure 3). All of these properties were shared by drr-1 (dietary restriction response), a gene encoding a novel protein with no obvious human homolog (see Table S1 and Figures 2 and 3).
The phenotype of the last DR clone, drr-2, differed from those of the other DR clones in that drr-2(RNAi) worms had a normal, well-fed appearance (see Table S1) and normal reproduction (see Figure 2). DRR-2 is a putative RNA-binding protein, suggesting that it plays a regulatory role in triggering the longevity response (but not the reproductive response) to DR. Expression of this gene was also reduced in eat-2 mutants (2-fold, Figure 3), suggesting that, like the other genes we identified, down-regulation of this gene in response to DR somehow causes lifespan extension. In general, it was striking that expression of all four DR genes was reduced in eat-2 mutants. This suggests that in C. elegans, DR elicits a concerted transcriptional (or mRNA turnover) response that can inhibit multiple lifespan-shortening gene activities and thereby extends lifespan.
We also asked whether the lifespan of eat-2(ad1116) mutants could be increased by respiratory-chain RNAi, and found that it could (D. Crawford and C. Kenyon, unpublished data). Since reducing respiratory chain activity during development is required for lifespan extension [21], whereas reducing food levels only during adulthood extends lifespan [44], it seems likely that mitochondrial respiratory chain components and DR do not extend lifespan in exactly the same way.
Integrin Signaling is Likely to Play an Evolutionarily Conserved Role in Lifespan Limitation
Loss-of-function mutations in β-integrin (myospheroid) extend lifespan in Drosophila [45], and we found that RNAi clones of pat-4/C29F9.7, which encodes integrin-linked kinase, and pat-6/T21D12.4, which encodes actopaxin, a protein known to bind to integrin-linked kinase [46], increased lifespan in C. elegans (see Table 1). Integrin signaling influences insulin-signaling pathways in mammals [47], yet the lifespan extension produced by these two clones was daf-16 independent (Table 3). In addition, the lifespan of daf-2(e1370) mutants was extended when these animals were grown on either pat-4 or pat-6 RNAi (Table 3). Thus, integrin signaling may comprise a novel, conserved lifespan regulatory pathway, though it could potentially function in the insulin/IGF-1 pathway downstream of daf-16—which is known to act cell-non-autonomously [48]—to influence lifespan. (Inactivation of pat-4 and pat-6 impaired the health of eat-2 mutants, making their relationship to DR difficult to interpret.) The finding that genes or pathways already known to influence the lifespan of another organism also affect C. elegans' lifespan is significant, as such ancient, evolutionarily conserved longevity pathways could potentially also influence human lifespan.
Genes That Might Function in Novel Pathways to Influence Longevity
Using genetic epistasis analysis, we were able to associate most of our new genes with known aging-regulatory pathways or processes. However, the roles of three genes remained unclear. Two of these, ril-1/C53A5.1 and ril-2/C14C10.3 (RNAi-induced longevity), encoded novel proteins with no obvious homologs, whereas rha-2/C06E1.10 encoded a DEAH RNA helicase, suggesting that it regulates gene expression or nucleic-acid metabolism. All three RNAi clones extended the lifespan of both daf-16 and eat-2 mutants (Table 3), suggesting that they may function in a novel pathway or pathways to influence longevity.
Summary
In this study, we identified many interesting new genes whose normal function is predicted to inhibit longevity. Although our screen did not reach saturation, an interesting picture emerged. Most of the genes we identified fell into one of three classes: genes that influence lifespan through DAF-16/FOXO (8/29), genes that influence respiration (12/29), and genes that appear to affect the response to DR (4/29). Two more genes affected integrin signaling, which was known to influence lifespan in flies. Like the insulin/IGF-1/FOXO system and the respiratory chain, most biological pathways and systems consist of many genes, and we failed to identify even one component of many such systems (e.g., the TGF-β signaling system). In fact, of the genes that had conserved sequence motifs, only one, the rha-2 RNA helicase homolog, could not be linked to a known pathway. These findings are thought-provoking because until now, we have had no way of knowing whether the longevity pathways we know about represent only the tip of the iceberg. Our findings suggest that, in contrast, at most only one or a few other large multigenic systems influence lifespan in C. elegans. In other words, we may now be aware of most of the major biological pathways in C. elegans that, when inhibited, can produce large extensions in lifespan.
Materials and Methods
Strains.
All strains were maintained as described previously [49]. CF1037: daf-16(mu86)I, DA1116: eat-2(ad1116)II, CF1041: daf-2(e1370)III, CB4037: glp-1(e2141)III, MQ887: isp-1(qm150)IV, AA86: daf-12(rh41rh411)X, CF512: fer-15(b26)II; fem-1(hc17)III. All strains displayed similar Unc phenotypes when grown on unc-52 RNAi bacteria (data not shown).
RNAi aging screen.
The systematic RNAi screen was carried out as described [23,24] with some modifications. Each RNAi bacteria colony was grown at 37 °C in LB with 10 μg/ml tetracycline and 50 μg/ml carbenicillin, and then seeded onto NG-carbenicillin plates supplemented with 100 μl 0.1 M IPTG. For our screen, we employed a sterile strain, CF512 (fer-15(b26); fem-1(hc17)) [5], to avoid transferring aging worms away from their progeny. Approximately 60 eggs of CF512 were added to the RNAi plates and allowed to develop to adults at 25 °C and kept at this temperature (Chromosome I, II, and first half of Chromosome X) or shifted to 20 °C for the rest of their life (Chromosome III, IV, V, and second half of Chromosome X). As a positive control, we used a previously described daf-2 clone (pAD48, [9]), and as a negative control, we used the corresponding empty vector (pAD12, [9]). DAF-2 functions in both neural and non-neural tissues to influence lifespan [48,50,51], and daf-2 RNAi has been shown previously to double the lifespan of the animal [9]. Developmental phenotypes were scored at d 1, and additional 100 μl 0.1 M IPTG was added on d 7. Viability of worms on each plate was scored at d 24 of adulthood (25 °C) or d 30 of adulthood (20 °C), at which time generally all worms on control plates were dead and 75–95% of worms grown on daf-2 RNAi (pAD48) were still alive. RNAi appears to remain effective on older plates, since 30-d-old and newly seeded unc-52 RNAi plates were equally effective at inducing an Unc phenotype (data not shown). Plates were censored due to contamination, progeny production, no bacterial growth, etc. (see Figure S1). We then carried out quantitative survival analysis using both CF512 and N2 with all of the bacterial strains that scored positively in the screen. Unexpectedly, RNAi knockdown of four of the genes we retested, inx-8 (innexin-8), ril-3 (F26F2.1), zig-6 (protein with immunoglobulin domains), and gei-9 (similarity to acyl-CoA dehydrogenase), significantly extended lifespan of CF512 but not N2 animals (see Table S1).
RNAi clone analysis.
The identity of all positive RNAi clones was verified by sequencing of inserts with an M13-forward primer, and, upon every start of a lifespan analysis, by PCR with T7 primers. About half of the genes we identified in our screen are located in operons. Therefore, we considered the possibility that the phenotypes we observed were influenced by RNAi knockdown of co-transcribed genes. This type of “intra-operonic inhibition,” though rare, has been observed [52,53]. However, none of these RNAi clones produced the phenotypes predicted for knockdown of other genes in the same operon (data not shown). All gene annotations were based on WormBase and/or WormPD, and protein accession numbers were from UniProt.
Lifespan analysis.
Lifespan analysis was conducted at 20 °C as described previously unless otherwise stated [10,54]. Eggs were added to plates seeded with RNAi bacteria, and animals were transferred approximately every week to newly seeded plates. At least 60 worms were used for each experiment. More animals were included in the analysis to ensure sufficient power when the lifespan extension of a particular RNAi clone was expected to be minimal. To reduce the chance of false negative results, all RNAi clones that failed to extend the lifespan of a particular mutant strain were retested at least once more with the same strain (except for daf-2(e1370), see Table S1). Because of the large number of lifespan experiments conducted, we did not always perform positive controls with CF512 or N2 animals exactly in parallel with each mutant we examined. However, these controls were running in overlapping time frames, where they consistently extended lifespan, generally (in 21 of 28 repeated experiments for N2/CF512 and in 43 of 57 repeated experiments when including data on all strains used) by magnitudes that did not differ by more than ten percentage points from one another (see Table S1). All positive RNAi clones extended lifespan in at least four independent trials, including analysis of all genetic mutants.
In all experiments, the pre-fertile period of adulthood was used as t = 0 for lifespan analysis. Strains were grown at 20 °C at optimal growth conditions for at least two generations before use in lifespan analysis. Statview 5.01 (SAS, Cary, North Carolina, United States) software was used for statistical analysis and to determine means and percentiles. In all cases, p values were calculated using the Log-rank (Mantel-Cox) method.
Dauer assays.
For RNAi experiments, daf-2(e1370) animals were cultured at 20 °C on plates seeded with various RNAi clones or vector control, and their F1 eggs were transferred to 22.5 °C. Following incubation for four days, the number of dauers was determined using a dissecting microscope. 30–50% of the animals on vector control became dauers at 22.5 °C. Approximately 50–100 animals were scored in each experiment. daf-2(e1370) worms grown on ddl-3 RNAi gave rise to almost no progeny; therefore, this RNAi clone was not assayed.
For addressing whether germline ablation affects dauer formation, the number of dauers induced by either daf-2(e1370) or daf-2(e1370); mes-1(bn7) at 22.5 °C was assayed. This mes-1 allele causes ~50% of animals to lack germ cells and live long [55]: 61 ± 20% of daf-2(e1370) mutants formed dauers (n = 204) and 57 ± 13% of daf-2(e1370); mes-1(bn7) double mutants (n = 267) formed dauers (p = 0.42). (The sterile and fertile daf-2; mes-1 double mutants formed dauers at equal frequency at 22.5 °C.)
Stress response assays.
For the thermotolerance analysis, synchronized N2 animals grown on control or RNAi bacteria were shifted to 35 °C as 3-d-old adults; for the oxidative stress analysis, synchronized animals were exposed to 300 mM paraquat (Sigma, St. Louis, Missouri, United States) as 5-d-old adults. Survival was scored every 2 to 3 h after the treatment. Statview 5.01 (SAS) software was used for statistical analysis and to determine means. In all cases, p values were calculated using the Log-rank (Mantel-Cox) method.
Progeny production assays.
N2 eggs were incubated at 20 °C on plates seeded with various RNAi clones, and 24 late L4 stage worms were picked for each treatment and transferred to fresh RNAi plates every 12 h for 4–5 d. After this period, the worms were transferred every 24 h. Worms that crawled off the plates, bagged, or exploded were censored. All plates were then incubated at 20 °C for about 2 d and shifted to 4 °C. The number of worms that developed was determined at the end of the experiment.
Quantitative RT-PCR analysis.
Total RNA was isolated from approximately 5,000 d 1 adult worms and cDNA was made from 4 μg of RNA using Superscript II RT (Invitrogen, Carlsbad, California, United States). eat-2(ad1116) animals were harvested 8–18 h after N2, due to their delayed development. TaqMan real-time qPCR experiments were then performed by the Biomolecular Resource Center at UCSF as described in the manual using ABI Prism7900HT (Applied Biosystems, Foster City, California, United States). Primers and probes designed specifically for act-1,
sams-1, rab-10,
drr-1, and drr-2 are listed below.
Primers:
Act-1–720F: 5′-CTACGAACTTCCTGACGGACAAG-3′
Act-1–821R: 5′-CCGGCGGACTCCATACC-3′
Sams-1–209F: 5′-TCCGTCGTGTCATCGAAAAG-3′
Sams-1–275R: 5′-TTGCAGGTCTTGTGGTCGAA-3′
Rab-10–514F: 5′-GCTAAGATGCCTGATACCACTGA-3′
Rab-10–585R: 5′-ACTCTGCCTCTGTGGTTGCA-3′
Drr-1–137F: 5′-GGATTCTTTGGTTTACTCTAATTGTTCA-3′
Drr-1–208R: 5′-TCCGCAGGGCGAAGATT-3′
Drr-2–530F: 5′-TGAAGCCCCGTACCACAGA-3′
Drr-2–596R: 5′-CTTGGTCTCCTCTTCTTCTTGCT-3′
Probes (All probes listed here were labeled with FAM at the 5′ end and Black hole Quencher at the 3′ end):
Act-1-T: 5′-AAACGAACGTTTCCGTTGCCCAGAGGCTAT-3′
Sams-1–230T: 5′-TTGGATTCACCGACTCCAGCATTGG-3′
Rab-10–541T: 5′-CAATCCCGCGATACGGTGAATCCA-3′
Drr-1–166T: 5′-TTAATTATTTCCGCGGCGGCAACG-3′
Drr-2–551T: 5′-CGCTGAGATCGAGGCTCGCAA-3′
Supporting Information
Figure S1 Summary of RNAi Longevity Screen
(45 KB PPT)
Click here for additional data file.
Figure S2 Comparison of daf-2 RNAi Clones
(72 KB PPT)
Click here for additional data file.
Table S1 Complete Lifespan Analysis Data of RNAi Clones That Extend Lifespan
(839 KB DOC)
Click here for additional data file.
Table S2 Lifespan Analysis of Library Clones Encoding Known (Non-Neuronal) Longevity Genes
(48 KB DOC)
Click here for additional data file.
Table S3 Pumping Rate and Body Length of N2 Animals Grown on Mitochondrial RNAi Clones
Found at DOI: 10.1371/journal.pgen.0010017.st003 (49 KB DOC)
Click here for additional data file.
We thank all Kenyon lab members for discussions and help, Kaveh Ashrafi for comments on the manuscript, and Julie Ahringer, Ravi Kamath, and Andrew Fraser for the RNAi library. MH was supported by the Danish Medical and Natural Science Research Councils. ALH was supported by the Canadian Institutes of Health Research. This work was supported by grants from the Ellison Foundation and the NIH to CK, who is an American Cancer Society Professor at UCSF and also a co-founder of the biotechnology company Elixir Pharmaceuticals.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. MH, ALH, AD, and CK conceived and designed the experiments. MH and ALH performed the experiments and analyzed the data. MH, ALH, and CK wrote the paper.
Abbreviations
DRdietary restriction
RNAiRNA interference
TPRtetratrico-peptide-repeat
==== Refs
References
Tatar M Bartke A Antebi A 2003 The endocrine regulation of aging by insulin-like signals Science 299 1346 1351 12610294
Murphy CT McCarroll SA Bargmann CI Fraser A Kamath RS 2003 Genes that act downstream of DAF-16 to influence the lifespan of Caenorhabditis elegans
Nature 424 277 283 12845331
Lee RY Hench J Ruvkun G 2001 Regulation of C. elegans DAF-16 and its human ortholog FKHRL1 by the daf-2 insulin-like signaling pathway Curr Biol 11 1950 1957 11747821
Melendez A Talloczy Z Seaman M Eskelinen EL Hall DH 2003 Autophagy genes are essential for dauer development and life-span extension in C. elegans
Science 301 1387 1391 12958363
Garigan D Hsu AL Fraser AG Kamath RS Ahringer J 2002 Genetic analysis of tissue aging in Caenorhabditis elegans: A role for heat-shock factor and bacterial proliferation Genetics 161 1101 1112 12136014
Hsu AL Murphy CT Kenyon C 2003 Regulation of aging and age-related disease by DAF-16 and heat-shock factor Science 300 1142 1145 12750521
Morley JF Morimoto RI 2004 Regulation of longevity in Caenorhabditis elegans by heat shock factor and molecular chaperones Mol Biol Cell 15 657 664 14668486
Apfeld J O'Connor G McDonagh T DiStefano PS Curtis R 2004 The AMP-activated protein kinase AAK-2 links energy levels and insulin-like signals to lifespan in C. elegans
Genes Dev 18 3004 3009 15574588
Dillin A Crawford DK Kenyon C 2002 Timing requirements for insulin/IGF-1 signaling in C. elegans
Science 298 830 834 12399591
Apfeld J Kenyon C 1999 Regulation of lifespan by sensory perception in Caenorhabditis elegans
Nature 402 804 809 10617200
Alcedo J Kenyon C 2004 Regulation of C. elegans longevity by specific gustatory and olfactory neurons Neuron 41 45 55 14715134
Hsin H Kenyon C 1999 Signals from the reproductive system regulate the lifespan of C. elegans
Nature 399 362 366 10360574
Gerisch B Weitzel C Kober-Eisermann C Rottiers V Antebi A 2001 A hormonal signaling pathway influencing C. elegans metabolism, reproductive development, and life span Dev Cell 1 841 851 11740945
Tissenbaum HA Guarente L 2001 Increased dosage of a sir-2 gene extends lifespan in Caenorhabditis elegans
Nature 410 227 230 11242085
Oh SW Mukhopadhyay A Svrzikapa N Jiang F Davis RJ 2005 JNK regulates lifespan in Caenorhabditis elegans by modulating nuclear translocation of forkhead transcription factor/DAF-16 Proc Natl Acad Sci U S A 102 4494 4499 15767565
Wang MC Bohmann D Jasper H 2003 JNK signaling confers tolerance to oxidative stress and extends lifespan in Drosophila
Dev Cell 5 811 816 14602080
Wang MC Bohmann D Jasper H 2005 JNK extends life span and limits growth by antagonizing cellular and organism-wide responses to insulin signaling Cell 121 115 125 15820683
Lakowski B Hekimi S 1998 The genetics of caloric restriction in Caenorhabditis elegans
Proc Natl Acad Sci U S A 95 13091 13096 9789046
Houthoofd K Braeckman BP Johnson TE Vanfleteren JR 2003 Life extension via dietary restriction is independent of the Ins/IGF-1 signalling pathway in Caenorhabditis elegans
Exp Gerontol 38 947 954 12954481
Feng J Bussiere F Hekimi S 2001 Mitochondrial electron transport is a key determinant of life span in Caenorhabditis elegans
Dev Cell 1 633 644 11709184
Dillin A Hsu AL Arantes-Oliveira N Lehrer-Graiwer J Hsin H 2002 Rates of behavior and aging specified by mitochondrial function during development Science 298 2398 2401 12471266
Lee SS Lee RY Fraser AG Kamath RS Ahringer J 2003 A systematic RNAi screen identifies a critical role for mitochondria in C. elegans longevity Nat Genet 33 40 48 12447374
Kamath RS Ahringer J 2003 Genome-wide RNAi screening in Caenorhabditis elegans
Methods 30 313 321 12828945
Kamath RS Fraser AG Dong Y Poulin G Durbin R 2003 Systematic functional analysis of the Caenorhabditis elegans genome using RNAi Nature 421 231 237 12529635
Simmer F Moorman C van der Linden AM Kuijk E van den Berghe PV 2003 Genome-wide RNAi of C. elegans using the hypersensitive rrf-3 strain reveals novel gene functions PLoS Biol 1 e12 14551910
Kamath RS Martinez-Campos M Zipperlen P Fraser AG Ahringer J 2001 Effectiveness of specific RNA-mediated interference through ingested double-stranded RNA in Caenorhabditis elegans
Genome Biol 2 Research0002 11178279
Tavernarakis N Wang SL Dorovkov M Ryazanov A Driscoll M 2000 Heritable and inducible genetic interference by double-stranded RNA encoded by transgenes Nat Genet 24 180 183 10655066
Dorman JB Albinder B Shroyer T Kenyon C 1995 The age-1 and daf-2 genes function in a common pathway to control the lifespan of Caenorhabditis elegans
Genetics 141 1399 1406 8601482
Priess JR Schnabel H Schnabel R 1987 The glp-1 locus and cellular interactions in early C. elegans embryos Cell 51 601 611 3677169
Monaco HL 2000 The transthyretin-retinol-binding protein complex Biochim Biophys Acta 1482 65 72 11058748
Palha JA 2002 Transthyretin as a thyroid hormone carrier: Function revisited Clin Chem Lab Med 40 1292 1300 12553433
Eneqvist T Lundberg E Nilsson L Abagyan R Sauer-Eriksson AE 2003 The transthyretin-related protein family Eur J Biochem 270 518 532 12542701
Lowell BB Shulman GI 2005 Mitochondrial dysfunction and type 2 diabetes Science 307 384 387 15662004
Wilkinson MG Pino TS Tournier S Buck V Martin H 1999 Sin1: An evolutionarily conserved component of the eukaryotic SAPK pathway Embo J 18 4210 4221 10428959
Li S Armstrong CM Bertin N Ge H Milstein S 2004 A map of the interactome network of the metazoan C. elegans
Science 303 540 543 14704431
Satyal SH Chen D Fox SG Kramer JM Morimoto RI 1998 Negative regulation of the heat shock transcriptional response by HSBP1 Genes Dev 12 1962 1974 9649501
Raizen DM Lee RY Avery L 1995 Interacting genes required for pharyngeal excitation by motor neuron MC in Caenorhabditis elegans
Genetics 141 1365 1382 8601480
Chiang PK Gordon RK Tal J Zeng GC Doctor BP 1996 S-adenosylmethionine and methylation FASEB J 10 471 480 8647346
Shibata Y Fujii T Dent JA Fujisawa H Takagi S 2000 EAT-20, a novel transmembrane protein with EGF motifs, is required for efficient feeding in Caenorhabditis elegans
Genetics 154 635 646 10655217
Miller R Buehner G Chang Y Harper JM Sigler R 2005 Methionine-deficient diet extends mouse lifespan, slows immune and lens aging, alters glucose, T4, IGF-1 and insulin levels, and increases hepatocyte MIF levels and stress resistance Aging Cell 4 119 125 15924568
Zimmerman JA Malloy V Krajcik R Orentreich N 2003 Nutritional control of aging Exp Gerontol 38 47 52 12543260
Orentreich N Matias JR DeFelice A Zimmerman JA 1993 Low methionine ingestion by rats extends life span J Nutr 123 269 274 8429371
Richie JP Jr. Komninou D Leutzinger Y Kleinman W Orentreich N 2004 Tissue glutathione and cysteine levels in methionine-restricted rats Nutrition 20 800 805 15325691
Klass MR 1977 Aging in the nematode Caenorhabditis elegans : Major biological and environmental factors influencing life span Mech Ageing Dev 6 413 429 926867
Goddeeris MM Cook-Wiens E Horton WJ Wolf H Stoltzfus JR 2003 Delayed behavioural aging and altered mortality in Drosophila beta integrin mutants Aging Cell 2 257 264 14570233
Lin X Qadota H Moerman DG Williams BD 2003
C. elegans PAT-6/actopaxin plays a critical role in the assembly of integrin adhesion complexes in vivo Curr Biol 13 922 932 12781130
Dedhar S Williams B Hannigan G 1999 Integrin-linked kinase (ILK): A regulator of integrin and growth-factor signalling Trends Cell Biol 9 319 323 10407411
Libina N Berman JR Kenyon C 2003 Tissue-specific activities of C. elegans DAF-16 in the regulation of lifespan Cell 115 489 502 14622602
Brenner S 1974 The genetics of Caenorhabditis elegans
Genetics 77 71 94 4366476
Apfeld J Kenyon C 1998 Cell nonautonomy of C. elegans daf-2 function in the regulation of diapause and life span Cell 95 199 210 9790527
Wolkow CA Kimura KD Lee MS Ruvkun G 2000 Regulation of C. elegans life-span by insulinlike signaling in the nervous system Science 290 147 150 11021802
Montgomery MK Xu S Fire A 1998 RNA as a target of double-stranded RNA-mediated genetic interference in Caenorhabditis elegans
Proc Natl Acad Sci U S A 95 15502 15507 9860998
Bosher JM Dufourcq P Sookhareea S Labouesse M 1999 RNA interference can target pre-mRNA: Consequences for gene expression in a Caenorhabditis elegans operon Genetics 153 1245 1256 10545456
Kenyon C Chang J Gensch E Rudner A Tabtiang R 1993 A C. elegans mutant that lives twice as long as wild type Nature 366 461 464 8247153
Strome S Martin P Schierenberg E Paulsen J 1995 Transformation of the germ line into muscle in mes-1 mutant embryos of C. elegans
Development 121 2961 2972 7555722
Kenyon C 2005 The plasticity of aging: insights from long-lived mutants Cell 120 449 460 15734678
|
16103914
|
PMC1183531
|
CC BY
|
2021-01-05 08:00:24
|
no
|
PLoS Genet. 2005 Jul 25; 1(1):e17
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010017
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 1741133310.1371/journal.pgen.001002105-PLGE-ED-0145plge-01-01-12EditorialIntroducing PLoS Genetics
EditorialFrankel Wayne N *Wayne N. Frankel is Editor-in-Chief of PLoS Genetics. E-mail: [email protected] 2005 25 7 2005 1 1 e21Copyright: © 2005 Wayne N. Frankel.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Citation:Frankel WN (2005) Introducing PLoS Genetics. PLoS Genet 1(1): e21.
==== Body
On behalf of our editorial team, it is my pleasure to welcome you to PLoS Genetics, a new open-access journal from the Public Library of Science (PLoS). Led by an internationally recognized editorial board with broad knowledge and expertise, PLoS Genetics is a journal that celebrates the research of the greater genetics and genomics community. As you see in this first issue, PLoS Genetics is unique—publishing outstanding articles that reflect the full breadth and interdisciplinary nature of this research, all free to read and to use in your own research and teaching.
How did PLoS Genetics come about? In 2004, when PLoS asked several of us in the genetics community about a need and desire for an open-access genetics/genomics journal, I replied with a resounding “yes!” And I was not alone—others had the same reaction that the time was right for a new genetics journal of high quality. Certainly the open-access element was key—following in the public-domain spirit of genetics and genomics data release, for example, by the Human Genome Project. And creating such a journal—building on the strong experience and reputation of PLoS Biology—seemed an opportunity not to miss.
What is the focus of PLoS Genetics? Our mainstay is primary research articles, of which there are eleven in our inaugural issue. We strive for high-quality, original contributions from a broad sweep of the research community, the common theme being novel or incisive applications of genetics and genomics tools to address important research problems in any area of biology. The format itself is flexible; we encourage authors to be creative about the most appropriate and efficient way to present their research. Because we make this high-level research fully and immediately available on our Web site and through PubMed Central, scientists can keep current, build on these findings, and respond by submitting work of their own.
Will PLoS Genetics have special features? Yes—the prospect of which was another key to my own interest in being involved! In addition to research articles, PLoS Genetics offers a venue to consider important issues of the day. Our Reviews editors—Elizabeth Fisher, Nicholas Katsanis, Marcy MacDonald, and Susan Rosenberg—both invite reviews and consider submitted ones that summarize a particularly interesting, hot, or forward-looking aspect of genetics and genomics research, and include the authors' unique perspective on it. Beginning in August, we will aim to publish at least one review each month. Every few months, we will also offer a special feature article. For example, in this issue, Interviews Editor Jane Gitschier presents her one-on-one interview with statistical geneticist Neil Risch (DOI: 10.1371/journal.pgen.0010014).
PLoS Genetics celebrates the research of the greater genetics and genomics community.
In future issues, our front section will offer a correspondence forum—intended for readers to comment and shed light on specific manuscripts or topics of the day, or for editors to reply to questions raised by readers, particularly students. Contributions to Correspondence will be published at the discretion of the editors. I am excited about launching this section. Our community currently does not have a forum such as this, and I believe that there are some important issues that could benefit from an open exchange of ideas.
How does the PLoS Genetics editorial process operate? The PLoS Genetics editorial team is drawn from the community it serves. As a result, all review—whether by associate editors assigned to papers or the peer reviewers they invite—is conducted by experts in a broad range of disciplines. Reviewers are instructed to be thorough and decisions are reached after consultation between associate editors and myself.
The 30 or so members of the PLoS Genetics Editorial Board serve as our associate editors. They are a talented and dedicated group who handle each paper individually, but they are also encouraged to work together. As is appropriate for a journal that has the community in mind, associate editors may consult with one another at any point during the review process using the journal's Web-based review system. I have observed so far that these exchanges are learned and insightful, and ultimately lead (I hope) to the fairest possible decision about a manuscript that still keeps the ambitious goals of PLoS Genetics in mind.
I also invite “guest editors” to oversee the review process for particular manuscripts when the expert on our board is either unavailable or has a conflict of interest with a submission, or when a manuscript simply “falls between the cracks” of our expertise. The guest editor feature reflects, in a way, the fact that we are a community journal. Some of you will be tapped for this sooner or later—consider yourselves warned!
Although we aim for high-caliber peer reviews, we also strive to keep the turnaround time as short as possible. Our goal is to return a response to authors within 30 days of submission, and we're off to a decent start. In just six months of submissions, our average turnaround time from receipt to first decision was 35 days.
How are things going so far? The response to PLoS Genetics has been greater than we imagined—well over 100 research articles submitted, more than 3,000 sign-ups for electronic table of contents (eTOC) alerts, and a steady stream of presubmission inquiries. Are we receiving the breadth of manuscripts we hoped for? Absolutely. One look at our July issue table of contents will show you studies that represent a variety of organisms (yeast—two kinds—plants, nematodes, flies, mice, cats—big and little—and humans) and areas of research (genome annotation, genetics, epigenetics, microarrays, cancer, glaucoma, neurodegeneration, and evolution). The August issue will add prokaryotes, fish, and other new dimensions to the journal. Several studies consider multiple organisms, matching well our goals of connecting the greater genetics and genomics community.
What is the vision for PLoS Genetics? The editorial team has guided the journal from an idea to the launch issue you see here, but what we become is ultimately determined by the will and participation of the genetics and genomics community. I hope that you will embrace and, in one way or another, become a part of PLoS Genetics. █
|
17411333
|
PMC1183532
|
CC BY
|
2021-01-05 07:59:36
|
no
|
PLoS Genet. 2005 Jul 25; 1(1):e21
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010021
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1631598752910.1186/1471-2105-6-163Research ArticleSignal transduction pathway profiling of individual tumor samples Breslin Thomas [email protected] Morten [email protected] Carsten [email protected] Carl [email protected] Complex Systems Division, Department of Theoretical Physics, University of Lund, Sölvegatan 14A, SE-223 62 Lund, Sweden2005 29 6 2005 6 163 163 14 3 2005 29 6 2005 Copyright © 2005 Breslin et al; licensee BioMed Central Ltd.2005Breslin et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Signal transduction pathways convey information from the outside of the cell to transcription factors, which in turn regulate gene expression. Our objective is to analyze tumor gene expression data from microarrays in the context of such pathways.
Results
We use pathways compiled from the TRANSPATH/TRANSFAC databases and the literature, and three publicly available cancer microarray data sets. Variation in pathway activity, across the samples, is gauged by the degree of correlation between downstream targets of a pathway. Two correlation scores are applied; one considers all pairs of downstream targets, and the other considers only pairs without common transcription factors. Several pathways are found to be differentially active in the data sets using these scores. Moreover, we devise a score for pathway activity in individual samples, based on the average expression value of the downstream targets. Statistical significance is assigned to the scores using permutation of genes as null model. Hence, for individual samples, the status of a pathway is given as a sign, + or -, and a p-value. This approach defines a projection of high-dimensional gene expression data onto low-dimensional pathway activity scores. For each dataset and many pathways we find a much larger number of significant samples than expected by chance. Finally, we find that several sample-wise pathway activities are significantly associated with clinical classifications of the samples.
Conclusion
This study shows that it is feasible to infer signal transduction pathway activity, in individual samples, from gene expression data. Furthermore, these pathway activities are biologically relevant in the three cancer data sets.
==== Body
Background
The interpretation of microarray data is facilitated by combining the data, or results of data analysis, with prior contextual knowledge, e.g., ontologies [1-4], pathways [5-7] and other annotation groups of interest [8]. By using prior knowledge about pathways, we aim at inferring cellular signaling pathway activity from tumor microarray data, on a sample-by-sample basis. Furthermore, we examine whether the pathway activity of individual samples is associated with clinical classifications of the samples. Our approach is in sharp contrast to establishing pathways from gene expression data (see e.g. [9]). What we do here is to project gene expresson data onto prior knowledge, in this case established pathway databases.
Signaling pathway activity scoring is a more direct measurement of biological processes than ontology mapping, which aims at finding over-representation of genes in various groups of contextual annotation. A cellular signaling pathway (see Fig. 1) is composed of a series of signaling molecules that convey information, typically from the outside of the cell to the nucleus. The initial step consists of extracellular signaling molecules, ligands, that activate receptors of the cell. These receptors then initiate intracellular signaling events, which eventually regulate the activity of various transcription factors. These transcription factors, in turn, regulate the expression levels of various genes, termed downstream targets of the pathway.
Figure 1 Example pathway. A simplified and partial view of the TNF-α pathway. A ligand (yellow) binds to a receptor (green) on the cell surface, triggering a cascade of events. Eventually, transcription factors (blue) activate or repress the expression of genes.
To characterize pathway activity, it would be desirable to have both proteomic and gene expression data. Gene expression data alone is not sufficient for assessing protein concentrations [10] and post-translational modifications of proteins. In the absence of proteomic data, one is thus forced to rely on aspects of the pathway that are detectable at the mRNA level. The foremost candidate for this is the downstream targets, which we will focus on here. It is of course also possible that the mRNA levels of effector proteins in a pathway change due to altered pathway activity. However, such effects are outside the scope of this paper. A further complication is that many pathways overlap, both in terms of having common transcription factors, and in terms of distinct transcription factors having common downstream targets. Our methods will not be able to distinguish very similar pathways, and in some sense this problem can be seen as a result of the ambiguities that follow when the full protein network is partitioned into separate pathways.
Cellular signaling pathways are subject to intense research, and current knowledge is compiled into databases such as STKE [11], TRANSPATH [12] and TRANSFAC [13]. These databases do not yet account for all pathways or transcription factors, but develop over time. Most pathway information utilized in this work is collected from TRANSPATH/TRANSFAC, where information about transcription factors and downstream targets is readily available. The only exception is the estrogen receptor pathway, which is taken from [14]. We analyze three microarray data sets in this study: Two breast cancer data sets [15,16], and one leukemia data set [17].
For the three data sets, we assess pathway activity from two related, but different, points of view. The first is to examine which pathways behave in a coherent way across the entire data set, i.e., which pathways have significantly co-expressed downstream targets. This is done both with and without accounting for the fact that downstream targets of a single transcription factor are correlated irrespective of pathway behavior. The second point of view is to assess pathway activity of individual samples, relative to the other samples in the same experiment, yielding an active or inactive status for each pathway in each sample. Finally, we relate the sample-wise pathway activity to clinical classifications of samples by way of contingency tables. Several pathways are found to be highly predictive of the clinical classifications.
Results
In this section we assess variation in transcription factor and pathway activity across the samples. We then proceed to probe pathway activity in individual samples. Finally, we study the association between pathway activity and the clinical classifications of the samples.
Co-expression of the downstream targets of a transcription factor
As a prelude to the study of pathways, we quantified the degree of correlation among downstream targets of single transcription factors. For this purpose we used the Group Correlation Score defined in Methods. The p-values were calculated using random reshuffling of the genes. Table 2 shows the most significant transcription factors in the van 't Veer data set. We see, as expected, that several transcription factors have significantly correlated downstream targets. For the data set of Sotiriou et al., 8 out of 42 transcription factors have a p-value below 0.1, and for the Golub et al. data set, the corresponding numbers are 6 out of 39. Among these three data sets, the p-values are noticeably better for data sets with more samples and genes. With this in mind we conclude that the downstream targets of transcription factors are co-expressed in the these data sets, albeit not to a high degree. The full lists for all 3 data sets can be found in Additional file 1.
Table 1 Transcription factor significance. The 15 most significant transcription factors (TF) in the van 't Veer et al. [15] data set, and their number of downstream targets (DT). p-values are based on the Group Correlation Score. In all 54 TFs were studied for this data set. ER pathway notation as in Table 1.
TF # of DT p-value
NF-κB 19 4e-04
RelA 11 8e-04
ER Complex(i) 50 2e-03
NF-κB1 10 2e-03
STAT1α 2 6e-03
C/EBPα 21 7e-03
ER-induced (v) 5 7e-03
ER 77 8e-03
STAT6 6 8e-03
ER(v) 7 2e-02
GATA-1 7 2e-02
Elk-1 4 6e-02
c-Rel 3 6e-02
STAT4 3 7e-02
SMAD-3 6 7e-02
Co-expression of the downstream targets of a pathway
Here we use the same Group Correlation Score as above, applied to the downstream targets of entire pathways rather than those of individual transcription factors. Table 1 shows the results for the van 't Veer data set, where 21 out of 29 pathways have a p-value below 0.05, which is significantly more than expected by chance. However, many of the downstream targets have common transcription factors, which might be the major cause of the co-expression. To eliminate such a contribution we used theExclusive Group Correlation Score, which considers only pairs of downstream targets lacking common transcription factors. The p-values for the Exclusive Group Correlation Score are also shown in Table 1. Although these p-values are larger, they are still significant; out of 20 pathways with more than one transcription factor, 12 have a Exclusive Group Correlation Score p-value below 0.05, which is still more than expected by chance. We conclude that the co-expression of downstream targets in a pathway can only in part be explained by the genes having common transcription factors. This co-expression at the pathway level justifies the view of pathways as functional units. Similar tables for the two other data sets are shown in Additional file 1. Both data sets have smaller p-values than expected by chance, albeit not as convincingly as the van 't Veer data set.
Table 2 Pathway significance. Pathways in the van 't Veer et al. [15] data set, ordered by significance and their number of transcription factors (TF) and downstream targets (DT). Also shown are Group Correlation Score (GCS) and Exclusive Group Correlation Score (EGCS) p-values. ER means both induced and repressed ER-pathway and (v) means that the pathway has been verified in a second experiment (see [14]).
Pathway # of TF # of DT GCS p-value EGCS p-value
IL-1 5 21 1e-04 4e-01
fMLP 9 27 4e-04 1e-01
TLR4 9 41 9e-04 1e-03
EDAR 6 41 3e-03 5e-02
ER-induced 1 50 3e-03 n/a
RANK 6 41 3e-03 5e-02
Oncostatin M 1 2 5e-03 n/a
PDGF 8 15 6e-03 2e-02
ER 1 77 7e-03 n/a
ER-induced(v) 1 5 7e-03 n/a
IL-4 – STAT6 1 6 8e-03 n/a
TGF-β network 7 23 1e-02 1e-02
EGF 12 53 1e-02 1e-02
ER (v) 1 7 2e-02 n/a
Insulin 7 45 2e-02 4e-02
VEGF 3 8 2e-02 2e-02
TNF-α 8 61 2e-02 5e-02
TPO 6 10 3e-02 2e-02
PRL 6 10 3e-02 2e-02
IFN 6 10 3e-02 2e-02
IL-10 2 7 5e-02 5e-02
IL-12 – STAT4 1 3 7e-02 n/a
c-Kit 4 87 8e-02 6e-02
ER-repressed 1 27 1e-01 n/a
B-cell antigen receptor 4 10 3e-01 3e-01
T-cell antigen receptor 4 10 3e-01 3e-01
Wnt pathway 2 8 4e-01 3e-01
ER-repressed (v) 1 2 6e-01 n/a
IL-2 – STAT5 2 4 8e-01 7e-01
Pathway assignments for individual samples
After having established that downstream target genes are co-expressed in some pathways, we proceeded to study the status of pathway activity in individual samples. To this end we employed the Group Sample Score, which for each pathway designates every sample in a data set as either active or inactive, with an associated p-value.
Table 3 shows p-values and pathway activity status, for six samples in the van 't Veer data set. For example, in the first sample the RANK pathway is designated as inactive with a p-value of 0.004, whereas the inactiveness of the ER-induced pathway cannot be considered significant. The full tables for all samples and pathways in all 3 data sets are provided in Additional file 1.
Table 3 Sample pathway activity. The individual sample pathway activity p-values and sign for each pathway and six of the van 't Veer breast cancer samples. Bold face indicates significant (i.e. p-value ≤ 0.05) pathway activity in the sample. ER notation as in Table 1.
1 2 3 4 5 6
ER-induced 0.7066(-) 0.0000(+) 0.4822(+) 0.1648(+) 0.4106(+) 0.0072(+)
EDAR 0.0052(-) 0.1514(-) 0.1294(-) 0.8566(+) 0.0010(-) 0.0008(-)
IL-1 0.0074(-) 0.0416(-) 0.0158(-) 0.3848(+) 0.0088(-) 0.0002(-)
RANK 0.0058(-) 0.1506(-) 0.1340(-) 0.8598(+) 0.0022(-) 0.0008(-)
TNF-α 0.0014(-) 0.1744(-) 0.2596(-) 0.4044(+) 0.0030(-) 0.0006(-)
EGF network 0.2868(-) 0.3030(-) 0.4522(-) 0.9360(-) 0.2138(-) 0.0000(-)
ER 0.9994(-) 0.0002(+) 0.3074(+) 0.7624(+) 0.7886(-) 0.0002(+)
ER-induced (v) 0.4940(-) 0.0136(+) 0.1670(+) 0.3060(+) 0.3310(+) 0.3750(+)
TLR4 0.1584(-) 0.2572(-) 0.2624(-) 0.7778(+) 0.1856(-) 0.0000(-)
fMLP 0.0122(-) 0.0432(-) 0.0004(-) 0.7188(+) 0.0362(-) 0.0268(-)
Insulin 0.9274(-) 0.0182(-) 0.3926(-) 0.6116(+) 0.7898(+) 0.0172(-)
ER (v) 0.9382(-) 0.0084(+) 0.0680(+) 0.0806(+) 0.3266(-) 0.0802(+)
TGFβ network 0.1102(+) 0.7054(+) 0.8274(-) 0.3528(-) 0.7902(-) 0.2550(-)
c-Kit 0.2892(-) 0.0282(-) 0.3000(-) 0.7884(-) 0.0470(-) 0.2398(-)
ER-repressed (v) 0.3518(+) 0.0566(+) 0.0588(+) 0.0426(+) 0.0034(-) 0.0536(+)
VEGF 0.6526(-) 0.3174(+) 0.4388(-) 0.9372(-) 0.8526(-) 0.8608(-)
IL-10 0.3984(-) 0.5984(+) 0.8164(-) 0.8188(+) 0.2442(-) 0.7698(+)
IFN 0.3934(-) 0.7678(+) 0.4792(-) 0.7824(+) 0.5702(-) 0.7430(-)
PRL 0.3984(-) 0.7726(+) 0.4794(-) 0.7840(+) 0.5782(-) 0.7344(-)
TPO 0.3852(-) 0.7500(+) 0.4736(-) 0.8056(+) 0.5588(-) 0.7388(-)
PDGF 0.9488(+) 0.4568(-) 0.6370(-) 0.5418(-) 0.6680(+) 0.4060(-)
Oncostatin M 0.2746(-) 0.2986(-) 0.1848(-) 0.1428(+) 0.7450(+) 0.0046(-)
T-cell antigen receptor 0.2092(-) 0.1006(-) 0.0094(-) 0.9426(+) 0.6404(-) 0.3304(+)
B-cell antigen receptor 0.1978(-) 0.1040(-) 0.0142(-) 0.9484(+) 0.6508(-) 0.3222(+)
IL-12 – STAT4 0.3458(+) 0.4402(+) 0.8902(+) 0.6424(+) 0.1982(-) 0.9878(-)
ER-repressed 0.5724(+) 0.4596(+) 0.3936(+) 0.1784(-) 0.0830(-) 0.0158(+)
IL-2 – STAT5 0.1816(-) 0.8828(-) 0.4774(-) 0.7768(-) 0.3486(+) 0.0376(-)
IL-4 – STAT6 0.8268(-) 0.1670(-) 0.9212(-) 0.3412(-) 0.1430(+) 0.1510(-)
Wnt 0.8484(-) 0.1054(-) 0.0878(-) 0.5868(-) 0.3756(-) 0.5084(-)
Table 4 shows the number of samples that are active and inactive at the 5% level, for every pathway in the van 't Veer data set. The table also contains the family-wise p-value, defined in Methods, which gives the probability of observing at least this total number of significant samples for a pathway. The family-wise p-value assumes that the samples are independent, which is only approximately true since the mean expression value of a gene across all samples is zero. Corresponding tables for the two other data sets can be found in Additional file 1. We note that the most significant pathways according to this measure are mostly the same as with the correlation based scores, although the p-values are numerically different.
Table 4 Pathway significance. This table shows the number of samples (out of 117) in the van 't Veer data set with pathway status active (+) or inactive (-) and with Group Sample Score p-value ≤ 0.05. Also shown are the corresponding family-wise p-values. ER notation as in Table 1.
Pathway + - family-wise p-value
ER-induced 30 37 2e-55
EDAR 30 30 6e-46
IL-1 25 34 1e-44
RANK 29 30 1e-44
TNF-α 29 29 2e-43
EGF network 29 29 2e-43
ER 28 29 4e-42
ER-induced(v) 21 34 1e-39
TLR4 28 27 1e-39
fMLP 24 31 1e-39
Insulin 22 21 6e-26
ER (v) 17 21 6e-21
TGFβ network 20 18 6e-21
c-Kit 20 14 3e-17
ER-repressed (v) 11 17 4e-12
VEGF 11 15 1e-10
IL-10 12 13 7e-10
IFN 11 14 7e-10
PRL 11 14 7e-10
TPO 10 14 3e-09
PDGF 9 12 4e-07
Oncostatin M 6 15 4e-07
T-cell antigen receptor 11 8 6e-06
B-cell antigen receptor 11 7 2e-05
IL-12 – STAT4 7 9 2e-04
ER-repressed 7 6 6e-03
IL-2 – STAT5 5 7 1e-02
IL-4 – STAT6 5 3 2e-01
Wnt 4 4 2e-01
In Figures 2 and 3, we show heatmaps for two of the pathways, ER-induced and TNF-α respectively. These two examples are illustrative; there is a substantial fraction of genes that are clearly upregulated in the active samples and downregulated in the inactive samples, as seen by the red upper right corner and green upper left corner. There are also many genes in the pathway that do not seem to be regulated, and even a few which act oppositely. The latter ones are possibly genes, that are repressed by the pathway. Similar figures for the entire genome (not shown) do not show large red or green upper corners, confirming the statistical analysis.
Figure 2 ER-induced pathway heatmap. Heatmap of the ER-induced pathway activity corresponding to Table 4. The vertical divisions correspond to samples where the ER-induced pathway is significantly upregulated (30 samples, red bar), its status undetermined (50 samples, black bar) or significantly downregulated (37 samples, green bar) respectively. The rows represent genes in the ER-induced pathway. The genes are shown in descending order according to scalar product with the vector ± (1 - p-value) for each sample, where the sign is the sign of the pathway activity of that sample. The expression values are log ratios normalized as described in Methods. Red and green represent up- and down-regulation respectively, and the precise color scheme is illustrated in the color key.
Figure 3 TNF-α heatmap. Heatmap of the TNF-α pathway activity corresponding to Table 4. Notations as in Figure 2.
Association between sample-wise pathway activity and clinical classifications
We analyzed the association between sample-wise pathway activity and clinical classifications using contingency tables. For every pathway and data set, we divided the samples into three groups: Samples where the pathway was active at a 5% significance level, samples where it was inactive at a 5% significance level, and insignificant samples referred to as undecided. For each data set, contingency tables of pathway activity versus clinical classifications were created, and χ2 p-values were calculated.
In the data set of Golub et al., the only available clinical classification is tumor type, i.e., ALL or AML. Table 5 shows the contingencies for the Insulin pathway and the I1-1 pathway. Seven out of 29 pathways have contingency tables with a χ2 p-value below 0.01.
Table 5 Contingency tables for the ALL/AML status versus the Insulin and IL-1 pathways in the leukemia data set of Golub et al. [17]. Active, non-active and undecided pathways are denoted +, - and U respectively.
ALL AML
Insulin pw(+) 1 5
Insulin pw(-) 15 0
Insulin pw(U) 11 6
p-value: 1e-05
ALL AML
IL-1 pw(+) 0 6
IL-1 pw(-) 19 0
IL-1 pw(U) 8 5
p-value: 5e-04
For the breast cancer data set of van 't Veer et al., we investigated six clinical classifications: metastasis status (0), estrogen receptor status (20), progesterone receptor status (12), lymph node status (12), BRCA mutations (15) and histological grade (8). The numbers in parentheses refer to the number of significant contingency tables at the 0.01 level. The total number of pathways was again 29. For metastasis status, only 97 out of the 117 samples were labeled in the original data set, and this may contribute to the low degree of association between this clinical classification and pathway activity. However, similar results were obtained for the data set of Sotiriou et al., which indicates that it may be difficult to obtain any association between the pathways analyzed in this work and breast cancer metastasis status. Table 6 shows the contingency between estrogen receptor status and the ER-induced pathway. As expected, there is a strong association between presence of the estrogen receptor protein, and the activity status of the ER-induced pathway. Somewhat more surprisingly, there are also strong associations between ER status and many other pathways. Similar results are obtained for the data set of Sotiriou et al., but with fewer significant associations.
Table 6 Contingency tables of estrogen receptor protein (binned at three levels: 0, 5–50, 60–100) versus the ER-induced and RANK pathways in the breast cancer data set of van't Veer et al. [15]. Same notation as in Table 5.
ERp low ERp med. ERp high
ER-ind pw(+) 0 5 25
ER-ind pw(-) 31 3 3
ER-ind pw(U) 8 16 26
p-value: 2e-14
ERp low ERp med. ERp high
RANK pw(+) 25 2 2
RANK pw(-) 1 5 24
RANK pw(U) 13 17 28
p-value: 1e-11
A general tendency of the contingency table analysis is illustrated in Table 7. Lowering the pathway activity p-value cutoff makes the association to clinical classifications more specific but less sensitive. The complete set of contingency tables for all three data sets can be found in Additional file 1.
Table 7 Contingency table of lymphocytic infiltration status versus the IL-12/STAT4 pathway in the van't Veer data set. The upper and lower tables are obtained with a pathway activity cutoff at 0.05 and 0.1 respectively. Same notation as in Table 5.
cutoff: 0.05 L+ L-
IL-12/STAT4 pw(+) 0 7
IL-12/STAT4 pw(-) 9 0
IL-12/STAT4 pw(U) 80 21
p-value: 2e-05
cutoff: 0.1 L+ L-
IL-12/STAT4 pw(+) 2 8
IL-12/STAT4 pw(-) 13 0
IL-12/STAT4 pw(U) 74 20
p-value: 2e-05
Conclusion
We have shown that downstream target genes of signal transduction pathways behave coherently in gene expression tumor data sets. First, we confirmed that downstream targets of transcription factors are correlated across samples. We then demonstrated that the same holds true for downstream targets of an entire pathway, even after discounting the correlations due to genes having a common transcription factor. The correlations for entire pathways were found to be more significant than those for individual transcription factors.
The presence of significant correlations confirms the expectation that gene expression is controlled by the activity of pathways. However, these correlations do not tell us in which samples a pathway is active or inactive. To reveal this, we devised the Group Sample Score. With this score we classified the samples into those where the pathway was significantly active, significantly inactive or undecided, respectively. As seen in Table 4, the number of significant samples is, for most pathways, much higher than the random expectation.
In many cases, the active/inactive pathway status was highly correlated with independent clinical classifications. This confirms the relevance of pathways for understanding of the underlying biology. Furthermore, the activity status of one or more pathways may be used to subdivide the samples into groups with distinct biological characteristics. Such a subdivision is feasible if, for instance, tumors of a certain clinical diagnosis are an agglomerate of several subtypes.
The Group Sample Score is natural if a pathway either induces all its downstream targets, or represses them. However, in most pathways some downstream genes are induced, while others are repressed. To account for a mixture of induction and repression, one should include a sign, or more generally a weight, to each term in the sum. Such a weight might even depend on the type of tissue and the environment. Since this information was not readily available for the studied pathways, all genes were weighted equally. Nevertheless, we obtained significant results, indicative of a dominant trend among the downstream genes. For the estrogen receptor (ER) pathway, we did have information about the sign, but instead of introducing a more general score for this pathway alone, we split the ER pathway into two parts, with induced and repressed genes, respectively. In the breast cancer data set of van 't Veer et al. [15], there were 50 genes in the induced part and 27 in the repressed. As seen in Tables 1 and 4, the ER-induced pathway was highly significant, whereas the repressed pathway was not. The full pathway was also highly significant, although to a lesser extent. The significance of the full pathway is thus due to the induced genes, which constitute a majority of the downstream targets. The situation is similar for other pathways and data sets.
It should be stressed that correlations, and the pathway activity status observed in a sample, are only defined relative to the other samples in the same data set. If a pathway were active in all samples, it would not show up in our significance test. The status of a pathway, as we define it, is given by the downstream genes, and the connection to ligands, receptors and other pathway components cannot be inferred from this analysis.
Table 2 shows that the most significant transcription factor in the breast cancer data set of van 't Veer et al. is NF-κB. This transcription factor is also the most one in the leukemia data set of Golub et al., whereas NF-κB1 is the most significant one in the data set of Sotirou et al Recently, NF-κB has been shown to be involved in the transformation from benign to malignant cells in inflammation-associated cancers. Pikarsky et al. [18] demonstrate this in a mouse model of human hepatocellular carcinoma, where the inflammatory mediator tumor-necrosis factor-α (TNF-α) is shown to play an important role as an activator of NF-κB. Greten et al. [19] find similar results in a mouse model of colitis-associated cancer.
Our current knowledge of pathways, and of downstream targets of transcription factors, is far from complete. However, we find that the results presented herein constitute a proof of concept for analyzing microarray gene expression in the context of signal transduction pathways.
Methods
Pathway information and UniGene clusters
Transcription factors for 23 pathways were extracted from TRANSPATH [12]. The downstream target genes of those transcription factors were obtained from TRANSFAC [13]. Since our study contains breast cancer data, we have augmented the pathway information with the Estrogen Receptor (ER) pathway compiled from [14], where 89 direct target genes were identified. 59 of them were induced by the ER complex and 30 were repressed. 8 out of the 89 genes were previously verified. We employ all six combinations of induced/repressed/all and verified/all, yielding 6 versions of the ER pathway. The 29 pathways employed are listed in Table 1. Downstream target genes were represented as UniGene IDs , using UniGene Hs build 171. For the analysis of the data sets, gene identifiers were converted into UniGene IDs and expression values of clones belonging to the same UniGene cluster were averaged.
Data sets
The following three publicly available data sets were analyzed:
1. The breast cancer data set of van 't Veer et al. [15], consisting of samples from 117 patients, of which 46 developed metastases. After UniGene merging the data set contains 20663 genes.
2. The breast cancer data set of Sotiriou et al. [16], consisting of 99 samples of different clinical classifications, with 4878 genes after UniGene merging.
3. The leukemia data set of Golub et al. [17], derived from bone marrow samples from 38 patients, 27 of which were diagnosed with Acute Myeloid Leukemia (AML) and 11 with Acute Lymphoblastic Leukemia (ALL). After Uni Gene merging and removal of genes with no variance across the samples, 4701 genes remain.
Normalization of microarray data
The data sets in [15,16] are given in the form of log ratios of expression values in the samples versus a reference. The data set in [17] is given in the form of Affymetrix average difference values. For the calculation of Group Correlation Score and Exclusive Group Correlation Score the affymetrix average difference values were logarithm transformed, since the Pearson correlation is very sensitive to single outlier samples. For the Group Sample Score the original differences were kept. We denote the expression value for gene g in sample s by xgs, with missing values allowed. We normalized the expression values in two steps. First, for each sample, the mean of all genes was subtracted, in order to ensure that no samples are up- or down-regulated on average. The transformed expression values satisfy:
Second, for each gene, the mean of all samples was subtracted from the expression values of that gene, yielding:
The second normalization implies that the expression value of a gene is measured relative to the same gene in other samples.
Pearson correlation p-values
To determine if a group of downstream target genes is significantly co-expressed, a total score of the group is needed. Two scores were used here, both based on the Pearson correlation of a pair of genes g and h:
where the sums exclude missing values and is the mean of expression values for gene g.
The Group Correlation Score is defined as the sum of squares of Pearson correlations among all pairs of genes in a group of genes:
where the sum runs over all genes in the group. The square ensures that both correlations and anti-correlations contribute to the score. We use the Group Correlation Score for the downstream target genes of a single transcription factor, as well as for those of an entire pathway.
The Exclusive Group Correlation Score, on the other hand, is only applicable for the downstream targets of a pathway. It is defined as
where the sum runs exclusively over pairs of genes g and h that do not share any transcription factor.
The p-value of a score is defined as the fraction of random cases, drawn under the null hypothesis, which achieve a higher score than the score tested. For both scores, GCS and EGCS, our null hypothesis is reshuffling of the genes on the microarray. This null hypothesis keeps the structure and overlap of all pathways fixed, but changes the identity of the genes.
Pathway activity for individual samples
For each sample, s, and pathway, PW, the Group Sample Score is defined as follows:
where the sum runs over all downstream target genes of the pathway.
The null hypothesis is again reshuffling of the genes from the microarray. We are interested in pathways both with high and low scores. Hence, we consider the p-values for the score being higher (p+) and lower (p-) than random, respectively, and the final p-value is given by two times the smaller of these two p-values:
p = 2·min(p+, p-).
The pathway is said to be active (+) if p+ <p-, and inactive (-) otherwise.
Family-wise p-value
If N independent hypotheses are tested simultaneously, the probability to obtain K or more p-values below q is given by a binomial distribution:
We refer to this probability as the family-wise p-value.
Authors' contributions
TB, MK, and CT implemented the algorithms employed. All authors contributed conceptually to the methods presented herein, as well as to the preparation of the manuscript.
Supplementary Material
Additional File 1
All additional files are in tab-delimited format. Tablel_Golub.csv Tablel_Sotirou.csv Table2_Golub.csv Table2_Sotirou.csv Table2_Veer.csv Table3_Golub.csv Table3_Sotirou.csv Table3_Veer.csv Table4_Golub.csv Table4_Sotirou.csv Contingency_Tables_Golub.csv Contingency_Tables_Sotirou.csv Contingency_Tables_Veer.csv
Click here for file
Acknowledgements
This work was in part supported by the Swedish Foundation for Strategic Research (TB and MK), the Swedish Research Council (CP) and the Swedish National Research School for Bioinformatics and Genoniics (CT).
==== Refs
Zeeberg BR Feng W Wang G Wang MD Fojo AT Sunshine M Narasimhan S Kane DW Reinhold WC Lababidi S Bussey KJ Riss J Barrett JC Weinstein JN GoMiner: a resource for biological interpretation of genomic and proteomic data Genome Biol 2003 4 R28 12702209 10.1186/gb-2003-4-4-r28
Berriz GF King OD Bryant B Sander C Roth FP Characterizing gene sets with FuncAssociate Bioinformatics 2003 19 2502 2504 14668247 10.1093/bioinformatics/btg363
Draghici S Khatri P Martins RP Ostermeier GC Krawetz SA Global functional profiling of gene expression Genomics 2003 81 98 104 12620386 10.1016/S0888-7543(02)00021-6
Mootha VK Lindgren CM Eriksson KF Subramanian A Sihag S Lehar J Puigserver P Carlsson E Ridderstrale M Laurila E Houstis N Daly MJ Patterson N Mesirov JP Golub TR Tamayo P Spiegelman B Lander ES Hirschhorn JN Altshuler D Groop LC PGC-lalpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes Nat Genet 2003 34 267 73 12808457 10.1038/ng1180
Zien A Kuffner R Zimmer R Lengauer T Analysis of gene expression data with pathway scores Proc Int Conf Intell Syst Mol Biol 2000 8 407 17 10977101
Rahnenführer J Domingues FS Maydt J Lengauer T Calculating the Statistical Significance of Changes in Pathway Activity From Gene Expression Data Statistical Applications in Genetics and Molecular Biology 2004 3
Pandey R Guru RK Mount DW Pathway Miner: extracting gene association networks from molecular pathways for predicting the biological significance of gene expression microarray data Bioinformatics 2004 20 2156 2158 15145817 10.1093/bioinformatics/bth215
Robinson MD Grigull J Mohammad N Hughes TR FunSpec: a web-based cluster interpreter for yeast BMC Bioinformatics 2002 3 35 12431279 10.1186/1471-2105-3-35
Segal E Shapira M Regev A Pe'er D Bothstein D Koller D Friedman N Module networks: identifying regulatory modules and their condition-specific regulators from gene expression data Nat Genet 2003 34 166 76 12740579
Gygi SP Rochon Y Franza BR Aebersold R Correlation between protein and mRNA abundance in yeast Mol Cell Biol 1999 19 1720 30 10022859
Gough NR Science's signal transduction knowledge environment: the connections maps database Ann N Y Acad Sci 2002 971 585 587 12438188
Krull M Voss N Choi C Pistor S Potapov A Wingender E TRANSPATH: an integrated database on signal transduction and a tool for array analysis Nucleic Acids Res 2003 31 97 100 12519957 10.1093/nar/gkg089
Wingender E Chen X Fricke E Geffers R Hehl R Liebich I Krull M Matys V Michael H Ohnhauser R Pruss M Schacherer F Thiele S Urbach S The TRANSFAC system on gene expression regulation Nucleic Acids Res 2001 29 281 283 11125113 10.1093/nar/29.1.281
Lin CY Strom A Vega VB Kong SL Yeo AL Thomsen JS Chan WC Doray B Bangarusamy DK Ramasamy A Vergara LA Tang S Chong A Bajic VB Miller LD Gustafsson JA Liu ET Discovery of estrogen receptor alpha target genes and response elements in breast tumor cells Genome Biol 2004 5 R66 15345050 10.1186/gb-2004-5-9-r66
van 't Veer LJ Dai H van de Vijver MJ He YD Hart AA Mao M Peterse HL van der Kooy K Marton MJ Witteveen AT Schreiber GJ Kerkhoven RM Roberts C Linsley PS Bernards R Friend SH Gene expression profiling predicts clinical outcome of breast cancer Nature 2002 415 530 536 11823860 10.1038/415530a
Sotiriou C Neo SY McShane LM Korn EL Long PM Jazaeri A Martiat P Fox SB Harris AL Liu ET Breast cancer classification and prognosis based on gene expression profiles from a population-based study Proc Natl Acad Sci U S A 2003 100 10393 10398 12917485 10.1073/pnas.1732912100
Golub TR Slonim DK Tamayo P Huard C Gaasenbeek M Mesirov JP Coller H Loh ML Downing JR Caligiuri MA Bloomfield CD Lander ES Molecular classification of cancer: class discovery and class prediction by gene expression monitoring Science 1999 286 531 537 10521349 10.1126/science.286.5439.531
Pikarsky E Porat RM Stein I Abramovitch R Amit S Kasem S Gutkovich-Pyest E Urieli-Shoval S Galun E Ben-Neriah Y NF-kappaB functions as a tumour promoter in inflammation-associated cancer Nature 2004 431 461 466 15329734 10.1038/nature02924
Greten FR Eckmann L Greten TF Park JM Li ZW Egan LJ Kagnoff MF Karin M IKKbeta links inflammation and tumorigenesis in a mouse model of colitis-associated cancer Cell 2004 118 285 96 15294155 10.1016/j.cell.2004.07.013
|
15987529
|
PMC1184060
|
CC BY
|
2021-01-04 16:27:25
|
no
|
BMC Bioinformatics. 2005 Jun 29; 6:163
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-163
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1761601417510.1186/1471-2105-6-176Research ArticleQuantitative analysis of EGR proteins binding to DNA: assessing additivity in both the binding site and the protein Liu Jiajian [email protected] Gary D [email protected] Department of Genetics, Washington University School of Medicine, 660 S Euclid, Box 8232, St. Louis, MO 63110, U.S.A2005 13 7 2005 6 176 176 13 4 2005 13 7 2005 Copyright © 2005 Liu and Stormo; licensee BioMed Central Ltd.2005Liu and Stormo; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Recognition codes for protein-DNA interactions typically assume that the interacting positions contribute additively to the binding energy. While this is known to not be precisely true, an additive model over the DNA positions can be a good approximation, at least for some proteins. Much less information is available about whether the protein positions contribute additively to the interaction.
Results
Using EGR zinc finger proteins, we measure the binding affinity of six different variants of the protein to each of six different variants of the consensus binding site. Both the protein and binding site variants include single and double mutations that allow us to assess how well additive models can account for the data. For each protein and DNA alone we find that additive models are good approximations, but over the combined set of data there are context effects that limit their accuracy. However, a small modification to the purely additive model, with only three additional parameters, improves the fit significantly.
Conclusion
The additive model holds very well for every DNA site and every protein included in this study, but clear context dependence in the interactions was detected. A simple modification to the independent model provides a better fit to the complete data.
==== Body
Background
Zinc finger proteins are the largest family of transcription factors in the human genome. The EGR sub-family of C2H2 zinc finger proteins has been extensively studied to determine the basis of DNA-protein binding specificity. The structure of the DNA-protein complex has been determined for the wild-type EGR1 (zif268) protein bound to its consensus site [1,2] and for several other variants of the interaction [3-5]. From the structure, the interaction appears very modular with each protein containing several zinc finger domains and each finger interacting with adjacent 3 base-pair (or overlapping 4 base-pair) segments of the binding site. Analysis of binding sites for this family of proteins suggested there were simple rules that relate the sequence of the zinc finger protein to its preferred binding site sequence [6], and that those rules could be used to design proteins with desired specificities [7,8]. Soon after, experimental techniques of in vitro randomization and selection were employed to greatly expand the collection of protein-DNA high affinity interactions [9-12]. Several reviews [4,13-18] have analyzed the protein-DNA crystal structures, summarized the results of the in vitro selection experiments, described rules for predicting high affinity protein-DNA interacting pairs and assessed the success of those rules for designing proteins to recognize particular sequences. Most of the recognition rules that have been developed are qualitative, specifying the amino acid and base-pair combinations that are preferred at each position in the binding sites [18]. Such rules can be effectively used to design proteins with preferred binding sites that are desired [19].
Despite the success of the qualitative recognition codes for designing proteins with desired preferred binding sites, the utility of such codes is still quite limited. If one compares the collection of known protein-DNA interacting pairs obtained in in vitro selection experiments, more than half of the fingers contain at least one amino acid/base-pair interaction that is not included in the code [20]. Furthermore, the code only predicts the preferred binding site for each protein sequence, or preferred protein for each DNA binding site. But it does not, by its qualitative nature, attempt to predict differences in affinities to similar sequences. Because all of these proteins bind with limited specificity, sites that are very similar to the preferred binding site can often bind with only slightly reduced affinity. Therefore predicting the quantitative binding specificities is important for a comprehensive view of their functions.
Several quantitative binding models have been developed, either specifically for the zinc finger proteins or for general protein-DNA interactions [20-26]. In many cases such codes can accurately predict the preferred binding sites as well as the qualitative codes, but the overall accuracy of the quantitative predictions is limited, undoubtedly for a combination of reasons. One reason is that there are limited data upon which to infer the model parameters using statistical approaches. Another reason is that many of the models are overly simplified, for instance assuming that each amino acid/base-pair contact is independent of any of the surrounding structure. We know, for instance, that the interactions of the protein and DNA are not completely additive [27,28], and it is also known that both intermolecular and intramolecular interactions contribute to protein-DNA recognition (24). But it has also been shown that models which are additive over the DNA positions can be a reasonably good approximations, at least for some proteins [29,30]. Most studies of additivity have focused on the DNA binding site, testing whether independent models for each base-pair fit the binding data well [29,31,32]. But equally important to the recognition codes is whether additivity holds within the protein. In one example from the EGR family, additivity within the protein was shown to be approximately additive (within 0.5 kcal) for one pair of mutated amino acids [33]. But very few studies have addressed the issue. Even though many variants of EGR family proteins have been used in SELEX and phage-display selection studies (see [20] for a summary), very few of the affinities have been quantified. Bulyk et al [28] did measure the affinity to each of 64 different binding sites for five different proteins, but the proteins were different at too many positions to be useful for determining additivity. One needs to have a set of single mutations and their double mutant combinations in order to determine whether the contributions to binding are independent or not. Several structural studies have highlighted the substantial rearrangements that can occur at the protein-DNA interface and can cause single amino acid or base-pair substitutions to influence the interactions at neighboring positions [3,15,34,35]. Such context effects may limit the predictive accuracy of simple recognition codes, although it is also possible that additivity can hold approximately even in the presence of such rearrangements. In the Mnt protein, a single amino acid change can alter the preferred binding site primarily at two adjacent positions, and more weakly over a longer distance [36,37]. Nevertheless, a complete quantitative analysis of the adjacent positions that were primarily affected showed that the interaction was largely additive for a wide variety of amino acid substitutions [30].
In this study we analyze the additivity of the interaction in both the DNA binding sites and in the interacting positions of the protein. We measure binding affinities for each of six different proteins, with single and double mutations compared to the wild-type protein, to each of six different DNA sites, also with single and double mutations from the wild-type binding site. We show that for any specific protein or DNA an additive model fits the data quite well. However, there are clear context effects such that no single interaction model fits all of the protein-DNA combinations. But only a small modification to the additive model, with just three additional parameters, improves the fit significantly.
Results and discussion
Figure 1 diagrams the direct interactions between the amino acids of finger 1 of the zif268 protein with the bases of the consensus binding site as determined by X-ray crystallography [1,2]. In order to study the additivity of the interaction on the side of protein, we constructed wild-type zif268 and five mutants where mutations occur in finger one. These five mutants include two single mutants of zif268 at position -1 in which arginine (R18) (referred to as RE) was replaced by glutamine (Q) (referred to as QE) and aspartic acid (D) (referred to as DE), separately, one single mutant at position +3 where glutamic acid (E21) was mutated to asparagine (N) (referred to as RN), and two corresponding double mutants (referred as to QN and DN, respectively). The six DNA sites used for this study were chosen primarily based on the qualitative code that represents the correlations between amino acids located at different positions and the DNA bases that they specify [4,15,34]. Specifically, the anticipated base specificity for amino acids arginine, glutamine and aspartic acid at position -1 are G, A and C at position 9 in the DNA sequence, respectively. The favorable bases for amino acids glutamic acid and asparagine at position +3 are C and A at position 8. The oligos used to generate the six DNA sites are shown in Table 1. They share common sequences except for the DNA bases that are recognized by the amino acids at the position of +3 and -1 of finger 1, referred as CG, CA, CC, AG, AA, and AC, respectively. We measured the affinity of each of six proteins to each of six DNA sites, and we use these data to analyze the additivity in both the protein and the DNA binding sites.
Figure 1 Amino acid-base contacts observed in co-crystal structures. The amino acid residues at -1, +2, +3, and +6 for zif268 are R, D, E and R, while the DNA bases at positions 7, 8, 9 and 10 for wild-type operator of zif268 are G, C, G and T.
Table 1 Oligos applied in this study. I: Synthesized DNA templates bearing either wild-type binding site (Zif_1) for zif268 or one of its variants (Zif_2 to Zif_6) used for generating DNA binding sites by PCR amplification, where KS-1 and SK-1 are two primers (low case). II: Oligos employed to construct five zif268 variants with QuickChange™ XL site-directed mutagenesis Kit (Stratagene) using pzif268 as a template.
I Zif_1 tcgaggtcgacggtatcGCGTGGGCGCtccactagttctagagcggccgccac
Zif_2 tcgaggtcgacggtatcGCGTGGGCACtccactagttctagagcggccgccac
Zif_3 tcgaggtcgacggtatcGCGTGGGCCCtccactagttctagagcggccgccac
Zif_4 tcgaggtcgacggtatcGCGTGGGAGCtccactagttctagagcggccgccac
Zif_5 tcgaggtcgacggtatcGCGTGGGAACtccactagttctagagcggccgccac
Zif_6 tcgaggtcgacggtatcGCGTGGGACCtccactagttctagagcggccgccac
KS-1 tcgaggtcgacggtatc
SK*-1 gtggcggccgctctagaact (SK-1 was fluorescent labeled with either FAM, HEX, TAMRA, ROX, or CY5)
II 18Q_plus 5' CGCCGCTTTTCTcagTCGGATGAGCTTACCCGCC
18Q_minus 5' GGCGGGTAAGCTCATCCGActgAGAAAAGCGGCG
18D_plus 5' CGCCGCTTTTCTgatTCGGATGAGCTTACCCGCC
18D_minus 5' GGCGGGTAAGCTCATCCGAatcAGAAAAGCGGCG
21N_plus 5' CGCCGCTTTTCTCGCTCGGATaacCTTACCCGCC
21N_minus 5' GGCGGGTAAGgttATCCGAGCGAGAAAAGCGGCG
18Q_21N_plus 5' CGCCGCTTTTCTcagTCGGATaacCTTACCCGCC
18Q_21N_minus 5' GGCGGGTAAGgttATCCGActgAGAAAAGCGGCG
18D_21N_plus 5' CGCCGCTTTTCTgatTCGGATaacCTTACCCGCC
18D_21N_minus 5' GGCGGGTAAGgttATCCGAatcAGAAAAGCGGCG
For each protein we determined the relative affinity of each different binding site compared to the wild type site (CG) using the QuMFRA assay (Table 2). For the wild-type protein, the relative affinities of CA, CC, and AG to the reference site CG in this study are 0.27, 0.082 and 0.15, respectively. These data are in good agreement with the relative affinities previously determined by Miller and Pabo (0.21, 0.ll and 0.20, respectively [34]). Table 2 shows only the wild-type protein (RE) binds preferentially to the wild-type binding site (CG), all of the other proteins preferring a different binding site sequence. The range of affinities varies considerably between the different proteins. RE has about a 25-fold difference between the highest and lowest sites, while QE only varies by about 2-fold between the highest and lowest. We also measured the absolute binding affinity of each protein to one of the DNA binding sites with a Scatchard analysis (Table 3). The Kd for wildtype zif268 binding to the DNA site CC is 3.0 × 10-8 M, which converts to a Kd for wildtype binding site CG of 2.5 × 10-9 M. This value is almost the same as that determined by Hamilton et al (2.2 × 10-9 M) [41] (previously reported values for this Kd range from 0.04 to 6.5 nM, depending on the binding condition used [33]). No similar data exist for the other proteins in our collection. Combining the data from Tables 2 and 3, we derive the association constant of each protein for each different DNA sequence, which differ by over 300-fold between the highest and lowest affinities (Table 4).
Table 2 Relative binding constants for six DNA binding sites for wild-type of zif268 and its 5 derivatives, where wild-type operator of zif268 was used as the reference. Each data were obtained from 5 or more independent examinations, inside of parenthesis are the standard deviations.
DNA\Prot RE(wt) QE DE RN QN DN
CG(wt) 1 1 1 1 1 1
CA 0.27(0.06) 1.50(0.54) 1.16(0.49) 0.36(0.14) 0.49(0.19) 1.21(0.33)
CC 0.082(0.076) 2.17(0.91) 1.91(0.83) 0.41(0.23) 0.53(0.36) 2.61(0.59)
AG 0.15(0.10) 1.30(0.34) 1.48(0.56) 1.29(0.28) 4.45(2.64) 14.5(5.18)
AA 0.064(0.017) 1.36(0.48) 2.25(1.30) 0.68(0.28) 2.47(1.34) 4.02(1.56)
AC 0.041(0.045) 1.93(1.01) 3.08(0.45) 0.94(0.26) 2.78(0.80) 11.8(4.44)
Table 3 Experimental determined association constants (106M-1) for individual indicated DNA binding site binding to its corresponding protein. Each value is the mean from 5 or more independent determinations and the standard deviations are shown in parenthesis.
DNA\Prot RE(wt) QE DE RN QN DN
CC 33(7) 6.4(1.7) 4.7(2.6) 33(14)
AG 33(18) 17(6)
Table 4 Absolute Ka(106M-1) for six DNA binding sites and six variants of zif268, derived from the combination of Table 2 and Table 3.
DNA\Prot RE QE DE RN QN DN
CG 406 3.0 2.5 81 7.4 1.2
CA 109 4.5 2.8 30 3.5 1.4
CC 33 6.4 4.7 33 3.9 3.1
AG 63 3.9 3.6 105 33 17
AA 26 4.0 3.7 56 18 4.8
AC 16 5.7 5.5 77 21 14
From the binding data we can assess the additivity of the interaction for both the protein and the DNA. In a perfectly additive interaction the binding energy for each sequence would be the sum of the independent contributions at each position. For example, for any protein j, the binding energy to any DNA sequence XY, would be the sum of the interactions with base X and base Y:
ΔGj(X8Y9) = ΔGj(X8) + ΔGj(Y9). (1)
The important assumption of the additive model is that the interaction energy at position 8, for example, doesn't depend on which base occurs at position 9. We do not expect additivity to hold precisely [30,27,28], but it can be a very good approximation, at least for some proteins [27,29]. Previously, studies of additivity have focused on whether the positions in the DNA binding site contribute independently to the binding of a particular protein. Using the data of Table 4 we can also determine whether the positions in the protein contribute additively to the binding of a particular DNA site. That is, we can reverse the symbols of equation 1 to refer to the binding of a particular DNA sequence, i, to a protein sequence UV:
ΔGi(U-1V3) = ΔGi(U-1) + ΔGi(V3). (2)
Of course, we have not measured affinities to all possible DNA sequences or for all possible protein sequences, but because we have both single and double mutants in both the protein and the DNA, and have measured the binding affinities of all combinations, we can determine how well additivity holds on both sides, the DNA and the protein, at least for this limited set of variants.
We cannot actually measure the binding affinities to single positions because they always occur in some context. But we can find the "best fit" values for the independent interactions, and then determine how well the total data fits the additive model using those values. One method to obtain the best fit independent parameters is to apply multiple linear regression to the total data [31,32]. However, we have argued previously [29] that a better criterion is to minimize the difference in total free energy between the observed data and the model.
The and values are those obtained as the best fit parameters (those which minimize M) for each position assuming independence. The ω refers to either the protein or the DNA, and α,β refer to the residues at the two interacting positions. The first term inside the sum represents the probability that each particular residue sequence will be bound, and so weights the energy differences by their contribution to the total free energy of the system. As can be seen in the last form of the equation, M is the "mutual information" between the positions, the amount of total information content in the data that cannot be explained by the best independent model. We use log2 so that the mutual information is measured in bits.
Given the best fit independent parameters we can calculate the specificity information, Ispec, of each position independently [42]. For example the specificity information for the protein or DNA ω at the first interacting position is
Ispec measures the amount of specificity in the interaction in bits; any non-specific protein or DNA would have Ispec = 0. Figure 2 shows sequence logos [43] for each of the six proteins and the six DNA sequences for which we have measured the affinity. We have added the symbol "M" to each one which shows the amount of mutual information in each interaction [44,27,30]. That is the amount of total free energy, or specificity information, which is not captured by the best fit additive model. Half of the total mutual information is displayed above each position.
Figure 2 Sequence logos for each of six zinc finger proteins and the six DNA sites for which we have measured the affinity. M in each logo is the mutual information content in each interaction. The label at the top of each logo represents the DNA site (for the top two rows) or the protein (bottom two rows). The amino acid order is reversed so that they are lined up with the bases they contact. For example, the logo labeled "ER" shows the specificity for the RE (wild type) protein. In the lower six panels the maximum value on the y-axis is 0.5 bits.
Several interesting results are evident in Figure 2. As stated above, the proteins vary considerably in their specificity, with RE (shown as "ER" in the figure) showing large discrimination between the different DNA sites, whereas QE and DE are fairly non-specific. The same holds for the different DNA sites, where CG is much more specific than CC or AC. It is interesting that every DNA site prefers R at position -1 of the protein, showing that it contributes to the total affinity of each protein as well as to the specificity of some proteins. The small degree of mutual information, the "M" in each logo, means that every interaction fits well with an additive model. Not only do the DNA positions contribute very additively, as has been shown previously for this family of proteins [29], but the contributions of the amino acids in the protein are also largely additive. The conclusion that additive models are good approximations to the true data holds for every DNA site and every protein included in the analysis. However, it is also true that there is not a single set of additive parameters that fit well for every case. This is consistent with the context effects previously noted for this family [15,34]. For example, R prefers to bind to G over A or C, but the magnitude of that preference is much larger if position +3 is an E instead of N. And an N at position +3 always prefers an A over C in the binding site, but that preference is much weaker with an R at position -1 than with a Q or D. Similarly, E at position +3 prefers a C very strongly in the context of an R, but is quite non-specific with either a Q or D at position -1. Similar effects, but of smaller magnitude, can be seen in the context effects of the DNA sites. These results show that additive models can be good approximations not only for the DNA sites in binding to any particular protein as has been seen before [29], but also for the proteins in binding to any particular DNA site. But the results also show that additivity for specific proteins and DNA sites is not sufficient to generate a general recognition code because context effects can still be important when both the DNA and protein can be variable. The small amounts of mutual information observed for any specific protein or DNA can be reinforced to give much larger amounts when measured over combinations of both components.
To get a more detailed view of the dependencies in the data, it is useful to reformat it as in Figure 3A. Those data are the same as in Table 4 except that it has been normalized to a sum of 1000. In an experiment where every protein and DNA was equally available for binding, those elements in the table are 1000-times the probability of picking that particular combination from all of those in the bound state. The data are arranged in a four-dimensional (4D) table, with one dimension for each of the two positions in the protein and the two positions in the DNA. For example, the 335 at the RE-CG element of the table corresponds to the wild-type association constant of 406 from Table 4 after normalization. From the data in Figure 3A it is easy to obtain different lower dimensional views by summing over the other dimensions. For example, Figure 3B shows the 2D view of the interaction of the amino acid at position -1 with the base-pair at position 9 obtained by summing over all of the combinations of E,N at protein position +3 and C, A at binding site position 8 (inside the bold lines of Figure 3A). Similarly, Figure 3C shows a 2D view of the interaction between the amino acid at position +3 and the binding site position 8. Those two 2D views are orthogonal and together cover the 4D space of Figure 3A. We also show the remaining 2D views in Figures 3D–G. The pairs in Figure 3D,E and 3F,G are also orthogonal and together cover the 4D space of the data. If the binding interaction was completely additive, the true data of 3A could be calculated as the (renormalized) outer product of any pair of orthogonal matrices. Such predictions are not too bad, but demonstrate limitations of the additive model (see below).
Figure 3 DNA binding specificities for six DNA sites for zif268 and its five derivatives. A: four-dimensional table representing binding specificities for all DNA sites and zinc finger proteins in this study. It is converted from Table 4 by normalization to a sum of 1000; B: 2D table of combinations for the interaction of the amino acid at position -1 with the base-pair at position 9; C: 2D table of combinations for the interaction of the amino acid at position +3 with the binding site position 8; D: 2D table of combinations between amino acids at position -1 and +3; E: 2D table of combinations between DNA bases at position 8 and 9; F: 2D table of combinations between amino acid position 3 and base position 9; G: 2D table of combinations between amino acid position -1 and base position 8.
Because the data in Figure 3 are in probabilities (if divided by 1000), the information specificity can be calculated more easily than in equation (4):
Ispec(α) = log2Nα - Hα (5)
where α is any of the positions or combination of positions, Hα is the Shannon entropy of the data at those positions and Nα is the number of entries in the data. For example, position -1 of the protein has three entries, R, Q and D, with overall probabilities of 0.852, 0.093 and 0.054, respectively, which gives Ispec(- 1) = 0.84 bits. The upper half of Table 5 shows the specificity information for each of the positions (along the diagonal) as well as the specificity information for each of the pairs of positions (from the data shown in Figure 3). If the two positions contribute independently to the total specificity then the information for the paired positions is just the sum of the information at the each position. In this case the mutual information between the positions is the amount of information in the pair that exceeds the sum of the individual positions:
Table 5 Information for the position dependence. The diagonal is the specificity information for each of positions -1, 3, 8, and 9. The upper half of the matrix is the specificity information for each of the pairs of positions, and the lower half is the mutual information between pairs of positions.
Position -1 3 8 9
-1 0.84 0.91 0.94 1.09
3 0.05 0.02 0.24 0.28
8 0.06 0.19 0.03 0.29
9 0.02 0.04 0.04 0.22
M(α,β) = Ispec(α,β) - (Ispec(α) + Ispec(β)) (6)
Those values are shown in the lower half of Table 5. From the standard model of interaction between the DNA and protein we would expect there to be very little mutual information for any of the 2D datasets of Figure 3D–G, and that expectation is met. But we do expect high mutual information for the datasets in Figure 3B and 3C because those are the interacting positions. Just as we get high mutual information for positions that interact in RNA structures [44], we expect to see compensating changes between the amino acids and base-pairs that interact. That expectation is met for the combination of protein position +3 and base-pair position 8 (Figure 3C) where there is a clear preference for E binding to C and for N binding to A. In that case the mutual information is 0.19 bits, which is the main contribution to the total information of that pair, 0.24 bits. However, protein position -1 and base-pair position 9 also interact but show little mutual information because R is the preferred amino acid for each different DNA sequence and G is the preferred base-pair for each different protein. That pair has high specificity information, 1.09 bits, but it is very additive with only 0.02 bits of mutual information.
The total specificity information in the complete data of Figure 3A is 1.46 bits. The sum of the information for the interacting pairs, -1,9 and 3,8, is 1.33 bits, which shows that the complete specificity is reasonably well fit by assuming independent contributions from those interacting positions, as in most recognition code models [18]. If one predicts the complete data of Figure 3A as the outer-product of the matrices of Figure 3B and 3C (not shown), the correlation coefficient between the observed and predicted binding energies is 0.87 (Model 1 of Figure 5), similar to what had been observed previously for data in which only the DNA site had been varied [29]. While that result is reasonably good overall, examination of the complete data in Figure 3A identifies one clear source of context dependence between the interacting positions. When protein position -1 is R and the base-pair at position 9 is either G or A, there is a clear preference for the specific combination of E with C and a weak preference for N with A. But for all other combinations of positions -1 and 9, there is a strong preference for N with A, but very little preference for E. That is, the preference of E for C depends on the R with G or A combination being adjacent. In the structure of zif268 with the wild-type DNA there is no hydrogen bound between the position +3 E and the C base-pair, but rather it interacts with the backbone and with the neighboring R amino acid [2,1]. Various qualitative codes for the interactions of this protein family do not include E as an acceptable amino acid at position +3 [4,15]. But in the compilation of SELEX and phage-display results used by Benos et al [20], the combination of RE-CG was much more frequent than expected from the individual or pair occurrences (p-value less than 0.001). That is consistent with our result that in general E contributes little to the specificity of the binding site at position 8 except in the case where the adjacent interaction is R with G or A. Such context dependencies are not included in the simple recognition code models, but we can easily add that to the basic model. In Figure 4 we show two different specificity tables for the interaction of positions +3 and 8. Figure 4A represents the general case, and Figure 4B is for the special case of R with G or A at positions -1 and 9. If we now predict the complete data using these models, combined with the general model for positions -1 and 9 in Figure 3B, we obtain the values shown in Figure 4C. The specificity information of this data is 1.44 bits, showing that it models quite accurately the complete data. The correlation coefficient for those predicted binding energies with the measured energies is 0.96, a significant improvement over the model without the context dependent parameters (Model 2 of Figure 5). This improvement is at the cost of only three additional parameters due to the separation into two distinct classes depending on whether or not position -1 is an R that interacts with G or A. The completely additive model has 8 free parameters for the interaction of positions -1 and 9 (the 9 values in Figure 3B minus 1 for the total fixed sum) and 3 free parameters for the interaction of positions +3 and 8 (from the 4 values in Figure 3C). By separating the matrix of Figure 3C into two separate cases, shown in Figure 4A,B, we need 3 additional parameters in the model, for a total of 14. The model is used to predict data with 35 free values (the 36 elements of Figure 3A minus 1 for the fixed sum), so the additional parameters are only a small reduction in the degrees of freedom remaining to assess the fitness of the model.
Figure 5 Scatter plot of the observed (Figure 3A) and predicted binding probabilities. Model2 is the two component model, so those points show the fit between Figure 3A and Figure 4C. Model1 is for the single component model obtained from the outer product of Figure 3B and Figure 3C (table of predicted probabilities not shown).
Figure 4 DNA binding specificities with the two component model. A: The 2D table of interactions for amino acid position 3 with base position 8 obtained from the data in Figure 3A for all cases except R with G or A (and normalized to a sum of 1000). B: The 2D table of interactions for amino acid position 3 with base position 8 for the cases with R and G or A (normalized to 1000). C: The predicted binding probabilities for the entire dataset using the two component model. The elements for the cases of R with G or A are obtained by the outer product of the matrix from B with the R/G,A elements of the matrix in Figure 3B. The rest of the elements are obtained from the outer product of A with the remaining elements of the matrix from Figure 3B.
The EGR family of proteins is an ideal case to study the effectiveness of a recognition code for protein-DNA interactions. The collection of crystal structures along with a large number of examples from selection experiments provides a wealth of information for determining the relationship between the protein sequence and the affinity for different DNA sequences. Simple qualitative models that predict the preferred interactions can be very effective and useful for designing new TFs [14,19]. Quantitative models, that predict relative binding affinities to multiple DNA sites, are more challenging but some success has been achieved by statistical approaches as well as by structure based approaches [20-26]. Most current models of this type assume independence of the contributions to binding between the positions in the interactions. In this work we show that additive models can be a good approximation for any particular EGR protein and also for binding to any particular DNA site; additivity holds well for both the DNA and protein side of the interaction. But we also show that there is not a universal set of parameters that work for all proteins or all DNA sites, rather there is context dependence in the interactions. However, at least in the cases studied here, a simple addition to the independent model that divides sites into two classes provides a much better fit. This holds promise that, even though additivity does not hold precisely, it may still be possible to determine an additive recognition code by identifying a small set of classes that cover the entire set of interactions. How many classes will be needed is unknown at this time. The 36 combinations in our study required only two classes to give a very good fit but this is still far from a comprehensive analysis. The total number of adjacent amino acid pairs is 400 and the number of di-nucleotide combinations is 16, so there are 6400 possible combinations of the two. Quantitative analyses that cover all possible combinations of even a single zinc finger are impossible at this time. But more thorough sampling of the space of high affinity interactions, followed by quantitative binding assays, will provide much valuable information regarding the nature of recognition codes. While a completely additive model for the interaction of the protein and DNA is not correct, it may be that only relatively minor modifications are needed to make significantly better predictions.
Conclusion
By determining the binding affinities of single and double mutants in both the DNA binding site and in the protein we were able to assess the degree of additivity in both halves of the interaction. Although only a limited number of combinations were tested, we find that for every DNA sequence and for every protein sequence an additive model is a good approximation to the real binding data. However, when all of the data are considered together there are clear context effects that are not well fit by a single additive model. A slightly more complex model does provide a good fit to the observed data, suggesting that quite simple may still be employed to predict quantitative binding interactions of proteins with DNA. Further data are needed to determine how well these findings generalize to more variations and to other protein families.
Methods
Construction of wild-type zif268 DNA binding domain (DBD) and its variants
A plasmid containing the DNA binding domain of wild-type zif268 was obtained from Gendaq Limited [38]. The portion of zif268 cDNA encoding the three zinc-finger DBD (cDNA nucleotides 996–1262, amino acids 331–420) was amplified by PCR and subcloned into expression vector pET-28a-c(+) (Novagen) to create His-tagged fusion protein. The resulting construct, denoted pzif268, was verified by DNA sequencing. Five zif268 mutants with alterations in the base-contacting residues in finger one of zif268 DBD were constructed with QuikChange™ XL site-directed mutagenesis Kit (Stratagene) using pzif268 as a template: 3 single substitution mutants R18Q, R18D, E21N, and two double substitution mutants R18Q/E21N and R18D/E21N. The mutagenic primers containing the desired mutations used to create the five mutants are shown in Table 1. The resulting plasmids p18Q, p18D, p21N, p18Q21N and p18D21N were verified by DNA sequencing. Hereafter, the proteins are referred to by their amino acids at positions -1 and +3: RE (wild-type), QE, DE, RN, QN and DN.
Expression and purification of His-tagged-zif268 fusion protein and its variants
E. coli BL21 cells bearing pzif268 or one of its derivatives were grown in 2xYT medium at 37°C with constant shaking at 250 rpm. IPTG was added to a final concentration of 1 mM when OD600 reached 0.6–1.0. Cells were harvested 3 hrs after IPTG induction by centrifugation at 4000 rpm for 20 min. The pellets were then resuspended in 15 ml of lysis buffer (50 mM Tris-HCl pH 8.0, 300 mM NaCl, 10 mM DDT and 1 tablet of protease inhibitor cocktail tablets (Roche) and lysed with sonication. The pellets were then separated by centrifugation at 6000 rpm for 20 min and insoluble material removed. The His-tagged fusion protein was purified with Ni-resin chromatography similar to those described previously [39]. The elutions were collected as 2 ml fractions. Fractions were analyzed on 12% SDS-PAGE gel, followed by silver staining. Finally the fractions were pooled and dialysed against dialysis buffer (30 mM Tris-HCl pH 8.0, 50 mM NaCl, 3 mM DTT) at 4°C, followed by concentration with a Centricon filter (Amicon) and kept at -80°C until usage. The protein concentration was determined with BioRad assay kit.
Multiple quantitative fluorescence relative affinity (QuMFRA) assay to determine the relative binding constants
The relative binding constants of each protein to different binding sites were determined by the QuMFRA assay [27] with some modifications. Double-strand oligonucleotide binding sites used in this study were generated by PCR reactions. In each PCR reaction, a synthesized oligo containing either the wild-type binding site (zif1) of zif268 or one of its variants (Table 1) was used as template and the two primers are KS and SK (Table 1). The SK primer was labeled with one of the following four fluorophores: FAM, HEX, TAMRA, or ROX [27]. The PCR products were dissolved in TS buffer (10 mM Tris-HCl pH 8.0, 50 mM NaCl) after purification and precipitated with 1/10 vol of 3M NaAc and equal volume of isopropanol. The concentration of DNA was determined using a method similar to those as described previously [40].
The competitive binding assay [27] was performed by mixing 4 different fluorophore-labeled DNA binding sites with a certain amount of His-tagged zinc finger protein in 1x reaction buffer (30 mM Tris-HCl pH 8.0, 50 mM NaCl, 0.1 mg/ml BSA, 3 mM DTT, 20 uM ZnSO4, polydI-dC 5 ug/ml), in which the fluorophore-labeled zif1 served as an internal reference in each reaction. The reaction was equilibrated for 1 hr on ice before being electrophoresed on a 10% polyacrylamide gel. Each of 4 fluorophore-labeled PCR products was also loaded individually onto the same gel. After electrophoresis, the gels were scanned by a Typhoon Variable Scanner (Molecular Dynamics, Sunnyvale, CA) to obtain the fluorescent intensities of the separated bands (bound and unbound) at 4 different emission wavelengths using the same machine settings as employed by Man and Stormo [27]. For each separated band, the resultant fluorescence intensities at four emission wavelengths make up the output vector . Using the fluorescence intensities of the 4 individual fluorophore-labeled DNA at each emission wavelength we obtain the emission matrix E [27]. The input mixture of the 4 DNAs in each band, represented as the vector , were computed by a program developed for this study using the Gaussian elimination algorithm from the following relationship:
From the amount of each DNA in the bound and unbound bands of each lane, the relative binding affinity can be calculated by the following formula, where the wild-type binding site of zif268 (zif1) serves as the reference:
Kb test/Kb ref = [P·D]test[D]ref/[D]test[P·D]ref
Kb test/Kb ref = IP-DtestIDref/IDtestIP-Dref
where IP-D and ID are the intensities of the specified DNAs in the bound and unbound bands, respectively.
Determination of the absolute binding constant of a zinc finger protein to a binding site by Scatchard analysis
Scatchard analysis [41] was applied here to examine the absolute association constant, Ka, of a zinc finger protein to a binding site. Specifically, a fixed amount of purified His-tagged zinc finger protein, [P]total, was mixed with increasing Cy5-labeled DNA generated by PCR reactions in 1x reaction buffer for 1 hr on ice. The bound and unbound DNA were separated by electrophoresis on a10% polyacrylamide gel, as above, and the gels were scanned by a Typhoon Variable Scanner using the excitation wavelength of 633 nm and emission wavelength of 670 nm. From the following relationship
it can be seen that the association constant for the particular combination of protein and DNA, Ka(P,D), can be obtained from a plot of at multiple DNA concentrations. At least five independent determinations were made for each protein.
Authors' contributions
JL performed all of the experiments, which GS helped to design. Both authors contributed to the analysis of the data and the writing of the paper.
Acknowledgements
We thank Gendaq for giving us DNA phage coding for zif268. We thank Takis Benos for help with subcloning and David Granas for some statistical analyses of the SELEX and phage-display data. This work was supported by NIH grant GM28755.
==== Refs
Pavletich NP Pabo CO Zinc finger-DNA recognition: crystal structure of a Zif268-DNA complex at 2.1 A Science 1991 252 809 17 2028256
Elrod-Erickson M Rould MA Nekludova L Pabo CO Zif268 protein-DNA complex refined at 1.6 A: a model system for understanding zinc finger-DNA interactions Structure 1996 4 1171 80 8939742 10.1016/S0969-2126(96)00125-6
Elrod-Erickson M Benson TE Pabo CO High-resolution structures of variant Zif268-DNA complexes: implications for understanding zinc finger-DNA recognition Structure 1998 6 451 64 9562555 10.1016/S0969-2126(98)00047-1
Choo Y Klug A Physical basis of a protein-DNA recognition code Curr Opin Struct Biol 1997 7 117 25 9032060 10.1016/S0959-440X(97)80015-2
Wolfe SA Nekludova L Pabo CO DNA recognition by Cys2His2 zinc finger proteins Annu Rev Biophys Biomol Struct 2000 29 183 212 10940247 10.1146/annurev.biophys.29.1.183
Desjarlais JR Berg JM Toward rules relating zinc finger protein sequences and DNA binding site preferences Proc Natl Acad Sci U S A 1992 89 7345 9 1502144
Desjarlais JR Berg JM Redesigning the DNA-binding specificity of a zinc finger protein: a data base-guided approach Proteins 1992 12 101 4 1603798 10.1002/prot.340120202
Desjarlais JR Berg JM Use of a zinc-finger consensus sequence framework and specificity rules to design specific DNA binding proteins Proc Natl Acad Sci U S A 1993 90 2256 60 8460130
Choo Y Klug A Selection of DNA binding sites for zinc fingers using rationally randomized DNA reveals coded interactions Proc Natl Acad Sci U S A 1994 91 11168 72 7972028
Choo Y Klug A Toward a code for the interactions of zinc fingers with DNA: selection of randomized fingers displayed on phage Proc Natl Acad Sci U S A 1994 91 11163 7 7972027
Desjarlais JR Berg JM Length-encoded multiplex binding site determination: application to zinc finger proteins Proc Natl Acad Sci U S A 1994 91 11099 103 7972017
Rebar EJ Pabo CO Zinc finger phage: affinity selection of fingers with new DNA-binding specificities Science 1994 263 671 3 8303274
Nagaoka M Sugiura Y Artificial zinc finger peptides: creation, DNA recognition, and gene regulation J Inorg Biochem 2000 82 57 63 11132639 10.1016/S0162-0134(00)00154-9
Pabo CO Peisach E Grant RA Design and selection of novel Cys2His2 zinc finger proteins Annu Rev Biochem 2001 70 313 40 11395410 10.1146/annurev.biochem.70.1.313
Wolfe SA Greisman HA Ramm EI Pabo CO Analysis of zinc fingers optimized via phage display: evaluating the utility of a recognition code J Mol Biol 1999 285 1917 34 9925775 10.1006/jmbi.1998.2421
Suzuki M Gerstein M Yagi N Stereochemical basis of DNA recognition by Zn fingers Nucleic Acids Res 1994 22 3397 405 8078776
Pabo CO Nekludova L Geometric analysis and comparison of protein-DNA interfaces: why is there no simple code for recognition? J Mol Biol 2000 301 597 624 10966773 10.1006/jmbi.2000.3918
Benos PV Lapedes AS Stormo GD Is there a code for protein-DNA recognition? Probab(ilistical)ly.. Bioessays 2002 24 466 75 12001270 10.1002/bies.10073
Liu Q Xia Z Zhong X Case CC Validated zinc finger protein designs for all 16 GNN DNA triplet targets J Biol Chem 2002 277 3850 6 11726671 10.1074/jbc.M110669200
Benos PV Lapedes AS Stormo GD Probabilistic code for DNA recognition by proteins of the EGR family J Mol Biol 2002 323 701 27 12419259 10.1016/S0022-2836(02)00917-8
Paillard G Lavery R Analyzing protein-DNA recognition mechanisms Structure (Camb) 2004 12 113 22 14725771 10.1016/j.str.2003.11.022
Suzuki M Brenner SE Gerstein M Yagi N DNA recognition code of transcription factors Protein Eng 1995 8 319 28 7567917
Mandel-Gutfreund Y Margalit H Quantitative parameters for amino acid-base interaction: implications for prediction of protein-DNA binding sites Nucleic Acids Res 1998 26 2306 12 9580679 10.1093/nar/26.10.2306
Gromiha M Siebers JG Selvaraj S Kono H Sarai A Intermolecular and intramolecular readout mechanisms in protein-DNA recognition J Mol Biol 2004 337 285 94 15003447 10.1016/j.jmb.2004.01.033
Kono H Sarai A Structure-based prediction of DNA target sites by regulatory proteins Proteins 1999 35 114 31 10090291 10.1002/(SICI)1097-0134(19990401)35:1<114::AID-PROT11>3.0.CO;2-T
Yoshida T Nishimura T Aida M Pichierri F Gromiha MM Sarai A Evaluation of free energy landscape for base-amino acid interactions using ab initio force field and extensive sampling Biopolymers 2002 61 84 95 11891631 10.1002/1097-0282(2001)61:1<84::AID-BIP10045>3.0.CO;2-X
Man TK Stormo GD Non-independence of Mnt repressor-operator interaction determined by a new quantitative multiple fluorescence relative affinity (QuMFRA) assay Nucleic Acids Res 2001 29 2471 8 11410653 10.1093/nar/29.12.2471
Bulyk ML Johnson PL Church GM Nucleotides of transcription factor binding sites exert interdependent effects on the binding affinities of transcription factors Nucleic Acids Res 2002 30 1255 61 11861919 10.1093/nar/30.5.1255
Benos PV Bulyk ML Stormo GD Additivity in protein-DNA interactions: how good an approximation is it? Nucleic Acids Res 2002 30 4442 51 12384591 10.1093/nar/gkf578
Man TK Yang JS Stormo GD Quantitative modeling of DNA-protein interactions: effects of amino acid substitutions on binding specificity of the Mnt repressor Nucleic Acids Res 2004 32 4026 32 15289576 10.1093/nar/gkh729
Lee ML Bulyk ML Whitmore GA Church GM A statistical model for investigating binding probabilities of DNA nucleotide sequences using microarrays Biometrics 2002 58 981 8 12495153 10.1111/j.0006-341X.2002.00981.x
Stormo GD Schneider TD Gold L Quantitative analysis of the relationship between nucleotide sequence and functional activity Nucleic Acids Res 1986 14 6661 79 3092188
Elrod-Erickson M Pabo CO Binding studies with mutants of Zif268. Contribution of individual side chains to binding affinity and specificity in the Zif268 zinc finger-DNA complex J Biol Chem 1999 274 19281 5 10383437 10.1074/jbc.274.27.19281
Miller JC Pabo CO Rearrangement of side-chains in a Zif268 mutant highlights the complexities of zinc finger-DNA recognition J Mol Biol 2001 313 309 15 11800559 10.1006/jmbi.2001.4975
Wolfe SA Grant RA Elrod-Erickson M Pabo CO Beyond the "recognition code": structures of two Cys2His2 zinc finger/TATA box complexes Structure (Camb) 2001 9 717 23 11587646 10.1016/S0969-2126(01)00632-3
Raumann BE Knight KL Sauer RT Dramatic changes in DNA-binding specificity caused by single residue substitutions in an Arc/Mnt hybrid repressor Nat Struct Biol 1995 2 1115 22 8846224 10.1038/nsb1295-1115
Silbaq FS Ruttenberg SE Stormo GD Specificity of Mnt 'master residue' obtained from in vivo and in vitro selections Nucleic Acids Res 2002 30 5539 48 12490722 10.1093/nar/gkf684
Isalan M Choo Y Rapid, high-throughput engineering of sequence-specific zinc finger DNA-binding proteins Methods Enzymol 2001 340 593 609 11494872
Liu J Zuber P The ClpX protein of Bacillus subtilis indirectly influences RNA polymerase holoenzyme composition and directly stimulates sigma-dependent transcription Mol Microbiol 2000 37 885 97 10972809 10.1046/j.1365-2958.2000.02053.x
Teare JM Islam R Flanagan R Gallagher S Davies MG Grabau C Measurement of nucleic acid concentrations using the DyNA Quant and the GeneQuant Biotechniques 1997 22 1170 4 9187773
Hamilton TB Borel F Romaniuk PJ Comparison of the DNA binding characteristics of the related zinc finger proteins WT1 and EGR1 Biochemistry 1998 37 2051 8 9485332 10.1021/bi9717993
Stormo GD Fields DS Specificity, free energy and information content in protein-DNA interactions Trends Biochem Sci 1998 23 109 13 9581503 10.1016/S0968-0004(98)01187-6
Schneider TD Stephens RM Sequence logos: a new way to display consensus sequences Nucleic Acids Res 1990 18 6097 100 2172928
Gorodkin J Heyer LJ Brunak S Stormo GD Displaying the information contents of structural RNA alignments: the structure logos Comput Appl Biosci 1997 13 583 6 9475985
|
16014175
|
PMC1184061
|
CC BY
|
2021-01-04 16:27:23
|
no
|
BMC Bioinformatics. 2005 Jul 13; 6:176
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-176
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-661598241910.1186/1471-2407-5-66Research ArticleDifferential gene expression profile reveals deregulation of pregnancy specific β1 glycoprotein 9 early during colorectal carcinogenesis Salahshor Sima [email protected] Jason [email protected] Runjan [email protected] Steven [email protected] James R [email protected] Ontario Cancer Institute, Department of Medical Biophysics, 610 University Ave., Toronto, Ontario, M5G 2M9, Canada2 Princess Margaret Hospital, Department of Laboratory Medicine and Pathobiology, University of Toronto, Canada3 Mount Sinai Hospital, Department of Surgery, Centre for Cancer Genetics, Samuel Lunenfeld Research Institute, Toronto, Canada2005 27 6 2005 5 66 66 21 2 2005 27 6 2005 Copyright © 2005 Salahshor et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
APC (Adenomatous polyposis coli) plays an important role in the pathogenesis of both familial and sporadic colorectal cancer. Patients carrying germline APC mutations develop multiple colonic adenomas at younger age and higher frequency than non-carrier cases which indicates that silencing of one APC allele may be sufficient to initiate the transformation process.
Methods
To elucidate the biological dysregulation underlying adenoma formation we examined global gene expression profiles of adenomas and corresponding normal mucosa from an FAP patient. Differential expression of the most significant gene identified in this study was further validated by mRNA in situ hybridization, reverse transcriptase PCR and Northern blotting in different sets of adenomas, tumours and cancer cell lines.
Results
Eighty four genes were differentially expressed between all adenomas and corresponding normal mucosa, while only seven genes showed differential expression within the adenomas. The first group included pregnancy specific β-1 glycoprotein 9 (PSG9) (p < 0.006). PSG9 is a member of the carcinoembryonic antigen (CEA)/PSG family and is produced at high levels during pregnancy, mainly by syncytiotrophoblasts. Further analysis of sporadic and familial colorectal cancer confirmed that PSG9 is ectopically upregulated in vivo by cancer cells. In total, deregulation of PSG9 mRNA was detected in 78% (14/18) of FAP adenomas and 75% (45/60) of sporadic colorectal cancer cases tested.
Conclusion
Detection of PSG9 expression in adenomas, and at higher levels in FAP cases, indicates that germline APC mutations and defects in Wnt signalling modulate PSG9 expression. Since PSG9 is not found in the non-pregnant adult except in association with cancer, and it appears to be an early molecular event associated with colorectal cancer monitoring of its expression may be useful as a biomarker for the early detection of this disease.
==== Body
Background
FAP is characterized by the development of hundreds to thousands of adenomas throughout the entire colon and rectum which, if left untreated, progress to colorectal cancer [1,2]. FAP, an inherited tumour predisposition, is caused by mutant alleles of the adenomatous polyposis coli (APC) gene and provides an opportunity to define critical early genetic events in the development of tumours [3]. Early development of a large number of colon adenomas in this disorder indicates that mutations in the APC gene can be rate-limiting in adenoma development. The majority of colorectal tumours are sporadic in origin, however, they exhibit close similarities to tumours resulting in inherited colorectal cancer syndromes. Most sporadic colon adenomas and carcinomas also harbour APC gene mutations [4]. The APC gene, which has been recognized as a gatekeeper of colorectal carcinogenesis, is one of the key components of the Wnt signalling pathway. Wnt signalling induces nuclear translocation of transcriptionally active β-catenin through interference with the β-catenin-destruction complex, composed of glycogen synthase kinase-3 (GSK-3α and β), Axin (Axin1 and 2) and APC. In the absence of a Wnt signal this complex efficiently earmarks cytoplasmic β-catenin for degradation through the ubiquitin/proteasome pathway [5,6].
To identify the possible differences between different adenomas that either predispose to cancer or result in benign growths, we compared variations in gene expression between different adenomas and normal mucosa from the same patient with a germline mutation in the APC gene. The approach was designed to identify very early changes that occur during adenoma formation and to detect aberrant regulation of genes required for adenoma-carcinoma progression. Microarray-based expression profiling revealed that gene expression patterns between different adenomas are very similar but are different from normal mucosa. We describe the increased expression of a specific member of the pregnancy specific glycoprotein family and show that induction of this gene is a very early event that does not appear to be dependent on activation of β-catenin.
Methods
Samples
Adenomatous polyps, tumours and matched adjacent normal mucosal tissue samples from 18 FAP cases (germline APC mutations detected by standard techniques), 60 sporadic colorectal cancer cases, five liver metastases and one normal placenta, were obtained from University Health Network (UHN) human tissue bank and the Familial GI Cancer Registry at Mount Sinai hospital, in compliance with each Institutional Review Board. Colorectal cancer cell lines; SW620, SW480, LoVo, RKO, SW1417, LS1034 and MCF12A were purchased from ATCC and grown in media recommended by the distributor. Total RNA samples from normal ovarian, prostate, colon, breast and placental tissues were purchased from Ambion and Clontech. RNA was extracted from cell lines and tissue samples using an RNAeasy kit (Qiagen). Tissues were processed for RNA extraction, in situ hybridization or immunohistochemistry analysis.
Microarray procedure and data analysis
cDNA microarrays consisting of 19,200 human gene clones were employed to explore the variation in gene expression between adenoma and normal mucosa. Microarray slides were obtained from the University Health Network Microarray Centre (UHN, Toronto, Canada). Protocols used for array hybridisation were as published on the UHN Microarray Centre web page with some modifications. Briefly, 5 μg total RNA extracted from normal mucosa or adenoma was labelled with Cy5 and, a reference total RNA pool was labelled with Cy3. The reference RNA used was composed of total RNA from 10 human cell lines (Stratagene) which hybridize to the maximum number of spots on the array. The signals obtained from reference RNA have been used for normalisation of experimental samples. Microarray hybridization was carried out in a hybridization chamber humidified with 2XSSC. Labeled cDNA was dissolved in 80 μl of the hybridization buffer, denatured at 95°C for three minutes in a thermal cycler, and applied on the microarray slides. Microarrays were incubated overnight at 37°C. Post-hybridization washing was performed by serial incubations in buffers with decreasing SSC and SDS concentrations, at 50°C. At least two replicates including dye switches were performed for each experiment to account for possible dye labelling and hybridization bias. Relative expression was assessed by a two-colour hybridization experiment. Slides were scanned using either an Axon GenePix 4000A (Axon) or ScanArray 4000 Scanner (Packard BioScience). The scanned 16-bit TIFF images were quantified using QuantArray software (Packard BioScience). The quantified data files were transferred to a GeneTraffic microarray database and analysis system (Iobion Informatics, Stratagene) with a complete annotation of experiments based on the current MIAME standards for microarray experiments . Each hybridization dataset was filtered and spots that did not pass the quality criteria in both channels were excluded from further analysis. The lowess subarray normalization which uses a local weighted smoother to generate an intensity-dependent normalisation function was applied to each hybridization. The normalised log2 ratios were used for statistical analysis. Data were analysed by both SAM software (Significance Analysis for Microarrays) and the statistical program integrated into GeneTraffic 2.8. Genes exhibiting a consistent 2-fold or more up- or down regulation with a p value of <0.05 were considered significant.
Semi-quantitative and quantitative RT-PCR
Two μg total RNA was reverse transcribed by SuperScript II reverse transcriptase (Stratagene). One μg of each cDNA were amplified using PSG9 specific primers (PSG9-1420) by AccuaPrime Taq DNA Polymerase system (Invitrogen) (Fig. 1). For semi-quantitative PCR, all samples were normalised to the level of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) before PSG9 specific amplification. For quantitative real-time PCR analyses, 500 ng cDNA was amplified using a SYBR Green PCR kit and either primer PSG9, GAPDH or Axin2 and a 7900 Sequence Detector system (Applied Biosystems). Primers were designed by either Oligo 6.4 (Oligo) or Primer Express software (Applied Biosystems). Location of all primers for PSG9 are indicated in figure 1a. Primer sequences used for semi-quantitative PCR were as follow: PSG9-1420F (5'-CCA GCC ACC CAA AGT TTC-3', PSG9-1420R (5'-GGG CAT TCA GAT AGA CAG CAA-3'), GAPDH-F (5'-GAC GCC TGC TTC ACC ACC TTC-3') and GAPDH-R (5'-CCG CTT CGC TCT CTG CTC C-3'). Primers used for quantitative RT-PCR were as follows: PSG9-Q40F (5'-TGG TGG CCT CCG CAG TAA-3'), PSG9-Q40R (5'-GTC TGG ACC ATA GAG GAC ATT TAG G-3'), GAPDH-F 5'-GAA GGT GAA GGT CGG AGT C-3' and GAPDH-R 5'-GAA GAT GGT GAT GGG ATT TC-3', Axin2F (5'-CCA CAC CCT TCT CCA ATC CA-3') and Axin2R (5'-TGG ACA CCT GCC AGT TTC TTT-3'). PCR conditions are available upon request.
Analysis of PSG9 sequence variants in tumors
PSG9 isoforms expressed in normal placenta, SW620 and LoVo colon cancer cell lines and two sporadic colorectal cancer cases, were PCR amplified. Different isoforms were separated on a 2% agarose gel and purified by PerfectPrep gel cleanup kit (Eppendorf). The purified DNA products were cloned into pCR®II-TOPO vector. Clones were sequenced in both directions using T7 and T3 primers using an ABI sequencing system (Applied Biosystems). Sequence variation was examined using GeneJockey II software (Biosoft).
Northern blot analysis
PSG9 expression patterns were analysed using random primed radiolabeled full-length PSG9 cDNA. A 2 kb fragment was digested from a cDNA clone (GeneBank Accession No. 196828) by PAC I and EcoRI restriction enzymes, gel purified and then radiolabeled with [α-32P]dCTP using a T7 Quik Prime system (Pharmacia Biotech). Northern blot analysis was performed based on standard techniques [7].
RNA in situ hybridisation
RNA in situ hybridization (RISH) was used for the examination of PSG9 and PSG2 mRNA expression. Nonradioactive RISH was applied to frozen or paraffin-embedded sections using digoxigenin (DIG)-labelled copy RNA (cRNA) probes. Different cDNA fragments corresponding to human PSG9 (GenBank 196828) were amplified by PCR using sense and antisense primers containing either T7, T3 or SP6 promoter sequences, respectively. PSG2 riboprobes were amplified from a human placental RNA pool (Clontech). The locations of primers and oligoprobes sequences are indicated in figure 1. For cRNA probe synthesis, purified PCR products were used. The transcripts were labelled with DIG-labelled nucleotides, DIG RNA labelling kit and either T7, T3 or SP6 RNA polymerase (Roche Applied Science) to produce DIG-labelled riboprobes. Primers used for amplification and synthesis of each cRNA probe were: PSG9-E2; Sense 5'-T GCC GAA GTC ACG ATT GAA G-3', Anti-sense: 5'-GGA TGC GTT GGA ATA TAC TGT TTC T-3', PSG9-E4-5; Sense: 5'-A TGT CTT AGC CTT CAC CTG TG-3', Anti-sense: 5'-AGT GCC GGT GGG TTA GAT T-3' PSG2; Sense: 5'-GTC CAG ACC TCC CCA GAA T-3', Anti-sense 5'-AGG CTG CTA TGT TGG ATT AAG GAG AG-3'. PCR conditions are available upon request. Seven-μm-thick sections of paraffin-embedded or fresh frozen tissue were cut, fixed in 1XPBS (phosphate-buffered saline) containing 4% paraformaldehyde. A standard in situ hybridization technique was used with some modifications. Images were analysed by light microscope (Leica).
Western blot and immunohistochemical analysis
To detect β-catenin and β-actin, cells were lysed in RIPA buffer (150 mM NaCl, 1% NP-40, 0.5% DOC, 0.1% SDS, 50 mM Tris pH 8.0, 1 mM EDTA). Twenty μg protein were loaded onto 8% SDS polyacrylamide gels, separated and transferred to a nitrocellulose filter by semidry transfer. Western blotting were performed using a standard protocol. The antibody dilutions were as follows; primary antibodies β-catenin (1:1000; Transduction Laboratories) and β-actin (1:2000; Abcam), secondary monoclonal antibody (1:5000; BioRad). Standard immunohistochemistry (IHC) was carried out on formalin-fixed, paraffin-embedded or fresh frozen sections.
Wnt signalling stimulation and reporter gene assay
RKO cells were treated over night with 50 nM Wnt3a recombinant (R&D systems) or 10 μM Kenpaullone (Calbiochem) an inhibitor of GSK3 [8] prior to RNA or protein extraction. RKO cells transfection was performed in six-well plates at the density of 1.5 × 105 cells/well with Lipofectamine 2000 reagent (Invitrogen). Cells were transfected with either Super8XTOP- or 8XFOPFlash and β-galactosidase constructs [9] and were assayed for luciferase activity 23 hrs post-transfection/treatment. The luciferase activity was measured and quantified in a luminometer using a chemiluminescent reporter gene assay system for the combined detection of luciferase and β-galactosidase as recommended by manufacturer (Applied Biosystems). The β-galactosidase was used to normalize luciferase units in each transfection.
Results
Gene expression profiling
Gene profiles of three adenomas and corresponding normal epithelial tissue from an FAP patient (case 640) with an APC germline mutation (n. 2092; T→G) were analysed using 19K human cDNA microarray chips. Statistical analysis using SAM and GeneTraffic 2.8 revealed eighty four transcripts to be represented at statistically significant different levels in all adenomas compared to normal (p < 0.05) (Table 1, 2, 3). Pregnancy specific β1 glycoprotein 9 (PSG9) showed a consistent two fold up-regulation in adenomas compared to normal mucosa and as the most statistically significant candidate in these experiments (p < 0.006) was selected for further analysis (Fig. 1b). The result was consistent between all adenomas on all the microarray chips tested. Differential PSG9 expression was further verified using quantitative RT-PCR (Fig. 1c). PSG9 is a member of the pregnancy specific glycoprotein family (PSGs) [10,11]. We also detected deregulation of a number of TGFβ (transforming growth factor β) regulated genes in our cDNA microarray gene expression screening. Similar to PSG9, TGFβ plays an essential role in normal placentation and is expressed by uterine epithelium from early in pregnancy (Table 1 and 2).
PSG family member expression
Seventeen PSG clones are represented on the 19K human chip used in these studies: four clones represent PSG1, one for PSG3, two PSG4, two PSG5 clones, two PSG6, three PSG11, and three clones are assigned as PSG9. To examine why only one of the PSG9 clones on the cDNA chip indicated differential expression between normal mucosa and polyps, we re-sequenced all the clones representing PSG9. The sequencing results showed that only clone 196828 contained a full-length PSG9 cDNA. This clone represents the largest transcript of PSG9. Because of the high homology (>90%) between different PSG genes and possible cross-hybridization and non-specific binding of primers to different PSGs and also to different transcripts of PSG9, all primers were designed based on the sequence of this clone (Fig. 1a).
Expression of PSG9 in normal and cancer cells
PSG isoforms were originally identified in the circulation of pregnant women [12] and PSG concentrations in the bloodstream increase exponentially until term. By the third trimester the concentration reaches 200–400 μg/ml [13]. Using monoclonal antibodies against PSG as a group and studying their mRNA expression level, it was shown that PSGs are expressed mainly by synthiotrophoblasts during pregnancy, while their expression in normal colon, could not be detected [14,15]. Other studies have reported that PSGs are not exclusively expressed in human placenta and a number of PSG cDNA clones have been isolated from fetal liver, salivary gland, testis, myeloid cells and intestine [16-18]. However, the expression level in these tissues appears to be very low. Because anti-PSG antibodies used in some studies can cross-react with other members of the CEA family, the expression pattern and level of individual PSG proteins and their variants is not well documented. Therefore, we examined the normal expression pattern of PSG9 in different tissues by using a multiple tissue northern blot, semi- and quantitative RT-PCR. High expression of PSG9 transcripts was detected in placenta after only few hours of exposure, while expression in other tissues, including normal colon mucosa was undetectable even after one week exposures (Fig 2a). This result was further verified by quantitative RT-PCR analysis of PSG9 transcripts in different normal tissues (Fig. 2b). PSG9 expression was also examined in a panel of colorectal cancer cell lines. The highest level of PSG9 was found in the SW480 cell line which, similar to placental cells, expresses at least three different PSG9 transcripts. The RKO colon cancer cell line exhibited the lowest level of PSG9 (Fig. 2c–d). We also examined the degree of PSG9 expression in a panel of RNAs extracted from different tumours. Northern blot analysis showed clear expression of three PSG9 transcripts in colon and rectal cancer, and two transcripts in uterine cancer (Fig. 3a). Primary sporadic colorectal tumours, liver metastasis and corresponding normal tissues were also examined for PSG9 expression by semi- and quantitative PCR. Deregulation of PSG9 variants could be detected in most tumours and liver metastases tested. Of note, the expression pattern varied between different tumours and also between different colon cancer cell lines, which may relate to the stage of differentiation or type of mutation in those cases (Fig. 3b). It is also possible that the type of APC mutation plays a role in the level and pattern of PSG9 transcription.
To determine the degree of PSG9 deregulation in colorectal cancer we employed RNA in situ hybridization. Digoxigenin (DIG)-labelled riboprobes were designed to recognize specifically PSG9 variant 1 (nt 776→1087) or all variants (nt 98→318) (Fig. 1a). Normal placental tissues were used as positive control where both PSG9 and another member of PSG family, PSG2 are highly expressed (Fig. 4b–c). Sense-probes were used as negative controls (Fig. 4a). PSG9 RNA transcripts were detected at higher levels and at earlier stages in FAP cases than sporadic colorectal cancer. Transcripts of PSG9 were detected in 78% (14/18) of adenomas from FAP cases with APC germline mutations (Fig. 4g). Of those, 50% also expressed PSG9 in the corresponding normal mucosa (Fig. 4d). However, the expression level was lower in the histologically normal tissue than in the corresponding adenomas and tumours (Fig. 4e–g). In all cases the expression level was much higher in tumours compared to corresponding normal mucosa or adenoma in FAP cases. The degree of PSG9 expression in the tumour samples appears to be related to differentiation, with highly differentiated cells expressing the highest levels. PSG9 expression was also detected in 70% (42/60) of sporadic colorectal cancers, while only 8% (5/60) of the corresponding normal mucosa showed even low levels of PSG9 expression (Fig. 4i, 4k). To examine whether PSG9 expression was specific for tumours, we examined expression of PSG2 in the same tumours which had exhibited deregulated PSG9 expression. PSG2 was expressed at very low levels in both tumour and normal mucosa from colorectal cancer cases. As indicated earlier, both PSG9 and PSG2 are highly expressed in placenta (Fig. 4b–c). In general, PSG9 could be detected at higher levels in FAP cases. Notably, PSG9 was expressed in the morphologically normal mucosa of FAP patients with APC germline mutations (Fig. 4d–e), while its expression was rarely detectable in normal mucosa of sporadic colorectal cancers using in situ hybridization (Fig. 4i). The expression pattern of PSG9 also varied between different tumours and cell lines although the significance of these differences is unclear (Fig. 2–4). Normal tissues, adenomas and tumours were also examined for nuclear β-catenin accumulation by immunostaining (Fig. 4h, j, l). Nuclear β-catenin could be detected in all sporadic tumours (Fig. 4l), while staining was less intense in the normal-appearing epithelium and adenomas in FAP cases where PSG9 was highly upregulated (Fig. 4g–h). In general, higher levels of β-catenin nuclear accumulation could be detected in sporadic cancer (Fig 4l) compared to FAP cases (Fig. 4h) while as expected no β-catenin stabilisation was observed in the corresponding normal tissues (Fig. 4j).
Next, we examined whether the PSG9 expressed in tumours was different from PSG9 expressed by placental cells during pregnancy. The coding sequences of all PSG9 variants of two colon cancer cell lines (SW620 and LoVo) and two sporadic colorectal cancer cases were cloned and sequenced. Two sequence variations were found in the non-coding region, while no changes were found in the coding region of PSG9. These results indicate that the PSG9 proteins expressed by placental cells and tumour cells have similar sequences and might have similar function, however, the level of expression of each PSG9 variant differed from those found in placenta.
Effect of Wnt signal activation on PSG9 expression in the absence of APC mutation
Nuclear localization of β-catenin has been shown to correlate with its transcriptional activity in cells. While SW480 cells with very high levels of β-catenin and APC mutations express high levels of PSG9, RKO cells which have previously been reported to express wild type APC and β-catenin [19,20] show no PSG9 expression (Fig. 2c and 5a–b). To investigate the potential role of β-catenin/Wnt3a up-regulation in the absence of APC mutations on PSG9 transcription, accumulation and nuclear localization of β-catenin in RKO cells was induced by treatment with recombinant Wnt3a, Kenpaullone, or using a stable RKO-β-catenin cell line which expresses constitutively a active mutant (βcat-S37A) form of β-catenin [21] (Fig. 5b). One day post-treatment, cells were prepared for luciferase assays and RNA extraction. Reporter gene assays were performed to examine the ability of these cells to respond to Wnt stimulation and quantitative PCR to investigate whether Axin2, a known Wnt signal downstream target gene was induced. Stimulation of the cells resulted in four-fold induction of luciferase activity (Fig. 5c), and a 2.4-fold increase in Axin2 transcript levels, indicating that these cells are able to respond to Wnt signaling and this signal induction can increase downstream target gene expression (Fig. 5d). We further examined the PSG9 expression level in these stimulated cells. No PSG9 induction in these cells could be detected (Fig. 5e). These results suggest that enhanced levels of β-catenin or Wnt signaling in cells, without concomitant defect in APC, is likely insufficient to induce PSG9 expression.
Discussion
In this study, we have shown that PSG9 is ectopically expressed in colorectal cancer and this is most likely APC dependent, since abnormal expression can be detected as early as in normal appearing epithelial cells and adenomas of the FAP cases which carry APC germline mutations, while corresponding normal tissue in sporadic colorectal cancer tumors lack PSG9 expression. Given the increased expression of PSG9 in the mucosal cells of FAP patients that displayed lower β-catenin stabilisation compared to sporadic colorectal tumours in our study, it is possible that PSG9 is not directly regulated by the β-catenin/Wnt signalling pathway and that other molecules that regulate PSG9 expression are altered as a consequence of APC mutation. Notably, deregulation of PSG9 is detectable as early as in mucosa that appears histologically normal in FAP cases with APC germline mutations, suggesting that the dose and gradient of APC is important in PSG9 regulation. Functional analysis of APC protein has revealed a broad spectrum of activities for this molecule [22,23]. In normal-appearing epithelial cells in FAP, PSG9 was expressed mostly on the apical surface, while in tumours, expression was detected from the top to the base of crypts. Early expression of PSG9 even before adenoma formation at the top of the crypts in FAP cases, suggests that transformation process starts in cells at the top of the crypts which then gradually move downward (Fig. 4d). These observations are consistent with the "top-down" morphogenesis model of colorectal cancer [24].
PSGs exhibit sequence similarity to the carcinoembryonic antigen (CEA) family which, in turn, is a member of the immunoglobulin (Ig) superfamily. The CEA gene family can be divided into three subgroups; the CEA subgroup (12 genes), PSG subgroup (11 genes) and a pseudogene subgroup (6 genes) [25]. CEA is a widely used tumour marker, the main clinical utility of which is in monitoring clinical course of colorectal carcinoma after surgical resection [26]. Contrary to its name, CEA is expressed in normal adult tissue, as well as during fetal development. The role of CEA in normal human physiology is not well understood. Based on its structure, a number of functions have been suggested, including intracellular cell adhesion, signal transduction or signal transduction regulation. PSGs are secreted proteins and, in contrast to CEA, most PSGs are not expressed in normal colon epithelial cells. The main site for PSG production is the placental syncytiotrophoblasts during pregnancy [14]. Although PSGs were discovered more than three decades ago, their function is still unknown and the receptor(s) for these proteins has yet to be identified. Most PSGs have an RGD (Arg-Gly-Asp) motif in a conserved region in the N-terminal domain which suggest that these genes may function as adhesion recognition signals for integrins [27].
Immune privilege of cancer cells
Several studies suggest that T-cells are capable of recognizing and responding to tumours in experimental conditions, yet most tumours are able to escape detection by the immune system. A common characteristic of cancer and the placenta is their ability to avoid immune reactivity. The capability of cancer cells to sidestep the body's immune reaction, is believed to be partly aided by the protective coating of the cells, which is largely composed of glycoproteins [28]. PSGs are heavily glycosylated proteins and the primary amino acid sequence is masked by the sugar modification which helps in avoid specific antibody recognition [29]. It has been suggested that PSG production by trophoblasts regulates maternal immune response to the fetus. It is also possible that PSG9 expressed by pre-malignant and cancer cells protect tumours from recognition by the body's immune defences. Another similarity between the respective milieu where both placenta and adenomas develop is their low-oxygen environment. Trophoblast invasion and placental development during the first trimester occurs in a low-oxygen environment, as the blood flow to the intervillous space is not yet established [30]. We found lower levels of beta hemoglobin (HBB) RNA in adenomas compared to normal mucosa in this study (Table 1) as well as in another gene profiling study we have performed (manuscript in preparation).
Role of PSG9 in cancer
Most of the PSG subgroup (11 genes) are expressed. However, the existence of allelic variants with a stop codon in the N-domain of some PSGs, indicates that some individuals may not express all members of the PSG family [31]. The possible involvement of PSGs in cancer and their genetic variation may in part explain phenotypic divergence that exists in cancer cases with otherwise identical germline mutations. The RGD sequence motif in the N-terminal domain of most PSGs is also present in a variety of extracellular matrix proteins that bind to integrin receptors such as fibronectin and vitronectin [32]. It has been hypothesized that the PSGs (like most Ig superfamily members) are involved in adhesion/recognition processes. Another possibility is that upregulation of PSG9 might favour tumour development by causing a reversion of the monolayered adult colonic epithelium to an embryonic multilayered arrangement.
Conclusion
Our results provide strong evidence that PSG9 deregulation in cells occurs early during adenoma-carcinoma formation. High levels and early deregulation of PSG9 in adenomas, as well as normal mucosa in some FAP cases, indicates the potent role of APC germline mutations that are often found in these cases. The precise role of PSG9 in carcinogenesis remains to be determined. However, early-onset over-expression of PSG9 in different types of cancer suggests that this gene may be considered as a valuable biochemical tumorigenesis marker. To elucidate whether the frequency of occurrence of elevated PSG9 could have clinical significance, further analysis of serum levels of PSG9 and also other PSGs are warranted. Since PSG9 is not found in the non-pregnant adult except in association with cancer, it may be useful as a biomarker for the early detection of cancers of various types.
Abbreviations
CRC; colorectal cancer, FAP; familial adenomatous polyposis, APC; adenomatous polyposis coli, CEA; carcinoembryonic antigen, PSG; pregnancy specific glycoprotein, SAM; statistical analysis of microarrays.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
SS; designed and performed the study, and drafted the manuscript. JG; assisted with the microarray data analysis. RC; was involved in the analysis and interpretation of histopathology and reviewed the final manuscript. SG; coordinated data/material collection from clinic and reviewed the final manuscript. JRW; conceived the study, participated in design and coordination and helped draft the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This work was supported by the National Cancer Institute of Canada funded by the Canadian Cancer Society and Canadian Institutes of Health Research. SS appreciates the support of the Helena Lam fellowship. We thank Drs. Vogelstein, Kinzler and Moon for cDNA constructs, Dr. Hung for stable cell line, and Dr. Redson for support.
Figures and Tables
Figure 1 Genomic structure of PSG9. The human PSG9 gene is located on chromosome 19q13.2 and encodes at least three variants. The exon structures of each isoform are indicated in the figure. The largest PSG9 variant, variant 1, consists of 1282 nucleotides and encodes for a protein of 427 amino acids. The location and size of each primer and riboprobes used in this study are indicated (a). Statistical analysis of microarray data using SAM program revealed up to two-fold over-expression of PSG9 in adenoma 1, 2 and 3 (A1 to A3) compared to normal (N). The log2 ratio of fluorescence intensity is shown on the y-axis while PSG9 expression in different samples is shown on the x-axis. At least four microarray replicates including dye switches (Cy3- or Cy5-labelled) were performed to account for possible labeling and hybridization bias (b). The data have further been verified in adenoma 1, 2 and normal tissue by Real Time PCR (c).
Figure 2 PSG9 is exclusively expressed by placental cells. Quantitative RT-PCR analysis of PSG9 transcript (a) and northern blot analysis of different normal tissues (1 μg mRNA/lane) probed with the labelled PSG9 cDNA (b) shows specific expression of PSG9 in placenta. (a-b). Colorectal cancer cell lines were also examined for PSG9 expression level by quantitative- (c) and semi-quantitative RT-PCR (d). Highest expression level of PSG9 was detected in SW480 cells and was lowest in the RKO cell line (c-d).
Figure 3 PSG9 is ectopically expressed by cancer cells. A multiple-tumour Northern blot (10 μg RNA/lane) revealed over-expression of PSG9 in colon, rectal and uterus cancer. At least three different transcripts were observed (a). Expression of PSG9 "isoform a" was examined in a panel of colorectal cancer cases. Forty-nine percent (15/27) of cases showed up-regulation of PSG9 (isoform a) in the tumours compared to corresponding normal tissue (b). However, the expression levels were different between different cases (T; tumour, A; adenoma).
Figure 4 RNA in situ hybridization and IHC analysis of colorectal cancer cases. Sections from sporadic/familial colorectal cancer and placenta (as positive control) were hybridised with Dig-labelled PSG9 RNA probes. Both PSG2 (b) and PSG9 (c) were expressed at a high level in placental tissue. Sense-probes were used as a negative control on placental tissue (a). In microscopic normal epithelial cells from FAP cases, PSG9 expression was detected at the top of crypt (d) (see discussion). PSG9 transcripts (shown as dark blue) were detected at very low levels in normal mucosa (e), adenomas (f), while high expression was detected in tumour cells from the same FAP case (g). In contrast to sporadic cases, PSG9 was detected in normal appearing mucosa in some FAP cases with APC germline mutations, suggesting that dose and level of APC have an impact on PSG9 levels in cells (e, i). A high level of PSG9 was detected in a sporadic case (k), while corresponding normal tissue was negative (i). Tumours and corresponding normal tissue were also examined for β-catenin stabilization by immunostaining (h, j, l). As expected, high levels of β-catenin were detected in all sporadic colorectal tumours (l), while the protein level was less intense in FAP cases (h) where PSG9 up-regulation could be measured (g).
Figure 5 PSG9 expression analysis in Wnt stimulated RKO cells. To determine whether induction of Wnt signaling in cells with wild type APC could induce PSG9 expression, RKO cells with wild type APC and β-catenin were stimulated with either Wnt3a or Kenpaullone (Kenp). After 23 hrs treatment RNA and protein were extracted and processed for PSG9 transcripts expression level by semi-quantitative RT-PCR (a) and β-catenin accumulation by western blot analysis (b). Neither of the treatments nor the RKO-β-cateninS37A (βcat-S37A) stable cell line which expressed constitutively active β-catenin in RKO cells caused expression of PSG9 (a). The SW480 colorectal cancer cell line was used as a positive control (a-b). The RKO cells responded to Wnt stimulation, since cells treated with Wnt3a showed 4-fold induction in luciferase activity compared to untreated cells (c). Axin2, a known downstream target of Wnt signaling, showed 2.4 fold up-regulation in expression as determined by quantitative PCR (d). No PSG9 transcript up-regulation was detected in these cells under these conditions (e). Each sample was analyzed in triplicate.
Table 1 Genes upregulated in all adenomas compared to normal mucosa (p < 0.05).
Name Description Gene ID Fold change
PSG9 pregnancy specific beta-1-glycoprotein 9 196828 2.59
PPP2CB protein phosphatase 2 (formerly 2A), catalytic subunit, beta isoform 469697 1.71
ITGA1 integrin, alpha 1 212078 1.95
NEDD8 neural precursor cell expressed, developmentally down-regulated 8 220392 1.74
ALAS2 aminolevulinate, delta-, synthase 2 (sideroblastic/hypochromic anemia) 201112 1.98
DCN decorin 197609 1.85
JAG1 jagged 1 (Alagille syndrome) 117734 2.21
DPP6 dipeptidylpeptidase 6 166550 1.86
RNASE4 ribonuclease, RNase A family, 4 201596 2.44
XBP1 X-box binding protein 1 213933 1.95
CDC42 cell division cycle 42 (GTP binding protein, 25 kDa) 214563 1.73
MGC8407 hypothetical protein MGC8407 179857 1.78
NUP98 nucleoporin 98 kDa 206345 2.02
CDKN1A cyclin-dependent kinase inhibitor 1A (p21, Cip1) 470149 1.56
IL1R2 interleukin 1 receptor, type II 470402 1.54
PIP5K1B phosphatidylinositol-4-phosphate 5-kinase, type I, beta 211877 1.64
TAGLN2 transgelin 2 152371 1.67
TAF7 TAF7 RNA polymerase II, TATA box binding protein (TBP)-associated factor 365930 1.81
RAB25 RAB25, member RAS oncogene family 149515 1.64
ANAPC7 anaphase-promoting complex subunit 7 487822 1.79
MTCH2 mitochondrial carrier homolog 2 (C. elegans) 138486 1.8
HCA112 hepatocellular carcinoma-associated antigen 112 207556 2.3
BS69 adenovirus 5 E1A binding protein 31337 1.69
MGC29898 hypothetical protein MGC29898 504671 1.65
LOC124245 hypothetical protein BC001584 194780 2.05
FVT1 follicular lymphoma variant translocation 1 115784 1.55
FLJ20202 FLJ20202 protein 142952 1.59
ARHGDIA Rho GDP dissociation inhibitor (GDI) alpha 32555 1.63
HEAB ATP/GTP-binding protein 212684 1.59
VRK3 vaccinia related kinase 3 488478 1.49
HLA-DPA1 major histocompatibility complex, class II, DP alpha 1 200735 1.48
WFDC2 WAP four-disulfide core domain 2 366323 1.85
KIAA0241 KIAA0241 protein 212216 1.83
GPR161 G protein-coupled receptor 161 489796 1.81
SLAMF8 SLAM family member 8 128371 1.99
ARG99 ARG99 protein 163160 1.7
LOC170394 hypothetical protein BC011630 154928 1.9
Differentially expressed genes identified in adenomas compared to corresponding normal tissue. In total, 84 genes showed significant differential expression in all three adenomas compared to normal-appearing epithelial cells (p < 0.05). Thirty-seven genes were up-regulated (Table 1), and 47 down-regulated (Table 2). PSG9 showed a consistent 2-fold over-expression in all adenomas compared to normal mucosa (p < 0.006).
Table 2 Genes downregulated in all adenomas compared to normal mucosa (p < 0.05).
Name Description Gene ID Fold change
AFP alpha-fetoprotein 428098 1.91
GFAP glial fibrillary acidic protein 382693 2.23
HBB hemoglobin, beta 148425 1.74
CLU clusterin (complement lysis inhibitor, SP-40,40, sulfated glycoprotein 2 343293 1.75
VDR vitamin D (1,25- dihydroxyvitamin D3) receptor 344295 2.09
MGP matrix Gla protein 502239 1.97
IGFBP2 insulin-like growth factor binding protein 2, 36 kDa 471830 1.74
PRKCE protein kinase C, epsilon 51986 3.39
COX4I1 cytochrome c oxidase subunit IV isoform 1 178644 1.89
APOB apolipoprotein B (including Ag(x) antigen) 300017 1.97
BIRC4 baculoviral IAP repeat-containing 4 138505 2.15
ITGAE integrin, alpha E (antigen CD103, human mucosal lymphocyte antigen 1; alpha polypeptide) 358848 2.3
GAS1 growth arrest-specific 1 341345 1.47
GATA3 GATA binding protein 3 148627 1.47
VEGFB vascular endothelial growth factor B 167296 2.01
ADIPOR2 adiponectin receptor 2 24286 1.98
RUFY2 RUN and FYVE domain containing 2 161523 2
LRP8 low density lipoprotein receptor-related protein 8, apolipoprotein e receptor 134714 2.1
SHMT2 serine hydroxymethyltransferase 2 (mitochondrial) 51583 1.96
TXK TXK tyrosine kinase 147839 1.7
LOC51315 hypothetical protein LOC51315 503300 1.39
UGCGL2 UDP-glucose ceramide glucosyltransferase-like 2 32716 3.02
MGC5178 hypothetical protein MGC5178 235090 1.73
IPO9 importin 9 233100 1.71
KBRAS2 I-kappa-B-interacting Ras-like protein 2 5091 2.05
SLC35F2 solute carrier family 35, member F2 42703 2.55
FLJ33761 hypothetical protein FLJ33761 23095 2.07
ZNF198 zinc finger protein 198 153735 1.75
PAG phosphoprotein associated with glycosphingolipid-enriched microdomains 487926 2
FLJ40432 hypothetical protein FLJ40432 23334 2.12
LOC114926 hypothetical protein BC013035 270038 1.64
LOC51277 Ras-associated protein Rap1 44081 1.92
FLJ20360 hypothetical protein FLJ20360 270110 1.61
LOC56931 hypothetical protein from EUROIMAGE 1967720 145011 1.8
LBP-32 leader-binding protein 32 195784 1.81
SATB2 SATB family member 2 26583 1.78
CTDSP1 CTD (carboxy-terminal domain, RNA polymerase II, polypeptide A) small phosphatase 1 131828 2.13
OSBPL2 oxysterol binding protein-like 2 305402 1.81
FLJ38991 hypothetical protein FLJ38991 31673 2.16
MGC12972 hypothetical protein MGC12972 256553 1.69
ZNF496 zinc finger protein 496 32065 1.81
FLJ35936 hypothetical protein FLJ35936 33409 2
C14orf141 chromosome 14 open reading frame 141 161520 1.68
SGPP2 sphingosine-1-phosphate phosphotase 2 46316 1.94
DKFZp313N0621 hypothetical protein DKFZp313N0621 258616 2.22
LOC285550 hypothetical protein LOC285550 250675 1.89
MGC52498 hypothetical protein MGC52498 683068 1.85
Table 3 Genes differentially expressed within different adenomas compared to corresponding normal mucosa.
Gene's name Description A1 A2 A3
GAS1 growth arrest-specific 1 ↓ ↑ ↓
IGFBP2 insulin-like growth factor binding protein 2 ↓ ↑ ↓
MAPRE3 microtubule-associated protein, RP/EB family, member 3 ↑ ↓ ↓
ZFYVE20 zinc finger, FYVE domain containing 20 ↑ ↑ ↓
FLJ11848 hypothetical protein FLJ11848 ↓ ↓ ↑
KIAA1750 KIAA1750 protein ↓ ↓ ↑
KIAA1941 KIAA1941 protein ↓ ↑ ↓
Differentially expressed genes identified in adenomas compared to corresponding normal tissue. In total, 84 genes showed significant differential expression in all three adenomas compared to normal-appearing epithelial cells (p < 0.05). Thirty-seven genes were up-regulated (Table 1), and 47 down-regulated (Table 2). PSG9 showed a consistent 2-fold over-expression in all adenomas compared to normal mucosa (p < 0.006).
==== Refs
Nishisho I Nakamura Y Miyoshi Y Miki Y Ando H Horii A Koyama K Utsunomiya J Baba S Hedge P Mutations of chromosome 5q21 genes in FAP and colorectal cancer patients Science 1991 253 665 669 1651563
Gryfe R Gallinger S Microsatellite instability, mismatch repair deficiency, and colorectal cancer Surgery 2001 130 17 20 11436007 10.1067/msy.2001.112738
Kinzler KW Nilbert MC Vogelstein B Bryan TM Levy DB Smith KJ Preisinger AC Hamilton SR Hedge P Markham A Identification of a gene located at chromosome 5q21 that is mutated in colorectal cancers Science 1991 251 1366 1370 1848370
Kinzler KW Vogelstein B Lessons from hereditary colorectal cancer Cell 1996 87 159 170 8861899 10.1016/S0092-8674(00)81333-1
Korinek V Barker N Morin PJ van Wichen D de Weger R Kinzler KW Vogelstein B Clevers H Constitutive transcriptional activation by a beta-catenin-Tcf complex in APC-/- colon carcinoma Science 1997 275 1784 1787 9065401 10.1126/science.275.5307.1784
Woodgett JR Judging a protein by more than its name: GSK-3 Sci STKE 2001 RE12 11579232
Ausubel MF Brent R Kingston RE Moore DD Seidman JG Smith JA Struhl K Current Protocoles in Molecular Biology 1994 1
Cohen P Goedert M GSK3 inhibitors: development and therapeutic potential Nat Rev Drug Discov 2004 3 479 487 15173837 10.1038/nrd1415
Veeman MT Slusarski DC Kaykas A Louie SH Moon RT Zebrafish prickle, a modulator of noncanonical Wnt/Fz signaling, regulates gastrulation movements Curr Biol 2003 13 680 685 12699626 10.1016/S0960-9822(03)00240-9
Plouzek CA Watanabe S Chou JY Cloning and expression of a new pregnancy-specific beta 1-glycoprotein member Biochemical & Biophysical Research Communications 1991 176 1532 1538 1840485 10.1016/0006-291X(91)90461-F
Tatarinov YS Hobes CA Immunological identification of a new b1-globulin in the blood serum of pregnant women Byull EKsp Biol Med 1970 69 66 68
Bohn H Detection and characterization of pregnancy proteins in the human placenta and their quantitative immunochemical determination in sera from pregnant women Arch Gynakol 1971 210 440 457 5001318 10.1007/BF01628222
Lin TM Halbert SP Spellacy WN Measurement of pregnancy-associated plasma proteins during human gestation J Clin Invest 1974 54 576 582 4853116
Zhou GQ Baranow V Zimmermann W Grunert F Erhard B Mincheva-Nilsson L Hammarström S Tompson J Highly specific monoclonal antibody demonstrates that pregnancy-specific glycoprotein (PSG) is limited to syncytiotrophoblast in human early and term placenta. Placenta 1997 18 491 501 9290143
Zimmermann W Webe B Ortlieb B Rudert F Schempp W Fiebig HH Shively JE Von Kleist S Thompson JA Chromosomal localization of the carcinoembryonic antigen gene family and differential expression in various tumors. Cancer Research 1988 48 2550 2554 3356015
Borjigin J Tease LA Barnes W Chan WY Expression of the pregnancy-specific beta 1-glycoprotein genes in human testis Biochemical & Biophysical Research Communications 1990 166 622 629 2302228 10.1016/0006-291X(90)90854-G
Horne CH Towle CM Milne GD Detection of pregnancy specific beta1-glycoprotein in formalin-fixed tissues. J Clin Pathol 1977 30 19 23 190273
Shupert WL Chan WY Pregnancy specific beta 1-glycoprotein in human intestine Molecular & Cellular Biochemistry 1993 120 159 170 8487756 10.1007/BF00926089
Da Costa LT He TC Yu J Sparks AB Morin PJ Polyak K Laken S Vogelstein B Kinzler KW CDX2 is mutated in a colorectal cancer with normal APC/beta-catenin signaling Oncogene 1999 18 5010 5014 10490837 10.1038/sj.onc.1202872
Sparks AB Morin PJ Vogelstein B Kinzler KW Mutational analysis of the APC/beta-catenin/Tcf pathway in colorectal cancer Cancer Res 1998 58 1130 1134 9515795
Deng J Miller SA Wang HY Xia W Wen Y Zhou BP Li Y Lin SY Hung MC beta-catenin interacts with and inhibits NF-kappa B in human colon and breast cancer Cancer Cell 2002 2 323 334 12398896 10.1016/S1535-6108(02)00154-X
Fearnhead NS Britton MP Bodmer WF The ABC of APC Human Molecular Genetics 2001 10 721 733 11257105 10.1093/hmg/10.7.721
Näthke IS The adenomatous polyposis coli protein: the Achilles heel of the gut epithelium Annu Rev Cell Dev Biol 2004 20 337 366 15473844 10.1146/annurev.cellbio.20.012103.094541
Shih IM Wang TL Traverso G Romans K Hamilton SR Ben-Sasson S Kinzle KW Vogelstein B Top-down morphogenesis of colorectal tumors. Proc Natl Acad Sci 2001 98 2640 2645 11226292 10.1073/pnas.051629398
Olsen A Teglund S Nelson D Gordon L Copeland A Gorgescu A Garrano A Hammarström S Gene organization of the pregnancy-specific glycoprotein region on human chromosome 19: Assembly and analysis of a 700 kb cosmid conting sapnning the region. Genomic 1994 23 659 668 10.1006/geno.1994.1555
Thomson DM Krupey J Freedman SO Gold P The radio immunoassay of circulating carcinoembryonic antigen of the human digestive system. Proc Natl Acad Sci USA 1969 64 161 167 5262998
Hammarström S The carcinoembryonic antigen (CEA) family: structures, suggested functions and expression in normal and malignant tissues Seminars in Cancer Biology 1999 9 67 81 10202129 10.1006/scbi.1998.0119
Currie GA Bagshawe KD The masking of antigens on trophoblast and cancer cells Lancet 1967 1 708 710 4163947 10.1016/S0140-6736(67)92183-6
Chen H Plouzek CA Liu JL Chen CL Chou JY Characterization of a major member of the rat pregnancy-specific glycoprotein family DNA Cell Biol 1992 11 139 148 1547019
Rodesch F Simon P Donner C Jauniaux E Oxygen measurements in endometrial and trophoblastic tissues during early pregnancy Obstet Gynecol 1992 80 283 285 1635745
Teglund S Olsen A Khan WN Frangsmyr L Hammarström S The pregnancy-specific glycoprotein (PSG) gene cluster on human chromosome 19: fine structure of the 11 PSG genes and identification of 6 new genes forming a third subgroup within the carcinoembryonic antigen (CEA) family. Genomics 1994 23 669 684 7851896 10.1006/geno.1994.1556
Rutherfurd KJ Chou JY Mansfield BC A motif in PSG11s mediates binding to a receptor on the surface of the promonocyte cell line THP-1 Mol Endocrinol 1995 Oct;9(10):1297-305.
|
15982419
|
PMC1184062
|
CC BY
|
2021-01-04 16:03:06
|
no
|
BMC Cancer. 2005 Jun 27; 5:66
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-66
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-701599626710.1186/1471-2407-5-70Research ArticleFamilial breast cancer: characteristics and outcome of BRCA 1–2 positive and negative cases Veronesi Andrea [email protected] Clelia de [email protected] Maria D [email protected] Davide [email protected] Martina [email protected] Cristina [email protected] Riccardo [email protected] Alessandra [email protected] Diana [email protected] Ettore [email protected] Mauro [email protected] Division of Medical Oncology C, Centro di Riferimento Oncologico, Aviano, Italy2 Division of Experimental Oncology I, Centro di Riferimento Oncologico, Aviano, Italy3 Epidemiology Unit Centro di Riferimento Oncologico, Aviano, Italy2005 4 7 2005 5 70 70 17 1 2005 4 7 2005 Copyright © 2005 Veronesi et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The clinical and pathological characteristics and the clinical course of patients with breast cancer and BRCA 1–2 mutation are poorly known.
Methods
From 1997, patients with breast cancer and a family history of breast or ovarian cancer were offered BRCA testing. The clinical and pathological features of patients with known BRCA status were retrospectively assessed and comparisons were made between cancers arising in BRCA positive and BRCA wild type (WT) patients respectively. Type of treatment, pattern of relapse, event (local relapse, contralateral breast cancer, metastases) free and overall survival were also compared in the two groups. Out of the 210 patients tested, 125 had been treated and followed-up at our Institution and were evaluated in this study.
Results
BRCA positive patients tended to be more often premenopausal (79% vs 65%) and to have positive lymphnodes (63% vs 49%), poorly differentiated tumours (76% vs 40% – p = 0.002 at univariate analysis, not significant at multivariate analysis) and negative estrogen receptors (43% vs 29%). Treatment was not different in the two groups. In the 86 BRCA-WT patients, the first event was a local relapse in 3 (3%), metachronous contralateral breast cancer in 7 (8%) and distant metastases in 16 (19%). In the 39 BRCA positive patients, the corresponding figures were 3 (8%), 8 (21%) and 3 (8%). There was no difference in event free survival, with a median of 180 months in both groups of patients. At 20 years, projected survival was 85% for BRCA positive patients and 55% for BRCA-WT, but this difference was not statistically significant.
Conclusion
Although BRCA positive patients have more frequently negative prognostic factors, their prognosis appears to be equal to or better than in patients with BRCA-WT.
==== Body
Background
The issue of familial breast cancer has raised much attention in recent years due to its numerous medical and social implications. Following to the identification of the tumour suppressor genes BRCA1 and BRCA2 and of the increase in breast cancer risk associated with mutations of these genes, several studies have examined the entity of this risk and the diagnostic and therapeutic procedures indicated to curtail the incidence of breast and ovarian cancer in this population [1].
Whether the cases of breast cancer arising in BRCA1-2 positive women are different from those occurring in the general population in terms of biologic characteristics, response to treatment and eventual outcome has not been fully elucidated, although most reports suggest that BRCA1-2 positive cases differ from sporadic cases without familiarity for more frequent adverse prognostic factors [2].
At our Institution, a programme of identification of healthy BRCA1-2 carriers is ongoing, starting with the determination of the BRCA status in patients with breast cancer and a significant family history. This caused the identification of a population of breast cancer patients with family history, with or without BRCA1-2 mutation.
With the aim of ascertaining the differences in the clinical and pathological characteristics and in the clinical course of these two groups of patients, we retrospectively evaluated and compared several clinico-pathologic and therapeutic issues and the outcome of patients with or without BRCA1-2 mutation.
The results of that study are the subject of the present report.
Methods
From 1997, patients with breast cancer presenting at our Institution were considered for eligibility to BRCA testing. Conditions of eligibility included a history among first degree relatives of: at least two cases of breast cancer <50 years; or at least 3 cases of breast cancer, any age; or at least two cases of bilateral breast cancer, any age; or at least two cases of ovarian cancer, any age; or at least one case of breast cancer <50 years plus one case of ovarian cancer, any age. Eligible patients were offered BRCA testing and consenting patients underwent genetic testing according to standard methods. In case of a positive test, genetic testing was also offered to healthy first degree relatives.
The clinical and pathological features of patients with known BRCA status were assessed and comparisons were made between cancers arising in BRCA positive and BRCA wild type (WT) patients respectively. Type of treatment, pattern of relapse, event (local relapse, contralateral breast cancer, metastases) free and overall survival were also compared in the two groups. In case of bilateral metachronous breast cancer, the primary tumour was considered to be the index cancer and the secondary tumour was classified as an event (contralateral breast cancer). Overall and progression free survival curves were estimated by the Kaplan-Meier technique and compared with the use of the two-sided log-rank test. The Cox proportional hazards model was fittted to assess the association with selected patient characteristics. The model included also terms for age (one-year age group) and tumour grade. A p value of less than 0.05 was required to reject the null hypothesis.
Out of the 210 patients tested, 125 had been treated and followed-up at our Institution and were evaluated in this study.
Results
Of the 125 patients with known BRCA status followed or treated at our Institution, 86 were BRCA-WT and 39 were BRCA positive (9 BRCA1, 30 BRCA2). The main characteristics of the patients at the time of diagnosis of breast cancer and the primary treatment received are reported in Table 1. BRCA positive patients tended to be more often premenopausal (82% vs 64%) and to have positive lymphnodes (63% vs 49%), poorly differentiated tumours (76% vs 40% – p = 0.002 at univariate analysis, not significant at multivariate analysis)) and negative oestrogen receptors (43% vs 30%). BRCA-WT tended to be more often multifocal or multicentric (47% vs 33%). Only a minority of patients had a HER-2 determination. Previous treatment was not different in the two groups, although, as expected from oestrogen receptors distribution, more BRCA-WT patients received adjuvant hormone therapy and more BRCA positive patients received adjuvant chemotherapy. Although numbers were small, there was no significant difference between BRCA1 and BRCA2 cases. In the 30 BRCA2 cases, average age was 40.3 years, 62% of patients were premenopausal, 7% had T3-T4 tumours, 67% were node positive, 69% had G3 and 29% oestrogen receptor negative tumours.
The median follow-up time was 69 months (range, 10–280 months). In the 86 BRCA-WT patients, the first event was a local relapse in 3 (3%), metachronous contralateral breast cancer in 7 (8% of the 84 patients with primary unilateral breast cancer) and distant metastases in16 (19%). In the 39 BRCA positive patients, the corresponding figures were 3 (8%), 8 (21% of the 38 patients with primary unilateral breast cancer) and 3 (8%). The difference in the frequency of contralateral breast cancer as first event approached statistical significance (p = 0.07).
Event free survival in BRCA-WT and BRCA positive patients is shown in Figure 1. There was no difference in event free survival, with a median of 180 months in both groups of patients. Overall survival is shown in Figure 2. At 20 years, projected survival was 85% for BRCA positive patients and 55% for BRCA-WT, but this difference was not statistically significant. There was no difference in event free survival and overall survival between BRCA1 and BRCA2 positive patients.
Discussion
Although the importance of family history as a risk factor for breast cancer is widely recognized, there is disagreement on its impact upon prognosis, with conflicting results reported in several series [3,4]. The discovery of the breast cancer susceptibility genes BRCA1-2 has allowed a clearer identification of genetically related cases.
Some studies suggest that mutations of the BRCA gene may be related, besides their impact on the susceptibility to breast and ovarian cancer, to distinctive biological characteristics and clinical course.
In general, the histopathologic features of BRCA associated cases are reported as being more unfavourable as compared to sporadic cases, with more high-grade, oestrogen receptor negative and rapidly proliferating tumours [5]. In a recent study, nodal status was found to be correlated with tumour size in BRCA-1 negative but not in BRCA-1 positive patients [6]. In some relatively small studies, the prognosis of patients with BRCA mutations was worse [7-9] or comparable [10,11] to that of sporadic cases. In these series, prognostic factors were generally worse than in sporadic cases. In a Finnish study [12], it was noted that BRCA1 breast cancer patients had a lower survival rate than sporadic cases, BRCA2 patients or patients with non BRCA1/2 related familial breast cancer. However, the difference was not statistically significant. It should be noted that the BRCA positive patient populations studied varied much, including Ashkenazi Jewish patients with node negative disease [7], early-onset disease [8,9], or familial breast cancer [11,12]. Also control groups were different, varying from BRCA negative cases in the same population to large population based registries. The clinical implications of these findings appear not to be fully understood and more data on the issue are necessary before conclusions can be drawn. In particular, the impact of BRCA alterations upon survival is unclear [13].
In the present series, we analyzed the clinical characteristics, treatment and outcome in a group of patients with familial breast cancer according to their BRCA status.
The first finding to be mentioned is that, even in a group of breast cancer with a family history, the incidence of BRCA mutations is relatively low (about 25%) in our population of Italian women. This further stresses the need for stringent selection criteria before offering the test to the individual women. Besides, as a consequence, the study population was not large enough to detect small differences between BRCA positive and negative case..
The characteristics of patients differed in the two groups, in that features linked to an aggressive course (premenopausal status, poorly differentiated tumours, ER negative, node positive) were or tended to be more frequent in BRCA (predominantly BRCA2) positive patients. Treatment was similar.
The first event was more frequently distant metastases in BRCA-WT patients and contralateral breast cancer in BRCA positive patients. Time to progression was superimposable.
When survival curves are examined, one should take into account that the mechanism of the study caused long survivors to be overrepresented, which translates into very high long term survival rates (median survival has not been reached at 20 years). It appears that, while event free survival is superimposable, the percentage of long term survivors is higher among BRCA positive patients. This would be in accordance with a protective effect of BRCA, which counteracts the unfavourable prognostic factors associated with BRCA positivity. No difference emerged in our series between BRCA1 and BRCA2 positive patients, but the small number of BRCA1 cases in this predominantly BRCA2 series precludes conclusions. These findings are not entirely comparable with most data reported so far, in that in our series all patients had a family history of breast cancer and patients with BRCA-WT were probably positive for other unknown mutated genes. This is in agreement with a possible protective effect of BRCA positivity even within otherwise genetically related breast cancer.
Conclusion
In conclusion, in this series it appeared that BRCA status interacted with known prognostic factors in determining the eventual outcome.
Whether BRCA may become a clinically useful prognostic or predictive factor can be ascertained only by large randomized trials. Prospective correlations within ongoing randomized studies of family history and BRCA status with outcome appear to be warranted.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
AV conceived the study and drafted the manuscript; CdG, MDM, DL and DC collected the data and were involved in the drafting and revision of the paper; RD and AV performed BRCA tests, collected the data and were involved in the drafting and revision of the paper; EB performed the statistical analysis; MB supervised laboratory work and was involved in the drafting and revision of the paper
Pre-publication history
The pre-publication history for this paper can be accessed here:
Figures and Tables
Figure 1 Event free survival.
Figure 2 Overall survival.
Table 1 Patient Characteristics and their corresponding hazard ratios*
CHARACTERISTICS OF PATIENTS BRCA-WT (n = 86) BRCA+ (n = 29) p-value# Hazard Ratio (95% CI)
Average age in years (range) 45.3 (22–82) 42.3 (26–62) _ _
Laterality (right/left/bilat)+ 35/41/5 15/13/1 0.68 1.3 (0.1–5.7)
Premenopausal 53/81 (65%) 23/29 (79%) 0.17 1.9 (0.5–6.7)
T3-T4 8/83 (10%) 3/23 (13%) 0.64 1.2 (0.3–6.5)
N+ 37/76 (49%) 17/27 (63%) 0.20 1.4 (0.5–4.6)
G3 31/77 (40%) 19/25 (76%) 0.002 2.6 (0.8–7.2)
Multifocal-multicentric 38/82 (46%) 8/24 (33%) 0.26 0.8 (0.1–3.7)
ER- 20/70 (29%) 10/23 (43%) 0.18 1.5 (0.4–5.1)
Pgr- 22/72 (31%) 8/23 (35%) 0.71 1.1 (0.2–6.7)
HER2 2+/3+ 11/31 (35%) 2/13 (15%) 0.18 0.6 (0.02–4.9)
*analyses stratified according to age (one-year age group) and tumour grade
# univariate chi-square
+ Hazard Ratio not computed for bilaterality
==== Refs
Blackwood MA Weber BL BRCA1 and BRCA2: from molecular genetics to clinical medicine J Clin Oncol 1998 16 1969 1977 9586917
Chappuis PO Nethercot V Foulkes WD Clinico-pathological characteristics of BRCA-1 and BRCA-2 related breast cancer Semin Surg Oncol 2000 18 287 295 10805950 10.1002/(SICI)1098-2388(200006)18:4<287::AID-SSU3>3.0.CO;2-5
Albano W Recabaren J Lynch H Natural history of hereditary cancer of the breast and colon Cancer 1982 50 360 363 7083143
Andersson D Badzioch M Survival in familial breast cancer patients Cancer 1986 58 360 365 3719528
Chang J Elledge RM Clinical management of women with genomic BRCA1 and BRCA2 mutations Breast Cancer Res Treat 2001 69 101 113 11759816 10.1023/A:1012203917104
Foulkes WD Metcalfe K Hanna W Lynch HT Ghadirian P Tung N Olopade O Weber B McLennan J Olivotto IA Sun P Chappuis PO Bégin LR Brunet J-S Narod SA Disruption of the expected positive correlation between breast tumor size and lymph node status in BRCA-1 related breast carcinoma Cancer 2003 98 1569 1577 14534871 10.1002/cncr.11688
Foulkes WD Chappuis PO Wong N Brunet JS Vesprini D Rozen F Primary node negative breast cancer in BRCA1 mutation carriers has a poor outcome Ann Oncol 2000 11 307 313 10811497 10.1023/A:1008340723974
Ansquer Y Gautier C Fourquet A Asselain B Stoppa-Lyonnet D Survival in early-onset BRCA1 breast cancer patients: Institute Curie Breast Cancer Group Lancet 1998 352 541 9716060 10.1016/S0140-6736(05)79248-5
Haffty BG Harrold E Khan AJ Pathare P Smith TE Turner BC Glazer PM Ward B Carter D Matloff E Bale AE Alvarez-Franco M Outcome of conservatively managed early-onset breast cancer by BRCA 1/2 status Lancet 2002 359 1471 1477 11988246 10.1016/S0140-6736(02)08434-9
Johannsson OT Ranstam J Borg A Olsson H Survival of BRCA1 breast and ovarian cancer patients: a population-based study from southern Sweden J Clin Oncol 1998 16 397 404 9469321
Verhhoog LG Brekelmans CTM Seynaeve C Dahmen G van Geel AN Bartels CC Tilanus-Linthorst MM Wagner A Devilee P Halley DJ van den Ouweland AM Meijers-Heijboer EJ Klijn JG Survival in hereditary breast cancer associated with germline mutations of BRCA2 J Clin Oncol 1999 17 3396 3402 10550133
Eerola H Vahteristo P Sarantaus L Kyyronen P Pirhonen S Blomqvist C Pukkala E Nevanlinna H Sankila R Survival of breast cancer patients in BRCA1, BRCA2, and non-BRCA1/2 breast cancer families: a relative survival analysis from Finland Int J Cancer 2001 93 368 372 11433401 10.1002/ijc.1341
Nicoletto MO Donach M De Nicolo A Artioli G Banna G Monfardini S BRCA-1 and BRCA-2 mutations as prognostic factors in clinical practice and genetic counselling Cancer Treatment Reviews 2001 27 295 304 11871866 10.1053/ctrv.2001.0233
|
15996267
|
PMC1184063
|
CC BY
|
2021-01-04 16:03:06
|
no
|
BMC Cancer. 2005 Jul 4; 5:70
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-70
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-891604865410.1186/1471-2407-5-89Research ArticleOff-trial evaluation of bisphosphonates in patients with metastatic breast cancer Liauw Winston [email protected] Eva [email protected] Anna [email protected] Ms Ruth [email protected] Matthew [email protected] Robyn [email protected] Clinical Trials Centre, St Vincent's Hospital, Sydney, Australia2 Cancer Care Centre, St George Hospital, Sydney, Australia3 Department of Medical Oncology, St Vincent's Hospital, Sydney, Australia4 Concord Hospital, Concord. Sydney, Australia5 Faculty of Medicine, University of New South Wales, Sydney, Australia2005 28 7 2005 5 89 89 8 3 2005 28 7 2005 Copyright © 2005 Liauw et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Bisphosphonate therapy has been readily accepted as standard of care for individuals with bone metastases from breast cancer. In this study we determined whether the proportion of patients experiencing a skeletal related event (SRE) in a clinical practice population was similar to that observed in phase III randomized controlled studies.
Methods
A retrospective chart review was conducted of 110 patients receiving intravenous bisphosphonates for advanced breast cancer. The proportion of patients experiencing at least one SRE after 12 months of therapy was determined. SRE included vertebral or non-vertebral fracture, cord compression, surgery and/or radiotherapy to bone.
Results
The proportion of patients who had an SRE was 30% (28 individuals) and the median time to first event was greater than 350 days. Non-vertebral events and radiotherapy were the most frequent type of SRE, while cord compression and hypercalcaemia were rare (1%). Most patients in the study had bone-only disease (58.2%) and most had multiple bone lesions. In the first 12 months the mean duration of exposure to intravenous bisphosphonates was 261 days and most patients remained on treatment until just before death (median 27 days).
Conclusion
This study suggests that the rate of clinically relevant SREs is substantially lower than the event rate observed in phase III clinical trials. We attribute this lower rate to observational bias. In the clinical trial setting it is possible that over-detection of skeletal events occurs due to the utilisation of regular skeletal survey or radionucleotide bone scan, whereas these procedures are not routine in clinical practice. Phase IV observational studies need to be conducted to determine the true benefits of bisphosphonate therapy in order to implement rationale use of bisphosphonates.
==== Body
Background
Randomised controlled studies have consistently demonstrated that the skeletal complications of metastatic breast cancer can be reduced by the regular administration of intravenous bisphosphonates [1-7]. Given the potential impact of a pathological fracture, it is not surprising that bisphosphonate therapy has been readily accepted as standard of care for metastatic bone disease [8-10]. However, this therapy is expensive, and since bisphosphonates have no impact on survival, their cost-effectiveness is primarily justified by the avoidance of radiotherapy or surgery [11,12].
To date, clinical trials of bisphosphonate therapy have failed to determine the optimal frequency of administration, timing of initiation or duration of use [9]. In practice, patients are treated on a three to four weekly basis for an indefinite period. Until their death, individuals with advanced cancer are therefore exposed to the risk of infusion-related adverse events, the possibility of nephrotoxicity, and the inconvenience of intravenous treatment. In an effort to promote rational prescribing of bisphosphonates, a number of medical and funding agencies have developed treatment guidelines which include rules for initiation and cessation of treatment [9,10]. Unfortunately, recent studies have demonstrated that adherence to these guidelines is universally poor [10,13,14]. We propose that prescribing habits are unlikely to change without evidence of the efficacy, impact on quality of life, and cost-effectiveness of bisphosphonates in routine clinical practice.
To the knowledge of the investigators, there has been no report of the outcomes of patients receiving bisphosphonates outside the conduct of a clinical trial. Given the impact of inappropriate prescribing, it seems unreasonable to assume that the benefits observed in randomized controlled studies will be perfectly replicated in a clinical practice population. Furthermore, individuals in the key randomized studies underwent regular skeletal surveys, and the skeletal-related events (SRE) identified in these trials represented the composite of clinically relevant as well as asymptomatic radiological changes [1-6]. Skeletal surveys are not used routinely in clinical practice, and thus only clinically relevant skeletal events are likely to be identified in this setting. In this study, we report the results of an audit of intravenous bisphosphonate use in patients with metastatic breast cancer in the setting of routine clinical practice.
Methods
Study cohort
110 women with metastatic breast cancer were included in this study. They had commenced intravenous pamidronate or zoledronic acid for the prophylaxis of skeletal complications between January 1998 and September 2003. Patients were excluded if they were receiving bisphosphonates therapy for osteoporosis or for the management of tumour-related hypercalcaemia. Individuals were identified through the pharmacy records of two large comprehensive cancer centers in Australia; St Vincent's Hospital, Darlinghurst, NSW, and St George Hospital, Kogarah, NSW. The dataset was extracted from medical and pharmacy records by three investigators (AL, RD, ES) and entered into a standardised data collection form. A random sample of forms was audited by an independent investigator (WL) and data queries were resolved by case review by three investigators (WL, ES, RW). Relevant clinical information as well as the details of bisphosphonates usage was recorded on each person for a period of 12 months from the time of commencement of bisphosphonate therapy.
The primary outcome measure used in this study was the proportion of patients experiencing at least one SRE in a 12 month period following the commencement of intravenous bisphosphonate therapy. SRE were defined as either a pathological fracture, a bony lesion requiring intervention (surgery or radiotherapy) for pain or prevention of skeletal complications, spinal cord compression or hypercalcaemia of malignancy. The total number of SRE did not include events occurring within 30 days of commencement of intravenous bisphosphonate therapy if they were related to the index presentation of bony disease. Simultaneous SRE (i.e. presenting on the same day) were counted as one event. Treatment of a single lesion with radiotherapy and surgery were not coded as separate SREs.
The study was conducted under guidelines consistent with the NHMRC National Statement of Ethical Conduct in Research Involving Humans.
Statistical analysis
Descriptive statistics were used to define the characteristics of the sample and Kaplan-Meier (KM) estimates were used to calculate the proportion with an SRE. The KM technique was used to estimate the time from commencement of bisphosphonate administration to the first SRE. Only SRE not cancer-related deaths were analysed as events. All data was analysed using SPSS statistical software V11.0 (SPSS Inc., Chicago, IL).
Results
Patient characteristics and bisphosphonate administration
The baseline clinical characteristics of the study cohort are shown in Table 1. At initiation of bisphosphonates therapy the mean age of the group was 57.2 ± 12.1 years and most patients (50%) had multiple bone metastases. On average, skeletal disease had developed 4.2 ± 4.4 years from the primary diagnosis of breast cancer and bisphosphonate therapy was commenced at a mean of 291 ± 497 days from the time of diagnosis of bone metastases. Bisphosphonates were commenced concurrently with chemotherapy (either alone or with hormone therapy) in 54.4% of patients, with hormonal therapy in 40.4% of patients, and without any systemic treatment in 5.3% of individuals. In the first 12 months the mean duration of exposure to intravenous bisphosphonates was 261 days, and a total of 60 patients (54.5%) remained on therapy beyond the 12 month observation period. The mean and median number of infusions of intravenous bisphosphonates in the first 12 months of follow-up was 9 and 12 respectively. Over the total duration of follow-up a mean of 16 cycles of IV bisphosphonate were received, with an upper range of 56 cycles. The reasons for discontinuation of bisphosphonates within the 12 month observation period were cancer death (24 patients, 22%); change to oral bisphosphonates (13 patients, 12%); intolerance (5 patients, 4.5%) or not specified (3 patients, 2.7%). A significant number of patients remained on therapy despite being close to death from cancer progression. In the 24 patients who died within the study period, the time from last documented bisphosphonate infusion to death was between 2 and 140 days with a median of 27 days. Although bisphosphonate treatment guidelines state that an abnormal bone scan without evidence of bone destruction does not justify the initiation of bisphosphonates, we found that 21% (23 individuals) of individuals in this audit had commenced treatment on this basis.
Skeletal complications
The proportion of patients experiencing at least one SRE within 12 months of commencing bisphosphonates was 30% (28 individuals). The characteristics of these events, and the proportion of patients experiencing them, are described in Table 2. Spinal cord compression and hypercalcaemia were rare events, occurring in only 1% of individuals within the study period. The median time to the first occurrence of a SRE was greater than 365 days (Figure 1). Within the 12 month observation period, nine of 28 individuals developed a second SRE at a mean of 223 days from bisphosphonate initiation, and of these six were still undergoing treatment with bisphosphonate at the time of the second SRE. Three individuals developed a fourth SRE and one a fifth SRE within 12 months of starting bisphosphonates.
Interestingly, when patients with bone scan only disease were excluded from the analysis the cumulative proportion of individuals developing a SRE within the first year of commencing bisphosphonates was 32% (23 of 87 patients). There was no statistically significant difference between subjects with bone scan-only detected metastases and the remainder of the cohort in the proportion of subjects with an SRE in the first 12 months (p = 0.77). This data suggests that bone scan only disease does not represent a better prognostic group of patients.
Discussion
Over 80% of women with metastatic breast cancer have bone metastases [15,16], yet a much smaller proportion of these will develop clinically apparent complications related to bone destruction. One large retrospective study performed before the introduction of bisphosphonates demonstrated that 29% of individuals develop a clinically significant SRE [15]. This figure is comparable to the findings of the current study (30%) yet different from the results of the randomized controlled trials of intravenous bisphosphonates [1-6]. In these latter studies, the proportion of individuals who experienced at least one skeletal complication at 12 months from bisphosphonate commencement was 43% in the treatment arm compared with 56% in the placebo arm. We propose that the discrepancy between the results of the randomized controlled studies and our current audit relates primarily to the use of regular radiographic skeletal surveys in the trial setting [4]. In practice, skeletal events are only identified on the basis of clinical suspicion and thus many bone lesions may be appropriately undetected and untreated. The corollary is of course that intensive exposure to skeletal surveys may result in the treatment of lesions which may never become clinically significant.
The baseline population risk of a skeletal event is a further factor which may explain the discrepancies between the current audit and the results of previous randomized controlled studies. Patients with metastatic disease limited to bone are four times more likely to fracture a long bone than those with concurrent liver and bony metastases, while individuals with extensive metastasis involving long bones are more likely to develop a fracture than those with solitary metastases [17]. In terms of these risk factors, our patient population was certainly at no lesser risk of fracture than the individuals enrolled in the studies of Hortobagyi et al [1,2] and Theriault [3]. The percentage of individuals with metastatic disease limited to bone was comparable at 58.2% (current audit), 60% (Hortobagyi studies) and 70% (Theriault et al,). Furthermore, only 9.1% of the subjects in the current study had a solitary bone metastasis, whereas 43% of patients in the Hortobagyi studies had an isolated lesion [1,2].
The endpoint used in this study, specifically the proportion of patients with more than one SRE is well accepted and provides readily assessable and comparable estimations of treatment effect. It however only captures information about the first event and does not measure the multiple skeletal events which sometimes occur in a single individual. While acknowledging that the impact of bisphosphonate therapy on recurrent events is important the best way to measure and analyse this data is controversial [18-21]. For this reason, we chose not to evaluate secondary endpoints such as total number of SRE, and event rate in the current audit.
Given the relative risk reduction in skeletal related events bisphosphonates clearly have a place in the management of women with metastatic bone disease. It is uncertain however what the absolute benefits of these reductions are, whether in terms of SRE, or economic endpoints. The rationale use of these agents is hindered by a lack of data concerning drug scheduling, duration of use, indications for initiation and cessation [9]. This audit highlights the fact that while these questions remain unanswered, clinicians will be reluctant to alter current prescribing habits. Most patients remained on bisphosphonates despite a declining performance status; the drugs were indefinitely administered on a three-four weekly basis and 67% of patients who experienced a second SRE continued bisphosphonate therapy.
Conclusion
The findings of this study provide an additional impetus to proceed with post-marketing evaluation of the use of bisphosphonates in clinical practice.
Competing interests
RW is a member of the Pharmaceutical Benefits Advisory Committee (PBAC), Commonwealth Department of Health and Ageing, Canberra, ACT, Australia. The views presented here are those of the authors and should not be understood or quoted as being made on behalf of the PBAC or its Scientific Committees. The other authors declare that they have no competing interests.
Authors' contributions
WL carried out statistical analysis, collected data and prepared draft versions of the manuscript. ES collected data and prepared draft versions of the manuscript. AL collected data. RD collected data. ML collected data and participated in the design of the study. RW conceived and coordinated the study, carried out statistical analysis and prepared the final versions of the manuscript. The final manuscript was approved by all authors.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The study was performed with the support of the South East Area Health Service and St Vincents Hospital.
Figures and Tables
Figure 1 Kaplan-Meier estimates of the time to first skeletal related event from the date of first intravenous bisphosphonate infusion. Censored at 12 months follow-up.
Table 1 Disease characteristics of patients
Number of patients Percent (%)
Sites of extra-osseous disease
None 64 58.2
Liver Only 14 12.7
Brain Only 1 0.9
Lung Only 3 2.7
Multiple 11 10.0
Other 15 13.6
Unknown 2 1.8
Number of bone lesions
1 10 9.1
2–5 40 36.4
>5 55 50.0
Unknown 5 4.5
Treatment prior to bisphosphonates
Nil 33 30.0
Chemotherapy +/- hormone 31 28.2
Hormone therapy alone 46 41.8
Hormone therapy prior to bisphosphonates
1st line 31 28.2
2nd line 19 17.3
3rd line 3 2.7
Not applicable 57 51.8
Median survival from start of bisphosphonates Days (95% CI)
All subjects 818 (644–911)
Subjects with bone metastases only 998 (653–1342)
Subjects with extra-osseous metastases 513 (339–686)
* based on available imaging, either bone scan, skeletal survey, CT or MRI.
Table 2 Proportion of patients having each type of SRE in the first 12 months of intravenous bisphosphonate therapy
Number of patients ¥Proportion with event (%)
Type of SRE
Pathological fracture 7 9%
Cord compression 1 1%
Other 20 23%
Site of SRE
Vertebral 9 10%
Non-vertebral 17 22%
Hypercalcaemia 1 1%
Treatment for SRE
Radiotherapy 17 21%
Surgery 3 4%
Medical+ 8 10%
¥Kaplan-Meier estimates of the proportion with a complication. + Medical includes change in analgesics, hormonal treatment or chemotherapy.
==== Refs
Hortobagyi GN Theriault RL Porter L Blayney D Lipton A Sinoff C Wheeler H Simeone JF Seaman J Knight RD Efficacy of pamidronate in reducing skeletal complications in patients with breast cancer and lytic bone metastases. Protocol 19 Aredia Breast Cancer Study Group N Engl J Med 1996 335 1785 1791 8965890 10.1056/NEJM199612123352401
Hortobagyi GN Theriault RL Lipton A Porter L Blayney D Sinoff C Wheeler H Simeone JF Seaman JJ Knight RD Heffernan M Mellars K Reitsma DJ Long-term prevention of skeletal complications of metastatic breast cancer with pamidronate. Protocol 19 Aredia Breast Cancer Study Group J Clin Oncol 1998 16 2038 2044 9626201
Theriault RL Lipton A Hortobagyi GN Leff R Gluck S Stewart JF Costello S Kennedy I Simeone J Seaman JJ Knight RD Mellars K Heffernan M Reitsma DJ Pamidronate reduces skeletal morbidity in women with advanced breast cancer and lytic bone lesions: a randomized, placebo-controlled trial. Protocol 18 Aredia Breast Cancer Study Group J Clin Oncol 1999 17 846 854 10071275
Rosen LS Gordon D Kaminski M Howell A Belch A Mackey J Apffelstaedt J Hussein M Coleman RE Reitsma DJ Seaman JJ Chen BL Ambros Y Zoledronic acid versus pamidronate in the treatment of skeletal metastases in patients with breast cancer or osteolytic lesions of multiple myeloma: a phase III, double-blind, comparative trial Cancer J 2001 7 377 387 11693896
Rosen LS Gordon D Kaminski M Howell A Belch A Mackey J Apffelstaedt J Hussein MA Coleman RE Reitsma DJ Chen BL Seaman JJ Long-term efficacy and safety of zoledronic acid compared with pamidronate disodium in the treatment of skeletal complications in patients with advanced multiple myeloma or breast carcinoma: a randomized, double-blind, multicenter, comparative trial Cancer 2003 98 1735 1744 14534891 10.1002/cncr.11701
Rosen LS Gordon DH Dugan WJ Major P Eisenberg PD Provencher L Kaminski M Simeone J Seaman J Chen BL Coleman RE Zoledronic acid is superior to pamidronate for the treatment of bone metastases in breast carcinoma patients with at least one osteolytic lesion Cancer 2004 100 36 43 14692022 10.1002/cncr.11892
Kohno N Aogi K Minami H Nakamura S Asaga T Iino Y Watanabe T Goessl C Ohashi Y Takashima S Zoledronic acid significantly reduces skeletal complications compared with placebo in Japanese women with bone metastases from breast cancer: a randomized, placebo-controlled trial J Clin Oncol 2005 23 3314 3321 15738536 10.1200/JCO.2005.05.116
Pavlakis N Stockler M Bisphosphonates for breast cancer Cochrane Database Syst Rev 2002 CD003474 11869664
Hillner BE Ingle JN Chlebowski RT Gralow J Yee GC Janjan NA Cauley JA Blumenstein BA Albain KS Lipton A Brown S American Society of Clinical Oncology 2003 update on the role of bisphosphonates and bone health issues in women with breast cancer J Clin Oncol 2003 21 4042 4057 12963702 10.1200/JCO.2003.08.017
Verma S Kerr-Cresswell D Dranitsaris G Charbonneau F Trudeau M Yogendran G Cesta AM Clemons M Bisphosphonate use for the management of breast cancer patients with bone metastases: A survey of Canadian Medical Oncologists Support Care Cancer 2004 12 852 858 15322969 10.1007/s00520-004-0671-9
Hillner BE Weeks JC Desch CE Smith TJ Pamidronate in prevention of bone complications in metastatic breast cancer: a cost-effectiveness analysis J Clin Oncol 2000 18 72 79 10623695
Body JJ Effectiveness and cost of bisphosphonate therapy in tumor bone disease Cancer 2003 97 859 865 12548587 10.1002/cncr.11139
Clemons M Enright K Cesta A Charbonneau F Chow E Warr D Kee-Cresswell D Chang J Yogendran G Trudeau M De Angelis C Cottrell W Dranitsaris G Do physicians follow systemic treatment and funding policy guidelines? Can J Clin Pharmacol 2004 11 e168 78 15300959
Enright K Clemons M Chow E Utilization of palliative radiotherapy for breast cancer patients with bone metastases treated with bisphosphonates-Toronto Sunnybrook Regional Cancer Centre experience Support Care Cancer 2004 12 48 52 14577020 10.1007/s00520-003-0548-3
Coleman R Rubens R The clinical course of bone metastases from breast cancer British Journal of Cancer 1987 55 61 66 3814476
Coleman RE Skeletal complications of malignancy Cancer 1997 80 1588 1594 9362426 10.1002/(SICI)1097-0142(19971015)80:8+<1588::AID-CNCR9>3.0.CO;2-G
Plunkett TA Smith P Rubens RD Risk of complications from bone metastases in breast cancer. implications for management Eur J Cancer 2000 36 476 482 10717523 10.1016/S0959-8049(99)00331-7
Cook RJ Lawless JF Interim monitoring of longitudinal comparative studies with recurrent event responses Biometrics 1996 52 1311 1323 8962455
Ghosh D Lin DY Nonparametric analysis of recurrent events and death Biometrics 2000 56 554 562 10877316 10.1111/j.0006-341X.2000.00554.x
Major PP Cook R Efficacy of bisphosphonates in the management of skeletal complications of bone metastases and selection of clinical endpoints Am J Clin Oncol 2002 25 S10 8 12562046 10.1097/00000421-200212001-00003
Coleman RE Bisphosphonates: clinical experience Oncologist 2004 9 14 27 15459426 10.1634/theoncologist.9-90004-14
|
16048654
|
PMC1184064
|
CC BY
|
2021-01-04 16:03:05
|
no
|
BMC Cancer. 2005 Jul 28; 5:89
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-89
|
oa_comm
|
==== Front
BMC Dev BiolBMC Developmental Biology1471-213XBioMed Central London 1471-213X-5-111597210510.1186/1471-213X-5-11Research ArticleFGF10/FGFR2b signaling plays essential roles during in vivo embryonic submandibular salivary gland morphogenesis Jaskoll Tina [email protected] George [email protected] Daniel [email protected] Frederic G [email protected] Saverio [email protected] Mohammad K [email protected] Michael [email protected] Laboratory Developmental Genetics, University of Southern California, Los Angeles, CA, USA2 Department of Surgery, and Division of Developmental Biology, the Saban Research Institute of Children's Hospital Los Angeles, Los Angeles, CA, USA3 School of Biological Sciences, University of East Anglia (UEA), Norwich – Norfolk, UK2005 22 6 2005 5 11 11 31 3 2005 22 6 2005 Copyright © 2005 Jaskoll et al; licensee BioMed Central Ltd.2005Jaskoll et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Analyses of Fgf10 and Fgfr2b mutant mice, as well as human studies, suggest that FGF10/FGFR2b signaling may play an essential, nonredundant role during embryonic SMG development. To address this question, we have analyzed the SMG phenotype in Fgf10 and Fgfr2b heterozygous and null mutant mice. In addition, although previous studies suggest that the FGF10/FGFR2b and FGF8/FGFR2c signaling pathways are functionally interrelated, little is known about the functional relationship between these two pathways during SMG development. We have designed in vivo and in vitro experiments to address this question.
Results
We analyzed Fgf10 and Fgfr2b heterozygous mutant and null mice and demonstrate dose-dependent SMG phenotypic differences. Hypoplastic SMGs are seen in Fgf10 and Fgfr2b heterozygotes whereas SMG aplasia is seen in Fgf10 and Fgfr2b null embryos. Complementary in vitro studies further indicate that FGF10/FGFR2b signaling regulates SMG epithelial branching and cell proliferation. To delineate the functional relationship between the FGF10/FGFR2b and FGF8/FGFR2c pathways, we compared the SMG phenotype in Fgfr2c+/Δ/Fgf10+/- double heterozygous mice to that seen in wildtype, Fgf10+/- (Fgfr2c+/+/Fgf10+/-) and Fgfr2c+/Δ (Fgfr2c+/Δ/Fgf10+/+) single heterozygous mutant littermates and demonstrate genotype-specific SMG phenotypes. In addition, exogenous FGF8 was able to rescue the abnormal SMG phenotype associated with abrogated FGFR2b signaling in vitro and restore branching to normal levels.
Conclusion
Our data indicates that FGF10/FGFR2b signaling is essential for the SMG epithelial branching and histodifferentiation, but not earliest initial bud formation. The functional presence of other endogenous signaling pathways could not prevent complete death of embryonic SMG cells in Fgf10 and Fgfr2b null mice. Though we were able to rescue the abnormal phenotype associated with reduced in vitro FGF10/FGFR2b signaling with exogenous FGF8 supplementation, our results indicate that the FGF10/FGFR2b and FGF8/FGFR2c are nonredundant signaling pathways essential for in vivo embryonic SMG development. What remains to be determined is the in vivo functional relationship between the FGF10/FGFR2b signal transduction pathway and other key signaling pathways, and how these pathways are integrated during embryonic SMG development to compose the functional epigenome.
==== Body
Background
Mouse submandibular gland (SMG) development is initiated with a thickening of the oral epithelium of the mandibular arch around embryonic day 11.5 (E11.5) and is best conceptualized in stages[1,2]. In the PreBud Stage, SMG development begins as thickening of the oral epithelium adjacent to the tongue. During the Initial Bud Stage, this thickening grows down into mandibular arch mesenchyme to form the initial SMG bud. With continued epithelial proliferation and downgrowth, the SMG primordium becomes a solid, elongated epithelial stalk terminating in a bulb. Repeated end-bud branching results in the formation of a network of epithelial branches and terminal buds (the Pseudoglandular Stage). These epithelial branches and terminal buds hollow out by cell apoptosis during the Canalicular and Terminal Bud Stages, respectively, to form the ductal system and presumptive acini, with mucin protein being produced by the presumptive acini.
