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
World J Surg OncolWorld Journal of Surgical Oncology1477-7819BioMed Central London 1477-7819-3-231586012910.1186/1477-7819-3-23ReviewManagement of leiomyosarcomas of the spermatic cord: the role of reconstructive surgery Enoch Stuart [email protected] Simon M [email protected] Douglas S [email protected] West Midlands Regional Centre for Plastic and Reconstructive Surgery, Selly Oak Hospital, University Hospital of Birmingham, – B29 6JD, UK2 Wound Healing Research Unit, University Department of Surgery, University of Cardiff/University Hospital of Wales, Cardiff, – CF14 4UJ, UK2005 28 4 2005 3 23 23 23 1 2005 28 4 2005 Copyright © 2005 Enoch et al; licensee BioMed Central Ltd.2005Enoch 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
Leiomyosarcomas (LMS) of the spermatic cord are extremely rare. Radical inguinal orchiectomy and high ligation of the cord is the standard primary surgical procedure. The extent of surrounding soft tissue excision required and the precise role of adjuvant radiotherapy, however, remains unclear. In addition, recurrence is a commonly encountered problem which might necessitate further radical excision of adjacent soft tissues.
Methods
This article reviews the pathophysiology of spermatic cord leiomyosarcomas (LMS), and discusses the various reconstructive surgical options available to repair the inguinal region and the lower anterior abdominal wall after excision of the tumour and the adjacent soft tissues.
Results
There is paucity of literature on LMS of spermatic cord. The majority of paratesticular neoplasms are of mesenchymal origin and up to 30% of these are malignant. In adults, approximately 10% of spermatic cord sarcomas are LMS. Approximately 50% of these tumours recur loco-regionally following definitive surgery; however, the incidence decreases if resection is followed by adjuvant radiotherapy.
Conclusion
It is therefore important to achieve negative histological margins during the primary surgical procedure, even if adjuvant radiotherapy is instituted. If extensive resection is required, either during the primary procedure or following recurrence, reconstructive surgery may become necessary. This article reviews the pathophysiology of spermatic cord LMS, the reasons for recurrence, and discusses the management options including the role of reconstructive surgery.
==== Body
Introduction
Tumours of the spermatic cord and paratesticular tissue are rare [1,2], and as such, their true incidence has never been established. Radical inguinal orchiectomy and high ligation of the cord is the standard primary surgical procedure. However, the extent of surrounding soft tissue excision required, including margins, and the role of adjuvant radiotherapy (RT), remains controversial. The paucity of literature in this area often makes treatment decisions difficult; prospective trials are precluded by the rarity of this tumour and no series is sufficiently large to accurately evaluate the most appropriate treatment option.
Recurrence is a commonly encountered problem, which might necessitate further radical excision of adjacent soft tissues. It hence becomes important to achieve negative histological margins during the primary surgery even if adjuvant radiotherapy (RT) is instituted. If extensive areas of adjacent soft tissues are resected, either during the primary procedure or following recurrence, reconstructive surgery may be indicated to repair the defect.
In addition to debating the above issues, this article reviews the pathophysiology of spermatic cord leiomyosarcomas (LMS), and discusses the various reconstructive surgical options available to repair the inguinal region and the lower anterior abdominal wall after excision of the tumour and the adjacent soft tissues.
Pathophysiology
The majority of paratesticular neoplasms are of mesenchymal origin and up to 30% of these are malignant [3]. In adults, approximately 10% of spermatic cord sarcomas are LMS with a peak incidence in the sixth and seventh decades [4]. LMS most likely originate from the smooth muscle of different areas such as the vas deference canal wall, blood vessels and cremaster muscle, and is thought to arise as a result of malignant degeneration from previously existing leiomyomatous tumours [5].
It has been estimated that approximately 50% of these tumours recur loco-regionally following definitive surgery [6-10], although the incidence decreases if resection is followed by adjuvant RT [8,9]. Pathologic features that convey a higher risk of local recurrence include large tumour size, inguinal location, narrow or positive margins and prior intralesional surgery [11]. Loco-regional relapse may occur in the cord, scrotum, or adjacent pelvis, with or without involvement of the regional lymph nodes [10]. If relapse occurs in the cord or scrotum, they often extend proximally through the internal inguinal ring into the pelvic cavity [12].
The patients may also present with signs and symptoms of distant metastasis, although dissemination may be seen as late as 15 or more years after resection of the primary tumour [13]. The common means of metastasis are by lymphatic spread either to the regional (pelvic) [12] or distant lymph nodes (para-aortic) [14], and by the haematogenous route, usually to the lungs and the liver [9,12]; metastasis to other regions such as the orbit [15], although rare, may also occur.
Management
Preoperative diagnosis of spermatic cord LMS is difficult and the tumour is usually described as a firm, gradually enlarging, and painless extra-testicular intra-scrotal mass [16]. It may be occasionally accompanied by a hydrocoele [1]. Scrotal ultrasound is a useful initial investigation to identify a sinister mass within the scrotum [17]. If a neoplasm is identified, magnetic resonance imaging is the investigation of choice to characterise and delineate the anatomical extent of the tumour [18]. Other modalities such as fluorodeoxyglucose positron emission tomography (FDG-PET) have been used in the diagnosis of dedifferentiated liposarcoma of the spermatic cord [19]; but it is not established in routine clinical practice. The role, if any, of fine needle aspiration cytology in the preoperative diagnosis of spermatic cord LMS has not been clearly established although they have been used to diagnose other types of spermatic cord sarcomas such as liposarcoma [20] and malignant fibrous histiocytoma [21,22].
Surgical excision and margins
Radical inguinal orchiectomy and high ligation of the cord is the standard primary surgical procedure [9,23,24]. The extent of surrounding soft tissue excision and the margins required, however, remains controversial. Simple excision is clearly inadequate for paratesticular sarcomas as repeat wide excision has revealed microscopic residual disease in 27% of completely excised cases [23]. Aggressive surgical strategies are therefore recommended in the management of spermatic cord LMS and a decrease in local recurrence has been observed in patients who underwent re-operative wide excision after a prior incomplete resection [25].
Malignant mesenchymal tumours such as LMS have a "pseudocapsule" [26] and, as such, there may be infiltration of tumour cells into the adjacent tissues; therefore, it might be difficult to precisely estimate the tumour margins. Due to this 'irregular' pattern of tumour growth and the anatomical constraints in the inguinal region, wide circumferential resection margins may be difficult to achieve [1,12]. Nevertheless, due to the tumours' high propensity for local recurrence, if negative histological margins are not achieved during primary surgery, re-excision should be considered mandatory, even if this involves sacrificing some areas of adjacent normal anatomy. Advances in microsurgery have made it possible to reconstruct significant anatomical defects in this region (discussed later).
The role of prophylactic lymph node dissection remains unclear. Although performed in some centres, the true incidence of para-aortic and pelvic nodal metastasis has never been documented [12], and hence there is insufficient evidence at present to suggest that prophylactic para-aortic and pelvic nodal dissection prevents relapse or improves the prognosis of patients with spermatic cord LMS [8]. It, however, has a role in other types of para-testicular sarcomas such as rhabdomyosarcomas, fibrosarcomas, and intermediate or high-grade malignant fibrous histiocytomas [7].
Surgically relevant anatomy
The inguinal region lies at the junction of the lower abdomen and the anterior thigh. The lower abdominal wall below skin and subcutaneous tissue is composed of three layers of muscle (external oblique, internal oblique and transversus abdominis) and their aponeurosis anterolaterally, and vertical rectus muscle covered in an aponeurotic sleeve medially (Figure 1). The rectus sheath differs in composition above and below the arcuate line, which lies one-third of the way from the umbilicus to the symphysis pubis (Figure 2).
Figure 1 Layers of the anterior abdominal wall
Figure 2 Formation of the rectus sheath in horizontal section, above [A] and below [B] the arcuate line
The inguinal canal allows the passage of the spermatic cord or round ligament (in females) without vascular compromise and without obstructing ductus deferens in the male. Below the inguinal ligament lies the femoral triangle. The boundaries of the femoral triangle are the inguinal ligament superiorly, the medial border of sartorius laterally, and the medial border of adductor longus medially. The femoral triangle contains the femoral vessels and nerve, sapheno-femoral junction and the deep inguinal lymph nodes.
Reconstruction of the inguinal region and the anterior lower abdominal wall after wide surgical resection
After wide excision of soft tissues, it is important to repair the defect in the inguinal region and the lower anterior abdominal wall (Figure 3). The structures that may require covering include the spermatic cord, femoral neurovascular bundle, abdominal wall muscles and sometimes-exposed bone. If the surgical resection involves removal of abdominal wall muscles, then reconstruction using mesh in addition to tissue coverage may be necessary. The principle of soft tissue reconstruction is based on the reconstructive ladder (Figure 4). Various reconstructive surgical options are available in the plastic surgeon's armoury to repair defects in this region (Table 1).
Figure 3 Defect created after removal of the tumour along with lower part of the abdominal wall including the inguinal canal, part of rectus sheath and external oblique aponeurosis, subcutaneous tissue and skin
Figure 4 Reconstructive ladder in wound closure
Table 1 reconstructive surgical options to cover defects in the inguinal region and the anterior lower abdominal wall
Coverage type Name of flap
Skin and fat flap • Deep inferior epigastric perforator (DIEP) flap
Fascio-cutaneous flaps • Antero – lateral thigh flap
• Groin flap
Muscle/musculo-cutaneous flaps • Rectus abdominis flap (TRAM/VRAM)
• Tensor fasciae latae flap
• Gracilis flap
• Rectus femoris flap
• Vastus lateralis flap
Free flap • Latissimus dorsi flap
Others • Sartorius muscle switch
• Omental flap
Local flaps to cover the inguinal region and the lower anterior abdominal wall include the muscle or musculo-cutaneous flaps such as rectus abdominis (transverse rectus abdominis muscle (TRAM) or vertical rectus abdominis muscle (VRAM) flaps, gracillis flap, tensor fasciae latae flap (Figure 5), rectus femoris flap and vastus lateralis flap. The tensor fasciae latae (TFL) flap is a frequently employed flap because it is very resilient and has a reliable vascular pedicle from the lateral femoral circumflex branch of the profunda femoris artery [27] (Figure 5). In addition, the division of TFL from its normal anatomical insertion does not have a significant effect on the function of the limb; for this reason, the rectus femoris and vastus lateralis flaps are not the first choice since their loss may lead to some limitation of movements at the hip or knee joints, and thus functional loss.
Figure 5 Some reconstructive options (tensor fasciae latae flap, gracilis flap, rectus abdominus flap, groin flap) available to repair defects in the inguinal region and the lower anterior abdominal wall
If the resection is not deep (no muscle loss), fascio-cutaneous flaps such as the groin flap and the antero-lateral thigh flap, or an abdominal wall skin and fat flap based on perforating vessels – deep inferior epigastric perforator (DIEP) flap – could be used. The use of a groin flap, however, may be limited in situations where the vessels, which originate in the femoral triangle, may be ligated as part of the resection. If femoral dissection is required leading to soft tissue loss, a sartorius muscle switch or an omental flap [28] could be used to cover exposed vessels in the femoral triangle.
In the event of no local flap options being available (very rare), a free flap may be used to achieve cover. This involves raising a distant flap, usually the latissimus dorsi muscle, disconnecting the blood supply and transferring the tissue to the groin and anastomosing the thoraco-dorsal vessels to recipient vessels in the inguinal region.
The use of well vascularised tissue to cover exposed femoral vessels is important as it minimises local vascular complications when adjuvant RT is instituted to the groin region.
The various reconstructive options discussed above are useful in different clinical (or operative) scenarios. With adequate planning and proper choice, satisfactory short- and long-term results can be obtained with all the above reconstructive procedures. The choice of one particular flap for reconstruction depends on various factors: (i) the anatomical location of the cover required; (ii) the extent and type of soft tissue loss (e.g., skin, muscle, etc.); (iii) tumour type – primary or recurrent; (iv) the ease of donor site coverage; (v) flap availability; and (vi) the experience of the surgeon in the usage of a particular flap. In primary tumours, the skin might not be involved and hence a muscle flap might be adequate. However, in recurrence, the overlying skin might be damaged due to RT. This may necessitate excision of the skin (along with underlying tissues) and coverage using a musculo-cutaneous or a fascio-cutanoeus flap (alternatively a muscle flap with split skin grafting can be used).
Radiotherapy and chemotherapy
Adjuvant RT has become an established method in the management of LMS of the spermatic cord [8,9,12,13]. Some authors recommend adjuvant RT only for high-grade LMS or to those patients believed to be at a high risk of local recurrence. However, due to the tumour's high propensity (approximately 50%) [8,9] of local recurrence following surgery alone [6,12], there is increasing consensus that LMS of all grades and histology should receive adjuvant RT [12]. Several studies have reported better results with combined modality treatment and the recurrence could be reduced to 10 – 20% by using adjuvant RT [29]. It needs to be emphasised at this point that adjuvant RT should supplement rather than substitute radical surgical excision and should be instituted only after complete clearance of the tumour is achieved surgically. The field for RT should include the inguinal canal, ipsilateral pelvic tissue [12], and the scrotum [8]. The role of pre-operative RT has not been clearly established, and, as such, there is no evidence to suggest that radiotherapy before surgical excision reduces the rate of recurrence or improves the overall prognosis.
There are no controlled studies at present that specifically addresses the role of adjuvant systemic chemotherapy in adult spermatic cord LMS [11], although they have a well-defined role in childhood rhabdomyosarcomas [30]. Adjuvant chemotherapy is currently not used routinely in the management of spermatic cord LMS though it has been suggested that it might have a role in abrogating the haematogenous metastatic potential in high grade sarcomas [12] and in patients with metastatic disease.
Prognosis
The prognosis of patients with LMS is highly variable. Kyle (1966) [31] suggested a probable five-year survival of 10–15% but more recently a five-year survival of 50–80% has been reported [9], possibly reflecting the advances in diagnosis and management of these tumours. The wide range in the five-year survival rate might be due to the variations in tumour stage and grade at the time of diagnosis as well as the diversity of therapies involved.
Conclusion
In the management of spermatic cord LMS, complete excision of the tumour with radical inguinal orchiectomy and high ligation of the cord should be the primary surgical procedure. Due to the high propensity for local recurrence, if the margins are not clear, re-excision should be considered in all cases. In addition, based on current evidence, the authors recommend that adjuvant RT should be instituted for all grades and types of spermatic cord LMS. There is insufficient evidence to advocate prophylactic nodal treatment in the management of these tumours.
Competing interests
The author(s) declare that they have no competing interests
Authors' contributions
SE – Conception and design, literature review with references, drafting the article and revising, preparation of images
SMW – Drafting the article and revising, preparation of images
DSM – Specialist plastic surgical input, supervision, and review, proof-reading and revising the manuscript
All authors read and approved the final manuscript for publication
==== Refs
Lioe TF Biggart JD Tumours of the spermatic cord and paratesticular tissue. A clinicopathologic study Br J Urol 1993 71 600 606 8518870
Donovan MG Fitzpatrick JM Gaffney EF Paratesticular leiomyosarcoma Br J Urol 1987 60 590 3427346
Bajaj P Agarwal K Niveditha SR Pathania OP Leiomyosarcoma arising from tunica vaginalis testis: a case report Indian J Pathol Microbiol 2001 44 145 146 11883131
Rao CR Srinivasulu M Naresh KN Doval DC Hazarika D Adult Paratesticular Sarcomas: A report of eight cases J Surg Oncol 1994 56 89 93 8007685
Padilla AJ Gonzalez CM Caldero MM Arano YA Torroella BF Fanlo AP Leiomyosarcoma of the spermatic cord Actas Urol Esp 1993 17 464 467 8368123
Blitzer PH Dozoretz DE Proppe KH Treatment of malignant tumours of the spermatic cord: A study of 10 cases and review of the literature J Urol 1981 126 611 614 7299919
Sclama AO Berger BW Cherry JM Young JD Jr Malignant fibrous histiocytoma of the spermatic cord: the role of retroperitoneal lymphadenectomy in management J Urol 1983 130 577 579 6310165
Catton CN Cummings BJ Fornasier V O'Sullivan B Quirt I Warr D Adult paratesticular sarcomas: a review of 21 cases J Urol 1991 146 342 345 1906946
Fagundes MA Zietman AL Althausen AF Coen JJ Shipley WU The management of spermatic cord sarcoma Cancer 1996 77 1873 1876 8646687 10.1002/(SICI)1097-0142(19960501)77:9<1873::AID-CNCR17>3.0.CO;2-X
Merimsky O Terrier P Bonvalot S Le Pechoux C Delord JP Le Cesne A Spermatic cord sarcomas in adults Acta Oncologica 1999 38 635 638 10427954 10.1080/028418699431249
Folpe AL Weiss SW Paratesticular soft tissue neoplasms Semin Diagn Pathol 2000 17 307 318 11202547
Ballo MT Zagars GK Pisters PW Le Pechoux C Delord JP Le Cesne A Spermatic cord sarcoma: outcome, patterns of failure, and management J Urol 2001 166 1306 1310 11547063 10.1097/00005392-200110000-00019
Schiller AL Teitelbaum SL Rubin E, Farber JL Smooth muscle tumours Pathology 1999 third Lippincort-Raven publishers, Philadelphia 1411
Banowsky LH Shultz GN Sarcoma of the spermatic cord and tunics: review of the literature, case report and discussion of the role of retroperitoneal lymph node dissection J Urol 1970 103 628 631 5443846
Bakri SJ Krohel GB Peters GB Farber MG Spermatic cord leiomyosarcoma metastatic to the orbit Am J Ophthalmol 2003 136 213 215 12834705 10.1016/S0002-9394(02)02279-1
Stein A Kaplun A Sova Y Zivan I Laver B Lurie M Lurie A Leiomyosarcoma of the spermatic cord: report of two cases and review of the literature World J Urol 1996 14 59 61 8646243
Secil M Kefi A Gulbahar F Aslan G Tuna B Yorukoglu K Sonographic features of spermatic cord leiomyosarcoma J Ultrasound Med 2004 23 973 976 15292568
Woodward PJ Schwab CM Sesterhenn IA From the archives of the AFIP: Extratesticular scrotal masses: radiologic-pathologic correlation Radiographics 2003 23 215 240 12533657
Kosuda S Wahl RL Grossman HB Demonstration of recurrent dedifferentiated liposarcoma of the spermatic cord by FDG-PET Ann Nucl Med 1997 11 263 266 9310177
Dalla Palma P Barbazza R Well-differentiated liposarcoma of the paratesticular area: report of a case with fine-needle aspiration preoperative diagnosis and review of the literature Diagn Cytopathol 1990 6 421 426 2292227
Bosch-Princep R Martinez-Gonzalez S Alvaro-Naranjo T Castellano-Megias VM Salvado-Usach MT Marti-Mestre J Fine needle aspiration and touch imprint cytology of a malignant fibrous histiocytoma of the spermatic cord: case report Acta Cytol 2000 44 423 428 10834004
Berardo MD Powers CN Wakely PE JrAlmeida MO Frable WJ Fine-needle aspiration cytopathology of malignant fibrous histiocytoma Cancer 1997 81 228 237 9292738 10.1002/(SICI)1097-0142(19970825)81:4<228::AID-CNCR5>3.0.CO;2-L
Catton C Jewett M O'Sullivan B Kandel R Paratesticular sarcoma: failure patterns after definitive local therapy J Urol 1999 161 1844 1847 10332450 10.1097/00005392-199906000-00033
Watanabe J Soma T Kawa G Hida S Koisi M Leiomyosarcoma of the spermatic cord Int J Urol 1999 6 536 538 10533906 10.1046/j.1442-2042.1999.00100.x
Coleman J Brennan MF Alektiar K Russo P Adult spermatic cord sarcomas: management and results Ann Surg Oncol 2003 10 669 675 12839852 10.1245/ASO.2003.11.014
Bowden L Booher RJ The principles and technique of resection of soft parts for sarcoma Surgery 1958 44 963 977 13624959
Roth DA Aston SJ, Beasley RW Thoracic and abdominal wall reconstruction Grabb and Smith's Plastic Surgery 1997 Fifth Thorne CHM. Lippincott-Raven publishers, Philadelphia 1029
Watkins RM Thomas JM The role of greater omentum in reconstructing skin and soft tissue defects of the groin and axilla Br J Surg 1985 72 925 926 3904914
Tran LM Mark R Meier R Calcattera TC Parker RG Sarcoma of the head and neck Cancer 1992 70 169 177 1606539
de Vries JD Paratesticular rhabdomyosarcoma World J Urol 1995 13 219 225 8528295
Kyle VN Leiomyosarcoma of the spermatic cord: a review of the literature and report of an additional case J Urol 1966 96 795 5923307
| 15860129 | PMC1097762 | CC BY | 2021-01-04 16:39:03 | no | World J Surg Oncol. 2005 Apr 28; 3:23 | utf-8 | World J Surg Oncol | 2,005 | 10.1186/1477-7819-3-23 | oa_comm |
==== Front
World J Surg OncolWorld Journal of Surgical Oncology1477-7819BioMed Central London 1477-7819-3-241586212310.1186/1477-7819-3-24Case ReportInflammatory myofibroblastic tumour of the gallbladder Behranwala Kasim A [email protected] Peter [email protected] Andrew [email protected] Cyril [email protected] Jeremy N [email protected] Gastrointestinal Surgery Unit, Royal Marsden NHS Trust, 203 Fulham Road, London SW3 6JJ, UK2 Department of Pathology, Royal Marsden NHS Trust, 203 Fulham Road, London SW3 6JJ, UK2005 29 4 2005 3 24 24 17 11 2004 29 4 2005 Copyright © 2005 Behranwala et al; licensee BioMed Central Ltd.2005Behranwala 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
Inflammatory myofibroblastic tumour (IMT) is a benign, nonmetastasizing proliferation of myofibroblasts with a potential for local infiltration, recurrence and persistent local growth.
Case report
We report a case of a 51 year-old female, who had excision of a gallbladder tumour. Histopathology showed it to be IMT of the gallbladder.
Conclusion
The approach to these tumours should be primarily surgical resection to obtain a definitive diagnosis and relieve symptoms. IMT has a potential for local infiltration, recurrence and persistent local growth.
inflammatory myofibroblastic tumourgallbladder tumour
==== Body
Introduction
Inflammatory myofibroblastic tumour (IMT) is a rare benign lesion that has been discussed in various organs and tissues. They are well recognised in lung and upper respiratory tract of children and young adults with a predilection for first and second decade. Intra-abdominal forms of the disease are reported to occur most frequently in the liver, followed by stomach, bowel, spleen, mesentery [1] and extrahepatic bile duct [2]. The clinical presentation will vary on the site involved. We report a case of IMT of the gallbladder, which has not been previously described. A case of inflammatory pseudotumour of the gallbladder and bile ducts with synchronous lesion in the lung, has been described, which subsided on high-dose prednisolone therapy [3]. Another case of chronic cholecystitis with features of xanthogranulomatous inflammation due to the presence of a prominent inflammatory infiltrate composed of plasma cells, lymphocytes, macrophages, foamy histiocytes and huge fibroblastic and myofibroblastic proliferation was described by Corsi A et al [4].
Case Report
A 51-year-old female presented with history of acute right upper abdominal pain, localised abdominal signs and raised inflammatory markers. Ultrasound scanning suggested acute cholecystitis. The patient was explored initially by laparoscopy but converted to an open operation at the referring hospital. An irresectable mass, which was thought to be an advanced gallbladder carcinoma, was found. Several needle biopsies were taken from the tumour but the gallbladder was not excised. Histology of the biopsies showed features of IMT of the gallbladder. She developed obstructive jaundice postoperatively, ERCP showed a distal bile duct stricture which was stented. CT scan showed a mass lesion in the gallbladder area with direct involvement of the liver and no metastatic disease elsewhere (Figure 1). In view of the histological diagnosis the patient underwent re-laparotomy through an extended subcostal incision. Operative findings included a tense, distended gallbladder containing stones, debris and pus with a segment of transverse colon densely adherent to the mass. An en-bloc cholecystectomy and limited transverse colectomy with primary anastomosis was performed. The mass was peeled off the first and second part of duodenum, common bile duct and transverse mesocolon. Intra-operative cholangiogram via the cystic duct stump showed no biliary leakage and the dye flowed freely into the duodenum with the biliary stent in situ.
Figure 1 CT scan showing the gallbladder inflammatory mass
On pathological examination, the tumour measured 12 cms. Microscopy of the resected mass showed the gallbladder wall to be replaced by proliferative spindle myofibroblastic cells arranged in fascicles, admixed with diffuse chronic inflammatory cells including lymphocytes, plasma cells and eosinophils with lymphoid aggregates. In places hyalinised fibrous stroma was seen. No pleomorphism or vascular invasion was evident. Mitoses were inconspicuous (Figure 2). The mass was seen to grow in an infiltrating pattern with entrapment of adipocytes and extending to the muscularis propria of colon from externally. The tumour extended to the resection margin. Four lymph nodes showed reactive changes. Immunhistochemistry showed positivity for SMA (Figure 3) and calponin. ALK-1 was equivocal. However desmin, caldesmon and CAM 5.2 were negative (Figure 4).
Figure 2 Photomicrograph showing features of inflammatory myofibroblastic tumour (Haematoxylin and Eosin × 50)
Figure 3 Photomicrograph showing immunohistocemical positive staining with smooth muscle actin (×100)
Figure 4 Photomicrograph showing immunohistocemical negative staining with desmin (× 100)
The postoperative period was uneventful. ERCP was done at four weeks after operation with replacement of the stent because of slight stricturing of the common hepatic duct. No local recurrence was detected at six months follow-up on CT scan.
Discussion
Inflammatory myofibroblastic tumour is a benign, non-metastasizing proliferation of myofibroblasts with potential for recurrence and persistent local growth, similar in some respect to the fibromatoses [5]. IMT is associated with constitutional symptoms and it has been variously termed plasma cell granuloma, inflammatory pseudotumour, inflammatory myofibrohistiocytic proliferation to reflect divergent views concerning its pathogenesis and level of malignancy. The attributes of a myofibroblast places it midway between a fibroblast and a smooth muscle cell and it appears capable of functional and phenotypic modulation.