Morphogenesis of complex organs such as the SMG is regulated by the functional integration of parallel and broadly related signaling pathways which regulate cell proliferation, apoptosis and histodifferentiation [3-6]. To understand the complex interactions within this dynamic signaling network, one must first determine the contribution of individual pathways and identify those which play essential, nonredundant roles during embryonic SMG initial bud formation, branching morphogenesis and histodifferentiation.
The FGF family, with at least 22 members, mediates diverse biological functions such as cell proliferation, branching morphogenesis and histodifferentiation by binding and activating four tyrosine kinase receptors (FGFR 1–4) [see reviews [7-9]]. Tissue-specific alternate splicing of the Fgfr1, Fgfr2 and Fgfr3 genes generates isoforms which differentially bind specific FGF ligands [8,10]. Ligand-receptor binding potentially activates multiple intracellular cascades, including the ERK/RAS/MAPK, P13K, and PLC-γ/PKC pathways [see reviews [8,9,11]].
Functional studies have demonstrated that embryonic SMG epithelial cell proliferation, branching morphogenesis, and histodifferentation are regulated through growth factor, cytokine, and transcription factor-mediated signaling pathways, including EGF, TGF-β, Shh, FGFs, and Eda [2,6,12-21]. Although FGFs have been implicated in embryonic SMG development [13,15,22-24], a more complete understanding of their precise roles will provide insight into the complex network of parallel and broadly-related signaling pathways which regulate SMG organogenesis.
Gene targeting studies have clearly shown the importance of the FGFR2b (FGFR2-IIIb) signaling pathway for embryonic morphogenesis [22,23,25]. Fgfr2b-/- null mice die at birth due to lung insufficiency and exhibit severe dysmorphic organs, including agenesis or dysplasia of the lungs, mammary glands, pancreas, thyroid, teeth, and limbs [22,23,25,26]. Although FGF1, FGF3, FGF7, and FGF10 bind with high affinity to FGFR2b [8,10], the phenotypic similarities between Fgf10 and Fgfr2b null mice [22-25,27] indicate that FGF10 is the major ligand for FGFR2b in vivo. Of particular interest is the absence of SMGs in E14.5 and older Fgfr2b null mice and newborn Fgf10 null mice [23-25]. However, whether this glandular absence is due to the lack of development of a SMG initial bud (i.e. agenesis) or subsequent aplasia of the initial bud was heretofore unknown.
Recently, Entesarian et al. [28] have shown the importance of FGF10 gene dosage for salivary gland development in humans. Individuals with autosomal dominant ALSG (aplasia of lacrimal and salivary gland) exhibit hypoplastic or absent parotid and submandibular glands. ALSG was mapped to 5p13.2-5q13.1 to include the FGF10 gene; heterozygous FGF10 mutations were identified in all family members with ALSG. Complementary study of adult Fgf10+/- mutant mice revealed that Fgf10 heterozygotes have absent parotid glands and smaller SMGs although other organs such as lungs, liver, spleen, pancreas, thyroid, limbs appeared normal.
Taken together, the literature suggests that FGF10/FGFR2b signaling plays an essential, nonredundant, dose-dependent role during embryonic SMG development. To address this postulate, we evaluated SMG development in Fgfr2b and Fgf10 heterozygous mutant and null mice and demonstrate dose-dependent differences in SMG phenotypes. In a complementary set of in vitro experiments, we confirm the importance of FGF10/FGFR2b signaling and demonstrate that enhanced FGF10/FGFR2b signaling significantly induces, and abrogated FGF10/FGFRb signaling significantly diminishes, SMG branching morphogenesis and cell proliferation.
Our previous analyses of mutant mice and functional in vitro studies indicate that the FGF8/FGFR2c signaling pathway is essential for embryonic SMG epithelial branching morphogenesis and histodifferentiation [2,13]. Little is known about the functional relationship between the FGF10/FGFR2b and FGF8/FGFR2c pathways during SMG development. Thus, we evaluated the SMG phenotype in Fgfr2c+/Δ/Fgf10+/- double heterozygous mice, compared them to Fgf10+/- (Fgfr2c+/+/Fgf10+/-) and Fgfr2c+/Δ (Fgfr2c+/Δ/Fgf10+/+) single heterozygous mutant, as well as wildtype (WT), littermates and demonstrate genotype-specific SMG phenotypes. Though we were able to rescue the abnormal phenotype associated with reduced in vitro FGF10/FGFR2b signaling with exogenous FGF8 peptide supplementation, our results indicate that the FGF10/FGFR2b and FGF8/FGFR2c are nonredundant signaling pathways essential for in vivo embryonic SMG development.
Results
To delineate the role of FGF10/FGFR2b signaling during embryonic SMG development, we evaluated the SMG phenotype in Fgf10 and Fgfr2b null and heterozygous mutant mice. The E12.5 normal SMG appears as an elongated solid cord of epithelium terminating in an end-bulb (i.e., Initial Bud Stage) (Fig. 1A). By contrast, the E12.5 Fgfr2b null SMG is severely hypoplastic (compare Fig. 1B to 1A), displaying an extremely small initial bud similar to the earliest Initial Bud Stage [1,2]. A similar phenotype is seen in Fgf10 null mice (compare Fig. 1C to 1A, B). Normally by E13.5, epithelial proliferation results in SMG primordia with end-bulbs characterized by several branches; SMGs from both Fgfr2b and Fgf10 null mice are entirely absent (compare Fig. 2B, C to 2A). On the other hand, Fgfr2b+/- and Fgf10+/- heterozygosity results in SMG branching hypoplasia compared to WT glands, with fewer ducts and terminal buds being seen in heterozygous mutant SMGs than in WT SMG (compare Fig. 3B to 3A, D to 3C). Taken together, our data indicate that FGF10/FGFR2b signaling plays an essential, dose-dependent role during in vivo embryonic SMG branching morphogenesis and histodifferentation, but not earliest initial bud formation. Thus, the pathology in null mice is aplasia, not agenesis.
Figure 1 Abnormal SMG phenotypes in E12.5 Fgfr2b and Fgf10 null mutants A. E12.5 WT (+/+) mouse. B. E12.5 Fgfr2b-/- (R2b-/-) null mouse. C. E12.5 Fg10-/- null mouse. In the WT mouse (A), the Initial Bud Stage SMG is seen in the mandible ventrolateral to the tongue. By contrast, the SMG bud (outlined in white) in Fgfr2b-/- (B) and Fg10-/- (C) mice is extremely small, resembling the earliest Initial Bud Stage SMG. Bar, 50 μm.
Figure 2 SMG aplasia in Fgfr2b and Fgf10 null mice. A. E13.5 WT mouse. B. E13.5 Fgfr2b-/- null mouse. C. E13.5 Fg10-/- null mouse. The E13.5 WT SMG (A) has achieved the Late Initial Bud Stage, with several branches being seen in the end-bulb epithelium; a small sublingual gland (SL) bud is seen lateral to the SMG bud. In Fgfr2b-/- (B) and Fg10-/- (C) embryos, no SMGs are found; undifferentiated mesenchyme is seen in the site normally occupied by SMG epithelia. PS-palatal shelf. Bar, 50 μm.
Figure 3 Hypoplastic SMGs are seen in Fgfr2b and Fgf10 heterozygous mice. A. Newborn WT Fgfr2b+/+ SMG. B. Newborn Fgfr2b+/- heterozygous SMG. C. Newborn WT Fgf10+/+ SMG. D. Newborn Fg10+/- heterozygous SMG. WT Fgfr2b+/+ (A) and Fgf10+/+ (C) mice exhibit Late Terminal Bud Stage SMGs consisting of ducts and terminal buds displaying distinct lumina. In contrast, fewer ducts and terminal buds are seen in Fgfr2b+/- (B) and Fg10+/- (D) heterozygous SMGs compared to WT littermates. Bar, 50 μm.
Enhanced FGF10/FGFR2b signaling in vitro induces embryonic SMG branching morphogenesis and epithelial cell proliferation
To further delineate the role of FGF10/FGFR2b signaling during embryonic SMG development, we used our well-defined organ culture system to analyze the effect of enhanced FGF10/FGFR2b signaling on embryonic SMG branching morphogenesis. Paired E13 (Initial Bud Stage) or E14 (Pseudoglandular Stage) SMG primordia were cultured for up to 3 days in the presence or absence of FGF10 peptide (200 ng/ml or 500 ng/ml). Since a notable difference in SMG branch number is usually seen among littermates, we compared the number of terminal buds in right and left glands (treated and control) from each embryo. Spooner ratios (end bud number/initial bud number) were determined for each explant, the data were then arcsin transformed, and the mean ratios compared by paired t-test. FGF10 supplementation induced a significant increase in branching morphogenesis (Fig. 4A, B, E): 81% for E13 + 3 [200 ng/ml (P < 0.01); 500 ng/ml (P < 0.001)] and 46% for E14 + 2 [500 ng/ml (P < 0.0001)].
Figure 4 Enhanced or abrogated FGF10/FGFR2b signaling modulates embryonic SMG branching morphogenesis in vitro. A-B. Enhanced signaling. Paired E13 SMG primordia were cultured for 3 days in the absence (A) or presence (B) of 500 ng/ml FGF10 peptide supplementation. FGF10 induced a significant increase in branching compared to control (CONT). C-D. Abrogated signaling. Paired E13 SMG primordia were cultured for 3 days in 10 ng/ml IgG-Fc (C) or FGFR2b-Fc chimera (D). FGFR2b-Fc chimera treated explant exhibits a significant decrease in branching compared to IgG-Fc control. Bar, 30 μm. E. Comparison of mean Spooner ratios in E13 + 3 and E14 + 2 explants. Exogenous FGF10 peptide supplementation induced a significant 81% increase in E13 +3 explants [200 ng/ml (P < 0.01); 500 ng/ml (P < 0.001)] and a significant 46% increase in E 14 + 2 explants (P < 0.0001) compared to controls. FGFR2b-Fc chimera-mediated interruption of E13 + 3 and E 14 + 2 SMGs significantly reduced branching morphogenesis by 22% [E13 +3: 10 ng/ml (P < 0.01); E 14 + 2: 5 ng/ml (P > 0.0001); 10 ng/ml (P < 0.01)] compared to controls.
Since cell proliferation is not required for early embryonic SMG epithelial branching [29], we determined if this FGF10-induced increase in branching is due to increased epithelial cell proliferation. We cultured E13+3 SMG primordia in the presence or absence of 500 ng/ml FGF10 peptide and calculated the epithelial cell proliferation index (the number PCNA-positive cells/total number of cells). Exogenous FGF10 induced a significant 78% (P < 0.0001) increase in epithelial cell proliferation compared to control (Fig. 5). Our results are similar to enhanced pancreatic epithelial cell proliferation and pancreatic hyperplasia in transgenic mice with persistent Fgf10 expression in developing pancreatic epithelia [30].
Figure 5 Enhanced or abrogated FGF10/FGFR2b signaling modulates cell proliferation. The cell proliferation index of E13 + 3 explants was calculated as the number of PCNA positive epithelial cells/total epithelial cells. Exogenous FGF10 peptide induced a significant 78% (P < 0.0001) increase, and FGFR2b-Fc chimera resulted in a significant 32% (P < 0.001) decrease, in cell proliferation compared to control.
Abrogated FGF10/FGFR2b signaling in vitro decreases embryonic SMG branching morphogenesis and cell proliferation
We interrupted FGF10/FGFR2b signaling in vitro by adding exogenous soluble FGFR2b-Fc chimera to the culture medium to competitively bind endogenous FGFR2b ligands. This exogenous receptor/ligand binding methodology has been successfully used to interrupt FGFR2b, FGFR2c, FGFR1b and FGFR1c signaling in vitro [13,15,21]. This experiment was designed to provide an in vitro model of the in vivo Fgf10+/- heterozygous mutant SMG phenotype (i.e. SMG hypoplasia). We conducted dose-response studies and determined the concentration of FGFR2b-Fc chimera which induces a clear biologic effect of reduced branching morphogenesis, thereby creating an in vitro model of the in vivo Fgf10 mutant heterozygote. Paired E13+3 or E14+2 SMG primordia were cultured in IgG-Fc (5 ng/ml or 10 ng/ml) or FGFR2b-Fc chimera (5 ng/ml or 10 ng/ml) and Spooner ratios were determined as described above. E13 + 3 FGFR2b-Fc-treated explants exhibit a significant 22% [10 ng/ml (P < 0.01)] decrease in branching morphogenesis compared to controls (Figs. 4C, D, E); a similar 22% reduction [5 ng/ml (P < 0.001); 10 ng/ml (P < 0.01)] was also seen in E14+2-treated explants (Figs. 4E). These results mimic the in vivo heterozygous mutants reported above.
Since exogenous FGF10 supplementation in vitro induced a significant increase in cell proliferation, we then evaluated the epithelial cell proliferation index in E13 +3 SMG primordia with or without abrogated FGF10/FGFR2b signaling. A significant 32% (P < 0.001) decrease in cell proliferation was seen in the presence of 10 ng/ml FGFR2b-Fc chimera compared to control (Fig. 5). Reduction in cell proliferation with FGFR2b-Fc treatment in vitro was also reported by Steinberg et al., [21]. These results indicate that FGF10/FGFR2b signaling modulates embryonic SMG epithelial cell proliferation and branching morphogenesis.
Relationship between FGFR2b and FGFR2c signaling pathways during SMG development
Our previous observation of hypoplastic SMGs in Fgfr2c+/Δ deficient mice suggested that FGFR2c signaling is necessary for embryonic SMG branching morphogenesis and histodifferentation [2]. The functional importance of this signaling pathway was confirmed in vitro; reduction of FGFR2c signaling in vitro resulted in a significant dose-dependent decrease in branching morphogenesis [13]. In addition, FGF8, a major FGFR2c ligand, has been shown to play an essential, nonredundant role during embryonic SMG development [13]. Like the hypoplastic SMG phenotype of Fgfr2c deficient mice, hypoplastic glands are seen in Fgf8 hypomorphic mice [13]. Importantly, SMG aplasia is seen in Fgf8 tissue-specific conditional mutant mice. To investigate the relationship between the FGFR2b and FGFR2c signal transduction pathways during in vivo SMG development, we evaluated the SMG phenotype in mice which are heterozygous for both the Fgf10 and Fgfr2c genes (i.e. Fgfr2c+/Δ/Fgf10+/- mutants). Mutant mice heterozygous for either Fgfr2c (Fgfr2c+/Δ/Fgf10+/+) or Fgf10 (Fgfr2c+/+/Fgf10+/-) exhibit hypoplastic SMGs compared to WT (Fgfr2c+/+/Fgf10+/+) littermates (compare Fig. 6B, C to 6A). In contrast, Fgfr2c+/Δ/Fgf10+/- double heterozygous mutant SMGs are smaller and exhibit fewer ducts and terminal buds than seen in single heterozygous Fgfr2c+/Δ/Fgf10+/+ or Fgfr2c+/+/Fgf10+/- mutant glands (compare Fig. 6D to 4B, C). These results suggest that, during embryonic SMG development, the FGFR2b and FGFR2c signal transduction pathways each induce at least some downstream targets that are idiosyncratic and not coincident.
Figure 6 Fgfr2c+/Δ/Fgf10+/- double heterozygous mutant SMG is severely hypoplastic. A. Newborn R2c+/+/Fgf10+/+ WT SMG. B. Newborn R2c+/Δ/Fgf10+/+ (Fgfr2c+/Δ/Fgf10+/+) single heterozygous mutant SMG. C. Newborn R2c+/+/Fgf10+/-(Fgfr2c+/+/Fgf10+/-) single heterozygous mutant SMG. D. Newborn R2c+/Δ/Fgf10+/-(Fgfr2c+/Δ/Fgf10+/-) double heterozygous mutant SMG. The Fgfr2c+/Δ/Fgf10+/+ (B) and Fgfr2c+/+/ Fgf10+/- (C) single heterozygous SMGs have fewer ducts and terminal buds than seen in WT littermates (A). A much smaller, severely hypoplastic gland is seen in Fgfr2c+/Δ/Fgf10+/- double heterozygous mice (D) compared to Fgfr2c+/Δ/Fgf10+/+ (B) and Fgfr2c+/+/Fgf10+/- (C) littermates. Bar, 50 μm.
FGF8 rescues SMGs with interrupted FGF10/FGFR2b in vitro
Given that FGF8-mediated signaling plays an essential and unique role during SMG development [13], as well as our observation in Fgfr2c+/Δ/Fgf10+/- double heterozygous mutant mice discussed above, we postulated that exogenous FGF8 peptide would rescue the abnormal SMG phenotype associated with abrogated FGFR2b signaling in vitro. To determine if exogenous FGF8 peptide supplementation could restore branching morphogenesis to normal, we interrupted SMG morphogenesis in vitro with FGFR2b-Fc chimera and attempted to "rescue" these explants and restore branching to the level seen in controls. We cultured paired E13 + 3 SMGs in 10 ng/ml FGFR2b-Fc chimera for an initial period of 3 hours and then in FGFR2b-Fc + 500 ng/ml FGF8 peptide or FGFR2b-Fc alone for a total of 3 days (Fig. 7A, B). A second experiment to confirm that exogenous FGFR2b-Fc chimera decreased SMG branching morphogenesis consisted of paired E13 + 3 cultured in the presence of 10 ng/ml FGFR2b-Fc or IgG-Fc. In this set of experiments, FGFR2b-Fc-treated explants exhibit a significant 20% (P < 0.001) decrease in branching compared to IgG-Fc controls (Fig. 7C). Addition of exogenous FGF8 peptide induced a significant 24% (P < 0.02) increase in branching compared to FGFR2b-Fc treatment alone, thus completely restoring branching morphogenesis to the level seen in controls (Fig. 7C).
Figure 7 Exogenous FGF8 supplementation in vitro rescues SMG branching morphogenesis. Paired E13 embryonic SMGs were preincubated for 3 hrs in 10 ng/ml FGFR2b-Fc chimera and then cultured for a total of 3 days with/without 500 ng/ml FGF8 peptide. A. E13+3 10 ng/ml FGFR2b-Fc-treated explant. B. E13+3 FGFR2b-Fc chimera + FGF8-treated explant. Bar, 30 μm. C. Comparison of Spooner ratios. A significant 20% (P < 0.001) decrease in branching morphogenesis with FGFR2b-Fc chimera abrogation was seen compared to IgG-Fc control. Exogenous FGF8 supplementation induced a significant 24% (P < 0.02) increase in branching to completely restore branching to the level seen in control.
Discussion
The FGF family of growth factors is critical to normal embryogenesis, regulating cell proliferation, survival and apoptosis [8]. Analyses of Fgfr2b and Fgf10 mutant and null mice clearly demonstrate that the FGF10/FGFR2b signal transduction pathway is essential for the development of branching organs, including the lung, mammary gland, lacrimal gland, pancreas, thyroid gland and salivary gland [23-28,31-33]. Although SMGs were absent from E14.5 or older Fgfr2b null mice and newborn Fgf10 null mice [22-25], the presence of an initial SMG bud in E12.5 Fgf10 and Fgfr2b null embryos (Fig. 1) indicates that this is a true aplasia, and not agenesis. Moreover, the absence of SMGs in E13.5 and older Fgfr2b-/- and Fgf10-/- mutants confirms that FGF10/FGFR2b signaling is essential for earliest initial epithelial branching and subsequent Pseudoglandular Stage and older SMG morphogenesis, but not earliest initial bud formation.
The observed initial SMG bud formation and subsequent aplasia is consistent with the pathogenesis seen in other organs, including the Fgfr2b null mammary gland [26], but differs from that seen in lung and pancreas [24,31,34]. Interestingly, Bellusci and colleagues [26] also detected genotype-specific phenotypic differences in mammary bud formation. A transient single initial mammary gland bud (bud 4) seen in E11.5 Fgfr2b-/- mice is absent in E12.5 mice whereas bud 4 is maintained (but does not branch) in Fgf10-/- mice. This result suggests that FGFR2b signaling is essential to maintain bud 4 and to induce the other mammary placodes whereas another FGFR2b ligand (probably FGF7) acts redundantly with FGF10 to maintain the mammary gland placode. Similarly, differences in pancreatic development were observed between Fgf10 and Fgfr2b null mice [31,35]. Taken together, these results clearly indicate that Fgfr2b and Fgf10 null mice demonstrate tissue-specific differences in affected organs.
Although Fgf10+/- heterozygous mice were described as being normal [24,34], lacrimal, parotid and submandibular gland aplasia or hypoplasia were recently reported in adult Fgf10+/- mice and in ALSG patients with FGF10 heterozygous mutations [28]. To determine if Fgfr2b and Fgf10 gene dosage plays an important role during embryonic SMG development, we evaluated newborn Fgfr2b+/- and Fgf10+/- and found SMG hypoplasia in both (Fig. 3). This is the first report of organ abnormality in Fgfr2b heterozygous mice. Our data indicate that SMG development is Fgf10 and Fgfr2b dose-dependent. Homozygous null mutants exhibit SMG aplasia while heterozygotes exhibit SMG hypoplasia. Moreover, the observations of normal lungs, livers, and limbs [23-25,28,34], but abnormal SMG phenotypes [28], in Fgfr2b+/- and Fgf10+/- mutant mice provides additional evidence of FGF10/FGFR2b tissue-specificity.
Relationship between FGF10/FGFR2b and FGF8/FGFR2c signaling pathways
It is critical to remember that FGF10 binding to FGFR2b is part of a much larger genetic network. Organogenesis is the programmed expression of regulatory genes coupled to downstream structural genes and epigenetic events. Specific signaling pathways are parallel and largely functionally redundant; that is, several pathways differentially and combinatorially compensate for the dysfunction of a given individual pathway. There are some pathways, however, that have unique and nonredundant functions. One has always to ask two key questions: Is our pathway of interest functionally redundant or nonredundant? Will a broadly related, not independent, pathway compensate for the dysfunction of our pathway of interest?
The observation of SMG aplasia in Fgf10 and Fgfr2b null mice indicates that the functional presence of other endogenous FGF/FGFR pathways (e.g., FGF8/FGFR2c, FGF/FGFR1) or other signaling pathways (e.g., TGF-α/EGF/EGFR, Eda/Edar; IGF-II/IGF-IR) could not prevent complete death of embryonic SMG cells in Fgf10-/- and Fgfr2b-/- mice. Interestingly, although SMG aplasia was also seen in Fgf8 (Fgf8C/N;AP2αIRESCre/+) conditional mutant mice in which Fgf8 expression was ablated from first branchial arch epithelium, a small initial SMG bud is still seen in E15.5 embryos [13]. By contrast, the Fgf10and Fgfr2b null SMG bud is much more transient, being seen in E12.5 and absent in E13.5 mice (present study). This suggests that the FGF10/FGFR2b and FGF8/FGFR2c signaling pathways are both essential for branching and that the FGF10/FGFR2b signal is necessary for epithelial bud maintenance at earlier stages than the FGF8/FGFR2c signal.
The absence of SMGs in both Fgf8C/N; AP2αIRESCre/+ conditional mutant [13] and Fgf10-/- null mice (present study) suggests that FGF8 and FGF10 probably induce some of the same downstream targets, although through different receptors (FGF10/FGFR2b and FGF8/FGFR2c). However, the mere fact of ontogenic arrest and SMG aplasia, as well as temporal differences, in Fgf8 conditional mutants and Fgf10 and Fgfr2b null mice indicate that FGF10/FGFR2b and FGF8/FGFR2c signaling pathways induce broadly-related, but unique and nonredundant downstream cascades in SMGs that cannot be compensated by normal function of the other under physiologic conditions. Our observation of smaller and more severely hypoplastic SMGs in Fgfr2c+/Δ/Fgf10+/- double heterozygous mice compared to either the Fgfr2c+/Δ or Fgf10+/- single heterozygous mutants further supports this conclusion.
It is well established that FGF/FGFR signaling can simultaneously activate multiple signaling cascades (e.g., ERK/RAS/MAPK, P13K, and PLC-γ/PKC to mediate epithelial cell proliferation, survival and histodifferentiation [8,9,11]. Inhibition of ERK/RAS/MAPK or P13K signaling significantly reduced embryonic SMG branching morphogenesis in vitro [17-19,21], whereas inhibition of PKC modestly increased morphogenesis [18]. A recent study of FGFR2b signaling in vitro suggests that FGF10/FGFR2b requires ERK activation to mediate cell proliferation and branching whereas FGF7/FGFR2b requires both ERK and P13K activation [21].
Although the importance of ERK/RAS/MAPK, P13K and PKC signaling during embryonic SMG development in vitro has been demonstrated, it is presently unclear which signaling cascades downstream of the FGF10/FGFR2b and FGF8/FGFR2c pathways are essential for epithelial branching morphogenesis, proliferation and survival during in vivo SMG development. This is important because in vitro and in vivo results may not necessarily coincide, suggesting the effect of differing physiologic conditions. Recently, Steinberg et al. [21] demonstrated that, in vitro, once the gland is formed, inhibition of FGF10 does not inhibit branching. This is the precise opposite of that seen in both the Fgf10-/- null mice and the Fgf10+/- heterozygotes in vivo (Fig. 1C, 2C, 3C), as well as FGF10 heterozygous mutant individuals with ALSG syndrome [28].
Fgf10 (Fgf10-/-) or Fgfr2b (Fgfr2b-/-) loss of function is ultimately epistatic to each other and to the epigenome under normal physiologic conditions (i.e. no other gene mutations nor untoward environments), the very reason they are critical to SMG morphogenesis. However, since the epistasis associated with declining Fgf10 or Fgfr2b function is a nonlinear emergent property of the complete functional epigenotype, it can be manipulated in vitro in the manner reported here. Exogenous FGF8 peptide can completely rescue and restore to normal the abnormal phenotype seen with abrogated FGF10/FGFR2 signaling in vitro (Fig. 7). This is not surprising since FGF8 has the ability to simultaneously activate similar, as well as unique and ligand-specific, intracellular cascades which control proliferation, survival, and differentiation. Rescue experiments have always to be only a proof of this principle, not a mimic of the in vivo condition. After all, Fgf10-/- and Fgfr2b-/- mutant SMGs are not rescued in vivo, the very essence of epistatic mutations. What remains to be determined is the in vivo functional relationship between the FGF10/FGFR2b signal transduction pathway and other key downstream signaling pathways, and how these pathways are integrated during embryonic SMG development to compose the functional epigenome.
Conclusion
Our results indicate that FGF10/FGFR2b signaling is essential for the SMG epithelial branching and histodifferentiation, but not earliest initial bud formation. The functional presence of other endogenous FGF pathways or other signaling pathways could not prevent complete death of embryonic SMG cells in Fgf10 and Fgfr2b null mice. Moreover, our analysis of Fgfr2c+/Δ/Fgf10+/- double heterozygous mice indicates that FGF10/FGFR2b and FGF8/FGFR2c signaling pathways induce broadly-related, but unique and nonredundant downstream cascades in SMGs. What remains to be determined is the in vivo functional relationship between the FGF10/FGFR2b signal transduction pathway and other key downstream signaling pathways, and how these pathways are integrated during embryonic SMG development to compose the functional epigenome.
Methods
Fgf10 and Fgfr2b mutant mice
Fgf10 and Fgfr2b mutant mice were generated on a C57Bl/6 background and genotyped by RT-PCR as previously described [23,26,34]. Wildtype mice (WT) (littermates or otherwise) were used as controls. WT (E11.5-E18.5), Fgf10-/- (E11.5-5-E14.5) and Fgfr2b-/- (E11.5-E15.5) embryos were collected. For analysis of heterozygous SMGs, WT and heterozygous mice were mated and newborn Fgf10+/-, Fgfr2b+/- and WT littermates were collected and their genotypes confirmed by RT-PCR. The embryos and newborn heads were fixed in 4% paraformaldehyde in PBS, and stored in 70% ethanol until further processing. E11.5-E14.5 heads and E15 and older SMGs were processed, embedded in paraplast and serial coronal sections were stained with hematoxylin and eosin as previously described [1]. A minimum of 3 SMGs per age per genotype were analyzed.
Generation of Fgfr2c+/Δ/Fgf10 +/- mice
Due to their neonatal lethality, Fgfr2-IIIc+/Δ mice are routinely generated by crossing males in which a copy of FgfR2-exon 9 (IIIc) has been flanked by loxP sites [36], with females carrying a PGK-Cre transgene [37]. However, to obtain the Fgfr2-IIIc +/Δ;Fgf-10 +/- allele, we used PGK-Cre females in to which we had previously introduced a heterozygous Fgf10 null allele (Fgf10 +/-) [27]. For simplicity, the Fgfr2-IIIc +/Δ;Fgf-10 +/- is called Fgfr2c +/Δ/Fgf-10 +/-. This cross resulted in the recovery of the following genotypes in the correct Mendelian ratios: Fgfr2c +/+/ Fgf10 +/+ (WT); Fgfr2c +/+/ Fgf10 +/-, Fgfr2c +/Δ/Fgf-10 +/+ and Fgfr2c +/Δ/Fgf-10 +/- [36]. Lines were maintained and crosses were performed in a C57/black 6 background. E17-newborn WT and mutant mice were generated, SMGs were fixed and processed as described above, and their genotypes confirmed by RT-PCR as previously described [27,36]. All animal studies were conducted with the approval of the appropriate committees regulating animal research.
Culture system
Timed-pregnant females [C57Bl/10 (B10.A)] were sacrificed on day 13 and day 14 of gestation, and embryos were dissected in cold PBS and staged according to Theiler [38]. E13 (Initial Bud Stage) and E14 (Pseudoglandular Stage) SMG primordia were cultured for up to 3 days using a modified Trowell method as previously described [6]. The medium consisted of BGJb (Life Technologies, Rockville, MD) supplemented with 1% BSA, 0.5 mg ascorbic acid/ml and 50 units penicillin/streptomycin (Life Technologies), pH 7.2, and replicate cultures were changed every day. Supplementation studies: paired E13 and E14 SMG primordia were cultured in the absence or presence of exogenous FGF10 peptide (200 or 500 ng/ml, R and D Systems) for 3 days (E13 + 3) or 2 days (E14 + 2), respectively; controls consisted of enriched BGJb alone. Because a notable difference in SMG epithelial branch number is seen between embryos within a given litter and among litters, we calculated the Spooner branch ratios (end bud number/initial bud number) for each explant as previously described [13] and compared the Spooner branch ratios in right and left glands (treated and control) from each embryo. Mean Spooner ratios were determined, the data were arcsin transformed to insure normality and homoscedasticity, and compared by paired t-test for all embryos studied [39]. In this set of experiments, 4–6 explants per treatment were analyzed.
Interruption studies
We interrupted FGF10/FGFR2b signaling using soluble FGFR2b-Fc chimera (R & D Systems, Inc). This method has been successfully used to interrupt FGFR2b, FGFR2c, FGFR1b and FGFR1c signaling during embryonic SMG development [13,15,21]. We conducted dose-response studies in which we cultured paired E13 + 3 or E14 + 2 SMG primordia in the presence of soluble FGFR2b-Fc chimera (5 or10 ng/ml) or control IgG-Fc (5 or 10 ng/ml IgG-Fc; R & D Systems, Inc), analyzed Spooner ratios as described above, and determined the optimal FGFR2b-Fc chimera concentration which results in a hypoplastic gland similar to that seen in the in vivo Fgf10 mutant heterozygotes. In this set of experiments, 4–8 explants per treatment were analyzed. Based on this set of experiments, we used 10 ng/ml FGFR2b-Fc chimera in all subsequent interruption experiments.
Cell proliferation assay
The cell proliferation index was determined as previously described [6]. Paired E13 + 3 were cultured in 10 ng/ml FGFR2b-Fc or 10 ng/ml IgG-Fc, fixed in 10% formalin, embedded in paraffin and serially-sectioned. The sections were stained with anti-PCNA for 1 hr using the Zymed mouse PCNA kit (South San Francisco, CA) and counterstained with hematoxylin as previously described [13]. The development time was adjusted according to experiment: 2–5 min for supplementation studies and 10–15 min for interruption studies. In these experiments, the cytoplasm appears blue and PCNA-positive cells appear brown. Three explants per group were analyzed. Control and treated explants were serially-sectioned and the midpoint of each explant identified. The section showing the explant's mid-point was selected, as well as the fourth section to the right and the fourth section to the left of the midpoint. This design insured that different buds were counted in each of the 3 sections/treatment. In each section, we photographed terminal bud clusters in the upper left and lower right at 400 × and counted a minimum of 3 buds/area.
Cell proliferation was quantitated as the ratio of PCNA-positive epithelial cells/total epithelial cells. The mean cell proliferation index was determined per section and the mean cell proliferation index of the three sections/explant was determined for each treatment. The data was arcsin transformed and the means ratios compared by t-test.
Rescue experiment
Paired E13 SMG primordia were cultured in 10 ng/ml FGFR2b-Fc chimera for an initial period of 3 hrs and then each pair was cultured in FGFR2b-Fc or FGFR2b-Fc + 500 ng/ml FGF8 (R and D Systems, Inc.) for 3 days. This FGF8 peptide concentration was determined to induce a significant increase in branching morphogenesis (data not shown). A concurrent control experiment was conducted as an internal control to verify that FGFR2b-Fc chimera supplementation interrupted branching morphogenesis; these controls consisted of E13 primordia cultured in 10 ng/ml FGFR2b-Fc chimera or 10 ng/ml IgG-Fc for 3 days. The explants were collected and mean Spooner ratios determined and compared as described above. Four explants per treatment were analyzed.
List of abbreviations
ALSG-aplasia of lacrimal and salivary gland
BSA-bovine serum albumen
CONT-control
FGF-Fibroblast growth factor
FGFR2b-FGF receptor 2-IIIb
FGFR2c-FGF receptor 2-IIIc
ERK-p44/p42 mitogen activating kinases (ERK-1/2)
MAPK-mitogen-activating protein kinases
P13K-phosphatidylinositol 3-kinase
PBS-phosphate-buffered saline
PKC-protein kinase C
PLC-γ-phospholipase C γ1
R2b-/--Fgfr2b-/-
R2c-Fgfr2c
RT-PCR-reverse transcriptase-polymerase chain reaction
SMG-submandibular salivary gland
WT-wildtype
Authors' contributions
TJ designed and coordinated the study, was involved in aspects of all experiments, and drafted the manuscript. GA prepared the histological sections for this study and performed some of the morphology and cell proliferation experiments. DW participated in morphological analyses and culture experiments and generated all figures. FGS generated the Fgf10 and Fgfr2b mutant mice and harvested and genotyped the embryos. SB assisted in the analyses of mutant mouse histopathology. MKH generated the Fgfr2c+/Δ/Fgf10 +/- single and double mutant mice, harvested and genotyped these embryos, and assisted in the analysis of their histopatholgy. MM participated in the design and coordination of this study, assisted in the analysis of histopathology, performed the statistical analysis, and helped draft the manuscript. All authors read and approved the final manuscript.
Acknowledgements
We gratefully acknowledge Dr. Clive Dickson in whose laboratory the double heterozygous mice were generated and maintained. This research was supported by a grant from American Lung Association (SB), NIH grants RO1 DE014535 (TJ/MM) and RO1 HL074832 (SB).
==== Refs
Jaskoll T Melnick M Submandibular gland morphogenesis: stage-specific expression of TGF-alpha, EGF, IGF, TGF-beta, TNF and IL-6 signal transduction in normal mice and the phenotypic effects of TGF-beta2, TGF-beta3, and EGF-R null mutations Anat Rec 1999 256 252 268 10521784 10.1002/(SICI)1097-0185(19991101)256:3<252::AID-AR5>3.0.CO;2-6
Jaskoll T Zhou YM Chai Y Makarenkova HP Collinson JM West JD Hajihosseini MK Lee J Melnick M Embryonic submandibular gland morphogenesis: stage-specific protein localization of FGFs, BMPs, Pax 6 and Pax 9 and abnormal SMG phenotypes in FGF/R2-IIIc+Δ, BMP7-/- and Pax6-/- mice Cells Tissues Organs 2002 170 83 98 11731698 10.1159/000046183
Davidson EH Rast JP Oliveri P Ransick A Calestanim C Yuh CH Minokawa T Armore G Hinman V Arenas-Mena C Otim O Brown CT Livi CB Lee PY Reville R Tust AG Pan ZJ Schilstra MJ Clarke PJC Arnone ML Rowen L Cameron RA McClay DR Hood L Bolour H A genomic regulatory network for development Science 2002 295 1669 1678 11872831 10.1126/science.1069883
Davidson EH McClay DR Hood L Regulatory gene networks and the properties of the development process Proc Natl Acad Sci 2003 100 1475 1480 12578984 10.1073/pnas.0437746100
Gardner TS Bernardo DD Lorenz D Collins JJ Inferring genetic networks and identifying compound mode of action via expression profiling Science 2003 301 102 105 12843395 10.1126/science.1081900
Melnick M Chen H Zhou Y Jaskoll T The functional genomic response of developing embryonic submandibular glands to NF-κB inhibition BMC Dev Biol 2001 1 15 11716784 10.1186/1471-213X-1-15
Goldfarb M Signaling by fibroblast growth factors: the inside story Sci STKE 2001 2001 PE37 11687709
Ornitz DM Itoh N Fibroblast growth factors Genome Biol 2001 2 reviews3005 11276432 10.1186/gb-2001-2-3-reviews3005
Tsang M Dawid IB Promotion and attenuation of FGF signaling through the Ras-MAPK pathway Sci STKE 2004 PE17
Ibrahimi OA Zang F Eliseenkova AV Itoh N Linhardt RJ Mohammadi M Biochemical analysis of pathogenic ligand-dependent FGFR2 mutations suggest distinct pathophysiological mechanisms for craniofacial and limb anomalies. Hum Mol Genet 2004 13 2313 2324 15282208 10.1093/hmg/ddh235
Schlessinger J Common and Distinct Elements in Cellular Signaling via EGF and FGF Receptors Science 2004 306 1506 1507 15567848 10.1126/science.1105396
Hoffman MP Kidder BL Steinberg ZL Lakhani S Ho S Klienman HK Larsen M Gene expression profiles of mouse submandibular gland development: FGFR1 regulates branching morphogenesis in vitro through BMP- and FGF- dependent mechanisms Development 2002 129 5767 5778 12421715 10.1242/dev.00172
Jaskoll T Witcher D Toreno L Bringas P Moon AM Melnick M FGF8 dose-dependent regulation of embryonic submandibular salivary gland morphogenesis Dev Biol 2004 268 457 469 15063181 10.1016/j.ydbio.2004.01.004
Jaskoll T Leo T Witcher D Ormestad M Astorga J Carlsson P Melnick M Sonic Hedgehog signaling plays an essential role during embryonic salivary gland Epithelial Branching Morphogenesis Dev Dyn 2004 229 722 732 15042696 10.1002/dvdy.10472
Jaskoll T Melnick M Embryonic salivary gland branching morphogenesis. In Branching Morphogenesis: Davies J. (Ed.) Georgetown: Landes Biosci 2004
Kashimata M Gresik E Epidermal growth factor system is a physiological regulator of development of the mouse fetal submandibular gland and regulates expression of the α6-integrin subunit Dev Dyn 1997 208 149 161 9022052 10.1002/(SICI)1097-0177(199702)208:2<149::AID-AJA2>3.0.CO;2-I
Kashimata M Sayeed S Ka A Onetta-Muda A Sakagami H Faraggiana T Gresik E The ERK-1/2 signaling pathway is involved in the stimulation of branching morphogenesis of fetal mouse submandibular glands by EGF Dev Biol 2000 220 183 196 10753509 10.1006/dbio.2000.9639
Koyama N Kashimata M Sakagami H Gresik EW EGF-stimulated signaling by means of P13K, PLCγ1, and PKC isozymes regulates branching morphogenesis of the fetal mouse submandibular gland Dev Dyn 2003 227 216 226 12761849 10.1002/dvdy.10309
Larsen M Hoffman MP Sakai T Neibaur JC Mitchell JM Yamada KM Role of PI 3-kinase and PIP3 in submandibular gland branching morphogenesis Dev Biol 2003 255 178 191 12618142 10.1016/S0012-1606(02)00047-7
Melnick M Jaskoll T Mouse submandibular gland morphogenesis: a paradigm for embryonic signal processing Crit Rev Oral Biol Med 2000 11 199 215 12002815
Steinberg Z Myers C Heim V Lathrop CA Rebustini IT Stewart JS Larsen M Hoffman MP FGFR2b signaling regulates ex vivo submandibular gland epithelial cell proliferation and branching morphogenesis. Development 2005 132 1223 1234 15716343 10.1242/dev.01690
Celli G LaRochelle WJ Makem S Sharp R Merlino G Soluble dominant-negative receptor uncovers essential roles for fibroblast growth factors in multi-organ induction and patterning. EMBO J 1998 17 1642 1655 9501086 10.1093/emboj/17.6.1642
DeMoerlooze L Spencer-Dene B Revest JM Hajihosseini MK Rosewell I Dickson C An important role for the III isoform of fibroblast growth factor receptor 2 (FGFR2) in mesenchymal-epithelial signaling during mouse organogenesis. Development 2000 127 483 492 10631169
Ohuchi H Hori Y Yamasaki M Harada H Sekine K Kato S Itoh N FGF10 acts as a major ligand for FGF Receptor 2 IIIb in mouse multi-organ development Biochem Biophys Res Commun 2000 277 643 649 11062007 10.1006/bbrc.2000.3721
Revest JM Spencer-Dene B Kerr K De Moerlooze L Roswell I Dickson C Fibroblast growth factor receptor-2-IIIb acts upstream of Shh and Fgf4 and is required for limb bud maintenance but not for the induction of Fgf8, Msx1, or Bmp4 Dev Biol 2001 231 47 62 11180951 10.1006/dbio.2000.0144
Mailleux AA Spencer-Dene B Dillion C Ndiaye D Savona-Baron C Itoh N Kato S Dickson C Thiery JP Bellusci S Role of FGF10/FGFR2b signaling during mammary gland development in the mouse embryo Development 2002 129 53 60 11782400
Min H Danilenko DM Scully SA Bolon B Ring BD Tarpley JE DeRose M Simonet WS Fgf-10 is required for both limb and lung development and exhibits striking functional similarity to Drosophila branchless Genes Dev 1998 12 3156 3161 9784490
Entesarian M Matsson H Klar J Bergendal B Olson L Arakaki R Hayashi Y Ohuchi H Falahat B Bolstad AI Jonsson R Wahren-Herlenius M Dahl N Mutations in the gene encoding fibroblast growth factor 10 are associated with aplasia of lacrimal and salivary glands Nat Genet 2005 37 125 127 15654336 10.1038/ng1507
Nakanishi Y Morita T Nogawa H Cell proliferation is not required for the initiation of early cleft formation in mouse embryonic submandibular epithelium in vitro. Development 1987 99 429 437 3115749
Hart A Papadopoulou S Edlund H Fgf10 maintains Notch activation, stimulates proliferation, and blocks differentiation of pancreatic epithelial cells. Dev Dyn 2003 228 185 193 14517990 10.1002/dvdy.10368
Bhushan A Itoh N Kato S Thiery JP Czernichow P Bellusci S Scharfmann R Fgf10 is essential for maintaining the proliferative capacity of epithelial progenitor cells during early pancreatic organogenesis Development 2001 128 5109 5117 11748146
Govindarajan V Ito M Makarenkova H Lang RA Overbeek PA Endogenous and Ectopic Gland induction by FGF-10 Dev Biol 2000 225 188 200 10964474 10.1006/dbio.2000.9812
Mailleux A Kelly R Veltmaat JM De Langhe SP Zaffran S Thiery JP Bellusci S Fgf10 expression identifies parabronchical smooth muscle cell progenitors and is required for their entry into the smooth muscle cell lineage. Development 2005 132 2157 2166 15800000 10.1242/dev.01795
Sekine K Ohuchi H Fujiwara M Yamasaki M Yoshizawa T Sato T Yagishita N Matsui D Koga Y Itoh N Kato S Fgf10 is essential for limb and lung formation. Nat Genet 1999 21 138 141 9916808 10.1038/5096
Pulkkinen MA Spencer-Dene B Dickson C Otonkoski T The IIIb isoform of fibroblast growth factor receptor 2 is required for proper growth and branching of pancreatic ductal epithelium but not for differentiation of exocrine or endocrine cells Mech Dev 2003 120 167 175 12559489 10.1016/S0925-4773(02)00440-9
Hajihosseini MK Wilson S De Moerlooze L Dickson C A splicing switch and gain-of-function mutation in FgfR2-IIIc hemizygotes causes Apert/Pfeiffer-syndrome-like phenotypes Proc Natl Acad Sci U S A 2001 98 3855 3860 11274405 10.1073/pnas.071586898
Lallemand Y Luria V Haffner-Krausz R Lonai P Maternally expressed PGK-Cre transgene as a tool for early and uniform activation of the Cre site-specific recombinase Transgenic Res 1998 7 105 112 9608738 10.1023/A:1008868325009
Theiler K The House Mouse 1989 New York, Springer-Verlag
Sokal R Rohlf FJ Biometry 1981 New York, Freeman
|
15972105
|
PMC1184065
|
CC BY
|
2021-01-04 16:30:32
|
no
|
BMC Dev Biol. 2005 Jun 22; 5:11
|
utf-8
|
BMC Dev Biol
| 2,005 |
10.1186/1471-213X-5-11
|
oa_comm
|
==== Front
BMC DermatolBMC Dermatology1471-5945BioMed Central London 1471-5945-5-71597515010.1186/1471-5945-5-7Research ArticlePruritus in hemodialysis patients Akhyani Maryam [email protected] Mohammad-Reza [email protected] Nasrin [email protected] Behnaz [email protected] Maryam [email protected] Department of dermatology, Tehran University of Medical Sciences, Razi Hospital, Tehran, Iran2 Department of nephrology, Tehran University of Medical Sciences, Dr. Shariati hospital, Tehran, Iran3 Department of pathology, Tehran University of Medical Sciences, Sina hospital, Tehran, Iran4 Department of pediatrics, Iran University of Medical Sciences, Hazrat Rasool Hospital, Tehran, Iran2005 24 6 2005 5 7 7 31 10 2004 24 6 2005 Copyright © 2005 Akhyani et al; licensee BioMed Central Ltd.2005Akhyani et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Pruritus is one of the most bothersome symptoms in patients on maintenance hemodialysis (HD), however little progress is seen in our understanding of its pathogenesis. The aim of this study was to evaluate the frequency of pruritus in HD patients in Tehran, Iran, and to correlate its presence and intensity with relevant clinical and laboratory parameters.
Methods
One hundred sixty-seven patients on maintenance HD at three out-patient HD units were enrolled in the study. Itch intensity was scored as mild, moderate and severe. Some relevant clinical and laboratory parameters (age, sex, xerosis, presence of neuropathy, duration of dialysis, history of atopy and laboratory findings including hematocrit, creatinine, urea, calcium, phosphorus, parathyroid hormone [PTH] and alkaline phosphatase) were evaluated.
Results
Pruritus was found in 41.9% of patients. The intensity of itching was mild, moderate and severe, in 51.4%, 11.4% and 37.7% of patients, respectively. In 22 patients (31.4%) pruritus intensified during and after dialysis. There was no significant difference in the serum levels of creatinine, blood urea nitrogen, calcium, phosphorus, alkaline phosphatase, PTH and hematocrit between patients with and without pruritus. Age, sex, xerosis, underlying renal disease, history of atopy and duration of haemodialysis were not significantly different between the two groups. However, neuropathy was significantly more common in the pruritic group (63.8% versus 42.1%) (pv = 0.006).
Conclusion
Clinical neuropathy was the only significant finding in the pruritic group in our study. This finding justifies further research on nerve function and neurotransmitters in hemodialysis patients and the introduction of new drugs targeting neuropathy.
==== Body
Background
Pruritus often constitutes a major problem for patients with end-stage renal disease (ESRD). Unfortunately, dialysis has only a slight impact on pruritus. Therefore it is quite frustrating that an ever-increasing number of ESRD patients on HD, are waiting for transplantation, while they suffer from this tormenting symptom. According to most sources, more than half of patients undergoing HD complain of varying degrees of pruritus. [1-4] The mechanism underlying pruritus is poorly understood; current theories include secondary hyperparathyroidism, divalent-ion abnormalities, histamine, allergic sensitization, proliferation of skin mast cells, iron deficiency anemia, hypervitaminosis A, xerosis, neuropathy and neurologic changes, opioid system involvement (understimulation of κ receptors or overexpression of μ receptors), cytokines, serum bile acids, nitric oxide, or some combination of these.[1,4-12] The quality of uremic pruritus varies between patients. In some it is persistent, extensive and intractable but in others it may be transitory and localized. [1] Pruritus usually begins about six months after the start of dialysis, [13] and some authors have shown a significant positive relationship with the duration of HD. [10] The aim of this study was to evaluate the frequency of pruritus in hemodialysis patients in Tehran, Iran, and its relationship with age, sex, xerosis, neuropathy, duration of dialysis, history of atopy and laboratory findings including hematocrit, creatinine, urea, calcium, phosphorus, alkaline phosphatase and PTH.
Methods
All patients (n = 167)on maintenance HD attending three HD units at Emam Khomeini, Dr Shariati and Sina Medical Complexes in Tehran, Iran, from 1998–1999 were enrolled in a cross-sectional study. Eighty-three were men and 84 were women. Their mean age was 47.95 +/-16.73 years with a range of 10–76 years. The routine of dialysis was 3 times per week (except holidays) for 4 hours. The mean of dialysis sessions per week are presented in table 1. Polysulfone membranes were used as dialyser, the dialysate was mainly acetate, and sodium hypochlorite was used for sterilization. Twenty nine of the patients were diabetic, while four patients suffered from systemic lupus erythematosus (SLE), two from lymphoma and one from multiple myeloma. The etiology in the remaining cases was unknown. None of the patients presented with primary skin disease.
Table 1 Comparison of clinical and laboratory parameters in HD patients
Clinical & laboratory parameters Pruritic (mean ± SD) Non pruritic (mean ± SD) p-value
Age (years) 51.54 ± 16.63 45.37 ± 16.41 0.1212
Time on treatment (months) 44.42 ± 44 41.41 ± 44.73 0.666
Dialysis per week 2.6 ± 0.55 2.57 ± 0.59 0.682
Hematocrit (%) (M = 39–52)(F = 36–46) 25.43 ± 6.99 24.95 ± 5.43 0.647
Serum calcium (mg/dl) (8.6–10.3) 8.705 ± 1.79 8.707 ± 1.44 0.994
Serum phosphorus (mg/dl) (2.7–4.5) 5.45 ± 1.72 5.74 ± 1.66 0.314
Serum alkaline phosphatase (u/l) (150–450 u/l) 322.01 ± 272.61 494.56 ± 750.76 0.080
BUN (mg/dl) (8–20) 131.31 ± 66.01 136.08 ± 68.61 0.673
Serum creatinine (mg/dl) (0.4–1.3) 9.43 ± 3.21 9.27 ± 3.29 0.777
Parathyroid hormone (pmole/l)(0.8–2.5) 0.9 ± 1.16 1.43 ± 0.85 0.252
The patients were questioned, by a physician, if they had pruritus during the previous two weeks, and the symptoms were classified according to intensity (mild, moderate, and severe), location (generalized, head and neck, trunk and limbs), type of pruritus (episodic, persistence) and duration of HD. The patients' symptoms were recorded before a mid-week HD session and a blood sample was collected. The following laboratory tests were performed: Serum calcium, phosphorus, alkaline phosphatase, PTH, BUN, creatinine and hematocrit.
Measurement of pruritus
The severity of pruritus was assessed subjectively and scored as follows:
Mild: Episodic and localized pruritus without disturbance in usual work and sleep.
Moderate: Generalized and continuous pruritus without sleep disturbance.
Severe: Generalized and continuous pruritus with sleep disturbance.
The major emphasis was on the disturbance in usual work and sleep.
Grading of xerosis
Xerosis was graded as follows:
0: Absent; 1: Mild (only on legs); 2: Moderate (all of the extremities); and 3: Severe (generalized).
Neuropathy
The diagnosis of clinical neuropathy was based on the presence of paresthesia, restless leg syndrome, decreased sensations of vibration, position, light touch, pain, decreased deep tendon reflexes or muscular force.
Definition of atopy
Positive history of asthma, atopic dermatitis (based on Hanifin-Rajka criteria [14]) or allergic rhinitis was defined as atopy.
Statistical analysis
Pruritus and non-pruritus groups were compared according to the variables. Data entry and coding was done through a database system developed with PE2 and analyzed by SPSS version 9. Student t-test and X2 score were used for statistical analysis. Statistical significance was defined as p < 0.05.
Results
Seventy patients (41.9%) (30 women and 40 men) complained of pruritus classified as mild, moderate and severe in 51.4% (n = 36), 11.4% (n = 8) and 37.1% (n = 26) of patients, respectively. The difference between the frequency of pruritus by gender was not significant (pv > 0.05). Pruritus was episodic in 60% (n = 42) and persistent in 40%(n = 28) of patients. Six patients had episodic but generalized pruritus with disturbance in work or sleep and their itch was categorized as moderate or severe. In 31.4% (n = 22) of patients pruritus intensified during and after dialysis.
Pruritus was generalized in 70 %(n = 49) of patients, limited to the trunk in 14.3% (n = 10), limbs 14.3% (n = 10), and head and neck 1.4 %(n = 1). There was history of atopy in 17 (24.3%) of pruritic and 20 (20.6%) of non-pruritic patients. The difference was not significant (pv > 0.05). Nine out of 21 diabetic patients, one out of 4 SLE patients and one lymphomatous patient complained of pruritus. There was no statistically significant difference between pruritus in diabetic and non-diabetic subjects (pv > 0.05). Forty seven out of 97 non-pruritic patients had a past history of pruritus during HD treatment. Comparison of clinical and laboratory parameters in HD patients with and without pruritus is shown in table 1. There was no significant difference among the two groups using t-test (pv > 0.05).
There wasn't any relationship between pruritus and xerosis in HD patients 2. Pruritus was alleviated in one patient after parathyroidectomy.
Table 2 Frequency of pruritus in HD patients according to severity of xerosis
Xerosis Pruritic Non pruritic
Absent 27 (38.6%) 33 (34%)
Mild 19 (27.1%) 40 (41.2%)
Moderate 19 (27.1%) 18 (18.6%)
Severe 5 (07.1%) 6 (06.2%)
Total 70 (100%) 97 (100%)
X2 pv = 0.26
Eighty-four patients (51.3%) suffered from peripheral neuropathy (Three patients with diabetic foot amputation were excluded). Fifty-eight had only subjective symptoms; decreased sensation of vibration, decreased deep tendon reflexes, decreased pain perception, decreased sensation of light touch, and decreased sensation of position were noted in 19, 11, 7, 5, and 4 cases, respectively. In the pruritic group, 44 patients (63.8%), and in the non-pruritic group, 40 cases (42.1%) had neuropathy. The difference was significant (p-value = 0.006).
Discussion
Pruritus is one of the most prevalent presentations of uremia and occurs in 10–85% of HD patients. [1,15-17] Difficulty in defining this very subjective symptom, the limited number of patients in most series, and the retrospective nature of some of the information, may be some of the reasons why such a wide range of figures is quoted. Recently, Urbonas et al noticed a decreasing trend in the prevalence of pruritus in HD patients and attributed it to more precise calculation of HD doses based on Kt/v or creatinine clearance measurements, introduction of new dialysers with larger surfaces as well as replacement of cuprophane fibers to more biocompatible ones made from polysulfone and amyl nitrite. [1,10,18]
Despite the long history of HD in Iran, to the best of our knowledge no study is published in the literature regarding the itch in Iranian HD patients. In our study, including 167 patients, pruritus was found in 41.9% of patients, severe in 37.1%, moderate in 11.4%, and mild in 51.4% of pruritic patients. Stahle-Backdahl et al, described pruritus in 60–65% of their cases which was severe in 8%, moderate in 24%, and mild in 34% of the examined patients. [19] Zucker and his colleagues reported pruritus in 48% of their HD patients at the time of the study in Tel-Aviv. [20] The figure reported from Poland by Szepietowski et al was 40.8% [10,21]. In Kato's series from Japan, 74% complained of pruritus [11]. Benchikhi et al reported a figure of 74.4% in HD patients from Moroc [22]. It is worth mentioning that a major drawback for studying and comparing results from different studies is the lack of a uniform way for assessment of this very subjective symptom. Recently, Yosipovitch et al has developed a comprehensive questionnaire for the assessment of pruritus and tested it in uremic patients and found it valid and reliable [23]. This questionnaire can provide data on many aspects of itching: the effect of daily life habits, physical activities and antipruritics on itch, the effect of itch on quality of life, temporal factors, location, severity, characteristics and exacerbating and relieving factors are all included. They found a significant correlation between the visual analogue scale and affective scores.
Uremic pruritus may be constant in 50% of cases with exacerbation or intermittent with spontaneous remissions. [24] In the majority of those affected, the itch is paroxysmal and may be localized (56%) or generalized (44%).[25] Seventy percent of our patients suffered from generalized pruritus as Moroccan patients.[22] Exacerbations of continuous itch are usually observed during HD or soon after, [1] which can be attributed to allergy to dialysate or dialysis membranes. [18] Exacerbation of pruritus during dialysis sessions was seen in 31.4% of HD patients in our study. In Yosipovitch series, too, itch appeared or aggravated in only 26% of patients during dialysis and the majority did not note an effect of dialysis on their itch. [23]
In Stahle-Backdahl et al's study, patients with pruritus tended to have been on dialysis treatment longer than those without pruritus.[2] Szepietowski et al showed a significant relationship between the total score of pruritus and duration of hemodialysis.[10] On the contrary, Altmeyer et al [26] described a significant improvement of itching in patients who had been on HD for long period of time: of the 23 patients with short term dialysis (2–3 years), 78% complained of pruritus, while it was seen in only 43% of 28 patients with long-term dialysis (>8 years). Murphy et al [27], however, could not confirm that pruritus decreases with progressive duration of dialysis treatment. In our study, we did not find any relationship between pruritus and duration of dialysis as seen in some previous studies. [7,19,26,27]
Xerosis is seen in the majority of patients on HD [1,4,7,8] and it may contribute to pruritus. [1,7,10,11,20] A relationship between the degree of xerosis and pruritus has been demonstrated in several studies, [1,10,20,28,29] although others failed to find an association between skin hydration and pruritus using a skin surface hygrometer. [11,30,31] Xerosis found in HD patients has been attributed to increased level of vitamin A in epidermis, atrophy of sebaceous and sweat glands and dysautonomia. [1,7,15] In our study the prevalence of xerosis in pruritic and non-pruritic patients was 61.4% and 66%, respectively, and we couldn't find any association between xerosis and pruritus.
Increased serum levels of magnesium, phosphorus and calcium have been proposed to be involved in uremic pruritus by some authors. [1,9,15,24] It has been suggested that an increased skin divalent ions concentration may lead to microprecipitation of calcium or magnesium phosphate, which may be the cause of pruritus. [1] On the other hand, the role of magnesium itself in the modulation of nerve conduction and release of histamine from mast cells was put forward. [1] While marked improvement of uremic pruritus with low dialysate calcium and magnesium has been reported [32,33], only a few studies showed a significant correlation between serum or skin divalent ion content and the presence of pruritus.[1,6] Recently, Momose et al, found increased calcium ion concentration in the deepest layers of the epidermis indicating a disrupted calcium ion gradient in the skin.[6] Like most studies [3,22,31], we couldn't find any association between serum calcium and phosphorus and uremic pruritus.
Hyperparathyroidism has been proposed by some as a cause of uremic pruritus.[8] Cases of disappearance of pruritus after parathyroidectomy support this theory.[1,4,8] This hormone can indirectly lead to metabolic alterations that subsequently cause pruritus.[1] On the other hand, a direct role for parathyroid hormone in causing uremic pruritus has been questioned because of the failure of intradermal injections of PTH analogs to cause pruritus, and because of negative immunohistochemical studies for PTH in skin biopsy specimens.[19]
Furthermore, no correlation between PTH levels and intensity of itching was found in most studies.[1,9,31] Although pruritus improved after parathyroidectomy in one of our cases, we couldn't find any relationship between serum PTH and pruritus in our study.
Iron deficiency anemia contributes to renal itch according to some authors, but we couldn't find any relationship between hematocrit level and uremic pruritus as shown in more recent studies[3,31].
Likewise, we couldn't find any association between HD pruritus and atopy. No previous research was recorded in the literature.
In our study there were also no significant difference between pruritic and non pruritic HD patients according to age, sex, underlying renal disease, serum alkaline phosphatase, BUN and creatinine similar to previous studies. [1,3,4,11,20-22,31]
Peripheral neuropathy affects up to 65% of patients starting dialysis. Some consider pruritus another manifestation of this neuropathy.[17] Contradictory data have been published about neurone-specific enolase fibers in the epidermis of HD patients. [2,34] Zakrzewska-Pniewska and Jedras found nervous dysfunction especially the somatic component related to pruritus in uremic patients. Pruritus was more frequent in patients with paresthesia in their study. Moreover the relationship between the severity of pruritus and the incidence of paresthesia was significant. [35]On the other hand, reports showing the efficacy of lidocaine [36,37], capsaicin [9,38,39], and gabapentin [40,41] in controlling uremic pruritus are in favor of a relationship between neuropathy and itching in HD patients. Yosipovitch et al reported paradoxical heat sensation an early sign of uremic sensory neuropathy. [42]
Neuropathy was significantly more frequent in patients with pruritus than without in our series, showing a figure of 63.8% versus 42.1% (pv < 0.05). These findings were not confounded by the presence of diabetics in the study group, as pruritus was not significantly different between diabetics and non-diabetics. Clinical neuropathy was significantly more frequent in HD patients with pruritus in Mesic's cases.[43] Winkelman et al, too, showed correlation between restless legs syndrome and pruritus. [44]
In conclusion, clinical neuropathy was the only significant finding in the pruritic group in our study. This finding justifies further research on nerve function and neurotransmitters in hemodialysis patients. Furthermore, therapeutic or preventive modalities targeting neuropathy may be effective in controlling the itch in hemodialysis patients.
Authors' contributions
MA conceived the study, participated in its design and coordination, and drafted the manuscript.
GMR participated in the design and conduct of the study.
SN and KB performed the data collection and the statistical analysis.
DM drafted the manuscript.
All authors read and approved the final manuscript.
Competing interests
The author(s) declare that they have no competing interests.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Urbonas A Schwartz RA Szepietowski JC Uremic pruritus-an update Am J Nephrol 2001 21 343 350 11684792 10.1159/000046272
Stahle-Backdahl M Uremic pruritus. Clinical and experimental studies Acta Derm Venereol (Stockh) 1989 145 1 38
Virga G Visentin I La Milia V Bonadonna A Inflammation and pruritus in hemodialysis patients Nephrol Dial Transplant 2002 17 2164 2169 12454228 10.1093/ndt/17.12.2164
Greaves MW Champion RH, Burton JL, Burns DA, Breathnach SM Pruritus Textbook of dermatology 1998 6 London: Blackwell science Ltd 623
Mettang T Pauli-Magnus C Alscher DM Uraemic pruritus – new perspectives and insights from recent trials Nephrol Dial Transplant 2002 17 1558 63 12198205 10.1093/ndt/17.9.1558
Momose A. Kudo S Sato M Saito H Nagai K Katabira Y Funyu T Calcium ions are abnormally distributed in the skin of hemodialysis patients with uraemic pruritus Nephrol Dial Transplant 2004 19 2061 15187190 10.1093/ndt/gfh287
Greaves MW Freedberg IM, Eisen AZ, Wolff K, Austen KF, Goldsmith LA, Katz SL Pathophysiology and clinical aspects of pruritus Fitzpatrick's Dermatology in general medicine 2003 6 USA: McGraw-Hill Companies, Inc 399 405
Massry SG Popovtzer MM Coburn JW Makoff DL Maxwell MH Kleeman CR Intractable pruritus as a manifestation of secondary hyperparathyroidism in uremia. Disappearance of itching after subtotal parathyroidectomy N Eng J Med 1968 279 697
Cho YL Liu HN Huang TP Tarng DC Uremic pruritus: Roles of parathyroid hormone and substance P J Am Acad Dermatol 1997 36 538 541 9092738
Szepietowski JC Sikora M Kusztal M Salomon J Magott M Szepitowski T Uremic pruritus: a clinical study of maintenance hemodialysis patients J Dermatol 2002 29 621 6 12432992
Kato A Hamada M Maruyama T Maruyam A Hishida A Pruritus and hydration state of stratum corneum in hemodialysis patients Am J Nephrol 2000 20 437 42 11146309 10.1159/000046196
Yozipovitch G Itching in the new millennium: Highlights of the Second International Workshop for the Study of Itch, Toyoma, Japan J Am Acad Dermatol 2004 51 625 7 15389199 10.1016/j.jaad.2004.02.018
Robertson KE Mueller BA Uremic pruritus Am J Health Syst Pharm 1996 15 2523
Hanifin JM Rajka G Diagnostic features of atopic dermatitis Acta Derm Venereol 1980 92 44 47
Ashmore SD Jones CH Newstead CG Daly MJ Chrystyn H Ondansetron therapy for uremic pruritus in hemodialysis patients Am J kidney Dis 2000 5 827 31 10793015
Stahl-Backdahl M Pruritus in hemodialysis patients Skin pharmacol 1992 1 14 20
Carmichael AJ Renal itch Bernhard JD In Itch: Mechanisms and management of pruritus 1992 Mc Graw-Hill, Inc 217 228
Mettang T Pauli-Magnus C The pathophysiological puzzle of uremic pruritus-Insights and speculation from therapeutic and epidemiological studies Perit Dial Int 2000 20 493 494 11117239
Stahle-Backdahl M Hagermark O Lines LE Pruritus in patients on maintenance hemodialysis Acta Med Scan 1988 224 55 60
Zucker I Yosipovitch G David M Gafter U Boner G Prevalence and characterization of uremic pruritus in patients undergoing hemodialysis: uremic pruritus is still a major problem for patients with end-stage renal disease J Am Acad Dermatol 2003 49 842 6 14576662 10.1016/S0190-9622(03)02478-2
Szepietowski JC Salomon J Uremic pruritus: still an important clinical problem J Am Acad Dermatol 2004 51 842 3 15523377 10.1016/j.jaad.2004.04.003
Benchikhi H Moussaid L Doukaly O Ramdani B Zaid D Lakhdar H Hemodialysis-related pruritus. A study of 134 Moroccans Nephrologie 2003 24 127 31 12814059
Yosipovitch G Zucker I Boner G Gafter U Shapira Y David M A questionnaire for the assessment of pruritus: validation in uremic patients Acta Dermatol Venereol 2001 81 108 111 10.1080/00015550152384236
Szepietowski J Selected elements of the pathogenesis of pruritus in hemodialysis patients: My own study Med Sci Monti 1996 2 343 347
Gilchrest BA Stern RS Steinman TI Brown RS Arndt KA Anderson WW Clinical features of pruritus among patients undergoing maintenance hemodialysis Arch Dermatol 1982 118 154 156 7065661 10.1001/archderm.118.3.154
Altmeyer P Kachel HG Junger M Koch KM Holzmann H Skin changes in long-term dialysis patients. Clinical study Hautarzt 1982 33 303 309 6809694
Murphy M Carmichael AJ Renal itch Clin Exp Dermatol 2000 25 103 106 10733630 10.1046/j.1365-2230.2000.00587.x
Morton CA Lafferty M Hau C Henderson I Jones M Lowe JG Pruritus and skin hydration during dialysis Nephrol Dial Transplant 1996 11 2031 2036 8918718
Giocoechea M de Sequera P Ochando A Andrea C Caramelo C Uremic pruritus: An unresolved problem in hemodialysis patients Nephron 1999 82 73 74 10224488 10.1159/000045371
Yosipovitch G Tur E Morduchowicz G Boner G Skin surface pH, moisture and pruritus in haemodialysis patients Nephrol Dial Transplant 1993 8 1129 1132 8272228
Ostlere LS Tylor C Baillod R Wright S Relationship between pruritus, transepidermal water loss, and biochemical markers of renal itch in hemodialysis patients Nephrol Dial Transplant 1994 9 1302 1304 7816295
Carmichael AJ Dickinson F McHugh MI Martin AM Farrow M Magnesium-free dialysis for uremic pruritus Br Med J 1988 297 584 85 3139220
Kyriasis J Glotsos J Dialysate calcium concentration of =or < 1.25 mmol/l: Is it effective in suppressing uremic pruritus? Nephron 2000 84 85 86 10644916 10.1159/000045546
Fantini F Baraldi A Sevignani C Spattini A Pincelli C Giannetti A Cutaneous innervation in chronic renal failure patients. An immunohistochemical study Acta Dermatol Venereol (Stockh) 1992 72 102 5
Zakrzewska-Pniewska B Jedras B Is pruritus in chronic uremic patients related to peripheral somatic and autonomic neuropathy? Study by R-R interval variation test (RRIV) and by sympathetic skin response (SSR) Neurophysiol Clin 2001 31 181 93 11488229
Tapia L Cheigh JS David DS Pruritus in dialysis patients treated with parenteral lidocaine N Eng J Med 1988 17 180 89
Tapia L Cheigh JS David DS Sullivan JF Saal S Reidenberg MM Stenzel KH Rubin AL Pruritus in dialysis patients treated with parenteral lidocaine N Engl J Med 1977 296 261 2 831109
Weisshaar E Dunker N Gollnick H Topical capsaicin therapy in humans with hemodialysis-related pruritus Neurosci Lett 2003 345 192 4 12842288 10.1016/S0304-3940(03)00511-1
Tarng DC Cho YL Liu HN Huang TP Hemodialysis-related pruritus: a double-blind, placebo-controlled, crossover study of capsaicin 0.025% cream Nephorn 1996 72 617 22
Gunal AI Ozalp G Yaldas TK Gunal SY Kirciman E Celiker H Gabapentin therapy for pruritus in hemodialysis patients: a randomized, placebo-controlled, double-blind trial Nephrol Dial Transplant 2004 19 3137 9 15575002 10.1093/ndt/gfh496
Manenti L Vaglio A Costantino E Danisi D Oliva B Pini S Prati E Testori A Gabapentin in the treatment of uremic itch: an index case and a pilot evaluation J Nephrol 2005 18 86 91 15772928
Yosipovitch G Yarnitsky D Mermelstein V Sprecher F Reiss J Witenberg C Hemli JA Boner G Paradoxical heat sensation in uremic polyneuropathy Muscle Nerve 1995 18 768 71 7783767 10.1002/mus.880180714
Mesic E Tabakovic M Habul V Atic M Lekic S Resic H Halilbasic A Trnacevic S Halilbasic A Clinical characteristics of uremic pruritus in hemodialysis patients Acta Med Croatica 2004 58 377 80 (abstract) 15756803
Winkelman JW Chertow GM Lazarus JM Restless legs syndrome in end-stage renal disease Am J Kidney Dis 1996 28 372 8 8804235
|
15975150
|
PMC1184066
|
CC BY
|
2021-01-04 16:29:46
|
no
|
BMC Dermatol. 2005 Jun 24; 5:7
|
utf-8
|
BMC Dermatol
| 2,005 |
10.1186/1471-5945-5-7
|
oa_comm
|
==== Front
BMC Fam PractBMC Family Practice1471-2296BioMed Central London 1471-2296-6-311605095810.1186/1471-2296-6-31Research ArticlePreconception care by family physicians and general practitioners in Japan Kitamura Kazuya [email protected] Michael D [email protected] Nobutaro [email protected] Department of General Medicine, Nagoya University Hospital 65 Tsurumai-cho, Showa-ku, Nagoya 466-8560 Japan2 Department of Family Medicine, University of Michigan Health System 1018 Fuller Street, Ann Arbor, Michigan, USA 48109-07082005 28 7 2005 6 31 31 30 8 2004 28 7 2005 Copyright © 2005 Kitamura et al; licensee BioMed Central Ltd.2005Kitamura et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Preconception care provided by family physicians/general practitioners (FP/GPs) can provide predictable benefits to mothers and infants. The objective of this study was to elucidate knowledge of, attitudes about, and practices of preconception care by FP/GPs in Japan.
Methods
A survey was distributed to physician members of the Japanese Academy of Family Medicine. The questionnaire addressed experiences of preconception education in medical school and residency, frequency of preconception care in clinical practice, attitudes about providing preconception care, and perceived need for preconception education to medical students and residents.
Results
Two hundred and sixty-eight of 347 (77%) eligible physicians responded. The most common education they reported receiving was about smoking cessation (71%), and the least was about folic acid supplementation (12%). Many participants reported providing smoking cessation in their practice (60%), though only about one third of respondents advise restricting alcohol intake. Few reported advising calcium supplementation (10%) or folic acid supplementation (4%). About 70% reported their willingness to provide preconception care. Almost all participants believe medical students and residents should have education about preconception care.
Conclusion
FP/GPs in Japan report little training in preconception care and few currently provide it. With training, most participants are willing to provide preconception care themselves and think medical students and residents should receive this education.
==== Body
Background
Appropriate preconception care provided by family physician/general practitioners (FP/GPs) can provide great benefits to mothers and infants. [1-4] In spite of many potentially helpful interventions prior to conception, there is little literature illustrating the effectiveness of preconception services delivery through primary care settings. Muchowski and Paladine present the evidence of effectiveness for components of preconception care that could be provided in primary care settings.[4] Korenbrot and colleagues conducted a systematic review and found no RCTs of prepregnancy interventions, though one RCT was conducted among women who had a negative pregnancy test.[5] Despite recent research demonstrating the importance of assessing primary care workers views on preconception care,[6] we found no published research on Japanese FP/GPs' approaches to the provision of preconception care to women of reproductive age.
Such research is needed because obstetricians in Japan usually cannot provide preconception counseling. Women infrequently present to OB/GYN physicians prior to conception unless they have gynecological problems. Many of our primary care colleagues consider women's health issues to be outside the realm of their practice, and they lack systematic education in preventive care during family/general medicine training. Despite a longstanding call for family physicians to provide preventive care in Japan,[7] we hypothesized that few FP/GPs are providing preconception counseling.
The best proven preconception intervention, taking folic acid supplements one to two months prior to conception, has been shown to prevent neural tube defects.[8] Unfortunately, many pregnant women do not have their first visit for prenatal care until eight weeks of pregnancy or later, even though fetal development is most vulnerable to development of neural tube defects during this time.[1] Epidemiological studies published over the last fifteen years document that prenatal supplementation with folic acid reduces the risk of neural tube defects, such as spina bifida and anencephaly. [9-11] In many countries, daily consumption of 0.4 mg of folic acid is recommended for reproductive-aged women, and 4 mg is recommended for those who previously had an affected fetus/infant.[12,13] The World Health Organization (WHO) recommends preconception care, including folic acid supplementation, for primary prevention of birth defects in developing and developed countries alike.[14]
Until recently, there has been no recommendation for Japanese women to take a folic acid supplement prior to conception. In December 2000, the Ministry of Health, Labour and Welfare (MHLW) formally recognized the importance of reproductive-aged women taking folic acid supplementation for the prevention of neural tube defects.[15] This recommendation was based on research showing that the rate of neural tube defects is reduced by about 72% when women take folic acid supplementation one to three months prior to conception. The MHLW recommended to the Japan Medical Association, Japan Society of Obstetrics and Gynecology, and Japan Pediatric Society that these organizations should provide their membership with adequate information about the value of taking folic acid to reproductive-aged women.[15]
In addition, there are other compelling topics to cover during preconception counseling based on theoretical considerations and indirect evidence. A particularly important topic in Japan is screening for immunity to rubella. The prevalence of rubella vaccination in Japan is only about 70%.[16] Rubella vaccination is not mandatory due in part to serious side effects that resulted from an MMR vaccination manufactured in Japan several years ago.[16] Antibody negative women can be safely vaccinated prior to pregnancy during preconception care.[17]
Excessive alcohol intake during pregnancy causes fetal alcohol syndrome.[18] Smoking during pregnancy is associated with low birth rate infants.[19] As post-partum hemorrhage is the most important cause of preventable maternal mortality in Japan,[20] prevention of anemia through early detection and treatment with iron supplementation merits consideration. As the average intake of calcium, especially in reproductive women in Japan, is lower than recommended in general, calcium intake is a particularly salient issue for Japanese women of childbearing age. Though controversial, exercise, and mechanisms to optimize pregnancy are among other potentially beneficial effects of preconception counseling. [1-4] Based on clinical experience with a large population of Japanese couples desiring pregnancy, but having difficulty conceiving and not willing to see an infertility specialist, we consider brief counseling on basal body temperature monitoring and timing of intercourse as topics relevant in Japan.
Given the importance of these issues in maternal child health, the purpose of this research was to elucidate the knowledge of, attitudes about, and practices of preconception care for reproductive-aged women by FP/GPs in Japan.
Methods
In this survey research, we distributed a structured questionnaire to physicians who were registered members of the Japanese Academy of Family Medicine (JAFM). The JAFM has membership based on interest, not criteria such as board certification or completion of family medicine residency training. As family medicine is still a young discipline in Japan, most members have trained in non-family medicine programs. The physician members include a diverse group: those trained in a family medicine training program in the US or Japan, those trained in another specialty or multiple specialties and became a general practitioner after entering practice, and those who trained in a general internal medicine program in Japan or the US.
The JAFM was established in 1986 and has taken a leadership role in establishing family medicine in Japan.[21] The academy had a membership of 460 during the research period. Most members of the JAFM are practicing physicians and/or teachers of family medicine. The membership also includes medical students, residents and paramedical staff who are interested in family medicine. We excluded from the analysis non-physician members, medical students, physicians not in active practice, and those who had resigned from the JAFM (Figure 1).
Figure 1 Selection of survey participants.
All participants were asked to complete a questionnaire with items addressing: 1) experiences in preconception education during medical school and residency training; 2) the frequency of providing preconception care in their practice; 3) their attitudes about providing preconception care; and 4) their perceptions of the need for preconception education for medical students and residents. The instrument was distributed with a cover letter requesting their participation. A reminder was sent to non-respondents twice at two-week intervals. A fourth mailing included the final request and a second copy of the instrument.
The components of preconception care we investigated are depicted in Table 1. Some are compelling based on best evidence (such as folic acid supplementation, smoking cessation),[5] while others were chosen based on circumstances particular to Japan as presented above.
Table 1 Preconception Care Interventions
Services Intervention
Folic acid supplementation Advise 0.4 mg of folic acid daily (4 mg if previous pregnancy with neural tube defect) three months prior to conception.
Smoking cessation Educate about the risks of smoking during pregnancy and counsel about smoking cessation.
Anemia screen Check hematocrit/CBC and recommend iron supplement if anemia is detected.
Testing for rubella antibody Check IgG rubella antibody before conception. If the test is negative, vaccinate and avoid conception for three months.
Alcohol restriction Screen for alcoholism by using a validated questionnaire, and counsel or refer if positive screen.
Restricting caffeine Restrict caffeine intake to less than 250 mg a day.
Exercise Advise regular to moderate exercise before and during pregnancy.
Calcium supplementation Assess calcium intake and as needed supplement for a target of 1200 mg daily.
Use of basal body temperature* Conduct basal body temperature every morning to identify the day of ovulation.
Timing of intercourse* Advise intercourse every other day around time of ovulation to maximize chances of conception.
*For couples having difficulty conceiving
We analyzed the data using SPSS (Statistical Package for Social Sciences). Simple statistics were calculated on the demographics data. Statistical significance of continuous measures was tested by two-tailed student t-test, and categorical data were tested by the chi-square test where appropriate. Physicians who reported their specialty as family medicine were compared for differences with physicians who reported their specialty as general internal medicine.
Results
Of the 460 members of the JAFM during the study period, we excluded the 74 non-physician and medical student members from the survey. Of the 386 physicians to whom we distributed instruments, 291 physician members responded. Sixteen letters were returned because of unknown addresses. We dropped 23 physicians from the analysis who reported that they were retired or had resigned from the academy. Thus, 268 of 347 eligible physicians (response rate 77.2%) were included in the analysis (Figure 1). The demographics of the participants are listed in Table 2. Most were men (86%) and the mean age was almost 40 years. Most physicians reported seeing reproductive-aged women in their offices.
Table 2 Participant Demographics (n = 268)
n (%)
Gender
Male 230 (85.8)
Age
Mean (range) 40.1 (25–73)
Period after graduating from med school (yrs)
Mean (range) 14.4 (1–26)
Specialty (total response = 308)*
General Internal Medicine 140 (45.5)
Family Medicine 102 (33.1)
Specialty in Internal Medicine 19 (6.2)
General Medicine 10 (3.2)
Surgery 8 (2.6)
Pediatrics 8 (2.6)
Psychiatry 6 (1.9)
Orthopedics 5 (1.6)
Others 10 (3.2)
Experiences of out-patient training in
Pediatrics 160 (59.7)
OB/GYN 99 (36.9)
Average number of patients per week
All 142 /wk
Reproductive women 11 /wk
Average number of patients by age
Child (0–15) 15.3 (8.9)
Adolescence (16–19) 9.5 (7.0)
Adult (20–64) 43.9 (35.0)
Elderly (65 and more) 69.5 (49.4)
*multiple responses were possible
With regard to educational experiences in preconception counseling during medical school and residency training (Table 3), most participants reported they had received little. The most common relevant training they reported receiving included: smoking cessation (71%), screening for anemia (64%), and blood testing for rubella antibody (58%). Few reported training experiences in providing folic acid supplementation (12%), timing of intercourse to maximize chances of conception (14%), or exercise during pregnancy (18%).
Table 3 Japanese family physicians' experiences in preconception education during medical school and residency training (n = 268)
n (%)
Smoking cessation 191 (71.3)
Testing for anemia 172 (64.2)
Blood testing for rubella antibody 156 (58.2)
Use of basal body temperature monitoring 125 (46.6)
Restricting alcohol intake 120 (44.8)
Increasing calcium intake 86 (32.1)
Restricting caffeine intake 52 (19.4)
Exercise during pregnancy 48 (17.9)
Timing of intercourse to maximize chances of conception 38 (14.2)
Folic acid supplementation 31 (11.6)
Preconception care practices of the respondents are depicted in Table 4. Many participants reported often or always addressing smoking cessation with reproductive-aged women (60%). Some reported they provide screening for anemia (35%) and counseling about restricting alcohol intake (27%). Few reported they addressed either calcium intake through foods/supplements (10%), or folic acid supplementation (4%).
Table 4 Japanese family physicians' self-reports of preconception care in their clinical practice (n = 268)
Never/Almost Never Sometimes Often/Always No Response
n (%) n (%) n (%) n (%)
Timing of intercourse to maximize chances of conception 222 (82.8) 20 (7.5) 9 (3.6) 17 (6.3)
Folic acid supplementation 217 (81.0) 24 (9.0) 11 (4.1) 16 (6.0)
Exercise during pregnancy 198 (73.9) 32 (11.9) 20 (7.5) 18 (6.7)
Testing for rubella antibody 180 (67.2) 55 (20.5) 21 (7.8) 12 (4.5)
Increasing calcium intake 168 (62.7) 61 (22.7) 27 (10.1) 12 (4.5)
Restricting caffeine intake 166 (61.9) 55 (20.5) 31 (11.6) 16 (6.0)
Use of basal body temperature monitoring 157 (58.6) 60 (22.4) 38 (14.2) 13 (4.9)
Restricting alcohol intake 107 (39.9) 74 (27.6) 72 (26.9) 15 (5.6)
Testing for anemia 56 (20.9) 112 (41.8) 94 (35.1) 6 (2.2)
Smoking cessation 37 (13.8) 63 (23.5) 162 (60.4) 6 (2.2)
Their attitudes about providing preconception care are depicted in Table 5. About two thirds of participants reported their willingness to provide preconception care about such topics as calcium intake (70%), blood testing for rubella antibody (69%), and restricting caffeine intake (64%). On the other hand, some expressed dissatisfaction with counseling about timing of intercourse to maximize chances of conception (46%), and use of basal body temperature monitoring (22%). More than 60% reported their willingness to provide folic acid supplementation, though one in four stated they would not provide it. Almost all participants think medical students (95%), and residents (91%), should have education in preconception care.
Table 5 Japanese family physicians' willingness to provide preconception care in their practice (n = 268)
Currently Provide Willing to Provide Would Not Provide No Response
n (%) n (%) n (%) n (%)
Smoking cessation 154 (57.5) 103 (38.4) 5 (1.9) 6 (2.2)
Screening for anemia 103 (38.4) 145 (54.1) 11 (4.1) 9 (3.4)
Restricting alcohol intake 90 (33.6) 143 (53.4) 26 (9.7) 9 (3.4)
Use of basal body temperature monitoring 61 (22.8) 137 (51.1) 59 (22.0) 9 (4.1)
Restricting caffeine intake 42 (15.7) 171 (63.8) 42 (15.7) 13 (3.8)
Blood testing for rubella antibody 41 (15.3) 186 (69.4) 30 (11.2) 11 (4.1)
Increasing calcium intake 37 (13.8) 18 (70.1) 34 (12.7) 9 (3.4)
Exercise during pregnancy 30 (11.2) 165 (61.6) 57 (21.3) 16 (6.0)
Folic acid supplementation 14 (5.2) 170 (54.1) 68 (25.4) 16 (6.0)
Timing of intercourse to maximize chances of conception 14 (5.2) 114 (42.5) 122 (45.5) 18 (6.7)
Though the instrument did not have open-ended questions, a number of respondents (n = 69, 26%) provided comments indicating they had been unaware of the recommendation for women to take folic acid supplementation and they were pleased to learn from the survey about this important issue. In contrast, a few (n = 11, 4%) stated they did not understand why FP/GPs should provide preconception care.
There were no statistical or clinically meaningful differences between the reports of physicians who reported their specialty as family medicine and physicians who reported their specialty as general internal medicine.
Discussion
The movement to establish family medicine in Japan started at least 20 years ago.[7] There has been much debate about whether Japanese family physicians should provide obstetric and other women's health care. There are few FP/GPs who provide OB care in Japan.[22] Our study reveals that few FP/GPs have educational experiences in the provision of preconception care, and few actually provide this care in their practices. However, it also reveals their willingness to provide preconception care in the future after appropriate educational experiences.
Japanese FP/GPs seem reluctant to inquire about human sexuality issues.[23] In the current study, some reported they could not ask patients about future pregnancy plans during a routine acute visit. Our results show FP/GPs in Japan are not accustomed to addressing preconception-related topics (timing of intercourse, folic acid supplementation, exercise during pregnancy), while they are familiar with more general topics (smoking cessation, screening for anemia, calcium intake). FP/GPs are in the unique position to provide health care services to male and female patients of all ages,[24,25] and they have many opportunities to discuss patients' concerns. For example, young parents do not often visit family physicians for their own health problems, but do come for their children. At these visits, family physicians can also discuss family planning. For this reason, FP/GPs in Japan need training in women's health care, even if they will not provide deliveries in the future. We hope these data will provide a catalyst for dialogue among Japan's FP/GPs about regular provision of preconception counseling.
Remarkably, only 10 % of participants reported knowledge of folic acid supplementation and few reported providing this care. Yamanaka surveyed pregnant women about the importance of folic acid.[26] This research revealed that only 8% reported they knew well about its importance, and 46% stated they did not know at all. Furthermore, the small percentage of participants who knew well about the importance of folic acid reported that they learned it from a newspaper, TV, or magazine, while only 16% learned it from medical professionals. Based on these data, Yamanaka emphasized that medical professionals should provide correct and concise information about folic acid.
As the MHLW only recently recommended folic acid supplementation to reproductive-aged women,[15] it is not too surprising that many FP/GPs do not know the importance of folic acid and few provide this care. The JAFM was not included in the list of organizations notified of the MHLW policy change. Based on these data it is clear that its membership is interested in women's health issues and should be included in notifications about women's health policy changes formulated at the government level.
About 70% of participants reported they are willing to screen women for rubella antibody, though almost 70% reported not providing this care in their practice. Given the historic mistrust of the rubella vaccine and the low rate of vaccination in Japan,[16] Japanese family doctors need to proactively address the topic. The lack of mandatory vaccination is a loophole in public health policy[16] and highlights why screening women of childbearing potential for rubella antibody is especially important. Delay in testing, and hence immunization, leads to an increased risk for congenital rubella, a highly serious disease. Non-immune pregnant women should post-pone rubella vaccination until after delivery.
Japan's FP/GPs must learn to provide preconception counseling in order to close this important gap in women's health. Fortunately, many of these respondents are willing to provide some level of preconception care even though they currently are not – presumably due in part to a lack of educational experience. Japan's FP/GPs need educational materials and clinical tools to encourage women to make an office visit and receive preconception care. Family medicine training in Japan heavily emphasizes adult and geriatric medicine with little emphasis on prenatal, newborn, children, adolescent, or women's health.[7,21,27] Initial efforts to disseminate information about preconception care are underway through the medical literature,[25] though other educational efforts will no doubt be needed. Data from Japanese women from Japan who are on temporary assignment in the United States, illustrate low levels of knowledge about prenatal folic acid supplementation and resistance to take supplements.[28]
A potential limitation of this study is selection bias. The participants are limited to physician members of the JAFM and the data might not reflect the current situation of all FP/GPs in Japan. This bias would likely favor the most motivated physicians and the estimates herein probably represent the upper limits of willingness to provide preconception care by the population of primary care providers in Japan. If this interpretation is correct, the need for public campaigns about the importance of preconception care and training of Japan's family physicians are even more imperative.
Despite compelling evidence of the effectiveness of folic acid supplementation and other preconception care for the reduction of serious birth defects at a very low cost, these data provide evidence that the WHO's message about the importance has not filtered down to the clinical level even in a developed country like Japan. Continued efforts to spread and diffuse the WHO's message are desperately needed for the advancement of maternal-child health.
Conclusion
Our study reveals that many Japan's FP/GPs have limited training in preconception care and few currently provide it. Most participants report their willingness to provide preconception care themselves and educational campaigns are needed to enhance preventive care provided by FP/GPs in Japan.
List of abbreviations used
CBC – complete blood count
FP/GPs – family physicians and general practitioners
IgG – immunoglobulin G
JAFM – Japanese Academy of Family Medicine
mg – milligrams
MHLW – Ministry of Health, Labour and Welfare
OB/GYN – obstetrics and gynecology
RCT – randomized controlled trial
SPSS – Statistical Package for the Social Sciences
WHO – World Health Organization
wk – week
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
KK contributed to the conception and study design, performed data analysis, interpretation, and draft the manuscript. MDF contributed to the conception and study design and critical revision of the manuscript. NB participated in the study design and critical revision of the manuscript. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We gratefully acknowledge all responding physician members of the Japanese Academy of Family Medicine.
==== Refs
Leuzzi RA Scoles KS Preconception counseling for the primary care physician Medical Clinics of North America 1996 80 337 374 8614177
Gjerdingen DK Fontaine P Preconception health care: a critical task for family physicians Journal of the American Board of Family Practice 1991 4 237 250 1927590
Brundage SC Preconception health care American Family Physician 2002 65 2507 2514 12086240
Muchowski K Paladine H An ounce of prevention: The evidence supporting periconception health care Journal of Family Practice 2004 53 126 133 14764295
Korenbrot CC Steinberg A Bender C Newberry S Preconception care: a systematic review Matern Child Health J 2002 6 75 88 12092984 10.1023/A:1015460106832
Heyes T Long S Mathers N Preconception care: Practice and beliefs of primary care workers Family Practice 2004 21 22 27 14760039 10.1093/fampra/cmh106
Ishibashi Y Why is family medicine needed in Japan Journal of Family Practice 1987 25 83 86 3598483
Lumley J Watson L Watson M Bower C Periconceptional supplementation with folate and/or multivitamins for preventing neural tube defects Cochrane Database of Systematic Reviews 2005 1
Berry RJ Li Z Erickson JD Li S Moore CA Wang H Mulinare J Zhao P Wong LY Gindler J Hong SX Correa A Prevention of neural-tube defects with folic acid in China. China-U.S. Collaborative Project for Neural Tube Defect Prevention New England Journal of Medicine 1999 341 1485 1490 10559448 10.1056/NEJM199911113412001
Czeizel AE Dudas I Prevention of the first occurrence of neural-tube defects by periconceptional vitamin supplementation New England Journal of Medicine 1992 327 1832 1835 1307234
Wald N Prevention of neural tube defects: Results of the Medical Research Council Vitamin Study Lancet 1991 338 131 137 1677062 10.1016/0140-6736(91)90133-A
Clark NA Fisk NM Minimal compliance with the Department of Health recommendation for routine folate prophylaxis to prevent fetal neural tube defects British Journal of Obstetrics and Gynaecology 1994 101 709 710 7947508
Centers for Disease Control and Prevention Recommendations for the use of folic acid to reduce the number of cases of spina bifida and other neural tube defects MMWR 1992 41 1 7
World Health Organization Services for the prevention and management of genetic disorders and birth defects in developing countries Community Genetics 1999 2 World Health Organization 196 201 14960842 10.1159/000016212
Ministry of Health Labor and Welfare The recommendation of taking folic acid for reproductive-aged women to decrease risk of neural tube defects (Japanese) 2000 2005 Tokyo, Japan
Miyazaki C MMR vaccine Pediatrics of Japan 2002 43 585 590
Centers for Disease Control and Prevention (CDC) Revised ACIP recommendation for avoiding pregnancy after receiving a rubella-containing vaccine MMWR 2001 50 1117 11794623
American Academy of Pediatrics Fetal alcohol syndrome and fetal alcohol effects Pediatrics 1993 91 1004 1006 8507280
Lumley J Oliver SS Chamberlain C Oakley L Interventions for promoting smoking cessation during pregnancy Cochrane Database of Systematic Reviews 2005 1
Nagaya K Fetters MD Ishikawa M Kubo T Koyanagi T Saito Y Sameshima H Sugimoto M Takagi K Chiba Y Honda H Mukubo M Kawamura M Satoh S Neki R Causes of maternal mortality in Japan Journal of the American Medical Association 2000 283 2661 2667 10819948 10.1001/jama.283.20.2661
Smith BW Demers R Garcia-Shelton L Family medicine in Japan Arch Fam Med 1997 6 59 62 9003172 10.1001/archfami.6.1.59
Okkes IM Polderman GO Fryer GE Yamada T Bujak M Oskam SK Green LA Lamberts H The role of family practice in different health care systems: a comparison of reasons for encounter, diagnoses, and interventions in primary care populations in the Netherlands, Japan, Poland, and the United States J Fam Pract 2002 51 72 73 11927068
Ban N Medical interview: How to evoke sex-related issues (Japanese) Journal of International Medicine 2001 11 697 700
Narato K Japanese Academy of Family Medicine What is a family physician (Japanese) The Introduction of Family Physicians and Primary Care Physicians 2001 Osaka , Premaid Press 18 22
Kitamura K Ban N Family physicians and preconception care (Japanese) Japanese Journal of Family Practice 2002 9 89 94
Yamanaka M Folic acid can prevent neural tube defect: The importance is not recognized (Japanese) Medical Tribune 2002 8 16
Kitamura K Fetters MD Ban N The experiences of Japanese generalist physicians in overseas faculty development programs Family Medicine 2002 34 761 765 12448647
Yeo SA Fetters MD Maeda Y Japanese couples' childbirth experiences in Michigan: Implications for care Birth 2000 27 191 198 11251502 10.1046/j.1523-536x.2000.00191.x
|
16050958
|
PMC1184067
|
CC BY
|
2021-01-04 16:29:13
|
no
|
BMC Fam Pract. 2005 Jul 28; 6:31
|
utf-8
|
BMC Fam Pract
| 2,005 |
10.1186/1471-2296-6-31
|
oa_comm
|
==== Front
BMC Fam PractBMC Family Practice1471-2296BioMed Central London 1471-2296-6-321607638710.1186/1471-2296-6-32Research ArticleAssessing, treating and preventing community acquired pneumonia in older adults: findings from a community-wide survey of emergency room and family physicians Krueger Paul [email protected] Mark [email protected] Caralyn [email protected] H Gayle [email protected] St. Joseph's Health System Research Network, Father Sean O'Sullivan Research Centre, Hamilton, Ontario, Canada2 Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Ontario, Canada3 St. Joseph's Lifecare Centre, Brantford, Ontario, Canada4 Department of Pathology and Molecular Medicine, McMaster University, Hamilton, Ontario, Canada5 Department of Family Relations and Applied Nutrition, University of Guelph, Guelph, Ontario, Canada2005 2 8 2005 6 32 32 7 3 2005 2 8 2005 Copyright © 2005 Krueger et al; licensee BioMed Central Ltd.2005Krueger et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Respiratory infections, like pneumonia, represent an important threat to the health of older Canadians. Our objective was to determine, at a community level, family and emergency room physicians' knowledge and beliefs about community acquired pneumonia (CAP) in older adults and to describe their self-reported assessment, management and prevention strategies.
Methods
All active ER and family physicians in Brant County received a mailed questionnaire. An advance notification letter and three follow-up mailings were used to maximize physician participation rate. The questionnaire collected information about physicians' assessment, management, and prevention strategies for CAP in older adults (≥60 years of age) plus demographic, training, and practice characteristics. The analysis highlights differences in approaches between office-based and emergency department physicians.
Results
Seventy-seven percent of physicians completed and returned the survey. Although only 16% of physicians were very confident in assessing CAP in older adults, more than half reported CAP to be a very important health concern in their practices. In-service training for family physicians was associated with increased confidence in CAP assessment and more frequent use of diagnostic tests. Family physicians who reported always requesting chest x-rays were also more likely to request pulse oximetry (OR 5.6, 95% CI 1.40 to 22.5) and recommend both follow-up x-rays (OR 5.4, 95% CI 1.7 to 16.6) and pneumococcal vaccination (OR 3.4, 95% CI 1.1 to 10.0).
Conclusion
The findings of this study provide a snapshot of how non-specialists from a non-urban Ontario community assess, manage and prevent CAP in older adults and highlight differences between office-based and emergency department physicians. This information can guide researchers and clinicians in their efforts to improve the management and prevention of CAP in older adults.
==== Body
Background
Community-acquired pneumonia (CAP) is an important threat to older adults. The majority of excess deaths and hospitalizations due to CAP occur in older adults, as reflected by over 44,000 hospitalizations for pneumonia and influenza annually in people aged 65 and older in Canada. The incidence of pneumonia increases dramatically in the very old, with rates increasing from 15 cases per 1,000 in those aged 60–74 years to 34 cases per 1,000 for those 75 years and older [1]. Clinical management decisions for CAP, such as timing of antimicrobials or site of care, can impact mortality and cost, particularly in the elderly [2,3]. Few studies, however, have examined provision of care for CAP in entire communities. A vaccination trial set in a township in Finland allowed researchers the opportunity to assess etiology, incidence, and risk factors for CAP [4-6]. However, health service delivery for CAP was not addressed. In this paper, we report the findings from a community-wide survey of family and emergency room (ER) physicians from one Ontario community to assess knowledge and beliefs about CAP and ascertain variation in management strategies.
Methods
Setting
This study was conducted in Brant County, which includes the city of Brantford and the amalgamated County of Brant. The population of Brant County in 2001 was 118,485, with 14% of the population aged 65 years and older [7]. There are two community hospitals, eight radiology centres, and approximately 80 family physicians who serve older adults in Brant County. Brant County was selected for this community-wide study because of its moderate size and population demographics. It is a predominantly English speaking community with 86% of the population reporting English as a first language.
Selection of physicians
Patients presenting with CAP in Brant County are first assessed either in a primary care office setting or in the hospital emergency department. Therefore, the target population for our survey included family and ER physicians. We created a comprehensive up-to-date list of all family and ER physicians practising in Brant County using several sources: hospital physician lists; the local telephone book; and The College of Physicians and Surgeons of Ontario's Doctor Search Internet database [8]. This list was then reviewed by research team members, two local family physicians, a hospital administrator and selected hospital staff to ensure accuracy and completeness.
Questionnaire
Using a framework described by Dillman, a questionnaire was developed to collect information about the assessment, management and prevention of CAP in older adults [9]. The questionnaire was pretested by two family physicians prior to implementation. An advance notification letter and three follow-up mailings were used to help maximize the response rate. An advance notification letter informed physicians of the survey's purpose and assured respondents of complete confidentiality. One week later the questionnaire, a return postage-paid envelope and a cover letter explaining the study's purpose was mailed to all physicians. An incentive (booklet of five $1 coupons for coffee) was also included in the package as a token of appreciation. Physicians were notified that they could opt out of receiving follow-up mailings by returning their blank questionnaire with a note stating they were not interested in participating. The following week a thank you/reminder letter was mailed to all physicians. Two weeks later a second package (without the incentive) was mailed to all physicians who had not yet responded. Two weeks later, a final package was sent by registered mail to non-responding physicians. This research received ethics approval from McMaster University.
Statistical analysis
Data from the mailed questionnaires were entered into and analyzed using SPSS 12.0.0 (SPSS, Inc., Chicago IL). Descriptive statistics were computed for all variables measured, including frequency counts and percentages, or means and standard deviations (as appropriate). We used the chi-square test or, when appropriate, Fisher's exact test to determine differences in categorical variables. Unadjusted odds ratios (ORs), 95% confidence intervals (CIs), and p-values are reported as appropriate. A probability level of <0.05 was used to determine statistical significance.
Results
Response rate
Of 98 eligible physicians, 75 (77%) returned completed questionnaires. The response rate varied by type of physician with 78% (63/81) of family physicians and 71% (12/17) of ER physicians responding.
Respondent characteristics
Of the 75 respondents, 28 (37%) classified their type of practice as "solo practice", 19 (25%) as "family physician group practice (with other family physicians)", 12 (16%) as ER, 6 (8%) as "family physician/specialist group practice (with other physician/dental specialists)", 5 (7%) as "acute & urgent care", 1 (1%) as "multi-disciplinary group practice (with independent practitioners other than MDs)", and 4 as "other" (2 walk-in clinic, 1 locum, and 1 hospitalist). Fifty-nine (79%) of the physicians reported their method of reimbursement as fee for service; 11 (15%) as salary (hospitalist and ER physicians); 3 (4%) as capitation and 2 (3%) as sessional payment (ER physicians). Seventy-two percent of respondents were male and length of time in practice ranged from one to 51 years (with a mean of 22 years). Forty-one percent of respondents reported having an undergraduate degree (other than medicine); 60% a CCFP; and 10% a graduate degree (e.g. MSc, MA).
Attitudes and knowledge about CAP
When asked to rate the importance of CAP as a health concern for older adults in their practices, 51% of physicians reported it to be very important. Although a higher percentage of ER physicians than family physicians reported CAP to be very important (58% vs 49% respectively) this difference was not statistically significant (p = 0.56). An additional 43% rated CAP as being fairly important in their practices. Only 16% of respondents reported being very confident in assessing CAP in older adults. ER physicians were significantly more likely (p = 0.02) than family physicians to respond that they were very confident in their assessments, 42% vs 11% respectively (OR 5.7, 95% CI 1.4 to 22.9, p = 0.019). An additional 75% reported being fairly confident in such assessments. Regarding education about CAP, 77% of respondents reported having attended continuing medical education (CME) events with a focus on CAP in older adults. In addition to attending CME events, respondents reported obtaining information related to assessing and treating CAP in older adults from a number of sources including: journal articles (92%), discussions with colleagues (79%), pharmaceutical representatives (55%), in-service training (12%) and the Internet (5%). Family physicians who received in-service training related to assessing and treating CAP in older adults were significantly more likely (p = 0.03) than those who did not receive such training to report being very confident in assessing CAP in older adults (OR 13.6, 95% CI 2.4 to 78.8).
Only 39% of respondents agreed or strongly agreed that their undergraduate medical education provided adequate training in assessing and treating CAP in older adults. This increased to 72% when asked about their postgraduate medical education.
Assessment and diagnostic testing
The questionnaire included an extensive list of 24 signs and symptoms that have been associated with CAP in older adults. Physicians were asked to report how frequently (always, usually, occasionally, rarely, never) they see these signs and symptoms in older adults clinically diagnosed with CAP. The most common symptoms or signs, always or usually reported, were fatigue (84%), abnormal breath sounds (81%), shortness of breath (72%), and productive cough (64%). The questionnaire also provided physicians with a list of patient characteristics and asked them to rate how important this information was when assessing and treating older adults suspected of having CAP. Characteristics that were reported as being very important to know included: other co-morbidities (73%), smoking status (61%), hydration (57%), social support (32%), age (25%), cognitive impairment (20%), alcohol consumption (19%), caregiver burden (17%), and physical disabilities (13%). ER physicians were significantly more likely (p = 0.04) to report age as being very important to know compared to family physicians (50% vs 21% respectively).
Physicians were also asked to indicate, from a list of tests, how often (always, usually, occasionally, rarely, never) they normally request each test when they suspect an older adult has CAP (Table 1). Although over 90% of family and ER physicians always or usually ordered chest radiographs, ER physicians were significantly more likely (p = 0.01) than family physicians to always request chest x-rays (92% vs 51% respectively). ER physicians were also more likely to always or usually order complete blood count (92%) and pulse oximetry (100%) than family physicians (52% and 24% respectively).
Table 1 Frequency that tests are normally requested when older adults are suspected of having CAP.
Family Physicians ER Physicians
Test Always % Usually % Combined % Always % Usually % Combined % P-value*
Chest X-ray 50.8 41.3 92.1 91.7 8.3 100 >0.05
CBC 14.3 38.1 52.4 58.3 33.3 91.6 <0.05
Pulse oximetry 14.3 9.5 23.9 91.7 8.3 100 <0.001
Sputum culture 1.6 11.3 16.1 0.0 33.3 33.3 >0.05
Blood culture 3.2 6.5 9.7 16.7 50.0 66.7 <0.001
Sputum gram stain 1.6 3.2 4.8 0.0 25.0 25.0 <0.05
Arterial blood gas 1.6 1.6 3.2 0.0 8.3 8.3 >0.05
*Comparison of "combined" percentages.
Family physicians who reported always requesting chest x-rays when older adults were suspected of having CAP were significantly more likely than those who did not always request chest x-rays to also report: usually or always requesting pulse oximetry (OR 5.60, 95% CI 1.4 to 22.5, p = 0.010); always requesting follow-up x-rays (OR 5.4, 95% CI 1.7 to 16.6, p = 0.003); and always recommending pneumococcal vaccine (OR 3.4, 95% CI 1.1 to 10.0, p = 0.027). Family physicians who reported usually or always requesting pulse oximetry when older adults were suspected of having CAP were significantly more likely than those who did not to also report: CAP as being a very important concern for older adults in their practice (OR 3.9, 95% CI 1.1 to 13.9, p = 0.032); always requesting chest x-rays when older adults are suspected of having CAP (OR 5.6, 95% CI 1.4 to 22.5, p = 0.010); usually or always using the Pneumonia Severity Index (PSI) to assist clinical judgement in admitting an older adult to hospital (OR 11.7, 95% CI 2.0 to 69.9, p = 0.007; obtaining information related to assessing and treating CAP in older adults via in-service training (OR 5.5, 95% CI 1.3 to 24.2, p = 0.029); and not having their primary method of reimbursement as fee-for-service (OR 11.9, 95% CI 1.1 to 125.0, p = 0.039).
Therapeutic management and site of care
Physicians were asked which antibiotics they normally prescribe for treating outpatient, immunocompetent, older adults with CAP. They were asked to list their first and second choices along with treatment duration. The most common antimicrobials reported by 57 family physicians were newer macrolides (by 43 or 75%), respiratory fluoroquinolones (by 6 or 11%), and beta-lactams (by 4 or 7%). The most common antimicrobials reported by 12 emergency department physicians were respiratory fluoroquinolones (by 6 or 50%), newer macrolides (by 3 or 25%), and beta-lactams (by 3 or 25%). There was very little variation in treatment duration by either ER or family physicians. Most reported prescribing these antimicrobials for seven to 10 days.
Ancillary therapy and follow-up strategies are summarized in Table 2. Use of analgesics and follow-up chest radiograph were commonly practised by the physicians surveyed. Family physicians who reported always requesting follow-up chest x-rays for older patients clinically diagnosed with CAP were significantly more likely than those who did not to: also report CAP as being a very important concern for older adults in their practice (OR 4.3, 95% CI 1.5 to 13.0, p = 0.007); and always requesting chest x-rays when older adults are suspected of having CAP (OR 5.4, 95% CI 1.7 to 16.6, p = 0.003).
Table 2 Frequency that physicians normally prescribe/recommend management strategies for older adults with clinically diagnosed CAP.
Family Physicians ER Physicians
Management Strategy Always % Usually % Combined % Always % Usually % Combined % P-value*
Analgesics / antipyretics 14.5 66.1 80.6 8.3 83.3 91.6 >0.05
Follow-up chest x-ray 38.1 39.7 77.8 8.3 50.0 58.3 >0.05
Follow-up appointment (24–48 hrs) 21.0 41.9 62.9 33.3 58.3 91.6 >0.05
Hydration 19.4 40.3 59.7 25.0 41.7 66.7 >0.05
Respiratory therapy 1.6 9.7 11.3 8.3 25.0 33.3 >0.05
Oxygen therapy 1.6 4.8 6.4 8.3 41.7 50.0 <0.001
Referral to ER 1.6 3.2 4.8 n/a n/a n/a -
Referral to physician specialist 0.0 4.8 4.8 0.00 16.7 16.7 >0.05
Hospital admission 0.0 3.3 3.3 0.0 25.0 25.0 <0.05
Home care services
Nursing care 0.0 10.0 10.0 0.0 10.0 10.0 >0.05
Nutrition assessment 0.0 3.4 3.4 0.0 10.0 10.0 >0.05
Homemaking 0.0 6.7 6.7 0.0 10.0 10.0 >0.05
IV antibiotic therapy 0.0 3.4 3.4 0.0 9.1 9.1 >0.05
*Comparison of "combined" percentages.
Physicians were asked how often they used the Pneumonia Severity Index (or similar decision tools) to assist their clinical judgement in admitting an older adult to hospital. Most family physicians (67%) reported rarely or never while half (50%) of ER physicians reported usually or always using the PSI. Some of the family physicians commented that they had no hospital privileges and others that they had never seen the PSI. Family physicians who reported usually or always using the PSI to assist clinical judgement in admitting an older adult to hospital were significantly more likely than those who did not to also report: obtaining information related to assessing and treating CAP in older adults via in-service training (OR 12.3, 95% CI 2.1 to 71.2, p = 0.008); and usually or always requesting pulse oximetry (OR 11.7, 95% CI 2.0 to 69.9, p = 0.007).
Preventing CAP in older adults
Physicians were asked how frequently (always, usually, occasionally, rarely, never) they recommended various prevention strategies to older adults in their practice setting (Table 3). Family physicians who reported always recommending annual influenza vaccine to older patients were significantly more likely than those who did not to also report always recommending pneumococcal vaccine to older patients (OR 1.7, 95% CI 1.2 to 2.4, p < 0.001). Family physicians who reported always recommending pneumococcal vaccine to older patients were significantly more likely than those who did not to also report: always requesting chest x-rays when older adults are suspected of having CAP (OR 3.4, 95% CI 1.1 to 10.0, p = 0.027); and always recommending an annual influenza vaccine to older patients (OR 4.2, 95% CI 2.6 to 6.7, p < 0.001)
Table 3 Strategies always or usually recommended by physicians to prevent pneumonia in their older adult patients.
Family Physicians ER Physicians
Prevention Strategy Always % Usually % Combined % Always % Usually % Combined % P-value*
Annual influenza vaccination 85.7 14.3 100 36.4 45.5 81.9 <0.05
Smoking cessation 84.1 15.9 100 54.5 36.4 90.9 >0.05
Pneumococcal vaccine 65.1 33.3 98.4 36.4 27.3 63.7 <0.01
Avoidance of tobacco smoke 48.4 35.5 83.9 36.4 54.5 90.9 >0.05
Frequent hand washing 20.6 44.4 65.0 18.2 18.2 36.4 >0.05
Nutritional programs 6.3 22.2 28.5 9.1 9.1 18.2 >0.05
Rehabilitation (OT and/or PT) 1.6 9.5 11.1 9.1 9.1 18.2 >0.05
*Comparison of "combined" percentages.
Discussion
This survey provides a snapshot of how non-specialists, who care for the vast majority of older patients with CAP, practise in a typical non-urban Ontario community. The results provide insight into and highlight differences in approaches between office-based and emergency department physicians.
Although no individual element of the history and physical examination possesses a high enough likelihood ratio to establish a clinical diagnosis of CAP [10], one of the signs and symptoms that respondents reported seeing most frequently in patients with CAP (abnormal breath sounds) is associated with high positive likelihood ratios (up to 8.6) [11]. In contrast, shortness of breath, one of the most frequently reported signs and symptoms, has a positive likelihood ratio of only 1.4 [12]. These findings when considered in combination with self-reported levels of confidence in assessing CAP, suggest that there are clinical knowledge gaps that could be improved through research and education.
Differences in diagnostic testing between family and emergency department physicians were predictable. Although both groups ordered chest radiographs, the vast majority of ER physicians ordered complete blood counts and pulse oximetry compared to about only half of the family physicians. Since a greater proportion of patients with pneumonia seen by ER physicians will be admitted to hospital, this reflects the fact that ER physicians apply management standards for the hospitalized patient to a greater extent than family physicians. This is also a reflection of the increased severity of illness seen in patients presenting in the emergency department as well as the greater resources available in the ER compared to office practise. Whether a chest radiograph should be ordered for suspected pneumonia in the office setting is, however, an unanswered question. One study that randomly allocated chest radiographs to 1,500 consecutive patients with chronic cough, found a beneficial change in care to only 3% of patients [13]. Whether the same applies to patients with suspected pneumonia is unknown.
It is also notable that while the vast majority (92%) of ER physicians reported always using pulse oximetry, only 24% of family physicians reported (always or usually) using this technology. This again is likely a reflection of hospital management standards being applied in the ER, reflecting illness severity. Pulse oximetry is more readily available in emergency departments. However, pulse oximetry may be useful in the office setting, particularly when assessing older patients with chronic lung disease for pneumonia [14].
Some of the differences in attitudes and knowledge about CAP are a direct reflection of different clinical experience. For example, ER physicians would be expected to be more comfortable with their assessments of CAP given that they are more familiar with patients presenting with severe illness and they are more familiar with the management guidelines. Since age plays an important role in the Pneumonia Severity Index (3), this may be why more ER physicians were more likely to report age as being very important. Other differences relate to the amount of time spent with patients. For example, the lower rates of counseling for smoking cessation and immunization among ER physicians reflect the limited amount of time spent with each patient.
One of the most rigorous studies on CAP involved the derivation and validation of the Pneumonia Severity Index [3]. This index, created using an analysis based on over 14,000 patients and validated in a cohort of over 38,000, provides an accurate assessment of prognosis with respect to patients presenting with CAP. We found that the use of this index was reported relatively infrequently by family and ER physicians in our study. Our finding that fewer family physicians (12%) than ER physicians (50%) reported always or usually using the PSI is perhaps not surprising given the greater severity of illness typically seen in the ER. Although most physicians reported not using the index, one feature of the index that physicians did indicate was very important when assessing older patients with CAP was other co-morbidities (73% of respondents). Age, however, is one of the most important predictors of death, but overall only 26% of respondents felt that it was a very important characteristic to know when assessing/treating older adults suspected of having CAP.
We found that practice patterns among family physicians, such as requesting pulse oximetry, ordering chest x-rays for older patients, and using the Pneumonia Severity Index, were clustered. Although we did not have a large enough sample size to conduct a multivariable analysis, these physicians did state that CAP was an important concern for older patients in their practice. It may be that these physicians had a disproportionate number of older patients compared to other family physicians.
Regarding therapy, both family and ER physicians followed recommended Canadian guidelines for empiric therapy of CAP, with family physicians prescribing newer macrolides and emergency department physicians prescribing respiratory fluoroquinolones [15].
Conclusion
The fact that the majority of family physicians (86%) reported always recommending influenza vaccine to patients compared to 36% in the emergency department raises the question as to whether the emergency department provides a good opportunity for immunization. The fact that physician practises tended to be clustered is also an interesting finding. For example, family physicians who ordered chest radiographs were more likely to order pulse oximetry, request follow up chest radiograph and to recommend pneumococcal vaccine.
The findings of this study provide a snapshot of how non-specialists from a non-urban Ontario community assess, manage and prevent CAP in older adults and highlight differences between office-based and emergency department physicians. Knowledge and beliefs about CAP were found to be associated with assessment, management and prevention strategies. An understanding of this connection between what physicians think and how they respond to CAP can guide researchers and clinicians in their ongoing efforts to improve the management and prevention of CAP in older adults.
Competing interests
The author(s) declare that they have no competing interests
Authors' contributions
PK had a major role in the conception and design of the study, supervised all aspects of the study's implementation, had a major role in data analysis and interpretation, and was the lead writer of this manuscript.
ML contributed to the conception and study design, participated in data analysis and interpretation, contributed to the writing of the manuscript, and provided editorial comments.
CK contributed to the implementation and acquisition of study data, participated in data analysis and interpretation, contributed to the writing of the manuscript and provided editorial comments.
GE contributed to the study design, implementation and acquisition of study data, critical review of the manuscript and provided editorial comments.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We gratefully acknowledge those who helped pretest the survey instrument and all the physicians who participated in this study. This survey was part of a larger study that was funded by a research grant from the Canadian Institutes of Health Research.
==== Refs
Canadian Institute of Health Information Health Report 1997
Meehan TP Fine MJ Krumholz HM Scinto JD Galusha DH Mockalis JT Weber GF Petrillo MK Houck PM Fine JM Quality of care, process, and outcomes in elderly patients with pneumonia JAMA 1997 278 2080 4 9403422 10.1001/jama.278.23.2080
Fine MJ Auble TE Yealy DM Hanusa BH Weissfeld LA Singer DE Coley CM Marrie TJ Kapoor WN A prediction rule to identify low-risk patients with community-acquired pneumonia N Engl J Med 1997 336 243 50 8995086 10.1056/NEJM199701233360402
Koivula I Sten M Makela PH Risk factors for pneumonia in the elderly Amer J Med 1994 96 313 20 8166149 10.1016/0002-9343(94)90060-4
Jokinen C Heiskanen Juvonen H Kallinen S Karkola K Korppi M Kurki S Ronnberg PR Seppa A Soimakallio S Incidence of community-acquired pneumonia in the population of four municipalities in eastern Finland Am J Epidemiol 1993 137 977 88 8317455
Jokinen C Heiskanen L Juvonen H Kallinen S Kleemola M Koskela M Leinonen M Ronnberg PR Saikku P Sten M Tarkiainen A Tukiainen A Pyorala K Makela PH Microbial etiology of community-acquired pneumonia in the adult population of 4 municipalities in eastern Finland Clin Infect Dis 2001 32 1141 54 11283803 10.1086/319746
Statistics Canada 2001 Community Profiles Catalogue No: 93F0053XIE Ottawa 2002
College of Physicians and Surgeons of Ontario Doctor Search
Dillman DA Mail and telephone surveys: the total design method 1978 NewYork: John Wiley & Sons
Metlay JP Fine MJ Testing strategies in the initial management of patients with community-acquired pneumonia Ann Intern Med 2003 138 109 18 12529093
Diehr P Wood RW Bushyhead J Krueger L Wolcott B Tompkins RK Prediction of pneumonia in outpatients with acute cough – a statistical approach J Chronic Dis 1984 37 215 25 6699126 10.1016/0021-9681(84)90149-8
Gennis P Gallagher J Falvo C Baker S Than W Clinical criteria for the detection of pneumonia in adults: guidelines for ordering chest roentgenograms in the emergency department J Emerg Med 1989 7 263 8 2745948 10.1016/0736-4679(89)90358-2
Bushyhead JB Wood RW Tompkins RK Wolcott BW Diehr P The effect of chest radiographs on the management and clinical course of patients with acute cough Med Care 1983 21 661 73 6350743
Zoorob RJ Campbell JS Acute dyspnea in the office Am Fam Physician 2003 68 1803 10 14620600
Mandell LA Marrie TJ Grossman RF Chow AW Hyland RH Canadian guidelines for the initial management of community-acquired pneumonia: an evidence-based update by the Canadian Infectious Diseases Society and the Canadian Thoracic Society. The Canadian Community-Acquired Pneumonia Working Group Clin Infect Dis 2000 31 383 421 10987698 10.1086/313959
|
16076387
|
PMC1184068
|
CC BY
|
2021-01-04 16:29:12
|
no
|
BMC Fam Pract. 2005 Aug 2; 6:32
|
utf-8
|
BMC Fam Pract
| 2,005 |
10.1186/1471-2296-6-32
|
oa_comm
|
==== Front
BMC GastroenterolBMC Gastroenterology1471-230XBioMed Central London 1471-230X-5-201596975710.1186/1471-230X-5-20Research ArticleThe presence of the proteolysis-inducing factor in urine does not predict the malignancy of a pancreatic tumour Teich Niels [email protected] Jörg [email protected] Herbert [email protected]össner Joachim [email protected] Volker [email protected] Helmut [email protected] Johann [email protected] Universität Leipzig, Medizinische Klinik und Poliklinik II, Leipzig, Germany2 Universität Heidelberg, Chirurgische Universitätsklinik, Heidelberg, Germany3 Charité Berlin, Medizinische Klinik und Poliklinik, Gastroenterologie, Hepatologie und Endokrinologie, Berlin, Germany2005 21 6 2005 5 20 20 10 2 2005 21 6 2005 Copyright © 2005 Teichs et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The proteolysis-inducing factor (PIF) was identified as a tumour product in various gastrointestinal cancers. A previous study in pancreatic cancer patients suggested PIF expression as a tumour marker, which is not related to tumour size. We hypothesized that PIF could be a useful marker to exclude benign pancreatic tumors, as chronic pancreatitis with a pancreatic mass.
Methods
Urine of patients with a pancreatic mass of uncertain malignancy was investigated for PIF expression by Western blot. Sufficient urine protein for analysis was available in 59 patients. The diagnosis was established by histology in 54 patients and by follow up in five patients with chronic pancreatitis. In addition, serum CA19-9 was measured.
Results
The sensitivity (specifity) for the detection of a malignant pancreatic tumour was 90% (75%) and 54% (71%) for CA19-9 and PIF, respectively. The sensitivity (specifity) for the distinction of pancreatic cancer from chronic pancreatitis was 89% (80%) and 57% (63%) for CA19-9 and PIF, respectively.
Conclusion
Evaluation of PIF in urine is of no diagnostic value in patients with a pancreatic mass of unknown malignancy.
==== Body
Background
Distinction between benign and malignant pancreatic tumours is still difficult, despite significant progress in imaging techniques. Patients with chronic pancreatitis are at increased risk to develop pancreatic cancer [1]. Pancreatic inflammation, as observed in chronic pancreatitis, can be mistaken on imaging as cancer and inversely. Serum carbohydrate antigen 19-9 (CA19-9) levels are elevated in 80% of pancreatic cancer patients, but can also be increased in 20% of patients with chronic pancreatitis [2]. An accurate and non-invasive test to differentiate pancreatic cancer from chronic pancreatitis is not available.
Cancer frequently induces cachexia, but not all cancer patients will develop cachexia. This is not necessarily a late phenomenon in tumour progression. It may be present at diagnosis and may be the leading symptom that induces search for a tumour. A study of pancreatic cancer patients revealed that patients had lost a median of about 14% of their usual body weight at the time of diagnosis, and that this weight loss was progressive, increasing to a median of 25% at the time of the last assessment [3]. The central role in this process of – predominant – skeletal muscle waste seems to play the proteolysis-inducing factor (PIF) [4].
PIF is a 24-kDa sulphated glycoprotein synthesized by cachexia-inducing murine and human tumours, which induces catabolism of myofibrillar proteins in skeletal muscle via a direct stimulation of the proteasome pathway in muscle cells [5]. Administration of PIF to normal mice leads to a rapid decrease in body weight, which is based primarily on a loss of skeletal muscle mass [4,6].
PIF is expressed in a variety of gastrointestinal cancers [7]. It was detected in the urine of 44 from 55 pancreatic cancer patients, who had a significantly greater total weight loss and rate of weight loss than patients whose urine did not contain PIF. Interestingly, PIF expression was not dependent on cancer stage, but seemed to be a qualitative marker of pancreatic cancer: even stage 2 tumours expressed PIF in 83 per cent [8].
PIF has not been investigated in benign pancreatic diseases. We therefore hypothesized that PIF production in patients with a pancreatic tumour would clearly indicate the malignant nature of the disease. Special attention was paid to a clear distinction to a pancreatic mass caused by chronic pancreatitis.
Methods
100 patients with a pancreatic mass of uncertain malignancy (detected by computed tomography, magnet resonance tomography, ultrasound and/or endoscopic retrograde cholangiopancreaticography) were investigated. Patients with UICC stage 4 (locally extended and metastasised disease) were not enrolled. After collection of urine from all study participants, we investigated 30 ml urine for the presence of PIF with a specific mouse monoclonal antibody in a Western blot setting as described recently [4,8]. In brief, urine protein was precipitated with ammonium sulphate and dialysed against water between 12 and 15 hours (overnight) with a molecular weight cut-off of 10 kD (Slide-A-Lyzer Dialysis Cassette). 5 μg of concentrated samples were separated with sodium dodecylsulphate polyacrylamide gel electrophoresis and subsequently blotted on nitrocellulose membranes. After incubation with 10 μg/ml of the mouse monoclonal PIF antibody (provided with courtesy by MJ Tisdale) and streptavidine horseradish peroxidase conjugate, the bands were detected using the Fluorescent ECL Plus system (Amersham). In addition, CA19-9 in serum was measured with an electrochemiluminescence immunoassay (CA19-9®Roche Diagnostics or ADVIA Centaur®, Bayer Healthcare), the upper limit of normal was 22 U/ml. The ethics committees of the three participating centres approved this protocol.
Results and discussion
The analysis failed in 41 patients due to insufficient amount of urinary protein after dialysis. In 59 patients, PIF Western blot was successfully performed (figure 1). This cohort was subsequently assessed: it included 31 male and 28 female patients (median age 59 years, range 33–89 years). The diagnoses of the patients were shown in figure 2. Final diagnosis was done by histology in 54 patients or follow up for at least one year in five patients with chronic pancreatitis, who did not undergo surgery.
The diagnostic values of PIF for the detection of a malignant pancreatic tumour and with special attention to pancreatic cancer are shown in table 1. Summarizing all 59 patients, the median CA19-9 values in patients with benign or malignant pancreatic tumours were 16.5 (standard deviation (SD): 141) and 477 (SD: 22173 U/l), respectively (p < 0.05). The median CA19-9 values in patients with chronic pancreatitis and pancreatic cancer were 15.0 (SD: 18) and 478.5 U/l (SD: 22022), respectively (p < 0.05). The analysis of CA19-9 in dependence on the presence of PIF in urine in PIF negative patients revealed 22.5 U/l (SD: 742) and 580.5 U/l in PIF positive patients (SD: 25203) (p < 0.05). PIF was detected in two patients with CA19-9 negative pancreatic cancer, but was not detectable in 16 pancreatic cancer patients (15 of them with elevated CA19-9).
This is the first attempt to evaluate the proteolysis-inducing factor as a diagnostic marker of pancreatic cancer in patients with a potentially resectable pancreatic tumour. We found a weak association with malignancy, but the diagnostic value to distinguish benign from malignant pancreatic tumours is lower than CA19-9. PIF seems to be of a rather limited importance to answer this question in every-day care.
A major drawback of our data is the high amount of patients with insufficient urinary protein after extraction and dialysis. For practical reasons in this multi-centre collaboration, we tried to investigate PIF in a small urine sample, as initially suggested by Todorov et al. [4]. Our problem could be potentially solved by the collection of higher urine volumes in further studies.
PIF was detected in 6 patients with chronic pancreatitis without evidence for a malignancy by 12 months follow up or histology. This raises the question about the origin of PIF in these patients. Although the majority of investigations found PIF expression only in tumour tissue, the expression of PIF in non-malignant tissue has been also reported [7]. The primary role of PIF seems to be in regulation of development [9]. After cloning of the cDNA for PIF, a human homologue – the human cachexia associated protein (HCAP) – has been identified, which is minimally expressed in normal tissues [10,11]. It can be speculated, that the ongoing inflammation in patients with chronic pancreatitis is capable to generate PIF. This could explain the low specifity of PIF to detect pancreatic malignancies in our study. Three patients with pancreatic cancer had normal CA19-9 values, two of them were PIF-positive. A sequential testing of PIF in the case of a CA19-9-negative pancreatic tumour may be beneficial but has to be considered cautiously based on these small patient numbers. However, even this approach would fail to detect all pancreatic cancer patients in our study.
Although our data indicate that PIF is not helpful as a diagnostic marker of pancreatic cancer, a benefit may be the early identification of patients who need nutritional intervention [12]. In previous studies the detection of PIF was associated with prior weight loss. PIF induces an increase of the ubiquitin – proteasome activity resulting in protein catabolism. Administration of the polyunsaturated fatty acid eicosapentaenoic acid (EPA) attenuates protein degradation by antagonizing the PIF induced up regulation of the ubiquitin – proteasome proteolytic pathway in cachectic tumour bearing mice [13]. A first randomised placebo – controlled trial in patients with pancreatic cancer suggest that a EPA enriched oral supplement has the potential to induce a net gain of weight, lean body mass and improvement of quality of life [14].
Conclusion
PIF is not superior to the established tumour marker CA19-9 to distinguish benign from malignant pancreatic tumours. Further investigations should clarify whether the onset of PIF expression in the long-term follow-up of chronic pancreatitis patients is associated with early malignancy and whether it precedes morphologic and clinical signs of pancreatic cancer. In future, the evaluation of PIF as an indicator for early nutritional intervention seems to be warranted.
Abbreviations
proteolysis-inducing factor PIF
pancreatic cancer PaCa
chronic pancreatitis CP
carbohydrate antigen 19-9 CA19-9
eicosapentaenoic acid EPA
standard deviation SD
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
N.T and J.O. had the initial idea and initiated the study. N.T. co-ordinated the laboratory analysis. All authors ascertained patients in this multicenter investigation, wrote and discussed the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We thank Professor M.J. Tisdale for kindly providing the anti-PIF-antibody. We furthermore thank S. Kistner and T. Herrmann for expert technical assistance.
Figures and Tables
Figure 1 Western blot analysis for PIF after urinary protein extraction. Lane 1: Rainbow® recombinant protein molecular weight marker (#RPN 800, Amersham Life Science); lane 2: healthy control person, lanes 3 and 5: chronic pancreatitis, PIF not detected, lanes 4 and 7: pancreatic cancer, UICC stage 1 and 2, respectively, PIF detected (24 kD, broken line); lanes 6: chronic pancreatitis, PIF detected; lane 8: positive control (pancreatic cancer, UICC stage 4)
Figure 2 Diagnoses in PIF positive and PIF negative patients
Table 1 Diagnostic values of PIF in comparison to CA19-9 to discriminate malignant vs. benign and pancreatic cancer (PaCa) vs. chronic pancreatitis (CP).
all malignant vs. all benign PaCa vs. CP
PIF CA19-9 PIF CA19-9
Sensitivity 54% 90% 57% 89%
Specifity 71% 75% 63% 80%
positive predictive value 73% 88% 74% 93%
negative predictive value 52% 80% 44% 73%
==== Refs
Malka D Hammel P Maire F Rufat P Madeira I Pessione F Levy P Ruszniewski P Risk of pancreatic adenocarcinoma in chronic pancreatitis Gut 2002 51 849 52 12427788 10.1136/gut.51.6.849
Satake K Takeuchi T Comparison of CA19-9 with other tumor markers in the diagnosis of cancer of the pancreas Pancreas 1994 9 720 724 7846015
Wigmore SJ Plester CE Richardson RA Fearon KCH Changes in nutritional status associated with unresectable pancreatic cancer Br J Cancer 1997 75 106 109 9000606
Todorov P Cariuk P McDevitt T Coles B Fearon K Tisdale M Characterization of a cancer cachectic factor Nature 1996 379 739 42 8602222 10.1038/379739a0
Lorite MJ Smith HJ Arnold JA Morris A Thompson MG Tisdale MJ Activation of ATP-ubiquitin-dependent proteolysis in skeletal muscle in vivo and murine myoblasts in vitro by a proteolysis-inducing factor (PIF) Br J Cancer 2001 85 297 302 11461093 10.1054/bjoc.2001.1879
Tisdale MJ The 'cancer cachectic factor' Support Care Cancer 2003 11 73 78 12560934
Cabal-Manzano R Bhargava P Torres-Duarte A Marshall J Bhargava P Wainer IW Proteolysis-inducing factor is expressed in tumours of patients with gastrointestinal cancers and correlates with weight loss Br J Cancer 2001 84 1599 1601 11401311 10.1054/bjoc.2001.1830
Wigmore SJ Todorov PT Barber MD Ross JA Tisdale MJ Fearon KCH Characteristics of patients with pancreatic cancer expressing a novel cancer cachectic factor Br J Surg 2000 87 53 58 10606911 10.1046/j.1365-2168.2000.01317.x
Caruik P Lorite MJ Todorov PT Field WN Wigmore SJ Tisdale MJ Induction of cachexia in mice by a product isolated from the urine of cachectic cancer patients Br J Cancer 1997 76 606 613 9303359
Schittek B Hipfel R Sauer B Bauer J Kalbacher H Stevanovic S Schirle M Schroeder K Blin N Meier F Rassner G Garbe C Dermcidin: a novel antibiotic peptide secreted by sweat glands Nat Immunol 2001 2 1133 1137 11694882 10.1038/ni732
Porter D Weremowicz S Chin K Seth P Keshaviah A Lahti-Domenici J Bae YK Monitto CL Merlos-Suarez A Chan J Hulette CM Richardson A Morton CC Marks J Duyao M Hruban R Gabrielson E Gelman R Polyak K A neural survival factor is a candidate oncogene in breast cancer Proc Natl Acad Sci USA 2003 100 10931 10936 12953101 10.1073/pnas.1932980100
Ockenga J Pirlich M Gastell S Lochs H Tumoranorexie – Tumorkachexie in der Gastroenterologie: Standards und Visionen Z Gastroenterol 2002 40 929 936 12436371 10.1055/s-2002-35411
Whitehouse AS Smith HJ Drake JL Tisdale MJ Mechanism of attenuation of skeletal muscle protein catabolism in cancer cachexia by eicosapentaenoic acid Cancer Res 2001 61 3604 3609 11325828
Fearon KC von Meyenfeldt MF Moses AG van Geenen R Roy A Gouma DJ Giacosa A van Gossum A Bauer J Barber MD Aaronson NK Voss AC Tisdale MJ Effect of a protein and energy dense N-3 fatty acid enriched oral supplement on loss of weight and lean tissue in cancer cachexia: a randomised double blind trial Gut 2003 52 1479 1486 12970142 10.1136/gut.52.10.1479
|
15969757
|
PMC1184069
|
CC BY
|
2021-01-04 16:03:27
|
no
|
BMC Gastroenterol. 2005 Jun 21; 5:20
|
utf-8
|
BMC Gastroenterol
| 2,005 |
10.1186/1471-230X-5-20
|
oa_comm
|
==== Front
BMC GenomicsBMC Genomics1471-2164BioMed Central London 1471-2164-6-851593875510.1186/1471-2164-6-85Research ArticleFunctional analysis and comparative genomics of expressed sequence tags from the lycophyte Selaginella moellendorffii Weng Jing-Ke [email protected] Milos [email protected] Clint [email protected] Department of Biochemistry, Purdue University, West Lafayette, IN 47907, USA2 Department of Botany and Plant Pathology, Purdue University, West Lafayette, IN 47907, USA3 current address, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY 11724, USA2005 6 6 2005 6 85 85 5 3 2005 6 6 2005 Copyright © 2005 Weng et al; licensee BioMed Central Ltd.2005Weng et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The lycophyte Selaginella moellendorffii is a member of one of the oldest lineages of vascular plants on Earth. Fossil records show that the lycophyte clade arose 400 million years ago, 150–200 million years earlier than angiosperms, a group of plants that includes the well-studied flowering plant Arabidopsis thaliana. S. moellendorffii has a genome size of approximately 100 Mbp, as small or smaller than that of A. thaliana. S. moellendorffii has the potential to provide significant comparative information to better understand the evolution of vascular plants.
Results
We sequenced 2181 Expressed Sequence Tags (ESTs) from a S. moellendorffii cDNA library. One thousand three hundred and one non-redundant sequences were assembled, containing 291 contigs and 1010 singletons. Approximately 75% of the ESTs matched proteins in the non-redundant protein database. Among 1301 clusters, 343 were categorized according to Gene Ontology (GO) hierarchy and were compared to the GO mapping of A. thaliana tentative consensus sequences. We compared S. moellendorffii ESTs to the A. thaliana and Physcomitrella patens EST databases, using the tBLASTX algorithm. Approximately 60% of the ESTs exhibited similarity with both A. thaliana and P. patens ESTs; whereas, 13% and 1% of the ESTs had exclusive similarity with A. thaliana and P. patens ESTs, respectively. A substantial proportion of the ESTs (26%) had no match with A. thaliana or P. patens ESTs.
Conclusion
We discovered 1301 putative unigenes in S. moellendorffii. These results give an initial insight into its transcriptome that will aid in the study of the S. moellendorffii genome in the near future.
==== Body
Background
Our understanding of biology has been greatly improved by studying genome structure and gene function of a broad sampling of model organisms such as Mus musculus (mouse), Drosophila melanogaster (fruit fly), Danio rerio (zebrafish), Caenorhabditis elegans (nematode), and Arabidopsis thaliana [1-5]. Comparative genomics has made it clear that orthologs of many proteins that act as signal transduction components, transcriptional regulatory factors, and metabolic enzymes can be identified between and among these model organisms [6]. As a result, the knowledge gained from comparative and evolutionary studies of these species can provide insights into homologous processes in a wide range of other organisms, varying from crop plants to humans [7]. Within plants however, most of the efforts in genomics have been focused on crop plants or economically important plants such as Oryza sativa (rice), Zea mays (maize), and Lycopersicon esculentum (tomato) [8-10]. Thus, coupled with the sequencing of the A. thaliana genome, these efforts have provided data on only a single branch of the plant evolutionary tree, namely members of the Monocotyledonae and Dicotyledonae, collectively termed the angiosperms and commonly known as flowering plants. As a result, the community of plant scientists has little sequence data on other plant lineages that could provide insights into common mechanisms of how plants develop and survive in a terrestrial environment, nor do they have any kind of evolutionary benchmarks that might reveal how angiosperms have come to dominate most world ecosystems [11].
Clear evidence for the existence of angiosperms is present in the fossil record of the lower Cretaceous (140 million years ago), and some evidence suggests their existence 60 million years earlier, around the same time that conifers and ginkgos arose [12]. In contrast, fossil evidence for the lycophytes is found in strata dated to approximately 420 million years ago [13]. Thus, this clade diverged very early from the lineage that led to all other vascular plants (Figure 1), and has existed on earth over twice as long as plants that are the most common subjects of current laboratory and agricultural research. As such, the study of lycophytes may provide novel insights into plant biology that would not be provided by research that focuses only on flowering plants.
Figure 1 A simplified version of the plant phylogenetic tree simplified and condensed from Pryer et al. [11]. The tree shows that lycophytes (highlighted) diverged from other vascular plant lineages soon after plants colonized the terrestrial environment. Representative species were chosen from sub-clades within the clades listed, and illustrate major developments in plant evolution including the colonization of land (land plants, L), the development of vasculature (vascular plants, V) and true leaves (euphyllophytes, E), and the evolution of flowers (flowering plants, F), and seeds (seed plants, S).
Selaginella is an extant genus of the lycophyte clade. It is sometimes referred to as a 'seed-free' plant to highlight the fact that it has not evolved flowers and seeds in the time since its divergence from other plant lineages. It has a number of characteristics that would make its study convenient for, and valuable to, the plant biology community [11,14]. For example, like many other species of Selaginella, S. moellendorffii (Figure 2) is a small diploid plant that can be easily grown in the laboratory. Further, it has an approximate genome size of 100 Mbp [14], smaller than that of A. thaliana, and among the smallest published genome sizes for 'seed-free' genera. Because of these attributes, S. moellendorffii was recently chosen as one of the non-crop plants for BAC library construction in a NSF funded Green Plant BAC library Project [15]. More importantly, the Department of Energy Joint Genome Institute (JGI) has officially announced that it will sequence the S. moellendorffii genome [16], making this species a target of extreme interest for research into comparative plant genomics, biochemistry, and development.
Figure 2 The morphology of S. moellendorffii. (a) A greenhouse grown S. moellendorffii. (b) A close up of an aerial branch of S. moellendorffii indicating the bulbils (white arrows) that can be used for clonal propagation and sporangia (black arrows) containing microspores and megaspores for sexual propagation.
Expressed sequence tag (EST) sequencing has been used as an efficient and economical approach for large-scale gene discovery [17]. It has also successfully provided frameworks for many genome projects [18,19]. Recently, a large number of ESTs have been generated from various plant species and deposited in GenBank, including both model and crop plants like A. thaliana, rice, wheat, and maize as well as species representative of clades other than angiosperms, such as gymnosperms, cycads, and mosses [20-23]. Although over 1000 ESTs from another Selaginella species S. lepidophylla, also known as the resurrection plant, have also been deposited in GenBank [20], no manuscript has been published reporting on their analysis. In this paper, we describe 2181 ESTs generated from a S. moellendorffii cDNA library. These ESTs were assembled into 1301 clusters, annotated using the BLASTX algorithm, surveyed for their abundance within the dataset, and classified into functional groups according to the Gene Ontology (GO) hierarchy. Finally, a comparative genomics approach was used for comparing S. moellendorffii ESTs with those of A. thaliana and Physcomitrella patens to look for genes unique to S. moellendorffii.
Results and Discussion
Generation of S. moellendorffii cDNA library and ESTs
To gain a broad coverage of S. moellendorffii transcripts, we collected and pooled whole S. moellendorffii plants for mRNA extraction and subsequent cDNA library construction. To enrich for full-length cDNA clones, double-stranded cDNA was size-fractionated before cloning. Based upon the average insert sizes of 35 cDNA clones chosen at random from the library, we estimate that the cDNA library has an average insert size of 850 bp. 2304 clones were sequenced from the 5' end of the cDNAs, which generated 2181 vector-trimmed EST sequences with an average sequencing read length of 640 bp.
Assembly of S. moellendorffii ESTs
To identify overlapping EST sequences, reduce sequencing error and produce non-redundant EST data for further functional annotation and comparative analysis, 2181 ESTs were assembled into clusters through stackPACK v2.2 clustering system [24]. Based upon regions of nucleotide identity, EST sequences were merged into contiguous consensus sequences (contigs). One thousand three hundred and one non-redundant EST clusters, putatively regarded as unigenes, were generated, consisting of 291 contigs and 1010 singletons. The cluster size varied from one to 105 copies of any given EST (Figure 3). Manual inspection of the assembled ESTs identified 10 clusters counted as unigenes that may actually represent non-overlapping sequence reads from cDNAs corresponding to four single genes. As an example, three unigenes were found to be best aligned to three different regions of the same protein in a BLASTX analysis (described in the following paragraph), suggesting we lack a complete transcript for their accurate assembly. Conversely, we also found that some clustered ESTs did not necessarily have identical sequences within their overlapping regions. In most of the cases, regions of sequence disagreement within the clusters tend to appear towards the ends of the EST reads, which is likely to be caused by errors generated during sequencing. In some other cases, it may due to failure to discriminate between gene family members during clustering, or allelic diversity in S. moellendorffii.
Figure 3 Distribution of S. moellendorffii ESTs by cluster size. ESTs were clustered into putative unigene sets using StackPack v. 2.2, and the number of cluster members of each size category was plotted relative to their abundance within the EST collection.
Annotation of S. moellendorffii ESTs
To annotate S. moellendorffii ESTs, the 1301 putative unigenes were translated dynamically in all 6 reading frames and searched for homology against the NCBI non-redundant (nr) protein database using BLASTX [25]. BLASTX hits with E-values less than 10-5 were taken to be significant. Among 1301 unigenes, 962 (74%) had BLASTX hits in the nr database, while the remaining 339 (26%) had hits with E-values greater than 10-5 or no hit. When a less permissive cutoff E-value of 10-10 was adopted, the numbers of unigenes with BLASTX hits and without BLASTX hits changed slightly to 891 (68%) and 410 (32%) respectively. Our dataset showed that the inferred translation products of most S. moellendorffii ESTs appear to be similar to proteins in other organisms but that there was also a percentage of ESTs that represented potential Selaginella- or lycophyte-specific genes. Interestingly, 15 ESTs had at least their top five BLASTX hits from non-plant organisms, including six from bacteria or cyanobacteria (SmoC-1_02_N06, SmoC-1_01_C17, SmoC-1_02_B19, SmoC-1_06_K12, SmoC-1_cn167, SmoC-1_03_D21), two from fungi (SmoC-1_06_O23, SmoC-1_02_H20), one from an insect (SmoC-1_06_K02), three from nematodes (SmoC-1_04_D10, SmoC-1_02_L08, SmoC-1_cn108), one from fish (SmoC-1_04_F24), and two from mammals (SmoC-1_02_H05, SmoC-1_03_F21). These data suggest that homologs have either not yet been identified or are absent from other plant lineages, although in one case (SmoC-1_06_O23), a more distantly related A. thaliana gene was returned by BLASTX, and in a further three cases, BLASTN analysis of the EST-others database identified potential homologs in P. patens (SmoC-1_02_N06, SmoC-1_06_K12) and S. lepidophylla (SmoC-1_cn167).
Highly represented S. moellendorffii ESTs
EST copy number can be used to approximate gene expression levels in an organism, although there are artifacts of cDNA library construction that may limit or over-represent certain transcripts [26]. Table 1 summarizes the first 32 most abundantly represented transcripts in the S. moellendorffii EST collection, having six or more EST copies in each cluster, with their identities putatively assigned by BLASTX analysis of the assembled contigs. As expected, a large number of the S. moellendorffii ESTs are photosynthesis-related genes, with 19 clusters containing 213 ESTs (9% of total sequenced ESTs) corresponding to genes involved in photosynthesis. There were seven clusters matching to core proteins of photosynthesis reaction centers, including four subunits of photosystem I (PSI-G, PSI-H, PSI-L, PSI-N), and three photosystem II proteins (PsbW, OEC23, CP22). There were four contigs corresponding to light-harvesting chlorophyll a/b-binding proteins, including one early light-induced protein. We also found ESTs for the RuBisCO small subunit, carbonic anhydrase, plastocyanin, one subunit of cytochrome b6f complex, ferredoxin and ferredoxin/NADP oxidoreductase, proteins involved in carbon fixation and photosynthetic electron transport. There were two putative anti-oxidative proteins found within S. moellendorffii ESTs: chloroplastic iron superoxide dismutase and catalase, presumably required for the decomposition of superoxide and hydrogen peroxide [27,28]. The BLASTX results show that all of these highly expressed S. moellendorffii photosynthetic genes had homologs in A. thaliana genome, consistent with previous observation that the photosynthesis machinery has been highly conserved throughout plant evolution.
Table 1 The most abundantly represented ESTs in the S. moellendorffii cDNA library.
Cluster Number of ESTs Top BLASTX hit in non-redundant protein database
Accession Number Best Identity Description E-value
1 SmoC-1_cn126 105 - Novel -
2 SmoC-1_cn125 46 - Novel -
3 SmoC-1_cn018 31 SP:P16031 Ribulose bisphosphate carboxylase small subunit) [Larix laricina] 8E-51
4 SmoC-1_cn121 25 SP:P04669 Ferredoxin, chloroplast precursor [Silene latifolia subsp. alba] 2E-26
5 SmoC-1_cn106 17 PIR:S16294 chlorophyll a/b-binding protein [Lycopersicon esculentum] 9E-99
6 SmoC-1_cn107 17 GB:AAM46780 latex plastidic aldolase-like protein [Hevea brasiliensis] 1E-164
7 SmoC-1_cn171 17 PIR:S31863 chlorophyll a/b-binding protein [Pinus sylvestris] 1E-106
8 SmoC-1_cn011 14 GB:AAC78107 photosystem-1 H subunit GOS5 [Oryza sativa] 8E-30
9 SmoC-1_cn233 13 SP:Q9SXW9 Plastocyanin, chloroplast precursor [Physcomitrella patens] 2E-37
10 SmoC-1_cn025 11 SP:P51118 glutamine synthetase cytosolic isoenzyme 1 [Vitis vinifera] 1E-152
11 SmoC-1_cn089 11 GB:AAG17036 S-adenosylmethionine synthetase [Pinus contorta] 7E-17
12 SmoC-1_cn195 11 SP:P11432 Early light-induced protein, chloroplast precursor (ELIP) [Pisum sativum] 1E-32
13 SmoC-1_cn023 9 SP:P82977 Subtilisin-chymotrypsin inhibitor [Triticum aestivum] 4E-11
14 SmoC-1_cn145 9 - Novel -
15 SmoC-1_cn179 9 SP:P30361 Cytochrome B6-F complex iron-sulfur subunit 1, chloroplast precursor [Nicotiana tabacum] 3E-74
16 SmoC-1_cn189 9 - Novel -
17 SmoC-1_cn006 8 GB:AAG59875 PSII subunit PsbW [Physcomitrella patens] 5E-13
18 SmoC-1_cn078 8 SP:O48560 Catalase 3 [Glycine max] 0
19 SmoC-1_cn211 8 SP:P23993 Photosystem I reaction center subunit XI, chloroplast precursor [Hordeum vulgare] 2E-55
20 SmoC-1_cn226 8 PDB:1EKJA Carbonic Anhydrase [Pisum Sativum] 2E-63
21 SmoC-1_cn019 7 REF:NP_175963 photosystem I reaction center subunit V, chloroplast, [Arabidopsis thaliana] 2E-34
22 SmoC-1_cn108 7 PIR:T23512 hypothetical protein K08H10.2a [Caenorhabditis elegans] 1E-12
23 SmoC-1_cn215 7 GB:AAB88617 ubiquitin conjugating enzyme [Zea mays] 3E-82
24 SmoC-1_cn218 7 SP:P27494 Chlorophyll a-b binding protein 36, chloroplast precursor [Nicotiana tabacum] 1E-127
25 SmoC-1_cn013 6 PIR:T06471 core protein [Pisum sativum] 1E-20
26 SmoC-1_cn016 6 SP:Q9SLQ8 Oxygen-evolving enhancer protein 2, chloroplast precursor [Cucumis sativus] 1E-79
27 SmoC-1_cn033 6 GB:AAM97011 expressed protein [Arabidopsis thaliana] 6E-40
28 SmoC-1_cn136 6 GB:AAO49652 photosystem I-N subunit [Phaseolus vulgaris] 2E-37
29 SmoC-1_cn139 6 DBJ:BAC66946 chloroplastic iron superoxide dismutase [Barbula unguiculata] 3E-69
30 SmoC-1_cn180 6 EMB:CAB71293 chloroplast ferredoxin-NADP+ oxidoreductase precursor [Capsicum annuum] 1E-139
31 SmoC-1_cn208 6 SP:P54773 Photosystem II 22 kDa protein, chloroplast precursor [Lycopersicon esculentum] 5E-61
32 SmoC-1_cn250 6 - Novel -
Non-redundant protein database includes all non-redundant GenBank CDS translations (GB)+ RefSeq Proteins (REF) +PDB + SwissProt (SP) + PIR + PRF. The identities of ESTs were putatively described by the top BLASTX hit (with lowest E-value) of the assembled EST contigs.
Three highly expressed S. moellendorffii transcripts corresponded to genes encoding enzymes of metabolism, including an aldolase-like protein, a putative glutamine synthetase cytosolic isoenzyme involved in nitrogen assimilation [29,30], and a putative S-adenosylmethionine synthetase required for the synthesis of the major methyl group donor involved in the methylation of a variety of biomolecules ranging from histones to secondary metabolites, and for the biosynthesis of ethylene [31,32].
Other relatively abundant ESTs included one encoding a putative subtilisin-chymotrypsin inhibitor, exhibiting 49% amino acid sequence identity with the wheat subtilisin-chymotrypsin inhibitor, which may play a role in plant defense by inhibiting the serine proteinases of pathogens [33]. Two transcripts that matched an A. thaliana expressed protein and Pisum sativum core protein may function as membrane channel proteins. Interestingly, one highly expressed EST matched with an E-value of 10-12 a C. elegans protein of unknown function, and is only more distantly related to an A. thaliana late embryogenesis abundant protein.
There were five highly expressed ESTs that did not yield significant matches using BLASTX (E>10-5). These are putative Selaginella-specific genes and may encode proteins with functions unique to Selaginella or lycophytes. The first two highly expressed ESTs in this project, represented by clusters SmoC1_cn126 and SmoC1_cn125, had 105 and 46 copies in their clusters respectively, but returned no BLASTX hits with the nr protein database or BLASTN hits with the NCBI EST-others database. To determine whether these sequences represented bona fide Selaginella genes, we amplified the corresponding sequences by PCR using genomic DNA as a template (data not shown). Both sequences amplified successfully, and both had introns, indicating that they were not derived from DNA contamination from prokaryotic symbionts. The rational translation of SmoC1_cn126 contig contains a three repeats of the motif "XXXGXXTCDKCAQTGVCTCGKN", which aligns with similar cysteine-rich motifs in proteins with epidermal growth factor repeats. Using a low BLASTX stringency (E = 0.002), SmoC1_cn125 matched to a Cynodon dactylon metallothionein-like protein (GB:AAS88721.1, 75% identical within a 20 amino acid motif). The other three highly expressed S. moellendorffii specific ESTs lack hints for functional annotation. The biological function of the proteins encoded by these genes, and the question of whether high transcript abundance is predictive of high protein expression will be a matter for future investigation.
Functional categorization of S. moellendorffii ESTs
The most sensitive method to find new members of known gene families among EST sequences is to search for homology of the translated ESTs to motifs extracted from a multiple alignment of known gene family members [18]. To functionally categorize S. moellendorffii ESTs using motif homology searches, we translated the 1301 unigenes in six reading frames and imported them into InterProScan [34], which aligned 491 clusters to InterPro entries (E<10-5). Mapping of InterPro entries to GO [35], assigned 343 out of 491 InterPro hits with 562 GO accession numbers. The 562 accession numbers further generated 964 individual GO mappings in the three major ontologies (biological processes, molecular functions and cellular components) [36]. The apparent discrepancies between these values arises from the fact that not all InterPro hits had available GO accession numbers associated with them, one InterProScan entry could be assigned to more than one GO accession numbers, and one GO accession number could be mapped under multiple parental categories [37].
Tables 2 and Figure 4 summarize the GO assignment of S. moellendorffii ESTs in terms of biological processes, molecular functions and cellular components, covering a broad range of the GO functional categories. Using the downloaded A. thaliana GO assignments from the TIGR A. thaliana Gene Index [38,39], we compared the distribution of GO categories between S. moellendorffii ESTs and A. thaliana tentative consensus sequences (TCs). Table 3 shows that the distribution patterns of GO assignments of S. moellendorffii and A. thaliana transcripts were generally similar, with a few exceptions in some categories. Besides the true differences in functional distribution of unigenes, some of the differences could be due to the difference in EST data sources between these two species. For example, in terms of biological processes, A. thaliana has a higher percentage in 'response to stimulus and stress' and 'development' than S. moellendorffii. Considering that among the A. thaliana ESTs in the TIGR database, some were generated from plants at specific developmental stages or from plants exposed to specific biotic or abiotic stimuli, it is very likely that ESTs from orthologs of these genes would be missing from the S. moellendorffii ESTs which were generated from normal mature plants.
Table 2 The GO categorization of S. moellendorffii ESTs by biological process, molecular function, and cellular component.
Gene Ontology term Representation Representation percentage
Biological process Metabolism 312 74%
Biosynthesis 64 15%
Protein metabolism 57 14%
Catabolism 22 5%
Nucleic acid metabolism 20 5%
Cell growth and/or maintenance 53 13%
Transport 44 10%
Response to stimulus and stress 19 5%
Photosynthesis 16 4%
Cell communication 15 4%
Signal transduction 12 3%
Homeostasis 3 1%
Development 1 <1%
Cell death 1 <1%
Molecular function Catalytic activity 132 36%
Hydrolase activity 36 10%
Transferase activity 25 7%
Oxidoreductase activity 22 6%
Kinase activity 12 3%
Binding 107 29%
Nucleotide binding 65 18%
Metal ion binding 20 5%
Transporter activity 64 18%
Electron transporter activity 16 4%
Carrier activity 12 3%
Structural molecule activity 40 11%
Translation regulator activity 10 3%
Signal transducer activity 4 1%
Chaperone activity 3 1%
Enzyme regulator activity 2 1%
Motor activity 1 <1%
Transcription regulator activity 1 <1%
Cellular component Intracellular 135 75%
Membrane 45 25%
Note that one gene product may be assigned to more than one GO terms, and one children term can fit into multiple parental categories. The representation means the number of non-redundant ESTs that can be mapped to a certain GO term. The representation percentage is based on the total number of GO mappings in each of the three major ontologies (biological process: 420, molecular function: 364, cellular component: 180).
Table 3 Comparison of GO assignments between A. thaliana ESTs and S. moellendorffii ESTs.
Gene Ontology term Categories Representation percentage
S. moellendorffii A. thaliana
Biological process Metabolism 74% 39%
Cell growth and/or maintenance 13% 13%
Response to stimulus and stress 5% 16%
Photosynthesis 4% <1%
Cell communication 4% 6%
Homeostasis 1% 1%
Development <1% 6%
Cell death <1% 1%
Molecular function Catalytic activity 36% 41%
Binding 29% 32%
Transporter activity 18% 8%
Structural molecule activity 11% 2%
Translation regulator activity 3% 1%
Signal transducer activity 1% 1%
Chaperone activity 1% 2%
Enzyme regulator activity 1% 1%
Motor activity <1% 1%
Transcription regulator activity <1% 7%
Cellular component Intracellular 75% 70%
Membrane 25% 19%
The GO assignments for A. thaliana ESTs were obtained from TIGR [38]. The percentage of GO assignments for A. thaliana was calculated based on the total numbers of GO mappings in each of the three major ontologies with the number of unknown terms deducted from them (biological process: 20185, molecular function: 23680, cellular component: 6309). The functional categories present in A. thaliana but not in S. moellendorffii were not listed in the table.
Figure 4 Representations of Gene Ontology (GO) mapping results for S. moellendorffii non-redundant ESTs. (a) Biological process (b) Molecular function (c) Cellular component.
The current GO annotations for plants are based solely on the annotated proteins of A. thaliana and O. sativa, both of which are angiosperms. Since the lycophyte clade diverged from other plant lineages 400 million years ago, and 200 million years before angiosperms, it is perhaps to be expected that a large proportion of S. moellendorffii genes could not be accurately assigned to GO categories in the database containing only angiosperm gene entries. We expect that the representation of plant species other than angiosperms will certainly benefit resources as InterPro and in turn will lead to further resolution within GO.
Comparative genomics of S. moellendorffii ESTs
One important objective of comparative genomics is to trace gene evolution including the emergence, development, and loss of orthologous genes in different organisms over evolutionary time [40]. To survey the S. moellendorffii ESTs in an evolutionary context, we used the S. moellendorffii unigene sequences as queries to search for homologous sequences in the A. thaliana and P. patens EST databases using tBLASTX algorithm (cut off E-value = 10-6). There were two reasons that we chose A. thaliana and P. patens ESTs as tBLASTX databases. First, A. thaliana and P. patens are representatives of the most diverged lineages of land plants, namely angiosperms and bryophytes. They flank Selaginella in the plant phylogenetic tree, and last shared a common ancestor over 400 million years ago [23], thus providing ample opportunity for the evolutionary divergence of individual genes and gene families. Second, the large quantities of A. thaliana and P. patens ESTs in GenBank (472,278 and 104,027 respectively) provide a substantial coverage of the transcriptome in these two species. Using them as BLAST databases makes it possible to do a relatively comprehensive genomic analysis even in the absence of the full genome sequence of P. patens.
Figure 5 summarizes the distribution of S. moellendorffii ESTs by tBLASTX results. Among 1301 non-redundant S. moellendorffii ESTs, 788 (61%) ESTs had homology with both A. thaliana and P. patens ESTs. These ESTs probably identify non-dispensable genes, which tend to be evolutionarily conserved in all land plants [41]. 168 (13%) ESTs had exclusive similarity with A. thaliana ESTs, and may represent the genes that evolved in land plants after the divergence of bryophytes, or those that were lost from the genomes of mosses. Table 4 shows the top 20 S. moellendorffii EST tBLASTX hits for A. thaliana ESTs that were not present within the P. patens EST database ranked by tBLASTX E-values. Among these, it is possible to identify candidates that might have contributed to the success of vascular plants, including those involved in functions such as lignification (SmoC-1_05_G17) [42], cell division control (SmoC-1_01_E02) [43], intracellular transport (SmoC-1_02_C05 and SmoC-1_05_G03) [44,45], responses to sulfur starvation (SmoC-1_03_C14) [46], dehydration (SmoC-1_06_M11), and viral infection (SmoC-1_06_P21) [47]. Only 8 (1%) S. moellendorffii ESTs had similarity only with P. patens ESTs. These ESTs may represent genes that arose early in plant evolution but were lost later after the divergence of the lycophytes. It should be noted, however, that all eight of these ESTs had relative low tBLASTX score (E-value around 10-10), limiting our certainty that the homologous ESTs in P. patens are true orthologs. Finally, there were 337 (26%) ESTs that had no tBLASTX match in the A. thaliana and P. patens EST databases. These ESTs may be Selaginella-specific genes, possibly having evolved only in lycophytes after their divergence from other lineages or having arisen after the divergence of bryophytes and later being lost in euphyllophytes.
Figure 5 A Venn diagram showing the distribution of S. moellendorffii EST tBLASTX matches by databases.The 1301 translated S. moellendorffii non-redundant ESTs were used as queries in homology searches against A. thaliana and P. patens EST databases, respectively. The two inner circles contain the numbers and percentages of S. moellendorffii ESTs that share tBLASTX similarity with A. thaliana or P. patens ESTs. The region between inner circles and outer circle represents S. moellendorffii ESTs without tBLASTX matches.
Conclusion
We sequenced 2181 ESTs from the lycophyte S. moellendorffii, putatively representing 1301 unigenes. Our data showed that a large proportion of the genes had homologous genes in the well-studied model plant A. thaliana and other plant species. By browsing the putative functional annotations of these ESTs, researchers will be able to choose S. moellendorffii genes of interest and compare them to their othologs in other species. We also found a substantial number of putative Selaginella-specific genes that do not share similarity with known genes, with some of them even representing very highly expressed genes. Considering the complexity of the plant kingdom and a time span more than 150 million years between the divergences of lycophytes and angiosperms, it will not be surprising to identify gene functions in S. moellendorffii that are not present in A. thaliana. When the draft genome sequence of S. moellendorffii is completed and released, this EST resource will also play an important role in the mapping and annotation of the genome. As a member of a clade that arose after the bryophytes and before all other vascular plants, S. moellendorffii will provide new opportunities in studying plant evolution, particularly those adaptations relating to fundamental traits that facilitated the transition of green plants to the land, such as lignification in vascular plants, root/stem/leaf organography, complex patterns of sporophyte branching, and the elaboration of reproductive structures.
Methods
Plant material and cDNA library Construction
S. moellendorffii was obtained from Plant Delights Nursery (Raleigh, NC). Plants were grown at 23°C in a greenhouse with a photoperiod of 16h light/8h dark. The cDNA library used in this study was made from RNA extracted from pooled tissue including stems, microphylls, strobilis, and rhizophores of S. moellendorffii plants. Briefly, fresh tissue was ground in liquid nitrogen and total RNA was extracted using the RNeasy Max Kit (QIAGEN, Valencia, CA), treated with RNase-free DNase, and precipitated in 2 M lithium chloride. Poly A+ RNA was isolated from total RNA using the Dynabeads mRNA Purification Kit (Dynal Biotech, Brown Deer, WI). The cDNA library was constructed from 1 μg mRNA using the Creator Smart cDNA Library Construction Kit (CLONTECH, Palo Alto, CA). After first-strand synthesis, the full length double stranded cDNAs were synthesized by primer-extension. Full length double stranded cDNAs were digested with Sfi I and size fractionated using a CHROMA SPIN-400 column (CLONTECH, Palo Alto, CA). cDNA-containing fractions were pooled, and ethanol precipitated. The cDNAs were then cloned into pDNR-LIB at Sfi I site, and electroporated into E. coli DH10B cells (Invitrogen, Carlsbad, CA). The library had an un-amplified titer of 1.6 × 106 colony-forming units mL-1 and a total complexity of 3.2 × 106 colonies. To estimate the average insert size of the library, plasmid DNAs were extracted from 35 randomly picked clones from the library, digested with Sfi I, and analyzed by agarose gel electrophoresis.
EST sequencing and dbEST submission
18,432 colonies from un-amplified S. moellendorffii cDNA library were arrayed into 48 384-well plates using Q-Pix multifunction colony picker (Genetix). Plasmid DNA was isolated from 2304 clones picked from the first six 384-well plates. Sequences of cDNAs were determined from their 5' end by conventional procedures using the big-dye terminators on the ABI 3730xl DNA analyzer (Applied Biosystems, Foster City, CA) at the Purdue Genomics Center using T7-ZL (5'-TAATACGACTCACTATAGGG-3') as the 5'-sequencing primer. The vector sequence was trimmed from the original EST sequences resulting in 2181 sequences. The 2181 ESTs have been submitted to GenBank dbEST under the accession numbers DN837577 to DN839757 [20].
EST clustering and homology search
2181 EST sequences were imported into the stackPACK v2.2 clustering system (Electric Genetics, Reston, VA) through WebPipe for clustering with default setting, and contig consensus sequences were generated from the clusters. One thousand three hundred and one non-redundant EST sequences were exported through WebReport in FASTA format. BLASTX analyses using the nr database were performed on the 1301 unigene sequences, using E-value of 10-5 as a cutoff threshold. The complete BLASTX annotation of 1301 S. moellendorffii unigenes can be viewed at [48].
Functional categorization of ESTs
To search for functional protein domains of translated ESTs, 1301 unigene sequences were merged into one FASTA file and imported into InterProScan, which was run on a local SUN unix server. BlastProDom, Coil, FPrintScan, HMMPIR, HMMPfam, HMMSmart, HMMTigr, ProfileScan, ScanRegExp, and Seg superfamily were selected as the database methods. All the sequences were translated in six reading frames and aligned to the entries in the selected databases. EST clusters which had positive InterProScan hits (E <10-5) were automatically assigned InterPro accession numbers. According to the mapping of InterPro entries to GO [35], GO accession numbers were assigned to EST clusters, which were used to classify ESTs into functional groups by molecular function, cellular component, and biological process. In comparison of the distribution of GO categories between S. moellendorffii ESTs and A. thaliana TCs, the GO assignments for A. thaliana ESTs were obtained from TIGR [38]. The Complete Interpro assignment and GO mapping of S. moellendorffii ESTs can be accessed in the supplemental data (see Additional file: 1).
Comparison of S. moellendorffii ESTs to A. thaliana and P. patens ESTs
472,278 A. thaliana ESTs and 104,027 P. patens ESTs retrieved from GenBank by searching 'Arabidopsis / Physcomitrella and gbdiv est' in NCBI Entrez [25] were saved to a local server. The 1301 S. moellendorffii unigenes were translated in six reading frames and searched for homology against the six-frame translations of A. thaliana ESTs and P. patens ESTs respectively using the BLAST algorithm. An E-value of 10-6 was set as stringency threshold. The complete result of S. moellendorffii unigenes tBLASTX against A. thaliana and P. patens ESTs can be viewed at [48].
Genomic PCR
To amplify the genomic sequences of the two most highly expressed ESTs (SmoC1_cn126 and SmoC1_cn125) in S. moellendorffii, PCR was performed using genomic DNA extracted from 50 mg fresh tissue of S. moellendorffii as described previously [49] as template and two pairs of PCR primers designed from their EST contig sequences: CC1170 (5'-CGAGCTCGTAGTGATAGTGTC -3') and CC1171 (5'-AACCATAGGAGAGGAAGACC-3') for SmoC1_cn126; CC1228 (5'-ATAGCTTAGCTGCTTTCTTCTC-3') and CC1229 (5'-ATACTACTCATGTCGCAGCTC -3') for SmoC1_cn125. PCR was performed using an initial 2 min denaturation at 94°C, followed by 25 cycles, each consisting of a 0.5 min denaturation at 94°C, a 0.5 min annealing at 50°C, and a 1 min extension at 72°C. These 25 cycles were followed by a 5 min extension at 72°C. PCR products were purified using QIAquick PCR Purification Kit (QIAGEN) and sequenced at Purdue Genomics Center.
Authors' contributions
JKW constructed the S. moellendorffii cDNA library, participated in the EST sequencing, carried out the bioinfomatic analysis of the ESTs, and performed the genomic PCR for two transcripts. MT participated in the S. moellendorffii cDNA library construction and provided comments on the manuscript. CC conceived the study and coordinated work. JKW and CC wrote the article. All authors read and approved the final manuscript.
Table 4 Top 20 S. moellendorffii EST tBLASTX hits for A. thaliana ESTs that are not present within the P. patens EST database.
Non-redundant EST tBLASTX E-value Best BLASTX Descriptor in A. thaliana Accession Number
1 SmoC-1_01_H05 1E-107 expressed protein REF:NP_194688
2 SmoC-1_02_C05 1E-99 oligopeptide transporter OPT family protein REF:NP_192815
3 SmoC-1_01_L23 2E-99 putative Mg-protoporphyrin IX chelatase REF:NP_196867
4 SmoC-1_05_G17 2E-99 putative caffeoyl-CoA 3-O-methyltransferase REF:NP_195131
5 SmoC-1_05_K13 5E-90 chloroplast membrane protein (ALBINO3) REF:NP_180446
6 SmoC-1_01_E02 1E-89 cullin family protein REF:NP_567243
7 SmoC-1_05_G03 7E-87 putative UDP-galactose/UDP-glucose transporter REF:NP_563949
8 SmoC-1_05_I19 6E-86 expressed protein REF:NP_566060
9 SmoC-1_02_N15 9E-86 nicotinate phosphoribosyltransferase family protein REF:NP_179923
10 SmoC-1_03_I01 7E-80 glycoside hydrolase family 77 protein REF:NP_181616
11 SmoC-1_cn293 4E-77 amine oxidase family protein REF:NP_181830
12 SmoC-1_03_C24 9E-73 uridylyltransferase-related protein REF:NP_564010
13 SmoC-1_02_P14 5E-70 expressed protein REF:NP_199542
14 SmoC-1_06_P21 2E-69 RNase L inhibitor protein-related REF:NP_196569
15 SmoC-1_05_G10 3E-69 expressed protein REF:NP_191746
16 SmoC-1_03_C14 1E-66 putative isoflavone reductase REF:NP_565107
17 SmoC-1_03_N06 6E-65 transducin / WD-40 repeat family protein REF:NP_190148
18 SmoC-1_06_M11 1E-63 dehydration stress-induced protein GB:AAM62648
19 SmoC-1_06_B20 1E-60 putative membrane protein REF:NP_849987
20 SmoC-1_05_O21 2E-60 paired amphipathic helix repeat-containing protein REF:NP_186781
The tBLASTX E-value of an EST varies with its BLASTX E-value in a small range (e.g. SmoC-1_01_H05 has a tBLASTX E-value of 1E-107 against its homologous A. thaliana EST and a BLASTX E-value of 2E-94 against the translated full length A. thaliana cDNA.). The homology ranking was based on the tBLASTX E-value. The identities of ESTs were putatively described by the A. thaliana protein with the lowest E-value in the BLASTX analysis.
Supplementary Material
Additional file 1
The Complete Interpro assignment and GO mapping of S. moellendorffii ESTs, Excel file.
Click here for file
Acknowledgements
This research was supported by a grant from the National Science Foundation to C.C. and a pilot project grant from the Department of Biochemistry, Purdue University. This is journal paper number 2005-17677 from the Purdue University Agricultural Experiment Station. We thank Dr. Jo Ann Banks for critically reading the manuscript.
==== Refs
Mouse Genome Sequencing Consortium Initial sequencing and comparative analysis of the mouse genome Nature 2002 420 520 562 12466850 10.1038/nature01262
Adams MD Celniker SE Holt RA Evans CA Gocayne JD Amanatides PG Scherer SE Li PW Hoskins RA Galle RF The genome sequence of Drosophila melanogaster Science 2000 287 2185 2195 10731132 10.1126/science.287.5461.2185
Grunwald DJ Eisen JS Headwaters of the zebrafish – emergence of a new model vertebrate Nat Rev Genet 2002 717 724 12209146 10.1038/nrg892
The C. elegans Sequencing Consortium Genome Sequence of the Nematode C. elegans: A Platform for Investigating Biology Science 1998 282 2012 2018 9851916 10.1126/science.282.5396.2012
Arabidopsis Genome Initiative Analysis of the genome sequence of the flowering plant Arabidopsis thaliana Nature 2000 408 796 815 11130711 10.1038/35048692
O'Brien SJ Menotti-Raymond M Murphy WJ Nash WG Wienberg J Stanyon R Copeland NG Jenkins NA Womack JE Marshall Graves JA The Promise of Comparative Genomics in Mammals Science 1999 286 458 481 10521336 10.1126/science.286.5439.458
Miller W Makova KD Nekrutenko A Hardison RC Comparative genomics Annu Rev Genomics Hum Genet 2004 5 15 56 15485342 10.1146/annurev.genom.5.061903.180057
Yu J Hu S Wang J Wong GK Li S Liu B Deng Y Dai L Zhou Y Zhang X A draft sequence of the rice genome (Oryza sativa L. ssp. indica) Science 2002 296 79 92 11935017 10.1126/science.1068037
Martienssen RA Rabinowicz PD O'Shaughnessy A McCombie WR Sequencing the maize genome Curr Opin Plant Biol 2004 7 102 107 15003207 10.1016/j.pbi.2004.01.010
Tanksley SD Ganal MW Prince JP de Vicente MC Bonierbale MW Broun P Fulton TM Giovannoni JJ Grandillo S Martin GB High density molecular linkage maps of the tomato and potato genomes Genetics 1992 132 1141 60 1360934
Pryer KM Schneider H Zimmer EA Banks JA Deciding among green plants for whole genome studies Trends in Plant Sci 2002 7 550 554 12475497 10.1016/S1360-1385(02)02375-0
Stewart WN Rothwell GW Paleobotany and the evolution of plants 1993 2 Cambridge University Press, Cambridge, UK
Kenrick P Crane PR The origin and early evolution of plants on land Nature 2003 389 33 39 10.1038/37918
Wang W Tanurdzic M Luo M Sisneros N Kim HR Weng JK Kudrna D Mueller C Arumuganathan K Carlson J Construction of a bacterial artificial chromosome library from the spikemoss Selaginella moellendorffii: A new resource for plant comparative genomics BMC Plant Biol 2005 5 10 15955246 10.1186/1471-2229-5-10
The Green Plant BAC Library Project
JGI Approved Community Sequencing Program Projects for 2005
Whitfield CW Band MR Bonaldo MF Kumar CG Liu L Pardinas JR Robertson HM Soares MB Robinson GE Annotated expressed sequence tags and cDNA microarrays for studies of brain and behavior in the honey bee Genome Res 2002 12 555 566 11932240 10.1101/gr.5302
Jongeneel CV Searching the expressed sequence tag (EST) databases: panning for genes Brief Bioinform 2000 1 76 92 11466975
Adams MD Kelley JM Gocayne JD Dubnick M Polymeropoulos MH Xiao H Merril CR Wu A Olde B Moreno RF Complementary DNA sequencing: expressed sequence tags and human genome project Science 1991 252 1651 1656 2047873
NCBI expressed sequence tag database
Kirst M Johnson AF Baucom C Ulrich E Hubbard K Staggs R Paule C Retzel E Whetten R Sederoff R Apparent homology of expressed genes from wood-forming tissues of loblolly pine (Pinus taeda L.) with Arabidopsis thaliana Proc Natl Acad Sci USA 2003 100 7383 7388 12771380 10.1073/pnas.1132171100
Brenner ED Stevenson DW McCombie RW Katari MS Rudd SA Mayer KF Palenchar PM Runko SJ Twigg RW Dai G Expressed sequence tag analysis in Cycas, the most primitive living seed plant Genome Biol 2003 4 R78 14659015 10.1186/gb-2003-4-12-r78
Nishiyama T Fujita T Shin-I T Seki M Nishide H Uchiyama I Kamiya A Carninci P Hayashizaki Y Shinozaki K Comparative genomics of Physcomitrella patens gametophytic transcriptome and Arabidopsis thaliana: implication for land plant evolution Proc Natl Acad Sci USA 2003 100 8007 8012 12808149 10.1073/pnas.0932694100
stackPACK
NCBI
McCarter JP Mitreva MD Martin J Dante M Wylie T Rao U Pape D Bowers Y Theising B Murphy CV Analysis and functional classification of transcripts from the nematode Meloidogyne incognita Genome Biol 2003 4 R26 12702207 10.1186/gb-2003-4-4-r26
McKersie BD Murnaghan J Jones KS Bowley SR Iron-superoxide dismutase expression in transgenic alfalfa increases winter survival without a detectable increase in photosynthetic oxidative stress tolerance Plant Physiol 2000 122 1427 1438 10759538 10.1104/pp.122.4.1427
Fita I Rossmann MG The active center of catalase J Mol Biol 1985 185 21 37 4046038 10.1016/0022-2836(85)90180-9
Mann AF Fentem PA Stewart GR Identification of two forms of glutamine synthetase in barley (Hordeum Vulgare) Biochem Biophys Res Commun 1979 88 515 521 37833 10.1016/0006-291X(79)92078-3
Oliveira IC Coruzzi GM Carbon and Amino Acids Reciprocally Modulate the Expression of Glutamine Synthetase in Arabidopsis Plant Physiol 1999 121 301 310 10482686 10.1104/pp.121.1.301
Yang SF Hoffman NE Ethylene biosynthesis and its regulation in higher plants Annu Rev Plant Physiol 1984 35 155 189 10.1146/annurev.pp.35.060184.001103
Lamblin F Saladin G Dehorter B Cronier D Grenier E Lacoux J Bruyant P Laine E Chabbert B Girault F Overexpression of a heterologous sam gene encoding S-adenosylmethionine synthetase in flax (Linum usitatissimum) cells: Consequences on methylation of lignin precursors and pectins Physiol Plant 2001 112 223 232 11454228 10.1034/j.1399-3054.2001.1120211.x
Poerio E Gennaro SD Maro AD Farisei F Ferranti P Parente A Primary structure and reactive site of a novel wheat proteinase inhibitor of subtilisin and chymotrypsin Biol Chem 2003 384 295 304 12675523 10.1515/BC.2003.033
Mulder NJ Apweiler R Attwood TK Bairoch A Barrell D Bateman A Binns D Biswas M Bradley P Bork P The InterPro Database, 2003 brings increased coverage and new features Nuc Acids Res 2003 31 315 318 10.1093/nar/gkg046
Mapping of InterPro entries to GO
Gene Ontology Consortium
Gene Ontology Consortium Creating the gene ontology resource: design and implementation Genome Res 2001 11 1425 1433 11483584 10.1101/gr.180801
TIGR Arabidopsis Gene Index
Quackenbush J Cho J Lee D Liang F Holt I Karamycheva S Parvizi B Pertea G Sultana R White J The TIGR Gene Indices: analysis of gene transcript sequences in highly sampled eukaryotic species Nucleic Acids Res 2001 29 159 164 11125077 10.1093/nar/29.1.159
Mirkin BG Fenner TI Galperin MY Koonin EV Algorithms for computing parsimonious evolutionary scenarios for genome evolution, the last universal common ancestor and dominance of horizontal gene transfer in the evolution of prokaryotes BMC Evol Biol 2003 3 2 12515582 10.1186/1471-2148-3-2
Jordan IK Rogozin IB Wolf YI Koonin EV Essential genes are more evolutionarily conserved than are nonessential genes in bacteria Genome Res 2002 12 962 968 12045149 10.1101/gr.87702. Article published online before print in May 2002
Guo D Chen F Inoue K Blount JW Dixon RA Downregulation of caffeic acid 3-O-methyltransferase and caffeoyl CoA 3-O-methyltransferase in transgenic alfalfa. impacts on lignin structure and implications for the biosynthesis of G and S lignin Plant Cell 2001 13 73 88 11158530 10.1105/tpc.13.1.73
Kipreos ET Lander LE Wing JP He WW Hedgecock EM cul-1 is required for cell cycle exit in C. elegans and identifies a novel gene family Cell 1996 85 829 839 8681378 10.1016/S0092-8674(00)81267-2
Koh S Wiles AM Sharp JS Naider FR Becker JM Stacey G An oligopeptide transporter gene family in Arabidopsis Plant Physiol 2002 128 21 29 11788749 10.1104/pp.128.1.21
Norambuena L Marchant L Berninsone P Hirschberg CB Silva H Orellana A Transport of UDP-galactose in plants: Identification and functional characterization of AtUTr1, an Arabidopsis thaliana UDP-galactose/UDP-glucose transporter J Biol Chem 2002 277 32923 32929 12042319 10.1074/jbc.M204081200
Petrucco S Bolchi A Foroni C Percudani R Rossi GL Ottonello S A maize gene encoding an NADPH binding enzyme highly homologous to isoflavone reductases is activated in response to sulfur starvation Plant Cell 1996 8 69 80 8597660 10.1105/tpc.8.1.69
Braz AS Finnegan J Waterhouse P Margis R A plant orthologue of RNase L inhibitor (RLI) is induced in plants showing RNA interference J Mol Evol 2004 59 20 30 15383904 10.1007/s00239-004-2600-4
Purdue University Selaginella Page
Edwards K Johnstone C Thompson C A simple and rapid method for the preparation of plant genomic DNA for PCR analysis Nucleic Acids Res 1991 19 1349 2030957
|
15938755
|
PMC1184070
|
CC BY
|
2021-01-04 16:32:49
|
no
|
BMC Genomics. 2005 Jun 6; 6:85
|
utf-8
|
BMC Genomics
| 2,005 |
10.1186/1471-2164-6-85
|
oa_comm
|
==== Front
BMC GenomicsBMC Genomics1471-2164BioMed Central London 1471-2164-6-921595524010.1186/1471-2164-6-92Research ArticleGenome-wide structural and evolutionary analysis of the P450 monooxygenase genes (P450ome) in the white rot fungus Phanerochaete chrysosporium : Evidence for gene duplications and extensive gene clustering Doddapaneni Harshavardhan [email protected] Ranajit [email protected] Jagjit S [email protected] Environmental Genetics and Molecular Toxicology Division, Department of Environmental Health, University of Cincinnati College of Medicine, Cincinnati, OH 45267-0056, USA2 The Center for Genome Information, Department of Environmental Health, University of Cincinnati College of Medicine, Cincinnati, OH 45267-0056, USA2005 14 6 2005 6 92 92 24 11 2004 14 6 2005 Copyright © 2005 Doddapaneni et al; licensee BioMed Central Ltd.2005Doddapaneni et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Phanerochaete chrysosporium, the model white rot basidiomycetous fungus, has the extraordinary ability to mineralize (to CO2) lignin and detoxify a variety of chemical pollutants. Its cytochrome P450 monooxygenases have recently been implied in several of these biotransformations. Our initial P450 cloning efforts in P. chrysosporium and its subsequent whole genome sequencing have revealed an extraordinary P450 repertoire ("P450ome") containing at least 150 P450 genes with yet unknown function. In order to understand the functional diversity and the evolutionary mechanisms and significance of these hemeproteins, here we report a genome-wide structural and evolutionary analysis of the P450ome of this fungus.
Results
Our analysis showed that P. chrysosporium P450ome could be classified into 12 families and 23 sub-families and is characterized by the presence of multigene families. A genome-level structural analysis revealed 16 organizationally homogeneous and heterogeneous clusters of tandem P450 genes. Analysis of our cloned cDNAs revealed structurally conserved characteristics (intron numbers and locations, and functional domains) among members of the two representative multigene P450 families CYP63 and CYP505 (P450foxy). Considering the unusually complex structural features of the P450 genes in this genome, including microexons (2–10 aa) and frequent small introns (45–55 bp), alternative splicing, as experimentally observed for CYP63, may be a more widespread event in the P450ome of this fungus. Clan-level phylogenetic comparison revealed that P. chrysosporium P450 families fall under 11 fungal clans and the majority of these multigene families appear to have evolved locally in this genome from their respective progenitor genes, as a result of extensive gene duplications and rearrangements.
Conclusion
P. chrysosporium P450ome, the largest known todate among fungi, is characterized by tandem gene clusters and multigene families. This enormous P450 gene diversity has evolved by extensive gene duplications and intragenomic recombinations of the progenitor genes presumably to meet the exceptionally high metabolic demand of this biodegradative group of basidiomycetous fungi in ecological niches. In this context, alternative splicing appears to further contribute to the evolution of functional diversity of the P450ome in this fungus. The evolved P450 diversity is consistent with the known vast biotransformation potential of P. chrysosporium. The presented analysis will help design future P450 functional studies to understand the underlying mechanisms of secondary metabolism and oxidative biotransformation pathways in this model white rot fungus.
==== Body
Background
The cytochrome P450 monooxygenases ("P450s") constitute a large superfamily of heme-thiolate proteins widely distributed in different life forms including prokaryotes (archaea, bacteria), lower eukaryotes (fungi, insects), and higher eukaryotes (plants and animals). P450s play an important role in the metabolism of a wide variety of endogenous and xenobiotic compounds. The current P450 nomenclature [1] is based on divergent evolution of the P450 superfamily. On the basis of sequence homology, all P450s can be categorized into two main classes, B ('bacterial') and E ('eukaryotic') [2]. Bacterial P450s with three component systems [an FAD-containing flavoprotein (NADPH or NADH-dependent reductase), an iron-sulphur protein, and the P450 hemeprotein] and the fungal P450nor (CYP55) [3] belong to the 'B'-class. All the other known P450s from distinct systems, including eukaryotic P450s and bacterial P450s such as P450BM3 (CYP102) from Bacillus megaterium [4], belong to the 'E'-class. The eukaryotic microsomal P450 system contains two components, the NADPH:P450 oxidoreductase (POR), a flavoprotein containing both FAD and FMN, and the P450 monooxygenase containing the heme domain. The prokaryotic (bacterial) soluble P450 monooxygenase P450BM3 (CYP102) exists as a single protein with both heme and flavin functional domains. Typically, the bacterial P450s are soluble and shorter (approximately 400 amino acids) in length, whereas the eukaryotic P450s average around 500–600 amino acids or larger and are membrane-bound. The amino acid (aa) sequence of these P450 monooxygenase proteins is extremely diverse, with levels of identity as low as 16% in some cases, but their structural fold has remained the same throughout evolution. The existing data suggest that divergence of the P450 superfamily into B and E classes, and further divergence into stable P450 groups within the E class, is very ancient and had occurred before the appearance of eukaryotes [5]. From the phylogenetic classification point of view, families have been identified based primarily on amino acid sequence similarity, with less than 40% similarity defining a family and 40–55% similarity defining a sub-family. Recently, the concept of "Clan" which represents higher order grouping of P450 families is gaining wider acceptance in the P450 community [6,7].
Among the different phyla, plants have the highest number of P450 sequences, followed by fungi. So far, more than 380 fungal P450s have been identified and the number is increasing with the sequencing of new fungal genomes . In comparison to the yeast forms (Saccharomyces and Candida), the higher order fungi have a larger number of P450 sequences in their genomes, with 41 P450 genes predicted in Neurospora crassa , 111 in Aspergillus spp ., 123 in Magnaporthe grisea , 110 in Fusarium graminearum and about 150 in Phanerochaete chrysosporium. Assuming the possibility that the present day P450s have evolved from a single ancestor P450 gene, the large disparity in the P450 gene count among the different phyla and genera is reflective of the differences in metabolic demand. According to Nelson [8], while plants have invested heavily on biochemistry for development of survival strategies thereby driving the P450s to expand rapidly, animals have invested on development of higher order sensory systems and hence carry comparatively fewer P450s. Fungi, which resemble plants in their sedentary habitat, appear to have driven P450s to diversify rapidly.
Phanerochaete chrysosporium , the model white rot fungus, has the extraordinary ability to degrade and mineralize (to CO2) lignin, the earth's most abundant aromatic polymer, and a wide range of toxic chemical pollutants. Lignin biodegradation occurs under nutrient-limited conditions when the fungus enters secondary metabolism. Only white rot fungi are known to possess the ability to completely degrade lignin [9],[10],[11]. The working hypothesis is that initial depolymerization of the lignin by extracellular peroxidases releases chemical compounds that are internalized and further degraded by diverse intracellular enzymes, including P450 monooxygenases. Further, several studies have shown that this fungus can degrade and mineralize a broad spectrum of aromatic, alicyclic, and aliphatic chemical pollutants under both nutrient-limited (ligninolytic) and nutrient-sufficient (non-ligninolytic) conditions. Lately, P450 monooxygenases have been shown to be involved in several of these biotransformations and this has led to an increasing interest in the characterization of cytochrome P450 systems in this model white rot fungus.
Initial cloning efforts from this laboratory identified the first complete P450 system genes in P. chrysosporium. These efforts led to the isolation [12] and functional characterization [14,15,17] of three full-length P450 genes pc-1, pc-2 and pc-3 that were assigned to CYP63 family, and the P450 oxidoreductase gene (POR ) that is responsible for electron transfer to the multiple P450 monooxygenases [16]. Subsequent whole genome sequencing by the Joint Genome Institute (JGI) of the US Department of Energy (US-DOE) has led to the identification of a large P450 contingent (P450ome) in P. chrysosporium [13]. The P450 genes constitute nearly 1% of the coding sequences in the genome of this basidiomycetous fungus [13]. Of the nearly 150 P450 monooxygenase genes identified in P. chrysosporium genome, 108 have been assembled full-length and tentatively annotated based on general overall sequence homology analysis [13]. To date, this is the highest number of P450s identified in any fungus, which appears to be one of the major underlying factors for its extraordinary catalytic potential. Understanding the structural and evolutionary aspects of such large family of P450s in conjunction with their physiological characteristics will help understand the functional versatility of this fungus. This study reports a detailed genome-wide structural and phylogenetic analysis of the P450ome of P. chrysosporium , coupled with cloning and characterization of new cDNAs for selected multigene P450 families, with a broader aim to facilitate the classification/nomenclature and to understand the functional diversity and evolution of the P450s in this wood-rotting model white rot fungus.