Myofibroblastic tumours fall into four main groups: the family of reactive fascitis like lesions, a group of benign lesions (e.g. Mammary myofibroblastoma, intranodal myofibroblastoma, angiomyofibroblastoma and dermatomyofibroma), the locally aggressive fibromatoses (either superficial or deep) which share features of fibroblasts and myofibroblasts in varying degree and finally sarcomas showing myofibroblastic differentiation (low grade lesions – infantile myofibroblastic sarcoma, inflammatory myofibroblastic tumours, low grade myofibroblastic sarcoma, inflammatory fibrosarcoma [6] and high grade lesions – malignant fibrous histiocytoma) [7]. An aberrant or exaggerated response to tissue injury without an established cause has generally been favoured as the pathogenesis of the inflammatory pseudotumour or IMT. An immunological pathogenesis remains a possibility. Tumours with myofibroblast as the principal cell type are designated as IMT. The IMT and inflammatory fibrosarcoma appear to have many overlapping clinical and pathological features. These tumours are histogenetically related and if they are separate entities, they are differentiated more by degrees than absolutes [8]. IMT of the gastrointestinal tract is extremely rare and differ clinically, histologically and immunohistochemically from inflammatory fibroid polyps though both have a prominent inflammatory infiltrate admixed with spindle-shaped fibroblasts/ myofibroblasts set in a collagenous, fibrovascular or myxoid stroma [9].
The inflammatory myofibroblastic tumours are well circumscribed but rarely encapsulated. They usually have a homogenous consistency although areas of haemorrhage, necrosis or calcification may be found. Multicentric lesions are rare. IMT is composed of fascicles of bland myofibroblastic cells admixed with a prominent inflammatory infiltrate consisting of lymphocytes, plasma cells, macrophages and eosinophils. The lack of atypia, hyperchromasia and abnormal mitotic figures are pointers towards a benign lesion. The spindle cells stain positively for smooth muscle actin and vimentin but are negative for S100, desmin, CD100, cytokeratin, CD35 and latent membrane protein. Differential diagnosis includes calcifying fibrous pseudotumour [10,11] inflammatory fibrosarcoma, follicular dendritic cell tumour and gastrointestinal autonomic nerve tumours. Calcifying fibrous pseudotumour is a benign fibrous lesion characterised by three distinctive features: a collection of dense, hyalinized collagen interspersed with benign-appearing spindle cells, psammomatous or dystrophic calcification, and a lymphoplasmocytic inflammatory infiltrate of variable intensity [11]. Inflammatory fibrosarcoma is histogenetically related but is regarded as malignant on the basis of the high rate of local recurrence, multiple peritoneal implants, locally aggressive behaviour, distant metastases and tumour related deaths. IMT and inflammatory fibrosarcoma may be two ends of a part of a neoplastic continuum of myofibroblastic proliferation with increasing cellular atypia and aggressiveness [8]. ALK immunostaining is detected in 36% to 60% of cases in a fibrillary or granular distribution in cytoplasm or nucleus, with occasional cell and nuclear membranous accentuation [12]. The presence of chromosomal aberrations indicates that IMT is a neoplastic proliferation [13]. IMT are clonal and a proportion (30 – 40%) has reproducible cytogenetic abnormalities involving the region of the anaplastic lymphoma kinase (ALK1) gene on chromosome 2 [10]. Inflammatory myofibroblastic tumours show a wide spectrum of cellular atypia and biological behaviour with p53 and MDM2 expression. However the alterations in the p53 pathway seem not to play a major role in the pathogenesis of inflammatory myofibroblastic tumour [14].
There is no regular vascular pattern thus there is variable contrast enhancement on CT scans. On magnetic resonance imaging, hepatic IMT appears as a mass or as an area of periportal soft tissue infiltration. The mass is hypointense on T1-W and slightly hyperintense relative to surrounding liver parenchyma on T2-W image [15]. The prognosis of this tumour is generally considered to be favourable with no reports of malignancy [16]. Twenty-two cases have been reported in the pancreatic region [17]. Coffin CM et al [5] have reported 84 cases with a median age of 9 years (3 months to 46 years) occurring at various sites (abdomen, retroperitoneum or pelvis (61 cases); head and neck including upper respiratory tract (12 cases); trunk (8 cases); and extremities (3 cases)) and ranging in size from 1 to 17 cms. Excision was performed in 69 cases and clinical follow-up in 53 cases revealed 13 patients (25%) had one or more recurrences at intervals of 1–24 months. Histological confirmation is always required before diagnosis and treatment, to differentiate it from malignant tumour. Misdiagnosis has led some patients to be inappropriately treated with chemotherapy [12]. The therapeutic approach to these tumours should rely primarily on surgical resection in order to obtain a definitive diagnosis as well as to relieve symptoms. IMT has a potential for local infiltration, recurrence and persistent local growth. Local recurrence may occur many years later and thus strict follow-up after surgery is required.
==== Refs
Sawant S Kasturi L Amin A Inflammatory myofibroblastic tumour Indian J Paediatr 2002 66 1001 1002
Buyukyavuz I Karnak I Haliloglu M Senocak ME Inflammatory myofibroblastic tumour of the extrahepatic bile ducts: an unusual cause of obstructive jaundice in children Eur J Paediatr Surg 2003 13 421 424 10.1055/s-2003-44736
Ikeda H Oka T Imafuku I Yamada S Yamada H Fujiwara K Hirata M Idezuki Y Oka H A case of inflammatory pseudotumour of the gallbladder and bile duct Am J Gastroenterol 1990 85 203 206 2405646
Corsi A Bosman C Chronic cholecystitis with features of diffuse inflammatory pseudotumour: a clinico-pathological case study and review of the literature Ital J Gastroenterol 1995 27 252 255 8541577
Coffin CM Watterson J Priest JR Dehner LP Extrapulmonary inflammatory myofibroblastic tumour (inflammatory pseudotumour). A clinicopathologic and immunohistochemical study of 84 cases Am J Surg Pathol 1995 19 859 872 7611533
Meis-Kindblom JM Kjellstrom C Kindblom LJ Inflammatory fibrosarcoma: update, reappraisal and perspective on its place in the spectrum of inflammatory myofibroblastic tumours Semin Diagn Pathol 1998 15 133 143 9606804
Fletcher CD Myofibroblastic tumours: an update Verh Dtsch Ges Pathol 1998 82 75 82 10095420
Coffin CM Dehner LP Meis-Kindblom JM Inflammatory myofibroblastic tumour, inflammatory fibrosarcoma and related lesions: an historical review with differential diagnostic considerations Semin Diagn Pathol 1998 15 102 110 9606802
Makhlouf HR Sobin LH Inflammatory myofibroblastic tumours (inflammatory pseudotumours) of the gastrointestinal tract: how closely are they related to inflammatory fibroid polyps? Hum Pathol 2002 23 307 315 11979371 10.1053/hupa.2002.32213
Nascimento AF Ruiz R Hornick JL Fletcher CD Calcifying fibrous 'pseudotumour': clinicopathologic study of 15 cases and analysis of its relationship to inflammatory myofibroblastic tumour Int J Surg Pathol 2002 10 189 196 12232572
Mourra N Bell S Parc R Flejou JF Calcifying fibrous pseudotumour: first case report in the gallbladder Histopathology 2004 44 84 86 14717677 10.1111/j.1365-2559.2004.01743.x
Coffin CM Patel A Perkins S Elenitoba-Johnson KS Perlman E Griffin CA ALK and p80 expression and chromosomal rearrangements involving 2p23 in inflammatory myofibroblastic tumour Mod Pathol 2001 14 569 576 11406658 10.1038/modpathol.3880352
Su LD Atayde-Perez A Sheldon S Fletcher JA Weiss SA Inflammatory myofibroblastic tumours: cytogenetic evidence supporting clonal origin Mod Pathol 1998 11 364 368 9578087
Yamamota H Oda Y Saito T Sakamoto A Miyajima K Tamiya S Tsuneyoshi M p53 mutation and MDM2 amplification in inflammatory myofibroblastic tumours Histopathology 2003 42 431 439 12713619 10.1046/j.1365-2559.2003.01611.x
Venkataraman S Semelka RC Braga L Danet IM Woolsey JT Inflammatory myofibroblastic tumour of the hepatobiliary system: report of MR imaging appearance in four patients Radiology 2003 227 758 763 12728186
Stringer MD Ramani P Yeung CK Capps SNJ Kiely EM Spitz L Abdominal inflammatory myofibroblastic tumours in children Br J Surg 1992 79 1357 1380 1486440
Yamamoto H Watanabe K Nagata M Tasaki K Inflammatory myofibroblastic tumour of the pancreas J Hepatobiliary Pancreat Surg 2002 9 116 119 12021906 10.1007/s005340200013
| 15862123 | PMC1097763 | CC BY | 2021-01-04 16:39:03 | no | World J Surg Oncol. 2005 Apr 29; 3:24 | utf-8 | World J Surg Oncol | 2,005 | 10.1186/1477-7819-3-24 | oa_comm |
==== Front
World J Surg OncolWorld Journal of Surgical Oncology1477-7819BioMed Central London 1477-7819-3-241586212310.1186/1477-7819-3-24Case ReportInflammatory myofibroblastic tumour of the gallbladder Behranwala Kasim A [email protected] Peter [email protected] Andrew [email protected] Cyril [email protected] Jeremy N [email protected] Gastrointestinal Surgery Unit, Royal Marsden NHS Trust, 203 Fulham Road, London SW3 6JJ, UK2 Department of Pathology, Royal Marsden NHS Trust, 203 Fulham Road, London SW3 6JJ, UK2005 29 4 2005 3 24 24 17 11 2004 29 4 2005 Copyright © 2005 Behranwala et al; licensee BioMed Central Ltd.2005Behranwala 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
Inflammatory myofibroblastic tumour (IMT) is a benign, nonmetastasizing proliferation of myofibroblasts with a potential for local infiltration, recurrence and persistent local growth.
Case report
We report a case of a 51 year-old female, who had excision of a gallbladder tumour. Histopathology showed it to be IMT of the gallbladder.
Conclusion
The approach to these tumours should be primarily surgical resection to obtain a definitive diagnosis and relieve symptoms. IMT has a potential for local infiltration, recurrence and persistent local growth.
inflammatory myofibroblastic tumourgallbladder tumour
==== Body
Introduction
Inflammatory myofibroblastic tumour (IMT) is a rare benign lesion that has been discussed in various organs and tissues. They are well recognised in lung and upper respiratory tract of children and young adults with a predilection for first and second decade. Intra-abdominal forms of the disease are reported to occur most frequently in the liver, followed by stomach, bowel, spleen, mesentery [1] and extrahepatic bile duct [2]. The clinical presentation will vary on the site involved. We report a case of IMT of the gallbladder, which has not been previously described. A case of inflammatory pseudotumour of the gallbladder and bile ducts with synchronous lesion in the lung, has been described, which subsided on high-dose prednisolone therapy [3]. Another case of chronic cholecystitis with features of xanthogranulomatous inflammation due to the presence of a prominent inflammatory infiltrate composed of plasma cells, lymphocytes, macrophages, foamy histiocytes and huge fibroblastic and myofibroblastic proliferation was described by Corsi A et al [4].
Case Report
A 51-year-old female presented with history of acute right upper abdominal pain, localised abdominal signs and raised inflammatory markers. Ultrasound scanning suggested acute cholecystitis. The patient was explored initially by laparoscopy but converted to an open operation at the referring hospital. An irresectable mass, which was thought to be an advanced gallbladder carcinoma, was found. Several needle biopsies were taken from the tumour but the gallbladder was not excised. Histology of the biopsies showed features of IMT of the gallbladder. She developed obstructive jaundice postoperatively, ERCP showed a distal bile duct stricture which was stented. CT scan showed a mass lesion in the gallbladder area with direct involvement of the liver and no metastatic disease elsewhere (Figure 1). In view of the histological diagnosis the patient underwent re-laparotomy through an extended subcostal incision. Operative findings included a tense, distended gallbladder containing stones, debris and pus with a segment of transverse colon densely adherent to the mass. An en-bloc cholecystectomy and limited transverse colectomy with primary anastomosis was performed. The mass was peeled off the first and second part of duodenum, common bile duct and transverse mesocolon. Intra-operative cholangiogram via the cystic duct stump showed no biliary leakage and the dye flowed freely into the duodenum with the biliary stent in situ.
Figure 1 CT scan showing the gallbladder inflammatory mass
On pathological examination, the tumour measured 12 cms. Microscopy of the resected mass showed the gallbladder wall to be replaced by proliferative spindle myofibroblastic cells arranged in fascicles, admixed with diffuse chronic inflammatory cells including lymphocytes, plasma cells and eosinophils with lymphoid aggregates. In places hyalinised fibrous stroma was seen. No pleomorphism or vascular invasion was evident. Mitoses were inconspicuous (Figure 2). The mass was seen to grow in an infiltrating pattern with entrapment of adipocytes and extending to the muscularis propria of colon from externally. The tumour extended to the resection margin. Four lymph nodes showed reactive changes. Immunhistochemistry showed positivity for SMA (Figure 3) and calponin. ALK-1 was equivocal. However desmin, caldesmon and CAM 5.2 were negative (Figure 4).
Figure 2 Photomicrograph showing features of inflammatory myofibroblastic tumour (Haematoxylin and Eosin × 50)
Figure 3 Photomicrograph showing immunohistocemical positive staining with smooth muscle actin (×100)
Figure 4 Photomicrograph showing immunohistocemical negative staining with desmin (× 100)
The postoperative period was uneventful. ERCP was done at four weeks after operation with replacement of the stent because of slight stricturing of the common hepatic duct. No local recurrence was detected at six months follow-up on CT scan.
Discussion
Inflammatory myofibroblastic tumour is a benign, non-metastasizing proliferation of myofibroblasts with potential for recurrence and persistent local growth, similar in some respect to the fibromatoses [5]. IMT is associated with constitutional symptoms and it has been variously termed plasma cell granuloma, inflammatory pseudotumour, inflammatory myofibrohistiocytic proliferation to reflect divergent views concerning its pathogenesis and level of malignancy. The attributes of a myofibroblast places it midway between a fibroblast and a smooth muscle cell and it appears capable of functional and phenotypic modulation.
Myofibroblastic tumours fall into four main groups: the family of reactive fascitis like lesions, a group of benign lesions (e.g. Mammary myofibroblastoma, intranodal myofibroblastoma, angiomyofibroblastoma and dermatomyofibroma), the locally aggressive fibromatoses (either superficial or deep) which share features of fibroblasts and myofibroblasts in varying degree and finally sarcomas showing myofibroblastic differentiation (low grade lesions – infantile myofibroblastic sarcoma, inflammatory myofibroblastic tumours, low grade myofibroblastic sarcoma, inflammatory fibrosarcoma [6] and high grade lesions – malignant fibrous histiocytoma) [7]. An aberrant or exaggerated response to tissue injury without an established cause has generally been favoured as the pathogenesis of the inflammatory pseudotumour or IMT. An immunological pathogenesis remains a possibility. Tumours with myofibroblast as the principal cell type are designated as IMT. The IMT and inflammatory fibrosarcoma appear to have many overlapping clinical and pathological features. These tumours are histogenetically related and if they are separate entities, they are differentiated more by degrees than absolutes [8]. IMT of the gastrointestinal tract is extremely rare and differ clinically, histologically and immunohistochemically from inflammatory fibroid polyps though both have a prominent inflammatory infiltrate admixed with spindle-shaped fibroblasts/ myofibroblasts set in a collagenous, fibrovascular or myxoid stroma [9].
The inflammatory myofibroblastic tumours are well circumscribed but rarely encapsulated. They usually have a homogenous consistency although areas of haemorrhage, necrosis or calcification may be found. Multicentric lesions are rare. IMT is composed of fascicles of bland myofibroblastic cells admixed with a prominent inflammatory infiltrate consisting of lymphocytes, plasma cells, macrophages and eosinophils. The lack of atypia, hyperchromasia and abnormal mitotic figures are pointers towards a benign lesion. The spindle cells stain positively for smooth muscle actin and vimentin but are negative for S100, desmin, CD100, cytokeratin, CD35 and latent membrane protein. Differential diagnosis includes calcifying fibrous pseudotumour [10,11] inflammatory fibrosarcoma, follicular dendritic cell tumour and gastrointestinal autonomic nerve tumours. Calcifying fibrous pseudotumour is a benign fibrous lesion characterised by three distinctive features: a collection of dense, hyalinized collagen interspersed with benign-appearing spindle cells, psammomatous or dystrophic calcification, and a lymphoplasmocytic inflammatory infiltrate of variable intensity [11]. Inflammatory fibrosarcoma is histogenetically related but is regarded as malignant on the basis of the high rate of local recurrence, multiple peritoneal implants, locally aggressive behaviour, distant metastases and tumour related deaths. IMT and inflammatory fibrosarcoma may be two ends of a part of a neoplastic continuum of myofibroblastic proliferation with increasing cellular atypia and aggressiveness [8]. ALK immunostaining is detected in 36% to 60% of cases in a fibrillary or granular distribution in cytoplasm or nucleus, with occasional cell and nuclear membranous accentuation [12]. The presence of chromosomal aberrations indicates that IMT is a neoplastic proliferation [13]. IMT are clonal and a proportion (30 – 40%) has reproducible cytogenetic abnormalities involving the region of the anaplastic lymphoma kinase (ALK1) gene on chromosome 2 [10]. Inflammatory myofibroblastic tumours show a wide spectrum of cellular atypia and biological behaviour with p53 and MDM2 expression. However the alterations in the p53 pathway seem not to play a major role in the pathogenesis of inflammatory myofibroblastic tumour [14].
There is no regular vascular pattern thus there is variable contrast enhancement on CT scans. On magnetic resonance imaging, hepatic IMT appears as a mass or as an area of periportal soft tissue infiltration. The mass is hypointense on T1-W and slightly hyperintense relative to surrounding liver parenchyma on T2-W image [15]. The prognosis of this tumour is generally considered to be favourable with no reports of malignancy [16]. Twenty-two cases have been reported in the pancreatic region [17]. Coffin CM et al [5] have reported 84 cases with a median age of 9 years (3 months to 46 years) occurring at various sites (abdomen, retroperitoneum or pelvis (61 cases); head and neck including upper respiratory tract (12 cases); trunk (8 cases); and extremities (3 cases)) and ranging in size from 1 to 17 cms. Excision was performed in 69 cases and clinical follow-up in 53 cases revealed 13 patients (25%) had one or more recurrences at intervals of 1–24 months. Histological confirmation is always required before diagnosis and treatment, to differentiate it from malignant tumour. Misdiagnosis has led some patients to be inappropriately treated with chemotherapy [12]. The therapeutic approach to these tumours should rely primarily on surgical resection in order to obtain a definitive diagnosis as well as to relieve symptoms. IMT has a potential for local infiltration, recurrence and persistent local growth. Local recurrence may occur many years later and thus strict follow-up after surgery is required.
==== Refs
Sawant S Kasturi L Amin A Inflammatory myofibroblastic tumour Indian J Paediatr 2002 66 1001 1002
Buyukyavuz I Karnak I Haliloglu M Senocak ME Inflammatory myofibroblastic tumour of the extrahepatic bile ducts: an unusual cause of obstructive jaundice in children Eur J Paediatr Surg 2003 13 421 424 10.1055/s-2003-44736
Ikeda H Oka T Imafuku I Yamada S Yamada H Fujiwara K Hirata M Idezuki Y Oka H A case of inflammatory pseudotumour of the gallbladder and bile duct Am J Gastroenterol 1990 85 203 206 2405646
Corsi A Bosman C Chronic cholecystitis with features of diffuse inflammatory pseudotumour: a clinico-pathological case study and review of the literature Ital J Gastroenterol 1995 27 252 255 8541577
Coffin CM Watterson J Priest JR Dehner LP Extrapulmonary inflammatory myofibroblastic tumour (inflammatory pseudotumour). A clinicopathologic and immunohistochemical study of 84 cases Am J Surg Pathol 1995 19 859 872 7611533
Meis-Kindblom JM Kjellstrom C Kindblom LJ Inflammatory fibrosarcoma: update, reappraisal and perspective on its place in the spectrum of inflammatory myofibroblastic tumours Semin Diagn Pathol 1998 15 133 143 9606804
Fletcher CD Myofibroblastic tumours: an update Verh Dtsch Ges Pathol 1998 82 75 82 10095420
Coffin CM Dehner LP Meis-Kindblom JM Inflammatory myofibroblastic tumour, inflammatory fibrosarcoma and related lesions: an historical review with differential diagnostic considerations Semin Diagn Pathol 1998 15 102 110 9606802
Makhlouf HR Sobin LH Inflammatory myofibroblastic tumours (inflammatory pseudotumours) of the gastrointestinal tract: how closely are they related to inflammatory fibroid polyps? Hum Pathol 2002 23 307 315 11979371 10.1053/hupa.2002.32213
Nascimento AF Ruiz R Hornick JL Fletcher CD Calcifying fibrous 'pseudotumour': clinicopathologic study of 15 cases and analysis of its relationship to inflammatory myofibroblastic tumour Int J Surg Pathol 2002 10 189 196 12232572
Mourra N Bell S Parc R Flejou JF Calcifying fibrous pseudotumour: first case report in the gallbladder Histopathology 2004 44 84 86 14717677 10.1111/j.1365-2559.2004.01743.x
Coffin CM Patel A Perkins S Elenitoba-Johnson KS Perlman E Griffin CA ALK and p80 expression and chromosomal rearrangements involving 2p23 in inflammatory myofibroblastic tumour Mod Pathol 2001 14 569 576 11406658 10.1038/modpathol.3880352
Su LD Atayde-Perez A Sheldon S Fletcher JA Weiss SA Inflammatory myofibroblastic tumours: cytogenetic evidence supporting clonal origin Mod Pathol 1998 11 364 368 9578087
Yamamota H Oda Y Saito T Sakamoto A Miyajima K Tamiya S Tsuneyoshi M p53 mutation and MDM2 amplification in inflammatory myofibroblastic tumours Histopathology 2003 42 431 439 12713619 10.1046/j.1365-2559.2003.01611.x
Venkataraman S Semelka RC Braga L Danet IM Woolsey JT Inflammatory myofibroblastic tumour of the hepatobiliary system: report of MR imaging appearance in four patients Radiology 2003 227 758 763 12728186
Stringer MD Ramani P Yeung CK Capps SNJ Kiely EM Spitz L Abdominal inflammatory myofibroblastic tumours in children Br J Surg 1992 79 1357 1380 1486440
Yamamoto H Watanabe K Nagata M Tasaki K Inflammatory myofibroblastic tumour of the pancreas J Hepatobiliary Pancreat Surg 2002 9 116 119 12021906 10.1007/s005340200013
| 15869711 | PMC1097764 | CC BY | 2021-01-04 20:47:05 | no | World J Surg Oncol. 2005 May 3; 3:25 | latin-1 | World J Surg Oncol | 2,005 | 10.1186/1477-7819-3-25 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1588497710.1371/journal.pbio.0030171Research ArticleAnimal BehaviorNeuroscienceInsectsFunction of a Fly Motion-Sensitive Neuron Matches Eye Movements during Free Flight Motion-Sensitive Cell Matches Eye MovementsKern Roland [email protected]
1
van Hateren J. H
2
Michaelis Christian
1
Lindemann Jens Peter
1
Egelhaaf Martin
1
1Department of Neurobiology, Faculty for BiologyBielefeld University, BielefeldGermany2Department of NeurobiophysicsUniversity of GroningenThe NetherlandsSrinivasan Mandyam V. Academic EditorAustralian National UniversityAustralia6 2005 17 5 2005 17 5 2005 3 6 e17110 1 2005 14 3 2005 Copyright: © 2005 Kern 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.
Fly Movie Theater Reveals Secrets of How Insects See the World
Sensing is often implicitly assumed to be the passive acquisition of information. However, part of the sensory information is generated actively when animals move. For instance, humans shift their gaze actively in a sequence of saccades towards interesting locations in a scene. Likewise, many insects shift their gaze by saccadic turns of body and head, keeping their gaze fixed between saccades. Here we employ a novel panoramic virtual reality stimulator and show that motion computation in a blowfly visual interneuron is tuned to make efficient use of the characteristic dynamics of retinal image flow. The neuron is able to extract information about the spatial layout of the environment by utilizing intervals of stable vision resulting from the saccadic viewing strategy. The extraction is possible because the retinal image flow evoked by translation, containing information about object distances, is confined to low frequencies. This flow component can be derived from the total optic flow between saccades because the residual intersaccadic head rotations are small and encoded at higher frequencies. Information about the spatial layout of the environment can thus be extracted by the neuron in a computationally parsimonious way. These results on neuronal function based on naturalistic, behaviourally generated optic flow are in stark contrast to conclusions based on conventional visual stimuli that the neuron primarily represents a detector for yaw rotations of the animal.
With the aid of virtual reality, the authors study the neuronal responses to visual motion associated with natural behavior.
==== Body
Introduction
In moving animals, retinal image flow differs from conventional visual stimuli used in the laboratory, by its characteristic dynamics that are largely determined by the animals' own actions and reactions. For instance, the retinal image flow on the eyes of humans steering a car, is determined by the direction and speed of the car, but also by the body, head, and eye movements of the driver. In particular, the eyes are rotated actively in a sequence of saccades towards interesting locations in the scene (review [1]). Many insects, such as blowflies, employ a similar saccadic viewing strategy (review [2]). They shift their gaze during free flight by saccadic turns of body and head, keeping gaze basically fixed between saccades [3–5]. This active viewing strategy generates retinal image flow with characteristic dynamical features and separates to a large extent the image flow resulting from rotational and translational movements of the animal. Because the rotational optic flow component does not depend on the distance between the eyes and environmental objects, whereas the translational flow component does, the saccadic flight strategy may help the nervous system to extract information about the spatial layout of the environment. So far, it is not clear to what extent neuronal processing matches the specific properties of the retinal image flow during natural behaviour and thus may be appropriate for gathering environmental information. It is possible that the mechanisms of neuronal information processing are specifically adapted to efficiently utilize image flow under behaviourally relevant conditions. This hypothesis is tested here by analysing the performance of an identified motion-sensitive neuron in the blowfly under stimulus conditions that approximate natural situations.
To circumvent the problems of recording neuronal responses in freely moving animals we took advantage of recent technological developments: In the blowfly, a model system for visual motion computation [6,7], body and even head movements were recorded during free flight [5,8]. The resulting behaviourally generated retinal image flow was reconstructed (Video S1) and replayed to blowflies with a panoramic visual stimulator that is sufficiently fast to show visual stimuli as experienced even during rapid saccadic turns [9]. During replay the activity of the so-called horizontal system equatorial cell (HSE) [10,11] was recorded intracellularly.