Results and Discussion
I. Structural analysis and characterization of P. chrysosporium P450ome
P450ome of P. chrysosporium
Our phylogenetic analysis coupled with the standard sequence homology criterion for P450 nomenclature revealed that P. chrysosporium P450s fall into 12 families and 23 sub-families (Figure 1). This family and sub-family classification is based on the amino acid sequence similarity using the existing criteria of less than 40% similarity defining a family and less than 55% similarity defining a sub-family as followed by the International P450 Nomenclature Committee. Earlier we had reported an initial phylogenetic grouping of 163 predicted P450 sequences (including even partial fragments) into 26 clusters based on overall general homology criterion [15]. Here we report the phylogenetic grouping of 126 P450 genes in 12 families and 23 sub-families using the P450 family/sub-family homology-based criterion. To eliminate the possibility of bias in grouping, only full-length or near full-length P450 sequences with more than 300 aa residues were used for constructing the tree in this study. Of the 12 families, the International P450 Nomenclature Committee so far has named only three families including the highly conserved families CYP51 and CYP61, and the newly designated P. chrysosporium family CYP63. Nevertheless, based on the < 40% homology criterion, nine other families, each with varying degree of similarity to the existing one or more P450 families, were identifiable and were arbitrarily designated as follows: CYP58/53, CYP62, CYP64, CYP67, CYP503, CYP505, CYP5031/547, CYP614/534, CYP617/CYP547. As compared to the recently reported analysis [13], we have added 18 P450 genes grouped as three new clusters in the tree. Individual genes from these clusters, when compared to the P450 sequences available at the P450 BLAST server , showed highest similarity to the P450 families CYP617/547, CYP614/534 and CYP5031/547, respectively. These three sets of CYP names thus represent the families newly added to our phylogenetic tree (Figure 1).
Figure 1 P450ome of the white rot basidiomycete Phanerochaete chrysosporium . The UPGMA tree is based on 126 full-length or near full-length P450s of P. chrysosporium P450ome and uses the three human CYP1 proteins (1A1, 1A2, and 1B1) for comparison. There are 12 families demarcated using the solid vertical line and 23 sub-families differentiated using the dashed vertical line.
Among the 12 P450 families, the CYP64 family has the highest number (more than fifty) of member genes, followed by CYP67 (sixteen), CYP503 (fourteen), CYP58/53 (ten), CYP63 (seven), CYP505 (seven), CYP614/534 (seven), CYP617/547 (six), CYP5031/CYP547 (two), whereas P450 families CYP51, CYP61 and CYP62 consist of only one member gene each.
The International Nomenclature Committee on cytochrome P450s has been annotating the new P450 sequences on the basis of the existing primary amino acid sequence similarity criteria. However, there have been instances where the 40% similarity rule in defining a family has been relaxed, such as in naming the mammalian CYP4 family [7]. Simultaneously, the usage of CYP "Clans" which are higher-order clusters of related families is gaining consensus [6,7]. In view of this, a clan-based classification of the P450 families in P. chrysosporium is presented and discussed in subsequent sections, with an aim to understand their evolution and functional significance.
Cloning of new P450 cDNAs
Our initial studies on cloning of cDNAs for the first cloned P450 gene pc-1 [12] and the related gene pc-3 [17], both belonging to CYP63 family, indicated a complex structural organization of P450 genes in P. chrysosporium. In an effort to further understand the structural organization of P450 genes in this system, the following additional cDNA sequences for the other CYP63 genes were isolated in the present study: pc-2 (1842 bp), pc-4 (430 bp), pc-5 (324 bp) and pc-6 (330 bp). The isolated cDNA sequences (full-length and partial) were then compared with the corresponding predicted coding sequences in the genome. This gene/cDNA analysis helped understand the gene features of the first multigene P450 family (CYP63) identified in this organism [12]. The analysis also helped identify an additional gene, pc-7, groupable under this family. Furthermore, we isolated a cDNA sequence representing the C-terminus half (the reductase domain) of one of the fused P450foxy-like genes, pc-foxy1. The cloned cDNA (681 bp), that spans 747 bp of the corresponding genomic region, helped identify the structural features of the P450foxy gene family (CYP505) of the P450ome.
Gene features (introns, exons and deduced coding regions) of the P. chrysosporium P450s
i). CYP63 family
The cloned CYP63 cDNA sequences revealed that the member genes carry 6–14 introns with conserved GT/AG splice junctions and encode 571 to 603 aa proteins. The deduced proteins contain transmembrane domains in their N-terminus region centered around 21–57 aa residues with a matrix score of more than 1000 as determined by TMpred analysis, indicating that these proteins are membrane bound. The seven CYP63 members (pc-1 through pc-7) are groupable into three sub-families, CYP63A (pc-1 through pc-4), CYP63B (pc-5 and pc-6), and CYP63C (pc-7). The three CYP63A member genes pc-1 (CY63A1), pc-2 (CY63A2) and pc-3 (CY63A3) are tandemly linked in that order with 322 bp and 469 bp intergenic regions. Their intron organization (number and position) and the number of amino acids encoded are nearly the same (Figure 2), suggesting that the three genes are paralogous genes and originated by tandem duplication. The typical P450 motifs including heme-binding region (HR2), helix-I, and helix-K showed high sequence conservation among the seven CYP63 member proteins (Figure 3).
Figure 2 Gene organization of the seven members of CYP63 family of P450s in the white rot fungus P. chrysosporium. Horizontal lines represent the predicted coding region (exons) and the vertical lines indicate intron locations. Full-length cDNAs isolated for pc-1, pc-2 and pc-3, and partial cDNAs isolated for pc-4 (430 bp), pc-5 (324 bp) and pc-6 (330 bp), are shown by the solid horizontal lines.
Figure 3 Alignment of amino acid sequences of the CYP63 family P450s in P. chrysosporium. The seven member proteins PC-1 through PC-7 were aligned using the GeneDoc program version 2.6.002 in conserved format. Relatively conserved heme-binding (HR2) region, helix-I and helix-K for the seven sequences are underlined with square-dotted line, solid line and round-dotted line, respectively. The alignment shading is based on the physicochemical properties group analysis, wherein each column in the alignment is assigned to one of the twelve predefined groups of physicochemical properties. These twelve groups represent three overlapping hierarchies of size, the electrical charge for the polar amino acids, and the aromaticity for non-polar amino acids.
ii). P450foxy gene family (CYP505)
The P. chrysosporium whole genome sequence revealed seven fused oxidoreductase-P450 proteins ("P450foxy" proteins), of which 5 are located on a 43 kb stretch of the genome scaffold 73. The process of deducing the amino acid sequences for these genes was guided in part by the pc-foxy1 cDNA cloned in this study, in conjunction with the gene characteristics of the known P450foxy sequence of Fusarium oxysporum [18], and the earlier reported P450 splicing pattern in P. chrysosporium [12]. Analysis of the P450foxy genes revealed 17 to 22 introns, of which 11 are conserved in all the seven member genes (Figure 4). The cloned pc-foxy1 cDNA fragment (681 bp) contained splicing site for a single Type-II intron (66 bp) with defined GT/AG splice junction. The reductase portion of the cloned pc-foxy1 cDNA contained two FAD and three NADPH functional domains and showed deduced aa homology to the distal region of the P450 oxidoreductase (POR) protein of this organism. The full-length (736 aa) functional POR earlier reported for P. chrysosporium has multiple domains, including 3 FMN, 3 FAD and 3 NADPH domains [16]. However, except for the proximal FAD domain, the degree of motif conservation was poor between the two reductase sequences (pc-foxy1 reductase component and the POR), suggesting that the reductase component of the P450foxy fusion genes has an independent descent and is not duplicated from the native POR gene.
Figure 4 Gene organization of the seven fused oxidoreductase-P450 (P450foxy-like) proteins. Horizontal bars represent the predicted coding regions (exons) and the vertical lines indicate intron locations. The unfilled part of the bar represents the cloned cDNA portion of the gene ug.73.17.1, whereas the solid bars indicate the predicted cDNAs.
The five P450foxy member genes on genome scaffold 73 show divergent transcription, with the three member genes (ug.73.17.1, pc.73.4.1 and pc.73.14.1) transcribing from the same strand, whereas the other two genes (ug.73.15.1 and ug.73.16.1) transcribing from the opposite strand of DNA. The deduced P450foxy proteins containing 624–1111 amino acid residues show high aa similarity (> 55%) among them and hence can be classified under the same sub-family. The general domain architecture of these proteins is the same as that of the earlier reported P450foxy protein from F. oxysporum [18], with an N-terminal P450 segment and a C-terminal reductase segment. All seven proteins have a full-length P450 segment with conserved P450 signature sequences. However, the reductase segment is full-length in only 6 of the seven members and is truncated in the member gene gx.187.5.1. These six member proteins contain a transmembrane domain, centered around 908–920 aa positions with a matrix score of more than 1000, as detected by TMpred analysis. The truncated member contains a transmembrane domain centered on amino acid residue 570. The above analysis points to the membrane-bound nature of these fused proteins similar to the homologous F. oxysporum protein and consistent with the membrane-bound nature of most of the eukaryotic P450 and POR proteins.
It is unusual that P. chrysosporium has only one P450 oxidoreductase (POR) gene to cater to such a large contingent of P450 monooxygenases for the electron supply, with the exception of P450foxy proteins (see below). This suggests a more significant role of alternate electron transfer proteins such as cytochromes b5 and b5 reductase in this organism; it will, however, require further functional characterization to prove this assumption. The P450foxy proteins appear to be self sufficient in their electron transfer mechanism considering the presence of a complete reductase partner (with all needed functional domains) as a part of the fusion protein. Such electron transfer function of the reductase component was experimentally demonstrated in the first known P450foxy fusion protein of the ascomycete fungus Fusarium oxysporum [18].
iii). Gene features of other P450ome genes
Among the other P450 families in the P450ome, the CYP64 family has the most genes (54), with member genes carrying 5–12 introns of size 21–268 bp, and encoding 350–534 aa size proteins. Detailed gene features of the other individual P450 families of the P. chrysosporium P450ome are listed in Table 1.
Typically, all the full-length assembled P450 genes were found to carry multiple exons frequently interrupted by small introns. In this analysis, we observed a relatively narrow size range for introns (21 to 620 bp) as compared to the exons; the shortest exon is 6 bp (2 aa) and the longest is 1048 bp (Table 1). Nearly half of the P450 genes show small exons (upto 67 bp), with 41 genes carrying 29 bp or shorter exons (Table 1). Interestingly, the P450ome genes contain microexons (encoding 2–10 aa) that are spread genome-wide and are not restricted to any particular P450 family. Such ubiquitous occurrence of the microexons in P. chrysosporium P450 genes points to their potential role in conferring functional diversity and emphasizes the fact that the functional divergence is not merely a product of gene duplication/translocation. In comparison to the P450 genes, the intron size range in the most intensively characterized peroxidase gene family of this organism is 49–78 bp, with 49 to 78 bp for the lignin peroxidase (LiP) sub-family and 50 to 72 bp for the manganese-dependent peroxidase (MnP) sub-family. Furthermore, as revealed by the whole genome sequencing, the presence of relatively small introns (average 117 bp) appears typical of this genome as compared to the higher eukaryotic genomes [13].
P450 gene clustering and tandem genes
The P. chrysosporium genome shows a total of 16 P450 gene clusters, with up to 11 member genes in a cluster. Of the 16 clusters, three clusters are located on scaffold 30, two each on scaffolds 20, 24, and 73, and one each on scaffolds 1, 16, 50, 53 59, 79 and 97. For convenience, the clusters have been assigned numbers 1 through 16 in the order they appear on the individual genomic scaffolds. Among these, cluster-5 has the highest number of tandem genes (eleven) followed by cluster-3 (five) and cluster-6 (four) (Table 2). The earlier cloned tandem CYP63 genes pc-1, pc-2 and pc-3 represent cluster-4.
There is a high sequence similarity (up to 84%) among the members of a given cluster and the cluster members fall under the same P450 family on the phylogenetic tree (Figure 1; Table 2). The number of introns and exons and their relative positions in the member genes of a cluster are conserved in 10 of the 16 clusters (Table 2). Two distinct patterns exist for the intron organization (number and position) within each cluster. In the case of small gene clusters (2–3 member genes), the number and relative position of introns as well as the size of the exons are conserved among the members, whereas the larger gene clusters (> 3 member genes) show varying degrees of dissimilarity in these characteristics. While the conserved gene characteristics suggest more recent duplications, the dissimilarity suggests either more distant duplication events or translocation events. Therefore, the P. chrysosporium genome shows both homogeneous and heterogeneous clusters of tandem genes. Tandem duplicates (paralogous genes), which initially are structurally and functionally identical, diverge with time due to mutations or translocations. It is plausible that one copy retains the original function, while the second copy either acquires a new function that is selected to meet the increased metabolic demand or gets deleted from the genome. In this context, it is noteworthy that member genes of cluster-4 (pc-1, pc-2 and pc-3), though closely spaced and structurally highly homogeneous, showed non-coordinate regulation of transcription and varying substrate inducibility [14,17]. While such substrate diversity among the members of tandemly linked P450 genes has been observed in the yeast Candida maltosa [19], this observation is new in the context of filamentous fungi.
Analysis of spatial organization of the P450 gene clusters in the P. chrysosporium genome revealed a variable pattern. For instance, the two P450 clusters on scaffold-20 are spread over a 220 kb genomic region separated by more than 200 kb, whereas the gene clusters on other scaffolds are more closely spaced. On scaffold-24, the clusters 5 and 6 are spread over a 62 kb of genomic region and are separated by less than 10 kb genomic region. Similarly, clusters 7 to 9 on scaffold-30 are spread over an 80 kb region with a 27–28 kb gap between them. The two clusters on scaffold-73 are separated by less than 10 kb DNA and are spread over a 30 kb genomic region. This analysis constitutes the first report on P450 clustering and spatial organization in filamentous fungi. Nevertheless, in fungi, it has been frequently observed that the genes coding for enzymes involved in secondary metabolism, such as those involved in the synthesis of ergot alkaloids [20], HC-toxin [21], and mycotoxins such as sterigmatocystin and aflatoxins [22,23], are heterologous clusters (containing P450 and other metabolic genes). Clustering of secondary metabolic genes has been proposed to favor their survival and dispersal, at least in part, via horizontal gene transfer [23,24]. The close association seen in some of the P. chrysosporium gene clusters might point to their co-ordinated regulation as observed in the case of the above secondary metabolic gene clusters in different fungi. However, experimental evidence is needed to extrapolate this assumption to the fungal P450 clusters. Our initial transcription-based analysis demonstrated lack of such co-ordinated regulation among the tandem CYP63 genes of cluster-4 in P. chrysosporium [14].
Alternative splicing and functional diversity
Alternative splicing is an important mechanism for regulation of gene expression, which expands the coding capacity of a single gene to allow production of different protein isoforms, often with diverse functions [25]. More than 50% of human genes are alternatively spliced [26]. Reports in fungi on alternative splicing are few [12,27], in comparison to humans where this mechanism has been well documented. We have experimentally identified two splice variants of the first characterized P450 gene pc-1 (CYP63A1) from P. chrysosporium [12] and predicted more such variants based on the in silico analysis. Although splice variants have not been identified so far in the other two tandemly arranged members pc-2 and pc-3 of this cluster, their similar gene organization (typically marked by multiple introns and exons as small as 4 to 10 aa length) suggests the existence of splice variants. Further, while validating our custom 70-mer microarray analysis data [15] on P450 gene transcription and induction using RT-PCR (wherein gene specific primers were chosen from different locations on the gene), we observed that the transcript quantification for a given sample in some cases varied with the location of the primer chosen (Unpublished data). This observed variation points to the existence of alternative splice variants. A recently proposed system of nomenclature for such splice variants suggests that the transcript name should include the exon involved in the splicing event [7]. It is noteworthy that nearly one-third of the P. chrysosporium P450ome shows microexons, which are likely candidates that promote alternative splicing events during transcription. Such alternative splicing helps increase the diversity of the transcriptome, and is likely to significantly contribute to the metabolic diversity of this organism.
II. Evolutionary analysis of P. chrysosporium P450ome
Fungal P450 clans in P. chrysosporium
Finding meaningful associations and evolutionary relationships among members of the rapidly expanding superfamily of the P450 proteins at the species level is becoming a challenge using the existing family level classification. For instance, the CYP6 family that is exclusive to insects forms a close cluster with the CYP3 and CYP5 families from mammals, CYP30 from clams, and CYP25 from C. elegans on the phylogenetic tree, indicating that these five families probably have evolved from a common ancestor with similar function before the deuterostome-protostome split [8]. However, this is not reflected in the family names, as the family classification is based on an arbitrary 40% amino acid similarity criterion. To explain such higher order groupings, the term "Clan" was recently introduced [6]. Typically, a clan represents a cluster of P450 families across species, grouped based on relationships that are beyond the family designations. There are 9 clans in vertebrates and 10 in plants.
A detailed phylogenetic analysis was carried out to understand the evolutionary relationship of P. chrysosporium P450 gene families and their relatedness with other fungal clans. In lower forms of fungi (yeasts), 4 P450 families (CYP51, 52, 53 and 61) have been characterized, of which CYP51 and 61 are conserved. In higher forms of fungi (filamentous fungi), 13 P450 families (CYP51, 53, 61, 65, 68, 505, 531, 532, 537, 539, 540, 548, and 552) are common as exemplified by the hitherto sequenced genomes of four ascomycetous fungal species: Neurospora crassa , Magnoporthe grisea , Fusarium graminearum and Aspergillus nidulans . Interestingly, based on homology analysis, the P450 families CYP51, 53, 61, and 505 are also present in P. chrysosporium (a basidiomycetous fungus) albeit with a widely varying degree of similarity. Further, clan level comparisons revealed that 12 P. chrysosporium P450 families (Figure 1) have resemblances in 11 fungal P450 clans and show varying degrees of structural similarities to the P450 genes from different ascomycetous fungi such as Aspergillus and Fusarium , suggesting that P. chrysosporium has acquired these P450 families as a part of vertical descent from a common ancestor followed by further diversification. It will be interesting to compare the hitherto uncharacterized zygomycetous P450ome to arrive at the ancestral P450ome that led to the current P450 diversity among these three major fungal groups (ascomycetes, zygomycetes and basidiomycetes).
CYP51 clan
One of the highly conserved and functionally well-characterized P450 gene families in fungi is the CYP51 family that encodes the cell wall ergosterol biosynthesis reaction, lanosterol 14-alpha demethylation in yeasts or eburicol 14-alpha demethylation in filamentous fungi, a target step for antifungal azole compounds. The current state of knowledge on P450 evolution in eukaryotes points to CYP51 as the ancestral P450, which is believed to have led to the evolution of all the present day P450 families [8]. In comparison to some of the ascomycetous fungi, which carry multiple CYP51 genes, there is a single CYP51 gene (located on scaffold-2) in P. chrysosproium that codes for a 549 aa protein. The CYP51 minimal evolution tree, based on the available CYP51s from 17 fungal species (Figure 5), depicts that basidiomycete CYP51s (including P. chrysosporium CYP51) form the closest association with CYP51s from the ascomycetous fungi as compared to those from the zygomycetous fungus Cunninghamella elegans and the hemi-ascomycetous fungus (budding yeast). Considering the fact that all known fungi except ascomycetes have a single CYP51 gene in their genome (Figure 5), it appears that the ascomycetous fungi required diversification of their CYP51 gene. This intragenome diversification of CYP51 in ascomycetes, such as in the genomes of A. nidulans , M. grisea , and F. graminearum , is reflected by high bootstrap values separating their multiple CYP51s. Likewise, analysis of the recently completed plant genomes has shown that the rice genome contains ten functional CYP51 genes and two pseudogenes [28], and that Arabidopsis has two CYP51 genes, in contrast to the popular belief that a single CYP51 gene exists in all phyla. The presence of multiple CYP51s might indicate their involvement beyond the sterol biosynthesis to take up new function(s) for the organism's survival. Since the ascomycete species that carry multiple CYP51s are parasitic in nature, multiple CYP51s/variants might have been acquired to give these species an edge in host-pathogen interaction or survival against antifungal treatment [29], unlike in saprophytic fungi such as P. chrysosporium.
Figure 5 Minimal evolution tree of the fungal CYP51 clan. The fungal species included are Aspergillus nidulans; Blumeria graminis, Botryotinia fuckeliana, Candida albicans, Cryptococcus neoformans, Cunninghamella elegans, Fusarium graminicola, Meloidogyne graminicola, Magnaporthe grisea, Neurospora crassa, Penicillium italicum, Phanerochaete chrysosporium, Saccharomyces cerevisiae, Schizosaccharomyces pombe, Tapesia yallundae, Uncinula necator, and Ustilago maydis.
CYP 61 clan
CYP61 codes for sterol 22 desaturase in Saccharomyces cerevisiae , which is involved in later stages of the ergosterol pathway in metabolizing Ergosta-5,7,24(28)-trienol to Ergosta-5,7,22,24(28)-tetraenol by introducing a C-22(23) double bond in the sterol side chain [30]. Since this gene acts late in the ergosterol pathway, it is considered to have evolved as a result of duplication and diversification of the CYP51 gene [8]. The P. chrysosporium genome contains one CYP61 gene. However, unlike its CYP51 (Figure 5), CYP61 of P. chrysosporium is phylogenetically as distant from euascomycetous CYP61s as it is from the hemi-ascomycetous CYP51 (Figure 6). These analyses also show diversification of CYP61 within the ascomycetous group at least in 2 subclusters (with high bootstrap values) that coincide with small genomes (yeast-like) versus larger genomes (filamentous fungi).
Figure 6 Minimal evolution tree of the fungal CYP61 clan. The fungal species included are Aspergillus nidulans; Candida albicans, Eremothecium gossypii, Fusarium graminearum, Magnaporthe grisea, Neurospora crassa, Phanerochaete chrysosporium, Saccharomyces cerevisiae, Schizosaccharomyces pombe, Symbiotaphrina buchneri, and Symbiotaphrina kochii .
CYP52 Clan
The CYP52 family of P450 proteins was originally identified in Candida species [31,32] with a role in carrying out terminal hydroxylation of n-alkanes and ω-hydroxylation of fatty acids. The varying substrate specificities of these genes have given this yeast its ability to modify a range of n-alkanes. Later, homologous genes have been identified in other yeasts [33,34] and filamentous fungal species (Figure 7). The earlier discussed CYP63 family of P. chrysosporium, which consists of seven members, pc-1 through pc-7, shows close architectural resemblance to the CYP52 family of yeasts and thus can be classified under the same clan. However, as evident from the minimal evolution tree (Figure 7), the white rot CYP63 members form a separate cluster on the tree from the CYP52 family members of the ascomycetous fungi and other CYP52-related fungal P450 families. Furthermore, despite high structural similarity among them, the individual member genes of the CYP63 cluster are separated by high bootstrap values, indicating significant divergence among them. This is corroborated by our experimental studies on substrate inducibility of the three tandemly linked CYP63 members pc-1, pc-2, and pc-3; the expression of these linked P450s was differentially induced (both in qualitative and quantitative terms) in response to linear alkanes and substituted alkanes. Moreover, their substrate-specificity seems to extend beyond alkanes/substituted alkanes as they also showed induction in the presence of various aromatics, including the polycyclic aromatic compounds such as PAHs [14,15]. These and other experimentally generated data on their differential physiological regulation [14] coupled with the presented phylogenetic analysis point to the evolution of the CYP63 family to acquire diverse functional roles while retaining its original alkane-degrading ability. While the n-alkanes are hydroxylated for assimilation (as carbon source) by Candida species, the P450 hydroxylation of alkanes by P. chrysosporium (which does not use them as sole carbon source) appears to be a co-metabolic activity required for catalyzing aliphatic oxidations during the natural lignin degradation process. This is further supported by the fact that there are structural resemblances between the inducer compounds (alkanes and aromatics) and the substructures present in its natural substrate lignin.
Figure 7 Minimal evolution tree of the fungal CYP52 clan. Abbreviations: An-Aspergillus nidulans ; Ca-Candida apicola ; Cm-Candida maltosa ; Ct-Candida tropicalis ; Fg-Fusarium graminearum ; Mg-Magnaporthe grisea ; Nc-Neurospora crassa ;Pc-Phanerochaete chrysosporium ; Rsp-Rhodotorula sp; Yl -Yarrowia lipolytica .
CYP64 clan
The CYP64 family, first identified in Aspergillus spp., was shown to include P450 proteins that catalyze various reactions involved in biosynthesis of aflatoxins and other such secondary metabolites in these species [35]. Genome-wide sequence similarity analysis and annotation of the P. chrysosporium P450ome revealed that 54 P450 genes fall under the CYP64 clan (Figure 8). The P. chrysosporium CYP64 family members are interspersed on the tree (as indicated by low bootstrap values on the critical branch nodes linking P. chrysosporium with other fungi) constructed based on the CYP64 clan genes from other fungal groups suggesting their multiple lineage or diversification. White rot fungi such as P. chrysosporium , which have a typical secondary metabolic switch in their growth/ biodegradation cycle, may have the hitherto uncharacterized role for its CYP64 clan proteins in secondary metabolism. Functional evaluation of one or more members from this group will bring into light the enzymatic or functional diversity of such a large number of CYP64-like genes in this fungus.
Figure 8 Minimal evolution tree of the fungal CYP64 clan. The unrooted phylogenetic tree is based on 69 P450 sequences (54 genes from P. chrysosporium and 15 genes from six other fungal species). Abbreviations: Af-Aspergillus flavus ;An-Aspergillus nidulans ; Ap-Aspergillus parasiticus ; Cc-Coprinus cinereus ;Fg-Fusarium graminearum ; Le-Lentinula edodes ; Mg-Magnaporthe grisea ; Pc-Phanerochaete chrysosporium; Um-Ustilago maydis.
CYP505 clan
The fatty acid hydroxylase P450BM3 (CYP102) of the bacterium Bacillus megaterium , containing a P450 monooxygenase gene fused with a P450 reductase gene, was the first identified fusion protein member of the P450 superfamily [4]. Later, a similar fused P450 gene coding for fatty acid hydroxylase was identified from the fungus Fusarium oxysporum [18]. The latter, named P450foxy, shows 40.6% and 35.3% amino acid similarity in its P450 and reductase domains to the corresponding domains in the bacterial P450 fusion protein P450BM3. The current hypothesis suggests that such fused proteins are of eukaryotic origin and their occurrence in the prokaryotic (bacterial) cells is due to horizontal gene transfer [18].
Interestingly, there are 7-fused P450foxy-like genes in the P. chrysosporium genome located on 3 different scaffolds. These seven P450s form a separate cluster from the known P450foxy proteins of other fungi (ascomycetes) on the minimal evolution tree (Figure 9). There are two distinct clusters among the ascomycetous fungi (Figure 9), which appear to be lineages of two ancestral fused proteins based on their phylogenetic distance; of these, one clusters more closely with the P. chrysosporium fusion proteins. The white rot fusion proteins form three sub-groups within their phylogenetic branch separated by reasonably high bootstrap values (89% separating subgroup 1 from subgroups 2 and 3, and 63% separating subgroup 2 from subgroup 3), indicating considerable divergence among these proteins. Nevertheless, these fused proteins, while forming three sub-groups within their phylogenetic branch, show high conservation of functional domain sequences and intron/exon organization, suggesting the involvement of gene duplication and translocation events in their formation. There are multiple lines of evidence to support this argument. First, as discussed earlier, five out of the seven fused proteins are placed within a distance of 43 kb region on scaffold 73. Second, the conserved intron positioning in all agrees perfectly with their common grouping on the tree (Figure 9). Third, multiple regions of high sequence similarity in the flanking regions around these genes have been identified using the BLAST program. For instance, when the 43 kb region of the scaffold 73 was blasted against itself, a near perfect inverse duplication region of 2 kb was identified; the region from base number 32876 to 34876 is identical to the regions spanning base numbers 27 to 801 and 2134 to 3348 on scaffold 73. This explains the awkward position of the pc.73.11.1 gene between the closely related genes ug.73.15.1 and pc.73.14.1 on this scaffold and indicates a duplicative inversion event (Figure 10). Similarly, there is high sequence similarity (E-value = 0) between the 1000 bp non-coding flanking region downstream of the gx.187.5.1 gene, spanning base numbers 24674 to 25673 on scaffold 187, and the 983 bp non-coding region upstream of gene pc.73.1.1 spanning base numbers 36423 to 37405 on scaffold 73. This points to an ectopic insertion of the gx.187.5.1 on scaffold 187 after duplication on scaffold 73. This finding of an extraordinary level of sequence identity for non-coding DNA regions provides further evidence to the proposed involvement of gene duplication and translocation events.
Figure 9 Minimal evolution tree of the fungal CYP505 clan. The unrooted phylogentic tree was constructed using only the P450 portion of the fused protein (P450foxy). Abbreviations: An-Aspergillus nidulans ; Ao-Aspergillus oryzae ; Fg-Fusarium graminearum ; Fo-Fusarium oxysporum ; Gm-Gibberella moniliformis ; Gz-Gibberella zeae ; Mg-Magnaporthe grisea ; Nc-Neurospora crassa ; Pc-Phanerochaete chrysosporium .
Figure 10 Proposed microevolution of six of the P450foxy proteins in P. chrysosporium by gene duplication and translocation events. Position and orientation of the genes are indicated with thick arrows above the horizontal scaffold bar. One or more initial tandem duplications were followed by translocation and ectopic integration (as indicated by thin arrows) of the fused genes leading to the present day organization. The regions of high sequence similarity on the scaffolds are shown in the same hatching.
Many questions regarding the origin and distribution of these fused P450 proteins in fungi remain unanswered. One can conclude from the phylogenetic tree (Figure 9) that more than one original fusion event happened in the ancestral fungus either before the ascomycetous-basidiomycetous split approximately 400 million years ago [36], or a second fusion event happened in ascomycetous fungi immediately after these two groups split. Either way, during the course of evolution, these proteins have diversified further, possibly due to more gene fusions or due to gene duplications (Figure 10). Assuming that the ancestral fungus carried two fusion proteins, the question as to what happened to the other gene lineage in basidiomycetous fungi arises; has it been lost immediately after these two groups split without further diversification or was it never there and the second lineage in the ascomycetous fungi originated after these groups had separated? However, there is one complication to this argument; if P450foxy proteins predate the ascomycetous-basidiomycetous split, then why are these proteins missing entirely in archiascomycotina and hemiascomycotina (fission and budding yeasts)? This is an open question and may be answered by analyzing the genomes of vesicular-arbuscular mycorrhizas (VAM) or chitrids, which predate these two groups' split [36]. The third possible reason for occurrence of these fused proteins in P. chrysosporium could be horizontal gene transfer from one or more of the ascomycetous fungi. When the P. chrysosporium P450foxy proteins were compared to the recently completed whole genome shotgun sequences of other basidiomycetes Coprinus cinereus , Cryptococcus neoformans , and Ustilago maydis , no P450 fused protein homologues were found indicating the unique presence of these proteins in P. chrysosporium, a member of the wood-rooting group of basidiomycetes. However, as more fungal genomes become available, especially from the basidiomycetous group, it will be clear if these fusion proteins are actually unique in P. chrysosporium (or wood-rotting basidiomycetes subgroup) among basidiomycetous fungi. Nevertheless, it appears that fused proteins are predominantly present across ascomycetous fungi, and their presumed exceptional occurrence in the basidiomycete P. chrysosporium could possibly be a result of a horizontal gene transfer, an event otherwise rare in fungal genome evolution [37,23,24]. Looking at the gene organization and flanking sequence homology of these P450 genes in P. chrysosporium in Figures 9 and 10, and their distribution on the genome, it appears that six of the seven genes have branched out from a single progenitor gene, while the origin of the seventh gene pc.17.40.1 remains unclear and could possibly be a result of independent transfer. Role of intragenomic duplication event in the origin of pc.17.40.1 is ruled out based on the fact that there is no flanking sequence similarity between this and the other six genes. Furthermore, P. chrysosporium has the largest contingent (7 member genes) of P450foxy proteins (CYP505) among the fungi (3–5 member genes) containing this family of proteins. These observations collectively point to a rapid evolution of this P450 fusion proteins family (CYP505) in P. chrysosporium, possibly to meet the metabolic demand for fatty acid hydroxylation in the ecological niches of this fungus.
Clan-level relationships of other P. chrysosporium P450 families
The P. chrysosporium genome contains one member of the benzoate 4-hydroxylase family (CYP53) and nine members homologous to CYP58, both clustering under a common family on the P. chrysosporium tree (Figure 1) and groupable into the CYP53 clan . Presence of a single CYP53 protein in P. chrysosporium contrasts with the multiple CYP53 proteins detected in ascomycetous fungi. A total of 170 sequences (including CYP53 and CYP58) that are groupable into 4 sub-classes (A, B, C and D) have been assigned to the CYP53 clan in different ascomycetous fungi . The P. chrysosporium CYP53 protein is groupable under the B sub-class. The CYP58 family proteins have been shown to be involved in the synthesis of a group of secondary metabolites (trichothecene mycotoxins) in Fusarium spp. [38]. These proteins are grouped under the CYP53 clan sub-class C. However, the P. chrysosporium CYP58 homologues form a cluster separate from other fungal CYP58 proteins on the minimal evolution tree (Figure 11). Hence, we cannot rule out the possibility that these CYP58 structural homologues have evolved as functional variants, acquiring both CYP53 and CYP58 activities.
Figure 11 Minimal evolution tree of the fungal CYP53 clan. Nine P. chrysosporium CYP58 homologues were compared with the known CYP58 sequences from other fungal species. Abbreviations: An-Aspergillus nidulans ; Fcul-Fusarium culmorum ; Fg-Fusarium graminearum ; Fpseudog-Fusarium pseudograminearum ; Fs-Fusarium sporotrichioides ; Mr-Myrothecium roridum ; Pc-Phanerochaete chrysosporium .
Phylogenetic analysis based on the fungal CYP503 clan revealed that 12 of the P. chrysosporium P450 genes show close relatedness to the CYP512A1 gene from another lignin-degrading basidiomycete, Coriolus versicolor , but are distant from the CYP54 family and other CYP503 clan proteins from ascomycetous fungi, as shown by a high bootstrap value of 96% (Figure 12). The CYP503 family of P450 proteins originally found in Gibberella fujikuroi encode for multifunctional ent -kaurene oxidases that catalyze oxidation steps in the gibberellin biosynthesis of the plant growth hormone gibberellin, a secondary metabolite in this fungus. This is suggestive of a possible role of P. chrysosporium CYP503 clan genes in secondary metabolism.
Figure 12 Minimal evolution tree of the fungal CYP503 clan. Fourteen P. chrysosporium P450 genes were compared with the known homologous P450 proteins from other fungi. Abbreviations: An-Aspergillus nidulans ; Cv-Coriolus versicolor ; Fg-Fusarium graminearum ; Fp-Fusarium proliferatum ; Fs-Fusarium sporotrichioides ; Gf-Gibberella fujikuroi ; Mg-Magnaporthe grisea ; Nc-Neurospora crassa ; Pc-Phanerochaete chrysosporium; Pb-Phoma betae .
Sixteen P. chrysosporium P450 genes relate to the CYP67 clan (Figure 13). The CYP67 family was originally constituted of plant-induced rust genes identified in the basidiomycetous fungus Uromyces fabae [39]. The diterpene aphidicolin synthesizing gene PbP450-1 from Phoma betae and the genes involved in sterigmatocystin biosynthesis in Emericella nidulans are also groupable in the same clan in addition to some P450s encoding secondary metabolic reactions in other ascomycetous fungi (Figure 13). This suggests that the sixteen P. chrysosporium homologues might as well be involved in similar reaction steps in the synthesis of its secondary metabolites.
Figure 13 Minimal evolution tree of the fungal CYP67 clan. P450 genes from P. chrysosporium were compared with the known homologous P450 proteins from other fungal species. Abbreviations: An-Aspergillus nidulans ; Ao-Aspergillus ochraceoroseus ; En-Emericella nidulans ; Fg-Fusarium graminearum ; Fs-Fusarium sporotrichioides ; Mg-Magnaporthe grisea ; Nc-Neurospora crassa ; Pc-Phanerochaete chrysosporium ; Pb-Phoma betae; Uf-Uromyces fabae .
Two P. chrysosporium P450 families, CYP617 and CYP5031, are groupable under the CYP547 clan. While the CYP547 clan proteins occur in all the higher ascomycetous fungi studied so far, two such proteins (CYP5031A1 and CYP5032A1) have also been identified recently in the basidiomycetous fungus U. maydis , of which CYP5031A1 shows closest relatedness (35% bootstrap value) to the two P. chrysosporium P450 proteins (Figure 14). On the other hand, the six other P. chrysosporium P450 proteins (pc.14.209.1, pc.16.161.1, PFF 311a, pc.142.11.1, pc.16.153.1 and pc.5.122.1), while showing highest BLAST homology to the CYP617 family of proteins, form a distinct group on the phylogenetic tree with a bootstrap value of 79% (Figure 14). It is possible that these six genes represent a unique family of P450 proteins, hitherto unidentified in other fungi. The same is true in the case of seven genes from the CYP614/534 clan and one gene (pc.96.21.1) from the CYP62 family of P. chrysosporium , which form independent clusters on the phylogenetic tree (Figures not shown).
Figure 14 Minimal evolution tree of the fungal CYP547 clan. Eight P450 genes from P. chrysosporium showing homology with three P450 families, CYP547, CYP617, and CYP5031, were compared with homologous P450 sequences from other fungal species and were found to qualify as members of the CYP547 clan. Abbreviations: An-Aspergillus nidulans ; Eg-Eremothecium gossypii ; Fg-Fusarium graminearum ; Mg-Magnaporthe grisea ; Nc-Neurospora crassa ; Pc-Phanerochaete chrysosporium; Um-Ustilago maydis.
Based on the clan-level comparison, it is evident that, although the P. chrysosporium P450ome has member counterparts among different fungal groups, in most cases, their independent clustering suggests significant sequence diversification and likely unique functionality. Such diversification emphasizes the need for assigning new family names to the P. chrysosporium P450s.
Conclusion
The recognized extraordinary catalytic diversity of the white rot fungus P. chrysosporium correlates with its enormous P450 repertoire (P450ome), which is one of the largest among lower eukaryotes. Our structural and phylogenetic analyses of the P450ome, meant to understand the genesis of such a large number of P450 genes and facilitate their classification/nomenclature, have provided important clues to the evolution of the enormous catalytic diversity in this fungus. Considering the fact that certain P450 families (such as CYP64) have diversified more extensively than others, it appears that the P. chrysosporium P450ome has evolved in specific directions to meet the metabolic demand in its environmental niches. While the ancestral genes like CYP51 and CYP61 have remained unchanged, possibly due to their minimal role in P. chrysosporium (and other saprophytic fungi) versus that in their parasitic cousins such as pathogenic Aspergilli, other P450 gene families with suggestive roles in secondary metabolism (such as CYP64) have evolved as multigene families and even exist as gene clusters in this fungus. Such family-specific evolution was warranted presumably due to an extensive demand for generation of a broad range of metabolites in the secondary metabolic switch required for degradation of complex natural substrates such as lignin. The presented analysis indicates that the progenitor P450 families, originally acquired as a part of vertically descending P450ome, have diversified rapidly via multiple genetic mechanisms such as tandem duplications, translocations, mutations, and possibly gene fusions, to give rise to majority of the multigene families in the P. chrysosporium P450ome. Consequently, this structural diversification within individual multigene P450 families, as experimentally demonstrated in the case of CYP63 family in our studies, seems to have led to the acquisition of novel functions hitherto unseen in their ancestral counterparts (CYP52 genes of yeasts in this case). In addition, our experimental evidence for the presence of alternatively spliced variants in the P. chrysosporium P450 transcriptome further explains the evolution of expanded substrate diversity in this organism. The P. chrysosporium P450ome forms a model to investigate extrapolation of the evolved P450 gene diversity to the known vast biodegradation potential and will help design the future functional studies to understand the individual P450 gene functions in order to dissect the P450 functional diversity in white rot fungi.
Methods
Fungal cultures and cDNA cloning
Phanerochaete chrysosporium strain BKM-F-1767 (ATCC 24725) was grown as shaken cultures for 4 days in defined low nitrogen (LN) medium (2.4 mM N, 1% glucose) as described previously [14]. Total RNA (500 ng) was isolated from frozen fungal mycelia using the TRI Reagent kit (Molecular Research Center, Cincinnati, OH, USA) and a cDNA pool was generated using the SMART™ PCR cDNA synthesis kit (Clontech, Palo Alto, CA, USA) per the manufacturer's instructions. Briefly, first-strand cDNA synthesis step was carried out in a 10 μl reaction volume using 200 units of MMLV reverse transcriptase and 1 μM each of the SMART III Oligonucleotide and the CDS III/3' PCR Primer, at 42°C for 1 hour. One-fifth of the first-strand reaction (2 μl) was then added to a 100 μl long distance (LD)-PCR reaction with 1 μM each of the 5' PCR Primer and the CDS III/3' PCR Primer, for the synthesis of a double-stranded (ds) cDNA pool. Amplification parameters included initial denaturation at 95°C for 1 min., followed by a two step-PCR protocol involving use of 95°C for 1 min. and 68°C for 6 min., for 24 cycles. Quality of the ds cDNA amplified was analyzed on a 1.1% Agarose/EtBr gel and the product was quantified using spectrophotometer. Gene-specific cDNA isolation involved use of 100 ng of this cDNA mix as the template, in conjunction with an appropriate pair of gene-specific primers listed in Table 3. The gene-specific cDNA synthesis reaction contained 100 ng each of the forward and the reverse primer in a 50 μl PCR reaction volume and the amplification included 37 cycles, each involving denaturation at 95°C for 30 sec, annealing at 60°C for 30 sec, and extension at 72°C for 1 min.
The cDNA amplicons generated were cloned using 2.1-TOPO vector (Invitrogen, Carlsbad, CA, USA) per the manufacturer's instructions. The recombinant plasmid DNA for amplicon sequencing was isolated and purified using QIAprep Spin Miniprep kit (Qiagen, Valencia, CA) per the manufacturer's specifications. The DNA sequencing was performed at the university's DNA core facility. The cloned cDNA sequences generated in this study have been submitted to the GenBank under the accession numbers-AY835607 (pc-2), AY321373 (pc-4), AY321374 (pc-5), AY835606 (pc-6) and AY835608 (pc-foxy1).
Sequence alignments and phylogenetic analysis
The P. chrysosporium P450 sequences used in the phylogenetic analysis were retrieved from the website of the Joint Genome Institute of US Department of Energy- (US-DOE) and from the P450 website . The deduced amino acid sequence for a given P450 was compared from the above two sources and a sequence with the longest aa stretch was selected. However, the gene number assigned in the published white rot genome was retained for uniformity and convenience. For the cloned cDNAs, Gene Runner program (version 3.05, Hastings Software, Inc. Hastings, NY, USA) was used to extract and analyze the corresponding gene sequences from the genome, design the primers for RT-PCR amplification, and deduce the amino acid sequences. P450 sequences for other fungi were obtained from the NCBI GenBank database and the P450 website. Sequences were aligned using the CLUSTALW program at the EMBL-EBI website . Alignment of the 126 sequences of the P. chrysosporium P450ome was generated using the following customized parameters that varied from the default parameters- Matrix-BLOSUM (Henikoff), and Gap Open Penalty -1. All the other multiple alignments were generated using the default alignment parameters including the Matrix-GONNET 250 and Gap Open Penalty -10. Phylogenetic trees were constructed using the MEGA 2.1 software [40]. The minimal evolution trees (Figures 5 to 10 and 11 to 14) were generated by heuristic search using the Close-Neighbor-Interchange (CNI) algorithm, with the Neighbor-Joining tree serving as the temporary tree. The topological distance (dT) was set at 2 for searching the minimal evolution tree. The P. chrysosporium P450ome (Figure 1) was constructed using the Unweighted Pair Group Method with Arithmatic Mean (UPGMA) method with gamma distance model. The alignment gaps and missing data sites were deleted and a bootstrap value based on 1000 replications was set for all the phylogenetic trees generated in this study. Protein sequences used for constructing the phylogenetic trees were obtained either from the GenBank (those shown with accession numbers) or from the P450 web site (those with preassigned CYP names).
Authors' contributions
HD performed the experiments, the analysis of the data, and the manuscript preparation. RC was involved in the phylogenetic analysis and interpretation of the evolutionary aspects of the paper. JSY who conceived of the study, was involved in its design and overall coordination, and helped in data interpretation and preparation of the manuscript. All authors have read and approved the final manuscript.
Supplementary Material
Additional file 1
Click here for file
Additional file 2
Click here for file
Additional file 3
Click here for file
Acknowledgements
This work was supported by the NIH's National Institute of Environmental Health Sciences (NIEHS) grant R01-ES10210 (JSY).
==== Refs
Nelson DR Koymans L Kamataki T Stegeman JJ Feyereisen R Waxman DJ Waterman MR Gotoh O Coon MJ Estabrook RW Gunsalus IC Nebert DW P450 superfamily: update on new sequences, gene mapping, accession numbers and nomenclature Pharmacogenetics 1996 6 1 42 8845856
Gotoh O Omura T, Ishimura Y, FujiiKuriyama Y Evolution and differentiation of P450 genes Cytochrome P450 Second Edition 1993 Tokyo: Kodansha 255 272
Kizawa H Tomura D Oda M Fukamizu A Hoshino T Gotoh O Yasui T Shoun H Nucleotide sequence of the unique nitrate/nitrite inducible cytochrome P450 cDNA from Fusarium oxysporum J Biol Chem 1991 266 10632 10637 2037602
Ruettinger RT Wen LP Fulco AJ Coding nucleotide, 5' regulatory, and deduced amino acid sequences of P450BM3, a single peptide cytochrome P450: NADPH-P450 reductase from Bacillus megaterium J Biol Chem 1989 264 10987 10995 2544578
Degtyarenko KN Structural domains of P450-containing monooxygenases systems Protein Eng 1995 8 737 747 8637843
Nelson DR Metazoan cytochrome P450 evolution Comp Biochem Physiol 1998 121 15 22
Nelson DR Zeldin DC Hoffman SM Maltais LJ Wain HM Nebert DW Comparison of cytochrome P450 (CYP) genes from the mouse and human genomes, including nomenclature recommendations for genes, pseudogenes and alternative splice variants Pharmacogenetics 2004 14 1 18 15128046 10.1097/00008571-200401000-00001
Nelson DR Cytochrome P450 and the individuality of species Arch Biochem Biophys 1999 369 1 10 10462435 10.1006/abbi.1999.1352
Eriksson KE Blanchette RA Ander P Timell TE Biodegradation of Lignin Microbial and enzymatic degradation of wood and wood components 1990 New York: Springer 225 333
Kirk TK Farrell RL Enzymatic "combustion": the microbial degradation of lignin Annu Rev Microbiol 1987 41 465 505 3318677 10.1146/annurev.mi.41.100187.002341
Orth AB Tien M Kuck Biotechnology of lignin degradation The Mycota II Genetics and Biotechnology 1995 Berlin: Springer-Verlag 287 302
Yadav JS Soellner MB Loper JC Mishra PK Tandem cytochrome P450 monooxygenase genes and splice variants in the white rot fungus Phanerochaete chrysosporium : cloning, sequence analysis, and regulation of differential expression Fungal Genet Biol 2003 38 10 21 12553932 10.1016/S1087-1845(02)00508-X
Martinez D Larrondo LF Putnam N Gelpke MD Huang K Chapman J Helfenbein KG Ramaiya P Detter JC Larimer F Coutinho PM Henrissat B Berka R Cullen D Rokhsar D Genome sequence of the lignocellulose degrading fungus Phanerochaete chrysosporium strain RP78 Nature Biotechnol 2004 22 695 700 15122302 10.1038/nbt967
Doddapaneni H Yadav JS Differential regulation and xenobiotic induction of tandem P450 monooxygenase genes pc-1 (CYP63A1) and pc-2 (CYP63A2) in the white rot fungus Phanerochaete chrysosporium Appl Microb Biotechnol 2004 65 559 565
Yadav JS Doddapaneni H Anzenbacher P, Hudecek J Genome-wide expression profiling and xenobiotic inducibility of P450 monooxygenase genes in the white rot fungus Phanerochaete chrysosporium proceedings of the 13th Internat Conf On Cytochromes P450 Biochemistry, Biophysics And Drug Metabolism: June 29-July 3, Prague (Czech Republic) 2003 Monduzzi Editore, Bologna, Italy 333 340
Yadav JS Loper JC Cytochrome P450 oxidoreductase gene and its differentially terminated cDNAs from the white rot fungus Phanerochaete chrysosporium Curr Genet 2000 37 65 73 10672447 10.1007/s002940050010
Doddapaneni H Subramanian V Yadav JS Physiological regulation, xenobiotic induction and heterologous expression of P450 monooxygenase gene pc-3, a new member of the CYP63 gene cluster in the white rot fungus Phanerochaete chrysosporium Curr Microbiol 2005 DOI: 10.1007/s00284-005-4480-2
Kitazume T Takaya N Nakayama N Shoun H Fusarium oxysporum fatty-acid subterminal hydroxylase (CYP505) is a membrane-bound eukaryotic counterpart of Bacillus megaterium cytochrome P450BM3 J Biol Chem 2000 275 39734 39740 10995755 10.1074/jbc.M005617200
Ohkuma M Muraoka S Tanimoto T Fujii M Ohta A Takagi M CYP52 (cytochrome P450alk) multigene family in Candida maltosa : identification and characterization of eight members DNA Cell Biol 1995 14 163 173 7865134
Tudzynski P Holter K Correia T Arntz C Grammel N Keller U Evidence for an ergot alkaloid gene cluster in Claviceps purpurea Mol Gen Genet 1999 261 133 141 10071219 10.1007/s004380050950
Ahn JH Walton JD Chromosomal organization of TOX2, the complex locus controlling host-selective toxin biosynthesis in Cochliobolus carbonum Plant Cell 1996 8 887 897 8672886 10.1105/tpc.8.5.887
Brown DW Yu JH Kelkar HS Fernandes M Nesbitt TC Keller NP Adams TH Leonard TJ Twenty-five coregulated transcripts define a sterigmatocystin gene cluster in Aspergillus nidulans Proc Natl Acad Sci USA 1996 93 1418 1422 8643646 10.1073/pnas.93.4.1418
Keller NP Hohn TM Metabolic pathway gene clusters in filamentous fungi Fung Genet Biol 1997 21 17 29 10.1006/fgbi.1997.0970
Watson JD Horizontal gene transfer and the evolution of secondary metabolite gene clusters in fungi: An hypothesis Fung Genet Biol 2000 30 167 171 10.1006/fgbi.2000.1224
Black DL Protein diversity from alternative splicing: a challenge for bioinformatics and post-genome biology Cell 2000 103 367 370 11081623 10.1016/S0092-8674(00)00128-8
Leipzig J Pevzner P Heber S The Alternative Splicing Gallery (ASG): bridging the gap between genome and transcriptome Nucleic Acids Res 2004 32 3977 3983 15292448 10.1093/nar/gkh731
Larrondo LF Gonzalez B Cullen D Vicuna R Characterization of a multicopper oxidase gene cluster in Phanerochaete chrysosporium and evidence of altered splicing of the mco transcripts Microbiology 2004 150 2775 2783 15289573 10.1099/mic.0.27072-0
Nelson DR Schuler MA Paquette SM Werck-Reichhart D Bak S Comparative genomics of rice and Arabidopsis. Analysis of 727 cytochrome p450 genes and pseudogenes from a monocot and a dicot Plant Physiol 2004 135 756 772 15208422 10.1104/pp.104.039826
Joseph-Horn T Hollomon DW Molecular mechanisms of azole resistance in fungi FEMS Microbiol Lett 1997 149 141 149 9141655 10.1016/S0378-1097(97)00043-8
Skaggs BA Alexander JF Pierson CA Schweitzer KS Chun KT Koegel C Barbuch R Bard M Cloning and characterization of the Saccharomyces cerevisiae C-22 sterol desaturase gene, encoding a second cytochrome P-450 involved in ergosterol biosynthesis Gene 1996 169 105 109 8635732 10.1016/0378-1119(95)00770-9
Sanglard D Loper JC Characterization of the alkane-inducible cytochrome P450 (P450alk) gene from the yeast Candida tropicalis : identification of a new P450 gene family Gene 1989 76 121 136 2663647 10.1016/0378-1119(89)90014-0
Zimmer T Iida T Schunck WH Yoshida Y Ohta A Takagi M Relation between evolutionary distance and enzymatic properties among the members of the CYP52A sub-family of Candida maltosa Biochem Biophys Res Commun 1998 251 244 247 9790939 10.1006/bbrc.1998.9450
Yadav JS Loper JC Multiple p450alk (cytochrome P450 alkane hydroxylase) genes from the halotolerant yeast Debaryomyces hansenii Gene 1999 226 139 46 9931473 10.1016/S0378-1119(98)00579-4
Iida T Sumita T Ohta A Takagi M The cytochrome P450ALK multigene family of an n-alkane-assimilating yeast, Yarrowia lipolytica : cloning and characterization of genes coding for new CYP52 family members Yeast 2000 16 1077 87 10953079 10.1002/1097-0061(20000915)16:12<1077::AID-YEA601>3.0.CO;2-K
Bhatnagar D Ehrlich KC Cleveland TE Molecular genetic analysis and regulation of aflatoxin biosynthesis Appl Microb Biotechnol 2003 61 83 93
Berbee ML Taylor JW Dating the evolutionary radiations of the true fungi Canadian J Bot 1993 71 1114 1127
Gonzalez P Barosso G Labarere J Molecular analysis of the split cox1 gene from the basidiomycota Agrocybe aegerita : relationship of its introns with homologous ascomycota introns and divergence levels from common ancestral copies Gene 1998 220 45 53 9767103 10.1016/S0378-1119(98)00421-1
Hohn TM Desjardins AE McCormick SP The Tri4 gene of Fusarium sporotrichioides encodes a cytochrome P450 monooxygenase involved in trichothecene biosynthesis Mol Gen Genet 1995 248 95 102 7651333
Hahn M Mendgen K Characterization of in planta-induced rust genes isolated from a haustorium-specific cDNA library Mol Plant Microbe Interact 1997 10 427 437 9150592
Kumar S Tamura K Jakobsen IB Nei M MEGA2: Molecular Evolutionary Genetics Analysis software Bioinformatics 2001 12 1244 1245 11751241 10.1093/bioinformatics/17.12.1244
|
15955240
|
PMC1184071
|
CC BY
|
2021-01-04 16:32:49
|
no
|
BMC Genomics. 2005 Jun 14; 6:92
|
utf-8
|
BMC Genomics
| 2,005 |
10.1186/1471-2164-6-92
|
oa_comm
|
==== Front
BMC GenomicsBMC Genomics1471-2164BioMed Central London 1471-2164-6-961597813110.1186/1471-2164-6-96Research ArticleBacillus thuringiensis Cry1Ca-resistant Spodoptera exigua lacks expression of one of four Aminopeptidase N genes Herrero Salvador [email protected] Tsanko [email protected] Petra L [email protected] William J [email protected] Maagd Ruud A [email protected] Business Unit Bioscience, Plant Research International B.V., Wageningen University and Research Center, P.O. Box 16, 6700 AA Wageningen, The Netherlands2 Laboratory of Virology, Department of Plant Sciences, Wageningen University, Binnenhaven 11, 6709 PD Wageningen, The Netherlands3 Department of Entomology and Plant Pathology, Auburn University, Auburn, Alabama 36849, USA2005 24 6 2005 6 96 96 17 3 2005 24 6 2005 Copyright © 2005 Herrero et al; licensee BioMed Central Ltd.2005Herrero et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Insecticidal toxins from Bacillus thuringiensis bind to receptors on midgut epithelial cells of susceptible insect larvae. Aminopeptidases N (APNs) from several insect species have been shown to be putative receptors for these toxins. Here we report the cloning and expression analysis of four APN cDNAs from Spodoptera exigua.
Results
Suppression Subtractive Hybridization (SSH) was used to construct cDNA libraries of genes that are up-and down-regulated in the midgut of last instar larvae of beet armyworm, S. exigua exposed to B. thuringiensis Cry1Ca toxin. Among the clones from the SSH libraries, cDNA fragments coding for two different APNs were obtained (APN2 and APN4). A similar procedure was employed to compare mRNA differences between susceptible and Cry1Ca resistant S. exigua. Among the clones from this last comparison, cDNA fragments belonging to a third APN (APN1) were detected. Using sequences obtained from the three APN cDNA fragments and degenerate primers for a fourth APN (APN3), the full length sequences of four S. exigua APN cDNAs were obtained. Northern blot analysis of expression of the four APNs showed complete absence of APN1 expression in the resistant insects, while the other three APNs showed similar expression levels in the resistant and susceptible insects.
Conclusion
We have cloned and characterized four different midgut APN cDNAs from S. exigua. Expression analysis revealed the lack of expression of one of these APNs in the larvae of a Cry1Ca-resistant colony. Combined with previous evidence that shows the importance of APN in the mode of action of B. thuringiensis toxins, these results suggest that the lack of APN1 expression plays a role in the resistance to Cry1Ca in this S. exigua colony.
==== Body
Background
During sporulation, B. thuringiensis (Bt) produces a crystal composed of one or more Cry proteins with toxicity against insects. The mode of action of the largest group of Cry proteins has been extensively studied and can be divided into four main steps: (i) Solubilization of the inclusion body to release the Cry proteins in their protoxin form, (ii) gut protease processing of these protoxins to an active toxin, (iii) binding of the active form to specific receptors in the midgut of the insect, (iv) membrane insertion, pore formation, and cellular lysis [1]. Any alteration in one of these steps could result in the development of resistance to one or several Cry proteins in a given insect population.
Development of insect resistance to insecticides is one of the most important problems of agriculture because it increases the costs of crop protection and reduces its productivity. More than 500 insect and mite species have been reported to develop resistance to one or more pesticides [2]. Although less in number, several cases of resistance have been also reported for B. thuringiensis insecticides in field populations as well as in laboratory selection experiments [3]. With the commercialization of transgenic crops expressing B. thuringiensis toxins (Bt-crops) the selection pressure has increased, with the consequent increased risk of resistance development. A high-dose/refuge strategy has been proposed and implemented in some cases to delay the development of insect resistance to Bt-crops. The effectiveness of this strategy mainly depends on the mode of inheritance of resistance and the initial frequency of Bt resistance alleles. Knowledge of the genes involved in Bt resistance will allow fast molecular screening for resistance gene frequencies in the field before or during use of a Bt crop. Furthermore, to determine if refuges or any other resistance management strategy are working, one should keep track of the frequency of resistance in field [4].
So far, the best-characterized mechanism of Bt resistance is the alteration of Cry protein binding to its receptors in the midgut of the target insect. This alteration has been reported in different populations of Plutella xylostella, Plodia interpunctella, Heliothis virescens, and Spodoptera exigua and also for several Cry proteins (Cry1Aa, Cry1Ab, Cry1Ac, and Cry1Ca) [3]. Aminopeptidase N (APN) and cadherin-like proteins have been characterized as candidates for Cry1 toxin receptors [1]. Gahan et. al, 2001 [5] studying the H. virescens (YHD2) resistant strain showed that the major gene for resistance to Cry1Ac in this strain was highly linked to the locus coding for a cadherin-like protein, and that this gene was disrupted in the resistant strain by the insertion of a retrotransposon. In 1994, Knight et al. [6] identified an APN as an insect receptor of the Cry1Ac protein. Since 1994 to today, more than 60 different APNs from different Lepidoptera have been sequenced and registered in databases showing the high diversity in isoforms. Other putative receptors for Cry toxins are proteins [7], glycoconjugates [8], and glycolipids [9].
Previous selection studies in the beet armyworm, S. exigua showed the ability of this insect to develop high levels of resistance (>75-fold at generation 35) to the Cry1Ca activated toxin [10]. Binding of this protein to brush border membrane vesicles (BBMV) from the midgut of the resistant insects was decreased 5-fold in affinity compared with the susceptible insects [10].
In our study, using Suppression Subtractive Hybridization (SSH), we compared S. exigua midgut gene expression between Cry1Ca-exposed and non-exposed susceptible insects and between resistant and susceptible non-exposed insects in order to identify genes that may be differentially expressed and therefore putatively involved in the insect response to toxin action or in resistance. Based on the SSH results, full cDNA sequences of four APNs were obtained. Northern blot analysis showed the lack of expression of one of the APNs in the insects from the resistant colony.
Results
Isolation of APN encoding cDNAs
In order to study genes that might be involved in the insect's response to intoxication by an active Cry toxin and/or in the mechanism of resistance to such a toxin, we took the approach of selection of cDNAs representing genes, which are differentially expressed in these different conditions. Suppression Subtractive Hybridization (SSH) was performed, in the first experiment with midgut cDNA from Cry1Ca-sensitive 5th instar S. exigua larvae that had been raised on either control diet or on diet containing a sublethal dose of Cry1Ca protein during the entire larval life stage. Thus two pools of subtracted cDNAs were obtained and cloned, one presumably containing cDNAs representing genes with increased expression (library 1) and one representing genes with decreased expression (library 2) in response to toxin exposure, respectively.
In a second experiment, SSH was performed with midgut cDNA pools from Cry1Ca-sensitive and Cry1Ca-resistant 5th instar larvae, both raised on diet without toxin. Thus two more pools of subtracted cDNAs were obtained and cloned, one representing genes with comparatively higher expression in sensitive larva (library 3) and one with lower expression in sensitive larva (library 4) as compared to resistant larva.
From each of the four libraries 96 clones were randomly picked and sequenced. Sequences were analyzed and overlapping or identical sequences were assembled into contigs. An in-depth analysis of the characterized sequences and the studies on the expression of the corresponding genes by microarray analysis will be published elsewhere. Homology searches with contig sequences resulted in the identification of a number of contigs encoding peptides with high homology to known lepidopteran Aminopeptidase N sequences and were selected for further studies.
In library 1 (higher expression in toxin-exposed larvae) we identified two single sequences of 375 and 436 base pairs, respectively. The first, designated contig 13 (APN2, following the classification of Nakanishi et al. [11]), encoded a peptide with high homology with Lymantria dispar APN2 [GenBank: AAD31184] amino acids 265 to 389 and almost equally high homology to a Bombyx mori APN [GenBank: BAA32140]. The second, contig 31 (APN4) showed highest homology to Spodoptera litura APN [GenBank: AAK69605] amino acids 173 to 296 and to a lesser extent to a second B. mori APN [GenBank: BAA33715]. The fact that the two encoded peptides were overlapping but not entirely identical in the overlap region, and were homologous to two different B. mori APNs, indicated that they might be fragments of two different APN cDNAs.
In library 3 (higher expression in sensitive larvae compared to resistant larvae) we identified 3 contigs with homology to APNs. Contig 82 (APN1) consisted of 6 identical copies of a 309 base pair fragment, while contigs 101 and 128 consisted of single fragments of 142 and 144 bp, respectively. All three sequences encoded peptides with high homology to different parts of Helicoverpa armigera APN1 [Genbank: AAK85538] (amino acids 19 to 123, 388 to 430, and 666 to 709 for contigs 82, 101, and 128, respectively) and to a lesser extent to a third B. mori APN [Genbank: AAC33301]. Although the three peptides were not overlapping, the high homology to single respresentatives of the APNs of H. armigera and B. mori indicated that they might be derived from one single S. exigua cDNA, different from the two cDNAs derived from library 1.
Overall results from SSH experiments provided the partial sequence of 3 different APNs (APN1, 2 and 4). Since previous work [11,12] suggested the presence of at least four different classes of APNs in Lepidoptera, a degenerate primer (Table 1) based on the known sequences of APN class 3 of other Lepidoptera was designed for the specific amplification of the S. exigua APN corresponding to class 3 in 3'RACE experiments. Sequences obtained from 3'RACE were employed for the design of a primer for the 5' end fragment amplification (Table 1).
Table 1 Sequence and localization of the primers employed for the isolation ofthe midgut aminopeptidases from S. exigua by RACE experiments
Gene name Primer sequence (5' to 3') Position in ORF
APN1 3'Race TGCGCTTCGAAGATTGGCTCACGATA 2018–2044
5'Race CTTCGATGAATGCCGAGACGCTGGAC 2122–2097
APN2 3'Race CTTTGCTGCTGGTGCTATGGAGAACTGG 906–932
5'Race ACAGGGCTGACTTCATTTCCGAACCATT 1067–1040
APN3 3'Race ATGCGTGAYGAYATGTACGGTAT 496–518
5'Race GCCAGGAGATAAGTTGACATTTTTGGGG 779–752
APN4 3'Race TGGCTTTCCATGTGAGCGATTTTGTGC 728–754
5'Race CACACCGATTTCTGCAGCATACGAGTGC 849–822
Aminopeptidase sequence analysis
Using the primers described in Table 1, overlapping fragments from the 5'-and 3'-ends of the four APNs were amplified by RACE, cloned in pGEM-T Easy and sequenced. Open reading frames (ORF) of 3063, 2880, 3015, and 2853 bp were obtained for APN1 [GenBank: AY218842], APN2 [GenBank: AY218843], APN3 [GenBank: AY218844], and APN4 [GenBank: AY218845], respectively (Fig. 1). Protein products of 1021, 960, 1005, and 951 amino acids were predicted from the APN1, APN2, APN3, and APN4 cDNA sequences, respectively.
Figure 1 Schematic representation of the four cloned APN cDNAs. Gray bars correspond to the respective ORF. White blocks inside the gray bars represent the cDNA fragments coming from the SSH experiments and numbers correspond to the contig name. The black lines over each bar correspond to the fragment of RNA probe used in the Northern blots.
The predicted proteins were aligned using ClustalX as described in material and methods (Figure 2). Percentage of identity among the different APNs ranged from 36.7 between APN1 and APN3 to 23.3 between APN2 and APN3. Analysis of the N-terminal region of the predicted proteins using the SignalP program [13] indicated the presence of a signal peptide sequence in all four APNs (residues enclosed by dashed lines in Figure 2). All four APNs showed the presence of a zinc binding motif (residues enclosed by a solid line in Figure 2) and the GAMEN motif (underlined) characteristic of the gluzincin aminopeptidases and involved in their aminopeptidase activity [14]. Analysis of the amino acid sequences using the GPI Predictor Server [15] revealed the presence of a putative Glycosylphosphatidylinositol (GPI)-anchor site in the C-terminal end of all four aminopeptidases. The predicted GPI-modification sites were localized in the amino acid positions 1000 (APN1), 938 (APN2), 983 (APN3), and 931 (APN4).
Figure 2 ClustalX aligment of the deduced amino acid sequences from the four Spodoptera exigua aminopeptidases. Dashed lines enclose the predicted signal peptides. Solid lines box the putative Zinc binding motives. The GAMEN motif is underlined. Predicted GPI-modification sites in the C-terminal sequence are indicated by the black triangles.
One allele each from all of the lepidopteran APNs deposited in GenBank so far (see methods section) were aligned together with these four S. exigua APN's using the ClustalX program. Phylogenetic tree construction showed that lepidopteran APNs group in five subfamilies (Figure 3). Alignment also shows that the four S. exigua APNs reported here each represent one of the larger four groups. Our APN3 is highly homologous to a S. exigua APN protein sequence deposited in GenBank during the course of our studies [GenBank: AAT99437]. Compared with the APN3 from the susceptible colony employed in this study, this APN showed 27 amino acid differences. Although degenerate primers were designed for the RACE of a S. exigua cDNA representative of the putative class 5, we were not able to obtain an amplification product corresponding to such an APN5. Bootstrapping analysis was performed in order to test the robustness of the generated tree. Bootstrap values of 1000 were obtained for all the main branches except for the one corresponding to the Class 3 APN. For this branch a bootstrap value of 720 was obtained, likely caused by the presence of the P. interpunctella APN [GenBank: AAC36148]. When this sequence was omitted from the alignment, bootstrap values of 1000 were obtained for all the five branches.
Figure 3 Phylogenetic tree derived from a ClustalX alignment of all published lepidopteran midgut aminopeptidases. Species name and GenBank accession number are shown for each protein. Numbers on the branches report the level of confidence as determined by bootstrap analysis (1000 bootstrap replicates). To clarify the figure, bootstrap values are shown only for the main branches. The scale bar indicates an evolutionary distance of 0.1 amino acid substitutions per position in the sequence. Species abbreviations: Se: Spodoptera exigua; Ms: Manduca sexta; Ld: Lymantria dispar; Hv: Heliothis virescens; Ha: Helicoverpa armigera; Hp: Helicoverpa punctigera; Bm: Bombyx mori; Sl: Spodoptera litura; Px: Plutella xylostella; Pi: Plodia interpunctella; Ep: Epiphyas postvittana
APN expression analysis
The cDNA fragments of APN2 and APN4 originated each as a single fragment from a library putatively representing genes that were up-regulated in response to toxin exposure. However, when we compared expression of these genes as well as of APN1 and APN3 between guts of larvae exposed or not exposed to Cry1Ca toxin, by Northern blotting, we found no evidence for regulation of their expression by toxin exposure (Figure 4A).
Figure 4 Northern blot analysis of APN expression in 5th instar midguts and whole neonate larvae of Spodoptera exigua. A. Comparison of expression of APNs in midguts from control (-) and toxin-exposed larvae (1Ca). An ATP synthase transcript (ATPsyn) was used as an internal control of RNA extraction and gel loading. B. Expression of APNs in midguts from fifth instar larvae (left panel) and whole neonate larvae (right panel) of the susceptible (S) and of the Cry1Ca-resistant (R) line. Higher molecular weight bands correspond to the different APN transcripts. Lower molecular weight bands correspond to the ATPsynthase transcripts.
The three cDNA contigs of APN1 were all derived from library 3, putatively representing genes that are expressed at a higher level in sensitive larvae compared to in resistant larvae. This suggested that APN1 was not, or at a lower level, expressed in the resistant line. In order to determine differences in expression levels of the APNs between the susceptible and resistant larvae, Northern blot analysis was performed with total RNA extracted from midguts of 5th instar larvae of both colonies. Northern blot analysis showed that all four APNs are expressed in susceptible larvae (Figure 4B, "S" lanes). In agreement with the SSH results, no expression of APN1 was detected in batches of RNA from resistant larvae (around 10 larvae/batch), while the other three APNs were expressed at levels similar to that in susceptible larvae (Figure 4B, "R" lanes). Similar results were obtained with different batches of RNA (at least three) from the resistant insects and employing higher concentrations of RNA or in RT-PCR experiments specific for APN1 (data not shown), indicating that lack of APN1 expression is virtually complete in the resistant colony. Since most bioassays and selections for resistance are performed with the more toxin-sensitive neonate larvae in stead of the 5th instar larvae used here, we also tested APN1 expression in (whole) neonate larvae from the sensitive and resistant colonies. Also here, APN1 expression was detectable in sensitive larvae, but not in resistant larvae (Fig. 4B, right panel).
Discussion
Based on the results from SSH experiments and in one case by using degenerate primers (APN3), sequences of four different APN cDNAs from S. exigua have been determined. Sequence analysis of the predicted proteins showed the common structural characteristics that other lepidopteran APNs have, such as signal peptides, zinc binding sites, GAMEN motifs, and GPI-anchor sites. Oltean et al. [12] suggested the distribution of the different lepidopteran APNs in at least 4 homology groups following phylogenetic analysis of the sequences. Nakanishi et al. [11] called these groups class 1 to 4. Currently, with the addition of new sequences to GenBank, a new analysis clearly reveals the presence of at least five different classes of APNs in Lepidoptera. Definition of the fifth class is not only supported by the phylogenetic analysis of the sequences but also by the presence of a sequence from Helicoverpa armigera in all 5 branches [GenBank: AAK85538, AAW72993, AAN04900, AAM44056, and AAK85539]. Amplification of a S. exigua cDNA representing class 5 was attempted without success. Possible explanations could be that the degenerate primers used for the amplification of this APN were designed based on the consensus of only two sequences, and the selected primer sequences may not be conserved in a S. exigua APN5. Alternatively, S. exigua may not have a gene for the class 5 APN, or its expression in the midgut is too low to be picked up as cDNA.
SSH comparison between the susceptible and the resistant colonies suggested a lower expression of APN1 in the resistant insects compared with the susceptible insects. This difference in the expression was confirmed by Northern blot analysis (Fig. 4B), showing clearly the absence of expression of this APN in the resistant insects, both in 5th instar larvae as well as in neonate larvae. APN has been described, together with cadherin-like proteins, as one of the proteins involved in the binding of Cry1A toxins to the midgut of several lepidopteran species [6,12,16-21]. Recently new evidence of the importance of APNs in the mode of action of Bt toxins was obtained by transformation of Drosophila melanogaster with the M. sexta class 1 APN [Genbank: CCA61452]. Expression of this APN in the midgut of the D. melanogaster larvae increased the susceptibility of these insects to Cry1Ac toxin [22].
In the case of Cry1Ca toxin, Luo et al. [23] using toxin affinity chromatography, detected and isolated a 106 kDa APN from M. sexta that bound to Cry1Ca. Although the entire sequence of this 106 kDa APN has not been published, its N-terminal sequence is nearly identical (9 out of 10 amino acids) to the sequence of the published M. sexta class 1 APN [GenBank: CAA61452] [24] and [GenBank: AAF07223] [25], and much more different from the M. sexta class 2, 3, and 4 APNs. This strongly suggests that at least one of the M. sexta Cry1Ca-binding proteins is a class 1 APN.
Agrawal et al. [26] cloned a Spodoptera litura class 4 APN cDNA, slapn [GenBank: AAK69605], and expressed it in insect cells, detecting binding of Cry1Ca toxin to the surface of the cells expressing this APN. More recent results showed that silencing of the expression of the slapn in S. litura by RNA interference (RNAi) through injection of a 756 bp dsRNA fragment, reduced the susceptibility of these insects to Cry1Ca [27].
So far, only one fully characterized gene has been associated with the development of resistance to Bt toxins in Lepidoptera. Gahan et al. [5] studied the Heliothis virescens YHD2 resistant strain and detected a strong linkage between the resistance to Cry1Ac toxin and the disruption of a cadherin gene. Unfortunately, since our S. exigua resistant line was lost during the development of this research, analysis of linkage between the resistance to Cry1Ca and the lack of APN1 expression could not be performed.
Data from SSH with resistant vs susceptible insects, support a possible role of the APNs in the mode of action of Bt toxins in Lepidoptera. Previous binding studies with the resistant line employed here, showed a 5-fold decrease in the binding affinity of Cry1Ca toxin to brush border membrane vesicles prepared with midguts from the resistant line. All these data together with previous work supporting the binding of Cry1Ca toxins to APN either in M. sexta or S. litura, strongly support the notion that lack of APN1 in S. exigua is responsible, at least in part, for resistance to Cry1Ca in this resistant line. Future work on the possible role of the APN in toxin binding, including heterologous expression or down-regulation by RNA interference strategies may further support this role.
Conclusion
We have cloned and characterized four different midgut cDNAs from S. exigua, representing 4 out of 5 putative classes of lepidopteran APNs and have shown that one of these was not expressed in the larvae of a Cry1Ca-resistant colony. Combined with other evidence of the importance of APN in the mode of action of B. thuringiensis toxins these results suggest that the lack of APN1 expression plays a role in the resistance to Cry1Ca in this S. exigua colony.
Methods
Insects and tissue isolation
Cry1Ca-sensitive (parental strain of the Cry1Ca-resistant colony) and Cry1Ca resistant S. exigua larvae were obtained from laboratory colonies maintained at Auburn University [10]. The Cry1Ca-resistant population was selected with trypsinized Cry1Ca at 320 μg/gram diet most generations, including the generation preceding that from which eggs were obtained for gene expression studies [10]. Eggs were hatched and larvae reared on artificial diet at 28°C. Midguts were pulled from early 5th instar larvae after cutting off the hindbody between the last two pairs of prolegs. Next, midguts were cut longitudinally and washed in phosphate-buffered physiological saline to remove gut contents. Midguts were stored at -80°C until further use.
Toxin isolation
Escherichia coli expressing Cry1Ca was grown for protoxin production and protoxin was extracted, solubilized and trypsin-treated as described before [28]. Trypsinized (activated) toxin was dialyzed overnight against 50 mM sodium hydrogencarbonate pH9, 150 mM sodium chloride and quantified by SDS-PAGE. For feeding studies toxin solutions were diluted with PBS (Phosphate Buffered-Saline) to 250 ng Cry1Ca per gram diet and mixed with autoclaved artificial agar-based diet, which had been cooled to 55°C. Control diets merely contained equal amounts of PBS.
mRNA isolation
Total RNA was extracted from whole 5th instar midguts using TriPure Isolation Reagent (Roche Diagnostics, Almere) according to the protocols provided by the supplier. mRNA was purified from total RNA using GenoPrep oligo-dT beads (GenoVision, Oslo). Concentration of mRNA was calculated from OD260 and mRNA quality checked by agarose gel electrophoresis.
Subtracted cDNA-fragment library construction
In order to isolate cDNA fragments representing mRNAs more abundant in midgut tissue of toxin-exposed or control diet-fed larvae, we used Suppression Subtractive Hybridization (SSH) [29] with the Clontech PCR-Select cDNA subtraction kit (Clontech, Palo Alto), according to the protocols provided by the supplier. Briefly, pools of mRNA were isolated as described above from 10 5th instar beet armyworm midguts (exposed or unexposed to Cry1Ca toxin). cDNA was synthesized with AMV reverse transcriptase (first strand) and DNA polymerase I, and E. coli DNA ligase in the presence of RNase H followed by T4 DNA polymerase (second strand). cDNAs were digested with RsaI, and appropriate linkers were ligated to the resulting cDNA fragments of the "tester" pool using T4 DNA ligase. cDNA fragment pools resulting from subtractive hybridization representing toxin-induced mRNAs ("exposed" pool as tester and "control" cDNA as driver) or for toxin-repressed mRNAs (reciprocal experiment) were further enriched with a "suppression PCR" step. Thus enriched cDNA pools were ligated into the vector pGEM-T Easy (Promega Benelux B.V., Leiden) and used for transformation of E. coli strain JM109.
In a second subtraction library construction experiment, cDNAs from 5th instar larval guts of the Cry1Ca-sensitive line and of the Cry1Ca-selected line (both grown on diet without toxin), were used for subtractive hybridization. Pools containing cDNAs representing genes that were putatively expressed at a higher level (when cDNA of resistant insects was used as "tester") or at lower level (when cDNA of sensitive insects was used as "tester") in the resistant line compared to the sensitive line were separately ligated into pGEM-T Easy and used for transformation of E. coli strain JM109.
Ampiflication of cDNA-5' and 3' fragments
For selected cDNA-fragments, both 3' as well as 5' cDNA fragments were amplified using the SMART RACE-kit from Clontech. For all cDNA-ends 5'-ready and 3'-ready cDNA pools were produced from reverse transcribed mRNA (using PowerScript reverse transcriptase) isolated from midguts of 5th instar susceptible S. exigua exposed to toxin as described above. cDNA specific primers used for the 5' and 3' cDNA amplifications are listed in Table 1. Amplified cDNA-end fragments were purified using a QIAquick PCR purification kit (Qiagen, Benelux B.V., Venlo) and ligated into pGEM-T Easy (Promega Benelux B.V., Leiden).
DNA sequencing and sequence analysis
DNA sequencing was performed by the dye-termination method using Bigdye terminator sequence mix (PE Applied Biosystems Benelux; Nieuwerkerk a/d IJssel) in an ABI 3700 automatic DNA sequencer. All oligonucleotides were obtained from Eurogentec (Brussels). DNA sequence homology searches were performed using the BlastX algorithm [30].
Comparison of the deduced amino acid sequences corresponding to the cloned APNs and phylogenetic reconstruction were performed using the ClustalX program [31]. A preliminary alignment was performed with the 61 different complete APNs protein sequences in the database. Based on the first alignment, redundant sequences, probably representing alleles from the same APN in the same insect species (identity higher than 95 %) were screened and a representative member for each class-species combination was selected for further analysis. A second phylogenetic reconstruction was performed with the 31 selected APNs and the four APNs isolated in current work. Phylogenetic reconstruction was obtained by the neighbor-joining method [32] together with bootstrap analysis using 1000 replicates. Kimura correction for multiple substitutions was applied [33].
Northern blot expression analyses
Total RNA (1 μg) extracted from midgut tissue or whole neonate larvae was heat-denatured and applied to a 1% agarose gel in 2.2 M formaldehyde-MOPS buffer. After electrophoresis, the RNA was transferred overnight onto nylon membrane (Roche Diagnostics, Almere) and then UV crosslinked for 60 seconds. The labeled probe was prepared as follows. cDNA fragments cloned in pGEM-T Easy were amplified by PCR using M13 primers. The PCR products were purified and used as template for in vitro transcription using the SP6 polymerase in the presence of Digoxigenin-UTP, following the procedure described in the DIG RNA labeling Kit (Roche Diagnostics, Almere). Probe fragments of the APNs corresponded to the 5' end of the cDNA of each gene. As a control for equal loading of the Northern blots a probe corresponding to a cDNA fragment with homology to an ATP synthase obtained from the first subtraction library was used, for which microarray experiments showed no differential expression over a wide range of conditions (results not shown). The sizes of the probes were approximately 1300 bases (APN1), 1000 bases (APN2), 850 bases (APN4), and 700 bases (APN3, and ATP synthase) (Fig. 1). Hybridization and detection was performed following the instructions of the Dig Northern Starter Kit (Roche Diagnostics, Almere).
Authors' contributions
SH participated in sequence analysis, expression studies, RACE PCR and phylogenetic analysis and manuscript preparation. TG participated in RACE PCR, sequence analysis and expression studies. PLB carried out the SSH experiments, sequence analysis and participated in RACE PCR. WJM performed insect selection, carried out the insect cultures and participated in the manuscript preparation. RAdM conceived of the study, participated in its design and coordination, and participated in manuscript preparation. All authors have read and approved the manuscript.
Acknowledgements
SH was supported by a postdoctoral fellowship of the Spanish Ministry of Education, Culture and Sports and by a European Community Marie Curie fellowship (contract MCFI-2002-00796). TG was supported by an IAC-fellowship from the Dutch Ministry of Agriculture, Nature Management and Fisheries. PLB. and RAdM were supported by the Dutch Ministry of Agriculture, Nature Management and Fisheries DWK program 347 "Biological Safety of Transgenic plants".
==== Refs
de Maagd RA Bravo A Crickmore N How Bacillus thuringiensis has evolved specific toxins to colonize the insect world Trends Genet 2001 17 193 199 11275324 10.1016/S0168-9525(01)02237-5
Georghiou G Lagunes-Tejada A The occurrence of resistance to pesticides in arthropods Food and Agriculture Organization of the United Nations, Rome 1991
Ferré J Van Rie J Biochemistry and genetics of insect resistance to Bacillus thuringiensis Ann Rev Entomol 2002 47 501 533 11729083 10.1146/annurev.ento.47.091201.145234
Tabashnik BE Breaking the code of resistance Nature Biotechnol 2001 19 922 924 11581654 10.1038/nbt1001-922
Gahan LJ Gould F Heckel DG Identification of a gene associated with bt resistance in Heliothis virescens Science 2001 293 857 860 11486086 10.1126/science.1060949
Knight PJK Crickmore N Ellar DJ The receptor for Bacillus thuringiensis CryIA(c) delta-endotoxin in the brush border membrane of the lepidopteran Manduca sexta is aminopeptidase N Mol Microbiol 1994 11 429 436 7908713
Hossain DM Shitomi Y Moriyama K Higuchi M Hayakawa T Mitsui T Sato R Hori H Characterization of a novel plasma membrane protein, expressed in the midgut epithelia of Bombyx mori, that binds to Cry1A toxins Appl Environ Microbiol 2004 70 4604 4612 15294792 10.1128/AEM.70.8.4604-4612.2004
Valaitis AP Jenkins JL Lee MK Dean DH Garner KJ Isolation and partial characterization of gypsy moth BTR-270, an anionic brush border membrane glycoconjugate that binds Bacillus thuringiensis Cry1A toxins with high affinity Arch Insect Biochem Physiol 2001 46 186 200 11304752 10.1002/arch.1028
Griffitts JS Haslam SM Yang T Garczynski SF Mulloy B Morris H Cremer PS Dell A Adang MJ Aroian RV Glycolipids as receptors for Bacillus thuringiensis crystal toxin Science 2005 307 922 925 15705852 10.1126/science.1104444
Moar WJ Pusztai-Carey M Faassen HV Bosch D Frutos R Rang C Luo K Adang MJ Development of Bacillus thuringiensis Cry1C resistance by Spodoptera exigua (Hubner) (Lepidoptera, Noctuidae) Appl Environ Microbiol 1995 61 2086 2092 16535038
Nakanishi K Yaoi K Nagino Y Hara H Kitami M Atsumi S Miura N Sato R Aminopeptidase N isoforms from the midgut of Bombyx mori and Plutella xylostella – their classification and the factors that determine their binding specificity to Bacillus thuringiensis Cry1A toxin FEBS Lett 2002 519 215 220 12023048 10.1016/S0014-5793(02)02708-4
Oltean DI Pullikuth AK Lee HK Gill SS Partial purification and characterization of Bacillus thuringiensis Cry1A toxin receptor a from Heliothis virescens and cloning of the corresponding cDNA Appl Environ Microbiol 1999 65 4760 4766 10543783
Nielsen H Engelbrecht J Brunak S von Heijne G Identification of prokaryotic and eukaryotic signal peptides and prediction of their cleavage sites Protein Eng 1997 10 1 6 9051728 10.1093/protein/10.1.1
Laustsen PG Vang S Kristensen T Mutational analysis of the active site of human insulin-regulated aminopeptidase Eur J Biochem 2001 268 98 104 11121108 10.1046/j.1432-1327.2001.01848.x
Eisenhaber B Bork P Eisenhaber F Prediction of potential GPI-modification sites in proprotein sequences J Mol Biol 1999 292 741 758 10497036 10.1006/jmbi.1999.3069
Jenkins J Dean D Binding specificity of Bacillus thuringiensis Cry1Aa for purified, native Bombyx mori aminopeptidase N and cadherin-like receptors BMC Biochemistry 2001 2 12 11722800 10.1186/1472-2091-2-12
Hua G Tsukamoto K Rasilo ML Ikezawa H Molecular cloning of a GPI-anchored aminopeptidase N from Bombyx mori midgut: a putative receptor for Bacillus thuringiensis CryIA toxin Gene 1998 214 177 185 9729121 10.1016/S0378-1119(98)00199-1
Luo K Sangadala S Masson L Mazza A Brousseau R Adang MJ The Heliothis virescens 170 kDa aminopeptidase functions as "receptor A" by mediating specific Bacillus thuringiensis Cry1A delta-endotoxin binding and pore formation Insect Biochem Molec Biol 1997 27 735 743 9443374 10.1016/S0965-1748(97)00052-0
Denolf P Hendrickx K Vandamme J Jansens S Peferoen M Degheele D van Rie J Cloning and characterization of Manduca sexta and Plutella xylostella midgut Aminopeptidase N enzymes related to Bacillus thuringiensis toxin binding proteins Eur J Biochem 1997 248 748 761 9342226 10.1111/j.1432-1033.1997.t01-1-00748.x
Yaoi K Kadotani T Kuwana H Shinkawa A Takahashi T Iwahana H Sato R Aminopeptidase N from Bombyx mori as a candidate for the receptor of Bacillus thuringiensis Cry1Aa toxin Eur J Biochem 1997 246 652 657 9219522 10.1111/j.1432-1033.1997.t01-1-00652.x
Valaitis AP Lee M Rajamohan F Dean DH Lee MK Brush border membrane aminopeptidase-N in the midgut of the gypsy moth serves as the receptor for the CryIA(c) delta-endotoxin of Bacillus thuringiensis Insect Biochem Molec Biol 1995 25 1143 1151 8580914 10.1016/0965-1748(95)00050-X
Gill M Ellar D Transgenic Drosophila reveals a functional in vivo receptor for the Bacillus thuringiensis toxin Cry1Ac1 Insect Molec Biol 2002 11 619 625 12421420 10.1046/j.1365-2583.2002.00373.x
Luo K Lu YJ Adang MJ A 106 Kda form of Aminopeptidase is a receptor for Bacillus thuringiensis Cry1C delta-endotoxin in the brush border membrane of Manduca sexta Insect Biochem Molec Biol 1996 26 783 791 10.1016/S0965-1748(96)00027-6
Knight PJK Knowles BH Ellar DJ Molecular cloning of an insect aminopeptidase N that serves as a receptor for Bacillus thuringiensis CryIA(c) toxin J Biol Chem 1995 270 17765 17770 7629076 10.1074/jbc.270.30.17765
Luo K McLachlin JR Brown MR Adang MJ Expression of a glycosyl phosphatidylinositol-linked Manduca sexta aminopeptidase N in insect cells Protein Expr Purif 1999 17 113 122 10497076 10.1006/prep.1999.1122
Agrawal N Malhotra P Bhatnagar RK Interaction of gene-cloned and insect cell-expressed Aminopeptidase N of Spodoptera litura with insecticidal crystal protein Cry1C Appl Environ Microbiol 2002 68 4583 4592 12200317 10.1128/AEM.68.9.4583-4592.2002
Rajagopal R Sivakumar S Agrawal N Malhotra P Bhatnagar RK Silencing of midgut aminopeptidase N of Spodoptera litura by double-stranded RNA establishes its role as Bacillus thuringiensis toxin receptor J Biol Chem 2002 277 46849 46851 12377776 10.1074/jbc.C200523200
Herrero S Gonzalez Cabrera J Ferré J Bakker PL de Maagd RA Mutations in the Bacillus thuringiensis Cry1Ca toxin demonstrate the role of domains II and III in specificity towards Spodoptera exigua larvae Biochem J 2004 384 507 513 15320864 10.1042/BJ20041094
Diatchenko L Lau YFC Campbell AP Chenchik A Moqadam F Huang B Lukyanov S Lukyanov K Gurskaya N Sverdlov ED Suppression subtractive hybridization: A method for generating differentially regulated or tissue-specific cDNA probes and libraries Proc Natl Acad Sci USA 1996 93 6025 6030 8650213 10.1073/pnas.93.12.6025
Altschul SF Madden TL Schaffer AA Zhang J Zhang Z Miller W Lipman DJ Gapped BLAST and PSI-BLAST: a new generation of protein database search programs Nucl Acids Res 1997 25 3389 3402 9254694 10.1093/nar/25.17.3389
Thompson JD Gibson TJ Plewniak F Jeanmougin F Higgins DG The CLUSTAL_X windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools Nucleic Acids Res 1997 25 4876 4882 9396791 10.1093/nar/25.24.4876
Saitou N Nei M The neighbor-joining method: a new method for reconstructing phylogenetic trees Mol Biol Evol 1987 4 406 425 3447015
Kimura M The Neutral Theory of Molecular Evolution 1983 Cambridge, England
|
15978131
|
PMC1184072
|
CC BY
|
2021-01-04 16:32:48
|
no
|
BMC Genomics. 2005 Jun 24; 6:96
|
utf-8
|
BMC Genomics
| 2,005 |
10.1186/1471-2164-6-96
|
oa_comm
|
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-5-491596976410.1186/1471-2334-5-49Case ReportTreatment failure in a typhoid patient infected with nalidixic acid resistant S. enterica serovar Typhi with reduced susceptibility to Ciprofloxacin: a case report from Cameroon Nkemngu Njinkeng J [email protected] Etienne DN [email protected] Anna L [email protected] St. John's Maternity and Hospital, Fiango, Kumba, Cameroon2 St. Joseph's medical Centre, Yoke-Muyuka, Cameroon3 University Teaching Hospital, University of Yaounde I, Yaounde, Cameroon4 Faculty of Health Sciences, University of Buea, Buea, Cameroon2005 21 6 2005 5 49 49 15 2 2005 21 6 2005 Copyright © 2005 Nkemngu et al; licensee BioMed Central Ltd.2005Nkemngu et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Fluoroquinolones or third generation cephalosporins are the drugs of choice for the treatment of typhoid fever. Treatment failure with fluoroquinolones has been reported in Asia and Europe. We report a case of ciprofloxacin treatment failure in typhoid fever in Cameroon.
Case presentation
A 29-year-old female patient with suspected typhoid fever from Kumba, Cameroon, yielded growth of Salmonella enterica serovar Typhi in blood culture. The isolate was resistant to nalidixic acid but sensitive to ciprofloxacin by disc diffusion test. However, the patient did not respond to treatment with ciprofloxacin, although the isolate was apparently susceptible to ciprofloxacin.
Conclusion
Treatment failure with ciprofloxacin in our case indicates the presence of nalidixic acid resistant S. enterica serovar Typhi (NARST) with reduced susceptibility to ciprofloxacin in Cameroon (Central Africa).
==== Body
Case presentation
A 29-year-old woman from Kumba, Cameroon, was admitted in January 2004 to St. John's Hospital and Maternity, Kumba, with a five-day history of fever, emesis, poorly localized abdominal discomfort, myalgias and hepatosplenomegaly. Her total leukocyte count was 1.7 × 109/l, (neutrophils 51%, lymphocytes 43%, monocytes 5%). Urinalysis was normal and thin and thick film examinations of the peripheral blood were negative for malaria. The patient also tested negative for HIV. A blood Widal test however, showed a titre of 80 against "O" (somatic) antigen and 160 against the "H" (flagella) antigen of Salmonella enterica serovar Typhi (recommended cut-off titre in our hospital: ≥ 1:80 and ≥ 1:160 for the "O" and "H" antigens respectively). Blood culture grew Salmonella enterica serovar Typhi. Two months prior to her illness, she had suffered from an attack of suspected typhoid fever and had been treated with chloramphenicol 500 mg every 6 hours for 14 days.
Antibiogram of the isolated S. enterica serovar Typhi was performed by disc diffusion techniques as recommended by NCCLS guidelines [1], Minimum inhibitory concentrations (MIC) of nalidixic acid and ciprofloxacin were determined by agar dilution method [2]. The antibiotic discs used included ampicillin 10 μg (Beecham), co-trimoxazole 1.25/23.75 μg (Roche), chloramphenicol 30 μg (Antibioticos SA), ciprofloxacin 5 μg (Bayer), nalidixic acid 30 μg (Sigma) and ceftriaxone 30 μg (Roche). The isolate was found resistant to nalidixic acid, ampicillin, co-trimoxazole and chloramphenicol, but susceptible to ceftriaxone and ciprofloxacin by disc diffusion test. The MICs of ciprofloxacin and nalidixic acid were 0.5 μg/ml and 32.0 μg/ml respectively. The patient remained febrile after 7 days of oral administration of 500 mg ciprofloxacin every 12 hours. Thereafter, the patient was administered 1 g ceftriaxone every 12 hours intravenously, which rendered her afebrile within four days. Treatment was continued for another 3 days. The patient did not relapse on follow-up.
Conclusion
The emergence of multi-drug-resistant Salmonella enterica serovar Typhi (MDRST) strains was first reported in the 80 s, in Asia. Sporadic cases of ciprofloxacin treatment failure in typhoid fever have been reported in Europe and more recently, in Asia [3,4]. Our report indicates that MDRST and nalidixic acid resistant Salmonella enterica serovar Typhi (NARST) strains are now appearing in Cameroon, Central Africa. NARST have also been reported in East Africa [5]. However, treatment failure with fluoroquinolones in patients affected by the NARST strains in East Africa has not been described, although several reports suggest that the clinical response to fluoroquinolones in patients infected with NARST may be inferior to the response in those infected with nalidixic acid-susceptible strains [[4-11], this report].
There may be single or multi-mutations in the quinolone-resistance-determining region of either DNA gyrase (gyrA or gyrB or both) or DNA topoimerase IV (parC and parE or both) or both enzymes, which cause resistance of Salmonella enterica serovar Typhi strains to fluoroquinolone [5,9]. Resistance may also be due to other mechanisms such as decreased permeability and active efflux of the antimicrobial agents. Previous studies have shown that MDRST strains in East Africa were related to earlier drug-susceptible isolates but were unrelated to MDRST isolates from Asia. [10]. MDRST and NARST isolates in Central African may be unrelated to those earlier reported in Asia, Europe and east Africa [3-11].
Ceftriaxone is an alternative drug in cases of quinolone resistant typhoid fever. However, there have been reports of high-level resistance to ceftriaxone (MIC= 64 mg/l) in both Salmonella enterica serovar Typhi and Paratyphi A [3]. Third generation cephalosporins are also expensive (a treatment course with parental Ceftriaxone is six times more expensive compared to oral ciprofloxacin in Cameroon), and regularly not available. The efficacy of azithromycin, which was recently shown to be an effective alternative treatment for uncomplicated enteric fever due to MDRST, needs to be confirmed in patients with typhoid fever due to a NARST strain [12]. Improved hygienic conditions and effective surveillance methods to monitor newly emerged MDRST and NARST strains in Africa and other enteric fever endemic regions are of utmost importance.
Authors' contributions
NJN conceived and coordinated of the study, drafted the manuscript, analyzed the microbial tests results. EDNA helped in the drafting of the manuscript and analysis of microbial test results. ALN helped to draft the manuscript. All authors read and approved the final manuscript.
Competing interests
The author(s) declare that they have no competing interests.
Financial support
None
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We thank T. Fosi-Mbantenkhu, and Ngang K. Che for critical reading of the manuscript. Written consent was obtained from the patient for publication of the study.
==== Refs
National Committee for Clinical Laboratory Standards Performance standards for antimicrobial disk susceptibility test National Committee for Clinical Laboratory Standards, Wayne, Pa 1999 Ninth information supplement M100-S9
National Committee for Clinical Laboratory Standards Methods for dilution antimicrobial susceptibility test for bacterial that grow aerobically Approved standard M7-A5 National Committee forclinical Laboratory Standards, Wayne, Pa 2000
Parry CM Hien TT Dougan G White NJ Farrar JJ Typhoid fever N Engl J Med 2002 347 1770 1782 12456854 10.1056/NEJMra020201
Butt Tariq Ahmad Rifat Nadeem Mahmood Abid Zaidi Sabeen Ciprofloxacin Treatment Failure in Typhoid Fever Case, Pakistan Emerging Infectious Diseases 2003 9 1621 1622 14720407
Kariuki S Revathi G Muyodi J Mwituria J Munyalo A Mirza S Hart CA Characterization of multidrug-resistant typhoid outbreaks in Kenya Clin Microbiol 2004 42 1477 1482 10.1128/JCM.42.4.1477-1482.2004
Hakanen A Kotilainen P Jalava J Siitonen A Detection of decreased fluoroquinolone susceptibility and validation of nalidixic acid screening test J Clin Microbiol 1999 37 3572 3577 10523554
Threlfall EJ Ward LR Decreased susceptibility to ciprofloxacin in Salmonella enterica serotype Typhi, United Kingdom Emerg Infect Dis 2001 7 448 450 11384525
Mehta G Randhawa VS Mohapatra NP Intermediate susceptibility to ciprofloxacin in Salmonella enterica serovar Typhi strains in India Eur J Clin Microbiol Infect Dis 2001 20 760 761 11757985 10.1007/s100960100603
Wain J Hoa NT Chinh NT Quinolone-resistant Salmonella typhi in Viet Nam: molecular basis of resistance and clinical response to treatment Clin Infect Dis 1997 25 1404 10 9431387
Kariuki S Gilks C Revathi G Hart CA Genotypic Analysis of Multidrug-Resistant Salmonella enterica Serovar Typhi, Kenya Emerg Infect Dis 2000 6 649 651 11076726
Hakanen A Kotilainen P Huovinen P Helenius H Siitonen A Reduced fluoroquinolone susceptibility in Salmonella enterica serotypes in travelers returning from Southeast Asia Emerg Infect Dis 2001 7 996 1003 11747728
Chinh NT Parry CM LY NT A Randomized Controlled Comparison of Azithromycin and Ofloxacin for Treatment of Multidrug-Resistant or Nalidixic Acid-Resistant Enteric Fever Antimicrobial Agents Chemother 2000 44 1855 1859 10.1128/AAC.44.7.1855-1859.2000
|
15969764
|
PMC1184073
|
CC BY
|
2021-01-04 16:28:15
|
no
|
BMC Infect Dis. 2005 Jun 21; 5:49
|
utf-8
|
BMC Infect Dis
| 2,005 |
10.1186/1471-2334-5-49
|
oa_comm
|
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-5-521598518310.1186/1471-2334-5-52Research ArticleTransmission dynamics of rabies virus in Thailand: Implications for disease control Denduangboripant Jessada [email protected] Supaporn [email protected] Boonlert [email protected] Nipada [email protected] Wirongrong [email protected] Apirom [email protected] Thiravat [email protected] Department of Biology, Faculty of Science, Chulalongkorn University, Bangkok, Thailand2 Molecular Biology Laboratory for Neurological Diseases, Chulalongkorn University Hospital, Bangkok, Thailand3 Queen Saovabha Memorial Institute, Thai Red Cross Society, Bangkok, Thailand4 Department of Biology, Faculty of Science, Chulalongkorn University, Bangkok, Thailand5 Department of Livestock Development, Ministry of Agriculture, Bangkok, Thailand6 Department of Disease Control, Ministry of Public Health, Nonthaburi, Thailand7 Molecular Biology Laboratory for Neurological Diseases, Chulalongkorn University Hospital, Bangkok, Thailand2005 29 6 2005 5 52 52 2 5 2005 29 6 2005 Copyright © 2005 Denduangboripant et al; licensee BioMed Central Ltd.2005Denduangboripant et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
In Thailand, rabies remains a neglected disease with authorities continuing to rely on human death statistics while ignoring the financial burden resulting from an enormous increase in post-exposure prophylaxis. Past attempts to conduct a mass dog vaccination and sterilization program have been limited to Bangkok city and have not been successful. We have used molecular epidemiology to define geographic localization of rabies virus phylogroups and their pattern of spread in Thailand.
Methods
We analyzed 239 nucleoprotein gene sequences from animal and human brain samples collected from all over Thailand between 1998 and 2002. We then reconstructed a phylogenetic tree correlating these data with geographical information.
Results
All sequences formed a monophyletic tree of 2 distinct phylogroups, TH1 and TH2. Three subgroups were identified in the TH1 subgroup and were distributed in the middle region of the country. Eight subgroups of TH2 viruses were identified widely distributed throughout the country overlapping the TH1 territory. There was a correlation between human-dependent transportation routes and the distribution of virus.
Conclusion
Inter-regional migration paths of the viruses might be correlated with translocation of dogs associated with humans. Interconnecting factors between human socioeconomic and population density might determine the transmission dynamics of virus in a rural-to-urban polarity. The presence of 2 or more rabies virus groups in a location might be indicative of a gene flow, reflecting a translocation of dogs within such region and adjacent areas. Different approaches may be required for rabies control based on the homo- or heterogeneity of the virus. Areas containing homogeneous virus populations should be targeted first. Control of dog movement associated with humans is essential.
==== Body
Background
Rabies is not high on the list of the World Heath Organization's list of important infectious diseases, and is also often overlooked by regional, national, and local public-health professionals. The dog is the primary reservoir and vector of rabies transmission in Thailand and developing countries [1].
To date, the evaluation of the importance of rabies has been determined solely by estimating the number of human deaths and statistics on dog rabies infectivity, which may not be a reliable indicator in developing countries [2]. For example, an accurate assessment of the burden of rabies will never be complete without including the financial burden incurred due to human rabies post-exposure prophylaxis (PEP) and animal control.
In Thailand, the substantial decline in human rabies deaths from almost 200 a decade ago to less than 20 in 2003, has occurred due to the huge and continuously escalating financial obligation in the annual budget required to supply rabies biologicals for human PEP. More than 400,000 patients received PEP in 2003, as compared to approximately 90,000 in 1991 [Ministry of Public Health (MOPH) annual report]. Moreover, annual human rabies deaths in Bangkok, where diagnostic facilities and neurologists are readily available, rose from less than 5 in 1990–1994 to 5–10 in 1995–2001 (MOPH annual report).
There are no reliable statistical analyses of dog populations that could be evaluated to determine the effectiveness of the current human rabies prevention methodologies used in Thailand. One quoted figure of 6 million dogs in Thailand is undoubtedly an underestimate of the actual population present within the country. The Division of Disease Control and Ministry of Agriculture reported that between 60 to 78% of the dog population was vaccinated (based on estimated total population) in Thailand between 1995 and 2000. Experience in Latin America has shown that vaccination of a critical percentage of dogs, on the order of 40–70%, at least in major urban areas, was sufficient to interrupt canine rabies transmission and resulted in diminished human rabies deaths [3]. However, this has not been the case in Thailand. The percentage of rabies infectivity of samples sent to diagnostic laboratories all over the country remains high, within the range of 30–40% (MOPH annual report).
A survey in 1999 by the Department of Livestock and the Bangkok Metropolitan Administration revealed that stray dog populations in Bangkok (an area of 1,565 sq km) have tripled in size, (from 40,756 in 1992 to 110,584 in 1999). Additionally, a 2002 survey suggested that dog populations were increasing, both in Bangkok and the countrywide, implying that the specific carrying capacity of canine habitats has not yet been saturated. Moreover, a substantial number of dogs, especially stray and community dogs, are not vaccinated.
Due to budget limitation, an intensive dog vaccination and sterilization program has been in place only in Bangkok City since June 2002. Seventy-two million baht (approximately US$ 1,800,000) were spent during the first 2 phases of the program (June 2002-September 2003), with the third phase (October 2003-September 2004) costing an additional 31 million baht (approximately US$ 775,000). Although there were no human rabies deaths in Bangkok in 2002, 3 deaths were reported in 2003. Preliminary assessment revealed that less than 20 percent of the estimated dog population was sterilized and vaccinated.
Without reliable data on dog ecology and surveillance of rabies infection in dogs and humans, it is not possible to develop a strategic plan for rabies prevention and control and to assess program success. Therefore, our objective was to use molecular biological techniques to characterize the presence and movement of rabies virus according to geographical locations in Thailand and use this information as baseline data to design and implement rabies prevention programs in the country. Areas with evidence of continuous gene flow, and presence of viruses of more than one genetic group or subclade, were characterized. The potential translocation of rabies virus from one area to another was evaluated in relation to natural barriers, transportation routes, human activity and socioeconomic factors.
Methods
Samples
Two hundred and thirty nine brain samples (7 humans, 7 cats, 216 dogs, 6 cattle, 1 water buffalo, 2 squirrels) from 56 provinces were obtained from 25 diagnostic laboratories all over Thailand between 1998 and 2002. Samples selected for analysis were chosen to be representative of the geographical location in each province down to the scale of small districts (subdivisions of a province). Samples were not available from 20 provinces. All samples were prescreened for evidence of rabies virus using the direct fluorescence antibody test and kept frozen at -80 degree C until genetic analysis was conducted.
Genetic analysis
Genetic typing was based on nucleotide sequence differences in cDNA obtained by direct one step RT-PCR amplification of the nucleocapsid (N) gene fragment from the samples. The amplified products of 414 bp (nt 1101 – 1506) were characterized by sequencing. RT-PCR and sequencing procedures were conducted as previously described [4]. One set of primers was used for RT-PCR sequencing reaction. GenBank accession numbers of the N sequences in this study were AY849022-AY849260 (see Additional file 1).
Twelve additional N sequences were retrieved from Genbank database to be outgroups for this study: a non-rabies lyssavirus Mokola Virus (S59448), 3 Australian Bat Lyssavirus (ABLV) isolates (NC003243, AF081020, AF418014), a rabies strain Pasteur Virus (PV) (M13215), 6 rabies viruses from other Asian countries (AY138550 from Sri Lanka, AY138551 Sri Lanka, AY138549 Sri Lanka, AF155039 China, AF374721 India, U22482 Iran) and 1 rabies isolate from Thailand (U22653). The sequences of all isolates were aligned together using program ClustalX [5]. Genetic relationships between these N gene sequences were calculated and a tree diagram was drawn using neighbor-joining (NJ) method, which was suitable to illustrate below species-level genetic relationships. These phylogenetic analyses were performed with program PAUP* version 4.0b10 [6]. Robustness of the tree was accessed with branch supporting-values from bootstrap (BS) statistic analyses (1,000 replicates). The collecting provinces and districts of all virus samples were mapped on the trees. Geographical locations of samples were mapped (Arcview 3.2, ESRI) and compared among the subgroups.
Results
Phylogenetic analyses of the 239 N rabies sequences collected from all over Thailand clearly demonstrated that all of the isolates formed a monophyletic group with 100% boostrap supporting values, separate from Mokola Virus, Australian Bat Lyssavirus (ABLV), Pasteur Virus (PV) and other Asian rabies viruses (Fig. 1). The N sequences from India and Sri Lanka were weakly grouped with those of Thai rabies virus with 51% bootstrap support. Neither the sequences of infected human nor non-canine animals (cats and other wildlife) were specifically clustered as a unique group, but rather paired with the dog rabies viruses analyzed in the study. Two major viral groups were clearly recognized from the tree and designated as TH1 and TH2 clusters, with 82% and 68% bootstrap supporting values, respectively.
Figure 1 Comparison between NJ tree of Thai rabies N genes and geographical distribution map. Neighbor-joining (NJ) tree based on 414 bp nucleotide sequences of the N genes of all 239 Thai rabies virus isolates compared with other 11 lyssavirus outgroups. Numbers along tree branches are >50% bootstrap supporting-value (1,000 replicates). The map of Thailand indicates geographical distributions of the 2 major phylogroups, TH1 and TH2, in a district-level (a subdivision of a province).
Considering sampling locations of viral isolates, both Thai rabies phylogroups were confined to certain geographical areas, though overlapping did occur in some areas (see the map of Thailand in Fig. 1). TH1 viruses were found mainly in the middle part of the country, from the lower northern region to the central region, Bangkok and surrounding provinces, and the upper southern region of Thailand. Based on BS supporting values on each branch, branch lengths (equally to numbers of substitutions/site), and the tree topology, the TH1 phylogroup was divided into 3 minor subgroups (Fig. 2): TH1A isolates were identified in Bangkok and outskirts as well as some other provinces in the central region; TH1B isolates were identified in the northern and central regions, and TH1C isolates were identified in Bangkok, the central, and the upper southern regions.
Figure 2 Comparison between NJ tree of TH1 rabies sequences and the distribution map. A comparison between the NJ tree of 60 N gene sequences of TH1 rabies viruses (with the TH2 isolate 703KKm added as an outgroup) and the Thailand map indicates geographical distributions of the subgroups TH1A, TH1B, and TH1C.
The TH2 rabies phylogroup was distributed in much wider areas than the TH1 phylogroup (Fig. 1), from the northern to the southern-most regions of the country. The distribution areas of TH2 group covered almost every province in the northeastern region, all main provinces in the upper central and the central regions, Bangkok and 5 surrounding provinces, the eastern and western regions, and nearly the entire area of southern Thailand. Using similar criteria as in the TH1 group, TH2 was divided into 8 subgroups: TH2A (Fig. 3) with samples from the northeastern region, TH2B (Fig. 3) with samples from the south and some northeastern provinces, TH2C (Fig. 4) from a few provinces scattered in the east, upper central, and northeast, TH2D (Fig. 4) from western provinces, TH2E (Fig. 4) had a much wider distribution-range including the far north, northeast, central including provinces around Bangkok, to the upper south, TH2F (Fig. 4) and TH2G (Fig. 5) were mainly located in the northeastern regions, and TH2H (Fig. 5) was found in the lower north to the upper central.
Figure 3 Comparison between NJ tree of TH2A and TH2B rabies sequences and the distribution map. A comparison between the bottom part of the NJ tree of TH2 rabies viruses (with the TH1 isolate 125SSktb added as an outgroup) and the Thailand map indicates geographical distributions of the subgroups TH2A and TH2B.
Figure 4 Comparison between NJ tree of TH2C, TH2D, TH2E, and TH2F rabies sequences and the distribution map. A comparison between the middle part of the NJ tree of TH2 rabies viruses and the Thailand map indicates geographical distributions of the subgroups TH2C, TH2D, TH2E, and TH2F.
Figure 5 Comparison between NJ tree of TH2G and TH2H rabies sequences and the distribution map. A comparison between the top part of the NJ tree of TH2 rabies viruses and the Thailand map indicates geographical distributions of the subgroups TH2G and TH2H.
Discussion
The use of molecular biological techniques to evaluate the epidemiology of viral diseases is being increasingly employed to complement conventional methods [[7-11], for examples of rabies epidemiology]. These techniques can give a clearer understanding of the origination and transmission patterns of viral epidemics. Eventually, data produced from molecular epidemiological studies could lead to a better understanding of and a more effective strategy to control the spread of infectious diseases.
Our study revealed that all of the currently identified Thai rabies viruses share a common origin that is genetically distant from the PV, ABLV, and Mokola outgroups. Additionally, the monophyletic tree of the Thai rabies viruses analyzed in this study was clearly distinguishable from other rabies N sequences from India, Sri Lanka, China, and Iran (Fig. 1). Thus, rabies viruses circulating in Thailand (or in Southeast Asia) could possibly have an exclusive evolutionary background that might be recognized as being unique, an hypothesis previously suggested by Susetya et al. [12]. It will be necessary to analyze additional sequences of rabies viruses circulating in neighboring countries adjacent to Thailand to confirm this hypothesis.
The NJ genetic distance tree also confirmed that the sequences obtained from non-canine sources (human, cats and other mammals) were very similar to those obtained from rabid dogs. No specific grouping of sequences from rabies virus isolated from non-canine species was identified. Instead, these rabies virus sequences were scattered across the tree. This finding was in accord to our expectation that the dog is a prime reservoir and transmitting vector for rabies and causes spillover to human and domestic animals and wildlife. Nevertheless, we are also aware that there may be other vectors, such as bats, and other lyssaviruses, besides genotype 1, circulating in Thailand. In fact, our recent survey in Thai bats indicated that as many as 7.5% of the bat population had evidence of lyssavirus infection by an as yet unidentified genotype(s) [13].
The 2 major groups found in our Thai rabies phylogeny were judged to be significant with high bootstrap supporting values. Notably, these 2 major lineages resembled the putative groups A and B found in our previous study [14] in which fewer numbers of samples from Bangkok and its surrounding provinces were analyzed. The 2 phylogroups we identified had certain trends in their geographical distributions. The distribution areas of TH1 group were only found in the central part of the country – from Nakhon Sawan province, down along Choa Praya river to the capital city of Bangkok, ending at Ranong province in the upper southern region (Fig. 2). On the other hand, those of TH2 group were spread across more than three-quarters of the entire country – from Phayao province in the north (Fig. 4), to Ubol Ratchathani in the northeast corner (Fig. 5), and to the southern Yala province along the Thailand-Malaysia border (Fig. 3).
Although there are some overlapping areas shared between the TH1 and TH2 phylogroups, the viral transmission dynamics and evolutionary background the sub-lineages may not be similar which could explain why both have different success levels in disease dispersals. It has been proposed that the degree of differences in compartmentalization mechanisms may influence the duration that each individual canine-associated rabies variant resides in certain geographical regions [15]. Relationships between dogs and humans within a community, dog population density, and relative dog-human population ratio are common explanations for such compartmentalization phenomenon [16,17]. Local geographical barriers such as rivers and mountains are other important factors considered to have strong influences on the inhibition of the spread of vector-borne diseases [18]. This inhibition effect caused by natural barriers could, however, be compromised by human transportation routes, for instance bridges, or roads and railways through mountains.
To estimate the epidemiological characteristics of the TH1 and TH2 Thai rabies groups, a phylogeographical concept was introduced to infer their transmission dynamics [19,20]. Comparison between the sampling localities and molecular phylogenetic-tree topology could determine how viruses in each and different group are genetically related. First, in Bangkok and the surrounding provinces both TH1 and TH2 were identified as occurring together (Fig. 1). This area is industrialized and highly populated. Hundreds of mainroads, highways, and railways have been built in the country heartland. Networks of transportation routes including bridges across waterways are an effective means for vector borne viral transmission. This is one explanation as to why some viral subgroups of both TH1 and TH2 were discovered along both sides of Choa Phraya and many other rivers (Fig. 6), as has been previously reported [14].
Figure 6 Comparison between geographical distribution of rabies viruses in Thailand and in Kanchanaburi province. Geographical distribution of all Thai rabies virus subgroups. Kanchanaburi province was magnified to show province geography and roadmap. Red areas in the Kanchanaburi map indicate the collecting localities of rabies hosts in a tambon (a subdivision of a district)-level. The map was retrieved from
Secondly, we suggest that transmission of rabies virus may be related to human activity, particularly human migration. Considering the phylogeographic areas of the 3 genetic subgroups of TH1 rabies virus (Fig. 2), genetic exchanges within the TH1B subgroup between Sukhothai in the north and Nonthaburi province near Bangkok, almost 500-kilometres apart, could not have been accomplished by migrations of animal virus-vectors alone. It is more likely that canine vectors of the TH1B genotype were translocated from areas around Bangkok to the north, and vice versa, simply by following movements of pet-owners via the national mainroads number 1 and 11 (Fig. 2). Similarly, the same translocation factors can be applied to a long-distance dispersal of the TH1C subgroup from central to southern Thailand, probably via the national mainroad number 4 (Fig. 2). Transmission dynamics of the TH1 subgroup might also have been influenced by a combination of factors including social and socioeconomic status, human and animal population density in addition to the availability of transportation routes.
The theory that the spread of canine rabies virus was instigated by pet-owner translocation via transportation routes was supported in this study by the results of the distribution pattern of each subgroup of TH2 (Figs. 3, 4, 5). Members of the TH2 group appeared to be scattered across the regions at a very distant range, and are unlikely to have occurred due to animal self-translocation. From our analyses, we propose that genetically linked viruses of each subgroup were localized in specific areas by utilizing transportation routes throughout Thailand (as shown in Fig. 3, 4, and 5), and areas that have more than one viral group present are apparently local transportation, for instance, Mueng district (Khon Kaen province), Pa Kham district (Buri Ram province) and Phimi district (Nakhon Ratchasima province) (shown as black areas in Fig. 3).
The most convincing support for the human-facilitated rabies distribution hypothesis we propose herein is the geographical distribution of the TH2B subgroup in which all, except a few samples from the northeast, were from the southern region of the country. This phylogeographic subgroup with a 1300–1600 km spreading range, had a very strong bootstrap supporting-value (89%) on the genetic tree. We propose that this inter-regional migration path of the TH2 subgroup is explained by a rural-to-urban viral transmission polarity. [12,16] The majority of people in the northeast have a relatively lower socioeconomic status than people living in other regions. Most of them are conventional crop farmers with low annual income [9,279 Baht (approximately US$ 230) average monthly household income versus national average of 13,736 Baht (approximately US$ 340), reported by National Statistical Office on 2002] and during the off-growing season they usually migrate to other regions to seek employment as common laborers. The strong economics in southern Thailand has been mainly supported by marine fishery as well as the rubber plant and oil palm agricultural industry, of which most workers originate from northeastern Thailand. Rabies virus infected canine pets accompanying the migratory workers from the northeastern therefore could be spread along their owners' travel routes. This would not only explain the northeast-to-south migration path of the TH2B viruses, but also could elucidate why most of the TH2 subgroups examined were closely linked with viral isolates from the northeastern region.
Selection of suitable areas using molecular epidemiological techniques should be considered as a powerful tool when planning disease control strategies. For decades, Thailand has invested vast sums of money and manpower on the effort to control and vaccinate the dog population in randomly selected districts and, recently, Bangkok capital city without success. Results of this research demonstrated that Bangkok and other metropolitan cities (such as Prathum Thani, Samut Sakhon, Nakhon Sawan, Khon Kaen, Ubol Ratchathani) contain various groups and subgroups of viruses, actively circulating to and from other surrounding provinces (Fig. 6). Therefore, developing a campaign for disease control in such city alone, without considering neighboring areas, is highly unlikely to be successful. We propose that the most appropriate place to initiate a rabies control campaign should be in a genetically isolated area, where there are either natural or artificial barriers to prevent further viral influxes.
On a national scale, we propose that rabies control can be successful if it is initiated in southern Thailand. This region contains only the TH2B rabies subgroup. Furthermore, it is an "island-like" area surrounding by Andaman Sea, Gulf of Thailand, and the Malaysian border. Influx from the TH1C subgroup has been restricted to an area around Ranong province, plausibly from high mountain ridges. Moreover, the majority of the population in southern Thailand are Muslims who do not keep pet dogs or feed stray dogs. Implementation of a rabies control in this region should therefore be effective in terms of a dog population reduction and vaccination campaign, and the enforcement of strict regulations regarding dog transfer.
In order to test this concept of "targeting a genetically defined area", a mass rabies control compaign should be conducted in a suitable-size province with a homogeneous virus population. The province of Kanchanaburi, 19,483 km sq and about 130 km westwards from Bangkok, appears to be a good choice since, according to our study, it contains only the TH2D rabies subgroups (Fig. 6) which are clustered mostly on the southernmost tambons (subdistricts; subdivisions of an district). The province is also island-like in that it is surrounded by the mountainous Thailand-Myanmar border and also has mountain ridges along the eastern boundary to other provinces. Any strategic plan in this region should also include recommendations to control pet-dog movements via the national mainroad number 323, the primary transportation route into the province. Moreover, in order to control rabies situation, at least 50 – 70% of dogs must be vaccinated. Currently available vaccine used in Thailand is injectable type, thus, requiring a capturing or restraining process which is extremely difficult especially in the case of community or stray dogs. Oral type vaccine such as that used in wildlife once proven of its safety and efficacy in dogs may be an alternative. Public participation in dog population control and vaccination needs to be created. Intensive and extensive educational activities should be carried out to increase understanding of the necessity to have rabies and dog population control program implemented [21]. Assessment of the success of such a program can be measured by a strict surveillance of rabies incidence in humans and animals and by analyzing genetic sequences of rabies virus as compared to others in adjacent provinces. This should also be correlated with transportation tracks on a local scale.
Conclusion
In conclusion, we have presented a novel approach to the development of a rabies control and prevention program through the utilization of genetic epidemiology. We believe that the implementation of such a disease control program utilizing existing information on the genetics of circulating rabies viruses in a country like Thailand could be successful if the campaign target areas have been carefully selected and limited to one circulating phylogroup of virus and the movement of dogs along human transportation routes into the area is strictly enforced.
Competing interests
The authors declare that they have no financial or personal relationships with other people or organizations that could inappropriately influence this research. All authors have access to all data in the study and held final responsibility for the decision to submit for publication.
Authors' contributions
JD participated in data analysis and interpretation, phylogenetic tree construction, and writing the paper. SW participated in data collection and analysis, PCR primer design, sequencing rabies genes, and writing the paper. BL participated in specimen collection, data analysis, and writing the paper. NR participated in data analysis, figure preparation, and writing the paper. WH participated in coordinating the study with the veterinarians throughout the country, data analysis, and writing the paper. AP participated in coordinating the study with the physicians throughout the country, data analysis, and writing the paper. TH participated in study design, data analysis and interpretation, and writing the paper. The first 3 authors (JD, SW, BL) contributed equally to this work. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
Table 1 Taxon list of N-gene sequences of Thai rabies virus used in this study
Click here for file
Acknowledgements
The authors are indebted to Deborah Briggs for her critical and useful comments on the manuscript. We also would like to thank these laboratories which supplied rabies specimens for this study: Southern Veterinary Research and Development Center, Nakhon Si Thammarat; Northern Veterinary Research and Development Center, Phitsanulok; Northern Veterinary Research and Development Center, Lampang; North Easthern Veterinary Research and Development Center, Khon Kaen; North Easthern Veterinary Research and Development Center, Surin; Regional Bureaus of Animal Health and Sanitary no.3, no.4, no.6, no.7, no.8, and no.9; Provincial Livestock Offices (PLOs): Chaiyaphum, Nakhon Ratchasima, Chai Nat, Phetchabun, Kamphaeng Phet, Surat Thani, Amnat Charoen, Si Sa Ket, Udon Thani, Sakon Nakhon, and Kalasin; Department of Medical Science, Ministry of Public Health; and Regional Medical Science Center, Nakhon Ratchasima. This work was supported in part by grants from Thailand Research Fund (DBG/01/2545) and National Science and Technology Development Agency.
==== Refs
Mitmoonpitak C Wilde H Tepsumetanon W Current status of animal rabies in Thailand J Vet Med Sci 1997 59 457 60 9234220 10.1292/jvms.59.457
Hemachudha T Laothamatas J Rupprecht CE Human rabies: a disease of complex neuropathogenetic mechanisms and diagnostic challenges Lancet Neurol 2002 1 101 9 12849514 10.1016/S1474-4422(02)00041-8
PAHO Epidemiological surveillance of rabies in the Americas 2000 Washington DC: Pan American Health Organization
Hemachudha T Wacharapluesadee S Lumlertdaecha B Orciari LA Rupprecht CE La-ongpant M Juntrakul S Denduangboripant J Sequence analysis of rabies virus in humans exhibiting encephalitic or paralytic rabies J Infect Dis 2003 188 960 6 14513414 10.1086/378415
Thompson JD Gibson TJ Plewniak F Jeanmougin F Higgins DG The ClustalX windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools Nucleic Acids Res 1997 25 4876 82 9396791 10.1093/nar/25.24.4876
Swofford DL PAUP*: phylogenetic analysis using parsimony (* and other methods), version 4 2002 Sunderland (MA): Sinauer Associates
David D Yakobson B Smith JS Stram Y Molecular epidemiology of rabies virus isolates from Israel and other Middle- and Near-Eastern countries J Clin Microbiol 2000 38 755 62 10655381
De Mattos CA De Mattos CC Smith JS Miller ET Papo S Utera A Osburn BI Genetic characterization of rabies field isolates from Venezuela J Clin Microbiol 1996 34 1553 58 8735118
Miksza LKC Wadford DA Schnurr PD Molecular epidemiology of enzootic rabies in California J Clin Virol 1999 14 207 19 10614858 10.1016/S1386-6532(99)00054-2
Nadin-Davis SA Casey GA Wandeler AI A molecular epidemiological study of rabies virus in central Ontario and western Quebec J Gen Virol 1994 75 2575 83 7931145
Teichman BFV Thomson GR Meredith CD Nel LH Molecular epidemiology of rabies virus in South Africa: evidence for two distinct virus groups J Gen Virol 1995 76 73 82 7844544
Susetya H Sugiyama M Inagaki A Ito N Oraveerakul K Traiwanatham N Minamoto N Genetic characterization of rabies field isolates from Thailand Microbiol Immunol 2003 47 653 9 14584612
Lumlertdacha B Boongird K Wanghongsa S Wacharapluesadee S Chanhome L Khawplod P Hemachudha T Kuzmin I Rupprecht CE Survey of bat lyssaviruses in Thailand Emerg Infect Dis
Lumlertdacha B Wachrapluesadee S Denduangboripant J Ruankaew N Hoonsuwan W Puanghat A Sakarasaeranee P Briggs D Hemachudha T Complex genetic structure of rabies virus in Bangkok city and its surroundings: implications for canine rabies control Trans Roy Soc Trop Med Hyg
Smith JS Jackson AC, Wunner WH Molecular epidemiology Rabies 2002 Amsterdam: Academic Press 79 111
De Mattos CC De Mattos CA Loza-Rubio E Aguilar-Setien A Orciari LA Smith JS Molecular characterization of rabies virus isolates from Mexico: implications for transmission dynamics and human risk Am J Trop Med Hyg 1999 61 587 97 10548293
Holmes EC Woelk CH Kassis R Bourhy H Genetic constraints and the adaptive evolution of rabies virus in nature Virology 2002 292 247 57 11878928 10.1006/viro.2001.1271
Smith JS Orciari LA Yager PA Seidel HD Warner CK Epidemiologic and historical relationships among 87 isolates as determined by limited sequence analysis J Infect Dis 1992 166 296 307 1634801
Holmes EC The phylogeography of human viruses Mol Ecol 2004 13 745 56 15012753 10.1046/j.1365-294X.2003.02051.x
Avise JC Phylogeography: the history and formation of species 2000 London: Harvard University Press
Hemachudha T Rabies and dog population control in Thailand: Success or failure? J Med Assoc Thailand 2005 88 120 123
|
15985183
|
PMC1184074
|
CC BY
|
2021-01-04 16:28:13
|
no
|
BMC Infect Dis. 2005 Jun 29; 5:52
|
utf-8
|
BMC Infect Dis
| 2,005 |
10.1186/1471-2334-5-52
|
oa_comm
|
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-5-571600883810.1186/1471-2334-5-57Research ArticleHospitalization for pertussis: profiles and case costs by age O'Brien Judith A [email protected] J Jaime [email protected] Caro Research Institute, 336 Baker Avenue, Concord, MA, USA2 Division of General Internal Medicine, McGill University, 687 Pine Avenue, West, Montreal, Quebec, Canada2005 11 7 2005 5 57 57 10 1 2005 11 7 2005 Copyright © 2005 O'Brien and Caro; licensee BioMed Central Ltd.2005O'Brien and Caro; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Pertussis, a highly contagious respiratory illness, affects people of all ages and can have serious clinical consequences. It has been reported that from 1997–2000, 20% of all pertussis cases required hospitalization in the US. This analysis examined demographics, case fatality rate, resource use and costs of hospital care related to pertussis by age.
Methods
ICD-9 codes (033.0, 033.9) were used to identify cases of pertussis in hospital discharge databases from roughly 1,000 US hospitals in 4 states (California, Florida, Maryland, Massachusetts). Data from 1996–1999 were examined by age group. Separate analyses were done for infants (<1 year) and children (1–11 years); however, adolescent and adult cases were combined into one group (12+ years), due to the small number of cases. Databases were used to determine demographics, health service utilization and care costs. Cost estimates include accommodations, ancillary and physician services, reported in 2002 US$.
Results
Of the 2,518 cases identified, 90% were infants. The inpatient case fatality rate was <1%. Of survivors, 99% were discharged home (6% with home health care); 1% required further sub-acute inpatient care. For the 2,266 infants, the mean LOS was 6 days at a cost of $9,586 per stay. Children (n = 191) had a mean LOS of 3.7 and cost of $4,729; adolescents/adults (n = 61, mean age 40 years) stayed on average 3.4 days with a cost of $5,683 per hospitalization.
Conclusion
Infants are responsible for the bulk of hospitalizations and generate higher inpatient costs. Costly hospital care occurs, however, in patients with pertussis at all ages.
==== Body
Background
Bordetella pertussis is a highly contagious respiratory illness that can have serious clinical consequences. It is endemic in the United States (US) and every 3 – 5 years an outbreak occurs [1]. It is most commonly thought of as a childhood disease, not ranking high on the list of differential diagnoses for adolescents or adults presenting with respiratory symptoms. While this may be understandable as the reported incidence of pertussis declined by 99.6% between the mid-1930s to 1970s [2] due to effective childhood vaccination programs, clinicians need to be aware that pertussis continues to be a problem that is on the rise. Since the late 1980s, the incidence of reported pertussis has been increasing in most age groups in the US, particularly among adolescents and adults [1,3-8]. The exception was among young children where the reported incidence has remained stable and likely reflects protection resulting from pertussis vaccinations administered as recommended [9]. In 2002, the greatest number of pertussis cases (n = 9,771) since 1964 were reported in the US resulting in an incidence of 3.4 per 100,000. In the period from 1990 to 2001, a 400% increase in the adult incidence was noted [10] and this is likely an underestimation [11]. Infants continue to be the most vulnerable group. In 2002, 21% of all reported cases were infants too young (age < 6 months) to have started the recommended vaccination program [9]. Nationwide in that year, the reported incidence for infants less than six months old was 108.8 compared to 1.2 per 100,000 among those aged 20 years and older [9]. According to the Centers for Disease Control (CDC), pertussis is the only disease for which childhood vaccination is recommended that has had an increase in incidence in the past 20 years [10].
There is a common misconception vaccinating children for pertussis protects them forever. That is not the case. Immunity achieved through childhood vaccinations starts to diminish after three years with little protection left after twelve years [12-15]. Boosters are not recommended after seven years of age [2] so older adolescents and adults are vulnerable. Furthermore, those who contract pertussis later in life may go undiagnosed, but their sustained cough spreads the infection among family, friends, associates and health care workers [2,6,7,13,15-17].
Although outpatient management with antibiotics, usually erythromycin, is the expected treatment for pertussis [1], hospitalization does occur for those with complications [18-21]. According to the CDC, from 1997 – 2000, 20% of all patients with reported pertussis required hospitalization. This rate was much higher (63%) in infants younger than 6 months [22]. The objective of this analysis was to examine cases of pertussis admitted to US hospitals over a period of four years. Case fatality rates, complications, resource use and inpatient care costs by age group were studied.
Methods
The cost of hospital care was determined by analyzing inpatient resource use and costs reported in all payer discharge databases from four states representing different areas of the US. Discharge and cost data from roughly 1,000 hospitals in California, Florida, Massachusetts and Washington State for the years 1996 through 1999 were analyzed to identify cases of pertussis, defined as those admitted with an International Classification of Diseases – 9th Revision – Clinical Modification (ICD-9) code of either 033.0 or 033.9 [23] as the principal diagnosis for all age groups. These databases contain merged discharge-level demographic, clinical and economic data for all hospital discharges reported in each state within a given year for patients of all ages. Data on length of stay (LOS), charges incurred, source of admission and discharge disposition were abstracted for each patient. Secondary diagnosis codes reported in the databases were examined for complications of pneumonia in whooping cough (ICD-9 code: 484.3), other pneumonias (ICD-9 codes: 480.0–484.1, 484.4–486), convulsions (ICD-9 codes: 780.31, 780.38), apnea (ICD-9 codes: 770.8, 786.03), encephalopathy (ICD-9 codes: 348.3, 348.5) and acute respiratory distress, failure or arrest (ICD-9 codes: 518.0, 518.81, 518.82, 786.09, 799.0, 799.1).
The acute care hospital costs reported include all accommodations (e.g., routine, intensive care unit, nursery), ancillary (e.g., pharmacy, laboratory, imaging), and physician services. As costs pertaining to physicians' services are not included in the discharge data, it was necessary to estimate an average cost of inpatient physician services. This estimate was derived by determining the type and frequency of visit or service, applying the appropriate unit cost for each and multiplying it by the proportion of inpatients using that service. Explicit data regarding type of visit and services are not provided; therefore, data elements from the discharge databases providing information pertinent to physician care, such as LOS, procedure codes, special care unit stays, emergency department use prior to admission, admission source and discharge status were used to establish a likely inpatient physician visit/service profile for each age group. The appropriate Physicians' Current Procedural Terminology (CPT-4) code was assigned to each visit, service or procedure [24] in the profile. A unit cost representative of diverse payers was established using fee schedules [25,26] and published payer ratios [27], and applied to each CPT-4 code in the profiles. Where physician fee data from a single state were used (i.e., Florida Medicaid physician fees), state costs were adjusted to national values using published ratios [28].
The costs presented here are meant to reflect the economic value of the resources consumed, regardless of who actually pays for them. They represent only direct medical costs – those associated with the delivery of the health care service itself. Indirect costs, such as those due to a parent's lost wages when caring for a child with pertussis, are not included. All cost estimates are reported in 2002 US dollars. Values from previous years were inflated using rates based on the medical care component of the US Consumer Price Index, supplied by the Federal Bureau of Labor Statistics for the appropriate years [29]. Hospital charges for inpatient accommodations and ancillary services were adjusted to costs using a cost-to-charge ratio. There is no standard or national cost-to-charge ratio for hospital care that is reflective of care provided to the scope of patients included in this analysis – those of all ages and covered by different insurers. In the absence of a standard value, 0.61 was used based on a figure calculated by the Commonwealth of Massachusetts Office of Health Care Finance and Policy for hospitals in Massachusetts. The discharge data from each state was analyzed individually first, then these were combined giving equal weight to each state to avoid states with substantially larger populations from being overrepresented.
Cases were analyzed by age groups defined initially as: infants (aged less than 1 year); children (1 through 11 years); adolescents (12 through 17 years) and adults (age greater than 22 years); however, as the number of adolescent and adult cases was insufficient to analyze as separate groups, they were combined.
Seasonal occurrence was explored in terms of the month of hospitalization, except for Florida that reported only the quarter. Analyses by calendar quarter were also carried out for all cases.
Incidence estimates for the states represented in this analysis, as well as nationwide, for the years 1996 – 1999 were obtained from the CDC [30-33] or the individual state Departments of Health [34-37]. When a discrepancy occurred, the data from the source with the higher number of cases reported for that year was used.
Results
A total of 2,518 cases admitted to hospital for pertussis during the four-year period were identified (Table 1). Females accounted for 51% of the admissions. The mean age was 2.3 years (range: 0 – 83 years), but 90% of cases were younger than 1 year (median: <1 year) and most of the rest (8%) were for children. Admission via the emergency department was noted in 45% and 3% spent time in an observation unit prior to hospitalization.
Table 1 Summary of hospital stay results for pertussis by age group
Infants Children Adolescents & Adults
Pertussis cases: n (%) 2,266 (90) 191 (8) 61 (2)
Age range (years) <1 1 – 11 12 – 83
Mean age (years) * 3 40
Males (%) 50 37 46
Emergency Department (%) 45 40 45
Observation Unit Stay (%) 0 0 3
Mean Total Length of Stay (days) 6 4 3
Mean Total Cost (per stay) $9,580 $4,729 $5,310
* Age in months cannot be provided as databases used for this analysis do not provide those data.
The third quarter (July through September) accounted for more admissions (33%) than the other quarters. For infants and children, spring and summer months had the most admissions, while among older patients, summer and fall were more active (Figure 1).
Figure 1 Admissions for pertussis by age group for years 1996 – 1999 by months of hospitalization.
The mean LOS was 6 days (median: 3.5, range: 1 – 110 days) and 13% of patients spent a portion of their stay in an intensive care unit. The mean cost of a hospital stay was estimated to be $9,130 per person (median: $4,600, range: $520 – $507,697). The cumulative cost over four years for all hospitalized pertussis cases examined in this study was estimated to be $29.4 million. Medicaid, a health insurance plan jointly funded by federal and state governments for those with low income who qualify, was the responsible payer for the majority (54%) of hospitalized cases. Managed care organizations were the next largest payer group, responsible for almost one third (31%) of the hospital stays.
Pneumonia was coded in 6% of cases and apnea or respiratory failure in 4%. Convulsions were recorded for 1%. Encephalopathy was recorded in less than 1% (0.3%) of all cases. The case fatality rate was less than 1% (n = 7). Six of the deaths occurred among infants, and the other was a patient aged 75 years. The vast majority (99%) went home (6% with other home health care services) and the rest (1%) returned to long-term or residential care facilities.
During the years 1996 to 1999, the four states reported 9,605 pertussis cases, representing one third of all cases reported in the US over that period [30-37]. Hospitalized cases were 26% of those reported in these four states. This proportion varied substantially by state (Figure 2).
Figure 2 Number and proportion of hospitalized compared to all reported pertussis cases by state for 1996 – 1999.
Infants
Those less than one year of age accounted for 90% (n = 2,266) of all pertussis admissions. Males and females were represented equally. As precise age data are not available in these data sets, the distribution by month of age cannot be determined. A stay in an observation unit prior to admission was noted in 3% of cases and 45% were treated in the emergency department prior to admission. The majority (59%) of hospitalizations for infants occurred in the months of April through September. Pneumonia was coded in 12% of infant cases; apnea, respiratory distress or failure in 16%; convulsions in less than 1%, and encephalopathy in 0.2%. The mean LOS was 6 days (median: 4.5, range: 1 – 110 days) and 14% of all infants spent an average of 8 days in an intensive care nursery or pediatric intensive care unit. The cost per hospital stay averaged $9,580 (median: $4,670, range: $515 – $496,712). The inpatient case fatality rate was less than 1%. All survivors went home, of whom 6% were referred for home health care services, with less than 1% discharged on continuing intravenous drug therapy. Medicaid was the responsible payer for 57% of the hospital stays.
Children
Children, aged 1 through 11 years, represented 8% (n = 191) of the admissions. The majority was female (63%) and 40% were seen in the emergency department prior to admission. The mean age for children was 3 years (median: 2). The average stay was 3.7 days (median 3, range 1 – 10 days); 15% had pneumonia coded; 5% convulsions and 8% apnea, respiratory distress or failure. Admissions for children were more in the summer months, (58%) in the months of June through September. The cost per hospital stay, estimated at $4,729 (median: $4,400, range: $537 – $49,945), was the lowest of the three age groups. This was not merely a reflection of the shorter stays but also a lower average per diem cost ($1,233) compared to infants ($1,347) and adolescents/adults ($1,612). No children died during hospitalization; all were discharged home (4% with home health care services). Medicaid (42%) and managed care organizations (39%) were responsible for most of the hospital stays.
Adolescents and adults
Those aged 12 and older represented only 2% (n = 61) of the hospitalized cases. Slightly more than half (54%) of them were female; mean age 40 years (median: 40, range: 12 – 83 years, Figure 3). Adults (mean age 50 years, median 49 years, range: 22 – 83) had the highest rate (48%) of admissions via the emergency department while the 17 adolescents (mean age: 13.5 years, range: 12–18 years) had the lowest rate (29%) of all groups, but 17% spent time in an observation unit prior to hospitalization.
Figure 3 Age distribution of adolescent/adult hospitalized pertussis cases.
One third (32%) of these admissions occurred between July and September; another 28% were admitted from October through December. The mean LOS was 3.4 days (median: 3, range: 1–11 days) and did not vary appreciably by age (adolescents, mean: 3, median: 3.2 days; adults, mean 3.4, median 3 days). In the small sample of adolescents, no one spent time in an intensive care unit, whereas 11% of the adults did, an average of one day. Pneumonia was noted in 11%; respiratory distress or failure in 8%; and convulsions in 2%. Among these patients, only one, aged 75 years died during hospitalization. The average cost per stay was $5,310 (median: $4,293, range: $685 – $27,625). All survivors went home, with 4% referred for home health care. Managed care organizations were the responsible payer for 43% of the patients in this group, the largest primary payer for both adolescents and adults. In the adult group, Medicare was the primary payer for 18% of the cases. Medicare is the federally funded health insurance program for those aged 65 years and older, as well as some younger patients who qualify because of disabilities.
Discussion
This study shows that hospitalization is a common consequence of pertussis severe enough to be diagnosed and reported – one quarter of the cases of pertussis in these four states were admitted. Almost all of these patients admitted with pertussis survived and returned home, yet these results show that the economic consequences of pertussis at any age can be considerable. This has budgetary implications for both the public and private sector, as the responsible payers for the majority of cases were Medicaid or managed care organizations. Caring for patients with pertussis who require hospitalization generates millions of dollars in health care expenses and yet this substantial amount is a conservative estimate as it reflects but one aspect of care. As the focus of this analysis was limited to examining direct medical care costs incurred during hospitalization, post-acute care and outpatient care costs were not included in the estimates. Most patients will be treated as outpatients and will use other health care services, such as office and emergency department visits, and antibiotics. Neither those costs nor indirect costs have been considered in this analysis; however, a study examining pertussis cases in one New York county over a six year period from 1989–1994 [38] reporting inpatient and outpatient care costs noted that the cost of hospitalization represented one of the two largest contributing costs associated with pertussis. Furthermore, this study and another [39] by the same author, also examined indirect costs related to pertussis and found that work-related (e.g., lost work days, decrease in productivity) costs were responsible for more than 60% of the overall cost of pertussis when both direct and indirect costs were considered. Thus, while the cost of hospital care provides a key component of pertussis-related care costs and must be considered in any economic analysis of the disease, it does not convey fully the economic impact of the disease.
The sample size for adolescents and adults in this analysis was small. There are several possible reasons for this. Pertussis in older patients often presents in an atypical manner without the characteristic whoop [2,6,7,13], which may contribute to misdiagnosis. In such circumstances, those cases would not be identified by the pertussis-related ICD-9 codes and therefore not captured, or counted, for this analysis. It is more likely, however, that the severity of pertussis in older age groups does not warrant inpatient management as often as required for infants and young children. In the 1993 pertussis outbreak in Cincinnati [5], 84% of the hospitalizations were among those less than 12 months old.
Pneumonia was noted as a complication in 6% for all cases and in 12% of infants in our analysis. For seizures, we report a 1% rate for all cases and <1% for infants; 0.3% and 0.2% rates of encephalopathy in all cases and among infants, respectively. These rates appear comparable to those reported by the CDC based on data from 1997–2000 [40]. Among all reported cases, pneumonia occurred in 5.2% of pertussis cases and in 11.8% of infants < 6 months. Seizures were reported by the CDC in 0.8% for all reported pertussis cases and 1.4% for infants; encephalopathy among 0.1% for all cases and 0.2% among infants. Rate reporting for other age groups is not done in a comparable manner as the range of ages within the child and adolescent/adult group differ; however, it does appear that the rates of 15% among children and 11% among adolescents/adults for pneumonia coded in our analysis are higher than rates reported for all ages over age one year by the CDC. The reason for this discrepancy is unknown. It may reflect an error in coding or that this analysis is based only on more severe hospitalized cases; whereas, the CDC is reporting complications rates for all reported cases of pertussis.
Using data from any large administrative database raises concern about potential coding errors. The coding error rate for pertussis in the state databases is unknown, but it is not uncommon for an infectious disease to be identified with less specific codes. This has been reported for other conditions, such as pneumonia [41]. Therefore, it is possible that pertussis cases could have been coded to more generic respiratory conditions, such as unspecified upper respiratory infection (ICD-9 code 465.9). In 1999, there were 5,449 hospitalizations in these databases coded with this non-specific principal diagnosis. It has been reported that 39% of university students [6] who were ultimately determined to have pertussis were first diagnosed by their primary care provider as having an upper respiratory tract infection. Another 48% in that study were misdiagnosed with bronchitis.
A large proportion of the reported cases in some of the states were hospitalized, possibly suggesting that only the most severe cases are being reported; however, this analysis showed a large disparity in these proportions among the four states. The reason for the disparity is not known and can not be determined based on the data available. It is more likely, however, that there are multiple factors contributing the variability seen in the rates of hospitalization among states rather than a differing clinical profile or admitting threshold. Among these factors could be how pertussis cases are reported in each state, the number of cases of infant versus adolescent/adult disease for a state, or that many are not being diagnosed correctly and therefore go unreported.
Based on the results of our analysis and those of others [2,8,11,12,14,15,21], it is evident that pertussis is neither a disease limited to childhood nor one that has been conquered through current immunization programs. The rise in pertussis incidence in the past several years highlights the growing vulnerability of those beyond the protection afforded by childhood immunization. While newer acellular pertussis vaccines may permit extension of immunity to an older population, until very recently booster doses of pertussis-containing vaccines were not available for those over the age of seven years [1].
While efforts have been made to make the clinical community more aware of the increasing occurrence of this disease in older patients, it has been shown that pertussis often goes unrecognized, undiagnosed and underreported in adolescents and adults [7,8,11,19]. Thus, the burden of illness reflected in the results from this study, particularly with regard to older patients may be much lower than in reality.
Clearly, the impact of immunization on the disease burden and costs will depend largely on the extent of herd immunity attained and this is not yet securely known for any particular level of coverage.
Conclusion
The results reported in this analysis, albeit limited to one level of management, illustrate how costly it can be to manage pertussis when hospitalization is required. These results provide decision makers concerned with the economic aspects of preventing and treating pertussis with recent and relevant cost estimates for a substantial cost component of managing pertussis; however, to assess the cost of an expanded vaccination program, all aspects of the economic burden of caring for patients with pertussis, including indirect costs, will need to be considered.
Competing interests
Caro Research, of which Judith A. O'Brien and J. Jaime Caro are shareholders, received an unrestricted grant from Aventis Pasteur SA for this work. Relevant staff from Aventis Pasteur SA were allowed to review and comment on this manuscript but were explicitly forbidden from exerting any editorial control. The authors are employees of Caro Research. Expenses for travel to present the findings at the 2004 International Congress of Infectious Disease were reimbursed.
Authors' contributions
JAO conceived of and designed the study, performed the analysis, and participated in drafting the manuscript. JC participated in the design of the study and participated in drafting the manuscript. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors wish to thank Piedad Duran and Clare Pitoniak-Morse of Caro Research Institute for their assistance with the data analysis.
==== Refs
Centers for Disease Control. Divisions of Bacterial and Mycotic Diseases. Diseases of Infection: Pertussis – Technical Information
Bass JW Stephenson SR The return of pertussis Pediatr Infec Dis J 1987 6 141 144 3562135
From the Centers for Disease Control. Resurgence of Pertussis – United States: 1993 JAMA 1994 271 340 341 8283572 10.1001/jama.271.5.340
Centers for Disease Control. Pertussis surveillance: United States, 1986–1988 JAMA 1990 263 1058 1059 2299773 2299773 10.1001/jama.263.8.1058
Christie CDC Marx ML Marchant CD Reising SF The 1993 epidemic of pertussis in Cincinnati: resurgence of disease in highly immunized population of children NEJM 1994 331 16 21 8202096 10.1056/NEJM199407073310104
Morgan-Mink CA Cherry JD Christenson P Lewis K Pineda E Shlian D Dawson JA Blumberg DA A search for Bordetella pertussis infection in university students Clin Infect Dis 1992 14 464 471 1554832
Cromer BA Goydos J Hackell J Mezzatesta J Dekker C Mortimer EA Unrecognized Pertussis Infection in Adolescents AJDC 1993 147 575 577 8488807
Nenning ME Shinefield HR Edwards KM Black SB Fireman BH Prevalence and Incidence of Adult Pertussis in an Urban Population JAMA 1996 275 1672 1674 8637142 10.1001/jama.275.21.1672
Centers for Disease Control. Summary of Notifiable Diseases, United States 2002 MMWR 2004 51 1 84 15123988
Centers for Disease Control. Pertussis Outbreak Among Adults at an Oil Refinery Illinois 2002 MMWR 2003 52 1 4
Herwaldt LA Pertussis in Adults Arch Intern Med 1991 151 1510 1512 1872655 10.1001/archinte.151.8.1510
Lambert HJ Epidemiology of a small pertussis outbreak in Kent County, Michigan Public Health Rep 1965 80 365 369 14279983
Mortimer EA Pertussis and Its Prevention: A Family Affair The Journal of Infectious Diseases 1990 161 473 479 2179422
Jenkinson D Duration of effectiveness of pertussis vaccine: evidence from a 10 year community study BMJ 1988 296 612 614 3126927
Mink CM Sirota NM Nugent S Outbreak of Pertussis in a Fully Immunized Adolescent and Adult Population Arch Pediatr Adolesc Med 1994 148 153 157 8118532
Kurt TL Pertussis Immunization for medical personnel Ann Intern Med 1972 76 1046 1047 5027580
Kurt TL Yeager AS Guenette S Dunlop S Spread of pertussis by hospital staff JAMA 1972 221 264 267 4338335 10.1001/jama.221.3.264
Marchant CD Loughlin AM Lett SM Todd CW Wetterlow LH Bicchieri R Higham S Etkind P Silva E Siber GR Pertussis in Massachusetts, 1981–1991:Incidence, Serologic Diagnosis, and Vaccine Effectiveness The Journal of Infectious Diseases 1994 169 1297 1305 8195607
Sutter RW Cochi SL Pertussis hospitalizations and Mortality in the United States, 1985–1988. Evaluation of the Completeness of National Reporting JAMA 1992 267 386 391 1309387 10.1001/jama.267.3.386
Gan VN Murphy TV Pertussis in Hospitalized Children A in J Dis Child 1990 144 1130 1134
Lee LH Pichichero ME Costs of Illness Due to Bordetella pertussis in Families ArchFam Med 2000 9 989 996 10.1001/archfami.9.10.989
CDC. Pertussis – United States, 1997–2000 MMWR 2002 51 73 6 11837909
Hart A Hopkins CA eds 2001 ICD-9-CM Code Book 2000 West Valley City, UT: St. Anthony Publishing Inc. and Ingenix, Inc
American Medical Association Physicians' Current Procedural Terminology 2001 Chicago, IL: American Medical Association
2002 Physician Fee Schedule Payment Amount File National/Carrier. Centers for Medicare and Medicaid Services
2001 Florida Medicaid physician fee schedule
Veririlli DK Zuckerman S Health Care Financing Review 1996 17 161 170 10158728
U.S. Department of Health and Human Services Centers for Medicare & Medicaid Services Office of Research, Development, and Information 2003 Baltimore, MD: Medicare and Medicaid Statistical Supplement, 2001. Health Care Financing Review. Pub. No. 03441
Bureau of Labor Statistics
Centers for Disease Control. Summary of Notifiable Diseases, United States 1996 MMWR 1997 45 1 87 9385877
Centers for Disease Control. Summary of Notifiable Diseases, United States 1997: MMWR 1998 46 1 87
Centers for Disease Control. Summary of Notifiable Diseases, United States 1998: MMWR 1999 47 1 93
Centers for Disease Control. Summary of Notifiable Diseases, United States 1999: MMWR 2001 48 1 104
Division of Communicable Control-Surveillance and Statistics Section CA Department of Health Services, Sacramento
Florida Communicable Disease Frequency Reports Florida Department of Health, Tallahasse
Division of Epidemiology and Immunization Bureau of Communicable Disease Control Chart1: Number of Confirmed Cases of Communicable Diseases Reported to the MDPH 1990–2001 Massachusetts Department of Public Health, Boston
Pertussis Incidence Rates. Communicable Disease Epidemiology Washington State Department of Health Olympia
Pichichero ME Treanor J Economic impact of pertussis Arch Pediatr Adolesc Med 1997 151 35 40 9006526 9006526
Lee LH Pichichero ME Costs of illness due to Bordetella pertussis Arch Fam Med 2000 9 989 996 11115197 10.1001/archfami.9.10.989
Centers for Disease Control Pertussis
O'Brien JA Patrick AR Caro JJ Costs of Managing CAP are Double for Elderly Patients Drug Benefit Trends 2003 15 32 47
|
16008838
|
PMC1184075
|
CC BY
|
2021-01-04 16:28:15
|
no
|
BMC Infect Dis. 2005 Jul 11; 5:57
|
utf-8
|
BMC Infect Dis
| 2,005 |
10.1186/1471-2334-5-57
|
oa_comm
|
==== Front
BMC ImmunolBMC Immunology1471-2172BioMed Central London 1471-2172-6-121596976710.1186/1471-2172-6-12Research ArticleModulation of p53 activity by IκBα: Evidence suggesting a common phylogeny between NF-κB and p53 transcription factors Dreyfus David H [email protected] Masayuki [email protected] Erwin W [email protected] Lucy Y [email protected] Division of Basic Sciences, Department of Pediatrics, National Jewish Medical Research Center, Denver, CO 80262 USA2 Department of Pediatrics, Yale University School of Medicine, New Haven, CT, USA3 Departments of Pediatrics and Developmental Biology, Postgraduate School, Tokyo Medical and Dental University, Tokyo, Japan4 The Webb-Waring Institute for Cancer, Aging, and Antioxidant Research and the Department of Medicine, the University of Colorado at Denver and Health Sciences Center, Denver CO 80262 USA; To whom correspondence should be addressed at The Webb-Waring Institute, UCDHSC, Box C321, 4200 East Ninth Ave., Denver, CO 80262 USA2005 21 6 2005 6 12 12 27 10 2004 21 6 2005 Copyright © 2005 Dreyfus et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
In this work we present evidence that the p53 tumor suppressor protein and NF-κB transcription factors could be related through common descent from a family of ancestral transcription factors regulating cellular proliferation and apoptosis. P53 is a homotetrameric transcription factor known to interact with the ankyrin protein 53BP2 (a fragment of the ASPP2 protein). NF-κB is also regulated by ankyrin proteins, the prototype of which is the IκB family. The DNA binding sequences of the two transcription factors are similar, sharing 8 out of 10 nucleotides. Interactions between the two proteins, both direct and indirect, have been noted previously and the two proteins play central roles in the control of proliferation and apoptosis.
Results
Using previously published structure data, we noted a significant degree of structural alignment between p53 and NF-κB p65. We also determined that IκBα and p53 bind in vitro through a specific interaction in part involving the DNA binding region of p53, or a region proximal to it, and the amino terminus of IκBα independently or cooperatively with the ankyrin 3 domain of IκBα In cotransfection experiments, κBα could significantly inhibit the transcriptional activity of p53. Inhibition of p53-mediated transcription was increased by deletion of the ankyrin 2, 4, or 5 domains of IκBα Co-precipitation experiments using the stably transfected ankyrin 5 deletion mutant of κBα and endogenous wild-type p53 further support the hypothesis that p53 and IκBα can physically interact in vivo.
Conclusion
The aggregate results obtained using bacterially produced IκBα and p53 as well as reticulocyte lysate produced proteins suggest a correlation between in vitro co-precipitation in at least one of the systems and in vivo p53 inhibitory activity. These observations argue for a mechanism involving direct binding of IκBα to p53 in the inhibition of p53 transcriptional activity, analogous to the inhibition of NF-κB by κBα and p53 by 53BP2/ASPP2. These data furthermore suggest a role for ankyrin proteins in the regulation of p53 activity. Taken together, the NFκB and p53 proteins share similarities in structure, DNA binding sites and binding and regulation by ankyrin proteins in support of our hypothesis that the two proteins share common descent from an ancestral transcriptional factor.
==== Body
Background
In this work we present evidence that the p53 tumor suppressor and NF-κB transcription factors could be related through common descent from a family of ancestral transcription factors regulating cellular proliferation and apoptosis. P53 and related proteins are transcription factors that regulate DNA repair and cellular apoptosis in response to stress and injury, notably those resulting in DNA damage [1-4]. Although it is a non-essential gene, loss of p53 function in humans through hereditary syndromes is associated with a markedly increased rate of malignancy. Furthermore, over 50% of malignancies have mutated p53 alleles [5]. These observations suggest p53 and related proteins function as a checkpoint for malignant transformation either by repairing DNA damage or by eliminating cells with irreparably damaged DNA [6-8]
P53 is a homotetrameric transcription factor that binds a consensus sequence 5' RRRRC(A/T)(T/A)GYYY-3' (where R indicates purine, A or G; and Y indicates pyrimidine, C or T)[9]. The consensus sequence is usually present as a dimer in p53 inducible gene promoters such as the p21WAF1 protein regulating cell cycle progression [10]. P53 protein can be divided into three functional domains, the amino-terminal activation domain encompassing amino acids 1–43, the core sequence-specific DNA-binding domain (amino acids 100–300), and the multi-functional carboxy-terminal domain (amino acids 300–393) [11,12]. Point mutations in p53 identified in malignant cells are clustered around volutionarily conserved regions in the DNA binding region of p53 and simultaneously eliminate both sequence-specific DNA binding and transcriptional activity [12-14].
P53 is regulated on multiple levels including post-translationally by modifications such as acetylation, phosphorylation, protein degradation, and protein-protein interaction [15,16]. Phosphorylation of p53 induces conformational changes that alter interactions with regulatory proteins such as MDM2, which in turn can regulate p53 stability, and can also activate site-specific DNA binding activity [17-22]. Additional cellular proteins that bind to p53 include proteins of the general transcription machinery such as CBP/p300 [23,24]. CBP/p300 binding to p53 regulates acetylation and p53 transcriptional activation [25,26]. P53 is also regulated through association with ankyrin repeat proteins such as p53 binding protein 1 (53 BP1) and p53 binding protein 2 (53BP2, now known to be a fragment of apoptosis stimulating protein of p53 or ASPP2) [27] and gankyrin [28,29]. Thus, it is likely that p53 is modulated by association with and/or modification by a variety of regulatory proteins including kinases, transcription factors, and ankyrin-containing proteins. In addition, viral proteins also bind to and modify p53 and may contribute to malignant transformation of infected cells by viruses such as papilloma [30,31], cytomegalovirus [32], and Epstein-Barr virus (EBV)[33,34].
NF-κB transcription factors also play a central role in the control of apoptosis [35-37]. NF-κB transcription factors bind to the consensus sequence 5'-GGRNNYYCC-3' in the promoters of both cellular and viral genes [38,39]. The RelA/p65 subunit of NF-κB is regulated by the ankyrin repeat protein IκBα which masks the nuclear localization signal of the p65/p50 NF-κB heterodimer [40,41]. NF-κB-inducing signals are transmitted from the cell surface to the cytoplasm resulting in site-specific phosphorylation at two sites in the N-terminus of IκBα [42-45], conjugation of ubiquitin molecules to IκBα, and subsequent degradation of ubiquitinated IκBα by the 26S proteasome complex [46-49]. Degradation of IκBα in turn unmasks the nuclear localization signal of p50/p65 followed by translocation of the active transcription factor to the nucleus. Other NF-κB subunits including a homodimer of the p50 subunit also bind IκBα [50]. IκBα deficient animals while viable, die of uncontrolled inflammation in infancy, and mice overexpressing IκBα display an abnormal immunologic repertoire suggesting that a major physiologic role of IκBα is to limit immune and inflammatory responses through a feedback pathway [35,51,52].
Interactions between p53 and NF-κB have been noted, for example both factors compete for a binding site in the regulatory factor CBP/p300 [53]. Transfection of a constitutively active form of IκBα protein can block p53 dependent cell death [54-56] and p53 regulatory factors can modulate NF-κB pathways [57-59]. Not only are p53 and NF-κB transcription coregulated under a variety of physiological conditions, but similarity has been noted between the crystal structures of p53 and NF-κB p50 [14,60-62]. Both proteins contain a similarly sited zinc atom that coordinates site specific DNA binding and similar secondary and tertiary organization, but no primary amino acid similarity was noted between the two proteins. Furthermore, p65 has been shown to bind the ankyrin protein p53-binding protein 2 (53BP2). 53BP2 has been shown to be a fragment of a larger protein, ASPP2, that promotes the apoptotis-inducing effects of p53 [27]. P65 apparently inhibits p53-mediated apoptosis by binding to 53BP2/ASSP2 [63]. As with p53, NF-κB transcription is altered by viral pathogens [64-67].
In this work, we present evidence for the specific binding of the NF-κB inhibitor and ankyrin protein, IκBα, to p53. A physical interaction between p53 and IκBα was also reported by Chang [68], as well as Zhou, et al. [69], and was shown to be regulated by proapoptotic and growth suppressing stimuli. Our studies show the binding interaction to involve both ankyrin and non-ankyrin sequences in the IκBα protein, and the DNA binding core domain of p53. We demonstrate that transient expression of IκBα is associated with NF-κB independent decreases in p53 mediated transcription of a p53 reporter gene in vivo. These observations were made in Akata cells, an EBV-genome positive lymphoblastoid cell line originally derived from a Burkitts lymphoma. Akata cells lack endogenous functional p53 [33]. We propose that the binding of the ankyrin protein IκBα to p53 is based upon the similarity in molecular structure of NF-κB and p53.
Results
Similarity between p53 and NF-κB transcription element binding sites
The hemi-dyad DNA consensus binding sites of p53 and NF-κB transcription factors are intriguingly similar as has been noted previously by Foo, et al., [70]; for p53, the binding site sequence is 5'-RRRRC(A/T)(T/A)GYYY-3' and for NF-κB the binding site sequence is 5'- GGGRNNYYCC-3' where N, R, and Y indicate any nucleotide, purine (A or G), and pyrimidine (C or T), respectively. Two changes in the nucleotide sequence of the C(A/T)(A/T)G core (underscored, above) of a p53 binding site is sufficient to generate the RNNY core (underscored, above) of a NF-κB binding site while sequences flanking the core are conserved. Depending on the specific sequences, these binding sites potentially encode a hairpin structure that could promote these nucleotide substitutions. Under these circumstances, only one mutagenic event would be necessary, since the second nucleotide exchange could occur by excision-repair following a mutagenic event in the first site. Based on this observation, we hypothesized that both regulatory factors shared descent from a common ancestral transcription factor, a proto-p53/NF-κB. An alternative possibility is that two independent families of transcription factors, proto-p53 and proto-NF-κB, converged to independently recognize a similar recognition sequence.
If both p53 and NF-κB descended from a common ancestral protein, they might retain the ability to bind to common proteins or proteins of related structures in addition to binding similar DNA sequences. With independent descent converging upon a similar DNA binding sequence no such common interactions would be expected. In support of descent from a common ancestral protein, the crystal structure of NF-κB p65 aligns with that of p53 (Figure 1B and 1D, discussed below) a relationship that would not be expected unless the two proteins were related through phylogeny (descent) as well as ontogeny (function). Furthermore, numerous p53 binding proteins have in common the ankyrin repeat structure, a feature shared with the IκB family of NF-κB regulatory proteins.
Structural alignment of p53 and p65 and ankyrin proteins, 53BP2 and IκBα
To test the hypothesis that p53 and p65 may share structural homology, the known crystal structures of both molecules were aligned using the program C3D4.1 [71]. P53 (Figure 1A) has previously been crystallized in association with the ankyrin protein 53BP2 (Figure 2C) [28]. Like IκBα (Figure 2A), 53BP2 is also an ankyrin protein and has been identified as a fragment of a larger protein known as ASPP2 (apoptosis stimulating protein of p53) [27]. The structure of p65 (Figure 1C) is taken from the crystal structure of the dimer comprised of p65 and p50 in association with IκBα [50]. P65 has also been crystallized as a homodimer in association with IκBβ [72].
We used the information content algorithm of C3D4.1 which uses a BLOSUM62 based matrix to calculate conservation between the two pairs of proteins: p53 and p65 (Figure 1E); and IκBα and 53BP2 (Figure 2E). The amino acids colored red in the aligned structures of 1D and 2D and the amino acid sequences of 1E and 2E depict regions of highest conservation while the grey areas are regions where there is no conservation. As shown by the regions of the two molecules colored red, we observed a surprising degree of structural alignment between the Rel homology domain of p65 and the p53 core domain, supposedly unrelated molecules (Figure 1D). The sequence alignment corresponding to the structures in 1D is depicted in 1E.
Using these same methods, we aligned the structures of the ankyrin proteins, IκBα [50] and 53BP2 (Figure 2D and 2E). The structure used for IκBα (Figure 2A) contains 71–280 of the wild-type protein. We omitted the SH3 domain of the 53BP2 structure (Figure 2C) since it is not relevant to this report. The IκBα protein contains six ankyrin motifs while the 53BP2 contains four. As expected, the two ankyrin proteins show a high degree of alignment. The fourth and 5th ankyrins of IκBα are in very close alignment to the second and third ankyrins of 53BP2 (Figure 2D). The sequences corresponding to these structures is shown in Figure 2E.
Specific binding between IκBα and p53 occurs in vitro
Purified IκBα protein was specifically labelled in vitro with [32P] using p90rsk which we have previously shown to quantitatively phosphorylate IκBα. [32P] -labeled IκBα was incubated with purified p53 (GST-p53) and precipitated either with glutathione/Sepharose (Seph-GSH) beads to bind the glutathione binding tag on GST-p53 or a p53-specific antibody (Ab2) and protein A/G beads. As shown in Figure 3, evidence of a specific association between purified bacterially-produced p53 and purified bacterially-produced wild-type IκBα was observed in vitro (Fig. 2A). The conformation-specific p53 antibody (Ab5) does not recognize bacterially produced p53 (unpublished observations), and Ab5 was used as a control for non-specific association between IκBα and antibody or Protein A/G beads. A c-Jun protein with a glutathione binding protein tag (GST-c-Jun) was used as a control for non-specific association between IκBα and glutathione beads. As shown, no association between control proteins or antibodies was evident.
Relatively decreased in vitro binding between IκBα and p53 results from deletion of the N-terminus of IκBα
Initial experiments conducted with both p53 and IκBα made in bacterial cells supported the hypothesis that the p53 protein could bind to IκBα. Post-translational modifications of the proteins when expressed in eukaryotic cells could affect the interactions noted, particularly with respect to relative binding affinities of mutant proteins. Furthermore, bacterially produced protein may differ in conformation to those produced in mammalian hosts. Overexpression of IκBα alleles in COS cells was used to confirm that binding to p53 was evident with IκBα produced in vivo and also to map the p53 binding to sites within IκBα. Only some alleles of IκBα can be overexpressed in COS cells including IκBα deleted of N-terminal regulatory sequences, denoted ΔN1(missing amino acids 2–36), ΔA2 (missing amino acids 110–136), and ΔC1 deleted of the PEST-containing non-ankyrin C- terminus (amino acids 264–317). A schematic of deletion mutants of IκBα used is depicted in Figure 3C.
After expression of IκBα in COS cells, whole cell lysates were co-precipitated with bacterially produced purified GST-p53 (Figure 3B). These experiments confirmed that binding between the proteins was not related to production of an aberrant form of the IκBα protein in bacterial cells since alleles of IκBα protein produced in COS cells were enriched by incubation and co-precipitation with GST-p53. In representative experiments shown, relatively less ΔN1 protein protein co-precipitated with GST-p53 from COS cell lysates as compared to the proportions of ΔA2 or ΔC1 IκBα proteins associating with p53 suggesting that binding between the proteins at least in part required an intact N-terminal region of IκBα. Relative co-precipitation of the C-terminally deleted form of IκBα was comparable to that of wild-type protein (wild-type data not shown), suggesting that the amino acids 264–317 of IκBα was not important for binding to p53.
The binding between IκBα and p53 is disrupted by an antibody to the DNA-binding region of p53
Experiments detailed above confirmed that a specific binding association occurs between IκBα and p53 in vitro. Further experiments were designed to approximate the region on p53 responsible for IκBαbinding by use of epitope-specific monoclonal antibodies. Glutathione-sepharose beads (G) were used to precipitate p53 (Figure 4A, lane 1, and 4B, lanes 1 and 5) and were found to co-precipitate IκBα. A monoclonal antibody directed at amino acids 46–55 in the amino terminus of p53 (Ab2) recognized purified bacterially produced p53 (4B, lanes 2 and 6) and also co-precipitated IκBα with p53 (4A, lane 2). In contrast, a different p53-specific monoclonal antibody directed at amino acids 212–217 within the DNA binding core of p53 (Ab3), while able to recognize and precipitate bacterially produced p53 protein (4B, lanes 3 and 7), did not co-precipitate IκBa (4A, lane 4). These observations suggested that the binding site between purified IκBα and p53 in part coincided with the Ab3 epitope, a region which is within the p53 DNA binding core. Alternatively, the antibody could hinder or disrupt IκBα binding to regions near the recognized. epitope or alter the structure of p53 into a conformation unfavorable for IκBα binding.
Overexpression of wild-type IκBα modulates p53 dependent transcription of a synthetic p53 reporter gene in vivo
In vitro studies supported the hypothesis that a direct physical association between p53 and IκBα can occur but they did not provide evidence that such interactions occur in vivo. We hypothesized that transient overexpression of IκBα in cells with co-transfected transcriptionally active p53 might reveal effects of IκBα upon p53 mediated transcription. Using forced overexpression of both proteins from identical viral CMV promoter elements rather than endogenous promoters would eliminate confounding transcriptional effects of IκBα upon transcription of p53 or vice versa. Any effect of NF-κB upon CMV would be normalized as well.
Akata cells were used because these B-lymphoblastoid cells are known to lack endogenous p53. Neither the wild-type nor a mutant protein is expressed that could complicate interpretation of results. P53 transiently transfected into these cells using dextran sulfate is transcriptionally active. As a control for non-specific effects of transfected plasmid DNA, effects of IκBα expression plasmids were compared to co-transfected plasmid containing an identical CMV promoter element (pCMV). Results shown were also normalized to expression of co-transfected pRL-SV40, a plasmid expressing a second form of luciferase as a control for cell viability and transfection efficiency. We found that wild-type IκBα was associated with decreased p53-dependent transcription of the p53 reporter gene pG(13)PyLuc in Akata cells (Figure 5). Increasing the relative ratio of transfected expression plasmids between IκBα and p53 from 1:1 (equivalent to 200 ng IκBα plasmid transfected) to 10:1 (equivalent to 2000 ng IκBα plasmid transfected) demonstrated a dose response effect that varied with different IκBα alleles. Immunoblotting for wt-p53, ΔC-p53, and IκBα alleles show relatively equivalent protein levels for these proteins in these extracts (Figure 5B–D).
Allele specific modulation of p53 transcription by IκBα are independent of the effects of the alleles on NF-κB transcription in vivo
The wild-type IκBα protein is subject to signal-dependent phosphorylation and subsequent degradation leading to release of inhibition of NF-κB. Deletion (e.g. as in the ΔN allele deleting amino acids 2–36) or mutation of the phosphorylation sites (S32 and S36) in the amino terminus of IκBα leads to a stable mutant that constitutively represses NF-κB transcription. Cotransfection at a 1 to 1 ratio of ΔN1 to p53 expression plasmid (200 ng each of p53 and IκBα plasmid transfected) had no significant effects upon p53 transcription (data not shown). Increasing the relative ratio of ΔN1 expression plasmid to p53 showed that at higher ratios, ΔN1 could significantly decrease p53-dependent transcription (Figure 5A, inhibition of the constitutively active ΔC-p53 is shown at 1600 ng ΔN1 plasmid. Similar results were obtained when the ratio of ΔN1 to wild-type p53 was increased to 8:1, data not shown). However, at the lower ratio of 4:1 shown for the wild-type p53 allele, in Figure 5A, there was little inhibition of transcriptional activity by ΔN1. The relatively poor inhibitory activity of ΔN1 towards p53 is consistent with its relatively poor binding affinity for p53 as shown in Figure 3B and described above.
The finding that the ΔN1 allele is only partially competent to inhibit p53 is consistent with our previous observation that deletion of the N-terminus of IκBα reduced its physical affinity for p53 (Figure 3). While the ΔN1 allele is a super-repressor of NF-κB, the ΔA5 allele does not repress NF-κB, and thus this allele could be used to ask whether p53 repression and NF-κB inhibition are separable characteristics of the IκBα molecule. Remarkably, ΔA5, which cannot repress NF-κB transcription, had very pronounced inhibitory effects upon transcription by p53 (Figure 5A), comparable to that of wild-type.
Mapping of interactions between IκBα and p53 in vivo to the ankyrin 3 domain of IκBα
Ankyrin regions in IκBα, shared with other members of the IκB gene family are the sites of binding to NF-κB transcription factor family members. Further experiments were performed to define the effects of deletion of additional ankyrin domains of IκBα upon p53 transcription in Akata cells. In these experiments, shown in Figure 5, IκBα plasmids were transfected at 800 ng (at a ratio of 4:1 relative to p53 plasmid). Deletion of either the second (ΔA2, amino acids 110–136, not shown) or fourth ankyrin domains (ΔA4, amino acids 182–208) had the most potent inhibitory effects upon p53 transcription. Deletion of amino acids 264–317 in the carboxyl terminus (ΔC1), distal to the 6th ankyrin domian, were similarly inhibitory. This region retains the 6th ankyrin domain but deletes a potentially regulatory acidic PEST region. Deletion of the ankyrin 1 domain of IκBα was not evaluated in this work.
In multiple experiments with wild-type p53, deletion of the ankyrin 3 (ΔA3) domain of IκBα (missing amino acids 143–169) resulted in a loss of inhibitory activity towards p53 transcription. Thus ΔA3 acts as a null allele for both p53 and NF-κB transcription in vivo, while other ankyrin deletion alleles of IκBα (ΔA2, ΔA4, ΔA5) act as null alleles for NF-κB but gain of function with respect to p53. Thus, our data are consistent with the hypothesis that the ankyrin 3 and N-terminus of IκBα are cooperatively or independently involved in repressing p53. We cannot, however, rule out an alternative possibility, that the deletion of the ankyrin 3 (ΔA3) of IκBα resulted in a protein with high turnover in Akata cells.
Relative effects of alleles of IκBα are conserved with constitutively active p53 and require transcriptionally active p53
P53 transcription is regulated in part by phosphorylation of the carboxyl terminus of the protein. Phosphorylation induces a conformational change in p53 so that an auto-inhibitory region of the carboxyl terminus no longer inhibits DNA binding [73-75]. To determine whether the effects of IκBα upon p53 transcription required the auto-inhibitory carboxyl terminus of p53, experiments were repeated with a truncated form of p53 containing the first 353 amino acids (Figure 5A, ΔC-p53) together with an 8-fold excess of IκBα plasmid (1600 ng). As expected, these experiments demonstrated increased p53 transcriptional activity measured as p53 -dependent transcription of pG(13)PY/Luc in Akata cells at the same plasmid concentrations compared to experiments using wild-type p53. Co-transfection of a fixed concentration of a plasmid expressing ΔC-p53 protein (200 ng), and IκBα alleles (1600 ng) demonstrated that the carboxyl terminus of p53 was not required for the association between p53-dependent transcription and IκBα expression. As in experiments with wild-type p53, deletion of the ankyrin 4 (ΔA4) domain of IκBα was associated with greater decreases in p53-dependent transcription than other alleles of IκBα. A transcriptionally inactive p53 mutated in the p53 DNA binding core (pC53-CSX3, V193A) gave very low levels of luciferase activity that were not altered by co-transfection of any IκBα alleles (data not shown). Thus, the effects of IκBα upon p53 transcription in Akata cells were found to be entirely dependent upon the presence of co-transfected transcriptionally active p53 but independent of the carboxyl-terminus regulatory sequences of p53.
Allele specific association between IκBα and p53 synthesized in a rabbit reticulocyte lysate (RRL) system correspond to allele specific modulation of transcription in vivo
Some alleles of IκBα including ΔA3 and ΔA4 were not stable in COS cells (unpublished observations). Therefore, COS cells could not be used to determine the correlation between IκBα and p53 binding in vitro (Figure 2B) and transcriptional effects in Akata cells (Figure 5). Thus, to overcome this technical problem, p53 and IκBα proteins were both separately synthesized in a rabbit reticulocyte lysate system (Figure 6). IκBα synthesized in the reticulocyte lysate system is functionally active in its effects upon NF-κB transcription factors and ubiquitin-dependent degradation. Likewise, but in contrast to bacterially synthesized p53, p53 synthesized in RRL system is functionally active as a site-specific DNA-binding protein and like IκBα is degraded in vitro by a ubiquitin-dependent pathway. Experiments were performed to determine whether RRL produced p53 that could interact with IκBα alleles likewise translated in vitro in RRL.
Coupled transcription/translation of a plasmid encoding wild-type p53 in RRL (lane 1, Figure 6B) resulted in several template-specific products that could be precipitated from lysates with p53-specific monoclonal antibodies (lanes 2 and 3, Figure 6B). These products appeared to be a mixture of full length and less-than full-length p53 proteins resulting from internal initiation sites in the p53 template. Negative control precipitation with IgG2A immunoglobulin (lane 4) or protein G/A beads (lane 5) resulted in undetectable levels of p53 recovery.
IκBα was precipitated with a polyclonal antiserum directed at the IκBα C-terminus in the absence (p53-) or presence of p53 (p53+) protein after both proteins were synthesized in the reticulocyte lysate system (Figure 6A). Deletion of the carboxyl terminus of IκBα did not interfere with binding between GST-p53 and IκBα synthesized in COS cells (Figure 2), suggesting that binding of antibody directed at the IκBα carboxyl terminus would not interfere with binding between p53 and IκBα. Visual inspection and gel densitometry of these co-precipitation experiments confirmed a specific association between wild-type IκBα and p53. The association of ΔA3 IκBα allele to p53 was significantly reduced (Figure 6A, p53+, lane 3) as determined by gel densitometry (Figure 6D, p < 0.05) relative to the association between the wild-type IκBα, ΔA4, and ΔA5 alleles of IκBα to p53. Gel densitometry was not sensitive enough to determine whether small differences in the binding of other IκBα alleles to p53 were significant. In the absence of labeled IκBα, bands corresponding to p53 were still evident (5A, lane 1, p53+) suggesting either non-specific binding or co-precipitation with endogenous unlabeled IκBα. There are small quantities of immuno-detectable IκBα in certain preparations of RRL (L. Ghoda, unpublished observations). Similar results were obtained in experiments in which an antibody against p53 (Ab2) were used to co-precipitate p53 and IκBα alleles (unpublished observations). Binding was also evident between IκBα and p53-related proteins smaller in size than full-length p53 although these interactions could not be reliably quantitated due to variable co-migrating protein species in the relevant portions of the gel.
In vivo binding of ΔA5-IκBα to wild-type p53
The melanoma cell line A2085 expressing endogenous wild-type p53 was stably transfected with the hemagluttinin (HA)-tagged ΔA5 allele of IκBα [76]. The ΔA5 allele was chosen because it has no effect on NF-κB activity but is one of the more potent alleles of IκBα in modulating p53-dependent transcription. Several independent stable clones as well as pooled transfectants were expanded and used for experiments for which results of the pooled transfectants (DA5) are shown here (Figure 7). Results were qualitatively similar with pooled vs cells expanded from single colonies. Pooled transformants resulting from antibiotic selection of vector transfected cells were used as controls (Con). Association of ΔA5-IκBα with p53 was evident when UV-irradiated cell extracts were immunoprecipated using anti-p53 antibody and the immunoprecipitates probed with anti-IκBα antibody (top panel). Likewise, association between transfected IκBα and p53 are evident when immunoprecipitates generated using anti-HA were probed with anti-p53 (bottom row). Immunoprecipitation with normal rabbit serum (ns) precipitated little or no IκBα or p53. Association between endogenous wild-type IκBα and p53 are also evident as seen by the upper band detectable with anti-IκBα western blotting in Con and DA5 lanes when extracts are precipitated with anti-p53. When extracts of irradiated cells are immunoprecipitated with anti-HA, and immunoblotted with anti-IκBα, Both endogenous and ΔA5 are evident. This is likely due to the precipitation of wt- IκBα with the p53/ ΔA5-IκBα complex since p53 exists as a tetramer.
Discussion
In this work we determined that IκBα and p53 bind in vitro through a specific interaction in part involving the DNA binding region of p53, or a region proximal to it, and the amino terminus of IκBα independently or cooperatively with the ankyrin 3 domain of IκBα. A physical interaction between p53 and IκBα has been noted by two other groups [68,69]. Generally, our data corroborate these previously reported interactions and provide further evidence for direct transcriptional modulation of p53 by IκBα. There are some notable differences in our observations from that of others, in particular, the binding site on IκBα was reported as existing in the non-ankyrin C-terminus by Chang [68] in an yeast two-hybrid system using a deletion construct missing amino acids 244-317. This construct in fact deletes the 6th ankyrin domain of IκBα, located at amino acids 244 – 263, retained in our C-terminal deletion construct (Δ264–317) which also retains p53 binding and regulation. When the data are taken in toto, it can be inferred that in fact the 6th ankyrin domain, retained in our construct but deleted in Chang's, may in fact contain another p53 contact point, the loss of which were observed by Chang but not by us.
We also found that deletion of the ankyrin 2, 4, or 5 domains of IκBα increased the inhibitory effect of IκBα on p53-dependent transcription in Akata cells. We speculate that the compacting of the structure by elimination of an ankyrin domain may promote better binding of the protein to p53 as it may convert IκBα to a structure closer in size to that of 53BP2 which only contains four ankyrins. Mutations in the ankyrin regions of IκBα have been characterized regarding their effects upon NF-κB-mediated transcription as well as nuclear translocation, with most deletions of ankyrins including deletion of ankyrin 5 leading to a null phenotype (confirmed in part in this work in Akata cells) with respect to NF-κB transcription. Nevertheless, over-expression of ΔA5-IκBα was more effective than wild-type at decreasing p53 transcription at all but the lowest concentrations examined. Thus, it was evident that over-expression of some alleles of IκBα could influence p53 transcription independently of their effects upon transcription by NF-κB. Deletion of the ankyrin 3 region of IκBα eliminated detectable interactions between IκBα and p53 in vivo and in vitro suggesting a critical role of the ankyrin 3 region in a specific in vivo interaction with p53. The combined results obtained using bacterially produced IκBα and p53 (Figure 2) as well as RRL produced proteins (Figure 6) suggest a correlation between in vitro co-precipitation in at least one of the systems and in vivo p53 inhibitory activity. These observations argue for a mechanism involving direct binding of IκBα to p53 in the inhibition of p53 transcriptional activity.
Our interpretation of these observations is that binding between IκBα and p53 occurs primarily between the p53 DNA-binding core or a region proximal to it, and the ankyrin 3 and N-terminal regions of IκBα. Binding results in decreased p53-dependent transcription when both IκBα and p53 are overexpressed in vivo. These results are consistent with those published by Zhou, et al. [69] where induction of p53 regulatable genes such as p21 and Mdm2 in response to doxorubicin was abrogated by expression of a degradation-resistant form of IκBα. These interactions are strikingly similar to the known interactions between IκBα and its physiologic ligand, NF-κB. We suggest that these interactions are a result of a conserved relationship, i.e., that of a common descent, of p65 and p53 from a primordial transcription factor which we term proto p53/NF-κB (Figure 8B). Not only do the two transcription factors bind ankyrin proteins but both NF-κB and p53 have similar DNA binding sites (Figure 8A). Furthermore, both factors bind a viral protein encoded by the EBV BZLF-1 open reading frame (also known as ZEBRA) and can transcription can be modulated by this viral protein when overexpressed in lymphoblastoid cell lines [64,77]. As shown in Figure 8C, we found that a cryptic ankyrin like region is present in the dimerization region of BZLF-1, coincident with the region required for binding to both p65 and p53. With the exception of ankyrin domains, BZLF-1 and IκBα are otherwise not similar in sequence, structure, or function.
Thus, ankyrin repeat domains appear to be key points of interaction between diverse proteins in the NF-κB and p53 superfamilies. A similar conclusion is suggested by the crystal structure of a complex between p53 and 53BP2 [28]. 53BP2 is a 1002 amino acid protein containing four ankyrin repeats that has been found to be a fragment of an even larger protein known as ASPP2. Residues TYSD located in the 4th ankyrin repeat of 53BP2/ASPP2 (Figure 2E, 133 TYsd) binds to the L2 loop of p53 immediately after the zinc ligand site in p53, while the non-ankyrin, SH3 domain of 53BP2 also contribute to the binding interactions. These four residues in 53BP2/ASPP2 are exactly aligned with an equivalent block in the sixth ankyrin repeat of IκBα (Figure 2E, 181 TYqq). It is interesting to note that this corresponds to the 6th ankyrin domain predicted to be a contact point for p53 as outlined above. Paradoxically, 53BP2/ASPP2 is exclusively cytoplasmic thus it may only inhibit and sequester p53 in the cytoplasm. IκBα thus may fall into a paradigm of endogenous and viral p53 regulatory proteins where ankyrin and non-ankyrin domains contribute to p53 binding and transcriptional modulation. Other ankyrin proteins that bind to either NF-κB or p53 could also bind and potentially modulate each other.
Conclusion
In conclusion, modulation of p53 transcription by IκBα and related host and viral proteins could play a role in the regulation of p53-dependent apoptosis in vivo. Both p53 and IκBα are members of multi-gene families, and both protein families regulate apoptosis. Previously, a cooperative relationship between p53-dependent apoptosis and NF-κB activation had been reported [59]. Our observations may represent the converse of this situation where inactivation of NF-κB (i.e. as a result of elevated IκBα levels) results in inhibition of p53 transcriptional activity. Alternatively, a situation where accumulation of IκBα free of NF-κB, as a result of phosphorylation of p65 by RSK1 and ensuing dissociation of NF-κB from IκBα, could result in the regulation of p53 by IκBα [78]. Another setting where our observations may play a significant physiological role is during herpesvirus infection. The Epstein-Barr nuclear antigen (EBNA1) contains a region enriched in gly-ala repeats, and this repeat sequence inhibits the 26S proteasome leading to inhibition of peptide presentation by the MHC Class I restricted pathway [79]. EBNA1 is expressed during viral latency, a condition where it is beneficial for the virus to inhibit apoptosis. Proteasome inhibition will lead to inhibition of all cellular protein degradation and the impact would be the greatest on those proteins with extremely short half-lives such as p53 and IκBα. Thus, the accumulation of IκBα under these circumstances may down regulate the activity of p53.
Decreased p53-dependent transcription could limit the ability of p53 to trigger cellular apoptosis in the presence of inflammation since levels of targets of NF-κB, including IκBα, are up-regulated. If, in fact, IκBα inhibits or dampens p53 function, cells could continue to proliferate and differentiate in the presence of an inflammatory event even if there is DNA damage or other cellular injury that would normally activate p53-dependent cell cycle arrest and apoptosis. Thus, a combination of inflammation and cellular damage could increase malignant transformation of cells since the threshold for p53 levels to trigger apoptosis may be elevated. It is notable that we know little about what thresholds regulate p53 function – in particular, what determines whether a cell repairs DNA and resumes transit through the cell cycle or dies. A molecule such as IκBα may be involved in influencing threshold-dependent cell fate decisions. The observation that truncated p53 proteins may bind to IκBα raises the possibility that short p53-related peptides could alter the binding between IκBα and p53, in turn altering a putative regulatory relationship between these proteins in vivo. Modulation of this regulatory interaction by pharmacologic or other means could potentially alter the balance between cellular proliferation and cell cycle arrest/apoptosis in the context of inflammation.
Methods
In vitro precipitation of p53 and IκBα
Bacterially produced 6xHis-tagged IκBα(His-IκBα) was labeled in vitro with 32P using [γ-32P]ATP and Xenopus laevis p90rsk (a gift of Dr. James Maller, Howard Hughes Memorial Institute and UCHSC, Denver CO) [80] and detected with polyclonal rabbit antiserum specific to the N- or C-terminus of IκBα as described (a gift of Dr. Warner Greene, the Gladstone Institute for Virology & Immunology, UCSF, San Francisco, CA) GST-p53 produced from a pGEX construct in bacteris was incubated with hexahistidine-tagged IκBα purified as described in [44] in IP buffer buffer containing 25 mM Hepes, pH 7.5, 75 mM KCl, 2.5 mM MgCl2, 0.1 mM EDTA, 0.15% NP-40, and 1 mM DTT. Protein complexes were purified using either glutathione conjugated Sepharose beads to bind GST (Pharmacia) for GST-p53 or Protein A/G conjugated Sephadex beads (Oncogene Science, CA). Proteins were separated by polyacrylamide gel electrophoresis (PAGE) on either 10% or 12% gels which were dried and autoradiographed on photographic film (Fuji) for detection of radiolabeled protein or transferred to supported nitrocellulose or PVDF membrane and probed with antibodies using the Renaissance system for western blotting (NEN, Boston, MA). p53 specific monoclonal antibodies Ab1, Ab2, Ab3, Ab5, and Ab6 were obtained from Oncogene Science [81].