HSE is a major output neuron of the visual system and belongs to an identified set of motion-sensitive neurons present in both the left and right third visual neuropil of the blowfly brain. These neurons are believed to extract parameters of self-motion from the optic flow field [10–12]. HSE responds in a directionally selective manner to visual wide-field motion; because of its input from many local motion-sensitive elements, it is depolarised by front-to-back motion in the ipsilateral visual field and hyperpolarised during motion in the reverse direction [6]. The graded depolarisations, although superimposed by spikes of variable amplitude, are still quite pronounced in the axon terminal of the cell [10,11,13] (Figure 1A). Such depolarisations were shown in other motion-sensitive neurons with the same type of mixed response to be transmitted to postsynaptic cells [14]. Because HSE also receives excitatory input from the contralateral eye during back-to-front motion via the identified H1 and H2 neurons, previous studies using conventional experimenter-defined stimuli concluded that its main functional role is to encode rotations around the vertical axis of the head (yaw). In addition, HSE also responds to binocular front-to-back motion [10,12,15], as occurs during forward flight.
Figure 1 Response Characteristics of HSE under Various Stimulus Conditions
(A) Individual response of the right HSE to rightward (preferred-direction [PD]) and leftward (null-direction [ND]) rotation of a vertical sinusoidal grating (wavelength 20 deg, velocity 40 deg/s, contrast 0.98, azimuth ±120 deg, elevation ±50 deg).
(B) Downward view of a flight trajectory (dotted line), with head position and orientation shown every 50 ms (time colour coded: start, red; finish, green).
(C) Head yaw velocity (black) and total optic flow (green) during the flight shown in (B). Positive velocities denote leftward turns, PD for the right HSE.
(D) Membrane potential of a right HSE in response to the flight shown in (B).
(E) Average membrane potential of the right HSE as determined across trials; the entire responses including spikes were taken into account (N = 6 cells, n = 17 responses).
(F) Saccade-triggered averages, calculated from the mean responses (N = 6, one to four repetitions) of right and left HSE to two trajectories for PD (black) and ND (red) saccades (63 different saccades). Vertical line denotes time of saccade peak velocity.
(G) Same as (E), but with rotations only (position of fly fixed in centre of cage; N = 6, n = 11). Broken lines denote resting potential; red or blue vertical lines indicate peak times of PD or ND saccades, respectively. Signals in (D), (E), and (G) are shifted backwards by 22.5 ms to account for response latencies; this value was determined by cross-correlation of the yaw velocity and the corresponding response traces. Signals in (E) and (G) are low-pass filtered (with a Gaussian standard deviation of 3 ms).
Responses to behaviourally generated optic flow, however, cast doubt on the presumed role of HSE. We provide evidence that HSE, rather than primarily representing a detector for yaw rotations of the animal, also encodes information about sideward translational optic flow. This feature could be significant from a functional point of view because only optic flow induced by translatory motion, but not by rotatory motion, contains information about the spatial layout of the environment. The animal's saccadic flight style is concluded to produce a match of the dynamical properties of retinal image flow with the visual motion pathway of the blowfly.
Results
The velocity profile of yaw rotations reflects the blowfly's saccadic flight style: The fly executes a series of saccadic turns in which the head shows peaks in yaw velocity approaching several thousand degrees per second (Figure 1B and 1C). Between saccades the gaze is kept basically stable and the resulting optic flow is much smaller than during saccades (Figure 1C). Figure 1D shows a recording from an HSE cell of a blowfly watching a movie, played back on the panoramic stimulator, consisting of the image flow experienced by another blowfly during the flight of Figure 1B. For most of the time during this stimulus, the HSE cell was depolarised relative to its resting potential. This depolarisation was interrupted by brief hyperpolarisations during saccades evoking null-direction motion (“null-direction saccades”) (Figure 1D and 1E, blue lines; Figure 1F, red trace). In contrast to expectations from responses to conventional stimuli, HSE did not consistently depolarise during saccades leading to preferred-direction motion (“preferred-direction saccades”), but often showed a slight response dip (negative-going response) relative to the ongoing depolarisation level (Figure 1D and 1E, red lines; Figure 1F, black trace). Similar response profiles were obtained for all ten behaviourally generated motion sequences that were tested. Hence, we may conclude that the overall depolarisation of HSE is not evoked by preferred-direction saccades, but by optic flow between saccades.
This finding was surprising because the strongest optic flow in the cell's preferred direction was generated during preferred-direction saccades rather than between saccades. The possibility that the angular velocities during saccades were too large to be perceived by the motion vision system can be excluded. When the neuron was stimulated exclusively with the original rotations without any superimposed translation, thereby mimicking a situation in which the environmental structures were at an infinite distance (Figure 1G), pronounced depolarising response peaks during preferred-direction saccades occurred. From the considerable difference between the responses of Figure 1E and 1G, it follows that the translational optic flow component has a major impact on the neuronal response profile, even though the translational component is much smaller than the rotational component evoked by saccades. Because blowflies keep their gaze stable between saccades, apart from small-amplitude, broad-band yaw rotations (Figures 1C, 2A, and 2B), they may gather useful information about the spatial layout of the outside world from the translational optic flow components that dominate at low frequencies in intersaccadic intervals.
Figure 2 Properties of the Intersaccadic Stimulus for the Trajectories Used for Figure 3A
(A) Probability density function of yaw velocity (red), sideward (black), and forward (blue) velocity. Forward is parallel to the frontal axis of the fly's head; sideward is perpendicular to the head's plane of symmetry. Sideward and forward velocities were converted to angular velocities by multiplying by the nearness (equal to the inverse of the distance [16]) averaged over the trajectories and over the receptive fields of the neurons (7.02 m−1).
(B) Power spectra of yaw velocity (red), sideward (black), and forward (blue) angular velocity. Typical relative standard error of the mean (SEM) = 20%.
To test this hypothesis, we analysed the intersaccadic-response segments by masking the saccadic segments of stimulus and response. Masks were obtained by gating a region surrounding each saccade that was large enough to include all parts of both saccadic stimulus and corresponding response (for details see Materials and Methods). To establish whether single HSE responses provide the animal with information on its self-motion parameters that could be accessed by simple filtering operations, we determined the optimal linear filters by estimating these parameters from the responses. The similarity between estimated and original self-motion parameters was quantified by the coherence that varies between zero (i.e., both signals are unrelated) and one (i.e., perfect reconstruction). Whereas the coherence of the intersaccadic yaw velocity and the neuronal response was significant only between approximately 20 Hz and 60 Hz, there was considerable coherence between sideward velocity and the neuronal responses at low frequencies (Figure 3A, results from ten flights and five HSE neurons). Surprisingly, the coherence was much smaller for the forward velocity although HSE responds well to constant-velocity front-to-back motion [10,11,15] (see Figure 1A). The coherence with the other self-motion parameters (upward velocity, pitch, and roll) was negligible. We conclude from these results that HSE might make use of the difference in frequency content of rotations and translations (see Figure 2B) to provide information on both optic flow components in adjacent frequency bands (Figure 3A). This is possible because the intersaccadic yaw velocities are smaller by an order of magnitude than during saccades (compare Figures 1C and 2A). Hence, the saccadic gaze strategy may be viewed as a specialisation that enables the extraction of translatory optic flow amidst rotatory optic flow that would otherwise dominate the response if smooth yaw rotations were used exclusively for steering.
Figure 3 Coherence of Stimulus Parameters with HSE Response, Intersaccadic Parts of Stimulus and Response Only
(A) Average coherence of the response of the right and left HSE (N = 5 cells) with yaw velocity (red), sideward (black), and forward (blue) velocity.
(B) Construction of control stimuli. Times i denote the start of an intersaccadic period, m is its midpoint, and s is the start of a saccade. For OT, each orientation coordinate (yaw shown) between consecutive midpoints is compressed into the saccadic period, leaving orientation constant between saccades. For OR, each position coordinate (x shown) is similarly compressed, leaving position constant between saccades.
(C) Average coherence of yaw velocity for the OR (right and left HSE, N = 2) control (red), and of sideward (black) and forward (blue) velocity for the OT (N = 3) control. Typical relative SEM = 10% for all coherences shown at (A) and (C).
The intersaccadic responses to rotation and translation during natural flight might interfere with each other in a complicated way. To check whether such an effect influences our conclusions, we designed two control stimuli (Figure 3B and Video S2). These control stimuli allowed us to study the response to rotation and translation separately whilst keeping the visual scene viewed by the fly virtually the same. In the only rotation (OR) control there was no translation between saccades, so to obtain basically the same trajectory of the eye, the intersaccadic translation of the original trajectories was added to the translation during saccades. In contrast, in the only translation (OT) control, the eye was assumed to translate without any superimposed rotation between saccades; the rotation between saccades was added to saccadic rotation. Coherence of yaw velocity with the response to OR stimuli and coherence of the translational velocities with the response to OT stimuli show similar frequency dependencies as the corresponding components of the original optic flow (Figure 3A and 3C), confirming our conclusion of a frequency separation of the rotational and translational components. Again, the coherence was much smaller for the forward velocity than for the sideward velocity. Hence, the low-frequency components of the neuronal responses between saccades encode mostly sideward motion, whereas information about yaw velocity dominates the high-frequency response components.
By combining the responses of the HSE cells from both halves of the brain, the specificity of the intersaccadic responses to the translational optic flow components can be enhanced. The summation of the responses almost exclusively signifies forward velocity (Figure 4A), whereas the difference between the responses almost exclusively signifies the sideward and yaw velocities (Figure 4B). The latter signals can be separated by low-pass and band-pass filtering, respectively. It is not known whether the blowfly actually uses such a processing scheme, but it is clear that the information on yaw, forward, and sideward velocity can be extracted by simple operations that can also be interpreted in neuronal terms.
Figure 4 Coherence of Stimulus Parameters with Combined HSE Responses, Intersaccadic Parts of Stimulus and Response Only
(A) Average coherence of yaw (red), sideward (black), and forward (blue) velocity with the summed responses of right and left HSE (N = 5).
(B) Same as (A) for the subtracted responses of right and left HSE. Typical relative SEM = 10% for all coherences shown at (A) and (B).
Because translational optic flow depends on the distance of the animal to objects in its environment [16–18], the dependence of the neuronal responses on translation velocity is likely to reflect the spatial relation of the animal to its surroundings. This prediction is supported by experiments in which the optic flow of a given flight trajectory was tested not only for the original flight arena. but also for virtual flight arenas of increasingly larger size (Figure 5). When enlarging the virtual arena, the overall response profile changed dramatically and became virtually indistinguishable from the response to the original rotations without any superimposed translation (compare bottom trace in Figure 5A and Figure 1F) when the size of the flight arena increased to more than approximately 2 m. Accordingly, the coherence between the difference of the responses of the right and the left HSE and sideward velocity dropped to chance levels with increasing distance of the fly to the arena walls (Figure 5B–5E). Hence, intersaccadic responses of HSE implicitly reflect distance information. For the translatory velocities of flies observed in the present experiments in the original 40-cm cage, all distances larger than about 1 m were effectively at infinity.
Figure 5 Influence of Wall Nearness
(A) Yaw velocity and average responses (n = 4, low-pass filtered with a Gaussian standard deviation of 3 ms) to behaviourally generated optic flow of a given flight trajectory in the original flight arena but also in virtual flight arenas with increasingly larger size.
(B–E) Coherence of yaw velocity (red) and sideward angular velocity (black) for different cage sizes: 40-cm edge length (original cage; B), 55 cm (C), 105 cm (D), and 235 cm (E). For (C–E) the flight was centred in the cage. Insets in (B–E): the original and the virtual arenas as seen from above.
Discussion
Here it is shown that, between saccades, the neuronal signals of an identified motion-sensitive visual interneuron of the blowfly provide information about translatory self-motion and thus, implicitly, about the spatial relation of the animal to its surroundings. This result was obtained by a novel experimental paradigm that made it possible, for the first time, to present in electrophysiological experiments what an animal has seen during free-flight manoeuvres. Although the behavioural free-flight sequences were obtained in a relatively small flight arena, there are preliminary results that blowflies under natural outdoor conditions [19] employ the same saccadic flight strategy as observed under laboratory conditions. However, whereas the position and orientation of the eyes of free-flying blowflies could be resolved with unprecedented resolution in the laboratory setting by using a magnetic coil technique [4,5], reconstruction of gaze with a similar precision is not easily possible from high-speed video data collected outdoors. Hence, the behaviourally generated optic flow sequences that were used for stimulation in the present study represent the currently most precise approximation of the visual input of freely flying blowflies.
Our conclusions obtained with behaviourally generated optic flow do not match previous conclusions based on conventional stimuli exclusively defined by the experimenter.
In contrast to the common view that the analysed HSE neuron mainly acts as a detector for the animal's self-rotation [12,20], our results show that, depending on the three-dimensional layout of the environment, its response may not be dominated by the most prominent turns of the animal that occur during saccades.
Although the cell experiences the largest optic flow during saccades, it may encode behaviourally relevant information especially between saccades. Because blowflies keep their gaze stable between saccades apart from small, broad-band yaw rotations, they may gather useful information about the outside world from the translational optic flow components that dominate at low frequencies in intersaccadic intervals. Indeed, between saccades, neuronal signals provide rich information about the spatial relation of the animal to its surroundings. It should be noted that distance is signalled only relative to the fly's own velocity, because retinal velocities evoked during translation are inversely proportional to distance and proportional to translation velocity. This implies that in walking flies, the visual surroundings should affect the responses of the HSE cell only when the fly is very close to environmental structures, just as has been found previously [21,22]. This implicit scaling of distance information by the actual speed of the animal may be a parsimonious and advantageous way to extract from optic flow behaviourally relevant information about the outlay of the environment, because, for instance, evasive actions evoked by obstacles in the path of locomotion need to be evoked only at a smaller distance when the animal moves slower.
Based on experimenter-designed motion stimuli, motion-sensitive neurons are conventionally expected to encode stimulus velocity. Indeed, stimulus velocity can be reconstructed faithfully from the responses of blowfly motion-sensitive neurons as long as the velocities and velocity changes are relatively small [23,24]. However, during saccades the visual motion system operates far beyond the linear range. At higher velocities and, in particular, for very transient motion stimuli, the responses of motion-sensitive neurons are no longer determined by pattern velocity alone, but acceleration and higher time-derivatives of velocity presumably also play a role in shaping the response profile [25–28]. Hence, to assess the functional significance of neuronal mechanisms it is important to analyse the system under its natural operating conditions. The limited linear operating range of motion vision is frequently regarded as a disadvantage because motion-sensitive neurons are implicitly expected to encode velocity in a linear way. In contrast to this view, our results suggest that the nonlinearities of the visual motion system may be essential for HSE to encode information about the spatial relation of the animal to its environment. This interpretation is corroborated by model simulations of the blowfly's visual motion pathway and of HSE responses to behaviourally generated optic flow (J. P. Lindemann et al., unpublished data). If the neuron encoded linearly the entire velocity range that the system encounters in behaviour, by far the largest responses would be generated during body saccades. This would leave only a small response range for encoding information about optic flow between saccades. This information would be strongly degraded by noise in the neuronal signals. Hence, because during saccades the motion vision system does operate outside its linear range, it appears to be able, between saccades, to encode useful information about translation and thus about the spatial relation of the animal to the outside world.
How can the time-dependent responses to complex dynamic stimuli a blowfly encounters in free flight be explained? All features of the HSE responses that are characterised here by electrophysiological techniques can be explained by a model of the computational mechanisms implemented by the neuronal circuits in the blowfly motion vision pathway (J. P. Lindemann et al., unpublished data). Although this model was originally proposed based on simple experimenter-designed stimuli, we showed that it also exploits the active saccadic gaze and flight strategy of blowflies in a similar way to its neuronal counterpart. By stepwise dissection of the model circuit, we could determine which of its components are essential for these remarkable features. Most relevant is the nonlinear velocity encoding of the mechanism of local motion detection, modelled by correlation-type movement detectors, as well as the nonlinear spatial integration properties of the HSE cell itself. The model study suggests that the complex time course of the responses to behaviourally generated optic flow is not significantly shaped by adaptive processes. Such processes were previously characterised in various neurons, not only with simple constant velocity stimuli [29–35], but also with experimenter-designed time-varying velocity fluctuations [28,36,37]. The only adaptive change required in the model was a slight decrease of the gain of the system during prolonged stimulation with behaviourally generated optic flow. This finding of our modelling study is in accordance with recent experimental evidence [33,38].
The close similarity between the present electrophysiological results and the results of our model simulations (J. P. Lindemann et al., unpublished data) indicates that down to the level of the lobula plate the responses are primarily stimulus-driven (“bottom up”). It appears then that feedback mediated by other sensory modalities, such as from the haltere system [39–42] or efference copies, related, for example, to intended neck muscle activity (“top down”), are not important at this level of optic flow processing. Notwithstanding, it is conceivable that haltere signals or efference copies are advantageous when it comes to reading out the lobula plate signals, in particular when splitting these up into saccadic and intersaccadic intervals.
Although our experiments demonstrate that the fly's nervous system might be capable of extracting both rotational and translational information from the combined output of the HSE cells, this should not be taken as evidence that it does so. It is not known, so far, what features are extracted by downstream circuitry from the signals of HSE cells and of other motion-sensitive output neurons of the visual system. Subtraction or addition of the outputs of HSE cells from both halves of the visual system as done in our analysis are merely simple processing schemes that make information on yaw, forward, and sideward velocity available. Nonetheless, these simple formal operations can be approximated by cellular computations through combining excitatory and/or inhibitory synapses. Moreover, because synapses are often found to act as frequency filters (e.g., [43–45]), it appears to be feasible, by neuronal mechanisms, to separate the information on sideward translation and yaw rotation that is inherent in different frequency bands in the HSE difference signal. It will be of interest in future studies to determine whether the proposed encoding scheme, which the current experiments show is possible, is actually employed by the fly's brain during flight. Moreover, it needs to be worked out how the saccadic flight and gaze control system interacts with the system that mediates compensatory optomotor responses by reducing the slip velocity between the animal and its environment that results from unintended course deviations [46–48]. Such deviations may result, for instance, from asymmetries of the fly's motor system. The different dynamical properties of both systems would explain why saccadic turns are not counteracted by the compensatory optomotor system: Actively induced fast saccadic turns are not impeded by the much slower optomotor system [48,49]. There is evidence for this interpretation from recent behavioural experiments both on walking [50] and flying blowflies (R. K., unpublished data).
Our results suggest that the computational design principles of the blowfly visual motion pathway are adapted to the active vision strategies of the animal and therefore allow the extraction of behaviourally relevant information. Information resulting mainly from sideward movements of the animal in the intersaccadic interval may be used to elicit saccades that prevent the animal from crashing into an obstacle. Indeed, there is evidence from combined behavioural and modelling analyses on Drosophila that image expansion in the lateral visual field may play this important role [51].
In conclusion, our results provide an example in which a novel functional role of a neuron emerges by probing the neuron with stimuli that are actively generated by the animals' own behaviour. The new role is revealed here because the behaviourally generated input has dynamical properties strongly differing from those of conventional experimenter-designed stimuli.
Materials and Methods
Stimulus generation and electrophysiology.
The position and orientation of the head of blowflies flying in a cage of 40 × 40 × 40 cm3, with images of herbage on its side walls, were recorded using magnetic fields driving search coils attached to the flies [5,8]. Because the fly's compound eye is an integral part of its head, and the visual interior of the cage is known, the visual stimulus encountered by the fly during a flight could be reconstructed. Reconstructions of ten flights of 3.45 s, originating from three different flies, were played back on a panoramic stimulus device [9] at a frame rate of 370 Hz. Proper spatial and temporal prefiltering prevented spatiotemporal aliasing during fast turns [9]. An approximation of the response of the contralateral HSE to the same flight was obtained by presenting a mirrored version of the reconstruction. Intracellular recordings were made from the HSE-cell in the right optic lobe of 1- to 2-d-old female blowflies of Calliphora vicina, following standard routines [52] and ensuring careful alignment of the flies' eyes. Results are based on HSE recordings from 14 flies.
Data analysis.
Coherence between stimulus and response was calculated as γ2b = |Psr|2/(PssPrr)
[53], where Psr is the cross spectral density of stimulus and response, Pss is the power spectral density of the stimulus, and Prr is that of the response. The filter Psr/Prr reconstructs stimulus from response, and Prs/Pss, response from stimulus. Spectra were calculated by periodogram averaging of 50% overlapping data segments, with each periodogram the discrete Fourier transform of a cos2-tapered zero-mean data segment of 256 ms, extended by zero-padding to 512 ms. Results were not strongly dependent on segment length. Before segmentation, the response was aligned with the stimulus by shifting it 22.5 ms backwards in time, the approximate latency under the conditions of these experiments. Results were not strongly dependent on shift size. Segments from all flights used as stimulus for a particular cell were included in the periodogram averaging. Bias in the coherence estimate was corrected [54] by γ2 = n/(n − 1)γ2b − 1/(n − 1)
, where n is the total number of segments. Coherence of the response with two parameters of the stimulus was obtained by first conditioning the second parameter with the first [55], i.e., s′2 = s
2 − (P
21/P
11)s
s
, where s
1 is the first parameter, and s
2 and s′2
is the original and conditioned second parameter, respectively; P
21 and P
11 are cross and power spectra of the second and first parameter. Conditioning removes from s
2 the second-order statistical dependence with s
1. With three stimulus parameters (e.g., yaw, sideward, and forward velocity (see Figure 3A), the conditioned third parameter is s′3 = s
3 − (P
32′/P
2′2′)s′2 − (P
31/P
11)s
1
, which removes from s
3 the second-order statistical dependence with both s
1 and s′2
. The order of evaluating parameters does not significantly affect the results for the stimulus parameters used in this study because they are almost uncorrelated.
Masks selecting saccadic segments in stimulus and response were obtained by gating (transmitting) a region surrounding saccades, here defined as peaks (≥500 deg/s) in the total angular velocity of the head. The region was large enough to include all parts of both saccadic stimulus and corresponding response. Saccades that were close together were merged to reduce boundary effects. Edges of the masks were tapered with a 12.5-ms cos2 taper to reduce spectral leakage biasing the coherence estimate at high frequencies. The intersaccadic mask, used for suppressing the saccadic stimulus and response, equals one minus the saccadic mask. Masked data consisted of gated data intermitted with blocks of zeroes. Although the mask shapes the power and cross spectra of the masked data, this occurs in a similar way for all spectra in the numerator and denominator of the definition of coherence. Consequently, the mask by itself does not generate coherence for uncorrelated data, which was checked in control computations with uncorrelated noise. The coherence of masked data includes the zero blocks, however, and therefore should be regarded as belonging to the entire masked signal, not just to its intersaccadic part.
The power spectra of Figure 2B were not calculated for the entire masked data because the mask dominates the shape of these spectra, producing a strong peak at the saccade rate, approximately 10 Hz. Instead, we used a routine for calculating the Fourier transform of gapped data [56] and obtained the power spectra by averaging the squared amplitude over segments. The routine ignores all data masked, defined here as points where the intersaccadic mask was smaller than 0.5.
The total optic flow for Figure 1C was calculated by projecting the optic flow onto the local response field of the right HSE cell, i.e., it is weighted according to the cell's local preferred directions and motion sensitivities.
Supporting Information
Video S1 Flight of 3.45 s, Shown Ten Times Slower than Real Time (at 25 fps Playback Speed)
The left panel shows a reconstruction of the flight in a schematic cage, with, for the sake of clarity, the fly rendered three times larger, relative to the cage, than it is in reality. The middle panel shows an enlarged view of the orientational movements of the fly's thorax (blue) and head (red) during the same flight [5,8]. The right panel shows the visual scene, viewed from the centre of the head, during the same flight. It shows a 180 deg fisheye projection, with the centre straight ahead, the far right of the image pointing at 90 deg to the right, and the upper part of the image pointing straight up. The section of 1.5 s length used for Video S2 starts at 1.11 s from the beginning, i.e., approximately at one third of the video. Note the visual consequences of the saccadic yaw changes of thorax and head, and the roll compensation of the head.
(9.9 MB ZIP).
Click here for additional data file.
Video S2 Illustration of the OT and OR Control Stimuli
The middle panel shows 1.5 s of the video corresponding to the original trajectory, shown 40 times slower than real time, for a patch of 30 deg × 30 deg in a direction in the horizontal plane at 45 deg to the right of the frontal axis of the eye (azimuth 45 deg, elevation 0 deg). This direction is in the middle of the receptive field of the right HSE. The letters S and I in the title bar signal the time course of the saccadic and intersaccadic masks used for analysing the saccadic and intersaccadic responses, respectively. The left panel shows the OT control in which between saccades (I on), all rotation is removed and only translational optic flow is presented to the fly. The right panel shows the OR control in which between saccades, all translation is removed and only rotational optic flow is presented to the fly. Note that the intersaccadic speeds for sideward translation (original and OT) and rotation (original and OR) are of the same order of magnitude (see the overlapping velocity distributions of yaw and sideward velocity in Figure 2A), whereas the rotations extend to higher temporal frequencies than the translations (see the power spectra of yaw and sideward velocity in Figure 2B). This is reflected in the coding of rotation and translation in the HSE neuron (see the coherences for yaw and sideward velocity in Figure 3A, 3C, and 4B).
(8.8 MB ZIP).
Click here for additional data file.
We thank N. Boeddeker, J. Grewe, K. Karmeier, H. P. Snippe, and A.-K. Warzecha for a critical reading of the manuscript. Supported by the Deutsche Forschungsgemeinschaft (DFG).
Competing interests. The authors have declared that no competing interests exist.
Author contributions. RK, JHvH, JPL, and ME conceived and designed the experiments. RK and CM performed the experiments. RK, JHvH, CM, and ME analyzed the data. JHvH and JPL contributed materials and analysis tools. RK, JHvH, and ME wrote the paper.
Citation: Kern R, van Hateren JH, Michaelis C, Lindemann JP, Egelhaaf M (2005) Function of a fly motion-sensitive neuron matches eye movements during free flight. PLoS Biol 3(6): e171.
Note Added in Proof
During the review of this paper, we (RK, JHvH, ME) submitted a related paper to the Journal of Neuroscience, which has since been published: van Hateren JH, Kern R, Schwerdtfeger G, Egelhaaf M (2005) Function and coding in the blowfly H1 neuron during naturalistic optic flow. J Neurosci 25: 4343-4352.
Abbreviations
HSEhorizontal system equatorial cell
ORonly rotation
OTonly translation
SEMstandard error of the mean
==== Refs
References
Findlay JM Gilchrist ID Active vision: The psychology of looking and seeing 2003 Oxford Oxford University Press 220
Land MF Collett TS Srinivasan MV Venkatesh S A survey of active vision in invertebrates From living eyes to seeing machines 1997 Oxford Oxford University Press 16 36
Land MF Head movement of flies during visually guided flight Nature 1973 243 299 300
Schilstra C van Hateren JH Blowfly flight and optic flow. I. Thorax kinematics and flight dynamics J Exp Biol 1999 202 1481 1490 10229694
van Hateren JH Schilstra C Blowfly flight and optic flow. II. Head movements during flight J Exp Biol 1999 202 1491 1500 10229695
Egelhaaf M Kern R Kurtz R Krapp HG Kretzberg J Neural encoding of behaviourally relevant motion information in the fly Trends Neurosci 2002 25 96 102 11814562
Borst A Haag J Neural networks in the cockpit of the fly J Comp Physiol A Neuroethol Sens Neural Behav Physiol 2002 188 419 437 12122462
Schilstra C van Hateren JH Stabilizing gaze in flying blowflies Nature 1998 395 654 9790186
Lindemann JP Kern R Michaelis C Meyer P van Hateren JH FliMax, a novel stimulus device for panoramic and high speed presentation of behaviourally generated optic flow Vision Res 2003 43 779 791 12639604
Hausen K Motion sensitive interneurons in the optomotor system of the fly. I. The horizontal cells: Structure and signals Biol Cybern 1982 45 143 156
Hausen K Motion sensitive interneurons in the optomotor system of the fly. II. The horizontal cells: Receptive field organization and response characteristics Biol Cybern 1982 46 67 79
Hausen K Egelhaaf M Stavenga D Hardie RC Neural mechanisms of visual course control in insects Facets of vision 1989 Berlin Springer Verlag 391 424
Haag J Borst A Active membrane properties and signal encoding in graded potential neurons J Neurosci 1998 18 7972 7986 9742164
Warzecha AK Kurtz R Egelhaaf M Synaptic transfer of dynamical motion information between identified neurons in the visual system of the blowfly Neuroscience 2003 119 1103 1112 12831867
Horstmann W Egelhaaf M Warzecha AK Synaptic interactions increase optic flow specificity Europ J Neurosci 2000 12 2157 2165
Koenderink JJ Optic Flow Vision Res 1986 26 161 180 3716209
Eckert MP Zeil J Zanker JM Zeil J Towards an ecology of motion vision Motion vision: Computational, neural, and ecological constraints 2001 Berlin Springer Verlag 333 369
Gibson JJ The perception of the visual world 1950 Boston Houghton Mifflin 235
Boeddeker N Lindemann JP Egelhaaf M Zeil J Zimmermann H Krieglstein K Analysis of neuronal responses in the blowfly visual system to optic flow under natural outdoors conditions Proceedings of the 6th meeting of the German Neuroscience Society. 30th Göttingen Neurobiology Conference; 2005 Feb 17–20 2005 1 Suppl Göttingen, Germany Neuroforum 2005 22B
Krapp HG Hengstenberg R Egelhaaf M Binocular contribution to optic flow processing in the fly visual system J Neurophysiol 2001 85 724 734 11160507
Kern R Petereit C Egelhaaf M Neural processing of naturalistic optic flow J Neurosci 2001 21 1 5
Kern R Lutterklas M Petereit C Lindemann JP Egelhaaf M Neuronal processing of behaviourally generated optic flow: Experiments and model simulations Network 2001 12 351 369 11563534
Bialek W Rieke F de Ruyter van Steveninck R Warland D Reading a neural code Science 1991 252 1854 1857 2063199
Haag J Borst A Encoding of visual motion information and reliability in spiking and graded potential neurons J Neurosci 1997 17 4809 4819 9169539
Srinivasan MV The impulse response of a movement-detecting neuron and its interpretation Vision Res 1983 23 659 663 6613007
Egelhaaf M Reichardt W Dynamic response properties of movement detectors: Theoretical analysis and electrophysiological investigation in the visual system of the fly Biol Cybern 1987 56 69 87
Egelhaaf M Borst A Transient and steady-state response properties of movement detectors J Opt Soc Am A 1989 6 116 127 2921651
Maddess T DuBois R Ibbotson MR Response properties and adaptation of neurones sensitive to image motion in the butterfly Papilio aegeus J Exp Biol 1991 161 171 199
Maddess T Laughlin SB Adaptation of the motion-sensitive neuron H1 is generated locally and governed by contrast frequency Proc R Soc Lond B 1985 225 251 275
Ruyter van Steveninck R de, Zaagman WH Mastebroek HAK Adaptation of transient responses of a movement-sensitive neuron in the visual system of the blowfly, Calliphora erythrocephala Biol Cybern 1986 54 223 236
Borst A Egelhaaf M Temporal modulation of luminance adapts time constant of fly movement detectors Biol Cybern 1987 56 209 215
Harris RA O'Carroll DC Laughlin SB Adaptation and the temporal delay filter of fly motion detectors Vision Res 1999 39 2603 2613 10492824
Harris RA O'Carroll DC Laughlin SB Contrast gain reduction in fly motion adaptation Neuron 2000 28 595 606 11144367
Reisenman C Haag J Borst A Adaptation of response transients in fly motion vision. I. Experiments Vision Res 2003 43 1291 1307 12726835
Borst A Reisenman C Haag J Adaptation to response transients in fly motion vision. II: Model studies Vision Res 2003 43 1309 1322 12726836
Brenner N Bialek W de Ruyter van Steveninck R Adaptive rescaling maximizes information transmission Neuron 2000 26 695 702 10896164
Fairhall AL Lewen GD Bialek W de Ruyter van Steveninck R Efficiency and ambiguity in an adaptive neural code Nature 2001 412 787 792 11518957
Heitwerth J Egelhaaf M Zimmermann H Krieglstein K A new role of motion adaptation in visual motion pathway of the blowfly Proceedings of the 6th meeting of the German Neuroscience Society. 30th Göttingen Neurobiology Conference; 2005 Feb 17–20 2005 1 Suppl Göttingen, Germany Neuroforum 2005 17B
Nalbach G Hengstenberg R The halteres of the blowfly Calliphora. II. Three-dimensional organization of compensatory reactions to real and simulated rotations J Comp Physiol A 1994 175 695 708
Nalbach G The halteres of the blowfly Calliphora I. Kinematics and dynamics J Comp Physiol A 1993 173 293 300
Chan WP Prete F Dickinson MH Visual input to the efferent control system of a fly's “gyroscope.” Science 1998 280 289 292 9535659
Sherman A Dickinson MH Sumation of visual and mechanosensory feedback in Drosophila flight control J Exp Biol 2004 207 133 142 14638840
Fortune ES Rose GJ Passive and active membrane properties contribute to the temporal filtering properties of midbrain neurons in vivo J Neurosci 1997 17 3815 3825 9133400
Fortune ES Rose GJ Short-term synaptic plasticity as a temporal filter Trends Neurosci 2001 24 381 385 11410267
Trussell LO Synaptic mechanisms for coding timing in auditory neurons Annu Rev Neurosci 1999 61 477 496
Götz KG The optomotor equilibrium of the Drosophila navigation system J Comp Physiol 1975 99 187 210
Heisenberg M Wolf R Vision in Drosophila 1984 Berlin Springer Verlag 250
Egelhaaf M Dynamic properties of two control systems underlying visually guided turning in house-flies J Comp Physiol A Neuroethol Sens Neural Behav Physiol 1987 161 777 783
Warzecha AK Egelhaaf M Intrinsic properties of biological motion detectors prevent the optomotor control system from getting unstable Phil Trans R Soc Lond B 1996 351 1579 1591
Kern R Egelhaaf M Optomotor course control in flies with largely asymmetric visual input J Comp Physiol A Neuroethol Sens Neural Behav Physiol 2000 186 45 55
Tammero LF Dickinson MH The influence of visual landscape on the free flight behavior of the fruit fly Drosophilia melanogaster J Exp Biol 2002 205 327 343 11854370
Warzecha AK Egelhaaf M Borst A Neural circuit tuning fly visual interneurons to motion of small objects. 1. Dissection of the circuit by pharmacological and photoinactivation techniques J Neurophysiol 1993 69 329 339 8459270
Theunissen F Roddey JC Stufflebeam S Clague H Miller JP Information theoretic analysis of dynamical encoding by four identified primary sensory interneurons in the cricket cercal system J Neurophysiol 1996 75 1345 1364 8727382
van Hateren JH Rüttiger L Sun H Lee BB Processing of natural temporal stimuli by macaque retinal ganglion cells J Neurosci 2002 22 9945 9960 12427852
Bendat JS Piersol AG Random data: Analysis and measurement procedures 2000 New York Wiley-Interscience 594
Scargle JD Studies on astronomical time series analysis. III. Fourier transforms, autocorrelation functions, and cross-correlation functions of unevenly spaced data Astrophys J 1989 343 874 887
| 15884977 | PMC1110907 | CC BY | 2021-01-05 08:21:23 | no | PLoS Biol. 2005 Jun 17; 3(6):e171 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030171 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1588497810.1371/journal.pbio.0030172Research ArticleBioinformatics/Computational BiologyGeneral MedicineRespiratory MedicineStatisticsBiochemistryBiophysicsCell BiologyEvolutionGenetics/Genomics/Gene TherapyInfectious DiseasesMicrobiologyMolecular Biology/Structural BiologyVirologyVirusesA Three-Stemmed mRNA Pseudoknot in the SARS Coronavirus Frameshift Signal A New Frameshifting Pseudoknot in SARSPlant Ewan P
1
Pérez-Alvarado Gabriela C
2
Jacobs Jonathan L
1
Mukhopadhyay Bani
1
Hennig Mirko
2
Dinman Jonathan D [email protected]
1
1Department of Cell Biology and Molecular Genetics, University of MarylandCollege Park, MarylandUnited States of America2Department of Molecular Biology and the Skaggs Institute for Chemical Biology, The Scripps Research InstituteLa Jolla, CaliforniaUnited States of AmericaWickens Marv Academic EditorUniversity of WisconsinUnited States of America6 2005 17 5 2005 17 5 2005 3 6 e17226 1 2005 14 3 2005 Copyright: © 2005 Plant 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.
New Frameshifting Pseudoknot Found in SARS Virus
Pseudoknots: RNA Structures with Diverse Functions
A wide range of RNA viruses use programmed −1 ribosomal frameshifting for the production of viral fusion proteins. Inspection of the overlap regions between ORF1a and ORF1b of the SARS-CoV genome revealed that, similar to all coronaviruses, a programmed −1 ribosomal frameshift could be used by the virus to produce a fusion protein. Computational analyses of the frameshift signal predicted the presence of an mRNA pseudoknot containing three double-stranded RNA stem structures rather than two. Phylogenetic analyses showed the conservation of potential three-stemmed pseudoknots in the frameshift signals of all other coronaviruses in the GenBank database. Though the presence of the three-stemmed structure is supported by nuclease mapping and two-dimensional nuclear magnetic resonance studies, our findings suggest that interactions between the stem structures may result in local distortions in the A-form RNA. These distortions are particularly evident in the vicinity of predicted A-bulges in stems 2 and 3. In vitro and in vivo frameshifting assays showed that the SARS-CoV frameshift signal is functionally similar to other viral frameshift signals: it promotes efficient frameshifting in all of the standard assay systems, and it is sensitive to a drug and a genetic mutation that are known to affect frameshifting efficiency of a yeast virus. Mutagenesis studies reveal that both the specific sequences and structures of stems 2 and 3 are important for efficient frameshifting. We have identified a new RNA structural motif that is capable of promoting efficient programmed ribosomal frameshifting. The high degree of conservation of three-stemmed mRNA pseudoknot structures among the coronaviruses suggests that this presents a novel target for antiviral therapeutics.
A new structural and conserved element is identified within the SARS virus genome. The element is important for gene expression, and might be a useful target for antiviral drugs.
==== Body
Introduction
Severe acute respiratory syndrome (SARS) first appeared in Guangdong Province, China, late in 2002. Its rapid transmission and high rates of mortality and morbidity resulted in a significant threat to global health by the spring of 2003, and the epidemic had a significant effect on the public health and economies of locales affected by SARS outbreaks. The rapid response of the World Health Organization is credited with containing this contagion by late June 2003, and only a few cases were reported during the winter cold season of 2003–2004. The severity of this crisis mobilized the scientific community as well: by March 24, 2003, scientists at the Centers for Disease Control and Prevention and in Hong Kong had announced that a new coronavirus had been isolated from patients with SARS (reviewed in [1]). The sequences from two isolates of SARS-CoV were published simultaneously on May 1, 2003 [2,3].
Coronaviruses are enveloped animal viruses that cause respiratory and enteric diseases. Analysis of the SARS-CoV genome revealed that, similar to all coronaviruses, the 70% (approximately) at the 5′ end of its large single (+) stranded RNA genome consists of two sizable genes called ORF1a and ORF1b. The 3′ ORF1b overlaps, and is out of frame with, its 5′ neighbor, ORF1a, and similar to other coronaviruses, a programmed −1 ribosomal frameshift (−1 PRF) was posited to be used by the virus to produce an ORF1a/1b fusion protein [2]. A wide range of RNA viruses use −1 PRF for the production of viral fusion protein (reviewed in [4–6]). In many such cases, e.g., the Retroviridae and Totiviridae, the efficiency of ribosomal frameshifting determines the stoichiometric ratio between structural and enzymatic proteins available for viral particle assembly, and even small changes in frameshift frequencies can have profound negative effects on virus propagation, thus targeting −1 PRF for antiviral therapies (reviewed in [7]). It has been shown that the SARS-Cov −1 PRF signal is able to promote efficient frameshifting in a rabbit reticulcyte system, and the −1 PRF signal reported in that publication consisted of a typical heptameric “slippery site” (UUU
AAAC), a 5-nt spacer, and a typical H-form mRNA pseudoknot containing two double-stranded RNA stems and two single-stranded loops [8]. Two very recently published papers have suggested that the SARS-CoV mRNA pseudoknot may contain a third stem-loop structure [9, 10]. In this work, we present computational, comparative genomic, molecular, biophysical, and genetic evidence demonstrating that the SARS-CoV frameshift signal includes a new type of highly ordered three-stemmed mRNA pseudoknot that likely contains a large number of noncanonical base interactions. Although total deletion of the third stem does not significantly alter frameshifting efficiency, its disruption significantly inhibits this process. The fact that this general structure appears to be conserved among the coronaviruses raises questions regarding its biological function.
Results
Computational Analysis of the SARS-CoV Frameshift Signal Suggests the Presence of a Three-Helix-Containing RNA Pseudoknot
−1 PRF signals typically have a tripartite organization. From 5′ to 3′, these are composed of a heptameric “slippery site,” a “spacer” region, and a stable mRNA secondary structure, typically an mRNA pseudoknot (reviewed in [11]). A previous analysis of the SARS-CoV −1 PRF signal demonstrated that a sequence spanning nucleotide positions 13392–13472 satisfied these three requirements and was able to promote efficient −1 PRF in rabbit reticulocyte lysates [8]. The −1 PRF signal presented in that study contained a typical mRNA pseudoknot composed of two double-helical, Watson–Crick basepaired stems connected by two single-stranded loops (Figure 1A).
Figure 1 Different Representations of the SARS-CoV Frameshift Signal
(A) Two-stemmed H-type mRNA pseudoknot proposed by Thiel et al. [8].
(B) Three-stemmed mRNA pseudoknot structure investigated in this study.
The presence of a long, 29-nt loop 2 seemed to be unusual, prompting us to subject the sequence from positions 13392–13472 to additional computational analyses in an effort to further define the structure of this mRNA pseudoknot. The nucleotide sequence suspected of featuring a −1 PRF signal between ORF1a and ORF1b was scanned by RNAMotif [12], using a pattern-based description capable of finding common −1 PRF signals in other RNA viruses. As expected, a so-called slippery site (UUU
AAAC) and a large H-type pseudoknot were identified—the two primary stimulating elements required for efficient ribosomal slippage. This analysis was coupled with Pknots [13], a software package that predicts the most thermodynamically stable structure for a given RNA sequence. The predicted structure for the SARS-CoV frameshift signal was extremely stable, with a calculated minimum free energy (MFE) of −26.68 kcal/mol. The surprising result was that the 29-nt sequence designated loop 2 by Thiel et al. [8] was predicted to form a third helix, nested within the sequences defined by stems 1 and 2 (Figure 1B). Though a small, internally nested third helix (helix-3) has been shown to be present in the HIV-1 group O frameshift signal [14], such an extensive basepairing pattern has not to our knowledge been heretofore demonstrated for any other viral frameshift signal. To determine the statistical significance of this finding, a distribution of MFE values taken from 500 randomly shuffled SARS-CoV frameshift signals was created. Each of the randomly shuffled sequences was folded using Pknots with the same parameters. The resulting normal distribution had mean MFE of −21.12 kcal/mol (standard deviation = 2.67, 500×), revealing that the predicted three-stemmed pseudoknot structure of the native sequence is highly significant with a z-score of −2.05 and a p value of 0.02 (one-tailed Student's t-test).
Phylogenetic Conservation of Predicted Three-Stemmed mRNA Pseudoknots in Coronaviruses
To address the question of whether the potential to form a three-stemmed mRNA pseudoknot is unique to the SARS-CoV, we searched for such structures in all of the known viral −1 PRF signals listed in the RECODE 2003 database [15], as well as the putative frameshift signals in all of the sequenced members of the Order Nidovirales (including coronaviruses and arteriviruses). The SARS-CoV frameshift signal itself is homologous to all of the nine other frameshift signals for coronaviruses whose genomes have been fully sequenced. A multiple sequence alignment of the ten coronavirus frameshift signals is presented in Figure 2. This shows that both stems 1 and 2 are highly conserved, with a strong conservation of base complementation in the cores of both stems 1 and 2 (blue and red sequences, respectively). This analysis also shows all of the coronavirus frameshift signals have the potential to form a third helix, although the structures and sequences are less well conserved (Figure 2, in green). In addition, the potential of sequences located approximately 200 nt downstream of the slippery site to form long-range “kissing loop” interactions with the 5′ half of stem 2 was previously noted for HCoV-229E [16] and TEGV [17]. This property was only conserved among all of the group 2 coronaviruses, not in any of the others (see Figure 2). The potential significance of this observation is discussed below. A phylogenetic tree of the −1 PRF signals constructed from the multiple sequence alignment is presented in Figure 3. As expected, the group 1 and group 2 coronaviruses cluster together, and neither the SARS-CoV nor the avian infectious bronchitis virus (AIBV) frameshift signals cluster with either group. Of particular interest, however, is that very similar mRNA pseudoknot structures are predicted to occur within groups—but not between them.
Figure 2 Multiple Sequence Alignment of the SARS-CoV −1 PRF with Nine Homologous Signals Found in Other Coronavirus Genomes
AIBV, avian infectious bronchitis virus; BCoV, bovine coronavirus; HCoV-229E, human coronavirus 229E; HCoV-HKU1; HCoV-NL63, human coronavirus NL63; HCoV-OC43, human coronavirus OC43; MHV, murine hepatitis virus; PEDV, porcine epidemic diarrhea virus; SARS, SARS coronavirus; TGV, transmissible gastroenteritis virus. Heptameric slippery sites are indicated in brown; dashes indicate gaps in the sequence alignments; basepairing positions involved in the consensus first, second, and third helices are denoted by blue, red, and green nucleotides, respectively. Downstream regions homologous to the kissing loop known to promote frameshifting in HCoV-229E [16,17]. HCoV-229R, HCoV-NL, PEDV, and TGV are also highlighted in red with the flanking stem-forming sequences underlined. Asterisks indicate perfectly conserved positions in primary sequence.
Figure 3 Phylogenetic Analyses of Coronavirus −1 PRF Signals
Unrooted tree constructed based on the multiple sequence alignment from Figure 2.
Nuclease Mapping of the SARS-CoV Frameshift Signal Is Consistent with the Presence of a Complex, Three-Helix-Containing RNA Pseudoknot Structure
In light of the computational findings, we conducted biochemical analyses of the SARS-CoV frameshift signal using a [32P] 5′ end labeled SP6 RNA polymerase product spanning nucleotides 13399–13475. RNase A cleaves preferentially at single-stranded pyrimidine bases, RNase T1 cuts at single-stranded guanosine residues, and RNase V1 cleaves double-stranded RNA. We also examined alkaline hydrolysis cleavage patterns at low concentrations of sodium hydroxide to identify exposed phosphodiester bonds. Representative autoradiograms of the reactions are shown in Figure 4A and 4B, and the predicted cleavage patterns mapped onto the pseudoknot structure are shown in Figure 4C.
Figure 4 Secondary Structure Mapping of the SARS-CoV Frameshift Signal
(A and B) The results of nuclease cleavage of RNA from nucleotides 13400–13470 of SARS-CoV. RNAs were 5′ end labeled with 32P and subjected to enzymatic digestion, as described in Materials and Methods. The three different concentrations of each nuclease are indicated by the triangles are described in Materials and Methods. C denotes undigested control, and OH− denotes hydrolysis ladders.
(C) Interpretation of nuclease digestion analyses mapped onto the proposed secondary structure of the SARS-CoV frameshift signal. Nuclease cleavage sites, proposed basepairs, and specific bases protected from nuclease attack are indicated.
The nuclease mapping data are generally consistent with the computational predictions, showing double-stranded regions corresponding to all three stems. Some notable deviations from the predicted structure were observed, however. These fell into three general classes. One class consisted of distortions in predicted helical A-RNA structures, typified by bases that were equally digested by both single- and double-strand-specific nucleases and by nearby bases that were refractory to nuclease attack. These clustered in the middle of stem 1 (13406–13410 and 13427), near the middle of stem 2 (13419–13420), and in the middle of stem 3 (13436–13439 and 13458–13461). Another major group consisted of bases located in regions predicted to link the three stems that were completely protected from nuclease attack. Specifically, these were G13405 and G13435 at the stem 1/stem 3 junction, G13414 and G13423–C13425 at the stem 1/stem 2 junction, and C13463 and U13464, which link stem 2 with stem 3. We also observed enhanced susceptibility of the three pyrimidines in the predicted loop 2 region (C13447, U13448, and U13451) to attack by both single- and double-strand-specific endonucleases, suggesting that this region is structurally dynamic under the conditions assayed. The ability of the bulged adenosine residue at position 13467 to be recognized by RNaseV1 demonstrates that it is involved in a basepairing interaction, whereas the opposite pertains with regard to A13446, A13452, and A13456. The three bases at the 5′ and 3′ terminal ends of the molecule could not be meaningfully resolved.
Two-Dimensional Nuclear Magnetic Resonance Analysis Confirms the Presence of Three Stems
Given the ambiguity of the nuclease mapping, homo- and heteronuclear two-dimensional (2D) nuclear magnetic resonance (NMR) experiments were used to confirm the predicted basepairing interactions of the three stems. The presence of 21 hydrogen-bonded guanine and uracil residues for the sequence including residues 13405–13472 of the SARS-CoV genome (Figure 5A) was evident from the imino region observed in the 2D 1H,1H-NOESY, 15N-HMQC, and quantitative J(N,N) HNN-COSY data. In this study, we have obtained sequential imino 1H and 15N assignments for the A-form helices of the frameshifting SARS pseudoknot, using a combination of information from 1H,1H-NOE and 1H,15N-HMQC spectra. The latter experiment distinguishes between uridine and guanosine iminos by the characteristic 15N chemical shift, whereas the NOESY yields sequential 1H,1H-NOEs connecting imino protons in helical stem regions.
Figure 5 NMR Data Were Collected at 25 °C at a Proton Resonance Frequency of 900 MHz
(A) Secondary structure of the SARS-CoV frameshift pseudoknot (residues 13405–13472). Different color coding was used to denote basepaired regions in stems 1 (cyan), 2 (green), and 3 (grey and blue). Only the last two digits of the wild-type sequence numbering are used for clarity.
(B) Imino region of a one-dimensional jump-return echo spectrum of SARS-CoV pseudoknot.
(C) Portion of a 2D 1H,1H-NOESY. Sequential imino-imino proton NOE assignment paths are shown by different colors for stem 1 (cyan), stem 2 (green), and stem 3 (black and blue).
(D) 2D Quantitative J(N,N) HNN-COSY spectrum showing interstrand 1H3–15N3(U) to 15N1(A) and 1H1–15N1(G) to 15N3(C) correlations. Data were collected on a uniformly 13C/15N-labeled sample. Red peaks correspond to diagonal resonances and are labeled with assignment information for the basepaired stem regions matching the color coding in (C). Green cross peaks are caused by scalar cross hydrogen bond 2h
J(N,N) couplings detected using a defocusing delay of 36 ms. Carrier positions were on water for 1H and 185 ppm.
Assignments were made for 8 bp in stem 1, 4 bp in stem 2, and 5 bp in stem 3 (Figure 5). As a result of fraying, the terminal basepairs of helical stem regions are not observed. Of those basepairs assigned, 16 are Watson–Crick-type basepairs and only one is a canonical wobble G:U basepair. This G:U basepair present in stem 3 can be inferred directly from the strong NOE correlation between the G38 and U59 imino protons (Figure 5C). The corresponding donor G38:N1 and U59:N3 imino nitrogens are evidently not engaged in G:C or U:A hydrogen bonds (Figure 5D).
The quantitative J(N,N) HNN-COSY contains a total of five correlations between the imino N3 nitrogens of uridines and the N1 nitrogens in adenines, indicative of canonical Watson–Crick-type basepairing interaction. A total of 11 correlations stemming from Watson–Crick G:C basepairs are observed between the imino N1 nitrogens of guanosines and the N3 nitrogens of cytidines. In summary, the complete sequential NOE walk connecting most of the basepaired imino protons unambiguously confirmed the presence of three stems corresponding to the secondary structure prediction shown (Figure 5A).
The Predicted SARS-CoV Frameshift Signal Functions Like Other −1 PRF Signals
To address the question of whether the predicted SARS-CoV −1 PRF signal functions similarly to −1 PRF promoting elements from other viruses, this sequence was cloned into bicistronic dual luciferase reporter constructs designed to assay programmed ribosomal frameshifting using in vitro and in vivo systems [18,19]. As a minor modification, instead of using a simple readthrough construct as the zero-frame control, the corresponding control contained the SARS-CoV −1 PRF signal with one additional base inserted 3′ of the Renilla luciferase sequence and 5′ of the −1 PRF signal. The resulting zero-frame reporter places the firefly luciferase ORF in frame with Renilla and inactivates the −1 PRF signal by moving it out of frame with regard to elongating ribosomes, while controlling for ribosomes dislodged from the reporter mRNAs by the mRNA pseudoknot. This seemed to alleviate the large errors observed by other groups using similar methodology (e.g., [9]).