Expression of IκBα alleles in cells
COS cells and Akata cells were transiently transfected using Superfectin (Qiagen, Chatworth, CA) with expression vectors for IκBα alleles in an expression plasmid regulated by a CMV promoter (pCMV5) obtained from Dr. S.C. Sun (Penn State, Hershey, PA). COS cells were lysed in IP buffer (see above) and incubated with p53-pGEX. Overexpression and coprecipitation of IκBα proteins in COS cells was complicated by the apparent instability of some IκBα alleles (ΔA3, ΔA4) these cells and lack of an association between other alleles (ΔA5) and p53-pGEX (unpublished observations). Following PAGE, ΔN1 IκBα protein was detected using a rabbit polyclonal antisera directed at the C-terminus of IκBα, while ΔC1 IκBα protein was detected using a rabbit polyclonal antisera directed at the N-terminus of IκBα in these experiments (antibodies to N- and C-termini of IκBα were a gift of Dr. W.C. Greene, the Gladstone Institute for Virology and Immunology and the University of California, San Francisco, CA).
Transient transfection of Akata cells
Akata cells were obtained from Dr. J. F. Jones (National Jewish Medical Research Center, Denver, CO). The identity of Akata cells was confirmed by expression of EBV lytic gene products following ligation of surface IgG (data not shown). Cells were cultured using standard conditions in RPMI medium (GIBCO-BRL) supplemented with 10% fetal calf serum, penicillin/streptomycin (100 U/ml) and L-glutamine (2 mM). As previously reported, endogenous p53 protein was not detected in Akata cells by Western blotting with a rabbit polyclonal antisera directed at the entire p53 coding sequence (Santa Cruz Biologicals, Santa Cruz, CA) [21,81].
Akata cells were transiently transfected with plasmids using dextran sulfate (Pharmacia). p53 protein transiently expressed by plasmids transfected into Akata cells could be detected in nuclear extracts of transfected cells (data not shown). Levels of transfected wild-type IκBα protein were not detected by Western blotting above a high background of endogenous IκBα proteins in Akata cells, although expression of IκBα protein could be inferred by effects on transcription of NF-κB reporter genes. Akata were grown to a density of 1 × 106 cells/ml and 1 ml of cells for each experiment were transfected with Dextran. After transfection, cells were incubated in fresh culture medium for 24 hours prior to determination of luciferase activity.
Luciferase assay
A luciferase reporter gene regulated by 13 tandem copies of a p53 response element, denoted pG(13)PyLuc, or p21 promoter reporter, denoted WWP/Luc, were transfected into Akata cells in experiments at concentrations indicated. As noted in the text the relative effects of different alleles of IκBα upon p53 dependent luciferase expression varied with the relative ratios of expression vector for IκBα to expression vector for p53. CMV-promoter driven p53 expression plasmids encoding wild-type p53 (pC53-SN3) and a DNA binding mutated p53 (pC53-CSX3 V193A) were obtained from Dr. B. Vogelstein, (Johns Hopkins University, Baltimore, MD). An expression plasmid encoding a carboxyl-terminus deleted p53 (pCEP4-353) was obtained from Dr. J. Pietenpol, Vanderbilt Univ., Memphis, TN). Plasmids were prepared using either the Promega (Madison, WI) or Qiagen (Chatsworth, CA) procedures with similar results.
A plasmid encoding a second form of luciferase (Renilla luciferase) driven by an SV40 promoter was used as an internal control for transfection efficiency and cell viability (pRL-SV40, Promega, Madison WI). Similar results were obtained without pRL transfection by normalization of luciferase activity to total cellular protein. Luciferase activity was measured using the Stop and Glo assay (Promega) and an Analytical Luminescence Laboratory luminometer (San Diego, CA). Each data point shown from luciferase assay experiments represents pooled data from at least three independent experiments. Standard error and significant differences (p < 0.05, indicated with an asterick, *) were as determined by the students t-test for multiple comparisons of data points using statistical software (SAS Institute Inc, Cary, NC).
Reticulocyte lysate expression and immunoprecipitation
For reticulocyte lysate expression and immunoprecipitation studies, IκBα expression vectors were digested with XbaI/HindIII and ligated into the XbaI/HindIII site of pBS KS(Stratagene) to place the open reading frames downstream of a T7 RNA polymerase promoter element. These IκBα constructs lack epitope tags. Vectors encoding p53 wild-type (SN3) and DNA binding mutant p53 protein (pBKS-273) for reticulocyte lysate expression were obtained from Dr. B. Vogelstein. 35S-Cysteine/methionine labelled IκBα and p53 proteins were produced in the TNT coupled transcription/translation system (Promega) using T7 RNA polymerase for wild-type p53 and IκBα and T3 for mutant p53. 10 μl of reticulocyte lysates, as indicated, were incubated in a total volume of 100 μl IP buffer for 2 hours at 4°C. Antibodies were added and incubated for an additional 1 hour, and proteins were precipitated with Protein G plus/A Sephadex (Oncogene Science). Beads were washed once with 500 μl IP buffer and proteins denatured in Laemmli sample buffer and separated on 10 or 12% PAGE gels dried and visualized by autoradiography. Despite possibly confounding variables due to the presence of the p65 (RelA) subunit of NF-κB and possibly other related factors in reticulocyte lysates (unpublished observations), a correlation was evident between the quantity of p53 coprecipitated with IκBα and the relative effect of individual alleles upon in vivo p53 transcription in Akata cells. In particular, there was a lack of a detectable interaction between the ankyrin 3 deletion mutant of IκBα and p53 in vivo or in vitro.
Abbreviations used
53BP2, p53 binding protein 2
ASPP2, apoptosis stimulating factor of p53 -2
EBNA, Epstein-Barr nuclear antigen
EBV, Epstein-Barr virus
Authors' contributions
DHD and MN contributed equally to this work. Richard Ruhlen is acknowledged for his technical assistance.
Acknowledgements
This work was supported in part by funding from the American Cancer Society (Junior Faculty Award and Research and Clinical Investigation Award) and NIH grant GM049055 to LYG; and NIH grants HL36577, HL61005, and AI42246, and by EPA grant R835702 awarded to EWG.
Figures and Tables
Figure 1 Alignment of p53 (PDB ID:1YCS; 1YCS_A d1) with p65 (PDB ID:1IKN; 1IKN_A d1). A. The structures of p53 molecule (taken from the crystal structure of p53 and p53 binding protein 2); B. The aligned structures of p53 and p65; C. The structure of the Rel homology domain of p65 (taken from the crystal structure of the p65/p50 heterodimer bound to IκBα); D. The aligned structures were colored according to information content based on a BLOSUM62 matrix to calculate conservation using the public domain program CN3D4.1 [70]. A spectrum of red to blue is used to denote the degree of conservation where red is the most conserved. E. The sequence alignment of p53 and p65 is depicted using the same coloring scheme as in the structure alignment in 1D.
Figure 2 Alignment of IκBα (PDB ID:1IKN; 1IKN_D d1) with p53 binding protein 2 (PDB ID:1YCS; 1YCS_B d1). (A) The structure of IκBα (taken from the crystal structure of the p50/p65 heterodimer bound to IκBα); (B) The aligned structures of IκBα and 53BP2; (C) The structure of 53BP2 (taken from the crystal structure of p53 bound to 53BP2); (D) The aligned structure in 2B colored for conservation according to information content as described above; (E) The sequence alignment of IκBα and 53BP2 is depicted using the same coloring scheme as in the structure alignment in 2D. A four amino acid region near the N-termini of each structure is colored yellow as a reference point.
Figure 3 P53 and IκBα proteins co-precipitate in vitro. A: Purified bacterially produced IκBα protein co-precipitated specifically with p53. Purified IκBα protein was specifically labeled with [γ-32P]ATP using p90rsk (lane 5). [32P]-labeled IκBα was incubated with purified p53 (GST-p53, lanes 1–3) or a control GST fusion protein (GST-c-Jun, lane 4). P53 was precipitated either by a glutathione binding tag on GST-p53 and glutathione Sepharose beads (Seph-GSH, lane 1) or a p53 specific antibody (Ab2, lane 2) and protein A/G Sephadex beads. IκBα protein (position indicated) was not precipitated by p53 specific Ab5 (lane 3) that does not recognize bacterially synthesized p53 protein, or by incubation with GST-c-Jun and precipitation with glutathione Sepharose beads (lane 4). Proteins were separated by PAGE and detected by autoradiography of [32P]-labeled protein. Electrophoresis of the input IκBα used in this experiment is also shown in lane 5. B: Relatively less ΔN1 protein than ΔC1 IκBα protein co-precipitated with GST-p53 from COS cell lysates. After expression of IκBα in COS cells, whole cell lysates were incubated with bacterially produced purified p53 (GST-p53). P53/ IκBα complexes were then precipitated with glutathione Sepharose beads and analyzed by PAGE/Western blotting. In the left panels, ΔN1 and ΔA2 proteins were detected using a rabbit polyclonal antisera directed at the C-terminus of IκBα, while in the right panels, ΔA2 and ΔC1 proteins were detected using a rabbit polyclonal antisera directed at the N-terminus of IκBα. C: Structures of wild-type IκBα and mutant constructs used in these studies.
Figure 4 A monoclonal antibody recognizing an epitope in the DNA binding domain of p53 (Ab3) interferes with IκBα binding to p53. Purified p53 with a glutathione binding protein epitope tag (GST-p53) and purified IκBα protein were incubated together in vitro. A. P53 was precipitated either with glutathione Sepharose (denoted G, lane 2) or with p53-specific monoclonal antibodies Ab2 (lane 3), Ab3 (lane 4), or Ab5 (lane 5) and Sephadex protein A/G beads. Precipitated proteins were separated by SDS-PAGE and detected by Western blotting with a rabbit polyclonal antiserum directed against the N-terminus of IκBα. Mobility of IκBα protein is indicated (lane 1). This also represents the total input IκBα. Similar quantities of murine immunoglobulin heavy chain (HC, lane 6) were precipitated by protein A/G beads and served as the negative control. B. GST-p53 was quantitatively precipitated in the absence (- IκBα) or presence (+ IκBα) of IκBα by glutathione Sepharose (denoted G, lanes 1,5), Ab2 (lanes 2,6), and Ab3 (lanes 3,7), but not Ab5 (lanes 4,8), as detected with a rabbit polyclonal p53 antiserum. This blot is essentially identical to that shown in panel A but for the antibody used in the Western blotting step. Since Ab5 did not precipitate GST-p53, it served as a negative control for non-specific association between IκBα protein and either antibody or protein A/G beads.
Figure 5 Transient transfection of IκBα alleles specifically blocks p53 transcription in EBV-positive Akata cells. A. 200 ng of a p53-dependent reporter plasmid encoding 13 copies of the p53 response element driving a firefly luciferase (pG(13)Py/Luc) were cotransfected into 2 × 105 cells with 1 ng of an SV40 promoter driven Renilla luciferase (pRLSV40) and 200 ng of a CMV-promoter driven wild-type p53 cDNA construct (wt-p53, denoted below the x-axis) or a C-terminally deleted transcriptionally active p53 (ΔC-p53, denoted below the x-axis). The effect of 800 ng of CMV-promoter driven IκBα alleles or a control construct that did not contain DNA encoding IκBα (to which all data were normalized) were used to determine the effect of IκBα on wild-type p53-mediated transcription of pG(13)Py/Luc while 1600 ng of CMV-promoter driven IκBα alleles (or control) were used in experiments where the ΔC-p53 allele was used. Immunoblotting controls for wt-p53, ΔC-p53 and IκBα alleles. B: Western blot analysis of wt-p53 and ΔC-p53 in the presence of transfected IκBα alleles. Extracts were derived from Akata cells as described for Figure 5A. None indicates transfection of empty CMV vector containing no IκBα allele. C: Western blot analysis for IκBα in the presence of transfected wt-p53. A subset of extracts were analyzed both with C-term and N-term directed IκBα antibodies. Endogenous IκBα is detected by these methods. D: Western blot analysis of IκBα in the presence of ΔC-p53. ΔA3 was analyzed on a separate gel due to space limitations and showed a band intensity similar to that of ΔA4 and ΔA5 (data not shown).
Figure 6 Rabbit reticulocyte lysate (RRL) produced p53 and IκBα proteins interact in vitro. A. Wild-type IκBα and mutant proteins were synthesized in rabbit reticulocyte lysates (lane 1, RRL no RNA; lane 2, ΔA2; lane 3, ΔA3; lane 4, ΔA4; lane 5, ΔA5; and lane 6, wild-type) and precipitated with a rabbit polyclonal antiserum against the C-terminus of IκBα using protein A/G sephadex beads either in the absence (p53-) or in the presence (p53+) of RRL synthesized, [35S]-labeled, p53. Equal volumes of RRL were used in all cases. For p53- reactions, RRL programmed with empty vector was incubated with [35S]-labeled methionine. A control reticulocyte lysate without IκBα protein template (lane 1) was included in these experiments. ΔA2 (lane 2), ΔA3 (lane 3), ΔA4 (lane 4) ΔA5 (lane 5) and wild-type (lane 6) IκBα proteins were co-incubated with p53, precipitated and analzyed by SDS-PAGE followed by radiography. B. Wild-type p53 protein translated in RRL (lane 1) produced both putative full-length p53 protein (denoted p53, mobility approximately 55 kD) and also at least two less-than-full length translation products (denoted * and **; lanes 1–3). Putative full-length and less-than full-length p53 translation products were precipitated by p53-specific monoclonal antibodies and protein A/G sephadex beads (lane 2, Ab1 recognizing the carboxyl terminus of p53; lane 3, Ab2 recognizing the amino terminus of p53). Less-than full-length proteins were more readily precipitated by Ab1 than Ab2. Control precipitation with IgG2A immunoglobulin (lane 4) or protein G/A beads (lane 5) did not precipitate p53. C. IκBα proteins were precipitated in similar quantities by IκBα antiserum. A shorter exposure of the gel than in panel A is shown. D. Densitometry of p53 protein precipitated in association with various IκBα alleles. Data shows significant precipitation of p53 with ΔA4, ΔA5 and wild-type, expressed as a ratio of label precipitating with IκBα allele to beads alone. Data was calculated by first normalizing to the background in each lane followed by calculating the ratio of label precipitating with IκBα to a corresponding area in the beads alone lane.
Figure 7 P53 associates with ΔA5-IκBα in UV-irradiated melanoma cells. A2085 cells stably transfected with the parent vector (Con) or ΔA5-IκBα allele (DA5) were irradiated for 30" with UVB and harvested 4 hrs later. The extracts were subjected to immunoprecipitation with non specific rabbit serum (ns), polyclonal rabbit anti-HA antibody (α-HA), or mouse monoclonal anti-p53 Ab2 antibody (α-p53) and probed by western analysis with mouse anti-IκBα antibody directed at the C-terminus of IκBα (top panel) or the Ab3 anti-p53 monoclonal antibody (bottom panel). Experiments were performed on stably transfected, pooled transfectants of controls (Con) or ΔA5-IκBα (DA5).
Figure 8 A model for molecular evolution of p53 and NF-κB from a common ancestral transcription factor, proto-p53/ NF-κB. A. P53 (first row) and NF-κB (fourth row) DNA binding sites share eight out of ten nucleotides as shown by the sequence depicted in the second row (consensus). R represents purine, A or G; Y represents pyrimidine, C or T. The red X denotes nucleotides where there is no match. The predicted DNA binding site sequence of the ancestral proto-p53/NF-κB is shown in the third row. B. An ancestral transcription factor proto-p53/ NF-κB, also regulated by an ankyrin protein, with a DNA binding site shown in A, above, could have been the precursor to both p53 and NF-κB. C. An ankyrin-like region in EBV protein BZLF-1 (ZEBRA) is shown in alignment with ankyrin motifs in IκBα, NF-κB p50, as well as invertebrate ankyrin like regulatory repeats from Drosophila Cactus and C. elegans Unc22.
==== Refs
Ko LJ Prives C p53: puzzle and paradigm Genes Dev 1996 10 1054 72 8654922
Lu H Fisher RP Bailey P Levine AJ The CDK7-cycH-p36 complex of transcription factor IIH phosphorylates p53, enhancing its sequence-specific DNA binding activity in vitro Mol Cell Biol 1997 17 5923 34 9315650
Ollmann M Young LM Di Como CJ Karim F Belvin M Robertson S Whittaker K Demsky M Fisher WW Buchman A Duyk G Friedman L Prives C Kopczynski C Drosophila p53 is a structural and functional homolog of the tumor suppressor p53 Cell 2000 101 91 101 10778859 10.1016/S0092-8674(00)80626-1
Shieh SY Ikeda M Taya Y Prives C DNA damage-induced phosphorylation of p53 alleviates inhibition by MDM2 Cell 1997 91 325 34 9363941 10.1016/S0092-8674(00)80416-X
Soussi T The p53 tumor suppressor gene: from molecular biology to clinical investigation Ann N Y Acad Sci 2000 910 121 37 discussion 137-9 10911910
Chipuk JE Green DR p53's believe it or not: lessons on transcription-independent death J Clin Immunol 2003 23 355 61 14601643 10.1023/A:1025365432325
Lakin ND Jackson SP Regulation of p53 in response to DNA damage Oncogene 1999 18 7644 55 10618704 10.1038/sj.onc.1203015
Sancar A Lindsey-Boltz LA Unsal-Kaccmaz K Linn S Molecular mechanisms of mammalian DNA repair and the DNA damage checkpoints Annu Rev Biochem 2004 73 39 85 15189136 10.1146/annurev.biochem.73.011303.073723
el-Deiry WS Kern SE Pietenpol JA Kinzler KW Vogelstein B Definition of a consensus binding site for p53 Nat Genet 1992 1 45 9 1301998 10.1038/ng0492-45
Mirza A Wu Q Wang L McClanahan T Bishop WR Gheyas F Ding W Hutchins B Hockenberry T Kirschmeier P Greene JR Liu S Global transcriptional program of p53 target genes during the process of apoptosis and cell cycle progression Oncogene 2003 22 3645 54 12789273 10.1038/sj.onc.1206477
Hulboy DL Lozano G Structural and functional analysis of p53: the acidic activation domain has transforming capability Cell Growth Differ 1994 5 1023 31 7848903
Unger T Mietz JA Scheffner M Yee CL Howley PM Functional domains of wild-type and mutant p53 proteins involved in transcriptional regulation, transdominant inhibition, and transformation suppression Mol Cell Biol 1993 13 5186 94 8355677
Kern SE Pietenpol JA Thiagalingam S Seymour A Kinzler KW Vogelstein B Oncogenic forms of p53 inhibit p53-regulated gene expression Science 1992 256 827 30 1589764
Cho Y Gorina S Jeffrey PD Pavletich NP Crystal structure of a p53 tumor suppressor-DNA complex: understanding tumorigenic mutations Science 1994 265 346 55 8023157
Brooks CL Gu W Ubiquitination, phosphorylation and acetylation: the molecular basis for p53 regulation Curr Opin Cell Biol 2003 15 164 71 12648672 10.1016/S0955-0674(03)00003-6
Xu Y Regulation of p53 responses by post-translational modifications Cell Death Differ 2003 10 400 3 12719715 10.1038/sj.cdd.4401182
Dohoney KM Guillerm C Whiteford C Elbi C Lambert PF Hager GL Brady JN Phosphorylation of p53 at serine 37 is important for transcriptional activity and regulation in response to DNA damage Oncogene 2004 23 49 57 14712210 10.1038/sj.onc.1207005
Gu J Chen D Rosenblum J Rubin RM Yuan ZM Identification of a sequence element from p53 that signals for Mdm2-targeted degradation Mol Cell Biol 2000 20 1243 53 10648610 10.1128/MCB.20.4.1243-1253.2000
Hupp TR Sparks A Lane DP Small peptides activate the latent sequence-specific DNA binding function of p53 Cell 1995 83 237 45 7585941 10.1016/0092-8674(95)90165-5
Katayama H Sasai K Kawai H Yuan ZM Bondaruk J Suzuki F Fujii S Arlinghaus RB Czerniak BA Sen S Phosphorylation by aurora kinase A induces Mdm2-mediated destabilization and inhibition of p53 Nat Genet 2004 36 55 62 14702041 10.1038/ng1279
Lohrum M Scheidtmann KH Differential effects of phosphorylation of rat p53 on transactivation of promoters derived from different p53 responsive genes Oncogene 1996 13 2527 39 9000127
Sluss HK Armata H Gallant J Jones SN Phosphorylation of serine 18 regulates distinct p53 functions in mice Mol Cell Biol 2004 24 976 84 14729946 10.1128/MCB.24.3.976-984.2004
Lill NL Grossman SR Ginsberg D DeCaprio J Livingston DM Binding and modulation of p53 by p300/CBP coactivators Nature 1997 387 823 7 9194565 10.1038/42981
Scolnick DM Chehab NH Stavridi ES Lien MC Caruso L Moran E Berger SL Halazonetis TD CREB-binding protein and p300/CBP-associated factor are transcriptional coactivators of the p53 tumor suppressor protein Cancer Res 1997 57 3693 6 9288775
Dornan D Shimizu H Perkins ND Hupp TR DNA-dependent acetylation of p53 by the transcription coactivator p300 J Biol Chem 2003 278 13431 41 12499368 10.1074/jbc.M211460200
Gu W Roeder RG Activation of p53 sequence-specific DNA binding by acetylation of the p53 C-terminal domain Cell 1997 90 595 606 9288740 10.1016/S0092-8674(00)80521-8
Samuels-Lev Y O'Connor DJ Bergamaschi D Trigiante G Hsieh JK Zhong S Campargue I Naumovski L Crook T Lu X ASPP proteins specifically stimulate the apoptotic function of p53 Mol Cell 2001 8 781 94 11684014 10.1016/S1097-2765(01)00367-7
Gorina S Pavletich NP Structure of the p53 tumor suppressor bound to the ankyrin and SH3 domains of 53BP2 Science 1996 274 1001 5 8875926 10.1126/science.274.5289.1001
Krzywda S Brzozowski AM Higashitsuji H Fujita J Welchman R Dawson S Mayer RJ Wilkinson AJ The crystal structure of gankyrin, an oncoprotein found in complexes with cyclin-dependent kinase 4, a 19 S proteasomal ATPase regulator, and the tumor suppressors Rb and p53 J Biol Chem 2004 279 1541 5 14573599 10.1074/jbc.M310265200
Crook T Tidy JA Vousden KH Degradation of p53 can be targeted by HPV E6 sequences distinct from those required for p53 binding and trans-activation Cell 1991 67 547 56 1657399 10.1016/0092-8674(91)90529-8
Scheffner M Werness BA Huibregtse JM Levine AJ Howley PM The E6 oncoprotein encoded by human papillomavirus types 16 and 18 promotes the degradation of p53 Cell 1990 63 1129 36 2175676 10.1016/0092-8674(90)90409-8
Doniger J Muralidhar S Rosenthal LJ Human cytomegalovirus and human herpesvirus 6 genes that transform and transactivate Clin Microbiol Rev 1999 12 367 82 10398670
Farrell PJ Allan GJ Shanahan F Vousden KH Crook T p53 is frequently mutated in Burkitt's lymphoma cell lines Embo J 1991 10 2879 87 1915267
Mauser A Saito S Appella E Anderson CW Seaman WT Kenney S The Epstein-Barr virus immediate-early protein BZLF1 regulates p53 function through multiple mechanisms J Virol 2002 76 12503 12 12438576 10.1128/JVI.76.24.12503-12512.2002
Baeuerle PA Henkel T Function and activation of NF-kappa B in the immune system Annu Rev Immunol 1994 12 141 79 8011280
Baichwal VR Baeuerle PA Activate NF-kappa B or die? Curr Biol 1997 7 R94 6 9081673 10.1016/S0960-9822(06)00046-7
Green DR Death and NF-kappaB in T cell activation: life at the edge Mol Cell 2003 11 551 2 12667438 10.1016/S1097-2765(03)00107-2
Kawakami K Scheidereit C Roeder RG Identification and purification of a human immunoglobulin-enhancer-binding protein (NF-kappa B) that activates transcription from a human immunodeficiency virus type 1 promoter in vitro Proc Natl Acad Sci U S A 1988 85 4700 4 3133660
Pierce JW Lenardo M Baltimore D Oligonucleotide that binds nuclear factor NF-kappa B acts as a lymphoid-specific and inducible enhancer element Proc Natl Acad Sci U S A 1988 85 1482 6 3125549
Nolan GP Ghosh S Liou HC Tempst P Baltimore D DNA binding and I kappa B inhibition of the cloned p65 subunit of NF-kappa B, a rel-related polypeptide Cell 1991 64 961 9 2001591 10.1016/0092-8674(91)90320-X
Beg AA Ruben SM Scheinman RI Haskill S Rosen CA Baldwin AS Jr I kappa B interacts with the nuclear localization sequences of the subunits of NF-kappa B: a mechanism for cytoplasmic retention Genes Dev 1992 6 1899 913 1340770
Alkalay I Yaron A Hatzubai A Orian A Ciechanover A Ben-Neriah Y Stimulation-dependent I kappa B alpha phosphorylation marks the NF-kappa B inhibitor for degradation via the ubiquitin-proteasome pathway Proc Natl Acad Sci U S A 1995 92 10599 603 7479848
DiDonato J Mercurio F Rosette C Wu-Li J Suyang H Ghosh S Karin M Mapping of the inducible IkappaB phosphorylation sites that signal its ubiquitination and degradation Mol Cell Biol 1996 16 1295 304 8657102
Ghoda L Lin X Greene WC The 90-kDa ribosomal S6 kinase (pp90rsk) phosphorylates the N-terminal regulatory domain of IkappaBalpha and stimulates its degradation in vitro J Biol Chem 1997 272 21281 8 9261139 10.1074/jbc.272.34.21281
Sun S Elwood J Greene WC Both amino- and carboxyl-terminal sequences within I kappa B alpha regulate its inducible degradation Mol Cell Biol 1996 16 1058 65 8622650
Grisham MB Palombella VJ Elliott PJ Conner EM Brand S Wong HL Pien C Mazzola LM Destree A Parent L Adams J Inhibition of NF-kappa B activation in vitro and in vivo: role of 26S proteasome Methods Enzymol 1999 300 345 63 9919536
Rodriguez MS Wright J Thompson J Thomas D Baleux F Virelizier JL Hay RT Arenzana-Seisdedos F Identification of lysine residues required for signal-induced ubiquitination and degradation of I kappa B-alpha in vivo Oncogene 1996 12 2425 35 8649784
Karin M Ben-Neriah Y Phosphorylation meets ubiquitination: the control of NF-[kappa]B activity Annu Rev Immunol 2000 18 621 63 10837071 10.1146/annurev.immunol.18.1.621
Scherer DC Brockman JA Chen Z Maniatis T Ballard DW Signal-induced degradation of I kappa B alpha requires site-specific ubiquitination Proc Natl Acad Sci U S A 1995 92 11259 63 7479976
Phelps CB Sengchanthalangsy LL Huxford T Ghosh G Mechanism of I kappa B alpha binding to NF-kappa B dimers J Biol Chem 2000 275 29840 6 10882738 10.1074/jbc.M004899200
Beg AA Sha WC Bronson RT Baltimore D Constitutive NF-kappa B activation, enhanced granulopoiesis, and neonatal lethality in I kappa B alpha-deficient mice Genes Dev 1995 9 2736 46 7590249
Klement JF Rice NR Car BD Abbondanzo SJ Powers GD Bhatt PH Chen CH Rosen CA Stewart CL IkappaBalpha deficiency results in a sustained NF-kappaB response and severe widespread dermatitis in mice Mol Cell Biol 1996 16 2341 9 8628301
Webster GA Perkins ND Transcriptional cross talk between NF-kappaB and p53 Mol Cell Biol 1999 19 3485 95 10207072
Egan LJ Eckmann L Greten FR Chae S Li ZW Myhre GM Robine S Karin M Kagnoff MF IkappaB-kinasebeta-dependent NF-kappaB activation provides radioprotection to the intestinal epithelium Proc Natl Acad Sci U S A 2004 101 2452 7 14983030 10.1073/pnas.0306734101
Holmes-McNary MQ Baldwin AS JrZeisel SH Opposing regulation of choline deficiency-induced apoptosis by p53 and nuclear factor kappaB J Biol Chem 2001 276 41197 204 11483591 10.1074/jbc.M010936200
Jung M Zhang Y Lee S Dritschilo A Correction of radiation sensitivity in ataxia telangiectasia cells by a truncated I kappa B-alpha Science 1995 268 1619 21 7777860
Perkins ND Oncogenes, tumor suppressors and p52 NF-kappaB Oncogene 2003 22 7553 6 14576816 10.1038/sj.onc.1207139
Rocha S Campbell KJ Perkins ND p53- and Mdm2-independent repression of NF-kappa B transactivation by the ARF tumor suppressor Mol Cell 2003 12 15 25 12887889 10.1016/S1097-2765(03)00223-5
Ryan KM Ernst MK Rice NR Vousden KH Role of NF-kappaB in p53-mediated programmed cell death Nature 2000 404 892 7 10786798 10.1038/35009130
Jeffrey PD Gorina S Pavletich NP Crystal structure of the tetramerization domain of the p53 tumor suppressor at 1.7 angstroms Science 1995 267 1498 502 7878469
Muller CW Harrison SC The structure of the NF-kappa B p50:DNA-complex: a starting point for analyzing the Rel family FEBS Lett 1995 369 113 7 7641872 10.1016/0014-5793(95)00541-G
Muller CW Rey FA Sodeoka M Verdine GL Harrison SC Structure of the NF-kappa B p50 homodimer bound to DNA Nature 1995 373 311 7 7830764 10.1038/373311a0
Yang JP Hori M Takahashi N Kawabe T Kato H Okamoto T NF-kappaB subunit p65 binds to 53BP2 and inhibits cell death induced by 53BP2 Oncogene 1999 18 5177 86 10498867 10.1038/sj.onc.1202904
Dreyfus DH Nagasawa M Kelleher CA Gelfand EW Stable expression of Epstein-Barr virus BZLF-1-encoded ZEBRA protein activates p53-dependent transcription in human Jurkat T-lymphoblastoid cells Blood 2000 96 625 34 10887127
Gutsch DE Holley-Guthrie EA Zhang Q Stein B Blanar MA Baldwin AS Kenney SC The bZIP transactivator of Epstein-Barr virus, BZLF1, functionally and physically interacts with the p65 subunit of NF-kappa B Mol Cell Biol 1994 14 1939 48 8114725
Morrison TE Mauser A Klingelhutz A Kenney SC Epstein-Barr virus immediate-early protein BZLF1 inhibits tumor necrosis factor alpha-induced signaling and apoptosis by downregulating tumor necrosis factor receptor 1 J Virol 2004 78 544 9 14671137 10.1128/JVI.78.1.544-549.2004
Sun SC Elwood J Beraud C Greene WC Human T-cell leukemia virus type I Tax activation of NF-kappa B/Rel involves phosphorylation and degradation of I kappa B alpha and RelA (p65)-mediated induction of the c-rel gene Mol Cell Biol 1994 14 7377 84 7935451
Chang NS The non-ankyrin C terminus of Ikappa Balpha physically interacts with p53 in vivo and dissociates in response to apoptotic stress, hypoxia, DNA damage, and transforming growth factor-beta 1-mediated growth suppression J Biol Chem 2002 277 10323 31 11799106 10.1074/jbc.M106607200
Zhou M Gu L Zhu N Woods WG Findley HW Transfection of a dominant-negative mutant NF-kB inhibitor (IkBm) represses p53-dependent apoptosis in acute lymphoblastic leukemia cells: interaction of IkBm and p53 Oncogene 2003 22 8137 44 14603254 10.1038/sj.onc.1206911
Foo SY Nolan GP NF-kappaB to the rescue: RELs, apoptosis and cellular transformation Trends Genet 1999 15 229 35 10354583 10.1016/S0168-9525(99)01719-9
Hogue CW Cn3D: a new generation of three-dimensional molecular structure viewer Trends Biochem Sci 1997 22 314 6 9270306 10.1016/S0968-0004(97)01093-1
Malek S Huang DB Huxford T Ghosh S Ghosh G X-ray crystal structure of an IkappaBbeta x NF-kappaB p65 homodimer complex J Biol Chem 2003 278 23094 100 12686541 10.1074/jbc.M301022200
Merrick BA Zhou W Martin KJ Jeyarajah S Parker CE Selkirk JK Tomer KB Borchers CH Site-specific phosphorylation of human p53 protein determined by mass spectrometry Biochemistry 2001 40 4053 66 11300786 10.1021/bi002045i
Bessard AC Garay E Lacronique V Legros Y Demarquay C Houque A Portefaix JM Granier C Soussi T Regulation of the specific DNA binding activity of Xenopus laevis p53: evidence for conserved regulation through the carboxy-terminus of the protein Oncogene 1998 16 883 90 9484779 10.1038/sj.onc.1201598
Sakaguchi K Sakamoto H Xie D Erickson JW Lewis MS Anderson CW Appella E Effect of phosphorylation on tetramerization of the tumor suppressor protein p53 J Protein Chem 1997 16 553 6 9246643 10.1023/A:1026334116189
Akslen LA Morkve O Expression of p53 protein in cutaneous melanoma Int J Cancer 1992 52 13 6 1500218
Dreyfus DH Nagasawa M Pratt JC Kelleher CA Gelfand EW Inactivation of NF-kappaB by EBV BZLF-1-encoded ZEBRA protein in human T cells J Immunol 1999 163 6261 8 10570319
Bohuslav J Chen LF Kwon H Mu Y Greene WC p53 induces NF-kappaB activation by an IkappaB kinase-independent mechanism involving phosphorylation of p65 by ribosomal S6 kinase 1 J Biol Chem 2004 279 26115 25 15073170 10.1074/jbc.M313509200
Levitskaya J Sharipo A Leonchiks A Ciechanover A Masucci MG Inhibition of ubiquitin/proteasome-dependent protein degradation by the Gly-Ala repeat domain of the Epstein-Barr virus nuclear antigen 1 Proc Natl Acad Sci U S A 1997 94 12616 21 9356498 10.1073/pnas.94.23.12616
Stefanovic D Erikson E Pike LJ Maller JL Activation of a ribosomal protein S6 protein kinase in Xenopus oocytes by insulin and insulin-receptor kinase Embo J 1986 5 157 60 3514207
Bonsing BA Corver WE Gorsira MC van Vliet M Oud PS Cornelisse CJ Fleuren GJ Specificity of seven monoclonal antibodies against p53 evaluated with Western blotting, immunohistochemistry, confocal laser scanning microscopy, and flow cytometry Cytometry 1997 28 11 24 9136751 10.1002/(SICI)1097-0320(19970501)28:1<11::AID-CYTO2>3.0.CO;2-K
|
15969767
|
PMC1184076
|
CC BY
|
2021-01-04 16:03:32
|
no
|
BMC Immunol. 2005 Jun 21; 6:12
|
utf-8
|
BMC Immunol
| 2,005 |
10.1186/1471-2172-6-12
|
oa_comm
|
==== Front
BMC ImmunolBMC Immunology1471-2172BioMed Central London 1471-2172-6-131597812710.1186/1471-2172-6-13Methodology ArticleStandardization of cytokine flow cytometry assays Maecker Holden T [email protected] Aline [email protected]'Souza Patricia [email protected] Janice [email protected] Eva [email protected] Claire [email protected] Peter [email protected] Josephine [email protected] Omu [email protected] Miguel [email protected] Alexandre [email protected] Ian [email protected] Ruth [email protected] Megan [email protected] Jennifer [email protected] Janet [email protected] Laurie [email protected] C Lorrie [email protected] Elizabeth [email protected] Maria A [email protected] Kara [email protected] Sandra [email protected] Sophia [email protected] Gailet [email protected] Hazel [email protected] Ellen [email protected] Josephine [email protected] Clive [email protected] Marcus [email protected] Nolwenn [email protected] Jean [email protected] Lynda [email protected] Timothy [email protected] Barry [email protected] Mario [email protected] Richard [email protected] Vernon C [email protected] Kent [email protected] Giuseppe [email protected] Jill [email protected] Helen [email protected] Rafick P [email protected] BD Biosciences, San Jose, USA2 Université de Montreal and CANVAC, the Canadian Network for Vaccines and Immunotherapeutics, Montreal, Canada3 National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, USA4 Chelsea and Westminster Hospital and IAVI, London, UK5 Uganda Virus Research Institute and IAVI, Entebbe, Uganda6 Kenya AIDS Vaccine Initiative (KAVI), University of Nairobi, Kenya7 Centre Hospitalier Universitaire Vaudois and EUROVAC, Lausanne, Switzerland8 University of Washington and HVTN, Fred Hutchinson Cancer Research Center, Seattle, USA9 Duke University Medical Center and HVTN, Durham, USA10 Vaccine Research Center, National Institutes of Health, Bethesda, USA11 University of California, San Francisco, USA12 Merck and Co., West Point, USA13 University of Pennsylvania, Philadelphia, USA14 Sanofi Pasteur, Lyon, France15 Massachusetts General Hospital, Boston, USA16 National Institute for Communicable Diseases, Johannesburg, South Africa17 Henry Jackson Foundation, Rockville, USA2005 24 6 2005 6 13 13 4 12 2004 24 6 2005 Copyright © 2005 Maecker et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Cytokine flow cytometry (CFC) or intracellular cytokine staining (ICS) can quantitate antigen-specific T cell responses in settings such as experimental vaccination. Standardization of ICS among laboratories performing vaccine studies would provide a common platform by which to compare the immunogenicity of different vaccine candidates across multiple international organizations conducting clinical trials. As such, a study was carried out among several laboratories involved in HIV clinical trials, to define the inter-lab precision of ICS using various sample types, and using a common protocol for each experiment (see additional files online).
Results
Three sample types (activated, fixed, and frozen whole blood; fresh whole blood; and cryopreserved PBMC) were shipped to various sites, where ICS assays using cytomegalovirus (CMV) pp65 peptide mix or control antigens were performed in parallel in 96-well plates. For one experiment, antigens and antibody cocktails were lyophilised into 96-well plates to simplify and standardize the assay setup. Results (CD4+cytokine+ cells and CD8+cytokine+ cells) were determined by each site. Raw data were also sent to a central site for batch analysis with a dynamic gating template.
Mean inter-laboratory coefficient of variation (C.V.) ranged from 17–44% depending upon the sample type and analysis method. Cryopreserved peripheral blood mononuclear cells (PBMC) yielded lower inter-lab C.V.'s than whole blood. Centralized analysis (using a dynamic gating template) reduced the inter-lab C.V. by 5–20%, depending upon the experiment. The inter-lab C.V. was lowest (18–24%) for samples with a mean of >0.5% IFNγ + T cells, and highest (57–82%) for samples with a mean of <0.1% IFNγ + cells.
Conclusion
ICS assays can be performed by multiple laboratories using a common protocol with good inter-laboratory precision, which improves as the frequency of responding cells increases. Cryopreserved PBMC may yield slightly more consistent results than shipped whole blood. Analysis, particularly gating, is a significant source of variability, and can be reduced by centralized analysis and/or use of a standardized dynamic gating template. Use of pre-aliquoted lyophilized reagents for stimulation and staining can provide further standardization to these assays.
==== Body
Background
Enzyme-linked immunospot (ELISPOT) and cytokine flow cytometry (CFC) (or more specifically, intracellular cytokine staining (ICS)) are popular methods for single-cell analysis of antigen-specific T cell cytokine production. T cell production of IFNγ, and increasingly also IL-2, is taken as a measure of vaccine immunogenicity in experimental vaccine trials. Of the two types of assays, ICS has the advantage of a highly multiparametric read-out (flow cytometry) that allows for precise phenotyping of the responding T cell populations. It has also recently been adapted to a 96-well plate configuration [1,2], allowing for higher throughput analysis similar to that used for ELISPOT. However, while the precision of ELISPOT assays across sites has been recently documented [3], similar studies for ICS assays have been lacking.
Numerous phase I and phase II clinical trials have been initiated using candidate prophylactic HIV vaccines (reviewed in [4]). Many of these trials use ICS as part of their immune monitoring. While most current HIV trials are not powered to determine efficacy, and cytokine production has not been validated as a surrogate marker of protection from HIV infection or progression, there is nevertheless a desire to measure immunogenicity of candidate vaccines as well as safety in early clinical trials [5]. Because many different groups are performing immune monitoring for these clinical trials, there is currently a lack of standardization that would allow accurate comparisons of immunogenicity across candidate vaccines in different clinical trials.
There is some published literature on the intra-and inter-assay precision of ICS assays in whole blood [6]. These values were determined to be about 8% and 20% C.V., respectively. Guidelines for performance of ICS assays have also been recently published [7]. However, there are no existing data documenting the precision of ICS between laboratories, or comparing the precision of ICS using different sample types (e.g., whole blood versus cryopreserved PBMC). In order to allow more meaningful comparisons between laboratories and prioritization of emerging vaccine candidates, and thereby accelerate HIV vaccine development, this ICS standardization study was undertaken.
The objectives of the study were three-fold: (1) to assess the reproducibility of ICS assays using different sample types (shipped whole blood vs. cryopreserved PBMC); (2) to determine the inter-laboratory precision of ICS assays among major HIV vaccine clinical research laboratories; and (3) to improve the concordance of methodologies used in these laboratories. To achieve these objectives, joint experiments (Figure 1) were devised using (1) whole blood activated at a central site, then fixed, frozen and shipped to participating laboratories for processing and analysis; (2) fresh whole blood drawn at a central site and shipped to participating labs for activation, processing, and analysis; and (3) cryopreserved PBMC shipped from a central site to participating labs for activation, processing, and analysis. In the latter case, this experiment was also repeated with a larger number of participating laboratories, using pre-formatted microtiter plates containing lyophilised stimuli and lyophilised staining antibodies. In each experiment, raw data files were also sent by the participating labs to a central site for analysis, which was done using a dynamic gating template and batch analysis [1] (Figure 2).
Results
Activated, fixed, and frozen whole blood
In the first experiment, whole blood from three cytomegalovirus (CMV)-seropositive donors was activated, fixed, and frozen by the method described in Nomura et al.[6]. The blood was incubated for 6 hours in the presence of brefeldin A, either with no stimulus, Staphylococcal enterotoxin B (SEB), or a mixture of overlapping peptides corresponding to the CMV pp65 protein [8-10]. Aliquots of the frozen activated whole blood were then shipped to 9 laboratories for processing and analysis. The results, as reported by each site and also as determined by central, automated analysis of the raw data files, are summarized in Figure 3. For data reported by each site, the mean inter-lab C.V. was 55% for CD4 T cell responses and 32% for CD8 T cell responses. This is higher than the inter-assay C.V. previously reported for ICS assays performed at a single site [6]. However, when the raw data was centrally analyzed, the inter-lab C.V. was reduced to 24% for both CD4 and CD8 T cell responses, very similar to the inter-assay C.V. previously reported [6]. Thus, a large proportion of the site-to-site variability could be explained by differences in gating of the ICS data.
Fresh whole blood
In a second experiment, whole blood from three CMV-seropositive donors was shipped overnight to 6 U.S. labs for activation, processing, and analysis. This experiment was conducted twice, since the first trial was compromised by shipping delays. The results of the second trial are shown in Figure 4. As in Figure 3, the inter-lab C.V.'s were higher for data reported by each site, although the reduction due to centralized analysis was less dramatic than in the first experiment. Feedback on gating differences was provided to the labs between the first and second experiment, so the smaller effect of centralized analysis could be attributed to a progression of the individual sites toward a more uniform gating scheme. Also, the relatively high C.V. for CD4 T cell responses (42% even after centralized analysis) could be due to the low mean response to CMV peptide mix in two of the donors (donors were not identical in the different experiments). Since the C.V. varies inversely with the mean of the sample population, comparison of C.V.'s between experiments performed on different donors are subject to this confounding variable.
Cryopreserved PBMC
In a third experiment, PBMC were isolated from 6 CMV-seropositive donors and cryopreserved. Replicate cryopreserved vials were sent to each of 7 sites, where they were thawed, rested overnight, stimulated, processed, and analyzed. The post-thaw viability and recovery of the PBMC samples from each site are shown in Figure 5. Mean viabilities were >82%, and mean recoveries were >75% for each sample (determined by trypan blue exclusion). In general, viabilities were quite consistent across labs, while recoveries varied more widely, both between labs and between samples. This could be due in part to imprecise filling of the vials when they were initially frozen, which would impact the apparent recoveries calculated upon thawing. Furthermore, while a common thawing protocol was provided, no attempt was made to standardize counting methods or other factors that may impact the reproducibility of viability and recovery calculations across labs.
The ICS results from cryopreserved PBMC are shown in Figure 6. The inter-lab C.V.'s for this experiment averaged slightly lower than those for the whole blood experiments (25–32% for manual analysis; 23–25% for centralized automated analysis). Like the fresh whole blood experiment, the improvement in C.V. from centralized analysis was relatively small. When outlier samples with unusually low responses were checked for viability and recovery, they were not necessarily low in these parameters as well. In fact, there was no obvious relationship of viability or recovery with response, perhaps because viabilities and recoveries were virtually all within generally acknowledged limits of acceptability (>80% viability, >50% recovery) [11,12].
Cryopreserved PBMC with preconfigured lyophilised reagent plates
To expand upon the results of the third experiment using cryopreserved PBMC, a fourth experiment with this sample type was carried out using an enlarged cohort of participating laboratories (table 1). In addition, a protocol refinement was introduced to attempt to further reduce inter-lab variability: Peptide stimuli together with brefeldin A were provided in lyophilised form in appropriate wells of a microtiter plate, to provide simplified assay set-up; and lyophilised staining antibody cocktails were provided in the corresponding wells of a second microtiter plate. These latter were rehydrated and added to the cells in the first microtiter plate after fixation and permeabilization of the cells. This experiment also sought to compare two different types of staining antibody cocktails: the two cocktails used in the previous three experiments (IFNγ FITC/CD69 PE/CD4 PerCP-Cy5.5/CD3 APC and IFNγ FITC/CD69 PE/CD8 PerCP-Cy5.5/CD3); and one that combined CD4 and CD8 staining in a single sample, as well as adding IL-2 staining (CD4 FITC/IFNγ +IL-2 PE/CD8 PerCP-Cy5.5/CD3 APC).
The results of this experiment are shown in Figure 7 (note the change to a linear scale in this and the following figures). A set of peptides consisting of epitopes from CMV, EBV, and influenza (CEF) [13] was used as a positive control, due to restrictions on international shipment of SEB. CEF was expected to induce CD8, rather than CD4 responses, and indeed the CD4 responses to this control were very low or negative. The CD8 responses, while positive, were considerably lower than those seen with SEB in the previous experiments. Responses to CMV pp65 peptide mix were also not high in the donors used in this experiment, but were detectable in both CD4 and CD8 compartments. Despite the lower response means, the average C.V. for this experiment was roughly similar to the previous experiment, when comparing the same staining antibody cocktails (cocktails 2 and 3). For unknown reasons, the average C.V. for cocktail 1 (CD4/IFNγ +IL-2/CD8/CD3) was higher than that for cocktails 2 and 3, although the mean percentage of cytokine-positive cells was not significantly different. The addition of IL-2 in this cocktail did not significantly increase the mean percentage of cytokine-positive cells, as very few cells responding to CMV produce IL-2 without IFNγ [14]. This would not be expected to be true for all types of responses, however.
When data for this experiment were centrally analysed (Figure 7B), the average C.V.'s were considerably reduced, much like in the first experiment (Figure 3). This could reflect the fact that new laboratory sites had been added that had not yet standardized their gating strategies with the existing sites; thus more benefit was realized by centralized analysis. The difference in average C.V. between cocktail 1 and cocktails 2 and 3 was preserved even after centralized analysis. The mean C.V. for cocktails 2 and 3 was now 18%, the lowest variability seen in any of the experiments. For comparison, the mean inter-lab C.V. of the percent CD4+ or CD8+ cells in the unstimulated samples from this experiment was 3% and 7%, respectively (data not shown).
Lyophilised control cells
As a positive control in Experiment 4, a set of PBMC were SEB-activated, processed, stained, and then lyophilised in certain wells of the lyophilised antibody plates. They were hydrated and transferred to the plate containing activated cells, along with the staining antibodies. These cells served as a control for instrument setup and gating, since all the activation and processing steps were done centrally. The results reported by the individual sites for these cells are shown in the left panel of Figure 7C. Surprisingly, the average C.V. (20.5%) was only slightly lower than that for the rest of Experiment 4, in which cells were activated and processed independently by each site. However, when the control cell data were centrally analysed using a dynamic gating template (right panel), the C.V.'s were reduced to 3–7%. This reinforces the notion that the vast majority of inter-lab variability is due to gating.
Spontaneous cytokine production in the three sample types
"Background" or spontaneous cytokine production (subtracted from all data in Figures 3, 4, 6, and 7) is plotted for all experiments in Figure 8. Backgrounds were generally low. For all experiments combined, the mean CD4 background was 0.02% and the mean CD8 background was 0.05%. 98% of CD4 samples and 84% of CD8 samples had backgrounds =0.1%. There were a significant number of CD8 samples that exhibited high spontaneous cytokine production. However, the mean CD8 background was significantly higher than the mean CD4 background only in the activated, fixed whole blood experiment (p < 0.0005). When centralized automated analysis was applied to the data, backgrounds were not usually reduced. This indicates that the gains in reproducibility seen with centralized analysis were not simply due to reductions in background.
While CD4 backgrounds were very similar between experiments, CD8 backgrounds varied. The median CD8 background in the PBMC experiments was significantly lower than that of the frozen activated whole blood experiment (p < 0.0001) or the fresh whole blood experiment (p < 0.05). The differences were significant after centralized analysis as well. However, this could be due to the fact that different donors were used in the four experiments, rather than being due to any inherent difference between assay types. In experiment 4, the CD4 backgrounds for cocktail 1 were significantly higher than those for cocktail 2 (p < 0.05, data not shown), while there was no significant difference for CD8 backgrounds. This could be due to the inclusion of IL-2 in cocktail 1, which would be expected to be produced by more CD4+ than CD8+ cells, and thus contribute selectively to the CD4 background.
Discussion
This study examined the reproducibility of ICS assays across sites using different assay formats. It was not designed to compare ICS with other immune monitoring assays, comparisons of which have been published [15-21]. The current study used 96-well plate-based protocols exclusively, as these were considered more convenient, and have recently been validated against tube-based protocols for both PBMC and whole blood [1].
Lyophilised reagents in plates were used for Experiment 4. These have been extensively compared to liquid reagents ([22] and Figure 9) and shown to be largely equivalent. In addition to convenience of assay set-up, the lyophilised reagent plates offer long reagent stability, even at room temperature (>1 year, data not shown), and a potential reduction in errors caused by incorrect reagent addition. Intra-plate variability using lyophilised reagents was determined to be <10% in ICS assays (data not shown).
There are some potential drawbacks to the use of 96-well plates. One of these is the possibility of well-to-well contamination during the assay. This was observed in an initial subset of Experiment 4 (data not shown), in which some sites received lyophilised plates with SEB as a positive control. Some of these sites experienced high backgrounds in the negative control wells adjacent to the SEB-containing wells. It was later determined that cross-contamination probably occurred during the initial distribution of the antigens on the plates, and this was compounded by the fact that the donors used were unusually sensitive to SEB stimulation (responses >30% of CD4+ and CD8+ T cells). When SEB was replaced with CEF as a positive control, no such problems were noted. This experience suggests that the choice and placement of positive control wells on a plate deserves consideration.
The current study was designed to determine inter-lab variability in ICS assays. As such, there were no data "filters" applied to exclude potentially erroneous data or outliers. However, improved precision of ICS results might be obtained if certain acceptance criteria were applied before data were taken as valid. For example, a minimum number of collected events could be specified (sites in this study were asked to collect 10,000–40,000 CD4+ or CD8+ T cells per sample, or 60,000 CD3+ cells). This number of events was designed to yield precision levels that would minimize event number as a factor in inter-lab reproducibility. There could also be acceptance criteria based upon the absolute level of background, or the degree of reproducibility between duplicate samples, if run (the current study did not use duplicate samples).
It is also possible to apply statistics to derive further meaning from numerical results. For example, statistical tests could be used to determine whether a given response can be discriminated from a given background, for a particular number of events collected [23,24]. This can be given by a power calculation as follows:
N = [2*Pav(1-Pav)(Zα +Zβ)2]/Δ2
where N is the number of events in each sample needed for significance, Pav is the average proportion (between the background and test samples), and Δ is the difference between these two proportions. The term (Zα +Zβ)2 is referred to as a power index, and varies depending upon the desired power and p value. For example, (Zα +Zβ)2 = 23.9 for 99% power and p < 0.005 [23].
In addition, a confidence interval could be derived around the difference of the test result and the negative control [24], in order to allow discrimination of significant differences between various samples. Other statistical methods have also been employed in order to determine cut-off values for positive responses in ICS [25,26]. No attempt was made in the current study to define which results were positive, as all data were reported objectively, and all donors were known to be CMV seropositive.
Examination of the data from Figures 3, 4, 6, and 7 suggests that samples with a low number of cytokine-positive cells had higher variability than samples with a high number of cytokine-positive cells. The relationship of response level and C.V. is summarized in Table 2 for all assays (CD4 and CD8, whole blood and PBMC) considered together. These data emphasize the difficulty of achieving precise results at response levels of less than 0.1% of CD4 or CD8 T cells. For these samples, collecting even more events than what was suggested would be expected to improve precision, per the discussion above.
The average C.V. across the four experiments is summarized in Table 3. These data are confounded by the fact that different donors and different laboratories participated in the four experiments. However, variability due to individual analysis can be excluded by comparing only centrally analysed data (bottom row of Table 3). Assuming no effect from the other confounding variables, we found that Experiment 4 (cryopreserved PBMC with lyoplates) yielded a significantly lower average C.V. than Experiment 2 (shipped whole blood) (p < 0.05). Also, the average C.V. of centrally analysed data from all experiments was significantly lower than that of individually analysed data (p < 0.0001). This highlights the amount of variability in each experiment that is due to gating differences between sites.
Mitigation of gating variability was achieved in these experiments by centralized analysis with a dynamic gating template (see Figure 2B). The dynamic gating template allowed for more automated, batch analysis of the data. Once such a template was created and optimized (see Materials and Methods section for description), it could also have been provided to individual sites in order to yield the same results. It is further possible that similar results could be achieved by manual analysis, provided it was done by a single operator. Standardization of gating techniques, in the absence of centralized analysis or dynamic gating templates, could also improve precision. The improvement in C.V. made by centralized analysis was most marked in the first experiment, and progressively less in experiments 2 and 3, perhaps because of standardization of gating among sites over time. Experiment 4 included many new sites, and the improvement in C.V. from centralized analysis was again more marked.
Because the C.V. varies as a function of the response level (Table 2), it is possible that differences in mean C.V. between assay formats were due to the number of low versus high responders in each experiment (since different donors were used in the four experiments). Also, the C.V. is highly sensitive to small changes in the mean, when the mean is a very low number. Therefore, an analysis of S.D. versus mean was also performed for the four experiments (Figure 10). This data confirms the data of Table 3, indicating that the three assay formats showed grossly similar reproducibility. However, when analysis variability was removed, cryopreserved PBMC assays appeared to be slightly more reproducible than shipped whole blood assays. This seemed especially apparent in experiment 4, where lyophilised reagents were used.
In addition to differences in reproducibility, the various assay formats have other benefits and drawbacks as well. Cryopreserved PBMC are much more amenable to peptide (and superantigen) stimulation than to whole protein stimulation [9]; while whole blood assays are equally amenable to stimulation with either type of antigen. Also, consistently good cryopreservation of PBMC at multiple clinical sites is difficult to achieve, but highly important for achieving reproducible results with PBMC [27,28] (DeLaRosa et al., manuscript in preparation). This could become less of a factor if a stabilizing matrix for preserving whole blood or PBMC function during shipping were discovered. All in all, the choice of assay format for a clinical trial will depend not only upon considerations of assay precision, but also upon the type of antigen(s) used and the capabilities of the participating clinical sites.
The use of lyophilised reagent plates appeared to reduce inter-lab variability. This conclusion cannot be drawn with certainty, because different participating laboratories and different donors were used between experiments 3 and 4. However, it is intriguing to note that, when centrally analysed data was compared (to remove gating as a source of variability), the mean C.V.'s of experiment 4 were the lowest of all four experiments (18%, Table 3). This is despite the fact that the donors and stimuli used in experiment 4 resulted in lower mean response levels, which should tend to increase the C.V. This is also borne out by the analysis of Figure 11B, where the results for experiment 4 appeared to be generally closer to the theoretical minimum SD than did the results for the other experiments.
With the possibility of achieving inter-laboratory C.V.'s of less than 20%, even with relatively low responses, ICS compares favourably to ELISPOT, for which interassay C.V.'s of 17–18% for PHA and 55–65% for Candida have been reported [29,30]. ICS is also comparable to cytokine ELISA, the latter having reported interassay C.V.'s of <25% [31,32]. Phenotypic staining, such as used for CD4 counting, can achieve higher precision levels than functional assays, and averaged around 10% C.V. in one multisite study [33]. For comparison, the inter-lab C.V. of the CD4+ or CD8+ cell percentages was around 5% in experiment 4 of the present study (data not shown). CD4 counting precision has also been shown to be dependent upon the number of events collected, gating, and use of automated analysis [33,34]. Since functional assays are subject to more variables than phenotypic staining, the ability to achieve precision levels such as those reported here should be considered favourable. ICS could thus be a viable tool for comparing immune responses even across clinical trials, provided the methodology was standardized.
Conclusion
ICS assays could be performed with inter-laboratory C.V.'s of approximately 20% at response levels of >0.5%, and C.V.'s of approximately 25–30% at response levels of 0.1–0.5%. The C.V. increased further at response levels of =0.1%. A significant portion of inter-laboratory variability could be eliminated by use of centralized analysis and/or a dynamic gating template.
Whole blood and cryopreserved PBMC showed grossly similar levels of reproducibility. However, when analysis variability was removed, cryopreserved PBMC processed with lyophilized reagents showed significantly better reproducibility than shipped whole blood. Shipped whole blood assays were also subject to data loss when samples were not delivered in a timely fashion.
Background cytokine production was mostly =0.05% for both CD4 and CD8 cells. While CD8 backgrounds were lower in cryopreserved PBMC than in whole blood, this could have been due to the use of different donors in the four experiments. With the high viabilities and recoveries obtained for cryopreserved PBMC in this study, there was no obvious relationship between viability/recovery and response.
The use of microtiter plates containing lyophilised reagents simplified the ICS protocol, and appeared to improve assay reproducibility. This format lends itself to international shipping of reagents (because there is no need for refrigeration), and also to larger clinical trials (because of the stability of the lyophilised reagents). It is also a way to reduce the chance of pipetting errors, because the plates are pre-formatted.
The results of this study indicate that ICS assays can be reasonably standardized between sites, but that considerations of sample format and expected response levels can influence the precision of the results. These data should guide comparisons of ICS results between different groups or in different clinical trials.
Methods
Whole blood preparation
Heparinized whole blood was collected from healthy CMV seropositive volunteers for experiments 1 and 2. For experiment 1, the blood was activated in 15 mL conical tubes according to the method of Nomura et al.[6]. Activated blood was treated with 2 mM final concentration of EDTA for 15 minutes at room temperature, then 10 volumes of FACS Lysing Solution (BD Biosciences, San Jose, CA) were added. After 10 minutes at room temperature, the tubes were frozen at -80°C, then shipped to participating laboratories on dry ice. The protocol used by each laboratory for handling these samples is provided in Additional File 1.
For experiment 2, 5 mL of heparinized whole blood was overnight shipped in an insulated container at ambient temperature to each participating lab. The protocol used by each lab for handling these samples is provided in Additional File 2.
PBMC preparation and cryopreservation
For experiments 3 and 4, PBMC from leukapheresis of CMV seropositive donors were isolated using Ficoll gradient separation. They were then cryopreserved according to a standard protocol (Disis et al., submitted for publication). These cryopreserved PBMC were shipped to participating labs using liquid nitrogen dry shippers. The protocol used by each lab for thawing and processing of these cells is provided as Additional Files 3 and 4.
Instrumentation and setup
The flow cytometry instrumentation used in this study included 12 BD FACS Caliburs (BD Biosciences), 3 BD LSRIIs (BD Biosciences), and 1 CyAn (Dako Cytomation, Fort Collins, CO). Instrument setup was at the discretion of the individual laboratory, and was either manual (using isotype control stained cells to set PMT voltages, and single-stained cells to set compensation) or automated (using BD FACSComp software and BD Calibrite beads (BD Biosciences)). In some labs, automated setup was followed by manual adjustment using stained cells as above.
Dynamic gating templates
Original FCS files from each site were sent to BD Biosciences for analysis using a dynamic gating template (Figure 2B). This template was built using "Snap-To Gating" and "Tethering" tools available in CellQuest Pro software (BD Biosciences). The shape of the snap-to gates is determined by a clustering algorithm, and this algorithm allows for their movement from sample to sample in a data-dependent manner. The size and amount of allowable movement of each snap-to gate was adjusted by inspection of a subset of the files to be used, with iterative changes being made until the template performed as desired. The template was then used, without further adjustment, on all the files of a given experiment. Since the template was generated in CellQuest Pro software, only files generated on FACS Calibur instruments were analyzable by this method.
Statistical analyses
The %CV was calculated as 100*SD/mean for each sample, from the percentage of cytokine-positive cells reported by each laboratory or derived from centralized analysis of that sample. The mean CV for each experiment was taken as the average of all the individual sample CVs. Statistical significance of differences in the average CV between experiments was calculated using a Kruskal-Wallis test, with Dunn's Multiple Comparison test to determine where significant differences were found. The significance of the difference between individually and centrally analyzed data was calculated by comparing the aggregate CVs of all samples from all experiments using a Wilcoxin signed rank test for matched pairs. A two-tailed Student t test was used to calculate significance of differences in background within or between experiments.
Authors' contributions
HTM compiled the data and drafted the manuscript. AR coordinated the study. P.D'S. and J.D. secured logistic and financial support. All authors helped design the study, supervised and/or carried out the experiments, and provided editorial comments and assistance.
Supplementary Material
Additional File 1
Protocol for fixed, activated whole blood (Experiment 1)
Click here for file
Additional File 2
Protocol for shipped whole blood (Experiment 2)
Click here for file
Additional File 3
Protocol for cryopreserved PBMC (Experiment 3)
Click here for file
Additional File 4
Protocol for cryopreserved PBMC with lyophilised antigen and antibody plates (Experiment 4)
Click here for file
Acknowledgements
The authors acknowledge the National Institute for Allergy and Infectious Diseases and CANVAC for financial and logistical support, and BD Biosciences for providing reagents. They also thank Doug Haney (BD Biosciences) for advice on statistical analysis.
Figures and Tables
Figure 1 Experimental design. (A) Schematic of protocol for Experiments 1–3, performed using liquid antigens and antibodies. (B) Schematic of protocol for Experiment 4, performed using lyophilised antigen and antibody plates.
Figure 2 Manual versus automated gating templates. (A) Representative manual analysis of a CEF-stimulated sample from Experiment 4. Sequential gates on small lymphocytes, CD3+ cells, and CD3+CD8+. cells are applied and the percent CD69+IFNg+ cells are determined from a plot gated on all of these regions. (B) Dynamic gating template for the same data file as above. Sequential dynamic gates ("Snap-To" gates) are applied as above, except that negative populations are also gated so as to provide a boundary for the movement of the positive region. The percent CD69+IFNg+ cells obtained is very similar to that obtained by manual gating in this example, since manual gating was performed so as to include CD3dim and CD8dim cells.
Figure 3 Results of Experiment 1 (fixed activated whole blood). IFNγ-positive cells in response to SEB or CMV pp65 peptide mix are expressed as a percentage of CD4+ or CD8+ T cells. Results from each site are indicated as a circle, with median responses for each sample (105, 950, and 1040) indicated by a horizontal bar. The C.V. for each sample is listed across the top of each panel, along with the mean C.V. for that set of samples.
Figure 4 Results of Experiment 2 (shipped whole blood). IFNγ-positive cells in response to SEB or CMV pp65 peptide mix are expressed as a percentage of CD4+ or CD8+ T cells. Results from each site are indicated as a circle, with median responses for each sample (105, 1040, and 1090) indicated by a horizontal bar. The C.V. for each sample is listed across the top of each panel, along with the mean C.V. for that set of samples.
Figure 5 Viabilities and recoveries for cryopreserved PBMC samples used in Experiment 3. Error bars represent SEM of the 6 sites participating.
Figure 6 Results of Experiment 3 (cryopreserved PBMC). IFNγ-positive cells in response to SEB or CMV pp65 peptide mix are expressed as a percentage of CD4+ or CD8+ T cells. Results from each site are indicated as a circle, with median responses for each sample indicated by a horizontal bar. Sample names are listed along the X axis. The C.V. for each sample is listed across the top of each panel, along with the mean C.V. for that set of samples.
Figure 7 Results of Experiment 4 (cryopreserved PBMC with lyophilized reagents). Cytokine-positive cells in response to CEF peptides or CMV pp65 peptide mix are expressed as a percentage of CD4+ or CD8+ T cells. Results from each site are indicated as a circle, with median responses for each sample indicated as a horizontal bar. Sample names are listed along the X axis. The C.V. for each positive sample is listed across the top of each panel, along with the mean C.V. for that set of samples. (A) Data reported by individual sites. (B) Data after centralized analysis using a dynamic gating template. Cocktail 1 consisted of CD4 FITC/IFNγ +IL-2 PE/CD8 PerCP-Cy5.5/CD3 APC. Cocktail 2 consisted of IFNγ FITC/CD69 PE/CD4 PerCP-Cy5.5/CD3 APC, and cocktail 3 consisted of IFNγ FITC/CD69 PE/CD8 PerCP-Cy5.5/CD3 APC. (C) Control cell data from Experiment 4. Activated, processed, and stained PBMC were lyophilised and run as controls by each site in Experiment 4. These cells had been stained with cocktail 1, 2, or 3. The left panel shows data reported by individual sites; the right panel, data after centralized analysis using a dynamic gating template. Most of the inter-lab variability was removed by this method of analysis.
Figure 8 Background cytokine-producing cells by experiment. Cytokine-positive cells in the absence of stimulus are expressed as a percentage of CD4+ or CD8+ T cells. Each symbol represents the background from a single sample processed by a single site. Medians are shown by a horizontal bar. Data from experiment 4 are for cocktails 2 and 3 only, to be comparable with experiments 1–3.
Figure 9 Comparison of liquid and lyophilised reagents. Comparative results are shown with backgrounds subtracted; no significant differences in backgrounds were seen with liquid versus lyophilised reagents (data not shown). (A) Data from one site that compared cocktail 1 (CD4 FITC/IFNγ +IL-2 PE/CD8 PerCP-Cy5.5/CD3 APC) in liquid and lyophilised form in Experiment 4. Black bars indicate liquid antibodies, grey bars indicate lyophilised antibodies. Error bars indicate SEM of duplicate wells. (B) Combined comparison of liquid antigen + liquid antibodies versus lyophilised antigen + lyophilised antibodies. Whole blood was activated with either SEB or pp65 peptide mix, and the percentage of IFNγ+ cells (CD4+ or CD8+) were compared with liquid versus lyophilised reagents (left panel). A similar comparison was made for the mean fluorescence intensity (MFI) of IFNγ+ cells (CD4+ or CD8+) (right panel). Similar results were obtained using PBMC (not shown).
Figure 10 S.D. versus mean for all assays. A linear relationship between S.D. and mean is expected based upon counting statistics [23]. This expected relationship (for a data set of 40,000 events) is shown by the solid black line. The actual data from the four experiments is shown in the symbols. Data from experiment 4 are for cocktails 2 and 3 only, to be comparable with experiments 1–3. The difference (in the Y dimension) between the data points and the solid line represents variability from sources other than counting statistics. Note that the data points for all three assay types cluster together, indicating that variability is similar for all three assay types. When individual analysis variability is removed (B), there is a slight tendency toward lower variability with cryopreserved PBMC (solid circles), and higher variability with shipped whole blood (open squares). The tendency toward lower variability is more pronounced in the experiment using cryopreserved PBMC with lyophilised reagents (panel B, open circles).
Table 1 Study participants and institutions.
Institution (Consortium) Participants Participated in:
Exp. 1 (fixed activated blood) Exp. 2 (shipped whole blood) Exp. 3 (cryo-preserved PBMC) Exp. 4 (cryo-preserved PBMC with lyoplates)
University of Montreal (CANVAC) Rafick Sekaly, Eva Roig, Claire Landry x x x
Chelsea and Westminster Hospital (IAVI) Jill Gilmour, Peter Hayes x x x
Uganda Virus Research Institute (IAVI) Josephine Birungi, Omu Anzala x
Centre Hospitalier Universitaire Vaudois (EUROVAC) Giuseppe Pantaleo, Alexandre Harari, Miguel Garcia x x x
Fred Hutchison Cancer Research Center (HVTN) Helen Horton ,Ruth Baydo, Ian Frank x x x x
Duke University (HVTN) Kent Weinhold, Janet Ottinger, Megan Baker, Jennifer Holbrook x x x
Vaccine Research Center, NIH Mario Roederer, Richard Koup, Laurie Lamoreaux x x x x
Merck and Co. Timothy Tobery, Lynda Tussey, Kara Punt x x x x
University of California, San Francisco Barry Bredt, Elizabeth Sinclair, Lorrie Epling x x x
BD Biosciences Vernon Maino, Holden Maecker, Maria Suni x x x x
Sanofi Pasteur Nolwenn Nougarede, Sophia El-Bahi x
National Inst. Communicable Diseases, South Africa Clive Gray, Hazel Maila x
Massachusetts General Hospital Marcus Altfeld, Gailet Alter x
University of Pennsylvania Jean Boyer, Sandra Calarota x
Henry Jackson Foundation Josephine Cox, Ellen Kuta x
Financial and Operational Support: Aline Rinfret, CANVAC; Patricia D'Souza, NIAID, NIH; Janice Darden, NIAID, NIH
Table 2 Percent C.V. by mean percent cytokine-positive T cells.
Mean % cytokine-positive cells = 0.1% 0.1 – 0.5% >0.5%
Number of samples in range 6 6 35
Average percent C.V. Individual (Manual) Analysis 71% 31% 21%
Central (Automated) Analysis 56% 25% 18%
Table 3 Percent C.V. by assay format.
Assay Type Fixed Activated Blood Shipped Whole Blood Cryopreserved PBMC Cryopreserved PBMC with lyophilised reagents
Number of samples 12 12 24 9
Average percent C.V. Individual (Manual) Analysis 44% 38% 28% 39%
Central (Automated) Analysis 24% 31% 23% 18%
==== Refs
Suni MA Dunn HS Orr PL deLaat R Sinclair E Ghanekar SA Bredt BM Dunne JF Maino VC Maecker HT Performance of plate-based cytokine flow cytometry with automated data analysis BMC Immunology 2003 4 9 12952557
Maecker HT Hawley TS, Hawley RG Cytokine flow cytometry Flow Cytometry Protocols Meth Molec Biol Walker J 2004 2nd Totowa, NJ , Humana Press 95 107
Cox JH Ferrari G Kalams SA Lopaczynski W Oden N Group ELISPOTS Results of an ELISPOT proficiency panel conducted in 11 laboratories participating in international human immunodeficiency virus type 1 vaccine trials. AIDS Res Hum Retroviruses 2005
McMichael AJ Hanke T HIV vaccines 1983-2003 Nat Med 2003 9 874 880 12835708
Pantaleo G Koup RA Correlates of immune protection in HIV-1 infection: what we know, what we don't know, what we should know Nat Med 2004 10 806 810 15286782
Nomura LE Walker JM Maecker HT Optimization of whole blood antigen-specific cytokine assays for CD4(+) T cells Cytometry 2000 40 60 68 10754518
Landay A Performance of single cell immune response assays NCCLS Standards and Guidelines 2004 Wayne, PA , NCCLS I/LA26 A
Kern F Faulhaber N Frommel C Khatamzas E Prosch S Schonemann C Kretzschmar I Volkmer-Engert R Volk HD Reinke P Analysis of CD8 T cell reactivity to cytomegalovirus using protein- spanning pools of overlapping pentadecapeptides Eur J Immunol 2000 30 1676 1682 10898504
Maecker HT Dunn HS Suni MA Khatamzas E Pitcher CJ Bunde T Persaud N Trigona W Fu TM Sinclair E Bredt BM McCune JM Maino VC Kern F Picker LJ Use of overlapping peptide mixtures as antigens for cytokine flow cytometry J Immunol Methods 2001 255 27 40 11470284
Kern F Bunde T Faulhaber N Kiecker F Khatamzas E Rudawski IM Pruss A Gratama JW Volkmer-Engert R Ewert R Reinke P Volk HD Picker LJ Cytomegalovirus (CMV) phosphoprotein 65 makes a large contribution to shaping the T cell repertoire in CMV-exposed individuals J Infect Dis 2002 185 1709 1716 12085315
Kleeberger CA Lyles RH Margolick JB Rinaldo CR Phair JP Giorgi JV Viability and recovery of peripheral blood mononuclear cells cryopreserved for up to 12 years in a multicenter study Clin Diagn Lab Immunol 1999 6 14 19 9874657
Weinberg A Zhang L Brown D Erice A Polsky B Hirsch MS Owens S Lamb K Viability and functional activity of cryopreserved mononuclear cells Clin Diagn Lab Immunol 2000 7 714 716 10882680
Currier JR Kuta EG Turk E Earhart LB Loomis-Price L Janetzki S Ferrari G Birx DL Cox JH A panel of MHC class I restricted viral peptides for use as a quality control for vaccine trial ELISPOT assays J Immunol Methods 2002 260 157 172 11792386
De Rosa SC Lu FX Yu J Perfetto SP Falloon J Moser S Evans TG Koup R Miller CJ Roederer M Vaccination in humans generates broad T cell cytokine responses J Immunol 2004 173 5372 5380 15494483
Kuzushima K Hoshino Y Fujii K Yokoyama N Fujita M Kiyono T Kimura H Morishima T Morishima Y Tsurumi T Rapid determination of Epstein-Barr virus-specific CD8(+) T-cell frequencies by flow cytometry Blood 1999 94 3094 3100 10556194
Moretto WJ Drohan LA Nixon DF Rapid quantification of SIV-specific CD8 T cell responses with recombinant vaccinia virus ELISPOT or cytokine flow cytometry Aids 2000 14 2625 2627 11101083
Asemissen AM Nagorsen D Keilholz U Letsch A Schmittel A Thiel E Scheibenbogen C Flow cytometric determination of intracellular or secreted IFNgamma for the quantification of antigen reactive T cells J Immunol Methods 2001 251 101 108 11292486
Pahar B Li J Rourke T Miller CJ McChesney MB Detection of antigen-specific T cell interferon gamma expression by ELISPOT and cytokine flow cytometry assays in rhesus macaques J Immunol Methods 2003 282 103 115 14604545
Sun Y Iglesias E Samri A Kamkamidze G Decoville T Carcelain G Autran B A systematic comparison of methods to measure HIV-1 specific CD8 T cells J Immunol Methods 2003 272 23 34 12505709
Whiteside TL Zhao Y Tsukishiro T Elder EM Gooding W Baar J Enzyme-linked immunospot, cytokine flow cytometry, and tetramers in the detection of T-cell responses to a dendritic cell-based multipeptide vaccine in patients with melanoma Clin Cancer Res 2003 9 641 649 12576430
Karlsson AC Martin JN Younger SR Bredt BM Epling L Ronquillo R Varma A Deeks SG McCune JM Nixon DF Sinclair E Comparison of the ELISPOT and cytokine flow cytometry assays for the enumeration of antigen-specific T cells J Immunol Methods 2003 283 141 153 14659906
Dunne JF Maecker HT Automation of cytokine flow cytometry assays J Assoc Lab Automation 2004 9 5 9
Motulsky H Intuitive Biostatistics 1995 Oxford , Oxford Univ. Press 199 200
Maecker HT Disis ML The role of immune monitoring in evaluating cancer immunotherapy Immunotherapy of Cancer Cancer Drug Discovery and Development 2005 Totowa, NJ , Humana Press
Trigona WL Clair JH Persaud N Punt K Bachinsky M Sadasivan-Nair U Dubey S Tussey L Fu TM Shiver J Intracellular staining for HIV-specific IFN-gamma production: statistical analyses establish reproducibility and criteria for distinguishing positive responses J Interferon Cytokine Res 2003 23 369 377 14511463
Sinclair E Black D Epling CL Carvidi A Josefowicz SZ Bredt BM Jacobson MA CMV antigen-specific CD4+ and CD8+ T cell IFNgamma expression and proliferation responses in healthy CMV-seropositive individuals Viral Immunol 2004 17 445 454 15357911
Weinberg A Wohl DA Brown DG Pott GB Zhang L Ray MG van der Horst C Effect of cryopreservation on measurement of cytomegalovirus-specific cellular immune responses in HIV-infected patients J Acquir Immune Defic Syndr 2000 25 109 114 11103040
Costantini A Mancini S Giuliodoro S Butini L Regnery CM Silvestri G Montroni M Effects of cryopreservation on lymphocyte immunophenotype and function J Immunol Methods 2003 278 145 155 12957403
Lathey J Preliminary steps toward validating a clinical bioassay: case study of the ELISpot assay Biopharm Intl 2003 March 42 50
Lathey J Sathiyaseelan J Matijevic M Hedley ML Validation of pretrial ELISspot measurements BioProcess Intl 2003 Sept. 34 41
Borg L Kristiansen J Christensen JM Jepsen KF Poulsen LK Evaluation of accuracy and uncertainty of ELISA assays for the determination of interleukin-4, interleukin-5, interferon-gamma and tumor necrosis factor-alpha Clin Chem Lab Med 2002 40 509 519 12113298
Findlay JW Smith WC Lee JW Nordblom GD Das I DeSilva BS Khan MN Bowsher RR Validation of immunoassays for bioanalysis: a pharmaceutical industry perspective J Pharm Biomed Anal 2000 21 1249 1273 10708409
Reimann KA O'Gorman MR Spritzler J Wilkening CL Sabath DE Helm K Campbell DE Multisite comparison of CD4 and CD8 T-lymphocyte counting by single- versus multiple-platform methodologies: evaluation of Beckman Coulter flow-count fluorospheres and the tetraONE system.The NIAID DAIDS New Technologies Evaluation Group Clin Diagn Lab Immunol 2000 7 344 351 10799444
Koepke JA Landay AL Precision and accuracy of absolute lymphocyte counts Clin Immunol Immunopathol 1989 52 19 27 2656016
|
15978127
|
PMC1184077
|
CC BY
|
2021-01-04 16:03:32
|
no
|
BMC Immunol. 2005 Jun 24; 6:13
|
utf-8
|
BMC Immunol
| 2,005 |
10.1186/1471-2172-6-13
|
oa_comm
|
==== Front
BMC MedBMC Medicine1741-7015BioMed Central London 1741-7015-3-111597514710.1186/1741-7015-3-11Research ArticleThe influence of long chain polyunsaturate supplementation on docosahexaenoic acid and arachidonic acid in baboon neonate central nervous system Diau Guan-Yeu [email protected] Andrea T [email protected] Eszter A [email protected] Vasuki [email protected] Peter W [email protected] J Thomas [email protected] Division of Nutritional Sciences, Cornell University, Ithaca, New York, USA2 College of Veterinary Medicine, Cornell University, Ithaca, New York USA3 Division of Pediatric Surgery, Department of Surgery, Tri-Service General Hospital (TSGH), National Defense Medical Center (NDMC), 325 Chenggung Rd, 2 Sec, Naihu, Taipei 114, Taiwan, Republic of China4 Dept of Nutritional Sciences, University of California, Berkeley, CA, USA5 Brandeis University, Foster Biomedical Laboratory, Waltham, MA, USA6 Dept of Obstetrics and Gynecology, University of Texas Health Science Center, San Antonio, TX, USA2005 23 6 2005 3 11 11 16 2 2005 23 6 2005 Copyright © 2005 Diau et al; licensee BioMed Central Ltd.2005Diau et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Docosahexaenoic acid (DHA) and arachidonic acid (ARA) are major components of the cerebral cortex and visual system, where they play a critical role in neural development. We quantitatively mapped fatty acids in 26 regions of the four-week-old breastfed baboon CNS, and studied the influence of dietary DHA and ARA supplementation and prematurity on CNS DHA and ARA concentrations.
Methods
Baboons were randomized into a breastfed (B) and four formula-fed groups: term, no DHA/ARA (T-); term, DHA/ARA supplemented (T+); preterm, no DHA/ARA (P-); preterm and DHA/ARA supplemented (P+). At four weeks adjusted age, brains were dissected and total fatty acids analyzed by gas chromatography and mass spectrometry.
Results
DHA and ARA are rich in many more structures than previously reported. They are most concentrated in structures local to the brain stem and diencephalon, particularly the basal ganglia, limbic regions, thalamus and midbrain, and comparatively lower in white matter. Dietary supplementation increased DHA in all structures but had little influence on ARA concentrations. Supplementation restored DHA concentrations to levels of breastfed neonates in all regions except the cerebral cortex and cerebellum. Prematurity per se did not exert a strong influence on DHA or ARA concentrations.
Conclusion
1) DHA and ARA are found in high concentration throughout the primate CNS, particularly in gray matter such as basal ganglia; 2) DHA concentrations drop across most CNS structures in neonates consuming formulas with no DHA, but ARA levels are relatively immune to ARA in the diet; 3) supplementation of infant formula is effective at restoring DHA concentration in structures other than the cerebral cortex. These results will be useful as a guide to future investigations of CNS function in the absence of dietary DHA and ARA.
==== Body
Background
Docosahexaenoic acid (DHA) is the most unsaturated fatty acid in mammalian tissue. It is found at particularly high concentration in retina [1] and cerebral cortex [2], concentrated mainly in serine and ethanolamine phosphoglycerides [3]. Early observations led to studies showing that low tissue DHA induced by dietary deficiency of n-3 fatty acids results in compromised retinal function as reflected by poor electroretinogram parameters [4-6] and altered cognitive function [7]. Several hypotheses have been proposed to explain the molecular role of DHA, including its high degree of molecular flexibility as a component of membrane phospholipids [8], improvement of G-protein-coupled signaling [9,10], more favorable interaction with integral membrane proteins [11], its ability as a free fatty acid to stabilize the electrical activity of ion channels [12], and, very recently, as a precursor for compounds protective of CNS function during ischemia [13].
The majority of DHA studies have focused on its role in the cerebral cortex. There are few data on DHA concentrations in most deep CNS structures at a gross level. For instance, there are no studies of DHA concentrations in the basal ganglia. These complex structures are involved in a wide array of integrative functions involving motor coordination, integration of visual signals, and psychiatric and personality phenomena [14]. The globus pallidus, caudate nucleus and putamen suffer specific, massive loss of neurons in Huntington's disease, resulting in chorea, dementia and psychiatric disturbances [15]. In Parkinson's disease, elimination of excessive output of the inner segment of the globus pallidus by surgical removal or lesion of the subthalamic nucleus has re-emerged as a treatment for elimination of tremors [16]. The mechanism of neural loss in these diseases remains elusive. Similarly, there are few data on the fatty acid compositions of the limbic regions, the thalamus or the midbrain.
Mammals obtain DHA nutritionally in two forms, either as preformed DHA, found primarily in marine foods, or via the dietary essential polyunsaturated fatty acid (PUFA) α-linolenic acid (18:3n-3), from which DHA is synthesized in various tissues. In vitro data suggest that in the CNS, glia and cerebral endothelial cells, but not neurons, synthesize DHA from 18:3n-3 and other precursor n-3 fatty acids [17]. Once formed, phospholipids containing DHA are rapidly incorporated into neurons, and are concentrated in synaptosomes where they influence neurotransmitter content [18].
There has been intense interest in DHA and its influence over function in the developing CNS. Randomized clinical trials have shown that infant formulas with DHA improve retinal and CNS function compared to formulas containing only 18:3n-3 [19]. All studies also include arachidonic acid (ARA) to prevent any possible negative effects on growth reported in early studies [20]. Virtually all clinical developmental studies in infants have examined visual and cognitive outcomes [19], and are limited to sampling of blood for assessment of tissue responses to supplementation, rather than the target CNS tissues.
Our first purpose in this report is to assess DHA and ARA concentrations in normal, four week old baboons within many CNS structures not previously studied. We also test the hypothesis that supplemental DHA and ARA in infant formula maintains CNS tissue concentrations similar to those that are found for breastfed baboons. We test these hypotheses in both term and preterm baboons at identical post-conceptual ages.
Methods
Animals
The Cornell Institutional Animal Care and Use Committee approved the care of animals and the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) approved the facility. Female pregnant baboons (Papio cynocephalus) obtained from the Southwest Foundation for Biomedical Research (San Antonio, TX) were used in this study. After confirmation of pregnancy, 29 pregnant baboons were transported to Cornell University (Ithaca, NY). The animals were involved in a series of studies reported recently [21-24] and involving only manipulation of dietary LCP and/or prematurity, as presented here. A complete veterinary examination was performed upon arrival in Ithaca. Pregnant baboons were housed in individual cages in sight of at least one other baboon.
Breastfed neonates were housed with their mothers in a controlled-access nursery where the temperature and humidity were 28°C and 50% respectively, and with a 14 h light and 10 h dark cycle, and consumed breastmilk exclusively during the study period. All other neonates, the "bottle-fed" groups, were removed and placed in a nursery and consumed commercially available infant formula without (minus, "-") or supplemented with (plus, "+") long chain polyunsaturates (LCP).
Study design and diets
A diagram showing the timeline for CS and for euthanasia is presented in Figure 1. All pregnant adult animals consumed a commercial chow, yams, and fruits. The breastfed (B) and term groups (T+ and T-) were delivered vaginally after spontaneous labor and averaged a birthweight of 863 grams, and there were no significant differences among groups. Further details are presented in Table 1.
Figure 1 Study timeline for the formula groups. Time is represented as days from conception (conceptual age) or adjusted to time of normal term birth. The preterm (P) groups were taken at -24 days of gestational age by Cesarean section, while the term (T) groups were born spontaneously. All groups were euthanized at 28 days adjusted age, including the breastfed group (B).
Table 1 Characteristics of baboon neonates.
Breastfed B Term, no LCP T- Term, LCP T+ Preterm, no LCP P- Preterm, LCP P+
Gender 2F, 3M 5F, 2M 2F, 5M 3F, 2M 3F, 2M
Gestational age at CSa (d) Spontaneous Labor Spontaneous labor Spontaneous labor 156 ± 5 154 ± 2
Age at euthanasia (d)b 28 28 28.6 208 ± 2 208 ± 2
Birth weight (g)c 875 ± 65A 821 ± 119A 894 ± 96A 632 ± 61B 608 ± 71B
Body weight at euthanasia (g)d 1152 ± 177 1147 ± 113 1107 ± 145 1071 ± 187 1035 ± 220
Brain weight at euthanasia (g)d 103 ± 4 98 ± 6 100 ± 11 93 ± 8 94 ± 9
Data are expressed as mean ± SD.
aCS, Cesarean section.
bReported as birth age for B and T groups and conceptual (gestational) age for the P groups. The normal gestational period for baboons is 180–182 days. Conceptual age for B and T at euthanasia was within 2 days of that for P+ and P-. Two animals in the T- group and one in the B group were 42 days old at euthanasia. Other than these, all animals were about 208 days post-conception at euthanasia.
cValues that do not share common capital letter superscripts are significantly different (p <0.05).
dNo significant differences among groups.
The concentrations of selected LCP in diets are presented in Table 2, and are available in detail in other publications [22,24]. Pregnant and, for the breastfed group, lactating, females consumed a conventional commercial primate diet for the entirety of the study. Normal primate meal contains fish meal as a source of protein, which also adds n-3 long chain polyunsaturated fatty acids (LCP). The breastmilk of these baboons, sampled at four weeks postpartum, contained about 0.68 ± 0.22% DHA, which falls at the higher end of the range of fish-eating human populations [25]. It also contained 20:5n-3 (0.34 ± 0.13%) and 22:5n-3 (0.51 ± 15%) as outlined in more detail elsewhere [22].
Table 2 Content of selected LCP in baboon breastmilk and in formulas, as % (w/w).
Breastmilk T-,P- T+ P+
DHA 0.68 ± 0.22 ND 0.30 0.61 ± 0.03
20:5n-3+22:5n-3 0.84 ± 0.16 ND 0.10 0.18 ± 0.10
ARA 0.62 ± 0.12 ND 0.55 1.21 ± 0.09
ND: not detected
The formulas with no LCP were identical for the term (T-) and preterm (P-) groups. The term group that was supplemented with LCP (T+) consumed formula fortified with 0.3% energy DHA and 0.6% energy ARA, as presented in detail previously [24]. For four of these animals, ARA was a component of phospholipid, while for three it was a component of triglyceride. Otherwise, diets were balanced and similar to commercial infant formula [24].
Preterm groups (P+ and P-) were delivered by Cesarian section (CS) at about 24 days preterm, with details also presented in Table 1. Preterm animals supplemented with LCP (P+) consumed formula fortified with 0.6% en DHA and 1.2% en ARA added as an encapsulated powder, and kindly provided by Mead-Johnson Nutritionals [22]. DHA and ARA in these diets were about twice those in the T+ diet, in an attempt to match the DHA content of breastmilk while holding the DHA/ARA ratio constant at about 2:1.
Neonatal tissue was collected at about 28 days adjusted age. This time point was chosen because previous data showed significant neurobehavioral effects within the first few weeks in term rhesus neonates [26]. The neonates were weighed, ketamine was injected intramuscularly, halothane general anesthesia was induced, and euthanasia was performed by exsanguination under continued halothane anesthesia. The brain and spinal cord were removed, wrapped in foil, frozen in liquid nitrogen and stored at -80°C until analysis. A transection between the brain and the spinal cord was made at the upper C1 level, that is, between the first cervical spine and the skull base. The brain weight listed in Table 1 is thus the sum of the brain, cerebellum and the upper part of the cervical spinal cord.
With the exception of cortical regions, CNS structures were dissected and analyzed in their entirety. For cortex, the top 3 mm of gray matter was carefully dissected from the underlying white matter.
Lipid extraction and analysis
Total lipids were extracted from tissue homogenates by methods reported previously [27]. Fatty acid methyl esters (FAME) were prepared using 14% BF3 in methanol (Sigma Chemical, St. Louis, Mo.). Butylated hydroxytoluene (BHT) was added to solvents as an antioxidant and a known quantity of freshly prepared heptadecanoic acid (99+%, Sigma Chemical, St Louis, MO) in chloroform was added as an internal standard to tissue samples just before extraction. FAME were dissolved in heptane with BHT and stored at -20°C until analysis.
FAME were analyzed using a Hewlett Packard 5890 series II gas chromatography with a BPX 70 column (60 m × 0.32 mm I.D. × 0.25 μm film; SGE, Austin, Tx) and H2 as carrier gas. Quantitative profiles were calculated using methyl-17:0 as an internal standard and an equal weight FAME mixture (68A; Nuchek Prep, Elysian, Mn) to derive response factors for each FAME. Chromatography conditions and calibration details have been reported previously [27]. The DHA and ARA concentrations are expressed as weight percentages of total fatty acids from 14 to 24 carbons.
Statistics
DHA data are presented as means ± SD. Weight percents of DHA in various CNS regions were tested for significant differences by Duncan's multiple range test with significance declared at p < 0.05. For treatment effects, the Fisher's least significant difference (Fisher's LSD) procedure was used. A one-way analysis of variance (Anova) was performed to test for equivalence of treatment means for either DHA or ARA in a specific CNS region. When significant at the p < 0.05 level, a t-test was performed on a pairwise basis. Statistics were calculated using functions provided in Excel 2000 for WinXP (Microsoft, Redmond, WA). Duncan's multiple range test was performed to compare the relative concentrations of DHA and ARA among CNS regions.
Results
CNS DHA concentrations
Figure 2 is a schematic presentation of DHA concentrations found in four week old baboon neonate CNS. The coloring reflects DHA concentrations running from white for highest to dark blue for lowest, as shown in the legend; regions colored gray were not analyzed. The globus pallidus has the highest DHA concentration at 15.8 ± 0.5% (w/w), while the optic nerve is lowest at 4.5 ± 0.4%. The legend presents the results of Duncan's multiple range test; regional DHA means that are not significantly different share a common line.
Figure 2 Baboon Neonate CNS DHA Map. Schematic diagram of four-week old baboon central nervous system with DHA concentrations color coded and ranked highest (white) to lowest (dark blue). Numbered lines in the parasagittal section refer to coronal sections; in this view, right side is shown with most of the right hemisphere removed for clarity. "Duncan" refers to Duncan's multiple range test; means sharing a line are not statistically different (p < 0.05). The colors each span 10% of the DHA concentration range.
Gray matter DHA was statistically different among cerebral cortex regions and averaged about 14%. This average is greater than we reported elsewhere for cortices of similar baboons [27]. In previous studies, cerebrum was sampled with an indeterminate contribution of white matter. In the present study, about 3 mm of gray matter was carefully removed from the underlying white corona radiation. Our results show that white matter is much poorer in DHA than gray matter. The precentralis gray matter is significantly richer in DHA (14.6 ± 0.9%) than the other cortex lobes, while the temporal has the least DHA (12.2 ± 0.7%).
The distribution of DHA in five coronal sections is presented in an inset of Figure 2. Half of the central regions of the brainstem and surrounding area are presented. Coronal section 1 shows the globus pallidus, putamen and caudate to have very high DHA concentrations (>14.3%), while the surrounding white matter of the coronal radiation is relatively low in DHA (5.5 ± 0.9%). The corona radiation was sampled in regions with only nerve fibers and devoid of neuronal soma, thus nuclei are represented in Figure 2 in gray since they were not sampled. Coronal sections 2–4 present views of the cingulate gyrus, thalamus nuclei, geniculate bodies, pulvinar, colliculi and hippocampus, all rich in DHA. Coronal section 5 shows the white matter of the corpus callosum and coronal radiation, both low in DHA.
Included in Figure 2 is a scale showing DHA values ranked from highest to lowest and statistical differences. Close examination shows that differences in DHA concentrations are relatively small in consecutively ranked regions, 0.6% or less, except for the discontinuity between the lateral geniculate (11.2 ± 1.6%) and corpus callosum (7.0 ± 1.3%). This dichotomy represents a demarcation between predominantly gray or white matter, the latter of which, as previously noted, is relatively poor in DHA.
CNS ARA concentrations
Figure 3 is a map of arachidonic acid concentrations in the same CNS regions as presented in figure 2, and was obtained from the same analyses. ARA concentrations range from a high of 13.7 ± 0.5% in the amygdala to a low of 6.8 ± 0.4% in the optic tract, to yield a range of about 2-fold in concentration compared to 3.5-fold for DHA. As with DHA, gray matter is generally richer in ARA than is white matter; however, in contrast to DHA, there is no discontinuity between gray and white matter ARA concentrations.
Figure 3 Baboon Neonate CNS ARA Map. Schematic diagram of four-week old baboon central nervous system with ARA concentrations color coded and ranked highest (white) to lowest (dark blue). See Figure 2 for key.
Influence of treatments on CNS DHA and ARA
The results of analyses of DHA and ARA concentrations as a function of treatments are presented in tables 3 and 4, respectively. For DHA, one way ANOVA indicated that 23 of the 26 regions showed treatment effects, while only 8 of the 26 regions showed significant ARA treatment effects.
Table 3 CNS DHA concentrations (mean ± SD, % w/w) for each treatment. One way Anova was significant (p < 0.05) for all structures except the spinal cord, internal capsule and medial geniculate. The right column shows the results of t tests. Gray matter falls into two classes: 1) LCP supplementation does not support DHA levels found for breastfed animals, detected for cerebral cortex and cerebellum; 2) LCP supplementation does support DHA levels, found in most other gray tissue (basal ganglia, limbic regions, thalamus and midbrain).
CNS Region Group Statistics
B T+ T- P+ P-
Gray,Class 1
Pre-Centralis 14.4 ± 0.9 12.6 ± 0.6 11.7 ± 1.0 13.0 ± 0.6 10.4 ± 0.7 [B>all; T+>P-; P+>T-,P-; T->P-]
Post-Centralis 14.1 ± 0.8 12.7 ± 0.8 11.7 ± 1.0 13.4 ± 0.6 10.4 ± 0.2 [B>T+,T-,P-; T+>T-,P-; P+>T-,P-; T->P-]
Occipital 13.9 ± 1.0 11.7 ± 0.9 10.7 ± 0.8 11.8 ± 0.1 8.7 ± 0.6 [B>all; T+>P+,P-; T->P+,P-; P+>P-]
Cingulate 13.4 ± 0.9 11.4 ± 0.7 10.7 ± 0.8 11.8 ± 0.8 9.5 ± 0.9 [B>all; T+>P-; P+>T-,P-; T->P-]
Frontal 12.7 ± 0.6 11.2 ± 0.8 10.6 ± 0.7 11.4 ± 0.4 8.4 ± 1.0 [B>all; T+>P+,P-; T->P-; P+>P-]
Temporal 12.2 ± 0.6 10.5 ± 1.1 10.3 ± 0.8 11.2 ± 0.5 9.0 ± 0.7 [B>all; T+,T-,P+>P-]
Cerebellum 12.6 ± 0.9 11.2 ± 0.7 10.6 ± 0.8 11.4 ± 0.2 7.7 ± 0.6 [B>all; T+>P-; T-,P+>P-]
Gray, Class 2
Globus Pallidus 15.8 ± 0.5 15.4 ± 0.6 12.7 ± 2.3 16.6 ± 0.3 13.4 ± 1.5 [B>T-,P-; T+>T-,P-; P+>T-,P-]
Putamen 14.7 ± 0.8 13.9 ± 0.9 13.5 ± 0.6 13.6 ± 1.3 11.6 ± 0.8 [B>T-,P-; T+,T-,P+>P-]
Caudate 14.3 ± 1.1 13.4 ± 0.8 13.0 ± 1.1 13.5 ± 0.8 11.3 ± 0.4 [B>T-,P-; T+,T-,P+>P-]
Superior Colliculus 15.2 ± 1.2 14.2 ± 1.1 13.6 ± 0.5 15.0 ± 0.4 13.4 ± 0.7 [B>T-,P-; P+>T-,P-]
Inferior Colliculus 14.4 ± 1.0 12.9 ± 1.4 13.4 ± 1.3 15.0 ± 0.6 13.4 ± 0.4 [B>T+; P+>T+,T-,P-]
Amygdala 12.0 ± 0.6 11.4 ± 0.6 10.8 ± 0.9 11.7 ± 0.5 9.4 ± 0.4 [B>T-,P-; T+,T-,P+>P-; P+>T-]
Hippocampus 13.0 ± 0.8 11.6 ± 0.5 10.6 ± 1.5 12.6 ± 0.5 9.0 ± 0.5 [B>T+,T-,P-; T+,T-,P+>P-; P+>T-]
Hypothalamus 12.0 ± 0.4 11.3 ± 0.6 11.0 ± 0.7 12.0 ± 0.8 10.6 ± 0.4 [B>T-,P-; T+>P-; P+>T-,P-]
Pulvinar 13.5 ± 0.6 12.7 ± 0.7 12.5 ± 1.0 14.6 ± 1.1 12.0 ± 0.4 [B>T-,P-; P+>all]
Thalamus DM 13.0 ± 1.0 12.4 ± 1.2 12.2 ± 0.9 13.5 ± 0.8 10.9 ± 1.3 [B>P-; T+,T-,P+>P-]
Thalamus VPL 12.9 ± 0.7 12.6 ± 0.9 12.2 ± 0.9 13.1 ± 1.7 10.3 ± 1.6 [B>P-; T+,T-,P+>P-]
Medial Geniculate 12.5 ± 1.3 12.8 ± 1.1 12.2 ± 0.9 12.6 ± 1.3 11.7 ± 0.8
Lateral Geniculate 11.2 ± 1.6 10.6 ± 0.9 9.5 ± 0.8 11.7 ± 0.7 8.8 ± 1.8 [B>T-,P-; T+>P-; P+>T-,P-]
White Matter
Corpus Callosum 7.0 ± 1.3 7.9 ± 0.7 6.4 ± 0.9 8.2 ± 0.7 6.3 ± 0.7 [T+>T-,P-; P+>B,T-,P-]
Internal Capsule 6.8 ± 1.0 7.0 ± 1.5 6.4 ± 0.7 7.9 ± 0.9 6.3 ± 1.4
Spinal Cord1 5.7 ± 0.3 6.0 ± 1.0 5.6 ± 0.5 6.2 ± 0.5 5.1 ± 0.6
Corona Radiation 5.5 ± 0.9 5.2 ± 0.9 4.4 ± 0.4 5.7 ± 0.8 4.7 ± 0.7 [B>T-; P+>T-,P-]
Optic Tract 4.8 ± 0.5 4.5 ± 0.4 4.3 ± 0.6 5.1 ± 0.3 3.9 ± 0.4 [B,T+>P-; P+>T+,T-,P-]
Optic Nerve 4.5 ± 0.5 4.0 ± 0.9 3.9 ± 0.4 4.8 ± 0.9 3.6 ± 0.3 [B>T-,P-; T+>P-; P+>all]
1Spinal cord is a mixture of white and gray matter.
Table 4 CNS ARA concentrations (mean ± SD, % w/w) for eight structures for which a one-way analysis of variance indicated at least one unequal mean (p < 0.05). Anova was not significant for the other 18 structures.
CNS Region Group Statistics
B T+ T- P+ P-
Occipital 13.2 ± 0.8 13.3 ± 0.7 13.6 ± 0.8 14.0 ± 0.5 14.3 ± 0.6 [P->B,T+]
Hypothalamus 11.7 ± 0.7 12.2 ± 0.6 11.4 ± 0.4 12.6 ± 0.7 12.9 ± 0.5 [T+,P+,P->T-; P+,P->B]
Superior Colliculus 10.1 ± 0.9 11.0 ± 0.8 10.6 ± 0.7 11.0 ± 0.6 11.7 ± 0.6 [T+,P+,P->B; P->T-]
Thalamus DM 10.0 ± 0.5 10.3 ± 0.6 9.6 ± 0.8 10.8 ± 0.6 10.7 ± 0.7 [T+,P+,P->T-; P+>B]
Corpus Collosum 9.9 ± 0.6 11.2 ± 0.5 10.2 ± 0.7 11.3 ± 0.5 11.4 ± 0.9 [T+,P+,P->B; T+,P+,P->T-]
Lateral Geniculate 9.5 ± 1.1 10.6 ± 0.6 9.9 ± 0.4 10.7 ± 0.5 10.5 ± 0.4 [T+,P+,P->B; P+>T-]
Inferior Colliculus 9.3 ± 0.5 10.5 ± 1.0 9.3 ± 0.7 10.3 ± 0.5 10.3 ± 0.4 [T+,P+,P->B,T-]
Optic Tract 6.8 ± 0.3 6.9 ± 0.4 7.2 ± 0.7 7.4 ± 0.3 7.7 ± 0.4 [P-,P+>B; P->T+]
We consider the randomized, breastfed control group (B) to be a gold standard against which the others are compared. With the exception of one region, the B group DHA concentration was always among the greatest, while the P- group was always among the lowest. More careful examination of the DHA statistical analysis reveals that the responses of DHA concentrations among gray matter regions fall into two distinct classes, as arranged in Table 3. For all lobes of the cerebral cortex, and for the cerebellum, supplementation with 0.3% (T+) or 0.6% (P+) DHA did not support tissue DHA concentrations similar to those of the breastfed group. This is indicated by the observation that the B group had significantly greater DHA concentration than the other groups, denoted in the table by "B>all". All other areas of gray matter investigated, including the basal ganglia, limbic regions, thalamus and midbrain, fall in a class 2, in which supplemental dietary DHA did restore DHA concentrations to breastfed levels. This is denoted in the table by "B>T-,P-", indicating that the B group DHA is not significantly different from the T+ and P+ group DHA. All the regions listed under class 2 have this pattern, and for two regions, inferior colliculus and hippocampus, the B group DHA is significantly greater than a supplemented group (T+). Another dominant trend among gray regions is significantly greater DHA concentrations in supplemented groups (T+, P+) compared to unsupplemented groups (T-, P-). We can conclude from these data that in gray matter, DHA supplementation increases DHA concentrations in most regions; with the critically important exceptions of the cerebral cortex and cerebellum, supplemental dietary DHA restores CNS DHA concentrations to breastfed levels.
Trends in white matter are not as consistent as the two classes of gray matter. Two of the six white matter regions investigated did not respond to treatments (internal capsule and spinal cord [includes white matter]). The B group DHA was significantly greater than the others in three regions but, uniquely, less than a supplemented group (P+) in the corpus callosum.
The trends for ARA are very different from those for DHA. As noted, treatments were ineffective at altering ARA concentrations in 18 regions. Table 4 shows the eight CNS regions that were significantly influenced by treatments. Remarkably, the B group is of lower ARA concentration than the P- group in every case, and this is significant in 7 of 8 regions. In every region, the P- group is among the set of highest ARA concentration, and B is among the lowest.
For both DHA and ARA, the effects of prematurity per se are not strong. For DHA, the T+ group is significantly greater than the P+ group in some cases, and similarly for the T-/P- comparison. Because the groups were adjusted age-matched, they were on formula for different periods: the term groups were on formula for 4 weeks, while the preterm groups were on formula for 7.5 weeks. For this reason alone it is reasonable to expect that the P- group would have lower DHA than the T- group. It should be borne in mind that the supplementation for P+ was twice that for T+, which partially compensates for this difference in exposure to formula DHA. However, in every case, the P+ DHA concentration exceeds that of the T- DHA concentration, showing that the mild prematurity studied here does not severely impair DHA accretion compared to an unsupplemented term group. In other words, DHA concentrations are greater in supplemented preterms than they are in unsupplemented terms, indicating that prematurity itself does not impair DHA accretion.
Table 5 presents results for four other fatty acids of interest in three selected CNS regions (occipital cortex, hippocampus, corpus callosum). Neither 18:0 (stearic acid) nor 22:4n-6 (adrenic acid) showed any statistically significant variation as a function of treatment in these regions; thus, we report means and standard deviations pooled for all treatments. Oleic acid (18:1n-9) was significantly different in two regions but the difference in means was not large, thus we also report pooled means. Importantly, 22:5n-6 (Osbond acid) was significantly elevated in the CNS regions of the groups that did not have dietary DHA, P- and T-. In 25 of 26 CNS regions (except thalamus DM), P- was significantly elevated, while T- was significantly elevated in 22 of 26 regions. Osbond acid generally rises when DHA falls and it is considered a marker of DHA insufficiency [22]. This fatty acid has been previously reported to increase dramatically in n-3 fatty acid deficiency [28] and to fall when deficient animals are supplied with dietary DHA [29]. These data extend our previous report [22] that 22:5n-6 levels rise throughout the CNS when DHA is supplied in the diet with 18:3n-3 levels that are considered to supply sufficient n-3 fatty acids for growth and development.
Table 5 Concentrations (mean ± SD) of selected fatty acids in three representative CNS regions. 18:0 (stearic acid) and 22:4n-6 (adrenic acid) were not significantly altered by treatments in these CNS regions or in most others. While 18:1n-9 (oleic acid) was significantly different in occipital cortex and hippocampus but not corpus callosum, the magnitude of the differences was small and overall means are reported. 22:5n-6 (Osbond acid) was significantly elevated in all regions (except thalamus DM), for the P- and T- groups, similar to the patterns shown for the regions in this table.
Occipital Corpus Callosum Hippocampus
18:0a 19.5 ± 0.23 18.4 ± 0.72 19.3 ± 0.40
18:1n-9 10.6 ± 0.30b1 15.9 ± 0.80 11.5 ± 0.30b2
22:4n-6c 5.60 ± 0.25 7.060 ± 0.29 5.89 ± 0.19
22:5n-6 B T+ T- P+ P-
Occipital 1.97 ± 0.23 2.26 ± 0.32 2.47 ± 0.40 1.85 ± 0.23 2.39 ± 0.16 T-,P->B; T+,T-,P->P+
Corpus Callosum 1.16 ± 0.22 1.79 ± 0.26 1.43 ± 0.27 1.41 ± 0.22 1.94 ± 0.33 T+,P->B; T->T+; T+>P+; P->T-,P+
Hippocampus 1.86 ± 0.30 2.27 ± 0.40 2.28 ± 0.41 1.88 ± 0.16 2.33 ± 0.14 T+,T-,P->B; T+,T-,P->P+
a,cNo significant differences among groups (B, T+, T-, P+, P-).
b1Significant group differences: B<T-,P+
b2Significant group differences: B<T-,P+, P-
Discussion
Our fatty acid data were obtained from macroscopically dissectible tissue and therefore contain a mix of cell types and constituents that differ from region to region. Bourre reported the DHA contents of cellular brain fractions of soya oil fed 15-day-old rats [30] as neurons (8.2%), astrocytes (10.6%), synaptosomes (8.5%), oligodendrocyte (5.1%) and myelin (5.8%). These data are qualitatively in accord with our DHA CNS results, considering that gray matter is composed largely of neurons and glia while white matter is composed principally of myelin and oligodendrocytes. However, literature data clearly suggest that DHA concentrations are not explained by a simple mix of cells and cell types. For instance, adult human dorsal-medial (DM) thalamus total cell density and glia/neuron indices are higher than the globus pallidus, though DHA concentrations are lower [31]. Similarly, neuron density (neurons/gram) in humans is in the order occipital lobe > precentralis > frontal lobe > hippocampus, a gradation that also is not consistent with our DHA data [32,33]. Data reported for neuron density of human occipital lobe > temporal lobe > pre-centralis > thalamus were also poorly correlated with our results [34]. Finally, mean neuron volume in humans in the order hippocampus > frontal > precentralis > occipital also failed to show correlation with tissue DHA [32].
A notable trend in the CNS regions richest in DHA is that they are all involved in motor function. The basal ganglia are all integrative centers for motor control and the globus pallidus, caudate and putamen are among the top six in DHA concentration. The cortical region with the highest DHA concentration is the precentralis, which has long been known as the motor cortex. The superior colliculus is a major relay point for saccades, the rapid adjustment of eye position. Of the six regions with the highest DHA, the only region that is not known as a motor region is the inferior colliculus, which plays a role in decoding spatial information in auditory signal processing. Even here a role in motor control system is possible and, notably, no clear function has emerged for the dorsal part of this structure [35].
Recent studies in humans and animals suggest that even a mild DHA deficit may impair motor function. Infant rhesus monkeys on high DHA formula had improved neuromotor scores compared to those on DHA-free diets in the first month of life [26]. Human infants on formula that induced low plasma DHA for the first four months of life scored lower on one year evaluations of motor development than those with better DHA status [36]. Bailey motor development index scores for preterm infants on DHA-containing formula were improved compared to those on formulas with no DHA [37].
Our treatments are a strong test of the hypothesis that LCP supplementation increases CNS LCP concentrations for DHA and ARA, the two LCP that have garnered the most attention for the developing human infant CNS. This is, in part, because the randomized breastfed group is a true gold standard for comparison, and because of the availability of CNS tissue for LCP analysis, neither of which are available in human studies. DHA supplementation at a level matched to that in the breastmilk of these control baboons maintained DHA levels in many but not all CNS regions. The cerebral cortex and cerebellum, both physically large regions requiring proportionately more DHA, were unable to maintain levels similar to controls. We have previously speculated [22] that this may be due to the presence of significant amounts of 20:5n-3 and 22:5n-3 in breastmilk (0.85%), which are thought to be much more efficient precursors of DHA than 18:3n-3 [38], quantitatively the most prominent omega-3 PUFA in breastmilk and formula. Another important consideration is that these regions have developmentally later peak growth profiles than basal and limbic regions, which develop early in gestation [39]. We can speculate that the massive requirement for DHA imposed by aggressive growth during this period renders these regions more vulnerable to DHA insufficiency, and that biosynthesis cannot keep up with demand. We thus accept the hypothesis that supplemental DHA improves CNS tissue status, with the proviso that current levels may be insufficient to maintain levels in cortex. Notably, current formula DHA levels in the USA are 0.15%–0.35% (w/w), which is about 1/4 to 1/2 that used in the P+ group, though the T+ group was similar to commercial formulas with a DHA concentration of about 0.3%. Finally, we note that previous studies of neonatal rhesus monkeys have shown an increase in cortex phosphatidylethanolamine DHA from about 15% at birth to 22% and 34% at 22 months in frontal and occipital cortex, respectively [28]. CNS demands for DHA clearly continue well beyond the perinatal period as the CNS matures.
The influence of our treatments on CNS ARA concentrations is radically different from that for DHA. Most of the CNS regions we investigated are nearly immune to supplementation at levels twice that in current infant formulas (1.2%, w/w, for the P+ group). In fact, the significant trends were, in some respects, opposite to what would have been expected from the simple model that more dietary ARA should result in more CNS ARA. In particular, the breastfed group tended to have the lowest ARA, while the group longest exposed to ARA-free formula, the P- group, tended to have the highest ARA concentrations. Thus, the hypothesis that ARA-supplemented formula always increases CNS ARA is rejected. CNS ARA concentrations are clearly under tighter control than DHA concentrations, a fact that may derive from ARA's role as precursor of potently bioactive eicosanoids. Very recent evidence suggests that DHA is also a precursor of oxygenated derivatives, docosanoids; however the bioactivity of these compounds and their biochemistry are not fully described [13].
Conclusion
In summary, these data reveal that DHA and ARA concentrations in the CNS are highly region-specific and are unexpectedly high in the deep CNS regions embedded in white matter of much lower DHA and ARA concentration. Supplementation improves DHA concentrations and in all but the cerebral cortex, maintains DHA at levels similar to that of breastfed controls. Prematurity per se does not severely impair accretion of DHA. ARA is much less sensitive to dietary manipulation than DHA, and the response of CNS regions to ARA supplementation is complex. Future research targeting functions controlled and mediated by CNS structures other than the cerebrum and visual system that are likely to be sensitive to DHA nutrition is indicated.
Abbreviations
ARA, arachidonic acid; DHA, docosahexaenoic acid; FAME, fatty acid methyl ester; LCP, long chain polyunsaturated fatty acids; PUFA, polyunsaturated fatty acids.
Competing interests
JTB is a consultant for Mead-Johnson Nutritionals. All other authors declare that they have no competing interests.
Authors' contributions
Participation was as follows: Project conception (GYD, PWN, JTB); Study design (GYD, VW, PWN, JTB); Live animal management (GYD, ATH, ESN, VW); Dissection and laboratory analysis (GYD); Interpretation of results (GYD, JTB), manuscript preparation (GYD, ATH, JTB). All authors approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We gratefully acknowledge the input of A. J. Armstrong, M.-C. Huang, X.Y. Ding, C. Li, A. C. Chao and A. Turpeinen with animals, and C. Garza for helpful discussions in the early stages of this work. Supported by NIH grant EY10208, GM49209, and GM71534; ATH acknowledges support from NIH training grant DK07158. NIH grant HD21350 supported infrastructure for the baboon colony. Mead-Johnson Nutritionals (Evansville, IN) and Numico (Friedrichsdorf, Germany) kindly provided diets and diet components.
==== Refs
Benolken RM Anderson RE Wheeler TG Membrane fatty acids associated with the electrical response in visual excitation Science 1973 182 1253 1254 4752217
Breckenreidge WC Gombos G Morgan IG The docosahexaenoic acid of the phospholipids of synaptic membranes, vesicles and mitochondria Brain Res 1971 33 581 583 5134944 10.1016/0006-8993(71)90143-0
Svennerholm L Distribution and fatty acid composition of phosphoglycerides in normal human brain J Lipid Res 1968 9 570 579 4302302
Neuringer M Connor WE Van Petten C Barstad L Dietary omega-3 fatty acid deficiency and visual loss in infant rhesus monkeys J Clin Invest 1984 73 272 276 6317716
Neuringer M Connor WE n-3 fatty acids in the brain and retina: evidence for their essentiality Nutr Rev 1986 44 285 294 3537864
Weisinger HS Vingrys AJ Bui BV Sinclair AJ Effects of dietary n-3 fatty acid deficiency and repletion in the guinea pig retina Invest Ophthalmol Vis Sci 1999 40 327 338 9950590
Neuringer M Reisbick S Janowsky J The role of n-3 fatty acids in visual and cognitive development: current evidence and methods of assessment J Pediatr 1994 125 S39 47 7965452
Farkas T Kitajka K Fodor E Csengeri I Lahdes E Yeo YK Krasznai Z Halver JE Docosahexaenoic acid-containing phospholipid molecular species in brains of vertebrates Proc Natl Acad Sci U S A 2000 97 6362 6366 10823917 10.1073/pnas.120157297
Mitchell DC Niu SL Litman BJ Optimization of receptor-G protein coupling by bilayer lipid composition I: kinetics of rhodopsin-transducin binding J Biol Chem 2001 276 42801 42806 11544258
Niu SL Mitchell DC Lim SY Wen ZM Kim HY Salem N JrLitman BJ Reduced G protein-coupled signaling efficiency in retinal rod outer segments in response to n-3 fatty acid deficiency J Biol Chem 2004 279 31098 31104 15145938 10.1074/jbc.M404376200
Eldho NV Feller SE Tristram-Nagle S Polozov IV Gawrisch K Polyunsaturated docosahexaenoic vs docosapentaenoic acid-differences inlipid matrix properties from the loss of one double bond J Am Chem Soc 2003 125 6409 6421 12785780 10.1021/ja029029o
Leaf A Kang JX Xiao YF Billman GE Voskuyl RA Functional and electrophysiologic effects of polyunsaturated fatty acids on exictable tissues: heart and brain Prostaglandins Leukot Essent Fatty Acids 1999 60 307 312 10471113 10.1016/S0952-3278(99)80004-0
Mukherjee PK Marcheselli VL Serhan CN Bazan NG Neuroprotectin D1: a docosahexaenoic acid-derived docosatriene protects human retinal pigment epithelial cells from oxidative stress Proc Natl Acad Sci U S A 2004 101 8491 8496 15152078 10.1073/pnas.0402531101
Ring HA Serra-Mestres J Neuropsychiatry of the basal ganglia J Neurol Neurosurg Psychiatry 2002 72 12 21 11784818 10.1136/jnnp.72.1.12
Naarding P Kremer HP Zitman FG Huntington's disease: a review of the literature on prevalence and treatment of neuropsychiatric phenomena Eur Psychiatry 2001 16 439 445 11777733 10.1016/S0924-9338(01)00604-6
Lozano AM Lang AE Pallidotomy for Parkinson's disease Adv Neurol 2001 86 413 420 11554004
Moore SA Cerebral endothelium and astrocytes cooperate in supplying docosahexaenoic acid to neurons Adv Exp Med Biol 1993 331 229 233 8333338
de la Presa Owens S Innis SM Docosahexaenoic and arachidonic acid prevent a decrease in dopaminergic and serotoninergic neurotransmitters in frontal cortex caused by a linoleic and alpha-linolenic acid deficient diet in formula-fed piglets J Nutr 1999 129 2088 2093 10539789
Gibson RA Chen W Makrides M Randomized trials with polyunsaturated fatty acid interventions in preterm and term infants: functional and clinical outcomes Lipids 2001 36 873 883 11724459
Carlson SE Werkman SH Peeples JM Cooke RJ Tolley EA Arachidonic acid status correlates with first year growth in preterm infants Proc Natl Acad Sci USA 1993 90 1073 1077 8430076
Sarkadi-Nagy E Wijendran V Diau GY Chao AC Hsieh AT Turpeinen A Lawrence P Nathanielsz PW Brenna JT Formula feeding potentiates docosahexaenoic and arachidonic acid biosynthesis in term and preterm baboon neonates J Lipid Res 2004 45 71 80 14523049 10.1194/jlr.M300106-JLR200
Sarkadi-Nagy E Wijendran V Diau GY Chao AC Hsieh AT Turpeinen A Nathanielsz PW Brenna JT The influence of prematurity and long chain polyunsaturate supplementation in 4-week adjusted age baboon neonate brain and related tissues Pediatr Res 2003 54 244 252 12736388 10.1203/01.PDR.0000072795.38990.F2
Diau GY Loew ER Wijendran V Sarkadi-Nagy E Nathanielsz PW Brenna JT Docosahexaenoic and arachidonic acid influence on preterm baboon retinal composition and function Invest Ophthalmol Vis Sci 2003 44 4559 4566 14507905 10.1167/iovs.03-0478
Wijendran V Huang MC Diau GY Boehm G Nathanielsz PW Brenna JT Efficacy of dietary arachidonic acid provided as triglyceride or phospholipid as substrates for brain arachidonic acid accretion in baboon neonates Pediatr Res 2002 51 265 272 11861929
Lauritzen L Hansen HS Jorgensen MH Michaelsen KF The essentiality of long chain n-3 fatty acids in relation to development and function of the brain and retina Prog Lipid Res 2001 40 1 94 11137568 10.1016/S0163-7827(00)00017-5
Champoux M Hibbeln JR Shannon C Majchrzak S Suomi SJ Salem N JrHigley JD Fatty acid formula supplementation and neuromotor development in rhesus monkey neonates Pediatr Res 2002 51 273 281 11861930
Su HM Bernardo L Mirmiran M Ma XH Corso TN Nathanielsz PW Brenna JT Bioequivalence of dietary alpha-linolenic and docosahexaenoic acids as sources of docosahexaenoate accretion in brain and associated organs of neonatal baboons Pediatr Res 1999 45 87 93 9890614
Neuringer M Connor WE Lin DS Barstad L Luck S Biochemical and functional effects of prenatal and postnatal omega 3 fatty acid deficiency on retina and brain in rhesus monkeys Proc Natl Acad Sci U S A 1986 83 4021 4025 3459166
Connor WE Neuringer M Lin DS Dietary effects on brain fatty acid composition: the reversibility of n-3 fatty acid deficiency and turnover of docosahexaenoic acid in the brain, erythrocytes, and plasma of rhesus monkeys J Lipid Res 1990 31 237 247 2139096
Bourre JM Pascal G Durand G Masson M Dumont O Piciotti M Alterations in the fatty acid composition of rat brain cells (neurons, astrocytes, and oligodendrocytes) and of subcellular fractions (myelin and synaptosomes) induced by a diet devoid of n-3 fatty acids J Neurochem 1984 43 342 348 6736955
Pakkenberg B Gundersen HJ Total number of neurons and glial cells in human brain nuclei estimated by the disector and the fractionator J Microsc 1988 150 1 20 3043005
Tower DB Structural and Functional Organization of Mammalian Cerebral Cortex:The Correlation of Neurone Density and Brain Size J Comp Neurol 1957 103 19 51
Sholl DA A Comparative study of the neuronal packing density in the cerebral cortex J Anat 1965 93 143 158 13641114
Pakkenberg B Gundersen HJ Neocortical neuron number in humans: effect of sex and age J Comp Neurol 1997 384 312 320 9215725 10.1002/(SICI)1096-9861(19970728)384:2<312::AID-CNE10>3.0.CO;2-K
Hudspeth AJ Kandel ER, Schwartz JH, Jessell TM Hearing Principles of Neural Science 2000 Fourth New York: McGraw-Hill 590 613
Voigt RG Jensen CL Fraley JK Rozelle JC Brown FR Heird WC Relationship between omega3 long-chain polyunsaturated fatty acid status during early infancy and neurodevelopmental status at 1 year of age J Hum Nutr Diet 2002 15 111 120 11972740 10.1046/j.1365-277X.2002.00341.x
O'Connor DL Hall R Adamkin D Auestad N Castillo M Connor WE Connor SL Fitzgerald K Groh-Wargo S Hartmann EE Jacobs J Janowsky J Lucas A Margeson D Mena P Neuringer M Nesin M Singer L Stephenson T Szabo J Zemon V Growth and development in preterm infants fed long-chain polyunsaturated fatty acids: a prospective, randomized controlled trial Pediatrics 2001 108 359 371 11483801 10.1542/peds.108.2.359
Pawlosky RJ Hibbeln JR Novotny JA Salem N Jr Physiological compartmental analysis of alpha-linolenic acid metabolism in adult humans J Lipid Res 2001 42 1257 1265 11483627
Clancy B Darlington RB Finlay BL The course of human events: predicting the timing of primate neural development Dev Sci 2000 3 57 66 10.1111/1467-7687.00100
|
15975147
|
PMC1184078
|
CC BY
|
2021-01-04 16:27:56
|
no
|
BMC Med. 2005 Jun 23; 3:11
|
utf-8
|
BMC Med
| 2,005 |
10.1186/1741-7015-3-11
|
oa_comm
|
==== Front
BMC Med EducBMC Medical Education1472-6920BioMed Central London 1472-6920-5-231598752610.1186/1472-6920-5-23Research ArticleElectronic learning can facilitate student performance in undergraduate surgical education: a prospective observational study Healy David Gerard [email protected] Fergal J [email protected] David [email protected] Patrick [email protected] Alfred Edward [email protected] Thomas [email protected] Enda W [email protected] John M [email protected]'Higgins Niall J [email protected] Arnold DK [email protected] Department of Surgery, Mater Misericordiae University Hospital, Eccles St, Dublin 7, Ireland2 Department of Surgery, St Vincent's University Hospital, Elm Park, Dublin 4, Ireland3 Conway Institute of Biomolecular & Biomedical Research, University College Dublin, Dublin 4, Ireland4 Centre for Health Informatics, University College Dublin, Dublin 4, Ireland2005 29 6 2005 5 23 23 12 3 2005 29 6 2005 Copyright © 2005 Healy et al; licensee BioMed Central Ltd.2005Healy et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Our institution recently introduced a novel internet accessible computer aided learning (iCAL) programme to complement existing surgical undergraduate teaching methods. On graduation of the first full cycle of undergraduate students to whom this resource was available we assessed the utility of this new teaching facility.
Method
The computer programme prospectively records usage of the system on an individual user basis. We evaluated the utilisation of the web-based programme and its impact on class ranking changes from an entry-test evaluation to an exit examination in surgery.
Results
74.4% of students were able to access iCAL from off-campus internet access. The majority of iCAL usage (64.6%) took place during working hours (08:00–18:00) with little usage on the weekend (21.1%). Working hours usage was positively associated with improvement in class rank (P = 0.025, n = 148) but out-of hours usage was not (P = 0.306). Usage during weekdays was associated with improved rank (P = 0.04), whereas weekend usage was not (P = 0.504). There were no significant differences in usage between genders (P = 0.3). Usage of the iCAL system was positively correlated with improvement in class rank from the entry to the exit examination (P = 0.046). Students with lower ranks on entry examination, were found to use the computer system more frequently (P = 0.01).
Conclusion
Electronic learning complements traditional teaching methods in undergraduate surgical teaching. Its is more frequently used by students achieving lower class ranking with traditional teaching methods, and this usage is associated with improvements in class ranking.
==== Body
Background
Medicine has become increasingly complex and the challenges faced by the medical education system are becoming even greater. Higher levels of technical and scientific knowledge are required as well as effective communication and management skills. Acquisition of this knowledge must be achieved within a finite time period. In addition the clinical opportunities for medical students are diminishing with decreasing length of hospital stay for many surgical procedures. This environment has led to the search for novel teaching methods to deliver undergraduate medical education.
Computer aided learning (CAL) offers distinct advantages over conventional teaching methods, including the potential for multimedia applications with a significant interactive content. Not only may text be presented, but in addition, tables, images, video and animation can be integrated into dynamic packages. In this manner multimedia education strategies offer potential strategic advantages over traditional paper based material. The system is very flexible allowing a lecturer to update and upload data in the package and to monitor an individual student's activity. An internet-based computer aided learning (iCAL) programme has the additional advantage of access from an off-site location at any hour of the day. Unfortunately the establishment of the infrastructure for iCAL is quite costly and the cost to benefit analysis of undergraduate iCAL has been questioned[1] The Department of Surgery at University College Dublin, Ireland, has recently introduced an iCAL package for undergraduate education. We wished to evaluate the effectiveness of this programme following the first complete cycle of undergraduate medical students.
Methods
Medical undergraduates at University College Dublin (UCD) have a forty two week training period in surgery. This is divided in two parts, a 28-week period and a 14-week period. Students are allocated clinical attachments to hospital surgical services and rotated though the spectrum of surgical specialities. During these rotations they participate in ward, theatre and outpatient activities but have no clinical responsibilities. Clinical "by the bedside" teaching with clinical lectures in small groups is provided. In addition a formal lecture programme is delivered and library facilities are provided for independent learning. This structure is the same in the initial 28 week and the later 14 week periods. The clinical sites are identical, and clinical rotations are organised so that all students get an exposure to all surgical specialties over the course of their training over the entire 42 week period. Whether an individual student goes through a particular rotation in the initial 28 weeks or the later 14 weeks, is a random allocation.
At the end of the first period of the programme, undergraduates have a surgical examination consisting of a multiple choice paper and a clinical examination. For the purposes of this paper this is referred to as the entry examination. At the end of the second period of the course students take the final exit examination in surgery. The exit exam also consists of a multiple choice exam and clinical exam with an additional written short answer examination.
iCAL-SURG was used during the second period of the surgery course, after students had taken their entry examination. The only fundamental methodology difference between these two training periods was the introduction of the iCAL programme. The questions presented through the iCAL system are identical in style to those presented in the short written examination and clinical examination problem solving scenarios.
The class was divided into quartiles on the basis of this entry examination. Utilisation of the iCAL programme by students was optional and was compared with absolute performance and change in class ranking in the final examination. One hundred and forty eight students completed the two year programme. Data on these students' utilisation of the iCAL platform was analysed.
Technical infrastructure
All students in UCD Medical Faculty are conversant with use of ICT, with CAL used to deliver up to 35% of preclinical courses. Clinical students have access to a suite of networked high specification multimedia PC's with T1 internet access. This facility is available to students from 8.00 to 18.00, weekdays. iCAL-SURG is also available anywhere with an internet connection. Online course material is delivered within the Blackboard Virtual Learning Environment. Courses are password protected and use of iCAL-SURG by individual students is monitored. Instructors can find the number of times each student has logged on to iCAL-SURG, the areas of the course accessed, the time and day of log-on. The number of times the site was accessed by the student is referred to as the "hit rate". It is not possible to record the content viewed or duration of student activity, once logged on.
iCAL-SURG
During this pilot phase, iCAL-SURG use was optional and did not form part of the assessment process for students. This phase concentrated on the COURSE DOCUMENTS and the ASSIGNMENTS areas in Blackboard. The COMMUNICATION area was used for both social and academic purposes but only accounted for 5.3% of use of iCAL-SURG and no assessment was made of this in relation to student performance.
In the COURSE DOCUMENTS area [Fig, 1,2] lecture notes are available prior to lectures, either as Powerpoint, Adobe Portable Document Format (PDF) or Microsoft Word files. The material available on the iCAL system not only included material presented at lectures but also additional supplementary material. Old examination question papers and sample answers are also available here. In the ASSIGNMENTS area each week the academic staff uploaded a short series of clinical questions with suggested answers made available made available [Fig. 3]. Students read these and studied the relevant areas in text books or in discussion with tutors and peers. Students could also take online OSCE examinations [Fig. 4,5]. This was presented as a series of photographs with related clinical details. Students were asked to answer the questions in a prepared downloaded Microsoft Word document, which they then uploaded to the "Digital Dropbox" in Blackboard. These submissions were reviewed by academic staff. On the following day sample answers were published [Fig. 4,5] which students could compare with their submissions.
Figure 1 Screenshot from iCAL-SURG.
Figure 2 Screenshot taken from some materials available in course materials.
Figure 3 This is an example of a short OSCE available on iCAL-SURG.
Figure 4 Screenshot of link to OSCE with instructions.
Figure 5 Screenshot of first station in a full online OSCE.
Statistical analysis
A Student t test was used for comparison between groups of two. Where more than two groups were involved an ANOVA was performed followed by Post-hoc analysis as appropriate, with a Fischer's least significant difference test. Data involving class rank was analysed non-parametrically with a Mann Whitney U Test for comparison between groups of two, a Kruskal Wallis where more groups were involved and a Spearman Rank Correlation for investigating correlations. Alpha was set at 0.05.
Results
Complete data on computer utilisation was available on 148 students. There were 58 males and 90 females. There were no significant gender differences in the utilisation of online access (P = 0.3) or final year examination results (P = 0.251). One of the criticisms of the programme from the start was the concern that students without independent computer and internet access would be disadvantaged. 74.4% of students were able to access iCAL from off-campus internet access. Despite this high availability of independent internet access, the majority of iCAL usage (64.6%) took place during daytime hours (08:00–18:00) [Fig. 6] with less then expected usage on the weekend (21.1%) [Fig. 7].
Figure 6 This shows the pattern of computer usage over the day. Data is presented as means with standard errors.
Figure 7 This shows the number of hits into the computer system each day. Use of the system on a Friday, Saturday or Sunday was significantly less than use on Monday, Tuesday and Thursday (ANOVA P < 0.001). * P < 0.05
To assess the educational benefits of the programme the class was ranked based on the entry test results. Class rank on the entry test was then compared to the class rank at the exit exam. Usage of the iCAL system during daytime hours was positively associated with improvement in class rank (P = 0.025) but out-of hours usage was not (P = 0.306). Usage during weekdays was associated with improved rank (P = 0.04), whereas weekend usage was not (P = 0.504). There was no significant difference in improvement in class rank between students with independent internet access and those dependant on access provided by the University Hospitals. We therefore felt that the computer facilities provided by the University were adequate for the needs of the students and that students without independent computer and internet access were not disadvantaged.
We were also interested in the level of usage of iCAL by different student abilities. Students with lower ranking on the entry test were found to subsequently use the computer system more frequently (P = 0.01). The lower quartiles of the class were the most frequent users of the programme (ANOVA, P = 0.042) [Fig. 8]. We then wanted to evaluate whether this usage translated into improvements in class rank. Usage of the iCAL system was positively correlated with improvement in class rank from the entry to the exit examination (P = 0.046).
Figure 8 Students have been divided into quartiles based on the performance in the entry surgical examination. A] This shows the usage of the surgical computer education package by students in different class rank quartiles. Significantly more use was made of the computer package by students in the lowest entry rank quartile (ANOVA P = 0.042). B] This shows the change in class ranking between entry examination and final examination in surgery. Initially low ranking students showed the greatest improvement in their class position (Kruskal Wallis P = 0.01). Mean and standard error. * P < 0.05, ** P < 0.01; compared to lowest class quartile at entry test.
Discussion
This cohort was the first group to complete their undergraduate surgical training since the introduction of this dynamic iCAL package. Of concern to us in establishing this program was the potential to competitively disadvantage students without independent computer and internet access. This in part motivated the decision not to include iCAL as part of the formal student assessment. However our study shows no such disadvantage. Instead, structured usage of iCAL during normal working time was associated with improved class performance rather then usage out of hours. This suggests that improvement in class rank with iCAL was not facilitated by studying late into the night but rather by a structured and disciplined approach. This conclusion is limited by the fact that the system cannot distinguish a student accessing the system from home during these hours or from the University computer suite. However as the students had other onsite daily activities during the day, with monitored attendance, it seems most likely that this activity occurred from University provided facilities.
One of the objectives of this study was to examine the patterns of usage of the iCAL system. Part of the motivation in using an internet accessible format was the expectation that students could utilise the programme as a distance learning tool. Although 35% of iCAL utilisation took place outside of the normal academic day, this usage was not associated with improvements in relative position in the class. This at first glance questions the benefit of an internet based infrastructure. However this technology has other advantages to a faculty using more then one teaching hospital, such as enabling the educators to upload data from multiple sites.
We were also interested in the utilisation of the programme by students of different class rank. It was interesting to us that the top two quartiles of the class, as ranked by the entry examination, did not utilise the iCAL programme as frequently as the lower two quartiles. The factors underlying this are uncertain. The method of assessing usage was the number of times an individual student logs into the system. The package does not measure the amount of time spent with the package or the volume of content covered by the student. It is conceivable that students at the lower end of the class took longer to completely digest the content of the web pages and required more frequent visits to the same sites. However if that is the case it appears to have been rewarding as the lowest 25% of the class on the entry test, who also used the iCAL package most, were the group who made the most progress up the class ranking. It must be observed that those at the top have little room to move up and those at the bottom little distance to fall in terms of class ranking. This could have influenced an outcome measure based on class rank. However the pattern was not only seen with the top and bottom quartile but is also seen in the central 25–75% of the class. Another group has reported finding similar to our experience observing that students performing poorly in the entry examination used the CAL programme most and were the individuals with the greatest improvement in grades, although class ranking was not reported[2] Some report the use of CAL as a remedial tool for those performing poorly in conventional evaluation and show improvement of scores following CAL instruction[3]
Criticisms have been raised that iCAL may not be suitable for students with negative attitudes to computers and a student preference for hardcopy material over computer screen presentation is reported[4,5] Some randomised trials of CAL versus lectures reported superior objective outcome measures in the CAL group although students reported a preference for the traditional lecture structure[6] However a positive attitude to CAL was reported when it was an addition to traditional teaching methods, rather then as a replacement[7] The general response of medical students is positive with the view that CAL is a novel and fun way to learn[8] Medical students have however expressed concern that the delivery of an education programme by computer will compromise the student trainer relationship if the computer supplants other forms of training[8] The direction chosen at our unit was not to replace lectures, but rather to complement them with this additional teaching resource. A negative attitude to computers and CAL may not be an obstruction to successful learning however, as knowledge is increased by usage of a CAL programme even among medical students who reported negative attitudes to computers and CAL[5] Not every report of CAL in medical education is positive and it appears that the style of teaching and stage of clinical development of the student is important to the outcome[9,10] While problem based teaching and clinical simulation are very effective methods of CAL in clinical years, it appears that in the early years of basic science instruction in medicine, a more didactic CAL delivery of material is superior to a problem based presentation[10,11]
Few randomised controlled trials have been performed in CAL medical education and many of those that have been performed look at small subunits of the overall medical corriculum[1] Of the trials that have been performed on CAL, the evidence suggests a high information assimilation rate among medical students[5,10-13] Seabra et al. wanted to explore the outcome of replacing lectures completely with a CAL programme and demonstrated that computer aided learning could achieve similar results to standard lectures[14] Some evidence suggests that although the gain in knowledge is similar following computer or traditional instruction, the time required to achieve these similar results is less when the student uses the computer aided instruction[15]
Conclusion
Our experience shows that electronic learning complements traditional methods in undergraduate surgical teaching. It offers advantages to the teaching staff in the speed and convenience of dissemination of information. For students the establishment of an internet based CAL system offers the advantage of distance learning access outside the working day. However our experience has been that such student utilisation of the programme was not associated with improvements in class rank. The iCAL programme was more frequently used by students with lower class ranking with traditional teaching methods, and iCAL usage by these students is associated with improvements in class ranking. The incorporation of iCAL programmes into undergraduate medical education offers distinct administrative advantages for the teaching staff and more learning opportunities for students with diverse abilities and learning preferences.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Greenhalgh T Computer assisted learning in undergraduate medical education. BMJ 2001 322 40 45 11141156 10.1136/bmj.322.7277.40
Holt RI Miklaszewicz P Cranston IC Russell-Jones D Rees PJ Sonksen PH Computer assisted learning is an effective way of teaching endocrinology. Clin Endocrinol 2001 55 537 542 10.1046/j.1365-2265.2001.01346.x
Karnath BM Thornton W DasCarlos M Pilot study of computer based self teaching in cardiac auscultation. Med Educ 2003 37 1048 1049 14629447 10.1046/j.1365-2923.2003.01643.x
Vogel M Wood DF Love it or hate it? Medical students attitudes to computer aided learning. Med Educ 2002 36 214 215 11879510 10.1046/j.1365-2923.2002.01181.x
Lynch TG Steele DJ Palensky JEJ Lacy N Duffy SW Learning preferences, computer attitudes and test performance with computer aided instruction. Am J Surg 2001 181 368 371 11438276 10.1016/S0002-9610(01)00575-X
William C Aubin S Harkin P Cottrell D A randomized, controlled, single-blind trial of teaching provided by a computer-based multimedia package versus lecture. Med Educ 2001 35 847 854 11555222 10.1046/j.1365-2923.2001.00960.x
Reid WA Harvey J Watson GR Luqmani R PJR H Arends MJ Medical student appraisal of interactive computer assisted learning programmes embedded in a general pathology course. J Pathol 2000 191 462 465 10918223 10.1002/1096-9896(2000)9999:9999<::AID-PATH655>3.0.CO;2-P
Steele DJ Palensky JEJ Lynch TG Lacy NL Duffy SW Learning preferences, computer attitudes, and student evaluation of computerised instruction. Med Educ 2002 36 225 232 11879512 10.1046/j.1365-2923.2002.01141.x
Lieberman G Abramson R Volkan K McArdle PJ Tutor versus computer: a prospective comparison of interactive tutorials and computer assisted instruction in radiology education. Acad Radiol 2002 9 40 49 11918358 10.1016/S1076-6332(03)80295-7
Devitt P Palmer E Computer aided learning: an overvalued education resource? Med Educ 1999 33 136 139 10211264 10.1046/j.1365-2923.1999.00284.x
Devitt PG Cehic D Palmer E Computers in medical education -2: use of a computer package to supplement the clinical experience in a surgical clerkship: an objective evaluation. Aust NZ J Surg 1998 68 457 460
Lipman AJ Sade RM Glotzbach AC Lancaster CJ Marshell MF The incremental value of internet based instruction as an adjunct to clasroom instruction: a propsective randomized study. Acad Med 2001 76 1060 1064 11597850
Gilbart MK Hutchison CR Cusimano MD Regehr G A computer-based trauma simulator for teaching trauma management skills. Am J Surg 2000 179 223 228 10827325 10.1016/S0002-9610(00)00302-0
Seabra D Srougi M Baptista R Nesrallah LJ Ortiz V Sigulem D Computer aided learning versus standard lecture for undergraduate education in urology. J Urol 2004 171 1220 1222 14767306 10.1097/01.ju.0000114303.17198.37
Shomaker TS Ricks DJ Hale DC A prospective, randomized, controlled study of computer assisted learning in parasitology. Acad Med 2002 77 446 449 12010707
|
15987526
|
PMC1184079
|
CC BY
|
2021-01-04 16:30:56
|
no
|
BMC Med Educ. 2005 Jun 29; 5:23
|
utf-8
|
BMC Med Educ
| 2,005 |
10.1186/1472-6920-5-23
|
oa_comm
|
==== Front
BMC Med EducBMC Medical Education1472-6920BioMed Central London 1472-6920-5-231598752610.1186/1472-6920-5-23Research ArticleElectronic learning can facilitate student performance in undergraduate surgical education: a prospective observational study Healy David Gerard [email protected] Fergal J [email protected] David [email protected] Patrick [email protected] Alfred Edward [email protected] Thomas [email protected] Enda W [email protected] John M [email protected]'Higgins Niall J [email protected] Arnold DK [email protected] Department of Surgery, Mater Misericordiae University Hospital, Eccles St, Dublin 7, Ireland2 Department of Surgery, St Vincent's University Hospital, Elm Park, Dublin 4, Ireland3 Conway Institute of Biomolecular & Biomedical Research, University College Dublin, Dublin 4, Ireland4 Centre for Health Informatics, University College Dublin, Dublin 4, Ireland2005 29 6 2005 5 23 23 12 3 2005 29 6 2005 Copyright © 2005 Healy et al; licensee BioMed Central Ltd.2005Healy et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Our institution recently introduced a novel internet accessible computer aided learning (iCAL) programme to complement existing surgical undergraduate teaching methods. On graduation of the first full cycle of undergraduate students to whom this resource was available we assessed the utility of this new teaching facility.
Method
The computer programme prospectively records usage of the system on an individual user basis. We evaluated the utilisation of the web-based programme and its impact on class ranking changes from an entry-test evaluation to an exit examination in surgery.
Results
74.4% of students were able to access iCAL from off-campus internet access. The majority of iCAL usage (64.6%) took place during working hours (08:00–18:00) with little usage on the weekend (21.1%). Working hours usage was positively associated with improvement in class rank (P = 0.025, n = 148) but out-of hours usage was not (P = 0.306). Usage during weekdays was associated with improved rank (P = 0.04), whereas weekend usage was not (P = 0.504). There were no significant differences in usage between genders (P = 0.3). Usage of the iCAL system was positively correlated with improvement in class rank from the entry to the exit examination (P = 0.046). Students with lower ranks on entry examination, were found to use the computer system more frequently (P = 0.01).
Conclusion
Electronic learning complements traditional teaching methods in undergraduate surgical teaching. Its is more frequently used by students achieving lower class ranking with traditional teaching methods, and this usage is associated with improvements in class ranking.
==== Body
Background
Medicine has become increasingly complex and the challenges faced by the medical education system are becoming even greater. Higher levels of technical and scientific knowledge are required as well as effective communication and management skills. Acquisition of this knowledge must be achieved within a finite time period. In addition the clinical opportunities for medical students are diminishing with decreasing length of hospital stay for many surgical procedures. This environment has led to the search for novel teaching methods to deliver undergraduate medical education.
Computer aided learning (CAL) offers distinct advantages over conventional teaching methods, including the potential for multimedia applications with a significant interactive content. Not only may text be presented, but in addition, tables, images, video and animation can be integrated into dynamic packages. In this manner multimedia education strategies offer potential strategic advantages over traditional paper based material. The system is very flexible allowing a lecturer to update and upload data in the package and to monitor an individual student's activity. An internet-based computer aided learning (iCAL) programme has the additional advantage of access from an off-site location at any hour of the day. Unfortunately the establishment of the infrastructure for iCAL is quite costly and the cost to benefit analysis of undergraduate iCAL has been questioned[1] The Department of Surgery at University College Dublin, Ireland, has recently introduced an iCAL package for undergraduate education. We wished to evaluate the effectiveness of this programme following the first complete cycle of undergraduate medical students.
Methods
Medical undergraduates at University College Dublin (UCD) have a forty two week training period in surgery. This is divided in two parts, a 28-week period and a 14-week period. Students are allocated clinical attachments to hospital surgical services and rotated though the spectrum of surgical specialities. During these rotations they participate in ward, theatre and outpatient activities but have no clinical responsibilities. Clinical "by the bedside" teaching with clinical lectures in small groups is provided. In addition a formal lecture programme is delivered and library facilities are provided for independent learning. This structure is the same in the initial 28 week and the later 14 week periods. The clinical sites are identical, and clinical rotations are organised so that all students get an exposure to all surgical specialties over the course of their training over the entire 42 week period. Whether an individual student goes through a particular rotation in the initial 28 weeks or the later 14 weeks, is a random allocation.
At the end of the first period of the programme, undergraduates have a surgical examination consisting of a multiple choice paper and a clinical examination. For the purposes of this paper this is referred to as the entry examination. At the end of the second period of the course students take the final exit examination in surgery. The exit exam also consists of a multiple choice exam and clinical exam with an additional written short answer examination.
iCAL-SURG was used during the second period of the surgery course, after students had taken their entry examination. The only fundamental methodology difference between these two training periods was the introduction of the iCAL programme. The questions presented through the iCAL system are identical in style to those presented in the short written examination and clinical examination problem solving scenarios.
The class was divided into quartiles on the basis of this entry examination. Utilisation of the iCAL programme by students was optional and was compared with absolute performance and change in class ranking in the final examination. One hundred and forty eight students completed the two year programme. Data on these students' utilisation of the iCAL platform was analysed.
Technical infrastructure
All students in UCD Medical Faculty are conversant with use of ICT, with CAL used to deliver up to 35% of preclinical courses. Clinical students have access to a suite of networked high specification multimedia PC's with T1 internet access. This facility is available to students from 8.00 to 18.00, weekdays. iCAL-SURG is also available anywhere with an internet connection. Online course material is delivered within the Blackboard Virtual Learning Environment. Courses are password protected and use of iCAL-SURG by individual students is monitored. Instructors can find the number of times each student has logged on to iCAL-SURG, the areas of the course accessed, the time and day of log-on. The number of times the site was accessed by the student is referred to as the "hit rate". It is not possible to record the content viewed or duration of student activity, once logged on.
iCAL-SURG
During this pilot phase, iCAL-SURG use was optional and did not form part of the assessment process for students. This phase concentrated on the COURSE DOCUMENTS and the ASSIGNMENTS areas in Blackboard. The COMMUNICATION area was used for both social and academic purposes but only accounted for 5.3% of use of iCAL-SURG and no assessment was made of this in relation to student performance.
In the COURSE DOCUMENTS area [Fig, 1,2] lecture notes are available prior to lectures, either as Powerpoint, Adobe Portable Document Format (PDF) or Microsoft Word files. The material available on the iCAL system not only included material presented at lectures but also additional supplementary material. Old examination question papers and sample answers are also available here. In the ASSIGNMENTS area each week the academic staff uploaded a short series of clinical questions with suggested answers made available made available [Fig. 3]. Students read these and studied the relevant areas in text books or in discussion with tutors and peers. Students could also take online OSCE examinations [Fig. 4,5]. This was presented as a series of photographs with related clinical details. Students were asked to answer the questions in a prepared downloaded Microsoft Word document, which they then uploaded to the "Digital Dropbox" in Blackboard. These submissions were reviewed by academic staff. On the following day sample answers were published [Fig. 4,5] which students could compare with their submissions.
Figure 1 Screenshot from iCAL-SURG.
Figure 2 Screenshot taken from some materials available in course materials.
Figure 3 This is an example of a short OSCE available on iCAL-SURG.
Figure 4 Screenshot of link to OSCE with instructions.
Figure 5 Screenshot of first station in a full online OSCE.
Statistical analysis
A Student t test was used for comparison between groups of two. Where more than two groups were involved an ANOVA was performed followed by Post-hoc analysis as appropriate, with a Fischer's least significant difference test. Data involving class rank was analysed non-parametrically with a Mann Whitney U Test for comparison between groups of two, a Kruskal Wallis where more groups were involved and a Spearman Rank Correlation for investigating correlations. Alpha was set at 0.05.
Results
Complete data on computer utilisation was available on 148 students. There were 58 males and 90 females. There were no significant gender differences in the utilisation of online access (P = 0.3) or final year examination results (P = 0.251). One of the criticisms of the programme from the start was the concern that students without independent computer and internet access would be disadvantaged. 74.4% of students were able to access iCAL from off-campus internet access. Despite this high availability of independent internet access, the majority of iCAL usage (64.6%) took place during daytime hours (08:00–18:00) [Fig. 6] with less then expected usage on the weekend (21.1%) [Fig. 7].
Figure 6 This shows the pattern of computer usage over the day. Data is presented as means with standard errors.
Figure 7 This shows the number of hits into the computer system each day. Use of the system on a Friday, Saturday or Sunday was significantly less than use on Monday, Tuesday and Thursday (ANOVA P < 0.001). * P < 0.05
To assess the educational benefits of the programme the class was ranked based on the entry test results. Class rank on the entry test was then compared to the class rank at the exit exam. Usage of the iCAL system during daytime hours was positively associated with improvement in class rank (P = 0.025) but out-of hours usage was not (P = 0.306). Usage during weekdays was associated with improved rank (P = 0.04), whereas weekend usage was not (P = 0.504). There was no significant difference in improvement in class rank between students with independent internet access and those dependant on access provided by the University Hospitals. We therefore felt that the computer facilities provided by the University were adequate for the needs of the students and that students without independent computer and internet access were not disadvantaged.
We were also interested in the level of usage of iCAL by different student abilities. Students with lower ranking on the entry test were found to subsequently use the computer system more frequently (P = 0.01). The lower quartiles of the class were the most frequent users of the programme (ANOVA, P = 0.042) [Fig. 8]. We then wanted to evaluate whether this usage translated into improvements in class rank. Usage of the iCAL system was positively correlated with improvement in class rank from the entry to the exit examination (P = 0.046).
Figure 8 Students have been divided into quartiles based on the performance in the entry surgical examination. A] This shows the usage of the surgical computer education package by students in different class rank quartiles. Significantly more use was made of the computer package by students in the lowest entry rank quartile (ANOVA P = 0.042). B] This shows the change in class ranking between entry examination and final examination in surgery. Initially low ranking students showed the greatest improvement in their class position (Kruskal Wallis P = 0.01). Mean and standard error. * P < 0.05, ** P < 0.01; compared to lowest class quartile at entry test.
Discussion
This cohort was the first group to complete their undergraduate surgical training since the introduction of this dynamic iCAL package. Of concern to us in establishing this program was the potential to competitively disadvantage students without independent computer and internet access. This in part motivated the decision not to include iCAL as part of the formal student assessment. However our study shows no such disadvantage. Instead, structured usage of iCAL during normal working time was associated with improved class performance rather then usage out of hours. This suggests that improvement in class rank with iCAL was not facilitated by studying late into the night but rather by a structured and disciplined approach. This conclusion is limited by the fact that the system cannot distinguish a student accessing the system from home during these hours or from the University computer suite. However as the students had other onsite daily activities during the day, with monitored attendance, it seems most likely that this activity occurred from University provided facilities.
One of the objectives of this study was to examine the patterns of usage of the iCAL system. Part of the motivation in using an internet accessible format was the expectation that students could utilise the programme as a distance learning tool. Although 35% of iCAL utilisation took place outside of the normal academic day, this usage was not associated with improvements in relative position in the class. This at first glance questions the benefit of an internet based infrastructure. However this technology has other advantages to a faculty using more then one teaching hospital, such as enabling the educators to upload data from multiple sites.
We were also interested in the utilisation of the programme by students of different class rank. It was interesting to us that the top two quartiles of the class, as ranked by the entry examination, did not utilise the iCAL programme as frequently as the lower two quartiles. The factors underlying this are uncertain. The method of assessing usage was the number of times an individual student logs into the system. The package does not measure the amount of time spent with the package or the volume of content covered by the student. It is conceivable that students at the lower end of the class took longer to completely digest the content of the web pages and required more frequent visits to the same sites. However if that is the case it appears to have been rewarding as the lowest 25% of the class on the entry test, who also used the iCAL package most, were the group who made the most progress up the class ranking. It must be observed that those at the top have little room to move up and those at the bottom little distance to fall in terms of class ranking. This could have influenced an outcome measure based on class rank. However the pattern was not only seen with the top and bottom quartile but is also seen in the central 25–75% of the class. Another group has reported finding similar to our experience observing that students performing poorly in the entry examination used the CAL programme most and were the individuals with the greatest improvement in grades, although class ranking was not reported[2] Some report the use of CAL as a remedial tool for those performing poorly in conventional evaluation and show improvement of scores following CAL instruction[3]
Criticisms have been raised that iCAL may not be suitable for students with negative attitudes to computers and a student preference for hardcopy material over computer screen presentation is reported[4,5] Some randomised trials of CAL versus lectures reported superior objective outcome measures in the CAL group although students reported a preference for the traditional lecture structure[6] However a positive attitude to CAL was reported when it was an addition to traditional teaching methods, rather then as a replacement[7] The general response of medical students is positive with the view that CAL is a novel and fun way to learn[8] Medical students have however expressed concern that the delivery of an education programme by computer will compromise the student trainer relationship if the computer supplants other forms of training[8] The direction chosen at our unit was not to replace lectures, but rather to complement them with this additional teaching resource. A negative attitude to computers and CAL may not be an obstruction to successful learning however, as knowledge is increased by usage of a CAL programme even among medical students who reported negative attitudes to computers and CAL[5] Not every report of CAL in medical education is positive and it appears that the style of teaching and stage of clinical development of the student is important to the outcome[9,10] While problem based teaching and clinical simulation are very effective methods of CAL in clinical years, it appears that in the early years of basic science instruction in medicine, a more didactic CAL delivery of material is superior to a problem based presentation[10,11]
Few randomised controlled trials have been performed in CAL medical education and many of those that have been performed look at small subunits of the overall medical corriculum[1] Of the trials that have been performed on CAL, the evidence suggests a high information assimilation rate among medical students[5,10-13] Seabra et al. wanted to explore the outcome of replacing lectures completely with a CAL programme and demonstrated that computer aided learning could achieve similar results to standard lectures[14] Some evidence suggests that although the gain in knowledge is similar following computer or traditional instruction, the time required to achieve these similar results is less when the student uses the computer aided instruction[15]
Conclusion
Our experience shows that electronic learning complements traditional methods in undergraduate surgical teaching. It offers advantages to the teaching staff in the speed and convenience of dissemination of information. For students the establishment of an internet based CAL system offers the advantage of distance learning access outside the working day. However our experience has been that such student utilisation of the programme was not associated with improvements in class rank. The iCAL programme was more frequently used by students with lower class ranking with traditional teaching methods, and iCAL usage by these students is associated with improvements in class ranking. The incorporation of iCAL programmes into undergraduate medical education offers distinct administrative advantages for the teaching staff and more learning opportunities for students with diverse abilities and learning preferences.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Greenhalgh T Computer assisted learning in undergraduate medical education. BMJ 2001 322 40 45 11141156 10.1136/bmj.322.7277.40
Holt RI Miklaszewicz P Cranston IC Russell-Jones D Rees PJ Sonksen PH Computer assisted learning is an effective way of teaching endocrinology. Clin Endocrinol 2001 55 537 542 10.1046/j.1365-2265.2001.01346.x
Karnath BM Thornton W DasCarlos M Pilot study of computer based self teaching in cardiac auscultation. Med Educ 2003 37 1048 1049 14629447 10.1046/j.1365-2923.2003.01643.x
Vogel M Wood DF Love it or hate it? Medical students attitudes to computer aided learning. Med Educ 2002 36 214 215 11879510 10.1046/j.1365-2923.2002.01181.x
Lynch TG Steele DJ Palensky JEJ Lacy N Duffy SW Learning preferences, computer attitudes and test performance with computer aided instruction. Am J Surg 2001 181 368 371 11438276 10.1016/S0002-9610(01)00575-X
William C Aubin S Harkin P Cottrell D A randomized, controlled, single-blind trial of teaching provided by a computer-based multimedia package versus lecture. Med Educ 2001 35 847 854 11555222 10.1046/j.1365-2923.2001.00960.x
Reid WA Harvey J Watson GR Luqmani R PJR H Arends MJ Medical student appraisal of interactive computer assisted learning programmes embedded in a general pathology course. J Pathol 2000 191 462 465 10918223 10.1002/1096-9896(2000)9999:9999<::AID-PATH655>3.0.CO;2-P
Steele DJ Palensky JEJ Lynch TG Lacy NL Duffy SW Learning preferences, computer attitudes, and student evaluation of computerised instruction. Med Educ 2002 36 225 232 11879512 10.1046/j.1365-2923.2002.01141.x
Lieberman G Abramson R Volkan K McArdle PJ Tutor versus computer: a prospective comparison of interactive tutorials and computer assisted instruction in radiology education. Acad Radiol 2002 9 40 49 11918358 10.1016/S1076-6332(03)80295-7
Devitt P Palmer E Computer aided learning: an overvalued education resource? Med Educ 1999 33 136 139 10211264 10.1046/j.1365-2923.1999.00284.x
Devitt PG Cehic D Palmer E Computers in medical education -2: use of a computer package to supplement the clinical experience in a surgical clerkship: an objective evaluation. Aust NZ J Surg 1998 68 457 460
Lipman AJ Sade RM Glotzbach AC Lancaster CJ Marshell MF The incremental value of internet based instruction as an adjunct to clasroom instruction: a propsective randomized study. Acad Med 2001 76 1060 1064 11597850
Gilbart MK Hutchison CR Cusimano MD Regehr G A computer-based trauma simulator for teaching trauma management skills. Am J Surg 2000 179 223 228 10827325 10.1016/S0002-9610(00)00302-0
Seabra D Srougi M Baptista R Nesrallah LJ Ortiz V Sigulem D Computer aided learning versus standard lecture for undergraduate education in urology. J Urol 2004 171 1220 1222 14767306 10.1097/01.ju.0000114303.17198.37
Shomaker TS Ricks DJ Hale DC A prospective, randomized, controlled study of computer assisted learning in parasitology. Acad Med 2002 77 446 449 12010707
|
15967034
|
PMC1184080
|
CC BY
|
2021-01-04 23:52:11
|
no
|
BMC Med Inform Decis Mak. 2005 Jun 20; 5:18
|
latin-1
|
BMC Med Inform Decis Mak
| 2,005 |
10.1186/1472-6947-5-18
|
oa_comm
|
==== Front
BMC Med Inform Decis MakBMC Medical Informatics and Decision Making1472-6947BioMed Central London 1472-6947-5-211597813810.1186/1472-6947-5-21Research ArticleDistribution of immunodeficiency fact files with XML – from Web to WAP Väliaho Jouni [email protected] Pentti [email protected] Mauno [email protected] Institute of Medical Technology, FI-33014 University of Tampere, Finland2 Department of Information Technology, University of Turku, FI-20520 Turku, Finland3 Research Unit, Tampere University Hospital, FI-33520 Tampere, Finland2005 26 6 2005 5 21 21 11 3 2004 26 6 2005 Copyright © 2005 Väliaho et al; licensee BioMed Central Ltd.2005Väliaho et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Although biomedical information is growing rapidly, it is difficult to find and retrieve validated data especially for rare hereditary diseases. There is an increased need for services capable of integrating and validating information as well as proving it in a logically organized structure. A XML-based language enables creation of open source databases for storage, maintenance and delivery for different platforms.
Methods
Here we present a new data model called fact file and an XML-based specification Inherited Disease Markup Language (IDML), that were developed to facilitate disease information integration, storage and exchange. The data model was applied to primary immunodeficiencies, but it can be used for any hereditary disease. Fact files integrate biomedical, genetic and clinical information related to hereditary diseases.
Results
IDML and fact files were used to build a comprehensive Web and WAP accessible knowledge base ImmunoDeficiency Resource (IDR) available at . A fact file is a user oriented user interface, which serves as a starting point to explore information on hereditary diseases.
Conclusion
The IDML enables the seamless integration and presentation of genetic and disease information resources in the Internet. IDML can be used to build information services for all kinds of inherited diseases. The open source specification and related programs are available at .
==== Body
Background
Biomedical information is often very complex. Deciphering the roles of genes in human health and disease is a grand challenge for many reasons, including impediments to defining phenotypes, difficulties in identifying and quantifying environmental effects, technical problems in generating genotypic information, and the difficulties of studying humans [1]. The completion of the draft sequence of the human genome [2,3] and advances in molecular biology provide new opportunities to increase our understanding of the role of genetic factors in human health and disease [1]. The number of identified genetic diseases has increased exponentially [4]. The new knowledge can be applied to the prevention, diagnosis and treatment of diseases. This far, the knowledge of genetics has had a large role in the health care of only a few patients and a small role in the health care of many [5]. The biomedical informatics holds great promise for developing informatics methods that will be crucial in the development of genomic medicine [6].
Most hereditary diseases are rare and the diagnosed patients for a condition are often randomly spread out in the world. One doctor usually has only a few patients with a disease. It is often difficult to find comprehensive and validated biomedical information related to rare diseases. In addition, it is more and more difficult to publish results in scientific journals only from a few cases even when they are interesting [7]. Still, all these pieces of information can contain clues to understanding the fundamental defects at molecular level and can help to develop targeted treatments. The scattering of the disease-related information to literature and Internet is a big obstacle especially for those interested in rare diseases. First of all, there may not be that much data for these diseases and secondly it may be very difficult to find and collect. Further, the user has often difficulties in assessing the quality of data.
There is an increasing need for tools and services capable of integrating information from a variety of sources. Clinicians and researchers could benefit from a more consolidated and unified view of the available biomedical data. Systems biology researchers need to integrate disparate information from multiple public sources to merge with their own experimental data to generate models of processes. Biomedical data mining attempts to extract information from biomedical databases by using e.g. automated natural language processing (NLP) techniques [8]. Processing of biomedical texts presents many challenges such as in the areas of terminology or ontology building, information extraction from texts, knowledge discovery from collections of documents, as well as sharing and integrating knowledge from factual and textual data bases, semantic annotation, etc. Without standardized nomenclature the information extraction (IE) about a particular subject from various resources is difficult. Due to ambiguity of terms, a search for a particular term often retrieves results for unrelated entities. Since there are also some technical problems arising from the diversity of computer hardware and software, there is a need for such a data form, that can be handled by any computer and which can be easily presented on any platform.
The Extensible Markup Language (XML) is a standard created by the World Wide Web Consortium (W3C) for characterizing the content and structure of documents [9]. It is designed to improve the functionality of the Web by enabling more flexible and adaptable information identification and presentation. XML allows to define tags and document structures for own context-specific use. It was derived from SGML (Standard Generalized Markup Language), the international standard for defining descriptions of the structure and content of different types of electronic documents [10]. XML is simpler than SGML, but it allows the use of richly structured documents over the Internet. Information encoded in XML is easy to read and understand, and easy to process by computers. In XML files, structured data are bounded by tags and attributes. XML tags, attributes and element structure provide context information that facilitates the interpretation of the meaning of content, thereby making it feasible to develop efficient search engines and agents and perform intelligent data mining, etc. The XML allows the separation of content, logic and presentation.
Beyond XML there are a number of additional specifications such as Document Object Model (DOM) [11], XML Schemas [12], XSL Transformations [13], and Resource Description Framework (RDF) [14]. XML will have a big role in integration and interoperation of biological databases. Some biomedical information models have been implemented using XML specifications [15,16], many of them being clinical models for electronic healthcare documents [17-19].
A unified data format of resources is required for comparison between similar diseases and reutilization of information. Here we present a new data model called fact file, which integrates biomedical information related to hereditary diseases into a Web and WAP accessible knowledge base. Our scope is wider than e.g. in gene oriented knowledge bases such as GeneCards [20], UniGene [21], or LocusLink [22]. The disease information sources are even more diverse than those for genetic information. The fact files concentrate on sharing and integrating biomedical knowledge from different sources. The presented data model can be applied to any hereditary disease.
The fact files were applied to build a comprehensive, validated knowledge base for primary immunodeficiencies (PIDs) called ImmunoDeficiency Resource (IDR) [23,24]. It is designed for different user groups such as researchers, physicians and nurses as well as patients and their families and the general public. The IDR is the major information source to immunodeficiencies in the Web. Fact files serve as the core of the IDR knowledge base.
Methods
The fact file data model and the Inherited Disease Markup Language (IDML) were developed to facilitate disease information integration, storage and exchange in the first place for immunodeficiencies, but in principle for any hereditary disease. The IDML is an XML specification and container for bioinformatical data on hereditary diseases. The fact file data model schema was defined according to W3C XML specification [12, see Additional files 1, 2, 3]. The fact file data model can be depicted as a tree structure graph where a <FactFile> element is a root (Figure 1). Fact files make use of the following specifications, standards and databases: HUGO nomenclature [25], RefSeq [22], Swiss-Prot [26] and SOURCE [27].
Figure 1 The tree diagram of main IDML elements of fact file data model. The operators "?" (optional), "*" (zero or more) and "+" (one or more) are used to denote cardinality indicating how many instances of an element type are permitted.
Stand-alone IDML fact files have been generated for each PID. The fact files are uniquely identified by an id attribute of FactFile root element. The major concepts in the first tier of the fact file hierarchy below the root level are general information, clinical information and molecular biology (Table 1). In addition, there are other resources, which contain links to related information providers. Each of these elements, in turn, comprises one or more additional levels of guideline constructs.
Table 1 The description of high-level concepts in the fact file document model
Element Description Content Content modela
FactFile The root element for IDML-based fact file document Elements (GeneralInformation, ClinicalInformation, MolecularBiology, Other)
GeneralInformation Describes the disease in general terms Elements (DiseaseName, Abbreviation*, AlternativeNames?, Description, Classification?, Omim*, CrossReferences?, Incidence?)
ClinicalInformation The short overview of characteristic clinical features Elements (ClinicalDescription?, Diagnosis?, TherapeuticOptions?, ResearchPrograms?)
MolecularBiology Molecular genetic elements Elements (GeneInformation?, AnimalModels?, ProteinInformation?, ExpressionPattern?)
Other Other related information Elements (Publications?, Societies?, OtherSites?)
a Content model consists of a set of parenthesis enclosing some combination of child element names and operators. The order operators "," (strict sequence) and "|" (choice) indicate how elements may be combined. The operators "?" (optional), "*" (zero or more) and "+" (one or more) are used to denote cardinality indicating how many instances of an element type are permitted?
The components of the fact file model are defined as IDML elements. According to XML, elements have distinct names and they are delimited with start and end tag, e.g. <DiseaseName>X-linked agammaglobulinemia</DiseaseName>. Elements may contain other elements or attributes, they may store text, or they may be empty. Elements may appear as often as required. Many IDML elements contain href attribute for hyperlinking to more detailed information by using globally unique idenfier URL (Unified Resource Locator). The element naming convention follows the approach used by Electronic Business XML (ebXML) core components [28]. The IDML specified element names are in upper camel case (UpperCamelCase) and attribute names are in lower camel case (lowerCamelCase) notations. The usage of acronyms has been avoided, but when they are used the capitalization remains (example: ReferenceDNA).
General information elements identify a particular genetic disease and describe pattern of heritance and frequency in general terms (Table 2). The <Abbreviation> element includes commonly used abbreviations and the <AlternativeNames> element lists known aliases and synonymous names for the disease. The <Description> element provides a short overview in general terms. The description text may contain several Glink -tagged words that can act as links to a glossary, which is an integral part of the IDR service. The <C lassification> element is used to classify a disease explicitly to a group of related diseases. It exploits the hierarchic structure of XML documents by nesting <Class> elements. Each <Class> element contains a unique identifier in level attribute and a class name in <Title> element. The <Omim> element links the fact file to the Online Mendelian Inheritance in Man (OMIM) knowledge base [4] and the <CrossReferences> element refers to the related fact files grouped in <Phenotype>, <Gene>, and <OtherRelatedDiseases> elements. The incidence element stores information about disease frequency in human populations.
Table 2 The description of IDML: GeneralInformation element
Element Description Content Content modela
DiseaseName Disease name Type String
Abbreviation Abbreviation for disease name Type String
AlternativeNames List of alternatively used disease names Elements (Name*)
Description General description of disease Mixed (Glink | Italic)*
Classification Classifies document explicitely in the fact files hierarchy Elements (Class)
Omim A collection of the related references to the OMIM database Elements (OmimReference+)
CrossReferences Refers to the related fact files Elements (PhenotypeRelatedDiseases?, OtherRelatedDiseases?, GeneRelatedDiseases?)
Incidence Description of incidence Type String
a See table 1
Clinical information elements provide a short overview of characteristic clinical features, diagnosis, treatment and research related to the disease (Table 3). The <ClinicalDescription> element stores text, that describes characteristic clinical features and the most important laboratory findings. The <Diagnosis> refers to data on diagnostic criteria and guidelines. It also refers to databases of laboratories performing clinical and/or genetic analyses for the disease including IDdiagnostics [29], the European Directory of DNA diagnostic Laboratories (EDDNAL, ) and GeneTests [30]. Detailed diagnostic guidelines are available for several IDs [31]. The <TherapeuticOptions> lists therapeutic interventions that are available. The <ResearchPrograms> includes important research and clinical trials related to the disease.
Table 3 The description of IDML: ClinicalInformation element
Element Description Content Content modela
ClinicalDescription Describes characteristic clinical features Mixed (Glink | Italic)*
Diagnosis A collection of diagnostic guidelines and laboratories Elements (DiagnosticRecommendations?, AdditionalInformation?, DiagnosticLaboratories?)
TherapeuticOptions A collection of available therapeutic options Elements (Option+)
ResearchPrograms A collection of related studies (Program+)
a See table 1
Molecular biology comprises the main genetic components on DNA, RNA and protein level, animal models, protein properties and expression patterns (Table 4). The <GeneInformation> elements store the basic information on gene names, aliases and synonyms. They provide also <ReferenceSequences> element that covers reference sequences on three levels and lists also other related sequences that are available from sequence databanks. Information on gene locus is stored in <ChromosomalLocation>, <Maps>, and <Markers> elements. The <Variations> element refers to related locus specific mutation and single nucleotide polymorphism (SNP) databases. We and others are maintaining a large number of immunodeficiency mutation databases [24,32]. The <OtherResources> element refers to the other genetic web services such as Ensembl [33], GENATLAS [34], GeneCards [20], UniGene [21], LocusLink [22], euGenes [35], GDB [36], GeneLynx [37] and SOURCE [27]. The <AnimalModels> element refers to the related transgenic animal studies.
Table 4 The description of IDML: MolecularBiology element
Element Description Content Content modela
GeneInformation Contains information on the gene name, aliases, reference sequences, chromosomal location, maps, markers, variations and other gene related resources Elements (Name?, Aliases?, ReferenceSequences?, OtherSequences?, ChromosomalLocation?, Maps?, Markers?, Variations?, OtherResources?)
AnimalModels A collection of related transgenic animal studies Elements (Animal*)
ProteinInformation Contains information on protein characteristic features, structures, domains, motifs and other protein resources Elements (ProteinDescription?, Structures?, Domains?, Motifs?, ProteinResources?)
ExpressionPattern Gene expression levels in a variety of cells and tissues Elements (Expression*)
a See table 1
The <ProteinInformation> element stores characteristic structural and functional properties of the protein. The <ProteinDescription> contains several subelements e.g. <Function>, <SubcellularLocation>, <CatalyticActivity>, which are inherited from the Swiss-Prot entry model [26]. The <Structures> element refers to solved protein structures available in Protein DataBank (PDB) [38]. The domain and motif elements describe conserved protein regions. Each <Domain>, <Motif> and further <ProteinResources> element includes links to related resources for example in Pfam [39], InterPro [40], ProDom [41], SMART [42] or PROSITE [43]. The <ExpressionPattern> stores information on gene or protein expression. This information is mainly from SOURCE [27], which is a web-based resource bringing together genetic information from different sources.
The last high level element <Other> stores various information in elements such as <Publications> and <Societies>, which is categorized by <GeneralSocieties> and <DiseaseSpecificSocieties> elements (Table 5). The <OtherSites> element refers to other related resources in the Internet.
Table 5 The description of IDML: Other element
Element Description Content Content modela
Publications A collection of related publications Elements (PubmedSearch?, Pubmed?)
Societies List of related general and disease specific societies Elements (GeneralSocieties?, DiseaseSpecificSocieties?)
OtherSites A collection of other related Web sites Elements (Site+)
a See table 1
The IDML schema version 1.0 (idml.xsd file), examples of IDML-document and documentation on the syntax are available at our web site . The IDML document type definition file (idml.dtd) is also available, althougth we prefer to use the IDML schema for validation. Many IDML elements are optional. The syntax allows one to put comments, both within and outside of the XML markup. The parser must pass internal comments to the application programs, which can then properly treat the information. IDML documents specify which version of the schema is to be used to validate their content, eliminating possible confusion when several versions exist. IDML is open access, however, a licence is needed for building other services. Contact the authors for details.
Results
The IDML model was implemented to describe primary immunodeficiencies, which is a group of over 100 hereditary diseases. IDs can be grouped as follows: combined B and T cell immunodeficiencies, deficiencies predominantly affecting antibody production, defects in lymphocyte apoptosis, other well-defined immunodeficiency syndromes, defects of phagocyte function, interferon-γ (IFNγ) associated immunodeficiencies, DNA breakage associated syndromes, defects of the complement cascade proteins, and defects of complement regulatory proteins. The disease information is stored in IDML-based fact files, which form the central repository for data retrieval of ImmunoDeficiency Resource (IDR) service [23,24,29] available at . The data flow diagram of IDR is shown in figure 2. In addition to information on fact files, the IDR contains several introductory texts and collections of immunology related data sources. The IDR pages are extensively hyperlinked to our on-line immunology glossary. More detailed description about the IDR web service has been published elsewhere [23,24].
Figure 2 The data flow diagram of IDR. Notations: A red open-ended rectangle represents data store, light yellow ovals represent processes, blue rectangles represent external entities, orange rounded rectangles shows transferred documents, green rounded rectangles represent destination devices and arrows shows the flow of information from its source to its destination. White rectangles show the separation of content, logic and presentation.
The ImmunoDeficiency Resource is a comprehensive knowledge base on immunodeficiencies. IDR is developed and maintained by IMT Bioinformatics group in collaboration with experts on individual immunodeficiencies. All the information in the IDR will be validated by expert curators. However, all changes, additions and corrections to the fact files are made by our group. IDR is designed for different user groups such as researchers, physicians and nurses as well as patients and their families and the general public. IDR contains fact files for practically all known PIDs. The numerous individual data items in IDR have been collected partly manually, usually with simple Perl scripts written for datamining from numerous local and Internet databases and services.
We selected Apache AxKit XML Application Server version 1.61 for implementation of the IDML-encoded web service. AxKit is an application and document server that uses XML processing pipelines to generate and process content and to deliver it to clients in a wide variety of formats, such as HTML, WML, PDF and plain text using either standard techniques of World Wide Web Consortium (XSLT) [13], or flexible custom codes (XPathScript XPS, eXtensible Server Pages XSP).
Similar XML application server called Cocoon, has been written in Java. We settled on AxKit, because it is built in Perl, which makes it easy to integrate with bioinformatic applications many of which are written in Perl. It is important to note that AxKit is not limited to XML source documents. Non-XML documents and data sources can be converted to XML when necessary. AxKit separates the content, logic and presentation. The content reuse was implemented with XInclude [44] and XPointer [45] techniques. The root element of IDML schema is <FactFiles> and according to W3C Recommendation "Namespaces in XML" [46] we declared a default namespace attribute in the root element xmlns:idml="" to avoid the problems of ambiquity and name collisions.
Each fact file is stored in an IDML file, that has a unique name and url address. When a fact file requests the pipeline it might look like this in diagramatic terms : Request > [XSP] > (XML) > [XSLT] > (HTML) > Browser, where processors are in square brackets and products in round brackets. The output of XSP pages is structured XML content, which can pipe through XSLT to produce HTML. The XSP feature is not currently in use in the IDR.
The information on fact files can be easily transformed and presented in any platform. It is easy to write platform or even browser and screen specific pages. We have implemented a transformation from IDML to WML for portable devices (such as mobile phones) with WAP compliance (Wireless Application Protocol). The fact files are available via bioinformatics related WAP service, BioWAP [47,48] practically anywhere, anytime.
New web techniques are developed continuously. During this project a number of new specifications and software appeared, requiring upgrading of the system many times. The separation of content and presentation enables to share the project for people who are responsible for information content and people who develop the knowledge management techniques. Once the data model was created, we have not had to touch it hardly at all in spite of technical improvements, content additions and deletions.
Discussion
As far as we know there are no other efforts to develop a markup language to describe connections between disease and genetic information. The IDML was designed with following purposes in mind. First, we wanted a markup that is able to present disease, clinical, diagnostic and genetic information and relations between them. Secondly, the data model structure had to be intuitive, hierarchical, flexible, but still machine and human readable. Sometimes the relatively large XML files can appear verbose for human readers, but hierarchically and logically organized structure in addition to semantic markup facilitate the interpretation of documents. Thirdly, an application and platform independent data format was needed. Its portability, extensibility and robustness are primary advantages for interoperating heterogeneous systems. The availability of open source and free tools for processing files in all major programming languages is important. The openness of source code as well as data formats and data itself allows better integration and interoperation between data resources. The IDML enables the seamless integration of genetic and disease information resources in the Internet. The data model is appropriate for the implementation of automated decision support systems such as diagnostic consultations. Fourthly, the data have to be unambiguous and validated.
A fact file is a user oriented user interface, which serves as a good starting point to explore information on hereditary diseases. For some time now, there has been many advanced search facilities in the Internet such as Google, that are able to find very fast web pages that contain given keywords. However, the web searches typically turn up innumerable completetely irrelevant "hits", requiring much manual filtering by the user. Navarro et al. lists some issues related to database searching and accessibility that can cause difficulties [49] including inaccurate and redundant search results, nomenclature issues, lack of internal access, non-availability of the source code, lack of customization and differing data formats. New methods are needed for improving search results.
There is an increasing number of biomedical data sources in the Internet. The Human Genome Initiative [2] and other genome research projects have generated enormous quantities of data. The genetic data is well organized in web accessible databases for example EMBL [50], GenBank [51], Swiss-Prot [26], etc. Several organizations offer public interfaces for obtaining biomedical information across a range of domains. They provide numerous tools and applications for genetic data retrieval and analysis for example with Sequence Retrieval System SRS [52] and BioPerl [53]. In addition to the sequence information, databases contain a lot of valuable information in annotations. There are also some genetic knowledge bases such as GeneCards and GeneLynx that comprise the essential information on genes. Swiss-Prot contains also some disease related annotations. The most comprehensive database on hereditary diseases is OMIM [4], which contains descriptions for known hereditary diseases.
Almost all pages in the Internet have been written in HyperText Markup Language (HTML) where it is used for style description. It provides some possibilities for simple description about a document. It is able to use special metatags that contain simple keywords or more advanced descriptions like Dublin Core Languages, but they are very little utilised and only the most sophisticated search engines can exploit them.
There are some efforts to integrate heterogeneous biomedical databases [15,54,55]. Some level of standardization is required for more automatic integration. Development of integration techniques is moving databases towards the Internet and XML-based systems [56]. In the future, Web services will use standard Internet protocols including SOAP, WSDL, and UDDI for interoperability with other resources. Thereby the flexible and expandable integration of diverse scientific tools will be achieved.
Conclusion
The XML-based language IDML and fact file data model were developed for integrating, storing and exchanging information on inherited diseases. The IDML language and fact file model are implemented in the IDR knowledge base. The fact files can be easily transformed from IDML to any format such as HTML or WML using either standard W3C techniques or flexible custom code. The content management as well as the exchange of presentation are facilitated by separating document content and presentation. The IDML-based information system was proved to be a viable and applicable specification for inherited diseases. Numerous downloads (altogether more than 250,000) from the IDR knowledge base during the last two years have proved the applicability and adaptability of the fact file model.
List of abbreviations
AxKit An XML Delivery Toolkit for Apache
BioWAP Bioinformatics service for portable devices
DOM Document Object Model
DTD Document Type Definition
ebXML Electronic Business XML
EDDNAL European Directory of DNA Diagnostic Laboratories
EMBL Genetic sequence database by European Molecular Biology Laboratory
GDB Genome Database
GenBank Genetic sequence database by National Center for Biotechnology Information
HTML Hypertext Markup Language
IDML Inherited Disease Markup Language
IDR ImmunoDeficiency Resource
IE Information extraction
IFNγ Interferon gamma
NLP Natural language processing
OMIM Online Mendelian Inheritance in Man
PID Primary Immunodeficiency
PDB Protein DataBank
PDF Portable Document Format
Pfam Protein Families Database
ProDom Protein Domain Database
PROSITE Database of Protein Families and Domains
RDF Resource Description Framework
SGML Standard Generalized Markup Language
SMART Simple Modular Architecture Research Tool
SNP Single nucleotide polymorphism
SOAP Simple Object Access Protocol
SOURCE Genomic resource in the Internet
Swiss-Prot Protein knowledgebase
UDDI Universal Description, Discovery, and Integration
URL Unified Resource Locator
W3C World Wide Web Consortium
WAP Wireless Application Protocol
WML Wireless Markup Language
WSDL Web Services Definition/Description Language
XML Extensible Markup Language
XPS XPathScript
XSL Extensible Style Language
XSLT XSL Transformations
XSP eXtensible Server Pages
Competing interests
The author(s) declare that they have no competing interests.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
An XML schema for IDML
Click here for file
Additional File 2
A DTD file for IDML
Click here for file
Additional File 3
An example file using IDML. A fact file for X-linked agammaglobulinemia.
Click here for file
Acknowledgements
Financial support from the European Union, the National Technology Agency of Finland and the Medical Research Fund of Tampere University Hospital is gratefully acknowledged.