The ability of the SARS-CoV sequence to promote −1 PRF was assessed using two different in vitro and in vivo assay systems each. The results of these experiments are shown in Figure 6A. In vitro, the SARS-CoV sequence was able to promote efficient −1 PRF in both wheat germ protoplasts (23.7% ± 1.9%) and rabbit reticulocytes (14.3% ± 3.7%). In vivo, the sequence was able to promote efficient −1 PRF in the Vero epithelial cell line (14.4% ± 0.6%), a finding that is important in light of the fact that the SARS-CoV infects lung epithelial cells. The sequence also promoted efficient −1 PRF in yeast cells, suggesting that this frameshift signal might be amenable to the molecular genetic toolbox available in the yeast system. To test this hypothesis, we examined the effects of a drug (anisomycin) and of a host cell mutant (mak8–1) that were previously shown to specifically affect L-A virus–directed −1 PRF in yeast cells [20–22]. The results of these experiments show that, similar to their effects on L-A-promoted −1 PRF, anisomycin was able to inhibit SARS-CoV-directed −1 PRF (21% inhibition, p = 5.04 × 10−8), whereas −1 PRF was stimulated in cells harboring the mak8–1 allele of RPL3 (25% stimulation, p = 5.0 × 10−5) (Figure 6B). These findings show that the SARS-CoV frameshift signal is amenable to analysis by the full array of yeast-based genetic, pharmacological, and molecular tools that we and others have developed. Interestingly, the absolute values for frameshifting in yeast (2.99% ± 0.06%) were significantly less than those observed in the other systems (ranging from approximately 15% to 25%), suggestive of differences between fungal and metazoan ribosomes that might be pharmacologically exploited. This is discussed in greater detail below.
Figure 6 Functional Characterization of the SARS-CoV Frameshift Signal
(A) The wild-type SARS-CoV frameshift signal promotes efficient frameshifting in vitro and in vivo. Programmed −1 ribosomal frameshifting was monitored in wheat germ and rabbit reticulocyte lysates in vitro, and in Vero epithelial cells and yeast in vivo, as described in Materials and Methods. Error bars denote the standard error.
(B) SARS-CoV-directed −1 PRF was monitored in wild-type yeast cells with or without anisomycin (20 μg/ml), or in isogenic RPL3 gene deletion cells expressing either the wild-type or mak8–1 alleles of RPL3 on an episomal plasmid [21]. Changes in −1 PRF efficiencies are shown as fold wild-type, and p-values are shown as described previously [25].
Structural Requirements for Efficient SARS-CoV Frameshifting Activity
Given that Vero cells more resemble the natural host of SARS-CoV than do yeast, a series of mutants of the SARS-CoV frameshift signal were developed to functionally dissect the mRNA pseudoknot in this cell type. Typically, mutagenesis experiments are constructed so as to change one or another side of a stem to disrupt basepairing, and then to combine the two mutants to re-form the stem (e.g., see [23,24]). The series of mutants that were created by oligonucleotide site-directed mutagenesis to address this question is shown in Figure 7. The S2 series of mutants were designed to examine the general requirement for stem 2, and the specific contribution of the bulged adenosine residue at position 13467 was designed to stimulate efficient −1 PRF. Similarly, the S3 mutant series were designed to examine the general requirement for stem 3, as well as the specific contribution of the bulged adenosine at position 13456. The complete data for these experiments, as formatted according to [25], are presented in Dataset S1.
Figure 7 Molecular Genetic Analyses of Stems 1 and 3
Constructs used to examine the contributions of stem structures and bulged adenosine residues to programmed −1 ribosomal frameshifting are depicted. Shading is used to indicate mutagenized bases. Programmed −1 ribosomal frameshifting promoted by the wild-type SARS-CoV −1 PRF signal was monitored in Vero, as described in the Materials and Methods. Standard deviations (S.D.) are indicated for each sample, as previously described [25]. The S2 series (above) examines the roles of structures and bases in stem 2. The S3 series (below) examines the roles of structures and bases in stem 3.
Six different stem 2 mutants were assayed for their ability to promote efficient −1 PRF (Figure 7). Not surprisingly, disruption of stem 2 (S2A and S2A′) precluded efficient −1 PRF. Unexpectedly, however, compensatory mutations that should promote re-formation of the basic stem 2 structure (S2B′) did not restore wild-type levels of frameshifting, suggesting the involvement of a primary mRNA sequence in this region in stimulating −1 PRF. However, the adenosine base at position 13465 in this construct had to be replaced with a guanosine to avoid creating a −1 frame termination codon. Though this substitution retains the potential to basepair with U13424, it is possible that the identity of the base at this position is critical. To examine this parameter, the base at this position was changed to guanosine in the context of an otherwise wild-type −1 PRF signal (A13465G). Though this mutation did not abrogate efficient −1 PRF, frameshifting efficiency was decreased by approximately 38% (p = 1.7 × 10−6). This result suggests that though this mutation was not the main cause of the dramatic reduction in −1 PRF observed with S2B′, the identity of the base at this position is important for maximizing −1 PRF efficiency. These observations are consistent with the hypothesis that both the general structure and base-specific sequence of stem 2 are required for efficient −1 PRF.
Bulged adenosine residues are known to stimulate assembly of higher-order RNA structures by helping to link helices together [26]. Two constructs were assayed to examine the requirement of the bulged residue at position 13467 for efficient −1 PRF. In mutant S2C′, A13467 was substituted with cytosine, whereas in construct S2C, the adenosine at position 13467 was removed from the middle of stem 2 and repositioned six bases downstream to maintain translational reading frame. Either replacing the A-bulge with cytosine (S2C′) or deleting it entirely (S2C) dramatically reduced frameshifting in Vero cells (>94%, p < 3.3 × 10−16), repressing −1 PRF to a similar extent as the mutants S2A, S2A′, and S2B′.
Similar to the approach described above, five mutants were constructed to investigate stem 3 (Figure 7). Constructs S3A, S3A′, S3B, and S3C′ were directed toward addressing the function of stem 3: in S3A, the guanine and cytosine residues in the 5′ half of stem 3 were mutated to cytosine and guanine, respectively, disrupting stem 3; the opposing mutations were made in the 5′ half of stem 3 in S3A′; and S3B harbored the compensatory mutations to allow re-formation of stem 3. Frameshifting with S3A was reduced by 68% (p = 2.61 × 10−24), and −1 PRF was significantly, although less dramatically, reduced in S3A′ (36% of wild-type, p < 1.07 × 10−19). Similar to the effects observed in stem 2, the presence of compensatory mutations in construct S3B did not rescue −1 PRF efficiency to near wild-type levels, again suggesting that both the general structure of stem 3 and specific sequences within it are required for maximal stimulation of −1 PRF.
Similar to stem 2, a bulged adenosine is predicted in stem 3 at position 13456, and the phylogenetic analysis showed that this base was conserved among all of the coronaviruses. Substitution of this base to cytosine (S3C′) promoted a moderate but significant reduction in −1 PRF (26% inhibition, p = 2.56 × 10−11). In addition, as no significant internal nested stems have been observed in other viral frameshift pseudoknots, and because deletion of sequence corresponding to this region did not dramatically affect −1 PRF in AIBV [27], the entire stem 3–forming region was deleted in construct ΔS3 to create a more typical two-stemmed H-type RNA pseudoknot. In Vero cells, this smaller pseudoknot, lacking the third nested helix, actually promoted a modest increase in frameshifting (9.2%, p = 2.15 × 10−3), demonstrating that stem 3 is not critical for promoting efficient −1 PRF per se.
An analogous series of constructs were also assayed in yeast (data not shown). In general, the trends were similar, though the actual baseline frameshifting efficiencies were lower. For example, in mutants S2A, S2A′, and S2B′, frameshifting was equally reduced by 85%–90%; the S3A and S3A′ mutations also resulted in moderate (35%–89%) decreases in yeast, and deletion of stem 3 (ΔS3) also presented a slight increase in −1 PRF (35%) in yeast cells. There were some notable contrasts, however: though the S2C′ and S2C constructs dramatically reduced frameshifting in Vero cells (95%), they only reduced −1 PRF by approximately 25%–33% in yeast. More strikingly, some of the mutations that resulted in a 25%–30% decrease in −1 PRF in Vero cells (A13465G, S3B, and S3C′) did not affect the overall rate of −1 PRF in yeast at all. The potential significance of these findings is discussed below.
Discussion
Though the first descriptions of the mRNA secondary structure stimulating −1 PRF [28] and of an RNA pseudoknot [29] were serendipitously published back to back in 1985, the two concepts were only functionally linked together 1 y later in studies of a coronavirus, AIBV [30]. Here, the coronaviruses have again revealed a new twist on mRNA pseudoknots and −1 PRF. The phylogenetic comparisons presented in this study reveal that stem 1 lengths and G:C compositions are highly conserved in all ten coronavirus sequences analyzed. Their relatively long G:C-rich composition presumably contributes significantly to the stability of these structures. In contrast, stem 2 structures are predicted to vary significantly between the different coronavirus groups. Specifically, the group 2 coronaviruses (HCoV-043C, HCoV-HKU1, BCoV, and MHV) have the longest and most stable predicted stem 2 structures, whereas the stem 2 regions of the group 1 coronaviruses (TGV, PDEV, HCoV-NL, and HCoV-229E) are anticipated to be the least stable. The stem 2 regions of SARS-CoV and AIBV appear to be intermediate between these two. Although the sequences in the stem 3/loop 3 region are not well conserved, a third stem independently predicted in the SARS-CoV −1 PRF signal [10] has been demonstrated in this work, and we predict third stems in other coronavirus frameshift signals. Similar structures are generally predicted to be able to form within groups.
Specifically, loop 3 is predicted to be long and positioned between stems 3 and 2 in the group 2 coronaviruses. In contrast, the group 1 viruses contain little or no loop 3 but, rather, have an extended loop 2 positioned between stems 1 and 3. The notable exception is TGV, in which the relative structure and orientation of stem 3 and loop 3 more resembles those observed in SARS-CoV and AIBV.
Structurally, the nuclease analyses, showing distortions of the regular helical structures in the stems, protection of specific bases from nuclease attack, and the apparent involvement of bases in loop 3 and of A13467 in basepairing interactions, suggest that the three stems fold back on one another to form a complex, globular RNA structure. Long-range interaction anchors mediated by adenosine residues such as those at positions 13456 (stem 3) and 13467 (stem 2), making contact with the shallow minor grooves of two stacked basepairs of A-form helical stems, are a recurring theme in RNA structural biology. For example, the crystal structure of the ribosome reveals that RNA has a remarkable propensity for contributing adenine bases to such A-minor interactions [31], thereby stimulating the assembly of higher-order RNA structures [26]. Mounting evidence suggests that stimulation of −1 PRF by mRNA pseudoknots requires specific noncanonical basepairing between helical stems and pseudoknot loop regions to set specifically required frameshift efficiencies. The structures of the few frameshift-promoting pseudoknots that have been determined at the atomic level are revealing that a large range of higher-order noncovalent interactions serve to promote stable, novel structures [32–36]. It is clear that the three-helix-containing mRNA pseudoknot described here represents a novel global architecture stimulating ribosomal frameshifting, and possibly a source of new structural motifs in the coming future. Experiments are currently underway to define this structure at the atomic level, using high-resolution NMR techniques. Elucidation of this novel mRNA structure will be of great utility in the rational development of therapeutic agents designed to interfere with SARS-CoV programmed −1 ribosomal frameshifting, and in furthering our understanding of how different pseudoknots stimulate translational recoding.
Molecular genetic analysis of stem 2 of the SARS-CoV pseudoknot, demonstrating that frameshifting was reduced in all cases, including our attempts to make complementary mutations, indicates that primary sequences as well as structures are important for maximal frameshifting. Possible reasons for the observed sequence specificity could include aberrant folding or disruption of an essential interaction required for formation of the complex tertiary mRNA structure. For example, the findings that changing the identities of the bulged adenosine residues in stems 2 and 3 from adenosine to cytosine (S2C′ and S3C′), or deleting the bulge in stem 2 altogether (S2C), abrogated the stimulatory effects of the pseudoknot support the notion that this structural property of bulged adenosine residues is functionally important in this context. In addition, water-nucleobase “stacking” in the form of H-π and lone pair-π interactions have been demonstrated at the junctions between the stems in the BWYV pseudoknot [34]. The corresponding regions of the wild-type SARS-CoV pseudoknot were refractory to nuclease attack and were disrupted in S2A, S2A′, S2B′, S3, S3A′, and S3B, possibly explaining the effects of all of these mutants on frameshifting. The changes made in S2C and S2C′ are also adjacent to this region, and inhibition of frameshifting is nearly as dramatic as with the S2A, S2A′, and S2B′ mutants.
Though alterations to stem 3 significantly reduced frameshifting levels, these effects were one to two orders of magnitude less than analogous mutations of stem 2. This is supported by the observation in another study that alteration of the sequence in stem 3 also promoted decreases in −1 PRF [9]. Further, complete deletion of stem 3 had only a minimal effect on frameshifting efficiency—an observation consistent with studies in AIBV, in which deletion of all but 5 nt between stems 1 and 2 did not significantly alter −1 PRF [27,37]. These findings demonstrate that the presence of stem 3 is not required for efficient frameshifting per se. However, its high degree of conservation among the coronaviruses and its location in the frameshift signal suggest that it plays a more complex role in programmed −1 ribosomal frameshifting as it relates to the viral life-cycle. A similar conclusion was drawn by the authors of another independent study that was performed concurrently with ours and that was published while this manuscript was under review [9].
If stem 3 is not required to promote efficient frameshifting, why then has it been so highly conserved among the coronaviruses? It may be that frameshifting levels in coronaviruses need to be regulated in a manner not supported by a two-stem pseudoknot. For example, the frameshift signal marks the boundary between proteins required during the immediate early phase of infection (e.g., ORF1a-encoded proteases used to prepare the cell for virus production) and those required for intermediate functions in the viral life cycle (i.e., ORF1b-encoded RNA-dependent RNA polymerase and helicase used in transcription of subgenomic mRNAs, [−] strand synthesis, and genome replication). One of the fundamental problems of in the biology of (+) RNA viruses regards the switch between translation and replication. An elegant model proposes that the −1 ribosomal frameshift in barley yellow dwarf virus plays a central role in remodeling the (+) strand from translation competent to replication competent: frameshifting enables synthesis of the replicase, which in turn is able to denature the frameshift-promoting cis-acting element, eventually clearing the (+) strand of ribosomes that could potentially block the replicase [23]. In coronaviruses, the idea of functional switching by RNA remodeling has been demonstrated for MHV [38], and similar functional elements are present in both SARS-CoV and BCoV [39].
In a previous study, frameshifting in HCoV-229E was shown to be stimulated by a short sequence approximately 200 nt downstream from the slippery site, and it was shown that efficient frameshifting is promoted by kissing-loop interactions [16]. A subsequent report also found this potential motif in the TGV genome [17]. The phylogenetic analysis presented here reveals the potential to form similar short imperfect stem 2 structures for the other two group 1 coronaviruses for which the sequence is known (HCoV-NL and PEDV). In contrast, similar interactions cannot be readily discerned in SARS-CoV, nor among the group 2 and group 3 coronaviruses. Nevertheless, the idea that viral sequences in the pseudoknot may interact either in cis with sequences on the (+) strand or in trans with either sequences in subgenomic mRNAs or on the (−) strand to modulate frameshifting remains an intriguing possibility.
A final finding of interest derives from the observed differences between yeast- and metazoan-derived frameshift assay systems. This represents a potentially exciting avenue of exploration, as it may be indicative of mRNA folding differences between the two systems or of differences in how yeast versus metazoan ribosomes interact with downstream mRNA structures. This could be a result of relative size differences in the ribosomes. Alternatively, the lower levels of frameshifting in yeast relative to wild-type in the Vero cells could reflect a higher sensitivity of these ribosomes to subtle changes in the frameshift signal. The normal levels of frameshifting in yeast promoted by the S3B and S3C′ mutants further support the notion that the reason for stem 3 may lie with some function other than programmed ribosomal frameshifting.
Materials and Methods
Computational analyses
The SARS-CoV −1 PRF signal was identified from the complete genome sequence, using a combined approach. First, a pattern matching descriptor of known −1 PRF signals was used in conjunction with RNAMotif [12] to identify the nucleotide sequence corresponding to the frameshift signal's slippery site. Second, Pknots [13] was employed to “fold” the sequence immediately downstream (3′) to the slippery site and to produce a predicted MFE value in kilocalories per mole for the sequence. The statistical significance of the predicted MFE value of the three-stemmed RNA pseudoknot was tested by generating 500 randomly shuffled sequences derived from the native sequence, refolding each of these, and calculating their MFE values using Pknots. This resulted in a normal distribution of MFE values, against which the native sequence could be compared and z-scores calculated. FASTA3 v3.4 [40] was used to initially identify sequences homologous to the SARS −1 PRF signal based on primary sequence similarity. The search space included 1,724 viral genome sequences downloaded using the National Center for Biotechnology Information's Entrez Taxonomy Browser [41]. The resulting pairwise alignments produced by FASTA3 were used to produce a multiple-sequence alignment using ClustalW v1.82 [42]. An unrooted phylogenetic tree was created from this alignment and visualized using Tree View v1.6.6 [43].
Strains, genetic methods, and programmed ribosomal frameshifting assays
Escherichia coli strain DH5α was used to amplify plasmids, and E. coli transformations were performed using the high-efficiency transformation method of Inoue et al. [44]. YPAD and a synthetic complete medium (H−) were used as described previously [45]. Yeast strain JD932 (MATa ade2–1 trp1–1 ura3–1 leu2–3,112 his3–11,15 can1–100) and the JD1228/JD1229 isogenic pairs in which the disrupted RPL3/TCM1 allele is complemented with pRPL3 or pmak8–1 (MATα ura3–52 lys2–801 trp1δ leu2= his3 RPL3::HIS3) [21] were used for in vivo measurements of −1 PRF. Yeast cells were transformed using the alkali cation method [46]. Dual luciferase assays for programmed ribosomal frameshifting in yeast were performed as previously described [19]. African green monkey Vero cells were cultured in DMEM with L-glutamine (BioWhittaker, Walkersville, Maine, United States) and 10% FBS at 37 °C in 5% CO2. Cells cultured without antibiotics were transformed with plasmid DNA, using Amaxa (Cologne, Germany) Nucleofector solution according to the manufacturer's instructions. Dual luciferase assays were performed the following day, using extracts from cells lysed with the Passive Lysis Buffer (Dual-Luciferase Reporter System, Promega, Fitchburg, Wisconsin, United States). Wheat germ and rabbit reticulocyte lysates from Ambion (Austin, Texas, United States) were used to monitor frameshifting in vitro, using synthetic mRNA transcripts (Ambion mMESSAGE mMACHINE transcription kit), generated with T7 polymerase either from plasmids that had been digested with SspI, Proteinase K treated, phenol/chloroform and chloroform extracted, and ammonium acetate precipitated, or from PCR amplicons encompassing the dual luciferase reporter cassettes. All assays were repeated until the data were normally distributed, enabling statistical analyses both within and between experiments [25]. At least three readings derived from lysates derived from a minimum of three different transfection plates were used.
Oligonucleotides, plasmid construction, and mutagenesis.
Oligonucleotides were synthesized and purified by IDT (Coralville, Iowa, United States). These are listed in Table 1. The SARS-sense and SARS-antisense oligonucleotides were annealed, gel purified, and ligated into BamHI- and SacI-digested p2luc [18], generating plasmid pJD435. The Renilla and firefly bicistronic elements were amplified by PCR using previously described primers [19], SpeI- and XhoI-digested, and cloned into p416ADH [47]. One additional base was introduced after the BamHI restriction site, using the Stratagene (La Jolla, California, United States) QuikChange XL kit to correct the reading frame. Sequence analysis revealed an additional point mutation in the firefly luciferase gene that was reverted by oligonucleotide site-directed mutagenesis. The resulting plasmid, pJD465, constituted the wild-type SARS-CoV −1 PRF yeast assay plasmid. A zero-frame control plasmid, pJD474, was constructed by adding one cytosine residue upstream of the BamHI restriction site of pJD465. Additional constructs with various mutations in the pseudoknot were made; the 5′ portion of stem 3 was changed from
GCGGCACAG to
CGCCGAGAC (pJD467, also known as S3A), and this was the template for mutagenesis to make the complementary mutation in the 3′ half of stem 3, CUGAUGUCGU to GUCU
ACGGCG (pJD479, S3B). The control construct with just the changes in the 3′ portion of stem 3 was made from pJD465 (pJD567, S3A′). pJD465 was also used as the template for mutagenesis to move the A13456 residue out of stem 3 and into loop 2 (pJD492, S3C) and to make the change A13456C (pJD544, S3C′), while pJD467 was used to eliminate stem 3 entirely (pJD469, ΔS3). Stem 2 was also subjected to mutagenesis: the 5′ portion of stem 2 was changed from
GCCCG to
CGGGC (pJD466, S2A), and this in turn was the template for mutagenesis to make the complementary sequence in the 3′ half of stem 2,
CAGGGC to
GACCCG (pJD480, S2B). pJD465 was used as the template to create a construct in which the bulged A13467 residue in stem 2 was eliminated by moving it 6 nt downstream (pJD491, S2C) or replaced by cytosine (pJD542, S2C′).
Table 1 Oligonucleotides Used in This Study
An additional set of plasmids was constructed from the parental plasmids described above that lacked the yeast-specific markers but contained the SV40 early promoter, T7 promoter, and SV40 late poly (A) signal. These were used for programmed ribosomal frameshifting analyses in epithelial cells, wheat germ, and rabbit reticulocyte lysates. The BamHI and EcoRI fragment from pJD465 was purified and ligated into BamHI- and EcoRI-digested p2luc [18] to generate the test plasmid pJD502. A zero-frame control plasmid (pJD464) was constructed by cloning the BamHI/EcoRI fragment from pJD465 into p2luci. Similarly, BamHI and EcoRI fragments from the yeast plasmids described above were cloned into p2luc to generate a complete plasmid set for analyses of –1 PRF in epithelial cells (pJD503/S2A, pJD538/S2A′, pJD541/S2B′, pJD504/S2C, pJD537/S2C′, pJD487/S3A, pJD536/S3A′, pJD488/S3B, pJD506/S3C, pJD539/S3C′, and pJD490/ΔS3). An additional construct (A13465G) was made to control for the change at this position from adenine to guanine that prevents the creation of a termination codon in S2B′ constructs, but is not involved in stem 2 basepairing (pJD540 for Vero cells and pJD545 as the yeast plasmid).
Nuclease analysis.
The SP6SARS and revSARS oligonucleotides were used to generate a PCR amplicon from which an RNA transcript was made using the Ambion MEGAscript SP6 kit. The RNA was treated with calf intestinal phosphatase and 5′ end labeled with [γ-32P]ATP, using T4 polynucleotide kinase. The labeled RNA was gel purified and then eluted with 0.5 M NH4Ac, 1 mM EDTA, 0.1% SDS. Nuclease treatment with RNase A (1.0–0.01 ng), RNase T1 (1.0–0.01 U), and RNase V1 (0.1–0.001 U) from Ambion was performed according to the manufacturer's instructions for 15 min at room temperature. Digested RNA was electrophoresed through a 10% polyacrylamide gel and analyzed using a Storm PhosphoImager (Sunnyvale, California, United States).
Preparation of RNA samples for NMR
A DNA construct (residues 13405–13472 of the SARS-CoV genome) was generated by PCR from pJD465 containing the wild-type SARS-CoV frameshift pseudoknot sequence. Two oligodeoxynucleotides (Invitrogen, Carlsbad, California, United States) were designed with a 5′ primer, including a T7 promoter sequence. The resulting PCR product was cloned into a pUC18 plasmid. To prepare milligram quantities of the SARS-CoV frameshift pseudoknot (residues 13405–13472), 7.5- to 20-ml in vitro transcription reactions with phage T7 polymerase from a linearized plasmid template were performed [48]. Unlabeled NTPs were purchased from Sigma Pharmaceuticals (South Croydon, United Kingdom), and labeled NTPs were purified from Methylophilus methylotropus (
ATCC 53528, American Type Culture Collection, Manassas, Virginia, United States) bacteria grown on labeled medium with 15N-ammonium sulfate and 13C-methanol [49]. After 4–5 h incubation at 37 °C, the reaction was spun down to remove traces of precipitated pyrophosphate. RNA transcripts were purified by anion-exchange FPLC with two HiTrap Q columns (Amersham Pharmacia, Piscataway, New Jersey, United States) equilibrated in 50 mM Tris (pH 8) at room temperature. The target RNA sample was eluted with an increasing sodium chloride gradient. Pure fractions were concentrated using a CentriPrep YM10 (Millipore, Billerica, Massachusetts, United States) concentrator, passed through a NAP25 column (Amersham Pharmacia) equilibrated with NMR buffer (20 mM potassium phosphate [pH 6.5], 200 mM potassium chloride, 0.5 mM EDTA [ethylene diamine tetraacetic acid disodium salt], 0.02% sodium azide, 5% deuterium oxide), and concentrated to 0.2–2 mM, using a CentriPrep YM10 concentrator (Millipore). The identity of the RNA product was verified by mass spectroscopy, as well as agarose and TBE-Urea-PAGE (Bio-Rad, Hercules, California, United States) gels.
NMR spectroscopy
All NMR spectra were recorded at 5 °C, 15 °C, and 25 °C on a Bruker Avance 900 MHz spectrometer (Rheinstetten, Germany) equipped with a standard 5-mm triple axis pulsed field gradient 1H/13C/15N probehead optimized for proton detection. NMR experiments were performed on samples of 500-μl volume containing 0.2–2 mM SARS-CoV frameshift pseudoknot RNA. Data were processed using NMRPipe [50] and analyzed using NMRVIEW [51]. One-dimensional imino proton spectra were acquired using a jump-return echo sequence. The observable iminos in aqueous solution are diagnostic for hydrogen-bonded guanine and uracil bases, which are protected from exchange with the solvent. Imino resonances were assigned sequence-specificity from water flip-back, WATERGATE 2D nuclear Overhauser effect spectroscopy (NOESY) [52] spectra (τmix = 200 ms), and a jump-return [53] 1H,15N-heteronuclear multiple quantum correlation (HMQC) [54]. Elucidation of basepairing and secondary structure was verified from scalar 2h
J(N,N) couplings through hydrogen bonds in the quantitative J(N,N) HNN correlation spectroscopy (COSY) data [55,56].
Supporting Information
Dataset S1 Molecular Genetic Analyses of Stems 2 and 3
The first page provides a summary of the final statistics for the frameshifting experiments shown in Figure 6. Subsequent pages show the raw data and subsequent analyses for all of the different constructs following the methodologies, as previously described [25].
(178 KB XLS).
Click here for additional data file.
Accession Numbers
The GenBank (http://www.ncbi.nlm.nih.gov/Genbank/) accession numbers for the sequences discussed in this paper are AIBV (NC_001451), BCoV (NC_003045), HCoV-229E (NC_002645) , HCoV-HKU1 (NC_006577), HCoV-NL63 (NC_005831), HCoV-OC43 (NC_005147), MHV (NC_001846), PEDV (NC_003436), SARS-CoV (NC_004718), and TGV (NC_002306).