==== Refs
Collins FS Green ED Guttmacher AE Guyer MS A vision for the future of genomics research Nature 2003 422 835 847 12695777 10.1038/nature01626
Lander ES Linton LM Birren B Nusbaum C Zody MC Baldwin J Devon K Dewar K Doyle M FitzHugh W Funke R Gage D Harris K Heaford A Howland J Kann L Lehoczky J LeVine R McEwan P McKernan K Meldrim J Mesirov JP Miranda C Morris W Naylor J Raymond C Rosetti M Santos R Sheridan A Sougnez C Stange-Thomann N Stojanovic N Subramanian A Wyman D Rogers J Sulston J Ainscough R Beck S Bentley D Burton J Clee C Carter N Coulson A Deadman R Deloukas P Dunham A Dunham I Durbin R French L Grafham D Gregory S Hubbard T Humphray S Hunt A Jones M Lloyd C McMurray A Matthews L Mercer S Milne S Mullikin JC Mungall A Plumb R Ross M Shownkeen R Sims S Waterston RH Wilson RK Hillier LW McPherson JD Marra MA Mardis ER Fulton LA Chinwalla AT Pepin KH Gish WR Chissoe SL Wendl MC Delehaunty KD Miner TL Delehaunty A Kramer JB Cook LL Fulton RS Johnson DL Minx PJ Clifton SW Hawkins T Branscomb E Predki P Richardson P Wenning S Slezak T Doggett N Cheng JF Olsen A Lucas S Elkin C Uberbacher E Frazier M Gibbs RA Muzny DM Scherer SE Bouck JB Sodergren EJ Worley KC Rives CM Gorrell JH Metzker ML Naylor SL Kucherlapati RS Nelson DL Weinstock GM Sakaki Y Fujiyama A Hattori M Yada T Toyoda A Itoh T Kawagoe C Watanabe H Totoki Y Taylor T Weissenbach J Heilig R Saurin W Artiguenave F Brottier P Bruls T Pelletier E Robert C Wincker P Smith DR Doucette-Stamm L Rubenfield M Weinstock K Lee HM Dubois J Rosenthal A Platzer M Nyakatura G Taudien S Rump A Yang H Yu J Wang J Huang G Gu J Hood L Rowen L Madan A Qin S Davis RW Federspiel NA Abola AP Proctor MJ Myers RM Schmutz J Dickson M Grimwood J Cox DR Olson MV Kaul R Shimizu N Kawasaki K Minoshima S Evans GA Athanasiou M Schultz R Roe BA Chen F Pan H Ramser J Lehrach H Reinhardt R McCombie WR de la Bastide M Dedhia N Blocker H Hornischer K Nordsiek G Agarwala R Aravind L Bailey JA Bateman A Batzoglou S Birney E Bork P Brown DG Burge CB Cerutti L Chen HC Church D Clamp M Copley RR Doerks T Eddy SR Eichler EE Furey TS Galagan J Gilbert JG Harmon C Hayashizaki Y Haussler D Hermjakob H Hokamp K Jang W Johnson LS Jones TA Kasif S Kaspryzk A Kennedy S Kent WJ Kitts P Koonin EV Korf I Kulp D Lancet D Lowe TM McLysaght A Mikkelsen T Moran JV Mulder N Pollara VJ Ponting CP Schuler G Schultz J Slater G Smit AF Stupka E Szustakowski J Thierry-Mieg D Thierry-Mieg J Wagner L Wallis J Wheeler R Williams A Wolf YI Wolfe KH Yang SP Yeh RF Collins F Guyer MS Peterson J Felsenfeld A Wetterstrand KA Patrinos A Morgan MJ Szustakowki J de Jong P Catanese JJ Osoegawa K Shizuya H Choi S Chen YJ Initial sequencing and analysis of the human genome Nature 2001 409 860 921 11237011 10.1038/35057062
Venter JC Adams MD Myers EW Li PW Mural RJ Sutton GG Smith HO Yandell M Evans CA Holt RA Gocayne JD Amanatides P Ballew RM Huson DH Wortman JR Zhang Q Kodira CD Zheng XH Chen L Skupski M Subramanian G Thomas PD Zhang J Gabor Miklos GL Nelson C Broder S Clark AG Nadeau J McKusick VA Zinder N Levine AJ Roberts RJ Simon M Slayman C Hunkapiller M Bolanos R Delcher A Dew I Fasulo D Flanigan M Florea L Halpern A Hannenhalli S Kravitz S Levy S Mobarry C Reinert K Remington K Abu-Threideh J Beasley E Biddick K Bonazzi V Brandon R Cargill M Chandramouliswaran I Charlab R Chaturvedi K Deng Z Di Francesco V Dunn P Eilbeck K Evangelista C Gabrielian AE Gan W Ge W Gong F Gu Z Guan P Heiman TJ Higgins ME Ji RR Ke Z Ketchum KA Lai Z Lei Y Li Z Li J Liang Y Lin X Lu F Merkulov GV Milshina N Moore HM Naik AK Narayan VA Neelam B Nusskern D Rusch DB Salzberg S Shao W Shue B Sun J Wang Z Wang A Wang X Wang J Wei M Wides R Xiao C Yan C Yao A Ye J Zhan M Zhang W Zhang H Zhao Q Zheng L Zhong F Zhong W Zhu S Zhao S Gilbert D Baumhueter S Spier G Carter C Cravchik A Woodage T Ali F An H Awe A Baldwin D Baden H Barnstead M Barrow I Beeson K Busam D Carver A Center A Cheng ML Curry L Danaher S Davenport L Desilets R Dietz S Dodson K Doup L Ferriera S Garg N Gluecksmann A Hart B Haynes J Haynes C Heiner C Hladun S Hostin D Houck J Howland T Ibegwam C Johnson J Kalush F Kline L Koduru S Love A Mann F May D McCawley S McIntosh T McMullen I Moy M Moy L Murphy B Nelson K Pfannkoch C Pratts E Puri V Qureshi H Reardon M Rodriguez R Rogers YH Romblad D Ruhfel B Scott R Sitter C Smallwood M Stewart E Strong R Suh E Thomas R Tint NN Tse S Vech C Wang G Wetter J Williams S Williams M Windsor S Winn-Deen E Wolfe K Zaveri J Zaveri K Abril JF Guigo R Campbell MJ Sjolander KV Karlak B Kejariwal A Mi H Lazareva B Hatton T Narechania A Diemer K Muruganujan A Guo N Sato S Bafna V Istrail S Lippert R Schwartz R Walenz B Yooseph S Allen D Basu A Baxendale J Blick L Caminha M Carnes-Stine J Caulk P Chiang YH Coyne M Dahlke C Mays A Dombroski M Donnelly M Ely D Esparham S Fosler C Gire H Glanowski S Glasser K Glodek A Gorokhov M Graham K Gropman B Harris M Heil J Henderson S Hoover J Jennings D Jordan C Jordan J Kasha J Kagan L Kraft C Levitsky A Lewis M Liu X Lopez J Ma D Majoros W McDaniel J Murphy S Newman M Nguyen T Nguyen N Nodell M Pan S Peck J Peterson M Rowe W Sanders R Scott J Simpson M Smith T Sprague A Stockwell T Turner R Venter E Wang M Wen M Wu D Wu M Xia A Zandieh A Zhu X The sequence of the human genome Science 2001 291 1304 1351 11181995 10.1126/science.1058040
Hamosh A Scott AF Amberger J Bocchini C Valle D McKusick VA Online Mendelian Inheritance in Man (OMIM), a knowledgebase of human genes and genetic disorders Nucleic Acids Res 2002 30 52 55 11752252 10.1093/nar/30.1.52
Guttmacher AE Collins FS Genomic medicine - a primer N Engl J Med 2002 347 1512 1520 12421895 10.1056/NEJMra012240
Maojo V Kulikowski CA Bioinformatics and medical informatics: collaborations on the road to genomic medicine? J Am Med Inform Assoc 2003 10 515 522 12925552 10.1197/jamia.M1305
Grivell L Mining the bibliome: searching for a needle in a haystack? New computing tools are needed to effectively scan the growing amount of scientific literature for useful information EMBO Rep 2002 3 200 203 11882534 10.1093/embo-reports/kvf059
Yandell MD Majoros WH Genomics and natural language processing Nat Rev Genet 2002 3 601 610 12154383
World Wide Web Consortium Extensible Markup Language (XML) 1.0 (Second Edition)
International Organization for Standardization Standard Generalized Markup Language (SGML)
World Wide Web Consortium Document Object Model (DOM) Level 1 Specification
World Wide Web Consortium XML Schema Part 0: Primer
World Wide Web Consortium XSL Transformations (XSLT) Version 1.0
World Wide Web Consortium RDF/XML Syntax Specification (Revised)
Mork P Halevy A Tarczy-Hornoch P A model for data integration systems of biomedical data applied to online genetic databases Proc AMIA Symp 2001 473 477 11825233
Achard F Vaysseix G Barillot E XML, bioinformatics and data integration Bioinformatics 2001 17 115 125 11238067 10.1093/bioinformatics/17.2.115
Coyle JF Mori AR Huff SM Standards for detailed clinical models as the basis for medical data exchange and decision support Int J Med Inf 2003 69 157 174 10.1016/S1386-5056(02)00103-X
McDonald CJ Huff SM Suico JG Hill G Leavelle D Aller R Forrey A Mercer K DeMoor G Hook J Williams W Case J Maloney P LOINC, a universal standard for identifying laboratory observations: a 5-year update Clin Chem 2003 49 624 633 12651816 10.1373/49.4.624
Lee KP Hu J XML Schema Representation of DICOM Structured Reporting J Am Med Inform Assoc 2003 10 213 223 12595410 10.1197/jamia.M1042
Rebhan M Chalifa-Caspi V Prilusky J Lancet D GeneCards: encyclopedia for genes, proteins and diseases
Schuler GD Pieces of the puzzle: expressed sequence tags and the catalog of human genes J Mol Med 1997 75 694 698 9382993 10.1007/s001090050155
Pruitt KD Maglott DR RefSeq and LocusLink: NCBI gene-centered resources Nucleic Acids Res 2001 29 137 140 11125071 10.1093/nar/29.1.137
Väliaho J Pusa M Ylinen T Vihinen M IDR: the ImmunoDeficiency Resource Nucleic Acids Res 2002 30 232 234 11752302 10.1093/nar/30.1.232
Väliaho J Riikonen P Vihinen M Novel immunodeficiency data servers Immunol Rev 2000 178 177 185 11213802 10.1034/j.1600-065X.2000.17807.x
Wain HM Lush MJ Ducluzeau F Khodiyar VK Povey S Genew: the Human Gene Nomenclature Database, 2004 updates Nucleic Acids Res 2004 Database issue D255 7 14681406 10.1093/nar/gkh072
Boeckmann B Bairoch A Apweiler R Blatter MC Estreicher A Gasteiger E Martin MJ Michoud K O'Donovan C Phan I Pilbout S Schneider M The SWISS-PROT protein knowledgebase and its supplement TrEMBL in 2003 Nucleic Acids Res 2003 31 365 370 12520024 10.1093/nar/gkg095
Diehn M Sherlock G Binkley G Jin H Matese JC Hernandez-Boussard T Rees CA Cherry JM Botstein D Brown PO Alizadeh AA SOURCE: a unified genomic resource of functional annotations, ontologies, and gene expression data Nucleic Acids Res 2003 31 219 223 12519986 10.1093/nar/gkg014
Eisenberg B Nicull D ebXML Technical Architecture Specification v1.0.4
Samarghitean C Valiaho J Vihinen M Online Registry of Genetic and Clinical Immunodeficiency Diagnostic Laboratories, IDdiagnostics J Clin Immunol 2004 24 53 61 14997034 10.1023/B:JOCI.0000018063.55963.0c
Tarczy-Hornoch P Shannon P Baskin P Espeseth M Pagon RA GeneClinics: a hybrid text/data electronic publishing model using XML applied to clinical genetic testing J Am Med Inform Assoc 2000 7 267 276 10833163
Conley ME Notarangelo LD Etzioni A Diagnostic criteria for primary immunodeficiencies. Clin Immunol 1999 93 190 197 10600329 10.1006/clim.1999.4799
Vihinen M Arredondo-Vega FX Casanova JL Etzioni A Giliani S Hammarström L Hershfield MS Heyworth PG Hsu AP Lähdesmäki A Lappalainen I Notarangelo LD Puck JM Reith W Roos D Schumacher RF Schwarz K Vezzoni P Villa A Väliaho J Smith CI Primary immunodeficiency mutation databases Adv Genet 2001 43 103 188 11037300
Birney E Andrews D Bevan P Caccamo M Cameron G Chen Y Clarke L Coates G Cox T Cuff J Curwen V Cutts T Down T Durbin R Eyras E Fernandez-Suarez XM Gane P Gibbins B Gilbert J Hammond M Hotz H Iyer V Kahari A Jekosch K Kasprzyk A Keefe D Keenan S Lehväslaiho H McVicker G Melsopp C Meidl P Mongin E Pettett R Potter S Proctor G Rae M Searle S Slater G Smedley D Smith J Spooner W Stabenau A Stalker J Storey R Ureta-Vidal A Woodwark C Clamp M Hubbard T Ensembl 2004 Nucleic Acids Res 2004 Database issue D468 70 14681459 10.1093/nar/gkh038
Frezal J Genatlas database, genes and development defects C R Acad Sci III 1998 321 805 817 9835018
Gilbert DG euGenes: a eukaryote genome information system Nucleic Acids Res 2002 30 145 148 11752277 10.1093/nar/30.1.145
GDB Human Genome Database
Lenhard B Hayes WS Wasserman WW GeneLynx: a gene-centric portal to the human genome Genome Res 2001 11 2151 2157 11731507 10.1101/gr.199801
Bernstein FC Koetzle TF Williams GJ Meyer EFJ Brice MD Rodgers JR Kennard O Shimanouchi T Tasumi M The Protein Data Bank: a computer-based archival file for macromolecular structures J Mol Biol 1977 112 535 542 875032
Bateman A Birney E Cerruti L Durbin R Etwiller L Eddy SR Griffiths-Jones S Howe KL Marshall M Sonnhammer EL The Pfam protein families database Nucleic Acids Res 2002 30 276 280 11752314 10.1093/nar/30.1.276
Mulder NJ Apweiler R Attwood TK Bairoch A Barrell D Bateman A Binns D Biswas M Bradley P Bork P Bucher P Copley RR Courcelle E Das U Durbin R Falquet L Fleischmann W Griffiths-Jones S Haft D Harte N Hulo N Kahn D Kanapin A Krestyaninova M Lopez R Letunic I Lonsdale D Silventoinen V Orchard SE Pagni M Peyruc D Ponting CP Selengut JD Servant F Sigrist CJ Vaughan R Zdobnov EM The InterPro Database, 2003 brings increased coverage and new features Nucleic Acids Res 2003 31 315 318 12520011 10.1093/nar/gkg046
Servant F Bru C Carrere S Courcelle E Gouzy J Peyruc D Kahn D ProDom: automated clustering of homologous domains Brief Bioinform 2002 3 246 251 12230033
Letunic I Copley RR Schmidt S Ciccarelli FD Doerks T Schultz J Ponting CP Bork P SMART 4.0: towards genomic data integration Nucleic Acids Res 2004 Database issue D142 4 14681379 10.1093/nar/gkh088
Sigrist CJ Cerutti L Hulo N Gattiker A Falquet L Pagni M Bairoch A Bucher P PROSITE: a documented database using patterns and profiles as motif descriptors Brief Bioinform 2002 3 265 274 12230035
World Wide Web Consortium XML Inclusions (XInclude) Version 1.0
World Wide Web Consortium XML Pointer Language (XPointer)
World Wide Web Consortium Namespaces in XML
Riikonen P Boberg J Salakoski T Vihinen M BioWAP, mobile Internet service for bioinformatics Bioinformatics 2001 17 855 856 11590108 10.1093/bioinformatics/17.9.855
Riikonen P Boberg J Salakoski T Vihinen M Mobile access to biological databases on the Internet IEEE Trans Biomed Eng 2002 49 1477 1479 12542244 10.1109/TBME.2002.805459
Navarro JD Niranjan V Peri S K. JC From biological databases to platforms for biomedical discovery Trends in Biotechnol 2003 21 263 268 10.1016/S0167-7799(03)00108-2
Kulikova T Aldebert P Althorpe N Baker W Bates K Browne P van den Broek A Cochrane G Duggan K Eberhardt R Faruque N Garcia-Pastor M Harte N Kanz C Leinonen R Lin Q Lombard V Lopez R Mancuso R McHale M Nardone F Silventoinen V Stoehr P Stoesser G Tuli MA Tzouvara K Vaughan R Wu D Zhu W Apweiler R The EMBL Nucleotide Sequence Database Nucleic Acids Res 2004 Database issue D27 30 14681351 10.1093/nar/gkh120
Benson DA Karsch-Mizrachi I Lipman DJ Ostell J Wheeler DL GenBank: update Nucleic Acids Res 2004 Database issue D23 6 14681350 10.1093/nar/gkh045
Zdobnov EM Lopez R Apweiler R Etzold T The EBI SRS server-new features Bioinformatics 2002 18 1149 1150 12176845 10.1093/bioinformatics/18.8.1149
Stajich JE Block D Boulez K Brenner SE Chervitz SA Dagdigian C Fuellen G Gilbert JG Korf I Lapp H Lehväslaiho H Matsalla C Mungall CJ Osborne BI Pocock MR Schattner P Senger M Stein LD Stupka E Wilkinson MD Birney E The Bioperl toolkit: Perl modules for the life sciences Genome Res 2002 12 1611 1618 12368254 10.1101/gr.361602
Lawrence R Barker K Integrating relational database schemas using a standardized dictionary: ; Las Vegas, Nevada, United States. Proceedings of the 2001 ACM symposium on Applied computing 2001 ACM Press 225 230
Stevens RD Robinson AJ Goble CA myGrid: personalised bioinformatics on the information grid Bioinformatics 2003 Suppl 1 I302 I304 12855473 10.1093/bioinformatics/btg1041
Das M Lawhead PB Information storage and management in large web-based applications using XML J Comput Small Coll 2003 18 72 79
|
15978138
|
PMC1184081
|
CC BY
|
2021-01-04 23:52:11
|
no
|
BMC Med Inform Decis Mak. 2005 Jun 26; 5:21
|
utf-8
|
BMC Med Inform Decis Mak
| 2,005 |
10.1186/1472-6947-5-21
|
oa_comm
|
==== Front
BMC Med Res MethodolBMC Medical Research Methodology1471-2288BioMed Central London 1471-2288-5-191591889810.1186/1471-2288-5-19Research ArticleNo role for quality scores in systematic reviews of diagnostic accuracy studies Whiting Penny [email protected] Roger [email protected] Jos [email protected] MRC Health Services Research Collaboration, Department of Social Medicine, University of Bristol, Bristol, UK2 Centre for Reviews and Dissemination, University of York, York, UK2005 26 5 2005 5 19 19 9 2 2005 26 5 2005 Copyright © 2005 Whiting et al; licensee BioMed Central Ltd.2005Whiting et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
There is a lack of consensus regarding the use of quality scores in diagnostic systematic reviews. The objective of this study was to use different methods of weighting items included in a quality assessment tool for diagnostic accuracy studies (QUADAS) to produce an overall quality score, and to examine the effects of incorporating these into a systematic review.
Methods
We developed five schemes for weighting QUADAS to produce quality scores. We used three methods to investigate the effects of quality scores on test performance. We used a set of 28 studies that assessed the accuracy of ultrasound for the diagnosis of vesico-ureteral reflux in children.
Results
The different methods of weighting individual items from the same quality assessment tool produced different quality scores. The different scoring schemes ranked different studies in different orders; this was especially evident for the intermediate quality studies. Comparing the results of studies stratified as "high" and "low" quality based on quality scores resulted in different conclusions regarding the effects of quality on estimates of diagnostic accuracy depending on the method used to produce the quality score. A similar effect was observed when quality scores were included in meta-regression analysis as continuous variables, although the differences were less apparent.
Conclusion
Quality scores should not be incorporated into diagnostic systematic reviews. Incorporation of the results of the quality assessment into the systematic review should involve investigation of the association of individual quality items with estimates of diagnostic accuracy, rather than using a combined quality score.
==== Body
Background
Quality assessment is as important in systematic reviews of diagnostic accuracy studies as it is for any other systematic review. One method of incorporating quality into a review is to use a quality score. Quality scores combine the individual items from a quality assessment tool to provide an overall single score. One of the main problems with quality scores is determining how to weight each item to provide an overall quality score. There is no objective way of doing this and different methods are likely to produce different scores that may lead to different results if these scores are used in the analysis.
There has been much discussion regarding the use of quality scores in the area of clinical trials[1-8]. Although this discussion has not been specific to diagnostic accuracy studies much of these discussions also apply to this topic area. Previous work illustrating the problems associated with quality scores has used different scales, which not only weighted items differently but also included different items[9]. It has been argued that it was the differences in the items covered by the tools that contributed to the differences found, rather than the use of a combined quality score[2,3,6]. The debate regarding quality scores remains and quality scores continue to be used as part of the quality assessment process in both therapeutic and diagnostic systematic reviews [10-14]. The Jadad scale, one of the most commonly used quality assessment tools for therapeutic studies, incorporates a quality score[15], as does one of the commonly used diagnostic quality assessment tools[16]. A recent review of existing quality assessment tools for diagnostic accuracy studies found that 12 of 67 tools (18%) incorporated a quality score[17]. A further review of how quality assessment has been incorporated into systematic reviews found that 16% of reviews that performed some form of quality assessment used quality scores as part of this assessment[18].
We are not aware of any work that has looked at the effect of using different weightings for the same quality assessment tool to produce an overall quality score or that has been done in the area of diagnostic accuracy studies. This project presents a practical example of the problems associated with the use of quality scores in systematic reviews. The aim is to use QUADAS, a quality assessment tool that we recently developed to assess the quality of diagnostic accuracy studies included in systematic reviews[19], to investigate the effect of different weightings on estimates of test performance.
Methods
Scoring methods
QUADAS does not incorporate a quality score. We therefore developed five different schemes for weighting QUADAS (Table 1) to produce an overall study quality score:
Table 1 QUADAS and scoring guide for each of the 5 schemes
QUADAS Item Scoring scheme
1 2 3 4 5
1 Was the spectrum of patients representative of the patients who will receive the test in practice? 1 2 2 3 10
2 Were selection criteria clearly described? 1 2 1 1 2
3 Is the reference standard likely to correctly classify the target condition? 1 2 3 2 10
4 Is the time period between reference standard and index test short enough to be reasonably sure that the target condition did not change between the two tests? 1 2 3 1 6
5 Did the whole sample or a random selection of the sample, receive verification using a reference standard of diagnosis? 1 2 3 3 9
6 Did patients receive the same reference standard regardless of the index test result? 1 2 3 2 7
7 Was the reference standard independent of the index test (i.e. the index test did not form part of the reference standard)? 1 2 3 1 7
8 Was the execution of the index test described in sufficient detail to permit replication of the test? 1 2 2 1 3
9 Was the execution of the reference standard described in sufficient detail to permit its replication? 1 2 2 1 2
10 Were the index test results interpreted without knowledge of the results of the reference standard? 1 2 3 3 8
11 Were the reference standard results interpreted without knowledge of the results of the index test? 1 2 3 3 6
12 Were the same clinical data available when test results were interpreted as would be available when the test is used in practice? 1 2 3 3 5
13 Were uninterpretable/ intermediate test results reported? 1 2 1 1 4
14 Were withdrawals from the study explained? 1 2 1 1 3
Total score 14 28 33 26 85
All scoring given above refer to the score which studies that answered "yes" to each question should be given. Studies that answered "no" or "unclear" were scored 0 for each scoring system with the exception of system 2 in which studies that scored "unclear" were given 1.
1. Equal weighting
All items were weighted equally and scored 1 for yes and 0 for no or unclear.
2. Equal weighting accounting for unclear
All items were weighted equally but scored 2 for yes, 1 for unclear and 0 for no.
3. Weighting according to item type
Items which aimed to detect the presence of bias were scored 3 for yes (items 3, 4, 5, 6, 7, 10, 11, 12), items which aimed to detect sources of variation between studies were scored 2 for yes (item 1) and items which were related to the quality of reporting were scored 1 for yes (items 2, 8, 9, 13, 14). All items were scored 0 for no or unclear.
4. Weighting based on the evidence
The evidence used in the development of QUADAS was used to determine item weighting[18]. Two systematic reviews of the diagnostic literature provided an evidence base for the development of QUADAS. The first was a review of evidence on factors that can lead to bias or variation in the results of diagnostic accuracy studies[20]. For each source of bias or variation, the number of studies that found that a particular source of bias or variation impacted on estimates of diagnostic accuracy was summarised. The second review considered all existing quality assessment tools designed for diagnostic accuracy studies[17]. The proportion of tools that covered each of a list of possible items relating to the quality of diagnostic accuracy studies was summarised. To estimate quality scores using this weighting scheme, items for which there was evidence of bias or variation from at least 5 studies or which were included in at least 75% of existing quality assessment tools were scored 3 for yes (items 1, 5, 10, 11, 12); items for which there was evidence of bias from at least 2 studies and which were included in at least 50% of existing quality assessment tools were scored 2 points for yes (items 3, 6). All other items were given 1 point for yes (items: 4, 7, 8, 9, 13, 14). All items were scored 0 for no or unclear.
5. Subjective scoring
Each item was scored from 1 – 10 based on one of the author's subjective opinion of its importance. This allowed items which the author considered to be of greater importance to receive a much greater weighting than items considered less important. For example items such as inclusion of an appropriate patient spectrum and the use of an appropriate reference standard were judged to be much more important than items such as reporting of selection criteria or details of the reference standard. This is reflected in the weightings given to these items.
These weighting schemes are summarised in Table 1. Each different weighting scheme was used to produce an overall quality score, giving a total of five different scores for each study. As the total maximum possible points differed across the scoring schemes, the scores were expressed as the percentage of the maximum possible points for each scoring scheme so that the quality scores could be compared across schemes.
Data set
We selected a data set consisting of 28 studies that looked at ultrasound for the diagnosis of vesico-ureteral reflux in children. These came from a systematic review on the diagnosis and further investigation of urinary tract infection (UTI) in children under 5[21]. The studies were selected as they provided a set of studies that were heterogeneous in terms of quality and individual study results. They provide two separate data sets within one larger data set as they can be split according to the type of ultrasound used: contrast-enhanced (16 studies) or standard ultrasound (12 studies). Although both types of study evaluated ultrasound and so involve similar quality issues, there were differences in accuracy between the ultrasound types: contrast-enhanced ultrasound is a much more accurate test for vesico-ureteral reflux in children than standard-ultrasound.
Thus we were able to investigate whether different quality scores have the same impact on two separate data sets. QUADAS was used in this review to assess the quality of studies. All studies had previously been coded using QUADAS as yes, no or unclear. This coding was carried out by one reviewer and checked by a second reviewer.
Analysis
Methods for investigating the effects of the quality scores on test performance We used three different methods to investigate the effects of quality scores on test performance. Each method was performed separately for the standard ultrasound studies and for the contrast-enhanced ultrasound studies. For each of the steps involving pooling of studies, standard SROC (summary receiver operating characteristic) methods were used to pool individual study results[22]. The SROC model was estimated by regressing D (log(DOR), where DOR is the diagnostic odds ratio) against S (logit (sensitivity) + logit (1-specificity)), weighting according to sample size, for each study. To account for zero cells in the 2 × 2 tables, 0.5 was added to every cell for all 2 × 2 tables as recommended by Moses et al.[22]. All analyses were carried out using STATA version 8 (StataCorp, College Station, Texas).
a. Ranking of studies
Studies were ranked according to quality score and we investigated whether the ranking of each study was different according to the method used to weight the quality scores. This allowed investigation of whether the use of a summary quality score in a table as an overall indicator of quality is appropriate.
b. Difference in estimates diagnostic accuracy between high and low quality studies
We stratified studies into "high" and "low" quality studies using the quality score. The median quality score was calculated for each scoring scheme. Studies with scores higher than the median score were classified as "high" quality studies, while studies with the median quality score or lower were classified as "low" quality studies. A relative diagnostic odds ratio (RDOR) was calculated for each of the different quality scores by dividing the pooled diagnostic odds ratio (DOR) for the high quality studies by that for the low quality studies.
c. Quality score as a possible source of heterogeneity
The effects of quality on test performance were investigated using meta-regression analysis. The SROC model was extended to include "quality score" as a continuous variable, assuming a linear relationship between quality score and log DOR. We calculated the RDOR for a 10 point increase in quality by multiplying the coefficient for the quality score obtained from the regression analysis by 10 and then anti-logging it.
Results
Table 2 summarises the results for the 28 studies included in this study. It presents the 2 × 2 table results for each study, the results of the quality assessment, and the summary quality scores produced using each of the five scoring schemes. Reading table 2 vertically per item allows readers to make some judgments about which items might contribute to variations in the scores. Figure 1 shows the results of the studies plotted in receiver operating characteristic (ROC) space, giving an indication of the heterogeneity between studies.
Table 2 Individual study results (2 × 2 data), results of the quality assessment, and quality scores using each of the five scoring schemes
Study details 2 × 2 Data QUADAS Results Quality score
TP FP FN TN Spectrum composition Selection criteria Reference standard Disease progression bias Partial verification bias Differential verification bias Incorporation bias Test execution details Reference execution details Test review bias Diagnostic review bias Clinical review bias Uninterpretable results Withdrawals Score 1 Score 2 Score 3 Score 4 Score 5
Standard US
Baronciani (1986) [24] 13 4 8 49 + + + ? + + + - - ? ? ? ? ? 43 64 45 46 53
Dura (1997) [25] 3 4 14 27 - + + + + + + + + + + ? + + 86 89 85 77 79
Evans (1999) [26] 2 10 17 84 - + + ? + + + + - ? ? ? + ? 50 68 48 42 49
Foresman (2001) [27] 24 43 25 47 - + + + + + + + - + + ? + + 79 82 79 73 76
Mage (1989) [28] 22 5 19 76 - - + ? + + + + - ? ? ? ? ? 36 57 42 35 42
Mahant (2002) [29] 14 30 21 97 - + + + + + + + - + + ? ? ? 64 75 73 65 68
Morin (1999) [30] 20 41 2 7 - + + + + + + + - ? ? ? ? ? 50 68 55 42 52
Muensterer (2002) [31] 35 76 34 241 - + + + + + + + - ? ? ? ? + 57 71 58 46 55
Oostenbrink (2000) [32] 21 20 16 83 + + + ? - ? + + - + ? ? ? - 43 61 42 42 47
Salih (1994) [33] 26 3 1 12 + - + + + + + - - ? ? ? ? - 43 57 52 46 58
Tan (1988) [34] 3 6 14 32 - - + + + + + + + - - ? ? ? 50 61 58 42 52
Verber (1988) [35] 8 9 20 25 + - + ? - + + + + ? ? ? ? - 43 61 45 38 46
Median Score 50 66 53 44 52
Contrast-enhanced US
Alzen (1994) [36] 20 6 2 73 - - + + + + + - - ? ? ? ? ? 36 54 45 35 46
Bergius (1990) [23] 56 2 14 176 + - + ? + + + + - + + + ? - 64 71 76 81 76
Berrocal (2001) [37] 94 29 10 307 - - + + + + + + - ? ? ? ? ? 43 61 52 38 49
Berrocal Frutos (2000) [38] 63 19 7 204 - + + + + + + + + + + ? + + 86 89 85 77 79
Haberlick (1997) [39] 21 10 9 114 - + + + + + + + - ? ? ? ? ? 50 68 55 42 52
Kessler (1982) [40] 13 0 4 38 - + + ? - + + + - + ? ? ? - 43 57 45 38 44
McEwing (2002) [41] 8 3 8 173 - + + + + + + + - + + ? ? + 71 79 76 69 72
Mentzel (2002) [42] 36 10 4 174 - - + + + + + + + + ? ? + + 71 79 73 62 69
Nakamura (2002) [43] 9 3 2 52 - - + ? + + + + - + + ? ? ? 50 64 61 58 59
Piaggio (2003) [44] 42 35 32 196 - + + ? + + + + + ? ? + ? + 64 79 64 58 56
Radmayr (2002) [44] 71 5 3 129 - - + + + + + + + + + ? + + 79 82 82 73 76
Schneider (1984) [45] 46 15 17 141 - + + + + + + + + + + ? ? ? 71 82 79 69 71
Siamplis (1996) [46] 15 4 3 154 - + + + + + + + - ? ? ? + ? 57 71 58 46 56
Valentini (2001) [47] 34 4 8 72 - - + + + + + + + - - ? ? ? 50 61 58 42 52
Uhl (2003) [48] 16 0 3 28 + + + + + + + - - + ? ? + ? 64 75 67 65 74
Von Rohden (1995) [49] 6 0 1 19 - + + + + + + + + + + ? + + 86 89 85 77 79
Median score 64 73 65 60 64
TP = true positives; FP = false positives; FN = false negatives; TN = true negatives + = yes; - = no; ? = unclear
Figure 1 Estimates of sensitivity and 1-specificity plotted in ROC space for standard and contrast enhanced ultrasound
a. Ranking of studies
The ranking of the studies using the different quality scores is summarised in Figure 2. For standard ultrasound, all scoring schemes ranked the same three studies as being the best studies, and ranked these in the same order. All scoring schemes also ranked the same study as being of the worst quality. For contrast enhanced ultrasound, scores 1, 2, 3 and 5 ranked the same two studies as being of the best quality. Score 4 ranked these two studies as having the second highest quality score. The study ranked as being the best quality study by score 4 was ranked as being of intermediate quality by the other scoring schemes. All scores ranked the same three studies as being of worst quality, with scores 1, 2, 3 and 4 ranking them in the same order. For both types of ultrasound the different scoring schemes ranked the more intermediate quality studies in different orders.
Figure 2 ranking of studies according to each different quality score
b. Difference in estimates of diagnostic accuracy between high and low quality studies
The RDOR comparing studies classified as "high" to those classified as "low" quality using each of the five scoring schemes is shown in Figure 3, separately for standard ultrasound and contrast enhanced ultrasound. For standard ultrasound, scores 1,2, and 3 gave RDORs suggesting that high quality studies produced lower estimates of diagnostic accuracy than low quality studies. In contrast, the results from schemes 4 and 5 suggested that there was no difference in estimates of the DOR between high and low quality studies. For contrast-enhanced ultrasound, scores 1, 3, 4 and 5 all classified the same set of studies as being of high and low quality. The RDORs for these quality scores suggested that high quality studies produce higher DORs than low quality studies. In contrast, scheme 2 produced an RDOR suggesting that high quality studies produce lower estimates of diagnostic accuracy than low quality studies.
Figure 3 Forest plots showing the RDOR in "high" quality studies compared to "low" quality studies for each of the five quality scoring schemes
c. Quality score as a possible source of heterogeneity
Figure 4 shows the RDORs for a 10 point increase in quality score for each of the five different quality scores, separately for standard and contrast-enhanced ultrasound. For standard ultrasound, all scoring schemes suggested that high quality studies produce lower DORs than low quality studies. For contrast-enhanced ultrasound, scores 1, 3, 4 and 5 suggested that higher quality studies produce higher DORs than lower quality studies, while score 2 suggested that they produced lower estimates. However, the confidence intervals around these estimates were wide and all included one.
Figure 4 Forest plots showing the RDOR for a 10 point increase in quality for each of the 5 quality scoring schemes
Discussion
This study has shown that using different methods of weighting individual items from the same quality assessment tool can produce different quality scores. Incorporating these quality scores into the results of a review can lead to different conclusions regarding the effect of study quality on estimates of diagnostic accuracy.
Although the ordering of studies using the different quality scores were broadly similar, there were some differences which could lead to different conclusions if they were used in a systematic review. For example, for the contrast enhanced ultrasound studies, if quality scoring scheme 4 or 5 was used then the study by Bergius and colleagues[23] would be considered to be one of the best quality studies. However, if scoring schemes 1, 2, or 3 were used then this study would be considered to be an average quality study. This suggests that quality scores should not be used as a summary indicator of quality in results tables in systematic reviews. Instead either the results of the whole quality assessment, or key components of the quality assessment, should be reported.
Stratifying studies into high and low quality studies according to quality score also varied according to the scoring scheme used. Although the confidence intervals for all comparisons were wide and all but one included one, the conclusions regarding the association of study quality and diagnostic accuracy differ according to the scoring scheme used. It is important to note that in practice a reviewer would only use one scoring scheme and so the results from the other scoring schemes would not be available to them: they would have to draw conclusions from the results for the single scoring scheme that they selected. For standard ultrasound, two of the schemes assessed produced an overall quality score that suggested no association between study quality and the diagnostic odds ratio. However, if the other three schemes were used then the conclusion would have been that high quality studies tend to produce lower estimates of diagnostic accuracy than low quality studies. Similarly for contrast-enhanced ultrasound, the conclusion for four of the scoring schemes was that high quality studies tend to produce higher estimates of diagnostic accuracy than low quality studies. In contrast, if the other scoring scheme had been used the conclusions would have been reversed. These results suggest that the use of quality scores to stratify studies into high and low quality studies should be avoided.
The inclusion of quality score as a continuous variable in the meta-regression showed fewer differences between scoring schemes. There were larger associations between quality score and the DOR for standard ultrasound than for contrast enhanced ultrasound. This would be expected as there was more heterogeneity between studies of standard ultrasound and so there was more variation that could have been explained by differences in quality. For standard ultrasound the direction of the association between study quality and test performance was the same for all scoring schemes. For contrast enhanced ultrasound the associations reported for quality scores were close to one with wide confidence intervals. This suggests very little association between quality score and diagnostic accuracy, although scoring scheme 2 again produced an association in the opposite direction to the other scoring schemes. The investigation of the association of an overall quality score with a summary effect estimate can be complicated. If no association is found between the two, this does not mean that quality does not affect the summary estimate. It may be that there is no association with any of the components of quality incorporated into the score; there may be associations with one or more components but that these have very little weight and are lost in the overall quality score; or it may be that there are association with two or more components but that these act in opposite directions cancelling each other out[7].
It is interesting to note that for the contrast enhanced ultrasound studies that it was generally scoring scheme 2 that produced different results to the other scoring schemes. All other scoring schemes scored studies that answered "unclear" to an item in the same way as studies that answered "no". Scoring scheme 2 scored these studies higher than those that answered "no". The difference between scoring scheme 2 and the other scoring schemes may therefore be related to the quality of reporting of studies: studies that were poorly reported and answered "unclear" to many of the QUADAS items would be rated higher using this scoring scheme than the other schemes.
The results of this study support the finding of Juni and colleagues that using summary scores to identify high quality studies is problematic[9]. We did not find such large differences between the different scoring schemes included in this study as Juni et al. This would be expected as we were using different methods of weighting the same quality assessment tool whereas they used different quality assessment tools, each of which not only weighted items differently but also included different items. In addition, we used only five different scoring schemes whereas Juni et al. used 25 different quality scales.
Our study was limited by the relatively few primary studies included: for standard ultrasound we included 12 studies, and for contrast-enhanced ultrasound we included 16 studies. The greater the number of studies included in a meta-analysis, the greater the power for detecting associations between study quality and estimates of diagnostic accuracy. If additional primary studies had been available, more precise estimates of the association between quality score and diagnostic accuracy would have been produced and the differences between these associations for the different scoring schemes could have been assessed in more detail. An additional limitation was the poor quality of the reporting of the studies. This resulted in a large proportion of "unclear" responses to the quality assessment.
A further limitation of this study was the lack of a gold standard against which to compare the quality scoring schemes. Lack of agreement between different scoring systems could be expected and does not necessarily invalidate all the scoring systems. The problem in this situation is determining which quality scoring scheme is the most valid. This is an inherent problem with using a quality score, and there is no reliable way of doing this.
Conclusion
This study, in the area of diagnostic systematic reviews, supports the evidence from previous work in the area of therapeutics suggesting that quality scores should not be incorporated into systematic reviews. Incorporation of the results of the quality assessment into the systematic review should involve a component approach, where the association of individual quality items with test accuracy are investigated individually, rather than using a combined quality score.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
Penny whiting contributed to the conception and design of the study, acquisition of data, analysis and interpretation of data, and drafted the manuscript. Roger Harbord and Jos Kleijnen contributed to the analysis and interpretation of data and the critical review of the manuscript for important intellectual content.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
No financial or material support was provided for this study. We would like to thank Marie Westwood for help in performing the quality assessment of the primary studies.
==== Refs
Juni P Altman DG Egger M Assessing the quality of controlled trials BMJ 2001 323 42 46 11440947 10.1136/bmj.323.7303.42
Assendelft JJ Koes BW van Tulder MW Bouter LM Scoring the quality of clinical trials [letter] JAMA 2000 283 1421 10732922 10.1001/jama.283.11.1421
ter Riet G Leffers P Zeegers M Scoring the quality of clinical trials [letter]. JAMA 2000 283 1421 10732923 10.1001/jama.283.11.1421
Berlin JA Rennie D Measuring the quality of trials: the quality of quality scales JAMA 1999 282 1083 5 10493209 10.1001/jama.282.11.1083
Juni P Egger M Scoring the quality of clinical trials [letter] JAMA 2000 283 1422 3 10732924
Klassen T Bias against quality scores 2001 2002
Greenland S Invited commentary: A critical look at some popular meta-analytic methods American Journal of Epidemiology 1994 140 290 296 8030632
Greenland S Quality scores are useless and potentially misleading American Journal of Epidemiology 1994 140 300 302
Juni P Witschi A Bloch RM Egger M The hazards of scoring the quality of clinical trials for meta-analysis JAMA 1999 282 1054 1060 10493204 10.1001/jama.282.11.1054
Seffinger MA Najm WI Mishra SI Adams A Dickerson VM Murphy LS Reinsch S Reliability of spinal palpation for diagnosis of back and neck pain: a systematic review of the literature Spine 2004 29 E413 E425 15454722 10.1097/01.brs.0000141178.98157.8e
Ejnisman B Andreoli CV Soares BG Fallopa F Peccin MS Abdalla RJ Cohen M Interventions for tears of the rotator cuff in adults Cochrane Database Syst Rev 2004 CD002758 14973989
Macdermid JC Wessel J Clinical diagnosis of carpal tunnel syndrome: a systematic review J Hand Ther 2004 17 309 319 15162113
Warren E Weatherley-Jones E Chilcott J Beverley C Systematic review and economic evaluation of a long-acting insulin analogue, insulin glargine Health Technol Assess 2004 8 1 72
Mowatt G Vale L Brazzelli M Hernandez R Murray A Scott N Fraser C McKenzie L Gemmell H Hillis G Metcalfe M Systematic review of the effectiveness and cost-effectiveness, and economic evaluation, of myocardial perfusion scintigraphy for the diagnosis and management of angina and myocardial infarction Health Technol Assess 2004 8 iii 207 15248938
Jadad AR Moore A Carroll D Jenkinson C Reynolds DJ Gavaghan DJ McQuay HJ Assessing the quality of reports of randomised clinical trials: is blinding necessary? Control Clin Trials 1996 17 1 12 8721797 10.1016/0197-2456(95)00134-4
Mulrow CD Linn WD Gaul MK Pugh JA Assessing quality of a diagnostic test evaluation J Gen Intern Med 1989 4 288 295 2760697
Whiting P Rutjes AWS Dinnes J Reitsma JB Bossuyt PM Kleijnen J A systematic review finds that diagnostic reviews fail to incorporate quality despite available tools Journal of Clinical Epidemiology 2005 58 1 12 15649665 10.1016/j.jclinepi.2004.04.008
Whiting P Rutjes AWS Dinnes J Reitsma JB Bossuyt PMM Kleijnen J Development and validation of methods for assessing the quality and reporting of diagnostic accuracy studies Health Technol Assess 2004 8 iii, 1 234
Whiting P Rutjes AWS Reitsma JB Bossuyt PM Kleijnen J The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews BMC Medical Research Methodology 2003 3 25 14606960 10.1186/1471-2288-3-25
Whiting P Rutjes AWS Reitsma JB Glas AS Bossuyt PM Kleijnen J Sources of Variation and Bias in Studies of Diagnostic Accuracy: A
Systematic Review Annals of Internal Medicine 2004 140 189 202 14757617
Whiting P Westwood M Ginnelly L Palmer S Richardson G Cooper J Watt I Glanville J Sculpher M Kleijnen J A systematic review of tests for the diagnosis and evaluation of urinary tract infection (UTI) in children under five years Health Technology Assessment 2005
Moses LE Shapiro D Littenberg B Combining independent studies of a diagnostic test into a summary ROC curve: data-analystic approaches and some additional considerations Stat Med 1993 12 1293 1316 8210827
Bergius AR Niskanen K Kekomaki M Detection of significant vesico-ureteric reflux by ultrasound in infants and children Z Kinderchir 1990 45 144 5 2197833
Baronciani D Bonora G Andreoli A Cambie M Nedbal M Dellagnola CA The Value of Ultrasound for Diagnosing the Uropathy in Children with Urinary-Tract Infections Rivista Italiana di Pediatria-Italian Journal of Pediatrics 1986 12 214 220
Dura TT Gonzalez MR Juste RM Gonzalez de DJ Carratala MF Moya BM Verdu RJ Caballero CO [Usefulness of renal scintigraphy in the assessment of the first febrile urinary infection in children] An Esp Pediatr 1997 47 378 382 9499305
Evans ED Meyer JS Harty MP Bellah RD Assessment of increase in renal pelvic size on post-void sonography as a predictor of vesicoureteral reflux Pediatr Radiol 1999 29 291 4 10199910 10.1007/s002470050591
Foresman WH Hulbert WCJ Rabinowitz R Does urinary tract ultrasonography at hospitalization for acute pyelonephritis predict vesicoureteral reflux? J Urol 2001 165 2232 4 11371951 10.1097/00005392-200106001-00004
Mage K Zoppardo P Cohen R Reinert P Ponet M [Imaging and the first urinary infection in children. Respective role of each test during the initial evaluation apropos of 122 cases] J Radiol 1989 70 279 283 2677332
Mahant S Friedman J MacArthur C Renal ultrasound findings and vesicoureteral reflux in children hospitalised with urinary tract infection Arch Dis Child 2002 86 419 420 12023172 10.1136/adc.86.6.419
Morin D Veyrac C Kotzki PO Lopez C Dalla VF Durand MF Astruc J Dumas R Comparison of ultrasound and dimercaptosuccinic acid scintigraphy changes in acute pyelonephritis Pediatr Nephrol 1999 13 219 222 10353409 10.1007/s004670050596
Muensterer OJ Comprehensive ultrasound versus voiding cysturethrography in the diagnosis of vesicoureteral reflux Eur J Pediatr 2002 161 435 437 12172827 10.1007/s00431-002-0990-0
Oostenbrink R van der Heijden AJ Moons KG Moll HA Prediction of vesico-ureteric reflux in childhood urinary tract infection: a multivariate approach Acta Paediatr 2000 89 806 10 10943962 10.1080/080352500750043693
Salih M Baltaci S Kilic S Anafarta K Beduk Y Color flow Doppler sonography in the diagnosis of vesicoureteric reflux Eur Urol 1994 26 93 7 7925538
Tan SM Chee T Tan KP Cheng HK Ooi BC Role of renal ultrasonography (RUS) and micturating cystourethrogram (MCU) in the assessment of vesico-ureteric reflux (VUR) in children and infants with urinary tract infection (UTI) Singapore Med J 1988 29 150 152 3041610
Verber IG Strudley MR Meller ST 99mTc dimercaptosuccinic acid (DMSA) scan as first investigation of urinary tract infection Arch Dis Child 1988 63 1320 1325 2849382
Alzen G Wildberger JE Muller-Leisse C Deutz FJ [Ultrasound screening of vesico-uretero-renal reflux] Klin Padiatr 1994 206 178 180 8051912
Berrocal T Gaya F Arjonilla A Lonergan GJ Vesicoureteral reflux: diagnosis and grading with echo-enhanced cystosonography versus voiding cystourethrography Radiology 2001 221 359 365 11687676
Berrocal Frutos T Gaya Moreno F Gomez Leon N Jaureguizar Monereo E Ecocistografia con contraste: una nueva modalidad de imagen para diagnosticar elreflujo vesicoureteral. [Cystosonography with echoenhancer. A new imaging technique for the diagnosis of vesicoureteral reflux] An Esp Pediatr 2000 53 422 30 11141363
Haberlik A Detection of low-grade vesicoureteral reflux in children by color Doppler imaging mode Pediatr Surg Int 1997 12 38 43 9035208
Kessler RM Altman DH Real-time sonographic detection of vesicoureteral reflux in children Am J Roentgenol 1982 138 1033 1036 6979203
McEwing RL Anderson NG Hellewell S Mitchell J Comparison of echo-enhanced ultrasound with fluoroscopic MCU for the detection of vesicoureteral reflux in neonates Pediatr Radiol 2002 32 853 858 12447589 10.1007/s00247-002-0812-6
Mentzel HJ Vogt S John U Kaiser WA Voiding urosonography with ultrasonography contrast medium in children Pediatr Nephrol 2002 17 272 276 11956881 10.1007/s00467-002-0843-0
Nakamura M Wang Y Shigeta K Shinozaki T Taniguchi N Itoh K Simultaneous voiding cystourethrography and voiding urosonography: an in vitro and in vivo study Clin Radiol 2002 57 846 849 12384112
Piaggio G gl' Innocenti ML Toma P Calevo MG Perfumo F Cystosonography and voiding cystourethrography in the diagnosis of vesicoureteral reflux Pediatr Nephrol 2003 18 18 22 12488985 10.1007/s00467-002-0974-3
Schneider K Jablonski C Wiessner M Kohn M Fendel H Screening for vesicoureteral reflux in children using real-time sonography Pediatr Radiol 1984 14 400 3 6390320
Siamplis D Vasiou K Giarmenitis S Frimas K Zavras G Fezoulidis I Sonographic detection of vesicoureteral reflux with fluid and air cystography. Comparison with VCUG Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 1996 165 166 9 8815950
Valentini AL Salvaggio E Manzoni C Rendeli C Destito C Summaria V Campioni P Marano P Contrast-enhanced gray-scale and color Doppler voiding urosonography versus voiding cystourethrography in the diagnosis and grading of vesicoureteral reflux J Clin Ultrasound 2001 29 65 71 11425090 10.1002/1097-0096(200102)29:2<65::AID-JCU1000>3.0.CO;2-I
Uhl M Kromeier J Zimmerhackl LB Darge K Simultaneous voiding cystourethrography and voiding urosonography Acta Radiol 2003 44 265 268 12751996 10.1034/j.1600-0455.2003.00065.x
Von Rohden L Bosse U Wiemann D [Reflux sonography in children with an ultrasound contrast medium in comparison to radiologic voiding cystourethrography] Paediat Prax 2004 49 49 58
|
15918898
|
PMC1184082
|
CC BY
|
2021-01-04 16:32:51
|
no
|
BMC Med Res Methodol. 2005 May 26; 5:19
|
utf-8
|
BMC Med Res Methodol
| 2,005 |
10.1186/1471-2288-5-19
|
oa_comm
|
==== Front
BMC Musculoskelet DisordBMC Musculoskeletal Disorders1471-2474BioMed Central London 1471-2474-6-281594387310.1186/1471-2474-6-28Research ArticleAgreement between diagnoses reached by clinical examination and available reference standards: a prospective study of 216 patients with lumbopelvic pain Laslett Mark [email protected] Barry [email protected] Hans [email protected] Charles N [email protected]Öberg Birgitta [email protected] Dept for Health and Society: Physiotherapy, Linköping University, Linköping, Sweden2 Institute of Information and Mathematical Sciences, Massey University, Albany, New Zealand3 Dept for Health and Society, Linköping University, SE-58183 Linköping, Sweden4 Louisiana State University Health Science Center, 2718 Cadiz St, New Orleans, LA 70115, USA5 SwedenDept for Health and Society, Linköping University, SE-58183 Linköping, Sweden2005 9 6 2005 6 28 28 9 12 2004 9 6 2005 Copyright © 2005 Laslett et al; licensee BioMed Central Ltd.2005Laslett et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The tissue origin of low back pain (LBP) or referred lower extremity symptoms (LES) may be identified in about 70% of cases using advanced imaging, discography and facet or sacroiliac joint blocks. These techniques are invasive and availability varies. A clinical examination is non-invasive and widely available but its validity is questioned. Diagnostic studies usually examine single tests in relation to single reference standards, yet in clinical practice, clinicians use multiple tests and select from a range of possible diagnoses. There is a need for studies that evaluate the diagnostic performance of clinical diagnoses against available reference standards.
Methods
We compared blinded clinical diagnoses with diagnoses based on available reference standards for known causes of LBP or LES such as discography, facet, sacroiliac or hip joint blocks, epidurals injections, advanced imaging studies or any combination of these tests. A prospective, blinded validity design was employed. Physiotherapists examined consecutive patients with chronic lumbopelvic pain and/or referred LES scheduled to receive the reference standard examinations. When diagnoses were in complete agreement regardless of complexity, "exact" agreement was recorded. When the clinical diagnosis was included within the reference standard diagnoses, "clinical agreement" was recorded. The proportional chance criterion (PCC) statistic was used to estimate agreement on multiple diagnostic possibilities because it accounts for the prevalence of individual categories in the sample. The kappa statistic was used to estimate agreement on six pathoanatomic diagnoses.
Results
In a sample of chronic LBP patients (n = 216) with high levels of disability and distress, 67% received a patho-anatomic diagnosis based on available reference standards, and 10% had more than one tissue origin of pain identified. For 27 diagnostic categories and combinations, chance clinical agreement (PCC) was estimated at 13%. "Exact" agreement between clinical and reference standard diagnoses was 32% and "clinical agreement" 51%. For six pathoanatomic categories (disc, facet joint, sacroiliac joint, hip joint, nerve root and spinal stenosis), PCC was 33% with actual agreement 56%. There was no overlap of 95% confidence intervals on any comparison. Diagnostic agreement on the six most common patho-anatomic categories produced a kappa of 0.31.
Conclusion
Clinical diagnoses agree with reference standards diagnoses more often than chance. Using available reference standards, most patients can have a tissue source of pain identified.
==== Body
Background
Different pathoanatomic conditions and mechanisms in the lumbar spine and pelvis region may produce low back pain (LBP), pelvic or lower extremity symptoms. The most frequent tissue sources of lumbopelvic and referred pain are the intervertebral discs, zygapophysial (facet) and sacroiliac joints [1]. Nerve root irritation, spinal stenosis, the hip joint, fractures, neoplasms and disorders of the vascular system or viscera are potential but less common sources of lumbopelvic or lower extremity pain. Psychosocial distress increases the complexity of diagnosis, confounds therapeutic endeavours and is a major factor determining disability [2,3]. During the last 20 years, advances in technology and clinical science have resulted in techniques that have improved our ability to identify the tissue origin of LBP and referred lower extremity symptoms. While controversy surrounds techniques such as provocation discography for diagnosis of discogenic pain [4-6], the value of diagnostic anaesthetic blocks to the lumbar zygapophysial joints (ZJ) and sacroiliac joints (SIJ) is more secure. A body of evidence supports the use of these diagnostic procedures in specific clinical circumstances using established methodological guidelines [7-13]. While it is commonly stated that no pathoanatomic explanation for symptoms is possible for about 80% of cases [2,14], some now argue that the use of recently improved diagnostic and clinical reasoning techniques makes diagnosis of the tissue origin of pain possible in 46–75% of cases [7]. Because these procedures are invasive, they cannot be justified in acute or subacute cases, since prognosis in these cases is good [7]. However there is value in such a diagnosis for chronic LBP. Between 5 and 10% of patients initially visiting a primary care physician for LBP will ultimately develop chronic LBP [15]. These patients have a high level of dissatisfaction with primary care management [16] yet continue to desire a diagnosis and explanation for persistent pain and disability [17]. Persistent discogenic pain may be treated with spinal fusion or intradiscal electrothermal annuloplasty (IDET), ZJ pain with intra-articular injections or medial branch neurotomy, and SIJ pain with intra-articular injections or surgical fusion. These procedures are invasive, and are not universally successful in returning the patient to pain-free full function. While psychosocial distress undoubtedly plays an important role in these less than ideal outcomes, poor pre-procedure selection and diagnosis, is a major contributing factor to failure [7].
The clinical examination (history and physical examination with or without imaging) is the basis upon which management rests. The diagnostic value of this examination is debatable for all but a minority of cases. A few symptomatic patho-anatomic entities such as disc herniations causing nerve root compression, symptomatic spinal stenosis, radiologically demonstrated fractures and neoplasms may be identified using the clinical examination in combination with advanced imaging studies. The current project was conceived to compare diagnoses derived from a detailed clinical examination by a physiotherapist, with expert diagnoses obtained using available reference standards for diagnosis of discogenic, facetogenic, SIJ, hip joint, nerve root pain and symptomatic spinal stenosis.
Methods
A prospective, blinded, reference standard-related design was utilized. The Louisiana Institutional Review Board approved the study and all included patients signed an informed consent form. A physiotherapist with 30 years experience as a manipulative therapist attended a specialist spinal diagnostic clinic in Louisiana, for blocks of 4–8 weeks between May 2001 and October 2002 and examined consecutive chronic LBP patients during these periods. Clinical examinations required between 30 and 60 minutes and were carried out immediately before the reference standard diagnostic tests. A radiologist with 20 years experience in fluoroscopically guided diagnostic injections and interpretation of advanced imaging techniques attempted to identify the tissue origin of chronic LBP, based on imaging and responses to diagnostic injections. These diagnoses were the reference standards against which diagnoses arrived at by the clinical (physiotherapy) examination were contrasted. Another therapist with 17 years clinical experience carried out examinations of 13 patients. Figure 1 presents a flow diagram describing patient recruitment, summary of examiner's diagnoses and reference standards employed.
Figure 1 Summary of patient recruitment, diagnoses by examiners and reference standard procedures used.
The great majority of possible causes of LBP are uncommon, even rare. Appendix 1 presents those painful clinical entities believed to be most common and the procedures used as diagnostic reference standards [see additional file 4].
In addition, three other categories were available to the examining clinicians – "illness behaviour" where the patient's behaviour and responses to questioning and examination suggested psychosocial distress of some kind, "others" (for the more uncommon causes of pain) and an "indeterminate" category where no conclusion could be reached.
At presentation, clinic staff collected medical history, demographic and questionnaire data. If informed consent was obtained, the physiotherapist examined the patient and the patient received scheduled diagnostic injection procedures in sequence.
Blinding
The physiotherapy examinations were conducted blind to the results of disability and self report questionnaires, the results of previous imaging studies and diagnostic injections. The physician was blinded to the results of the physiotherapy examination and diagnostic conclusions.
Diagnostic classification
The clinical reasoning by which the physiotherapist reached a diagnosis has been presented elsewhere in detail [18,19]. Very briefly, discogenic pain was concluded when centralisation, peripheralisation [20-22] or directional preference [23] were reported by the patient during an examination with repeated standardised end range test movements [24], or if the dominant or primary pain was located in the exact midline of the lumbar spine. ZJ pain was recorded if the clinical criteria specified by Revel et al (1998) [25] were satisfied in the absence of centralisation. SIJ pain was recorded if three or more stress SIJ tests [26] provoked familiar pain in the absence of centralization [27,28]. Nerve root pain was recorded when referred pain was provoked with nerve tension tests. Symptomatic spinal stenosis was recorded when the patient reported a clear pattern of intermittent claudication which was relieved by sitting or a flexed spinal posture [29]. Hip joint pain was recorded if passive movements of the hip provoked familiar pain more readily than SIJ provocation or lumbar tests [30,31]. Diagnosis of instability presented a problem since no reference standard exists [7,8,32,33]. The radiologist's diagnosis was based on the observation of paradoxical motion on flexion / extension radiographs [34,35] and the physiotherapy clinical diagnosis was based on clinical criteria gleaned from post graduate course material and published opinion [32,36,37].
Some data on the diagnostic accuracy of the clinical tests used in the physiotherapy clinical examination was available prior to commencement of the study. A summary for key tests is presented in Appendix 2 [see additional file 5].
The physiotherapist and physician recorded their diagnoses on standardised forms. In January 2003 a meeting of researchers and experienced clinicians was convened at the Auckland University of Technology to recommend the method by which the diagnosis data were to be entered into an electronic database for analysis.
Data analysis
Data were entered and stored in an electronic database (Minitab version 14.12 © 2001). Two variables were constructed to record agreement between physiotherapy and reference standard diagnoses. "Exact agreement" was recorded when the two diagnoses were exactly the same in all respects, including multiple diagnoses. "Clinical agreement" was recorded when the physical therapy diagnosis was included within the reference standard diagnosis (e.g. if the physiotherapy diagnosis was ZJ pain and the reference standard diagnosis was ZJ pain and symptomatic spinal stenosis). All cases of 'exact' agreement are included within 'clinical' agreement. Confidence intervals (CI) for proportions were calculated using recommended methods [38]. Agreement between physiotherapy diagnoses and reference standard diagnoses was estimated using the kappa statistic, which accounts for chance agreements. Kappa values of zero reflect chance agreement, values less than 0.0 reflect agreement worse than chance with -1.0 representing perfect disagreement. Values greater than 0.0 reflect agreement better than chance agreement with +1.0 representing perfect agreement [39]. The kappa statistic for multiple categories was calculated using Confidence Interval Analysis Software © T Bryant 2000 [40]. Chance agreement is strongly influenced by the prevalence of the diagnostic category. The more prevalent the disorder, the more likely the physiotherapist would correctly match the reference standard diagnosis on the basis of chance. To estimate agreement while accounting for the influence of disease prevalence and the number of possible diagnoses, the rule-of-thumb "proportional chance criterion" (PCC) was used. The PCC is commonly used in discriminant analysis to judge whether a classification method is better than guessing. The PCC is the expected proportion of correct classifications [41], and equals the sum of squared prevalences. Standard errors (SE) for PCC and agreement were calculated as (sqr(p(1-p)/n)) and 95% confidence intervals calculated as: proportion ± (1.96 * SE) [42].
To simply the evaluation of agreements on patho-anatomic diagnoses, both physiotherapy and reference diagnoses were condensed to six categories: disk, facet (z joint), sacroiliac joint, nerve root, hip joint and spinal stenosis. Where more than one diagnosis was provided, the case was included in each diagnostic category. For example: If the physiotherapist diagnosis was: disk and nerve root, the counts for both disk pain and nerve root pain were incremented by one. If the reference standard diagnosis was disc, nerve root and spinal stenosis, each of these diagnostic categories were incremented by one. Consequently there were many more diagnoses than patients.
Results
During the study period, 296 patients were invited to participate in the project and 78 were excluded. Reasons for exclusion were: 53 declined to participate, ten had no injection or other procedure carried out because of insufficient pain on the day of examination, nine were excluded because of time constraints, three patients were deemed incompetent to understand study procedure, blinding was compromised in two cases, and contradictory diagnostic reporting of diagnostic injection results occurred in one case. Table 1 presents demographic, historical and profile data for included patients (n = 216) [see additional file 1]. Figure 1 presents patient recruitment patterns in the project.
Table 2 presents a cross-tabulation of physiotherapist diagnoses with diagnoses reached by the interventional radiologist [see additional file 2]. The radiologist came to a single diagnostic conclusion in 144 cases (66%) and more than one in 72 (34%) cases, with two cases having three diagnoses. The physiotherapist reached a single diagnostic conclusion in 163 (76%) of cases and two conclusions in 53 cases. Based on reference standard / expert opinion diagnoses, the chance of the physiotherapist correctly guessing the diagnosis (PCC) was 13% with 95% confidence intervals of 9% and 18% (standard error = 0.023). Exact agreement (standard error, 95% CI) achieved was 32% (0.032, 26%, 38%) and clinical agreement 51% (0.034, 45%, 58%).
Many of the diagnostic categories contained only one case, so diagnoses were grouped under nine general labels: disc, Z-joint, sacroiliac joint, nerve root, hip joint, spinal stenosis, "other", "illness behaviour" and "Indeterminate". A total of 368 diagnoses were provided by physiotherapy and reference standard clinicians through multiple diagnoses. "Illness behaviour" was the sole diagnostic conclusion or included in 79 (21.5%) of 368 conclusions by the reference standard clinician. The physiotherapy examiner used this description for 76 cases. "Indeterminate" was included in 84 cases by the reference standard clinician and in 91 cases by the physiotherapist. No reference standard for illness behaviour was established a priori in this study and "indeterminate" is the absence of a diagnosis. The primary objective was to evaluate agreement on patho-anatomic diagnoses. To evaluate agreement on diagnosis for the six main patho-anatomic diagnoses: discogenic pain, Z-joint pain, sacroiliac joint pain, nerve root (radicular) pain, hip joint pain and spinal stenosis, Table 3 was truncated (represented by the dashed lines), by removing the columns and rows for the other categories [see additional file 3]. The category "Other" was removed from this analysis as it contained some uncommon pathologies: a rhabdomyosarcoma affecting the psoas and hip, a symptomatic spondylolysis, bone graft donor site pain, nerve root irritation from surgical hardware, and vascular claudication secondary to peripheral artery disease. After removal of the columns and rows representing non-patho-anatomic diagnoses and 'Others", 137 patho-anatomic physiotherapy and reference standard diagnoses. The chance that the physiotherapist would correctly guess the diagnostic category (PCC, standard error, 95% CI) was 33%, 0.039, 26%, 41%). Agreement achieved (standard error, 95%CI) was 57% (0.041, 48%, 64%). Kappa statistic (standard error, 95% CI) for this table is 0.31 (0.067, 0.18, 0.44).
Discussion
To our knowledge, this is the first study to estimate agreement between diagnoses based on blinded clinical (physiotherapy) examinations, and a range of diagnoses using available reference standards and other classification categories, in patients presenting with pain presumably of lumbopelvic origin. Overall agreement was better than could be expected on the basis of chance and ranged from 32% to 57% depending on agreement criteria and complexity of the diagnostic categorization analyzed. The proportion of patients indeterminate to reference standard diagnostic methods was 23%. However, the categories "illness behaviour", "indeterminate" and "instability" are non-pathoanatomic i.e. 'non-specific' in origin. There were 73 cases (34%) that the reference standard clinician classified using only these descriptions, and may be considered patho-anatomically 'non-specific'. This figure corresponds with claims that 46–75% of LBP patients have an identifiable tissue origin of pain using a reductionist approach to diagnosis [7].
The kappa statistic measures strength of agreement between examiners discounting chance agreement. The achieved level of 0.31 is considered 'fair agreement' [43]. However the context of this project is important. The patients in this sample were referred for invasive diagnostic testing and were typically chronic with high levels of distress and disability. Most patients had failed multiple attempts at treatment, and many had seen a number of general and specialist clinicians without a satisfactory diagnosis being provided. This was anticipated prior to commencement of the data collection phase of the project, and it was accepted that psychosocial distress would impact on the ability of physiotherapy and reference standard clinicians to make a tissue specific diagnosis. Some 87 patients (40%) had "illness behaviour" or "indeterminate" as the sole description, or as a part of the diagnostic classification provided by the reference standard clinician. The single greatest cause of "indeterminacy" in reference standard classifications was "illness behaviour". A similar pattern emerged for the physiotherapy clinical classifications (Tables 2 and 3). Indeterminacy and illness behaviour combined, accounted for as many patients as the largest pathoanatomic diagnostic group (disc). There was no reference standard for "illness behaviour" established prospectively, so any agreement between the physiotherapist and physician is merely interesting. Cases that were diagnostically indeterminate without apparent "illness behaviours" were also a large group, constituting 28/216 and 32/216 for the interventional radiologist and physiotherapist respectively. This is a result that might be expected in a tertiary referral environment, but was due in part because a proportion of patients could not receive the full battery of interventional and clinical tests needed to arrive at a patho-anatomic diagnosis. Restrictive terms of referral and low patient tolerance to discomfort were the primary reasons for failure of patients to receive all appropriate tests. Time constraints limited the physiotherapy examination procedure in less than 5% of cases included in the analysis. Cost containment was not a factor limiting reference standard or physiotherapy examination. The physiotherapy examination was provided free and any additional interventional procedures over and above those recommended on referral were also provided free.
The data provides information about agreement on diagnosis between non-invasive and inexpensive clinical methods of diagnosis carried out by a physiotherapist and a radiologist using invasive and expensive diagnostic technology, over and above chance. Across the whole spectrum of possible reasons for patients attending a tertiary referral diagnostic clinic estimated exact agreement is 19% (32% versus 13%) over and above chance. In practical terms, exact agreement is not required or expected when examining complex patients. 'Clinical agreement' is a more appropriate measure in that it is a measure of the degree to which the physiotherapist can reach accord with at least one part of the diagnosis. Clinical agreement is also estimated to be 19% better than chance agreement. For patho-anatomic categories, agreement over chance is 24% (57% versus 33%). The greater agreement on the six patho-anatomic categories may be a result of clearer diagnostic criteria for categorization by both clinicians. Some discordance between the physiotherapy and reference standard diagnoses resulted from one examiner being able to reach a diagnosis where the other was unable to carry out the examination(s) needed. Restrictions inherent in some referrals, meant that with some cases, the physiotherapist was able to reach a diagnosis, but the radiologist could not use the appropriate procedure necessary to provide a diagnosis for comparison. Conversely, some patients could not tolerate some parts of the physiotherapy clinical examination, whereas a clear diagnostic result was possible using interventional diagnostic procedures. These cases of unilateral indeterminacy resulted in disagreements, whereas agreement may have been possible if both clinicians could have fully examined the patients.
It was anticipated was that discogenic pain cases would be more numerous than other tissue sources of pain with 85 (39%) receiving this diagnosis and 59 (27%) having this as the sole diagnosis. This proportion is in concord with previous results [44]. Fewer ZJ and SIJ cases were identified than expected with significant consequences. Estimates of agreement between the physiotherapy and reference standard diagnoses for the less frequent diagnoses, especially SIJ pain, were compromised. The numbers for these less frequently occurring conditions are too low to enable conclusions to be drawn from the results. The identification of hip joint pain among patients referred as LBP sufferers can be expected [45,46], but it was not anticipated that they would be more prevalent than SIJ cases (Table 2).
In this sample 22 cases (10%) received two patho-anatomic diagnoses (by available reference standards), and two had three identified pain generators (Table 2). To our knowledge only one other study has reported multiple sources of nociceptive input to low back pain with about 3% having co-existent disc and ZJ pain [47]. In the current study, only two patients had discogenic and facetogenic pain (1%).
An issue for this study concerns generalizability. The patients were chronic and distressed and nearly 30% of included patients had a history of lumbar spinal surgery with persistent or recurrent pain. Physiotherapists do see patients in all stages and degrees of LBP and our results are arguably generalizable to more chronic patient populations. We did not exclude patients with a history of spinal surgery, so this study may be more representative of tertiary care patient samples than other back pain studies. Generalizability to acute and subacute LBP populations is not appropriate, but at the conceptual level at least, it may be argued that acute and subacute patients would be less complex than patients in this study. Consequently our results may represent the lower bound of potential agreement between physiotherapy diagnostic conclusions and reference standard diagnoses.
Another issue involving generalizability concerns the examining physiotherapists. Are the diagnostic techniques and procedures used, representative of a special interest "craft" group and not within the normal domain of practicing clinicians? Physiotherapists have a wide range of special interests generally and there are many schools of thought within musculoskeletal practice. In the last 15 years, an 'extended practice' role for physiotherapists in orthopaedic musculoskeletal practice has emerged [48] requiring more advanced training than is required to conduct the examination used in this study [49-51]. In this study the examination techniques of concern are the McKenzie repeated movement's examination, the provocation SIJ tests and the tests used by Revel et al (1998). The McKenzie system has been in the public domain since 1981 [20] and is the most widely used system among therapists in North America for examining and treating LBP patients [52]. It has been formally studied for inter-examiner reliability [53] with satisfactory results among trained clinicians [54,55] and for validity [22,23,56]. The provocation SIJ tests have been in the public domain for many decades [30] and the principal author has examined them for reliability [26] and validity [28] with satisfactory results. The criteria used to identify patients suitable for screening ZJ blocks ("Revel's criteria") are simple and well documented [25], although recently these authors' results have not been replicated [57,58]. The feasibility, reliability and construct validity of the clinical reasoning and classification system has been evaluated [19,59]. It is our contention that the system may be generalized to a proportion of clinicians, or can be learned readily enough at post-graduate level.
The use of discography, ZJ and SIJ blocks as reference standards may be criticized on the basis that false positives may compromise their status as "gold standards" [60,61]. It is recognized that perfect gold standards do not exist for the diagnosis of discogenic, facetogenic pain or pain arising from the SIJ [62]. However, modern methodology accounts for these potential errors and we believe that no alternative satisfactory standards exist [7].
Implications for clinical practice and future research
A comprehensive history and physical examination similar to that utilized in the current project has some potential to predict the diagnostic conclusions reached by an experienced physician using sophisticated and invasive technologies. Most of the agreement calculated was made up from identifying pain originating from the lumbar discs and the hip joints. Previous research has shown that pain arising from the ZJ joints cannot be characterized by clinical examination variables [57,58,63], and the results of this study support this conclusion. Although the current data does not permit a direct test of the validity of the clinical examination in relation to pain known to arise from SIJs, recent previous work has indicated that the clinical reasoning process and the examination techniques have some validity and clinical utility [28].
The diagnosis of instability is fraught with conceptual and terminological difficulties without widely accepted clinical diagnostic criteria [8,32,64]. In this study some historical cues and clinical findings suggestive of this condition [36] were documented and some results will be reported (Laslett, Oberg et al Submitted October 2004)
Nerve root pain was the second most common tissue-based diagnosis made by both physiotherapist and radiologist after discogenic pain. Although conceptually and clinically distinct, radiculopathy (clinical evidence of sensory and/or motor deficit) and radicular pain (lancinating pain in a myotomal distribution secondary to axonal stimulation rather than nociceptive stimulation) were combined in the diagnosis of "nerve root pain". It was differentiated from somatic referred pain believed to result from central nervous system convergence [8]. In this study, pain arising from the lumbar nerve roots was not consistently diagnosed using the examination procedures. Two factors may account for this inconsistency; a) the reference standard for diagnosis (epidural blockade) is typically used as a therapy but is less favoured as a diagnostic tool [8,65]. In the context of this study the type of epidural injection (caudal, translaminar or transforaminal) and response criteria were not standardized. b) The criteria for clinical diagnosis were rather loose and consisted of dominant pain in the lower extremity aggravated by either the straight-leg-raise or femoral nerve tension tests. Further studies are required to improve the precision and clarity of the criteria for clinically classifying nerve root pain. Definitional distinctions between radiculopathy, radicular pain and somatic referred pain are clear, but it is unknown whether recommended clinical criteria can reliably distinguish between these concepts. This study did not attempt to explore these issues.
On the surface, agreement does appear to be weak even though better than chance. However, this does not concern us greatly. This is a pragmatic report of the overall performance of a low tech clinical examination and diagnoses achieved, compared to highly sophisticated and predominantly invasive procedures with a complex group of patients. The high proportion of cases deemed to be diagnostically indeterminate or displaying confounding illness behaviours (in the opinion of the clinicians) attests to the complexity. The 50% agreement on patho-anatomic categories achieved in this study was about what was hoped for prior to commencement of these studies. As a consequence of this work, certain significant modifications to the clinical examination can be made and subsequent studies may demonstrate an improvement on what we achieved here. 'Exact agreement' is a very demanding requirement when multiple pathologies are present. 'Clinical agreement' as described in the paper is really all that can be expected of a low tech clinical examination of patients with low back pain – a symptom commonly described as 'non-specific' in the low back pain literature.
Conclusion
Using available reference standard technique, two thirds of patients received a pathoanatomic diagnosis with multiple pain generators identified in 10% of cases. Diagnoses of the tissue origin of chronic LBP or referred lower extremity symptoms by experienced physiotherapy clinicians agreed with available reference standard diagnoses 19–24% over and above expected chance agreement.
Competing interests
The authors declare that they have no competing interests. All authors declare that their contributions to this paper have been independent of influence from funding agencies
Authors' contributions
Concept and research design provided by ML, CA, BÖ and HT. Project management provided by ML. Facilities and equipment provided by CA. Writing provided by ML, BÖ and BM with manuscript review by HT. Data analysis, statistical support and manuscript review provided by BM. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
Flow diagram of patient recruitment patterns in the project.
Click here for file
Additional File 2
Table 1. Demographic, medical and psychometric profile of chronic low back pain patients.
Click here for file
Additional File 3
Table 2. Cross tabulation for physiotherapist and reference standard / expert opinion diagnoses
Click here for file
Additional File 4
Cross-tabulation of reference standard / expert opinion and physiotherapy diagnostic groups
Click here for file
Additional File 5
Appendix 1. Description of reference standards used in pathoanatomic diagnoses in low back pain
Click here for file
Acknowledgements
This project was supported by grants from:
1. The International Spine Injection Society
2. The New Zealand Society of Physiotherapists Scholarship Fund
3. The New Zealand Manipulative Physiotherapists Education Trust Fund.
Thanks to Professor Peter McNair and Maynard Williams of Auckland University of Technology, David Coombes PT and Paula van Wijmen PT for advice on data handling and checking. Thanks to Sharon B Young PT for examining 13 patients.
==== Refs
Bogduk N The anatomical basis for spinal pain syndromes J Manipulative Physiol Ther 1995 18 603 605 8775022
Spitzer WO Scientific approach to the assessment and management of activity-related spinal disorders. A monograph for clinicians. Report of the Quebec Task Force on Spinal Disorders Spine 1987 12
Waddell G The back pain revolution 1998 Edinburgh: Churchill Livingstone
Walsh TR Weinstein JN Spratt KF Lehmann TR Aprill CN Sayreh H Lumbar discography in normal subjects J Bone Joint Surg (Am) 1990 72 1081 1088 2384508
Carragee EJ Tanner CM Khurana S Hayward C Welsh J Date E Truong T Rossi M Hagle C The rates of false-positive lumbar discography in select patients without low back symptoms Spine 2000 25 1373 1381 10828919 10.1097/00007632-200006010-00009
Bogduk N An analysis of the Carragee data on false-positive discography International Spinal Injection Society Scientific Newsletter 2001 4 3 10
Bogduk N McGuirk B Medical management of acute and chronic low back pain Pain research and clinical management 13
Merskey H Bogduk N Classification of chronic pain: descriptions of chronic pain syndromes and definitions of pain terms 1994 Seattle: IASP Press
Guyer RD Ohnmeiss DD Contemporary Concepts in spine care. Lumbar discography Spine 1995 20 2048 2059 8578384
Bogduk N Modic MT Controversy. Lumbar discography Spine 1996 21 402 404 8742222
Dreyfuss PH Dreyer SJ Vaccaro A Lumbar zygapophysial (facet) joint injections The Spine Journal 2003 3 50 59 10.1016/S1529-9430(02)00450-3
Fortin JD Dwyer AP West S Pier J Sacroiliac Joint: Pain referral maps upon applying a new injection/arthrography technique. Part 1: Asymptomatic volunteers Spine 1994 19 1475 1482 7939978
Fortin JD Aprill C Pontieux RT Pier J Sacroiliac Joint: Pain referral maps upon applying a new injection/arthrography technique. Part II: Clinical evaluation Spine 1994 19 1483 1489 7939979
Pruitt SD Von Korff M Turk DC, Gatchel RJ Improving the management of low back pain: a paradigm shift for primary care Psychological approaches to pain management: a practitioner's handbook 2002 2 New York: The Guilford Press 301 316
Andersson GBJ Pope MH Snook S Occupational low back pain: Assessment, treatment and prevention 1991 St Louis, MO: Mosby Year Book
Cherkin DC Deyo RA Berg AO Evaluation of a physician education intervention to improve primary care for low back pain II Spine 1991 16 1173 1178 1836677
McPhillips-Tangum CA Cherkin DC Rhodes LA Markham C Reasons for repeated medical visits among patients with chronic back pain J Gen Intern Med 1998 13 289 295 9613883 10.1046/j.1525-1497.1998.00093.x
Laslett M Van Wijmen P Low back and referred pain:diagnosis and a proposed new system of classification NZ Journal of Physiotherapy 1999 27 5 14
Petersen T Laslett M Thorsen H Manniche C Ekdahl C Jacobsen S Diagnostic classification of non-specific low back pain. A new system integrating patho-anatomic and clinical categories Physiotherapy Theory and Practice 2003 19 213 237
McKenzie RA The Lumbar Spine: Mechanical Diagnosis and Therapy 1981 Waikanae: Spinal Publications Ltd
Donelson R Silva G Murphy K Centralisation phenomenon – Its usefulness in evaluating and treating referred pain Spine 1990 15 211 213 2141186
Donelson R Aprill C Medcalf R Grant W A prospective study of centralization of lumbar and referred pain. A predictor of symptomatic discs and anular competence Spine 1997 22 1115 1122 9160470 10.1097/00007632-199705150-00011
Donelson R Grant W Kamps C Medcalf R Pain response to sagittal end range spinal motion: A multi-centered, prospective, randomized trial Spine 1991 16 S206 S212 1830700
McKenzie RA May S Mechanical diagnosis and Therapy: The lumbar spine 2002 Waikanae, New Zealand: Spinal Publication Ltd
Revel M Poiraudeau S Auleley GR Payan C Denke A Nguyen M Chevrot A Fermanian J Capacity of the clinical picture to characterize low back pain relieved by facet joint anesthesia. Proposed criteria to identify patients with painful facet joints Spine 1998 23 1972 1977 9779530 10.1097/00007632-199809150-00011
Laslett M Williams M The reliability of selected pain provocation tests for sacroiliac joint pathology Spine 1994 19 1243 1249 8073316
Young SB Aprill CN Laslett M Correlation of clinical examination characteristics with three sources of chronic low back pain The Spine Journal 2003 3 460 465 14609690 10.1016/S1529-9430(03)00151-7
Laslett M Young SB Aprill CN McDonald B Diagnosing painful sacroiliac joints: a validity study of a McKenzie evaluation and sacroiliac joint provocation tests Aust J Physiother 2003 49 89 97 12775204
Fritz JM Delitto A Welch WC Erhard RE Lumbar spinal stenosis: A review of current concepts in evaluation, management, and outcome measurements Arch Phys Med Rehab 1998 79 700 708 10.1016/S0003-9993(98)90048-X
Cyriax J Textbook of Orthopaedic Medicine Diagnosis of Soft Tissue Lesions 1975 one London: Balliere Tindall
Ombregt L Bisschop P ter Veer HJ Van de Velde T A system of orthopaedic medicine 1995 London: WB Saunders Company Ltd
Hicks GE Fritz JM Delitto A Mishock L Interrater reliability of clinical examination measures for identification of lumbar segmental instability Arch Phys Med Rehabil 2003 84 1858 1864 14669195 10.1016/S0003-9993(03)00365-4
Nachemson A Instability of the lumbar spine. Pathology, treatment, and clinical evaluation Neurosurg Clin N Am 1991 2 785 790 1821757
Pitkanen MT Manninen HI Lindgren KA Sihvonen TA Airaksinen O Soimakallio S Segmental lumbar spine instability at flexion-extension radiography can be predicted by conventional radiography Clin Radiol 2002 57 632 639 12096864 10.1053/crad.2001.0899
Ehara S Shimamura T Paradoxical motion in spondylolisthesis due to two-segment instability Arch Orthop Trauma Surg 1997 116 435 436 9266060
O'Sullivan PB Lumbar segmental 'instability': clinical presentation and specific stabilizing exercise management Manual Therapy 2000 5 2 12 10688954 10.1054/math.1999.0213
Richardson C Jull G Hodges P Hides J Therapeutic exercise for spinal segmental stabilization in low back pain Scientific basis and clinical approach 1999 London: Churchill Livingstone
Altman DG Machin D Bryant TN Gardner MJ Statistics with confidence 2000 Bristol: British Medical Journal
Landis RJ Koch GG The measurement of observer agreement for categorical data Biometrics 1977 33 159 174 843571
Bryant TN Confidence Intervals Analysis 20 Build 41 2000 Bristol, BMJ Books
Huberty CJ Issues in the use and interpretation of discriminant analysis Psychological Bulletin 1984 95 156 171 10.1037//0033-2909.95.1.156
Freund JE Modern elementary statistics 1988 London: Prentice-Hall International
Sackett DL Haynes RB Guyatt GH Tugwell P Clinical epidemiology: a basic science for clinical medicine 1991 Boston: Little, Brown & Company
Schwarzer AC Aprill CN Derby R Fortin J Kine G Bogduk N The relative contributions of the disc and zygapophyseal joint in chronic low back pain Spine 1994 19 801 806 8202798
Schwarzer AC Aprill C Derby R Fortin JD Kine G Bogduk N The prevalence and clinical features of internal disc disruption in patients with chronic low back pain Spine 1995 20 1878 1883 8560335
Offierski CM Macnab I Hip-Spine Syndrome Spine 1983 8 316 321 6623198
Fogel GR Esses SI Hip Spine syndrome: management of coexisting radiculopathy and arthritis of the lower extremity The Spine Journal 2003 3 238 241 14589205 10.1016/S1529-9430(02)00453-9
Hourigan PG Weatherley CR Initial assessment and follow-up by a physiotherapist of patients with back pain referred to a spinal clinic J R Soc Med 1994 87 213 214 8182677
Hourigan PG Weatherley CR The physiotherapist as an orthopaedic assistant in a back pain clinic Physiotherapy 1995 81 546 548
Hockin J Bannister G The extended role of a physiotherapist in an out-patient orthopaedic clinic Physiotherapy 1994 80 281 284
Daker-White G Carr AJ Harvey I Woolhead G Bannister G Nelson I Kammerling M A randomised controlled trial. Shifting boundaries of doctors and physiotherapists in orthopaedic outpatient departments J Epidemiol Community Health 1999 53 643 650 10616677
Battie MC Cherkin DC Dunn R Clol MA Wheeler KJ Managing Low Back Pain : Attitudes and Treatment Preferences of Physical Therapists Phys Ther 1994 74 219 226 8115455
Razmjou H Kramer JF Yamada R Inter-tester reliability of the McKenzie evaluation in mechanical low back pain J Orthop Sports Phys Ther 2000 30 368 383 10907894
Kilpikoski S Airaksinen O Kankaanpaa M Leminen P Videman T Alen M Interexaminer reliability of low back pain assessment using the McKenzie method Spine 2002 27 E207 E214 11935120 10.1097/00007632-200204150-00016
Aina A May S Clare H The centralization phenomenon of spinal symptoms – a systematic review Manual Therapy 2004 9 134 143 15245707 10.1016/j.math.2004.03.004
Laslett M Oberg B Aprill CN McDonald B Centralization as a predictor of provocation discography results in chronic low back pain, and the influence of disability and distress on diagnostic power The Spine Journal
Manchikanti L Pampati V Fellows B Baha GA The inability of the clinical picture to characterize pain from facet joints Pain Physician 2000 3 158 166 16906195
Laslett M Oberg B Aprill CN McDonald B Zygapophysial joint blocks in chronic low back pain: A test of Revel's model as a screening test BMC Musculoskeletal Disorders 2004 5 43 15546487 10.1186/1471-2474-5-43
Petersen T Olsen S Laslett M Thorsen H Manniche C Ekdahl C Jacobsen S Inter-tester reliability of a new diagnostic classification system for patients with non-specific low back pain Aust J Physio 2004 50 85 91
Carragee EJ Tanner CM Yang B Brito JL Truong T False-positive findings on lumbar discography. Reliability of subjective concordance assessment during provocative disc injection Spine 1999 24 2542 2547 10626318 10.1097/00007632-199912010-00017
Schwarzer AC Aprill CN Derby R Fortin J Kine G Bogduk N The false-positive rate of uncontrolled diagnostic blocks of the lumbar zygapophysial joints Pain 1994 58 195 200 7816487 10.1016/0304-3959(94)90199-6
Saal JS General principles of diagnostic testing as related to painful lumbar spine disorders: a critical appraisal of current diagnostic techniques Spine 2002 27 2538 2545 12435989 10.1097/00007632-200211150-00027
Schwarzer AC Aprill C Derby R Fortin JD Kine G Bogduk N Clinical features of patients with pain stemming from the lumbar zygapophysial joints. Is the lumbar facet syndrome a clinical entity? Spine 1994 15 1132 1137 8059268
Bogduk N Clinical anatomy of the lumbar spine and sacrum 1997 London: Churchill Livingstone
Weinstein SM Herring SA Derby R Contemporary concepts in spine care. Epidural steriod injections Spine 1995 20 1842 1846 7502144
Roland M Morris R A study of the natural history of back pain. Part I: Development of a reliable and sensitive measure of disability in low-back pain Spine 1983 8 141 150 6222486
Zung WWK A self-rating depression scale Arch Gen Psych 1965 12 63 70
Main CJ The modified somatic perception questionnaire (MSPQ) J Psychosom Res 1983 27 503 514 6229628 10.1016/0022-3999(83)90040-5
Roland M Morris R A study of the natural history of low back pain. Part II: Development of guidelines for trials of treatment in primary care Spine 1983 8 145 150 6222487
Main CJ Wood PL Hollis S Spanswick CC Waddell G The distress and risk assessment method. A simple patient classification to identify distress and evaluate the risk of poor outcome Spine 1992 17 42 52 1531554
Spangfort EV The lumbar disc herniation. A computer-aided analysis of 2,504 operations Acta Orthop Scand 1972 142 1 95
Hakelius A Prognosis in sciatica Acta Orthop Scand 1970 129 1 70
Kosteljanetz M Bang F Schmidt-Olsen S The clinical significance of straight-leg-raising (Lasègue sign) in the diagnosis of prolapsed intervertebral disc Spine 1988 13 393 395 3406846
Andersson GBJ Deyo RA History and examination in patients with herniated lumbar discs Spine 1996 21 10S 18S 9112321
Bogduk N Commentary on A prospective study of centralization and lumbar and referred pain: A predictor of symptomatic discs and anular competence The Pain Medicine Journal Club Journal 1997 3 246 248
Schwarzer AC Derby R Aprill CN Fortin J Kine G Bogduk N Pain from the lumbar zygapophysial joints: a test of two models J Spinal Disord 1994 7 331 336 7949701
Katz JN Dalgas M Stucki G Lipson SG Degenerative lumbar spinal stenosis. Diagnostic value of the history and physical examination Arthritis & Rheumatism 1995 38 1236 1241 7575718
Fritz JM Erhard RE Delitto A Welch WC Nowakowski PE Preliminary results of the use of a two-stage treadmill test as a clinical diagnostic tool in the differential diagnosis of lumbar spinal stenosis J Spinal Disorders 1997 10 410 416
Dong GX Porter RW Walking and cycling tests in neurogenic and intermittent claudication Spine 1989 14 965 969 2551051
Spivak JM Current concepts review: Degenerative lumbar spinal stenosis Journal of Bone & Joint Surgery 1998 80-A 1053 1066 9698011
Baber YF Robinson AH Villar RN Is diagnostic arthroscopy of the hip worthwhile? A prospective review of 328 adults investigated for hip pain J Bone Joint Surg Br 2000 81 600 603 10463728 10.1302/0301-620X.81B4.8803
Stamathis G Arndt R-D, Horns JW, Gold RH Arthrography of the hip Clinical arthrography 1985 2 Baltimore: Williams & Wilkins 129 172
|
15943873
|
PMC1184083
|
CC BY
|
2021-01-04 16:32:04
|
no
|
BMC Musculoskelet Disord. 2005 Jun 9; 6:28
|
utf-8
|
BMC Musculoskelet Disord
| 2,005 |
10.1186/1471-2474-6-28
|
oa_comm
|
==== Front
BMC Musculoskelet DisordBMC Musculoskeletal Disorders1471-2474BioMed Central London 1471-2474-6-341597515110.1186/1471-2474-6-34Research ArticleBone mineral density, body mass index and cigarette smoking among Iranian women: implications for prevention Baheiraei Azam [email protected] Nicholas A [email protected] John A [email protected] Nguyen D [email protected] Tuan V [email protected] Bone and Mineral Research Program, Garvan Institute of Medical Research, St Vincent's Hospital, University of New South Wales, Sydney, Australia2 Department of Nuclear Medicine, St Vincent's Hospital, Sydney, Australia2005 24 6 2005 6 34 34 15 12 2004 24 6 2005 Copyright © 2005 Baheiraei et al; licensee BioMed Central Ltd.2005Baheiraei et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
While risk factors of osteoporosis in Western populations have been extensively documented, such a profile has not been well studied in Caucasians of non-European origin. This study was designed to estimate the modifiable distribution and determinants of bone mineral density (BMD) among Iranian women in Australia.
Methods
Ninety women aged 35 years and older completed a questionnaire on socio-demographic and lifestyle factors. BMD was measured at the lumbar spine (LS) and femoral neck (FN) using DXA (GE Lunar, WI, USA), and was expressed in g/cm2 as well as T-score.
Results
In multiple regression analysis, advancing age, lower body mass index (BMI), and smoking were independently associated with LS and FN BMD, with the 3 factors collectively accounting for 30% and 38% variance of LS and FN BMD, respectively. LS and FN BMD in smokers was 8% lower than that in non-smokers. Further analysis of interaction between BMI and smoking revealed that the effect of smoking was only observed in the obese group (p = 0.029 for LSBMD and p = 0.007 for FNBMD), but not in the overweight and normal groups. Using T-scores from two bone sites the prevalence of osteoporosis (T-scores ≤ -2.5) was 3.8% and 26.3% in pre-and post-menopausal women, respectively. Among current smokers, the prevalence was higher (31.3%) than that among ex-smokers (28.6%) and non-smokers (7.5%).
Conclusion
These data, for the first time, indicate that apart from advancing age and lower body mass index, cigarette smoking is an important modifiable determinant of bone mineral density in these Caucasians of non-European origin.
==== Body
Background
Osteoporosis is a common disorder in the elderly population, and represents one of the most significant public health problems in the world, predisposing to fractures with minimal or no antecedent trauma. These fractures are, in turn, associated with increased morbidity [1], reduced quality of life [2], mortality [3], and high health care costs [4].
Bone mineral density (BMD) measurement is considered an effective predictor of fracture risk, such that each standard deviation lower in BMD is associated with at least a 2-fold increase in age adjusted fracture risk. [5-7]. Therefore, a useful approach in assessing the importance of aetiological factors for osteoporosis is an investigation of the distribution and determinants of BMD. Although determinants of bone mineral density in Western populations have been extensively studied, such a profile has not been well documented in Caucasians of non-European origin.
Body weight or body mass index (BMI) is known to be positively associated with BMD[8,9]. Lifestyle factors such as low calcium intake, lack of physical activity, and smoking adversely affect bone mineral density and increase the risk of osteoporosis and its related fractures[10]. These factors also play an important role in the determination of peak bone mass and subsequent bone loss during the post-menopausal period. Among the modifiable risk factors of osteoporosis, cigarette smoking is considered one of the deleterious factors because cigarette smokers also have increased risk of fracture[11,12]. Nevertheless, the interactive effect of smoking on BMD has not been well studied. A recent study in a Caucasian population suggested that the effect of smoking was modified by body mass index, such that non-obese smokers had lower BMD than obese-smokers[13]. Iranian women on the average have a relatively high BMI[14,15], and it is not known whether such an interaction effect between smoking and BMI is present in this population.
The present study was designed to examine the modifiable distribution and determinants of bone mineral density among Iranian Australian women.
Methods
Subjects and setting
This study was designed as a cross-sectional investigation. All women were recruited via a media campaign using newsletters, noticeboards in community halls as well as word of mouth at community centres as part of a larger study to examine osteoporosis prevention in Iranian women. Inclusion criteria for the study were Iranian women and aged 35 years or older. The exclusion criteria were: current or past occurrence of any medical conditions known to affect bone metabolism such as Paget's disease and stroke; current pregnancy; and/or a history of breastfeeding within the last year. Also excluded were women who had been taking any medication affecting bone such as hormones, calcium, and glucocorticoids. In total, 96 women participated in the current study. Six women, who did not meet study's criteria on the basis of diseases or history of taking medications affecting bone, were excluded from the analysis. This study was approved by the University of New South Wales's Human Research Ethics Committee and written informed consent was obtained from each participant.
Data collection and measurements
Socio-demographic characteristics and lifestyle risk factors
Each woman completed a modified structured questionnaire [16] on socio-demographic and lifestyle risk factors. Income was included to be assessed, however, most participants refused to obtain information about their income level. Reproductive factors such as menopausal status and years since menopause were also provided for each participant. Menopause was defined as previous natural or surgical cessation of menstruation for more than 12 months. Calcium intake was calculated as the sum of current intake of main dairy products (milk, yogurt, and cheese) and was then converted to milligrams of calcium per day. Calcium contents for dairy products were provided from the product information in Australia [17]. Exercise was dichotomized as "yes" for current regular exercising, or "no" for not exercising. Amongst those who exercised, total amount of time spent per week was recorded. Current alcohol use was recorded as "yes" for drinking alcohol (beer, wine and liquor), or "no" for no intake of alcohol. Smoking habits were assessed based on previous and current cigarette smoking. Smoking status was dichotomized as "yes" for smoking, or "no" for never smoking. In addition, amongst those who smoked, dose and duration of smoking was recorded.
Anthropometric data
Weight (kg) and height (cm) were measured with light indoor clothing without shoes at the time of bone densitometry measurements. Weight was recorded to the nearest tenth of a kg using an electronic scale and standing height was measured to the nearest centimeter with a stadiometer. Body mass index (BMI) was calculated as body weight in kilograms divided by height in meters squared. According to the World Health Organization (WHO) recommended classification system, overweight and obese individuals were classified as having a BMI between 25 and 29, and equal to or greater than 30 kg/m2, respectively[18].
Bone density measurement
BMD was measured at the lumbar spine (LS) (L2-L4, anterior-posterior position) and femoral neck (FN) using dual-energy X-ray absorptiometry (DXA) with a Lunar Prodigy densitometer (GE Lunar, WI, U.S.A.). Areal BMD was expressed in g/cm2 and in standard deviations from the young normal mean (T-score), based on the Australian Reference Population. The sample of women was grouped into 3 groups based on the WHO recommended criteria: osteoporosis if T-score ≤ -2.5; osteopenia if -2.5 < T-score ≤-1.0; and normal if T-score >-1.0 [19].
Data analysis
To determine the magnitude of association between the potential risk factors (e.g., menopausal status, height, weight, dairy calcium intake, smoking, exercise, and alcohol use) and osteoporosis risk. Bone mineral density was considered the primary outcome, and was treated as a continuous variable. Individual risk factors were first considered in a simple linear regression analysis to estimate the strength of association between individual risk factor and BMD. In the subsequent analysis, all risk factors were simultaneously considered in a multiple linear regression analysis using the backward elimination algorithms, to screen for independent significant factors. Residual analysis performed to ensure that the usual assumptions of the regression model (i.e. normality, homogeneity and independence) were met. The entry of significance level (p value) was set to 0.10 to arrive at the most robust model.
In further analysis, differences between the pre-menopausal and post-menopausal groups were tested by unpaired t-test for the normally distributed variables, or the Mann-Whitney U test for non-normally distributed variables, and Chi-square test for categorized data. The analysis was performed with the SAS statistical analysis system[20] and SPSS for Windows statistical software [21].
Results
Characteristics of study subjects
The study population consisted of 90 women aged 48.5 ± 8.3 yr (mean age ± SD; range: 35 to 77 yr). Approximately 42 % of the women had education within high school. The majority of the women were married (78%) and performing home duties or not employed (56%). Their average duration of residence in Australia was about 10 years, with 75% of subjects having resided in Australia for at least 5 years. The mean age (SD) at immigration was about 39 ± 9.4 yr (range: 18 to 65 yr). The median (SD) dairy calcium intake in the women was 407 ± 283 mg/day. The Twenty-three women (26%) exercised regularly. Approximately 26% of women smoked cigarettes during their lifetime. Although cigarette smoking was common in these subjects, alcohol use was not frequent with about 11% of the women reporting drinking any kind of alcohol. Using the BMI criteria, 2.2% of subjects were underweight; 25.6% of women were in the healthy weight range; 35.6% were over-weight; and 36.7% were obese.
Forty two percent (n = 38) of women were post-menopausal, with the duration of post-menopause being between 1 and 32 years. Post-menopausal women had significantly higher age and parity and lower height, lumbar spine and femoral neck BMD, but no significant differences were found between the pre-and post-menopausal women in weight, BMI, dairy calcium intake, exercise, smoking status, duration of smoking, and alcohol use (Table 1).
Table 1 Clinico-demographic characteristic of study subjects
Pre-menopause Post-menopause p value
N 52 38
Age (years)* 43.6 ± 4.7 55.18 ± 7.4 <0.001a
Height (cm)* 157.7 ± 5.5 155 ± 5.7 0.027a
Weight (kg)* 70.8 ± 16.4 68.9 ± 10.3 0.530a
BMI (kg/m2)* 28.5 ± 6.9 28.7 ± 4.3 0.869a
LSBMD (g/cm2)* 1.19 ± 0.15 1.04 ± 0.16 <0.001a
FNBMD (g/cm2)* 0.97 ± 0.12 0.87 ± 0.11 <0.001a
Dairy calcium intake (mg/day)* 410 ± 262 498 ± 306 0.147a
Age at menopause* - 47.9 ± 4.02 -
Parity† 2 (2, 3) 3 (2, 4) 0.004b
Regular exercise§ 12 (23.1) 11(28.9) 0.528c
Smoking status
Current smokers 15.4 (8) 21.1 (8) 0.487c
Ex-smokers 21.2 (11) 23.7 (9) 0.775c
Duration of smoking (years)§ 0.155c
≤ 5 4 (33.3) 1 (9.1)
> 5 8 (66.7) 10 (90.9)
Alcohol use§ 7 (13.5) 3 (7.9) 0.407c
*Mean ± SD; †median (interquartile range); § n (%).
aunpaired t-test, bMann-Whitney U test, cChi-square test.
BMI, body mass index; LSBMD, lumbar spine bone mineral density; FNBMD, femoral neck bone mineral density.
Determinants of BMD
In simple linear regression analysis, age, height, weight, BMI, menopausal status, smoking habits, duration of smoking, and cigarette dose were each significantly associated with LS and FN BMD (Table 2). However, in the multiple linear regression, advancing age, lower BMI and smoking were independent predictors of LS and FN BMD (Table 3). After adjusting for age and BMI, smokers had 0.087 g/cm2 (8 %) and 0.075 g/cm2 (8 %) lower in LS and FN BMD, respectively, than non-smokers. The 3 factors collectively accounted for 30% and 38% of the variation in LS and FN BMD, respectively.
Table 2 Univariate association between individual risk factors and bone mineral density
Factor LSBMD (g/cm2) FNBMD (g/cm2)
β ± SE a R2b β ± SEa R2b
Age (per 1 year) -0.009 ± 0.002* 0.20 -0.008 ± 0.001* 0.25
Height (per -5 cm) 0.04 ± 0.012* 0.07 0.030 ± 0.010* 0.07
Weight (per -5 kg) 0.02 ± 0.005* 0.10 0.015 ± 0.005* 0.14
BMI (per -5 kg/m2) 0.035 ± 0.015* 0.05 0.030 ± 0.010* 0.07
Post-menopause -0.153 ± 0.033* 0.19 -0.103 ± 0.026* 0.15
Smoking (current and ex-smokers) -0.100 ± 0.040* 0.06 -0.086 ± 0.030* 0.08
Duration of smoking (per 5 years) -0.037 ± 0.015* 0.06 -0.031 ± 0.011* 0.07
Cigarette dose (per 10 cig/day) -0.054 ± 0.019* 0.08 -0.038 ± 0.014* 0.07
Dairy calcium intake (per 300 mg/day) 0.020 ± 0.03 0.001 0.020 ± 0.03 0.02
Regular exercise (yes) 0.017 ± 0.042 0.002 0.044 ± 0.031 0.02
Alcohol use (yes) 0.004 ± 0.058 0.001 -0.021 ± 0.044 0.003
BMI, body mass index; LSBMD, lumbar spine bone mineral density; FNBMD, femoral neck bone mineral density.
* P < 0.05 (statistically significant)
a Values are regression coefficients ± SE describing the change in bone mineral density (g/cm2) associated with the unit change in the risk factor
b Coefficient of determination: the proportion of variation in bone mineral density explained by the variation in a risk factor
Table 3 Association between age, body mass index, smoking, and bone mineral density:Results of multiple linear regression analysis
Determinant LSBMD (g/cm2)a FNBMD (g/cm2)a
Age (per 1 year) -0.008 ± 0.002** -0.007 ± 0.001**
BMI (-5 kg/m2) 0.006 ± 0.003* 0.005 ± 0.002**
Smoking (yes) -0.087 ± 0.035* -0.075 ± 0.025**
R2* 0.30 0.38
*0.01 <p < 0.05; **0.0001 <p < 0.01
a Values are regression coefficients ± SE describing the change in BMD (g/cm2) associated with one year advancing age, 5 kg/ m2 decrease in BMI and smoking status.
BMI, body mass index; LSBMD, lumbar spine bone mineral density; FNBMD, femoral neck bone mineral density.
* R2, coefficient of determination: a measure of the proportion of variation in BMD explained by the variation in the risk factors. The variables included in the initial regression analysis were: age, menopause status, BMI, smoking status, duration of smoking, cigarette dose, calcium intake, exercise, and alcohol use
As expected, advancing age was negatively associated with BMD in the both sites (LS: r = 0.45, p = 0.0001; FN: r = 0.50, p = 0.0001). Nevertheless, there was a significant positive correlation between BMI and LS and FN BMD (LS: r = 0.22, p = 0.033; FN: r = 0.26, p = 0.012). Current smokers had significantly lower lumbar spine and femoral neck BMD than non-smokers. However, there was no significant difference between ex-smokers and non-smokers in both BMD sites (Fig. 1). Among smokers, there was no significant linear correlation between cigarette dose and BMD (p = 0.14 for LSBMD and p = 0.64 for FNBMD) and duration of smoking and BMD (p = 0.76 for LSBMD and p = 0.86 for FNBMD).
Figure 1 Mean and standard error of lumbar spine (upper panel) and femoral neck (lower panel) bone mineral density (g/cm2) by smoking status.
Further analysis of interaction between BMI and smoking revealed that the effect of smoking was only observed in the obese group (p = 0.029 for LSBMD and p = 0.007 for FNBMD), but not in the overweight and normal groups (Fig. 2). This interaction effect was not affected by the dose of cigarette or duration of smoking. Moreover, there was a non-statistically significant interaction between age and smoking, as both smokers and non-smokers appeared to have a similar age-BMD association (Fig. 3).
Figure 2 Mean and standard error of lumbar spine (upper panel) and femoral neck (lower panel) bone mineral density (g/cm2) by body mass index and smoking status.
Figure 3 Interaction between age and smoking status on lumbar spine (upper panel) and femoral neck (lower panel) bone mineral density (g/cm2).
Prevalence of low bone density
Twenty-five women (27.8%) were osteopenic (T-score -1 to -2.49) at the lumbar spine and 32 (35.6%) at the femoral neck. Using the WHO T-score-based definition of osteoporosis, the proportion of women with osteoporosis was 12.2% (n = 11) at the lumbar spine and 2.2% (n = 2) at the femoral neck. When the two measures were considered simultaneously, the prevalence of osteoporosis was 13.3%. In post-menopausal women, the prevalence of osteoporosis (T-score ≤ -2.5) was 23.7% (n = 9) at the lumbar spine and 5.3% (n = 2) at the femoral neck. Among smokers, the prevalence was 30.4% (7/23) which was significantly higher (p < 0.01) than that among non-smokers (7.5%, 5/67).
Discussion
Osteoporosis is recognized as a public health problem in the world, but its epidemiology in non-Western populations remains poorly understood. This study represents an original contribution to the study of osteoporosis in Iranian women in Australia. Apart from advancing age and body mass index, it was found that cigarette smoking was an important modifiable determinant of bone mineral density.
These results are consistent with previous studies which indicated that advancing age was associated with lower BMD. In this study, each year increase in age was estimated to cross-sectionally "decrease" 0.8% in LS and FN BMD. This estimate is relatively consistent with longitudinal studies in Western Caucasian women which suggest an annual decrease of 1%[22,23].
Consistent with previous studies[8,9], this study found that lower BMI was associated with lower BMD. Moreover, the effect of smoking was significant in obese women. The prevalence of current cigarette smoking in this population (17.8%) was surprisingly higher than Iranian women in Iran which was around 3.6%[24]. BMD in smokers was 8% lower than in non-smokers, after adjusting for age and BMI. This difference is clinically significant, because each SD reduction in BMD is associated with a 2-fold increase in age adjusted fracture risk.
BMD measurements in ex-smokers were intermediate between current and non-smokers. It seems the effect of smoking varies linearly with the intake of cigarette and may suggest that smoking cessation have a positive effect on both LS and FN BMD. This makes cigarette smoking one of the important modifiable lifestyle risk factors of osteoporosis in Iranian Australian women.
Although some studies found an association between calcium intake and BMD, the present study found no such association. The results also revealed a non-significant relationship between exercise and BMD. One of the reasons for these results may be that the questionnaire only reflected the present situation, not permanent lifestyle. In addition, frequency and type of exercise on bone density could not be evaluated, because few of the subjects exercised regularly.
Alcohol use was not significantly associated with bone mineral density. Although cigarette smoking was common in this sample, alcohol use was not frequent and only a few (11%) women reported drinking any kind of alcohol, with the majority drinking only monthly or rarely. With such a low prevalence of use, it is perhaps not surprising that the study was unable to detect a significant effect of alcohol use on BMD.
Using T-scores from two bone sites the prevalence of osteoporosis (T-scores ≤ -2.5) was 3.8% (n = 2) and 26.3% (n = 10) in pre-and post-menopausal women, respectively. The most notable observation in this study is that osteoporosis was more likely to be identified at the lumbar spine than the femoral neck. This finding is consistent with a previous study among Iranian women in Iran [25]. However, the prevalence of osteoporosis at the femoral neck seemed to be relatively lower in the Iranian women compared with most Asian and other Caucasian populations [26-29]. These results suggest that BMD measured at the femoral neck requires further investigation in Iranian women.
The present study's findings must be interpreted within the context of a number of strengths and weaknesses. This study is the first attempt to address an important public health problem amongst Iranian women in an Australian setting. Some aspects of acculturation should be taken into account in the interpretation of these findings. Since 75% of women have settled in Australia for at least 5 years, change of lifestyle factors such as sun exposure and diet cannot be ruled out, and the present results may not be generalizable to Iranian women in Iran. The association between risk factors and BMD as observed in this study cannot be interpreted as a causal relationship because the study was a cross-sectional investigation. Because the women in this study were not sampled from an age-stratified scheme, the average T-scores and prevalence of osteoporosis could have been affected by the actual age group distribution, and this represents a potential limitation for generalizing the results to the general population.
Conclusion
These data suggest that, apart from advancing age and lower BMI, cigarette smoking is an important modifiable determinant of bone mineral density in the Iranian Australian women. These findings can potentially contribute toward the development of more effective public health strategies for the health promotion and osteoporosis prevention in this population.
Further studies are required to investigate the effect of changing environmental exposures which can influence osteoporosis prevalence and fracture risk in this population.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
Conception and study design: AB, JE, TN
Data collection: AB, NP
Drafting manuscript: AB
Data analysis: AB, NN, TN
Review of manuscript for important intellectual content: AB, NP, JE, TN
All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The Iranian Ministry of Health and Medical Education is thanked for the award of a scholarship to AB. The authors acknowledge the assistance of Ms Fiona McGrath in the measurement of bone densitometry. We wish to express our thanks to A/Prof Jan Ritchie for her comments. We would like to thank Mr. Hamid Atighpour for his assistance with the data collection. We also thank the Iranian women for their participation in this study.
==== Refs
Melton LJ 3rd Adverse outcomes of osteoporotic fractures in the general population J Bone Miner Res 2003 18 1139 1141 12817771
Hansen LB Vondracek SF Prevention and treatment of nonpostmenopausal osteoporosis Am J Health Syst Pharm 2004 61 2637 54 15646699
Cummings SR Melton LJ Epidemiology and outcomes of osteoporotic fractures Lancet 2002 359 1761 67 12049882 10.1016/S0140-6736(02)08657-9
Randell A Sambrook PN Nguyen TV Lapsley H Jones G Kelly PJ Eisman JA Direct clinical and welfare costs of osteoporotic fractures in elderly men and women Osteoporos Int 1995 5 427 432 8695963 10.1007/BF01626603
Marshall D Johnell O Wedel H Meta-analysis of how well measures of bone mineral density predict occurrence of osteoporotic fractures BMJ 1996 312 1254 59 8634613
Melton LJ Atkinson EJ O'Fallon WM Wahner HW Riggs BL Long term fracture risk prediction by bone mineral density assessed at different skeletal sites J Bone Miner Res 1993 8 1227 33 8256660
Nguyen TV Sambrook PN Kelly PJ Jones G Lord S Freund J Eisman J Prediction of osteoporotic fractures by postural instability and bone density BMJ 1993 307 1111 15 8251809
Felson DT Zhang Y Hannan MT Anderson JJ Effects of weight and body mass index on bone mineral density in men and women: the Framingham study J Bone Miner Res 1993 8 567 73 8511983
Nguyen TV Center JR Eisman JA Osteoporosis in elderly men and women: effects of dietary calcium, physical activity, and body mass index J Bone Miner Res 2000 15 322 31 10703935
Nguyen TV Center JR Eisman JA Osteoporosis: underrated, underdiagnosed and undertreated Med J Aust 2004 180 S18 22 14984358
Gerdhem P Obrant KJ Effects of cigarette smoking on bone mass as assessed by dual-energy X-ray absorptiometry and ultrasound Osteoporos Int 2002 13 932 36 12459935 10.1007/s001980200130
Ward KD Klesges RC A meta-analysis of the effects of cigarette smoking on bone mineral density Calcif Tissue Int 2001 68 259 70 11683532
Jones G Scott FS A cross-sectional study of smoking and bone mineral density in premenopausal parous women: effect of body mass index, breastfeeding, and sports participation J Bone Miner Res 1999 14 1628 633 10469293
Maddah M Eshraghian MR Djazayery A Mirdamadi R Association of body mass index with educational level in Iranian men and women Eur J Clin Nutr 2003 57 819 23 12821881 10.1038/sj.ejcn.1601615
Mirmiran P Mohammadi F Sarbazi N Allahverdian S Azizi F Gender differences in dietary intakes, anthropometrical measurements and biochemical indices in an urban adult population: the Tehran Lipid and Glucose Study Nutr Metab Cardiovasc Dis 2003 13 64 71 12929618
Doheny M Sedlak C Osteoporosis Preventive Behavior Survey, Kent State University, 1995 2000
National Food Authority The Supplement to NOTTAB95M 1995 Canberra: Australian Government Publishing Service
World Health Organization Preventing and managing the global epidemic: report of the WHO consultation on obesity 1998 Geneva, WHO
World Health Organization Assessment of fracture risk and its application to screening for postmenopausal osteoporosis. Report 843 1994 Geneva: World Health Organization
SAS Institute SAS/STAT: User's Guides 1990 SAS Institute, Cary, NC
SPSS Inc SPSS for Windows, Release 11.0.0 2001 Chicago: IL: SPSS, Inc
Jones G Nguyen T Sambrook P Kelly PJ Eisman JA Progressive loss of bone in the femoral neck in elderly people: longitudinal findings from the Dubbo osteoporosis epidemiology study BMJ 1994 17:309 691 695 7950520
Nguyen TV Sambrook PN Eisman JA Bone loss, physical activity, and weight change in elderly women: the Dubbo Osteoporosis Epidemiology Study J Bone Miner Res 1998 13 1458 67 9738519
Ahmadi J Khalili H Jooybar R Namazi N Mohammadagaei P Prevalence of cigarette smoking in Iran Psychol Rep 2001 89 339 41 11783559
Rassouli A Milanian I Moslemi-Zadeh M Determination of serum 25-hydroxyvitamin D (3) levels in early postmenopausal Iranian women: relationship with bone mineral density Bone 2001 29 428 30 11704493 10.1016/S8756-3282(01)00591-9
Looker AC Orwoll ES Johnston CC JrLindsay RL Wahner HW Dunn WL Calvo MS Harris TB Heyse SP Prevalence of low femoral bone density in older U.S. adults from NHANES III J Bone Miner Res 1997 12 1761 8 9383679
Iki M Kagamimori S Kagawa Y Matsuzaki T Yoneshima H Marumo F Bone mineral density of the spine, hip and distal forearm in representative samples of the Japanese female population: Japanese Population-Based Osteoporosis (JPOS) Study Osteoporos Int 2001 12 529 37 11527049 10.1007/s001980170073
Wu XP Liao EY Huang G Dai RC Zhang H A comparison study of the reference curves of bone mineral density at different skeletal sites in native Chinese, Japanese, and American Caucasian women Calcif Tissue Int 2003 73 122 32 14565593 10.1007/s00223-002-1069-7
Limpaphayom KK Taechakraichana N Jaisamrarn U Bunyavejchevin S Chaikittisilpa S Poshyachinda M Taechamahachai C Havanond P Onthuam Y Lumbiganon P Kamolratanakul P Prevalence of osteopenia and osteoporosis in Thai women Menopause 2001 8 65 9 11201518 10.1097/00042192-200101000-00011
|
15975151
|
PMC1184084
|
CC BY
|
2021-01-04 16:32:04
|
no
|
BMC Musculoskelet Disord. 2005 Jun 24; 6:34
|
utf-8
|
BMC Musculoskelet Disord
| 2,005 |
10.1186/1471-2474-6-34
|
oa_comm
|
==== Front
BMC Musculoskelet DisordBMC Musculoskeletal Disorders1471-2474BioMed Central London 1471-2474-6-361598752510.1186/1471-2474-6-36Research ArticleImmunologic testing of xeno-derived osteochondral grafts using peripheral blood mononuclear cells from healthy human donors Hetherington Vincent J [email protected] Jill S [email protected] Douglas S [email protected] Oleg S [email protected] Paul V [email protected] Daniel [email protected] Ohio College of Podiatric Medicine, 10515 Carnegie Avenue, Cleveland, Ohio, 44106 USA2 Cellular Technology Limited, 10515 Carnegie Avenue, Cleveland, Ohio, 44106 USA3 Centerpulse Orthopedics, Ltd., Postfach 65, CH-8404 Winterthur, Switzerland2005 29 6 2005 6 36 36 12 1 2005 29 6 2005 Copyright © 2005 Hetherington et al; licensee BioMed Central Ltd.2005Hetherington et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
One means of treating osteoarthritis is with autologous or allogeneic osteochondral grafts. The purpose of this study was to evaluate the innate immunological response in humans toward xeno-derived osteochondral grafts that have been partially or entirely treated by the photooxidation process.
Methods
The antigens tested included bovine, porcine, ovine and equine osteochondral samples that have been treated in successive steps of photooxidation. ELISPOT assays were used to evaluate the production of IL-1, IL-4, IL-6, IL-10, IL-12 and TNF-α by human monocytes in response to the antigens.
Results
Results indicated vigorous production of IL-1, IL-6, IL-10 and TNF-α in response to untreated bovine, porcine and equine specimens. This indicates that these samples are perceived as foreign, or stimulatory, by the human monocytes. There was no induction of IL-4 or IL-12, which is required for Th2 and Th1 immunity, respectively. In contrast, the processed bovine, porcine and equine samples did not induce significant activation of cells of the innate immune system. This occurred after the first step in processing (after cleaning in increasing strengths of ethanol). This suggests that the processing steps dramatically, if not completely, negated the immunostimulatory properties of the test sample. The results for the ovine samples indicate a reverse response.
Conclusion
The findings of the study suggest that photooxidized bovine, porcine or equine samples have the potential to be used as an osteochondral graft. Although the first step in processing reduced the immunological response, photooxidation is still necessary to retain the structure and mechanical integrity of the cartilage, which would allow for immediate joint resurfacing.
==== Body
Background
There are various methods for repairing cartilage defects in patients with osteoarthritis or other degenerative joint disease. One such method is mosaicplasty, which consists of the use of osteochondral grafting with autogenous [1-3] or allogeneic [4-6] implants. Concerns exist, however, with the use of both autogenous grafts due to the need to harvest the osteochondral implants from a joint or part of a joint that is otherwise healthy, and for allogeneic materials with respect to disease transmission. Both of these graft types also have a limited availability.
There is the potential for using photooxidized xenogeneic osteochondral scaffolds in place of autogenous or allogeneic grafts. Photooxidation is a light-mediated process that results in the cross-linking of collagen fibers. The result is a material that is relatively non-immunogenic and resistant to chemical and enzymatic degradation, while maintaining the mechanical properties of the tissue, as demonstrated with bovine [7-10] and porcine [11,12] tissue. The photooxidized cartilage has been shown to have no viable chondrocytes after treatment. [10]
Animal studies have included analysis of photooxidized bovine pericardial tissue for replacement of the mitral valve [13] and aortic valve. [14] In addition, the benefits of photooxidation in vein grafting have been demonstrated. [15,16] Additional animal studies have demonstrated good biocompatibility with integration of xeno-derived osteochondral scaffolds. [10,17,18] Kawalec et al. [17] examined the use of photooxidized bovine osteochondral plugs in a rabbit model. The grafts were implanted into the patellar groove of both knees for 12 weeks. Histology indicated a chronic inflammatory reaction to the photooxidized implant. Active bone remodeling was observed in the bone portion of the implant, while the cartilage layer remained intact and undamaged, but covered by a thin layer of fibrous tissue. Akens et al. [18] implanted photooxidized xenogeneic osteochondral plugs into the femoral condyle of ovine knees and compared the findings to untreated xenogeneic and autogeneic grafts after six, 12 and 18 months. Macroscopically, results indicated that most grafts possessed a well-maintained cartilage surface at all time points. Histologically, sections from all three experimental groups showed evidence of cystic lesions at six months, in addition to evidence of bone remodeling. The number of sections containing cysts decreased at 12 months for the photooxidized samples, but remained high for the untreated xenogeneic and autogeneic samples. This was also observed at 18 months. Cartilage fusion between graft and host was observed for the photooxidized implants, but not for the untreated xenogeneic and autogeneic implants.
The purpose of this study was to evaluate the innate immunological response in humans toward xeno-derived osteochondral grafts that have been partially or entirely treated by the photooxidation process. The long-term goal of this research is to evaluate the potential for using photooxidized xeno-derived osteochondral grafts in the repair of articular cartilage defects.
Methods
Antigen preparation
The grafts tested in this project were made from bovine, porcine, ovine or equine bone and cartilage. Samples were prepared by removing an osteochondral plug approximately 8 mm in diameter by 7 mm in length. After retrieval, the samples were placed in a phosphate-buffered saline (PBS) solution to prevent them from drying out prior to processing. There were five groups of grafts, with the exception of the equine grafts, of which there were two groups, based on the specific steps performed during processing, as described in Table 1. Samples from each group were then frozen at -80°C, ground to a fine powder averaging 2–5 μm and subsequently lyophilized and packaged prior to testing. The antigens were dissolved in water at a concentration of 1 mg/ml and were used for the ELISPOT assays at 10 μg/ml.
Table 1 Experimental groups. Five groups of antigens, based on specific steps performed during processing
Processing steps Bovine Group Porcine Group Ovine Group Equine Group
Group 1 – untreated B1 P1 O1 E1
Group 2 – after cleaning in increasing strengths of ethanol B2 P2 O2 N/A
Group 3 – after cleaning in ethanol and methylene chloride B3 P3 O3 N/A
Group 4 – after photooxidation process B4 P4 O4 N/A
Group 5 – after sterilization with Gamma radiation B5 P5 O5 E5
Human PBMC donors
Peripheral blood mononuclear cells (PBMC) were obtained from ten healthy donors that were 30–50 years of age, and ten donors that were 50–70 years of age. There was an equal distribution of males and females. Heparinized blood was obtained by venipuncture. The study was performed under the approval of the Institutional Review Board at the Ohio College of Podiatric Medicine.
Cell separations
The monocytes were isolated from PBMC by negative selection using the RosetteSep enrichment cocktail (Stemcell Technologies, Vancouver, BC, Canada). The enrichment cocktail was mixed with the blood to constitute a 50 μg/ml final concentration, followed by 20 minutes incubation at room temperature. Subsequently, the cells were separated using standard Ficoll-Hypaque density gradient centrifugation. The ensuing isolated monocyte enriched cell fraction was of >92% purity.
ELISPOT assays
Human cytokine ELISPOT assays for the detection of Interleukin-1 (IL-1), IL-4, IL-6, IL-10, IL-12 and Tumor Necrosis Factor-alpha (TNF-α) were performed as described previously [19]. Briefly, ImmunoSpot plates (Cellular Technology Limited, Cleveland, OH) were coated overnight at 4°C with the cytokine-specific capture antibody in PBS at concentrations specified below. The plates were blocked with 1% bovine serum albumin (10 g/l in PBS: PBS-BSA) for one hour and washed three times with PBS. Monocytes were plated in triplicates in complete RPMI 1640 medium at 5 × 104 cells per well. Antigens were added at 10 μg/ml to the final concentration. Wells containing the cells in the complete medium were used as a negative control. For the bovine and porcine samples tested with blood from the 30–50 year old age group, lipopolysaccharide (LPS) at a concentration of 2 μg/ml was used as a positive control. For all other assays, phytohemaglutinin (PHA) stimulation at a final concentration of 10 μg/ml was used as a positive control.
The cells were cultured in an incubator at 37°C. The IL-1, IL-4, IL-6 and TNF-α assays cultured for 24 hours, while the IL-10 and IL-12 assays were harvested after 48 hours. Subsequently, the plates were washed three times with PBS, then three times with PBS-Tween (0.5 %), and the detection antibodies (in PBS-1% BSA-0.05%Tween) were added at concentrations specified below. After an overnight incubation at 4°C, plates were washed four times with PBS-Tween, and the Streptavidin-Horse Radish Peroxidase (HRP) conjugate (Dako Corp., Carpenteria, CA) was added at a 1:2000 dilution in PBS-BSA for 2 hours at room temperature. Afterwards, plates were washed three times with PBS-Tween followed by washing three times with PBS.
The spots were visualized using the HRP-substrate 3-amino-9-ethylcarbazole (AEC) (Pierce Pharmaceuticals, Rockford, IL). The AEC stock solution was prepared by dissolving 10 mg AEC in 1 ml N,N-dimethyl formamide (Fisher Scientific, Fair Lawn, NJ). For the actual development, 1 ml of the AEC stock solution was freshly diluted into 30 ml of 0.1 M sodium-acetate buffer (pH 5.0), filtered (0.45 μm), and mixed with 15 μl H2O2. Two hundred μl of the AEC solution were plated per well. The reaction was stopped by rinsing with distilled water when spots became clearly visible macroscopically (10 to 45 minutes, dependent on the cytokine). The plates were air-dried overnight before subjecting them to image analysis using a Series 2 ImmunoSpot Analyzer (Cellular Technology, Ltd., Cleveland, OH). Each measurement was made in triplicate, and averaged to obtain the final value.
For human ELISPOT assays, the anti- IL-4, IL-6, IL-10, IL-12 and TNF-α antibodies that were used were purchased from BD Pharmingen (San Diego, CA), while anti-IL-1 antibodies were received from Pierce Biotechnology (Rockford, IL). Capture antibodies were used as follows: anti-IL-1(cat. #M-421B-E) at 3 μg/ml, IL-4 (cat. #554515) at 5 μg/ml, IL-6 (cat. #554543) at 1.5 μg/ml, IL-10 (cat. #554705) at 5 μg/ml, IL-12 (cat. #555065) at 5 μg/ml, and TNF-α at 2 μg/ml (cat. #551220). For detection, biotinylated antibodies were used: IL-1 (cat. #M-420B-E) at 0.3 μg/ml, IL-4 (cat. #554483) at 1.5 μg/ml, IL-6 (cat. #554546) at 1 μg/ml, IL-10 (cat. #554499) at 1 μg/ml, IL-12 (cat. #554660) at 0.6μg/ml, and TNF-α at 1 μg/ml (cat. #554511).
Results
Bovine samples
A representative photograph of individual ELISA spot wells is shown in Figure 1. Each black spot represents a cytokine-producing cell.
Figure 1 ELISA spot wells. Representative photograph of individual ELISA spot wells. Each black spot represents a cytokine-producing cell. This example represents IL-10 production by PBMC of healthy donors.
The results of the ELISA spot testing for bovine samples tested with cells from donors 30–50 years of age are shown in Figure 2. Untreated samples of bovine bone and cartilage (group B1) induced vigorous production of IL-1, IL-6, IL-10 and TNF-α; however no induction of IL-4 or IL-12 was seen. In contrast, the processed bone and cartilage (groups B2–B5) did not induce significant activation of cells of the innate immune system. The elevated levels of IL-6 seen in all five groups were at a level similar to that found in the media alone. For the positive controls, stimulation with LPS resulted in a response on the order of that seen with group B1.
Figure 2 Bovine 30–50 years old. Production of cytokines by purified PBMC of healthy donors from 30–50 years old group (bovine).
The results of the ELISA spot testing for bovine samples tested with cells from donors 50–70 years of age are shown in Figure 3. The findings were very similar to that produced by the 30–50 year old group. The exception was the decreased production of IL-1 and IL-6 in response to untreated samples of bone and cartilage (group B1). For the positive control, stimulation with PHA resulted in cells producing IL-1 and IL-6 too numerous to count, while the number of cells producing IL-10 and TNF-α were on the order of that produced by group B1.
Figure 3 Bovine 50–70 years old. Production of cytokines by purified PBMC of healthy donors from 50–70 years old group (bovine).
Porcine samples
The cytokine signature produced by testing porcine samples with cells from the 30–50 year old group (Figure 4) is similar to the findings from the bovine samples tested with cells from the same age group. In response to the untreated porcine bone and cartilage (group P1), there was a vigorous production of IL-1, IL-6, IL-10 and TNF-α, while no induction of IL-4 or IL-12 was seen. As with the bovine samples, the processed specimens (groups P2–P5) did not induce the significant response seen with the untreated samples. For the positive control, stimulation with LPS resulted in a response on the order of that seen with group P1.
Figure 4 Porcine 30–50 years old. Production of cytokines by purified PBMC of healthy donors from 30–50 years old group (porcine).
The results for the porcine samples tested with cells from the individuals that were 50–70 years of age (Figure 5) were similar to those from the bovine samples tested with cells from the same group. A strong response was evoked towards the untreated specimens (group P1), with respect to IL-10 and TNF-α. While less than that produced by the younger age group, the production of IL-1 and IL-6 was still significant. As with the bovine results, the processed specimens (groups P2–P5) did not induce a vigorous response. For the positive controls, stimulation with PHA resulted in cells producing IL-1 and IL-6 too numerous to count, while the number of cells producing IL-10 and TNF-α were on the order of that produced by group P1.
Figure 5 Porcine 50–70 years old. Production of cytokines by purified PBMC of healthy donors from 50–70 years old group (porcine).
Ovine samples
The results for the ovine samples tested with cells from the 30–50 year old group (Figure 6) were the reverse of those from the bovine and porcine samples. Unprocessed samples (group O1) did not induce production of cytokines. Only after sterilization with Gamma radiation (Group O5) was there a slight increase in the production of IL-1 and IL-6. The production of IL-10 was much weaker than that resulting from bovine and porcine samples. For the positive controls, stimulation with PHA resulted in an average number of cells producing IL-1 and IL-6 too numerous to count, while the average number of cells producing IL-10 and TNF-α was also elevated in comparison to the negative control and experimental groups.
Figure 6 Ovine 30–50 years old. Production of cytokines by purified PBMC of healthy donors from 30–50 years old group (ovine).
The cytokine signature for ovine samples tested with cells from the 50–70 years of age group (Figure 7) was similar to that from the younger age group. There was no cytokine response to the untreated samples (group O1). After the photooxidized process (group O4), there was an increase in production of IL-1 and a vigorous production of IL-6 and TNF-α. After sterilization with Gamma radiation (group O5), there was an even more vigorous production of IL-6 and TNF-α. Again, the production of IL-10 was much weaker than that observed with bovine and porcine samples. For the positive controls, stimulation with PHA resulted in elevated responses for IL-1, IL-6, IL-10 and TNF-α, with respect to the negative control and experimental groups.
Figure 7 Ovine 50–70 years old. Production of cytokines by purified PBMC of healthy donors from 50–70 years old group (ovine).
Equine samples
Only two equine samples were tested: untreated specimens (group E1) and after Gamma radiation (group E5). The results for the equine samples tested with cells from the 30–50 year old age group are shown in Figure 8. In response to the untreated samples, there was a slight increase in the production of IL-1, IL-6 and TNF-α. This response was absent after sterilization with Gamma radiation. There was no production of IL-10 for either group. For the positive controls, stimulation with PHA resulted in cells producing IL-1 and IL-6 too numerous to count. The number of cells producing IL-10 was two orders of magnitude greater than group H1 and those producing TNF-α was one magnitude of order greater than that produced by group H1.
Figure 8 Equine 30–50 years old. Production of cytokines by purified PBMC of healthy donors from 30–50 years old group (equine).
The cytokine signature for equine samples tested with cells from the older age group (Figure 9) was similar to that from the younger age group. In response to untreated samples, there was production of IL-1, IL-6 and TNF-α. This response did not appear after sterilization with Gamma radiation. It should be noted that there was an absence in the production of IL-10 for both groups. For the positive controls, stimulation with PHA resulted in elevated responses for IL-1, IL-6, IL-10 and TNF-α, with respect to the negative control and experimental groups.
Figure 9 Equine 50–70 years old. Production of cytokines by purified PBMC of healthy donors from 50–70 years old group (equine).
Discussion
The cytokine signature produced by the untreated bovine and porcine samples is consistent with vigorous activation of cells of the innate immune system. This pertains primarily to cytokines involved in induction of a humoral immune response (IL-10), but not of cytokines that are required for Th1 immunity (IL-12) or Th2 immunity (IL-4). The IL-1 and TNF-α production seem to favor mixed Th0-type T cell responses. This indicates that the untreated bovine and porcine samples are perceived as foreign, or stimulatory, by the human cells. This is consistent in both age groups, and most individuals tested behaved comparably. Therefore, in spite of genetic differences in the human test population, rather uniform response patterns were seen, suggesting that the findings will broadly apply to the human population.
In contrast, the processed bovine and porcine osteochondral samples did not induce the production of cytokines. This was observed after the first processing step, after cleaning in increasing strengths of ethanol. This suggests a highly improved biocompatibility, with respect to inducing a rejection T cell/antibody response and inducing sterile granulomatous rejections in the sense of a foreign body reaction. It can be expected that these non-stimulatory samples can be considered well suited for transplantation. Again, this was observed for both age groups.
The cytokine signature for equine samples tested with cells from both age groups is similar to that observed with bovine and porcine samples. The exception with the untreated samples is the weak production of IL-10, which is a potent B cell differentiation factor. Thus, the equine samples may be less likely to induce antibody responses.
The reverse was seen in response to ovine samples. The untreated samples did not produce a response, while the specimens that were photooxidized, and to a greater extent those that were sterilized with Gamma irradiation resulted in a response of IL-1, IL-6 and TNF-α. This suggests a mixed Th0-type T cell response. There was a much weaker production of IL-10, which indicates the absence of a humoral response. These findings were observed with both age groups.
It should be noted that the graft material tested in this study was a prepared material that was subjected to a series of proprietary steps of cleaning, photooxidation and sterilization in order to develop a graft or scaffold that avoids human antigenicity and rejection reactions. The antigenic composition of the processed material (groups 2–5) is fundamentally different from the natural, unprocessed material (group 1). It was intended to test each graft as a whole, as it is proposed to be used in clinical applications, rather than breaking down the graft into its components. The grafts that have been photooxidized are not live material, and have been rendered hypo-immunogenic to an extent that they are accepted in a series of recipient animal species without the need for immune suppression. [10,17,18] Photooxidized materials have been used clinically in cardiac surgery, and, to the best of our knowledge, with the graft material tested as a whole and not its potential constituents. [11-16]
Previous work has established that photooxidized cartilage is accepted by the host better than untreated cartilage. [17,18] It is likely that the oxidation process chemically modifies the protein components in such a way that these structures stimulate cells of the innate immune system to induce, for example, prostaglandins. Frequently, sterile inflammation of this sort enhances wound repair and therefore enhances the vascularization and integration of the grafted tissue, rather than rejection. Because the clinical acceptance of these grafts is common we assume that such reactions are also induced in this model.
Monocytes are not constitutive parts of the joint, however these cells are rapidly recruited to the joint after tissue injury, including surgical implantation of grafts. Unlike chondrocytes that are not professional antigen presenting cells, monocytes can phagocytose, process and present graft antigens and initiate an immune response. This initial phase was evaluated in this study. Once the T cells are primed, they can interact with chondrocytes and other non-professional antigen presenting cells in the joint to exert the effector arm of the rejection reaction.
The conclusions reached in this study were drawn using in vitro experiments. The isolation of monocytes and their tissue culture plates might affect the functionality of the cells and modify them relative to their in vivo activation state. These constraints apply to any in vitro work. However, the purpose of this study was to establish whether monitoring the activity of such cells in vitro would provide insights on additional activation of them by stimulation with compounds such as the tissue culture material that was used. It is clearly shown that beyond whatever activation occurs in tissue culture, the graft materials tested induce a variable level of activation. Because activation of such cells of the innate immune system is key to their immunological activity, this study shows that sensitive monitoring of monocytes activity in response to such materials can be done in vitro.
Conclusion
Grafts are rejected by a series of immune recognition events that include cells of the innate immune system (dendritic cells, macrophages) as well as the downstream activation of lymphocytes. Activation of cells of the innate immune system is a prerequisite for inducing an immune response by lymphocytes. Moreover, the type of cytokines induced in cells of the innate immune system by the antigen will define the (Th1/Th2) type of T cell response that evolves. This study demonstrated a reduction in the innate immune response to processed bovine, porcine and equine bone and cartilage, indicating a potential for use as a xeno-derived osteochondral scaffold. The effect was observed to occur after the first step in processing (after cleaning with increasing strengths of alcohol).
Although the first step in processing reduced the immunological response, photooxidation is still necessary to retain the structure and mechanical integrity of the cartilage, which would then allow for immediate joint resurfacing. The next phase of the study is to evaluate the downstream activation of the immune system caused by xeno-derived osteochondral grafts in an active system model, using an animal model. This will be followed by evaluation of the implant and immune reactions when a photooxidized implant is placed in the joint environment. Subsequent evaluation for the detection of viral or other residual infectious materials, such as α-galactase, is needed to alleviate concerns about xenotransplantation.
Competing interests
The research project was funded by Centerpulse Orthopedics Ltd. It is not known whether they will gain financially from the publication of this paper.
Authors' contributions
VJH helped conceive the study, participated in its design and coordination, assisted with analysis of findings, and helped to draft the manuscript. JSK participated in the design of the study, coordinated the blood withdrawal, assisted with data analysis, and drafted the manuscript. DD carried out blood withdrawal and assisted with analysis of the findings. OST coordinated the cell separations and ELISPOT assays, assisted with data analysis and helped to draft the manuscript. PVL participated in the design of the study, coordinated cell separations and ELISPOT assays and assisted with data analysis. DN carried out the antigen preparation. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
Antigens and funding for this study were provided by Centerpulse Orthopedics, Ltd.
==== Refs
Sammarco GJ Makwana NK Treatment of talar osteochondral lesions using local osteochondral graft Foot Ankle Int 2002 23 693 698 12199381
Jakob RP Franz T Gautier E Mainil-Varlet P Autologous osteochondral grafting in the knee: indication, results, and reflections Clin Orthop 2002 401 170 184 12151894
Mendecino RW Catanzariti AR Hallivis R Mosaicplasty for the treatment of osteochondral defects of the ankle joint Clin Podiatr Med Sur 2001 18 495 513
Kim CW Jamali A Tontz W JrConvery FR Brage ME Bugbee W Treatment of post-traumatic ankle arthrosis with bipolar tibiotalar osteochondral shell allografts Foot Ankle Int 2002 23 1091 1102 12503799
Gross AE Aubin P Cheah HK Davis AM Ghazavi MT A fresh osteochondral allograft alternative J Arthroplasty 2002 17 50 53 12068404 10.1054/arth.2002.32447
Fitzpatrick PL Morgan DA Fresh osteochondral allografts: a 6–10-year review Aust NZ J Surg 1998 68 573 579
Moore MA Bohachevsky IK Cheung DT Boyan BD Chen W-M Bickers RR McIlroy BK Stabilization of pericardial tissue by dye-mediated photooxidation J Biomed Mater Res 1994 28 611 618 8027101 10.1002/jbm.820280511
Moore MA Chen W-M Phillips RE Bohachevsky IK McIlroy BK Shrinkage temperature versus protein extraction as a measure of stabilization of photooxidized tissue J Biomed Mater Res 1996 32 209 214 8884497 10.1002/(SICI)1097-4636(199610)32:2<209::AID-JBM9>3.0.CO;2-X
McIlroy BK Robinson MD Chen W-M Moore MA Chemical modification of bovine tissues by dye-mediated photooxidation J Heart Valve Dis 1997 6 416 423 9263875
Akens MK von Rechenberg B Bittmann P Nadler D Zlinszky K Auer JA In-vitro studies of a photo-oxidized bovine articular cartilage J Vet Med A Physiol Pathol Clin Med 2002 49 39 45 11913825
Moore MA Adams AK Calcification resistance, biostability, and low immunogenic potential of porcine heart valves modified by dye-mediated photooxidation J Biomed Mater Res 2001 56 24 30 11309787 10.1002/1097-4636(200107)56:1<24::AID-JBM1064>3.0.CO;2-Q
Moore MA Phillips RE JrMcIlroy BK Walley VM Hendry PJ Evaluation of porcine valves prepared by dye-mediated photooxidation Ann Thorac Surg 1998 66 S245 S248 9930457 10.1016/S0003-4975(98)01118-7
Svendsen CA Kreykes NS Butany J Bianco RW In-vivo assessment of a photofixed bovine pericardial valve J Heart Valve Dis 2000 9 813 820 11128791
Schoen FJ Pathologic findings in explanted clinical bioprosthetic valves fabricated from photooxidized bovine pericardium J Heart Valve Dis 1998 7 174 179 9587858
Chanda J Kuribayashi R Liu KX Shibata Y Inhibitory effect of photooxidation on intimal and medical thickening of saphenous vein Ann Thorac Surg 1998 66 449 454 9725383 10.1016/S0003-4975(98)00444-5
Liu KX Yamamoto F Sekine S Goto Y Seki K Kondoh K Fu Y Inhibitory effect of methylene blue-induced photooxidation on intimal thickening of vein graft Ann Thorac Surg 1999 68 84 88 10421120 10.1016/S0003-4975(99)00448-8
Kawalec JS Hetherington VJ Nadler D Photooxidized bovine cartilage implants for the replacement of articular cartilage: histological evaluation 6th World Biomaterials Congress Trans 2000 204
Akens MK von Rechenberg B Bittmann P Nadler D Zlinszky K Auer JA Long-term in-vivo studies of a photo-oxidized bovine osteochondral transplant in sheep BMC Musculoskeletal Disorders 2001 2 9 11747477 10.1186/1471-2474-2-9
Helms T Boehm BO Asaad RJ Trezza RP Lehmann PV Tary-Lehmann M Direct visualization of cytokine-producing recall antigen-specific CD4 memory T cells in healthy individuals and HIV patients J Immunology 2000 164 3723 3732 10725731
|
15987525
|
PMC1184085
|
CC BY
|
2021-01-04 16:32:05
|
no
|
BMC Musculoskelet Disord. 2005 Jun 29; 6:36
|
utf-8
|
BMC Musculoskelet Disord
| 2,005 |
10.1186/1471-2474-6-36
|
oa_comm
|
==== Front
BMC OphthalmolBMC Ophthalmology1471-2415BioMed Central London 1471-2415-5-161598752110.1186/1471-2415-5-16Research ArticleA selective cyclic integrin antagonist blocks the integrin receptors αvβ3 and αvβ5 and inhibits retinal pigment epithelium cell attachment, migration and invasion Hoffmann Stephan [email protected] Shikun [email protected] Manlin [email protected] Marianne [email protected] Peter [email protected] Stephen J [email protected] David R [email protected] Doheny Eye Institute, Departments of Ophthalmology, Keck School of Medicine, University of Southern California 1355 San Pablo Street, Los Angeles 90033, CA, USA2 Department of Pathology, Keck School of Medicine, University of Southern California, 2011 Zonal Ave HMR 209, Los Angeles, CA 90033, USA3 Department of Ophthalmology, University of Leipzig, Liebigstrasse 10–14, 04103 Leipzig, Germany4 Berufsgenossenschaftliche Kliniken Bergmannsheil, University of Bochum, Department of Internal Medicine I, Bürkle-de-la-Camp-Platz 1, D-44789 Bochum, Germany2005 29 6 2005 5 16 16 9 3 2005 29 6 2005 Copyright © 2005 Hoffmann et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Proliferative vitreoretinopathy (PVR) is a leading cause of blindness after failed retinal reattachment surgery. PVR is characterized by the proliferation, migration and contraction of retinal pigmented epithelial cells (RPE), and these cellular responses are influenced by the expression and function of integrin receptors. The effect of a cyclic integrin antagonist containing the amino acid sequence Arg-Gly-Asp-D-Phe-Val (RGDfV), specific for the integrin receptors αvβ3 and αvβ5, was investigated on basic fibroblast growth factor (bFGF), platelet derived growth factor-BB (PDGF-BB), and serum induced human RPE proliferation, migration, invasion and attachment to the extracellular matrix. Furthermore, the effects of bFGF and PDGF-BB regulated expression of integrins αvβ3 and αvβ5 on RPE cells was examined.
Methods
The effect of a cyclic integrin antagonist and a control peptide (0.01 μg/ml to 300 μg/ml) was investigated on serum or cytokine (bFGF or PDGF-BB pretreatment) induced human fetal RPE cell proliferation by H3-thymidine uptake. The effect of the cyclic integrin antagonist on RPE cell attachment onto different extracellular matrices (laminin, collagen IV, fibronectin), RPE cell invasion stimulated by PDGF-BB or serum, and migration stimulated by PDGF-BB, vascular endothelial growth factor (VEGF) or serum was explored. PDGF-BB and bFGF modulation of the integrin receptors αvβ3 and αvβ5 was evaluated by flow cytometry.
Results
The integrin antagonist did not inhibit DNA synthesis stimulated by serum, bFGF, or PDGF-BB treatment. RPE attachment onto fibronectin was inhibited in a concentration range of 1–10 μg/ml (p < 0.05). Attachment of the RPE cells onto collagen IV and laminin was inhibited in a range of 3–10 μg/ml (p < 0.05). Serum and PDGF-BB stimulated migration was inhibited by the cyclic integrin antagonist in a concentration range of 1–10 μg/ml (p < 0.05). Furthermore, the cyclic integrin antagonist inhibited PDGF-BB stimulated RPE cell invasion through fibronectin (3μg/ml: 66% inhibition, p < 0.001). In each of these experiments, the control peptides had no significant effects. PDGF-BB and bFGF pretreatment of RPE cells increased the expression of integrin receptors αvβ3 (bFGF: 1.9 fold, PDGF-BB: 2.3 fold) and αvβ5 (bFGF: 2.9 fold, PDGF-BB: 1.5 fold).
Conclusion
A selective inhibition of the integrin receptors αvβ3 and αvβ5 through a cyclic integrin antagonist is able to inhibit RPE cell attachment, migration and invasion. Since these steps are of importance for the progression of PVR, a cyclic integrin antagonist should be further evaluated for the treatment of this disease.
==== Body
Background
Integrins are a family of heterodimeric, non-covalently bound cell surface receptors, which mediate cell-cell and cell to extracellular matrix (ECM) adhesion. They are transmembrane glycoproteins consisting of a larger α and a smaller β subunit. In mammals, 18 α- and 8 β integrin genes encode polypeptides that combine to form 24 αβ heterodimeric receptors [1]. By mediating cell-cell and cell-matrix contact, the integrins are involved in a spectrum of physiologic and pathologic processes. Integrin expression is modulated by several cytokines, growth factors and the extracellular matrix in different cell types [2].
The integrins αvβ3 and αvβ5 have been shown to be of particular significance for the progression of neovascularization. Both integrin receptors contain RGD sequences containing the amino acid sequence Arg-Gly-Asp. Vitronectin receptor integrins of the αv – chain subfamily interact with their target proteins via the tripeptide sequence RGD (NH2-arginine-glycine-aspartic acid-COOH) [3,4]. The blockage of these RGD receptor sequences induces endothelial cell apoptosis by inhibiting their binding to the ECM [5,6]. In addition fibroblast attachment onto fibrinogen, a substep involved in wound healing, is inhibited by blockage of the αvβ3 integrin receptor [7]. Therefore, the integrin αvβ3 plays a critical role during wound repair as an adhesion receptor and as a signaling receptor [8]. As an adhesion receptor, it has the exceptional ability to bind vitronectin, fibronectin and fibrinogen, the three major proteins present in the wound provisional matrix [9].
In proliferative vitreoretinopathy (PVR), a disease of the posterior eye characteristically induced by trauma or failed retinal reattachment surgery, integrins are also of importance for the progression of the disease [10,11]. PVR, an exaggerated wound healing process, is characterized by the proliferation, migration and contraction of retinal pigment epithelial cells (RPE), fibroblasts, glial cells and macrophages. PVR membranes are formed on, or under the sensory retina; their contraction may lead to retinal detachment and blindness. RPE cells play a dominant role in the pathobiology of the disease [12]. Growth factors including basic fibroblast growth factor (bFGF), platelet derived growth factor-BB (PDGF-BB), and vascular endothelial growth factor (VEGF), and the ECM molecule fibronectin, regulate proliferation, migration and invasion of the RPE cell and are expressed in PVR membranes [12,13]. Fibronectin is an adhesive substrate for both the integrin receptors αvβ3 and αvβ5 [14]. Several studies have identified integrins αvβ3, αvβ5, integrin alpha subunits 2,3,4,5,6, V, and integrin beta subunits 1, 2, 3 on RPE cells and cells within PVR membranes [15-17]. The ECM-cell interactions are thought to be mediated largely through this family of cell surface receptors. Because cell-ECM interaction is important in the process of wound healing, one could predict that integrins would be actively involved in the pathogenesis of PVR. For PVR, it was shown in vivo, that tractional retinal detachment could be inhibited [11] by blockage of integrin-ECM binding by an RGD containing disintegrin.
Peptide antagonists specific for individual integrins have been developed, including cyclic RGD peptides selective for the αv integrin [18], the αvβ3 integrin [19], and the αvβ5 integrin [18-21]. Sequence alterations in the backbone of the RGD-containing cyclic peptides that result in conformational differences in the interatomic distance between the Cβ atoms of Arg and Asp, have been shown to correlate with selective peptide recognition [18,22,23].
In this study we investigated the modulatory effect of the cytokines PDGF-BB and bFGF on the expression of the RGD containing integrin receptors αvβ3 and αvβ5 on human fetal RPE cells. Furthermore, the effect of a specific cyclic integrin antagonist that blocks the integrin receptors αvβ3 and αvβ5, was investigated for its effects on RPE attachment, proliferation, migration and invasion. A specific cyclic integrin antagonist was chosen for its higher affinity and stronger binding capacity to the αvβ3 and αvβ5 integrins compared to a non-cyclic integrin antagonist [24,18].
Methods
All cell culture solutions, media and ECM molecules (fibronectin, laminin and vitronectin) were purchased from Sigma (Sigma, St. Louis, MO).
Isolation and culture of human RPE cells
Human RPE cells were isolated from fetal donor eyes (gestation time over 22 weeks) which were obtained from the Anatomic Gift foundation (Woodbine, GA) as has been described [25]. RPE cells were seeded onto laminin-coated 6 well plates (Fisher Scientific, Tustin, CA) using DMEM with 10 % fetal bovine serum, 1 % penicillin/streptomycin and glutamine. Cells from passage 3 to 5 were used in all experiments.
Synthesis of the cyclic integrin antagonist and its control peptide
The cyclic integrin antagonist RGDfV, specific for the inhibition of integrins αvβ3 and αvβ5 [18-21] and the control peptide RADfV were synthesized as described [6] at the USC/ Norris Cancer Peptide synthesis facility. The peptides were dissolved in Hank's Balanced Salt Solution (HBSS), and stored at -80°C. HPLC analysis revealed a purity > 99% of the synthesized peptide.
Flow cytometry
RPE cells (30,000 cells/well) were treated for 5 days with the cytokines bFGF and PDGF-BB at a concentration of 10 ng/ml in 6 well plates. Cells were detached, fixed with 4 % paraformaldehyde, blocked with 5 % goat serum for 30 minutes, centrifuged at 4°C, and incubated with a monoclonal mouse antibody against αvβ3 or αvβ5 (5 μg/ml) (Chemicon, CA) for one hour. After washing, a fluorescein conjugated secondary antibody (Chemicon, CA) directed against the mouse antibodies was added in a dilution of 1:100 for one hour. The expression of the integrin receptors αvβ3 or αvβ5 was measured using a fluorescent-activated cell sorter (FACStar plus, Becton Dickinson, Mountain View, CA). Forward and side light scatter was used to gate the desired scattered events (RPE cells) from dead cells and debris. The fluorescent index was determined by multiplying the percentage of cells by their mean fluorescence. At least 10,000 cells were evaluated/experiment and all experiments were performed in triplicate.
Cell counting and viability studies
Subconfluent, bFGF (10 ng/ml) or PDGF-BB (10 ng/ml) pre-stimulated RPE cells were seeded at a density of 8 × 104 cells per well in 6 well plates containing DMEM with 10 % FBS and the cyclic integrin antagonist or the control peptide (0.01 μg/ml, 0.1 μg/ml, 1 μg/ml, 30 μg/ml and 300 μg/ml). The cells were exposed during the proliferation assay to 10% FBS or the cytokines bFGF and PDGF-BB (10 ng/ml). After 3 days of culture, the cell number was determined using a hemocytometer. The experiments were performed six times, each in triplicate. Cell viability was estimated by observation of cell morphology and by exclusion of 0.4 % trypan blue solution [26].
3H-thymidine uptake
RPE cells (10,000 cells/well) were seeded into DMEM with 10 % FBS in 24 well plates with the cyclic integrin antagonist or the control peptide (0.01 μg/ml, 0.1 μg/ml, 1 μg/ml, 30 μg/ml and 300 μg/ml). To determine a possible stronger effect of the cyclic integrin antagonist on RPE cell proliferation, cells were pre-stimulated for 5 days with bFGF (10 ng/ml) or PDGF-BB (10 ng/ml) (R&D Systems, Minneapolis, MN) to induce an upregulation of the integrin receptors αvβ3 or αv β 5. After the pre-incubation, cells were seeded in DMEM with 10 % FBS, with the cyclic integrin antagonist or control peptide, and bFGF (10 ng/ml) or PDGF-BB (10 ng/ml). Cells were treated for three days with the peptides, then 3H-thymidine uptake was measured as described previously [25]. Results are presented as the percentage of control values.
Migration assay
Migration was performed in a modified Boyden chamber (Falcon, Becton Dickinson Labware, Franklin Lakes, NJ). Subconfluent RPE cells, which have been pre-exposed to the cyclic integrin antagonist peptide (0.001 ng/ml, 0.01 ng/ml, 0.1 ng/ml, 1 ng/ml, 3 ng/ml and 10 ng/ml) or the control peptide (10 μg/ml) over night were seeded into DMEM with 1 % FBS into the upper part of a Boyden chamber (50,000 cells/chamber). The cyclic integrin antagonist was added directly to the upper part of the chamber (0.001 ng/ml, 0.01 ng/ml, 0.1 ng/ml, 1 ng/ml, 3 ng/ml, or 10 ng/ml). To evaluate nonspecific activity of the integrin antagonist, a control peptide (10 μg/ml) was used. The lower part of the chamber was filled with DMEM with 1 % FBS containing the chemoattractant PDGF-BB (10 ng/ml) or the ECM molecule fibronectin (30 ng/ml). RPE cells were incubated for 8 hours in a humidified CO2-incubator. Then the cells in the upper part were scraped away and the membrane was fixed with 1/2 strength Karnovsky's solution and stained with 0.25 % Richardson's solution. RPE cells on the lower side of the membrane were counted with the help of an inverted microscope and a grid (Carl Zeiss, Jena, Germany) in five random high power fields (200 X) in triplicate.
Cell attachment
Ninety-six well plates were coated with the ECM molecules laminin, fibronectin or collagen IV (25 μg/ml). RPE cells were pre-incubated with the cyclic integrin antagonist (concentrations as above) or control peptide (10 μg/ml) over night. Then 15 × 103 cells, suspended in 100 μl DMEM with the cyclic or control peptide were seeded in each well and allowed to attach for 60 minutes. A tetrazolium MTT [3-(4,5-dimethylthiazolyl-2)-2,5-diphenyltetrazolium bromide] solution (20 μl/well) (Boehringer Mannheim, Indianapolis, IN, USA) was added for 5 hours of incubation. Then the supernatants were decanted, 150 μl of 100% DMSO added, and the cells were placed on a shaker for 10 minutes. Absorbance of the cells was measured at 550 nm with a Dynatech MR 600 spectrophotometric microplate reader.
Invasion assay
RPE cell invasion was examined in a modified Boyden chambers. The insert membrane was coated overnight with fibronectin (25 μg/ml). Subconfluent RPE cells were pretreated overnight with the cyclic integrin antagonist (3 μg/ml) or the control peptide (3 μg/ml). An RPE cell suspension (50,000 cells/0.1 ml) in DMEM with 0.4 % FBS and 3 μg/ml cyclic integrin antagonist or control peptide was added into the upper part of the chamber. 600 μl DMEM with 0.4% FBS and PDGF-BB (10 μg/ml) was added into the lower part of the chamber. Invasion was measured with and without PDGF-BB stimulation after 5 hours at 37°C (95 % air/ 5 % Co2). After PBS washing and methanol fixation (10 minutes at 4°C), counterstaining with hematoxylin was performed. Cells in the upper chamber were wiped away. RPE cells that invaded the lower surface of the membrane were quantified by cell counting (320 × magnification).
Statistical analysis
All experiments were repeated in triplicate. Standard deviations and averages were calculated. The data obtained were analyzed for significance using one way analysis of variance (ANOVA). Data with a p-value of less then 0.05 were considered to be statistically significant.
Results
The cyclic integrin antagonist does not inhibit RPE cell proliferation
The RPE cells were stimulated to incorporate thymidine by either 10% serum or by 10 ng/ml bFGF or 10 ng/ml PDGF-BB (p < 0.01). Neither the integrin antagonist, nor control peptide showed any significant effect on serum, PDGF-BB or bFGF stimulated DNA synthesis. (data not shown). In addition, cell counting showed no effects of the integrin antagonist or the control peptide on serum or cytokine stimulated proliferation. Furthermore, a lack of toxicity for the integrin antagonist and the control peptide in the concentration range tested (0.01 μg/ml, 0.1 μg/ml, 1 μg/ml, 30 μg/ml and 300 μg/ml) was determined by use of the trypan blue exclusion assay (data not shown).
bFGF and PDGF-BB upregulate surface expression of αvβ3 and αvβ5
Pre-stimulation of RPE cells for 5 days with the cytokines PDGF-BB and bFGF (10 ng/ml) increased the expression of the integrin receptors αvβ3 and αvβ5 (Fig. 1; representative flow cytometric study). Incubation with 10 ng/ml bFGF upregulated the expression of αvβ3 (1.9 fold) and strongly enhanced the expression of αvβ5 (2.9 fold). PDGF-BB (10 ng/ml) strongly enhanced αvβ3 (2.3 fold), and less strongly upregulated surface expression of αvβ5 (1.5 fold).
The cyclic integrin antagonist inhibits RPE attachment onto ECM
RPE cell attachment onto fibronectin was inhibited significantly in the range of 1 to 10 μg/ml of cyclic integrin antagonist (p < 0.05) (Fig. 2A). A concentration of 10 μg/ml control peptide showed no effect. RPE cell attachment was inhibited 18 % with 1 μg/ml (p = 0.03), 23 % with 3 μg/ml (p < 0.005) and 27 % with 10 μg/ml (p < 0.001) onto fibronectin. An ID50 of 18.5 μg/ml cyclic integrin antagonist was calculated for the RPE cell attachment onto fibronectin. The attachment onto collagen IV was inhibited 25 % with 3 μg/ml (p < 0.001), and 29 % with 10 μg/ml (p < 0.001). Lower concentrations showed no significant effect. An ID50 of 17.2 μg/ml cyclic peptide was calculated for RPE cell attachment onto collagen IV. Attachment onto laminin was inhibited by the cyclic integrin antagonist in the range of 3–10 μg/ml (3 μg/ml, 27 % inhibition (p < 0.001); 10μg/ml: 23 % inhibition (p < 0.001). An ID50 of 21.7 μg/ml was calculated for the inhibition of RPE cell attachment onto laminin.
The cyclic integrin antagonist inhibits RPE cell migration
The ECM component fibronectin (30 ng/ml) induced chemotaxis when placed as a soluble agent in the lower part of the Boyden chamber (Fig. 2B). This migratory response was inhibited by the cyclic integrin antagonist but not by the control peptide. RPE cell migration was inhibited in a concentration range of 1–10 μg/ml (1μg/ml, 18 % inhibition (p < 0.001); 3 μg/ml, 23 % inhibition (p < 0.001); 10 mg/ml, 27 % inhibition (p < 0.001)). An ID50 of 18.5 μg/ml integrin antagonist was calculated for the serum induced migration. The migratory response induced by PDGF-BB (10 ng/ml) (93 % stimulation of migration) was diminished by the cyclic integrin antagonist. (1μg/ml, 26 % inhibition (p < 0.001); 3 μg/ml, 66 % inhibition (p < 0.001); and 10μg/ml, 74 % inhibition (p < 0.001)). An ID50 of 5.8 μg/ml was calculated for the inhibition of PDGF-BB induced migration of RPE cells. VEGF induced a migratory response of RPE cells by 39 %. This migratory response was inhibited by the cyclic integrin antagonist in a concentration range of 3–10 μg/ml (p < 0.01).
The cyclic integrin antagonist inhibits PDGF-BB stimulated invasion through fibronectin
PDGF-BB strongly stimulated RPE cell invasion (77 %, (p < 0.001)) through fibronectin coated onto the lower surface of a membrane insert in a modified Boyden chamber. This invasion was inhibited by the cyclic integrin antagonist (3 μg/ml, 66 % inhibition (p < 0.001)). The control peptide had no significant effect on invasion (Fig. 2C). An ID50 of 2.3 μg/ml was calculated for the effects of the cyclic integrin antagonist on PDGF-BB induced RPE invasion through fibronectin.
Discussion
These data provide support for the suggestion that inhibition of the integrin receptors αvβ3 and αvβ5 by a specific cyclic integrin antagonist is feasible for inhibiting RPE cell attachment, migration and invasion and may serve as an adjunctive therapy for disorders such as PVR. Integrins have been shown to influence cell signaling pathways that promote cell proliferation in concert with cytokines [27]. Interestingly, our data show that cyclic integrin inhibition of αvβ3 and αvβ5 had no effect on serum, PDGF-BB or bFGF induced proliferation of RPE. Since we also show that bFGF and PDGF-BB upregulate both αvβ3 and αvβ5, and that proliferation is not inhibited in these stimulated cells, we can infer that the lack of effect of the cyclic peptide on proliferation is not due to lack of integrin expression. Consistent with the lack of effect on proliferation, we found that the cyclic integrin inhibitor did not induce significant cell death. This suggests that in the conditions studied here, the effect of integrin inhibition by the cyclic peptide on survival and proliferation of RPE cells may differ from previously described effects on endothelial cells [28], specific tumor cells [29] and mammary epithelial cells [30].
Cell attachment of RPE cells onto fibronectin, laminin and collagen IV was inhibited by the cyclic integrin antagonist. Because substrate attachment and interaction is important for activated RPE function [31], it is reasonable to predict that anti-adhesive therapy holds promise for the therapy of PVR. The cyclic integrin antagonist has been shown previously to inhibit specific ligand binding or cell adhesion events mediated by αvβ3 or αvβ5 [18]. Blockage of RPE attachment onto laminin, collagen IV and fibronectin is of specific interest, because of the strong expression of these matrix proteins in PVR [12]. The ECM fibronectin reacts with the αvβ3 receptor with high specificity and affinity, supporting cell adhesion to this matrix protein. The αvβ5 receptor binds in a similar way to fibronectin [32]. The blockage of αvβ3 and αvβ5 thus provides a reasonable explanation for the decreased attachment of RPE onto collagen IV, fibronectin and laminin in the presence of the inhibitory cyclic peptide.
RPE cell migration is another important process in the development of PVR. Without migration, RPE cells would not gain access to the vitreous and form vitreoretinal membranes. For RPE cells, it was shown previously that monoclonal antibody inhibition of integrin subunit β 1 inhibits RPE cell migration from wound edges in organ culture [14]. Integrin signaling results in actin cytoskeleton polymerization, which in turn regulates cell shape and motility [33]. The αvβ3 receptor appears to be involved in pathways that regulate intracellular pH and intracellular Ca2+ [34] and may therefore be involved in migration in response to motility factors and extracellular matrix components [35]. The same range of concentration of the cyclic peptide inhibited migration induced by both soluble fibronectin and PDGF-BB. This result may in part be due to the finding that PDGF-BB induced migration is mediated by signaling pathways that are also used by integrin receptors [36-38].
The cyclic integrin antagonist had a particularly strong effect on RPE invasion into fibronectin. This may be due to the concurrent effects of the peptide on the inhibition of adhesion, migration and proteolytic activity since each of these steps may play a critical role in the invasion of the ECM.
The cyclic integrin antagonist described here blocks both αvβ3 and αvβ5. In contrast, highly selective integrin antagonists such as LM609 block only αvβ3. Therefore, a stronger inhibitory effect of the cyclic integrin antagonist can be expected as has been shown in tumors [36]. While increased specificity could result in lower toxicity, we did not find that the cyclic integrin antagonist inhibited thymidine incorporation, or that it increased trypan blue uptake in the concentrations tested.
Conclusion
The cyclic integrin antagonist RGDfV inhibits RPE migration, attachment and invasion, and should be further investigated as a potential adjuvant pharmacological treatment for PVR.
Competing interests
The author(s) declare that they have no competing interests
Authors' contributions
All authors have been involved in the experimental design, data collection and manuscript preparation. The final manuscript has been read and approved by all authors.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
Supported in part by grant EY02061 and by core grant EY03040 from the National Institute of Health and a grant from the German research association (WI 880\9-1).
Figures and Tables
Figure 1 A. Expression of the integrin receptors αvβ3 and αvβ5 after pretreatment with the cytokine bFGF in a concentration of 10 ng/ml for 5 days.B. Expression of the integrin receptors αvβ3 and αvβ5 after pretreatment with the cytokine PDGF-BB in a concentration of 10 ng/ml for 5 days.
Figure 2 A. RPE attachment on fibronectin, laminin and collagen IV. The cells were pretreated before the assay with the cyclic integrin antagonist, or the control peptide (10 μg/ml). Untreated RPE cells were used in addition as a standard (ST). B. RPE cell migration induced by PDGF-BB, VEGF and fibronectin. The effect of the cyclic integrin antagonist or the control peptide (cp) on migration is shown. (ST = untreated RPE cells) C. Effect of the cyclic integrin antagonist, (3 ug/ml) or the control peptide on RPE invasion through fibronectin. Cell invasion through fibronectin (F) was stimulated with PDGF-BB (10 ng/ml). Nonspecific effects were excluded by the use of a cyclic control peptide. DMEM without PDGF-BB was used as a control in the assay.
==== Refs
Hynes RO Integrins: bidirectional, allosteric signaling machines Cell 2002 110 673 687 12297042 10.1016/S0092-8674(02)00971-6
Kim LT Yamada KM The regulation of expression of integrin receptors Proc Soc Exp Biol Med 1997 214 123 131 9034129
Smith JW Cheresh DA Integrin (alpha v beta 3)-ligand interaction. Identification of a heterodimeric RGD binding site on the vitronectin receptor J Biol Chem 1990 265 2168 2172 1688848
Ruoslahti E Pierschbacher MD New perspectives in cell adhesion: RGD and integrins Science 1987 238 491 497 2821619
Brooks PC Montgomery AM Rosenfeld M Reisfeld RA Hu T Klier G Cheresh DA Integrin alpha v beta 3 antagonists promote tumor regression by inducing apoptosis of angiogenic blood vessels Cell 1994 79 1157 1164 7528107 10.1016/0092-8674(94)90007-8
Hammes HP Brownlee M Jonczyk A Sutter A Preissner KT Subcutaneous injection of a cyclic peptide antagonist of vitronectin receptor-type integrins inhibits retinal neovascularization Nat Med 1996 2 529 533 8616710 10.1038/nm0596-529
Gailit J Clarke C Newman D Tonnesen MG Mosesson MW Clark RA Human fibroblasts bind directly to fibrinogen at RGD sites through integrin alpha(v)beta3 Exp Cell Res 1997 232 118 126 9141628 10.1006/excr.1997.3512
Hannigan GE Dedhar S Protein kinase mediators of integrin signal transduction J Mol Med 1997 75 35 44 9020382 10.1007/s001090050084
Clark RA Tonnesen MG Gailit J Cheresh DA Transient functional expression of alphaVbeta 3 on vascular cells during wound repair Am J Pathol 1996 148 1407 1421 8623913
Robbins SG Brem RB Wilson DJ O'Rourke LM Robertson JE Westra I Planck SR Rosenbaum JT Immunolocalization of integrins in proliferative retinal membranes Invest Ophthalmol Vis Sci 1994 35 3475 3485 8056523
Yang CH Huang TF Liu KR Chen MS Hung PT Inhibition of retinal pigment epithelial cell-induced tractional retinal detachment by disintegrins, a group of Arg-Gly-Asp-containing peptides from viper venom Invest Ophthalmol Vis Sci 1996 37 843 854 8603869
Charteris DG Proliferative vitreoretinopathy: pathobiology, surgical management, and adjunctive treatment Br J Ophthalmol 1995 79 953 960 7488586
Jerdan JA Pepose JS Michels RG Hayashi H de Bustros S Sebag M Glaser BM Proliferative vitreoretinopathy membranes. An immunohistochemical study Ophthalmology 1989 96 801 810 2662102
Elner SG Elner VM The integrin superfamily and the eye Invest Ophthalmol Vis Sci 1996 37 696 701 8603855
Anderson DH Johnson LV Hageman GS Vitronectin receptor expression and distribution at the photoreceptor-retinal pigment epithelial interface J Comp Neurol 1995 360 1 16 7499556 10.1002/cne.903600102
Anderson DH Guerin CJ Matsumoto B Pfeffer BA Identification and localization of a beta-1 receptor from the integrin family in mammalian retinal pigment epithelial cells Invest Ophthalmol Vis Sci 1990 31 81 93 2137116
Weller M Wiedemann P Bresgen M Heimann K Vitronectin and proliferative intraocular disorders. II. Expression of cell surface receptors for fibronectin and vitronectin in periretinal membranes Int Ophthalmol 1991 15 103 108 1708746 10.1007/BF01046429
Pfaff M Tangemann K Muller B Gurrath M Muller G Kessler H Timpl R Engel J Selective recognition of cyclic RGD peptides of NMR defined conformation by alpha IIb beta 3, alpha V beta 3, and alpha 5 beta 1 integrins J Biol Chem 1994 269 20233 20238 8051114
Healy JM Murayama O Maeda T Yoshino K Sekiguchi K Kikuchi M Peptide ligands for integrin alpha v beta 3 selected from random phage display libraries Biochemistry 1995 34 3948 3955 7535098 10.1021/bi00012a012
Koivunen E Gay DA Ruoslahti E Selection of peptides binding to the alpha 5 beta 1 integrin from phage display library J Biol Chem 1993 268 20205 20210 7690752
Koivunen E Wang B Ruoslahti E Isolation of a highly specific ligand for the alpha 5 beta 1 integrin from a phage display library J Cell Biol 1994 124 373 380 7507494 10.1083/jcb.124.3.373
Ruoslahti E RGD and other recognition sequences for integrins Annu Rev Cell Dev Biol 1996 12 697 715 8970741 10.1146/annurev.cellbio.12.1.697
Ponce ML Nomizu M Kleinman HK An angiogenic laminin site and its antagonist bind through the alpha(v)beta3 and alpha5beta1 integrins Faseb J 2001 15 1389 1397 11387236 10.1096/fj.00-0736com
Cox D Aoki T Seki J Motoyama Y Yoshida K The pharmacology of the integrins Med Res Rev 1994 14 195 228 8189836
Hoffman S Gopalakrishna R Gundimeda U Murata T Spee C Ryan SJ Hinton DR Verapamil inhibits proliferation, migration and protein kinase C activity in human retinal pigment epithelial cells Exp Eye Res 1998 67 45 52 9702177 10.1006/exer.1998.0491
Freshny RI Animal cell culture: a practical approach. 1992 2 nd Ed. Oxford, England, UK , IRL-Press
Clark EA Brugge JS Integrins and signal transduction pathways: the road taken Science 1995 268 233 239 7716514
Meredith JEJ Fazeli B Schwartz MA The extracellular matrix as a cell survival factor Mol Biol Cell 1993 4 953 961 8257797
Taga T Suzuki A Gonzalez-Gomez I Gilles FH Stins M Shimada H Barsky L Weinberg KI Laug WE alpha v-Integrin antagonist EMD 121974 induces apoptosis in brain tumor cells growing on vitronectin and tenascin Int J Cancer 2002 98 690 697 11920637 10.1002/ijc.10265
Boudreau N Sympson CJ Werb Z Bissell MJ Suppression of ICE and apoptosis in mammary epithelial cells by extracellular matrix Science 1995 267 891 893 7531366
Tezel TH Del Priore LV Reattachment to a substrate prevents apoptosis of human retinal pigment epithelium Graefes Arch Clin Exp Ophthalmol 1997 235 41 47 9034841 10.1007/BF01007836
Akiyama SK Nagata K Yamada KM Cell surface receptors for extracellular matrix components Biochim Biophys Acta 1990 1031 91 110 1689589
Sastry SK Horwitz AF Integrin cytoplasmic domains: mediators of cytoskeletal linkages and extra- and intracellular initiated transmembrane signaling Curr Opin Cell Biol 1993 5 819 831 8240826 10.1016/0955-0674(93)90031-K
Leavesley DI Schwartz MA Rosenfeld M Cheresh DA Integrin beta 1- and beta 3-mediated endothelial cell migration is triggered through distinct signaling mechanisms J Cell Biol 1993 121 163 170 7681432 10.1083/jcb.121.1.163
Kohn EC Felder CC Jacobs W Holmes KA Day A Freer R Liotta LA Structure-function analysis of signal and growth inhibition by carboxyamido-triazole, CAI Cancer Res 1994 54 935 942 8313384
Friedlander M Brooks PC Shaffer RW Kincaid CM Varner JA Cheresh DA Definition of two angiogenic pathways by distinct alpha v integrins Science 1995 270 1500 1502 7491498
Brooks PC Klemke RL Schon S Lewis JM Schwartz MA Cheresh DA Insulin-like growth factor receptor cooperates with integrin alpha v beta 5 to promote tumor cell dissemination in vivo J Clin Invest 1997 99 1390 1398 9077549
Hinton DR He S Graf K Yang D Hsueh WA Ryan SJ Law RE Mitogen-activated protein kinase activation mediates PDGF-directed migration of RPE cells Exp Cell Res 1998 239 11 15 9511719 10.1006/excr.1997.3873
|
15987521
|
PMC1184086
|
CC BY
|
2021-01-04 16:03:50
|
no
|
BMC Ophthalmol. 2005 Jun 29; 5:16
|
utf-8
|
BMC Ophthalmol
| 2,005 |
10.1186/1471-2415-5-16
|
oa_comm
|
==== Front
BMC Public HealthBMC Public Health1471-2458BioMed Central London 1471-2458-5-721598516010.1186/1471-2458-5-72Study ProtocolHousehold cost-benefit equations and sustainable universal childhood immunisation: a randomised cluster controlled trial in south Pakistan [ISRCTN12421731] Andersson Neil [email protected] Anne [email protected] Noor [email protected] Khalid [email protected] Joe [email protected] Robert J [email protected] Peter [email protected] Beverley [email protected] Centro de Investigación de Enfermedades Tropicales (CIET), Universidad Autónoma de Guerrero, Acapulco, México2 Community Information and Epidemiological Technology (CIETcanada), Institute of Population Health, 1 Stewart Street, Ottawa, Outario, K1N GN5, Canada3 CIET in Pakistan, Islamabad, Pakistan4 Faculty of Medicine, Institute of Population Health, Ottawa, Canada5 CIET international, New York, USA2005 28 6 2005 5 72 72 12 5 2005 28 6 2005 Copyright © 2005 Andersson et al; licensee BioMed Central Ltd.2005Andersson et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Household decision-makers decide about service use based largely on the costs and perceived benefits of health interventions. Very often this leads to different decisions than those imagined by health planners, resulting in under-utilisation of public services like immunisation. In the case of Lasbela district in the south of Pakistan, only one in every ten children is immunised despite free immunisation offers by government health services.
Methods/design
In 32 communities representative of Lasbela district, 3344 households participated in a baseline survey on early child health. In the 18 randomly selected intervention communities, we will stimulate discussions on the household cost-benefit equation, as measured in the baseline. The reference (control) communities will also participate in the three annual follow-up surveys, feedback of the general survey results and the usual health promotion activities relating to immunisation, but without focussed discussion on the household cost-benefit equations.
Discussion
This project proposes knowledge translation as a two-way communication that can be augmented by local and international evidence. We will document cultural and contextual barriers to immunisation in the context of household cost-benefit equations. The project makes this information accessible to health managers, and reciprocally, makes information on immunisation effects and side effects available to communities. We will measure the impact of this two-way knowledge translation on immunisation uptake.
==== Body
Background
Despite billions of dollars spent on childhood immunisation, some countries have never reached universal childhood immunisation (UCI), and many more have been unable to sustain it. An estimated 1.5 million deaths under 5 years of age can be prevented by vaccination each year, measles making up 37% H. influenzae b-related disease 30%, and pertussis 19%[1]. In many countries, there is dramatic underutilisation of the offer of free immunisation. One reason for this is a difference between understanding of the costs and benefits of vaccination in international public health circles, and what primary decision-makers for children know about immunisation and its costs.
There is a gap between the way public health specialists understand immunisation benefits, and the cost-benefit equations that household decision-makers apply to their children's immunisation. In settings like Pakistan, where an expanded program of immunisation is offered free of charge, uptake is largely determined by access to services and the attendant cost-benefit assessments by parents and caregivers. These assessments could be influenced by access to knowledge and conditioned by gender and social inequalities.
Although communication of risks and benefits by service providers can influence health-seeking behaviour [2], current approaches to health communication do not always achieve the expected results. Efforts frequently produce an increase in knowledge without a corresponding change in attitudes or behaviour[3]. This is at least partly because conventional risk communication presumes to inform an uninformed public and to reduce irrational thinking. Questioning the value of these conventional one-way knowledge transfer (KT) initiatives, more holistic perspectives take account of social and cultural influences[4]. Impact studies of communication strategies to increase vaccination in the USA, Russia and Mozambique all highlight the need of multi-channel targeting of multiple groups – families, communities, health practitioners and opinion makers, such as community and religious leaders[5-7].
Current devolution reforms in Pakistan are expected to make service delivery more effective and to support institutionalised participation of community members[8,9]. Service delivery is now a district function, yet the provinces are still responsible for planning and monitoring health services [9]. The expanded programme of immunisation (EPI) remains a federal responsibility. The 2002 Pakistan Integrated Household Survey claimed an increase in immunisation coverage, mostly in the urban areas. The proportion of fully immunised children aged 12–23 months rose from 49% to 53%, and even this small gain was not consistent countrywide. In Balochistan, where the project will be located, immunisation coverage fell from 34% in 1998–9 to 24% in 2000–1, mostly attributed to diversion of resources to the polio eradication campaigns[10].
Balochistan is Pakistan's largest province (347,000 km2) and has the lowest population density (6.5 million, 19/km2). It covers more than 40% of the land area, but has less than 5% of the national population. The disease burden in Pakistan is still largely poverty-related, much of it preventable by immunisation[11,12].
Lasbella, in the south of Balochistan is one of the province's poorest districts. In 2003 there were around 50,000 women of childbearing age in this district. There were 11,594 infants under the age of one year, and reflecting the very high infant mortality, 8,140 children aged 12–23 months. The new district administration employs 39 full time vaccinators, but still only achieves symbolic coverage. The problem does not appear to be one of supervision or management. The vaccinators report to a superintendent, who in turn reports to the Deputy District Health Officer, who reports to the Executive District Officer for Health.
Statement of the research issue and approach
We believe that health care consumers make rational decisions from a cost-benefit perspective[13,14]. Based on their own knowledge, mothers and guardians weigh up the costs and the benefits of immunisation: how much time will it take, how much will it cost, will it work, will there be side effects and what will happen if I do not go. We propose that immunisation uptake is largely determined by this cost-benefit equation.
But the rationality of such decisions is a "bounded rationality" [15]. This concept enables health planners to see such factors as social norms and attitudes not as obstacles to rationality but rather mechanisms that facilitate fast and accurate decisions, and more efficient learning. In addition, the weigh-up of risks may be sharply discounted between the time of the immunization and the time of possible infection. Also, gender and poverty probably affect the household cost-benefit equation. The poor, who typically have less access to services and less information about services, almost certainly weigh up the costs and benefits in a different way than do the rich [16]. Diseases like measles and pertussis may be an inconvenience for the well-nourished, whereas for the malnourished, they can be a question of life or death. Costs of not vaccinating (disease burden, care and funerals) are borne disproportionately by the poor; in a single epidemic, these diseases can destroy a household economy[17].
The household cost-benefit equation is, therefore, a lens through which to view immunisation and the obstacles to immunisation
A focus on these equations gives value to the way ordinary people see immunisation, allowing their views to be taken systematically into account. It gives primacy to household decisions, and recognises that communication about immunisation is a two-way street. We propose to test the importance of this household cost-benefit equation that decision-makers for children derive from their knowledge, attitudes, social norms, intentions, sense of agency and degree of socialisation about immunisation.
The hypothesis is that this dynamic equation can be influenced by two-way knowledge translation (KT), and based on this culture-appropriate exchange, that people will adjust their household cost-benefit equations and their uptake of immunisation. A corollary of the household cost-benefit equation is accessible to planners and health service managers: cost-gains. Derived from the same data used by communities for their cost-benefit equations, cost-gains offers a common language for interaction between health services and communities. If local health managers understand how primary decision makers assess immunisation – their cost-benefit equations – this can improve their engagement with communities and lead, in turn, to sustainable UCI.
Methods
Objective 1. Identify the barriers and information imbalances that reduce childhood immunisation, in particular increasing understanding of the household cost-benefit equations underlying uptake of immunisation
Randomised controlled cluster trials are fairly widely used in developed countries, and have been introduced in several developing countries[18-20].
Randomisation
From the latest Pakistan census, 32 enumeration areas were randomly selected to represent the population of Lasbela district. Once the baseline survey was completed, clusters were randomised into intervention and control, with precautions about the usual biases of this design[21]. In each cluster, interviewers contacted homes of 100 children under the age of 24 months of age (a total of 3344 households). Three successive cycles that examine successive cohorts of children in this age group, and the same number of households in each site (not necessarily the same households) will preserve the proportional representation.
Design of survey instruments
Questionnaires were adapted from international EPI standards. Additional fact-finding tools will produce qualitative evidence – key informant interviews, service worker questionnaires, protocols for institutional reviews and focus group discussions. The results of each round might identify additional stakeholder-driven issues and priorities and be used to refocus the subsequent survey cycles.
Survey content
In addition to baseline data about the coverage with and obstacles to immunisation, we enriched the standard KAP approach[22] with a behaviour change model adapted by CIET to measure youth responses to risk. The beyond-KAP approach, "cascada", refers to conscious knowledge about immunisation and its side effects, attitudes to childhood immunisation, social norms (what neighbours do) and positive or negative deviation from those norms, intentions to change or to vaccinate in the future, agency (expectancy of self-efficacy or collective efficacy) and discussion about immunisation, its benefits and side effects. The outcome of this "cascada" is the action, immunisation. We will document perceived and real costs of immunisation and non-immunisation, and the household weigh up of costs and benefits.
Piloting
Eight rounds of piloting in non-sample sites included testing new sections of the instrument, testing the instrument for flow, and then testing the instrument as a whole in order to finalise the process. The pilot exercises assisted in refining the instruments, testing for clarity and ensuring proper translation.
Ethical review
Two review panels, one at the University of Ottawa and a panel in the south of Pakistan registered with the US Government's Office of Human Research Protections, deliberated the ethical issues and approved the study.
Objective 2. Formulate and implement knowledge transfer based on household cost-benefit equations, compared with health information in reference (control) communities
Intervention
We will update and translate available knowledge on immunisation and combine this with data from the baseline as an intervention focussed on the household cost-benefit equation. Mass media channels can increase awareness and knowledge, but interpersonal channels seem to work better in changing attitudes and behaviour[23,24]. A combination of communication channels could, therefore, include mass media appeals, reminder systems and engagement through trusted sources, and addressing risks and benefits of vaccination in an understandable manner[25-28]. Upon consideration of the evidence, people will hopefully adjust their household cost-benefit equations. We will measure this adjustment in the subsequent cycles.
Analysis
We will estimate the impact of this knowledge transfer on changing beliefs and practices of decision-makers for children and, as a consequence, immunisation uptake. Differences between intervention and reference sites will be analysed for independence from age, sex, household employment, community size, remoteness and other factors. Risk analysis will rely on the Mantel-Haenszel procedure and contrasts reported as the odds ratio (OR) or risk difference (RD)[29-31]. In the analysis, each site or cluster is treated as a mini-universe characterised by certain social dynamics, history, culture and collective practices. Because qualitative data are coterminous (they coincide with the same population) with the individual questionnaires it is easy to link quantitative and qualitative data[32]. An unconditional logistic regression model will be developed where appropriate, using a step-down approach from a saturated model[33].
We also seek to express a planning-appropriate perspective that fits the household cost-benefit perspective. 'Gain' is the theoretical proportion of the entire population that stands to benefit from the removal of an obstacle or the universalisation of an intervention. It is calculated by multiplying the risk difference (risk among exposed minus risk among unexposed) by the proportion requiring intervention (those exposed, if the interest is to remove risk factors).
Objective 3. Measure the impact of the KT on coverage and attitudes about immunisation
Focus groups to identify strategies
After the preliminary analysis, field teams will return to the communities to hold gender stratified focus groups. Key findings from the household interviews will be shared with the groups to generate additional insights, including how best to let similar communities know about the findings. Focus groups typically involve 8–12 participants. In a quiet location, groups are limited to a maximum of one hour in recognition of the value of participants' time. Participants are reassured about confidentiality (no identifiers are recorded). A trained facilitator runs the group, prompts to provoke discussion, and encourages participants to express opinions. A second member of the team records the content and manages the time.
Communications
Communication strategies (see below) to share the outcomes of the measurement with stakeholders in the intervention communities will be heavily conditioned by the outcome of the focus groups. The core concept is to socialise the household cost-benefit equation. Strategies are likely to include work with local elected representatives, community and religious leaders, service workers and community action groups such as citizen community boards (CCBs)[34].
Repeat survey
Each year for three years, the measurement will be repeated in both intervention and control sites. Much of the key instrument content will be unchanged, to detect time trends. Responses will provide substrate for the next round of household cost-benefit equations and cost-gain analysis, which in turn will feed into the next round of intervention.
Objective 4. Develop an evidence-based and gendered systems approach to increasing equity in immunisation, rooted in community knowledge, capable of building on local health protection cultures and of informing evidence-based decision-making to improve the health of populations and strengthen health systems through immunisation services
A lot of effort has gone into the supply side of immunisation (vaccine purchase, training health workers and logistics). There is a need to focus systematically on the demand for immunisation – which is probably a direct function of the quality of information people have.
Parallel to the community-based knowledge transfer intervention, the team will work with the district authorities in Lasbela. We will build capacity to improve immunisation rates in the selected district, reaching health care workers, community leaders and policy makers. Research teams will be trained in community-based research, enhancing the capacity for ongoing monitoring of immunisation and other key public services.
The cost-gains approach offers a bridge between planner and community views. Proving the value of this parameter could support a paradigm shift in resource allocation, from a system based on reconciliation of competing sectoral claims without a comparable evidence base to cost-gain planning.
The main selling point of a new knowledge transfer approach to sustainable immunisation is that it must work. immunisation coverage must increase measurably and people in key positions must know about this. Mainstreaming the household cost-benefit equation begins with respectful dialogue with local health and political authorities about immunisation concepts, service delivery, effects and side effects in the communities. The central activity is then to demonstrate by measuring and communicating, in reiterative cycles, the effect on cost and benefit assessments. Comparisons between sites with different levels or types of intervention will be almost as informative as the longitudinal picture emerging by following each site over four years. Evaluation is thus built into the project.
In addition to the overall concept of evidence-based immunisation support, and the household cost-benefit equation, relevant procedures and tools include:
1. Protocols for the survey (household beyond-KAP, key informants and institutional review) and sample selection, processing results through double data entry, epidemiological analysis and interpretation in community-based focus groups;
2. Procedures for achieving policy level buy-in, including aggregation tools (like customised epidemiological mapping freeware) that allow compounding of local experiences into regional and national pictures; a menu of methods and practical examples of communication tools for opening evidence-based dialogue that can increase community ownership of immunisation.
Discussion
Engagement of communities in evidence-based planning is a novel approach to immunisation, an intervention driven perhaps more than any other by an inappropriate if well-intentioned ethos of "we know what is good for you". The idea of a household cost-benefit equation is common sense: people weigh things up before their health choices. This study hopes to answer questions about what it takes to enter a dialogue that influences these equations. The year-to-year shift in knowledge, attitudes, subjective norms, intention, sense of agency, ability to discuss, and ultimately, uptake of immunisation will be evident from the results of the successive cycles.
Best practice cases – communities that increase immunisation – will be identified and held up as positive examples of what is possible under prevailing conditions. The direct beneficiaries will be the public, but indirect beneficiaries will be planners and policy-makers in federal, provincial and district governments. Improved understanding of immunisation risks and cost implications of higher immunisation uptake may be applicable in many other countries. Proof of cost-gains as the planners' corollary of the household cost-benefit equation could open new horizons for evidence-based service-public interaction.
Competing interests
The author(s) declare that they have no competing interests.
Contributions of authors
NA designed the study, developed the methodology and wrote the proposal. NA designed CIET map freeware, included in the proposal to combine mapping and epidemiology for decision makers. He is responsible for overseeing the project.
AC reviewed the protocol and is responsible for overseeing the project in Pakistan.
NoorA conducted the pilot studies, contributed to the instrument design and supervises the fieldwork in Balochistan.
KS coordinates the project in Pakistan.
JL reviewed the protocol and contributed to the overall study plan.
RJL co-developed the concept of household cost-benefit equations, and contributed to the proposal and development of instruments.
PT reviewed the protocol and will contribute to the systematic review.
BS helped write the proposal and is responsible for the systematic review.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This initiative was funded by the Canadian International Development Agency in collaboration with the International Development Research Centre, the Canadian Institutes of Health Research and Health Canada (File: 102172-007) through an open competition of Research Initiative Operational Research Grants for the Canadian International immunisation Initiative (CIII2).
==== Refs
WHO State of the World's Vaccines and immunisation
Bennett P Calman K Risk Communication and Public Health 2001 London: Oxford University Press
Committee on Risk Perception and Communication, National Research Council Improving risk communication 1989 Washington, DC
Leask J Vaccination and risk communication: Summary of a workshop J Paediatr Child Health 2000 38 124 128 12030991 10.1046/j.1440-1754.2002.00791.x
Division of Health Promotion and Disease Prevention, Institute of Medicine Overcoming Barriers to immunisation: a Workshop Summary 1994 Washington, DC
Porter R Steinglass R Kaiser J Olkhovsky P Rasmuson M Dzhatdoeva FA Fishman B Bragina V Role of Health Communications in Russia's Diphtheria immunisation Program J Infect Dis 2000 220 227 10.1086/315566
UNICEF, WHO/AFRO, USAID Social Mobilization and Communication Support for immunisation in Mozambique: A Joint Lessons Learned Study byUNICEF, WHO/AFRO, and USAID Mozambique 1999
National Reconstruction Bureau (NRB), Government of Pakistan Local Government Plan Islamabad 2000
Government of Pakistan The SBNP Local Government Ordinance 2001 Islamabad 2001
Federal Bureau of Statistics, Government of Pakistan Pakistan Integrated Household Survey (PIHS) Round 4: 2001–02 Islamabad 2002
Ministry of Finance, Government of Pakistan Accelerating growth and reducing poverty (Poverty Reduction Strategy Paper) Islamabad 2003
UNDP in Pakistan Pakistan National Human Development Report 2003: poverty, growth and governance Islamabad 2003
Kapiriri L Arnesen T Norheim OF Is cost-effectiveness analysis preferred to severity of disease as the main guiding principle in priority setting in resource poor settings? The case of Uganda Cost Eff Resour Alloc 2004 2 1 14711374 10.1186/1478-7547-2-1
Brock DW Separate spheres and indirect benefits Cost Eff Resour Alloc 2003 1 4 23 12773217 10.1186/1478-7547-1-4
Gigerenzer G Selten R Gigerenzer G, Selten R Rethinking Rationality Bounded Rationality: The Adaptive Toolbox 2001 Cambridge Massachusetts, MIT Press 1 12
The World Bank Group immunisation coverage inequalities: An Overview of Socio-Economic and Gender Differentials in Developing Countries 2001 Washington, DC
Andersson N Ledogar RJ Paredes S Who pays for measles? The economic arguments for sustained immunisation Health Policy Plan 1992 7 352 363
Graham A Moore L Sharp D Diamond I Improving teenagers' knowledge of emergency contraception: cluster randomised controlled trial of a teacher led intervention BMJ 2002 324 1179 1183 12016180 10.1136/bmj.324.7347.1179
Moher M Yudkin P Wright L Turner R Fuller A Schofield T Mant D Cluster randomised controlled trial to compare three methods of promoting secondary prevention of coronary heart disease in primary care BMJ 2001 322 1338 1342 11387182 10.1136/bmj.322.7298.1338
Kroeger A Villegas Avila Morison L Insecticide impregnated curtains to control domestic transmission of cutaneous leishmaniasis in Venezuela: cluster randomised trial BMJ 2002 325 810 813 12376442 10.1136/bmj.325.7368.810
Puffer S Torgerson D Watson J Evidence for risk of bias in cluster randomised trials: review of recent trials published in three general medical journals BMJ 2003 327 785 789 14525877 10.1136/bmj.327.7418.785
Conner M Norman P Conner M, Norman P The role of social cognition in health behaviours Predicting health behaviour 2001 1 Philadelphia: Open University Press 1 22
Rogers EM Diffusion of innovations 1983 3 New York: The Free Press
Wallack L Atkin C, Wallack L Mass media and health promotion: Promise, Problem, and Challenge Mass communication and public health: Complexities and Conflicts 1990 Newbury Park (CA): Sage Publication 52 59
Grilli R Freemantle N Minozzi S Domenighetti G Finer D Mass media interventions: effects on health services utilisation Cochrane Database Syst Rev 2000 2 CD000389 10796539
Szilagyi P Vann J Bordley C Chelminski A Kraus R Margolis P Rodewald L Effect of patient reminder/recall interventions on immunisation rates: A review JAMA 2000 284 1820 1827 11025835 10.1001/jama.284.14.1820
Freimuth V Linnan HW Potter P Communicating the Threat of Emerging Infections to the Public Emerg Infect Dis 2000 6 337 47 10905966
Atkinson W Pickering L Schwartz B Weniger B Iskander J Watson J General Recommendations on immunisation: Recommendations of the Advisory Committee on immunisation Practices (ACIP) and the American Academy of Family Physicians (AAFP) MMWR Recomm Rep 2002 51 1 36 11848294
Mantel N Haenszel W Statistical aspects of the analysis of data from retrospective studies of disease J Natl Cancer Inst 1959 22 719 748 13655060
Mantel N Chi-square tests with one degree of freedom: extensions of the Mantel Haenszel procedure J Am Stat Assoc 1963 58 690 700
Miettinen OS Estimability and estimation in case-referent studies Am J Epidemiol 1976 103 226 235 1251836
Andersson N Andersson N Mesoanalysis: Quantifying Qualitative Data from Communities and Services Evidence-Based Planning: the Philosophy and Methods of Sentinel Community Serveillance 1996 Washington DC: EDI/ World Bank 51 65
Hosmer DW Lemeshow S Applied logistic regression 1989 New York: John Wiley & Sons
National Reconstruction Bureau (NRB), Government of Pakistan Guidelines for Citizen Community Boards Islamabad 2002
|
15985160
|
PMC1184087
|
CC BY
|
2021-01-04 16:28:57
|
no
|
BMC Public Health. 2005 Jun 28; 5:72
|
utf-8
|
BMC Public Health
| 2,005 |
10.1186/1471-2458-5-72
|
oa_comm
|
==== Front
BMC PsychiatryBMC Psychiatry1471-244XBioMed Central London 1471-244X-5-291604281610.1186/1471-244X-5-29Research ArticleThe influence of methylphenidate on the power spectrum of ADHD children – an MEG study Wienbruch Christian [email protected] Isabella [email protected] Susanne [email protected] Hermann [email protected] Clinical Psychology, University of Konstanz, Konstanz, Germany2 Pratice for Child and Adolescent Psychiatry and Psychotherapy, Ettlingen, Germany2005 26 7 2005 5 29 29 21 4 2005 26 7 2005 Copyright © 2005 Wienbruch et al; licensee BioMed Central Ltd.2005Wienbruch et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The present study was dedicated to investigate the influence of Methylphenidate (MPH) on cortical processing of children who were diagnosed with different subtypes of Attention Deficit Hyperactivity Disorder (ADHD). As all of the previous studies investigating power differences in different frequency bands have been using EEG, mostly with a relatively small number of electrodes our aim was to obtain new aspects using high density magnetoencephalography (MEG).