We want to thank Drs. Wenxia Song, David Mosser, and Deborah Taylor and the members of their laboratories for their generous help with cell culture. This work was supported by a grant to JDD from the National Institutes of Health (GM58859).
Competing interests. The authors have declared that no competing interests exist.
Author contributions. EPP, GCPA, MH, and JDD conceived and designed the experiments. EPP and GCPA performed the experiments. EPP, GCPA, JLJ, MH, and JDD analyzed the data. JLJ and BM contributed reagents/materials/analysis tools. MH and JDD wrote the paper.
Citation: Plant EP, Pérez-Alvarado GC, Jacobs JL, Mukhopadhyay B, Hennig M, et al. (2005) A three-stemmed mRNA pseudoknot in the SARS coronavirus frameshift signal. PLoS Biol 3(6): e172.
Abbreviations
−1 PRFprogrammed −1 ribosomal frameshift
2Dtwo-dimensional
MFEminimum free energy
NMRnuclear magnetic resonance
SARSsevere acute respiratory syndrome
==== Refs
References
Lai MM SARS virus: The beginning of the unraveling of a new coronavirus J Biomed Sci 2003 10 664 675 14631105
Marra MA Jones SJ Astell CR Holt RA Brooks-Wilson A The genome sequence of the SARS-associated coronavirus Science 2003 300 1399 1404 12730501
Rota PA Oberste MS Monroe SS Nix WA Campagnoli R Characterization of a novel coronavirus associated with severe acute respiratory syndrome Science 2003 300 1394 1399 12730500
Brierley I Ribosomal frameshifting on viral RNAs J Gen Virol 1995 76 1885 1892 7636469
Farabaugh PJ Programmed translational frameshifting Microbiol Rev 1996 60 103 134 8852897
Gesteland RF Atkins JF Recoding: Dynamic reprogramming of translation Annu Rev Biochem 1996 65 741 768 8811194
Dinman JD Ruiz-Echevarria MJ Peltz SW Translating old drugs into new treatments: Identifying compounds that modulate programmed -1 ribosomal frameshifting and function as potential antiviral agents Trends Biotechnol 1998 16 190 196 9586242
Thiel V Ivanov KA Putics A Hertzig T Schelle B Mechanisms and enzymes involved in SARS coronavirus genome expression J Gen Virol 2003 84 2305 2315 12917450
Baranov PV Henderson CM Anderson CB Gesteland RF Atkins JF Programmed ribosomal frameshifting in decoding the SARS-CoV genome Virology 2005 332 498 510 15680415
Ramos FD Carrasco M Doyle T Brierley I Programmed -1 ribosomal frameshifting in the SARS coronavirus Biochem Soc Trans 2004 32 1081 1083 15506971
Plant EP Jacobs KLM Harger JW Meskauskas A Jacobs JL The 9-angstrom solution: How mRNA pseudoknots promote efficient programmed -1 ribosomal frameshifting RNA 2003 9 168 174 12554858
Macke TJ Ecker DJ Gutell RR Gautheret D Case DA RNAMotif, an RNA secondary structure definition and search algorithm Nucleic Acids Res 2001 29 4724 4735 11713323
Rivas E Eddy SR A dynamic programming algorithm for RNA structure prediction including pseudoknots J Mol Biol 1999 285 2053 2068 9925784
Baril M Dulude D Steinberg SV Brakier-Gingras L The frameshift stimulatory signal of human immunodeficiency virus type 1 group O is a pseudoknot J Mol Biol 2003 331 571 583 12899829
Baranov PV Gurvich OL Hammer AW Gesteland RF Atkins JF Recode 2003 Nucleic Acids Res 2003 31 87 89 12519954
Herold J Siddell SG An ‘elaborated' pseudoknot is required for high frequency frameshifting during translation of HCV 229E polymerase mRNA Nucleic Acids Res 1993 21 5838 5842 8290341
Eleouet JF Rasschaert D Lambert P Levy L Vende P Complete sequence (20 kilobases) of the polyprotein-encoding gene 1 of transmissible gastroenteritis virus Virology 1995 206 817 822 7856095
Grentzmann G Ingram JA Kelly PJ Gesteland RF Atkins JF A dual-luciferase reporter system for studying recoding signals RNA 1998 4 479 486 9630253
Harger JW Dinman JD An in vivo dual-luciferase assay system for studying translational recoding in the yeast Saccharomyces cerevisiae
RNA 2003 9 1019 1024 12869712
Dinman JD Ruiz-Echevarria MJ Czaplinski K Peltz SW Peptidyl transferase inhibitors have antiviral properties by altering programmed -1 ribosomal frameshifting efficiencies: Development of model systems Proc Natl Acad Sci U S A 1997 94 6606 6611 9192612
Meskauskas A Harger JW Jacobs KLM Dinman JD Decreased peptidyltransferase activity correlates with increased programmed -1 ribosomal frameshifting and viral maintenance defects in the yeast Saccharomyces cerevisiae
RNA 2003 9 982 992 12869709
Peltz SW Hammell AB Cui Y Yasenchak J Puljanowski L Ribosomal protein L3 mutants alter translational fidelity and promote rapid loss of the yeast killer virus Mol Cell Biol 1999 19 384 391 9858562
Barry JK Miller WA A -1 ribosomal frameshift element that requires base pairing across four kilobases suggests a mechanism of regulating ribosome and replicase traffic on a viral RNA Proc Natl Acad Sci U S A 2002 99 11133 11138 12149516
Dinman JD Icho T Wickner RB A -1 ribosomal frameshift in a double-stranded RNA virus forms a Gag-pol fusion protein Proc Natl Acad Sci U S A 1991 88 174 178 1986362
Jacobs JL Dinman JD Systematic analysis of bicistronic reporter assay data Nucleic Acids Res 2004 32 e160 e170 15561995
Ferre-D'Amare AR Zhou K Doudna JA Crystal structure of a hepatitis delta virus ribozyme Nature 1998 395 567 574 9783582
Brierley IA Rolley NJ Jenner AJ Inglis SC Mutational analysis of the RNA pseudoknot component of a coronavirus ribosomal frameshifting signal J Mol Biol 1991 220 889 902 1880803
Jacks T Power MD Masiarz FR Luciw PA Barr PJ Characterization of ribosomal frameshifting in HIV-1 gag-pol expression Nature 1988 331 280 283 2447506
Puglisi JD Wyatt JR Tinoco I A pseudoknotted RNA oligonucleotide Nature 1988 331 283 286 3336440
Brierley IA Dingard P Inglis SC Characterization of an efficient coronavirus ribosomal frameshifting signal: Requirement for an RNA pseudoknot Cell 1989 57 537 547 2720781
Nissen P Ippolito JA Ban N Moore PB Steitz TA RNA tertiary interactions in the large ribosomal subunit: The A-minor motif Proc Natl Acad Sci U S A 2001 98 4899 4903 11296253
Kim YG Su L Maas S O'Neill A Rich A Specific mutations in a viral RNA pseudoknot drastically change ribosomal frameshifting efficiency Proc Natl Acad Sci U S A 1999 96 14234 14239 10588689
Nixon PL Cornish PV Suram SV Giedroc DP Thermodynamic analysis of conserved loop-stem interactions in P1-P2 frameshifting RNA pseudoknots from plant Luteoviridae Biochemistry 2002 41 10665 10674 12186552
Sarkhel S Rich A Egli M Water-nucleobase “stacking”: H-pi and lone pair-pi interactions in the atomic resolution crystal structure of an RNA pseudoknot J Am Chem Soc 2003 125 8998 8999 15369340
Giedroc DP Cornish PV Hennig M Detection of scalar couplings involving 2′-hydroxyl protons across hydrogen bonds in a frameshifting mRNA pseudoknot J Am Chem Soc 2003 125 4676 4677 12696863
Nixon PL Rangan A Kim YG Rich A Hoffman DW Solution structure of a luteoviral P1-P2 frameshifting mRNA pseudoknot J Mol Biol 2002 322 621 633 12225754
Napthine S Liphardt J Bloys A Routledge S Brierley I The role of RNA pseudoknot stem 1 length in the promotion of efficient -1 ribosomal frameshifting J Mol Biol 1999 288 305 320 10329144
Goebel SJ Hsue B Dombrowski TF Masters PS Characterization of the RNA components of a putative molecular switch in the 3′ untranslated region of the murine coronavirus genome J Virol 2004 78 669 682 14694098
Goebel SJ Taylor J Masters PS The 3′ cis-acting genomic replication element of the severe acute respiratory syndrome coronavirus can function in the murine coronavirus genome J Virol 2004 78 7846 7851 15220462
Pearson WR Flexible sequence similarity searching with the FASTA3 program package Meth Mol Biol 2000 132 185 219
Wheeler DL Chappey C Lash AE Leipe DD Madden TL Database resources of the National Center for Biotechnology Information Nucleic Acids Res 2000 28 10 14 10592169
Thompson JD Higgins DG Gibson TJ CLUSTAL W: Improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice Nucleic Acids Res 1994 22 4673 4680 7984417
Page RD TreeView: An application to display phylogenetic trees on personal computers Comput Appl Biosci 1996 12 357 358 8902363
Inoue H Nojima H Okayama H High efficiency transformation of Escherichia coli with plasmids Gene 1990 96 23 28 2265755
Dinman JD Wickner RB Translational maintenance of frame: Mutants of Saccharomyces cerevisiae with altered -1 ribosomal frameshifting efficiencies Genetics 1994 136 75 86 8138178
Ito H Fukuda Y Murata K Kimura A Transformation of intact yeast cells treated with alkali cations J Bact 1983 153 163 168 6336730
Mumberg D Muller R Funk M Yeast vectors for the controlled expression of heterologous proteins in different genetic backgrounds Gene 1995 156 119 122 7737504
Milligan JF Groebe DR Witherell GW Uhlenbeck OC Oligoribonucleotide synthesis using T7 RNA polymerase and synthetic DNA templates Nucleic Acids Res 1987 15 8783 8798 3684574
Batey RT Battiste JL Williamson JR Preparation of isotopically enriched RNAs for heteronuclear NMR Meth Enzymol 1995 261 300 322 8569501
Delaglio F Grzesiek S Vuister GW Zhu G Pfeifer J NMRPipe: A multidimensional spectral processing system based on UNIX pipes J Biomol NMR 1995 6 277 293 8520220
Johnson BA Blevins RA NMRView: A computer program for the visualization and analysis of NMR data J Biomol NMR 1994 4 603 614 22911360
Lippens G Dhalluin C Wieruszeski JM Use of a water flip-back pulse in the homonuclear NOESY experiment J. Biomol. NMR 1995 5 327 331 22911506
Sklenar V Bax A Spin-echo water suppression for the generation of pure-phase two-dimensional NMR-spectra J Magn Reson 1987 74 469 479
Bax A Griffey RH Hawkins BL Correlation of proton and N-15 chemical-shifts by multiple quantum NMR J Magn Reson 1983 55 301 315
Pervushin K Ono A Fernandez C Szyperski T Kainosho M NMR scaler couplings across Watson-Crick base pair hydrogen bonds in DNA observed by transverse relaxation optimized spectroscopy Proc Natl Acad Sci U S A 1998 95 14147 14151 9826668
Dingley AJ Grzesiek S Direct observation of hydrogen bonds in nucleic acid base pairs by internucleotide (2)J(NN) couplings J Am Chem Soc 1998 120 8293 8297
| 15884978 | PMC1110908 | CC BY | 2021-01-05 08:21:24 | no | PLoS Biol. 2005 Jun 17; 3(6):e172 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030172 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1588209310.1371/journal.pbio.0030186Research ArticleDevelopmentGenetics/Genomics/Gene TherapyNeuroscienceMus (Mouse)Neuronal Migration and Ventral Subtype Identity in the Telencephalon Depend on SOX1 Neuronal Identity and Migration in the ForebrainEkonomou Antigoni
1
¤aKazanis Ilias
1
Malas Stavros
1
¤bWood Heather
1
¤cAlifragis Pavlos
1
Denaxa Myrto
2
Karagogeos Domna
2
Constanti Andrew
3
Lovell-Badge Robin
4
Episkopou Vasso vasso. [email protected]
1
1Mammalian Neurogenesis Group, MRC Clinical Sciences CentreImperial College School of Medicine, Hammersmith Hospital Campus, LondonUnited Kingdom2Medical School and Institute of Molecular Biology and Biotechnology, University of CreteHeraklionGreece3Department of Pharmacology, The School of PharmacyLondonUnited Kingdom4Division of Developmental Genetics, National Institute of Medical ResearchLondonUnited KingdomJessell Thomas M. Academic EditorColumbia UniversityUnited States of America6 2005 17 5 2005 17 5 2005 3 6 e1865 1 2005 24 3 2005 Copyright: © 2005 Ekonomou 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.
Migration and Fate Specification in the Ventral Striatum
Little is known about the molecular mechanisms and intrinsic factors that are responsible for the emergence of neuronal subtype identity. Several transcription factors that are expressed mainly in precursors of the ventral telencephalon have been shown to control neuronal specification, but it has been unclear whether subtype identity is also specified in these precursors, or if this happens in postmitotic neurons, and whether it involves the same or different factors. SOX1, an HMG box transcription factor, is expressed widely in neural precursors along with the two other SOXB1 subfamily members, SOX2 and SOX3, and all three have been implicated in neurogenesis. SOX1 is also uniquely expressed at a high level in the majority of telencephalic neurons that constitute the ventral striatum (VS). These neurons are missing in Sox1-null mutant mice. In the present study, we have addressed the requirement for SOX1 at a cellular level, revealing both the nature and timing of the defect. By generating a novel Sox1-null allele expressing β-galactosidase, we found that the VS precursors and their early neuronal differentiation are unaffected in the absence of SOX1, but the prospective neurons fail to migrate to their appropriate position. Furthermore, the migration of non-Sox1-expressing VS neurons (such as those expressing Pax6) was also affected in the absence of SOX1, suggesting that Sox1-expressing neurons play a role in structuring the area of the VS. To test whether SOX1 is required in postmitotic cells for the emergence of VS neuronal identity, we generated mice in which Sox1 expression was directed to all ventral telencephalic precursors, but to only a very few VS neurons. These mice again lacked most of the VS, indicating that SOX1 expression in precursors is not sufficient for VS development. Conversely, the few neurons in which Sox1 expression was maintained were able to migrate to the VS. In conclusion, Sox1 expression in precursors is not sufficient for VS neuronal identity and migration, but this is accomplished in postmitotic cells, which require the continued presence of SOX1. Our data also suggest that other SOXB1 members showing expression in specific neuronal populations are likely to play continuous roles from the establishment of precursors to their final differentiation.
The mechanisms whereby neurons establish their final identity are largely unknown. This study shows that the transcription factor SOX1 plays an important part in this process.
==== Body
Introduction
The telencephalon is subdivided into dorsal (pallial) and ventral (subpallial) territories, which give rise to the cerebral cortex and the underlying basal ganglia, respectively. The embryonic subpallium consists of large protrusions—the ganglionic eminences. Several distinct types of neurons originate in the ganglionic eminences, and some migrate as far as the olfactory bulb, hippocampus, and neocortex [1–3], while others contribute more locally. The majority of neurons of the lateral ganglionic eminence (LGE) form the dorsal and ventral striatum (VS). The VS includes the caudate, putamen, nucleus accumbens, and olfactory tubercle (OT), which control various aspects of motor, cognitive, and emotional functions [4,5]. Little is known about the molecular mechanisms that control the emergence of various groups of neurons with distinct identities in this region.
Gene-expression studies and loss-of-function mutations in homeodomain transcription factors such as PAX6 [6,7] and GSH2/1 [8–13] confirm fate-mapping findings [14–16] that the majority of the VS neurons are specified within the progenitor domain of the LGE. The proneural basic helix-loop-helix (bHLH) factor MASH1 also marks the precursors of early-born neurons in the LGE progenitor domain, and its loss in the mouse leads to a deficit of both precursors and neurons of the telencephalon, including loss of VS neurons [17,18]. Therefore, GSH2 and MASH1 control VS precursor patterning and specification, but as they are not expressed in postmitotic cells it remained unknown to what extent they are involved in the emergence of neuronal subtypes in the ventral telencephalon, and whether different transcription factors with neuron-specific expression are required.
The SOX proteins constitute a family of transcription factors [19,20] that regulate transcription through their ability to bind to specific DNA sequences via their HMG box domains [21–24]. There are 20 Sox genes in mammals, and at least half are expressed in the developing nervous system [20,24]; however, their role in neural development is poorly understood. SOX1, SOX2, and SOX3 constitute the SOXB1 subfamily and share more than 95% identity within their HMG boxes and significant homology outside [25,26]. All three proteins are expressed in the neuroepithelium throughout central nervous system (CNS) development [25,27], and as they tend to be down-regulated upon neural differentiation they have been used as markers for neural stem cells and precursors [28,29]. Several studies suggest that SOXB1 factors function in stem cells and precursors to maintain broad developmental potential [30] and neural stem cell identity [30–32] by counteracting neurogenesis. Contradictory evidence, however, suggests that SOX1 promotes neurogenesis and cell cycle exit [33]. However, mice that are null for Sox1 [34] or Sox3 [35], or mice with one Sox2 allele deleted and the other hypomorphic [36], exhibit phenotypes associated with the loss of or functional deficit of only specific neuronal populations. As these SOXB1 factors are expressed in both precursors and neurons that are affected in these mutant mice, it was not known whether their function is required in precursors, postmitotic cells, or both.
We have previously shown that SOX1 is essential for the terminal differentiation of lens fibers and the activation of γ-crystallins [37], and for the development of VS neurons, the lack of which is associated with epilepsy [34]. Here, we show that absence of SOX1 has no effect on the generation, proliferation, and patterning of neuronal precursors. This is probably due to functional compensation by SOX2 and SOX3, which are co-expressed with SOX1 in precursors. Moreover, mice lacking only the neuron-specific expression of Sox1 in the ventral telencephalon still fail to develop VS neurons, revealing its requirement within these neurons. Consistent with this, maintenance of Sox1 expression in neurons of the ventral telencephalon is sufficient to direct them to the VS, confirming the adequacy of SOX1 function in postmitotic cells for their migration and identity. Therefore, VS-specific neuronal migration and subtype identity most likely is initiated in precursors but is completed in postmitotic cells by transcription factors such as SOX1.
Results
SOX1 Is Essential for the Histogenesis of the VS
To generate a detailed map of Sox1 expression in the developing and adult brain of mice, and to perform comparative studies between homozygotes and heterozygotes, we generated a novel targeted allele referred to as Sox1
βgeo. This contains an insertion of β-galactosidase-neo (βgeo) fusion protein in-frame with the SOX1 open reading frame (Figure 1A). Mice homozygous for Sox1
βgeo are null for SOX1 and exhibit the same phenotype as the previously described mice, which carry a deletion of the SOX1 coding region (Sox1
M1) [34,37], namely, lens defects and epileptic seizures. Staining for β-galactosidase activity (X-gal staining) in Sox1
βgeo/+ heterozygous embryos matches that for the wild-type allele as revealed by whole-mount in situ hybridization and SOX1 antibody staining (Figure 1B–1E).
Figure 1 The Mouse Sox1
βgeo Allele Reveals the Requirement of SOX1 in the Development of VS Neurons
(A) Strategy for targeting of the Sox1 locus by insertion of βgeo. Restriction enzymes: RV, EcoRV; K, KpnI; E, EcoRI; S, SpeI; B, BamHI. Yellow boxes indicate βgeo, green, SOX1 exon, and blue lines indicate fragments appearing in Southern blots of EcoR1-digested genomic DNA, hybridized with the external probe, which is shown with red lines.
(B–E) X-gal and SOX1 antibody staining of Sox1
βgeo/+. Comparison of Sox1
βgeo expression visualized by X-gal staining (B and D) and the endogenous wild-type Sox1 gene visualized by whole-mount in situ (C) and SOX1 antibody staining (E). (B and C) show E9-stage embryos and (D and E) show coronal sections of newborn ventral telencephalon.
(F–M) 100-μm coronal sections (Vibratome) were stained with X-gal to identify cells with Sox1 promoter activity. (F–I) show Sox1
βgeo/+ forebrain sections from E13 to birth (P0) showing normal migration of Sox1-expressing cells from the VZ to the site of the OT, including striatal bridges. (J–M) show sections of Sox1
βgeo/M1 forebrain, showing absence of X-gal staining in the OT and the striatal bridges. Red arrowheads show the anterior commissure.
Scale bar = 500 μm for (B) and (C) and 300 μm for (D–M).
To elucidate the role of SOX1 in the formation of the VS, we compared the expression pattern of Sox1 in heterozygous (Figure 1F–1I) and homozygous (Figure 1J–1M) brains from embryonic day 14 (E14) to postnatal day 0 (P0) using X-gal staining. Throughout much of the CNS, X-gal staining in the Sox1
βgeo/βgeo homozygotes is double the intensity observed in heterozygotes (data not shown). To perform comparative histological studies, we equalized the levels of X-gal staining in homozygous mice with those of heterozygous animals by generating homozygous mice that harbor two different Sox1-null alleles: βgeo (Sox1
βgeo) and the previously described M1-targeted allele (Sox1
M1) [37], which does not express β-galactosidase.
Our analysis shows (via X-gal staining) that in the developing forebrain, Sox1 is expressed throughout the ventricular zone (VZ) and subventricular zone (SVZ) and in neurons around the anterior commissure region, where the prospective nucleus accumbens forms (red arrowheads in Figure 1K and 1M), and in the striatal bridges that link this intermediate cluster of cells with the prospective OT region toward the pial surface. X-gal-positive neurons start populating the OT area as early as E14 and continue to accumulate at least until birth (Figure 1F–1I). In the Sox1
βgeo/M1-null mutants, the X-gal staining pattern of the VZ/SVZ is indistinguishable from that of the Sox1
βgeo/+ heterozygotes, and there is no obvious deficit of X-gal-positive cells around the anterior commissure. On the other hand, both the striatal bridges and the OT layers are absent in the Sox1-null brain at all developmental stages (compare Figure 1F–1I to 1J–1M). It is unlikely that neurons die en masse in this region, because an apoptosis assay did not reveal any evidence of increased cell death in the mutant (data not shown). In addition, X-gal staining is increased throughout the ventral telencephalon in the mutant postnatal brain (red arrowheads in Figure 2), suggesting that the Sox1-null cells are not correctly specified and contribute to other brain regions. Interestingly, although neurons that form the core of the nucleus accumbens express Sox1 highly, they form normally (Figure 2F and red arrowheads in Figure 1) and do not depend on SOX1 for their development. Therefore, SOX1 is required for histogenesis of the OT throughout its development.
Figure 2 Ectopic Distribution of Sox1-Null Neurons
X-gal staining of mouse forebrains at P16. (A and B) show intact forebrain viewed from the ventral surface, and (C–F) show 150-μm coronal Vibratome sections for Sox1
βgeo/+ mice (A, C, and E) and Sox1
βgeo/M1 mice (B, D, and E). Red arrows indicate the width of the OT. Red arrowheads indicate increased X-gal staining at more medial and posterior areas of the brain in (B), and in the striatum and septum in (D) and (F). White arrowheads indicate islands other than the medial islands of Calleja. an, accumbens nucleus; I, II, III, cell layers of the OT; ICjM, medial islands of Calleja; lot, lateral olfactory tract; lsn, lateral septal nucleus; ob, olfactory bulb; PC, olfactory (piriform) cortex; S, striatum; sb, striatal bridge Scale bar = 500 μm.
Normal Precursor Proliferation and Neurogenesis but Loss of OT Neuronal Differentiation in the Absence of SOX1
Studies with conflicting results suggest that SOX1 either, like SOX2 and SOX3, counteracts neurogenesis [32] or, unlike SOX2 and SOX3, promotes neurogenesis [33]. To examine whether the loss of SOX1 affects general neural differentiation in the area of the striatum, we used an anti-βIII-tubulin (TuJ1) antibody, which is a marker for immature neurons [38], at E13, a critical time of differentiation in the LGE. TuJ 1 immunocytochemistry did not reveal any obvious general differentiation problems in the Sox1 mutants (Figure 3A and 3B), suggesting that loss of SOX1 alone is not sufficient to compromise general neuron differentiation and maturity. The differentiation and distribution of specific mature neurons was examined in our previous study at adult stages with the expression of striatal markers such as preproenkephalin and Gad65/67 [34], and in our current study, at embryonic stages with additional markers such as Brn4 [39] (Figure 3C and 3D) and Robo [40] (Figure 3E and 3F). This analysis revealed a differentiation defect restricted in the region of the nucleus accumbens/OT, and not the rest of the striatum.
Figure 3 Normal Precursor Proliferation and Neurogenesis but Loss of OT Neuronal Differentiation in the Absence of SOX1
Coronal brain sections from the ventral telencephalon of wild-type (+/+) and Sox1-null (−/−) embryos. TuJ1 immunolabeling (A and B) at E13 shows no difference in early neuronal differentiation embryos; in situ hybridization at E16 for Brn4 (C and D) and Robo (E and F) shows absence of differentiation in the mutant at the prospective OT area. Red arrow in wild-type brain sections indicates OT. Telencephalic sections of wild-type (G, I, and K) and Sox1-null mutant (H, J, and L) embryonic brains were harvested 1 h after BrdU injection at E13 (G and H), E14 (I and J), and E15 (K and L) to detect actively dividing cells of the VZ/SVZ. Positive cells were visualized with anti-BrdU immunofluorescence (G–J) or with DAB staining (K and L). (K and L) show dorsal LGE area at high magnification. No differences were detected in the proliferation precursors at all stages examined, and no ectopic proliferation was observed in the mutant brains. Measurements and statistical analysis of BrdU-positive cells were performed on the DAB-stained sections, showing no significant differences (see Table S1). Scale bar = 300 μm (A and B), 300 μm (C–F), 500 μm (G–J), 500 μm (K and L).
As Sox1 expression is associated with dividing cells throughout the neuroepithelium, we examined whether a proliferation defect could partially account for the cell deficit in the ventral telencephalon. We used 5-bromo-2′-deoxyuridine (BrdU) to label all proliferating precursors in wild-type and Sox1-null embryos at E13–E15, and harvested their brains 1 h later. Most of the dividing cells were found in the VZ/SVZ, and there was no increase or ectopic proliferation (Figure 3G–3L; Table S1). Therefore, SOX1 is unlikely to be required for proliferation or the exit of precursors from the cell cycle in the LGE. Collectively, the above data show that SOX1 is not essential for the proliferation of precursors and general neuronal differentiation. However, it is required specifically for the differentiation and/or migration of VS neurons.
Early- and Late-Born OT Neurons Fail to Migrate in the Absence of SOX1
To investigate the possibility that neurons migrate in the Sox1-null OT regions, but are not visible because they do not express Sox1
βgeo and other differentiation markers, we exposed embryos to BrdU. This way, we permanently marked all proliferating precursors independently of Sox1 or other striatal-marker gene expression and followed them at later embryonic stages and after birth (Figure 4). Labeling the precursors at different embryonic days also provided information on the birthdates of the OT neurons, which were previously unknown for the mouse. Specifically, birth of ventral striatal neurons commences early at E13 and continues until birth (Figure 4A, 4C, 4E, and 4G; data not shown) and is consistent with data from the rat [41,42]. Furthermore, in wild-type embryos, BrdU exposure between E13 and E16 with examination of embryos either 72 h later (Figure 4A; data not shown) or after birth (Figure 4C, 4E, and 4G) showed that early-born neurons migrate more laterally than those born later.
Figure 4 Failure of Neurons to Migrate to the VS in the Absence of SOX1
This figure shows BrdU labeling of proliferating cells in the developing forebrain. Immunohistochemistry was performed on 5-μm coronal sections, cut at the level of the OT.
(A and B) Sections at E17, after BrdU injection at E14. White arrowheads in (A and B) indicate streams of migrating cells.
(C–H) Sections at P16, after BrdU injection at E13 (C and D), E14 (E and F), or E16 (G and H). The DAB reaction product (C–H) was viewed under dark-field illumination. “II” is layer II of the OT, and the red bracket indicates the olfactory cortex. Note E13-born neurons contribute laterally to the olfactory (piriform) cortex, and medially to the layer II of the OT and the striatal bridges (red arrow). E14-born neurons contribute to more medial VS structures than E15- and E16-born cells, which contribute almost exclusively to the medial islands of Calleja (red arrowheads).
Scale bar = 300 μm (A and B), 1 mm (C–H).
In the Sox1-null embryos, the presence and distribution of cells labeled with BrdU between E13 and E16 shows that the olfactory cortex is largely normal (bracket in Figure 4D), but in the VS area the number of labeled cells was found to be greatly reduced (Figure 4B, 4D, 4F, and 4H). Furthermore, in the Sox1-null postnatal brain the striatal mantle is more densely populated by BrdU-labeled cells (Figure 4D–4H), consistent with the general increase of X-gal staining observed throughout the striatum (see Figure 2). The above data suggest that in the absence of SOX1, early- and late-born neurons fail to migrate to the appropriate position to form the ventral areas of the striatum.
The Generation and Patterning of LGE Precursors Is Normal in the Absence of SOX1
It is known that the majority of VS neurons derive from precursors that are born in the LGE [10,15,43]. To investigate whether the defect is in the patterning of precursors, we examined the expression pattern of various transcription factors that mark LGE progenitors and are known to have a role in OT neuronal specification (Figures 5, S1, and S2).
Figure 5 Normal Generation and Patterning of LGE Precursors in the Absence of SOX1
(A and B) Immunocytochemistry and on coronal brain sections of dorsal and ventral telencephalic markers in wild-type (+/+) and Sox1-null (−/−) embryos. PAX6 and GSH2 immunocytochemistry in the dorsal LGE at E12 shows no difference at the expression boundary in the absence of SOX1; the arrows point at the stream of PAX6-positive cells emanating from the boundary.
Double immunostaining for SOX1/PAX6 in wild-type brain (C), and for β-galactosidase/PAX6 (D) in the Sox1
βgeo/− brain, at the VS area at E15. Note the presence of the PAX6-positive neurons in the area of the VS in the Sox1
βgeo/− brain.
(E and F) MASH1 immunocytochemistry in the LGE of wild-type (E) and Sox1-null brain (F), at E13. No changes are detected.
(G and H) The distribution of Dlx1-expressing cells, as detected by in situ hybridization, is similar in both wild-type and mutant brains.
Scale bar = 300 μm (A and B), 200 μm (C–F), 150 μm (G and H).
The homeodomain transcription factors PAX6 and GSH2 are expressed, respectively, in the pallial and subpallial precursor domains of the dorsal LGE. The boundary between them has been shown to be essential for the patterning of VS precursors [10,11]. Specifically, GSH2-null mice do not form early-born OT neurons, and the precursors of the dorsal LGE are lost as PAX6 expands ventrally into the LGE. However, loss of both PAX6 and GSH2 restores dorsoventral patterning and partially rescues OT formation [9–11]. Using double antibody immunohistochemistry in embryos at E12–E16, we found that loss of SOX1 has no effect on the expression of GSH2/PAX6 and the boundary in the dorsal LGE (Figures 5A, 5B, and S1). In addition, at this boundary, PAX6-expressing postmitotic cells form a stream (arrow in Figure 5A) that extends laterally to the VS (Figures 5A, 5C, S2E, and S2F) [6,7,44]. In the absence of SOX1, the stream of PAX6-positive postmitotic cells is normal (arrow in Figures 5B and S2F), but to characterize the PAX6- and SOX1-expressing neurons in the region of the OT, we used double antibody immunohistochemistry. Specifically, for the wild-type brain sections we used PAX6 and SOX1 antibodies, but to visualize the Sox1-expressing cells in the Sox1
βgeo/M1-null brain sections we used an antibody for β-galactosidase.
Our data showed that PAX6 and SOX1 proteins were co-expressed in progenitors (Figures S2E and S2F), but in postmitotic cells of the LGE this expression became mutually exclusive (Figure 5C and 5D). Pax6-expressing neurons were clustered laterally to those expressing Sox1, at the border between the OT and olfactory cortex (Figures 5C, S2A, and S2C). In Sox1-null mice, the postmitotic Pax6-expressing cells were distributed throughout the VS area (Figures 5D, S2B, and S2D), suggesting structural disorganization. It is unlikely that these ectopically localized Pax6-expressing neurons are mis-specified Sox1-null neurons, because they should be expressing both PAX6 and βgeo from the mutant Sox1
βgeo allele, but they do not (Figure 5D).
The LGE structure consists of neural progenitors with radial glial characteristics, having fibers that extend from the VZ to the pial surface. These cells also provide the substrate for the migration of neurons [45,46]. Staining with X-gal (see Figure 1) and β-galactosidase antibody in mice carrying the Sox1
βgeo allele allowed visualization of the cytoplasmic compartment of the SOX1-expressing progenitors, which have radial glial morphology (Figures S2F and S3). We examined the morphology of radial glia in the Sox1-null LGE using the RC2 antibody [47], but we did not find any difference from wild-type (Figure S4). Therefore, we conclude that the abnormal distribution of Pax6-expressing neurons in the LGE of the Sox1-null mice is unlikely to be caused by abnormal morphology of the radial glial fibers or substantial loss of radial glia-like precursors.
In the ventral telencephalon, the bHLH transcription factor MASH1 [18] and the homeodomain factor DLX1 mark LGE precursors [8]. Ablation of MASH1 in mice causes loss of specific subpopulations of precursors and striatal neurons that contribute to the OT and nucleus accumbens [18]. In addition, Gsh2-null mice, which also fail to develop the OT, exhibit reduced Dlx1 expression in LGE precursors [9]. We therefore examined the expression of these two genes in Sox1-null embryos, but found no difference (Figure 5E–5H), indicating that there is no deficit of early or late LGE precursors in the absence of SOX1. Collectively, the above data show that SOX1 is not required for patterning, generation, and maintenance of LGE precursors.
Sox1 Expression from the Endogenous Sox2 Promoter Can Be Tolerated In Vivo
It has been shown that all three SoxB1 genes are expressed [25,27] and share similar functions in neural precursors [31,32]. However, in the postmitotic cells of the mantle and the VS area, antibody staining for each of the genes at E15 indicated that SOX2- and SOX3-positive neurons represent a very small population compared to that expressing SOX1 (Figure 6A–6C). Therefore, it is likely that the other SoxB1 genes compensate for the loss of Sox1 in precursors, whereas they cannot do so in the LGE postmitotic cells. Nevertheless, it remained unknown whether SOX1 functions solely in precursors for VS fate specification or in postmitotic cells for maintaining this fate and the emergence of specific subtype identity and migration. To address this, we generated mice that express Sox1 mainly in precursors and not in LGE neurons. We took advantage of the fact that Sox2 is co-expressed with Sox1 in precursors but it is down-regulated in LGE neurons, and generated mice that express Sox1 from the endogenous Sox2 promoter. We confirmed the overlap of Sox1 and Sox2 expression in the VZ/SVZ of the LGE by staining serial coronal telencephalic sections with antibodies for each of the two genes and counter-staining with the nuclear stain TOTO at E14 (Figure S5), and by performing double anti-SOX1 and -SOX2 immunohistochemistry at E13 (Figure 6D–6L).
Figure 6 SOX2 and SOX3 Down-Regulation in LGE Neurons and SOX1/SOX2 Co-Expression in LGE Precursors
Immunofluorescence of coronal sections at LGE levels in (A–C) E15- and (D–L) E13-stage wild-type embryos visualized on a confocal microscope: antibody staining for (A, D, G, and J) SOX1 (red), (B, E, H, and K) SOX2 (green), (C) SOX3 (green), (D–L) double SOX1 (red) and SOX2 (green), and (F, I, and L) merged. In the OT area and the LGE mantle, there are more neurons expressing SOX1 (A and J) than SOX2 (B and K) and SOX3 (C). Note the extensive co-expression of the SOX1 and SOX2 in precursors (D–I). (G–I) are higher magnifications of the areas within the rectangles. Scale bar = 300 μm.
The replacement of the SOX2 open reading frame with that of SOX1 was achieved by targeting the Sox2 allele (Figure 7A). The new allele, Sox2
R, was engineered to express not only Sox1 but also βgeo via an internal ribosomal entry site (IRES). Furthermore, the coding region of SOX1 in the Sox2
R allele was flanked by LoxP sites, which can be deleted using Cre-mediated recombination [48]. In this way, we generated a mouse line carrying another allele, termed Sox2
βgeo2 (but referred to hereafter as Sox2
βgeo), which expresses only the βgeo reporter gene from the Sox2 promoter (Figure 7A) and not SOX1. Like the Sox2
βgeo/+ heterozygotes, Sox2
R/+ mice were viable, fertile, and phenotypically normal, indicating that SOX1 over-expression in precursors, as well as ectopic expression in other locations where Sox2 is uniquely expressed, does not cause any obvious developmental abnormality.
Figure 7 The Sox2
R Allele Delivers SOX1 in Sox2-Specific Expression Sites
(A) Strategy for targeted replacement of the SOX2 coding region with that of SOX1 and IRES-βgeo. Restriction enzymes: S, SalI; E, EcoRI; Sm, SmaI; X, Xho. Green boxes indicate Sox1; black arrowheads indicate LoxP sites; yellow boxes indicate IRES βgeo; blue lines indicate fragments appearing in Southern blots of EcoR1-digested genomic DNA, hybridized with the external probe, which is shown with red lines. Black arrows show the locus after recombination, homologous and Cre-mediated where is indicated.
(B–E) SOX1 immunostaining of frontal sections from E10 embryos. (B and D) Sox2+/+ and (C and E) Sox2
R/+ showing the ectopic expression of SOX1 in the diencephalon (arrowheads) and the nasal pit (np) at E13.
X-gal staining of embryos showed that both targeted Sox2 alleles (Sox2
βgeo and Sox2
R) express βgeo in the CNS, but to verify that SOX1 protein was produced from the Sox2
R allele, we used SOX1 antibody staining. We found that SOX1 protein was ectopically present at sites where SOX2 normally shows unique expression—for example, in the floor plate of the diencephalon (arrowheads in Figure 7B and 7C) and in the sensory placodes (arrows in Figure 7D and 7E). The intensity of the immunostaining at ectopic sites was comparable to the staining in areas with expression from two wild-type Sox1 alleles (VZ/SVZ), indicating that the level of expression was similar to the wild-type allele. Therefore, the Sox2
R allele produces SOX1 ectopically in all neurons uniquely positive for SOX2 and increases the endogenous level of SOX1 in precursors and neurons that express both genes, without causing an obvious defect in mice. The fact that ectopic expression does not cause any obvious phenotype suggests either that the partner factors required for SOX1 target specificity [19] are absent in those cells uniquely expressing Sox2 or that the two proteins are interchangeable, sharing target genes. Further experiments are required to clarify this.
Sox1 Over-Expression in Precursors Does Not Increase VS/OT Neuronal Fate Specification
To investigate more subtle defects due to the over-expression of Sox1 in precursors of the Sox2
R/+ mice, we examined several litters (n > 10) of mice and visualized the migration of the Sox1/Sox2-positive neurons in the VS via X-gal staining. Newborn mice carrying Sox2
R/+ with two (Sox2
R/+, Sox1
+/+; Figure 8A) or one (Sox2
R/+, Sox1
M1/+; Figure 8B) Sox1 endogenous wild-type alleles were compared with those carrying Sox2
βgeo/+, Sox1
+/+ that do not express Sox1 ectopically (Figure 8C). All the above mice have only one wild-type Sox2 allele. We also compared the Sox1-βgeo and Sox2-βgeo neurons of the ventral telencephalon in Sox1
βgeo/+ (Figure 8C and 8D) and Sox2
R/+ (Figure 8A) mice, respectively. In this area of the brain, the Sox2-positive neurons are far fewer than those positive for Sox1. Therefore, for comparison purposes, we used thin tissue sections (80 μm) with short X-gal staining (3 h) for Sox1-βgeo, and thicker sections (100–150 μm) with a long staining period (48 h) for Sox2-βgeo. Consistent with the antibody staining data (see Figure 6B), Sox2-βgeo neurons contribute to the OT, indicating that they are a subset of the VS neuronal population. Therefore, the ectopic expression of Sox1 in LGE neurons is expected to be very limited. More importantly, the migration and the number of LGE neurons expressing βgeo via the Sox2 promoter were found to be the same regardless of the number of endogenous Sox1 alleles or the ectopic presence of Sox1 (compare Figure 8A, 8B, and 8C). To further investigate the differentiation of the VS neurons, we used the striatal-specific markers dopamine and cAMP-regulated phosphoprotein (DARPP-32) at postnatal stages and found them to be unaffected in Sox2
R/+ mice (Figure 8E and 8F). The data therefore indicate that the over-expression of SOX1 in precursors does not increase OT neuronal specification.
Figure 8 The Distribution of VS Neurons Is Unaffected in Mice Over-Expressing SOX1 from the Sox2
R Allele
(A–D), X-gal staining of coronal sections from the ventral telencephalon of P0 mice. Note that there is no difference in the distribution of Sox2-expressing OT neurons with SOX1 (A) or without SOX1 (C), and in Sox2
R/+ mice with two wild-type Sox1 alleles (A) or one (B). Comparison of the number and distribution of neurons expressing Sox2
βgeo in (C) and Sox1
βgeo in (D) shows overlapping expression. (A–C) show 150-μm sections, and (D) shows a 80-μm section.
(E and F) DARPP-32 immunostaining of coronal sections from the ventral telencephalon of P10 Sox2
R/+ and Sox1
βgeo/+ single heterozygous mice, showing no difference in the generation and migration of OT neurons. AC, anterior commissure. Scale bar = 100 μm.
Sox1 Expression in Precursors Cannot Rescue OT Neuron Development
To address directly whether SOX1 function is essential in precursors, we crossed Sox1
M1/+, Sox2
R/+ mice with Sox1
βgeo/+, Sox2
+/+ mice and examined whether offspring carrying Sox2
R/+ without any wild-type Sox1 alleles (Sox1
M1/βgeo) could develop OT. In these Sox1
R/+ embryos that carry no endogenous Sox1 functional allele (termed here HoHe), SOX1 is expected to be expressed only via the Sox2
R allele in precursors and become down-regulated in postmitotic LGE cells. However, βgeo expression from the endogenous Sox1 mutant allele marks precursors and OT prospective neurons. We followed the Sox1
M1/βgeo prospective OT neurons with X-gal to determine whether they were capable of contributing to the OT in HoHe embryos (Figure 9). The Sox2
R allele also expresses βgeo; however, in LGE postmitotic cells, Sox2-βgeo expression is much less than that of Sox1-βgeo and is not very visible by short (3 h) X-gal staining (only a slight increase of X-gal staining is seen in the VS; Figure 9C compared to Figure 9B).
Figure 9
Sox1 Expression in Precursors Is Not Sufficient for the Emergence of OT/VS Neurons
(A–C) X-gal staining of coronal sections from the ventral telencephalon showing Sox1-βgeo-expressing OT-prospective neurons at E16-stage embryos with one wild-type Sox1 allele, Sox1
βgeo/+, in (A), and none, Sox1
βgeo/M1, in (B), and HoHe (Sox1
βgeo/M1, Sox2
R/+) in (C). Note the absence of X-gal-stained neurons in the area of the VS (red arrowheads), indicating failure of the Sox2
R allele to rescue OT neuron development in the HoHe embryos. Expression of βgeo from the Sox2
R allele in the HoHe (C) may account for the slight increase of X-gal staining compared to (B).
(D and E) SOX1 immunostaining at E16 embryos performed on the other halves of the brains of (A and C), respectively. Note that the level of SOX1 expression in the precursors (yellow arrows) is the same, whether it is expressed from the Sox2
R allele in the HoHe (E) or from one of the Sox1 wild-type alleles in the Sox1 single heterozygotes (D). Note in the HoHe (E), this expression is not sufficient for the development of OT neurons.
(F and G) DARPP32 immunostaining of coronal brain sections from Sox1
+/+
Sox2
R/+ (F) and Sox1
M1/M1
Sox2
R/+ (G) P10 mice indicating loss of VS neurons.
Red arrowheads indicate OT; red arrows indicate anterior commissure. OC, olfactory cortex. Scale bar = 500 μm (A–E), 1 mm (F and G).
Each brain was split into left and right hemispheres, and coronal sections of the left were used for short staining with X-gal (Figure 9A–9C) whereas sections of the right were stained with SOX1 antibody (Figure 9D and 9E). The hemispheres stained for X-gal showed characteristic staining of OT neurons in heterozygous Sox1
βgeo/+ mice (red arrowheads in Figure 9A), but the hemispheres of the Sox1-null embryos (Sox1
βgeo/M1) with Sox2
R/+ (Figure 9B) or without (Figure 9C) did not. This indicates that Sox1-null embryos do not develop OT despite the presence of the Sox2
R allele and SOX1 protein in progenitors. To verify the presence of SOX1 protein in the precursors of the HoHe mice, we examined the other hemisphere that was stained with SOX1 antibody. In the Sox1
βgeo/+ embryos, we found SOX1 present in the VZ (yellow arrows in Figure 9D and 9E) and the OT neurons (red arrowheads in Figure 9D). In Sox1
βgeo/M1 (null) embryos, SOX1 expression was completely absent (data not shown); in the HoHe embryos, SOX1 protein was present in the VZ (yellow arrow in Figure 9E) and in very few neurons of the LGE (Figure 9E). HoHe mice, like the Sox1-nulls, are born with small eyes, and around weaning age develop seizures associated with lethality, which, if anything, is increased compared to that of Sox1-null mice (data not shown). In the brain of P10 HoHe mice, we used staining with DARPP-32 antibody (which is a SOX1-independent striatal marker) to investigate the recovery of OT neurons, and found staining in the striatal mantle but not in the VS (Figure 9G compared to Figure 9F). We therefore concluded that SOX1 expression in precursors is not sufficient to rescue VS/OT neuron fate specification, and that the continued presence of SOX1 in postmitotic cells is required for their identity.
Sox1/Sox2 Expression in Neurons Is Sufficient for Their Migration to the VS
We have shown that in mice carrying two (Sox1
+/+), or one (Sox1M1
/+), Sox1 wild-type alleles (see Figure 8A, 8B, and 8C), the migration of the Sox2-positive LGE neurons is not overtly different from that observed in mice carrying the Sox2
R/+ allele. However, it remained unknown whether the Sox1/Sox2 double-positive LGE neurons migrated to the VS when both Sox1 endogenous alleles were missing (HoHe mice). We used X-gal staining to follow these neurons in several litters (n > 10), including HoHe mice, which have two Sox1
M1 alleles and thus βgeo expression exclusively driven by the Sox2
R allele. We found that in the LGE of these mice, the double-positive neurons are generated and migrate to the OT area (compare Figure 10A and 10B), but this area is compacted in the absence of the majority of the OT/SOX1 neurons. The above data show that the continued expression of Sox1 in neurons of the LGE is sufficient to direct their migration to the OT in the absence of endogenous Sox1.
Figure 10
Sox1 Expression in Postmitotic LGE Cells Is Sufficient for Neuronal Migration in the VS
X-gal staining of coronal sections from the ventral telencephalon of P0 mice indicating the migration of neurons expressing SOX1 from the Sox2
R allele in the presence (A) and absence (B; HoHe) of endogenous Sox1 wild-type genes. Black arrow points at the striatal bridges forming in HoHe mice. Scale bar = 100 μm.
Discussion
The specification of neurons in the ventral telencephalon has been shown to depend on several transcription factors that are expressed mainly in proliferating precursors. However, it was unknown to what degree specification in precursors included the emergence of neuronal subtype identity in the ventral telencephalon, and whether expression of additional transcription factors was required. We showed that the differentiation and migration of early- and late-born neurons that constitute the VS require SOX1 expression not only in precursors but also in postmitotic cells. Furthermore, in this region, the migration and organization of other neurons such as those expressing Pax6 also depend on the presence of SOX1-positive VS neurons. The finding that SOX1 functions in neurons to control migration and identity is novel and suggests that the other SOXB1 factors, in addition to their roles in precursors, have similar functions in neurons.
Identity and Migration of Neurons in the VS
The development of subtype identity and migration of neurons in the ventral telencephalon has not been well characterized. The expression of differentiation markers reveals neurons in both VS and dorsal striatum, but we showed that SOX1 specifically marks a large population of VS neurons that form the principal layer II of the OT, the islands of Calleja, and the nucleus accumbens (see Figures 1 and 2). In addition, we showed that the neurons expressing Sox1 are born continuously from E13 until the first postnatal week and that these migrate to a ventrolateral region of the telencephalon, with later-born neurons positioned progressively to more medial positions (see Figure 4). In the absence of SOX1, the majority of neurons of the VS fail to develop. All Sox1-expressing neurons of the OT and the islands of Calleja require SOX1 for their development, but it is essential only for the shell of the nucleus accumbens, although the core also expresses it. While neurons of the adjacent striatal mantle and the olfactory cortex that do not express Sox1 develop normally in its absence, other groups of neurons within the VS appear to be disorganized. Specifically, we identified a distinct population of neurons located lateral to the OT at the border with the olfactory cortex that expresses Pax6, but not Sox1. In the absence of SOX1, these neurons migrate into more medial positions, occupying the space of the missing OT neurons (see Figure 5). These are not mis-specified Sox1-null neurons because they do not express βgeo. This indicates that Sox1-expressing OT neurons play a non-cell-autonomous role in the organization of other neurons in this region, including the production of essential signals for migration. Most likely, the disorganization of the VS in the absence of SOX1 results in abnormal local neuronal connectivity, which in turn leads to the abnormal (epileptiform) electrophysiological behavior observed in the SOX1-deficient animals [34].
SOX1 Function in Precursors
SOXB1 factors share considerable homology in both their DNA binding and C-terminal transcriptional activation domains, and they are co-expressed in precursors. It is therefore possible that in the LGE precursors, the role of SOX1 in the specification of OT/VS neurons is redundant. However, as SoxB1 genes have a broad expression in the neuroepithelium, we have to assume that their specific function at different areas of the VZ, and particularly the VZ of the LGE, is controlled by the presence of LGE-specific partner factors. SoxNeuro and Dichaete, the two Drosophila orthologs of the vertebrate SoxB1 group genes, also show overlapping functions during neural development [49]. Furthermore, in Drosophila, these two genes have been shown to genetically interact with the dorsoventral patterning genes ind (intermediate neuroblast defective) and vnd (ventral nerve chord defective) [50,51]. The vertebrate orthologs of ind and vnd are Gsh1/2 [52,53] and Nkx2.2 [54], respectively. In the mouse, Gsh2 is expressed in the VZ/SVZ, and like Sox1, its loss results in a reduction of VS neurons. As target gene specificity of SOX proteins depends on partnering with other transcription factors [20], our work, along with the data from Drosophila, supports the hypothesis that in the LGE precursors GSH1/2 may act as partners for SOXB1 factors to initiate ventral telencephalic neuronal identity.
The neurons of the VS area occupy approximately a quarter of the striatal mass [55], and migrate there over a period of at least 10 d (E13 to first postnatal week). The LGE precursors that generate the OT/VS in the LGE are expected to have an equivalent representation during this period of development. In Gsh2-null mice, which are missing early-born OT neurons, there is a deficit of precursors in the LGE, readily seen by the reduced expression of Dlx1/2 in LGE precursors [9–11]. In Sox1-null mice, the neuronal deficit is more severe than that in Gsh2 mutants, as it includes both early- and late-born OT/VS neurons. However, BrdU labeling and LGE precursor-specific marker analysis, including Gsh2, Dlx1/2, Mash1, and Pax6, in Sox1-null brains at different stages did not show any deficit in precursors. The increase of X-gal-stained (Sox1-βgeo; see Figure 2) and BrdU-labeled neurons (see Figure 4D and 4F) in the area of the septum and the striatum supports the notion that the VS/OT neurons are born but lack VS subtype identity to migrate toward ventral positions. The normal expression of TuJ1, a marker of immature neurons, excluded the possibility that loss of SOX1 delays or enhances differentiation. Therefore, in the absence of SOX1, the precursors are there and generate neurons, but these fail to migrate to the VS because they assume different identity and position. The finding that Sox1-null neurons contribute widely to different areas argues that the presence of SOX1 provides neurons with ventral identity and the ability to migrate to ventral regions.