Methods
35 children (6 female, 29 male) participated in this study. Mean age was 11.7 years (± 1.92 years). 17 children were diagnosed of having an Attention-Deficit/Hyperactivity Disorder of the combined type (ADHDcom, DSM IV code 314.01); the other 18 were diagnosed for ADHD of the predominantly inattentive type (ADHDin, DSM IV code 314.0). We measured the MEG during a 5 minute resting period with a 148-channel magnetometer system (MAGNES™ 2500 WH, 4D Neuroimaging, San Diego, USA). Power values were averaged for 5 bands: Delta (D, 1.5–3.5 Hz), Theta (T, 3.5–7.5 Hz), Alpha (A, 7.5–12.5 Hz), Beta (B, 12.5–25 Hz) and Global (GL, 1.5–25 Hz).). Additionally, attention was measured behaviourally using the D2 test of attention with and without medication.
Results
The global power of the frequency band from 1.5 to 25 Hz increased with MPH. Relative Theta was found to be higher in the left hemisphere after administration of MPH than before. A positive correlation was found between D2 test improvement and MPH-induced power changes in the Theta band over the left frontal region. A linear regression was computed and confirmed that the larger the improvement in D2 test performance, the larger the increase in Theta after MPH application.
Conclusion
Main effects induced by medication were found in frontal regions. Theta band activity increased over the left hemisphere after MPH application. This finding contradicts EEG results of several groups who found lower levels of Theta power after MPH application. As relative Theta correlates with D2 test improvement we conclude that MEG provide complementary and therefore important new insights to ADHD.
==== Body
Background
Attention Deficit Hyperactivity Disorder (ADHD) is characterized by difficulties concentrating, completing assigned tasks, keeping track of things, waiting one's turn or sitting still. Three subtypes are classified in the DSM IV [1] ADHD of the predominantly inattentive type, ADHD of the predominantly hyperactive type and a combined type. The prevalence of ADHD is estimated to lie between 3 and 5% of all school children with a stronger tendency for boys to be diagnosed [1] However, Scahill and Schwab-Stone [2] investigated data from 13 studies and found prevalence to vary between 2 and 14.9%, depending on diagnostic tools and community sample. An increase in prevalence has been observed throughout the last years, which might be related to a change in diagnostic criteria and the introduction of ADHD predominantly hyperactive type in the DSM IV. All ADHD subtypes are generally treated the same way, the prescription of Methylphenidate (MPH) [3]. MPH has shown to be effective in 75–90% of ADHD children [4]. In line with increasing prevalence estimates, the usage of MPH has increased several fold during the last years in the USA [5] as well as in Germany [6].
It has been stated by several authors, that ADHD is related to cortical hypoarousal [7-10]. The mechanism behind this possible hypoarousal is not yet clarified. However, evidence from SPECT studies [11,12] and the mere fact that MPH – a psychostimulant – is an effective treatment of ADHD symptoms suggest there is a deficit in the dopaminergic neurotransmitter system. As SPECT studies (e.g. [13]) have shown, ADHD patients seem to have a higher number of DAT receptors, which are responsible for dopamine re-uptake, and in consequence have less dopamine available in the synaptic gap. MPH is a potent blocker of DAT receptors.
It has previously been shown that some children respond more to MPH than others, or that children with different sub-diagnoses react different. Clarke and colleagues [14,15] compared EEG power in different frequency bands of good and poor responders to MPH and found a cortical activation profile suggesting that good responders are more cortically hypoaroused than poor responders. The authors assume that MPH is most effective for children who are cortically hypoaroused. Loo and co-workers [16] conclude from their results that there are different electrophysiological correlates to MPH for good responders and poor responders. They also compared EEG power in different frequency bands and found that reponders showed a decrease in Theta and Alpha activity, as well as an increase in Beta activity, while poor responders showed the opposite pattern. Clarke and colleagues [17-19] investigated cortical differences between ADHD children of the combined type (ADHDcom) and the predominantly inattentive type (ADHDin). Generally, they found ADHDcom children having higher slow wave activity than ADHDin children. The authors concluded from their results that children with ADHDcom are more cortically hypoaroused than children with ADHDin. They hypothesized that ADHDcom might be related to frontal lobe dysfunctions, while children with ADHDin may have other forms of CNS functioning.
Due to the findings that MPH is more effective for some children than for others and since the cortical profiles of ADHD subtypes seem to differ, it is advisable to study the effects of MPH on cortical processing more closely. One desirable outcome of these studies might be to identify cortical indicators in order to differentiate between children who are good responders and those who are poor responders before they are treated with amphetamines. Another value lies in elucidating etiological factors of ADHD. Differential effects for the different subtypes can help understanding the underlying cause of the disorder.
The present study was dedicated to investigate the influence of MPH on cortical processing. All of the previous studies investigating power differences in different frequency bands have been using EEG, mostly with a relatively small number of electrodes. Our aim was to obtain new aspects using high density magnetoencephalography (MEG). MEG comprises several advantages over EEG. First of all, the magnetic fields measured are not as biased by low skull conductivity as electrical potentials. Second, MEG is reference-free. Unless EEG analysis is done using average reference (which only is reliable if the recording of the reference electrode is flawless), there will always be an influence on cortical effects produced by the reference type chosen. Third, MEG using magnetometers by definition mostly reflects cortical activity. Subcortical activity is often too weak to be detected. Thus, the complexity of the detected signals is reduced. Fourth, MEG mainly reflects cortical activity from structures, which have a tangential orientation to the surface of the head. Thus, the activity measured most likely stems from circumscribed cortical structures in the walls of the gyri and sulci, whereas potential differences measured by EEG can originate from both radial and tangential oriented fibers from the whole brain. Thus it is rather likely that MEG shows complementary but similar information than EEG.
Methods
Subjects
35 children (6 female, 29 male) participated in this study. Mean age was 11.7 years (± 1.92 years). 17 children were diagnosed of having an Attention-Deficit/Hyperactivity Disorder of the combined type (ADHDcom, DSM IV code 314.01); the other 18 were diagnosed for ADHD of the predominantly inattentive type (ADHDin, DSM IV code 314.0). Diagnoses were made by a paediatrician specialized in child psychiatry. All children and parents gave their written informed consent to participate according to the World Medical Association Declaration of Helsinki – Ethical Principles for Medical Research Involving Human Subjects [20].
Procedure
We measured the effect of Methylphenidate (MPH) on the MEG during a 5 minute resting period (subjects being relaxed but awake). The behavioural performance with and without medication was measured by a highly demanding attention test (D2 test of attention, [21]). Dosage of methylphenidate was based on the body weight of the child (0.1–0.5 mg/kg/day). To ensure that medication and not the mere administration of a pill had an effect, we chose a placebo design. Placebo and methylphenidate were applied by a pediatrician in form of pills that looked identical (placebo dosage was matched with MPH dosage).
The overall-design was the following (Fig. 1): due to feasibility it was decided to run the whole procedure within one day. Therefore it was not possible to counterbalance the application time of placebo and MPH, since MPH takes several days to be untraceable in the blood. In order to have an objective measure of the concentration of MPH in the blood serum, blood samples were taken from the children an hour after drug administration. The blood serum was separated right after being taken and was then deep frozen. MPH serum concentrations were measured in an external professional laboratory.
Figure 1 Study design.
MEG recording
Recording was done with a 148-channel magnetometer (MAGNES™ 2500 WH, 4D Neuroimaging, San Diego, USA). A subject specific headframe coordinate system was defined by means of 3 anatomical landmarks also called 'head fiducials' (left and right preauricular and nasion). These head fiducials, five coils and the subject's head shape were digitized with a Polhemus 3Space® Fasttrack prior to each measurement. The subject's head position relative to the pickup coils of the sensor was estimated before and after each measurement to ensure that no large movements occurred during the measurement. As the position of the pickup coils are known in device coordinates, this procedure also allows their transformation into the headframe coordinate system.
The children were lying supine in a comfortable position in a magnetically shielded room (Vakuumschmelze Hanau). They were instructed to lie still for 5 minutes and to fixate a point at the ceiling in order to keep eye movements minimal. Continuous data sets were recorded with a real high-pass filter of 0.1 Hz and a sampling rate of 678.17 Hz (bandwidth 200 Hz). Real time noise reduction procedures, i.e. the subtraction of the signal measured by 8 reference channels, multiplied with sets of weight factors for each of the 148 magnetometers, were applied during acquisition before and directly after analog/digital (AD) conversion. This noise reduction procedure affects signals of interest originating in the brain or the body much less than the presence of a cancellation coil in a standard gradiometer detector.
For artifact control, eye movements (EOG) were recorded from four electrodes attached to the left and right outer canthus and above and below the right eye. A Synamps amplifier (NEUROSCAN™) served for the recording of the EOG. A video camera installed inside the chamber allowed monitoring the children's behaviour and compliance at any time throughout the experiment.
D2 test
Immediately after the MEG recording, each child performed the D2 test in a quiet room. The test involves finding and marking the letter "d" within a string of letters ("d" and "p"), only when 2 dashes are arranged either individually or in pairs above and below "d". A high amount of attention is necessary to perform the task successfully, since not only the letter "d" is orthographically similar to the letter "p", but because there are many distractor letters "d" with more than 2 dashes. Additionally, a time limit is set for finding as many D2s with as little errors as possible.
Data analysis MEG
Global noise was filtered offline from the MEG data by subtracting external, non-biological noise recorded by 11 MEG reference channels. Before subtraction, reference channels were multiplied with individually calculated fixed weight factors. Again, noise reduction procedures have no or little influence on the biological signal because the distance of the reference set to the subjects head is relatively large (mean = 25.8 cm, std = 6.00 cm, min = 15.5 cm, max = 36.5 cm) compared to the distance between sensors and head, which is usually much smaller. The data was then split into epochs of 2500 ms length and was corrected for eye and cardiac artefacts by subtracting the moving average cardiac and vertical EOG signal from the data. All epochs with an MEG level > 3.5 pT between the minimum and maximum on one or more MEG channels after artefact correction were rejected. A fast fourier transformation (FFT) was computed for all epochs.
For each subject the average power was calculated across channels for 6 cortical regions (frontal, temporal and occipital; left and right, respectively). In EEG, electrode positions are comparable across subjects, since they are usually defined in relation to fixed anatomical landmarks on the head. This is not true for MEG sensor coils. Therefore averages of MEG measures of the same coil across subjects add additional variance. This variance is closely related to the variance of sensor coil positions between subjects. In order to ensure that the same cortical regions were covered, subject specific channel-groups were selected. We defined 6 landmarks in the headframe-based coordinate system. In the second step, we determined 6 corresponding channels for each subject and each measurement that were closest (smallest Euklidean distance) to the previously defined landmarks. These 6 channels served as centre channels of our subject specific channel groups, consisting of either 15 (occipital) or 20 (frontal, temporal) channels. Channels were selected by being nearest neighbours to the centre channel of the respective channel group. Within these channel-groups the power values were averaged for 5 bands (see Fig. 2) and normalized to the size of the frequency bin: Delta (D, 1.5–3.5 Hz), Theta (T, 3.5–7.5 Hz), Alpha (A, 7.5–12.5 Hz), Beta (B, 12.5–25 Hz) and Global (GL, 1.5–25 Hz). The power values of the Delta, Theta, Alpha, and Beta frequency bands were normalized to the global power yielding relative power values. Additionally, T/A and T/B ratios were calculated.
Figure 2 Power band definition.
Data analysis D2 test
The total number of correctly marked items was used to determine the individual child's attention level. Raw values were expressed in percentiles (derived from age-matched norm samples), in order to achieve age-independent test scores. Improvement of attention was determined by subtracting the test score after placebo application from the test score after MPH application. Further, the subjects were divided into good responders (ADHDg) and poor responders (ADHDp): a child was classified as ADHDg, if the improvement was larger than 30 percentiles (this was slightly more than 1 standard deviation).
Statistical analysis
To see if D2 test performance improved after medication, a one-way repeated measure ANOVA (analysis of variance) was calculated. D2 test score was dependent variable, TIME (pre, post) was repeated measure.
To quantify the influence of medication on the power bands in the different cortical regions, a mixed model analysis was computed using the statistical package SAS®9. Covariance parameters were estimated with the restricted maximum likelihood method (REML). Relative power values (D, T, A, B, T/A, T/B) as well as the global power (GL), were defined as dependent variables. TIME (pre, post), HEMISPHERE (left, right), REGION (fontal, temporal, occipital) and either DSMtype (combined, inattentive) or Response (responders, non-responders) were fixed effects. Depending on which fixed effect was used in the analysis, the factor Patient either nested in DSMtype or Response was used as random factor. Variance structure was variance components (VC). Post-hoc testing was performed following Tukey-Kramer. In cases of non-significant post-hoc tests, uncorrected p-values are reported.
Finally, all power values (all bands and all regions) after placebo were subtracted from the respective power values after MPH application. This gave us a measure of MPH induced power changes. Correlations were calculated between the power changes, MPH blood serum concentration, age in months and D2 test improvement.
Only significant main effects, interactions and post-hoc tests are reported. For the mixed model analysis, only significant interactions with D2-response or DSM-type are reported, since the aim was to reveal cortical differences between the different DSM-subtypes and children, who respond well to MPH compared to those who do not profit as much. All plots show standard errors.
Results
D2-test
A main effect for TIME was found (F(1,34) = 86.87, p < 0.001). D2-test performance was significantly higher after the application of MPH (73.1 percentiles) than before MPH (41.2 percentiles).
In order to investigate, which cortical region would be most affected by MPH application, an analysis over all ADHD children was performed, no matter what subtype they were diagnosed. The analysis yielded the significant interaction TIME*REGION (F(1,33) = 4.46, p = 0.015) for the dependent variable Global Power. As can be seen in figures 3 and 4, MPH effects were only found in frontal regions (p = 0.05) with higher amplitudes after MPH than before MPH. No differences were observed in temporal or occipital regions. Thus, further analysis was restricted to frontal channel groups.
Figure 3 Difference of global power – MPH minus Placebo. The difference of the grand average global and relative power (delta, theta, alpha, beta, and gamma) between the MPH condition and the placebo condition calculated over the 35 subjects. It can clearly be seen that the global power is strongest over frontal regions (global power is given in ft/vHz. Delta, Theta, Alpha, Beta and Gamma are relative measures, normalized to the Global power. Theta/Alpha and Theta/Beta are ratios of the relative power, which is identical to ratios of the power).
Figure 4 Interaction TIME*REGION for global power. MPH effects were only found in frontal regions (p = 0.05) with higher amplitudes after MPH than before MPH. No differences were observed in temporal or occipital regions.
Results frontal channels
Global power
The main effect TIME (F(1,33) = 7.53, p = 0.0098) was found. Global power amplitude was higher after MPH (23.9 ft/Hz^/2) than before MPH (22.7 ft/Hz^/2).
The interaction DSMtype*HEMISPHERE (F(1,33) = 7.79, p = 0.009) was revealed. Figure 5 displays that ADHDin children showed a hemispheric asymmetry with higher global power amplitudes in the left hemisphere compared to the right hemisphere (p = 0.04) independent of the administration of MPH. They also had higher global power amplitudes left hemispheric than ADHDcom children (p = 0.07).
Figure 5 Interaction DSMtype*HEMISPHERE for GL. Independent of the administration of MPH ADHD children of the inattentive type showed a hemispheric asymmetry with higher global power amplitudes in the left hemisphere compared to the right hemisphere (p = 0.04). They also had higher global power amplitudes left hemispheric than ADHD children of the combined type (p = 0.07).
Relative delta
The interaction DSMtype*HEMISPHERE (F(1,33) = 4.61, p = 0.039) was revealed. However, no differences were found in the post hoc analysis when p-values were adjusted following Tukey-Kramer. Looking at unadjusted p-values (p = 0.04), it was found that ADHDin children had lower relative delta band amplitudes in the right hemisphere (1.64) compared to the left hemisphere (1.66). No hemispheric differences were found for ADHDcom children.
Relative theta
The interaction TIME*HEMISPHERE (F(1,33) = 6.15, p = 0.018) was found. Figure 6 shows that in the left hemisphere relative Theta band amplitudes were higher after MPH than before MPH (p = 0.04). The interaction Response*TIME (F(1,33) = 6.16, p = 0.018) was revealed. As can be seen in figure 7, ADHDg children had higher amplitudes in the relative Theta band after MPH than before MPH (p = 0.02).
Figure 6 Interaction TIME*HEMISPHERE for relative Theta. In the left hemisphere relative Theta band amplitudes were higher after MPH than before MPH (p = 0.04).
Figure 7 Interaction RESPONSE*TIME for relative Theta. ADHD children who responded to MPH had higher amplitudes in the relative Theta band after MPH than before MPH (p = 0.02).
Relative alpha
The main effect TIME (F(1,33) = 7.09, p = 0.012) was found. Amplitudes in the relative Alpha band were lower after MPH (1.01) than before MPH (1.03). The interaction DSMtype*TIME (F(1,33) = 4.9, p = 0.03) was revealed. ADHDcom children had lower amplitudes in the relative Alpha band after MPH than before MPH (p = 0.009, see fig. 8).
Figure 8 Interaction DSMtype*TIME for relative Alpha. ADHD children of the combined type had lower amplitudes in the relative Alpha band after MPH than before MPH (p = 0.009).
The main effect Response (F(1,33) = 6.27, p = 0.017) was found. ADHDg children (1.06) had higher relative Alpha amplitudes than ADHDp children (0.98).
Relative beta
The interaction TIME*HEMISPHERE (F(1,33) = 4.77, p = 0.036) was revealed. However, no differences were found in the post-hoc analysis.
The interaction Response*TIME (F(1,33) = 5.18, p = 0.029) was found. Again, effects did not prove to be significant in the post-hoc analysis.
Theta/alpha ratio
The main effect TIME (F(1,33) = 7.24, p = 0.01) was revealed. The Theta/Alpha ratio was higher after MPH (1.32) than before MPH (1.28).
The main effect Response (F(1,33) = 4.58, p = 0.0399) was revealed. The Theta/Alpha ratio was lower for ADHDg children (1.24) than for ADHDp children (1.37).
Theta/beta ratio
The interaction TIME*HEMISPHERE (F(1,33) = 6.74, p = 0.014) was found. However, no differences were found in the post hoc analysis when p-values were adjusted following Tukey-Kramer. Looking at unadjusted p-values, it was revealed that the Theta/Beta ratio was higher after MPH than before MPH in the left hemisphere (p = 0.02, see fig. 9).
Figure 9 Interaction TIME*HEMISPHERE for T/B ratio. The Theta/Beta ratio was higher after administration of MPH than before MPH in the left hemisphere only (p = 0.02).
The interaction Response*TIME (F(1,33) = 4.75, p = 0.037) was revealed. Again, no differences were found in the post hoc analysis when p-values were adjusted following Tukey-Kramer. Looking at unadjusted p-values, it was found that ADHDg children had a higher Theta/Beta ratio after MPH than before MPH (p = 0.02, see fig. 10).
Figure 10 Interaction RESPONSE*TIME for T/B ratio. It was found that ADHD children who responded well to MPH had a higher Theta/Beta ratio after MPH than before MPH (p = 0.02).
Correlations and linear regressions
A correlation was found between D2 test improvement and MPH-induced power changes in the relative Theta band left frontal (r = .37, p < .05). A linear regression was computed and confirmed that the larger the improvement in D2 test performance was, the larger was the increase in Theta after MPH application (t = 2.27, p = 0.03), see fig. 11. No other correlations were found between any MPH-induced power band changes, MPH blood serum concentration and D2 test improvement.
Figure 11 Relationship between D2 test improvement and relative Thetapower increase. A correlation was found between D2 test improvement and MPH-induced power changes in the relative Theta band left frontal (r = .37, p < .05). A linear regression was computed and confirmed that the larger the improvement in D2 test performance was, the larger was the increase in T after MPH application (t = 2.27, p = 0.03).
Discussion
In the present study, main effects induced by medication were found in frontal regions. This result is consistent with etiological hypotheses of ADHD, as well as the working mechanism of MPH. MPH-influence on frontal lobe acitivation in ADHD subjects was also found in a SPECT study by Lou and colleagues [11]. They found ADHD subjects having reduced bloodflow in frontal regions as well as enhanced bloodflow in motor areas. After application of MPH, this pattern normalized. Niedermeyer [7,8] interprets these findings as support of the "lazy frontal lobe" hypothesis underlying ADHD. He argues that the prefrontal cortex is not only involved in allocation and sustaining attention (e.g. [22,23]), but also in inhibiting motor activity (augmented motor activity being characteristic for ADHD). Langleben and colleagues [12] performed a SPECT study with ADHD children who were on and off MPH. When the subjects were not taking MPH, bloodflow was higher in the motor, premotor, and the anterior cingulate cortices. The authors concluded that brief discontinuation of MPH treatment is associated with increased motor and anterior cingulate cortical activity. Thus, it appears that if the prefrontal cortex is underactivated, both attentional processes and the inhibition of the motor cortex will be diminished.
MPH acts upon the prefrontal cortex via the neurotransmitter dopamine. Involvement of the dopaminergic system has been suggested in patients suffering from ADHD since the symptoms can be successfully treated with MPH, a potent blocker of the dopamine transporter (DAT) [24]. MPH is known to influence the dopaminergic system by blocking dopamine reuptake and in consequence enhancing the availability of dopamine in the synaptic gap [13]. Dopamine is densely distributed in the prefrontal cortex as well as the striatum and acts mainly on inhibitory neurons. By increasing the availability of dopamine MPH seems to enhance the inhibitory effect on motor activity.
Our aim was to find cortical differences between children who were diagnosed with different subtypes of ADHD as well as children who were good or poor responders to MPH. We were also interested in MPH-induced changes that were common to all ADHD children. First of all, global power (amplitude power of the combined frequency band from 1.5 to 25 Hz) increased with MPH. This may be taken as further evidence of the hypoarousal model of ADHD (e.g. [9,10]). The hypoarousal model assumes that ADHD results from cortical underarousal (compare "lazy frontal lobe" hypothesis). If MPH increases the amplitude of the global power band, one might hypothesize that it counteracts cortical underarousal. Other studies supporting the hypoarousal model found decreased bloodflow especially in prefrontal areas [11,25]. In the study performed by Lou and colleagues (see above), this underarousal could be remediated by MPH.
In the present study, Alpha activity decreased in both hemispheres with MPH. Alpha activity has been related to attentional processes (e.g. [26-28]). I.e. synchronized Alpha activity can be found in the EEG when subjects are relaxed and inattentive. Alpha activity lessens when attention is directed towards a stimulus [26,27,29,30]. Klimesch and colleagues [26] argue that during Alpha desynchronization, different neural populations start oscillating with different frequencies which in consequence leads to the disappearance of the dominant Alpha rhythm. Our results cannot be directly compared to the findings described above, since we did not investigate Alpha activity related to vigilance tasks. However, we found decreased Alpha power after MPH application. Knowing that MPH is used to treat ADHD symptoms like excess motor arousal and inattentiveness, it is not surprising to find decreased Alpha activity after MPH application given that Alpha activity relates to attentiveness. The Alpha effect in the present study is in line with the results of an EEG study by Loo and co-workers [31], who also reported decreased Alpha power after MPH. The authors link this effect to an increase in cortical arousal. A similar effect of decreased Alpha activity recorded over left fonto-central sites in the EEG after MPH was reported by Swartwood et al. [32]. In contrast to these findings, the same group found increased Alpha activity after MPH over the left frontal pole in the same study. However, the authors take this contradictory result as "difficult to interpret". Clarke et al. [33] reported increased Alpha activity after MPH application for children diagnosed with ADHD of the predominantly inattentive type. The authors take this as part of a normalization of the EEG, since unmedicated ADHD children have been reported to have lower levels of Alpha activation compared to controls. The contradictory findings concerning the effect of MPH on Alpha activity are difficult to explain. Yet, knowing from studies on attention (see above) that higher levels of attentiveness are related to a decrease in Alpha activity, an MPH-induced decrease in Alpha power seems more plausible than an Alpha increase.
In the present study, Theta band activity increased left hemispherically after MPH application. This finding contradicts the results of Clarke and co-workers [33], Swartwood and colleagues [32] and Loo et al. [31]. All of them found lower levels of Theta power after MPH application. Generally, higher slow wave activity has been reported in ADHD children compared to controls (e.g. [34,35]). This was interpreted as an indicator of maturational lag in brain functioning (e.g. [35-37]), since slow wave activity normally decreases from childhood to adulthood (e.g. [38]). In a study by Chabot and Serfontein [39] EEG measures of ADHD children were compared to a normative database. Their results disagreed with the maturational lag hypothesis, since the EEG profile of ADHD children did not resemble the EEG profile of children of any age. Another possibility to interpret increased slow wave activity is again offered by the hypoarousal model. Bresnahan and colleagues [40] hypothesized that increased slow wave activity in ADHD subjects might be an effect of decreased dopamine functioning which is in turn the origin of cortical underarousal. In line with this, Clarke and co-workers [33], as well as Loo et al. [31] interpret the MPH-induced decrease of Theta power with an increase in cortical arousal. Swartwood and colleagues [32] assume that MPH blocks slow-wave activity. Interestingly, Loo and colleagues [16] found differential MPH-induced effects on Theta-activity depending on the DAT1 risk allele status of the ADHD children. In an eyes-open resting condition, children who carried the DAT1 10R allele (considered the "risk" allele) showed a focal increase in left parietal Theta power. Children, who carried the DAT1 9R allele showed a decrease. Unfortunately, this effect was not discussed by the authors. It seems, however, that Theta activity cannot solely be related to drowsiness and hypoarousal, otherwise MPH should not increase its power as in Loo et al.'s or the present study, especially since Theta increase was positively correlated with D2-test improvement. Theta band activity has also been investigated in connection to working memory processes. For instance, "functional" Theta activity was found in an EEG study investigating visual word encoding [41,42]. Event-related Theta activity was largest for words that could later be recalled. The authors assume that theta synchronization is selectively related to the encoding of new information. Interestingly, Theta power was largest left hemispheric. This corresponds to our finding of a left hemispheric increase in Theta power after MPH application. Larson and colleagues [43] found that long-term potentiation in the hippocampus is optimal when the stimulation pattern mimics theta rhythm. But also Theta oscillations generated in frontal brain regions play an active role in memory maintenance [44]. Aftanas and colleagues [45] found a relation between Theta synchronization and an emotionally positive state and internalized attention. This effect was particularly prominent in left prefrontal regions. Gevins et al. [46] related the midline theta rhythm to intense concentration.
In conclusion, the results described above suggest that Theta activity does not necessarily mirror cortical underarousal. It can also reflect information processing, consolidation and attention. The subjects in our study did not have to perform any task. They only rested in the MEG with open eyes. Thus, it is unlikely that the Theta increase found corresponds to information encoding or memory processes. However, it is possible, that MPH increases the functional aspect of the Theta rhythm rather than increasing underarousal or drowsiness. Again, the increase in Theta power was correlated with an increase in behavioural performance in the attention test D2. Since we defined children to be MPH good responders or non-responders based on their increase in D2-test performance, it is not surprising that it was only the children who responded well to MPH who showed an increase in Theta power and Theta/Beta ratio.
In the present study, the Theta/Alpha ratio also increased with MPH application (mostly in the left hemisphere), as did the Theta/Beta ratio. The increase in the Theta/Beta ratio is of course a consequence of an increase in relative Theta and a concurrent decrease in relative Alpha power. Thus, it does not reveal any new information. Presumably, the same is true for the increase in the Theta/Beta ratio. Although a significant interaction was found between Beta Power and MPH status, no significant differences were found in the post hoc tests. This implies that Beta power did not change to a great degree with MPH application and the increase in the Theta/Beta ratio is very likely a result of an increase in Theta power alone.
We did find differences between the two ADHD subtypes. Children with ADHD of the predominantly inattentive type had higher global power amplitudes in the left hemisphere than children with ADHD of the combined type. If global power activity reflects cortical arousal in our study (see above), one might hypothesize, that children with ADHD of the combined type are more hypoaroused than children with ADHD of the predominantly inattentive type. Clarke and colleagues [19] found the opposite result (combined > inattentive). They stated that higher levels of global power amplitude reflect cortical underarousal and consequently concluded that children with ADHD of the predominatly inattentive type are less hypoaroused than children with ADHD of the combined type. Yet, as described above, we found an MPH-induced increase in global power activity that was accompanied by an increase in behavioural performance. Thus, in our study, higher global power does not seem to mirror cortical hypoarousal, but in fact the opposite. Therefore, we might also conclude from our data that children with ADHD of the predominantly inattentive type are the ones being less hypoaroused. Another characteristic of children with ADHD of the inattentive type was a hemispheric asymmetry in global power and relative delta activity with more power in the left hemisphere. No asymmetries were found for children with ADHD of the combined type. Characteristic for the latter group was an MPH-induced decrease in Alpha power. Although the decrease in Alpha power became statistically significant for all ADHD children (see above), the effect seemed mainly to be driven by the children with ADHD of the combined type.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
CW designed the study, carried out the data acquisition, data processing and statistics, drafted the manuscript and has given final approval of the version to be published.
IP made substantial contribution to study design, carried out data acquisition and statistics and drafted the manuscript.
SB carried out the subject selection, performed the neuropsychological testing, and contributed during data acquisition.
HK carried out subject diagnostics, medical treatment, and contributed to data acquisition and the preparation of the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This study was supported by grants from the Volkswagenstiftung and the University of Konstanz. The authors like to thank U. Lommen, C. Wolf, and B. Awiszus for their support during the data acquisition.
==== Refs
Diagnostic and Statistical Manual of Mental Disorders - Fourth Edition (DSM-IV). American Psychiatric Association, Washington D.C. 1994
Scahill L Schwab-Stone M Epidemiology of ADHD in school-age children Child Adolesc Psychiatr Clin N Am 2000 9 541 55, vii 10944656
Diagnosis and Treatment of Attention Deficit Hyperactivity Disorder. NIH Consens Statement Online 1998 Nov 16-18; 16(2): 1-37. [http://odp.od.nih.gov/consensus/cons/110/110_statement.htm]
Mental Health: A report of the surgeon general [http://www.surgeongeneral.gov/library/mentalhealth/home.html]
United Nations International Narcotics Control Board Report of the International Narcotics Control Board. United Nations Publication Sales No. E.96XI.1 1995
Schwabe U Paffrath D Arzneiverordnungs-Report 2001 2001 , Springer
Niedermeyer E Frontal lobe functions and dysfunctions Clin Electroencephalogr 1998 29 79 90 9571295
Niedermeyer E Frontal lobe disinhibition, Rett syndrome and attention deficit hyperactivity disorder Clin Electroencephalogr 2001 32 20 23 11202137
Satterfield JH Cantwell DP Satterfield BT Pathophysiology of the hyperactive child syndrome Arch Gen Psychiatry 1974 31 839 844 4441251
Barry RJ Clarke AR Johnstone SJ A review of electrophysiology in attention-deficit/hyperactivity disorder: I. Qualitative and quantitative electroencephalography Clin Neurophysiol 2003 114 171 183 12559224 10.1016/S1388-2457(02)00362-0
Lou HC Henriksen L Bruhn P Focal cerebral hypoperfusion in children with dysphasia and/or attention deficit disorder Arch Neurol 1984 41 825 829 6331818
Langleben DD Acton PD Austin G Elman I Krikorian G Monterosso JR Portnoy O Ridlehuber HW Strauss HW Effects of methylphenidate discontinuation on cerebral blood flow in prepubescent boys with attention deficit hyperactivity disorder J Nucl Med 2002 43 1624 1629 12468511
Volkow ND Wang GJ Fowler JS Logan J Franceschi D Maynard L Ding YS Gatley SJ Gifford A Zhu W Swanson JM Relationship between blockade of dopamine transporters by oral methylphenidate and the increases in extracellular dopamine: therapeutic implications Synapse 2002 43 181 187 11793423 10.1002/syn.10038
Clarke AR Barry RJ McCarthy R Selikowitz M EEG differences between good and poor responders to methylphenidate and dexamphetamine in children with attention-deficit/hyperactivity disorder Clin Neurophysiol 2002 113 194 205 11856625 10.1016/S1388-2457(01)00736-2
Clarke AR Barry RJ McCarthy R Selikowitz M Croft RJ EEG differences between good and poor responders to methylphenidate in boys with the inattentive type of attention-deficit/hyperactivity disorder Clin Neurophysiol 2002 113 1191 1198 12139997 10.1016/S1388-2457(02)00147-5
Loo SK Specter E Smolen A Hopfer C Teale PD Reite ML Functional effects of the DAT1 polymorphism on EEG measures in ADHD J Am Acad Child Adolesc Psychiatry 2003 42 986 993 12874502 10.1097/01.CHI.0000046890.27264.88
Clarke AR Barry RJ McCarthy R Selikowitz M EEG-defined subtypes of children with attention-deficit/hyperactivity disorder Clin Neurophysiol 2001 112 2098 2105 11682348 10.1016/S1388-2457(01)00668-X
Clarke AR Barry RJ McCarthy R Selikowitz M Electroencephalogram differences in two subtypes of attention-deficit/hyperactivity disorder Psychophysiology 2001 38 212 221 11347867 10.1017/S0048577201981764
Clarke AR Barry RJ Bond D McCarthy R Selikowitz M Effects of stimulant medications on the EEG of children with attention-deficit/hyperactivity disorder Psychopharmacology (Berl) 2002 164 277 284 12424551 10.1007/s00213-002-1205-0
www.wma.net
Brickenkamp R Aufmerksamkeits- Belastungstest d2 1994 Göttingen, Hogrefe
Posner MI Petersen SE The attention system of the human brain Annu Rev Neurosci 1990 13 25 42 2183676 10.1146/annurev.ne.13.030190.000325
Knight RT Grabowecky MF Scabini D Role of human prefrontal cortex in attention control Adv Neurol 1995 66 21 34; discussion 34-6 7771302
Dresel S Krause J Krause KH LaFougere C Brinkbaumer K Kung HF Hahn K Tatsch K Attention deficit hyperactivity disorder: binding of [99mTc]TRODAT-1 to the dopamine transporter before and after methylphenidate treatment Eur J Nucl Med 2000 27 1518 1524 11083541 10.1007/s002590000330
Zametkin AJ Nordahl TE Gross M King AC Semple WE Rumsey J Hamburger S Cohen RM Cerebral glucose metabolism in adults with hyperactivity of childhood onset N Engl J Med 1990 323 1361 1366 2233902
Klimesch W Doppelmayr M Schimke H Pachinger T Alpha frequency, reaction time, and the speed of processing information J Clin Neurophysiol 1996 13 511 518 8978623 10.1097/00004691-199611000-00006
Klimesch W Doppelmayr M Pachinger T Ripper B Brain oscillations and human memory: EEG correlates in the upper alpha and theta band Neurosci Lett 1997 238 9 12 9464642 10.1016/S0304-3940(97)00771-4
Babiloni C Miniussi C Babiloni F Carducci F Cincotti F Del Percio C Sirello G Fracassi C Nobre AC Rossini PM Sub-second "temporal attention" modulates alpha rhythms. A high-resolution EEG study Brain Res Cogn Brain Res 2004 19 259 268 15062863 10.1016/j.cogbrainres.2003.12.010
Vazquez Marrufo M Vaquero E Cardoso MJ Gomez CM Temporal evolution of alpha and beta bands during visual spatial attention Brain Res Cogn Brain Res 2001 12 315 320 11587900 10.1016/S0926-6410(01)00025-8
Fu KM Foxe JJ Murray MM Higgins BA Javitt DC Schroeder CE Attention-dependent suppression of distracter visual input can be cross-modally cued as indexed by anticipatory parieto-occipital alpha-band oscillations Brain Res Cogn Brain Res 2001 12 145 152 11489617 10.1016/S0926-6410(01)00034-9
Loo SK Teale PD Reite ML EEG correlates of methylphenidate response among children with ADHD: a preliminary report Biol Psychiatry 1999 45 1657 1660 10376129 10.1016/S0006-3223(98)00250-9
Swartwood MO Swartwood JN Lubar JF Timmermann DL Zimmerman AW Muenchen RA Methylphenidate effects on EEG, behavior, and performance in boys with ADHD Pediatr Neurol 1998 18 244 250 9568922 10.1016/S0887-8994(97)00205-1
Clarke AR Barry RJ McCarthy R Selikowitz M Brown CR Croft RJ Effects of stimulant medications on the EEG of children with Attention-Deficit/Hyperactivity Disorder Predominantly Inattentive type Int J Psychophysiol 2003 47 129 137 12568943 10.1016/S0167-8760(02)00119-8
Lazzaro I Gordon E Li W Lim CL Plahn M Whitmont S Clarke S Barry RJ Dosen A Meares R Simultaneous EEG and EDA measures in adolescent attention deficit hyperactivity disorder Int J Psychophysiol 1999 34 123 134 10576397 10.1016/S0167-8760(99)00068-9
Clarke AR Barry RJ McCarthy R Selikowitz M EEG analysis in Attention-Deficit/Hyperactivity Disorder: a comparative study of two subtypes Psychiatry Res 1998 81 19 29 9829647 10.1016/S0165-1781(98)00072-9
Satterfield JH Cantwell DP Saul RE Lesser LI Podosin RL Response to stimulant drug treatment in hyperactive children: prediction from EEG and neurological findings J Autism Child Schizophr 1973 3 36 48 4740584 10.1007/BF01537553
Matsuura M Okubo Y Toru M Kojima T He Y Hou Y Shen Y Lee CK A cross-national EEG study of children with emotional and behavioral problems: a WHO collaborative study in the Western Pacific Region Biol Psychiatry 1993 34 59 65 8373939 10.1016/0006-3223(93)90257-E
Clarke AR Barry RJ McCarthy R Selikowitz M Age and sex effects in the EEG: development of the normal child Clin Neurophysiol 2001 112 806 814 11336896 10.1016/S1388-2457(01)00488-6
Chabot RJ Serfontein G Quantitative electroencephalographic profiles of children with attention deficit disorder Biol Psychiatry 1996 40 951 963 8915554 10.1016/0006-3223(95)00576-5
Bresnahan SM Anderson JW Barry RJ Age-related changes in quantitative EEG in attention-deficit/hyperactivity disorder Biol Psychiatry 1999 46 1690 1697 10624551 10.1016/S0006-3223(99)00042-6
Klimesch W Doppelmayr M Russegger H Pachinger T Theta band power in the human scalp EEG and the encoding of new information Neuroreport 1996 7 1235 1240 8817539
Klimesch W Doppelmayr M Schimke H Ripper B Theta synchronization and alpha desynchronization in a memory task Psychophysiology 1997 34 169 176 9090266
Larson J Wong D Lynch G Patterned stimulation at the theta frequency is optimal for the induction of hippocampal long-term potentiation Brain Res 1986 368 347 350 3697730 10.1016/0006-8993(86)90579-2
Jensen P Longer term effects of stimulant treatments for Attention-Deficit/Hyperactivity Disorder J Atten Disord 2002 6 Suppl 1 S45 56 12685518
Aftanas LI Golocheikine SA Human anterior and frontal midline theta and lower alpha reflect emotionally positive state and internalized attention: high-resolution EEG investigation of meditation Neurosci Lett 2001 310 57 60 11524157 10.1016/S0304-3940(01)02094-8
Gevins A Smith ME McEvoy L Yu D High-resolution EEG mapping of cortical activation related to working memory: effects of task difficulty, type of processing, and practice Cereb Cortex 1997 7 374 385 9177767 10.1093/cercor/7.4.374
|
16042816
|
PMC1184088
|
CC BY
|
2021-01-04 16:33:03
|
no
|
BMC Psychiatry. 2005 Jul 26; 5:29
|
utf-8
|
BMC Psychiatry
| 2,005 |
10.1186/1471-244X-5-29
|
oa_comm
|
==== Front
J CarcinogJournal of Carcinogenesis1477-3163BioMed Central London 1477-3163-4-101601881310.1186/1477-3163-4-10ResearchDifferential regulation of somatostatin receptors 1 and 2 mRNA and protein expression by tamoxifen and estradiol in breast cancer cells Rivera Juan A [email protected] Haydar [email protected] Ujendra [email protected] Fraser Laboratories For Diabetes Research, Department of Medicine, Royal Victoria Hospital, McGill University Health Centre, Montreal, Quebec, H3A 1A1, Canada2005 14 7 2005 4 10 10 31 7 2004 14 7 2005 Copyright © 2005 Rivera et al; licensee BioMed Central Ltd.2005Rivera et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Somatostatin (SST) inhibition of hormone hypersecretion from tumors is mediated by somatostatin receptors (SSTRs). SSTRs also play an important role in controlling tumor growth through specific antiproliferative actions. These receptors are well expressed in numerous normal and tumor tissues and are susceptible to regulation by a variety of factors. Estradiol, a potent trophic and mitogenic hormone in its target tissues, is known to modulate the expression of SST and its receptors. Accordingly, in the present study, we determined the effects of tamoxifen, a selective estrogen receptor (ER) modulator (SERM), and estradiol on SSTR1 and SSTR2 expression at the mRNA and protein levels in ER-positive and -negative breast cancer cells. We found that SSTR1 was upregulated by tamoxifen in a dose-dependent manner but no effect was seen with estradiol. In contrast, SSTR2 was upregulated by both tamoxifen and estradiol. Combined treatment caused suppression of SSTR1 below control levels but had no significant effect on SSTR2. Treatment with SSTR1-specific agonist was significantly more effective in suppressing cell proliferation of cells pre-treated with tamoxifen. Taking these data into consideration, we suggest that tamoxifen and estradiol exert variable effects on SSTR1 and SSTR2 mRNA and protein expression and distributional pattern of the receptors. These changes are cell subtype-specific and affect the ability of SSTR agonists to inhibit cell proliferation.
breast cancer cellsestradiolcell proliferationsomatostatin receptorstamoxifen
==== Body
Introduction
Somatostatin (SST) is a regulatory neuropeptide produced and secreted by neuroendocrine and inflammatory cells [1]. It inhibits secretory and proliferative responses in a number of target cells. In tumors, SST not only blocks hormone hypersecretion but also causes variable degrees of tumor shrinkage. This antiproliferative effect involves cytostatic (growth arrest) and cytotoxic (apoptosis) actions. The biological effect of SST is mediated directly through a family of five specific high affinity G protein-coupled receptors termed SSTR1-5. These receptors are present in most normal and tumor cells [2]. SST also acts indirectly by reducing the synthesis and secretion of local and systemic growth promoting factors. Other actions of SST include inhibition of angiogenesis, promotion of vasoconstriction, and modulation of immune cell function [1,2].
Most solid tumors, including breast, prostate, colon, pancreas, brain and liver cancer, express variable levels of SSTRs and are, therefore, amenable to therapy with SST analogues [1-3]. The anti-proliferative effects of SST and its analogue octreotide (OCT) in breast and other cancers have been clearly demonstrated [4-8]. In animal models of breast cancer, OCT enhanced the anti-neoplastic effect of tamoxifen (Tam) or castration [6]. In addition, SST analogues cause a potent reduction in circulating levels of growth hormone (GH) and its mediator hormone IGF-1, a potent mitogen in breast and other cancers [9]. Clinical studies, however, have failed to demonstrate a clinically significant benefit of OCT administration in addition to standard therapy with tamoxifen in patients with advanced breast cancer [10,11]. Although most tumors in patients with primary breast cancer express SSTRs [12], lower expression levels in more aggressive tumors may account for the failure of OCT with Tam in these settings [13]. Therefore, upregulation of SSTR expression in aggressive tumors may be a desirable therapeutic goal.
Some of the factors known to upregulate SSTR expression are also cell proliferation promoters. This presumably works as a way of endogenous counter regulation [1]. In several tissues, the sex hormone estradiol (E2) acts as a potent proliferative and mitogenic factor and is also known to increase the expression of SST and its receptors [14-18]. Previous studies have described the variable effects of E2 on SSTR-mRNA in a number of cell lines [14-18]. In breast cancer cells, treatment with E2 caused up-regulated SSTR2-mRNA expression, while Tam had variable effects depending on the cell line used [14,15]. Visser-Wisselaar et al [16] reported induction of SSTR2 and SSTR3 expression by estrogen in transplantable rat prolactin-secreting pituitary tumor cells. Similarly, Djordijevic et al [17] reported positive regulation of SSTR2 and SSTR3 alongside inhibition of SSTR1-mRNA by E2 in primary cultures of female rat pituitary cells expressing all five SSTRs. In contrast, Kimura et al [18] showed upregulation of SSTR1, SSTR2, and SSTR3, and drastic downregulation of SSTR5 in pituitary cells of ovariectomized rats treated with E2 for one month. From these studies, it is clear that the precise regulatory role of estrogens on SSTRs and its implications in cell growth control need further elucidation. Additionally, the role of Tam has not been investigated in detail.
Tam, an ER modulator, potently inhibits the growth of ER-positive breast cancer cells [19]. Cancer cells from tissues not classically considered estrogen-dependent (e.g. thyroid, skin, pancreas, liver, glia and meninges) also show an inhibitory proliferative response to Tam [20-24]. Some of the growth inhibitory properties of Tam are related to its ability to modify the expression of cell growth regulators. One such regulator, transforming growth factor β1 (TGF-β1), an inhibitor of cell proliferation, is upregulated by Tam in both ER-positive and ER-negative cells [23,25]. In contrast, Tam reduces circulating levels of IGF-1 [26] and interferes with the IGF-1 receptor signalling pathway in breast cancer cells by reducing its phosphorylation and by inhibiting the induction of its substrate IRS-1 [27,28].
It has been previously shown that SSTR1 and SSTR2 are highly expressed in breast cancer cells [29] and exert an inhibitory role on tumor cell proliferation and migration [30-35]. Accordingly, in the present study, we compare the effect of Tam and E2, alone or in combination, on SSTR1 and 2 expressions in breast cancer cells at the mRNA and protein levels. We report that Tam and E2 exhibited contrasting effects on SSTR1 in ER-positive cells. However, their actions on SSTR2 expression in ER-negative cells were similar. We further discuss our findings in terms of the potential clinical implications of such interactions.
Materials and methods
Reagents
All the culture cells were obtained from ATCC. Tamoxifen, Estradiol and 3-(4, 5-dimethylthiazolyl-2)-2, 5-diphenyltetrazolium bromide (MTT) were purchased from SIGMA (Sigma-Aldrich Canada Ltd). Anti rabbit fluorescein isothiocyanate (FITC)-conjugated secondary antibodies were obtained from Jackson Laboratory. Cell culture medium was from Invitrogen and FBS was purchased from Wisent, Canada. All other reagents were purchased from various suppliers as indicated.
Cell lines
ZR-75-1 and T47D cells were maintained in RPMI Medium 1640 (Gibco BRL) at standard conditions (humidified atmosphere, at 37°C, with 5% CO2). MDA-MB-231 cells were maintained in Leibovitz's L-15 medium at 37°C. Culture media were supplemented with 10% fetal bovine serum (FBS) and standard antifungal-antibacterial treatment. Culture media were replaced 24 hours prior to treatment with phenol red-free media supplemented with 10% dextran-coated charcoal-treated FBS. Experiments were conducted on cells between passages 4 to 8.
Cell Treatment
At ~70% confluency, cells were treated with increasing concentrations (10-10 M, 10-8 M and 10-6 M) of Tam and β-estradiol (E2), alone or in combination. Time course experiments (6, 12, 24, 48, and 96 h) showed that changes in mRNA levels were already maximal at 24 h and sustained thereafter. Thus a 24–36 h exposure time was used for these experiments. In order to determine the expression of SSTR1 and SSTR2 by immunocytochemistry, cells were treated with 10-6 M of the corresponding compound for 30–36 h. For cell proliferation experiments, cells were exposed to Tam, E2, both or none for 30 h. 0.5 μM of the nonpeptide SSTR1-specific agonist L-797, 591 [36] or vehicle was added to the medium, and proliferation rate were assessed after an additional 42 h.
RT-PCR
Total RNA was extracted using TRIzol Reagent (Invitrogen) according to the manufacturer's instructions. Potential contaminating genomic DNA was degraded by incubation with RQ1 ribonuclease-free deoxyribonuclease (Promega Corp.) for 30 minutes at 37°C in the presence of RNAse inhibitor (Invitrogen). 10 μg of total RNA was reverse transcribed (RT) using M-MLV-reverse transcriptase (Life technologies Inc), and then amplified using specific primers for SSTR1, SSTR2 and β-actin as an internal control. Primers used were as follows:
SSTR1-forward 5' TATCTGCCTGTGCTACGTGC 3' (nt 714–733)
SSTR1-reverse 5' GATGACCGACAGCTGACTCA 3' (nt 911–930)
SSTR2-forward 5' ATCTGGGGCTTGGTACACAG 3' (nt 600–619)
SSTR2-reverse 5' CTTCTTCCTCTTAGAGGAGC 3' (nt 728–747)
β-actin-forward 5' ATCATGAA GTGTGACGTGGAC 3' (nt 885–905)
β-actin-reverse 5' AACCGACTGCTGTCA CCTTCA 3' (nt 1325–1345)
For PCR amplification 4 μl of RT products (cDNA) were mixed in 100 μl total volume of PCR-buffer (Invitrogen) containing 2% DMSO, 2.25 mM MgCl2, 50 μM of each dNTP, 15 pmol of SSTR-primers, and 3 pmol of β-actin primers.
After initial denaturation at 94°C for 10 min, Taq DNA polymerase (Invitrogen) (2.5 U/reaction) was added and samples were subjected to 38 cycles of denaturation at 94°C for 1 min, annealing at 58°C for 45 sec, and extension at 72°C for 45 sec, followed by a 10 min final extension at 72°C. 20 μl of the PCR products were fractioned by electrophoresis in 1.5% agarose gels containing ethidium bromide and visualized under UV light. The specificity of the amplified SSTR products and the corresponding observed bands were confirmed by Southern blot hybridization using specific 32P-dCTP-random primer-labelled SSTR cDNAs. AlphaEaseFC (Alpha Innotech Corporation, San Leandro, CA) was used for optical density (OD) measurements of the product. OD of target band was corrected for the corresponding β-actin-band (OD ratio) and then normalized against the OD ratio of non-treated cells (OD index).
Immunocytochemistry
Expression of SSTRs in ZR-75-1, T47D and MDA-MB-231 were determined by immunocytochemistry using rabbit polyclonal antibodies for hSSTR1 and hSSTR2, diluted as previously described [37,38]. Briefly, cells were grown in 24-well plates to 60% confluency, and then treated with Tam or E2, alone or in combinations. Cells were fixed with 4% paraformaldehyde and incubated overnight at 4°C with primary antibodies diluted (1:300) in PBS. After three subsequent washes in PBS, cells were incubated at room temperature for 1 h with FITC-conjugated secondary antibodies (goat anti-rabbit IgG). The specificity of immunofluorescence was determined in the absence of hSSTR-specific antibodies.
Cell Proliferation Assay
Cell proliferation was assessed 72 h post-treatment by the MTT method. Cells were seeded at 1000 cells/well in 96-well plates containing 100 μl of standard medium. Treatments were started, as described above, 24 h later. At the end of the treatment period, 25 μl MTT solution, 5 mg/ml in PBS was added to each well. Following 2 h incubation at 37°C, 100 μl stop solution (50% dimethyl-formamide, 50% H2O, 20% SDS, pH 4.7) was added to each well and cells were incubated for an additional 20 h to solubilize the crystallized dye. Optical densities of the solutions, in each well, were determined by spectrophotometer.
Statistical Analysis
Results are expressed as means ± SE. Statistical significance was determined using the Student's unpaired t test.
Results
Effects of Tamoxifen and Estradiol on SSTR1 and SSTR2 mRNA and Protein Expression in ZR-75-1 Cells
In order to assess the effects of Tam and E2 on SSTR1 and SSTR2 expression, we first determined dose-dependent changes in mRNA expression. In the ER-positive breast cancer cell line ZR-75-1, Tam treatment resulted in a significant increase in SSTR1-mRNA, up to 2.2-fold in comparison to non-treated cells, at high concentrations (10 nM and 1 μM) (Fig. 1A). In contrast, E2 treatment significantly decreased SSTR1-mRNA expression by >25%. Treatment with both Tam and E2 resulted in a further decrease in SSTR1 mRNA levels. In contrast, both Tam and E2 increased SSTR2 mRNA expression. However, the effect of Tam was stronger (4.5 fold increase at 1 μM) in addition to being dose-dependent (Fig. 1B). In this case, simultaneous treatment with Tam and E2 had no significant effect on SSTR2 mRNA expression as compared with control (Fig. 1B).
Figure 1 Concentration dependent changes in SSTR1 (A) and 2 (B) mRNA in ZR-75-1 cells treated with Tam and E2 alone or in combination. Cells were treated with Tam and E2 for 24 h (see Materials and Methods for details). mRNA levels were determined by RT-PCR and expressed as OD ratio. Note a concentration-dependent increase in SSTR1 mRNA levels with Tam and decreased expression in the presence of E2 or combined treatment. SSTR2 mRNA levels increased in response to Tam or E2 but not with combined treatment. Data presented are from three independent experiments performed in triplicate. A representative ethidium bromide-stained gel image is shown. * p < 0.05
We further extended our study and determined the cellular distribution of SSTR1 and SSTR2 by indirect immunofluorescence using SSTR1- and SSTR2-specific antibodies. As shown in Fig. 2, in ZR-75-1 cells, the changes observed after a 30–36 h treatment were comparable to changes described at the mRNA level. In control cells, SSTR1 and SSTR2 immunoreactivity had patchy distributions on the cell surface. Upon Tam treatment, SSTR1 and SSTR2 immunoreactivity significantly increased and the receptors were more uniformly expressed on the cell surface (Fig. 2B and 2F). Treatment with E2 alone or in combination with Tam resulted in either decreased expression or no significant changes in SSTR1- and SSTR2-like immunoreactivity at all time-points (Fig. 2C,D,G and 2H).
Figure 2 Photomicrographs illustrating the immunohistochemical localization of SSTR1 and 2 in ZR-75-1 breast cancer cells. Cells were treated with 1 μM of Tam and E2, alone or in combination, for 30–36 h and labelled with anti-rabbit SSTR1 and 2 antibodies followed by FITC-conjugated goat anti-rabbit IgG. Note a significant increase in SSTR1 and 2-like immunoreactivity in response to Tam (B and F) and decreased staining in the presence of E2 (C and G). Upon combined treatment SSTR1 is less than control (D) but SSTR2 expression is comparable to the control (H). Scale bar = 25 μM.
Effects of Tamoxifen and Estradiol on SSTR1 and SSTR2 mRNA and Protein Expression in T47D Cells
To assess whether these effects were similar in other ER-positive cell lines, we studied T47D cells. Under estrogen-free conditions, non-treated T47D cells express mRNA for both SSTR1 and SSTR2 (Fig. 3A and 3B). SSTR1 mRNA exhibited a biphasic response to Tam treatment, slightly decreasing with 0.1 nM Tam while increasing by up to ~300% with 1 μM (Fig. 3A). In contrast, E2 down regulated SSTR1 mRNA. Combined treatment caused a greater decrease in SSTR1 mRNA levels compared to E2-treated cells. In comparison, SSTR2-mRNA in T47D cells increased with Tam or E2. However, the combined treatment did not significantly change SSTR2 mRNA expression compared to control cells (Fig. 3B). In order to examine the cellular distribution of SSTR1 and 2 in response to Tam and E2, alone or in combination, on receptor expression at the protein level, T47D cells were also studied by indirect immunofluorescence. In control cells (Fig. 4), both receptors are expressed as membrane proteins. SSTR1 and 2, immunoreactivity decreased in T47D cells upon 30–36 h treatment with either Tam or E2, alone or in combination. Furthermore, a punctate pattern of receptor immunoreactivity was detected in all treated cells (Fig. 4).
Figure 3 Semiquantitative analysis of SSTR1 (A) and SSTR2 (B) mRNA in T47D cells treated with Tam and E2, alone or in combination, for 24 h. mRNA levels were estimated by RT-PCR and expressed as OD ratio. SSTR1 gene expression increased in the presence of Tam but decreased with E2 and combined treatments in a concentration-dependent manner. SSTR2 gene expression increased in the presence of Tam as well as with E2 but not with the combined treatment. A representative ethidium bromide-stained gel image is shown. Data presented are from three independent experiments performed in triplicate. * p < 0.05.
Figure 4 Photomicrographs depicting the immunohistochemical localization of SSTR1 and 2 in T47D breast cancer cells. Cells were treated with 1 μM Tam and E2, alone or in combination, for 30–36 h and processed for localization of SSTR1 and SSTR2 as described in Fig. 2. Note the significant changes in distributional pattern of SSTR 1 (A–D) and 2 (E–H) like immunoreactivity upon treatment. In control cells, both receptors are uniformly expressed as membrane proteins; however, upon treatment, a punctated receptor-like immunoreactivity was noticed at the cell surface. Scale bar = 25 μM.
Effects of Tamoxifen and Estradiol on SSTR1 and SSTR2 Expression in MDA-MB-231 Breast Cancer Cells
We next examined the effects of Tam and E2 in ERα-negative breast cancer cells MDA-MB-231. At baseline conditions, cells appeared to express a low amount of SSTR1-mRNA in comparison to SSTR2 (Fig. 5). Tam treatment significantly increased SSTR1-mRNA at a 10 nM concentration (Fig. 5A). The increase observed in SSTR2 with Tam was modest and did not reach statistical significance (Fig. 5B). Similar changes occurred when cells were treated with E2: both SSTR1 and SSTR2-mRNA were upregulated; however, in this case, the effect on SSTR2 was stronger. The combined treatment resulted in a trend towards down-regulation of SSTR1-mRNA while there was no significant change in SSTR2-mRNA compared to control.
Figure 5 Concentration-dependent changes in SSTR1 and 2 mRNA in MD-MB231 cells treated with Tam or E2, alone or in combination, for 24 h. mRNA levels were determined by RT-PCR and expressed as OD ratio. No significant changes in receptor expression were seen except for SSTR1 in presence of 10 nM of Tam and SSTR2 at the lower concentration of E2. Data presented are from three independent experiments performed in triplicate. A representative ethidium bromide-stained gel image is shown. * p < 0.05.
Similarly, very mild SSTR1 and SSTR2 immunoreactivity was observed in MDA-MB-231 cells under basal conditions (Fig. 6). The distribution followed a diffuse punctate pattern for both receptors. SSTR1 and SSTR2 were upregulated after Tam treatment while SSTR2 immunoreactivity was also upregulated by E2. Combined treatment was also associated with modest increases in SSTR1 and SSTR2 immunoreactivity.
Figure 6 Photomicrographs illustrating the immunohistochemical localization of SSTR1 (A–D) and SSTR2 (E–H) in the ERα negative MDA-MB-231 breast cancer cells. Cells were treated with 1 μM Tam and E2, alone or in combination, for 30–36 h and processed for localization of SSTR1 and SSTR2 (see Fig. 2 for details). MDA-MB-231 cells exhibited weak expression of both receptors. Note the apparent intracellular localization of the receptors like immunoreactivity. Scale bar = 25 μM.
SSTR1 Selective Agonist Accelerates the Tamoxifen-Induced Inhibition of ZR-75-1 Cell Proliferation
As illustrated in Fig. 7, the SSTR1-specific agonist potentiated the antiproliferative effects of Tam in ZR-75-1 cells. Cells were pretreated with Tam and E2, alone or in combination, for 30 h and then exposed to SSTR1-selective agonist for 42 h. Upon treatment with Tam alone, a significant reduction in cell proliferation was observed in ER-positive breast cancer cells. Combining SSTR1-specific agonist with Tam further suppressed proliferation when compared with Tam alone. It is worth noting that treatment with SSTR1 agonist alone had no significant effect on cell proliferation. Interestingly, when both Tam and E2 were given simultaneously followed by SSTR1 agonist, there was a mild but significant decrease in cell proliferation rate. Furthermore, in T47D and MDA-MB-231 cells, pretreatment with Tam followed by SSTR1-specific agonist resulted in mild, yet non-significant, additional inhibition of proliferation compared with Tam alone (data not shown).
Figure 7 SSTR1 selective agonist enhanced the antiproliferative effects of Tam in ZR-75-1 cells. Cells were pre-treated with Tam or E2 for 30 h. after the initial 30 h. Subsequently, a non-peptide SSTR1-specific agonist (R1 agonist) was added, as indicated, and cells were cultured for an additional 42 h. Cell proliferation was determined by MTT assay as described in Material and Methods. Data are presented as proliferation index in comparison to control as an arbitrary unit.
Discussion
In the present study, we demonstrate the roles of Tam and E2 on SSTR1 and SSTR2 mRNA and protein expression in ER-positive and ER-negative cells. This is the first study that not only systematically analyses the regulation of SSTR by E2 and Tam in breast cancer cells at both the mRNA and protein levels, but also demonstrates how this translates into changes in cancer cell proliferation control with somatostatin analogs. Our first observation was the contrasting effects of E2 and Tam on SSTR1 in ER-positive cells. An inhibitory effect of E2 on SSTR1-mRNA has been previously reported in rat anterior pituitary cells [17]. SSTR1 has been shown to inhibit cell proliferation and migration of tumor cells [30-33]. Therefore, it is most likely that its down-regulation by E2 plays a role in the proliferative effect of E2 in susceptible tissues. Likewise, the observed up-regulation of SSTR1 by Tam may represent an important mechanism whereby Tam exerts an inhibitory effect on tumor progression. In CCL39 human fibroblasts expressing SSTR1, SST inhibited activation of Rho (a key regulator of the actin-based cell cytoskeleton), thereby inhibiting the assembly of focal adhesions and actin stress fibres, and impairing cell migration [30]. This effect was not seen in cells expressing only SSTR2. Moreover, SSTR1-expressing cells from GH- and PRL-secreting adenomas and medullary thyroid cancer exhibited inhibition of hormone secretion in addition to reduced cell viability in response to SSTR1-specific agonist [31,32].
SSTR2 mRNA up-regulation by both Tam and E2 is in agreement with previous studies in T47D cells but differs from what has been previously shown in ZR-75-1 and MDA-MB-231 cells [14,15]. Xu et al [14] observed no effect of E2 in MDA-MB-231 cells, while we found that these cells responded to nanomolar concentrations of Tam and E2 with upregulation of SSTR1 and SSTR2 mRNA. This discrepancy can be explained by the different methods used: nuclease protection assay by Xu et al, and RT-PCR by us, the latter being more sensitive to small changes. Furthermore, Xu et al showed in ZR-75-1 cells that only E2 caused SSTR2 upregulation while Tam opposed this effect. It has been shown that exposure to E2 can increase ERβ mRNA expression in breast cancer cells [39]; conversely, prolonged E2 deprivation can induce loss of ER expression in some breast cancer cell lines [40]. Therefore, we speculate that differences in the E2-free pretreatment period (48 h in Xu et al vs. 24 h in our experiments) could have altered ER expression at the time of treatment. Additionally, the use of Tam in our experiments vs. the active metabolite OH-tamoxifen (a product of liver metabolism of Tam in vivo) in the experiment of Xu et al may be responsible for these differences.
Given the fact that in ERα-negative cells, E2 and Tam displayed similar effects, our data provide evidence that the upregulation of SSTR2-mRNA by ER agonists is, at least in part, independent of ERα. Contrary to previous beliefs, the classical ER-negative breast cancer cell line MDA-MB-231 expresses low levels of ERβ but no ERα [39]. Kimura et al [41] have recently shown estrogen sensitive sequences in the promoter region of SSTR2 gene. Further studies are required to delineate the exact molecular mechanisms involved. Given that the affinity of Tam for the ER is lower than that of its metabolite OH-tamoxifen or E2 [42], it is not surprising that simultaneous treatment with equimolar concentrations of E2 and Tam resulted in down-regulation of SSTR1 as seen when E2 was given alone. It is remarkable; however, that the effect was stronger with the combined treatment and that, in the case of SSTR2, the combined E2 and Tam treatment resulted in loss of the upregulatory effect of either compound alone. These effects probably involve interactions between ERs and their coregulators, and may include E2 upregulation of ERβ expression and induction of the formation of ERα/ERβ heterodimers with decreased transcriptional activity [43,44].
Of interest, we detected surprising changes in the pattern of subcellular distribution of the receptors as a result of the treatments used. Basically, Tam changed the pattern of expression of SSTR1 and SSTR2 in ZR-75-1 cells from patchy to a more homogeneous cell surface distribution. In contrast, both Tam and E2 altered the homogeneous distribution of receptors in T47D cells to a more irregular patchy distribution. These changes in SSTR1 and SSTR2 immunoreactivity may involve receptor homo- or heterodimerization, receptor complex dissociation and/or internalization [45,46]. These could be a specific effect of ER activation or the result of changes in the fluidity and/or composition of the cell membrane. The ability of Tam to affect membrane stability by decreasing its fluidity has been previously shown. There are reports of Tam decreasing synthesis of glycosphingolipids by blocking synthesis of their precursor, glycosylceramide, thereby causing intracellular ceramide accumulation and membrane disruption [47].
In summary, our data show that SSTR1 and SSTR2 expression in breast cancer cells is modulated by Tam and E2 in a cell-specific manner. SSTR1 expression is upregulated by Tam while E2 and combined treatment preferentially downregulate this receptor in ER-positive cells. In ERα-negative cells, both ligands caused upregulation of SSTR1. The changes in subcellular distribution of SSTR1 and SSTR2 were favourable in only one of the ER-positive cell lines investigated (ZR-75-1). These cells also showed a stronger growth inhibitory response to SSTR1-agonist following pretreatment with Tam. Therefore, SSTR1 activation appears to be a potentially important mechanism for tumor cell growth control. For this reason, enhancing SSTR expression in tumor cells could make them more susceptible to growth inhibition by SSTR agonists, particularly SSTR1 agonists, as anti-neoplastic treatment. Our study suggests that SERMs like Tam may have such effects. These results have important implications in our understanding of the role of SSTRs in ER responsiveness of breast cancer. Moreover, it might not be premature to anticipate that, based on our results, some breast cancer patients may benefit from SSTR1-selective agonist therapy in combination with Tam, in an estrogen-depleted milieu.
Abbreviations
E2, estradiol; ER, estrogen receptor; FBS, fetal bovine serum; GH, growth hormone; IGF-1, insulin-like growth factor 1; IRS, insulin responsive substrate 1; mRNA, messenger ribonucleic acid; MTT, 3-(4,5-dimethlthiazolyl-2)-2, 5-diphenyltetrazolium bromide; RT-PCR, reverse transcriptase polymerase chain reaction; SERM, selective estrogen receptor modulator; SST, somatostatin; SSTR, SST receptor; Tam, tamoxifen; TGF, transforming growth factor; OD, optical density.
Acknowledgements
This work was supported by grants from the Canadian Institutes of Health Research # MOP-6196 and MOP-10411(UK). We thank Dr. S. P. Rohrer and Dr. J.M. Schaeffer (Merck Research Laboratories, Rahway, NJ) for the kind gift of SSTR1 selective agonist, Maria Correia for secretarial help and Heather Watt for the critical reading of this manuscript. J.A. Rivera was supported by a CIHR postdoctoral fellowship and McGill University Health Centre Research Institute.
==== Refs
Patel YC Somatostatin and its receptor family Frontiers in Neuroendocrinology 1999 20 157 98 10433861 10.1006/frne.1999.0183
Patel YC Molecular pharmacology of somatostatin receptor subtypes J Endocrinol Invest 1997 20 348 367 9294784
Reubi JC Waser B Schaer JC Laissue JA Somatostatin receptor sst1-sst5 expression in normal and neoplastic human tissues using receptor autoradiography with subtype-selective ligands Eur J Nucl Med 2001 28 836 46 11504080 10.1007/s002590100541
Nelson J Cremin M Murphy RF Synthesis of somatostatin by breast cancer cells and their inhibition by exogenous somatostatin and sandostatin Br J Cancer 1989 59 739 42 2567608
Pagliacci MC Tognellini R Grignani F Nicoletti I Inhibition of human breast cancer cell (MCF-7) growth in vitro by the somatostatin analog SMS 201–995: effects on cell cycle parameters and apoptotic cell death Endocrinology 1991 129 2555 2562 1935786
Weckbecker G Tolcsvai L Stolz B Pollak M Bruns C Somatostatin analogue octreotide enhances the antineoplastic effects of tamoxifen and ovariectomy on 7, 12-Dimethylbenz(a)anthracene-induced rat mammary carcinoma Cancer Res 1994 54 6334 6337 7987824
de Herder WW Lamberts SW Somatostatin and somatostatin analogues: diagnostic and therapeutic uses Curr Opin Oncol 2002 14 53 57 11790981 10.1097/00001622-200201000-00010
Schally AV Comaru-Schally AM Nagy A Kovacs M Szepeshazi k Plonowski A Varga JL Halmos G Hypothalamic hormones and cancer Front Neuroendocrinol 2001 22 248 291 11587553 10.1006/frne.2001.0217
Sachdev D Yee D The IGF system and breast cancer Endocr Relat Cancer 2001 8 197 209 11566611 10.1677/erc.0.0080197
Ingle JN Suman VJ Kardinal CG Krook JE Mailliard JA Veeder MH Loprinzi CL Dalton RJ Hartmann LC Conover CA Pollak MN A randomized trial of tamoxifen alone or combined with octreotide in the treatment of women with metastatic breast carcinoma Cancer 1999 85 1284 1292 10189133 10.1002/(SICI)1097-0142(19990315)85:6<1284::AID-CNCR10>3.0.CO;2-P
Bajetta E Procopio G Ferrari L Martinetti A Zilembo N Catena L Alu M Della TS Alberti D Buzzoni R A randomized, multicenter prospective trial assessing long-acting release octreotide pamoate plus tamoxifen as a first line therapy for advanced breast carcinoma Cancer 2002 94 299 304 11900215 10.1002/cncr.10239
van Eijck CH Krenning EP Bootsma A Oei HY van Pel R Lindemans J Jeekel J Reubi JC Lamberts SW Somatostatin-receptor scintigraphy in primary breast cancer Lancet 1994 343 640 643 7906813 10.1016/S0140-6736(94)92637-9
Alberini JL Meunier B Denzler B Devillers A Tass P Dazord L Le Simple T Laissue J de Jong R Le Cloirec J Reubi JC Bourguet P Somatostatin receptor in breast cancer and axillary nodes: study with scintigraphy, histopathology and receptor autoradiography Breast Cancer Res Treat 2000 61 21 32 10930087 10.1023/A:1006447325077
Xu Y Song J Berelowitz M Bruno JF Estrogen regulates somatostatin receptor 2 messenger ribonucleic acid expression in human breast cancer cells Endocrinology 1996 137 5634 5640 8940394 10.1210/en.137.12.5634
Xu Y Berelowitz M Bruno JF Dexamethasone regulates somatostatin receptor subtype messenger ribonucleic acid expression in rat pituitary GH4C1 cells Endocrinology 1995 136 5070 5075 7588243 10.1210/en.136.11.5070
Visser-Wisselaar HA Van Uffelen CJC Van Koetsveld PM Lichtenauer-Kaligis EGR Waaijers AM Uitterlinden P Mooy DM Lamberts SWJ Hofland LJ 17-β-estradiol-dependent regulation of somatostatin receptor subtype expression in the 7315b prolactin secreting rat pituitary tumor in vitro and in vivo Endocrinology 1997 138 1180 1189 9048625 10.1210/en.138.3.1180
Djordjijevic D Zhang J Priam M Viollet C Gourdji D Kordon C Epelbaum J Effect of 17β-estradiol on somatostatin receptor expression and inhibitory effects on growth hormone and prolactin release in rat pituitary cells Endocrinology 1998 139 2272 2277 9564833 10.1210/en.139.5.2272
Kimura N Tomizawa S Arai KN Kimura N Chronic treatment with estrogen up-regulates expression of sst2 messenger ribonucleic acid (mRNA) but down-regulates expression of sst5 mRNA in rat pituitaries Endocrinology 1998 139 1573 1580 9528936 10.1210/en.139.4.1573
Perry RR Kang Y Greaves B Effects of tamoxifen on growth and apoptosis of estrogen-dependent and independent human breast cancer cells Ann Surg Oncol 1995 2 238 245 7641021
Gelmann EP Tamoxifen for the treatment of malignancies other than breast and endometrial carcinoma Semin Oncol 1997 24 65 70
Hoelting T Siperstein AE Duh QY Clark OH Tamoxifen inhibits growth, migration, and invasion of human follicular and papillary thyroid cancer cells in vitro and in vivo J Clin Endocrinol Metab 1995 80 308 313 7829632 10.1210/jc.80.1.308
Tavassoli M Soltaninia J Rudnicka J Mashanyare D Johnson N Gaken J Tamoxifen inhibits the growth of head and neck cancer cells and sensitizes these cells to cisplatin induced-apoptosis: role of TGF-β1 Carcinogenesis 2002 23 1569 1575 12376463 10.1093/carcin/23.10.1569
Pollack IF DaRosso RC Robertson PL Jakacki RL Mirro JR JrBlatt J Nicholson S Packer RJ Allen JC Cisneros A Jordan VC A phase I study of high-dose tamoxifen for the treatment of refractory malignant gliomas of childhood Clin Cancer Res 1997 3 1109 1115 9815790
Holownia A Braszko JJ Tamoxifen cytotoxicity in hepatoblastoma cells stably transfected with human CYP3A4 Biochem Pharmacol 2004 67 1057 1064 15006542 10.1016/j.bcp.2003.10.027
Butta A MacLennan K Flanders KC Sacks NPM Smith I McKinna A Dowsett M Wakefield LM Sporn MB Baum M Induction of transforming growth factor beta 1 in human breast cancer in vivo following tamoxifen treatment Cancer Res 1992 52 4261 4264 1322240
Ho GH Ji CY Phang BH Lee KO Soo KC Ng EH Tamoxifen alters levels of serum insulin-like growth factors and binding proteins in postmenopausal breast cancer patients: a prospective paired cohort study Ann Surg Oncol 1998 5 361 367 9641459
Guvakova MA Surmacz E Tamoxifen interferes with the insulin-like growth factor I receptor (IGF-IR) signalling pathway in breast cancer cells Cancer Res 1997 57 2606 10 9205064
Molloy CA May FE Westley BR Insulin receptor substrate-1 espression is regulated by estrogen in the MCF-7 human breast cancer cell line J Biol Chem 2000 275 12565 12571 10777546 10.1074/jbc.275.17.12565
Froidevaux S Eberle AN Somatostatin analogs and radiopeptides in cancer therapy Biopolymers 2002 66 161 183 12385036 10.1002/bip.10256
Buchan AM Lin CY Choi J Barber DL Somatostatin, acting at receptor subtype 1, inhibits Rho activity, the assembly of actin stress fibers, and cell migration J Biol Chem 2002 277 28431 28438 12045195 10.1074/jbc.M201261200
Zatelli MC Tagliati F Piccin D Taylor JE Culler MD Bondanelli M degli Uberti EC Somatostatin receptor subtype 1-selective activation reduces cell growth and calcitonin secretion in a human medullary thyroid carcinoma cell line Biochem Biophys Res Commun 2002 297 828 834 12359227 10.1016/S0006-291X(02)02307-0
Zatelli MC Piccin D Tagliati F Ambrosio MR Margutti A Padovani R Scanarini M Culler MD degli Uberti EC Somatostatin receptor subtype 1 selective activation in human growth hormone (GH)- and prolactin (PRL)-secreting pituitary adenomas: effects on cell viability, GH, and PRL secretion J Clin Endocrinol Metab 2003 88 2797 2802 12788890 10.1210/jc.2002-021825
Aavik E Luoto NM Petrov L Aavik S Patel YC Hayry P Elimination of vascular fibrointimal hyperplasia by somatostatin receptor 1,4 selective agonist FASEB J 2002 16 724 726 11923215
Lahlou H Saint-Laurent N Esteve JP Eychene A Pradayrol L Pyronnet S Susini C sst2 somatostatin receptor inhibits cell proliferation though Ras-, Rap1-, and B-Raf-dependent ERK2 activation J Biol Chem 2003 278 39356 39371 12878607 10.1074/jbc.M304524200
Guillermet J Saint-Laurent N Rochaix P Cuvillier P Lavade T Achally AV Pradayrol L Buscail L Susini C Bousquet C Somatostatin receptor subtype 2 sensitizes human pancreatic cancer cells to death ligand-induced apoptosis Proc Natl Acad Sci USA 2003 100 155 160 12490654 10.1073/pnas.0136771100
Rohrer SP Schaeffer JM Identification and characterization of subtype selective somatostatin receptor agonists J Phsiol Paris 2000 94 211 215 10.1016/S0928-4257(00)00215-1
Kumar U Laird D Srikant CB Escher E Patel YC Expression of the five somatostatin receptor (SSTR1-5) subtypes in rat pituitary somatotrophes: quantitative analysis by double-layer immunofluorescence confocal microscopy Endocrinology 1997 138 4473 4476 9322965 10.1210/en.138.10.4473
Kumar U Sasi R Suresh S Patel A Thangaraju M Metrakos P Patel SC Patel YC Subtype-selective expression of the five somatostatin receptors (SSTR1-5) in human pancreatic islet cells Diabetes 1999 48 77 85 9892225
Vladusic EA Hornby AE Guerra-Vladusic FK Lakins J Lupu R Expression and regulation of estrogen receptor β in human breast tumors and cell lines Oncol Reports 2000 7 157 167
Schafer JM Lee ES O'Regan RM Yao K Jordan VC Rapid development of tamoxifen-stimulated mutant p53 breast tumors (T47D) in athymic mice Clin Cancer Res 2000 6 4373 4380 11106256
Kimura N Tomizawa S Arai KN Osamura RY Kimura N Characterization of 5'-flanking region of rat somatostatin receptor sst2 gene: transcriptional regulatory elements and activation by Pitx1 and estrogen Endocrinology 2001 142 1427 1441 11250922 10.1210/en.142.4.1427
Martel C Provencher L Li X St Pierre A Leblanc G Gauthier S Merand Y Labrie F Binding characteristics of novel nonsteroidal antiestrogens to the rat uterine estrogen receptors J Steroid Mol Biol 1998 64 199 205 10.1016/S0960-0760(97)00192-1
Hall JM McDonnell DP The estrogen receptor beta-isoform (ER beta) of the human estrogen receptor modulates ER alpha transcriptional activity and is a key regulator of the cellular response to estrogens and antiestrogens Endocrinology 1999 140 5566 5578 10579320 10.1210/en.140.12.5566
Bateau-Lozano H Angelino M Laredo B Melanin J Perrot-Applanat M Transcriptional regulation of vascular endothelial growth factor by estradiol and tamoxifen in breast cancer cells: a complex interplay between estrogen receptors alpha and beta Cancer Res 2002 62 4977 4984 12208749
Rocheville M Lange DC Kumar U Sasi R Patel RC Patel YC Subtypes of the somatostatin receptor assembles as functional homo-and heterodimers J Biol Chem 2000 275 7862 7869 10713101 10.1074/jbc.275.11.7862
Csaba Z Dournaud P Cellular biology of somatostatin receptors Neuropeptides 2001 35 1 23 11346306 10.1054/npep.2001.0848
Mandlekar S Knog A-NT Mechanisms of tamoxifen-induced apoptosis Apoptosis 2001 6 469 477 11595837 10.1023/A:1012437607881
|
16018813
|
PMC1184089
|
CC BY
|
2021-01-04 16:39:20
|
no
|
J Carcinog. 2005 Jul 14; 4:10
|
utf-8
|
J Carcinog
| 2,005 |
10.1186/1477-3163-4-10
|
oa_comm
|
==== Front
Cardiovasc UltrasoundCardiovascular Ultrasound1476-7120BioMed Central London 1476-7120-3-151595816910.1186/1476-7120-3-15ResearchA comparison of ultrasound measurements to assess carotid atherosclerosis development in subjects with and without type 2 diabetes Pollex Rebecca L [email protected] J David [email protected] Andrew A [email protected] Aaron [email protected] Anthony JG [email protected] Bernard [email protected] Stewart B [email protected] Robert A [email protected] Robarts Research Institute, London, Ontario, Canada2 Department of Medicine, University of Western Ontario, London, Ontario, Canada3 Department of Medicine, University of Toronto and Leadership Sinai Centre for Diabetes, Mount Sinai Hospital, Toronto, Ontario, Canada4 Thames Valley Family Practice Research Unit, University of Western Ontario, London, Ontario, Canada2005 15 6 2005 3 15 15 17 5 2005 15 6 2005 Copyright © 2005 Pollex et al; licensee BioMed Central Ltd.2005Pollex et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Subjects with type 2 diabetes are at an increased risk of vascular complications. The use of carotid ultrasound remains an attractive, non-invasive method to monitor atherosclerotic disease progression and/or response to treatment in patients with type 2 diabetes, with intima-media thickness routinely used as the gold standard to detect pathology. However, alternative measurements, such as plaque area or volume, may represent a potentially more powerful approach. Thus, the objective of this study was to compare the traditional intima-media thickness measurement against the novel total plaque volume measurement in analyzing carotid atherosclerosis development in individuals with type 2 diabetes.
Methods
The case-control study included 49 Oji-Cree adults with diabetes or impaired glucose tolerance, aged 21–69, and 49 sex- and age-matched normoglycemic subjects. At baseline, metabolic variables were measured, including body mass index, waist circumference, total cholesterol:high density lipoprotein ratio, plasma triglycerides, plasma glucose, and serum insulin. Carotid ultrasound measurements, 7 years later, assessed carotid arterial intima-media thickness and total plaque volume.
Results
At baseline, the two groups were well matched for smoking habits, hypertension, body mass index, and waist circumference. Differences were noted in baseline measurements of total cholesterol:high density lipoprotein (P = 0.0006), plasma triglycerides (P < 0.0001) and fasting glucose (P < 0.0001). After seven years, carotid ultrasound scans revealed that total plaque volume measurements (P = 0.037), but not intima-media thickness measurements, were higher in subjects with diabetes/impaired glucose tolerance compared to the normoglycemic controls. Correlation between intima-media thickness and total plaque volume was moderate. Based on our study findings, to achieve power levels >0.70 when comparing intima-media thickness measurements for diabetics versus non-diabetics, thousands of study subjects are required. For comparing total plaque volume measurements, only hundreds of study subjects are required.
Conclusion
The development of atherosclerotic plaque is greater in subjects with diabetes/impaired glucose tolerance. Total plaque volume appears to capture the atherosclerotic disease burden more effectively in subjects with type 2 diabetes, and would be an appropriate outcome measure for studies aimed at changing the diabetic milieu.
==== Body
Background
Macrovascular disease is the predominant contributor to morbidity and mortality in patients with type 2 diabetes [1]. Non-invasive assessment of arterial morphology using carotid ultrasound (US) represents an attractive tool to monitor progression and/or response to treatment in patients with type 2 diabetes. The traditional endpoint is intima-media thickness (IMT). Recently, however, newer measurements such as assessment of plaque area or volume represent a potentially more powerful approach, since this evaluates all plaques in the carotid system, and predicts clinical events somewhat more strongly than does IMT [2]. Thus, the objective of this case-control study was to compare the traditional IMT measurement versus the novel total plaque volume (TPV) measurement in analyzing carotid atherosclerosis development in individuals with type 2 diabetes.
Methods
Study Sample
Study participants were from the Oji-Cree community of Sandy Lake, Ontario, an isolated reserve located at the 55th parallel of latitude, in the subarctic boreal forest of central Canada. Baseline demographic, clinical, and biochemical attributes were gathered during the Sandy Lake Health and Diabetes Project of 1993–1995, a prevalence study of type 2 diabetes [3]. Seven hundred and twenty eight members of this community (72% of the total population) aged 10 years and above, participated in the original survey. In a follow-up study initiated in 2001 [4], 278 adults free of coronary heart disease had US assessment of the carotid arteries. Of these, 161 had participated in the original prevalence study and had baseline measurements. For the current analysis, 49 subjects, 46 with type 2 diabetes and 3 with impaired glucose tolerance (IGT), were selected and matched for sex and age (± 3 years) with a normoglycemic control subject. Of the subjects with diabetes, 43.5% were receiving oral medication, and 4.4% were receiving insulin. For simplicity, from this point forward the 46 subjects with type 2 diabetes and 3 with IGT will be referred to as the "diabetic" group. Signed informed consent was obtained from all participants and study approval was granted by both the Sandy Lake First Nation Band Council and the institutional review boards of the University of Toronto and the University of Western Ontario.
Clinical and biochemical baseline analysis
Body weight, height, waist circumference, and blood pressure were measured by standardized procedures [3]. Hypertensive individuals were defined as subjects having either a blood pressure reading >140/90 mmHg or taking anti-hypertensive medication. Measurements of fasting blood analytes, including triglycerides, insulin, lipoproteins, and total cholesterol were performed as described [3].
Ultrasound examination
All subjects were examined using an HDI 5000 scanner equipped with Sono-CT compound imaging and a L12-5 transducer (Advanced Technology Laboratories, Bothell, Washington) that had been flown to the community and housed within the Diabetes Research Center. Common carotid US images for all participants were gathered over a 4-week period and from this data, IMT, total plaque area (TPA) and TPV measurements were determined. TPA was strongly correlated with TPV in these subjects (r = 0.921, P < 0.0001), and thus for simplicity, only TPV measurements were compared against IMT.
IMT measurement
IMT was determined as previously described [5,6]. Briefly, a single observer, blinded to subjects' vascular risk, measured combined thickness of intima and media of the far wall of both common carotid arteries. Images were recorded from an anterolateral longitudinal view. The still images were analyzed using computerized edge-detection software (Prowin™) [7]. Using a step-wise algorithm, conditional sets of "edges" (consisting of lumen-intima and media-adventitia echoes) were located within the image and then tested for "edge strength", with the subsequent deletion of weak edge points. Once all acceptable edge points were identified, boundary gaps were filled by linear interpolation. The distance between lumen-intima and media-adventitia boundaries was then measured to calculate IMT. Mean IMT was computed from 120 measurements over a 10 mm span ending 5 mm proximal to the transition between the common carotid and bulb regions. Intra- and inter-operator coefficients of variation of 3.0 and 3.1%, respectively and intra- and inter-operator intraclass correlations were both 0.97 [n = 50] (both P < 0.01).
TPV measurement
TPV was determined as previously described [5,6]. Briefly, 3D ultrasound images were acquired with a mechanical linear scanning system and analyzed with L3Di visualization software [Life Imaging Systems Inc., London, Ontario]. Plaque volumes were measured using manual planimetry: each 3D image was 'sliced' transversely at an inter-slice distance of 1 mm, moving from one plaque edge to the other. Plaque boundaries were traced using a mouse driven cross-haired cursor. Slice areas were summed and multiplied by inter-slice distance to calculate plaque volume. For this analysis, TPV was defined as the sum of all plaque volumes on one side between the clavicle and angle of the jaw. Intra- and inter-observer reliability were 0.94 [n = 40] and 0.93 [n = 40], respectively (both P < 0.01).
Statistical analysis
SAS version 8.2 (SAS Institute, Cary, NC) was used for all statistical comparisons. Data are presented as means ± SE. The distribution of BMI, plasma total cholesterol, triglycerides, high density lipoprotein (HDL), and serum insulin, were non-normal in this data set, and thus were logarithmically transformed (natural log) and subjected to analysis of normality. IMT and TPV were also normalized using the inverse transformation of IMT and the square root transformation of TPV. The transformed variables were used for parametric statistical analyses, but the untransformed values are presented in Table 1. For continuous variables, differences between the groups were tested by the Student's t test; categorical variables were tested by γ2 analysis. Statistical significance was taken at nominal P < 0.05 for all comparisons. Correlation analysis between IMT and TPV was performed using Pearson correlation analysis.
Table 1 Clinical and biochemical attributes of Oji-Cree at baseline and carotid measurements after 7 years
Diabetic subjects Non-diabetic subjects P-value
number/females 49/26 49/26
attributes at screening
age (years) 40.3 ± 1.8 40.4 ± 1.8 NS (0.96)
duration of diabetes (years) 2.20 ± 0.62 - -
current smokers (%) 18.4 10.2 NS (0.25)
hypertensive (%) 36.7 32.7 NS (0.67)
antihypertensive treatment (%) 16.3 8.2 NS (0.22)
body mass index (kg/m2) 29.6 ± 0.5 29.5 ± 0.6 NS (0.75)
waist circumference (cm) 101 ± 1.4 99.8 ± 1.5 NS (0.63)
TC:HDL ratio 4.95 ± 0.18 4.10 ± 0.17 0.0006
plasma triglycerides (mmol/L) 2.25 ± 0.14 1.53 ± 0.11 <0.0001
plasma glucose (mmol/L) 10.79 ± 0.63 5.53 ± 0.07 <0.0001
serum insulin (pmol/L) 157 ± 12 133 ± 9 NS (0.16)
time elapsed since screening (years) 7.34 ± 0.10 7.33 ± 0.10 NS (0.96)
mean IMT (μm) 795 ± 19 789 ± 26 NS (0.49)
mean TPV (mm3) 109.9 ± 23.0 64.0 ± 17.0 0.037
Data are means ± SE unless otherwise indicated.
Abbreviations: TC, total cholesterol; HDL, high density lipoprotein; IMT, intima-media thickness; TPV, total plaque volume; NS, not significant
Hypothetical sample sizes were calculated using the online calculator for normal power calculations (normal distribution 2-sample equal variances) found at the UCLA Department of Statistics website [8]. This statistical tool calculates the sample size for two-sided tests of hypotheses on normal means, when the common population standard deviation is known, using the following formula:
where n1 and n2 are the sample sizes of the two groups, uα/2 and uβ are the lower limits of the cumulative standard normal probability integrals, σ is the known common standard deviation, δ0 is the least favourable non-negative difference consistent with the test hypothesis, and δ1 is difference in the population means [9].
Using the normalized transformed means from the case-control study, the mean standard deviation as the common standard deviation (SD), and a significance level of 0.05, sample sizes were calculated. Transformed means and standard deviations were 1.29 (diabetic) vs 1.33 (non-diabetic), SD 0.237, for IMT (inverse transformation), and 7.81 (diabetic) vs 4.94 (non-diabetic), SD 6.71, for TPV (square root transformation). Power was tested at 0.70, 0.80 and 0.90.
Results
Clinical and biochemical attributes of Oji-Cree subjects at baseline
As presented in Table 1, at the initial time of screening, the average age of the study participants was 40.3 ± 1.8 years for subjects with diabetes, and 40.4 ± 1.8 years for the control subjects. 53.1% were females. Subjects with diabetes were relatively newly diagnosed, with an average diabetes duration of 2.20 ± 0.62 years. Comparing the baseline characteristics of the diabetic and non-diabetic subjects, there was no significant difference in terms of smoking, hypertension, body mass index (BMI), waist circumference, or serum insulin concentrations. However, both TC:HDL ratio and plasma triglyceride and glucose concentrations were elevated in the diabetic subjects (P = 0.0006, P < 0.0001, and P < 0.0001, respectively).
Carotid plaque measurements after 7 years
Following 7 years, mean IMT values were slightly elevated for subjects with diabetes (795 ± 19 μm) vs non-diabetic subjects (789 ± 26 μm), however the difference was not significant (P = 0.49). Mean TPV measurements, however, were significantly higher in subjects with diabetes (109.9 ± 23.0 mm3 vs 64.0 ± 17.0 mm3, P = 0.037). The simple Pearson correlation coefficient between untransformed IMT and TPV was 0.524 (P < 0.0001) (Figure 1); the simple Pearson correlation coefficient between transformed values (1/IMT and square root of TPV) was approximately the same (r = -0.556, P < 0.0001).
Figure 1 Correlation between carotid arterial total plaque volume (TPV) and intima-media thickness (IMT). Subjects with type 2 diabetes or IGT are represented with black dots (n = 49), while non-diabetic control subjects are represented with white dots (n = 49). The Pearson correlation coefficient (r) = 0.524.
Comparison of sample sizes required for hypothetical studies
Using the mean IMT and TPV results from our case-control study as primary endpoints, we determined the sample sizes that would be required for a hypothetical study to achieve a statistical power of 0.70, 0.80 and 0.90. As shown in Table 2, to detect a 6 μm difference in IMT, a total of 1202 subjects would be required for a 0.70 power level, 1528 subjects for a 0.80 power level, and 2044 subjects for a 0.90 power level. To detect a 45.9 mm3 difference in TPV, a total of 136 subjects would be required for a 0.70 power level, 172 subjects for a 0.80 power level, and 230 subjects for a 0.90 power level.
Table 2 Comparison of sample sizes required for hypothetical trials with mean IMT or TPV as primary endpoints
Endpoint Significance Level Power Ndiabetic Nnon-diabetic N total
IMT (inverse): 6 μm difference
0.05 0.70 601 601 1202
0.05 0.80 764 764 1528
0.05 0.90 1022 1022 2044
TPV (square root): 45.9 mm3 difference
0.05 0.70 68 68 136
0.05 0.80 86 86 172
0.05 0.90 115 115 230
Abbreviations: IMT, intima-media thickness; TPV, total plaque volume
Discussion
We report: 1) elevated TPV for diabetic subjects vs non-diabetic subjects following a 7 year period (P = 0.037); 2) increased sensitivity of the TPV measurement in comparison to IMT measurements for diabetic subjects.
A previous study by Hunt et al., convincingly showed that early atherogenesis is present before the onset of diabetes, and thus is not solely dependent on the clinical manifestation of diabetes [10], but rather, both conditions (diabetes and cardiovascular disease) originate from a "common soil" of pro-inflammatory and pro-atherogenic risk factors [11]. Our study examined atherosclerosis burden after the diagnosis of diabetes had been made and found that diabetic subjects had TPV measurements that were 1.7-fold higher than non-diabetic subjects (P = 0.037). Similar observations have been made previously, such as in the Insulin Resistance Atherosclerosis Study (n = ~1200), where subjects with diabetes had increased carotid wall thickness at baseline (~70 μm increase in common carotid, ~130 μm increase in internal carotid) [12], and, over a five year time period, had IMT progression rates approximately twice as high as non-diabetic subjects (7.2 ± 1.9 vs 3.8 ± 1.3 μm/year) [13]. The Bruneck Study (n = 826) found that type 2 diabetes was a strong independent predictor (OR = 5.0, P < 0.001) of US-determined, advanced stenotic atherosclerosis, defined by >40% lumenal narrowing [14].
Significant differences were also noted in the lipid profile of diabetic subjects vs controls, with a greater TC:HDL ratio and elevated triglycerides observed for those with diabetes. Glucose intolerance has been previously reported as an independent predictor of both triglycerides and HDL cholesterol [15]. This worsening of lipids with glucose intolerance may potentially explain the differences between the two groups in terms of plaque volume progression.
While a significant difference was found for TPV, no significant difference was found for IMT between diabetic and non-diabetic subjects, although IMT tended to be greater for diabetic subjects. The lack of a significant difference undoubtedly is related to the low number of subjects, but it is apparent that it may also be due to the relative insensitivity of carotid IMT as a surrogate marker for atherosclerosis in patients with type 2 diabetes. The potential increased sensitivity for TPV was reflected by our finding of a significant difference for a relatively small study sample. An important feature for enhanced sensitivity is found in the wider dynamic scale ranges for TPV compared to IMT: ~90% of the IMT measurements fall within a relatively narrow 0.55–1.0 mm range, whereas ~60% of TPV values fall within a range of 5–500 mm3. Thus, the dynamic range of measurements varied by ~100-fold for TPV compared to ~2-fold for the IMT. Furthermore, the quantity being measured (mm3 for TPV vs mm for IMT) is much larger for TPV, so that in relation to the resolution of the ultrasound method, TPV is much easier to quantify both accurately and reliably.
In designing studies it may be worthwhile to consider using TPV in addition to the traditional IMT measurement, as a primary endpoint, due to its potentially greater sensitivity and discrimination, which may have the benefit of greater statistical power, allowing for the use of a small sample size. For example, to observe a significant difference in IMT using values similar to those observed in this case-control study, (ie. 6 μm difference in mean IMT) (Table 2), thousands of subjects are required to achieve a statistical power of even 0.70. In contrast, only a few hundred subjects are required to observe a statistical power of 0.90 for the TPV difference seen in our study (45.9 mm3). Performing studies with over a thousand subjects [10,13] imposes limitations and difficulties, which could be minimized by using a more sensitive technique such as TPV measurement. Studies on the effects of therapy on atherosclerosis using TPV as the outcome have already been effectively carried out with much smaller sample sizes. Ainsworth et al. [16] have shown significant differences between active atorvastatin and placebo in 3 months, with a sample size of only 20 per group.
IMT and TPV, however, are not interchangeable. The correlation between IMT and TPV in this study, although statistically significant, was moderate (r<0.7) and, as has been noted previously, these different US-derived measures of carotid artery morphology likely represent distinct attributes of atherosclerosis [5]. IMT may reflect wall hyperplasia or hypertrophy related to hypertension [17] and TPV may reflect the later stages of plaque formation and the total carotid disease burden in a subject [5]. This may be more relevant for the disease process of diabetes. It is also important to note that in our previous work [5] IMT correlated better with hypertension and age, and as blood pressure and age were balanced between the groups, this might explain the lack of difference observed here for IMT. It may be the case that IMT would capture the atherosclerotic disease burden more effectively in hypertensive diabetics, and would be a more appropriate outcome measure for studies aimed at improving hypertension in diabetics. These implications must be taken into consideration when designing and analysing a study, and more research is needed to provide a complete understanding of how these US measures fit into the pathophysiology of atherosclerotic disease.
Considering that by nature atherosclerotic plaque is not evenly distributed along the arterial wall, it is logical to develop methods that will attempt to quantify the total plaque burden more accurately. With the relatively larger volumes being measured for TPV assessment vs IMT or plaque area measurements, there are the accompanying benefits of more statistical power and less patients required per study, and also the potential benefit of less time required to observe significant differences between study groups. However, at present, using TPV as an assessment tool remains a labour-intensive task and has the additional disadvantage of not yet being widely used and standardized. Furthermore, in studies among children or other very young subjects, ultrasound evaluation of the arteries may be limited to IMT simply because plaque would not yet be developed in most cases. However, some plaque can be identified in most subjects above age 35 or 40 [2].
Conclusion
In conclusion, this case-control study showed significantly greater carotid TPV measurements in subjects with type 2 diabetes vs subjects without type 2 diabetes, after a 7-year period. This supports previous reports of increased carotid ultrasound analytes for persons with type 2 diabetes, but additionally, this study highlights the effectiveness of TPV as a marker for atherosclerotic disease burden in diabetes, and encourages the further use and development of this robust measurement. Currently the determination of TPV requires labour-intensive manual tracing, which has proven to be an accurate and reliable measurement [18,19]. Semi-automated methods are now in development. However, for TPV to be used as a universally implemented research/clinical tool, further studies will be needed to clarify the relationships among IMT, TPV and perhaps coronary plaque volume measured by intravascular ultrasound.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
RLP participated in the design of the study, analysis of the data, and writing of the manuscript. JDS, AAH, AF, AJGH, BZ, and SBH provided patients and data for the study, and assisted with manuscript revisions. RAH participated in the design of the study and writing of the manuscript. All authors read and approved the final manuscript.
Acknowledgements
The authors acknowledge Khalid Al-Shali for his work in measuring the plaque volume of the study subjects. The authors also gratefully acknowledge the chief, council and community members of Sandy Lake First Nation and the Sandy Lake community surveyors (Ken Goodwin, Edith Fiddler, Louisa Kakegamic, Tina Noon, Madeline Kakegamic, Elda Anishinabie, Annette Rae, Connie Kakegamic, and Mary Mamakeesic), whose partnership and co-operation was essential in the design and implementation of this project.
==== Refs
Centers for Disease Control and Prevention, US Department of Health and Human Services Diabetes: Diabling, Deadly, and on the Rise 2005 Atlanta, GA
Spence JD Ultrasound measurement of carotid plaque as a surrogate outcome for coronary artery disease Am J Cardiol 2002 89 10B 15B; discussion 15B-16B 11879661 10.1016/S0002-9149(01)02327-X
Hanley AJG Harris SB Barnie A Gittelsohn J Wolever TMS Logan A Zinman B The Sandy Lake Health and Diabetes Project: design, methods and lessons learned Chronic Dis Canada 1995 16 149 156
Hanley AJG Harris SB Mamakeesick M Goodwin K Fiddler E Hegele RA McLaughlin JR Zinman B Complications of type 2 diabetes among Native Canadians: increasing our understanding of prevalence and risk factors Canadian J Diabetes 2003 27 455 463
Al-Shali K House AA Hanley AJ Khan HM Harris SB Mamakeesick M Zinman B Fenster A Spence JD Hegele RA Differences between carotid wall morphological phenotypes measured by ultrasound in one, two and three dimensions Atherosclerosis 2005 178 319 325 15694940 10.1016/j.atherosclerosis.2004.08.016
Hegele RA Al-Shali K Khan HM Hanley AJG Harris SB Mamakeesick M Zinman B Fenster A Spence JD House AA Carotid ultrasound in one, two and three dimension Vasc Dis Prevention 2005 2 87 92 10.2174/1567270052774386
Selzer RH Hodis HN Kwong-Fu H Mack WJ Lee PL Liu CR Liu CH Evaluation of computerized edge tracking for quantifying intima-media thickness of the common carotid artery from B-mode ultrasound images Atherosclerosis 1994 111 1 11 7840805 10.1016/0021-9150(94)90186-4
UCLA Department of Statistics
Mace AE Sample-Size Determination 1974 Huntington, NY , Robert E. Krieger Publishing Company 77 80
Hunt KJ Williams K Rivera D O'Leary DH Haffner SM Stern MP Gonzalez Villalpando C Elevated carotid artery intima-media thickness levels in individuals who subsequently develop type 2 diabetes Arterioscler Thromb Vasc Biol 2003 23 1845 1850 12958039 10.1161/01.ATV.0000093471.58663.ED
Stern MP Diabetes and cardiovascular disease. The "common soil" hypothesis Diabetes 1995 44 369 374 7698502
Wagenknecht LE D'Agostino RBJ Haffner SM Savage PJ Rewers M Impaired glucose tolerance, type 2 diabetes, and carotid wall thickness: the Insulin Resistance Atherosclerosis Study Diabetes Care 1998 21 1812 1818 9802726
Wagenknecht LE Zaccaro D Espeland MA Karter AJ O'Leary DH Haffner SM Diabetes and progression of carotid atherosclerosis: the insulin resistance atherosclerosis study Arterioscler Thromb Vasc Biol 2003 23 1035 1041 12702517 10.1161/01.ATV.0000072273.67342.6D
Bonora E Kiechl S Oberhollenzer F Egger G Bonadonna RC Muggeo M Willeit J Impaired glucose tolerance, Type II diabetes mellitus and carotid atherosclerosis: prospective results from the Bruneck Study Diabetologia 2000 43 156 164 10753036 10.1007/s001250050024
Harris SB Zinman B Hanley A Gittelsohn J Hegele R Connelly PW Shah B Hux JE The impact of diabetes on cardiovascular risk factors and outcomes in a native Canadian population Diabetes Res Clin Pract 2002 55 165 173 11796183 10.1016/S0168-8227(01)00316-3
Ainsworth CD Blake CC Tamayo A Fenster A Spence JD Measurement of change in carotid plaque volume: A 3-dimensional ultrasound tool for rapid evaluation of new therapies Stroke 2005
Fujii K Abe I Ohya Y Ohta Y Arima H Akasaki T Yoshinari M Iida M Risk factors for the progression of early carotid atherosclerosis in a male working population Hypertens Res 2003 26 465 471 12862203 10.1291/hypres.26.465
Fenster A Landry A Downey DB Hegele RA Spence JD 3D ultrasound imaging of the carotid arteries Curr Drug Targets Cardiovasc Haematol Disord 2004 4 161 175 15180488 10.2174/1568006043336311
Landry A Spence JD Fenster A Measurement of carotid plaque volume by 3-dimensional ultrasound Stroke 2004 35 864 869 15017019 10.1161/01.STR.0000121161.61324.ab
|
15958169
|
PMC1184090
|
CC BY
|
2021-01-04 16:38:31
|
no
|
Cardiovasc Ultrasound. 2005 Jun 15; 3:15
|
utf-8
|
Cardiovasc Ultrasound
| 2,005 |
10.1186/1476-7120-3-15
|
oa_comm
|
==== Front
Cardiovasc UltrasoundCardiovascular Ultrasound1476-7120BioMed Central London 1476-7120-3-161595817310.1186/1476-7120-3-16ResearchQuantification of resting myocardial blood flow velocity in normal humans using real-time contrast echocardiography. A feasibility study Malm Siri [email protected] Sigmund [email protected] Frode [email protected] Kjetil [email protected] Stig [email protected] Terje [email protected] Department of Circulation and Medical Imaging, Faculty of Medicine, Norwegian University of Science and Technology, Trondheim, Norway2 GE Healthcare Technologies, Ultrasound R&D, Trondheim, Norway2005 16 6 2005 3 16 16 30 4 2005 16 6 2005 Copyright © 2005 Malm et al; licensee BioMed Central Ltd.2005Malm et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Real-time myocardial contrast echocardiography (MCE) is a novel method for assessing myocardial perfusion. The aim of this study was to evaluate the feasibility of a very low-power real-time MCE for quantification of regional resting myocardial blood flow (MBF) velocity in normal human myocardium.
Methods
Twenty study subjects with normal left ventricular (LV) wall motion and normal coronary arteries, underwent low-power real-time MCE based on color-coded pulse inversion Doppler. Standard apical LV views were acquired during constant IV. infusion of SonoVue®. Following transient microbubble destruction, the contrast replenishment rate (β), reflecting MBF velocity, was derived by plotting signal intensity vs. time and fitting data to the exponential function; y (t) =A (1-e-β(t-t0)) + C.
Results
Quantification was feasible in 82%, 49% and 63% of four-chamber, two-chamber and apical long-axis view segments, respectively. The LAD (left anterior descending artery) and RCA (right coronary artery) territories could potentially be evaluated in most, but contrast detection in the LCx (left circumflex artery) bed was poor. Depending on localisation and which frames to be analysed, mean values of were 0.21–0.69 s-1, with higher values in medial than lateral, and in basal compared to apical regions of scan plane (p = 0.03 and p < 0.01). Higher β-values were obtained from end-diastole than end-systole (p < 0.001), values from all-frames analysis lying between.
Conclusion
Low-power real-time MCE did have the potential to give contrast enhancement for quantification of resting regional MBF velocity. However, the technique is difficult and subjected to several limitations. Significant variability in β suggests that this parameter is best suited for with-in patient changes, comparing values of stress studies to baseline.
Contrast echocardiographymyocardial perfusionreal-time imagingcontrast replenishment rate
==== Body
Background
MCE is an emerging technique for non-invasive evaluation of myocardial perfusion and coronary heart disease (CAD) [1-11]. Recent advances in multipulse technology have made real-time MCE feasible with low acoustic power [12-17], giving minimal contrast destruction and frame rates that facilitate evaluation of scan plane and wall motion. However, technical difficulties concerning tailored ultrasound equipment, imaging techniques, data-analysis and interpretation still remain to be solved.
The majority of MCE studies have reported data relying on visual assessment somewhat limited by its subjective approach [18]. Wei and coworkers pioneered a method for more objective quantification of MBF with contrast microbubbles administered as constant intravenous infusion [5]. From the time course of video intensity during progressively prolonged pulsing intervals, both MBF velocity and myocardial blood volume (MBV) could be assessed. The product of these two parameters was shown to correlate well with radiolabeled microsphere-derived MBF [5,17,19,20]. This quantitative approach has also been applied to real-time MCE techniques [14-16,19]. A strong linear correlation between the rate of signal intensity (SI) rise and volumetric flow has been reported, both at rest and during hyperemia [14,16,22]. On the other hand, steady state SI has not been found to correlate as well with flow measurements [14,17], indicating that the microbubble replenishment rate might be the major MCE perfusion parameter.
The quantification of replenishment rates is often limited to selected myocardial regions due to imaging problems [19,22]. To our knowledge there are limited human studies reporting resting replenishment rates for all standard myocardial segments measured in different cardiac phases. The aim of this study was 1) to evaluate the feasibility of a very low-power real-time MCE technique for visualising the perfusion in normal human myocardium, and 2) to quantify the MBF velocity, β, of all myocardial segments of the apical scan views by using the destruction-replenishment approach.
Methods
Study subjects
Twenty study subjects were enrolled; ten healthy male volunteers (age 24 ± 3) and ten patients (age 55 ± 5), five of them female. The study subjects were not screened for echocardiographic image quality, the only inclusion criteria being an age above 18 and confirmed normal left ventricular regional and global systolic function by conventional echocardiography. The ten patients had undergone coronary angiography due to chest pain, with the findings of open and normal coronary arteries and normal left ventricular end-diastolic pressures. The healthy volunteers were assumed to have normal coronary anatomy and myocardial perfusion, due to the abscence of CAD risk factors and symptoms, and normal findings on standard echocardiography. All the subjects were in sinus rhythm. Exclusion criteria were pregnancy or lactation, known allergy to the contrast agent, significant valve diseases or shunts, severe pulmonary hypertension, and severe extra-cardiac disease. All the subjects gave their written informed consent to the participation. The study conformed to the declaration of Helsinki, and the Regional Commitee of Medical Ethics approved the protocol.
Contrast agent
The ultrasound contrast agent SonoVue® (Bracco, Milan, Italy) was used, consisting of microspheres of sulphur hexafluoride gas (SF6) stabilised by a phospholipid monolayer in an aqueous solution. SonoVue® was infused continuously by a manually rotated volume pump through a 20G vial in a proximal forearm vein. There were slight individual changes of the infusion rate (70–100 ml/ hour) to optimize the myocardial opacification and minimize the far-field attenuation. Once steady state was reached and the recording started, the infusion rate was held constant in every individual study.
MCE technique
Imaging was performed with Vivid 7™ (GE Vingmed Ultrasound, Horten, Norway) with a M3S matrix array transducer. The contrast-specific application, Coded Harmonic Angio™, is a very low power, real-time technique based on pulse inversion combined with power Doppler, operating at a frame rate of 20 Hz. With this choice of application and contrast agent, the optimal agent-tissue-ratio was achieved with a mechanical index (MI) as low as 0.04–0.05. The signal amplitudes were color-coded by the Angio mode and displayed as overlays on fundamental tissue grey-scale images. The focus was set basally, close to the mitral valve plane. The depth was set to let the left ventricle fill the image sector, and color gain was adjusted to reduce the signal-to-noise ratio to the point that hardly any noise was observed within tissue and cavity. The time gain compensation was adjusted to obtain homogenous SI and to reduce the noise from the myocardium, the epi-/ pericardium and the mitral valves. After the initial adjustments all settings were held constant in every individual study.
Baseline imaging was acquired in tissue harmonic mode for confirmation of normal anatomy and wall motion. MCE was performed in the apical four-chamber, two-chamber and long-axis views. Standard views were at times slightly modified, i.e. by centralising the lateral or anterior walls in the scan sector, to optimize the contrast detection and avoid attenuation and shadowing. When the myocardial contrast opacification reached a steady state, a 'flash' of 15 frames of high MI (1.0), timed to cover at least the entire systole, was applied for transient microbubble destruction. This was followed by immediate, automatic return to low MI continuous imaging of microbubble replenishment in end-expiration (See Additional file: 1 Movie demonstrating a real-time destruction-replenishment loop of the LV apical long-axis view). The procedure was repeated twice for every scan view. Fifteen cardiac cycles of every destruction-replenishment sequence, at least 10 after 'flash', were captured and stored digitally as raw-data.
Image analysis
The MCE data were analysed off-line on a PC workstation. Analyses of the cineloops were performed blinded in random order using EchoPAC PC™ (GE Vingmed Ultrasound, Horten, Norway). Measurement of mean signal intensity (dB) was done in manually placed, equally sized and shaped regions of interest (ROI) in the 16 standard myocardial segments [23], plus the two apical segments of the apical long-axis view. The ROIs were large, avoiding high intensity signals from the cavity and the epi-and endocardium. When necessary, their position was slightly adjusted to compensate for the translation of the heart. The depth of the ROI position was not changed. Finally, all ROIs were 'anchored' for each frame.
The myocardial SI was plotted against time (t) and fitted to the exponential function: y (t) =A (1-e-β(t-t0)) + C, where y is SI at any time during the contrast replenishment, A is the plateau SI corresponding to MBV, β is the rate of SI rise reflecting the mean bubble velocity or MBF velocity, and C is the intercept at the origin reflecting the background intensity level [5]. The introduction of t0 simply reflects that the analysis software allowed one to choose where to set t = 0. To further compensate for a possible non-zero initial value after flash, the constant C was added, implicating that the curve fitting was relatively independent of background myocardial SI. The ROIs were positioned and anchored before the curve fitting was applied. Segmental values of A and β were derived from the replenishment cycles by careful frame-by-frame analysis. Separate quantitative analysis was performed both for all-frames, for selected end-systolic (end of T-waves) and end-diastolic (close to peak R-wave) frames.
The myocardial segments were assigned to the coronary artery perfusion territories (Figure 1), and the feasibility for evaluating perfusion at a territorial level was assessed. Because the LV wall motion was normal, any lack of myocardial contrast opacification was considered to be due to attenuation or inadequate detection, and the current segment was excluded from the quantitative analysis. Since the healthy volunteers all had normal regional and global LV function, it seemed acceptable to make this assumption even if coronary angiography was not performed.
Figure 1 The different coronary artery beds and their representation in myocardial segments of the LV apical views, given a balanced coronary circulation. LAD = left anterior descending artery; LCx = left circumflex artery; LV = left ventricle; RCA = right coronary artery. Courtesy of Asbjorn Stoylen, dept. of Circulation and Medical Imaging, Norwegian University of Science and Technology, Trondheim, Norway.
Statistics
Continuous variables are presented as mean ± 1SD. Comparison between groups was performed with linear regression analysis (ANOVA), a posthoc analysis was done using Bonferroni's correction. Differences were considered statistically significant at p less than 0.05 (two-sided) with a power of 0.80.
Results
Some visible myocardial contrast enhancement was obtained in all views of all the study subjects. Precontrast myocardial tissue SI was negligible, but in spite of careful adjustment of the time-gain compensation we noticed relatively strong precontrast signals from the mitral valve and basal epi-and pericardium. On average 6 minutes of infusion time was spent to aquire repeated replenishment cineloops in the apical views. Myocardial contrast first appeared around one minute after the infusion was started, and steady-state SI was reached after a mean period of 2.5 minutes. After the 'flash' an almost complete disappearance of myocardial color signals was observed, leaving the myocardium dark. Real-time visual grading of myocardial SI during post-destruction wash-in was difficult due to cardiac contraction, translation and cyclic changes of myocardial SI, both between systole and diastole and from beat to beat. By reviewing selected end-systolic frames, refilling was first observed in the mid-septum progressing to full opacification in 3 to 5 heartbeats. However, when observed in end-diastole the refilling clearly appeared faster, yet variable. Selected end-systolic images of destruction-refill sequences of the apical LV views are presented in Figure 2.
Figure 2 Some selected end-systolic images from destruction-replenishment sequences. A. 4-chamber view, B. 2-chamber view, C. Apical long-axis view.
Feasibility of quantitative analysis
All 360 myocardial segments were evaluated regardless of baseline image quality. Since the LV wall motion was normal, any contrast defect was considered to be due to attenuation or inadequate detection, and the current segment was thus excluded from the quantitative analysis. Following this, the myocardial opacification was regarded as sufficient for quantification in 98 of 120 (82%) of the four-chamber view segments (Table 1). The septum filled in completely during wash-in, while enhancement of the lateral wall was more pathcy. In two-chamber and long-axis views, the number of feasible segments was lower; 59 of 120 (49%) and 76 of 120 (63%), respectively (Table 1) Thus, a total of 233 of 360 (65%) of myocardial segments were feasible for quantification. For healthy, young normals and patients the feasibility was 118 (66%) and 115 (64%) of segments, respectively. The most frequent dropouts were observed in the mid and basal segments of the lateral and anterior wall, and in the basal segment of the inferolateral wall. In these we only obtained myocardial opacification in half of the study subjects.
Table 1 Absolute values of segmental myocardial contrast replenishment rate, (s-1)
Scan view Myocardial segment N End-systolic analysis End-diastolic analysis All frames analysis
Apical Four-Chamber Basal septum 20 0.41 ± 0.11 0.63 ± 0.14 0.56 ± 0.17
Mid septum 20 0.36 ± 0.12 0.46 ± 0.15 0.47 ± 0.16
Apical septum 20 0.33 ± 0.14 0.43 ± 0.14 0.42 ± 0.18
Apical lateral 18 0.29 ± 0.10 0.43 ± 0.18 0.30 ± 0.12
Mid lateral 11 0.27 ± 0.11 0.44 ± 0.18 0.37 ± 0.12
Basal lateral 9 0.32 ± 0.13 0.55 ± 0.12 0.37 ± 0.14
Apical Two-chamber Basal inferior 14 0.43 ± 0.17 0.59 ± 0.19 0.44 ± 0.12
Mid inferior 15 0.36 ± 0.09 0.54 ± 0.11 0.46 ± 0.15
Apical inferior 13 0.35 ± 0.10 0.43 ± 0.14 0.34 ± 0.15
Apical anterior 6 0.34 ± 0.10 0.45 ± 0.15 0.60 ± 0.29
Mid anterior 5 0.27 ± 0.18 0.31 ± 0.18 0.21 ± 0.15
Basal anterior 6 0.24 ± 0.21 0.39 ± 0.11 0.35 ± 0.22
Apical Long-axis Basal inferolat. 7 0.42 ± 0.14 0.69 ± 0.12 0.61 ± 0.17
Mid inferolat. 11 0.36 ± 0.15 0.52 ± 0.16 0.40 ± 0.18
Apical inferolat. 13 0.31 ± 0.13 0.42 ± 0.18 0.44 ± 0.16
Apical anterosept. 18 0.31 ± 0.11 0.48 ± 0.23 0.33 ± 0.15
Mid anterosept. 15 0.36 ± 0.15 0.41 ± 0.19 0.47 ± 0.11
Basal anterosept. 12 0.45 ± 0.19 0.58 ± 0.20 0.48 ± 0.16
Values are mean ± 1SD. N = number of study subjects in which myocardial contrast opacification of the current segment was regarded as feasible for quantification.
Evaluated on a territorial level, perfusion in the LAD area could be assessed in all 20 subjects. Segments usually assigned to the RCA could be evaluated in 15 subjects, whereas the LCx supply area was analysable in only half of the subjects.
MCE parameters
Beta-values were derived by curve fitting in the 233 feasible segments (Figure 3). Depending on localisation and which frames of the heart cycle to be analysed, we found mean values of β ranging from 0.21 to 0.69 s-1 with SDs of 0.09 to 0.29 s-1 (Table 1). Segmental mean values of A ranged from 6.01 to 12.29 dB with SDs of 2.1 to 4.9 dB. Mean end-systolic β-values was found to be higher in medial than lateral parts of the scan plane (0.37 ± 0.13 vs. 0.32 ± 0.14 s-1, p = 0.03), and at greater depths (basal; 0.45 ± 0.16 s-1 vs. apical; 0.36 ± 0.14 s-1, p < 0.01). The A parameter similarly was found to be higher in medial (9.89 ± 2.7 dB) than lateral regions (7.99 ± 3.6 dB, p < 0.01), while it was significantly lower in basal than apical segments (7.87 ± 3.0 vs. 9.13 ± 2.74, p < 0.01).
Figure 3 ECG-triggered end-systolic analysis (EchoPAC PC™) of contrast replenishment in the LV apical long-axis view. Time-intensity plots are fitted to the exponential function A (1-e-β(t-t0)) + C. Typical attenuation artefact is seen in basal inferolateral wall. A = peak plateau signal intensity reflecting myocardial blood volume; B(C) = intercept at the origin; k (β) = rate of signal intensity rise (microbubble replenishment rate) reflecting myocardial blood flow velocity; LV = left ventricle.
By using the software's capability to perform an off-line ECG triggering, we did separate analysis from end-systolic and end-diastolic images (Figure 3). Significantly higher β-values were obtained when end-diastolic frames were analysed compared to end-systolic ones (0.49 ± 0.16 s-1 and 0.35 ± 0.13 s-1 respectively, p < 0.001). The β-values from all-frames analysis were lying between (0.43 ± 0.17 s-1), significantly different from both end-systolic and end-diastolic values (p = 0.002 and p = 0.001, respectively). Approximately the same level of differences was found between cardiac phases for A-values.
There were no significant differences in mean end-systolic β between four-chamber, two-chamber and long-axis views (0.34 ± 0.12 vs 0.35 ± 0.15 vs. 0.36 ± 0.14 s-1, respectively) nor between healthy volunteers and patients (0.36 ± 0.14 vs. 0.35 ± 0.13 s-1).
Hemodynamic and safety parameters
There were no significant changes in the study subjects' blood pressure, heart rate or rythm during performance of the MCE examinations. Each subject received a total dose of 9.5 ml of SonoVue®, and none of them experienced any adverse effect in the observation period, nor were any observed.
Discussion
Despite recent advances in contrast-specific imaging, our study demonstrates some of the difficulties with and the still limited ability of low-power real-time MCE for quantitative assessment of regional myocardial perfusion. The contrast detection was better in the segments with good precontrast myocardial image quality, the lateral and anterior walls being the poorest with frequent dropouts. The MCE imaging problems were reduced, but not eliminated by adjustment of infusion rates and by carefully repositioning the wall of interest more centrally in the scan sector. Nevertheless, two thirds of segments were feasible for quantitative analysis, and by assigning segments to the coronary territories, the LAD and RCA areas could potentially be evaluated. On the other hand, limited contrast detection made assessment of the LCx area difficult in more than half of the subjects.
Our study was not designed to test specific machine settings or examination variables. Nevertheless, we observed decreasing A values moving distally and laterally in the scan sector. β was similarly found to be lower in lateral than more axial parts, but in opposite to A it significantly increased at greater depths of the sector. This is in accordance with results from in vitro flow phantom and animal models [14,16], but has previously not been reported in human studies.
The exact mechanism for this spatial variability is uncertain. Variations in beam elevation and non-uniformity of the sound field are probably contributing factors [6,24,25]. With a phased array transducer, the acoustic energy delivered is lower laterally than in the centre of the sector, due to smaller effective aperture and some directivity of the elements [26]. Consequently, the backscatter from the microbubbles is inhomogenous. In addition attenuation and other artefacts decreasing backscatter are often more pronounced in the lateral regions. Non-uniformity of the effective regional 'flash' energy level might lead to variable bubble destruction, possibly altering the refilling parameters as well.
Far field attentuation likely contributed to the lowering of A in basal parts of sector. The finding of higher β in basal compared to apical segments could be due to the narrower beam elevation in the focal zone (which was set close to the mitral valve plane), resulting in a thinner scan plane with myocardial contrast destruction, and thereby faster replenishment of contrast from adjacent areas not affected by destruction.
The contrast refilling measurements were also influenced by the selection of frames from the cardiac cycle. We assessed significantly higher β-values from end-diastolic compared to end-systolic measurements. Corresponding values from all-frames analysis were lying in-between. Even the first diastolic frames after the 'flash' displayed some contrast signals, while the earliest systolic frame did not. This might be due to early bubble refilling through larger intramyocardial arterioles that are patent in diastole, but collapsed during systolic contraction [27,28]. By applying ROIs of uniform shape and size, the risk of capturing strong signals from the pericardium and the cavity seems to be greater in diastole, when the myocardium is thinner and localized more laterally in sector. Assessment of end-systolic frames seems preferable, despite more coronary flow in diastole, because the myocardium is thicker with less risk of 'contamination' from contrast in the blood pool.
These findings are in accordance with results from experimental studies [14,16]. Leong-Poi et al also found that MBF derived from images obtained in end-systole accurately reflected radiolabeled microsphere-derived MBF, whereas the correlation was poor, with significant overestimation of MBF, when the end-diastolic ones were used [16].
Insufficient contrast detection with too low agent-to-tissue signal ratio still remains a basic problem in quantifying the perfusion by pulse inversion Doppler. One disadvantage of the low-power real-time mode compared to destructive intermittent imaging, is the weaker backscatter from the non-destructive microbubble behaviour. A larger amount of contrast agent is required, which must be balanced against the degree of far field attenuation. Furthemore, some tissue harmonics are always present, and the movement artefacts will probably not be completely removed by the addition of power Doppler.
Relatively low temporal resolution was another anticipated problem. However, in our experience the applied frame rate of 20 Hz gave sufficient image quality to allow anatomical orientation and wall motion assessment.
An advantage of real-time MCE is the very short data aquisition time. Information that would require approximately a minute with intermittent imaging was recorded during seconds, making it possible to obtain all refill frames during one breathhold. The mean contrast infusion time of 6 minutes allowed us to aquire several replenishment loops from each scan view in every subject. But despite short data acquisition time, the procedure is still demanding. Given a beam elevation of only 3–4 mm at focus, it is mandatory to maintain a stable probe position during the entire destruction-replenishment period. Of the same reson, it is important to put an effort in avoiding off-axis views, particularly whenever tilting the scanplane to centralize a wall for improving contrast detection. Out-of plane imaging would possibly contribute to increased variability of the segmental quantitative parameters. Furtermore, the patients must be cooperative and able to hold their breath for up to ten seconds, preferably in end-expiration, to avoid lung interference and cardiac translation. Even some of our young, healthy study subjects experienced difficulties in this regard.
Study limitations
The sample size of this study was small, and the patients were selected from baseline image quality. The coronary anatomy was not known in all study subjects, and we did not use other methods to ensure normal perfusion in our patients. However, due to the selection criteria it is highly unlikely that subclinical CAD could explain our results.
The main limitation of this study is that we did not apply any manipulations to change blood flow, i.e. adding of vasodilator to obtain maximal hyperaemia. Regional resting perfusion is normally variable, both between segments and study subjects, and this variability is further reinforced by imaging and technical problems with the applied destruction-replenishment approach. In the presence of a non-critical coronary stenosis, normal resting blood flow is maintained by arteriolar vasodilation. And patients with critical lesions often have collateral circulation, making assessment of resting perfusion complex. Hence, the interpretation of replenishment kinetics is particularly difficult at rest. Since normal perfusion is expected to increase at least 3-fold during hyperaemia, giving faster contrast replenishment after transient destruction, the evaluation of stress-to-rest ratios (the MBF velocity reserve) would be a more appropriate quantitative approach to assess normal or abnormal perfusion.
Interestingly, a simplified algorithm using qualitative evaluation of MBF velocity (actually time to complete myocardial opacification after destruction) from a single stress MCE perfusion study, was recently shown to detect CAD in patients with normal left ventricular function at rest, avoiding the need for resting MCE studies [11]. However, in that study triggered imaging was used, and it still seems to be a challenge maintaining a stable probe position for good image alignment during stress studies.
Conclusion
Our study indicated that a very low-power real-time MCE could provide contrast opacification in multiple myocardial segments of the LV apical views. However, the acquistion of flash-replenishment loops adequate for quantification is limited by imaging and technical problems. The interpretation of regional resting replenishment curves is in addition complicated by great variability in normal perfusion. Our data support previous findings in experimental studies, which established that the absolute values of MBF velocity are highly influenced by ultrasound field geometry and which cardiac phase to be analyzed. The MBF velocity need to be interpreted with regard to the imaging technique used and the segments from which they are obtained, and due to the great variability is is probably best applied by analysis of relative changes, that is comparing values during stress with baseline.
competing interests
The author(s) declare that they have no competing interests.
Supplementary Material
Additional file 1
Movie demonstrating a real-time destruction-replenishment loop of the LV apical long-axis view
Click here for file
Acknowledgements
Dr. Siri Malm was supported by a Research Fellowship grant from the Norwegian Council for Cardiovascular Diseases. The autors are grateful to Bracco International providing the contrast agent used in this study. We also thank GE Vingmed Ultrasound, Norway, for providing the research ultrasound machine and software. No other financial payments have been made to the project leader or associates, and no obligations to or other connection with the sponsors exists.
==== Refs
Skyba DM Jayaweera AR Goodman NC Ismail S Camarano G Kaul S Quantification of myocardial perfusion with myocardial contrast echocardiography during left atrial injection of contrast. Implications for venous injection Circulation 1994 90 1513 1521 8087957
Porter TR Li S Kiltzer K Deligonul U Correlation between quantitative angiographic lesion severity and myocardial contrast intensity during a continuous infusion of perfluorocarbon-containing microbubbles J Am Soc Echocardiogr 1998 11 702 710 9692527
Kaul S Senior R Dittrich H Raval U Khattar F Lahiri A Detection of coronary artery disease with myocardial contrast echocardiography: Comparison with 99mTc-Sestamibi single-photon emission computed tomography Circulation 1997 96 785 792 9264483
Marwick TH Brunken R Meland N Brochet E Baer FM Binder T Flachskampf F Kamp O Nienaber C Nihoyannopoulos P Pierard L Vanoverschelde JL van der Wouw P Lindwall K the Nycomed NC100100 Investigators Accuracy and feasibility of contrast echocardiography for detection of perfusion defects in routine practice: comparison with wall motion and technetium-99m sestamibi single-photon emission computed tomography J Am Coll Cardiol 1998 32 1260 1269 9809934 10.1016/S0735-1097(98)00373-8
Wei K Jayaweera AR Firoozan S Linka A Skyba DM Kaul S Quantification of myocardial blood flow with ultrasound-induced destruction of microbubbles administered as a constant venous infusion Circulation 1998 97 473 483 9490243
Senior R Kaul S Soman P Lahiri A Power Doppler contrast echocardiography: a new technique for assessing myocardial perfusion Am Heart J 2000 139 245 251 10650297
Villanueva FS Gertz EW Csikari M Pulido G Fisher D Sklenar J Detection of coronary artery stenosis with power Doppler imaging Circulation 2001 103 2624 2630 11382734
Leistad E Ohmori K Peterson A Christensen G DeMaria AN Quantitative assessment of myocardial perfusion during graded coronary artery stenoses by intravenous myocardial contrast echocardiography J Am Coll Cardiol 2001 37 624 31 11216989 10.1016/S0735-1097(00)01127-X
Wei K Detection and quantification of coronary stenosis severity with myocardial contrast echocardiography Prog Cardiovasc Dis 2001 44 81 100 11568821 10.1053/pcad.2001.26444
Heinle S Noblin J Goree-Best P Mello A Ravad G Mull S Mammen P Graybrun PA Assessment of myocardial perfusion by Harmonic Power Doppler Imaging at rest and during adenosine stress. Comparison with 99mTc-Sestamibi SPECT Imaging Circulation 2000 102 55 60 10880415
Wei K Crouse L Weiss J Villanueva F Schiller NB Naqvi T Siegel R Monaghan M Goldman J Aggarwal P Feigenbaum H DeMaria A Comparison of usefulness of dipyridamol stress myocardial contrast echocardiography to Technetium-99m sestamibi single-photon emission computed tomography for detection of coronary artery disease (PB127 Multicenter Phase 2 Trial Results) Am J Cardiol 2003 91 1293 1298 12767419 10.1016/S0002-9149(03)00316-3
Porter TR Xie F Silver M Kricsfeld D Oleary E Real-time perfusion imaging with low mechanical index pulse inversion Doppler imaging J Am Coll Cardiol 2001 37 748 753 11693747 10.1016/S0735-1097(00)01204-3
Tiemann K Lohmeier S Kuntz S Köster J Pohl C Burns P Real-time contrast echo assessment of myocardial perfusion at low emission power: First experimental and clinical results using power pulse inversion imaging Echocardiography 1999 16 799 809 11175224
Lafitte S Masugata H Peters B Togni M Strachan M Yao B Kwan OL DeMaria AN Accuracy and reproducibility of coronary flow rate assessment by real-rime contrast echocardiography: In vitro and in vivo studies J Am Soc Echocardiogr 2001 14 1010 1019 11593206 10.1067/mje.2001.112908
Masugata H Peters B Lafitte S Strachan GM Ohmori K DeMaria AN Quantitative assessment of myocardial perfusion during graded coronary stenosis by real-time myocardial contrast echo refilling curves J Am Coll Cardiol 2001 37 262 269 11153750 10.1016/S0735-1097(00)01046-9
Leong-Poi H Le E Rim SJ Sakuma T Kaul S Wei K Quantification of myocardial perfusion and determination of coronary stenosis severity during hyperemia using real-time myocardial contrast echocardiography J Am Soc Echocardiogr 2001 14 1173 1182 11734784 10.1067/mje.2001.115982
Masugata H Peters B Lafitte S Strachan GM Kohno M DeMaria Comparison of microbubble agents that produce different myocardial signal intensity for quantification of myocardial blood flow by myocardial contrast echo Am J Cardiol 2001 88 714 718 11564409 10.1016/S0002-9149(01)01828-8
Von Bibra H Bone D Niklasson U Eurenius L Hansen A Myocardial contrast echocardiography yields best accuracy using quantitative analysis of digital data from pulse inversion technique: Comparison with second harmonic imaging and harmonic power Doppler during simultaneous Dipyridamol stress SPECT studies Eur J Echocardiography 2002 3 271 282 10.1053/euje.2002.0166
Murthy TH Li P Locvicchio E Baisch C Dairywala I Armstrong WF Vannan M Real-time myocardial blood flow imaging in normal human beings with the use of myocardial contrast echocardiography J Am Soc Echocardiogr 2001 14 698 705 11447415 10.1067/mje.2001.111156
Köster J Schlosser T Pohl C Lentz C Lohmeier S Veltman C Blood flow assessment by ultrasound-induced destruction of echocontrast agents using harmonic power Doppler imaging: which parameters determine contrast replenishment curves? Echocardiography 2001 18 1 8 11182774 10.1046/j.1540-8175.2001.00001.x
Masugata H Lafitte S Peters B Strachan GM DeMaria AN Comparison of real-time and intermittent triggered myocardial contrast echocardiography for quantification of coronary stenosis severity and transmural perfusion gradient Circulation 2001 104 1550 1556 11571251
Wei K Ragosta M Thorpe J Coggins M Moos S Kaul S Noninvasive quantification of coronary blood flow reserve in humans using myocardial contrast echocardiography Circulation 2001 103 2560 2565 11382724
Schiller NB Shah PM Crawford M DeMaria A Devereux R Feigenbaum H Gutgesell H Reichek N Sahn D Schnittger I Silvermann NH Tajik AJ Recommendations for quantification of the left ventricle by two-dimensional echocardiography. American Society of Echocardiography committee on standards, subcommittee on quantitation of two-dimensional echocardiograms J Am Soc Echocardiogr 1989 2 358 367 2698218
Weyman AE Principles and practice of echocardiography Chap 1 Philadelphia: Lea & Febiger 3 27
Porter TR Xie F Li S Kricsfeld D Deligonul U Effect of transducer stand-off on the detection, spatial extent, and quantification of myocardial contrast defects caused by coronary stenosis J Am Soc Echocardiogr 1999 12 951 956 10552356
Angelsen BAJ Principles of medical ultrasound imaging and measurements Ultrasound Imaging 2000 II 1 Trondheim: Emantec AS 1.3 1.99
Krams R Sipkema P Westerhof N Coronary oscillatory flow amplitude is more affected by perfusion pressure than ventricular pressure Am J Physiol 1990 258 H1889 1898 2193545
Hiramatsu O Kimura A Yada T Yamamoto T Ogasawara Y Goto M Tsujioka K Kajiya F Phasic characteristics of arterial inflow and venous outflow of right ventricular myocardium in dogs Am J Physiol 1992 262 H1422 1427 1590447
|
15958173
|
PMC1184091
|
CC BY
|
2021-01-04 16:38:31
|
no
|
Cardiovasc Ultrasound. 2005 Jun 16; 3:16
|
utf-8
|
Cardiovasc Ultrasound
| 2,005 |
10.1186/1476-7120-3-16
|
oa_comm
|
==== Front
CytojournalCytoJournal1742-6413BioMed Central London 1742-6413-2-121598750210.1186/1742-6413-2-12ReviewFine-needle aspiration of the thyroid: an overview Nguyen Gia-Khanh [email protected] Mark W [email protected] Jody [email protected] Tina [email protected] Darcy [email protected] Department of Laboratory Medicine and Pathology, University of Alberta Hospital, Edmonton, Alberta, Canada2 Department of Medicine (Endocrinology and Metabolism), University of Alberta Hospital, Edmonton, Alberta, Canada2005 29 6 2005 2 12 12 19 5 2005 29 6 2005 Copyright © 2005 Nguyen et al; licensee BioMed Central Ltd.2005Nguyen et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Thyroid nodules (TN) are a common clinical problem. Fine needle aspiration (FNA) of the thyroid now is practiced worldwide and proves to be the most economical and reliable diagnostic procedure to identify TNs that need surgical excision and TNs that can be managed conservatively. The key for the success of thyroid FNA consists of an adequate or representative cell sample and the expertise in thyroid cytology. The FNA cytologic manifestations of TNs may be classified into seven working cytodiagnostic groups consisting of a few heterogenous lesions each to facilitate the differential diagnosis. Recent application of diagnostic molecular techniques to aspirated thyroid cells proved to be useful in separating benign from malignant TNs in several cases of indeterminate lesions.
==== Body
Fine-needle aspiration (FNA) for cytologic evaluation of thyroid cancer was originally used by Martin and Ellis at New York Memorial Hospital for Cancer and Allied Diseases in 1930 [1]. However, this diagnostic procedure was subsequently found to have a limited value, and it was then discontinued at the above-mentioned institution [2,3]. The thyroid FNA was not further developed and did not gain acceptance in the United States for nearly 50 years until the early 1980s when its diagnostic value was firmly demonstrated by Scandinavian investigators [4-8]. The 1974 report by Crockford and Bain [9] and the 1979 paper of Miller and Hamburger [10] were apparently the first North American publications attesting to the value of thyroid FNA. This method of clinical investigation now is practiced worldwide and has become the cornerstone in the management of thyroid nodules (TN) [11-25].
Indication and Goal of Thyroid FNA
Thyroid nodular lesions are a common clinical problem. In the United States, 4 to 7% of the adult population have a palpable TN [13]. The incidence of thyroid cancer in a clinically solitary TN or in a multinodular goiter is equal and about 5% in non-endemic areas [26]. TNs constitute the main indication for FNA, and the goal of this diagnostic procedure is to detect thyroid neoplasms for surgical resection and to identify non-neoplastic lesions that may be managed conservatively [23]. This method of clinical investigation has reduced the number of diagnostic thyroid surgeries for TNs by 60–85%, and the difference in rates of thyroid surgery reflect the cytodiagnostic accuracy rates among different medical centers [24].
Contraindications and Complications of Thyroid FNA
The main contraindication to thyroid FNA is bleeding diathesis, as the formation of a large hematoma at the biopsy site may cause compression of the trachea and respiratory distress [13,23]. Therefore, a bleeding time, PT and PTT should be ordered to screen this condition in all patients prior to thyroid FNA. This diagnostic procedure, if properly performed, is almost free of complications. Subcutaneous hematoma at the biopsy site, accidental puncture of the trachea and local infection are rare complications [13]. Hematoma may be prevented by local pressure of the overlying skin at the biopsy site [13]. Tracheal injury is manifested by minimal and transient hemoptysis. Seeding of thyroid cancer cells along the needle tract is also an exceedingly rare complication with FNA [13].
Procuration and Preparation of Cell Samples
1. Procurement of cell samples
Obtaining an adequate or satisfactory cell sample for cytologic evaluation is not simple, and interpreting thyroid cytology is challenging and requires expertise [13,23]. To perform thyroid FNA, the TN is identified by palpation, and a 22- to 25-gauge and 4.5-cm-long needle is commonly used to procure cell samples from at least three different areas of any TN. Usually, only dermal anesthesia is required. Depending on personal preferences FNA of a TN may be performed either with or without a syringe [13]. However, for cystic thyroid lesions, the cyst contents should be evacuated first by FNA with a syringe. The gland is then carefully examined by palpation. If a residual nodule is found, it should be aspirated. If the TN is difficult to identify by palpation the patient should be referred to a radiologist for FNA under ultrasonographic guidance [13,22-24]. Since the thyroid is rich in capillary blood vessels the needle aspirate usually contains a large amount of peripheral blood that may be reduced by limiting the biopsy procedure to about five seconds or by using the FNA technique without aspiration [13].
2. Preparation of cell samples
For cytological evaluation, smears should be appropriately prepared and stained. Depending on the amount and nature of the thyroid needle aspirates one of the following preparation techniques is used: (a). A small drop of thyroid aspirate is put near the frosted end of a glass slide and is quickly and gently smeared by a cover slip. (b). A small drop of thyroid aspirate is put on a glass slide and gently crushed with a second slide that is then separated vertically from the first one. (c). A small or medium-sized drop of thyroid aspirate is put near the frosted end of a slide that is placed on a table. A second slide is used to spread the aspirated material in the same manner used to prepare a peripheral blood smear. (d) Cytospin smears should be prepared from the liquid contents of all cystic thyroid lesions. (e). Excess of aspirated material should be used for preparation of a cell block that may show diagnostic tissue fragments on sectioning. It is important that a small drop of aspirated material is used for smear preparation, as if a large drop of aspirate material is used, an unevenly thick smear may be obtained, and at the end of the slide a thick and bloody cell film may be formed. This will obscure the cellular details of underlying thyroid cells and tissue fragments, making their evaluation extremely difficult, if not impossible.
3. Routine staining methods
Depending on personal preference, either air-dried and Romanowsky-stained smears or ethanol-fixed and Papanicolaou-stained smears are prepared. For Papanicolaou staining, the smears must be fixed quickly before drying with 95% ethanol or with a commercial spray fixative. A delay in fixation will result in air-dried artefactual changes with loss of cellular details. Air-dried smears for staining with one of the Romanowsky modified methods (Wright stain, May-Grunwald-Giemsa or Diff-Quik method) now are widely used, as air-drying artifactual changes can be avoided. However, nuclear details in Romanowsky-stained smears are not as well-visualized as in wet-fixed and Papanicolaou-stained smears. A parallel use of air-dried and wet-fixed smears is usually recommended, as these two staining methods are complementary [13,22,23]. Fixation of aspiration smears in Carnoy solution for 3–5 minutes may be used to lyse red blood cells prior to staining with the Papanicolaou method.
Specimen Adequacy
Obtaining an adequate cell sample is a prerequisite to the success of thyroid cytology. Therefore, immediate microscopic assessment of the needle aspirate by a pathologist or a cytotechnologist is desirable. If the first sample is judged inadequate for cytological evaluation, the TN can be re-aspirated immediately. If a rapid evaluation is not available, multiple FNAs of different areas of the TN should be performed.
The range of inadequate or unsatisfactory specimens reported in the literature ranges from 2–21% (means 17%) [15]. Currently, criteria for specimen adequacy vary from institution to institution. Some investigators require that an adequate sample should contains five to six groups of well-preserved and well-visualized follicular cells with each group containing 10 or more cells [12]. One group requires multiple punctures of the TN to be evaluated, with at least six properly prepared smears and a minimum of 8–10 tissue fragments of well-preserved follicular epithelium on each of two slides [25]. Another group requires 10 clusters of follicular cells with at least 20 cells in each cluster [13]. The Papanicolaou Society of Cytopathology Task Forces on Standard of Practice does not specify any numbers and groups of thyroid follicular epithelial cells for specimen adequacy [23]. Two practical exceptions to these adequacy criteria are applied: (a) a benign colloid nodule may be suggested if a large amount of thick colloid material is present, regardless of the number of follicular epithelial cell clusters [23]; or, (b) if a cell sample contains one or two small clusters of malignant or highly atypical cells, it should be reported as malignant or suspicious for malignancy and not as unsatisfactory or inadequate for cytodiagnosis [23]. Thyroid FNA under ultrasonographic guidance achieved higher rates of adequate cell samples, in the range of 79–99.3% (mean, 91%) [21,27-36]. Ultrasound-guided thyroid FNA proved to be useful in sampling TNs smaller than 2 cm in greatest dimension, complex or solid-cystic TNs [27-36] and abnormal thyroid beds [35,36].
Cytodiagnosis and Its Limitations
The cytodiagnosis of TNs by FNA is complex for the following reasons [26]:
a. overlap of cytological patterns between neoplastic and non-neoplastic lesions.
b. overlap of cytological features between various neoplasms.
c. coexistence of non-neoplastic and neoplastic processes and multiple malignancies in the same gland.
For a practical diagnostic approach, the cytological findings of thyroid lesions may be divided into seven main groups, as recommended by the Papanicolaou Task Force on Standard of Practice [23]. These groups are heterogenous and consist of both neoplastic and non-neoplastic lesions that may show either similar or specific cytological manifestations [23]. A non-diagnostic group is added as some TNs yield inadequate or non-specific cytological findings. The above-mentioned groups with their commonly encountered thyroid lesions are tabulated in Table 1.
Table 1 Cytodiagnostic Groups with Commonly Encountered Thyroid Nodular Lesions*.
1. Benign colloid nodule - Solitary colloid nodule
- Prominent nodule in MNG
- Macrofollicular adenoma
2. Cellular microfollicular lesion - Microfollicular adenoma
- Low-grade follicular carcinoma
- Hyperplastic microfollicular lesions in HT or MNG
3. Hurthle cell lesion - Hurthle cell adenoma
- Hurthle cell carcinoma
- Hyperplastic Hurthle cell nodule in HT or MNG
4. Primary malignant tumor - Papillary carcinoma
- High-grade microfollicular carcinoma
- Insular carcinoma
- Medullary carcinoma
- Anaplastic carcinoma
- Lymphoma
5. Cystic lesions - Benign colloid nodule
- Papillary carcinoma
- Other thyroid neoplasms
6. Thyroiditis - Acute thyroiditis
- Hashimoto thyroiditis
- Subacute thyroiditis
7. Other lesions - Graves disease
- Metastatic cancer
8. Non-diagnostic category -
* HT, Hashimoto thyroiditis; MNG, multinodular colloid goiter
1. Benign Colloid Nodule
This group includes solitary benign colloid nodules and prominent benign colloid nodules in a multinodular colloid goiter. These lesions arecharacterized by abundant, thick colloid material with cracking or bubble pattern (Figs. 1 and 2) and sheets of benign follicular epithelial cells in honeycomb arrangement (Fig. 3). Clusters of slightly hyperplastic Hurthle cells may be present [12,22,23,25]. The cytological differential diagnosis between a benign colloid nodule and a macrofollicular adenoma of the thyroid is extremely difficult if not impossible, as the two lesions usually show abundant, thick colloid and similar follicular cells [22,23].
Figure 1 Thick, deep blue colloid material with cracking pattern in FNA of a benign colloid nodule (Diff-Quik stain, × 250).
Figure 2 Thick, deep blue colloid material with bubble pattern in FNA of a benign colloid nodule (Diff-Quik stain, × 250) view).
Figure 3 A monolayered sheet of benign follicular epithelial cells with honeycomb pattern in FNA of a benign colloid nodule (Diff-Quik stain, × 400).
2. Cellular Microfollicular Lesion
This group includes hyperplastic microfollicular nodules in a multinodular colloid goiter or Hashimoto thyroiditis, a microfollicular adenoma, and a well-differentiated follicular carcinoma. These lesions are the most challenging ones to diagnose cytologically [22-25]. They are commonly reported as a microfollicular lesion or tumor with a recommendation for surgical excision [13,22-24]. FNA from a microfollicular lesion usually reveals abundant follicular cells in clusters, acini and small monolayered sheets (Figs 4 and 5). The individual cells show scanty, ill-defined cytoplasm and oval nuclei with regular nuclear contours and inconspicuous or prominent nucleoli.
Figure 4 Cellular microfollicular lesion showing in FNA cells with round nuclei arranged in acini and small monolayered sheet (Papanicolaou stain, 4 × 160 and 5 × 400).
Figure 5 Cellular microfollicular lesion showing in FNA cells with round nuclei arranged in acini (Papanicolaou stain, 4 × 160 and 5 × 400).
Cellular microfollicular lesions of the thyroid fall into the diagnostic category of indeterminate or suspicious lesions [14,15,22], and in one large series 14% of microfollicular lesions were malignant [12].
3. Hurthle Cell Lesion
Diagnosis of Hurthle cell lesions is a challenge in thyroid cytology. A hyperplastic Hurthle cell nodule in a Hashimoto thyroiditis or in a multinodular colloid goiter and a Hurthle cell neoplasm display similar cytologic findings [22-25,37,38]. The presence of numerous lymphocytes or a large amount of thick colloid material in the needle aspirate may indicate a hyperplastic Hurthle cell nodule in Hashimoto disease or a multinodular colloid goiter, respectively [38]. Hurthle cell adenoma and carcinoma usually show similar cytologic findings that are characterized by sheets and clusters of polygonal epithelial cells with abundant, granular, eosinophilic or basophilic cytoplasm, oval nuclei with regular nuclear contours and conspicuous or inconspicuous nucleoli [22-25] (Figs. 6 and 7). The presence of syncytial clusters of Hurthle cells with or without prominent nuclei [25] and abundant naked tumor cell nuclei has been reported to be a feature of Hurthle cell carcinoma [38].
Figure 6 Hurthle cells with abundant, granular cytoplasm and round, central or eccentrically located nuclei in FNA of a Hurthle cell lesion (Papanicolaou stain, × 400).
Figure 7 Hurthle cells in loose, monolayered sheet and singly in FNA of a Hurthle cell lesion (Diff-Quik stain, × 400).
When a Hurthle cell lesion is detected by FNA, surgical excision is usually indicated for further histologic study [38]. Thyroid Hurthle cell lesions fall into the cytodiagnostic category of indeterminate lesions or suspected malignant lesions [14,15,22], and 13% of Hurthle cell lesions were malignant in one large series [12].
4. Primary Malignant Lesions
This group includes papillary, high-grade follicular, insular, medullary and anaplastic carcinomas, and lymphoma. These lesions commonly show distinctive cytologic features that permit a correct identification in the majority of cases [13,24,25]. An insular carcinoma, or poorly differentiated carcinoma yields small cells in clusters, similar to those of a high-grade microfollicular carcinoma [39].
4a. Papillary carcinoma (PC). The conventional PC is characterized in FNA by the presence of thick or thin papillary tissue fragments with fibrovascular cores, sheets of tumor cells showing focal nuclear crowding and overlapping, irregular nuclear contours, intranuclear cytoplasmic inclusions (INCI) and nuclear grooves (NG). Psammoma bodies and metaplastic squamous cells may also be present [13,22,24,25] (Figs. 8, 9, 10, 11, 12, 13). These nuclear changes are recognized with less difficulty in Papanicolaou-stained cell samples, but they may be difficult to identify in cell samples stained with the Romanowsky staining method [13,22,23]. However, a presence of minute true papillary tissue fragments with fibrous vascular cores even without the identifiable above-mentioned nuclear changes is indicative of a PC. These papillary tissue fragments should be differentiated from thick and large follicular epithelial cell clusters with vascular transgression that may be found in FNA from different types of non-papillary epithelial neoplasms of the gland [40].
Figure 8 Thick branching papillary tissue fragment with fibrovascular core in FNA of a conventional papillary carcinoma (Papanicolaou stain, × 100).
Figure 9 Thin branching papillary tissue fragment with fibrovascular core in FNA of a conventional papillary carcinoma (Papanicolaou stain, × 100).
Figure 10 A sheet of tumor cells showing focal nuclear crowding with several cells displaying nuclear grooves in FNA of a conventional papillary carcinoma (Papanicolaou stain, × 400).
Figure 11 A loose sheet of tumor cells showing minimal nuclear crowding and two cells with intranuclear cytoplasmic inclusions in FNA of a conventional papillary carcinoma (Diff-Quik stain, × 400).
Figure 12 Two psammoma bodies in a smear showing a small amount of colloid material. A small aggregate of poorly preserved follicular cells is seen beside one psammoma body (Papanicolaou stain, × 400).
Figure 13 A loose cluster of metaplastic squamous cells seen in FNA of a conventional papillary carcinoma (Papanicolaou stain, × 400).
- Micro- and macrofollicular PCs constitute a diagnostic challenge. A microfollicular PC may show in FNA follicular cells forming acini similar to those seen in the aforementioned cellular microfollicular lesions, and a macrofollicular PC may be easily mistaken for a macrofollicular adenoma or a benign colloid nodule cytologically, as nuclear changes characteristic for a thyroid PC may not be seen [41-43] (Fig. 14).
Figure 14 Papillary carcinoma, microfollicular variant showing in FNA cells in acinar arrangement. A tumor cell with an intranuclear cytoplasmic inclusion is noted (Papanicolaou stain, × 400).
- Hyalinizing trabecular adenoma is indistinguishable from a PC cytologically, as these two lesions yield cells with similar nuclear features [44]. Recent molecular studies have suggested that this tumor is actually an encapsulated trabecular variant of thyroid PC [45].
- Other PC carcinoma subtypes. Tall-cell PC is characterized by the presence of tall tumor cells with well-defined, granular cytoplasm and nuclei with NGs and single or multiple INCIs, making at least 30% of the aspirated cells [46-51] (Fig. 15). Columnar-cell variant shows no classic cytologic features of thyroid PC, but presence of clusters of columnar cells with palisading nuclei and the absence of classic nuclear changes of thyroid PC are cellular features of this neoplasm [52]. Diffuse sclerosing variant can be confidently suggested when abundant squamous cells admixed with lymphocytes, follicular epithelial cells with nuclear features of papillary carcinoma and a few psammoma bodies are noted [53,54] (Fig. 16).
Figure 15 Papillary carcinoma, tall-cell variant showing in FNA a sheet of pleomorphic cells with some cells with elongated configuration and cytoplasmic tails. A tumor cell with intranuclear cytoplasmic inclusion is present (Diff-Quik stain, × 400).
Figure 16 Papillary carcinoma, diffuse sclerosing type showing in FNA a sheet of metaplastic squamous cells, scattered lymphocytes and a psammoma body (Papanicolaou stain, × 400).
4b. A high-grade follicular carcinoma and insular carcinoma are characterized by sheets and acinar clusters of pleomorphic epithelial cells with prominent nucleoli [22,24,25].
4c. A medullary carcinoma shows in FNA a mixture of single and clustered polygonal cells and spindle tumor cells that may display INCIs [22,24,25] (Figs. 17 and 18). The tumor cells cytoplasm may show intracytoplasmic pink azurophil granules that are well-visualized by MGG or Diff-Quik stain and stain positively with calcitonin antibody. Amyloid material that stains positively with Congo red may be seen (Fig. 19).
Figure 17 Medullary carcinoma showing in FNA dyshesive plasmacytoid tumor cells with eccentrically located round nuclei and intracytoplasmic azurophil granules (Diff-Quik stain, × 400).
Figure 18 Medullary carcinoma showing in FNA loosely clustered spindle-shaped tumor cells with scanty, ill-defined cytoplasm (Papanicolaou stain, × 400).
Figure 19 A fragment of orange and granular amyloid material seen in FNA of a thyroid medullary carcinoma (Papanicolaou stain, × 400).
4.d. A naplastic thyroid carcinoma consists of two main histologic variants: Giant cell and spindle cell-subtypes. Depending on the histologic subtype, an anaplastic thyroid carcinoma may display in FNA pleomorphic large, bizarre cancer cells with prominent nucleoli or spindle cancer cells admixed with a variable amount of necrotic debris (Figs. 20 and 21).
Figure 20 Anaplastic carcinoma, giant-cell type showing in FNA single and clustered large, bizarre malignant cells with pleomorphic nuclei and prominent nucleoli (Papanicolaou stain, × 400).
Figure 21 Anaplastic carcinoma, spindle-cell type showing in FNA dyshesive spindle- shaped malignant cells with scant, ill-defined cytoplasm (Papanicolaou stain, × 400).
4.e. A primary thyroid non-Hodgkin lymphoma is usually of large cell type and yields in FNA cells similar to those of a lymph node involved by the same neoplastic process. A thyroid Hodgkin disease is characterized by Reed-Sternberg cells admixed with benign lymphoid cells and eosinophils [13,24,25].
5. Cystic Lesion
Benign cysts account for the majority of thyroid cystic lesions. They are formed as the result of hemorrhagic degeneration of a benign colloid nodule. FNA from a benign colloid cyst may show colloid material admixed with benign follicular epithelial cells and hemosiderin laden macrophages. However, any thyroid neoplasm may undergo hemorrhagic necrosis and become a cystic lesion [13,22-25]. Of the thyroid neoplasms, PC tends to undergo marked hemorrhagic degenerative change. FNA from the tumor commonly shows a large amount of blood and the cystic lesion tends to recur rapidly [23]. Cytological examination of the aspiration smears usually reveals a large amount of blood and rarely tumor cells. However, sections from the cell block prepared from the needle aspirate may show diagnostic papillary tissue fragments with fibrovascular cores or nuclear features of a PC [23] while that of a benign colloid nodule will show no true papillary tissue fragments with fibrovascular cores or nuclear features of a thyroid PC (Figs. 22 and 23).
Figure 22 Papillary tissue fragments with thin fibrovascular cores covered with epithelial cells displaying nuclear crowding and occasional intranuclear cytoplasmic inclusions seen in a cell block section prepared from the needle aspirate of a papillary carcinoma with hemorrhagic cystic degenerative change (hematoxylin and eosin stain, × 250).
Figure 23 Section from a cell block prepared from the needle aspirate of a benign colloid nodule with hemorrhagic cystic degenerative change showing papillary tissue fragments covered with epithelium displaying no nuclear changes characteristic for a papillary carcinoma (hematoxylin and eosin stain, × 250).
6. Thyroiditis
Hashimoto thyroiditis and subacute thyroiditis commonly have fairly distinctive clinical findings. Rarely, these lesions may present as a nodular lesion mimicking a thyroid neoplasm. Hashimoto thyroiditis is characterized by the presence of numerous benign lymphoid cells admixed with benign follicular cells and Hurthle cells, (Figs. 24 and 25). A subacute thyroiditis may yield clustered epithelioid cells, scattered lymphocytes and a few multinucleated giant cells containing up to one hundred nuclei [13,22-25,37] (Figs. 26 and 27). It should be born in mind that Hashimoto thyroiditis may harbor hyperplastic follicular and Hurthle cell nodules, and these two nodules are cytologically indistinguishable from a cellular follicular neoplasm and a Hurthle cell neoplasm, respectively [37,38]. Surgical excision of these lesions is usually required for histologic confirmation.
Figure 24 Hashimoto thyroiditis showing in FNA numerous lymphoid cells admixed with a sheet of follicular epithelial cells (Papanicolaou stain, × 100).
Figure 25 A sheet of follicular epithelial cells with oncocytic change admixed with benign lymphoid cells seen in FNA of a Hashimoto thyroiditis (Papanicolaou stain, × 400).
Figure 26 A syncytial cluster of epithelioid cells with carrot-shaped nuclei seen in FNA of a subacute thyroiditis (Diff-Quik stain, × 400).
Figure 27 A large multinucleate giant cell present in FNA of a subacute thyroiditis (Diff-Quik stain, × 400).
7. Other Lesions
Graves disease may rarely present as a nodular thyroid lesion [55]. It yields non specific cytologic findings [13].
Metastatic cancers to the thyroid are common in patients with advanced cancers arising from other body sites [25]. However, metastatic cancer to the thyroid gland presenting as a palpable TN is uncommon. For unknown reasons, renal cell carcinoma is the most common metastatic neoplasm to the thyroid, and cases of clinically occult renal cell carcinoma presenting initially as a large thyroid mass have been documented [25]. Cytodiagnosis of metastatic cancer to the thyroid is relatively straightforward as metastatic cancer usually displays a cytologic pattern distinctive from those of a primary thyroid carcinoma [25]. However, a cytological differential diagnosis between a metastatic renal cell carcinoma of clear cell type and a primary thyroid carcinoma with clear cell change may be difficult, and immunocytochemical staining of aspirated tumor cells with thyroglobulin antibody will be helpful to identify the aforementioned primary thyroid cancer.
8. Non-Diagnostic Category
The lesions in this category are highly diversified and may be any lesions listed in the above seven categories. In this category the FNA yields non-diagnostic or inadequate cellular materials. In one study, cystic thyroid lesions yielded non-diagnostic cell samples at initial FNA in about 50% of cases [12]. In the Mayo Clinic experience, repeating the FNA in the cases with initial non-diagnostic needle aspirates revealed diagnostic material in 30 to 80% of cases [12,15]. Other investigators found that thyroid re-FNA was of limited value [59]. If the re-aspiration is still non-diagnostic, ultrasound-guided FNA should be performed. Ultrasound-guided FNAs yield adequate cytologic materials in about 91% of cases [27-36]. Patients with no specific risk factors for thyroid malignancy and a non-diagnostic FNA who refuse a re-biopsy may be managed conservatively. While patients in the high-risk group should have their TNs removed for histologic study, an increase in nodule volume alone is not a reliable predictor of malignancy, as most solid and benign TNs grow in size [57].
Diagnostic Accuracy and Errors
In a review of seven large series totaling 18,183 thyroid FNAs, Gharib and Goellner found that the biopsy technique had a sensitivity rate varying from 65 to 98% (mean 83%), and that its specificity rate varied from 72 to 100% (mean 92%) [15]. The false-negative rate varied from 1 to 11.5% (mean, 5.2%), and the false-positive rate varied from 0 to 7.7% (mean, 2.9%) [15]. The overall cytodiagnostic accuracy rate of thyroid FNA approached 95% according to some reported series [13].
Adjunctive Diagnostic Value of Ancillary Techniques
Ultrafast Papanicolaou stain selectively swells the nuclei of papillary thyroid carcinoma, making their nuclear grooves disappear and making the swollen nuclei look like "watery grapes", while this staining method has no effect on nuclei of a follicular adenoma [21]. This artifactual change is due to the disorganization of nuclear lamins and permits a confident distinction between a follicular adenoma and a follicular variant papillary carcinoma [21]. Immunostaining with thyroid peroxidase antibody has been reported to be of value in distinguishing these two lesions, as malignant and benign follicular cells commonly stain negatively and positively with this antibody, respectively [58].
Ploidy determination has no value in distinguishing a follicular adenoma from a follicular carcinoma [59-62] and immunostaining for p53, Ki-67 and Bcl-2 has no value in separating benign from malignant Hurthle cell tumors [63].
Genetics-molecular studies have been extensively carried out on tissue samples of different types of thyroid neoplasm since the past decade [64]. However, only a few genetics-molecular studies on thyroid cells obtained by FNA have been recently published. Human telomerase reverse transcriptase (hTERT) gene expression, using reverse transcriptase-polymerase chain reaction, has been identified as a promising diagnostic marker in distinguishing benign from malignant tumors in materials obtained by FNA. It was found that 90 and 92.8% of thyroid carcinomas were positive for hTERT while 35 and 61.5% of benign thyroid nodules were positive for hTERT, respectively [65,66]. Among the thyroid tumors with positive hTERT, there were eight of eight papillary, two of two Hurthle cell and three of four follicular carcinomas [65]. BRAF point mutation and RET/thyroid PC rearrangements were found in 38% of thyroid PCs and refined the diagnosis of thyroid PC in five of fifteen cell samples that were considered either indeterminate or insufficient by cytology. No mutation was found in FNAs of follicular adenomas and non-toxic nodular goiters [67]. These molecular markers were of adjunctive diagnostic value when the FNA diagnosis of TN was equivocal [65-67].
Powerful molecular techniques including microarray analysis and molecular profiling may have a significant role in the future evaluation of TNs, while providing impetus for further insight into the molecular pathogenesis of both benign and malignant TNs [68-71]. Moreover, such techniques may allow deeper insight into both loss and gain of function of unidentified genes by examining panels of genes rather than one or a limited number of potential gene candidates. By analysis of cancer gene profiles for a cohort of 62 thyroid samples, Finley et al [68] were able to distinguish between benign and malignant thyroid tumors. They reported a sensitivity of 91.7% and specificity of 96.2% for the detection of thyroid carcinomas of various types, including thyroid PC and its follicular variant and follicular carcinoma [68]. Distinction of benign and malignant thyroid tumors and molecular classification of follicular thyroid tumors by gene profiling suggests that these powerful techniques may have significant diagnostic potential when used with FNA cytology [69,70]. Molecular profiling may also permit the distinction between primary and metastatic malignancies when dealing with multiple suspicious nodules at various sites. Using material retrieved by FNA, Schoedel et al [71], compared loss of heterozygosity (LOH) patterns and demonstrated genetic kinship of multifocal carcinomas in the thyroid and a separate nodule in the lung, supporting a diagnosis of metastatic thyroid carcinoma to the lung rather than an independent lung neoplasm.
At present, techniques such as microarray analysis are limited by the amount of RNA that can be retrieved from a sample, thereby often limiting analysis to surgically resected samples. However, refinement of these techniques may make them applicable to FNA, with extraction of RNA from a cell block from which molecular analysis of FNA material may have significant diagnostic benefit.
Acknowledgements
Co-editors of Cytojournal Vinod B. Shidham, MD, FRCpath, FIAC, and Barbara F. Atkinson, MD thank the academic editor: Zubair W. Baloch, MD, PhD.
Hospital of the University of Pennsylvania Pathology, 6 Founders, 3400 Spuce St, Philadelphia, PA 19104, USA Email: [email protected].
==== Refs
Martin HE Ellis EB Biopsy by needle puncture and aspiration Am Surg 1930 92 169 181
Stewart FW The diagnosis of tumors by aspiration Am J Pathol 1933 9 801 812
Frazell EL Foote FW Papillary carcinoma of the thyroid: a review of 25 years of experience Cancer 1958 11 895 922 13585342
Soderstrom N Puncture of goiters for aspiration biopsy. A primary report Acta Med Scand 1952 144 237 244 13007429
Eihorn J Franzen S Thin needle biopsy in the diagnosis of thyroid disease Acta Radiol 1962 58 321 336
Persson PS Cytodiagnosis of thyroiditis Acta Med Scand (suppl) 1967 483 8 100
Ljunberg O Cytologic diagnosis of medullary carcinoma of the thyroid gland with special regard to the demonstration of amyloid in smears of fine needle aspirates Acta Cytol 1972 16 253 255 4112681
Lowhagen T Sprenger E Cytologic presentation of thyroid tumours inaspiration biopsy smear. A review of 60 cases Acta Cytol 1974 18 192 197 4525519
Crockford PM Bain GO Fine needle aspiration biopsy of the thyroid Can Med Assoc J 1974 110 1029 1032 4406333
Miller JM Hamburger JI Kini SR Diagnosis of thyroid nodules by fine needle aspiration and needle biopsy JAMA 1979 241 481 486 366189 10.1001/jama.241.5.481
Jayaram G Fine-needle aspiration cytologic study of solitary thyroid nodules. Profiles of 308 cases with histologic correlation Acta Cytol 1985 29 967 973 3866460
Goellner JR Gharib H Grant CS Johnson DA Fine-needle aspirationcytology of the thyroid, 1980–1986 Acta Cytol 1987 31 587 590 3673463
Nguyen GK Ginsberg J Crockford PM Fine-needle aspiration biopsy cytology of the thyroid. Its value and limitations in the diagnosis and management of solitary thyroid nodules Pathol Annu 1991 25 63 91 2014145
Mazzaferri EL Management of a solitary thyroid nodule N Engl J Med 1993 328 553 559 8426623 10.1056/NEJM199302253280807
Gharib H Goellner JR Fine-needle aspiration biopsy of the thyroid; an appraisal Ann Int Med 1993 118 282 289 8420446
Komatsu M Hanamura N Tsuchiya S Preoperative diagnosis of follicular variant of papillary carcinoma of the thyroid: discrepancy between image and cytologic diagnosis Rat Med 1994 12 293 299
Danese D Centanni M Farsetti A Andreoli M Diagnosis of thyroid carcinoma J Exp Clin Cancer Res 1997 16 337 347 9387911
Giovagnoli MR Pisani T Drusco A Fine needle aspiration biopsy in the preoperative management of patients with thyroid nodules Anticancer Res 1998 18 3741 3745 9854487
Wong CK Wheeler MH Thyroid nodules: rational management World J Surg 2000 24 934 941 10865037 10.1007/s002680010175
Werga P Wallin G Skoog L Hamberger B Expanding role of fine-needle aspiration cytology in thyroid diagnosis and management World J Surg 2000 24 907 912 10865034 10.1007/s002680010163
Yang GC Liebeskind D Messina AV Ultrasound-guided fine-needle aspiration of the thyroid assessed by Ultrafast Papanicolaou stain: data from 1135 biopsies with a two- to six-year follow-up Thyroid 2001 11 581 589 11442006 10.1089/105072501750302895
Orell S Philips J Broadsheet number 57. Problems in fine needle biopsy of the thyroid Pathology 2000 32 191 198 10968393
Papanicolaou Society of Cytopathology Task Force on Standard of Practice (Suen KC, Chair) Guidelines of the Papanicolaou Society of Cytopathology for the examination of fine-needle aspiration specimens from thyroid nodules Mod Pathol 1996 9 710 715 8782212
DeMay RM The Art & Science of Cytopathology Aspiration cytology 1996 Chicago, ASCP Press 703
Kini SR Guide to clinical aspiration biopsy Thyroid 1997 New York, Igaku-Shoin
McCall A Jarosz H Lawrence AM Paloyan E The incidence of thyroid carcinoma in solitary cold nodules and multinodular goiter Surgery 1986 100 1128 1132 3787469
Cochand-Priollet B Guillausseau PJ Chagnon S The diagnostic value of fine-needle aspiration biopsy under ultrasonography in nonfunctional thyroid nodules: a prospective study comparing cytologic and histologic findings Am J Med 1994 97 152 157 8059781 10.1016/0002-9343(94)90025-6
Danese D Sciacchitano S Farsetti A Diagnostic accuracy of conventional versus sonography-guided fine-needle aspiration biopsy of thyroid nodules Thyroid 1998 8 15 21 9492148
Sabel MS Haque D Velasco JM Staren ED Use of ultrasound-guided fine needle aspiration biopsy in the management of thyroid disease Am Surg 1998 64 738 741 9697903
Tollin SR Mery GM Jelveh N The use of fine-needle aspiration biopsy under ultrasound guidance to assess the risk of malignancy in patients with a multinodular goiter Thyroid 2000 10 235 241 10779138
Braga M Cavalcanti TC Collaco LM Graf H Efficacy of ultrasound-guided fine-needle aspiration biopsy in the diagnosis of complex thyroid nodules J Clin Endocrinol Metab 2001 86 4089 4091 11549630 10.1210/jc.86.9.4089
Hatada T Okada K Ishii S Utsunomiya J Evaluation of ultrasound-guided fine-needle aspiration biopsy for thyroid nodules Am J Surg 1998 175 133 136 9515530 10.1016/S0002-9610(97)00274-2
Mittendorf EA Tamarkin SW McHenry CR The results of ultrasound-guided fine-needle aspiration biopsy for evaluation of nodular thyroid disease Surgery 2002 132 648 653 12407349 10.1067/msy.2002.127549
Court-Payen M Nygaard B Horn T US-guided fine-needle aspiration biopsy of thyroid nodules Acta Radiol 2002 43 131 140 12010289
Tambouret R Szyfelbein WM Pitman MB Ultrasound-guided fine-needle aspiration biopsy of the thyroid Cancer 1999 87 299 305 10536356 10.1002/(SICI)1097-0142(19991025)87:5<299::AID-CNCR10>3.0.CO;2-M
Krishnamurthy S Bedi DG Caraway NP Ultrasound-guided fine-needle aspiration biopsy of the thyroid bed Cancer 2001 93 199 205 11391607 10.1002/cncr.9029
Nguyen GK Ginsberg J Crockford PM Villanueva RR Hashimoto's disease. Needle aspiration cytology: diagnostic accuracy and pitfalls Diagn Cytopathol 1997 16 531 536 9181321 10.1002/(SICI)1097-0339(199706)16:6<531::AID-DC12>3.0.CO;2-J
Nguyen GK Husain M Akin MRM Diagnosis of benign and malignant Hurthle cell lesions of the thyroid by fine-needle aspiration biopsy cytology Diagn Cytopathol 1999 20 261 265 10319225 10.1002/(SICI)1097-0339(199905)20:5<261::AID-DC3>3.0.CO;2-E
Nguyen GK Akin MRM Cytopathology of insular carcinoma of the thyroid Diagn Cytopathol 2001 25 325 330 11747225 10.1002/dc.2164
Nguyen GK Fine-needle aspiration cytology of oncocytic papillary thyroid carcinoma Diagn Cytopathol 2000 23 402 405 11074646 10.1002/1097-0339(200012)23:6<402::AID-DC8>3.0.CO;2-6
Mesonero CE Jugle JE Wilbur DC Nayar R Fine needle aspiration of macrofollicular and microfollicular subtypes of follicular variant papillary carcinoma of the thyroid Cancer 1998 84 235 244 9723599 10.1002/(SICI)1097-0142(19980825)84:4<235::AID-CNCR9>3.0.CO;2-L
Zacks JF de las Morelas A Beazley RM O'Brian MJ Fine-needle aspiration cytology diagnosis of colloid nodules versus follicular variant papillary carcinoma of the thyroid Diagn Cytopathol 1998 18 87 90 9484634 10.1002/(SICI)1097-0339(199802)18:2<87::AID-DC1>3.0.CO;2-Q
Balock ZW Gupta PK Yu GH Follicular variant of papillary carcinoma. Cytologic and histologic correlation Am J Clin Pathol 1999 111 216 222 9930143
Akin MRM Nguyen GK Needle aspiration biopsy cytology of hyalinizing trabecular adenomas of the thyroid Diagn Cytopathol 1999 20 90 94 9951605 10.1002/(SICI)1097-0339(199902)20:2<90::AID-DC10>3.0.CO;2-N
Cheung CC Boemer SL MacMillan CM Hyalinizing trabecular tumor of the thyroid: a variant of papillary carcinoma proved by molecular genetics Am J Surg Pathol 2000 24 1622 1626 11117782
Ohori NP Schoedel KE Cytopathology of high-grade papillary thyroid carcinomas: tall cell variant, diffuse sclerosing variant and poorly differentiated papillary carcinoma Diagn Cytopathol 1999 20 19 23 9884822 10.1002/(SICI)1097-0339(199901)20:1<19::AID-DC5>3.0.CO;2-R
Harach HR Zusman SB Cytopathology of tall-cell variant of thyroidpapillary carcinoma Acta Cytol 1992 36 895 899 1449028
Gamboa-Dominguez A Candanedo-Gonzalez F Uribe-Uribe NO Angeles-Angeles A Tall-cell variant of papillary thyroid carcinoma: A cytohistologic correlation Acta Cytol 1997 18 672 676 9167681
Bocklage T DiTomasso JP Ramzy I Ostrowski ML Tall-cell variant of papillary thyroid carcinoma: cytologic features and differential diagnostic considerations Diagn Cytopathol 1997 17 25 29 9218899 10.1002/(SICI)1097-0339(199707)17:1<25::AID-DC5>3.0.CO;2-O
Solomon A Gupta PK LiVolsi VA Baloch ZW Distingishing tall-cellvariant of papillary thyroid carcinoma from usual variant of thyroid papillary carcinoma in cytologic specimens Diagn Cytopathol 2002 27 143 148 12203860 10.1002/dc.10156
Nair M Kapila K Karak AK Verma K Papillary carcinoma of the thyroid and its variants: a cytohistologic correlation Diagn Cytopathol 2002 24 167 173 11241899 10.1002/1097-0339(200103)24:3<167::AID-DC1035>3.0.CO;2-3
Ylagan LR Dehner LP Huettner PC Lu D Columnar cell variant of papillary thyroid carcinoma. Report of a case with cytologic findings Acta Cytol 2004 48 73 77 14969185
Caruso G Tabarri B Lucchi I Tison V Fine-needle aspiration cytology in a case of diffuse sclerosing carcinoma of the thyroid Acta Cytol 1990 34 352 354 2343691
Kumarasinghe MP Cytopathologic features of diffuse sclerosing variant of papillary carcinoma of the thyroid. Report of two cases in children Acta Cytol 1998 42 983 986 9684590
Carnell NE Valente WA Thyroid nodules in Graves' disease: classification, characterization, and response to treatment Thyroid 1998 8 571 576 9709909
Merchant SH Izquierdo R Khurana KK Is repeated fine-needle aspiration cytology useful in the management of patients with benign thyroid disease? Thyroid 2000 10 489 492 10907992
Alexander EK Hurwitz S Heering JP Natural history of benign solid and cystic thyroid nodules Ann Intern Med 2003 138 315 318 12585829
Christensen L Blichert-Toft M Brandt M Thyroperoxidase (TPO) immunostaining of solitary cold thyroid nodule Clin Endocrinology (Oxf) 2000 53 161 169 10.1046/j.1365-2265.2000.01035.x
Harlow SO Duda RB Bauer KD Diagnostic utility of DNA content flow cytometry in follicular neoplasms of the thyroid J Surg Oncol 1992 50 1 6 1573886
Oyama T Vickery AL JrPreffer FI Colvin RB A comparative study of flow cytometry and histologic findings in thyroid follicular carcinomas and adenomas Human Pathol 1994 25 271 275 8150457 10.1016/0046-8177(94)90199-6
Czyz W Joensuu H Pylkkanen Klempi PJ P53 protein, PCNA staining, and DNA content in follicular neoplasms of the thyroid J Pathol 1994 174 267 274 7884588 10.1002/path.1711740406
Ryan JJ Hay ID Grant CS Flow cytometric DNA measurements in benign and malignant Hurthle cell tumors of the thyroid World J Surg 1988 12 482 487 3420931 10.1007/BF01655427
Muller-Hocker J Immunoreactivity of p53, Ki-67, and Bcl-2 in oncocytic adenomas and carcinomas of the thyroid Human Pathol 1999 30 926 933 10452505 10.1016/S0046-8177(99)90246-0
Baloch ZW LiVolsi VA Mills SE, Carter D, Greeson JK, Oberman HA, Reuter V, Stoler MH Pathology of thyroid and parathyroid disease Sternberg's Diagnostic Surgical Pathology 2004 4 Lippincott Williams & Wilkins 558 619
Liou MJ Chan EC Lin JD Liu FH Chao TC Human telomerase reversetranscriptase (hTERT) gene expression in FNA samples from thyroid neoplasms Cancer Lett 2003 191 223 227 12618337 10.1016/S0304-3835(02)00678-X
Umbrich CB Conrad GT Clark DP Human telomerase reverse transcriptase gene expression and surgical management of suspicious thyroid tumors Clin Cancer Res 2004 10 5762 5768 15355904
Salvatore G Giannini R Faviana P Analysis of BRAF point mutation and RET/PTC rearrangement refines the fine-needle aspiration diagnosis of papillary carcinoma J Clin Endocrinol Metab 2004 89 5175 5180 15472223 10.1210/jc.2003-032221
Finley DJ Zhu B Barden CB Fahey TJ III Discrimination of benign and malignant thyroid nodules by molecular profiling Ann Surg 2004 240 425 436 15319714
Mazzanti C Zeiger Ma Costouros NG Using gene expression profiling to differentiate benign versus malignant thyroid tumors Cancer Res 2004 64 2898 2903 15087409 10.1158/0008-5472.CAN-03-3811
Barden CB Shister KW Zhu B Classification of follicular thyroid tumors by molecular signature: results of gene profiling Clin Cancer Res 2003 9 1792 1800 12738736
Schoedel KE Finkelstein SD Swalsky PA Ohori NP Molecular profiling of primary and metastatic neoplasms in the lung using cytologic material obtained by fine-needle aspiration: report of two cases Diagn Cytopathol 2004 30 342 346 15108233 10.1002/dc.20047
|
15987502
|
PMC1184092
|
CC BY
|
2021-01-04 16:36:22
|
no
|
Cytojournal. 2005 Jun 29; 2:12
|
utf-8
|
Cytojournal
| 2,005 |
10.1186/1742-6413-2-12
|
oa_comm
|
==== Front
CytojournalCytoJournal1742-6413BioMed Central London 1742-6413-2-81594385810.1186/1742-6413-2-8Case ReportPreoperative diagnosis of a mediastinal granular cell tumor by EUS-FNA: a case report and review of the literature Bean Sarah M [email protected] Mohamad A [email protected] Isam A [email protected] Robert J [email protected] Darshana N [email protected] Department of Pathology, Division of Anatomic Pathology, the Department of Medicine, University of Alabama at Birmingham, Birmingham, Alabama, USA2 Department of Pathology, Division of Gastroenterology and Hepatology, University of Alabama at Birmingham, Birmingham, Alabama, USA3 Department of Cardiothoracic Surgery at the University of Alabama at Birmingham, Birmingham, Alabama, USA2005 8 6 2005 2 8 8 22 3 2005 8 6 2005 Copyright © 2005 Bean et al; licensee BioMed Central Ltd.2005Bean et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
We report the first case of a posterior mediastinal granular cell tumor initially diagnosed on cytologic material obtained via endoscopic ultrasound-guided fine needle aspiration (EUS-FNA) in a 51-year-old male with a prior history of colon cancer. Aspirates obtained were cellular and composed of polygonal cells with abundant granular cytoplasm and small, round dark nuclei. An immunoperoxidase stain performed on the cell block for antibodies to S-100 protein showed strong, diffuse staining of the cytoplasmic granules. Electron microscopy performed on the cell block revealed numerous cytoplasmic lysosomes. This is the first case report in the English literature of a definitive preoperative diagnosis of a mediastinal granular cell tumor utilizing material obtained via EUS-FNA.
granular cell tumorEUS-FNAcytologymediastinum
==== Body
Background
Granular cell tumors are uncommon generally benign soft tissue neoplasms first described in 1926 by Abrikossoff [1]. Granular cell tumors commonly occur in the tongue, gastrointestinal tract, mediastinum, skin, breast, as well as other sites [2-5]. Mediastinal granular cell tumors are, however, exceptionally rare. Fewer than ten cases have been previously described [6-11]. Historically, the diagnosis of mediastinal granular cell tumors has been made on histologic material. Cytologic material was available for only one of the previously described mediastinal granular cell tumors [6]. However, a definitive diagnosis was not rendered based upon the FNA specimen. In this report, we describe the first case report of a preoperative, definitive diagnosis of a mediastinal granular cell tumor diagnosed on a cytologic EUS-FNA specimen. Confirmatory ancillary studies including immunohistochemistry and electron microscopy were performed on the cell block obtained via EUS-FNA.
Case Presentation
A 4.9 × 3.1 cm peritracheal posterior mediastinal mass was incidentally discovered on a computed tomography (CT) scan of a 51 year-old gentleman status-post left hemicolectomy in October 2003 for a T3N1Mx, stage III colon cancer. He had moderate gastroesophageal reflux and mild dysphagia. A barium swallow revealed compression and displacement of the proximal thoracic esophagus. Esophageal mucosal ulcerations were not observed. Integrated positron emission tomography(PET)/CT scan using fluorodeoxyglucose (FDG) showed no uptake in the lesion. Due to the benign clinical features of the mass, the differential diagnosis included benign neoplasms such as a leiomyoma or a benign thyroid neoplasm. The possibility of lymphadenopathy secondary to metastatic adenocarcinoma was remote. Endoscopic ultrasound was performed using the UC-30P (Olympus, America, Melville, NY) and revealed a 52- × 39-mm hypoechoic, well-circumscribed posterior mediastinal mass in the thoracic inlet approximately 22 cm from the incisor (Figure 1). After using color Doppler ultrasound, EUS-FNA of the mediastinal mass was performed. Adequate material was obtained for on-site rapid cytopathologic interpretation. The patient underwent a posterior-lateral thoracotomy with entry over the fourth rib. A firm, fixed mass in the posterior mediastinum near the esophagus entering the base of the neck was identified. A 4.0 × 2.8 × 2.5 cm well-circumscribed soft tissue mass was completely resected.
Figure 1 Endoscopic ultrasound image of a 52 x 39 mm hypoechoic mass in the upper mediastinum (Olympus UC-30P).
Materials and methods
For cytologic evaluation, air-dried smears and alcohol-fixed smears were prepared in the endoscopy suite by the cytopathologist and stained with Diff-Quick and Papanicolaou, respectively. Additional material was collected in Hanks balanced salts solution from which a cell block was prepared. Representative histologic sections of the mass and the cell block were fixed in 10% neutral buffered formalin, processed, embedded in paraffin, sectioned to 4 μm thick sections and stained with hematoxylin and eosin.
Immunoperoxidase stains were performed on deparaffinized sections from the cytologic cell block as well as histologic sections in an autostainer. An avidin-biotin-peroxidase complex was used to detect the following antibodies: S-100 (Ventana; predilute), CD68 (Ventana; predilute), thyroglobulin (Cell marque; predilute), and TTF-1 (Dako; 1:50). Electron miscroscopy was performed on the paraffin-embedded cytologic cell block.
The Institutional Review Board at the University of Alabama at Birmingham approved the use of human tissue in this case report.
Pathologic Findings
The cytologic smears revealed a cellular aspirate composed of loosely cohesive groups of polygonal cells and single cells (Figure 2). The cells had a low N:C ratio with centrally-located small, round, dark nuclei. The cytoplasm contained abundant eosinophilic granules. A background of fine granular debris was also noted. Classification of the lesion was deferred at immediate assessment performed at the time of the procedure. It was noted, however, that diagnostic material had been obtained for ancillary laboratory studies. Immunoperoxidase stains were performed on the cell block with the following results: S-100 (Ventana; predilute) positive, CD68 (Ventana; predilute) positive, thyroglobulin (Cell marque; predilute) negative, and TTF-1 (Dako; 1:50) negative (Figure 3). Positive and negative controls for each stain were appropriately reactive. The morphologic findings and immunohistochemical staining pattern supported a diagnosis of granular cell tumor. Electron microscopy revealed sporadic cells with abundant cytoplasmic pleomorphic electron-dense lysosomes and a relative paucity of organelles characteristic of granular cell tumor (Figure 4).
Figure 2 Smears revealed loosely cohesive polygonal cells with small, round, dark nuclei and granular cytoplasm (Diff-Quick, 200X).
Figure 3 Immunocytochemistry performed on the cell block revealed that the cytoplasmic granules stain strongly for S-100 (400X).
Figure 4 Electron microscopy revealed abundant cytoplasmic lysosomes
Representative sections of the resected mass were submitted for histologic evaluation. The tumor consisted of sheets of cells similar to those seen in the fine needle aspiration (Figure 5). Frozen section diagnosis of the tumor was granular cell tumor. An immunoperoxidase stain for antibodies to S-100 protein (Ventana; predilute) was performed on a representative section of the formalin-fixed, paraffin embedded tumor. The tumor cells demonstrated diffuse, strong positive cytoplasmic staining with antibodies to S-100 protein (Figure 6), supporting the diagnosis of granular cell tumor.
Figure 5 Histologic sections of the resected tumor revealed sheets of polygonal cells with small, round, dark nuclei and granular cytoplasm (200X).
Figure 6 Immunohistochemistry performed on tissue section revealed that the cytoplasmic granules stain diffusely positive for S-100 (200X).
Discussion
Mediastinal granular cell tumors are very rare tumors. Prior to this report, only seven mediastinal granular cell tumors have been described in the literature [11]. The average age at diagnosis of those seven patients (including our patient) is 33 years. In all, four men and four women have been diagnosed with mediastinal granular cell tumors. All of the lesions were located in the posterior mediastinum. Four of the patients were asymptomatic, while three patients reported cough, dyspnea, wheezing, and/or chest pain. Our patient experienced mild dysphagia. For this reason, the granular cell tumor was resected. All of these lesions have been surgically resected. In six of the seven reported cases a diagnosis of granular cell tumor was rendered based upon histologic materials.
In 1997, Smith et al. described fine needle aspiration cytology findings of a mediastinal granular cell tumor in a 53 year-old woman [6]. While they described the cytologic findings of a granular cell tumor, they failed to make a definitive, preoperative diagnosis of granular cell tumor utilizing ancillary studies on cytology material. A definitive diagnosis of granular cell tumor was made only after surgical resection and ancillary studies (immunohistochemistry and electron microscopy) were performed on a tissue section.
In the past, mediastinal lesions were sampled utilizing mediastinoscopy, transtracheal biopsy, or CT-guided biopsy. Superior and posterior mediastinal masses have been difficult to approach using those sampling modalities. EUS-FNA, however, has emerged as a safe, effective method of tissue sampling in the mediastinum [12]. Here we describe the first case of a mediastinal granular cell tumor definitively diagnosed based upon EUS-FNA material prior to surgical resection.
The cytologic findings of granular cell tumors have been previously described in organs such as esophagus, breast, tongue, bronchus, and skin [13-15]. Typically smears are composed of large round or polygonal cells with indistinct cytoplasmic borders, abundant PAS positive and S-100 positive cytoplasmic granules, and small round dark eccentric nuclei. In our case, the smears obtained by EUS-FNA were cellular with groups of loosely cohesive sheets and single polygonal to round cells with granular cytoplasm and round, dark nuclei without nucleoli in a background of fine granular debris. Given the anatomic location of the lesion and the morphologic findings, the differential diagnosis included a Hurthle cell neoplasm, benign iatrogenic squamous cells (secondary to trans-esophageal aspiration approach), and granular cell tumor. Metastatic adenocarcinoma, although unlikely, should also be included in the differential diagnosis. The conspicuous lack of nucleoli made a Hurthle cell neoplasm less likely. Furthermore, immunoperoxidase stains for thyroglobulin and TTF-1 were negative, excluding the possibility of a thyroid neoplasm. Iatrogenic squamous cells must be included in the differential diagnosis since the approach of the needle was trans-esophageal. The aspirated cells had distinct polygonal cell borders, a feature of squamous cells. Upon evaluation of the cells at a higher magnification, cytoplasmic granules rather than dense, waxy cytoplasm were apparent, thus, excluding the possibility of squamous cells. Lack of cytoplasmic mucin, columnar cell shape, and nucleocytoplasmic polarity excluded the possibility of a metastatic adenocarcinoma. An S-100 immunoperoxidase stain demonstrated diffuse positive staining of the cytoplasmic granules, consistent with a granular cell tumor. The characteristic morphologic findings in addition to the immunohistochemical staining pattern are diagnostic of a granular cell tumor [16]. The diagnosis was corroborated by findings on electron microscopy and on the surgical resection specimen.
This is the first case report of a preoperative diagnosis of a mediastinal granular cell tumor by EUS-FNA. While granular cell tumors are rare tumors, it is important to include it in the differential diagnosis of mediastinal lesions. The preoperative diagnosis of a granular cell tumor by EUS-FNA will tremendously benefit the patient by preventing further tissue sampling procedures and possibly even surgical resection.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
All authors made substantial contributions to the intellectual content and/or presentation of the manuscript. SMB (cytopathology fellow) is the first author, and she wrote the intial version of the manuscript under the guidance of DJ (cytopathologist) who is a senior author. DJ and IE (cytopathologists), diagnosed the cytopathological aspects of the case. ME is the endoscopist who performed the EUS-FNA and revised the manuscript. RC is the surgeon who resected the mass and revised the manuscript.
Acknowledgements
Co-editors of CytoJournal Vinod B. Shidham, MD, FRCPath, FIAC and Barbara F. Atkinson, MD thank: the academic editor- Jan F. Silverman, MD Chair & Professor, Pathology & Lab Medicine, Allegheny Gen Hosp, 320 East North Ave. Pittsburgh, PA 15212 (Email: [email protected]) for organizing and completing the peer-review of this manuscript.
Due to the archival nature of the case as well as the absence of any potentially identifying patient information, the Institutional Review Board at the University of Alabama at Birmingham granted a waiver of patient authorization for this case report.
==== Refs
Abrikossoff AI Uber Myome, ausgehend von der quergestneiften willkurlichen Muskulatur Virchows Arch Pathol 1926 260 215 23
Ordonez NG Mackay B Granular Cell Tumor: A Review of the Pathology and Histogenesis Ultrastruct Pathol 1999 23 207 222 10503740 10.1080/019131299281545
Johnston MJ Helwig EB Granular cell tumors of the gastrointestinal tract and perianal region: A study of 74 cases Dig Dis Sci 1981 26 807 816 6169495 10.1007/BF01309613
McSwain GR Colpitts R Kreutner A O'Brien PH Spicer S Granular cell myoblastoma Surg Gynel Obst 1980 150 703 10
Lack EE Worsham GF Callihan MD Crawford BE Klappenback S Rowden G Chun B Granular cell tumor: A clinicopathologic study of 110 patients J Surg Oncol 1980 13 301 16 6246310
Smith AR Gilbert CF Strausbauch P Silverman JF Fine Needle Aspiration Cytology of a Mediastinal Granular Cell Tumor with Histologic Confirmation and Ancillary Studies Acta Cytol 1998 42 1011 1016 9684595
Rosenbloom PM Barrows GH Kmetz DR Canty TG Granular Cell Myoblastoma Arising From the Thoracic Sympathetic Nerve Chain J Pediatr Surg 1975 10 819 822 171368 10.1016/0022-3468(75)90391-7
Harrer WV Patchefsky AS Malignant Granular-cell Myoblastoma of the Posterior Mediastinum Chest 1972 61 95 96 5049514
Aisner SC Chakravarthy AK Joslyn JN Coughlin TR Bilateral Granular Cell Tumors of the Posterior Mediastinum Ann Thorac Surg 1988 46 688 689 2848464
Robinson JM Knoll R Henry DA Intrathoracic Granular Cell Myoblastoma South Med J 1988 81 1453 1457 2847328
Machida E Haniuda M Eguchi T Kurai M Yamanda T Amano J Ota H Granular Cell Tumor of the Mediastinum Intern Med 2003 42 178 181 12636238
Arluk GM Coyle WJ EUS and fine-needle aspiration in the evaluation of mediastinal masses superior to the aortic arch Gastrointest Endosc 2001 53 793 797 11375594 10.1067/mge.2001.115338
Prematilleke IV Piris J Shah KA Fine needle aspiration cytology of a granular cell tumor of the esophagus Cytopathology 2004 15 119 23 15056174 10.1111/j.1365-2303.2004.00126.x
Akatsu T Kobayashi H Uematsu S Tamagawa E Shinozak H Kase K Kobayashi K Otsuka S Mukai M Kitajima M Granular cell tumor of the breast preoperatively diagnosed by fine-needle aspiration cytology: Report of a case Surg Today 2004 34 760 763 15338349 10.1007/s00595-004-2784-7
Wieczorek TJ Krane JF Domanski HA Akerman M Carlen B Misdraji J Granter SR Cytologic findings in granular cell tumors with emphasis on the diagnosis of malignant granular cell tumor by fine-needle aspiration Cancer Cytopathol 2001 93 398 408
Liu K Madden JF Olatidoye BA Dodd LG Features of benign granular cell tumor on fine needle aspiration Acta Cytol 1999 43 552 557 10432874
|
15943858
|
PMC1184093
|
CC BY
|
2021-01-04 16:36:22
|
no
|
Cytojournal. 2005 Jun 8; 2:8
|
utf-8
|
Cytojournal
| 2,005 |
10.1186/1742-6413-2-8
|
oa_comm
|
==== Front
Health Qual Life OutcomesHealth and Quality of Life Outcomes1477-7525BioMed Central London 1477-7525-3-411601880110.1186/1477-7525-3-41ResearchAn investigation into the psychometric properties of the Hospital Anxiety and Depression Scale in patients with breast cancer Rodgers Jacqui [email protected] Colin R [email protected] Rachel C [email protected] Kate [email protected] Mark [email protected] School of Neurology, Neurobiology and Psychiatry, Faculty of Medical Sciences, University of Newcastle, Newcastle upon Tyne, Tyne and Wear, NE17RU, UK2 The Nethersole School of Nursing, Faculty of Medicine, The Chinese University of Hong Kong, Esther Lee Building, Chung Chi College, Shatin, New Territories, Hong Kong, China3 Northern Centre for Cancer Treatment, Newcastle General Hospital, Newcastle upon Tyne, UK2005 14 7 2005 3 41 41 25 4 2005 14 7 2005 Copyright © 2005 Rodgers et al; licensee BioMed Central Ltd.2005Rodgers et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
To determine the psychometric properties of the Hospital Anxiety and Depression Scale (HADS) in patients with breast cancer and determine the suitability of the instrument for use with this clinical group.
Methods
A cross-sectional design was used. The study used a pooled data set from three breast cancer clinical groups. The dependent variables were HADS anxiety and depression sub-scale scores. Exploratory and confirmatory factor analyses were conducted on the HADS to determine its psychometric properties in 110 patients with breast cancer. Seven models were tested to determine model fit to the data.
Results
Both factor analysis methods indicated that three-factor models provided a better fit to the data compared to two-factor (anxiety and depression) models for breast cancer patients. Clark and Watson's three factor tripartite and three factor hierarchical models provided the best fit.
Conclusion
The underlying factor structure of the HADS in breast cancer patients comprises three distinct, but correlated factors, negative affectivity, autonomic anxiety and anhedonic depression. The clinical utility of the HADS in screening for anxiety and depression in breast cancer patients may be enhanced by using a modified scoring procedure based on a three-factor model of psychological distress. This proposed alternate scoring method involving regressing autonomic anxiety and anhedonic depression factors onto the third factor (negative affectivity) requires further investigation in order to establish its efficacy.
==== Body
Background
A diagnosis of breast cancer is often accompanied by a significant and profound experience of psychological distress, the most commonly presenting symptoms being those of anxiety and depression [1]. Indeed, prevalence rates of clinically relevant levels of anxiety and depression in cancer patients have been estimated to be up to 45% [2-4]. It has been observed that psychological symptoms often decrease over time, further it has also been observed in the clinical presentation of breast cancer that up to 30% of these patients will continue to experience clinically relevant levels of anxiety and depression at follow-up [5].
The role of psychological variables, particularly those of anxiety and depression in disease progression and clinical outcome has received attention from the research community. For example, Walker et al. [6] found in a study of women with advanced breast cancer that anxiety and depression, as assessed by self-report measure, were significant predictors of the patients' response to chemotherapy in terms of clinical and pathological outcomes. Importantly, Walker and colleagues [6] identified that anxiety and depression were independent predictors of outcome, and therefore recommended that psychological factors need to be assessed and evaluated within the overall context of treatment.
The predictive account of the relevance of psychological factors is further supported by the findings of other studies. Hopwood et al. [7], found that high levels of anxiety and depression were associated with higher mortality rates in cancer patients. Ratcliffe et al. [8], found that high levels of depression were associated with higher mortality rates in patients with Hodgkin's disease and non-Hodgkin's Lymphoma.
Given the relevance of anxiety and depression to clinical outcome in individuals with a diagnosis of cancer, techniques and tools that reliably and consistently measure these important psychological dimensions would be welcomed within the therapeutic assessment and monitoring battery. Indeed, the need for application of psychometrically robust affective assessment tools to the clinical oncology setting is pressing due to inadequate training of non-specialist clinicians and nurses in recognising and screening for symptoms of psychological distress [9]. This is particularly important given the possible prognostic advantages offered by effectively identifying those individuals who may be anxious and depressed following diagnosis and treatment and then targeting specific interventions at these patients to reduce psychological sequelae [6].
In summary, there is convincing clinical evidence to suggest that a psychometrically robust, accurate, easily administered and patient acceptable affective state assessment tool could be of great benefit in assessing levels of anxiety and depression in patients with cancer.
The Hospital Anxiety and Depression Scale (HADS) [10] is a widely used self-report instrument designed as a brief assessment tool of the distinct dimensions of anxiety and depression in non-psychiatric populations [11,12]. It is a 14-item questionnaire that consists of two sub-scales of seven items designed to measure levels both of anxiety and depression. The ease, speed and patient acceptability of the HADS has led to it being applied to a wide variety of clinical populations where significant anxiety and depression may co-exist with the manifestation of physical illness [6,13-21].
The HADS has also been used widely in the clinical oncology setting as a screening and research tool [22-28]. Interestingly, conclusions drawn from investigations that have explored the utility of the HADS in the clinical oncology setting have yielded contradictory findings. A number of studies have suggested that the HADS reliably measures anxiety and depression in cancer patients [23,27,28] and should be adopted as a routine clinical tool for screening for psychological distress [29-31]. However, a number of other investigations in this area have suggested that the HADS may not be a suitable instrument to assess patients with cancer [24,32]. A general criticism of the HADS in cancer screening has been issues relating to the instruments poor sensitivity (ability to detect true cases) and specificity (ability to detect true non-cases) when tested against a 'gold standard', typically, a structured clinical interview [24,32].
However, a further issue concerns the method of scoring the HADS in relation to the HADS anxiety (HADS-A) and depression (HADS-D) sub-scales. A number of oncology studies [23,26,33-35] have suggested the HADS total score (all-14 items) should be used as a global measure of 'psychological distress'. This approach is against the recommendations of the original developers of the HADS [10] and this practice is further reproached in the HADS administration manual [36]. Razavi and colleagues [26] however, based their recommendation on a psychometrically robust rationale for using the HADS total score to assess cancer patients. Based on a number of psychometric criteria, including factor analysis and sensitivity/specificity criteria this study found just one single-factor emerged, identified as a single dimension of global psychological distress. This represents a good rationale for using the HADS as a unitary measure because it suggests that, in this population, the HADS could not discriminate between anxiety and depression.
However, Razavi et al.'s [26] findings of a single-dimension of global psychological distress have not been replicated in other studies examining cancer. Moorey et al. [37] found support for the bi-dimensional (anxiety and depression) underlying structure of the HADS in cancer patients. Interestingly, Moorey [37] did find some inconsistencies in their analysis with the HADS-A item 'I can sit at ease and feel relaxed' loading onto the HADS-D sub-scale. A further study examining anxiety and depression in patients with malignant melanoma [22] found the HADS to have an underlying three-factor structure. Lloyd-Williams [24] conducted an investigation into the utility of the HADS in terminally ill cancer patients and found a four-factor underlying dimensional structure.
Interestingly, a recent international consensus statement on depression and anxiety in oncology recommended the use of the HADS for screening cancer patients [38], however the recommendation was made on the explicit basis that the HADS 'assesses anxiety and depression as 2 dimensions scored separately' [38].
The factor inconsistencies observed in the HADS are not specific to studies involving cancer patients. Psychometric anomalies in the factor structure of the HADS have been observed in a diverse variety of clinical populations including depression [39], coronary heart disease [17], chronic fatigue syndrome [21], end-stage renal disease [16] and pregnancy [14]. A recent review [11] of studies that have investigated the underlying factor structure of the HADS found that nearly half reported factor structures inconsistent with the two-dimensional anxiety and depression model proposed by Zigmond and Snaith [11]. Despite the international use of the HADS in a vast multitude of clinical populations, the lack of systematic structural evaluation of the instrument in target clinical groups has been highlighted as an important psychometric concern.
Dunbar [40], conducted a confirmatory factor analysis of the HADS in a non-clinical population and found support for the three-factor tripartite model proposed by Clark & Watson [41]. This was a theoretically important observation since Clark & Watson's [41] three-factor tripartite model represents a development of the conceptualisation of anxiety and depression within a coherent and evidenced-based model. In addition their model is based upon a theoretically rich psychological account of anxiety and depression which is consistent with clinical observations of the disorders. Interestingly a number of recent psychometric investigations of the HADS have identified a three-factor underlying structure to the HADS in clinical populations [17,39].
Importantly, a recent investigation [21] into the psychometric properties of the HADS in individuals with chronic fatigue syndrome (CFS) tested Clark & Watson's three-factor tripartite model [41] and found it to provide a significantly better fit to the data than the bi-dimensional model proposed by Zigmond & Snaith [10]. McCue's [21] study extended the observations of Dunbar et al. [40] of support for the tripartite model to a clinical population. The relevance of this is that these findings suggest that a three-factor underlying structure to the HADS may have clinical implications since this model would be predicted by a coherent theoretical development, that of Clark & Watson [41], in the understanding of anxiety and depression within a clinical context. Interestingly, a number of studies have identified a third factor in the HADS using exploratory factor analysis, the researchers having then deciding to reject the third factor as meaningless and subsequently 'forcing' a two-factor solution [42,43]. It is likely that these researchers were not expecting to find a third factor since this would be inconsistent with Zigmond & Snaith's fundamental premise of bi-dimensionality of the HADS [10] and therefore chose to ignore the third factor in favour of an anticipated two-factor solution. A more recent study [20] used exploratory factor analysis and found an initial three-factor structure to the HADS in patients with end-stage renal disease. Martin and colleagues [20] then 'forced' a two-factor solution to their data and then compared the forced solution with the initial three-factor solution.
These investigators found the three-factor initial solution to be a much superior fitting underlying factor structure to the HADS compared to the 'forced' two-factor solution. It therefore seems possible that some researchers are in many instances rejecting an 'unexpected' three-factor structure in favour of the anticipated bi-dimensional structure. This is understandable in the absence of a credible theoretical perspective that would explain the manifestation of a three-factor dimensional structure to the HADS. Nonetheless, as has been established earlier, an alternative theoretical account does exist that would, in principle, predict a three-factor underlying structure to the HADS; the tripartite model of Clark & Watson [41].
However, it is important to note, that a departure from the bi-dimensional model of anxiety and depression supporting the HADS would suggest that the use of the HADS-A and HADS-D sub-scales for screening purposes would be seriously undermined since this is the fundamental rationale for using the HADS in clinical practice [38]. Conclusions drawn from HADS-A and HADS-D sub-scales would be unreliable, since the instrument would not in reality be measuring anxiety and depression and therefore clinical decision-making based on such scores would be fundamentally flawed [14,21]. See Table 1 for a summary of the models.
Table 1 Characteristics of each factor model tested
Model No. Factors Population n Extraction method FLI1** FLI2 FLI3
Zigmond et al(1983) 2 Medical 100 None 1,3,5,7,9,11,13 2,4,6,8,10,12,14 ----
Moorey et al. (1991) 2 Cancer 568 PCA 1,3,5,9,11,13 2,4,6,7,8,10,12,14 -----
Dunbar et al (2000) 3 Non-clin 2,547+ CFA 1,5,7,11 2,4,6,7,8,10,12,14 3,9,13
Friedman et al. (2001)* 3 Depressed 2,669 PCA 1,7,11 2,4,6,8,10,12,14 3,5,9,13
Razavi et al. (1990) 1 Cancer 210 PCA All items --------- ---------
Brandberg et al. (1992) 3 Cancer 273 PCA 3,5,9,13 2,4,6,8,10,12 1,7,11,14
*The three-factors are correlated in this model. +Based on CFA of three independent samples of N = 894, 829 and 824, the total cohort in this study is 2,547.
#PCA: Principal Components Analysis; CFA: Confirmatory Factor Analysis. **FLI: Factor Loading Items. The HADS items loading on each model tested.
To date, no study has been conducted that has examined the factor structure of the HADS in cancer patients by comparing competing factor structures predicted by theoretical and evidenced-based accounts of psychological distress. There is a good rationale for pursuing this in cancer patients. Given that the HADS-A and HADS-D sub-scales have been demonstrated to have predictive outcome potential in the clinical oncology setting [6] establishing the best and most appropriate factor structure of the HADS in this group of clients may be a clinically useful way of improving the predictive capacity and reliability of the instrument [40]. The first step towards this goal is to establish the best factor structure and then undertake longitudinal research to establish the predictive value of that structure.
Most previous factor analyses of the HADS have used exploratory factor analysis (EFA) techniques, though there are a small number of recent and notable exceptions to this approach that have applied a more theoretically and clinically relevant methodology to data called confirmatory factor analysis [20,21,30,40].
This study seeks to determine the appropriateness of using the HADS as a two-dimensional instrument in women with breast cancer by examining the instrument's underlying factor structure using both EFA and CFA. The study will test the hypothesis that the HADS comprises a two-factor (anxiety and depression) underlying factor structure in women with breast cancer.
Methods
Design
The study used a cross-sectional design. To address the research questions exploratory factor analysis (EFA), confirmatory factor analysis (CFA) and reliability analysis methods were used using a pooled HADS data set from all participants. Relevant clinical details were also recorded.
Statistical analysis
Reliability analysis
A reliability analysis of the HADS total all-items, and HADS anxiety (HADS-A) and HADS depression (HADS-D) sub-scales, was conducted to ensure that the measures satisfied the criteria for clinical and research purposes using the Cronbach coefficient alpha statistical procedure [44]. A Cronbach's alpha reliability statistic of 0.70 is considered as the minimum acceptable criterion of instrument internal reliability [45,46].
Exploratory factor analysis
Exploratory factor analysis was performed on the full 14-item HADS. The criterion chosen to determine that an extracted factor accounted for a reasonably large proportion of the total variance was based on an eigenvalue greater than 1. A maximum likelihood factor extraction procedure was chosen on the basis that this approach is particularly useful in extracting psychologically meaningful factors [17,14,47]. A further advantage of using the maximum likelihood approach is that a chi-square statistic can be generated to determine if the number of extracted factors offers a statistically good fit to the model tested. An oblimin non-orthogonal factor rotation procedure was chosen [47] due to the possibility that extracted factors may be correlated. The arbitrary determination of a significant item factor loading was set at a coefficient level of 0.30 or greater, this level based on a rationale of maximising the possible number of items loading on emerging factors in order to generate a more complete psychological interpretation of the data set, this being a level consistent with investigators who have utilised exploratory factor analysis [14,17,48].
Confirmatory factor analysis
Confirmatory factor analysis was conducted using the Analysis of Moment Structures (AMOS) version 4 statistical software package [49]. Seven models derived from clinical and empirical research were tested. These were Zigmond & Snaith's original two-factor model [10], Moorey et al.'s two-factor model [37], Razavi et al.'s single-factor model [26], Clark and Watson's three-factor tripartite model [41], Clark and Watson's three-factor hierarchical tripartite model [41] Friedman et al.'s three-factor correlated model [39] and Brandberg et al.'s three-factor correlated model [22].
For all models, independence of error terms was specified and the maximum likelihood method of estimation was used. Factors were allowed to be correlated where this was consistent with the particular factor model being tested. Multiple goodness of fit tests [50] were used to evaluate the seven models, these being the Comparative Fit Index (CFI) [51], the Akaike Information Criterion (AIC) [52], the Consistent Akaike Information Criterion (CAIC) [53] and the Root Mean Squared Error of Approximation (RMSEA). A CFI greater than 0.90 indicates a good fit to the data [54]. A RMSEA with values of less than 0.08 indicates a good fit to the data, while values greater than 0.10 suggest strongly that the model fit is unsatisfactory. The AIC and CAIC are useful fit indices for allowing comparison between models [40]. The Chi-square goodness of fit test was also used to allow models to be compared and to determine the acceptability of model fit. A statistically significant χ2 indicates a proportion of the variance in the model remains unexplained by the model tested [50].
Comparison with normative data
Comparison with the most contemporary normative HADS data in breast cancer patients [55] was conducted using the one-sample t-test.
Procedure
An information sheet and consent form was posted to patients approximately three weeks prior to their routine clinic follow-up. Participants were either seen at home or at clinic by one of the researchers (RM) and completed a pack of questionnaires including the HADS. Participants also completed a short neurocognitive test battery. The study took 45 minutes to complete.
Participants
110 women who had undergone adjuvant treatment for breast cancer, and were at least 6 months post-chemotherapy, participated in the study. Patients with a history of major psychiatric illness were excluded. Women were recruited from three treatment groups: chemotherapy alone, hormonal therapy alone, and a combination of chemotherapy and hormonal therapy.
Socio-demographic and treatment characteristics of the participant groups are presented in Table 2. A significant group effect of age was observed, F(2,107) = 3.09, p = 0.05, with women in the hormone therapy alone group being significantly older than women in the chemotherapy alone group (Bonferroni post-hoc test, p = 0.04). No other statistically significant differences were observed between groups, all further group comparisons of socio-demographic, baseline treatment data and HADS-A and HADS-D scores being conducted using analysis of co-variance (ANCOVA) controlling for age.
Table 2 Demographic and clinical data mean scores/levels with standard deviations in parentheses and accompanying F and p values of group comparisons.
Group type
Variable Chemotherapy alone Chemotherapy and hormone Hormone alone F p
HADS-A 7.33 (3.99) 6.48 (3.96) 7.35 (4.90) 0.08* 0.92
HADS-D 3.11 (3.73) 2.90 (2.30) 4.35 (3.35) 1.95* 0.15
Length of time since treatment ended 2.49 (2.09) 2.53 (1.56) 1.83 (1.43) 1.69* 0.19
Townsend index of deprivation -0.78 (2.76) -0.94 (3.10) 0.37 (3.46) 1.20* 0.30
Age 52.52 (8.20) 55.24 (6.86) 57.95 (9.06) 3.09# 0.05
*Analysis of co-variance (ANCOVA) controlling for age, F(2,106) degrees of freedom.
#Analysis of variance (ANOVA), F(2,107) degrees of freedom.
The data was drawn from a larger study exploring neurocognitive and behavioural outcomes following breast cancer treatment. Ethical approval was obtained from Newcastle and North Tyneside Health Authority Joint Ethics Committee. Participants were recruited through the Northern Centre for Cancer Treatment and the Royal Victoria Infirmary, Newcastle upon Tyne, UK. Written informed consent was obtained from all participants prior to the commencement of the study.
Results
The mean scores of participant's ratings on the HADS-A were 7.43 (SD 4.14) and HADS-D was 3.25 (SD 2.97). Using Snaith & Zigmond's interpretation of HADS-A and HADS-D scores of 8 or over, 51 participants (46.4%) demonstrated possible clinically relevant levels of anxiety and 11 patients (10.0%) possible clinically relevant levels of depression [10]. Adopting Snaith & Zigmond's higher threshold for sensitivity of HADS-A and HADS-D scores of 11 or over, 24 participants (21.8%) demonstrated probable clinically relevant levels of anxiety and 3 participants (2.7%) probable clinically relevant levels of depression [36].
Reliability analysis
Calculated Cronbach's alpha of the HADS (all 14 items), HADS-A and HADS-D sub-scales was 0.85, 0.79 and 0.87 respectively, exceeding Kline's criterion for acceptable instrument internal reliability [45].
Comparison with normative data
No statistically significant differences were observed between HADS-A (t(109) = 0.18, p = 0.85) and HADS-D (t(109) = 0.16, p = 0.87) mean scores of the current study compared to those of Osborne et al. [55].
Exploratory factor analysis
The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy and the Bartlett Test of Sphericity (BTS) were conducted on the data prior to factor extraction to ensure that the characteristics of the data set were suitable for the factor analysis to be conducted. KMO analysis yielded an index of 0.86, and in concert with a highly significant BTS, χ2(df = 91) = 635.36, p < 0.001, confirmed that the data distribution satisfied the psychometric criteria for the factor analysis to be performed. Following factor extraction and oblimin rotation, three factors with eigenvalues greater than 1 emerged from analysis of the complete HADS and accumulatively accounted for 59.82% of the total variance. The factor loadings of the individual HADS items in relation to the three-factor solution are reproduced in Table 2.
Factor scores on each extracted factor for each participant were calculated using regression. In contrast with the Bartlett and Anderson-Rubin methods of factor score calculation, the regression method was chosen since this technique does not assume the extracted factors are orthogonal and also minimises any sum of squares discrepancies between true and estimated factors over individuals. Factor one proved to be highly statistically significantly, but negatively correlated with factor two, r = -0.48, p < 0.001. Factor one was significantly positively correlated with factor three, r = 0.45, p < 0.001. Factor two was observed to be highly statistically and negatively correlated with factor three, r = -0.63, p < 0.001. The chi-square goodness of fit test, χ2(df = 52) = 57.18, p = 0.29, was not statistically significant suggesting that the three-factor solution extracted provided a good fit to the data. A forced two-factor solution was then specified, however, the emergent factor solution failed to provide a good fit to the data, χ2(df = 64) = 85.62, p = 0.04. The forced two-factor solution accounted for only 45.08% of the total variance.
Confirmatory factor analysis
The factor models tested and accompanying fit indices are shown in Table 3. The χ2 goodness of fit analyses for all models were statistically significant (p < 0.05) indicating a proportion of the variance was unexplained by each model. Examination of the fit indices for each model revealed that the best fit to the data is Clark and Watson's [41] three-factor tripartite model, their being little difference between correlated and hierarchically correlated versions of the model (Figure 1). The second closest fit to the data was provided by Friedman et al.'s three factor model [39]. The third closest fit to the data was found to be Brandberg et al.'s [22] three-factor correlated model. Zigmond and Snaith's original two-factor model [10] offered the fourth best fit to the data, while the two-factor model of Moorey et al. [37] provided the fifth best fit. The worst fit to the data was furnished by the single factor model of Razavi et al. [26](Table 4).
Table 3 Factor loadings of HAD Scale items following maximum likelihood factor extraction with oblimin rotation
HAD Scale item Factor 1 Factor 2 Factor 3
Anxiety sub-scale
(1) I feel tense or wound up 0.17 -0.30 0.45
(3) I get a sort of frightened feeling as if something awful is about to happen 0.16 -0.80 -0.08
(5) Worrying thoughts go through my mind 0.24 -0.55 0.16
(7) I can sit at ease and feel relaxed 0.26 -0.10 0.61
(9) I get a sort of frightened feeling like 'butterflies' in the stomach -0.18 -0.79 0.04
(11) I feel restless as if I have to be on the move -0.06 0.01 0.53
(13) I get sudden feelings of panic 0.03 -0.82 0.04
Depression sub-scale
(2) I still enjoy the things I used to enjoy 0.72 -0.04 -0.02
(4) I can laugh and see the funny side of things 0.50 -0.11 0.12
(6) I feel cheerful 0.45 -0.15 0.15
(8) I feel as if I am slowed down 0.56 0.07 0.18
(10) I have lost interest in my appearance 0.35 -0.01 -0.05
(12) I look forward with enjoyment to things 0.88 -0.08 -0.11
(14) I can enjoy a good book or TV programme 0.58 0.07 0.01
*Bold indicates that item loading on a factor is 0.30 or above
Table 4 Factor structure of the HADS determined by testing the fit of models derived from factor analysis.
Model χ2 df p RMSEA CFI CAIC AIC
Zigmond and Snaith two-factor 121.77 (76) 0.001 0.07 0.92 287.08 179.77
Moorey et al. two-factor 132.16 (76) <0.001 0.08 0.90 297.47 190.16
Friedman three-factor correlated 101.79 (74) 0.018 0.06 0.95 278.50 163.79
Dunbar et al. three-factor tripartite 96.16 (73) 0.036 0.05 0.96 278.57 160.16
Dunbar et al. three-factor hierarchical tripartite 96.27 (73) 0.035 0.05 0.96 278.68 160.27
Razavi single-factor 212.22 (77) <0.001 0.13 0.77 371.83 268.22
Brandberg et al. three-factor 116.11 (74) 0.001 0.07 0.93 292.83 178.11
The best fit to the data is provided by the three-factor tripartite model and the three-factor hierarchical tripartite model of Clark & Watson (1991) based on Dunbar et al. (2000).
Figure 1 Clark & Watson's (1991) Tripartite model applied to HADS data. Note: Figures represent standardised parameter estimates.
Discussion
This study has yielded interesting and clinically pertinent observations regarding the HADS in relation to psychological screening in women with breast cancer. The finding of relatively high levels of anxiety (mean = 7.43) and low levels of depression (mean = 3.25) is entirely consistent with the most recent investigation reporting HADS normative anxiety (mean = 7.50) and depression (mean = 3.30) data in a relatively large (N = 731) population of women with breast cancer [55]. This finding is suggestive that the HADS-A and HADS-D sub-scales appear to be pathology specific and sensitive.
Estimations of internal reliability revealed Cronbach's alpha's of the HADS (all items) and the HADS-A and HADS-D sub-scales to be all statistically acceptable, indeed, these observations being entirely consistent with previous research into the psychometric properties of this instrument (Bjelland et al., 2002). The HADS-A and HADS-D sub-scales were found to be positively and statistically significantly correlated, an observation that is again consistent with previous research [11]. Taken together, the consistency of HADS-A and HADS-D scores between this study and normative breast cancer HADS scores, the good internal reliability of HADS-A and HADS-D sub-scales and confirmation of the anticipated significant positive correlation between HADS-A and HADS-D sub-scales suggests that the HADS has achieved a number of the psychometric credentials required to confer it's acceptability as a reliable and valid screening tool of anxiety and depression for use in women with breast cancer.
However, the results of the EFA and CFA add a further dimension to the debate over the psychometric integrity of this instrument in this clinical population and, indeed, provide compelling evidence that the assumed bi-dimensional anxiety and depression underlying structure of the HADS should be reviewed, particularly in patients with breast cancer.
The EFA of the HADS revealed an initial three-factor underlying structure which provided a good fit to the data. When compared to a forced two-factor solution, the initial three-factor model provided a better fit to the data, the two-factor forced solution offering a statistically poor fit to the data. This is a clinically pertinent observation since not only does this finding reveal that the HADS does not measure two distinct dimensions of anxiety and depression in this population, it informs the growing evidence base which has increasingly suggested that the HADS is not a reliable measure of anxiety and depression when used within the context of a wide range of pathology [14,20-22,26,39,56].
Examination of individual item loadings is illuminating. It was observed that the HADS-A sub-scale items 1. 'I feel tense or wound up', 7. 'I can sit at ease and feel relaxed' and 11. 'I feel restless as if I have to be on the move' loaded on extracted factor 3. This separation of HADS-A items has been observed previously in factor analysis of cancer patient data.
Brandberg et al. [22], in a study of patients with malignant melanoma (skin cancer), found a three factor structure to the HADS and identified a 'restlessness' factor comprising items 1, 7, 11 and 14. Item 14. 'I can enjoy a good book or TV programme' was not found to load on to the 'restlessness' factor reported by Brandberg and colleagues [22] in the current study, though with this exception, the loading of HADS-A items on this 'restlessness' factor is identical. Items 3. 'I get a sort of frightened feeling as if something awful is about to happen', 5. 'Worrying thoughts go through my mind', 9. 'I get a sort of frightened feeling like 'butterflies' in the stomach' and 13. 'I get sudden feelings of panic' loaded onto extracted factor three. This observation, is consistent, indeed identical, with that factor extracted and observed by Brandberg et al. (1992) and termed 'anxiety'. Item 1. 'I feel tense or wound up' was observed to also load onto factor 2, however it should be emphasised that this item loads more heavily on extracted factor 3. All the HADS-D items loaded onto factor 1, this extracted factor being consistent with the depression sub-scale these items are designed to measure. The findings from the EFA would suggest that the HADS is comprised of three underlying factors, these being depression, anxiety and restlessness.
The CFA both supports the findings of the EFA and provides further evidence to support the notion that the HADS is comprised of an underlying three-factor structure in breast cancer patients. It should be noted that, though the χ2 analysis of all models tested were statistically significant, indicating a significant proportion of the variance of the model tested to be unexplained in the data, it is readily acknowledged that trivial variations in the data can lead to significant χ2 test results [57] and therefore the usefulness of the test within the realm of CFA is that it provides an index of comparatively how well a model fits the data. The three-factor models tested proved to provide better fits to the data than the two two-factor models tested on virtually all indices of model fit. The single factor model tested revealed the poorest fit to the data of all the models.
Clark & Watson's [41] three-factor tripartite and three-factor hierarchical tripartite models [40] provided the best fit to the data, examination of the RMSEA and CFI fit tests revealing that these three-factor models satisfied the criteria for a good fit to the data. Interestingly, the participant population in Dunbar et al.'s study [40] was drawn from a non-clinical population and the basis for the study was to test a strong contemporary theoretically-based account of anxiety and depression, that of Clark & Watson [41]. The second best fit to the data (Clark & Watson's model fit being virtually identical will be treated as a single best fit model) was provided by Friedman et al.'s three-factor model [39]. Friedman et al.'s study was an EFA on HADS data from a psychiatric population, individuals being treated for depressive disorder [39]. This finding gives an indication to the possibility that the three-factor best model fit observed in the current study may not, essentially be related to the presenting pathology, since breast cancer and depressive disorder represent two distinct and aetiologically unrelated clinical presentations, therefore the superior (compared to the competing two-factor models tested) three-factor model fit of Friedman's model [39] may, in fact, be tapping into the basic fundamental factor structure of the HADS.
This observation would be supported by the findings of the model fit of Brandberg's three-factor model [22], superior to that of the two-factor models. Observation of an underlying three-factor structure to the HADS has been observed in a number of other studies investigating a broad spectrum of clinical and non-clinical populations [16,20,21]. There have also been a number of instances in studies of psychiatric disorder where a three-factor underlying structure to the HADS has been initially observed and has then been dismissed by the authors in favour of the (presumably) expected two factor solution [42,43]. Arguably and retrospectively, these studies suggest further support for a three-factor underlying structure to the HADS.
Brandberg et al. [22] commented that, in spite of finding support for a three-factor underlying structure to the HADS, there was not a need for a revision of the instrument, rather, it was suggested that further studies of the instrument should be conducted. It is now over ten years since Brandberg and colleagues study [22] and the accumulating evidence base concerning the factor structure of the HADS raises credible clinical issues regarding the utility of this instrument across a range of pathologies. The findings from the CFA in the current study revealed that the two-factor models tested [10,37] offered a poorer fit to the data compared to the three-factor models. However, it should be stressed that examination of the RMSEA and CFI of both these two-factor models revealed that they offered an acceptable fit to the data. This is a noteworthy observation since other studies which have found support for the three-factor model of the HADS have found evidence that two-factor models offer a very poor fit to the data [20,21]. In summary, the CFA findings from the current study support a three-factor underlying factor structure to the HADS, though poorer fitting two-factor models still provide an acceptable, albeit less so, fit to the data.
Two questions remain, firstly what is the HADS measuring within the context of a three-factor model and secondly, should the HADS be continued to be used as a bi-dimensional screening tool for the detection of individuals experiencing anxiety and depression?
The best fit to the data was provided by Clark & Watson's tripartite and hierarchical tripartite three-factor models [41], there being very little difference between the models statistically establishing that both models are measuring fundamentally the same constructs. According to Clark & Watson's [41] formulation of anxiety and depression, the three factors observed in the HADS would represent distinct dimensions of negative affectivity, autonomic anxiety and anhedonic depression. These theoretically derived models have been shown to provide a best fit to the data in two previous research investigations that have focused on both a non-clinical populations [40], and a clinical population of individuals with chronic fatigue syndrome [21]. Furthermore, Crawford et. al [58] in a study evaluating the reliability ad validity of the Positive and Negative Affect Schedule (PANAS) and its relationship with other measures of depression and anxiety including the HADS have recently provided further support for tripartite theory of anxiety and depression.
It must be acknowledged that a number of limitations will inevitably apply to the current study. It must be noted that the sample size for the study was borderline for conducting SEM with AMOS but was adequate by a number of conventional criteria. One must also take into account the suggestion that differing methodologies used across studies to undertake factor analysis may account for the differences found, see Martin [58] for a full discussion of these issues. Additionally the low mean depression scores for the sample, whilst consistent with other studies with similar populations, might result in the presence of a floor effect, thus limiting the variance within the sample. This may have resulted for the fact that in order to avoid the short term acute sequelae associated with intensive treatment all participants were at least two years from treatment at the time of the investigation.
This study has extended the observations of Dunbar et al. [40] and McCue et al. [21] to a further population with distinct pathology. It has been suggested by Dunbar [40] that by using the hierarchical tripartite model, the autonomic anxiety and anhedonic depression factors would be of greater value in discriminating between anxiety and depression than simply using HADS anxiety and depression sub-scale scores. Brandberg et al. [22] noted that the HADS-D sub-scale was the most useful for clinical purposes, though the rationale was not stated, it seems plausible to assume that this was because of the 'split' HADS-A sub-scale observed in their factor analysis. This observation is entirely consistent with that of Dunbar [40] who suggests a convincing rationale why the HADS is not a highly discriminative instrument in some populations is because the HADS-A and HADS-D sub-scale scores are contaminated by overlap between the three underlying factors. A method of significantly increasing discriminability has been suggested by Dunbar et al., [40] involving regressing autonomic anxiety and anhedonic depression factors scores on to the negative affectivity (third factor) sub-scale scores.
A further study would be required to establish the efficacy and desirability of this approach since comparison of factor derived scores would need to be compared against a gold standard such as a formal structured clinical interview schedule. Using this approach receiver operating characteristic (ROC) curves could be calculated to evaluate any relative improvement of regressed autonomic anxiety and anhedonic depression scores compared to HADS-A and HADS-D scores.
The RMSEA and CFI statistics revealed that the two-factor models tested offered acceptable fits to the data, with Zigmond & Snaith's original two-factor formulation [10] offering a slightly better fit to the data to that of Moorey et al.'s modified two-factor model [37]. It is worthy of comment that Zigmond & Snaith's [10] model was superior to that of Moorey et al.'s model [37], in spite of the latter researchers using a clinical cohort comprised exclusively of cancer patients. This observation offers further support to conclusions drawn from the three-factor models tested that the underlying factor structure of the HADS is relatively stable and the impact of pathology on the factor structure of the instrument may be relatively minor. One of the central tenet for supporting using the HADS is that it is easy to use, this of course, includes scoring the instrument. Whether, any significant benefits in discriminability that may be identified in using the regressed scores as suggested by Dunbar et al. [40] may be off-set by an increase in sophistication in terms of calculating regressed factor scores in clinical practice.
Obviously, this is an area for future investigation, however it is worth noting that a wide variety of health professionals use the HADS in clinical practice on an everyday basis and it is these individuals who may feel reluctant or lack the time to calculate regressed scores for the HADS unless there is a large improvement to be found in the instruments accuracy by doing so. A simple scoring algorithm would be a fundamental requirement if the approach suggested by Dunbar and colleagues [40] was to move from the arena of academic and clinical research into the natural environment for the HADS, everyday clinical practice.
On balance, and incorporating the above limitations of ensuring that the HADS remains an easy to use clinical screening instrument, it is suggested that HADS remains a useful screening instrument in the clinical oncology environment and may be scored and interpreted in the recommended manner [10,36]. However, further clinical research work is recommended in this area to determine if scoring the instrument as a three-factor measure offers any worthwhile benefits in case detection that may offset a more complicated scoring procedure. No evidence at all was forthcoming to suggest that the HADS should be used as a one-dimensional model of global psychological distress, the single factor model providing a very poor fit to the data. Based on this observation it is suggested that a total HADS score should not be used in this clinical context.
Conclusion
In conclusion, a compromise is suggested based on the clinical research observations of the current study and the clinical context of everyday professional practice where the HADS is used as a screening instrument of choice. The HADS was found to have an underlying three-factor structure in breast cancer patients. The possibility that improved accuracy in case detection may be found by using a three factor model to score the HADS is balanced by a potential decrease in the ease of use of the instrument because a more complex scoring system will be required.
This issue can be settled by future research in this area to determine the magnitude of any worthwhile clinical gains in scoring the HADS as a three-factor instrument. Currently however, it is suggested that the HADS can be continued to be used and scored in the traditional way, since the two-factor models tested still provided an acceptable fit to the data. However, it is recommended that for screening purposes with breast cancer patients, verification of borderline level scores should be established by a structured diagnostic clinical interview. Those using the HADS in clinical practice may also wish to consider using further measures of negative affectivity and autonomic anxiety, since these are currently poorly represented in the HADS. The possibility that the HADS, or a derivative of the HADS, may be more usefully developed as a three-dimensional rather than bi-dimensional tool consistent with advances in psychological models of anxiety and depression [41] should not be ruled out.
Authors' contributions
JR conceived of the study, participated in the design of the study, assisted in the analysis of the data and drafting of the manuscript. CM participated in the design of the study and performed the statistical analysis and drafted the manuscript. RM collected data, and contributed to the statistical analysis and interpretation of the results and the drafting of the manuscript. KK – participated in the conception and design of the study and the drafting of the manuscript. MV – provided access to participants, aided with the design of the study and participated in drafting the manuscript. All authors read and approved the final manuscript
Acknowledgements
We would like to thank all of the individuals who participated in the study. We would also like to acknowledge the support of Newcastle Hospitals Special Trustees.
==== Refs
Glanz K Lerman C Psychosocial impact of breast cancer: a critical review Behav Med 1992 14 204 212
Grassi L Indelli M Marzola M Maestri A Santini A Piva E Boccalon M Depressive symptoms and quality of life in home-care-assisted cancer patients J Pain and Symptom Manage 1996 12 300 307 8942125 10.1016/S0885-3924(96)00181-9
McDaniel JS Musselman DL Porter MR Reed DA Nemeroff CB Depression in patients with cancer Arch Gen Psychiat 1995 52 89 99 7848055
Minagawa H Uchitomi Y Yamawaki S Ishitani K Psychiatric morbidity in terminally ill cancer patients. A prospective study Psycho-oncol 1997 7 112 120
Howard R Harvey P A longitudinal study of psychological distress in women with breast symptoms J Health Psychol 1998 3 215 226
Walker LG Heys SD Walker MB Ogston K Miller ID Hutcheon AW Sarkar TK Ah-See AK Eremin O Psychological factors can predict the response to primary chemotherapy in patients with locally advanced breast cancer Euro J Cancer 1999 35 1783 1788 10.1016/S0959-8049(99)00169-0
Hopwood P Howell A Maguire P Screening for psychiatric morbidity in patients with advanced breast cancer: validation of two self-report questionnaires Brit Jour Cancer 1991 64 353 356 1892763
Ratcliffe MA Dawson AA Walker LG Eysenck Personality Inventory L-scores in patients with Hodgkin's disease and non-hodgkin's lymphoma Psycho-oncol 1995 4 39 45
Maguire G Granviolle-Grossman Improving recognition and treatment of affective disorders in cancer patients Recent Advances in Clinical Psychology 1992 Edinburgh, Churchill Livingstone
Zigmond AS Snaith RP The hospital anxiety and depression scale Acta Psychiatrica Scandinavica 1983 67 361 370 6880820
Bjelland I Dahl AA Haug TT Neckelmann D The validity of the Hospital Anxiety and Depression Scale. An updated literature review J Psychosomatic Res 2002 52 69 77 10.1016/S0022-3999(01)00296-3
Herrmann C International experiences with the Hospital Anxiety and Depression Scale – a review of validation data and clinical results J Psychosomatic Res 1997 42 17 41 10.1016/S0022-3999(96)00216-4
Barczak P Kane N Andrews S Congdon AM Clay JC Betts T Patterns of psychiatric morbidity in a genito-urinary clinic. A validation of the Hospital Anxiety Depression scale (HADS) Brit J Psychiat 1988 152 698 700 3167448
Karimova G Martin CR A psychometric evaluation of the Hospital Anxiety and Depression Scale during pregnancy Psychology, Health and Medicine 2002 8 89 103
Lewin RJ Thompson DR Martin CR Stuckey N Devlen J Michaelson S Maguire P Validation of the Cardiovascular Limitations and Symptoms Profile (CLASP) in chronic stable angina J Cardiopulmonary Rehab 2002 22 184 191 10.1097/00008483-200205000-00010
Martin CR Thompson DR Prediction of quality of life in patients with end-stage renal disease Brit J Health Psychol 2000 5 41 55 10.1348/135910700168757
Martin CR Thompson DR A psychometric evaluation of the Hospital Anxiety and Depression Scale in coronary care patients following acute myocardial infarction Psychology, Health and Medicine 2000 5 193 201
Martin CR Bowman GS Thompson DR The effect of a coordinator on cardiac rehabilitation in a district general hospital Coronary Health Care 2000 4 17 21 10.1054/chec.1999.0061
Martin CR Thompson DR The Hospital Anxiety and Depression Scale in patients undergoing peritoneal dialysis: internal and test re-test reliability Clinical Effectiveness in Nursing 2002 6 77 79 10.1054/cein.2002.0252
Martin CR Tweed AE Metcalfe MS A psychometric evaluation of the Hospital Anxiety and Depression Scale in patients diagnosed with end-stage renal disease Brit J Clin Psychol 2004 43 51 64 15005906 10.1348/014466504772812968
McCue P Martin C Buchanan T Rodgers J Scholey A An investigation into the psychometric properties of the Hospital Anxiety and Depression Scale in individuals with Chronic Fatigue Syndrome Psychology Health and Medicine 2003 8 427 441
Brandberg Y Bolund C Sigudardottir V Sjoden PO Sullivan M Anxiety and depressive symptoms at different stages of malignant melanoma Psycho-oncol 1992 1 71 78
Ibbotson T Maguire P Selby P Priestman T Wallace L Screening for anxiety and depression in cancer patients: the effects of disease and treatment EJ Cancer 1994 30A 37 40
Lloyd-Williams M Friedman T Rudd N An analysis of the validity of the Hospital Anxiety and Depression scale as a screening tool in patients with advanced metastatic cancer J Pain Symptom Manage 2001 22 990 996 11738161 10.1016/S0885-3924(01)00358-X
Nordin K Berglund G Glimelius B Sjoden PO Predicting anxiety and depression among cancer patients: a clinical model Euro J Cancer 2001 37 376 384 10.1016/S0959-8049(00)00398-1
Razavi D Delvaux N Farvacques C Robaye E Screening for adjustment disorders and major depressive disorders in cancer in-patients Brit J Psychia 1990 156 79 83
Carroll BT Kathol RG Noyes R JrWald TG Clamon GH Screening for depression and anxiety in cancer patients using the Hospital Anxiety and Depression Scale Gen Hosp Psychiat 1993 15 69 74 10.1016/0163-8343(93)90099-A
Hopwood P Stephens RJ Depression in patients with lung cancer: prevalence and risk factors derived from quality-of-life data J Clin Oncol 2000 18 893 903 10673533
Fossa SD Dahl AA Short Form 36 and Hospital Anxiety and Depression Scale. A comparison based on patients with testicular cancer J Psychosomatic Res 2002 52 79 87 10.1016/S0022-3999(01)00308-7
Johnston M Pollard B Hennessey P Construct validation of the hospital anxiety and depression scale with clinical populations J Psychosomatic Res 2000 48 579 584 10.1016/S0022-3999(00)00102-1
Walker LG Walker MB Heys SD Lolley J Wesnes K Eremin O The psychological and psychiatric effects of rIL-2 therapy: a controlled clinical trial Psycho-oncol 1999 6 290 301 10.1002/(SICI)1099-1611(199712)6:4<290::AID-PON283>3.3.CO;2-7
Hall A A'Hern R Fallowfield L Are we using appropriate self-report questionnaires for detecting anxiety and depression in women with early breast cancer? Euro J Cancer 1999 35 79 85 10.1016/S0959-8049(98)00308-6
Chaturvedi SK Clinical irrelevance of HADS factor structure Brit J Psychiat 1991 159 298 1812880
Lewis G Wessley S Comparison of the General Health Questionnaire and the Hospital Anxiety and Depression Scale Brit J Psychiat 1990 157 860 864 2289095
Ramirez AJ Richards MA Jarrett SR Fentiman IS Can mood disorder in women with breast cancer be identified preoperatively? Brit J Cancer 1995 72 1509 1512 8519668
Snaith RP Zigmond AS The Hospital Anxiety and Depression Scale Manual NFER:Nelson, Windsor 1994
Moorey S Greer S Watson M Gorman C Rowden L Tunmore R Robertson B Bliss J The factor structure and factor stability of the hospital anxiety and depression scale in patients with cancer Brit J Psychiat 1991 158 255 259 1812841
Ballenger JC Davidson JRT Lecrubier Y Nutt DJ Jones RD Berard RMF Consensus statement on depression, anxiety and oncology J Clin Psychol 2001 62 64 67
Friedman S Samuelian JC Lancrenon S Even C Chiarelli P Three-dimensional structure of the Hospital Anxiety and Depression Scale in a large French primary care population suffering from major depression Psychiat Res 2001 104 247 257 10.1016/S0165-1781(01)00309-2
Dunbar M Ford G Hunt K Der G A confirmatory factor analysis of the Hospital Anxiety and Depression scale: comparing empirically and theoretically derived structures Brit J Clin Psychol 2000 39 79 94 10789030 10.1348/014466500163121
Clark L A Watson D Tripartite model of anxiety and depression: Psychometric evidence and taxonomic implications J Abnormal Psychol 1991 100 316 336 10.1037//0021-843X.100.3.316
Dagnan D Chadwick P Trower P Psychometric properties of the Hospital Anxiety and Depression Scale with a population of members of a depression self-help group Brit J Med Psychol 2000 73 129 137 10759056 10.1348/000711200160255
Mykletun A Stordal E Dahl AA Hospital Anxiety and Depression (HAD) scale: factor structure, item analyses and internal consistency in a large population Brit J Psychiat 2001 179 540 544 11731359 10.1192/bjp.179.6.540
Cronbach LJ Coefficient alpha and the internal structure of tests Psychometrika 1951 16 297 334
Kline P The Handbook of Psychological Testing 1993 Routledge, London
Kline P A Psychometrics Primer 2000 Free Association Books London
West R Computing for Psychologists 1991 Harwood Academic Char
Meadows K Steen N McColl E Eccles M Shiels C Hewison J Hutchinson A The Diabetes Health Profile (DHP): a new instrument for assessing the psychosocial profile of insulin requiring patients – development and psychometric evaluation Qual Life Res 1996 5 242 254 8998493 10.1007/BF00434746
Arbuckle JL Wothke W AMOS 40 Users Guide, Smallwaters, Chicago 1999
Bentler PM Bonett DG Significance tests and goodness of fit in the analysis of covariance structures Psychol Bulletin 1980 88 588 606 10.1037//0033-2909.88.3.588
Bentler PM Comparative fit indexes in structural models Psychol Bulletin 1988 107 238 246
Akaike H Factor analysis and the AIC Psychometrika 1987 52 317 332
Bozdogan H Model selection and Akaike's information criterion (AIC): The general theory and its analytical extensions Psychometrika 1987 52 345 370
Kline RB Principles and Practice of Structural Equation Modeling, Guilford, New York 1998
Osborne RH Elsworth GR Hopper JL Age-specific norms and determinants of anxiety and depression in 731 women with breast cancer recruited through a population-based cancer registry Euro J Cancer 2003 39 755 762 10.1016/S0959-8049(02)00814-6
Martin CR Thompson DR Utility of the Hospital Anxiety and Depression Scale in patients with end-stage renal disease on continuous ambulatory peritoneal dialysis Psychology, Health and Medicine 1999 4 369 376 10.1080/135485099106117
Hu LT Bentler PM Hoyle RH Evaluating model fit Structural Equation Modelling: Concepts, Issues and Applications 1995 Thousand Oaks, CA: Sage
Crawford CR Henry JM The Positive and Negative Affect Schedule (PANAS): Construct validity, measurement properties and normative data in a large non-clinical sample Br J Clin Psychol 2005 43 245 265 15333231 10.1348/0144665031752934
Martin CR What does the Hospital Anxiety and Depression Scale (HADS) really measure in liaison psychiatry settings? Curr Psychiat Reviews 2005 1 69 73 10.2174/1573400052953510
|
16018801
|
PMC1184094
|
CC BY
|
2021-01-04 16:38:13
|
no
|
Health Qual Life Outcomes. 2005 Jul 14; 3:41
|
utf-8
|
Health Qual Life Outcomes
| 2,005 |
10.1186/1477-7525-3-41
|
oa_comm
|
==== Front
Immun AgeingImmunity & ageing : I & A1742-4933BioMed Central London 1742-4933-2-101598242010.1186/1742-4933-2-10ResearchImmune response to pneumococcal polysaccharides 4 and 14 in elderly and young adults. I Antibody concentrations, avidity and functional activity Kolibab Kris [email protected] S Louise [email protected] Anne K [email protected] Sadik [email protected] Sandra [email protected] George M [email protected] MA Julie [email protected] Department of Medicine, Medical College of Ohio, Toledo, OH, USA2 Respiratory Diseases Branch, Division of Bacterial and Mycotic Diseases, National Centers for Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, GA 30333, USA2005 27 6 2005 2 10 10 20 4 2005 27 6 2005 Copyright © 2005 Kolibab et al; licensee BioMed Central Ltd.2005Kolibab et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Streptococcus pneumoniae is a serious worldwide pathogen and the focus of numerous vaccine development projects. Currently the most widely accepted surrogate marker for evaluating the efficacy of a given vaccine is to utilize ELISA. Measurement of antibody concentration by ELISA without reduction in cross-reactive antibodies causes an overestimation of antibody concentration and therefore protection, this is most notable in the aged, an at risk group for this infection. We compared the immune response to the pneumococcal polysaccharides (PPS) 4 and 14 of 20 young to 20 elderly adults. Pre-and post-vaccination IgG antibody concentrations and antibody avidity against PPS4 and PPS14 were measured using two different enzyme-linked immunosorbant assay (ELISA) absorption protocols. All sera were pre-absorbed with either cell-wall polysaccharide (CPS), or CPS and serotype 22F polysaccharide.
Pre- and post-vaccination IgG antibody concentrations for serotype 4, but not 14, were significantly lowered with the additional absorption with serotype 22F polysaccharide in both age groups. Young and elderly demonstrated a significant increase from pre- to post-immunization antibody concentration, using either absorption method; and opsonophagocytic antibody titers in response to both PPS4 and PPS14. The correlation coefficients between ELISA and opsonophagocytic assays were improved by additional absorption with serotype 22F in response to serotype 4, but not serotype 14 in all age groups. Opsonophagocytic antibody titers in a sub-group of elderly (>77 years of age) were significantly lower than the opsonophagocytic antibody concentrations in young adults.
These results suggest the importance of eliminating cross-reactive antibodies from ELISA measurements by absorption of serum and an age-related impairment in the antibody response to pneumococcal polysaccharides.
==== Body
Background
Streptococcus pneumoniae is a major human pathogen responsible for serious infections in all age groups worldwide. Pneumococcal infections are estimated to cause around 40,000 deaths per year [1], and are a leading causes of death in the United States [2]. The highest incidence of pneumococcal infections occurs at the extremes of age, in the very young and in the elderly. The mortality rate of invasive pneumoccoal disease increases with age and is estimated to be 20% in those 65 years of age and 40% in those 85 years of age or older. Moreover, S. pneumoniae, is the most common cause of community acquired pneumonia among the elderly [3] and the fifth leading cause of death in those 65 years and older [4,5].
Pneumococcal capsular polysaccharide (PPS) is a major virulence factor of S. pneumoniae and reduces phagocytosis in the host [6,7]. The currently licensed pneumococcal vaccine consists of 23 purified PPS serotypes, which account for 76–90% of IPD causing strains [8,9]. The capsular PPS vaccine is based on the observation that antibodies against the capsule protect against disease by enhancing complement dependent phagocytosis [6]. Pneumococcal vaccination is recommended for all persons 65 years and older in the United States, however, efficacy studies indicate decreased protection among this target population despite protective levels of anti-pneumococcal antibodies [10,11]. The underlying causes of the reduced vaccine efficacy merit investigation.
Reports of the diminished vaccine efficacy in the elderly may reflect either poor functionality of vaccine induced anti-PPS specific antibodies or inconsistent antibody measurements. Protection is measured by in vitro analysis of opsonophagocytosis by anti-polysaccharide antibodies and serves as a surrogate marker of clinical protection [12]. However, the opsonophagocytic assay is labor intensive and difficult to perform and therefore impractical, when analyzing large numbers of samples. Therefore, alternate methods for measuring pneumococcal antibody concentrations, namely ELISA, have been developed. More practical than opsonophagocytic assays, the polysaccharide-specific antibody concentrations measured by ELISA do not necessarily correlate with functional antibody activity. The development of an ELISA method that accurately reflects functional antibody activity and can serve as a surrogate marker of protection is of utmost importance.
The enzyme-linked immunosorbant assay (ELISA) method used to measure PPS-specific antibody concentrations has changed considerably over the past 20 years. The presence of cell wall polysaccharide (CPS), covalently linked to the type-specific polysaccharide, led to poor correlations between antibody concentration and pneumococcal vaccine efficacy [13,14]. The realization that the measurement of anti-CPS antibodies caused an overestimation of anti-PPS antibody concentration led to a modification of the pneumococcal ELISA procedure to include an absorption step, removing cell wall polysaccharide directed antibodies [13,15]. Recent studies have demonstrated that the additional absorption of sera with CPS and serotype 22F PPS significantly increases the correlation between antibody concentration and opsonophagocytic activity compared to CPS absorption alone [16].
Previously, a study performed in elderly adults demonstrated a significant decrease in opsonophagocytic activity in response to PPS4 despite normal antibody concentrations [17]. The decrease in functional antibody activity increased with age and was serotype specific. The immune response to PPS14, as determined by antibody concentration and opsonophagocytic antibody activity, was well conserved in the elderly. However, in this study, an older version of the PPS-ELISA relying solely on CPS absorption was used to determine antibody concentration. Therefore, non-specific cross-reactive antibodies may have been responsible for the discrepancy between antibody concentration and opsonophagocytic activity to PPS4.
The aim of this study was to re-assess the response of elderly (>65 years old) adults to PPS4 and PPS14 employing the recently modified PPS-specific ELISA. We have undertaken the present study to further elucidate the nature of the elderly immune response to PPS in an effort to define correlates of protection in the ageing population. Should the elderly response contain a significant proportion of low avidity, cross-reactive antibodies as previously suggested [17] one would expect that absorption with PPS 22F would have a greater impact on antibody concentration and avidity measurements in the elderly than young.
Results
All volunteers were considered immune competent based on health history, complete blood cell counts, total immunoglobulin levels and quantitative B-cell and T-cell subset studies. There was no quantitative difference in number of B-cells or CD4+ and CD8+ T cells in the young versus elderly participants (Table 1).
Table 1 Measurement of T and B cell parameters in young and elderly volunteers.
CD4+ T cells GMN (95% CI) CD8+ T cells GMN (95% CI) B cells GMN (95% CI) Total IgG GMC (95% CI)
Normal Range 430 – 1185 180 – 865 25 – 285 591–1540
Young 879 (758–1000) 573 (487–659) 235 (156–314) 838 (757–919)
Elderly 817 (692–942) 396 (303–489) 164 (93–235) 993 (956–1030)
GMC = geometric mean concentration (ug/ml), GMN geometric mean number per ml. There was no significant difference (P>0.05) between young and elderly for all values.
Anti-pneumococcal Antibody Concentrations
Anti-PPS4 and PPS14 antibody concentrations were determined by the two ELISA methods, absorption with either CPS or CPS + PPS22F. Geometric mean IgG antibody concentrations (μg/ml) and 95% CI are shown in Table 2.
Table 2 Specific anti-PPS IgG antibody levels in young and elderly.
Pre-immune serum GMC (95% CI)
CPS CPS + 22F P value
PPS4 Young 1.5 (0.59–2.41) 1.22 (0.77–1.66) 0.033
Elderly 1.94 (0.85–3.03) 1.39 (0.97–1.81) 0.008
PPS14 Young 3.86 (2.54–5.18) 4.02 (2.10–5.94) 0.370
Elderly 4.77 (1.61–7.93) 4.73 (2.42–7.04) 0.367
Post-immune serum GMC (95% CI)
CPS CPS + 22F P value
PPS4 Young 4.03 (1.68–6.38) 3.45 (1.62–5.28) 0.036
Elderly 3.95 (0.73–7.17) 3.50 (1.01–5.99) 0.142
PPS14 Young 18.41 (7.24–29.58) 18.43 (8.46–28.40) 0.708
Elderly 11.43 (4.44–18.42) 10.22 (3.35–17.09) 0.013
Geometric mean concentration (GMC, μg/ml) of specific anti-PPS IgG antibody in pre-and post-immunization serum from young and elderly volunteers by two ELISA methods. CPS – serum adsorbed with CPS prior to ELISA. CPS+22F – serum absorbed with PPS22F and CPS prior to ELISA. P values shown in bold indicate significant difference between absorption methods by two tailed paired Student's T-test.
In response to PPS4 a significant increase in pre-to post-immunization antibody concentration was detected by both ELISA methods in both age groups (p range 0.003 – 0.005). When both ELISA methods were compared a significant reduction in antibody concentration was observed following CPS + PPS22F absorption in young volunteers' pre-and post-immune sera (18.6 %, p = 0.033 and 14%, p = 0.036 respectively) and elderly volunteers' pre-immune (28.4, p = 0.008) but not post-immune sera (11.4%, p = 0.142). There was no significant difference in the antibody levels detected between young and elderly volunteers in pre-or post-immunization sera using either absorption method, although young pre-immune sera tended to demonstrate lower geometric mean concentrations (GMC). However, more young than elderly demonstrated a two-fold or greater increase pre-post immunization (83% compared to 50%).
In response to PPS14, a similar significant increase in pre-to post-immunization antibody concentration was detected in all groups (p range 0.001–0.006). When both ELISA methods were compared there was no significant difference in IgG concentrations in young sera using either method. However, in the elderly the addition of PPS22F absorption significantly reduced the IgG concentration detected in post-(10.6%, p = 0.013) but not pre-immunization sera (0.8%, p = 0.367). There was a tendency towards lower antibody concentrations in the pre-immune sera of the young in comparison to the elderly, although this was not significant. The post-immune anti-PPS14 antibody concentrations were also not significantly different between young and elderly groups. However, more young than elderly demonstrated a two fold or greater increase in serum IgG concentrations, 90% and 55% respectively.
Our study was designed to compare the immune reaction to PPS in young adults (18 – 30 years old) to elderly (>65 years of age). However, several studies [17,18] indicate that the oldest elderly (>77 years of age) demonstrate a significant decrease in the immune response to PPS. We thus analyzed the results obtained for a subgroup of elderly >77 years old within our study population. In this small subgroup (n = 6) additional absorption with PPS22F resulted in a substantial but not significant decrease in measured antibody concentration of 28% in pre-immune and 15.6% in post-immune sera in comparison to young adults. A significant increase from pre-to post-immunization antibody concentration in response to PPS4 was only found when sera were absorbed with CPS (p = 0.05) but not when sera were absorbed with CPS + PPS22F (p = 0.069). There was no significant increase in pre-to post-immunization antibody levels against PPS14 for either method (p = 0.094 for both). Although, the post-immunization anti-PPS14 IgG GMC of the young adult group was not significantly different from the elderly, it was significantly higher than the sub-group of elderly >77 years old using both absorption methods (p = 0.035 and p = 0.022). There was a reduction in the proportion of this subgroup that demonstrated a two-fold or greater increase in IgG levels pre-post immunization, in response to both PPS4 (29%) and PPS14 (43%) than either all elderly (50% and 55% respectively) and young (83% and 90% respectively).
Immunoglobulin Isotype and Subclass
The predominant immunoglobulin isotype expressed in response to both PPS4 and PPS14 consisted of IgG. Despite a significant increase from pre-to post-immunization, the GMC of the PPS-specific IgM and IgA antibody concentrations in response to both PPS did not exceed one microgram per millilitre in young or elderly. Moreover, there was no significant difference in either IgM or IgA antibody concentrations in young versus elderly. Additional absorption with PPS22F reduced both IgM and IgA concentrations in young and elderly alike, but not significantly so.
As shown in Table 3, the IgG response to PPS4 and PPS14 was dominated by the IgG2 subclass in young and elderly adults as previously reported [19,20]. In response to PPS4, 67.5 to 80.5 % of the total IgG antibody concentration measured consisted of IgG2, with no significant difference between young and elderly or pre-to post-immunization. Similarly, of the pre-and post-immune total PPS14-specific IgG antibody, 63.7 to 73% of the response consisted of IgG2 with no significant difference between age groups.
Table 3 Anti-PPS specific IgG isotype levels in young and elderly.
Pre-immune, GMC (95% CI) Post-immune GMC (95% CI) P value
IgG1
PPS4 Young 0.13 (0.06) 0.35 (0.18) 0.001
Elderly 0.14 (0.09) 0.33 (0.38) 0.051
PPS14 Young 0.87 (0.13) 3.59 (0.33) 0.001
Elderly 0.93 (0.44) 1.19 (0.95) 0.015
IgG2
PPS4 Young 1.03 (0.37–1.69) 5.14 (0.64–9.64) 0.010
Elderly 0.83 (0.04–1.62) 1.96 (0.98–2.94) 0.005
PPS14 Young 2.18 (1.86–2.50) 12.33 (4.51–20.15) 0.002
Elderly 1.36 (0.62–2.10) 4.78 (2.37–7.19) 0.004
Geometric mean concentration (GMC μg/ml) of anti-PPS antibody in pre-and post-immunization serum from young and elderly volunteers measured following absorption with CPS. P values shown in bold indicate significant difference between pre and post immunization by two tailed paired students T-test
PPS-specific IgG1 antibodies were also readily detectable and increased significantly post-vaccination in both age groups. Overall, IgG1 antibodies represented 9.5 to 11.2% of total IgG antibodies in response to PPS4 and 18.7 to 25.4% of total IgG antibodies measured in response to PPS14. There was no significant difference detected in either IgG1 antibody concentration or percent usage between young and elderly.
Additional absorption of sera with PPS22F did not affect either pre-or post-immunization IgG1 PPS4 or PPS14 concentrations in either age group or the IgG2 response to PPS14. However, in response to PPS4, absorption with PPS22F significantly reduced pre-and post-immunization IgG2 concentrations in both young and elderly (17 to 27% with p = 0.019 to 0.025).
Opsonophagocytic Activity
Opsonophagocytic activity was measured in pre-and post-immunization sera against pneumococcal serotypes 4 and 14. Geometric mean titers (GMT) and ranges are shown in Table 4.
Table 4 Opsonophagocytic activity in young and elderly.
Pre-immune GMT (95% CI) Post-immune GMT (95% CI) P value
PPS4 Young 10.2 (4.43–15.97) 93.6 (-19.63–206.89) 0.001
Elderly 14.9 (2.74–27.12) 147.0 (-16.63–310.70) 0.001
PPS14 Young 16.7 (-33.03–66.36) 245.8 (82.50–409.05) 0.0004
Elderly 26.9 (-104.78–158.60) 265.03 (115.74–414.32) 0.0003
Geometric mean titer of serotype specific opsonophagocytic activity in pre-and post-immunization serum, from young and elderly volunteers. Significance of increase from pre to post-immunization was calculated using two tailed paired students T-test.
In response to serotypes 4 and 14 both young and elderly demonstrated a significant increase pre-to post-immunization (p range 0.0003–0.001). There was no significant difference between the opsonophagocytic titers in young and elderly for either polysaccharide (p range 0.203–0.754). However, when the eldest subgroup was analyzed separately there was a significant increase in opsonophagocytic activity only in response to serotype 14 (p = 0.04) but not in response to serotype 4 (p = 0.214). The post immunization titers for the eldest subgroup for serotype 4 (GMT 35.92) and 14 (90.51) were significantly lower than in young donors (p = 0.03 and 0.02), all elderly donors (p = 0.01 for both) and the younger group of elderly (p = 0.002, 0.001).
Antibody Avidity Measurements
Antibody avidity was determined using sodium thiocyanate (NaSCN) in both ELISA methods as described. Geometric mean concentrations (GMC, M NaSCN) and ranges are given in Table 5. The addition of PPS22F to the assay did not significantly change the avidity measurements, pre-or post-immunization, in young or elderly (p range 0.053 – 0.715) except in young anti-PPS14 post immunization sera (p = 0.04). In response to PPS4 there was no significant difference between young and elderly in any group analyzed (p range 0.32 – 0.93). In response to PPS14 a significant difference was detected between young and elderly pre-immune sera when absorbed with CPS alone (p = 0.03) and post-immune sera when absorbed with CPS and PPS22F (p = 0.03).
Table 5 Avidity of anti-PPS antibodies from young and elderly.
Pre-immune serum GMC (95% CI)
CPS CPS + 22F P value
PPS4 Young 0.19 (0.07–0.32) 0.21 (0.04–0.38) 0.091
Elderly 0.22 (0.13–0.31) 0.22 (0.12–0.31) 0.646
PPS14 Young 0.28 (0.21–0.36) 0.27 (0.19–0.35) 0.711
Elderly 0.43 (0.27–0.59) 0.42 (0.27–0.56) 0.196
Post-immune serum GMC (95% CI)
CPS CPS+22F P value
PPS4 Young 0.35 (0.22–0.49) 0.40 (0.25–0.55) 0.053
Elderly 0.43 (0.23–0.63) 0.46 (0.26–0.65) 0.014
PPS14 Young 0.47 (0.33–0.62) 0.41 (0.29–0.54) 0.618
Elderly 0.68 (0.50–0.85) 0.66 (0.48–0.84) 0.715
Geometric Mean Concentration (GMC, M NaSCN) representing serum avidity by two different ELISA methods. CPS – serum adsorbed with CPS prior to ELISA. CPS+22F – serum absorbed with PPS22F and CPS prior to ELISA. P values shown in bold indicate a significant difference between absorption methods calculated using paired Student's T-test.
Anti-PPS4 antibodies demonstrated a significant increase in avidity pre-to post-vaccination in elderly volunteers by both methods (p = 0.01, 0.003) but not in the young (p = 0.147, 0.257). Young anti-PPS14 antibodies showed a significant increase pre-to post-immunization when sera were absorbed with CPS alone (p = 0.015) but not CPS and PPS22F (p = 0.061). Avidity of elderly anti-PPS14 antibodies pre-and post-immunization was significantly different by both methods (p = 0.007, 0.002).
In the oldest subgroup of elderly anti-PPS4, but not anti-PPS14, antibodies, when absorbed with CPS and 22F demonstrated a significant increase pre-to post-immunization (p = 0.03). There was no significant difference between avidity measurements in this subgroup compared to either young, all elderly or the youngest elderly.
Correlation between Opsonophagocytic Antibody Activity and Antibody Concentrations
Correlation coefficients of association (r values) were determined between log2 IgG antibody levels and log2 opsonophagocytic titers for PPS4 and PPS14 by absorbing sera with CPS or CPS and PPS22F (Table 6). In response to PPS4, the correlation coefficient increased following CPS+ PPS22F absorption in young adults from an r value of 0.64 (CPS absorption) to 0.81 (CPS+ PPS22F absorption). A similar increase from 0.69 with CPS and 0.79 with CPS+PPS22F absorption was noted in the elderly. The sub-group of elderly >77 years of age showed a correlation coefficient of 0.73 following absorption with CPS that increased to 0.87 following absorption with CPS and PPS22F. Thus, all age groups demonstrated a significant correlation between antibody concentration and opsonophagocytic antibody response that improved by additional absorption with CPS and PPS22F.
Table 6 Correlation between opsonophagocytic titers and antibody concentration.
CPS CPS + PPS22F P value
PPS4 Young 0.64 0.81 0.28
Elderly 0.69 0.79 0.51
PPS14 Young 0.90 0.88 0.78
Elderly 0.75 0.72 0.85
Correlation coefficients of association (r values) between post-immunization opsonophagocytic titer and antibody concentration following absorption with CPS or CPS + 22F in young and elderly volunteers. There was no significant difference between absorption techniques by Fisher's Z transformation.
In addition, all age groups demonstrated a relatively high correlation between antibody concentration and functional antibody activity in response to PPS14. Additional absorption with PPS22F did not improve the correlation coefficients of association between IgG antibody concentrations and opsonophagocytic titers for PPS14. All age groups showed a significant correlation between antibody concentrations and opsonophagocytic antibody activity as previously reported in adults [16].
Correlation between Opsonophagocytic Antibody Activity and Antibody Avidity
We also determined the correlation coefficients of association (r values) between concentration of NaSCN necessary to reduce ELISA optical density by 50% and log2 opsonophagocytic titers for PPS4 and PPS14. In response to PPS4, all groups showed a relatively poor correlation between avidity and opsonophagocytic activity (Table 7) that improved marginally after absorption with CPS and PPS22F. In the sub-group of elderly >77 years the correlation of association was very poor, however the r value improved markedly from 0.069 to 0.25 by additional absorption with PPS22F.
Table 7 Correlation between opsonophagocytic titers and antibody avidity.
CPS CPS + PPS22F P value
PPS4 Young 0.39 0.44 0.86
Elderly 0.52 0.53 0.97
PPS14 Young 0.51 0.49 0.94
Elderly 0.64 0.59 0.81
Correlation coefficients of association (r value) between post-immunization opsonophagocytic antibody titer and antibody avidity following absorption with CPS or CPS + 22F in young and elderly volunteers. There was no significant difference between absorption techniques by Fisher's Z transformation.
In response to PPS14, the correlation coefficients of association were better than those in response to PPS4 (Table 7). The additional absorption with PPS22F however, did not significantly affect the correlation coefficients of association.
Discussion
Recent improvements in the ELISA method used to measure anti-pneumococcal polysaccharide-specific antibodies, aimed at removing non-specific antibodies, have resulted in a significant increase in the correlation between antibody concentration and opsonophagocytic activity [16]. Although this methodology has been extensively evaluated in healthy adults [16] the effect of additional serum absorption with PPS22F on the measured antibody response in elderly adults has not been determined. The presence of non-serotype specific, poorly functional antibodies may in fact explain the discrepancy between measured antibody concentrations and functional antibody activity in the elderly.
The results of the present study indicate that both young and elderly adults respond to pneumococcal vaccination with a significant increase in PPS-specific antibody concentration, although the fold increase in antibody concentration was lower in elderly. Additional absorption with PPS22F resulted in a significant reduction in antibody concentration to PPS4 but did not affect the antibody concentration measurements to PPS14. It is likely that PPS14 does not contain cross-reactive antigens or contaminants as described for other pneumococcal polysaccharides [16,21,22]. The decrease in measured anti-PPS4 antibody resulted in an increased correlation between antibody concentration and functional antibody activity, consistent with previous studies [16]. Although the decrease in anti-PPS4 antibody was more pronounced in pre-immune sera, it was not significantly different between young and all elderly and did not affect fold-increase in antibody concentration post-immunization. The results of previous studies [18,23] however, indicate that an impaired immune response to PPS is limited to the oldest elderly. Although small in number (n = 6), separate analysis of elderly > 77 years of age showed a decrease in significance of the increase from pre-post when the sera was absorbed with PPS22F. In addition fewer volunteers in this age group demonstrated a two fold or greater increase in antibody concentration pre-post than either the young or elderly. These results suggest that previous studies may have shown an overestimation of antibody concentration and that this may be particularly critical in the oldest elderly population. The additional absorption with PPS22F thus revealed a potential impairment in quantitative antibody response to PPS4 in the oldest age group.
Several studies have demonstrated that reduced or absent functional antibody activity is directly related to low antibody avidity and not necessarily a result of low antibody concentration [24,25]. Specifically, it has been demonstrated that post-vaccination sera from elderly with low opsonophagocytic activity correlate with low IgG antibody avidity [23]. Additional absorption with PPS22F, which serves to remove non-specific antibodies with presumably low avidity, should in principle result in increased avidity measurement. In response to PPS4, absorption of sera with CPS and PPS22F increased antibody avidity in young and elderly adults, although this increase was only significant in the young. However, the moderate correlation between opsonophagocytic antibody titers and antibody avidity in both young and elderly, did not improve significantly following absorption with PPS22F. These data suggest that high avidity antibodies are not necessarily functional, or conversely, opsonophagocytic antibodies are not necessarily of high avidity. In contrast, in response to PPS14, there was a strong correlation between opsonophagocytic antibody titers and antibody avidity in all age groups, including elderly > 77. Similarly, Romero-Steiner, et al. [23] found a significant correlation between serum opsonophagocytic activity and antibody avidity, particularly in response to PPS14. In most elderly individuals, specifically those > 77 years of age, low opsonophagocytic antibody activity was directly related to low antibody avidity.
Opsonophagocytic activity is dependent on both antibody avidity and antibody concentrations. In the elderly, particularly those > 77 years old, lack of correlation between functional antibody activity and antibody avidity was generally attributable to low antibody concentrations. In contrast, some of the young volunteers with low opsonophagocytic activity demonstrated moderate to high antibody concentrations with moderate or high antibody avidity. These data suggest that young adults in particular, may generate high levels of high avidity antibodies, not directed at opsonophagocytic epitopes. Based on these data, we propose that at least two different mechanisms may be responsible for the discrepancy between antibody concentration and functional antibody activity, namely, poor antibody avidity and antibodies directed at non-opsonophagocytic epitopes present on the pneumococcal polysaccharide.
While overall there was no statistical difference between the response in young and elderly there was a trend towards a reduced response which increased with age. Less of the eldest subgroup of elderly responded to the vaccine with a two fold or greater concentration of antibody than either all elderly or young. Immunosenescence is an ongoing process with age; as the patient ages the immune system continues to decline. Thus, in studies aimed at detecting the most differences with ageing, future studies should aim to enroll older elderly, i.e. those over 75 years of age.
Conclusion
Our study indicates that additional absorption of serum with PPS22F results in an improved correlation between measured antibody concentration and functional antibody activity in young and elderly adults. The removal of non-specific antibodies may, however, be of particular importance in assessing PPS-specific antibody concentration in elderly and may uncover previously un-noted differences in quantitative antibody response.
Methods
Human Volunteers/Vaccination
Healthy young adults (<30 years old) and elderly volunteers (>65 years old) were asked to participate in this study. Elderly volunteers were recruited from the community as well as from the general internal medicine clinic at Medical College of Ohio. Young volunteers were recruited from the student population at MCO. Each individual was questioned concerning prior pneumococcal vaccination, medications, previous illness and present health. In addition, we obtained complete blood count (CBC), comprehensive chemistry profile including renal and liver functions and serum albumin, total B cells, T cell subsets and total IgG, IgM and IgA levels in all study candidates. Individuals previously immunized with the pneumococcal vaccine and any individual considered to be immunocompromised on basis of medication (chemotherapy, steroid preparations, immunosuppressive agents including anti-TNFa agents), previous/present illness (previous pneumococcal disease, splenectomy, auto-immune disease, end-stage renal or liver disease, HIV positivity, organ transplant or cancer) and those with abnormal CBC, chemistries, B or T cells or immunoglobulin levels did not qualify. Informed consent was obtained from all participants using protocols reviewed and approved by Institutional Review Board of Medical College of Ohio.
Pneumococcal Polysaccharide ELISA
ELISA was used to determine the anti-PPS specific human antibodies, isotypes and subclasses, in all volunteers before and 6 weeks after vaccination [23,26]. Comprehensive details regarding this procedure are found in the web document: . US reference Pneumococcal Antiserum 89SF, courtesy of Dr Carl Frasch CBER/FDA, was used as a control throughout.
Briefly, Nunc microtiter plates were coated with PPS4 or PPS14 (ATCC, Rockville, Md.). Human sera were absorbed with 5 μg/ml of cell wall polysaccharide (CPS) or with 5 μg/ml of CPS and 10 μg/ml of serotype 22F polysaccharide (ATCC) and incubated for 30 minutes [26]. Serial dilutions of absorbed sera were added to the wells and incubated. Horse-radish peroxidase labeled anti-human isotype and subclass specific conjugates were added and reaction developed with σ-phenylenediamine (Sigma, St. Louis), absorbance was read at 490 nm.
Antibody Avidity ELISA
Avidity of IgG antibodies to PPS4 and PPS14 was determined by sodium thiocyanate (NaSCN), a chaotropic agent that interferes with antigen-antibody interaction [27]. The molarity of NaSCN required to elute 50% of bound antibody was used as a measure of avidity, the relative strength of antigen-antibody binding. ELISA was performed as previously described [23]. Microtiter plates were set up with NaSCN molarity ranges from 4 M to 0.0625 M. Controls consisted of uninhibited serum samples and blanks. All serum samples were tested in duplicate. Avidity was expressed as the molarity of NaSCN required to inhibit 50% of binding of uninhibited serum.
Opsonophagocytic Assays
Opsonophagocytic assays were performed as previously described [23,28] to measure the functional antibodies against PPS4 and PPS14. Briefly, viable pneumococci were allowed to pre-opsonize with serially diluted heat-inactivated sera. Newborn rabbit serum (Pel-Freez, Brown Deer, WI) was added as source of complement (12.5%). Washed, differentiated cells were added at a 400:1 effector: target cell ratio. After incubation, to allow phagocytosis to occur, an aliquot was plated and incubated overnight at 37°C with 5% CO2. The opsonophagocytic titer was determined as the reciprocal of the dilution with >50% killing when compared to the control wells without serum.
Statistics
Geometric mean concentrations of IgG reactive with PPS4 and PPS14 were calculated for each group. Pre-to post-immunization values were compared using paired Student's T test, young and elderly were compared using two sample, unpaired Student's T test. Correlation coefficients and p values were calculated using Fisher's Z transformation using log2 base of opsonophagocytic titers and antibody concentrations. P values less than 0.05 were considered to be significant. Statistical calculations were performed with use of SPSS software 11.5.1.
Competing interests
None of the authors of this paper have a commercial or other association that might cause conflict of interest.
Authors' contributions
KK: performed ELISAs, collated data, wrote initial draft of manuscript
SLS: adapted assays for laboratory, critical revision of manuscript, final revision of manuscript
AKS: performed opsonophagocytic assays
SK: provided key statistical help and advice
SRS: consultant regarding opsonophagocytic assay, provided cell lines and controls
GMC: consultant regarding ELISA assay, provided critical controls
MAJW: conceived of the study, critical revision of manuscript
Acknowledgements
This work was supported by Public Service Grant, AG15978, from the National Institute of Aging.
==== Refs
Vlasich C Pneumococcal infection and vaccination in the elderly Vaccine 2001 19 2233 2237 11257339 10.1016/S0264-410X(00)00451-5
Sisk JE Moskowitz AJ Whang W Lin JD Fedson DS McBean AM Plouffe JF Cetron MS Butler JC Cost-effectiveness of vaccination against pneumococcal bacteremia among elderly people Jama 1997 278 1333 1339 9343464 10.1001/jama.278.16.1333
Ruiz-Gonzalez A Falguera M Nogues A Rubio-Caballero M Is Streptococcus pneumoniae the leading cause of pneumonia of unknown etiology? A microbiologic study of lung aspirates in consecutive patients with community-acquired pneumonia Am J Med 1999 106 385 390 10225239 10.1016/S0002-9343(99)00050-9
Marrie TJ Community-acquired pneumonia in the elderly Clin Infect Dis 2000 31 1066 1078 11049791 10.1086/318124
Jackson LA Neuzil KM Yu O Benson P Barlow WE Adams AL Hanson CA Mahoney LD Shay DK Thompson WW Effectiveness of pneumococcal polysaccharide vaccine in older adults N Engl J Med 2003 348 1747 1755 12724480 10.1056/NEJMoa022678
Bruyn GA Zegers BJ van Furth R Mechanisms of host defense against infection with Streptococcus pneumoniae Clin Infect Dis 1992 14 251 262 1571441
Austrian R Some observations on the pneumococcus and on the current status of pneumococcal disease and its prevention Rev Infect Dis 1981 3 S1 17 7025155
Butler JC Breiman RF Campbell JF Lipman HB Broome CV Facklam RR Pneumococcal polysaccharide vaccine efficacy. An evaluation of current recommendations JAMA 1993 270 1826 1831 8411526 10.1001/jama.270.15.1826
Robbins JB Austrian R Lee CJ Rastogi SC Schiffman G Henrichsen J Makela PH Broome CV Facklam RR Tiesjema RH Considerations for formulating the second-generation pneumococcal capsular polysaccharide vaccine with emphasis on the cross-reactive types within groups J Infect Dis 1983 148 1136 1159 6361173
Rubins JB Janoff EN Pneumococcal disease in the elderly: what is preventing vaccine efficacy? Drugs Aging 2001 18 305 311 11392439
Hirschmann JV Lipsky BA The pneumococcal vaccine after 15 years of use [see comments] Arch Intern Med 1994 154 373 377 8117169 10.1001/archinte.154.4.373
Musher DM Chapman AJ Gorce A Jonsson S Briles D Banghu RE Natural and vaccine-related immunity to Streptococcus pneumoniae. J Infect Dis 1986 154 245 256 3722865
Siber GR Priehs C Madore D Standardization of antibody assays for measuring the response to pneumococcal infection and immunization. Pediatr Infect Dis J 1989 8 S84 S91 2648302
Koskela M Serum antibodies to pneumococcal C polysaccharide in children: response to acute pneumococcal otitis media or to vaccination Pediatr Infect Dis J 1987 6 519 526 3615065
Musher DM Luchi MJ Watson DA Hamilton R Baughn RE Pneumococcal polysaccharide vaccine in young adults and older bronchitics: determination of IgG responses by ELISA and the effect of adsorption of serum with non-type-specific cell wall polysaccharide. J Infect Dis 1990 161 728 735 2319166
Concepcion NF Frasch CE Pneumococcal type 22f polysaccharide absorption improves the specificity of a pneumococcal-polysaccharide enzyme-linked immunosorbent assay Clin Diagn Lab Immunol 2001 8 266 272 11238206 10.1128/CDLI.8.2.266-272.2001
Romero-Steiner S Musher DM Cetron MS Pais LB Groover JE Fiore AE Plikaytis BD Carlone GM Reduction in functional antibody activity against Streptococcus pneumoniae in vaccinated elderly individuals highly correlates with decreased IgG antibody avidity Clin Infect Dis 1999 29 281 288 10476727
Rubins JB Puri AK Loch J Charboneau D MacDonald R Opstad N Janoff EN Magnitude, duration, quality, and function of pneumococcal vaccine responses in elderly adults J Infect Dis 1998 178 431 440 9697723
Anttila M Voutilainen M Jantti V Eskola J Kayhty H Contribution of serotype-specific IgG concentration, IgG subclasses and relative antibody avidity to opsonophagocytic activity against Streptococcus pneumoniae Clin Exp Immunol 1999 118 402 407 10594558 10.1046/j.1365-2249.1999.01077.x
Lottenbach KR Mink CM Barenkamp SJ Anderson EL Homan SM Powers DC Age-associated differences in immunoglobulin G1 (IgG1) and IgG2 subclass antibodies to pneumococcal polysaccharides following vaccination Infect Immun 1999 67 4935 4938 10456954
Coughlin RT White AC Anderson CA Carlone GM Klein DL Treanor J Characterization of pneumococcal specific antibodies in healthy unvaccinated adults Vaccine 1998 16 1761 1767 9778753 10.1016/S0264-410X(98)00139-X
Yu X Gray B Chang S Ward JI Edwards KM Nahm MH Immunity to cross-reactive serotypes induced by pneumococcal conjugate vaccines in infants J Infect Dis 1999 180 1569 1576 10515817 10.1086/315096
Romero-Steiner S Musher DM Cetron MS Pais LB Groover JE Fiore AE Plikaytis BD Carlone GM Reduction in functional antibody activity against Streptococcus pneumoniae in vaccinated elderly individuals highly correlates with decreased IgG antibody avidity [see comments] Clin Infect Dis 1999 29 281 288 10476727
Sun Y Hwang Y Nahm MH Avidity, potency, and cross-reactivity of monoclonal antibodies to pneumococcal capsular polysaccharide serotype 6B Infect Immun 2001 69 336 344 11119522 10.1128/IAI.69.1.336-344.2001
Usinger WR Lucas AH Avidity as a determinant of the protective efficacy of human antibodies to pneumococcal capsular polysaccharides Infect Immun 1999 67 2366 2370 10225896
Wernette CM Frasch CE Madore D Carlone G Goldblatt D Plikaytis B Benjamin W Quataert SA Hildreth S Sikkema DJ Kayhty H Jonsdottir I Nahm MH Enzyme-linked immunosorbent assay for quantitation of human antibodies to pneumococcal polysaccharides Clin Diagn Lab Immunol 2003 10 514 519 12853378 10.1128/CDLI.10.4.514-519.2003
Pullen GRFMGHCS Antibody avidity determination by ELISA using thiocyanate elution. Journal of Immunological Methods 1986 86 83 87 3944471 10.1016/0022-1759(86)90268-1
Romero-Steiner S Libutti D Pais LB Dykes J Anderson P Whitin JC Keyserling HL Carlone GM Standardization of an opsonophagocytic assay for the measurement of functional antibody activity against Streptococcus pneumoniae using differentiated HL-60 cells Clin Diagn Lab Immunol 1997 4 415 422 9220157
|
15982420
|
PMC1184095
|
CC BY
|
2021-01-04 16:36:31
|
no
|
Immun Ageing. 2005 Jun 27; 2:10
|
utf-8
|
Immun Ageing
| 2,005 |
10.1186/1742-4933-2-10
|
oa_comm
|
==== Front
J NanobiotechnologyJournal of Nanobiotechnology1477-3155BioMed Central London 1477-3155-3-71599240410.1186/1477-3155-3-7ResearchDirect microcontact printing of oligonucleotides for biochip applications Thibault C [email protected] Berre V [email protected] S [email protected]évisiol E [email protected]çois J [email protected] C [email protected] LAAS-CNRS, 7, avenue du Colonel Roche 31077 TOULOUSE Cedex 42 Biochips Platform Genopole Toulouse, UMR-CNRS 5504 & INRA 792, 135, avenue de Rangueil, 31077 TOULOUSE Cedex 43 Laboratoire de Biotechnologie & Bioprocédés, UMR-CNRS 5504 & INRA 792, 135, avenue de Rangueil, 31077 TOULOUSE Cedex 42005 1 7 2005 3 7 7 11 4 2005 1 7 2005 Copyright © 2005 Thibault et al; licensee BioMed Central Ltd.2005Thibault et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
A critical step in the fabrication of biochips is the controlled placement of probes molecules on solid surfaces. This is currently performed by sequential deposition of probes on a target surface with split or solid pins. In this article, we present a cost-effective procedure namely microcontact printing using stamps, for a parallel deposition of probes applicable for manufacturing biochips.
Results
Contrary to a previous work, we showed that the stamps tailored with an elastomeric poly(dimethylsiloxane) material did not require any surface modification to be able to adsorb oligonucleotides or PCR products. The adsorbed DNA molecules are subsequently printed efficiently on a target surface with high sub-micron resolution. Secondly, we showed that successive stamping is characterized by an exponential decay of the amount of transferred DNA molecules to the surface up the 4th print, then followed by a second regime of transfer that was dependent on the contact time and which resulted in reduced quality of the features. Thus, while consecutive stamping was possible, this procedure turned out to be less reproducible and more time consuming than simply re-inking the stamps between each print. Thirdly, we showed that the hybridization signals on arrays made by microcontact printing were 5 to 10-times higher than those made by conventional spotting methods. Finally, we demonstrated the validity of this microcontact printing method in manufacturing oligonucleotides arrays for mutations recognition in a yeast gene.
Conclusion
The microcontact printing can be considered as a new potential technology platform to pattern DNA microarrays that may have significant advantages over the conventional spotting technologies as it is easy to implement, it uses low cost material to make the stamp, and the arrays made by this technology are 10-times more sensitive in term of hybridization signals than those manufactured by conventional spotting technology.
microcontact printingelastomeric stampDNA immobilisationbiochipsdetection of mutations
==== Body
Background
DNA microarrays have rapidly evolved to become one of the essential tools to investigate expression or mutation of thousands of genes simultaneously. Two main technology platforms for manufacturing DNA chips have emerged. The first platform uses the immobilization of prefabricated DNA or oligonucleotides by spotting on functionalized glass slides using metal pins as originally developed by Brown and collaborators (see ), or by a non-contact method using piezoelectric liquid handling [1]. The second platform rests on the direct in-situ synthesis of oligonucleotides (between 20 to 70 mers in general) on glass slides or silicon surfaces, as developed by Affymetrix or Agilent [2]. A typical characteristic of these techniques is the sequential nature of the process. One molecule is deposited after another or one base is added to the previous one, with the consequence that each array is made as an original with a reduced throughput, although Affymetrix microarrays manufacturing involves combinatorial processes that allow multiple microarrays (around 96) to be synthesized in parallel in matters of hours. Nevertheless, these technology platforms needs sophisticated equipment, leading to high density arrays that can be too expensive for production and utilization of simple-customized-DNA arrays.
There is a need for alternative patterning methods that must be very simple, reproducible, cost-effective, and eventually transferable to any laboratories for their own problematic. The microcontact printing (μCP) could fulfill this requirement as it is a printing technology that uses cheap elastomeric stamps made usually of polydimethylsiloxane (PDMS) and which exhibits relief patterns at the micron and nanoscale [3]. These stamps let to parallel deposition of molecules on a target surface, in the same manner as the printing of a page of book instead of a letter being written individually to compose the page. Previous works demonstrated that proteins can be deposited on a substrate surface by microcontact printing (μCP) [4,5]. More recently, Lange et al. [6] showed that μCP technique can be used to deposit DNA molecules with a PDMS surface of the stamp chemically modified to enable the DNA molecules to stick on the stamp. This functionalization step strongly restricted the speed of this technology, as it takes several hours from the conversion of the CH3 terminated surface of the PDMS into an aminated surface to complete inking of the stamps prior to printing the target surface.
In this paper, we demonstrate that μCP can be used to fabricate DNA biochips directly without any surface modification of the stamps. We show that inking and contact times of less than 30 seconds give high quality and high resolution arrays by μCP. According to our new variant of the process, the stamp is simply inked with the molecules of interest, dried under a nitrogen stream and then printed manually onto the substrate surface (see Fig. 1). It is foreseen that this technology platform will be highly competitive for high throughput analysis of gene expression and mutation detection analyses. Moreover, this technique can be easily implemented for sub-micron patterns as demonstrated previously [6] and in this work.
Figure 1 Principe of microcontact printing of DNA molecules. (1) Inking of the stamp with the oligonucleotide solution, a 1 cm2 stamp is loaded with a 2 to 20 μl droplet of solution for a given time (2) drying of the stamp under Nitrogen stream, (3) manual contact between the inked PDMS stamp and the glass slide, (4) probe molecules are transferred on the slide along patterns that correspond to the relief structures of the PDMS stamp.
Results and Discussion
The two main steps of μCP are the adsorption of the biomolecules on the stamp (inking process) and the transfer from the stamp to a target surface (contact printing). It is important that the retention of molecules on the stamp surface does not prevent their subsequent transfer to the slide, and that the inking and the contact time were as short as possible for optimizing the high throughput of the technique. In a recent work [6], this compromise was obtained by a specific chemical treatment of the elastomeric poly(dimethylsiloxane) material (PDMS) of the stamp after molding. In contrast to this report, we found that untreated PDMS stamp that has a strong hydrophobic surface after curing, easily adsorbs a sufficient amount of DNA molecules within few seconds while allowing their subsequent deposition by contact on microscope glass slides or silicon. The printing process works for untreated glass or silicon surfaces, but real bioassays were carried out on treated glass surfaces enabling strong binding of the probe molecules. During the contact, the purpose is to transfer efficiently and as quick as possible the molecules from the stamp surface to the slide without affecting the size of the patterns. A specific chemistry on the surface of the slide is also important for the attachment of the probes after taking away the stamp from the surface. We also verified that stamps could be reused several times after cleaning in deionized water. The experiments detailed below aim at investigating the influence of several parameters including the surface chemistry of the slide, the inking and the contact time of the stamp, and to demonstrate the potentiality of this technique for actual biochips.
Surface chemistry and high uniformity of DNA printing on target surfaces
Experiments reported in this paper were carried out using two different type of glass slides that differed by their surface functionalization: positively charged amine glass slides (Ultra Gap, Dow corning) and dendrislides, which are glass slides that have been functionalized with nanometric spherical dendrimeric particles bearing aldehydes reactive group at the periphery for covalent attachment of the 5'-NH2 probes [7,8]. These two types of functionalized slides were printed for 15 sec with a stamp that has been incubated for 30 sec with a 10 μM solution of 35-mers 5'-NH2 probe in Na-phosphate buffer at pH 9.0. Hybridisation was achieved using a 15-mer 5'Cy5 target complementary to the 35-mer 5'-NH2 probe. As shown on Fig. 2, the micronic features of the stamp (squares, disks, gears, crosses, spirals, ...) were clearly noticeable on both types of glass slides. However, we observed systematically a greater signal to noise ratio, a better uniformity and edge definition of the spots with dendrislides (Fig. 2B) than with electrostatic slides (Figure 3A). This result is consistent with our previous report that the functionalization of surface with dendrimers reduces the non specific adsorption of fluorescent material [8]. In addition, the "donut" formation of spots frequently obtained after deposition of DNA molecules by contact spotting was no longer observed since the μCP is a "dry" deposition technique. This enables a better treatment of the fluorescence images for quantitative analysis. The upper part of Fig. 2C shows few lines on the array that exhibit a pitch of 4 μm which could only be seen as very small red spots because the fluorescent scanner cannot resolve the features. A magnification on conventional features (i.e. squares and disks) is shown in Fig. 2D. On this image, the contour of the patterns was mainly blurred by the pixel size of the scanner. In order to allow Atomic Force Microscopy (AFM) characterization, submicronic features were printed on silicon surface instead of glass slides to minimize the surface roughness. These patterns consisted in a periodic array of 500 nm wide lines at a pitch of 1 μm. As shown in Fig. 3, the 500 nm wide lines are clearly visible and the printed oligonucleotides appear as small aggregates that could be distinguished from the smooth surface of the silicon substrate. It is worth noticing that in this case the surface of the sample could not be rinsed after printing, because the untreated silicon surface does not provide strong adhesion of DNA molecules. Edge roughness and small aggregates visible on the image can be possibly attributed to residues coming from the buffer solution.
Figure 2 Comparison between two types of slides. Fluorescence images of printed micronic patterns. Stamp was incubated with a 35-mers probe oligonucleotide for 30 sec, then put in contact for 15 sec with two types of microscope glass slides. A, electrostatic slide (ultra Gap, corning), B, dendrislide (home made slide). Slides were then incubated with a 15-mer 5'-Cy5 labeled oligonucleotide. C and D are a zoom area of B.
Figure 3 Example of DNA printing at the submicronic scale. AFM image (taping mode) of 30-mers 5'-GCATGCTTAGTTGCTATTATCAAAATA-3', corresponding to BCK2 yeast gene printed on an untreated silicon surface. The pitch of the periodic array of lines is 1 μm. Note that the chemical surface states of the silicon was not really controlled: rough native oxide.
Inking time
In our first trial, the molded PDMS stamps were incubated at room temperature in the oligonucleotides solution for different times ranging from 30 sec to 1 hr, and then printed on a dendrislide after drying. Under these conditions, a very high and saturating fluorescent intensity was obtained independently of the inking time, likely because the amount of transferred fluorescent DNA molecules to the surface was already very high at the shortest inking time tested. It was even possible to observe deleterious effects for excessive inking times due to excess fluorescent material deposited at the periphery of the stamp (data not shown). These results indicated that the PDMS surface was saturated with DNA molecules in less than 30 sec of inking. We therefore reduced the inking time to a period that is easily compatible with a handling procedure of the stamps, i.e. 15 sec.
To explain the excellent performance of this technique to print DNA probes, we suggest that a hydrophobic interaction takes place between the PDMS surface of the stamp and single strand DNA molecules, since the PDMS surface is highly hydrophobic, and the DNA strand can also exhibit hydrophobic properties through its bases content, even though it is an hydrophilic molecule. Moreover, hydrophobic interactions are 10 to 100 times stronger and have a longer range of action than the Van der Waals interactions [9,10]. On the other hand, a fast and efficient transfer of the DNA probes from the stamp to the slide required that the interacting forces between the oligonucleotides and the PDMS surface must be weaker than those occurring between the oligonucleotides and the surface of the slide. This was verified in our experiments for both positively charged and hydrophobic dendrimeric activated surface slides. As a consequence, preserving the hydrophobicity of the PDMS stamp is clearly a key point in order to reduce the inking times for DNA printing and to favor the subsequent transfer of the molecules to either a positive charged or a hydrophobic surface. This is the main difference between our work and that of Lange et al [6]. In this latter work, the adsorption of DNA probes on the stamp was mainly based on electrostatic interactions with the consequence of long inking period (45 min.). In addition, as the surface treatment of PDMS is known to be unstable on air, our process, which does not involve any surface modification after molding, should be more reproducible and should allow the reusability of the stamp (see below). It is worth to note that similar results were obtained using long single DNA molecules or double stranded PCR fragments. However, as can be seen in Fig. 4, the signal intensity was significantly lower with stamped PCR products than with oligonucleotides. This observation was actually not specific to this technique since the same results were observed using conventional fabrication of arrays by mechanical spotting (V. Le Berre, unpublished data).
Figure 4 Comparison between oligonucleotides and PCR fragments. Fluorescent images of typical micrometric printed features. Stamp was incubated for 30 sec with a 500 bp PCR fragment (dsDNA) of the yeast HSP12 gene (A) or with a 20-mer oligonucleotide of the same yeast gene (B), then set in contact manually for 15 sec with a dendrislide. Hybridisation was carried out with HSP12 complementary Cy5-labelled oligonucleotide. Values of fluorescence intensity were measured at 635 nm with the GenePix 4000B from axon at 600 PMT. Mean intensity at 635 of 12 features on two experiments – Background was 2120 for A and 4119 for B.
Contact time and successive prints
To identify the transfer mechanisms of the molecules from the stamp surface to the slide, we investigated the influence of the contact time and the evolution of fluorescent signals after successive prints with the same stamp loaded with a fluorescent 35-mer 5'-labelled Cy5-oligonucleotide-3'NH2 (5'Cy5-TTAGCGCATTTTGGCATATTTGGGCGGACAACTT-NH2-3'). On the same slide, consecutive stamping steps were performed with a contact time of 15 sec, 1 min or 2 min, which took in total 2 to 20 min to pattern a dendrislide with 10 successive prints. To evaluate the change in fluorescence intensity along the successive print, the total intensity subtracted from the local background of specific features on the patterned slide were integrated and compared to the total intensity from the first print which was set arbitrarily at 100%. As shown on Fig. 5, this change followed an exponential decay up to the 4th stamping, and surprisingly, this decay was dependent of the contact time. The following equation
Figure 5 Fluorescence signal variation for successive prints. Variation of the fluorescence intensity for successive prints and for three different contact times (15 seconds, 1 minute and 2 minutes) between the stamp and the slide. Stamp was incubated with a 35-mer 5'-labelled Cy5 oligonucleotide for 30 sec than put in contact with the dendrislides. The value of fluorescence intensity (fluorescent – background) was measured at 635 nm with Genepix scanner under 600 PMT optical excitation. Each point represents an average of 4 independent experiments. Fittings of the data points with an exponential linear regression (solid lines), exhibits good agreement as attested by the reported correlation factors R.
-dN/dn = kN
where N is the number of molecules deposited on the slide at print number n, could be used to determine the characteristic of k, a kind of sticking coefficient of the molecules on the surface. The extracted values for k turned out to be dependent upon the contact time, with k increasing as the contact time decreased (k = 1.36 for t = 15 s, k = 0.67 for t = 1 min, k = 0.57 for t = 2 min). This result indicated that longer the contact time, slower was the depletion of the stamp in biomolecules. This behavior is suggestive of a slow diffusion of the molecules retained inside the cavity of the PDMS stamp to its relief structures that are in contact with the slides, as depicted in Fig. 6. It is therefore expected to observe a slower decrease of the fluorescence intensity for increasing contact times because there is more time for the biomolecules to migrate to the surface. In addition, we calculated that the k coefficient roughly changes with the inverse of the square root of the contact time, which is consistent with a diffusion limited deposition mechanism. Accordingly, the exponential decay of the fluorescence signal was no longer valid after 4 successive printing steps (Fig. 6). For n > 4, the number of molecules initially adsorbed on the relief structures of the PDMS stamp has been largely depleted in previous prints. However, a low fluorescence intensity that decrease very slowly from the 5th to the 7th print was still measured. This suggested a slow diffusion of molecules from the edges of the pattern to the slides during the contact. In that case, the number of printed molecules should be higher at the periphery of the features than in the center. The fluorescence images of the 5th to the 7th print for a contact time of 2 min nicely confirmed this assumption (Fig. 7). Essentially the rims of the specific features were recognizable likely because the remaining molecules had enough time to migrate from the edges of the relief printing of the stamp to the glass surface during the contact time. Thus, at shorter contact times, the fluorescence images were even worse (not shown), and hence the intensity values were lower (see Fig. 5).
Figure 6 Proposed mechanism for the diffusion of oligonucleotides during stamping. This picture shows schematically the possible migration direction of the oligonucleotides on the stamp surface during contact. This flow could explain the preferential deposition of molecules at the rim of the patterns.
Figure 7 Comparison between first and last print with the same stamp. (A) shows the fluorescent image of the patterns transferred at the first print, and (B) shows the printing patterns after 5th (B1), 6th (B2) and 7th print (B3). Stamps were inked with a 15-mer 5'-labelled Cy5 oligonucleotide for 30 sec and then set in contact for 2 min with the dendrislide. The well defined features is shown in (A) whereas only the rims of the patterns were detected after the 4th print (B).
As a conclusion of this section, we clearly identified some problems related to diffusion of biomolecules during stamping that may hamper the production of high quality arrays by successive stamping without re-inking. On the other hand, taking into account that the loading of the stamp is very fast and that high quality deposition by μCP of DNA molecules takes less than 15 sec to give optimal fluorescence signals, it appears more favorable to re-ink the stamp during 15 – 30 sec after each print, which is eventually faster than consecutive print.
Comparison between μCP deposition and contact deposition using metal pins
In order to compare μCP with a conventional spotting method, we performed a dedicated experiment in which the fluorescence intensity of DNA array was determined as a function of the concentration of the DNA probe used to manufacture the slides by the two techniques. To allow a direct comparison between the two methods, spots of 60 μm diameter size made with different concentration of 20-mer oligonucleotides from HSP12 were spotted with a commercial spotter (VersArray ChipWriter Pro, Biorad company) on a dendrislide, and disks of the same dimension were printed by μCP under the same condition. The arrays were then hybridized with the complementary labeled molecules. Fig. 8 shows the evolution of the fluorescence intensity in arbitrary units as a function of the initial concentration of the probe. From a range of 0.1 to 10 μM, the fluorescence signal was 5 to 10-fold higher when the deposition was performed by μCP than by a conventional spotter. This significant difference could be explained by the fact that deposition with a dry stamp in which the DNA molecules are delivered at the interface between the elastomeric material and the slide surface could offer uniform layers of densely packed molecules. Conversely, the deposition of a liquid droplet on the slide surface, which is let to evaporate, may give irregular layers of dispersed molecules. Alternatively or complementary to this explanation, it is possible to consider that the probes printed on the surface by μCP are better organized than by spotting, enabling a greater amount of targets accessible to the probes. In any case, for a given signal/noise ratio, the amount of probe molecules is significantly lower to get the same hybridization signals using μCP as compared to the spotting technology. This could be in the future a reasonable advantage of this technique taking into account the prohibitive price of DNA probe molecules. Moreover, this printing procedure is versatile and gives also excellent results with longer DNA molecules or double stranded PCR fragments.
Figure 8 Comparison between μCP deposition and contact deposition using metal pins. Evolution of the fluorescence intensity in arbitrary units as a function of the concentration of the solution containing the probe molecules. 60 μm diameter spots of 20-mer oligonucleotides from HSP12, were deposited using a commercial Spotter (VersArray ChipWriter Pro, BIO-RAD) and then hybridized with the complementary labeled molecules. Disks and square of the same dimension were printed by μCP and treated exactly in the same conditions.
Mutation detection
Having demonstrated that oligonucleotides can be successfully printed in multiple copies, yielding uniform patterns, we investigated the possibility to manufacture an array bearing short oligonucleotides of a given gene by μCP for detecting a single mutation as it can be made with the DNA microarray technology [11,12]. We printed 5 different 20-mer oligonucleotides from HSP12, encoding a protein chaperone in yeast [13]. These probes differed from each other by a single or a double base mutation at positions proximal to the 5' or 3' end or in the middle of the sequence. These oligonucleotides were then hybridized with Cy5-labelled cDNA prepared from total yeast RNA (see method section for additional details) in the automatic hybridization room. We compared the hybridization intensity of the target molecules on the printed patterns with that from the perfectly matching target sequence to the 20-mer oligonucleotide probe. We observed that whatever the position and nature of the mutation, the hybridization signal was considerably reduced for mutated sequences. As expected, the position of the mutation along the sequence of the probe molecule strongly influenced the hybridization ratio (Fig 9). This experiment was repeated 4 times independently and yielded highly reproducible data with a statistical deviation of <1%. Altogether, these results were very similar to those obtained using microarrays fabricated with dendrislides by a conventional spotting method [7]. This indicates that the quality of the arrays printed by μCP with respect to hybridization assay is largely equivalent to arrays produced by conventional deposition techniques.
Figure 9 Mutation detection. Comparison of the hybridization signal intensity of the target molecules on 5 different printed patterns differing by only single or double mutations. "Mutation" 1 corresponds to the exact match of the target molecule and serves as a reference. Five 20-mer oligonucleotides probes were printed at 10 μM in Na-Pi buffer 0.3 M, pH 9.0 on a dendrislide. These oligonucleotides were part of the yeast HSP12 sequence, and varied from each other by a single or two mutations proximal to the 5' end or 3' end or in the middle of the sequence. The 20-mer sequences from HSP12 are noted as follows: 1: NH2 5'-AATATGTTTCCGGTCGTGTC-3'; 2: NH2 5'-AATATGTTTCAGGTCGTGTC-3'; 3: NH2 5'-AATATGTTTCCGGTCGTGTA-3'; 4: NH2 5'-AATATGATTCCGGACGTGTC-3'; 5: NH2 5'-AATAAGTTTCCGGTCGTGTC-3'; Hybridisation was carried out with Cy5-labelled oligonucleotide (Cy5 5'-GACACGACCGGAAACATATT 3'). Values of fluorescence intensity were measured at 635 nm with the GenePix 4000B from axon at 600 PMT and correspond to an average of 4 experiments. Statistics errors are less than 0.4% for the 4 experiments.
Conclusion
In this work, we demonstrated that μCP is a new potential technology platform to pattern DNA microarrays at a relatively high speed, high resolution and high reproducibility. Two additional features which may provide significant advantages of this technology over the conventional spotting technologies are: (i) the simplicity of the μCP associated with the low cost of the material employed to make the stamp, and (ii) the arrays made by μCP technology provide 10-times higher fluorescence intensity after hybridization compare to those manufactured by conventional spotting technology. With these advantages in mind, our next step will be the fabrication of a dedicated automatic X, Y, Z controlled tool for printing different probe molecules with a high throughput. In the future, μCP may help to simplify, accelerate and improve the fabrication of microarrays and increase significantly their reliability and accessibility in i.e. clinical applications.
Methods
Stamp fabrication
The first step of fabrication consists in generating a silicon master. This was achieved by proximity U.V. photolithography on a Si [100] wafer coated with positive resist (AZ 1529), and pattern transfer by deep Reactive Ion Etching (1.4 μm deep). For submicronic patterns, Electron beam lithography on PMMA (PolyMethylMetAcrylate) was used instead of UV photolithography and the etch depth was limited to 100 nm. To enable simple demoulding of this master, an anti-adhesive treatment is carried out using silanisation in liquid phase with OTS (octadecyltrichlorosilane). The final step consists to cure the PDMS pre-polymer solution containing a mixture (10:1 mass ratio) of PDMS oligomers and a reticular agent from Sylgard 184 Kit (Dow Corning) on the silicon master. The PDMS was thermally cured at 120°C for 90 min or for 12 hr at 80°C (both methods giving similar results of stamping). A silicon master can be reused more than 50 times and each stamp can be used for a large number of prints (>100).
Surface chemistry of the substrate
Two kinds of microscope glass slides were used for spotting and printing the probes. Using "electrostatic" glass slides that are positively charged amine glass slides (Ultra Gap, Dow corning), the printed/spotted probes were cross-linked onto the amine surface by UV light at 300 mJ. With dendrislides (home made slide bearing generation 4 dendrimers, see [7], and our web site: ), a covalent attachment of the probes on the glass surface through aldehyde function of the dendrimers was performed [8,9]). After spotting, the dendrislides were allowed to dry overnight at room temperature. The reduction of the imines function formed between probes and dendrimer was carried out by immersion of the slides into a solution containing NaBH4 at 3.5 mg/ml for 3 hr at room temperature under agitation. The DNA slides were washed three times in water during 2 min, at room temperature and then dried under a stream of nitrogen.
Stamping process
Stamps were incubated with 2–20 μl of a 10 μM oligonucleotide solution made in Na-phosphate buffer 0.3 M, pH 9 for only 30 sec (unless mentioned differently), and then blown dried under a stream of nitrogen. Then, the stamp was printed manually onto the substrate surface and left in place during a controlled contact time. A 35-mer 5'-labelled Cy5-oligonucleotide-3'NH2 (5'Cy5-TTAGCGCATTTTGGCATATTTGGGCGGACAACTT-NH2-3'), a 35-mer 5'-amino modified (5'NH2-GTGATCGTTGTATCGAGGAATACTCCGATACCATT) and 70-mer 5'NH2 oligonucleotides corresponding to yeast HSP12 gene (from Qiagen/Operon yeast set) were used in spotting and printing experiments. The PCR fragment was a 500 bp amplified fragment on HSP12 gene using universal primers as described elsewhere [7].
Preparation of labeled targets
The target was a 15-mer 5'-labelled Cy5 oligonucleotide (Cy5-AATGGTATCGGAGTA) complementary to the 35-mer probes (5'NH2-GTGATCGTTGTATCGAGGAATACTCCGATACCATT). Other targets were prepared from total yeast RNA as a template by incorporation of fluorescent-labeled Cy5 or Cy3-dCTP during first-stand cDNA synthesis. The labeling reaction and cDNA purification was carried out with 15 μg total RNA using the LabelStar Kit from Qiagen following the manufacturer's instruction.
Hybridization
In initial experiments, the hybridization was carried out in an hybridization cassette (Corning Inc), according to the standard protocol used in the lab for microarray technology in the presence of 20 μl solution containing 16.5 μl Dig Easy buffer (Roche Diagnostic), 1 μl of denatured salmon sperm DNA and 2.5 μl of labeled target and covered with a 2.2 cm2 cover slip to achieve a uniformed hybridization reaction during 15 min. After hybridization, the slides were washed for 2 min in 2 × SSC/0.1% (v/v) SDS; 2 min in 0.2 × SSC/0.1% (v/v) SDS and 2 min in 0.2 SSC at room temperature, and then dried under a nitrogen stream. In experiments reported on Figure 8, hybridization was carried out with an automatic hybridization room (Discovery from Ventana Medical System, Inc). Prehybridization was carried out with a freshly prepared solution of 1% BSA, 2 × SSC, 0.2% SDS during 1 h 30 at 42°C. After automatic washing according to manufacturer instruction, the slides were hybridized for 8 hr in a 200 μL of ChipHybeTM buffer (Ventana Medical System, Inc) containing 20 μl of labeled and purified cDNA.Fluorescence imaging. Fluorescent images were captured with the laser scanner GenePix 4000 B from Axon at appropriate sensitivity levels of photomultiplier (PMT). The scanner run and collects data in 5 μm steps, then averages the data into 10 μm pixels. For correct data treatment, only features bigger than 10 μm were used.
Authors' contributions
C.T. and V.L carried out the technological and biological part of the work and wrote the first draft of the manuscript. E.T. carried out the chemical part of the study. JF and CV conceived of the study, participated in the design of the experiments, and finalized the writing of the manuscript. All authors read and approved the final manuscript.
Acknowledgements
This work was supported in part by the EC-funded project NaPa (Contract n° NMP4-CT-2003-500120, to C.V.) and by Genopole Toulouse Midi-Pyrénées (to J.F.). The content of this work is the sole responsibility of the authors.
==== Refs
Hughes TR Mao LM Jones AR Burchard J Marton MJ Shannon KW Expression profiling using microarrays fabricated by an ink-jet oligonucleotide synthesizer Nat Biotechnol 2001 19 342 347 11283592 10.1038/86730
Lipshutz RJ Fodor SPA Gingeras TR Lockhart DJ High density synthetic oligonucleotide arrays Nat Genet 1999 14 1675 1680
Renault JP Bernard A Bietsch A Michel B Bosshard HR Delamarche E Kreiter M Hecht B Wild UP Fabricating Arrays of Single Protein Molecules on Glass Using Microcontact Printing J Phys Chem B 2003 107 703 711 10.1021/jp0263424
Michel B Bernard A Bietsch A Delamarche E Geissler M Juncker D Kind H Renault JP Rothuizen H Schmid H Schmidt-Winkel P Stutz R Wolf H Printing meets lithographiy: Soft approaches to high-resolution patterning IBM J Res & Dev 2001 45 697 719
Malaquin L Vieu C Clivia M Sotomayor Torres Using PDMS as Thermocurable Resist for a Mold Assisted Imprint Process Alternative Lithography "unleashing the potentials of Nanotechnology" 2003 Chapter 8 Kluwer Academic Publishers Boston/Dordrecht/London 169 199
Lange S Benes V Kern D Hörber H Bernard A Microcontact Printing of DNA Molecules Anal Chem 2004 76 1641 1647 15018562 10.1021/ac035127w
Le Berre V Trevisol E Dagkessamanskaia A Sokol S Caminade AM Majoral JP Meunier B François J Dendrimeric coating of glass slides for sensitive DNA microarrays analysis Nucleic Acids Research 2003 31 e1 8 12527790 10.1093/nar/gng088
Trévisiol E Leberre V Leclaire J Pratviel G Caminade AM Majoral JP François J Meunier B Dendrislides, Dendrichips: a Simple Chemical functionalization of glass slides with Phosphorus Dendrimers as an effective Mean for the Preparation of Biochips New J Chem 2003 27 1713 1719 10.1039/b307928g
Huajian G Yong K Daxiang C Cengiz S Spontaneous Insertion of DNA Oligonucleotides into carbon Nanotubes Nano Letters 2003 3 471 473 10.1021/nl025967a
Fain B Xia Y Levitt M Determination of Optimal Chebychev-expanded hydrophobic discrimination function for globular protein IBM Journal of research and development 2001 45 525 532
Hacia JG Resequencing and mutational analysis using oligonucleotide microarrays Nature Genet 1999 21 42 47 9915500 10.1038/4469
Hacia JG Collins FS Mutational analysis using oligonucleotide microarrays J Med Genet 1999 36 730 736 10528850
Praekelt UM Meacock PA HSP12, a new small heat shock gene of Saccharomyces cerevisiae: analysis of structure, regulation and function Mol Gen Genet 1990 223 97 106 2175390 10.1007/BF00315801
|
15992404
|
PMC1184096
|
CC BY
|
2021-01-04 16:38:05
|
no
|
J Nanobiotechnology. 2005 Jul 1; 3:7
|
utf-8
|
J Nanobiotechnology
| 2,005 |
10.1186/1477-3155-3-7
|
oa_comm
|
==== Front
Proteome SciProteome Science1477-5956BioMed Central London 1477-5956-3-51594147510.1186/1477-5956-3-5MethodologyEnhanced detergent extraction for analysis of membrane proteomes by two-dimensional gel electrophoresis Churchward Matthew A [email protected] R Hussain [email protected] John C [email protected] Kimberly K [email protected] Jens R [email protected] Dept. of Physiology and Biophysics, University of Calgary Faculty of Medicine, 3330 Hospital Drive NW, Calgary AB, T2N 4N1, CANADA2 Dept. of Biochemistry and Molecular Biology, University of Calgary Faculty of Medicine, 3330 Hospital Drive NW, Calgary AB, T2N 4N1, CANADA3 Hotchkiss Brain Institute, University of Calgary Faculty of Medicine, 3330 Hospital Drive NW, Calgary AB, T2N 4N1, CANADA2005 7 6 2005 3 5 5 23 1 2005 7 6 2005 Copyright © 2005 Churchward et al; licensee BioMed Central Ltd.2005Churchward et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The analysis of hydrophobic membrane proteins by two-dimensional gel electrophoresis has long been hampered by the concept of inherent difficulty due to solubility issues. We have optimized extraction protocols by varying the detergent composition of the solubilization buffer with a variety of commercially available non-ionic and zwitterionic detergents and detergent-like phospholipids.
Results
After initial analyses by one-dimensional SDS-PAGE, quantitative two-dimensional analyses of human erythrocyte membranes, mouse liver membranes, and mouse brain membranes, extracted with buffers that included the zwitterionic detergent MEGA 10 (decanoyl-N-methylglucamide) and the zwitterionic lipid LPC (1-lauroyl lysophosphatidylcholine), showed selective improvement over extraction with the common 2-DE detergent CHAPS (3 [(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate). Mixtures of the three detergents showed additive improvements in spot number, density, and resolution. Substantial improvements in the analysis of a brain membrane proteome were observed.
Conclusion
This study demonstrates that an optimized detergent mix, coupled with rigorous sample handling and electrophoretic protocols, enables simple and effective analysis of membrane proteomes using two-dimensional electrophoresis.
==== Body
Background
Historically, the proteomic analysis of hydrophobic membrane proteins has been considered to be difficult within the bounds of conventional protocols for two-dimensional gel electrophoresis (2-DE). The nature of first dimension isoelectric focusing (IEF) requires that proteins be thoroughly solubilized as they are subjected to an electric field in which they migrate to their isoelectric point, by definition the state of lowest possible net charge and thus lowest solubility in aqueous environments. In addition to being highly hydrophobic, many integral membrane proteins tend to be very large: human Ca2+ channels have 24 transmembrane helices and are typically > 200 kDa [1], and tyrosine kinase receptors are frequently > 100 kDa [2,3]. This leads to two major problems in the preparation of membrane protein samples for 2-DE. First, effectively extracting membrane proteins into a detergent that is IEF compatible. Second, maintaining protein solubility throughout loading onto IPG strips and the subsequent first dimension IEF separation. Although highly efficient membrane protein extractions are routinely carried out with a detergent such as SDS for one-dimensional PAGE, SDS is incompatible with IEF due to the charged head group. To overcome this, SDS solubilized samples often undergo solvent or acid precipitation to remove or reduce SDS and lipids. Despite these harsh treatments and even subsequent treatment of the precipitate with a strong base [4], delipidation by solvent extraction is often cited as enhancing protein recovery [5,6] without discussion of the loss or modification of proteins during precipitation. For example, highly hydrophobic proteins (such as proteolipids) and proteins with particular post-translational modifications (such as palmitoylation) are capable of partitioning into the solvent phase [7-9], and TCA treatment can cause acid hydrolysis of proteins or alter post-translational modifications. Additionally, some early general problems with effectively separating hydrophobic proteins by 2-DE have led to widespread general disregard for the analysis of membrane proteins, particularly in the development of alternate proteomic approaches [4,5,10,11].
Since membrane proteins comprise approximately 30% of human proteins[12], and may account for substantially more cellular functions, the focus on soluble proteins in so-called 'full' proteomic analyses is somewhat concerning. There is evidence that optimization of extraction conditions by alteration of buffers, chaotropes, and detergents is sufficient to reliably achieve high-resolution maps of membrane proteins [13-16]. To this end we have sought simple alternatives to optimize the detergent conditions used to extract proteins from native membranes by systematic analysis of the solubilization properties of a wide range of commercially available non-ionic and zwitterionic detergents and a range of natural and synthetic detergent-like lipids [17-19]. Using proven synthetic detergents, together with more native lipophilic agents, we find that combinations of these reagents generally improve the resolution of membrane proteomes analyzed by 2-DE, providing for select improvements in the yields of specific proteins. Optimization of conditions for particular samples remains a key to any successful analysis [20-22].
Results & Discussion
1D SDS-PAGE of RBC membrane
Analysis of RBC extracts using 1D SDS-PAGE allowed for the rapid screening of a large number of extraction reagents (including glycerols, lipids, fatty acids, and isoprenoids), providing results that could be interpreted qualitatively based on the selective increase and decrease of protein banding patterns relative to control extractions with CHAPS or SDS (Fig. 1). For example, band III, a large protein with multiple transmembrane spanning domains [23] could be clearly distinguished in SDS extracts at an apparent MW of ~110 kDa, compared to those made with CHAPS. Another band, at apparent MW of 28 kDa, was also observed in the SDS but not the CHAPS extract. Based on these simple criteria, detergents were selected that gave improved banding patterns over the CHAPS control extraction. An initial working series of effective detergents was thus identified for further testing (Table 1), and these were then used to extract RBC membrane samples for subsequent analysis by 2-DE. Notably LPC, the N-methylglucamide detergents MEGA 8, 9, and 10 and the sulfobetaine-based detergents ASB-14 and SB 3–10 showed improvements in the 1D banding pattern relative to the CHAPS control. A selection of natural source lipids were also tested, including lysophsphatidylglycerol (LPG), lysophosphatidylethanolamine (LPE), lysophosphatidylcholine (LPC, from egg), lysophosphatidylserine (LPS, from bovine brain), and cardiolipin (bovine heart). Although generally comparable, and of some selective use in extractions, most of these natural lipids proved to be of limited general usefulness as they are charged at neutral pH, and thus inherently incompatible with IEF. This limitation does not obviate the potential application of these lipids as extraction agents for use with alternate protein separation paradigms.
Table 1 Summary of detergents tested using systematic 1D SDS-PAGE analysis. Overall extraction efficacy analyzed by 1D SDS-PAGE or 2-DE separation is expressed qualitatively relative to SDS extraction (for 1D analysis) or CHAPS extraction (for 2-DE). + indicates compatibility but poor perfomance, ++ indicates similar or slightly worse than CHAPS extraction, +++ indicates performance equal to or better than CHAPS, – indicates incompatibility.
Detergent 1D-PAGE 2-DE Comments & Rationale
SDS +++ - IEF incompatible
CHAPS ++ +++ Poor extraction of hydrophobic and high molecular weight proteins
trans, trans-farnesol +++ ++ Natural isoprenoid
MEGA-8a +++ ++ Group of nonionic detergents commonly used for protein purification [35,36]
MEGA-9b +++ ++
MEGA-10c +++ +++
amidosulfobetaine-14 (ASB-14) +++ ++ Sulfobetaine-based detergents reported to improve membrane protein extraction [24-26]
Zwittergent® 3–10/SB 3–10d +++ +++
LPC (synthetic, lauroyl chain)e +++ +++ Zwitterionic lysophospholipid
LPC (egg, mixed chain)e ++ ++ Zwitterionic lysophospholipid
LPS (bovine brain)f ++ - Anionic lysophospholipid, incompatible with IEF
LPE (egg, mixed chain)g - - Zwitterionic lysophospholipid; low solubility in high urea buffer
LPG (egg, mixed chain)h ++ - Anionic lysophospholipid, incompatible with IEF
LPA (egg, mixed chain)i ++ - Anionic lysophospholipid, incompatible with IEF
cardiolipin (bovine heart) ++ - Anionic lipid, incompatible with IEF, low solubility in high urea buffer
5,7-docosadiynoic acid - - Synthetic fatty acid; low solubility in high urea buffer
lauric acid +++ ++ Medium chain fatty acid; low solubility in high urea buffer
free fatty acids (mixed) - - Mixed natural fatty acids; low solubility in high urea buffer
DODAPj +++ - Cationic lipid used as a transfection reagent [37], IEF incompatible
1-oleoyl-sn-glycerol + - Uncharged monoacylated lipid, very low solubility in high urea buffer
C12E8k + - Nonionic detergent used to study membrane proteins [38]
DL-α-O-benzylglycerol + - Amphipathic cyclic glycerol conjugate
tryptophol ++ - Amphipathic heterocyclic metabolite of tryptophan; ionizable at low pH
a octanoyl-N-methylglucamide
b nonanoyl-N-methylglucamide
c decanoyl-N-methylglucamide
d N-decyl-N,N-dimethyl-3-ammonio-1-propanesulfonate
e L-α-lysophosphatidylcholine
f L-α-lysophosphatiylserine
g L-α-lysophosphatidylethanolamine
h L-α-lysophosphatidylglycerol
i L-α-lysophosphatidic acid
j 1,2-dioleoyloxy-3-(dimethylamino)propane
k octaethylene glycol monododecyl ether
Figure 1 Composite of 1D SDS-PAGE analyses of RBC ghost membranes extracted with A) 4% CHAPS, B) 2% SDS, C) 2% LPC, D) 2% lauric acid, E) 2% trans, trans-farnesol, F) 2% MEGA 8, G) 2% MEGA 9, H) 2% MEGA 10, I) 2% 1,2 dioleoyloxy -3-(dimethylamino)propane, J) 2% SB 3–10 (Sigma), K) 2% SB 3–10 (Calbiochem), L) C12E8, M)1-oleoyl-sn-glycerol, N) DL-α-O-benzylglycerol. Arrows indicate notable differences between extractions including 1, the multiple transmembrane spanning protein band III.
2-DE analysis of Red Blood Cell membrane
Initially, red blood cell (RBC) membranes were extracted with a 2-DE buffer containing 4% total CHAPS (our standard concentration) or 1–2% total detergent, due to the relatively lower solubility of most test detergents compared to the highly soluble CHAPS. Extraction of RBC membranes with synthetic LPC (lauroyl chain, Sigma) and the zwitterionic detergent MEGA 10 under these conditions resulted in areas of selective improvement in resulting 2-DE patterns relative to the control extracts in CHAPS (data not shown). 2-DE of samples extracted with SB 3–10 yielded similar protein maps to CHAPS, however this detergent was difficult to solubilize into high urea buffer, as previously reported [24,25]. However, contrary to earlier reports [24,26], samples extracted with ASB-14 did not show improvement over either CHAPS or SB 3–10 (data not shown). In order to both appropriately account for more general effects of detergent concentration and take advantage of the high solubility and efficient solubilizing properties of CHAPS, both LPC and MEGA 10 were used to extract RBC membranes (Fig. 2), mouse brain membranes (Fig. 3), and mouse liver membranes (Fig. 5) as mixtures of 3% CHAPS : 1% alternate detergent, for a total 4% detergent.
Figure 2 2-DE of RBC ghost membranes extracted with A) 4% CHAPS, B) 3% CHAPS : 1% LPC, C) 3% CHAPS : 1% MEGA 10, D) 3% CHAPS : 0.5% LPC : 0.5% MEGA 10. Extractions were carried out in buffer with 8 M urea, 2 M thiourea, protease inhibitor cocktail, and the indicated detergent for 1 hour on ice. Gels are representative of three independent experiments. Roman numerals indicate areas of improvement including i, the multiple transmembrane spanning protein band III.
Figure 3 2-DE of mouse brain membranes extracted with A) 4% CHAPS, B) 3% CHAPS : 1% LPC, C) 3% CHAPS : 1% MEGA 10, D) 3% CHAPS : 0.5% LPC : 0.5% MEGA 10. Extractions were carried out as for Fig. 2. Gels are representative of three independent experiments. Areas defined with Roman numerals are shown in Fig. 4.
Figure 5 2-DE gels of mouse liver membranes extracted with A) 4% CHAPS, B) 3% CHAPS : 1% LPC. Extractions were carried out as for Fig. 2. Gels are representative of three independent experiments. Arrow indicates specific differences between gels.
RBC membranes extracted with the 4% total detergent mixtures of CHAPS and LPC showed general evidence of improved spot densities (Fig. 2B), as well as specific improvements in terms of reduced horizontal streaking in the intermediate molecular weight region (Fig. 2A). Automated spot detection and quantitative comparative analysis using Progenesis Workstation software identified specific changes in the protein pattern. Specific areas of the gels showed improvement relative to parallel 4% CHAPS control gels (Fig. 2Bi–iv). In particular, a prominent spot corresponding to band III was clearly observed [15,16], as well as a 2.2 ± 0.1 – fold increase in the density of a string of spots relative to the CHAPS extract (Fig. 2Bii). Three additional unique spots were observed when extracted with 3% CHAPS : 1% LPC (Fig 2Biii–iv). RBC membrane samples extracted with 3% CHAPS : 1% MEGA 10 (Fig. 2C) yielded protein maps of generally equivalent resolution to both the 3% CHAPS : 1% LPC and the 4% CHAPS maps, but did not resolve the protein band III as effectively as 3% CHAPS : 1% LPC.
In order to examine the overlapping effects of both these test detergents, but ameliorate the observed losses of protein, 0.5% of each was mixed with 3% CHAPS and tested in the extraction and 2-DE analysis of RBC membrane proteins (Fig. 2D). Extraction with 3% CHAPS : 0.5% LPC : 0.5% MEGA 10 does not yield the extent of differences identified in the 3% CHAPS : 1% LPC extracted condition, although the maps show general improvements over the control CHAPS condition that correlate with the improvements seen in the two individual detergent extractions (Figs. 2B, C). The density of the indicated string of proteins was increased an average of 1.7 ± 0.2 – fold over CHAPS (Fig. 2Dii). In general then, the addition of LPC to the extraction buffer enhances both protein recovery and resolution in the subsequent 2D protein maps.
Our initial findings using the RBC membrane as a model system lead us to expand the analyses to additional tissue types. Mouse brain membranes [27] and mouse liver membranes were chosen due to their availability, and broad international interest in improved analyses of these tissue proteomes.
2-DE analysis of mouse brain membrane
Adult mouse brain membrane samples were subjected to the final four extraction conditions (Fig. 3) in order to further test the results obtained in RBC membranes (Fig. 2). Overall the results were quite similar to those obtained in the tests on RBC membranes. Extraction of mouse brain membranes with 3% CHAPS : 1% LPC (Fig. 3B) showed improvement of spot number, density and resolution compared to extraction with 4% CHAPS alone (Fig. 3A); quantitative analysis indicated specific areas of significant improvement (Fig. 4). Automated analysis identified 13 ± 3 novel spots that were reproducibly detected primarily in the low molecular weight and basic extreme regions of the gel (Fig. 4B; blue arrows indicate novel spots). Additionally, 5 spots were identified that significantly increased in volume an average 7.0 ± 3.4 -fold, and increased in density 2.8 ± 0.9-fold compared to the 4% CHAPS condition (Fig. 4B; green arrows indicate increased recovery). Of the 15 ± 2 novel spots detected in the 3% CHAPS : 1% MEGA 10 condition (Fig. 4C), most were also observed in the 3% CHAPS : 1% LPC condition. Overall, of the same 5 spots showing increased recovery, the volume increased 5.8 ± 2.5-fold, while density increased 3.2 ± 0.9-fold (Fig 4C; green arrows). Extraction of mouse brain membrane with 3% CHAPS : 0.5% LPC : 0.5% MEGA 10 (Fig. 3D) showed an additive effect on spot number. Spots recovered in both 3% CHAPS : 1% LPC and 3% CHAPS : 1% MEGA 10 were also detected in the combined extraction system. 13 ± 1 novel spots were detected relative to control, and the 5 previously identified spots increased in volume 6.4 ± 0.4-fold and density was increased 2.6 ± 0.6-fold (Fig 4D; green arrows). The nature of the recovery of these protein spots in 3% CHAPS : 0.5% LPC : 0.5% MEGA 10 reveals the specific action of the two detergents – LPC and MEGA 10 working in concert. Only one selective loss of a protein spot was observed in relation to this recovery of unique spots (Fig. 4Ai); this loss is the result of a specific action shared by the two alternate detergents as opposed to a result of the difference in CHAPS concentration during extraction since this protein was not recovered even after extraction with 5% total detergent (4% CHAPS : 0.5% LPC : 0.5% MEGA 10) (data not shown). This loss implies some specific action of the alternate detergents that prevent the extraction of this particular protein, or possibly an alteration in the electrophoretic mobility of this protein in the first dimension by means of increasing or decreasing the number of exposed ionizable residues. Together, the results of the RBC membrane and mouse brain membrane extractions show that simple combinations of zwitterionic detergents (CHAPS and MEGA 10) with a zwitterionic lipid (LPC) are generally more effective at extracting membrane proteins and maintaining protein solubility during first dimension IEF than are standard CHAPS-based extraction conditions.
Figure 4 Enlargement and contrast of selected regions after 2-DE of mouse brain membranes (see areas defined in Fig 3). Areas i-vi show selective increases in spot number, resolution, and density. Samples were extracted with A) 4% CHAPS, B) 3% CHAPS : 1% LPC, C) 3% CHAPS : 1% MEGA 10, D) 3% CHAPS : 0.5% LPC : 0.5% MEGA 10. Results are representative of three independent experiments. Green arrows indicate spots showing increased volume and density, red arrows indicate decrease, blue arrows indicate novel spots.
Additional 2-DE Analyses
Interestingly, extracting mouse liver membranes with the same detergent combinations described above resulted in protein maps that were highly similar, with very limited improvements. Automated analysis indicated almost complete overlap of the resulting 2-DE protein patterns (Fig 5), with the specific and substantial recovery of one additional protein spot. We interpret the marked similarity in these liver protein profiles, relative to the differences seen in the RBC and brain samples, to be due to variability between tissues in terms of relative homogenization/extraction efficiency and compatibility with our current buffer system.
To control for possible differences arising from the changing CHAPS concentration in these test extraction buffers, mouse brain membranes were also extracted with 5% total detergent (5% CHAPS or 4% CHAPS : 0.5% LPC : 0.5% MEGA 10) and analyzed in parallel with membranes extracted with 4% total detergent. No significant difference in overall spot pattern or specific differences as described above was observed between the 5% and the 4% total detergent mixtures (data not shown), indicating that the differences described here are specifically attributable to the addition of LPC and MEGA 10 as solubilizing agents. Indeed, overall, membrane protein patterns were generally of somewhat lower resolution when the CHAPS concentration or total detergent concentration was increased to 5%.
Protein Quantification
During initial experiments we found total protein load to be the most significant variable confounding quantitative analyses. As such, great care was taken to ensure that the analyses meaningfully tested protein extraction and solubilization efficiency, in isolation from complicating variables. Simply, the goal was to compare reagents and conditions, not to compare different final total protein loads by 2-DE. Initially many protein samples were quantified using a modified Folin total protein assay (RC DC Protein Assay kit, BioRad). Colourimetric assays of this type (eg. Bradford, Lowry, BCA, and so forth) perform acceptably under many circumstances requiring routine normalization of a series of very similar samples. However one of several limitations of such total protein assays is a marked sensitivity to interfering substances, including components of typical IEF solubilization solutions such as detergents, reducing agents, and urea. In our experiments, detergents and detergent concentrations were systematically altered and combined. Not unexpectedly, we observed substantial variability in the results of the total protein assay, depending upon the solubilizing reagents present. The complications of applying systematic corrective controls, or of preparing separate standard curves for each of the solubilization conditions tested, simply increased the potential for error. Regardless, separate standard curves are not even feasible in the case of the RC DC assay, as urea causes a saturating false positive signal.
We have found that the EZQ Protein Quantitation kit (Molecular Probes) is insensitive to the nature and concentrations of detergent in all samples tested. In this assay format, the immobilized protein sample is washed exhaustively with methanol to remove components of the solubilization solution prior to addition of the fluorescent protein detection reagent. Thus, the chemistry of the assay proceeds in the absence of potentially confounding contaminants. In extensive comparisons, there were no significant differences in standard protein assay curves regardless of the type or quantity of detergent included (data not shown). Additionally, the method proved quite sensitive (routine detection of 0.030 μg of total protein/spot, or 15 μg/ml); this is fully 10-fold more sensitive and requires 4-fold less material than the RC DC Assay. Thus, as the chemistry of the assay was not altered under our different experimental conditions, we are confident that the improvements observed in our final protein maps were truly the result of differences in extraction and solubilization efficiency, and not artifacts generated by erroneous total protein assays leading to inconsistent total protein IEF loads between different test conditions. Although the EZQ protein assay certainly has its caveats, not least of which is cost, it does offer distinct benefits that support its utility in these and other ongoing proteomic analyses.
Conclusion
In order to optimize recovery of hydrophobic proteins for 2-DE, we have sought a simple, direct solution to the problem of protein extraction and solubility during IEF. The systematic screening and combination of commercially available detergents offers a direct, inexpensive, and convenient method for optimizing the conditions of IEF without entering into the complexities of a systematic synthesis of new detergents based on specific base molecules, or the potential losses or modification of proteins associated with solvent extraction techniques. Coupled with our ability to effectively analyze membrane proteomes using 2-DE [27] the resulting findings should also prove of use in defining optimized combinations of extraction reagents for use with alternate protein separation protocols.
Based on the hypothesis that highly lipophilic molecules (albeit at lower total concentrations than can be achieved with the more standard detergents), might better mimic native lipid-membrane protein interactions and thus improve protein solubilization, we found that LPC can substantially augment the extraction of membrane proteins from different sources. This finding does not obviate the need for optimization of extraction and 2-DE conditions for different samples, but does provide a powerful, widely available and reasonably priced alternative that can be readily tested in parallel with more routine solubilization reagents. Rigorous testing of protein assays ensured that these findings reflect a true effect on extraction and protein solubility, rather than an artifact of inconsistent protein loads between different 2-DE analyses. Notably, LPC and MEGA 10 provided particularly marked improvements in the resolution of the mouse brain membrane proteome.
Methods
Reagents
L-α-lysophosphatidylcholine lauroyl, urea, tris acetate, lauric acid, pH 3–10 ampholytes, ammonium persulfate, decyl-N,N-dimethyl-3-ammonio-1-propanesulfonate (SB 3–10), amidosulfobetaine-14 (ASB-14), DL-α-O-benzylglycerol, tributylphosphine (TBP), HEPES, sodium orthovanadate, staurosporine, cantharidin, and components of the broad spectrum protease inhibitor cocktail [28] were purchased from Sigma (St. Louis, Missouri). IPG strips (pH 3–10), 30% acrylamide/bisacrylamide solution, low melting agarose, Sypro Ruby, 10×TGS running buffer, RC DC Protein Assay Kit, bovine γ-globulin, and SDS were from BioRad (Hercules, California). EZQ Protein Quantitation Kit was from Molecular Probes (Eugene, OR), Zwittergent® 3–10 was from Calbiochem (La Jolla, California), and CHAPS was from Anatrace (Maumee, Ohio). 1,2-dioleoyloxy-3-(dimethylamino)propane, 5,7-docosadiynoic acid, and 1-oleoyl-sn-glycerol were from Toronto Research Chemicals (Toronto, Ontario). Bovine brain L-α-lysophosphatidylserine, egg L-α-lysophosphatidylcholine, egg L-α-lysophosphatidylethanolamine, egg L-α-lysophosphatidylglycerol, egg L-α-lysophosphatidic acid, bovine heart cardiolipin, and free fatty acids were from Doosan Serdary Research (Toronto, Ontario). Thiourea was from Fisher Scientific (Hampton, New Hampshire), and PBS, DTT, octanoyl-N-methylglucamide (MEGA 8), nonanoyl-N-methylglucamide (MEGA 9), decanoyl-N-methylglucamide (MEGA 10), TEMED, glycerol, 40% acrylamide solution, and octaethylene glycol monododecyl ether (C12E8) were from Bio Basic Inc. (Markham, Ontario). Narrow range ampholytes (pH 2.5–4, 3.5–5, 5–7, 7–9, and 8–9.5) were from Fluka (Buchs, Switzerland), and tryptophol and trans, trans-farnesol was from Aldrich (St. Louis, Missouri). All other chemicals were of at least analytical grade.
Red Blood Cell membrane preparation
Packed RBC were obtained from Canadian Blood Services, (Calgary, AB) and washed 3× with isotonic buffer (20 mM sodium phosphate pH 7.4, 0.9% NaCl). RBC ghosts were prepared according to the method of Chernomordik [29] with slight modifications. Cells were lysed osmotically in hypotonic lysis buffer (5 mM sodium phosphate pH 7.4, protease inhibitor cocktail [28], 5 mM DTT) for 20 minutes on ice. The lysate was flash frozen in a dry ice / ethanol bath, thawed, and membranes were collected by centrifugation (3000×g, 20 min, 4°C). Pellets were washed with wash buffer (20 mM sodium phosphate pH 8.5, protease inhibitor cocktail, 5 mM DTT) until supernatants were clear, and then subjected to a second round of hypotonic lysis and freeze-thaw. After washing until supernatants were clear, membranes were collected by centrifugation (3000×g, 40 min, 4°C), suspended in a minimal volume of wash buffer, and stored at -80°C. Before extraction, membrane isolates were washed with PBS containing protease inhibitors (PBS-PI) and pelleted (3 hours, 120 000×g, 4°C).
Membrane preparations from mammalian tissues
Membranes were isolated as previously described [27]. Briefly, mouse brains or livers were flash frozen after dissection and stored at -80°C until needed. For the isolation of all cellular membranes, we applied a simple physical separation / fractionation protocol. Briefly, frozen tissues were thawed in hypotonic lysis buffer (20 mM HEPES pH 7.4, protease inhibitor cocktail, 10 mM sodium orthovanadate, 4 μM staurosporin, 4 μM cantharidin) and manually homogenized on ice with a polyethylene pestle in a 1.5 mL microcentrifuge tube. The homogenate was subjected to one round of freeze-thaw (-80°C), before being combined with an equal volume of 2×PBS to restore isotonicity. Membranes were collected by ultracentrifugation (3 hours, 120 000×g, 4°C), and were washed twice; pellets were resuspended in PBS-PI for each wash and collected by ultracentrifugation, as described above.
Detergent Extractions
Detergent extraction buffers were prepared for 1D (7 M urea, 2 M thiourea, 9 mM Tris acetate pH 7.0, protease inhibitor cocktail, and detergent as indicated) or 2-DE (IEF buffer 1 containing 8 M urea, 2 M thiourea, protease inhibitor cocktail, and detergent as indicated [27]). Membrane pellets were resuspended by pipetting and vortexing. Extractions were incubated for 1 hour on ice, with periodic vortexing. Any insoluble material was separated by ultracentrifugation as previously described. Solubilized samples were assayed for total protein content using the EZQ Protein Quantitation Kit (Molecular Probes, Eugene, OR).
Protein Quantification
Total protein was assayed using either the EZQ Protein Quantitation Kit or the RC DC Protein Assay Kit (BioRad, Hercules, CA). The RC DC assay was carried out according to manufacturers instructions in 96-well plates and absorbance was measured using the Wallac Victor2 Multilabel HTS Counter (PerkinElmer Life Sciences, Boston, MA). EZQ Protein Quantitation was carried out essentially according to manufacturers instructions except fluorescence was recorded by imaging on the Proexpress multiwavelength fluorescent imager (PerkinElmer, Boston MA) and spot fluorescence was quantified using ImageQuant 5.2 software (Molecular Dynamics, Sunnyvale, CA).
1D SDS-PAGE
1D SDS-PAGE was performed in mini gel format using the BioRad Protean II Electrophoresis system, essentially as described [30] with minor modifications [31]. Samples were normalized to 2 mg/ml in the appropriate extraction buffer, and then diluted 1:1 (v/v) with 2 × SDS sample buffer [30]. 10 μg total protein was loaded per well on 12.5%T separating gels with 5%T stacking gels, buffered with 375 mM Tris (pH 8.8) as described [31]. Gels were run at 125 V for 10 min to stack proteins, and then the voltage was reduced to 90 V to completion [32].
2-DE
Samples for IEF were normalized to 2 mg/ml with the appropriate IEF buffer, then combined 1:1 (v/v) with an ampholyte-containing IEF buffer (8 M urea, 2 M thiourea, 1% pH 3–10 broad range ampholytes, 0.2% each narrow range ampholytes (pH 2.5–4, 3.5–5, 5–7, 7–9, and 8–9.5) and detergent as indicated [27]), to introduce a working concentration of ampholytes to the sample.
Samples were sequentially reduced and alkylated essentially according to Herbert et al. [33,34] with some minor modifications. Briefly, the sample was reduced by the addition of TBP and DTT to final concentrations of 2.3 mM and 45 mM DTT, respectively, and incubated for 1 hour at 25°C. The reduced sample was then alkylated with 230 mM acrylamide monomer for 1 hour at 25°C. Immediately following alkylation, the sample was loaded onto IPG strips for passive hydration at 25°C (12 hours). IEF was carried out at 15°C using the BioRad Protean IEF Cell; voltage was ramped linearly to 4000 V (2 hours) and IEF was carried out at 4000 V (constant) for 37500 Vhours. After focusing, IPG strips were equilibrated essentially according to the manufacturer's instructions by sequential immersion in equilibration buffer (6 M urea, 2% SDS (w/v), 20 % glycerol (w/v), and 375 mM Tris pH 8.8) containing 130 mM DTT for 10 minutes, followed by equilibration buffer with 350 mM acrylamide monomer for 10 minutes. Following equilibration, IPG strips were loaded onto 12.5%T separating gels with 5%T stacking gels (buffered as described for 1D) and sealed in place with an agarose overlay (0.5% low melting agarose, 0.1% SDS and 375 mM Tris pH 8.8). SDS-PAGE was otherwise carried out as described for 1D SDS-PAGE.
Image analysis
After electrophoresis, gels were fixed in 10% methanol, 7% acetic acid for 1 hour, washed thoroughly with water and stained with Sypro Ruby overnight. Gels were visualized using the Proexpress multiwavelength fluorescent imager (PerkinElmer, Boston MA). Quantitative image analysis was performed using Progenesis Workstation 2004 (Nonlinear Dynamics, Cambridge, UK). Parallel sets of gels were warped and matched by automated analysis, and volumes were normalized to a single spot consistent in size, shape, density and location across all gels.
Abbreviations used
LPC (12:0 L-α-lysophosphatidylcholine), LPG (L-α-lysophosphatidylglycerol), LPE (L-α-lysophosphatidylethanolamine), LPS (L-α-lysophosphatidylserine), MEGA 8 (octanoyl-N-methylglucamide), MEGA 9 (nonanoyl-N-methylglucamide), MEGA 10 (decanoyl-N-methylglucamide), SB 3–10 (N-decyl-N,N-dimethyl-3-ammonio-1-propanesulfonate), ASB-14 (amidosulfobetaine-14), C12E8 (octaethylene glycol monododecyl ether), RBC (red blood cell)
Competing interests
The author(s) declare that they have no competing interests.
Authors' Contributions
MAC supervised the SDS-PAGE analysis, designed and carried out the 2-DE analyses, data acquisition and interpretation and prepared the manuscript. RHB contributed to the experimental design, sample preparation, 2-DE analyses, and participated in drafting the manuscript. JCL and KKH designed and carried out the SDS-PAGE analysis. JRC conceived and planned the study, and assisted in interpretation of data and final preparation of the manuscript. All authors read and approved the final manuscript.
Acknowledgements
The authors would like to thank Tiffany Rice, Tammy Wilson, and Dr. V. Wee Yong for kindly providing mouse tissues. We would also like to thank Jeff Lamb, Dr. Sina Ahmadi Pirshahid, Marlies Ernst, and all the members of the Coorssen lab for helpful discussions and advice. J.R.C. acknowledges support from the Canadian Institutes of Health Research, the Canada Foundation for Innovation, the Alberta Heritage Foundation for Medical Research, the Alberta Network for Proteomics Innovation, and the Heart and Stroke Foundation of Canada.
==== Refs
Perez-Reyes E Wei XY Castellano A Birnbaumer L Molecular diversity of L-type calcium channels. Evidence for alternative splicing of the transcripts of three non-allelic genes J Biol Chem 1990 265 20430 20436 2173707
Strausberg RL Feingold EA Grouse LH Derge JG Klausner RD Collins FS Wagner L Shenmen CM Schuler GD Altschul SF Zeeberg B Buetow KH Schaefer CF Bhat NK Hopkins RF Jordan H Moore T Max SI Wang J Hsieh F Diatchenko L Marusina K Farmer AA Rubin GM Hong L Stapleton M Soares MB Bonaldo MF Casavant TL Scheetz TE Brownstein MJ Usdin TB Toshiyuki S Carninci P Prange C Raha SS Loquellano NA Peters GJ Abramson RD Mullahy SJ Bosak SA McEwan PJ McKernan KJ Malek JA Gunaratne PH Richards S Worley KC Hale S Garcia AM Gay LJ Hulyk SW Villalon DK Muzny DM Sodergren EJ Lu X Gibbs RA Fahey J Helton E Ketteman M Madan A Rodrigues S Sanchez A Whiting M Madan A Young AC Shevchenko Y Bouffard GG Blakesley RW Touchman JW Green ED Dickson MC Rodriguez AC Grimwood J Schmutz J Myers RM Butterfield YS Krzywinski MI Skalska U Smailus DE Schnerch A Schein JE Jones SJ Marra MA Generation and initial analysis of more than 15,000 full-length human and mouse cDNA sequences Proc Natl Acad Sci U S A 2002 99 16899 16903 12477932 10.1073/pnas.242603899
Anzai T Shiina T Kimura N Yanagiya K Kohara S Shigenari A Yamagata T Kulski JK Naruse TK Fujimori Y Fukuzumi Y Yamazaki M Tashiro H Iwamoto C Umehara Y Imanishi T Meyer A Ikeo K Gojobori T Bahram S Inoko H Comparative sequencing of human and chimpanzee MHC class I regions unveils insertions/deletions as the major path to genomic divergence Proc Natl Acad Sci U S A 2003 100 7708 7713 12799463 10.1073/pnas.1230533100
Nandakumar MP Shen J Raman B Marten MR Solubilization of trichloroacetic acid (TCA) precipitated microbial proteins via naOH for two-dimensional electrophoresis J Proteome Res 2003 2 89 93 12643547 10.1021/pr025541x
Mastro R Hall M Protein delipidation and precipitation by tri-n-butylphosphate, acetone, and methanol treatment for isoelectric focusing and two-dimensional gel electrophoresis Anal Biochem 1999 273 313 315 10469505 10.1006/abio.1999.4224
Aivaliotis M Corvey C Tsirogianni I Karas M Tsiotis G Membrane proteome analysis of the green-sulfur bacterium Chlorobium tepidum Electrophoresis 2004 25 3468 3474 15490440 10.1002/elps.200406079
Hagopian K Preparative electrophoretic method for the purification of a hydrophobic membrane protein: subunit c of the mitochondrial ATP synthase from rat liver Anal Biochem 1999 273 240 251 10469495 10.1006/abio.1999.4219
Ruppert C Kavermann H Wimmers S Schmid R Kellermann J Lottspeich F Huber H Stetter KO Muller V The proteolipid of the A(1)A(0) ATP synthase from Methanococcus jannaschii has six predicted transmembrane helices but only two proton-translocating carboxyl groups J Biol Chem 1999 274 25281 25284 10464251 10.1074/jbc.274.36.25281
Skalidis G Trifilieff E Luu B Selective extraction of the DM-20 brain proteolipid J Neurochem 1986 46 297 299 3940288
Zuo X Speicher DW A method for global analysis of complex proteomes using sample prefractionation by solution isoelectrofocusing prior to two-dimensional electrophoresis Anal Biochem 2000 284 266 278 10964409 10.1006/abio.2000.4714
Vuong GL Weiss SM Kammer W Priemer M Vingron M Nordheim A Cahill MA Improved sensitivity proteomics by postharvest alkylation and radioactive labelling of proteins Electrophoresis 2000 21 2594 2605 10949135 10.1002/1522-2683(20000701)21:13<2594::AID-ELPS2594>3.3.CO;2-B
Wallin E von HG Genome-wide analysis of integral membrane proteins from eubacterial, archaean, and eukaryotic organisms Protein Sci 1998 7 1029 1038 9568909
Rabilloud T Blisnick T Heller M Luche S Aebersold R Lunardi J Braun-Breton C Analysis of membrane proteins by two-dimensional electrophoresis: comparison of the proteins extracted from normal or Plasmodium falciparum-infected erythrocyte ghosts Electrophoresis 1999 20 3603 3610 10612287 10.1002/(SICI)1522-2683(19991201)20:18<3603::AID-ELPS3603>3.0.CO;2-V
Santoni V Kieffer S Desclaux D Masson F Rabilloud T Membrane proteomics: use of additive main effects with multiplicative interaction model to classify plasma membrane proteins according to their solubility and electrophoretic properties Electrophoresis 2000 21 3329 3344 11079553 10.1002/1522-2683(20001001)21:16<3329::AID-ELPS3329>3.0.CO;2-F
Luche S Santoni V Rabilloud T Evaluation of nonionic and zwitterionic detergents as membrane protein solubilizers in two-dimensional electrophoresis Proteomics 2003 3 249 253 12627377 10.1002/pmic.200390037
Tastet C Charmont S Chevallet M Luche S Rabilloud T Structure-efficiency relationships of zwitterionic detergents as protein solubilizers in two-dimensional electrophoresis Proteomics 2003 3 111 121 12601803 10.1002/pmic.200390019
Kern R Joseleau-Petit D Chattopadhyay MK Richarme G Chaperone-like properties of lysophospholipids Biochem Biophys Res Commun 2001 289 1268 1274 11741332 10.1006/bbrc.2001.6093
Nekrasova E Sosinskaya A Natochin M Lancet D Gat U Overexpression, solubilization and purification of rat and human olfactory receptors Eur J Biochem 1996 238 28 37 8665947 10.1111/j.1432-1033.1996.0028q.x
Perez-Gil J Cruz A Casals C Solubility of hydrophobic surfactant proteins in organic solvent/water mixtures. Structural studies on SP-B and SP-C in aqueous organic solvents and lipids Biochim Biophys Acta 1993 1168 261 270 8323965
Navarre C Degand H Bennett KL Crawford JS Mortz E Boutry M Subproteomics: identification of plasma membrane proteins from the yeast Saccharomyces cerevisiae Proteomics 2002 2 1706 1714 12469340 10.1002/1615-9861(200212)2:12<1706::AID-PROT1706>3.0.CO;2-K
Santoni V Molloy M Rabilloud T Membrane proteins and proteomics: un amour impossible? Electrophoresis 2000 21 1054 1070 10786880 10.1002/(SICI)1522-2683(20000401)21:6<1054::AID-ELPS1054>3.3.CO;2-#
Witzmann F Jarnot B Parker D Dodecyl maltoside detergent improves resolution of hepatic membrane proteins in two-dimensional gels Electrophoresis 1991 12 687 688 1752256 10.1002/elps.1150120919
Lux SE John KM Kopito RR Lodish HF Cloning and characterization of band 3, the human erythrocyte anion-exchange protein (AE1) Proc Natl Acad Sci U S A 1989 86 9089 9093 2594752
Rabilloud T Gianazza E Catto N Righetti PG Amidosulfobetaines, a family of detergents with improved solubilization properties: application for isoelectric focusing under denaturing conditions Anal Biochem 1990 185 94 102 2344051 10.1016/0003-2697(90)90261-7
Navarrete R Serrano R Solubilization of yeast plasma membranes and mitochondria by different types of non-denaturing detergents Biochim Biophys Acta 1983 728 403 408 6337628
Chevallet M Santoni V Poinas A Rouquie D Fuchs A Kieffer S Rossignol M Lunardi J Garin J Rabilloud T New zwitterionic detergents improve the analysis of membrane proteins by two-dimensional electrophoresis Electrophoresis 1998 19 1901 1909 9740050 10.1002/elps.1150191108
Butt RH Coorssen JR Postfractionation for enhanced proteomic analyses: Routine electrophoretic methods increase the resolution of standard 2D-PAGE J Proteome Res 2005 4 982 91 15952746 10.1021/pr050054d
Coorssen JR Blank PS Tahara M Zimmerberg J Biochemical and functional studies of cortical vesicle fusion: the SNARE complex and Ca2+ sensitivity J Cell Biol 1998 143 1845 1857 9864359 10.1083/jcb.143.7.1845
Chernomordik LV Sowers AE Evidence that the spectrin network and a nonosmotic force control the fusion product morphology in electrofused erythrocyte ghosts Biophys J 1991 60 1026 1037 1760502
Laemmli UK Cleavage of structural proteins during the assembly of the head of bacteriophage T4 Nature 1970 227 680 685 5432063 10.1038/227680a0
Cannon-Carlson S Tang J Modification of the Laemmli sodium dodecyl sulfate-polyacrylamide gel electrophoresis procedure to eliminate artifacts on reducing and nonreducing gels Anal Biochem 1997 246 146 148 9056200 10.1006/abio.1997.2002
Coorssen JR Blank PS Albertorio F Bezrukov L Kolosova I Backlund PSJ Zimmerberg J Quantitative femto- to attomole immunodetection of regulated secretory vesicle proteins critical to exocytosis Anal Biochem 2002 307 54 62 12137779 10.1016/S0003-2697(02)00015-5
Herbert B Galvani M Hamdan M Olivieri E MacCarthy J Pedersen S Righetti PG Reduction and alkylation of proteins in preparation of two-dimensional map analysis: why, when, and how? Electrophoresis 2001 22 2046 2057 11465505 10.1002/1522-2683(200106)22:10<2046::AID-ELPS2046>3.0.CO;2-C
Herbert BR Molloy MP Gooley AA Walsh BJ Bryson WG Williams KL Improved protein solubility in two-dimensional electrophoresis using tributyl phosphine as reducing agent Electrophoresis 1998 19 845 851 9629925 10.1002/elps.1150190540
Lundahl P Greijer E Lindblom H Fagerstam LG Fractionation of human red cell membrane proteins by ion-exchange chromatography in detergent on Mono Q, with special reference to the glucose transporter J Chromatogr 1984 297 129 137 6548478 10.1016/S0021-9673(01)89036-1
Poole RC Halestrap AP Reconstitution of the L-lactate carrier from rat and rabbit erythrocyte plasma membranes Biochem J 1988 254 385 390 3178766
Semple SC Klimuk SK Harasym TO Dos SN Ansell SM Wong KF Maurer N Stark H Cullis PR Hope MJ Scherrer P Efficient encapsulation of antisense oligonucleotides in lipid vesicles using ionizable aminolipids: formation of novel small multilamellar vesicle structures Biochim Biophys Acta 2001 1510 152 166 11342155
Martin DW Active unit of solubilized sarcoplasmic reticulum calcium adenosinetriphosphatase: an active enzyme centrifugation analysis Biochemistry 1983 22 2276 2282 6222767 10.1021/bi00278a034
|
15941475
|
PMC1184097
|
CC BY
|
2021-01-04 16:37:15
|
no
|
Proteome Sci. 2005 Jun 7; 3:5
|
utf-8
|
Proteome Sci
| 2,005 |
10.1186/1477-5956-3-5
|
oa_comm
|
==== Front
Reprod Biol EndocrinolReproductive biology and endocrinology : RB&E1477-7827BioMed Central London 1477-7827-3-241596322610.1186/1477-7827-3-24ResearchThe placental RCAS1 expression during stillbirth Wicherek Lukasz [email protected] Marek [email protected] Artur [email protected] Tadeusz J [email protected] Krystyna [email protected] Tomasz [email protected] Andrzej [email protected] Magdalena [email protected] Department of Gynecology and Infertility of Jagiellonian University, 23 Kopernik Str, 31–501 Krakow, Poland2 Ist Department of Gynecology of the Medical University in Lublin, 16 Staszica Str, 20–081 Lublin Poland3 Radiology Department of Jagiellonian University, 19 Kopernik Str, 31–501 Krakow, Poland4 Department of Pathomorphology Jagiellonian University, 17 Grzegórzecka, 31–531 Krakow, Poland5 ENT Department of Jagiellonian University, 2 Sniadeckich Str, 31–531 Krakow, Poland2005 17 6 2005 3 24 24 13 3 2005 17 6 2005 Copyright © 2005 Wicherek et al; licensee BioMed Central Ltd.2005Wicherek et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background: Independently of the fetal death cause the beginning and course of stillbirth is closely related with the growing cytotoxic activity at the maternal-fetal interface. RCAS1 participates in the inhibition of maternal immune response during pregnancy. The alterations of RCAS1 protein expression in placental cells seem to determine the beginning of the labor and participate in the placental abruption. The aim of the present study was to investigate RCAS1 expression in placentas obtained following stillbirths or normal term births. Methods: RCAS1 expression was evaluated by Western blot method with the use of monoclonal anti-RCAS1 antibody in 67 placental tissue samples. Pregnant women were divided into four groups according to the mode of labor onset – spontaneous or induced, and the type of labor, stillbirth or labor at term. Placental beta-Actin expression was chosen as a control protein. Relative amounts of placental RCAS1 were compared with the use of Student's t-test, whereas beta-Actin control data were compared with the use of Mann-Whitney U test. Results: The average relative amount of RCAS1 was significantly lower in women with induced stillbirths than in women with induced labor at term. Similarly, significantly lower RCAS1 placental levels were observed in patients with spontaneous stillbirths than in women with spontaneous labor at term. Significant differences in RCAS1 expression were also observed with the respect to the beginning of the stillbirth: spontaneous and induced. Lowest RCAS1 placental levels were observed in women with spontaneous stillbirth. Conclusions: These preliminary results indicate that the alterations of RCAS1 expression in the human placenta may be involved in the changes of maternal immune system that take place during stillbirth.
==== Body
Background
Dynamic fetal growth observed between 20th and 28th week of pregnancy is accompanied by fetal maturation which enables further uncomplicated extra-uterine growth. The fetal development is possible due to the phenomenon of maternal immunological tolerance to fetal antigens. Severe disturbances of the fetal growth might in some cases result in fetal intrauterine death. Some of the most common fetal death causes constituting about 50% of all cases include placental maturation disorders and fetal congenital malformations [1]. Stillbirth may be considered a form of complicated vaginal labor. Independently of the fetal death cause the beginning and course of stillbirth is closely related to the growing cytotoxic activity at the maternal-fetal interface. However, this activity does not have to arise simultaneously with the fetal death. In some situations fetal death may not be accompanied by immediate clinical features of the onset of labor. Molecular changes in membrane proteins expressed by trophoblast cells usually lead to the beginning of normal term delivery but may also be found in stillbirth. These proteins are neccessary for the development of maternal immune tolerance phenomenon. Cellular membrane proteins expressed by trophoblast cells that seem to be responsible for suppressing CTLs, dNK and NK cells are: Fas/Fas-L, killer inhibitory receptors family (KIRs), tumor necrosis factor receptors family (TNF) and others [2-4]. Recently, a novel factor called RCAS1 has been described [4-6]. RCAS1 is a type II membrane protein, expressed in extra-villous cytotrophoblast, villi-histiocytes, uterine endometrium and in various human cancer cells [7-12]. This protein acts as a ligand for a putative receptor that may be present on normal peripheral lymphocytes such as T, B, and NK cells. RCAS1 inhibits the growth of receptor expressing cells in vitro and in vivo and induces apoptotic cell death [13]. Main functions of RCAS1 include avoiding of immune recognition and evading immune surveillance by inhibition of maternal immune attack against fetal antigens [4,5]. The alterations of RCAS1 protein expression in placental cells seem to determine the onset of labor and participate in the placental abruption [5,6]. The aim of our study was to evaluate the changes of RCAS1 placental level during the stillbirths.
Methods
2.1. The subjects
Relative amount of RCAS1 content was estimated in 67 placental tissue samples taken from normal term deliveries and from stillborn fetuses. Informed consent for the use of placental tissue was obtained from all patients. The approval from the Ethical Committee of the Jagiellonian University in Krakow (KBET/379/13/2003) for this research program was also granted. Stillborns were defined as dead-born fetuses either upon completion of 23 weeks of gestation, or when the fetal weight was over 400 grams. Autopsy was performed in Pathomorphology Department of Jagiellonian University in all cases. Based on the autopsy report review we included cases of intrauterine fetal death caused by fetal congenital malformations. According to the onset of stillbirth patients were devided into four groups. The first group included 6 women in whom labor was induced following the recognition of the intrauterine fetal death. The second group consisted of 5 women with stillbirths with spontaneous onset of labor. Two control groups consisted of pregnant women with normal term birth (37–43 completed week' gestation), and were selected according to the onset of labor, that was spontaneous (group III) or induced (group IV). The former group consisted of 31 pregnant women with spontaneous onset of vaginal deliveries preceded by the regular uterine contractions during first and second stage of the labor. The latter group included 25 women with induced vaginal term deliveries. Maternal characteristics of the women are shown in Table 1. Patients with preterm deliveries, chorioamnionitis, hypertension, diabetes mellitus and multiple pregnancies were excluded. All placentas were histologically examined by an experienced pathologist.
Table 1 The characteristics of subjects
Pregnant women (n = 67) Maternal age ± SD (y) Gestational age ± SD (wk) Nulliparous (%) Birth weight ± SD (g) Mean Apgar ± SD
Induced stillbirths (n = 6) 28.5 (± 5.24) 25.23 (± 1.36) 33.3 612.6 (± 172.2) -
Spontaneous stillbirths (n = 5) 27 (± 5.91) 25.57 (± 1.43) 40 650 (± 145.4) -
Induced labor at term (n = 25) 28.07 (± 4.55) 38.96 (± 1.83) 60 3290 (± 467) 9.34 (± 2.02)
Spontaneous labor at term (n = 31) 28.16 (± 6.04) 39.68 (± 2.36) 58 3138 (± 649) 9.5 (± 1.22)
SD – standard deviation; (n – number of tissue samples)
2.2 Preparation of tissue extracts
Tissue samples (app. 0.5 × 0.5 × 0.5 cm) were obtained from the central part of the placenta collected following delivery and stillbirth and were immediately frozen. The 0.5 cm thick tissue samples contained villous lamina (extra-villous cytotrophoblast cells and syncytiotrophoblast) and cells from maternal decidua compacta (the external part of decidua basalis). The specimens were mixed with complete proteinase inhibitor cocktail (Roche, Germany) and homogenized on ice-bath in glass-glass Potter-Elvejhem homogenizer. The resulting suspensions were mixed with equal volume of SDS sample lysis buffer (4% SDS, 20% glycerol, 125 mM Tris-HCl pH 6.8) and boiled on water bath by 5 minutes. The chilled samples were then centrifuged at 16,000 g for 15 min at room temperature. The collected supernatants were used for further analysis.
2.3. Western blotting
Total protein content in the obtained supernatants was measured using BCA assay kit and different sample volumes (usually in the range of 2–10 μl) equivalent to 50 μg of total protein were then loaded on SDS-PAGE tris-tricine peptide-separating gels. Prestained broad range molecular weight proteins standard (Bio Rad, USA) was used in gel marker lane. Following electrophoresis the gels were electrotransferred on Immobilon-P polyvinylidene difluoride (PVDF) 0.45 μm membrane (Millipore, USA) in the buffer containing 10 mM 3-[cyclohexyloamino]-1-propanesulfonic acid (CAPS) pH 11, supplemented with 10% methanol. The obtained membrane blots were blocked overnight by gentle agitation in 5% bovine albumin in TST buffer (0.01 M Tris-HCl, pH 7.4, 0.9%NaCl), 0.5% Tween-20). All described procedures were performed at room temperature. Albumin solution was discarded and the membranes were then agitated for 2 hrs in TST with the mouse monoclonal anti-RCAS1 IgM-class antibody, 1: 4 000 dilution (Medical and Biological Laboratories, Japan). The membranes were then subjected to 4 cycles of washings in TST, 5 minutes each and immersed for agitation in the 1 : 2 000 dilution of SIGMA biotinylated anti-mouse IgM μ-chain specific antibodies for a 2 hours period. After 4 cycles of washings the membranes were then treated for 2 hours in 1 : 2 000 dilution of ExtrAvidin alkaline phosphatase conjugate (SIGMA, USA) and finally washed 2 times in TST and 2 times in TST without Tween-20. Color reaction was developed with the use of Fast Red TR/Naphtol AS-MX tablet set (SIGMA, USA). A sufficient bands intensity was obtained usually following 5 min period of developing. Obtained immunoblotts were then rinsed with distilled water and air-dried. Detailed description of the tissue preparation and semi-quantitative assessment of RCAS1 and control beta-Actin relative amounts in the tissue samples using Western blot technique has been presented in our previous reports [5,6]. The RCAS1 antigen was identified as a 32 kDa band and β-actin represented a 42 kDa band [6,10,11].
2.4. Computer-aided image analysis
Fluor-S MultiImager (BioRad, USA) was used to scan immunoblotted membranes and a QuantityOne software (BioRad, USA) was used to quantitate band intensities. All calculations were performed on RCAS1 antigen band having molecular mass of about 32 kDa [6,10,11]. The intensities of this band were expressed in arbitrary relative units and one unit (U) was defined as band intensity produced in the reference sample. This reference sample was randomly chosen but was strictly the same on all blots and was applied always in the same amount. Typical procedure of RCAS1 quantitation was as follows: scanned immunoblot membrane contained one lane of molecular mass standards, while one lane of sample used as a reference to calibrate RCAS1 amount and 12 lanes containing unknown samples. The location of 32 kDa RCAS1 bands in reference and unknown lanes was identified according to the lane containing molecular mass standard. The 32 kDa RCAS1 band intensities in reference and in unknown lanes were then calculated and expressed in pixel number units. These units were divided by pixel number of reference lane band which resulted in relative intensity units "U". The resulting intensity of reference lane band was always 1.00 U while the intensities of bands from 12 unknown sample lanes on the same membrane changed according to RCAS1 level (e.g., if RCAS1 amount in a given sample was 2 times higher than this in reference sample the result was 2.00 U. If the RCAS1 amount was 2 times lower then the result was 0.50 U). As mentioned earlier, because all immunoblots contained the same RCAS1 quantity standard and all lanes were loaded with the same amount of total protein (50 μg), the determined values were highly repetitive and independent of experiment conditions. The RCAS1 total amount in examined tissue samples was recognized relatively, which was necessary because no RCAS1 standard is available. The result always shows relative amount of RCAS1 32 kD antigen in 50 μg of total sample protein [5,6].
2.4. Statistical analysis
The distribution of the subjects was analyzed with the use of Shapiro-Wilk's test. RCAS1 were compared with the use of Student's t-test whereas β-Actin control data were compared with the use of Mann-Whitney U test. Differences between studied groups were considered significant at p < 0.05.
Results
The presence of RCAS1 expression was evaluated in all placental tissue samples derived from vaginal deliveries at term and stillbirths. The relative amount of β-Actin in all groups were found to be identical (table 2), this indicate that equal loading of proteins was performed and allow a comparative study between RCAS1 expression in each group.
Table 2 The comparison of RCAS1 average relative placental amount assessed by Western blot method during vaginal delivery according to the course of the labor.
RCAS1 placental relative amount (mean ± SD) β-Actin placental relative amount (mean ± SD)
Spontaneous stillbirths (n = 5) 0.2102 (± 0.0601) 1.1383 (± 0.1438)
Induced stillbirths (n = 6) 0.3334 (± 0.1387) 1.0924 (± 0.1316)
Induced labor at term (n = 25) 0.7131 (± 0.2896) 1.1559 (± 0.7445)
Spontaneous labor at term (n = 31) 0.4775 (± 0.2728) 1.1206 (± 0.5316)
SD-standard deviation; (n – number of tissue samples)
The average relative amount of RCAS1 was significantly lower in placental tissue samples obtained from induced stillbirths than from samples collected following induced labor at term (p = 0.004). Also, average RCAS1 expression observed in placental tissue samples from women with spontaneous stillbirths was significantly lower than placental protein expression in samples from women with term labor with spontaneous beginning (p = 0.03). Significantly higher RCAS1 placental levels were found in women with induced labor when compared to women with spontaneous onset of labor at term (p < 0.001). Significant differences in RCAS1 expression were also observed within the group of stillbirths with the respect to the beginning of the stillbirth: spontaneous and induced (p = 0.06).
Discussion
In order to investigate the possible role of RCAS1 in initiation of stillbirth, the expression levels of this protein were determined and compared between groups of patients with spontaneous and induced onset of stillbirth. The RCAS1 expression observed during spontaneous stillbirth was lower than the level of this protein found in induced stillbirth. Immune activity alterations noted during gestation have a local range and concern the endometrium-associated lymphoid tissue. Trophoblast cells participate in this phenomenon. Syncytiotrophoblast is terminally differentiated to functional trophoblast. The ratio of syncytiotrophoblast to cytotrophoblast increases with advancing gestation and at term syncytiotrophoblast occupies almost the entire placenta. The differentiation from cytotrophoblast to syncytiotrophoblast is identical to placenta development [14]. The cells of syncytiotrophoblast and cytotrophoblast play an important role in the function of the feto-maternal unit. The decrease of immune tolerance at term leads to the beginning of spontaneous labor, earlier appearance of this phenomenon may lead to miscarriage or preterm labor [15].
The activation of endometrium-associated lymphoid tissue is provided mainly by lymphocyte-mediated cytotoxicity. Normal endometrium is infiltrated predominantly by T lymphocytes, NK cells and macrophages. Increased number of B lymphocytes which constitute a small percentage during physiological cycle changes in endometrium is observed only in endometritis [16]. The analysis of mononuclear cells distribution in deciduas of term placentae revealed significantly increased number of dNK, NK and macrophages [17].
Our previous studies confirmed the observation of decreased RCAS1 expression level during labor of spontaneous onset when compared to induced labor at term. The decreased RCAS1 expression level seems to confirm the observation of higher dNK, NK activity, which might be clinically observed in the beginning of spontaneous labor, preterm labor or stillbirths [5]. Clinically asymptomatic onset of stillbirth indicates that that fetal demise might not be immediately recognized by maternal immune system. Lower placental RCAS1 expression in women with spontaneous beginning of stillbirth in comparison to induced stillbirths observed in our study seems to confirm this hypothesis. A correlation between RCAS1 expression and the infiltration by dNK cells in the human uterus found in early stage of pregnancy was observed by Oshima et al. [4]. In this study, in cases without maternal rejection RCAS1 was expressed mainly in trophoblast. In contrast, in cases with maternal rejection the expression of RCAS1 decreased strikingly. These changes were also accompanied by marked infiltration and activation of dNK cells. Normal placentas at term were selected as the controls. Immunohistochemical localization revealed the presence of RCAS1 in normal placental tissues at term. Positive staining for RCAS1 was found predominantly in extravillous cytotrophoblast and syncytiotrofoblast [4]. Therefore, it is possible that the onset of stillbirth may also require the decrease of immune tolerance which may reflect an increasing cytotoxic activity.
Conclusion
Our results indicate that alterations in RCAS1 expression may be involved in the changes of maternal immune system that take place during stillbirth.
List of abbreviations
RCAS1 (receptor-binding cancer antigen expressed on SiSo cells); CTLs (cytotoxic T lymphocytes); dNK (decidual natural killer); KIRs (killer inhibitory receptors); TNF (tumor necrosis factor).
Acknowledgements
We wish to thank Professor R. Klimek, Professor A. Reron and Professor J Marcinkiewicz for advice and helpful discussions and for the friendly words of support. We also thank dr P. Mak for biochemical support.
==== Refs
Noronha L Kasting G Martins VD Nones RB Sepulcri Rde P Carvalho DS Sampaio GA Serapiao MJ Bleggi-Torres LF Intrauterine and perinatal mortality: compartive analysis of 3904 necropsies Hospital the Clinicas, Curitiba from 1960 to 1996 J Pediatr Rio J 2000 76 213 221 14647672
Balkundi DR Hanna N Hileb M Dougherty J Sharma S Labor-associated changes in Fas Ligand expression and function inhuman placenta Pediatr Res 2000 47 301 308 10709727
King A Burrows T Loke YW Human Uterine Natural Killer Cells Nat Immun 1996 15 41 52 9032767
Ohshima K Nakashima M Sonoda K Kikuchi M Watanabe T Expression of RCAS1 and FasL in human trophoblasts and uterine glands during pregnancy: the possible role in immune privilege Clin Exp Immunol 2001 123 481 486 11298137 10.1046/j.1365-2249.2001.01461.x
Wicherek L Dutsch-Wicherek M Mak P Klimek M The role of RCAS1 and oxytocinase in immune tolerance during pregnancy Fetal Diagn Ther 2005 20 420 425 16113565
Wicherek L Dutsch-Wicherek M Mak P Klimek M Skladzien J Dubin A Comparative analysis of RCAS1 level in neoplasms and placenta Acta Biochim Pol 2003 50 1187 1194 14740005
Nakashima M Sonoda K Watanabe T Inhibition of cell growth and induction of apoptotic cell death by the human tumor-associated antigen RCAS1 Nat Med 1999 5 938 942 10426319 10.1038/11383
Sonoda K Kaku T Hirakawa T Kobayashi H Amada S Sakai K Nakashima M Watanabe T Nakano H The clinical significance of tumor-associated antigen RCAS1 expression in the normal, hyperplastic, and malignant uterine endometrium Gynecol Oncol 2000 79 424 429 11104614 10.1006/gyno.2000.5981
Dutsch-Wicherek M Popiela TJ Wicherek L Modrzejewski M Tomaszewska R Skladzien J Papaspyrou S RCAS1 expression in laryngeal and pharyngeal cancer with local lymph node metastasis 5th European Congress of Oto-Rhino-Laryngology Head and Neck Surgery 2004 Bologna: Medimond S.r.l 345 348
Tsuchiya F Ikeda K Tsutsumi O Hiroi H Momoeda M Taketani Y Muramatsu M Inoue S Molecular cloning and characterization of mouse EBAG9, homolog of a human cancer associated surface antigen: expression and regulation by estrogen Biochem Biophys Res Commun 2001 284 2 10 11374862 10.1006/bbrc.2001.4892
Engelsberg A Hermosilla R Karsten U Schulein R Dorken B Rehm A The Golgi protein RCAS1 controls cell surface expression of tumor-associated O-linked glycan antigens J Biol Chem 2003 278 22998 30007 12672804 10.1074/jbc.M301361200
Popiela TJ Wicherek L Dutsch-Wicherek M Tomaszewska R Rudnicka-Sosin L Klimek M Nowak W The presence of RCAS1 expression in breast cancer of advanced stage [abstract] Int J Gynecol Cancer 2004 223
Matsushima T Nakashima M Ohshima K Abe Y Nishimura J Nawata H Watanabe T Muta K Receptor binding cancer antigen expressed on SiSo cells, a novel regulator of apoptosis of erythroid progenitor cells Blood 2001 98 313 321 11435298 10.1182/blood.V98.2.313
Nomura M Tsukahara S Ando H Katsumata Y Okada M Itakura A Nomura S Kikkawa F Nagasaka T Mizutani S Differential distribution of placental leucine aminopeptidase/oxytocinase and aminopeptidase A in human trophoblasts of normal placenta and complete hydatidiform mole Placenta 2002 23 631 639 12361682 10.1053/plac.2002.0861
Szekeres-Barto J Varga B Pacsa AS Immunologic factors contributing to the initiation of labor-lymphocytes reactivity in term labor and threatened preterm delivery Am J Obstet Gynecol 1986 155 108 112 3460338
Disep B Innes BA Cochrane HR Tijani S Bulmer JN Immunohistochemical characterization of endometrial leucocytes in endometritis Histopathology 2004 45 625 632 15569054 10.1111/j.1365-2559.2004.02052.x
Sindram-Trujillo AP Scherjon SA van Hulst-van Miert PP Kanhai HH Roelen DL Claas FH Comparison of decidual leukocytes following spontaneous vaginal delivery and elective cesarean section in uncomplicated human term pregnancy J Reprod Immunol 2004 62 125 137 15288188 10.1016/j.jri.2003.11.007
|
15963226
|
PMC1184098
|
CC BY
|
2021-01-04 16:37:12
|
no
|
Reprod Biol Endocrinol. 2005 Jun 17; 3:24
|
utf-8
|
Reprod Biol Endocrinol
| 2,005 |
10.1186/1477-7827-3-24
|
oa_comm
|
==== Front
RetrovirologyRetrovirology1742-4690BioMed Central London 1742-4690-2-431600017410.1186/1742-4690-2-43EditorialSmall philanthropy and big science: the RETROVIROLOGY prize and Stephen P. Goff Jeang Kuan-Teh [email protected] The National Institutes of Health, Bethesda, MD, USA2005 6 7 2005 2 43 43 4 7 2005 6 7 2005 Copyright © 2005 Jeang; licensee BioMed Central Ltd.2005Jeang; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Stephen P. Goff wins the 2005 RETROVIROLOGY prize.
==== Body
Earlier this year, I announced that Retrovirology would inaugurate an annual prize to recognize the achievements of a deserving mid-career retrovirologist [1]. There are two reasons why we feel such an award is timely. First, retrovirology is an increasingly robust and important field of science. It seems wanting that no award exists to recognize exclusively excellence in basic retrovirus research. Second, although there are many scientific prizes and awards, very few are targeted only to mid-career scientists. In principle, the awarding of a prize carries two worthy aims. Prizes reward individuals for past achievements, and prizes also bring visibility and encourage future advances in particular research problems/fields. For the latter goal, choosing an outstanding mid-career individual who is on the peak of his/her productivity curve and who is poised for many years of "best" work ahead is more than appropriate. Hence, while the roster of winners of many awards frequently reads the same as other awards, the RETROVIROLOGY prize intends to highlight fresh scientists for their ongoing and accelerating research achievements.
A few wealthy international philanthropies frequently and deservedly capture the limelight with their research largess. On the other hand, every department Chairman and Dean of schools recognize the unsung importance of the many small local donors who fund the student scholarships, the travel stipends, and the stop-gap institutional research grants that keep the daily lifeblood of science flowing. The inaugural RETROVIROLOGY prize is supported through a donation from the Ming K. Jeang Foundation, an educational foundation based in Houston, Texas, which has provided for scholarships at Houston schools, at the University of Arizona, and at the Johns Hopkins University School of Medicine. Our inaugural award is named the M. Jeang RETROVIROLOGY prize.
For 2005, the Editors of Retrovirology selected Stephen P. Goff as the first recipient of the RETROVIROLOGY prize. Stephen Goff is the Higgins Professor of Biochemistry and Molecular Biophysics at the College of Physicians and Surgeons, Columbia University, USA. He was one of the first investigators to clone a functional copy of a retroviral genome, and to use recombinant DNA methods to study viral replication. Over the last two decades Dr. Goff has developed and exploited the Moloney murine leukemia virus as a genetic system. One of his most important results was the definition of the functional domains of the viral pol gene, and, especially, the seminal discovery of a viral function, now termed the integrase function, required for the integration of viral DNA into the host genome. Dr. Goff was also the first to develop plasmids that express enzymatically active reverse transcriptase (RT) of both mouse and human viruses. Fine-structure mutational analysis proved that reverse transcriptase contains separate, and separable, DNA polymerase and RNAse H domains.
Dr. Goff has used the yeast two-hybrid system to detect and characterize protein-protein interactions between viral proteins, and also between viral and host gene products. He discovered that cyclophilin A, a host prolyl isomerase, binds specifically to the HIV-1 Gag, and regulates both infectivity and escape from host restrictions. Dr. Goff has additionally pioneered the use of somatic cell genetics to identify host genes that affect retrovirus replication. He has identified a novel zinc-finger antiviral protein (ZAP) which directly binds specific sequences in viral RNAs in the cytoplasm and targets these RNAs for degradation.
Goff has also been active in the study of retroviral oncogenes, notably the v-abl gene carried by the Abelson murine leukemia virus. Goff was the first to prepare molecular clones of the viral genome and the cellular protooncogene, and to map the structure and chromosomal location of c-abl. The Goff laboratory has identified several novel Abl-binding proteins that regulate Abl signalling. He has shown that one such bridging molecule, PST-PIP1, directs phosphatases to act on Abl and thereby negatively regulate its tyrosine kinase activity.
In choosing the prize winner, the Editors considered more than just scientific excellence, and sought to identify the rare individual who is both an outstanding researcher and a selfless mentor. In the big picture of science, the successes of a scientist's students and students-of-students in generations hence likely hold much larger impact than even the most stellar singular career. Fittingly, at mid-career, Stephen Goff has already trained several highly successful scientists.
To understand Steve better as a scientist and a teacher, I had an opportunity in a question-and-answer session to solicit his views on several issues of interest. His answers to my questions are presented below.
KTJ: Did you always want to be a scientist or was there something else that you wanted to do when you were young?
SPG: I knew I wanted to do science very early in life, beginning with grade school. My father was a contractor, carpenter, boat-builder, mechanic, and clock repairman, and had a fabulous workshop. I grew up taking machinery apart and putting it back together and have always enjoyed watching complicated machines at work. I was particularly fascinated by chemistry in high school and with the idea that molecules were just very small machines. My older brother Chris went into molecular biology before me, and it was completely natural to me to follow in his footsteps. I'm still fascinated with these molecular machines that support our lives.
KTJ: What attracted you to go into retrovirus research?
SPG: I was first attracted to work on the small DNA viruses at Stanford as a graduate student in Paul Berg's lab because of the amazing techniques to manipulate DNA that were just being developed – SV40 provided the first chance to isolate and manipulate the complete genome of a simple organism (even before recombinant DNA). I liked the notion that one could write or modify a genetic program, much as one could write a computer program, and watch the results in action. As I looked around for possible subjects of study for my postdoc, I was very attracted by the RNA tumor viruses and their complex life cycle, at that time only known in outline, and their potential for gene therapy. A meeting with David Baltimore in 1977 convinced me that his lab would be an amazing place to apply recombinant DNA methods to the study of these viruses, and this cemented my choice. I've been working on these viruses ever since.
KTJ: Who had the most influence on your career and how?
SPG: Many people have shaped my thinking and the direction of my research. Paul Berg taught me how to design experiments, how to think carefully and fully about data, and how to present results clearly in writing and at meetings. David Baltimore taught me how to be incisive and efficient at work, and to focus on the most important issues – and to be bold. Many people in the field have been and continue to serve as role models: Howard Temin, Frank Lilly, Harold Varmus, and John Coffin have been particularly important.
KTJ: What do you see as some of the important questions in retrovirology in the near and long term future?
SPG: I think one obvious growth area is the interaction of viruses with the host. New tools – mainly genetic ones – have made it possible to identify and characterize the interactions of viral proteins with specific cellular machinery, and to learn how the viruses exploit that machinery for intracellular movement, macromolecular synthesis, and assembly. Another related area is the innate or intrinsic immunity to retrovirus infection that is just now being uncovered. I think one of the original justifications for the study of viruses – their ability to spotlight exciting or critical aspects of cell biology – is now being powerfully validated.
KTJ: What do you consider as your most important scientific contribution, and what is currently the most exciting research in your laboratory?
SPG: Broadly, I would suppose that my major contribution has been the development of the murine leukemia virus as a genetic system. Within that area, I think the characterizations of viral mutants defective in the protease, the integrase, and p12 gag are likely the most important results, both for their implications for basic virology and for the applied aspects of antiviral drug development. I am also proud of cloning the v-abl oncogene and c-abl protooncogene, which has ultimately led to the antitumor drug Gleevec, and of helping Liz Robertson in the development of mouse knock-out technology used in the isolation of abl KO mice.
KTJ: Young scientists who are yet established can sometimes feel that their manuscripts and/or their grants have been unfairly reviewed. What is your advice to them?
SPG: Many of us, young and old, often feel that our work has been unfairly reviewed and deserves reconsideration. It is important to develop a thick skin and keep trying – both in terms of resubmissions and in submitting to alternative journals. I encourage budding authors to rebut negative reviews as directly as possible and to request that the editors stand in judgement of the issues. When those avenues fail for a given journal, I would encourage young scientists to move on to the next logical journal, and given the expanding number of new journals there is little chance that an important result will go unpublished through prejudice or bias. I have great faith in the peer review process. In terms of grant reviews, competition is now at an almost unprecedented high. For new investigators working to establish their laboratories, this is a very difficult time to find funding, but here too persistence is key. Don't be discouraged and keep applying – quality will be rewarded.
KTJ: There are great scientists who do great science, and then there are great scientists who do great science and train great scientists. What are your secrets to training great scientists?
SPG: I hope that I will be remembered as someone who trained a large number of active and productive scientists – at last count I have graduated 25 students and trained about the same number of fellows. I try to give my students considerable freedom to explore new avenues, to fail and succeed, and so to learn by experience what is worthwhile and what is too risky. I try to encourage optimism, self-motivation, and give some sense of the excitement we all have in what we do.
KTJ: If one remembers you by the scientists that you have trained, who would you wanted to be remembered by and why?
SPG: I suppose one is most attached to the earliest people in the lab, those who helped establish us as an ongoing concern. In my case, the list would have to include such early students as John Colicelli, Naoko Tanese, and Pam Schwartzberg, and my first postdoc Monica Roth, all of whom are running active labs today. All are superb scientists and I consider myself blessed to have had them working with me – I would be proud to be remembered as having helped train them. But any such list is wildly incomplete, and I feel the same way about everyone who has passed through my laboratory. All are like family.
KTJ: How do you see your role with your former trainees after they have left your lab?
SPG: I feel it is critically important to set my outgoing people on a productive course, with at least one project for the fellows immediately ready to go – with the hope of quick and easy experiments. The pressure for rapid publications from a new laboratory is so intense today that successful trainees need to hit the ground running and do meaningful experiments from their first day in their new home. We have established a pattern where outgoing fellows are given the opportunity to exploit a new gene that they have identified in my lab and can work to develop an understanding of its mode of action, its partners, and its significance in the life cycle. I am always thrilled to read new papers from my former trainees as they expand the boundaries of virology.
KTJ: Someone told me once that every 10 years he would like to reinvent himself. What do you see yourself doing 10 years from now that might be different from what you are doing today? When do you think you would stop doing research and how will you know when the time is right?
SPG: I would be perfectly happy to be doing science ten years from now in much the same way I am now – realizing that the actual science we do is changing every day. The great thing about our field is how rapidly technology opens new doors and expands our capabilities, and I like to be early in applying new methods. I think I will stop running a laboratory when I no longer feel able to contribute to technology development, to effectively apply new methods to problems, and to be nimble. My students will probably let me know.
KTJ: In my office there is a quotation by Edmond Burke 'All that is necessary for the triumph of evil is that good men do and say nothing'. If there is one thing that you think needs to be said or done in science (or in society at large), what would that be?
SPG: A major problem for scientists is that the power and contributions of science to our society are grossly underappreciated by the nonscientific public at large. The gap between scientists and nonscientists is growing wider and is undercutting support for our work. We must fight this trend and point out to Congress all the benefits of their support. We as scientists must try to communicate the value and excitement of our work to our nonscientific friends.
KTJ: Would you want your children to become basic research scientists? Why or why not?
SPG: I think this is a wonderful time to enter the field of basic research, with the most amazing opportunities for new findings just coming available – even if funding in the next few years looks to be tight. But being a successful basic scientist requires a love of experimentation and an innate drive to learn about the way things work, and it is a demanding life. Neither of my two children is inclined toward basic science, and without that natural bent I would never push them into my field. I want them to be successful and happy in whatever way they choose to contribute to society.
KTJ: With the understanding that it will be a long time in the future, what would you like your headstone to read?
SPG: Maybe "He made it work;" or "He fixed it;" or"He liked solving puzzles."
Figure 1 Left, photograph of Stephen P. Goff; right, an engraved crystal trophy and a cash honorarium were presented to Dr. Goff in honor of his selection as the first RETROVIROLOGY prize recipient.
Acknowledgements
I thank Greg Towers, Andrew Lever, Ben Berkhout, Monsef Benkirane, Michael Lairmore, Masa Fujii, and Mark Wainberg for suggestions and advice; and Ya-hui Chi, Rieuwert Hoppes and Stephen Goff for preparation of pictures.
==== Refs
Jeang KT Life after 45 and before 60: the Retrovirology Prize Retrovirology 2005 2 26 15833114 10.1186/1742-4690-2-26
|
16000174
|
PMC1184099
|
CC BY
|
2021-01-04 16:36:39
|
no
|
Retrovirology. 2005 Jul 6; 2:43
|
utf-8
|
Retrovirology
| 2,005 |
10.1186/1742-4690-2-43
|
oa_comm
|
==== Front
RetrovirologyRetrovirology1742-4690BioMed Central London 1742-4690-2-451601417110.1186/1742-4690-2-45ResearchKu protein as a potential human T-cell leukemia virus type 1 (HTLV-1) Tax target in clastogenic chromosomal instability of mammalian cells Majone Franca [email protected] Roberto [email protected] Daniela [email protected] Yoichi [email protected] Kuan-Teh [email protected] Department of Biology, University of Padua, Padua, Italy2 Laboratory of Molecular Microbiology, NIAID, NIH, Bethesda, Maryland, 20892-0460, USA2005 13 7 2005 2 45 45 13 6 2005 13 7 2005 Copyright © 2005 Majone et al; licensee BioMed Central Ltd.2005Majone et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
The HTLV-1 Tax oncoprotein rapidly induces cytogenetic damage which can be measured by a significant increase in the number of micronuclei (MN) in cells. Tax is thought to have both aneuploidogenic and clastogenic effects. To examine the cellular target for Tax which might mechanistically explain the clastogenic phenomenon, we tested the ability of Tax to induce MN in rodents cells genetically defective for either the Ku80 protein or the catalytic subunit of DNA protein kinase (DNAPKcs). We found that cells genetically mutated in Ku80 were refractory to Tax's induction of MN while cells knocked-out for DNAPKcs showed increased number of Tax-induced MN. Using a cytogenetic method termed FISHI (Fluorescent In Situ Hybridization and Incorporation) which measures the number of DNA-breaks in cells that contained unprotected 3'-OH ends, we observed that Tax increased the prevalence of unprotected DNA breaks in Ku80-intact cells, but not in Ku80-mutated cells. Taken together, our findings suggest Ku80 as a cellular factor targeted by Tax in engendering clastogenic DNA damage.
==== Body
Background
We previously demonstrated that expression of the HTLV-I Tax oncoprotein rapidly induces cytogenetic damage which is reflected in a significant increase in the prevalence of micronuclei (MN) in cells [1-4]. To further characterize the phenomenon of Tax associated clastogenic DNA-damage, we wished to examine the status DNA-breaks in the nucleus and in MN in the presence or absence of Tax [4]. Using a cytogenetic method termed FISHI (Fluorescent In Situ Hybridization and Incorporation), DNA-breaks in the nucleus and in MN with centric or acentric DNA fragments could be characterized for the presence or absence of free 3'-OH ends. In our definition, free 3'-OH ends represent breaks which are accessible to the in situ addition of digoxigenin (DIG) -labeled dUTP using terminal deoxynucleotidyl transferase. On the other hand, an absence of accessible 3'-OH ends suggests that the breaks are protected and masked by a protein complex. In vivo, unprotected free 3'-OH ends may progress to larger lesions leading to increasingly serious chromosomal lesions which may eventually sow the seed for cellular transformation.
In earlier studies, we had observed that Tax increased the frequency of MN containing centric DNA fragments with unprotected free 3'-OH ends and that Tax decreased the frequency of MN containing DNA fragments with incorporation-inaccessible (i.e. protected) 3'-OH ends. Based on an increase in free 3'-OH containing ends/breaks, we hypothesized that Tax interfered with a protective cellular mechanism(s) that may normally recruit a protein complex to newly created DNA breaks. Subsequent to the publication of our report [4], Gabet et al. [5] showed that in some settings Tax can repress the expression of the catalytic subunit of human telomerase (hTert).
Telomerase is a ubiquitously-expressed multi-protein complex composed of a catalytic subunit (hTert), two associated proteins (TP-1 and HSP 90), and a highly conserved RNA (hTR) component of ~400 nucleotides. hTert acts as a reverse transcriptase, and normally catalyses the addition of short repetitive sequences to the ends of chromosomes using an RNA-template embedded within the hTert holoenzyme. Telomerase is expressed in proliferating stem cells, in germ cells, in activated lymphocytes and in many neoplastic cells such as gastric and colorectal carcinoma, breast tumours and adrenal tumours [6,7], and in some pre-neoplastic growths [8]. It is generally assumed that telomerase is silent in most primary somatic cells. Interestingly, because of the manner by which eukaryotic cells replicate DNA, when a cell does not have active telomerase, telomeres at the ends of chromosomes shorten progressively after every cellular division. Once the telomeric repeats have reached a critically abbreviated state, further cell division cannot ensue. This constraint may explain the senescence seen for normal somatic cells.
Telomeric repeats at the ends of chromosomes also appear to serve an end-protective function. Chromosomal ends which lack telomeric repeats are labile for end-to-end chromosome fusion and exonucleolytic degradation which can progress to further genetic rearrangements/damages. Provocatively, such gross rearrangements/damages can, at a low frequency, fortuitously alter the genome in a way to actually induce telomerase activity in the genetically altered cells. Once induced, such telomerase activity could endow the cells with the capacity to proliferate indefinitely, and this event could represent a first step towards malignant transformation [9].
We previously hypothesized [4] that proteins such as Ku, Sir, and the DNA protein kinase catalytic subunit (DNAPKcs) which are normally found at telomeric ends of chromosomes could be recruited rapidly to de novo interstitial chromosomal breaks. We had proposed that de novo interstitial breaks may be recognized by hTert and be stabilized by the transient addition of telomeric repeats which could then recruit Ku, Sir and DNAPKcs proteins [4]. Of note, Ku and DNAPKcs are also components of the non-homologous end-joining (NHEJ) DNA repair pathway. NHEJ is important for the repair of double-stranded DNA breaks. Knock-out mice and cultured cells deficient for one or more components of the Ku-DNAPKcs complex show genome instability phenotypes [10-16].
Because Tax interferes with the stability of de novo DNA breaks [4] and because Ku and DNAPKcs proteins apparently contribute protection to DNA breaks, we wish to understand how Tax influences double stranded DNA-breaks in cells (e.g. hamster xrs-6 cells) which are either genetically mutated for the Ku80 protein [10,11] or knocked out for the DNAPKcs gene (e.g. mouse embryo DNAPKcs -/- fibroblasts) [12]. We reasoned that if Tax acts to subvert the Ku protein, then cells (i.e. xrs-6) already lost for Ku80 would not incur increased DNA-break instability when Tax is over-expressed. On the other hand, if Tax targets DNAPKcs function, then we would expect that DNAPKcs-/- cells would not show enhanced frequency of cytogenetic damage when Tax is over-expressed, while xrs-6 cells would. Here, we used xrs-6 cells, DNAPKcs-/- cells, and the technique of in situ DIG-dUTP incorporation to distinguish between Ku80 and DNAPKcs as a DNA-break stabilizing factor targeted by Tax.
Results
MN induction by Tax in hamster and mouse cells
Clastogenic and aneuploidogenic agents increase the frequency of micronuclei (MN) because they disturb genome stability control mechanisms [1,4]. The frequency of MN can be viewed as being proportional to the cell's (in)efficiency at maintaining its genomic integrity. The NHEJ (Non-Homologous End Joining) pathway is one of the major pathways which eukaryotes use to repair double-stranded DNA breaks. Ku and DNAPKcs subunits are important NHEJ protein components.
To check Tax's effect in cells impaired for NHEJ, we first monitored the ambient frequency of micronuclei in hamster xrs-6 cells which have a mutated Ku80 gene [10,17]. We observed that MN frequency was significantly higher in xrs-6, than control CHO (Chinese hamster ovary) cells (Fig. 1). To the extent that MN reflects DNA-damage, this result suggests that under normal tissue culture conditions xrs-6 cells have a higher proclivity for cytogenetic damage. We next investigated mouse embryo fibroblasts (MEFs) engineered to be DNA-PKcs-/- [12]. We found that DNAPKcs-/- cells had a ten fold higher ambient frequency of MN when compared to wild type MEFs (DNA-PKcs+/+); and we also saw that DNAPKcs heterozygous MEFs (DNA-PKcs+/-) showed a five fold increase in MN over control MEFs (Fig. 2). Taken together, the results in figures 1 and 2 argue that both DNAPKcs and Ku proteins are important for the normal genomic homeostasis that prevents MN. Inactivation of either of these two NHEJ components appears to predispose the cell to increased cytogenetic damage.
Figure 1 Frequency (%) of micronuclei containing cells in xrs-6 and CHO cell cultures without or with transfection by Tax. *** indicates significantly different value (P < 0.001, G test) from that found in CHO cells. ** indicates significantly different value (P < 0.01, G test) from that found in CHO cells.
Figure 2 Frequency (%) of micronuclei in primary cultures of mouse embryo fibroblasts with indicated genotypes of DNAPKcs +/+, DNAPKcs +/-, or DNAPKcs -/- assayed without or with transfection of a Tax plasmid. *** indicates significantly different value (P < 0.001, G test) from that in DNAPKcs +/+ cells, with or without transfect with Tax plasmid. ** indicates significantly different value (P < 0.01, G test) from that in DNAPKcs +/+ cells without transfection with Tax plasmid.
We next compared MN frequencies in CHO, xrs-6, DNAPKcs+/+ and DNAPKcs-/- cells after transfection with a Tax-expression plasmid. Interestingly, after Tax transfection, the frequency of micronuclei in the xrs-6 cells did not significantly change from that seen in the same cells without Tax (Fig. 1). By contrast, Tax-transfected CHO cells showed a three fold increase in MN compared to mock transfected cells (Fig. 1). When we checked DNAPKcs+/+ and DNAPKcs-/- cells, we also found that both cell types showed increases in micronuclei after Tax-expression (Fig. 2).
We interpret the above results to mean that in Ku-intact cells (i.e. DNAPKcs+/+, DNAPKcs-/-, and CHO cells), Tax can increase cytogenetic damage above ambient levels. By contrast, Tax does not increase the extent of genetic damage in Ku defective cells (i.e. xrs-6 cells) (Fig. 1, 2). The two findings can be explained if Ku80 is specifically targeted by Tax. If so, because xrs-6 cells are already lost for Ku80, its already high baseline level of MN cannot be further aggravated by Tax. On the other hand, Tax could target the still intact Ku function in DNAPKcs+/+, DNAPKcs-/-, and CHO cells to increase MN numbers.
DIG(digoxigenin)-dUTP incorporation in nuclei and MN of hamster and mouse cells
We next investigated the status of DNA breaks in the nuclei and MN of xrs-6, DNA-PKcs-/- and control cells using the previously described in situ DIG-dUTP incorporation assay [4]. This method incorporates in situ a tagged-dUTP which can be used to identify and quantify broken and unprotected 3'-OH DNA ends. We were curious to compare how Tax affects the protection of 3'-OH DNA ends in Ku80-/- (i.e. xrs-6) and DNAPKcs-/- cells.
We found that the frequency of incorporated DIG-dUTP in nuclei and MN was significantly increased in xrs-6 cells compared to control CHO cells (Fig. 3). Under normal culturing conditions, xrs-6 cells showed robust and numerous in situ DIG-dUTP signals in nuclei and MN (Fig. 4A). These findings suggest that loss of Ku-function significantly increases the prevalence of unprotected freely accessible 3'-OH DNA ends. Interestingly, when we transfected Tax into xrs-6 cells, no further increase in DIG-dUTP incorporation in either the nuclei or MN was apparent (Fig. 4B). Thus, Tax expression in cells already lost for Ku80 failed to change further the number of unprotected 3'OH-DNA ends.
Figure 3 Comparison of the frequency of in situ incorporation of digoxigenin (DIG)-dUTP in nuclei of hamster and mouse cells in the absence or presence of Tax. *** indicates significantly different value (P < 0.001, G test) from that found in the respective control (comparison between the paired columns). ## or ### indicates significantly different value (P < 0.01, or P < 0.001, G test) from that of the respective controls in the absence of Tax.
Figure 4 Visualization of in situ incorporation of DIG-dUTP in xrs-6 cells in the absence (A) or presence (B) of transfected Tax. Counterstaining with propidium iodide is shown as red fluorescence while incorporation of DIG-dUTP is shown as yellow-green fluorescence. Multiple views show that in situ incorporation signals in nuclei and micronuclei do not increase substantially after transfection with a Tax-expressing plasmid.
We also checked DNAPKcs-/- MEFs. These cells are knocked out for the DNAPKcs gene but have intact Ku80 protein. Here, we found that the ambient incorporation of DIG-dUTP into DNAPKcs-/- nuclei and MN was low (Fig. 3; Fig. 5A). Indeed, the DIG-dUTP incorporation frequency in DNAPKcs-/- cells was not significantly different from that in control DNAPKcs+/+ or in DNAPKcs+/- heterozygote cells (Fig. 3). After transfection with a Tax-plasmid, both DNAPKcs +/+ (Fig. 3) and DNA PKcs-/- (Fig. 3; Fig. 5B) showed significant increases in the incorporation of DIG-dUTP into nuclei and MN. Unlike xrs-6 cells, DNAPKcs-/- and DNAPKcs+/+ cells have intact Ku80; we interpret their DIG-dUTP incorporation results to mean that Tax targeted the Ku80 protein in these cells and that such targeting increased the number of DIG-dUTP accessible unprotected 3'OH DNA ends.
Figure 5 Visualization of in situ incorporation of DIG-dUTP in PKcs-/- cells in the absence (A) or presence (B) of transfected Tax plasmid. Counterstaining with propidium iodide is shown as red fluorescence while incorporation of DIG-dUTP is shown as yellow-green fluorescence. Multiple views show that in the presence of the Tax (B) the incorporation signals are far greater than those in the absence of Tax (A). Note that many MN are seen to contain in situ incorporation signals.
Reduced Ku80 expression in HTLV-1 transformed cells
The above findings suggested Ku80 as a Tax-target. To ask if Tax affects Ku80 in HTLV-1 transformed human cells, we investigated the expression of this protein in Jurkat, MT-4, and C81 cells (Fig. 6). Jurkat is a spontaneously transformed T-cell line unrelated to HTLV-1; while both MT-4 and C81 cells are HTLV-1 transformed cells that highly express Tax. Using anti-Ku antibody which recognizes both the Ku70 and Ku80 proteins, we found that constitutive expression of Ku80 was reduced in both cells that express Tax, MT-4 and C81 (Fig. 6, lanes 5 and 9), when compared to Jurkat (Fig. 6, lane 1). Interestingly, when cells were treated with mitomycin C (a DNA-damaging agent), Ku80 expression remained inducible in both MT-4 and C81 cells. Thus targeting of Ku80 by Tax appears not to be an irreversible process.
Figure 6 Reduced constitutive expression of Ku80 in MT-4 and C81-6645 (C81) cells compared to Jurkat cells. Total cell lysates were prepared from the indicated cell lines and probed with anti-serum which recognizes both Ku70 and 80 proteins. Where indicated the cells were also treated with 1 μM mitomycin C (MMC) for the stated time period before harvesting. Note that constitutively reduced Ku80 expression remains inducible by MMC in the two Tax expressing cell lines (MT-4 and C81).
Discussion
Tax has been reported to cause both aneuploidogenic and clastogenic effects. Here we have explored the clastogenic property of Tax. We posed a simple question: in cells respectively defective for either Ku80 or DNAPKcs, which cell type remains responsive to Tax-induction of MN and DIG-dUTP incorporation? Based on our results that cells genetically mutated in Ku80 were no longer responsive to Tax's induction of MN and DIG-dUTP incorporation, we posit that Ku80, but not DNAPKcs, is a functional Tax target.
Both Ku and DNAPKcs are important for NHEJ. The current thinking is that Ku protein binds to DNA discontinuously and in a sequence independent manner, carrying out a DNA-protective role [18]. Once bound to DNA, Ku proteins recruit and activate the catalytic DNAPKcs subunit which can phosphorylate Ku and other neighboring DNA-bound proteins [19]. It has also been reported that DNAPKcs self-phosphorylates to inactivate the holo-kinase complex and then dissociates itself from Ku and the DNA. In this manner, the helicase activity of Ku is inactivated, allowing base pairing to occur between micro-homologous regions. DNAPKcs further recruits the XRCC4/ligase IV protein, which provides the DNA-ligase function needed to complete repair [20]. This intimate interplay between DNAPKcs and Ku explains why an absence of one or the other protein results in increased cytogenetic aberrations in cells.
Ku and DNAPKcs are commonly found at the telomeric ends of chromosomes. One view is that these proteins with other factors assemble a telomeric "cap" which contributes to the stability of chromosome ends [21]. Of note, there is evidence which suggests that telomeric repeats may also be transiently added to de novo interstitial chromosomal breaks leading to their stabilization and preventing further exacerbation of damage [22]. Accordingly, DNA-ends or DNA-breaks not capped by telomeric sequences and their associated proteins are unstable and labile to aberrant fusions [23,13]. Interestingly, studies have shown that upon DNA damage, PARP-1 (a nuclear enzyme which catalyzes the polyADP-ribosylation of target proteins in response to DNA damage) and Ku proteins are rapidly activated and compete for binding to DNA-ends [24], suggesting a general activity conferred by these proteins in stabilizing damaged DNA [25]. PARP-1 and Ku proteins can be co-immunoprecipitated [26], indicating that the two DNA end-sensing molecules interact in response to DNA strand breakages. Moreover, Ku function can be modulated by PARP-1 [27,28]. Thus, PARP-1 polyADP-ribosylates itself and also Ku70/80, and the polyADP-ribosylated Ku 70/80 is reduced in its DNA binding affinity, and becomes attenuated in its ability to stimulate Werner syndrome (WRN) exonuclease [28].
Our current data add the viral Tax oncoprotein to the list of complex interactors with Ku. We report here that cells genetically knocked out for Ku80 are refractory to the induction by Tax of MN and DIG-dUTP incorporation. Interestingly, in cells intact for Ku80, Tax expression reduced the ambient expression of this protein. It remains to be resolved how Tax mechanistically affects Ku80-expression; however, adding our current to our previous demonstration that Tax interferes with the protective cellular mechanisms used normally for stabilizing DNA breaks [4,29], we propose that Ku80 likely represents a crucial DNA end-protective protein targeted by Tax. Targeting of DNA end-protective proteins by oncoproteins may attenuate the functions of these factors and could lead to increased DNA structural instability and progression of damage. Progression of DNA structural damage may ultimately contribute to and mechanistically explain the process of cellular transformation. Our views on the implications of protecting de novo DNA-breaks with telomeric-caps for cellular transformation are in part consistent with recent findings that telomeric fusion to breaks reduces oncogenic translocations and tumor formation [30].
Materials and methods
Cells and transfection
Hamster xrs-6 (genetically mutated for Ku 80) cells, CHO wild type cells, mouse embryo fibroblasts knocked out for the PKcs gene, and PKcs +/- or PKcs+/+ MEFs, were cultured as monolayers in Dulbecco's minimal essential medium (Invitrogen) supplemented with 10% fetal calf serum. Where indicated, cells were transiently transfected using calcium phosphate with a wild-type Tax expression plasmid (HPx). The cells were surveyed 48 hours later for cytogenetic effects.
Micronuclei (MN) assay
For MN assay, suspensions of cells were prepared by trypsinization of cultured cells in log-phase. Cells were divided into 40 mm dishes with each dish receiving 8 × 10 5 cells in 10 ml of medium. The cells were collected 48 h later by trypsinization and were washed in phosphate-buffered saline and fixed for 15 minutes in paraformaldehyde (1% in PBS) for in situ incorporation analysis. Interphase preparations were obtained following the procedures previously described [1].
Fluorescence in situ incorporation
Fluorescence in situ incorporation was carried out using terminal transferase (TdT) which catalyses the addition of deoxyribonucleotide triphosphates to the 3'-OH ends of single or double-stranded DNA. To the substrates of TdT, digoxigenin-11-dUTP (the digoxigenin is bound to position 5 of the pyrimidine by an arm of 11 carbon atoms) was added to the 3'-OH ends. Antibody detection of DIG-dUTP labelling employed a specific antibody linked to fluoresceine, a fluorochrome which when stimulated at 494 nm wavelength emits a green signal (λ = 523 nm). The experimental protocol for fluorescent in situ incorporation used 2 washes with HBS (NaCl 280 mM, Na2PO4 × 7H2O, 1.5 mM, Hepes 50 mM). The TdT incorporation reaction of DIG-11-dUTP used the following: 10 μl of a solution (Boheringer) containing potassium cocodylate 1 M, Tri-HCl 125 mM (pH 6.6, 4°C), Bovine serum albumin (BSA) 1.25 mg/ml, CoCl2 10 mM; 0.2 μl of a solution (Boheringer) containing TdT (25 units/μl), EDTA 1 mM, 2 mercaptoethanol 4 mM, glycerol 50% (v/v) (pH 6.6, 4°C); 1 μl of DIG-11-dUTP (1 mM) mixture (Boheringer). Distilled water was added to a final volume of 50 μl. The cells were incubated in this solution at 37°C for 1 hour in an HBS-moist environment. At the end of the incubation the slides were immersed into a basin containing 0.1% Triton X-100 and 0.5% BSA in HBS to equilibrate the slides with anti-DIG-11-dUTP (1:50) labelled with FITC (Boheringer). Equilibration was conducted at room temperature for 30 minutes in an HBS moist environment. The slides were subsequently washed 3 times for 5 minutes each with the same HBS solution. The slides were then counterstained with propidium iodide (0.3 μg/ml).
Scoring of the slides
Fluorescent microscopy was performed on a Zeiss microscope with different filters and equipped with an HBO 100 mercury lamp (Osram, Munchen, Germany). Photographs were taken on Kodak Ektachrome 166 ASA film. To determine the number of MN per nucleus in slides, for each experimental point, 3000 cells were counted, using at least two independent slides for each experimental point. Differences between data from spontaneous and Tax induced cytogenetic effects were tested for significance using the G test [31].
Competing interests
The author(s) declare that they have no competing interests.
Acknowledgements
We thank Claudio Friso and Renzo Mazzaro (Department of Biology, Padua) for technical assistance in the preparation of figures, members of the Jeang laboratory for critical reading of manuscript, and Anthony Elmo for preparation of manuscript.
==== Refs
Majone F Semmes OJ Jeang KT Induction of micronuclei by HTLV-I Tax Virology 1993 193 456 459 8438579 10.1006/viro.1993.1145
Semmes OJ Majone F Cantemir C Turchetto L Hjelle B Jeang KT HTLV-I and HTLV-II Tax: differences in induction of micronuclei in cells and transcriptional activation of viral LTRs Virology 1996 217 373 379 8599225 10.1006/viro.1996.0126
Jeang KT Majone F Semmes OJ, Hammarskjöld ML Aneuploidogenic and clastogenic DNA damages induced by the HTLV-1 Tax protein Molecular pathogenesis of HTLV-1 1999 Arlington, Va, USA: ABI Professional Publications 43 48 10.1006/mpat.1999.0284
Majone F Jeang KT Clastogenic effect of the human T-cell leukemia virus type I Tax oncoprotein correlates with unstabilized DNA breaks J Biol Chem 2000 275 32906 32910 10969065 10.1074/jbc.C000538200
Gabet A Mortreux F Charneau P Riou P Duc Dodon M Wu Y Jeang KT Wattel E Inactivation of hTERT transcription by Tax Oncogene 2003 22 3734 3741 12802280 10.1038/sj.onc.1206468
Mannelli M Gelmini S Arnaldi G Becherini L Bemporad D Crescioli C Pazzagli M Mantero F Serio M Orlando C Telomerase activity is significantly enhanced in malignant adrenocortical tumors in comparison to benign adrenocortical adenomas J Clin Endocrinol Metab 2000 85 468 470 10634426 10.1210/jc.85.1.468
Boltze C Mundschenk J Unger N Schneider-Stock R Peters B Mawrin C Hoang-Vu C Roessner A Lehnert H Expression profile of the telomeric complex discriminates between benign and malignant pheochromocytoma J Clin Endocrinol Metab 2003 88 4280 4286 12970299 10.1210/jc.2002-021299
Zhu X Kumar R Mandal M Sharma N Sharma HW Dhingra U Sokoloski J Hsiao R Narayanan R Cell cycle dependent modulation of telomerase activity in tumor cells Proc Natl Acad Sci USA 1996 93 6091 6095 8650224 10.1073/pnas.93.12.6091
Musutomi K Hahn W Telomerase and tumorigenesis Cancer Lett 2003 194 189 197 12757977 10.1016/S0304-3835(02)00706-1
Jeggo PA Kemp LM X-ray-sensitive mutants of Chinese hamster ovary cell line, isolation and cross-sensitivity to other DNA-damaging agents Mutat Res 1983 112 313 327 6197643
Taccioli GE Gottlieb TM Blunt T Priestley A Demengeot J Mizuta R Lehmann AR Alt FW Jackson SP Jeggo PA Ku80: Product of the XRCC5 gene and its role in DNA repair and V(D)J recombination Science 1994 265 1442 1445 8073286
Kurimasa A Ouyang H Wang S Cordon Cardo G Li G Catalytic subunit of DNA dependent protein kinase impact on lymphocyte development and tumorigenesis Proc Nat Acad Sci USA 1999 96 1403 140 9990036 10.1073/pnas.96.4.1403
Bailey SM Meyne J Chen DJ Kurimasa A Li GC Lehnert BE Goodwin EH DNA double-strand break repair proteins are required to cap the ends of mammalian chromosomes Proc Natl Acad Sci USA 1999 96 14899 14904 10611310 10.1073/pnas.96.26.14899
Ferguson DO Sekiguchi J Frank K Gao Y Sharpless NE Gu Y Manis J Depinho RA Alt FW The interplay between NHEJ and cell cycle checkpoint factors in development genomic instability and and tumorigenesis Cold Spring Harbor Symposia on Quantitative Biology 2000 LXV 395 403 10.1101/sqb.2000.65.395
Pastink A Eeken JC Lohman PH Genomic integrity and the repair of double-strand DNA breaks Mutat Res 2001 480–481 37 50
Sekiguchi J Ferguson DO Yang E Frank K Gu Y Nussennzweig A Alt FW Genetic interactions between ATM and non homologous end joining factors in genomic stability and development Proc Natl Acad Sci USA 2001 23 3243 3248 11248063 10.1073/pnas.051632098
Getts RC Stamato TD Absence of a Ku-like DNA end binding activity in the xrs double-strand DNA repair-deficient mutant J Biol Chem 1994 269 15981 15984 8206892
Rathmel WK Chu G Involvment of the Ku autoantigen in the cellular response to DNA double-strand breaks Proc Natl Acad Sci USA 1994 91 7623 7627 8052631
Chu G Double strand break repair J Biol Chem 1997 272 24097 24100 9305850 10.1074/jbc.272.39.24097
Carlsson P Wateman ML Jones KA The hLEF/TCF 1 alpha HMG protein contains a context dependent transcriptional activation domain that induces the TCR alpha enhancer in T cells Genes Dev 1993 7 2418 2430 8253387
Weaver DT Telomeres: moonlighting by DNA repair proteins Curr Biol 1998 8 492 494 10.1016/S0960-9822(98)70315-X
Wilkie AOM Lamb J Harris PC Finney RD Higgs DR A truncated human chromosome 16 associated with α thalassaemia is stabilized by addition of telomeric repeat (TTAGGG)n Nature 1990 346 868 871 1975428 10.1038/346868a0
Boulton SJ Jackson SP Identification of a Saccharomyces cerevisiae Ku80 homologue: roles in DNA double strand break rejoining and in telomeric maintenance Nucleic Acids Research 1996 24 4639 4648 8972848 10.1093/nar/24.23.4639
D'Silva I Pelletier JD Lagueux J D'Amours D Chandhry MA Weinfeld M Lees-Miller SP Poirier GG Relative affinities of poly(ADP-ribose) polymerase and DNA-dependent protein kinase for DNA strand interruptions Biochem Biophys Acta 1999 1430 119 126 10082940
Herceg Z Wang ZQ Functions of poly (ADP-ribose) polymerase (PARP) in DNA repair, genomic integrity and cell death Mutat Res 2001 477 97 110 11376691
Ariumi Y Masutani M Copeland TD Mimori T Sugimura T Shimotohno K Ueda K Hatanaka M Noda M Suppression of the poly(ADP-ribose) polymerase activity by DNA-dependent protein kinase in vitro Oncogene 1999 18 4615 4625 10.1038/sj.onc.1202823
Tong WM Cortes U Hande MP Ohgaki H Cavalli LR Landsdorp PM Haddad BR Wang ZQ Synergistic role of Ku80 and Poly(ADP-ribose) Polymerase in suppressing chromosomal aberrations and liver cancer formation Cancer Research 2002 62 6990 6996 12460917
Li B Navarro S Kasahara N Comai L Identification and biochemical characterization of a Werner syndrome protein complex with Ku70/80 and PARP-1 J Biol Chem 2004 279 13659 13667 14734561 10.1074/jbc.M311606200
Jeang KT Giam G Majone F Aboud M Life, death, and Tax: role of HTLV-1 oncoprotein in genetic instability and cellular transformation J Biol Chem 2004 279 31991 31994 15090550 10.1074/jbc.R400009200
Qi L Strong MA Karim BO Huso DL Greider CW Telomere fusion to chromosome breaks reduces oncogenic translocations and tumour formation Nature Cell Biol 2005 7 706 711 15965466 10.1038/ncb1276
Sokal P Rolph FJ Biometry 1991 San Francisco: Freemann
|
16014171
|
PMC1184100
|
CC BY
|
2021-01-04 16:36:38
|
no
|
Retrovirology. 2005 Jul 13; 2:45
|
utf-8
|
Retrovirology
| 2,005 |
10.1186/1742-4690-2-45
|
oa_comm
|
==== Front
Respir ResRespiratory Research1465-99211465-993XBioMed Central London 1465-9921-6-701601179910.1186/1465-9921-6-70ResearchThe beta2 integrin CD11c distinguishes a subset of cytotoxic pulmonary T cells with potent antiviral effects in vitro and in vivo Beyer Marc [email protected] Hongwei [email protected] Nina [email protected] Sandra [email protected] Cordula [email protected] Peter JM [email protected] Jürgen [email protected] Department of Respiratory Medicine, NHLI, Imperial College London, Norfolk Place, London, UK2 Klinik für Kinder- und Jugendmedizin, St. Josef-Hospital, Ruhr-Universität Bochum, Alexandrinenstr. 3, 44791 Bochum, Germany2005 12 7 2005 6 1 70 70 14 9 2004 12 7 2005 Copyright © 2005 Beyer et al; licensee BioMed Central Ltd.2005Beyer et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The integrin CD11c is known as a marker for dendritic cells and has recently been described on T cells following lymphotropic choriomeningitis virus infection, a systemic infection affecting a multitude of organs. Here, we characterise CD11c bearing T cells in a murine model of localised pulmonary infection with respiratory syncytial virus (RSV).
Methods
Mice were infected intranasally with RSV and expression of β2 integrins and T lymphocyte activation markers were monitored by flow cytometry. On day 8 post RSV infection CD11c+ CD8+ and CD11c- CD8+ T cells were assessed for cytokine production, cytotoxic activity and migration. Expression of CD11c mRNA in CD8+ T cells was assessed by quantitative PCR.
Results
Following RSV infection CD11c+ CD8+ T cells were detectable in the lung from day 4 onwards and accounted for 45.9 ± 4.8% of CD8+ T cells on day 8 post infection, while only few such cells were present in mediastinal lymph nodes, spleen and blood. While CD11c was virtually absent from CD8+ T cells in the absence of RSV infection, its mRNA was expressed in CD8+ T cells of both naïve and RSV infected mice. CD11c+, but not CD11c-, CD8+ T cells showed signs of recent activation, including up-regulation of CD11a and expression of CD11b and CD69 and were recruited preferentially to the lung. In addition, CD11c+ CD8+ T cells were the major subset responsible for IFNγ production, induction of target cell apoptosis in vitro and reduction of viral titres in vivo.
Conclusion
CD11c is a useful marker for detection and isolation of pulmonary antiviral cytotoxic T cells following RSV infection. It identifies a subset of activated, virus-specific, cytotoxic T cells that exhibit potent antiviral effects in vivo.
==== Body
Background
Beta2 integrins, which are restricted to leukocytes, consist of a common β-chain (CD18) and the distinct α-chains CD11a (LFA-1), CD11b (Mac-1/ CR3) or CD11c (p150,95/ CR4) [1]. While CD11a is expressed widely on leukocytes, expression of CD11b and CD11c was thought to be confined to cells of myeloid origin and these molecules have been used as markers to define certain cell populations, e.g. CD11b for macrophages [2] and CD11c for dendritic cells [3]. β2 integrins are of critical importance for the development of functional immune responses; a mutation in the CD18 gene, resulting in decreased expression of β2 integrins and in defective migration of granulocytes, causes an immune defect termed leukocyte-adhesion deficiency [1]. CD11a/LFA-1 has been shown to be involved in T cell activation [4], T cell recruitment [5] and target cell killing [6] by binding to its ligand intercellular adhesion molecule-1 (CD54) thus mediating adhesive cell-cell-interactions. Over the past ten years several groups have reported that CD11b is expressed on activated cytotoxic T cells [7,8] and that it is also involved in T cell migration into inflamed tissues [9]. CD11c expression on T cells has been detected initially on a population of intestinal intraepithelial lymphocytes [10] and more recently on CD8+ T cells following systemic viral infections [11]. The studies regarding β2 integrins on T cells were mostly conducted in mouse-models of infection with LCMV, a virus that induces systemic immune responses involving many different organs [12]. Here, we studied CD11c+ T cells during a localised infection of the lung with RSV, monitoring their distribution and comparing the pulmonary to the systemic immune response. We hypothesized that CD11c is a marker of antigen-specific cytotoxic T cells, which is expressed following activation, rather than being transferred from APC during cell-cell interactions. Such a transfer from APC to T cells has been described for the co-stimulatory molecule CD80 [13]. In addition, we sought to define a function of CD11c when expressed on T cells following RSV infection.
Materials and methods
Virus and animals
Human RSV, type A2 from ATCC (Rockville, MD) free of mycoplasma contamination was used. The virus was cultured on HEp-2 cells from ATCC in Dulbecco's modified Eagle Medium (Invitrogen, Paisley, UK) containing 5% heat inactivated fetal calf serum and 1 % Penicillin / Streptomycin both from Sigma (Gillingham, UK).
Female BALB/c AnNCrl mice, 8 to 12 weeks of age, free of specific pathogens, were obtained from Charles River Laboratories (Margate, UK) and kept under specific pathogen free conditions. All experimental animals used in this study were under a protocol approved by the Home Office, London, UK. Mice were infected under light anesthesia with isoflurane by intranasal inoculation of RSV (5 × 105 PFU in 70 μl). Controls were untreated or mock infected with RSV, inactivated by irradiation with UV-light. RSV infection was confirmed by plaque assay as described previously [14]. Infection could be demonstrated in all infected animals tested but not in controls.
Experimental Protocols
Mice were infected with RSV. On days 0 (naive controls), 4, 6, 8, 10, and 21 post infection mice were sacrificed and lungs, spleens, mediastinal lymph nodes (MLN) and blood cells were harvested. For secondary RSV infection mice were infected 28 days after primary infection. Allergic airway inflammation was induced by intraperitoneal injection of OVA in Alum on days 0 and 14. On days 21 to 23 mice were challenged intranasally with 75 μl 1% OVA in PBS. Mice were sacrificed 24 hours after the last OVA challenge. Organs were disrupted using a steel mesh, and red blood cells were lysed with ACK lysis buffer. For isolation of CD8+ cells lungs were minced, incubated in collagenase (Sigma) solution and mononuclear cells were prepared using a Ficoll (Biochrom) gradient (density = 1.09 g/l). Cell numbers were assessed by counting in a Neubauer chamber.
Flow cytometry
Fc-receptors were blocked with anti-CD16/32-mAb (2.4G2), cells were incubated with appropriate antibodies for 20 min at 4°C, washed, and suspended in staining buffer. A FACScalibur flow cytometry cell analyzer (Becton Dickinson, Oxford, UK) was used for data acquisition and WinMDI software (Scripps institute, La Jolla, USA) for analysis. The following antibodies (Becton Dickinson) were used: anti-CD11c (HL3), anti-CD3 (145-2C11), anti-CD69 (H1.2F3), anti-CD54 (3E2), and anti-F4/80 (Serotec, Oxford, UK) were FITC-conjugated; anti-CD3 (145-2C11), anti-CD4 (RM4-5), anti-CD8 (53-6.7), anti-CD11b (M1/70), anti-CD11c (HL3), anti-CD40 (3/23), anti-CD62L (MEL-14), and anti-CD80 (16-10A1) were PE-conjugated; anti-CD8α (53-6.7) was CyChrome-conjugated; biotinylated anti-CD11c (N418, Serotec), anti-CD11a (M17/4), and anti-CD11c (HL3) were stained with streptavidin-CyChrome (Becton Dickinson). Rat IgG2a (R35-95), rat IgG2b (R35-38), mouse IgG2b (49.2) and hamster IgG (G235-2356) were used as isotype controls. RSV-M2-pentamers (ProImmune, Oxford, UK) were used according to manufacturer's instructions.
For intracellular cytokine staining, lung cell suspensions were cultured at a concentration of 106/ml medium in 24 well plates for 4 h in the presence of PMA (50 ng/ml), ionomycin (500 ng/ml) (both Sigma, Gillingham, UK) and Golgi Plug (1 μl/ml) (Becton Dickinson). Following fixation and permeabilization cells were stained with anti-IFN γ (XMG1.2) or isotype control.
Isolation of CD8+ cells
Mononuclear cells from lungs were pooled and incubated with anti-CD8-coated magnetic beads from Miltenyi (Bergisch-Gladbach, Germany) according to manufacturer's instructions. Cells were sorted by Auto-MACS (Miltenyi). Purity of CD8+ cells assessed by FACS was at least 90%. For sorting of CD8+ cells into CD11c+ and CD11c- populations, cells were labeled with anti-CD11c-Fitc and anti CD8-PE-antibodies and sorted on a FACSDiva (Becton Dickinson).
In vitro Cytotoxicity assay
The assay was performed as described [15]. Briefly, P815 cells (ATCC) were used as target cells and labeled with 2 μM RSV-M2-peptide (ABC Synthesis, London, UK), washed extensively and cultured in 96-well U-bottom plates (104 cells/well in 100 μl). Sorted populations of CD8+CD11c+ or CD8+CD11c- were labeled with CFSE (1 μM, Invitrogen, UK) and added to the target cells in the concentrations indicated, resulting in a final volume of 200 μl/well. In some experiments effector cells were pre-incubated with 5 μg/ml anti-CD11c-mAb (HL3; BD) or isotype control-mAb (20 minutes, 4°C) prior to co-culture with target cells. Cells were incubated for 4 hr (37°C, 95% CO2), washed, and incubated with PBS/BSA/0.1% sodium-azide containing 7-Aminoactinomycin D (7-AAD) (20 μg/ml, Sigma, UK). Afterwards cells were washed and fixed in 1% PFA containing Actinomycin D (10 μg/ml, Sigma, UK) and analyzed directly on a FACScalibur (BD). Staining of CFSE-negative cells for 7-AAD was analyzed and specific lysis calculated with the formula:
%specific lysis = 100 × (% 7-AAD+ sample - % 7-AAD+ basal)/(100 - %7-AAD+ basal).
In vivo Cytotoxicity assay
CD11c+CD8+ or CD11c-CD8+ cells isolated by FACSDiva (Becton Dickinson) from lungs harvested on dpi 8, were adoptively transferred (2.5 × 105 cells per recipient) by intra-tracheal application 1 hour after RSV infection of recipient mice. On dpi 4 RSV titers in the lungs were determined by plaque assay as described [14].
Migration assay
Lungs of RSV infected mice were harvested on dpi 8, CD8+ cells were isolated and labeled with 1 μM CFSE. Cells were incubated with 20 μg/ml of anti-CD11c antibodies (HL3, BD or N418, Serotec) or isotype controls for 30 minutes at 4°C and washed afterwards. Subsequently, cells were adoptively transferred by intravenous injection into recipient mice, which were either naïve or RSV infected (dpi4). After 3 or 24 hours lungs, perfused with PBS via the right ventricle, and spleens were harvested, cell suspensions were generated and analyzed by FACS.
Real-time PCR
RNA was isolated from purified cells using RNeasy kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions, including a step for on column DNA digestion. The isolated RNA was transcribed into cDNA using Omniscript kit (Qiagen). Primers and probes for CD11c (Applied Biosystems, Warrington, UK) and murine GAPDH (Qiagen) were used according to manufacturers' instructions. Samples were done as duplicates and analyzed on an ABI Prism 7000 light cycler PCR machine (Applied Biosystems).
Statistical analysis
Data are expressed as mean ± standard deviation unless indicated otherwise. Groups were compared by students t-test. P values for significance were set at < 0.05.
Results
RSV infection results in increases in numbers of CD11c+ CD8+ T cells in the lung
Recently, we have reported a population of CD8+CD11c+ T cells in the lung ten days after RSV infection [14]. To study the kinetics of these cells during infection, BALB/c mice were infected with RSV, lungs were harvested at several time points after infection and CD3-positive lymphocytes were monitored for expression of CD11c by FACS. In naïve (Figure 1a, c) as well as mock-infected mice (data not shown) only as little as 0.7 ± 0.3% of pulmonary CD8+ T cells expressed CD11c. Following RSV infection, numbers of CD11c+ CD8+ T cells increased significantly until dpi 8 when 45.9 ± 4.8% of CD8+ T cells expressed CD11c and declined thereafter. Interestingly, expression of CD11c was hardly increased on CD3+ CD8- cells (Figure 1a), consisting mostly of CD4+ T cells (data not shown).
Figure 1 Expression of CD11c on pulmonary CD8+ T cells following RSV infection. Following primary RSV infection lungs, spleens, MLN and blood were harvested from BALB/c mice at the time points indicated. Cells were stained with antibodies to CD3, CD8, and CD11c, CD11b or F4/80. Dot plots show expression of CD11c and CD8 on cells gated for CD3+ lymphocytes harvested from A) lungs, B) spleen, MLN and blood. Numbers indicate the percentage of CD8+ cells. Quadrant settings were adjusted that less than 1% of CD8+ cells stained positive for the isotype control of anti-CD11c-mAb. C) Changes in expression of CD11c, CD11b and F4/80 on CD3+ CD8+ cells from lungs, spleen, MLN, and blood following RSV infection. Cells were stained with antibodies as indicated and analyzed by FACS. The percentage of CD3+8+ cells staining positive for CD11c (black squares), CD11b (open circles) or F4/80 (triangles) after subtraction of isotype controls is shown as mean ± SEM (n = 6, *p < 0.05, significant difference versus dpi 0).
To analyze if CD11c is also expressed on T cells in other organs than the lung during RSV infection, CD8+ T cells were monitored in MLN, spleen and blood. In all of these organs small increases in CD11c+ CD8+ T cells were detected from dpi 4 onwards but their percentage returned to baseline rapidly. Only in MLN, numbers of CD8+ T cells expressing CD11c remained significantly elevated on dpi 10 (Figure 1c). However, numbers of CD8+ CD11c+ cells in the lungs were 8 – 12-fold higher than in any other organ at any time point assessed after RSV infection. Next, we asked if expression of CD11c on CD8+ T cells is unique to primary infection. Analysis of T cells from mice which were re-infected with RSV on dpi 28 showed that 31.9 ± 1.4% of pulmonary CD8+ T cells were positive for CD11c already 4 days after secondary infection. Again CD11c expression was much lower in MLN, spleen or blood and was not increased on CD3+CD8- cells (data not shown). Since both primary and secondary RSV infection resulted in appearance of CD8+ CD11c+ T cells we asked if CD11c expression on T cells is a general phenomenon in lung inflammation irrespective of etiology. To this end, allergic airway inflammation was induced in mice using a protocol, which leads to airway eosinophilia, as well as antigen-specific Th2 cytokine production by CD4+ lung cells (data not shown). In this setting, only 3.8 ± 1.0% of CD8+ T cells expressed CD11c and no changes were observed in the CD3+ CD8- population. To examine if other markers, expressed predominantly on cells of the myeloid lineage, are also increased on T cells during RSV infection we analyzed expression of CD11b and F4/80 on T cells from lungs, MLN, spleen and blood. Kinetics of CD11b expression on T cells paralleled CD11c expression, with lower percentages of CD11b+ CD8+ T cells in the lungs on dpi 8 and 10 (Figure 1c). Expression of the macrophage marker F4/80 was only detectable in significant amounts on splenic T cells on dpi 8 (Figure 1c).
Activation markers are expressed differentially on CD11c+ CD8+ and CD11c- CD8+ T cells
Having observed striking increases of CD8+ T cells expressing CD11c in the lung following RSV infection we asked if these cells are phenotypically different regarding their activation status compared to CD11c- CD8+ T cells. We analyzed pulmonary CD8+ cells almost all of which are T cells (>99% CD3+ in both naïve and infected mice) for surface markers known to be regulated during activation. CD69 was expressed on the majority of CD11c+ T cells but only on a small proportion of CD11c- T cells (Figure 2a). CD54, was up regulated on CD11c+ T cells, while expression on CD11c- T cells was comparable to pulmonary CD8+ T cells from naïve mice (Figure 2b). CD62L is expressed at high levels on naïve and memory T cells and down-regulated on effector T cells. On most CD11c+ T cells CD62L was detectable at low levels, while around two thirds of the CD8+ CD11c- population was CD62Lhi. Further, while only a small proportion of CD8+ T cells in MLN expressed CD11c, expression profiles of CD69 (CD11c+ 45.5 ± 2.3% versus CD11c- 7.7 ± 3.7%) and CD62Lhigh (CD11c+ 30.3 ± 8.5% versus CD11c- 87.1 ± 0.6%) on CD8+ T cells (97% of CD11c+ and 99% of CD11c- cells were CD3+) indicated that these cells already begin to express CD11c following activation in the draining lymph nodes.
Figure 2 Differential expression of surface markers on pulmonary CD11c+ and CD11c- CD8+ T cells. Lungs of naive mice or on day 8 after RSV infection were harvested. Cells were stained with antibodies for CD8, CD11c, and the surface markers indicated. A) Dot plots show expression of CD11c versus the indicated marker on cells gated for lymphocyte size and CD8 expression after RSV infection. Numbers indicate the percentage of cells in corresponding quadrants. Quadrants were set according to isotype controls except for CD11a, CD54 and CD62L. Since these three molecules are expressed to a considerable degree on naïve CD8 T cells, this expression was used for quadrant settings to highlight differences in CD11+ CTL. B) Histograms show expression of CD54, CD62L, and CD69 on pulmonary CD8+ cells from naïve (dotted line) or RSV infected mice (CD11c+ cells: black line, CD11c- cells: filled histogram).
Since most activated CD8+ T cells expressed CD11c, a surface molecule of myeloid cells including APC, we explored the expression of molecules involved in APC T cell-interaction on pulmonary CD8+ T cells 8 days after RSV infection. While CD40 was detectable neither on CD11c+ nor on CD11c- CD8+ T cells, expression of CD80 was confined mostly to CD11c+ T cells (Figure 2a). In addition, we measured co-expression of the β2 integrins CD11c and CD11a or CD11b on CD8+ T cells following RSV infection. Compared to CD11c- CD8+ T cells, CD11a was up regulated on CD11c+ CD8+ cells and CD11b was co-expressed on a subpopulation of CD11c+ cells (Figure 2a).
CD11c mRNA is expressed by both activated and resting CD8+ T cells
Surface molecules such as CD80 can be transferred from APC to T cells by cell-cell contact [13]. Since both CD11c and CD80 are expressed widely on murine APC we asked whether the presence of CD11c on CD8+ T cells is due to expression by T cells rather than to membrane transfer from APC during T cell priming. CD8+ cells were purified from lungs and spleens of RSV-infected or naïve mice, RNA was isolated and the expression of CD11c analyzed by quantitative PCR. CD11c+ pulmonary cells served as a positive control while LA-4 cells, a pulmonary epithelial cell line, and P815 cells, a mastocytoma cell line, were used as negative controls. In two separate experiments, expression levels of CD11c normalized to GAPDH were nearly equal in CD11c+ (relative expression = 1.05 ± 0.06, n = 4) and CD8+ (relative expression = 1.08 ± 0.03) cells from the lung following RSV infection. Surprisingly, CD11c mRNA could also be detected in comparable amounts in CD8+ cells isolated from spleens of naïve mice (relative expression = 1.12 ± 0.08). CD11c mRNA was neither detectable in P815 cells nor in LA-4 cells. These findings suggest that CD11c mRNA is constitutively expressed in CD8+ T cells.
CD11c+ CD8+ cells are IFNγ producing, virus-specific, potent cytotoxic T cells
To characterize the function of CD11c+ T cells in RSV infection we determined their capacity to produce IFNγ. Pulmonary cell suspensions, harvested 8 days after RSV infection, were incubated with medium or PMA/ionomycin and CD8+ T cells were analyzed for intracellular IFNγ production. In un-stimulated cells IFNγ was detectable in less than 3% of CD11c+ or CD11c- CD8+ cells while following stimulation the percentage of IFNγ-producing CD11c+ CD8+ cells was significantly higher than in CD11c- cells (38.4 ± 5.0% versus 25.5 ± 3.0%, respectively, n = 4, p < 0.05). The crucial function of CTL in viral infections is to induce apoptosis and lysis of infected cells to eliminate replicating virus. To determine if CD11c+ T cells are more effective CTL during RSV infection than their CD11c- counterparts we initially analyzed induction of apoptosis in RSV M2 peptide-labeled P815 cells. CD8+ lung cells were isolated by MACS and enriched for CD11c+ by FACS sorting, resulting in about 80% of CD11c+ CTL in the positive population, while the negative fraction still contained about 25% of CD11c+ CD8+ T cells. These CD11c+ and CD11c- CD8+ T cell populations were used as effector cells in a cytotoxicity assay. Loss of membrane integrity in apoptotic CFSE- target cells enables detection of dying cells by staining of DNA with the fluorescent dye 7-AAD (Figure 3a). CD11c+ T cells were significantly more efficient in inducing apoptosis in target cells than CD11c- T cells (Figure 3b). Pre-incubation of effector cells with anti-CD11c-mAb had no effect on the cytotoxic activity of either CD11c+ (Figure 3b) or CD11c- cells (data not shown). To assess the frequency of RSV-specific CTL directly we stained lung cells with RSV M2-pentamers, which recognize cells specific for this dominant CTL epitope. About 38% of CD11c+ CTL were M2-Pentamer+ and at 81% the majority of T cells expressing the RSV-M2 specific TCR was found to be CD11c+ (Figure 3c). In contrast, only about 10% of CD11c- CD8+ T cells were specific for RSV M2-peptide. These results show that CD11c+ CD8+ T cells, which include the majority of RSV specific CTL, are highly effective in inducing target cell apoptosis in vitro. To compare antiviral efficiency of CD11c+ and CD11c- CD8+ T cells in vivo, these cells were isolated by FACS sorting from lungs on dpi 8 and adoptively transferred one hour after RSV infection of recipient mice. Transfer of low numbers (2.5 × 105) of CD11c+ CTL (purity > 83%) significantly reduced lung RSV titers on dpi 4 (Figure 3d). Transfer of the same numbers of CD11c- CD8+ T cells (purity > 99%) did not have any effect, but transfer with tenfold higher numbers of CD11c- CD8+ T cells also reduced RSV titers in the lung significantly (data not shown).
Figure 3 Target cell lysis, binding of RSV M2-pentamer and reduction in viral titers by CD8+ T cells following RSV infection. Lungs were harvested on day 8 after RSV infection and CD8+11c+ or CD8+11c- cells were isolated. A) CFSE labeled effector and unlabeled P815 target cells were incubated at the indicated ratios for 4 hours. Subsequently the percentage of unlabeled cells, staining positive for 7-AAD, was determined by FACS and B) specific lysis was calculated (mean ± SD, n = 4, significant difference of * CD11c+ (open squares) versus CD11c- cells (open circles) or # anti-CD11c-mAb pretreated CD11c+ (filled triangles) versus CD11c- cells, p < 0.05) C) Freshly isolated lung cells were stained with anti-CD8, anti-CD11c and RSV M2-pentamer. Dot plots show expression of CD11c without or with RSV M2-pentamer staining on cells gated for CD8 expression. D) Isolated CD8+11c+ or CD8+11c- cells (2.5 × 105) were adoptively transferred to the lungs 1 hour after RSV infection of recipients. Controls only received PBS intra-tracheally after infection. Lung RSV titers were determined by plaque assay on dpi 4. Depicted are mean ± SD from 2 independent experiments (n = 6, * significant difference, p < 0.05, CD11c+ (black bar) versus PBS group (open bar) and CD11c- cells (hatched bar).
Preferential recruitment of CD11c+ CD8+T cells to the lung
As CD11c expression is associated with activation of virus-specific T cells but does not seem to have a specific, non-redundant function in cytotoxic activity, we asked if CD11c may be involved in migration of these cells to the lung. CD8+ cells containing 49.9 ± 1.2 % CD11c+ cells were isolated from lungs of mice 8 days after RSV infection, labeled with CFSE and injected intravenously into naïve mice, to assess their migratory behavior. Lungs and spleens of recipient mice were harvested 3 hours after cell transfer and the percentage of CD11c+ cells in the population of CFSE+ CD8+ cells, was analyzed by FACS (Figure 4a). In lungs of naïve mice 52.9 ± 13.6% of transferred CD8+ cells were CD11c+, while only 12.2 ± 3.6% stained positive for CD11c in the spleen. Total numbers of CD11c+ CD8+ T cells migrating to the lung (10.4 ± 8.4 × 103, n = 5, p < 0.05) were also significantly higher than those reaching the spleen (6.3 ± 4.5 × 103) in naïve mice. While the percentages of CD11c+ cells of CFSE labeled CD8+ T cells in the organs did not change significantly when cells were transferred into RSV infected mice (Figure 4b), absolute numbers of CD11c+ CD8+ T cells migrating to the lung increased significantly to 20.0 ± 7.2 × 103, (n = 5, p < 0.05). In contrast, migration to the spleen did not change significantly if recipients were RSV infected (2.9 ± 2.8 × 103). Further, no significant changes were observed when cells were harvested 24 instead of 3 hours after cell transfer (data not shown). Pre-incubation of CD8+ cells with two different anti-CD11c-mAbs (clone HL3 or clone N418) did not affect recruitment into the lung.
Figure 4 Irrespective of infection, CD11c+ T cells are recruited to the lung more effectively that to the spleen. Lungs were harvested on day 8 after RSV infection; CD8+ cells were isolated by magnetic cell sorting and labeled with CFSE. Subsequently cells were adoptively transferred by intravenous injection into naive or RSV infected mice. 3 hours after transfer lungs and spleens of recipient mice were harvested and expression of CD11c on transferred cells was analyzed by FACS. A) Dot plots show gating of transferred CFSE+CD8+ cells and staining of gated cells with isotype control or anti-CD11c antibody. B) The bars indicate percentage of transferred CSFE+ CD8+ cells staining positive for CD11c in lungs and spleens after pre-incubation of transferred cells with isotype control (open bars) or anti-CD11c-mAb (hatched bars) (mean ± SD, n = 4, * significant difference, p < 0.05, lung naïve and lung RSV versus spleen naïve and spleen RSV).
Discussion
Expression of β2 integrins on T cells following viral infections has been described in lymphatic organs [7,8,11] but knowledge about integrin expression on T cells following respiratory infections is scarce. Previously, we have described a population of CD8+ T cells in the lung expressing the β2 integrin CD11c following RSV infection [14]. To characterize these cells in more detail, we have performed a time course analysis of CD3+ cells displaying CD11c, thus excluding cells of myeloid origin. In mediastinal lymph nodes, where CD11c+ cell numbers increase during inflammation [17,18], CD8α expressing dendritic cells [19,20] could be mistaken for CD11c+ CD8+ T cells if analyzed based on CD11c and CD8α only. Following RSV infection, CD11c expression on lung CD8+ T cells increased with up to 50% of CD8 T cells positive for CD11c on dpi8, while, in contrast to LCMV infection, percentages of CD11c+ CTL in spleen and blood were low. CD11c up-regulation on mucosal T cells has previously only been described in the intestine where a subpopulation of intraepithelial T cells expresses CD11c constitutively [10]. In the lung, expression of CD11c on T cells was restricted to CD8+ T cells and was paralleled by increases in CD11b. In contrast to CD11b and CD11c, expression of F4/80 was hardly detectable on pulmonary T cells following RSV infection. Expression of F4/80 may be dependent on both mouse strain and organ analyzed: F4/80+ T cells seem to be more prevalent in C57Bl/6 than in BALB/c mice [11] and were detectable in significant amounts only on splenic CD8+ T cells, as shown here and by Lin and colleagues [11]. In contrast to viral infections, allergic airway inflammation elicited by sensitization to OVA, resulted only in a small population of CD11c+ CD8+ T cells in the lung, demonstrating that up-regulation of CD11c is not a consequence of T cell activation and migration per se or of an inflammatory environment, but that it is dependent on the underlying pathology. Interestingly, reactivation of antigen-specific CD8+ memory T cells by OVA results in CD11c up-regulation when cells are primed with an OVA-expressing virus [21]. This suggests that MHC class I-restricted antigen presentation as well as co-stimulatory factors induced by viral infection, e.g. IFNα, favor expression of CD11c on CD8+ T cells. Several studies have demonstrated up-regulation of CD11a [22], expression of CD11b [7,8,23] or CD11c [11] on T cells following viral infections. Here, we demonstrate that following RSV infection a population of CTL co-expresses CD11ahi, CD11b and CD11c. Furthermore CD11c mRNA was detectable in resting as well as activated CD8+ T cells. Contamination by CD11c+ non-T cells cannot be excluded completely but is very unlikely to account for CD11c mRNA detection in CD8 T cells since no major difference in CD11c mRNA quantity was detected between CD8+ and CD11c+ cell populations. CD11c mRNA seems to be expressed constitutively in CD8+ T cells while expression of CD11c protein on the cell surface requires activation. This notion is supported further by the finding that activation markers were up regulated on CD11c+ T cells both in the lung and in the draining MLN. Interestingly, CD11b mRNA is also expressed in resting as well as activated human T cells but CD11b becomes detectable on the cell surface only following stimulation of T cells [24]. Taken together these observations make it tempting to speculate that there may be a common mechanism of post-transcriptional regulation of β2 integrin expression following T cell activation.
Assessing functional properties of CD11c+ CD8+ T cells, we found that these were more efficient than CD11c- CD8+ T cells in IFNγ production, target cell lysis in vitro and induction of viral clearance in vivo. This is likely due to the fact that the CD11c+ CTL population contains the majority of activated T cells specific for the RSV M282–90-peptide, the immuno-dominant CTL epitope of RSV in BALB/c mice. Thus, we show an association of adhesion molecule expression and high cytotoxic activity of pulmonary CD8+ T cells during a respiratory viral infection, parallel to findings in lymphoid organs in LCMV infection [8,11]. A subset of CD11c- CD8+ T cells expressed high levels of the activation marker CD69 and low levels of CD62L. Cytotoxic effects induced by CD11c- CD8+ cells could be due to this subpopulation of activated CD11c- CTL. We cannot exclude this, since we did not assess cytotoxic effects of isolated subpopulations of CD11c- CD8+ T cells. We believe though, that contamination with CD11c+ CTL is a more likely explanation for cytotoxic activity observed when CD11c- cells were used as effectors. In contrast to a study implicating CD11c on human T cell lines in target cell binding [25] we could not detect an inhibitory effect of anti-CD11c-mAb on the cytotoxic activity of T cells. In addition to target cell binding, β2 integrins are involved in migration of leukocytes and a role for CD11b in recruitment of T cells into inflamed areas has been shown by antibody blocking experiments [9]. In our model, the percentage of CD11c+ cells within the CD8+ T cell population differed between the organs assessed, being lowest in MLN and spleen, higher in the blood, where a short-lived increase of these cells was noted during the acute phase of infection, and highest in the lung. This indicates that CD11c marks effector T cells, which are recruited to the site of infection. An analogous pattern of tissue distribution has been described for VLA-4hi T cells following intra-cerebral LCMV infection [8]. Interestingly, following adoptive transfer we observed efficient recruitment of CD11c+ CD8+ T cells to the lung in naïve mice, while only a small percentage of these cells were recruited to the spleen. This preferential recruitment of CD11c+ CD8+ T cells to the lung was even more pronounced in RSV infected recipients. Pre-treatment of transferred cells with anti-CD11c-mAb did not reveal a unique function of CD11c for homing of activated CTL to the lung. These observations suggest that CD11c is a marker for activated CTL, which are recruited preferentially to the lung, while the CD11c molecule itself may not be directly involved in the process of T cell recruitment to the lung. Functional redundancy of β2 integrins in target cell binding or migration into effector sites is a possible explanation of our results. CD11a has been shown to be important for target cell binding [26], recruitment of activated lymphocytes into the lung [5,27] and generation of an effective cytotoxic T cell response during primary RSV infection [28].
Conclusion
We demonstrate that the β2 integrin, CD11c, identifies a subset of activated, virus-specific, cytotoxic CD8+ T cells. These cells are preferentially recruited to the lung in RSV infection and exhibit potent antiviral effects in vivo. CD11c is expressed by CD8+ T cells, but this molecule may not be directly involved in effector functions. Thus, following respiratory viral infections CD11c is a useful marker for detection and isolation of activated, cytotoxic, pulmonary T cells, provided distinction from myeloid cells is taken into account.
Abbreviations
7-AAD 7-Aminoactinomycin D
APC antigen presenting cells
CFSE 5-(and-6)-carboxyfluorescein diacetate, succinimidyl ester
CTL cytotoxic T lymphocytes
GAPDH Glycerol aldehyde phosphate dehydrogenase
IFN interferon
LCMV lymphotropic choriomeningitis virus
MHC major histocompatibility complex
MLN mediastinal lymph nodes
OVA chicken ovalbumin
RSV respiratory syncytial virus
Authors' contributions
MB: Experimental design, cell isolation, cytotoxicity assay, adoptive cell transfers and flow cytometry, preparation of manuscript
HW: Designed and conducted experiment assessing viral clearance after T cell transfers.
NP: Quantitative PCR and adoptive cell transfers
SD: Animal infection, cell isolation, flow cytometry
CK-R: Cell isolation, flow cytometry
PJO: Experimental design
JS: Experimental design, preparation of manuscript
All authors have read and approved the final manuscript.
Acknowledgements
This work was supported by Grants 067454 from the Wellcome Trust, 01GC9802 from Bundesministerium für Bildung und Forschung and Schw 597/2-1 from Deutsche Forschungsgemeinschaft. We thank Niga Nawrolny for technical assistance with FACS sorting.
==== Refs
Larson RS Springer TA Structure and Function of Leukocyte Integrins Immunol Rev 1990 114 181 217 2196220
Springer T Galfré G Secher DS Milstein C Mac-1: a macrophage differentiation antigen identified by monoclonal antibody Eur J Immunol 1979 9 301 306 89034
Metlay JP Witmer-Pack MD Agger R Crowley MT Lawless D Steinman RM The distinct leukocyte integrin of mouse spleen dendritic cells as identified with new hamster monoclonal antibodies J Exp Med 1990 171 1753 1760 2185332 10.1084/jem.171.5.1753
Bachmann MF McKall-Faienza K Schmits R Bouchard D Beach J Speiser DE Mak TW Ohashi PS Distinct roles for LFA-1 and CD28 during activation of naive T cells: adhesion versus costimulation Immunity 1997 7 549 557 9354475 10.1016/S1074-7613(00)80376-3
Hamann A Jablonski-Westrich D Duijvestijn A Butcher EC Baisch H Harder R Thiele H Evidence for an accessory role of LFA-1 in lymphocyte-high enothelium interaction during homing J Immunol 1988 140 693 699 3276776
Stinchcombe JC Bossi G Booth S Griffiths GM The immunological synapse of CTL contains a secretory domain and membrane bridges Immunity 2001 15 751 761 11728337 10.1016/S1074-7613(01)00234-5
McFarland HI Nahill SR Macaiszek JW Welsh RM CD11b (Mac-1): A marker for CD8+ cytotoxic T cell activation and memory in virus infection J Immunol 1992 149 1326 1333 1500720
Andersson EC Christensen JP Marker O Thornsen AR Changes in cell adhesion molecule expression on T cells associated with systemic virus infection J Immunol 1994 152 1237 1245 7507962
Nielsen HV Christensen JP Andersson EC Marker O Thornsen AR Expression of CD8+ T Cells Type 3 Complement Receptor on Activated Facilitates Homing to Inflammatory Sites J Immunol 1994 153 2021 2030 8051407
Huleatt JW Lefranqois L Antigen-driven induction of CD11c on intestinal intraepithelial lymphocytes and CD8+ T cells in vivo J Immunol 1995 154 5684 5693 7751620
Lin Y Roberts TJ Sriram V Cho S Brutkiewicz RR Myeloid marker expression on antiviral CD8+ T cells following an acute virus infection Eur J Immunol 2003 33 2736 2743 14515257 10.1002/eji.200324087
Khanolkar A Fuller MJ Zajac AJ T cell responses to viral infections: lessons from lymphocytic choriomeningitis virus Immunol Res 2002 26 309 321 12403369 10.1385/IR:26:1-3:309
Hwang I Huang J Kishimoto H Brunmark A Peterson PA Jackson MR Surh CD Cai Z Sprent J T cells can use either T cell receptor or CD28 receptors to absorb and internalize cell surface molecules derived from antigen-presenting cells J Exp Med 2000 191 1137 1148 10748232 10.1084/jem.191.7.1137
Beyer M Bartz H Hörner K Doths S Koerner-Rettberg C Schwarze J Sustained increases in numbers of pulmonary dendritic cells following respiratory syncytial virus infection J Allergy Clin Immunol 2004 113 127 133 14713917 10.1016/j.jaci.2003.10.057
Lecoeur H Fevrier M Garcia S Riviere Y Gougeon M A novel flow cytometric assay for quantitation and multiparametric characterization of cell-mediated cytotoxicity J Immunol Methods 2001 253 177 187 11384679 10.1016/S0022-1759(01)00359-3
Prabhu Das MR Zamvil SS Borriello F Weiner HL Sharpe AH Kuchroo VK Reciprocal expression of co-stimulatory molecules, B7-1 and B7-2, on murine T cells following activation Eur J Immunol 1995 25 207 211 7531145
Legge KL Braciale TJ Accelerated migration of respiratory dendritic cells to the regional lymph nodes is limited to the early phase of pulmonary infection Immunity 2003 18 265 277 12594953 10.1016/S1074-7613(03)00023-2
van Rijt LS Prins J Leenen PJM Thielemans K de Vries VC Hoogsteden HC Lambrecht B Allergen-induced accumulation of airway dendritic cells is supported by an increase in CD31hi Ly-6Cneg bone marrow precursors in a mouse model of asthma Blood 2002 100 3663 3671 12393720 10.1182/blood-2002-03-0673
Vremec D Shortman K Dendritic Cell Subtypes in Mouse Lymphoid Organs Cross-Correlation of Surface Markers, Changes with Incubation, and Differences Among Thymus, Spleen, and Lymph Nodes J Immunol 1997 159 565 573 9218570
Belz GT Smith CM Kleinert L Reading P Brooks A Shortman K Carbone FR Heath WR Distinct migrating and nonmigrating dendritic cell populations are involved in MHC class I-restricted antigen presentation after lung infection with virus PNAS 2004 101 8670 8675 15163797 10.1073/pnas.0402644101
Kim S Schluns KS Lefrancois L Induction and visualization of mucosal memory CD8 T cells following systemic virus infection J Immunol 1999 163 4125 4132 10510347
Harrington LE van der Most R Whitton JL Ahmed R Recombinant vaccinia virus-induced T-cell immunity: quantitation of the response to the virus vector and the foreign epitope J Virol 2002 76 3329 3337 11884558 10.1128/JVI.76.7.3329-3337.2002
Christensen JE Andreasen SO Christensen JP Thornsen AR CD11b expression as a marker to distinguish between recently activated effector CD8+ T cells and memory cells Int Immunol 2001 13 593 600 11282998 10.1093/intimm/13.4.593
Wagner C Hänsch GM Stegmaier S Denefleh B Hug F Schoels M The complement receptor 3, CR3 (CD11b/CD18), on T lymphocytes: activation-dependent up-regulation and regulatory function Eur J Immunol 2001 31 1173 1180 11298342 10.1002/1521-4141(200104)31:4<1173::AID-IMMU1173>3.0.CO;2-9
Keizer GD Borst J Visser J Schwarting R de Vries JE Figdor CG Membrane glycoprotein ~150.95 of human cytotoxic T cell clones is involved in conjugate formation with target cells J Immunol 1987 138 3130 3136 3106475
Kupfer A Singer SJ Cell biology of cytotoxic and helper T cell functions: immunofluorescence microscopic studies of single cells and cell couples Annu Rev Immunol 1989 7 309 337 2523714 10.1146/annurev.iy.07.040189.001521
Thatte J Dabak V Williams MB Braciale TJ Ley K LFA-1 is required for retention of effector CD8 T cells in mouse lungs Blood 2003 101 4916 4922 12623847 10.1182/blood-2002-10-3159
Rutigliano JA Johnson TR Hollinger TN Fischer JE Aung S Graham BS Treatment with Anti-LFA-1 Delays the CD8+ Cytotoxic-T-Lymphocyte Response and Viral Clearance in Mice with Primary Respiratory Syncytial Virus Infection J Virol 2004 78 3014 3023 14990720 10.1128/JVI.78.6.3014-3023.2004
|
16011799
|
PMC1184101
|
CC BY
|
2021-01-04 16:23:27
|
no
|
Respir Res. 2005 Jul 12; 6(1):70
|
utf-8
|
Respir Res
| 2,005 |
10.1186/1465-9921-6-70
|
oa_comm
|
==== Front
Respir ResRespiratory Research1465-99211465-993XBioMed Central London 1465-9921-6-831604579610.1186/1465-9921-6-83ResearchIs neutrophil elastase the missing link between emphysema and fibrosis? Evidence from two mouse models Lucattelli Monica [email protected] Barbara [email protected] Eleonora [email protected] Silvia [email protected] Benedetta [email protected] Piero A [email protected] Giuseppe [email protected] Department of Physiopathology & Experimental Medicine, University of Siena, 53100 Siena, Italy2005 26 7 2005 6 1 83 83 29 4 2005 26 7 2005 Copyright © 2005 Lucattelli et al; licensee BioMed Central Ltd.2005Lucattelli et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The separation of emphysema from fibrosis is not as clear-cut as it was thought in early studies. These two pathologies may be present at the same time in human lungs and in mice either instilled with elastolytic enzymes or bleomycin or exposed to cigarette-smoke. According to a current view, emphysema originates from a protease/antiprotease imbalance, and a role for antiproteases has also been suggested in the modulation of the fibrotic process. In this study we investigate in experimental animal models of emphysema and fibrosis whether neutrophil elastase may constitute a pathogenic link between these two pathologies.
Methods
This study was done in two animal models in which emphysema and fibrosis were induced either by bleomycin (BLM) or by chronic exposure to cigarette-smoke. In order to assess the protease-dependence of the BLM-induced lesion, a group mice was treated with 4-(2-aminoethyl)-benzenesulfonyl fluoride hydrochloride, a serine proteinase inhibitor active toward neutrophil elastase. Lungs from each experimental group were used for the immunohistochemical assessment of transforming growth factor-β (TGF-β) and transforming growth factor-α (TGF-α) and for determination of the mean linear intercept as well as the percent volume densities of fibrosis and of emphysematous changes. Additionally, the lungs were also assessed for desmosine content and for the determination of elastase levels in the pulmonary interstitium by means of immunoelectron microscopy.
Results
We demonstrate that in BLM-treated mice (i) the development of elastolytic emphysema precedes that of fibrosis; (ii) significant amount of elastase in alveolar interstitium is associated with an increased expression of TGF-β and TGF-α; and finally, (iii) emphysematous and fibrotic lesions can be significantly attenuated by using a protease inhibitor active against neutrophil elastase.
Also, in a strain of mice that develop both emphysema and fibrosis after chronic cigarette-smoke exposure, the presence of elastase in alveolar structures is associated with a positive immunohistochemical reaction for reaction for both TGF-β and TGF-α.
Conclusion
The results of the present study strongly suggest that neutrophil elastase may represent a common pathogenic link between emphysema and fibrosis. Proteases and in particular neutrophil elastase could act as regulatory factors in the generation of soluble cytokines with mitogenic activity for mesenchymal cells resulting either in emphysema or in fibrosis or both.
==== Body
Background
Lung emphysema and fibrosis are generally considered to be two diseases that totally differ in their morphological aspects and pathogenic mechanisms.
However, both these pathologies may be present at the same time in lungs of mice after cigarette-smoke exposure [1], in animals instilled intratracheally with BLM or other substances [2-5], as well as in human lungs [6]. In particular, in smokers and ex-smokers, centrilobular emphysema may be associated with some subsets of idiopathic interstitial pneumonias (IIP), namely "desquamative interstitial pneumonia" (DIP), "respiratory bronchiolitis-associated lung disease" (RB-ILD), and finally "idiopathic pulmonary fibrosis" (IPF) also characterized by the histological pattern of "usual interstitial pneumonia" (UIP). This has been clearly outlined in a recent document of the ATS/ERS defining the clinical manifestations, pathology and radiologic features of patients with IIP [7]. These data all together suggest a common pathway in the development of these two pathologies.
According to a current view, pulmonary emphysema originates from an imbalance between elastinolytic proteases and their naturally occurring inhibitors. In particular, neutrophil elastase (NE), and other elastolytic proteases, such as cathepsin G, and macrophage elastase are thought to be the main causative factors of tissue damage in this condition. This hypothesis is based on a mixture of evidence from animal models, broncho-alveolar lavage fluid (BALF) data, in vitro experiments, and from the high incidence of emphysema in homozygous subjects with a deficiency of αl-proteinase inhibitor (α1-PI) [8,9].
Recently, it has been reported that α1-PI, the secretory leukocyte protease inhibitor (SLPI), as well as the synthetic inhibitor of leukocyte elastase ONO-5046, significantly attenuate the fibrotic response to BLM in rodents [10-12]. In man, the inactivation of proteolytic enzymes may also be a critical event in normal repair, and as demonstrated in infants with respiratory distress syndrome, lack of antiprotease activity is associated with chronicity and development of fibrosis [13].
Thus, these studies suggest a significant role for the antiprotease screen not only in the development of pulmonary emphysema but also in the modulation of fibrotic lesions. This is further supported by studies carried out in BLM-challenged mice either with a genetic deficiency in α1-PI [4], or with a targeted deletion of the NE gene [14].
As previously reported by us, BLM administration induces alveolitis and fibrosis in α1-PI deficient mice. It also results in enlargement of air spaces that may be due either to loss of alveolar septa and/or retraction forces caused by the fibrotic process [4].
In this study we investigate whether NE may constitute a pathogenic link between emphysema and fibrosis. This was done in two animal models in which these two pathologies were induced either by BML or chronic exposure to cigarette smoke. In order to assess the protease-dependence of the BLM-induced lung lesion under our experimental conditions, a group mice was treated with 4-(2-aminoethyl)-benzenesulfonyl fluoride hydrochloride, a serine proteinase inhibitor fully active against NE [15].
Methods
Animals and Animal Studies
Balb/c, C57 Bl/6 and DBA/2 mice were supplied by Charles River (Calco, Italy). C57 Bl/6J pa/pa (pallid) mice were from our colony and were originally obtained from Jackson Laboratory (Bar Harbor, ME, USA). All animal experiments were conducted in conformance with the "Guiding Principles for Research Involving Animals and Human Beings" and approved by the Local Ethical Committee of the University of Siena.
Bleomycin Study
The study was carried out by using strains of mice with different levels of serum antielastase screen (Balb/c, C57 Bl/6 and pallid, with normal, intermediate and low screen, respectively) [16]. Male mice, weighing 20 to 25 g (8–10 wk old) were treated under ether anaesthesia with a single intratracheal instillation of 0.1 μg BLM (Rhone-Poulenc Rorer, Milano, Italy) in saline solution (50 μl) or with the same amount of saline. At 3, 7, 14, 21, 28 and 35 days after the treatment, animals were killed by pentobarbital sodium overdose and exsanguinated by cutting the abdominal aorta. Lungs were used for immunohistochemistry of TGF-β and TGF-α and for determination of the mean linear intercept (Lm) [17] as well as the percent volume densities of fibrosis Vv(f) and emphysematous changes Vv(e). Details of the morphometric assessment are given below. Lungs were also assessed for desmosine [18] and for determination of the elastase burden by immunoelectron microscopy [19].
Animal treatment with a synthetic serine-proteinase inhibitor
In order to assess the protease-dependence of the BLM-induced lung lesion under our experimental conditions two groups of ten C57 Bl/6 mice were treated with either 4-(2-aminoethyl)-benzenesulfonyl fluoride hydrochloride (Merck), a serine proteinase inhibitor, or with the vehicle (controls). This inhibitor (2.4 μg/μl saline) was continuously delivered at a rate of 0.5 μl/hr for two weeks by means of osmotic pumps (Alzet 2002, Alza Corporation, Palo Alto, CA). The pumps were implanted subcutaneously according the manufacturer's instructions, 24 hr before BLM-treatment. At 7 and 14 days after instillation with BLM, the lungs were excised, processed for histology, and used for morphometrical assessment of emphysematous and fibrotic lesions as well as for TGF-β and TGF-α immunohistochemistry.
Cigarette Smoke Study
Two months old male mice of the strain DBA/2 were exposed to either the smoke of 3 cigarettes/day (commercial Virginia filter cigarettes: 12 mg of tar and 0.9 mg of nicotine), 5 days/week, or to room air (controls), for various periods of time (from 1 to 6 months) as previously described in detail [20]. Groups of 8 animals were sacrificed during the chronic exposure period at various time intervals. The lungs were excised and processed for histology. Histological sections were stained with hematoxylin-eosin and Masson's trichrome. Morphometric assessment of emphysema included the determination of the Lm [17] and of the internal surface area (ISA) estimated by the Lm method at postfixation lung volume [21]. Fibrosis was assessed by point counting as Vv(f) as described below. Tissue sections were also stained for TGF-β, TGF-α and NE.
Methodologies
Lung desmosine assay
To quantitate lung elastin, the lungs of each group were weighed, homogenised (1:5, w:v) and hydrolysed in 6 N HC1 before biochemical determinations. Desmosine was analysed on hydrolysates by means of an enzyme-linked immunosorbent assay essentially according to Cocci et al. [18]. Briefly, rabbit antiserum to desmosine-hemocyanin conjugate (Abl) (Elastin Product Company, Inc, Owensville, MO) was incubated with desmosine standard (0–30 ng) (Elastin Product Company, Inc, Owensville, MO) or with adequately diluted hydrolysates for 16 h at room temperature. At the same time, microtiter plates (Sigma) were incubated with 0.5 μg of desmosine-albumin conjugate (Elastin Product Company, Inc, Owensville, MO) in 0.05 M sodium carbonate buffer pH 9.6 at 4°C. After incubation, wells were washed five times with 0.05% Tween 20 in 0.15 M PBS, pH 7.2 and saturated with 0.05% Tween 20 in 0.15 M PBS, 1% BSA pH 7.2 for 1 h at room temperature. Eightfold aliquots of AbI-standard or AbI-sample solutions were then added to the wells for a 2 h incubation at room temperature. Wells were then in succession incubated with anti-rabbit IgG (1:2000) (Sigma) for 2 h at room temperature and with peroxidase-antiperoxidase complex (1:200) (Sigma) for 1 h at room temperature. 2,2'-Azino-bis (3-ethyl-benz-thiazoline-6-sulfonic acid) solution (Sigma) was then added to the wells. After incubation for 1 h at room temperature, absorbance was read at 405 nm. Data were expressed as μg/lung.
Morphology and Morphometry
The lungs from the different groups of mice were fixed by intra-tracheal inflation with buffered formalin (5%) at a constant pressure of 20 cm H2O at least for 24 hours. All lungs were then dehydrated, cleared in toluene, and embedded in paraffin. Sagittal sections (7 μm) of each pair of lungs were cut and stained with hematoxylin/eosin and Masson's trichrome stain. Morphometric assessment consisted of the determination by point counting, of the percent volume densities of fibrosis Vv(f) and of the emphysematous changes Vv(e) according to the stereological principle of Glagoleff and Weibel [22]: Vv = Pp, where Vv is the volume density and Pp the fraction of points superimposed a defined structural change. Point counting was performed at 100 × by determining 20 random fields per slide and using a multipurpose grid to count 45 points per field for a total of 900 points per slide. Fibrosis was defined as: "inflammatory and mesenchymal cell infiltration within the alveolar septa and alveolar spaces with deposition of extracellular matrix", and emphysematous changes were defined as: "abnormal enlargement of air spaces with loss of alveolar septa, and with or without thickening of the alveolar walls" [4].
The morphometric assessment of emphysema was also performed in all animals by determining the average inter-alveolar distance (mean linear intercept: Lm) [17]. For the determination of the Lm for each pair of lungs, 40 histological fields were evaluated both vertically and horizontally. Care was taken to avoid histological fields containing large bronchi, major vessels and areas of fibrosis.
Immunohistochemistry
For the immunohistochemical studies, tissue sections (8 μm) were stained for TGF-β, TGF-α and NE by an immunoperoxidase method. The sections were pre-treated with 3% hydrogen peroxide to inhibit the activity of the endogenous peroxidase. For antigen retrieval, the sections were heated in a microwave for 20 min in citrate buffer 0.01 M, pH 6.0, and allowed to cool slowly to room temperature. All the sections were incubated with 3% bovine serum albumin for 30 min at room temperature to block non-specific antibody binding. They were then incubated overnight at 4°C with the primary antibodies (Ab). The primary polyclonal Ab used were: rabbit Ab to mouse TGF-a diluted 1:50 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA), rabbit Ab to mouse TGF-α diluted 1:20 (Insight Biotechnology LTD., Wembley, England). For elastase detection we used rabbit Ab to human neutrophil elastase (cross-reacting with mouse neutrophil elastase) diluted 1:500 (Calbiochem-Novabiochem, San Diego, CA). All the sections were rinsed and incubated with sheep anti-rabbit IgG for 30 min at room temperature. The staining was revealed by adding peroxidase-antiperoxidase complex, prepared from rabbit serum. Detection was accomplished by incubating in diamino-benzidine freshly dissolved in 0.03% H2O2 in 50 mM Tris-HCl pH 7.6. As negative controls for the immunostaining the primary antibodies were replaced by non-immunised rabbit serum. The sections were counterstained with hematoxylin.
Determination of elastase burden by immunoelectron microscopy
The immunogold method (post-embedding technique) was used to localize elastase in thin lung sections prepared for electron microscopy using anti-mouse neutrophil elastase (MNE) antibodies obtained as previously described [19]. Lung tissue blocks (5/animal) were taken from two animals from each group. The blocks, 1 to 2 mm in thickness, were fixed for 3 hours in 4% paraformaldehyde and 0.1% glutaraldehyde in 0.1 M phosphate buffer (pH 7.2), dehydrated in acetone, and embedded in epoxy resin (Araldite) without postfixation in OsO4. Ultrathin sections (600 Å thick) were picked up on nickel grids and pretreated with phosphate-buffered saline containing 1% ovalbumin for 5 minutes. The grids were then floated on a drop of diluted anti-MNE antibodies (1:2600) for 48 h at 4°C; the grids were thoroughly rinsed for 10 min with a mild spray of phosphate-buffered saline and then distilled water and transferred onto 15 μl drops of a protein A-gold particles (15 nm) (E-Y-LABS, San Mateo, CA) solution diluted 1:8 in phosphate-buffer saline. The sections were then washed, dried, stained with uranyl acetate-lead citrate, and examined in a Philips 300 electron microscopy. Ten to 12 micrographs (final magnification, × 12,000) were taken for each grid. To exclude false-positive labeling, a series of control studies (including also the use of nonimmune rabbit serum or of BSA instead of ovalbumin) were carried out as previously described in detail [19]. The density of gold particles per square micrometer of lung tissue was determined for each of the micrographs with a superimposed quadratic lattice grid. A total of 50 micrographs was thus analyzed for each animal, and the average of gold particle density on lung connective tissue of each group was calculated.
Statistical Analysis
The significance of differences of the mean values was calculated using one-way ANOVA (F-test). A p value of less than 0.05 was considered significant.
Results
Bleomycin Study
Emphysema and Fibrosis after Bleomycin
The kinetics of the emphysematous and fibrotic changes obtained in the three strains of mice are reported in Fig 1 A and 1 B. In particular, we present the percent values of lung volume densities of fibrotic (Vv(f)) and emphysematous (Vv(e)) changes obtained at the various time points. In BLM-resistant Balbc mice with a normal antiprotease protection, negligible foci of cellular infiltration and fibrosis and no areas of air-space enlargements were detected during the period of the study. On the contrary, a progressive increase of areas of air space enlargement and fibrosis were seen in BLM-prone C57 Bl/6 and pallid mice with a mild and a marked deficiency of α1-PI, respectively.
Figure 1 Emphysema and fibrosis after bleomycin challenge. The volume densities of emphysematous changes (Vv(e)) (A) and fibrosis (Vv(f)) (B) were quantitated by morphometry (point counting) on hematoxylin/eosin or Masson's trichrome stained lung sections, at various times after bleomycin. Data from 10 animals for each time point are given as mean ± SEM of per cent lung volume densities. *: p < 0.01 versus respective untreated controls (0 days).
At 3 and 7 days after BLM, C57 Bl/6 and pallid mice showed appreciable morphologic emphysema with spotty areas of inflammatory cell infiltration in the absence of fibrotic changes (Figs 2 and 3). At these times the anatomical emphysema was associated with a significant increase of the Lm (Fig. 4 A) a significant decrease in lung desmosine content (Fig. 4 B) and a high neutrophil elastase burden (Fig. 4 C).
Figure 2 Histological appearance of C57 BI/6 mouse lungs 3 and 7 days following bleomycin challenge. (A) Histologic section from the lung of a C57 Bl/6 mouse treated with saline showing a normal parenchyma. Representative histologic sections of C57 Bl/6 mice at 3 (B) and 7 (C and D) days after bleomycin treatment showing appreciable morphologic emphysema but not fibrosis. Scattered inflammatory cells are present through lung parenchyma (E). (E) Shows a higher magnification of (D). (A-C): Hematoxylin-eosin stain, scale bar represents 400 μm. (D) and (E): Masson's trichrome stain, original magnification × 40 and × 100, respectively. Scale bars represent 400 μm and 100 μm, respectively.
Figure 3 Histological appearance of pallid mouse lungs 7 days following bleomycin challenge. (A) Histologic section from the lung of a pallid mouse treated with saline showing a normal parenchyma. Lung sections of pallid mice at 7 days after bleomycin showing appreciable emphysema (B) and spotty areas of inflammatory cell infiltration without fibrosis (C). (A) and (B): Hematoxylin-eosin stain, scale bar represents 400 μm. (C): Masson's trichrome stain, scale bar represents 100 μm.
Figure 4 Mean linear intercepts, lung desmosine and elastase burden in various strains of mice following bleomycin challenge. (A) Mean linear intercepts (Lm) in Balb/C, C57 Bl/6 and pallid mice after bleomycin challenge. Data are from 10 animals for each time point and are given as mean ± SD. *: p < 0.01 versus respective saline-treated group. (B) Lung desmosine content in Balb/C, C57 Bl/6 and pallid mice after bleomycin treatment. Data from 10 animals for each time point are given as mean ± SD and represent per cent change over respective saline-treated controls (Balb/C: 2.50 ± 0.28 μg/lung; C57 BI/6: 2.48 ± 0.30 μg/lung; pallid: 2.44 ± 0.32 μg/lung). *: p < 0.01 versus respective saline-treated group. (C) Lung elastase burden in Balb/C, C57 Bl/6 and pallid mice at 7 days after bleomycin treatment. Data are given as mean ± SD of the number of gold particles per μm2. *: p < 0.01 versus respective saline-treated group.
At 14 days after BLM, air spaces enlargements affected 6,68 ± 3.11 % and 12.71 ± 4.17 % (p < 0.01) of lung in C57Bl/6 and pallid mice, respectively (Fig. 1 A). At this time, the lungs of C57 Bl/6 and pallid mice showed also large areas of fibrosis which involved 24.11 ± 8.81 and 36.84 ± 11.47 % of the lungs, respectively (Fig. 1 B). Both lesions were widely spread and intermixed (Fig. 5 A). Nevertheless, areas of emphysema could also be detected in lung lobes without any fibrotic reaction. Additionally, in several areas the emphysematous changes were distant from the fibrotic foci (Fig. 5 B and 5 C).
Figure 5 Histological appearance of pallid and C57 Bl/6 mouse lungs 14 days following bleomycin challenge. Representative lung histologic sections of a pallid (A) and a C57BI/6 (B) mouse at 14 days after bleomycin. Fibrotic and emphysematous areas are widely spread and intermixed. Emphysema is often located quite distant from the fibrotic reaction (A-B).(C) shows a higher magnification of (B). (A): Hematoxylin-eosin stain, scale bar represents 400 μm. (B) and (C): Masson's trichrome stain, scale bars represent 400 μm and 100 μm, respectively.
At later times (21 days onward), the volume density of emphysematous and fibrotic changes markedly increased (Fig. 1 A and 1 B). From a morphological point of view, the emphysematous lesions appeared to be mainly of paracicatricial type (Fig. 6 A and 6 B). Nevertheless, large areas of emphysema could also be found in lung lobes without obvious fibrosis or situated adjacent to the fibrotic lesions (Fig. 6 C).
Figure 6 Histological appearance of C57 BI/6 and pallid mouse lungs 21 days following bleomycin challenge. Histologic sections from lungs of a C57BI/6 (A and C) and pallid (B) mouse at 21 days after bleomycin. The emphysematous lesions within or adjacent to the fibrotic areas appear to be mainly of the paracicatricial type (A and B). Nevertheless, several areas of emphysema can be detected quite distant from the fibrotic zones (C). (A) and (C): Hematoxylin-eosin stain, scale bar represents 400 μm. (B): Masson's trichrome stain, scale bar represents 400 μm.
No significant changes in terms of emphysema or fibrosis were seen after BLM treatment in Balb/c mice with normal levels of serum αl-PI, by morphological, morphometrical and biochemical analysis.
Immunohistochemistry
Lungs of mice from the three strains were also analysed for cytokine expression by immuno-histochemistry. A significant change of some cytokines related to the NE activity was observed in mice with a mild (C57 Bl/6) and a marked deficiency (pallid) of αl-PI, after BLM administration. In particular, TGF-β was detected in subpleural foci of cellular proliferation at 7 days (Fig. 7 A) when an increase of the elastase burden could also be demonstrated. Also at 7 days, an evident staining for TGF-α was observed in subpleural and peribronchiolar areas (Fig. 7 B).
Figure 7 Immunohistochemical reaction for TGF-β and TGF-α 7 days following bleomycin challenge. Lung parenchyma of a C57BI/6 mouse at 7 days after bleomycin treatment. (A) Immunohistochemical reaction for TGF-β. Counterstained with hematoxylin, scale bar represents 40 μm. (B) Immunohistochemical reaction for TGF-α. Counterstained with hematoxylin, scale bar represents 25 μm.
Effects of a serine proteinase inhibitor treatment on BLM-induced lesions
The treatment of animals with 4-(2-aminoethyl)-benzenesulfonyl fluoride hydrochloride, a serine-proteinase inhibitor, significantly prevented the BLM-induced lesions in C57 Bl/6 mice. In particular, treated animals showed 14 days after BLM treatment no areas of emphysema and only trivial foci of fibrotic reaction (Fig. 8 A, and 8 B). The Lm values (40.75 ± 0.91 μm) and the lung (Vv(e)) (0.93 ± 0.89 %) in these mice were not significantly different from those observed in control mice (Lm: 39.71 ± 0.80 μm; Vv(e): 0.26 ± 0.74 %). In addition the lung (Vv(f)) (14.13 ± 6.21 %) in mice receiving BLM plus proteinase inhibitor was significantly lower (p < 0.01) than that observed in mice receiving only BLM (24.11 ± 9.41 %).
Figure 8 Histological appearance of C57 BI/6 lung receiving 4-(2-aminoethyl)-benzenesulfonyl fluoride hydrochloride 21 days after bleomycin challenge. Representative histological section of a C57BI/6 mouse, receiving 4-(2-aminoethyl)-benzenesulfonyl fluoride hydrochloride and bleomycin, at 14 days after the treatment. No appreciable areas of emphysema are detectable in the lung parenchyma (A). Few trivial foci of fibrosis can be appreciated in some areas (B). (B) shows a higher magnification of (A). (A) and (B): Masson's trichrome stain, scale bars represent 400 μm and 50 μm, respectively.
No immunological reaction for TGF-α and a faint positive staining TGF-β was found 7 days after BLM administration in animals treated with 4-(2-aminoethyl)-benzenesulfonyl fluoride hydrochloride (data not shown).
Cigarette Smoke Study
The results of the morphometric assessment at various time points are shown in Fig. 9. Already at 3 months of smoke exposure the DBA/2 mice showed overt emphysema (Fig. 10 B) characterized by significant changes both of the Lm (+ 19 %, p < 0.01) and of the internal surface area (ISA) (-16 %, p < 0.01) (Fig.9). At this time, immunohistochemical examination revealed a positive reaction for mouse NE on the alveolar septa (Fig. 10 C). Of interest, the first foci of fibrosis were seen after 4 months of smoke exposure (Fig. 11 A) and their severity progressively increased with time reaching at 6 months a (Vv(f)) value of 5.53 ± 2.11 %. At 6 months after smoke exposure, the fibrotic lesions consisted mainly of subpleural foci. In some areas, the fibrotic reaction was seen in the lung parenchyma associated or not with foci of emphysema (Fig. 11 B).
Figure 9 Mean linear intercepts and lung internal surface areas of DBA/2 mice at various time-points during chronic cigarette smoke exposure. Mean linear intercept (LM) (A) and internal surface area /ISA) (B) of the lungs of DBA/2 mice at various time-intervals during chronic exposure to cigarette smoke. Data from 8 animals for each time point are given as mean ± SD. * p < 0.05 versus air-exposed controls.
Figure 10 Lung histology and immunohistochemical reaction for neutrophil elastase in DBA/2 mice 3 months after chronic cigarette smoke exposure. Lung parenchyma from an air-exposed (A) and a smoke-exposed DBA/2 (B) mouse, at 3 months. Hematoxylin-eosin stain, scale bar represents 400 μm. (C) and (D): Immunohistochemical reaction for neutrophil elastase on alveolar septa of DBA/2 mouse at 3 months after cigarette smoke exposure in absence (C) or in presence of the primary antibody (D). Scale bar represents 25 μm.
Figure 11 Histological appearance of DBA/2 lungs 4 and 6 months after chronic cigarette smoke exposure. Lung parenchyma from smoke-exposed DBA/2 mice, at 4 (A) and 6 months (B). The first foci of subpleural fibrosis are seen from 4 months of smoke exposure (A). After 6 months of cigarette smoke exposure, disseminated foci of severe emphysema and evident areas of subpleural fibrosis are present (B). (A) and (B): Hematoxylin-eosin stain, Scale bars represent 25 μm and 400 μm. respectively.
A positive immunohistochemical reaction for TGF-β and TGF-α could be demonstrated starting from 3 months of smoke exposure onwards (Figs. 12 A and 12 B). In general, these cytokines were detected in foci of cellular proliferation and subsequently (from 4 months onward) in subpleural and parenchymal areas of fibrosis.
Figure 12 Immunohistochemical reaction for TGF-β and TGF-α 3 months after chronic cigarette smoke exposure. Lung parenchyma of a smoke-exposed DBA/2 mouse, at 3 months. (A) Immunohistochemical reaction for TGF-β. Counterstained with hematoxylin, scale bar represents 40 μm. (B) Immunohistochemical reaction for TGF-α. Counterstained with hematoxylin, Scale bar represents 15 μm.
Discussion
Although lung emphysema and fibrosis may result from two distinct and apparent opposite processes, they may coexist either in different areas, or in a same area of the lung of humans and animals. The development and the degree of these morphological responses that generally follow, or exacerbate, an acute or chronic inflammation can be influenced by many individual factors, such as cytokine production, variation in collagen synthesis and deposition, antiprotease screen and antioxidant status [23-30].
The findings reported in this paper strongly suggest that NE may represent a common factor affecting the development of both emphysema and fibrosis.
In particular, we demonstrate that in mice BLM-treated which are genetically deficient in αl-PI (i) emphysema and fibrosis may coexist either in different areas, or in a same lung area; (ii) the development of emphysema precedes that of fibrosis; (iii) the development of emphysematous lesions, shortly after BLM administration, is preceded by an alveolar elastolytic burden and is matched by a marked decrease in lung desmosine content; and finally, (iiii) an evident staining for TGF-β and TGF-α is observed when an increased neutrophil elastase burden can also be demonstrated. Similarly, we found in lungs of mice after cigarette-smoke exposure that (i) emphysema and fibrosis may be present in the same lung; (ii) the development of the emphysematous lesions occur at earlier time points than that of the fibrotic foci, and (iii) a positive immunohistochemical reaction for neutrophil elastase is associated with a positive reaction for TGF-β and TGF-α two major fibrogenic cytokines (i.e. [31,32] in foci of cellular proliferation, and in areas of fibrosis.
Taken all together these results indicate that the air-space enlargements observed in mice with a genetic deficiency of serum αl-PI, early after BLM, represent areas of "true" emphysema caused by a proteolytic attack and characterised by lung desmosine loss. The strong immunoelectron microscopical reaction for NE found on alveolar septa of αl-PI deficient mice early after BLM and in DBA/2 mice after cigarette smoke suggests that NE may represent a common factor affecting the development of both emphysema and fibrosis.
This hypothesis is further supported by the data obtained in BLM-treated C57 B1/6J mice in which both emphysema and fibrosis were significantly attenuated by the use of a serine proteinase inhibitor active against NE. Of interest, in these animals no immunological reaction for TGF-α and only a faint positive staining for TGF-β could be demonstrated. Although there is no ideal animal model, including the BLM one, that mimics human idiopathic pulmonary fibrosis, the data reported here support a role for proteases, and in particular for NE, in both these two pathologies.
It is well know that proteases released by inflammatory cells recruited at the site of inflammation may be involved in the intracellular as well as extracellular route of catabolism of interstitial proteins. These proteases may play an important role in tissue injury and repair by degrading the components of the extracellular matrix [33]. During the reparative responses they can remove scar tissue influencing in this way the morphological end-point.
It is generally accepted that NE plays a role in the development of emphysematous lesions. It acts on a large variety of substrates, in particular elastin, collagens, fibronectin, laminin and proteoglycans [23]. Recent studies suggest that this enzyme may modulate the fibrotic response also by interacting with the cytokine network. NE can activate or inactivate by proteolytic cleavage several cytokines, receptors and polypeptide growth factors implicated in inflammation and reparative phases of the fibrotic response [34-37].
In this regard, NE constitutes an important factor for the generation of soluble TGFα [32], a potent mitogenic cytokine for mesenchymal cells. In fact, TGFα is activated by the cleavage of the membrane precursor pro-TGFα by elastase [38,39].
Additionally, NE modulates TGF-β bioactivity [35] either directly by releasing TGF-β1 from the extracellular matrix [40] or indirectly via MMP-12 [41].
Of interest, we found that TGFα and TGFβ immunoreactivity was significantly high in mice that develop foci of subpleural fibrosis after cigarette-smoke or BLM treatment when a positive reaction for mouse NE on the alveolar septa was found. Although the association between the development of foci of subpleural fibrosis and a positive reaction for mouse NE on the alveolar septa does not prove a causal relationship, the data presented here strongly support the hypothesis that NE may represent a common pathogenic link between emphysema and fibrosis.
Conclusion
In conclusion the data reported in this paper strongly suggest a significant role for the NE not only in the development of pulmonary emphysema, but also in the modulation of fibrotic lesions. The results may also offer an explanation for the antifibrotic activity of some protease inhibitors (i.e. αl-PI, SLPI and the synthetic inhibitor ONO-5046) that are all active against NE.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
ML played a major role in the design of the study, acquisition, analysis and interpretation of data and drafting the manuscript. BB and BL performed the histological and morphometrical analyses, and contributed to the interpretation of the results. EC and SF performed histochemical and biochemical analyses, carried out animal studies and participated in the interpretation of data. PAM performed some of the morphometrical analysis and contributed to the interpretation of data. GL conceived and coordinated the study, participated in the design of the study, analysis and interpretation of data and drafting the manuscript. All authors read and approved the final manuscript.
Acknowledgements
This work was supported by a grant from MIUR (Rome, Italy) and partially by a PAR grant (University of Siena, Italy)
==== Refs
Bartalesi B Cavarra E Fineschi S Lucattelli M Lunghi B Martorana PA Lungarella G Different lung responses to cigarette smoke in two strains of mice sensitive to oxidants Eur Respir J 2005 25 15 22 15640318 10.1183/09031936.04.00067204
Lucey EC Ngo HQ Agarwal A Smith BD Snider GL Goldstein R Differential expression of elastin and αl (I) collagen mRNA in mice with bleomycin-induced pulmonary fibrosis Lab Invest 1996 74 12 20 8569173
Yaekashiwa M Nakayama S Ohnuma K Sakai T Abe T Satoh K Matsumoto K Nakamura T Takahashi T Nukiwa T Simultaneous or delayed administration of hepatocyte growth factor, equally represses the fibrotic changes in murine lung injury induced by bleomycin Am J Respir Crit Care Med 1997 156 1937 1944 9412578
Cavarra E Martorana PA Bartalesi B Fineschi S Gambelli F Lucattelli M Ortiz L Lungarella G Genetic deficiency of αl-PI in mice influences lung responses to bleomycin Eur Respir J 2001 17 474 480 11405528 10.1183/09031936.01.17304740
Niewoehner DE Hoidal JR Lung fibrosis and emphysema: divergent responses to a common injury? Science 1982 217 359 360 7089570
Lang MR Fiaux GW Gillooy M Stewart JA Hulmes DJS Lamb D Collagen content of alveolar wall tissue in emphysematous and non-emphysematous lungs Thorax 1994 49 319 326 8202900
American Thoracic Society American Thoracic Society/European Respiratory Society International Multidisciplinary Consensus Classification of the Idiopathic Interstitial Pneumonias Am J Respir Crit Care Med 2002 165 277 304 11790668
Tuder RM Voelkel NF Voelkel NF, MacNee W The pathobiology of chronic bronchitis and emphysema Chronic obstructive pulmonary disease 2002 Hamilton, Ontario, Canada: BC Decker Inc 90 113
Snider GL Martorana PA Lucey EC Lungarella G Voelkel NF, MacNee W Animal models of emphysema Chronic obstructive pulmonary disease 2002 Hamilton, Ontario, Canada: BC Decker Inc 237 256
Nagai A Aoshiba K Ishihara Y Inano H Sakamoto K Yamaguchi E Kagawa J Takizawa T Administration of αl-proteinase inhibitor ameliorates bleomycin-induced pulmonary fibrosis in hamsters Am Rev Respir Dis 1992 145 651 658 1372163
Mitsuhashi H Asano S Nonaka T Hamamura I Masuda KI Kiyoki M Administration of truncated secretory leukoprotease inhibitor ameliorates bleomycin-induced pulmonary fibrosis in hamsters Am J Respir Crit Care Med 1996 153 369 374 8542145
Taoka Y Maeda A Hiyama K Ishioka S Yamakido M Effects of neutrophil elastase inhibitor on bleomycin-induced pulmonary fibrosis Am J Respir Crit Care Med 1997 156 260 265 9230758
Merritt TA Cochrane CG Holcomb K Bohl B Hallman M Strayer D Edwards DK Gluck L Elastase and α1-proteinase inhibitor activity in tracheal aspirates during respiratory distress syndrome: role of inflammation in the pathogenesis of bronchopulmonary dysplasia J Clin Invest 1983 72 656 666 6603478
Dunsmore S Roes J Chua FJ Segal AW Mutsaers SE Laurent GJ Evidence that neutrophil elastase-deficient mice are resistant to bleomycin-induced fibrosis Chest 2001 120 35s 36s 11451907 10.1378/chest.120.1_suppl.S35
Delclaux C Delacourt C D'Ortho MP Boyer V Lafuma C Harf A Role of gelatinase B and elastase in human polymorphonuclear neutrophil migration across basement membrane A J Respir Cell Mol Biol 1996 3 288 295
Gardi C Cavarra E Calzoni P Marcolongo P de Santi MM Martorana PA Lungarella G Neutrophil lysosomal dysfunction in mutant C57B1/6J mice: interstrain variations in content of lysosomal elastase, cathepsin G and their inhibitors Biochem J 1994 299 237 245 8166647
Thurlbeck WM Measurement of pulmonary emphysema Am Rev Respir Dis 1967 95 752 764 5337140
Cocci F Miniati M Monti S Cavarra E Gambelli F Battolla L Lucattelli M Lungarella G Urinary desmosine excretion is inversely correlated with the extent of emphysema in patients with chronic obstructive pulmonary disease Int J Biochem Cell Biol 2002 34 594 604 11943590 10.1016/S1357-2725(02)00015-8
Cavarra E Martorana PA Gambelli F de Santi MM van Even P Lungarella G Neutrophil recruitment into the lung is associated with increased lung elastase burden, decreased lung elastin, and emphysema in α1-proteinase inhibitor-deficient mice Lab Invest 1996 75 273 280 8765327
Cavarra E Bartalesi B Lucattelli M Fineschi S Lunghi B Gambelli F Ortiz LA Martorana PA Lungarella G Effects of cigarette smoke in mice with different levels of αl-proteinase inhibitor and sensitivity to oxidants Am J Respir Crit Care Med 2001 164 886 890 11549550
Thurlbeck WM Internal surface area of non emphysematous lungs Am Rev Respir Dis 1967 95 765 773 6023510
Weibel ER Stereological principles for morphometry in electron microscopy cytology Int Rev Cytol 1969 26 235 302 4899604
McElvaney NG Crystal RG Crystal RG, West JB, Weibel ER, Barnes PJ Inherited susceptibility of the lung to proteolytic injury The Lung 1997 Philadephia, USA: Lippincott-Raven 2537 2553
Selman M King TE JrPardo A Idiopathic pulmonary fibrosis: prevailing hypotheses about its pathogenesis and implications for therapy Ann Intern Med 2001 134 136 151 11177318
Schrier DJ Kunkel RG Phan SH The role of strain variation in murine bleomycin-induced pulmonary fibrosis Am Rev Respir Dis 1983 127 63 66 6185026
Baecher-Allan CM Barth R PCR analysis of cytokine induction profiles associated with mouse strain variation in susceptibility to pulmonary fibrosis Reg lmmunol 1993 5 207 217
Reid PT Sallenave JM Cytokines in the pathogenesis of chronic obstructive pulmonary disease Curr Pharm Des 2003 9 25 38 12570672 10.2174/1381612033392440
Noble PW Idiopathic pulmonary fibrosis. New insights into the classification and pathogenesis Am J Respir Cell Mol Biol 2003 29 S27 S31 14503550
Morse D The role of heme oxygenase-1 in pulmonary fibrosis Am J Respir Cell Mol Biol 2003 29 S82 S86 14503562
Phan SH Fibroblast phenotypes in pulmonary fibrosis Am J Respir Cell Mol Biol 2003 29 S87 S92 14503563
Madtes DK Busby TP Standjord TP Clark JG Expression of transforming growth factor-alpha and epidermal growth factor receptor is increased following bleomycin-induced lung injury in rats Am J Respir Cell Mol Biol 1994 11 540 551 7524566
Sime PJ Marr RA Gouldie D Xing Z Hewlett BR Graham FL Gauldie J Transfer of tumor necrosis factor-alpha to rat lung induces severe pulmonary inflammation and patchy interstitial flbrogenesis with induction of transforming factor-beta 1 and myofibroblasts Am J Pathol 1998 153 825 832 9736031
Lucattelli M Cavarra E de Santi MM Tetley TD Martorana PA Lungarella G Collagen phagocytosis by lung alveolar macrophages in animal models of emphysema Eur Respir J 2003 22 728 734 14621076 10.1183/09031936.03.00047603
Libert C Van Molle W Brouckaert P Fiers W αl-Antitrypsin inhibits the lethal response to TNF in mice J Immunol 1996 157 5126 5129 8943423
Niehorster M Tiegs G Schade UF Wendel A In vivo evidence for protease-catalysed mechanism providing bioactive tumor necrosis factor α Biochem Pharmacol 1990 40 1601 1603 2222515 10.1016/0006-2952(90)90461-S
Porteu F Brockhaus M Wallach D Engelmann D Nathan CF Human neutrophil elastase releases a ligand-binding fragment from the 75-kDa tumor necrosis factor (TNF) receptor. Comparison with the proteolytic activity responsible for shedding of TNF receptors from stimulated neutrophils J Biol Chem 1991 266 18846 18853 1655765
Buczek-Thomas JA Lucey EC Stone PJ Chu CL Rich CB Carreras I Goldstein RH Foster JA Nugent MA Elastase mediates the release of growth factors from lung in vivo Am J Respir Cell Mol Biol 2004 31 344 350 15191913 10.1165/rcmb.2003-0420OC
Pandiella A Massague J Cleavage of the membrane precursor for transforming growth factor alpha is a regulated process Proc Natl Acad Sci USA 1991 88 1726 1730 2000380
Meyer-Hoffert U Wingertszahn J Wiedow O Human leukocyte elastase induces keratinocyte proliferation by epidermal growth factor receptor activation J Invest Dermatol 2004 123 338 345 15245434 10.1111/j.0022-202X.2004.23202.x
Taipale J Lohi J Kovanen PT Kestki-Oja J Human mast cell chymase and leukocyte elastase release latent transforming growth factor-beta 1 from the extracellular matrix of cultured human epithelial cells J Biol Chem 1995 270 4689 4696 7876240 10.1074/jbc.270.9.4689
Morris DG Huang X Kaminski N Wang Y Shapiro SD Dolganov G Glick A Sheppard D Loss of integrin αvβ6-mediated TGF-β activation causes MMP-12-dependent emphysema Nature 2003 422 169 173 12634787 10.1038/nature01413
|
16045796
|
PMC1184102
|
CC BY
|
2021-01-04 16:23:24
|
no
|
Respir Res. 2005 Jul 26; 6(1):83
|
utf-8
|
Respir Res
| 2,005 |
10.1186/1465-9921-6-83
|
oa_comm
|
==== Front
Respir ResRespiratory Research1465-99211465-993XBioMed Central London 1465-9921-6-861605353010.1186/1465-9921-6-86ResearchDetection of reactive oxygen species in isolated, perfused lungs by electron spin resonance spectroscopy Weissmann Norbert [email protected] Nermin [email protected] Beate [email protected] Vedat [email protected]äfer Rolf U [email protected]ütte Hartwig [email protected] Hossein A [email protected] Ralph T [email protected] Christian [email protected] Akylbek [email protected] Bakytbek [email protected] Werner [email protected] Friedrich [email protected] Justus-Liebig University, Department of Internal Medicine II, Klinikstrasse 36, 35392 Giessen, Germany2 Charite, Department of Internal Medicine, Humboldt-University, 13353 Berlin, Germany3 ALTANA Pharma, 78467 Konstanz, Germany2005 31 7 2005 6 1 86 86 2 5 2005 31 7 2005 Copyright © 2005 Weissmann et al; licensee BioMed Central Ltd.2005Weissmann et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The sources and measurement of reactive oxygen species (ROS) in intact organs are largely unresolved. This may be related to methodological problems associated with the techniques currently employed for ROS detection. Electron spin resonance (ESR) with spin trapping is a specific method for ROS detection, and may address some these technical problems.
Methods
We have established a protocol for the measurement of intravascular ROS release from isolated buffer-perfused and ventilated rabbit and mouse lungs, combining lung perfusion with the spin probe l-hydroxy-3-carboxy-2,2,5,5-tetramethylpyrrolidine (CPH) and ESR spectroscopy. We then employed this technique to characterize hypoxia-dependent ROS release, with specific attention paid to NADPH oxidase-dependent superoxide formation as a possible vasoconstrictor pathway.
Results
While perfusing lungs with CPH over a range of inspired oxygen concentrations (1–21 %), the rate of CP• formation exhibited an oxygen-dependence, with a minimum at 2.5 % O2. Addition of superoxide dismutase (SOD) to the buffer fluid illustrated that a minor proportion of this intravascular ROS leak was attributable to superoxide. Stimulation of the lungs by injection of phorbol-12-myristate-13-acetate (PMA) into the pulmonary artery caused a rapid increase in CP• formation, concomitant with pulmonary vasoconstriction. Both the PMA-induced CPH oxidation and the vasoconstrictor response were largely suppressed by SOD. When the PMA challenge was performed at different oxygen concentrations, maximum superoxide liberation and pulmonary vasoconstriction occurred at 5 % O2. Using a NADPH oxidase inhibitor and NADPH-oxidase deficient mice, we illustrated that the PMA-induced superoxide release was attributable to the stimulation of NADPH oxidases.
Conclusion
The perfusion of isolated lungs with CPH is suitable for detection of intravascular ROS release by ESR spectroscopy. We employed this technique to demonstrate that 1) PMA-induced vasoconstriction is caused "directly" by superoxide generated from NADPH oxidases and 2) this pathway is pronounced in hypoxia. NADPH oxidases thus may contribute to the hypoxia-dependent regulation of pulmonary vascular tone.
==== Body
Background
Reactive oxygen species (ROS) play an important role in biological systems. While it is largely accepted that ROS-mediated oxidative damage occurs under pathophysiological conditions, recent evidence also favors a role for ROS as signaling molecules in physiological processes [1-6]. In the respiratory system, ROS are involved in pathological states such as hyperoxia-induced lung injury, sepsis and ischemia-reperfusion, but may also play a role in the development of hypoxia-and non-hypoxia-induced pulmonary hypertension [7-9]. In the context of regulatory processes of the lung, a growing body of evidence is emerging indicating that ROS may contribute to signaling events such as underlying hypoxic pulmonary vasoconstriction (HPV), an essential mechanism matching lung perfusion to ventilation in order to optimize pulmonary gas exchange [10,11]. The question of whether ROS generation is decreased or paradoxically increased during alveolar hypoxia, and the sources from which ROS may be derived under these conditions, are controversial [8,12]. In fact, substantial evidence exists suggesting both decreases as well as increases, in lung ROS generation during alveolar hypoxia [13-15]. Both mitochondria and NADPH oxidases have been proposed as source(s) of ROS generation underlying HPV [12,13,15-21].
One possible reason for the discrepancies among the different concepts of hypoxia-dependent ROS generation in the lung may be the inadequacies of the methods applied. For example, most of the evidence provided for increased ROS generation on the cellular level is based on measurements with fluorochromes, such as dihydrorhodamin 123 or 2',7'-dichlorofluorescin diacetate, which have been shown to be ROS generators themselves under various conditions [22,23]. In isolated perfused lungs, luminol and lucigenin, cytochrome c reduction, as well as spin trapping with 5,5-dimethyl-l-pyrrorine-N-oxide or sodium 3,5-dibromo-4-nitrosobenzenesulfonate [14,24] have been used for detection of ROS. These methods are, however, also prone to pitfalls due to autoxidation of the substrates, artificial ROS generation by redox cycling [22,24,25] or the fact that the product interferes with ascorbate in the tissue [26]. The ROS measurements have also been performed in rabbit lungs with the spin probe sodium 3,5-dibromo-4-nitrosobenzenesulfonatesodium [27]. However, in this investigation, the spin probe was added to the effluent of the lung, which did not allow detection of ROS release directly at its source in the lung vasculature [27].
To overcome several of these problems and to investigate in particular the role of hypoxia in NADPH oxidase-derived superoxide release and its role in the regulation of pulmonary vascular tone, we established a method combining ESR spectroscopy with the spin trapping technique for measurement of superoxide release from isolated perfused and ventilated rabbit and mouse lungs. Superoxide was detected by the cyclic hydroxylamine l-hydroxy-3-carboxy-2,2,5,5-tetramethylpyrrolidine (CPH) which was recently introduced for quantitative ROS measurements in biological systems [28]. Employing this agent, ROS formation can be quantified by detection of the corresponding nitroxide radical 3-carboxy-proxyl (CP•) by ESR spectroscopy [29,30]. After having proven the feasibility of CPH for ROS detection in the isolated perfused and ventilated rabbit lungs, with particular attention being paid to the autoxidation of CPH, we investigated the oxygen-dependence of intravascular ROS release in the intact lungs. Moreover, we quantified CPH oxidation over a range of inspired oxygen concentrations, while stimulating NADPH oxidases with phorbol-12-myristate-13-acetate (PMA). These experiments were performed in the absence and presence of superoxide dismutase (SOD) to estimate the proportion of superoxide-induced CPH oxidation. We thus demonstrated the feasibility of the ESR technique, and obtained interesting new insights into the oxygen-dependence of baseline versus PMA/NADPH oxidase-dependent superoxide generation in the intact lung vasculature.
Methods
Chemicals and reagents
l-hydroxy-3-carboxy-2,2,5,5-tetramethylpyrrolidine (CPH) was purchased from L-Optik (Berlin, Germany). Krebs-Henseleit buffer contained 125.0 mM NaCl, 4.3 mM KC1, 1.1 mM KH2PO4, 2.4 mM CaCl2, 1.3 mM MgCl2 and 275 mg glucose per 100 ml. FeCl2, deferoxamine (DFO), diethyldithiocarbamate (DETC), phorbol-12-myristate-13-acetate (PMA), and Cu/Zn-superoxide dismutase (SOD) were obtained from Sigma (Deisenhofen, Germany). Apocynin was from Merck Biosciences (Schwalbach, Germany).
In-vitro CPH experiments
For studying CPH characteristics and the effect of metal chelating agents on background ESR signals, experiments were performed in glass tubes containing 1 mM CPH solution prepared in Krebs-Henseleit buffer at room temperature. The iron chelator DFO was added in the concentrations 20 μM or 2 mM to the solution and measurements were run every 30 min for a total period of 5 h.
Isolated rabbit lung experiments
The technique of isolated lung perfusion and ventilation has been described previously in detail [31,32]. Briefly, pathogen-free New Zealand white rabbits of either sex (body weight 2.5–3.2 kg) were deeply anesthetized with ketamine (30–50 mg/kg body weight) and xylazine (6–10 mg/kg body weight), and anticoagulated with heparin (1000 U/kg body weight). The lungs were excised while perfused with Krebs-Henseleit buffer through cannulas in the pulmonary artery and the left atrium. After the lungs were rinsed with at least 1 1 of buffer fluid for removal of blood, the perfusion circuit was closed for recirculation (total system volume 150 ml) and left venous pressure set at 2 mmHg to secure West zone III conditions for perfusion. In parallel with the onset of artificial perfusion, room air ventilation was changed to a mixture of 5.3 % CO2, 21.0 % O2 and the balance N2 (tidal volume, 30 ml; frequency, 30 strokes/min) with the use of positive end-expiratory pressure of 1 cm H2O. The pH was adjusted to 7.35 – 7.40 by addition of NaHCO3. The isolated lungs were placed in a temperature-equilibrated housing chamber and the whole system was heated to 38.5°C.
Krebs-Henseleit buffer was incubated with 5 μM diethyldithiocarbamate overnight to allow sedimentation. Briefly before the experiments, 20 μM DFO and NaHCOs were added to the supernatant.
Lungs included in the study were those that i) had a homogeneous white appearance with no signs of hemostasis, edema or atelectasis, ii) revealed constant mean pulmonary artery and peak ventilation pressure in the normal range, and iii) were isogravimetric during the initial perfusion period of 55 min.
For normoxic and hypoxic ventilation maneuvers, a gas-mixing chamber (KM 60-3/6 MESO, Witt, Witten, Germany) was employed for step changes in the ventilator O2 content. 5.3 % CO2 were used throughout and the percentage of N2 was balanced accordingly.
After the initial steady state period with normoxic ventilation (21 % O2), CPH (1 mM) was added to the buffer fluid. Five minutes later, one of three different protocols was initiated:
1) A 2.5 or 3.0 h period of normoxic ventilation with 21 % O2, in the case of a 3 h ventilation period followed by a bolus application of PMA into the pulmonary artery, resulting in a concentration of 1 μM in the recirculating buffer fluid. Control experiments received a bolus of saline instead of PMA,
2) Consecutive 30 min periods of ventilation with different O2 concentrations (21, 16, 10, 5, 2.5 or 1 % O2 in a randomized fashion) for a total period of 3 h, or
3) One 30 min-period of ventilation at either 21, 16, 10, 5, 2.5 or 1 % O2, followed by a bolus application of PMA into the pulmonary artery, resulting in a concentration of 1 μM in the recirculating buffer fluid. Control experiments received a bolus of NaCl instead of PMA.
In a portion of the experiments a fiber oxygenator (Hilite 1000, Stolberg, Germany) was used instead of the lung for oxygenation of the buffer fluid.
Isolated mouse lung experiments
Mouse lung experiments were performed in a protocol analogous to the isolated rabbit lung experiments but in an in-chest preparation as previously described [33]. Lungs were perfused with 0.5 mM CPH at a flow rate of 2.0 ml/min for 120 min at normoxic ventilation (21 %O2), followed by a bolus application of PMA into the buffer fluid, resulting in a concentration of 10 μM. For these investigations either C57/BL6 mice (= wildtype control) or mice lacking the NADPH oxidase subunit gp91phox (= p91phox-/-). Mice were obtained from The Jackson Laboratory (Bar Harbor, Maine, USA).
ESR measurements
Oxidation of the spin probe CPH by superoxide forms the nitroxide CP radical. The triple-line spectrum of CP radical was detected by ESR spectroscopy, using a MS 100 spectrometer (Magnettech, Berlin, Germany). The ESR measurements were performed in field scan with the following settings: microwave frequency 9.78 GHz, modulation frequency 100 kHz, modulation amplitude 2 G, microwave power 18 mW. All samples, both from the in vitro experiments or from the venous outflow of the isolated lung, were made in 50 μl glass capillary tubes and measured immediately at room temperature. The ESR amplitude is in proportion to the amount of CP•, reflecting the interaction of ROS with CPH [29,34,35]. Thus, the quantity of trapped ROS was directly calculated from the ESR spectrum of the probe, while the contribution of superoxide radical to the formation of CP was determined in parallel experiments performed in the presence of SOD in the buffer fluid (150 U/ml). The first sample was taken 5 min after CPH addition to the buffer fluid (time set at zero), followed by further sampling every 5 or 30 minutes, as appropriate. The values after PMA addition were assessed every minute in the respective experiments. In the isolated mouse lung samples were taken every 2 min. For quantification, the second-field component of the ESR spectrum was used. To standardize values, the amplitude of this component was divided through the receiver gain.
Statistical analysis
Data are given as mean ± standard error (SEM). For comparison of two groups, a two-tailed t-test was employed. For multiple comparisons, analysis of variance was used, followed by the Student-Newman-Keuls post hoc test when differences were indicated. Statistical significance was assumed when p < 0.05.
Results
When CPH (1 mM) was dissolved in Krebs-Henseleit buffer ESR spectroscopy resulted in a triple band spectrum (Fig. 1A). The ESR signal increased to 87.06 ± 11.17 arbitrary units, (AU, n = 8) within 5 h when incubated in-vitro at room temperature (Fig. 1B). The presence of the iron chelating agent DFO at a concentration of 20 μM or 2 mM, from the beginning of the experiment, reduced the ESR amplitude ~8-fold to 11.66 ± 0.44 (n = 5) or 9.28 ± 0.57 (n = 4) AU within 5 h, with no major difference being observed between the high and the low concentration of DFO (Fig. 1B). Pre-incubation of the buffer with the copper chelator DETC (5 μM) further reduced the increase in signal (data not shown). In line with these in vitro data, ESR signal intensity increased when isolated rabbit lungs were perfused under normoxic ventilation with CPH (1 mM) for 2.5 h (Fig. 2). Perfusion with FeCl2 (1 μM) resulted in a markedly higher ESR signal intensity (Fig. 2). Addition of H2O2 to the buffer fluid (8 μmol/min, started after 1.5 h), in the presence of FeCl2, induced a strong increase in ESR signal intensity, reaching a value of 707.14 ± 103.15 AU (n = 4) within 1 h of lung perfusion (Fig. 2).
Figure 1 The effect of the iron chelating agent deferoxamine (DFO) on CP• nitroxide signal intensity in vitro. (A) Typical ESR spectrum of CP• nitroxide resulting from the reaction of the hydroxylamine spin probe CPH with ROS. The height of the first field component of the triple-line spectrum was used for quantification of signal intensity. (B) In-vitro incubation of CPH (1 mM) in Krebs-Henseleit buffer. Signal intensity is given in arbitrary units (AU). Data are shown for CPH oxidation in the absence (-DFO) or in the presence of either 20 μM or 2 mM deferoxamine (DFO). Asterisks indicate significant differences when compared to the-DFO group.
Figure 2 The ESR signal intensity in isolated perfused and ventilatedrabbit lungs during baseline conditions and in the presence of FeCl2/H2O2. Lungs were either perfused with Krebs-Henseleit buffer containing 1 mM CPH only (control) or in the presence of FeCl2 (1 μM, added at 0.5 h, FeCl2/H2O2). After 1.5 h, H2O2 was added to the buffer fluid by continuous infusion (8 μmol/min) in the FeCl2-perfused lungs. Bar indicates significant differences between the FeCl2/H2O2 group and control.
Performing a sequential mode of 30 min periods of ventilation maneuvers with different inspiratory oxygen concentrations (1 % – 21 % O2) revealed oxygen-dependent changes in ESR signal intensity. The ESR signals demonstrated a linear increase during the last 20 min of each maneuver with samples taken every 5 min. This time-dependent increase in the ESR signal intensity was highest when ventilating the lungs with 21 % O2 and lowest when ventilating the lungs with 2.5 % O2 (Fig. 3A). There was a tendency towards increased values found when the lungs were ventilated with 1 % O2, compared with 2.5 % O2. Parallel experiments in the presence of SOD (150 U/ml), revealed that a portion of the signal was derived from superoxide. Figure 3B depicts the effect of the ventilation maneuvers with the different oxygen concentrations on pulmonary artery pressure (PAP). Hypoxic ventilation induced an increase in PAP, which was highest at 1 % and lowest at 16 % O2.
Figure 3 Hypoxia-dependent superoxide release and vasoconstrictor responses in isolated perfused and ventilated rabbit lungs. During a total 3 h period of perfusion, the ventilator gas supply was switched to different oxygen concentrations every 30 min, using 1 %, 2.5 %, 5 %, 10 %, 16 % or 21 % O2 in a randomized mode. (A) Superoxide release. The rate of increase in ESR signal intensity turned out to be linear during the last 20 min of each ventilation period. Therefore, changes in the ESR signal intensity/min during the last 20 min of each ventilation period are given. In the +SOD group, 150 mU/ml superoxide dismutase (SOD) was present throughout the experiments. * significant differences between the +SOD and the -SOD group for the respective oxygen concentration. ** significant differences as compared to 21 % O2. (B) Vasoconstrictor response. The maximum increases in PAP (ΔPAP) are given for the different oxygen concentrations and are referenced to baseline (=normoxic) PAP values, which were 8.8 ± 0.6 mmHg
In Fig. 4A, the time-dependent increase of CP• formation is illustrated for long-term normoxic ventilation (3 h, 21% O2) of isolated lungs. After 3 h, a bolus of PMA, resulting in a concentration of 1 μM in the recirculating buffer fluid, was injected into the pulmonary artery. This application induced a rapid increase in the ESR signal intensity, which was almost entirely blocked by the addition of SOD. As the time-dependent increase in ESR signal intensity was linear before and after addition of PMA, increase rates, pre-and post-intervention, may be directly compared (Fig. 4B). While almost fully suppressed by SOD, a five-fold increase in CP• formation per unit time occurred in response to the PMA stimulation. For controls, we performed corresponding experiments, in which the lung was replaced by a fiber oxygenator, in order to exclude a direct effect of PMA on CPH oxidation. No response to PMA was observed in the absence of the lung vasculature either in the presence or in the absence of 1 μM FeCl2 (Fig. 4B).
Figure 4 Superoxide release from isolated perfused rabbit lungs afteraddition of phorbol-12-myristate-13-acetate (PMA) during normoxic ventilation. (A) ESR signal intensity during normoxic ventilation of rabbit lungs in the presence (+SOD) and the absence of SOD (-SOD). After 3 h, PMA was injected into the pulmonary artery, resulting in a concentration of 1 μM in the recirculating buffer fluid. The increase in ESR signal intensity was linear before and after addition of PMA. The insert shows the PMA effect with higher time resolution. (B) Changes in the increase rate of the ESR signal intensity (%) by comparison of values prior to and after addition of PMA to isolated rabbit lungs. In the +SOD group, SOD was present throughout the experiments. In two separate sets of experiments, a fiber oxygenator was used instead of the lung for oxygenation of the recirculating buffer fluid ("fiber oxygenator"). The fibre oxygenator experiments were performed either in the absence ("fiber oxygenator") or in the presence of 1 μM FeCl2 ("fiber oxygenator + FeCl2") in the buffer fluid. * significant difference between the +SOD and the -SOD group
The PMA-induced increase in ESR signal intensity in the isolated lung experiments was inhibited by the NADPH oxidase inhibitor apocynin but not by rotenone, an inhibitor of mitochondrial complex I (Fig. 5A). Moreover, in mice lacking the NADPH oxidase subunit gp91phox, the PMA-induced increase in the ESR signal intensity was prevented, when compared to wildtype mice (Fig. 5B). Corresponding to the SOD-inhibitable increase in ESR signal intensity in the perfused rabbit lungs, PMA application induced a vasoconstrictor response, which was also largely attenuated in the presence of SOD as well as in the presence of apocynin (table 1).
Figure 5 Effects of the NADPH oxidase and mitochondrial inhibitors, as well as of phagocytic NADPH oxidase gene deletion on the ESR signal intensity in isolated perfused lungs after addition of phorbol-12-myristate-13-acetate (PMA) during normoxic ventilation. (A) Effect of the NADPH oxidase inhibitor apocynin (500 μM) and the inhibitor of mitochondrial complex I, rotenone (350 nM) on the increase rate in ESR signal intensity after addition of PMA. Each inhibitor was added to the buffer fluid 30 min before addition of PMA. In the +SOD group, SOD was present throughout the experiments. * significant difference as compared to control. (B) Comparison of the increase rate in ESR signal intensity after addition of PMA in wildtype (WT) and gp91phox-deficient (gp91phox-/-) mice. Lungs were perfused for 120 min prior to PMA addition (10 μM) either in the presence or absence of 150 U/ml SOD. * significant difference as compared to WT. Data are given as changes in the increase rate of the ESR signal intensity after PMA addition, as compared to the values before PMA addition (set as 100%). Data are from n = 4–5 experiments for each group.
Table 1 Baseline pulmonary artery pressure (PAP) and phorbol-12-myristate-13-acetate (PMA)-induced changes in PAP after 3 hours of normoxic ventilation. Pulmonary artery pressure values (PAP) are given for time points directly before and 6 min after addition of PMA to the buffer fluid. In addition, the increase in PAP per minute (ΔPAP/min) after PMA addition is indicated. Data are shown for experiments in the absence (-SOD) and the presence of 150 U/ml superoxide dismutase (+SOD). The PMA was added after 3 h of normoxic ventilation. Values of the +SOD and -SOD group correspond to experiments in Fig. 4. In the apocynin group this agent was present in the perfusate at a concentration of 500 μM. In experiments without addition of PMA, no significant change in PAP was observed within 6 min after PMA addition. Asterisk indicates significant difference of the -SOD as compared to the +SOD or apocynin group.
directly before PMA [mmHg] 6 min after PMA ΔPAP/min after PMA n
-SOD 10.4 ± 0.4 15.2 ± 1.1 0.66 ± 0.18 5
+SOD 8.9 ± 0.7 9.9 ± 0.7 0.15 ± 0.02 4 *
apocynin 8.2 ± 0.4 8.3 ± 1.0 0.03 ± 0.01 4 *
To investigate the effect of inspired oxygen concentration on the PMA-induced increment in CP• formation, we performed short-term ventilation periods of 30 min at varying O2 concentrations, followed by bolus application of PMA (1 μM) into the pulmonary artery. The increases in the ESR signal intensity/min post-PMA was compared to that pre-PMA application (Fig. 6A). The highest increase in ESR signal intensity/time was observed when the lungs were ventilated with 5 % O2, within a total range of 1 – 21 % O2. Parallel experiments in the presence of SOD, revealed that the major portion of the increase in the ESR signal was due to superoxide release. Again, replacement of the lungs by a fiber oxygenator, to equilibrate the buffer fluid with the different oxygen concentrations did not change the increase rate in ESR signal intensity after PMA application, either in the absence or in the presence of 1 μM FeCl2 (Fig. 6A). PMA application induced an increase in PAP (Fig. 6B). This increase was highest when lungs were ventilated with 5 % O2.
Figure 6 Oxygen-dependency of PMA-induced changes in ESR signal intensity, lung superoxide release, and pulmonary artery pressure. Lungs were ventilated for 30 minutes with either 1 %, 2.5 %, 5 %, 10 %, 16 % or 21 % O2, followed by injection of PMA into the pulmonary artery, resulting in a concentration of 1 μM in the recirculating buffer fluid. (A) Changes in the increase rate of the ESR signal intensity after PMA addition, as compared to the values before PMA addition (set as 100 %) are given. Experiments were performed in the presence (+SOD) or the absence (-SOD) of SOD. In the control group, lungs were replaced by a fiber oxygenator for equilibration of the buffer fluid with the different oxygen concentrations. The fibre oxygenator experiments were performed either in the absence or in the presence of 1 μM FeCl2 in the buffer fluid. * significant differences between +SOD and -SOD groups. ** significant differences between different oxygen concentrations. (B) Increase in pulmonary artery pressure per minute (ΔPAP/min) within 7 min after PMA addition to the buffer fluid. In experiments without addition of PMA no significant change in PAP was observed. Data are from n = 4–9 experiments for each group.
Discussion
Methodological aspects
Because ROS formation has been implicated in a variety of lung diseases, the detection of ROS release from the pulmonary circulation is thought to be important [7-9]. However, current methods for the detection of lung vascular ROS formation lack specificity due to autoxidation of the substrates employed, the lack reliability due to artificial ROS generation by redox cycling [22,24,25], and a lack sensitivity of the techniques employed. The ESR technology may overcome several of these shortfalls. In a previous study, Katz and colleagues used spin trapping and ESR technology for ROS detection in isolated perfused rabbit lungs. However, in that investigation, the spin probe was added to the lung effluent, and was therefore unable to detect ROS release at its source within the vasculature, which would be desirable due to the short half-life of ROS [27]. We therefore favored the strategy to add the spin probe directly to the buffer fluid, and employed the cyclic hydroxylamine spin probe CPH, as this is an effective ROS scavenger allowing quantitative measurement of ROS formation [29,30]. Even under conditions of hypoxia-induced vasoconstriction the spin probe is expected to reach all parts of the lungs as these were perfused at constant flow under zone III conditions [31]. The reaction rate constant of CPH with ROS is 3.2 × 103 M-1 s-1, which is approximately 10-fold higher than the reaction rate constant of nitrone spin traps such as DMPO or DEPMPO [30,36]. The incorporation of CPH into the cell membrane is low, permitting analysis of extracellular ROS release, while the high stability of the formed nitroxides allowed quantitative measurement of the reaction product in the lung effluent [30,35].
CPH was, however, reported to be autoxidized through traces of transition metals, such as iron and copper, which can be eliminated by chelators [37]. Therefore, we first characterized CPH properties incubated in vitro in Krebs-Henseleit-buffer, which is routinely used for perfusion of the isolated lungs. Measurement of aliquots from this solution resulted in the typical triple-band spectrum [29,30]. The prominent reduction in the increased ESR signal intensity observed within a five-hour incubation period, in the presence of the iron chelator DFO, confirmed autoxidation of CPH catalyzed by traces of transition metals in the buffer fluid. This interpretation is also supported by the observation that the increase in signal intensity during perfusion of CPH through the lung is further enhanced by the addition of FeCl2 to the buffer fluid. No major difference was found between low and high DFO concentrations, and preincubation of the buffer with the copper chelator DETC further reduced CPH autoxidation. We therefore conducted all subsequent experiments in the presence of 20 μM DFO after preincubation of the buffer with DETC (5 μM). The DETC inhibits Cu, Zn -SOD. However, inhibition was only observed when much higher concentrations (>1 mM) were applied compared with the concentration employed in our study [38]. These findings are supported by other investigations from our laboratory, demonstrating that HPV in isolated rabbit lungs is not altered even at DECT concentrations 20 times higher than those employed in the present study [21].
When the DFO-and DETC-treated buffer was employed for perfusion of isolated and ventilated rabbit lungs, we first demonstrated that CPH is suitable for detection of ROS by the admixture of H2O2 to the buffer fluid, which resulted in a prominent increase in ESR signal intensity. This increase is most likely related to the generation of hydroxyl radicals (OH•) via Fenton reaction [39-41]. Although these data demonstrate the suitability of the technique, it became apparent that continuous oxidation of CPH circulating through the lungs occurs, which may depend on ongoing ROS release, or may reflect autoxidation independent of lung ROS liberation.
Oxygen-dependency of superoxide release
To investigate the dependency of the ESR signal increase on the inspired oxygen concentration, successive 30 min ventilation periods with different O2 concentrations of the inspired gas were performed. We found a decreased rate of ESR signal enhancement at lower oxygen concentrations, with a minimum at 2.5 % O2, whereas 1 % O2 caused again some increase. Addition of SOD to the buffer fluid reduced the rate of ESR signal increase at all oxygen concentrations, proving that at least a portion of the ESR signal detected is caused by intravascular superoxide release from the lungs. Several enzyme systems may be responsible for oxygen-dependent release of superoxide into the lung vasculature, including the NADPH-oxidases, xanthine-oxidase, nitric oxide synthases and arachidonic acid oxygenase pathways, as well as the mitochondrial respiratory chain [9,13,14,20]. Moreover, non-superoxide-derived ROS may add to the baseline increase in CPH oxidation, not suppressed by SOD [35]. As extracellular SOD is not thought to enter the intracellular milieu [42,43], intracellular trapping of superoxide by CPH may further add to this finding. However, as only low penetration of cell membranes was described for CPH this is less likely [34,44,45].
Ventilation of the lungs with oxygen concentrations lower than 21 % resulted in increased pulmonary artery pressure, with the changes in proportion to the relative O2 decrease being highest between 10.0 and 2.5 %, as previously reported for isolated rabbit lungs [31]. This hypoxia-induced vasoconstriction is known as a key feature of the lung, matching perfusion to ventilation in order to optimize pulmonary gas exchange by excluding poorly or non-ventilated areas from blood flow. Among other concepts, lung ROS generation has been suggested to be involved in this mechanism [15,46]. However, there is ongoing discussion about whether an increase or a decrease in ROS occurs during hypoxia, and from what source hypoxia-induced ROS may be derived [12,14,17,18,20,21]. As possible sources for both a decrease and an increase, mitochondria or NADPH-oxidases have been suggested [8,12]. With regard to these aspects it has not yet been determined whether hypoxia can indeed cause a paradoxical increase in NADPH oxidase-dependent superoxide release in the pulmonary circulation. The fact that extracellular SOD inhibited the ESR signal, but not the PAP increase under hypoxic conditions [8,32] can be explained by intracellular superoxide production, e.g. in the vascular smooth muscle cells, underlying the mechanism of hypoxic vasoconstriction, but not being a major contributor to the intravascular superoxide leak. Thus, the current focus on the intravascular compartment for ROS trapping may not be a suitable approach for analyzing ROS formation that underlies hypoxic pulmonary vasoconstriction. It may, however, be appropriate to measure ROS formation under conditions primarily affecting the endothelial cells, such as endothelial injury and ischemia-reperfusion conditions. Therefore, we targeted key enzyme systems that may be involved in endothelial ROS generation.
NADPH oxidase-dependent superoxide release and pulmonary vasoconstriction
To provoke a more pronounced ROS formation in the vascular compartment and to investigate the oxygen-dependence of NADPH oxidase-derived superoxide release, we employed intravascular administration of PMA, which was previously shown to stimulate NADPH oxidases via activation of protein kinase C [27]. Indeed, PMA induced a prominent increase in the rate of CPH oxidation, mostly attributable to intravascular superoxide release, as evident from a total block of this increase in the presence of SOD. Replacement of the lung by a fiber oxygenator to mimic oxygenation of the buffer fluid, as would occur in the lung, assured that no lung-independent oxidation of CPH was provoked by PMA, neither in the absence nor in the presence of FeCl2. Thus an overlapping effect of metal ions primarily being responsible for the oxygen-dependent effects seen in the presence of the lung as e.g. results from a Fenton reaction, can be excluded. The PMA-induced increase in the ESR signal was illustrated in our study to be attributable to the suggested pathway of NADPH oxidase stimulation, because it was prevented a) by the NADPH oxidase inhibitor apocynin as well as b) in mice lacking the NADPH oxidase subunit gp91phox (Nox-2). In contrast, rotenone, a mitochondrial complex I inhibitor, did not affect the PMA induced ROS release. This indicated that mitochondria-derived superoxide does not play a role in the oxygen-dependent ROS release induced by PMA. This finding is of particular interest, given the recent reports of mitochondria as possible sources of superoxide release [17]. Moreover, PMA caused an immediate pulmonary artery pressor response, which was also largely blocked by SOD, suggesting a direct vasoconstrictor effect of superoxide generated by PMA addition. This suggestion is in line with the inhibition of the vasoconstrictor response by the NADPH oxidase inhibitor apocynin. The fact that PMA stimulation of the lung induces a vasoconstrictor response via superoxide challenges previous studies suggesting that the PMA-induced vasoconstrictor response involves a Ca2+ sensitization by inhibition of myosin light chain phosphatase (for review see [47]). The superoxide-induced vasoconstriction in this pathway may involve intracellular calcium mobilization by enhancing cyclic ADP-ribose production [48], activation of RhoA/Rho kinase [49], or inactivation of NO [50] by superoxide. To investigate the oxygen-dependence of the PMA-induced superoxide release, we then stimulated the lungs with PMA in the presence of different oxygen concentrations. Most interestingly, we detected peak PMA-evoked lung superoxide release when lungs were ventilated with 5 % O2. This peak in superoxide release correlated with the maximum PMA-evoked vasoconstrictor effect. The NADPH oxidases of endothelial cells, which have been shown to contain all NADPH oxidase subunits needed for superoxide generation as well as leukocytes are resident in the intravascular compartment, and are suggested as a possible source of the PMA-induced superoxide release. The ESR technology was not suitable for detecting significant hypoxia-dependent changes in superoxide release in unstimulated isolated rabbit lungs. However, since i) hypoxia caused an increased release of NADPH-dependent superoxide release when lungs were challenged with PMA and ii) that superoxide caused a vasoconstriction; it is tempting to speculate that such mechanisms may contribute to the regulation of HPV. Data from our laboratory have repetitively suggested that an NADPH oxidase-dependent increase in lung ROS release contributes to the initiation of HPV [8,21]. Thus, it is interesting that many studies investigating HPV in isolated lungs the pulmonary circulation was primed with angiotensin II to yield a sufficient hypoxic vasoconstrictor response [51-53]. Angiotensin II has also been shown to activate NADPH oxidases and thus a possible interference of an angiotensin II-induced superoxide release in HPV has to be taken into account. The increase in NADPH oxidase-dependent superoxide release may not only play a role under physiological conditions, but may also contribute to pathophysiological pathways in conditions of ischemia-reperfusion, implicating not only a role of superoxide during reperfusion phase, but also in the ischemic phase.
Three distinct effects have to be taken into consideration regarding the mechanism of PMA-induced increase in intravascular superoxide release. First, a direct effect of hypoxia on the activity of NADPH oxidases. Although such an effect is not known for the NADPH oxidases from phagocytic cells, this may not be excluded for vascular NADPH oxidases that are composed of specific isoforms [54]. Thus, regulation of electron flux through the NADPH oxidase rather than oxygen itself (even at low concentrations) may be the rate-limiting step in superoxide formation. This suggestion is supported by the notion that NADPH oxidase-dependent superoxide release is stimulated by anoxia in cultured aortic endothelial cells [55]. Second, the increased superoxide release may be related to alterations in the activation mechanism upstream of NADPH oxidase (e. g. protein kinase c) or may be related to alterations in protein kinase c-independent mechanisms of NADPH oxidase assembly [55]. Third, alterations to the intravascular or cellular oxygen scavenging capacity may also contribute to an increased superoxide release in hypoxia.
Conclusion
In summary, we have established a new method to detect intravascular ROS release in intact lungs, combining ESR technology with lung perfusion in the presence of the spin probe CPH. This technique may be useful in elucidating the role of ROS in different physiological and pathophysiological conditions where ROS are suggested to be involved and to correlate ROS release to vasoreactivity. Focusing on hypoxia-dependent ROS release, we demonstrated that i) that PMA-induced vasoconstriction is caused by superoxide generated from NADPH oxidases (rather than by H2O2 or by a protein kinase c-dependent increase in Ca2+ sensitivity) and ii) that this pathway is pronounced in hypoxia. The NADPH oxidases thus may contribute the hypoxia-dependent regulation of pulmonary vascular tone.
Abbreviations
AU: arbitrary units
CP•: 3-carboxy-proxyl radical
ΔPAP: change in pulmonary artery pressure
DETC: diethyldithiocarbamate
DFO: deferoxamine
ESR: electron spin resonance
HPV: hypoxic pulmonary vasoconstriction
CPH: l-hydroxy-3-carboxy-2,2,5,5-tetramethylpyrrolidine
PAP: pulmonary artery pressure
ROS: reactive oxygen species
PMA: phorbol-12-myristate-13-acetate
SOD: Cu/Zn-superoxide dismutase
Authors' contributions
Conception and design: N. Weissmann, N. Kuzkaya, H. Schütte, H. A. Ghofrani, R. T. Schermuly, W. Seeger, F. Grimminger.
Analysis and interpretation of the data: N. Weissmann, N. Kuzkaya, R. U. Schäfer, H. Schütte, H. A. Ghofrani, R. T. Schermuly, W. Seeger, F. Grimminger, B. Fuchs, V. Tiyerili, C. Schudt, A. Sydykov, B. Egemnazarow.
Drafting of the article: N. Weissmann, N. Kuzkaya, W. Seeger, F. Grimminger. Critical revision of the article for important intellectual content: N. Weissmann, N. Kuzkaya, H. Schütte, H. A. Ghofrani, R. T. Schermuly, W. Seeger, F. Grimminger, B. Fuchs, V. Tiyerili, A. Sydykov, B. Egemnazarow, C. Schudt, R. U. Schäfer.
Final approval of the article: N. Weissmann, N. Kuzkaya, H. Schütte, H. A. Ghofrani, R. T. Schermuly, W. Seeger, F. Grimminger, B. Fuchs, V. Tiyerili, A. Sydyjov, B. Egemnazarow.
Provision of study materials: N. Weissmann, N. Kuzkaya, H. Schütte, H. A. Ghofrani, R. T. Schermuly, C. Schudt, R. U. Schäfer, W. Seeger, F. Grimminger,.
Statistical expertise: B. Fuchs, V. Tiyerili, A. Sydykov, B. Egemnazarow, N. Weissmann, N. Kuzkaya, R. U. Schäfer.
Performance of experiments: B. Fuchs, V. Tiyerili, A. Sydykov, B. Egemnazarow, N. Weissmann, N. Kuzkaya, R. U. Schäfer.
Obtaining of funding: N. Weissmann, N. Kuzkaya, H. Schütte, W. Seeger, F. Grimminger. Administrative, technical, or logical support: N. Weissmann, N. Kuzkaya, H. Schütte, C. Schudt, W. Seeger, F. Grimminger.
Collection and assembly of data: N. Weissmann, N. Kuzkaya, H. Schütte, H. A. Ghofrani, R.T. Schermuly, W. Seeger, F. Grimminger, B. Fuchs, V. Tiyerili, A. Sydykov, B. Egemnazarow, R. U. Schäfer.
N. Weissmann and N. Kuzkaya contributed equally to this work
Portions of the doctoral thesis of Vedat Tiyerili and Rolf Ulrich Schäfer are incorporated into this report.
Acknowledgements
This work was supported by the Deutsche Forschungsgemeinschaft, SFB 547, projects B6 and B7. The authors thank Dr. Skatchkov for helpful discussions and Rory Morty for linguistic editing of the manuscript.
Rolf Ulrich Schäfer is supported by a predoctoral fellowship from ALTANA Pharma.
==== Refs
Cai H Harrison DG Endothelial dysfunction in cardiovascular diseases: the role of oxidant stress Circ Res 2000 87 840 844 11073878
Fukai T Siegfried MR Ushio-Fukai M Cheng Y Kojda G Harrison DG Regulation of the vascular extracellular superoxide dismutase by nitric oxide and exercise training J Clin Invest 2000 105 1631 1639 10841522
Griendling KK Harrison DG Dual role of reactive oxygen species in vascular growth Circ Res 1999 85 562 563 10488060
Harrison DG Cellular and molecular mechanisms of endothelial cell dysfunction J Clin Invest 1997 100 2153 2157 9410891
Munzel T Harrison DG Evidence for a role of oxygen-derived free radicals and protein kinase C in nitrate tolerance J Mol Med 1997 75 891 900 9428622 10.1007/s001090050181
Sauer H Wartenberg M Hescheler J Reactive oxygen species as intracellular messengers during cell growth and differentiation Cell Physiol Biochem 2001 11 173 186 11509825 10.1159/000047804
Midorikawa J Maehara K Yaoita H Watanabe T Ohtani H Ushiroda S Maruyama Y Continuous observation of superoxide generation in an in-situ ischemia-reperfusion rat lung model Jpn Circ J 2001 65 207 212 11266196 10.1253/jcj.65.207
Weissmann N Grimminger F Olschewski A Seeger W Hypoxic pulmonary vasoconstriction: a multifactorial response? Am J Physiol Lung Cell Mol Physiol 2001 281 L314 317 11435204
Zhang H Slutsky AS Vincent JL Oxygen free radicals in ARDS, septic shock and organ dysfunction Intensive Care Med 2000 26 474 476 10872143 10.1007/s001340051185
Fishman AP Hypoxia on the pulmonary circulation. How and where it acts Circ Res 1976 38 221 231 1260964
Voelkel NF Mechanisms of hypoxic pulmonary vasoconstriction Am Rev Respir Dis 1986 133 1186 1195 3334415
Sylvester JT Hypoxic pulmonary vasoconstriction: a radical view Circ Res 2001 88 1228 1230 11420297
Archer SL Huang J Henry T Peterson D Weir EK A redox-based O2 sensor in rat pulmonary vasculature Circ Res 1993 73 1100 1112 8222081
Archer SL Nelson DP Weir EK Simultaneous measurement of O2 radicals and pulmonary vascular reactivity in rat lung J Appl Physiol 1989 67 1903 1911 2532193
Chandel NS Maltepe E Goldwasser E Mathieu CE Simon MC Schumacker PT Mitochondrial reactive oxygen species trigger hypoxia-induced transcription Proc Natl Acad Sci U S A 1998 95 11715 11720 9751731 10.1073/pnas.95.20.11715
Chandel NS McClintock DS Feliciano CE Wood TM Melendez JA Rodriguez AM Schumacker PT Reactive oxygen species generated at mitochondrial complex III stabilize hypoxia-inducible factor-1 alpha during hypoxia: a mechanism of O2 sensing J Biol Chem 2000 275 25130 25138 10833514 10.1074/jbc.M001914200
Chandel NS Schumacker PT Cellular oxygen sensing by mitochondria: old questions, new insight J Appl Physiol 2000 88 1880 1889 10797153 10.1063/1.1303764
Marshall C Mamary AJ Verhoeven AJ Marhall BE Pulmonary artery NADPH-oxidase is activated in hypoxic pulmonary vasoconstriction Am J Respir Cell Mol Biol 1996 15 633 644 8918370
Waypa GB Chandel NS Schumacker PT Model for hypoxic pulmonary vasoconstriction involving mitochondrial oxygen sensing Circ Res 2001 88 1259 1266 11420302
Waypa GB Schumacker PT O(2) sensing in hypoxic pulmonary vasoconstriction: the mitochondrial door re-opens Respir Physiol Neurobiol 2002 132 81 91 12126697 10.1016/S1569-9048(02)00051-4
Weissmann N Tadic A Hanze J Rose F Winterhalder S Nollen M Schermuly RT Ghofrani HA Seeger W Grimminger F Hypoxic vasoconstriction in intact lungs: a role for NADPH oxidase- derived H(2)O(2)? Am J Physiol Lung Cell Mol Physiol 2000 279 L683 690 11000128
Munzel T Afanas'ev IB Kleschyov AL Harrison DG Detection of superoxide in vascular tissue Arterioscler Thromb Vasc Biol 2002 22 1761 1768 12426202 10.1161/01.ATV.0000034022.11764.EC
Zou L Clanton TL Detection of reactive oxygen and nitrogen species in tissues using redox-sensitive fluorescent probes Methods Enzymol 2002 352 307 325 12125357
Archer SL Nelson DP Weir EK Detection of activated O2 species in vitro and in rat lungs by chemiluminescence J Appl Physiol 1989 67 1912 1921 2532194
Brubacher JL Bols NC Chemically de-acetylated 2',7'-dichlorodihydrofluorescein diacetate as a probe of respiratory burst activity in mononuclear phagocytes J Immunol Methods 2001 251 81 91 11292484 10.1016/S0022-1759(01)00308-8
Sanders SP Bassett DJ Harrison SJ Pearse D Zweier JL Becker PM Measurements of free radicals in isolated, ischemic lungs and lung mitochondria Lung 2000 178 105 118 10773136 10.1007/s004080000013
Katz SA Venkatachalam M Crouch RK Heffner JE Halushka PV Wise WC Cook JA Catalase pretreatment attenuates oleic acid-induced edema in isolated rabbit lung J Appl Physiol 1988 65 1301 1306 3182500
Slepneva IA Glupov VV Sergeeva SV Khramtsov VV EPR detection of reactive oxygen species in hemolymph of Galleria mellonella and Dendrolimus superans sibiricus (Lepidoptera) larvae Biochem Biophys Res Commun 1999 264 212 215 10527867 10.1006/bbrc.1999.1504
Dikalov S Grigor'ev IA Voinov M Bassenge E Detection of superoxide radicals and peroxynitrite by l-hydroxy-4-phosphonooxy-2,2,6,6-tetramethylpiperidine: quantification of extracellular superoxide radicals formation Biochem Biophys Res Commun 1998 248 211 215 9675114 10.1006/bbrc.1998.8936
Dikalov S Skatchkov M Bassenge E Spin trapping of superoxide radicals and peroxynitrite by l-hydroxy-3-carboxy-pyrrolidine and l-hydroxy-2,2,6, 6-tetramethyl-4-oxo-piperidine and the stability of corresponding nitroxyl radicals towards biological reductants Biochem Biophys Res Commun 1997 231 701 704 9070876 10.1006/bbrc.1997.6174
Weissmann N Grimminger F Walmrath D Seeger W Hypoxic vasoconstriction in buffer-perfused rabbit lungs Respir Physiol 1995 100 159 169 7624617 10.1016/0034-5687(94)00133-K
Weissmann N Winterhalder S Nollen M Voswinckel R Quanz K Ghofrani HA Schermuly RT Seeger W Grimminger F NO and reactive oxygen species are involved in biphasic hypoxic vasoconstriction of isolated rabbit lungs Am J Physiol Lung Cell Mol Physiol 2001 280 L638 645 11238003
Weissmann N Akkayagil E Quanz K Schermuly RT Ghofrani HA Fink L Hanze J Rose F Seeger W Grimminger F Basic features of hypoxic pulmonary vasoconstriction in mice Respir Physiol Neurobiol 2004 139 191 202 15123002 10.1016/j.resp.2003.10.003
Dikalov S Skatchkov M Bassenge E Quantification of peroxynitrite, superoxide, and peroxyl radicals by a new spin trap hydroxylamine l-hydroxy-2,2,6,6-tetramethyl-4-oxo-piperidine Biochem Biophys Res Commun 1997 230 54 57 9020059 10.1006/bbrc.1996.5880
Dikalov S Skatchkov M Fink B Bassenge E Quantification of superoxide radicals and peroxynitrite in vascular cells using oxidation of sterically hindered hydroxylamines and electron spin resonance Nitric Oxide 1997 1 423 431 9441913 10.1006/niox.1997.0139
Rosen GM Finkelstein E Rauckman EJ A method for the detection of superoxide in biological systems Arch Biochem Biophys 1982 215 367 378 6284047 10.1016/0003-9861(82)90097-2
Navarro JA Granadillo VA Rodriguez-Iturbe B Garcia R Salgado O Romero RA Removal of trace metals by continuous ambulatory peritoneal dialysis after desferrioxamine B chelation therapy Clin Nephrol 1991 35 213 217 1855329
Misra HP Reaction of copper-zinc superoxide dismutase with diethyldithiocarbamate J Biol Chem 1979 254 11623 11628 227874
Samuni A Goldstein S Russo A Mitchell JB Krishna MC Neta P Kinetics and mechanism of hydroxyl radical and OH-adduct radical reactions with nitroxides and with their hydroxylamines J Am Chem Soc 2002 124 8719 8724 12121116 10.1021/ja017587h
Urbanski NK Beresewicz A Generation of *OH initiated by interaction of Fe2+ and CU+ with dioxygen; comparison with the Fenton chemistry Acta Biochim Pol 2000 47 951 962 11996118
Winterbourn CC Toxicity of iron and hydrogen peroxide: the Fenton reaction Toxicol Lett 1995 82–83 969 974 10.1016/0378-4274(95)03532-X
Ishikawa M Oxygen radicals-superoxide dismutase system and reproduction medicine Nippon Sanka Fujinka Gakkai Zasshi 1993 45 842 848 8371013
Vig E Gabrielak T Leyko W Nemcsok J Matkovics B Purification and characterization of Cu, Zn-superoxide dismutase from common carp liver Comp Biochem Physiol B 1989 94 395 397 2591201 10.1016/0305-0491(89)90362-3
Ross AH McConnell HK Permeation of a spin-label phosphate into the human erythrocyte Biochemistry 1975 14 2793 2798 168918 10.1021/bi00684a001
Swartz HM Sentjurc M Morse PD 2nd Cellular metabolism of water-soluble nitroxides: effect on rate of reduction of cell/nitroxide ratio, oxygen concentrations and permeability of nitroxides Biochim Biophys Acta 1986 888 82 90 3741890 10.1016/0167-4889(86)90073-X
Paky A Michael JR Burke-Wolin TM Wolin MS Gurtner GH Endogenous production of superoxide by rabbit lungs: effects of hypoxia or metabolic inhibitors J Appl Physiol 1993 74 2868 2874 8396109
Savineau JP Marthan R Modulation of the calcium sensitivity of the smooth muscle contractile apparatus: molecular mechanisms, pharmacological and pathophysiological implications Fundam Clin Pharmacol 1997 11 289 299 Review 9263758
Zhang AY Yi F Teggatz EG Zou AP Li PL Enhanced production and action of cyclic ADP-ribose during oxidative stress in small bovine coronary arterial smooth muscle Microvasc Res 2004 67 159 67 15020207 10.1016/j.mvr.2003.11.001
Bailey SR Mitra S Flavahan S Flavahan NA Reactive oxygen species from smooth muscle mitochondria initiate cold-induced constriction of cutaneous arteries Am J Physiol Heart Circ Physiol 2005 289 1 H243 50 15764673 10.1152/ajpheart.01305.2004
Gryglewski RJ Palmer RM Moncada S Superoxide anion is involved in the breakdown of endothelium-derived vascular relaxing factor Nature 1986 320 454 456 3007998 10.1038/320454a0
Archer SL Reeve HL Michelakis E Puttagunta L Waite R Nelson DP Dinauer MC Weir EK O2 sensing is preserved in mice lacking the gp91 phox subunit of NADPH oxidase Proc Natl Acad Sci U S A 1999 96 14 7944 7949 10393927 10.1073/pnas.96.14.7944
Voelkel NF Allard JD Anderson SM Burke TJ cGMP and cAMP cause pulmonary vasoconstriction in the presence of hemolysate J Appl Physiol 1999 86 5 1715 1720 10233139
Reeve HL Michelakis E Nelson DP Weir EK Archer SL Alterations in a redox oxygen sensing mechanism in chronic hypoxia J Appl Physiol 2001 90 6 2249 2256 11356790
Brandes RP Kreuzer J Vascular NADPH oxidases: molecular mechanisms of activation Cardiovasc Res 2005 65 1 16 27 Review 15621030 10.1016/j.cardiores.2004.08.007
Schafer M Schafer C Ewald N Piper HM Noll T Role of redox signaling in the autonomous proliferative response of endothelial cells to hypoxia Circ Res 2003 92 9 1010 1015 12690038 10.1161/01.RES.0000070882.81508.FC
|
16053530
|
PMC1184103
|
CC BY
|
2021-01-04 16:23:27
|
no
|
Respir Res. 2005 Jul 31; 6(1):86
|
utf-8
|
Respir Res
| 2,005 |
10.1186/1465-9921-6-86
|
oa_comm
|
==== Front
World J Surg OncolWorld Journal of Surgical Oncology1477-7819BioMed Central London 1477-7819-3-351596323510.1186/1477-7819-3-35Case ReportSecretory carcinoma of the breast containing the ETV6-NTRK3 fusion gene in a male: case report and review of the literature Arce C [email protected] D [email protected] DG [email protected] MA [email protected]ñnas-Gonzalez A [email protected] A [email protected]érez V [email protected]ón D [email protected] F [email protected] Division of Internal Medicine, Instituto Nacional de Cancerología, Mexico2 Division of Clinical Research, Instituto Nacional de Cancerología, Mexico3 Division of Pathology, Instituto Nacional de Cancerología, Mexico4 Unidad de Investigacion Biomédica en Cancer, Instituto de Investigaciones Biomedicas, Universidad Nacional Autonoma de Mexico e Instituto Nacional de Cancerología, Mexico5 Genetic Pathology Evaluation Center of the Departments of Pathology, British Columbia Cancer Agency Vancouver Canada6 General Hospital and University of British Columbia and the Prostate Centre at the Vancouver General Hospital, Vancouver, British Columbia, Canada2005 17 6 2005 3 35 35 22 3 2005 17 6 2005 Copyright © 2005 Arce et al; licensee BioMed Central Ltd.2005Arce et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Secretory carcinoma (SC) of the breast is a rare and indolent tumor. Although originally described in children, it is now known to occur in adults of both sexes. Recently, the tumor was associated with the ETV6-NTRK3 gene translocation.
Case presentation
A 52-year-old male was diagnosed with secretory breast carcinoma and underwent a modified radical mastectomy. At 18 months the tumor recurred at the chest wall and the patient developed lung metastases. He was treated concurrently with radiation and chemotherapy without response. His tumor showed the ETV6-NTRK3 translocation as demonstrated by fluorescent in situ hybridization (FISH).
Conclusion
SC is a rare slow-growing tumor best treated surgically. There are insufficient data to support the use of adjuvant radiation or chemotherapy. Its association with the ETV6-NTRK3 fusion gene gives some clues for the better understanding of this neoplasm and eventually, the development of specific therapies.
==== Body
Background
SC of the breast is one of the rarest types of breast cancer accounting for less than 1% of all breast cancers. This entity was initially termed "Juvenile breast cancer" by McDivitt and Stewart, based on the fact that the average age of the seven patients described in their series was nine year-old with range of three to fifteen years [1]. Subsequently, more cases in children [2-6] and adults [7-12] were described. Therefore, it was recommended that the descriptive term SC replace the designation "juvenile carcinoma". Secretory breast carcinomas have a characteristic balanced translocation, t(12;15), that creates a ETV6-NTRK3 gene fusion. The finding of a fusion transcript in SC and the demonstration that ETV6-NTRK3 could transform murine mammary epithelial cell lines has challenged widely accepted beliefs on breast carcinogenesis [13]. This specific translocation is associated with congenital fibrosarcoma and mesoblastic nephroma, two morphologically similar pediatric mesenchymal tumors with no epithelial features [14]. The biological consequence of this translocation is the fusion of the dimerization domain of a transcriptional regulator (ETV6) with a membrane receptor tyrosine kinase (NTRK3) that activates the Ras-Mek1 and PI3K-Akt pathways which are important for breast cell proliferation and survival [15,17]. In only a few cases of secretory carcinoma the presence of the translocation has been confirmed. In the seminal report by Tognon 12 out of 13 cases tested positive [13] whereas Makretsov et al., found 3 out of 4 confirmed secretory carcinomas cases positive and at the same time screened 481 invasive breast carcinomas of which only one gave positive signal. This tumor was later confirmed to be a secretory carcinoma [16].
Case report
A 52 year-old male presented to our institution having undergone local excision of a left breast tumor one month previously. The tumor had measured 7 × 5 cm. The mass had been present for 10 years. At physical examination there was evidence of recent surgery and the patient had a 1 cm ipsilateral axillary lymph node. Serum tumor markers and other routine blood test were normal. The liver ultrasonography, chest X-ray and bone scan were negative for metastases.
He underwent a modified radical mastectomy. Residual tumor measuring 2.8 cm × 2.6 cm was present. On macroscopic examination the tumor was firm and circumscribed (Figure 1a). Microscopy showed the classical features of secretory carcinoma with a microcystic pattern (Figure 1b) with abundant intra and extracellular secretory material. No tumor was present at the surgical margins. Colloidal iron staining highlighted the secretory material (Figure 1c). On immunohistochemistry, the tumor cells were positive for S-100 protein (Figure 1d) but negative for estrogen and progesterone receptor and HER2 (Dako, Carpinteria, CA). 2 of 24 resected lymph nodes were positive for metastatic carcinoma.
Figure 1 Tumor was grossly firm and circumscribed (1a). The histological pattern was microcystic (1b) with abundant intra and extracellular secretory material as showed by the colloidal iron stain which was diffuse and strongly positive (1c). The tumor was positive for S-100 protein (1d).
The case was investigated for the t(12;15) ETV6-NTRK3 translocation using two complementary probe sets [16]. A t(12;15) translocation fusion probe assay (Fig. 2a) and a chromosome 15 NTRK3 gene split-apart assay (Fig. 2b) were used to detect the t(12;15) translocation. All BAC clones used in this study were obtained from the BACPAC Resources Centre at the Children's Hospital Oakland Research Institute. All probes were labeled by nick translation with the use of the manufacturer's recommended protocol (Vysis, Downer's Grove, Illinois). BAC clones RP11-434C1 and RP11-407P10 telomeric to ETV6 on 12p were labeled with spectrum orange. On chromosome 15, RP11-114I9 and RP11-730G13, centromeric to NTRK3 on 15q were labeled with spectrum green and clone RP11-247E14, telomeric to NTRK3 was labeled with spectrum orange. Six-micrometer tissue sections were baked overnight at 60C and then subjected to FISH with a modified protocol (Vysis, Downers Grove, IL) [14]. FISH signals were analyzed with a Zeiss Axioplan fluorescent microscope equipped with a COHU-CCD camera. Images were captures with Metasystems ISIS software (MetaSystems Group Inc., Belmont MA) with seven focal planes stacked for the analysis.
Figure 2 FISH images confirming the presence of the t(12;15). In 2a, the presence of the ETV6-NTRK3 fusion is demonstrated by the close proximity of a red signal (ETV6 from chromosome 12) with a green signal (NTRK3 from chromosome 15) in each cell. In 2b, each cell shows separation of red and green probes flanking the NTRK3 gene from chromosome 15.
In view of the nodal metastasis it was decided to treat the patient with six courses of adjuvant 5-fluorouracil, adriamycin and cyclophosphamide (FAC). The patient abandoned treatment after the second course. The patient returned to clinic eighteen months later with two hard nodules in the surgical resection area measuring 8 × 8 cm and 4 cm × 4 cm, (one ulcerated), and three left axillary subcutaneous nodules, two measuring 2 × 2 cm and one 3 × 3 cm (Figure 3a).
Figure 3 Disease recurrence at the chest wall (3a) and lung metastasis (3b). There were two hard nodules in the chest wall and three ipsilateral axillary nodules. In the lung several nodular metastases were present as well as a right sided pleural effusion.
A chest CT scan identified pulmonary metastases with a right pleural effusion (Fig. 3b). The effusion was drained via percutaneous thoracentesis. He then began treatment with concurrent radiation (total dose of 60 Gy) and UFT (Tegafur-Uracil) to the chest followed by systemic UFT as a single agent for 3 months. Post-treatment, there was no change in the pulmonary disease and there was minor response of chest-wall and axillary disease (Figure 4a and 4b).
Figure 4 Disease in the chest wall (4a) and lung (4b) after radiation and concurrent chemotherapy. There was only minor response in the chest wall disease and essentially no change in the lung.
Discussion
Secretory carcinoma is a very rare type of breast carcinoma. Lamovec and Bracko [18] reported 4 cases of SC in a retrospective series of 7038 breast carcinoma cases, and Botta et al [19] found one case of SC among 3000 breast carcinoma cases.
The age at presentation varies from 3 to 87 years with a median age of 25 years [1,5-11]. The male-female ratio is approximately 1:6 [20,21]. The case presented herein is extremely unusual as SC, particularly metastatic SC, has rarely been reported in males. Literature search identified only 16 other cases of SC in males. Our case was older than the average age reported for secretory carcinoma in males which is 17 years. Only the case reported by Kuwabara was older than our case (66 years) [22,23]. Table 1 summarizes the main clinical features of the cases of SC reported in males.
Table 1 Data on 17 males with Secretory Breast Cancer
Author Year Age Duration of symptoms Size (cm) Axillary status Treatment Hormone Receptors ETV6-NTRK3 Follow-up
Simpson31 1969 5 ND ND - (clinical) LE NE NE NED 4y
Tavassolli7 1980 9 ND ND - (clinical) LE NE NE NED 1.75y
Kari5 1985 3 1 mo 1.5 + (1/4) SM+ALNS NE NE ND
Roth32 1988 23 21y 2.0 - (0/21) MRM NE NE NED 4y
Krausz10 1989 24 Many years 4.0 ND SM + RT (axilla) NE NE DOD 20y
Serour21 1992 17 4 y 1.5 - (0/3) WLE + ALND ER- PR+ NE NED 5y
Lamovec18 1994 20 ND 1.2 - (0/?) MRM ER+ PR+ NE NED 1y
Pohar-Marinsek33 1994 20 6–7 y 1.2 - (clinical) SM ER+ PR+ NE NED 6 m
Kuwabara34 1988 66 3 y 3.0 + (2/?) MRM ER- PR+ NE NED 8 m
Vesoulis35 1998 33 10 y 1.5 ND MRM ER+ PR+ NE ND
Kameyama36 1998 50 ND 3.0 - (0/?) MRM ER+ NE ND
Chevallier37 1999 9 14 m 2.0 - (0/?) LE + ALND ER- PR- NE NED 45 m
Yildirim38 1999 11 1 y 1.5 + (1/18) MRT + CT+ RT ER - NE NED 12 m
Bhagwandeen39 1999–2000 9 1 m 1.2 - (0/15) MRM ER- PR- NE NED 20 m
De Bree22 2001 17 2 y 2.0 - (0–14) MRM ER- PR- NE NED 9 m
Grabellus40 2005 46 Male-female transexual ND 4.0 ND LE ER- PR- PRESENT ND
This case 2005 52 10 y 7 + 2/24 MRM + CT ER- PR- PRESENT AWD 25 m
ND: no defined, LE: Local excision, MRM: Modified Radical Mastectomy, CT: chemotherapy, RT Radiotherapy, NE: not examined, NED: not evidence of disease, AWD: Alive with disease, ER: estrogen receptor, PR progesterone receptor. SM simple mastectomy, ALNS: axillary lymph node sampling, ALND: axillary lymph node dissection, WLE: wide local excision, DOD: Died of disease.
SC's can demonstrate several histological patterns including, solid, microcystic, and ductal, with many tumors containing all three patterns [24]. The tumor cells are polygonal with granular eosinophilic cytoplasm. Atypia is minimal or absent and mitotic activity is low [20]. A typical finding is the presence of intracellular and extracellular secretions [7]. This secretory material is periodic acid-Schiff and alcian blue positive [24,25]. In this tumor there was no expression of steroid receptors or HER2. A recent study has reported than only 4 and 2 out of 13 cases expressed the estrogen and progesterone receptor respectively and only two were HER2 positive [26]. In the current case, the tumor had the t(12;15) ETV6-NTRK3 fusion gene (Fig. 4).
The most frequent clinical presentation is of an asymptomatic mobile mass, which is usually subareolar. The tumor size varies from 1 cm to 16 cm with an average diameter of 3 cm. [7,16]. Our patient had a mass of 7 cm × 5 cm. As the patient reported that the lesion had been present for at least 10 years, it had behaved in a slow growing, indolent fashion. This is supported by other reported cases [20]. In this regard, Biallo et al., have reported a MIB1 labeling index of 11.4% (range: <1 to 34%) [26].
Surgery is considered the primary treatment of secretory carcinoma, however, due to scarcity of reported cases no published guidelines for surgical management exist. There are however, many cases reported of patients who had suffered a local recurrence, therefore mastectomy appears to be a sound surgical choice [1,5,7,9-11,22,24,25]. There are no data however on conservative treatment but this option could be explored particularly in cases where breast development has not yet occurred. In regard to the management of the axilla, the overall incidence of axillary lymph node infiltration is around 30% in children and adults regardless of gender [21,24], hence axillary lymph node dissection is advocated by some authors for tumors ≥2 cm [7,24]. Nevertheless, sentinel node biopsy, may be useful for secretory carcinomas. A recent report on a 9-year-old girl treated with simple mastectomy and axillary sentinel lymph node biopsy shows that this is feasible [27].
Postoperative radiotherapy [19,21] and adjuvant chemotherapy [7,28] have been used on at least two occasions. There is at present insufficient evidence to recommend either approach in the management of secretory carcinoma.
Local recurrence after a long disease-free interval has been described in numerous cases; [1-3,5,9,11,28] however these occurred in patients that underwent conservative surgery. This is the second case reported with chest wall recurrence after mastectomy. In the other case the patient was treated with wide local excision and she is alive at 11-month follow-up. In contrast, our case also presented distant recurrence [12].
Distant metastases from secretory carcinoma are extremely rare with only four cases reported [20]. Our case is the fifth case who developed distant metastases, this, despite having only two positive lymph nodes at resection. Another recently reported patient remained disease free at a follow-up of 13 months despite having 12 out of 14 positive nodes and not having received adjuvant chemotherapy [23].
There are several reported cases of patients with secretory breast carcinoma with distant metastases who were treated with either single agent or combination chemotherapy without success. Among the drugs reported are 5-FU, vindesine, mitomycin and prednisone, adriamycin, epirubicin, cyclophosphamide, carboplatin, and even newer active agents such as docetaxel. These data clearly show that this neoplasm is not chemosensitive, as all of the patients treated with chemotherapy showed disease progression while on treatment [7,10,20] and [29]. In our case, despite using UFT alone and with concomitant radiotherapy there was no response. These observations are in contrast with reports on the high chemosensitivity to common agents (vincristine, cyclophosphamide, adriamycin, dactinomycin and ifosfamide) for congenital fibrosarcomas and mesoblastic nephromas, two other neoplasms associated with the translocation ETV6-NTRK3 [30]. This suggests that secretory breast carcinoma, due perhaps to its slow growing behavior, acquires additional genetic alterations than ultimately confer chemoresistance. It will be very useful to establish cancer cell lines from this tumor type to study whether the chemoresistance is a general phenomenon or drug specific.
Conclusion
Secretory carcinoma is a rare slow-growing tumor that is best approached by surgical treatment. There are insufficient data to support the use of adjuvant radiation and/or chemotherapy. Its association to the ETV6-NTRK3 fusion gene gives some clues for the better understanding of this neoplasm and may eventually lead to the development of specific therapies.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
C A-S, and D C-P conceived the study and wrote the manuscript; DG H and MA M performed the FISH analysis and participated in the discussion and writing of the manuscript; V P-S, did the pathological analysis; A A, and F L-M cared for the patient; D G-R critically reviewed the manuscript; AD-G participated in the discussion and writing of the manuscript.
Acknowledgements
The authors want to thank the patient for providing his consent for the publication of this report.
==== Refs
McDivitt RW Stewart FW Breast Carcinoma in Children JAMA 1966 195 144 146 10.1001/jama.195.5.388
Oberman HA Stephens PJ Carcinoma in the breast in childhood Cancer 1972 30 370 374
Byrne MP Fahey MM Gooselaw JG Breast cancer with axillary metastasis in an eight and one half years old girl Cancer 1973 31 726 728 4693600
Masse SR Rioux A Beauchesne C Juvenile carcinoma the breast Hum Pathol 1981 12 1044 1046 7319493
Karl SR Ballantine TVN Zaino R Juvenile secretory carcinoma of the breast J Pediatr Surg 1985 20 368 371 4045662
Fergunson TS Mc Carty KS Filston HC Juvenile secretory carcinoma and juvenile papillomatosis: diagnsis and treatment J Pediatr Surg 1987 22 637 640 3612460
Tavassoli FA Norris HJ Secretory carcinoma of the breast Cancer 1980 45 2404 2413 6445777
Oberman HA Secretory carcinoma of the breast in adult Am J Surg Pathol 1980 4 465 470 7435774
Akhtar M Robinson C Ashraf M Godwin JT Secretory carcinoma of the breast in adults: light and electron microscipic study of three cases with review of literature Cancer 1983 51 2245 2254 6189573
Krausz T Jenkins D Grontoft O Pollock DJ Azzopardi JG Secretory carcinoma of the breast in adult: emphasis on late recurrences and metastasis Histopathology 1989 14 25 36 2925177
Rosen PP Cranor ML Secretory carcinoma of the breast Arch Pathol Lab Med 1991 115 141 144 1992979
Mies C Recurrent secretory in residual mammarye after mastectomy Am J Surg Pathol 1993 17 715 721 8391221
Tognon C Knezevich SR Huntsman D Roskelley CD Melnyk N Mathers JA Becker L Carneiro F MacPherson N Horsman D Poremba C Sorensen PH Expression of the ETV6-NTRK3 gene fusion as a primary event in human secretory breast carcinoma Cancer Cell 2002 2 367 376 12450792 10.1016/S1535-6108(02)00180-0
Argani P Ladanyi M Recent advances in pediatric renal neoplasia Adv Anat Pathol 2003 10 243 60 12973047 10.1097/00125480-200309000-00001
Euhus DM Timmons CF Tomlinson GE ETV6-NTRK3 – Trk-ing the primary event in human secretory breast cancer Cancer Cell 2002 2 347 348 12450787 10.1016/S1535-6108(02)00184-8
Makretsov N He M Hayes M Chia S Horsman DE Sorensen PH Huntsman DG A fluorescence in situ hybridization study of ETV6-NTRK3 fusion gene in secretory breast carcinoma Genes Chromosomes Cancer 2004 40 152 157 15101049 10.1002/gcc.20028
Hughes-Davies L Huntsman D Ruas M Fuks F Bye J Chin SF Milner J Brown LA Hsu F Gilks B Nielsen T Schulzer M Chia S Ragaz J Cahn A Linger L Ozdag H Cattaneo E Jordanova ES Schuuring E Yu DS Venkitaraman A Ponder B Doherty A Aparicio S Bentley D Theillet C Ponting CP Caldas C Kouzarides T EMSY links the BRCA2 pathway to sporadic breast and ovarian cancer Cell 2003 115 523 535 14651845 10.1016/S0092-8674(03)00930-9
Lamovec J Bracko M Secretory carcinoma of the breast: light microscopical, inmunohistocheminal flow cytometric study Mod Pathol 1994 7 475 479 8066076
Botta G Fessia L Ghiringello B Juvenile milk protein secreting carcinoma Virchows Arch A Pathol Anat Histopathol 1982 395 145 152 10.1007/BF00429608
Herz H Goldstein D Metastasis secretory breast cancer. Non-responsiveness to chemotherapy: case report and review of the literature Ann Oncol 2000 11 1343 1347 11106125 10.1023/A:1008387800525
Serour F Gilad A Kopolovic J Krispin M Secretory breast cancer in childhood and adolescence: report of a case and review of the world literature Med Pediatr Oncol 1992 20 341 344 1608359
de Bree E Askoxylakis J Giannikaki E Chroniaris N Sanidas E Tsiftsis DD Secretory carcinoma of the male breast Ann Surg Oncol 2002 9 663 667 12167580 10.1245/aso.2002.9.7.663
Kavalakat AJ Covilakam RK Culas TB Secretory carcinoma of breast in a 17-year-old male World J Surg Oncol 2004 2 17 15175103 10.1186/1477-7819-2-17
Richard G Hawk JC IIIBaker AS JrAustin MR Multicentric adult secretory breast carcinoma: DNA flow cytometric findings, pronostic features and review of the world literature J Surg Oncol 1990 44 238 244 2200927
Nguyen G Neifer R Aspiration biopsy cytology of secretory carcinoma of the breast Diagn Cytopathol 1987 3 234 237 2822366
Diallo R Schaefer KL Bankfalvi A Decker T Ruhnke M Wulfing P Jackisch C Luttges J Sorensen PH Singh M Poremba C Secretory carcinoma of the breast: a distinct variant of invasive ductal carcinoma assessed by comparative genomic hybridization and immunohistochemistry Hum Pathol 2003 34 1299 305 14691916 10.1016/S0046-8177(03)00423-4
Bond SJ Buchino JJ Nagaraj HS McMasters KM Sentinel lymph node biopsy in juvenile secretory carcinoma Pediatr Surg 2004 39 120 121 10.1016/j.jpedsurg.2003.09.042
Sullivan JJ Magee HR Donald KJ Secretory (juvenile) carcinoma of the breast Pathology 1977 9 341 346 593733
Krohn M Trams G Brandt G Secretory breast cancer – a special morphological entity predominantly in children and young women – a case report Geburtshilfe Frauenheilkunde 1989 49 299 301
McCahon E Sorensen PH Davis JH Rogers PC Schultz KR Non-resectable congenital tumors with the ETV6-NTRK3 gene fusion are highly responsive to chemotherapy Med Pediatr Oncol 2003 40 288 92 12652616 10.1002/mpo.10272
Simpson JS Barson AJ Breast tumors in infants and children: a 40-year review of cases at a children's hospital Can Med Assoc J 1969 101 100 2 5794134
Roth JA Discafani C O'Malley M Secretory breast carcinoma in a man Am J Surg Pathol 1988 12 150 4 3341512
Pohar-Marinsek Z Golouh R Secretory breast cancer in a man diagnosed by fine needle aspiration biopsy Acta Cytol 1994 38 446 50 8191840
Kuwabara H Yamane M Okada S Secretory breast carcinoma in a 66 year old man J Clin Pathol 1998 51 545 7 9797736
Vesoulis Z Kashkari S Fine needle aspiration of secretory breast carcinoma resembling lactational changes Acta Cytol 1998 42 1032 6 9684599
Kameyama K Mukai M Iri H Kuramochi S Yamazaki K Ikeda Y Hata J Secretory carcinoma of the breast in a 51-year old male Pathol Int 1998 48 994 7 9952346
Chevallier A Boissy C Rampal A Soller C Turc-Carel C Thyss A Michiels JF Secretory carcinoma of the breast. Report of a case in a 9-year-old boy Clin Exp Pathol 1999 47 88 91 10398580
Yildrim E Turhan N Pak I Berberoglu U Secretory breast carcinoma in a boy Eur J Surg Oncol 1999 25 98 9 10188866
Bhagwandeen BS Fenton L Secretory carcinoma of the breast in a 9-year-old boy Pathology 1999 31 166 8 10399175 10.1080/003130299105386
Grabellus F Worm K Willruth A Schmitz KJ Otterbach F Baba HA Kimmig R Metz KA ETV6-NTRK3 gene fusion in a secretory carcinoma of the breast of male-to-female transsexual Breast 2005 14 71 4 15695086 10.1016/j.breast.2004.04.005
|
15963235
|
PMC1184104
|
CC BY
|
2021-01-04 16:39:04
|
no
|
World J Surg Oncol. 2005 Jun 17; 3:35
|
utf-8
|
World J Surg Oncol
| 2,005 |
10.1186/1477-7819-3-35
|
oa_comm
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.