Emergence of VS Neuron Identity
To test the role of Sox1 expression in neurons and to determine whether ventral identity emerges in postmitotic cells, we limited expression of Sox1 largely to precursors of LGE neurons. Sox2
R/+ mice express Sox1 from one of the Sox2 alleles. When the Sox2
R allele is present in animals with no endogenous Sox1 wild-type alleles (HoHe), SOX1 expression mimics that of SOX2—being present in VS/OT precursors but largely absent from the neurons they give rise to. HoHe mice also fail to develop the majority of VS/OT neurons (see Figure 9) and exhibit an equally severe phenotype to that of Sox1-null mice in the OT. As these mice reproduce faithfully the Sox1-null phenotype without any evidence of a partial rescue, it is unlikely that this is the result of incomplete expression of Sox1 from the Sox2 promoter in the precursors. However, to exclude the possibility that the failure of OT/VS neuron development in HoHe mice was due to a low level of expression of SOX1 protein in precursors, we used one hemisphere of the brain to assay OT development and the other for SOX1 antibody staining, linking in each animal the phenotype with the presence of SOX1 protein in precursors. We found no difference in the extent and level of expression of SOX1 protein in the VZ/SVZ of embryos with one copy of Sox1, whether it is expressed from the Sox2 locus in HoHe (see Figure 9E) or the endogenous Sox1 allele in Sox1
βgeo/+ heterozygotes (see Figure 9D). Therefore, the emergence of VS/OT identity requires Sox1 expression in postmitotic cells. Consistent with the above findings, the small population of LGE postmitotic cells in HoHe mice that maintain SOX1 expression from the Sox2
R allele migrate to the VS. However, the number of these neurons is small and cannot rescue the deficit in the area of the VS. In conclusion, although specification of neuronal identity is initiated in precursors, emergence of neuronal subtype and ventral migration require the continued presence of SOX1. Our findings suggest that in other brain areas, subtype identity and migration may also be controlled by the expression of transcription factors in postmitotic cells.
The current study, along with our previous one showing that SOX1 expression in the lens of the mouse is responsible for terminal differentiation and the expression of γ-crystallin genes [37], has revealed that SOX1 has important functions in postmitotic cell differentiation at two distinct sites. It is possible that the other SOXB1 factors have similar roles in postmitotic cells in which their expression is maintained.
Materials and Methods
Gene targeting.
The βgeo gene was inserted into the Sox1 single exon as previously described [37]. The resulting targeted locus produces a fusion protein consisting of the first 50 amino acids from SOX1 (which excludes the HMG box) followed by ten amino acids that are encoded by a synthetic linker sequence, and the βgeo sequence (including a polyadenylation signal). Tissue culture was carried out as described before [37], omitting the addition of gancyclovir for negative selection. The targeting vector did not contain any other selectable markers or promoters, and the targeting frequency was 1/52. As Sox1 is not normally expressed in embryonic stem (ES) cells, we used the minimum level of G418 for selection. Positive recombinants were identified by Southern blotting, using a 3′ 1-kb external probe on an EcoRI digest. Three ES cell clones were obtained, and one was successfully passed through to the germ line. All anatomical investigations were performed on mice of mixed genetic background. Sox1
βgeo/+ mice were mated with mice that were heterozygous for the previously described [37] Sox1 deletion (Sox1
M1) and did not express β-galactosidase.
For the Sox2 replacement vector, the 5′ and 3′ homologies used were the same as described before [30]. A SmaI-XhoI 2.2-kb Sox1 fragment containing the SOX1 open reading frame was flanked by LoxP cassettes followed by the NotI-SalI IRES–βgeo-polyA fragment (plasmid gift from Dr. A. Smith, University of Edinburgh). The replacement vector was linearized with SalI and electroporated into ES cells. Positive recombinants were identified by Southern blotting as described before [30]. Several targeted ES cells were isolated at a frequency of 1/20 and gave germ-line transmission of the mutation. Heterozygous animals carrying the Sox2Sox1
βgeo allele (referred to in the text as Sox2
R) expressed Sox1 and βgeo where Sox2 is normally expressed.
Deletion of the SOX1 coding region from the Sox2Sox1
βgeo allele was achieved via pronuclear injection of a supercoiled plasmid expressing Cre-recombinase (gift from Dr. K. Rajewsky, Harvard Medical School). Although in the text we refer to this new allele as Sox2
βgeo, it is officially named Sox2
βgeo2 to distinguish it from the one previously described [30].
X-gal staining and in situ hybridization.
For β-galactosidase staining, fetal, newborn, or adult mouse brains were processed as previously described [30]. Detection of Sox1 mRNA was performed in whole embryos, as described previously [27,34]. The probes that were used on embryonic brain slices were generated by RT-PCR from embryonic brain cDNA. The position of the probes was Dlx1 (nt 41–573), Brn4 (nt 199–541), and Robo (nt 8–779). All fragments were cloned into a suitable cloning vector (pGET-easy, Promega, Madison, Wisconsin, United States), and were re-amplified using a sense oligonucleotide and an oligonucleotide upstream of either the T7 or SP6 sites. The resulting products were gel-purified, and 40 ng was used for probe synthesis. The brains were processed as described before [27,34].
BrdU labeling
A 25 mg/ml solution of BrdU (Sigma, St. Louis, Missouri, United States) was made in PBS warmed to 37 °C. The solution was sterilized through a 0.2-μm syringe filter and injected into the peritoneal cavity of pregnant mice (0.1 ml per 25 g of body weight, to give a final dosage of 0.1 mg/g). Brains were harvested 1 or 72 h after the injection, or on P16. The brains were fixed in 4% PFA in PBS at 4 °C overnight, embedded in paraffin wax, and cut into 5-μm thick sections. The sections were processed for immunohistochemistry as previously described [18].
Immunohistochemistry.
The source of antibodies and the dilutions used are as follows: PAX6, 1:10 (gift from Dr. J. Briscoe); GSH2, 1:5,000 (gift from Dr. K. Campbell); β-galactosidase 1:2,000 (Cappel); TuJ1, 1:1,000 (Novus Biologicals, Littleton, Colorado, United States); and SOX1, 1:500; SOX2, 1:500; and SOX3, 1:500 (gift from Dr. T. Edlund). MASH1, 1:2 (gift from Dr. F. Guillemot) [18], RC2 [56], and DARPP-32 [57] were used as previously described. For single or double immunofluorescence, embryonic tissue was fixed either for 1 h in 4% PFA in PBS at room temperature or for 15 min in MEMFA as described before [58]. The brains were then washed in PBS, cryoprotected overnight in 30% sucrose in PBS at 4 °C, embedded in OCT (Raymond Lamb), and cut in 10-μm and 15-μm sections using a cryostat. Sections were rehydrated in PBS, blocked in 4% goat serum and 0.1% Triton X-100 in PBS, and incubated overnight at 4 °C with primary antibodies diluted in the blocking solution. After incubation, the slides were washed in PBS and incubated with fluorescent secondary antibodies (FITC- or TRITC-labeled, 1:200 in blocking solution) for 1 h at room temperature.
Slides for double immunolabeling were first immunostained for SOX1 as described above and visualized with Alexa568 goat anti-rabbit antibody (1:500, Molecular Probes, Eugene, Oregon, United States), and then incubated with unlabeled anti-rabbit secondary antibody (1:100, Dako, Glostrup, Denmark) for 1 h at room temperature to block existing unlabeled anti-SOX1 antibody. Subsequently, the slides were incubated with anti-SOX2 antibody in blocking solution without Triton X-100 for 48 h at room temperature and visualized with Alexa488 goat anti-rabbit antibody (1:500, Molecular Probes). The cross-reactivity of the SOX1 and SOX2 antibodies when using immunohistochemistry was excluded by looking at tissues where SOX1 (lens; [37]) or SOX2 (see Figure 7B–7E) are uniquely present, and in Western blots where each antibody recognizes a different size band (data not shown). After incubation with the secondary antibodies, the slides were washed in PBS and the sections were mounted with Vectashield mounting medium (Vector Laboratories, Burlingame, California, United States) and observed under either a fluorescent or a confocal microscope.
Supporting Information
Figure S1 The GSH2/PAX6 Boundary Is Unaffected throughout Development in the Absence of SOX1
Similar to earlier stages (E12; see Figure 5), the expression of GSH2 (green) and PAX6 (red) protein in wild-type (A, C, and E) and mutant brains (B, D, and F) is the same at E14 and E15, as shown by DAPI (blue) nuclear stain. Cx, cortex; lge, lateral ganglionic eminence; mge, medial ganglionic eminence. Scale bar = 200 μm for (C) and (D).
(10 MB TIF).
Click here for additional data file.
Figure S2 Abnormal Distribution of VS Pax6-Expressing Neurons in the Absence of SOX1
Ventral telencephalic region of coronal brain sections stained with antibodies: PAX6 (brown) at E16 (A and B) and adult (C and D); SOX1 (green) and PAX6 (red) at E14 (E); and SOX1 and β-galactosidase (green) at E15 (F). Note Pax6-expressing cells are excluded from the OT region in the ventral telencephalon in wild-type (Sox1
+/+) mice but not in the mutant(Sox1
−/−). Arrows indicate ectopically localized Pax6-expressing cells. PAX6 and SOX1 are co-expressed in precursors but not in postmitotic cells in both mutant and wild-type. Scale bar = 500 μm for (C) and (D).
(10 MB TIF).
Click here for additional data file.
Figure S3
Sox1-Expressing VZ Precursors Have Radial Glial Morphology
Detail from immunofluorescence with β-galactosidase antibody of an E15 coronal brain section from mice carrying the Sox1
βgeo allele. In the cortical ventricular zone, this staining reveals the cytoplasmic compartment of the SOX1-expressing progenitors, which have radial glial morphology.
(3.34 MB TIF).
Click here for additional data file.
Figure S4 The Distribution of Radial Glia in the LGE Is Unaffected in the Absence of SOX1
Coronal brain sections of wild-type (+/+) and Sox1-null (−/−) mice at E16. Immunostained with RC2 antibody (a radial glia marker) showing no differences.
(2.5 MB TIF).
Click here for additional data file.
Figure S5 Widespread Presence of SOX1 and SOX2 Proteins in Nuclei of the LGE VZ
Immunofluorescence (green) of E14-stage wild-type coronal brain sections at the level of the LGE stained with SOX1 (C and G) and SOX2 (D and H) antibodies and visualized under a confocal microscope. (A, B, and E–H) are stained with TOTO red-fluorescent nuclear stain. (C, E, and G) and (D, F, and H) show high magnification of the area indicated in the rectangle in (A) and (B), respectively.
(10 MB TIF).
Click here for additional data file.
Table S1 No Difference in the Number of LGE Dividing Precursors in the Absence of SOX1
BrdU-positive cells were counted using the Openlab image analysis program (Improvision, Coventry, United Kingdom). Measurements were performed in the area of the LGE (VZ/SVG) and of the pallial VZ. All data are represented as mean ± standard error of the mean. Cell counts were done in at least three different slides (sections) from each brain and in at least three separate optical fields in each slide (n = 4). To correct for tissue thickness and to obtain a better estimate of the proliferation within the LGE VZ/SVZ, the numbers of BrdU-positive cells were expressed as a ratio of the total number of LGE VZ/SVZ cells that were counted after hematoxylin staining, and as a ratio to the BrdU-positive cells of the pallial VZ/SVZ. Comparisons were made between wild-type and mutant mice using the unpaired Student's t-test (p < 0.05).
(21 KB DOC).
Click here for additional data file.
Accession Numbers
The GenBank (http://www.ncbi.nlm.nih.gov/Genebank/) accession numbers for the entities discussed in this paper are Brn4 (NM 008901), Dlx1 (NM 010053), Robo (MMU 17793), Sox1 (MN 009233), and Sox2 (MN 011443).
We thank for antibodies Drs. K. Campbell (Children's Hospital Research Foundation, Cincinnati, Ohio), T. Edlund (University of Umea, Sweden), and F. Guillemot (National Institute for Medical Research Medical Research [NIMR MRC], London). For comments on the manuscript we thank J. Briscoe (NIMR MRC) and J. Corbin (Georgetown University, Washington, DC). We are grateful to Z. Webster for the generation of transgenic mice and M. Delahaye for technical support with the mice. This work was supported by the MRC, the Wellcome Trust (grant 062197 to AC), and a Marie Curie Fellowship of the European Community Program (contract QLGA-CT-2001–50880 to AE).
Competing interests. The authors have declared that no competing interests exist.
Author contributions. AE, IK, SM, HW, and VE conceived and designed the experiments. AE, IK, SM, HW, PA, MD, and VE performed the experiments. AE, IK, SM, HW, and VE analyzed the data. AE, IK, SM, HW, DK, AC, RL, and VE contributed reagents/materials/analysis tools. VE wrote the paper.
¤a Current address: Stem Cell Biology Laboratory, Wolfson Centre for Age-Related Diseases, King's College, London, United Kingdom
¤b Current address: The Cyprus Institute of Neurology and Genetics, Nicosia, Cyprus
¤c Current address: Nature Reviews Neuroscience, London, United Kingdom
Citation: Ekonomou A, Kazanis I, Malas S, Wood H, Alifragis P, et al. (2005) Neuronal migration and ventral subtype identity in the telencephalon depend on SOX1. PLoS Biol 3(6): e186.
Abbreviations
bHLHbasic helix-loop-helix
BrdU5-bromo-2′-deoxyuridine
CNScentral nervous system
DARPP-32dopamine and cAMP-regulated phosphoprotein
E[number]embryonic day [number]
ESembryonic stem
IRESinternal ribosomal entry site
LGElateral ganglionic eminence
OTolfactory tubercle
P[number]postnatal day [number]
SVZsubventricular zone
TuJ1anti-βIII-tubulin antibody
VZventricular zone
VSventral striatum
βgeoβ-galactosidase-neo
==== Refs
References
Anderson S Mione M Yun K Rubenstein JL Differential origins of neocortical projection and local circuit neurons: Role of Dlx genes in neocortical interneuronogenesis Cereb Cortex 1999 9 646 654 10498283
Anderson SA Marin O Horn C Jennings K Rubenstein JL Distinct cortical migrations from the medial and lateral ganglionic eminences Development 2001 128 353 363 11152634
Parnavelas JG The origin and migration of cortical neurones: New vistas Trends Neurosci 2000 23 126 131 10675917
Deacon TW Pakzaban P Isacson O The lateral ganglionic eminence is the origin of cells committed to striatal phenotypes: Neural transplantation and developmental evidence Brain Res 1994 668 211 219 7704606
Heimer L Basal forebrain in the context of schizophrenia Brain Res Brain Res Rev 2000 31 205 235 10719150
Stoykova A Fritsch R Walther C Gruss P Forebrain patterning defects in Small eye mutant mice Development 1996 122 3453 3465 8951061
Fernandez AS Pieau C Reperant J Boncinelli E Wassef M Expression of the Emx-1 and Dlx-1 homeobox genes define three molecularly distinct domains in the telencephalon of mouse, chick, turtle and frog embryos: Implications for the evolution of telencephalic subdivisions in amniotes Development 1998 125 2099 2111 9570774
Anderson SA Qiu M Bulfone A Eisenstat DD Meneses J Mutations of the homeobox genes Dlx-1 and Dlx-2 disrupt the striatal subventricular zone and differentiation of late born striatal neurons Neuron 1997 19 27 37 9247261
Corbin JG Gaiano N Machold RP Langston A Fishell G The Gsh2 homeodomain gene controls multiple aspects of telencephalic development Development 2000 127 5007 5020 11060228
Toresson H Potter SS Campbell K Genetic control of dorsal-ventral identity in the telencephalon: Opposing roles for Pax6 and Gsh2 Development 2000 127 4361 4371 11003836
Yun K Potter S Rubenstein JL Gsh2 and Pax6 play complementary roles in dorsoventral patterning of the mammalian telencephalon Development 2001 128 193 205 11124115
Yun K Garel S Fischman S Rubenstein JL Patterning of the lateral ganglionic eminence by the Gsh1 and Gsh2 homeobox genes regulates striatal and olfactory bulb histogenesis and the growth of axons through the basal ganglia J Comp Neurol 2003 461 151 165 12724834
Szucsik JC Witte DP Li H Pixley SK Small KM Altered forebrain and hindbrain development in mice mutant for the Gsh-2 homeobox gene Dev Biol 1997 191 230 242 9398437
Jimenez D Lopez-Mascaraque LM Valverde F De Carlos JA Tangential migration in neocortical development Dev Biol 2002 244 155 169 11900465
de Carlos JA Lopez-Mascaraque L Valverde F Dynamics of cell migration from the lateral ganglionic eminence in the rat J Neurosci 1996 16 6146 6156 8815897
Bayer SA Altman J Russo RJ Dai XF Simmons JA Cell migration in the rat embryonic neocortex J Comp Neurol 1991 307 499 516 1856333
Horton S Meredith A Richardson JA Johnson JE Correct coordination of neuronal differentiation events in ventral forebrain requires the bHLH factor MASH1 Mol Cell Neurosci 1999 14 355 369 10588390
Casarosa S Fode C Guillemot F Mash1 regulates neurogenesis in the ventral telencephalon Development 1999 126 525 534 9876181
Wilson M Koopman P Matching SOX: Partner proteins and co-factors of the SOX family of transcriptional regulators Curr Opin Genet Dev 2002 12 441 446 12100890
Kamachi Y Uchikawa M Kondoh H Pairing SOX off: With partners in the regulation of embryonic development Trends Genet 2000 16 182 187 10729834
Schepers GE Teasdale RD Koopman P Twenty pairs of sox: Extent, homology, and nomenclature of the mouse and human sox transcription factor gene families Dev Cell 2002 3 167 170 12194848
Pevny LH Lovell-Badge R Sox genes find their feet Curr Opin Genet Dev 1997 7 338 344 9229109
Wegner M From head to toes: The multiple facets of Sox proteins Nucleic Acids Res 1999 27 1409 1420 10037800
Bowles J Schepers G Koopman P Phylogeny of the SOX family of developmental transcription factors based on sequence and structural indicators Dev Biol 2000 227 239 255 11071752
Collignon J Sockanathan S Hacker A Cohen-Tannoudji M Norris D A comparison of the properties of Sox-3 with Sry and two related genes, Sox-1 and Sox-2 Development 1996 122 509 520 8625802
Uchikawa M Kamachi Y Kondoh H Two distinct subgroups of Group B Sox genes for transcriptional activators and repressors: Their expression during embryonic organogenesis of the chicken Mech Dev 1999 84 103 120 10473124
Wood HB Episkopou V Comparative expression of the mouse Sox1, Sox2 and Sox3 genes from pre-gastrulation to early somite stages Mech Dev 1999 86 197 201 10446282
Aubert J Stavridis MP Tweedie S O'Reilly M Vierlinger K Screening for mammalian neural genes via fluorescence-activated cell sorter purification of neural precursors from Sox1-gfp knock-in mice Proc Natl Acad Sci U S A 2003 100 Suppl 1 11836 11841 12923295
Ying QL Stavridis M Griffiths D Li M Smith A Conversion of embryonic stem cells into neuroectodermal precursors in adherent monoculture Nat Biotechnol 2003 21 183 186 12524553
Avilion AA Nicolis SK Pevny LH Perez L Vivian N Multipotent cell lineages in early mouse development depend on SOX2 function Genes Dev 2003 17 126 140 12514105
Graham V Khudyakov J Ellis P Pevny L SOX2 functions to maintain neural progenitor identity Neuron 2003 39 749 765 12948443
Bylund M Andersson E Novitch BG Muhr J Vertebrate neurogenesis is counteracted by Sox1–3 activity Nat Neurosci 2003 6 1162 1168 14517545
Kan L Israsena N Zhang Z Hu M Zhao LR Sox1 acts through multiple independent pathways to promote neurogenesis Dev Biol 2004 269 580 594 15110721
Malas S Postlethwaite M Ekonomou A Whalley B Nishiguchi S Sox1-deficient mice suffer from epilepsy associated with abnormal ventral forebrain development and olfactory cortex hyperexcitability Neuroscience 2003 119 421 432 12770556
Rizzoti K Brunelli S Carmignac D Thomas PQ Robinson IC SOX3 is required during the formation of the hypothalamo-pituitary axis Nat Genet 2004 36 247 255 14981518
Ferri AL Cavallaro M Braida D Di Cristofano A Canta A Sox2 deficiency causes neurodegeneration and impaired neurogenesis in the adult mouse brain Development 2004 131 3805 3819 15240551
Nishiguchi S Wood H Kondoh H Lovell-Badge R Episkopou V Sox1 directly regulates the gamma-crystallin genes and is essential for lens development in mice Genes Dev 1998 12 776 781 9512512
Menezes JR Luskin MB Expression of neuron-specific tubulin defines a novel population in the proliferative layers of the developing telencephalon J Neurosci 1994 14 5399 5416 8083744
Alvarez-Bolado G Rosenfeld MG Swanson LW Model of forebrain regionalization based on spatiotemporal patterns of POU-III homeobox gene expression, birthdates, and morphological features J Comp Neurol 1995 355 237 295 7608343
Yuan W Zhou L Chen JH Wu JY Rao Y The mouse SLIT family: Secreted ligands for ROBO expressed in patterns that suggest a role in morphogenesis and axon guidance Dev Biol 1999 212 290 306 10433822
Bayer SA Neurogenesis in the rat primary olfactory cortex Int J Dev Neurosci 1986 4 251 271 3455589
Bayer SA Neurogenesis in the anterior olfactory nucleus and its associated transition areas in the rat brain Int J Dev Neurosci 1986 4 225 249 3455588
Rallu M Corbin JG Fishell G Parsing the prosencephalon Nat Rev Neurosci 2002 3 943 951 12461551
Puelles L Kuwana E Puelles E Bulfone A Shimamura K Pallial and subpallial derivatives in the embryonic chick and mouse telencephalon, traced by the expression of the genes Dlx-2, Emx-1, Nkx-2.1, Pax-6, and Tbr-1 J Comp Neurol 2000 424 409 438 10906711
Malatesta P Hack MA Hartfuss E Kettenmann H Klinkert W Neuronal or glial progeny: Regional differences in radial glia fate Neuron 2003 37 751 764 12628166
Anthony TE Klein C Fishell G Heintz N Radial glia serve as neuronal progenitors in all regions of the central nervous system Neuron 2004 41 881 890 15046721
Misson JP Edwards MA Yamamoto M Caviness VS Identification of radial glial cells within the developing murine central nervous system: Studies based upon a new immunohistochemical marker Brain Res Dev Brain Res 1988 44 95 108 3069243
Rajewsky K Gu H Kuhn R Betz UA Muller W Conditional gene targeting J Clin Invest 1996 98 600 603 8698848
Overton PM Meadows LA Urban J Russell S Evidence for differential and redundant function of the Sox genes Dichaete and SoxN during CNS development in Drosophila
Development 2002 129 4219 4228 12183374
Buescher M Hing FS Chia W Formation of neuroblasts in the embryonic central nervous system of Drosophila melanogaster is controlled by SoxNeuro Development 2002 129 4193 4203 12183372
Zhao G Skeath JB The Sox-domain containing gene Dichaete/fish-hook acts in concert with vnd and ind to regulate cell fate in the Drosophila neuroectoderm Development 2002 129 1165 1174 11874912
Valerius MT Li H Stock JL Weinstein M Kaur S Gsh-1: A novel murine homeobox gene expressed in the central nervous system Dev Dyn 1995 203 337 351 8589431
Hsieh-Li HM Witte DP Szucsik JC Weinstein M Li H Gsh-2, a murine homeobox gene expressed in the developing brain Mech Dev 1995 50 177 186 7619729
Pabst O Herbrand H Arnold HH Nkx2–9 is a novel homeobox transcription factor which demarcates ventral domains in the developing mouse CNS Mech Dev 1998 73 85 93 9545546
Paxinos G Franklin KBJ The mouse brain in stereotaxic coordinates 2001 San Diego Academic Press 264
Denaxa M Chan CH Schachner M Parnavelas JG Karagogeos D The adhesion molecule TAG-1 mediates the migration of cortical interneurons from the ganglionic eminence along the corticofugal fiber system Development 2001 128 4635 4644 11714688
Ouimet CC Miller PE Hemmings HC Walaas SI Greengard P DARPP-32, a dopamine- and adenosine 3′:5′-monophosphate-regulated phosphoprotein enriched in dopamine-innervated brain regions. III. Immunocytochemical localization J Neurosci 1984 4 111 124 6319625
Pevny LH Sockanathan S Placzek M Lovell-Badge R A role for SOX1 in neural determination Development 1998 125 1967 1978 9550729
| 15882093 | PMC1110909 | CC BY | 2021-01-05 08:21:24 | no | PLoS Biol. 2005 Jun 17; 3(6):e186 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030186 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030199SynopsisBioinformatics/Computational BiologyGeneral MedicineRespiratory MedicineStatisticsBiochemistryBiophysicsCell BiologyEvolutionGenetics/Genomics/Gene TherapyInfectious DiseasesMicrobiologyMolecular Biology/Structural BiologyVirologyVirusesNew Frameshifting Pseudoknot Found in SARS Virus Synopsis6 2005 17 5 2005 17 5 2005 3 6 e199Copyright: © 2005 Public Library of Science.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.
A Three-Stemmed mRNA Pseudoknot in the SARS Coronavirus Frameshift Signal
Pseudoknots: RNA Structures with Diverse Functions
==== Body
Viruses, small organisms that hijack our cellular machinery to replicate their genomes and make new viruses, constantly threaten human health. Not only are we unable to control infections caused by old enemies such as the cold virus, but we are also continually challenged by new enemies, like the coronavirus that causes severe acute respiratory syndrome (SARS-CoV). Vaccines provide one way to deal with viruses, but subtle differences between how host and viral proteins are made may also provide targets for new antiviral therapeutics.
SARS, a life-threatening respiratory illness, first appeared in late 2002. By February 2003, Guangdong Province, China, was in the grip of a SARS epidemic, and public-health officials were predicting that millions of people might become infected. The rapid implementation of effective containment efforts averted this new threat to human health, and, in the end, only 8,098 people became ill. SARS-CoV, a single-stranded RNA virus, was isolated in March 2003 and its genome sequenced by May 2003. Since then, researchers have intensively studied the virus, hoping to identify targets for antiviral therapeutics. Jonathan Dinman and colleagues now describe a new RNA structural motif in the SARS-CoV genome that may provide such a target.
During protein synthesis, molecular machines called ribosomes move along mRNA molecules, translating nucleotide triplets into amino acids. In human cells, the ribosomes usually hook onto the start of an mRNA and decode each triplet in turn. However, viral mRNAs often contain special signals that tell the ribosomes to change register or “frameshift.” This allows viruses to coordinate gene expression from overlapping coding sequences, and it ensures that the correct ratios of enzymatic and structural proteins are made.
A three-stemmed pseudoknot in SARS messenger RNAs could provide a target for antiviral therapeutics (Image: Jonathan Dinman and the National Institute of Allergy and Infectious Diseases)
ORF1a and ORF1b are overlapping, out-of-frame coding sequences within the SARS-CoV genome. Each encodes a polyprotein—a large protein that is cleaved into smaller, functional proteins. Polyprotein 1a is translated directly from ORF1a; the fused polyprotein 1a/1b is produced by programmed −1 ribosomal frameshifting in which the ribosome slips back one nucleotide at a special signal within the mRNA. Like other frameshift signals, the SARS-CoV signal contains a pseudoknot, a stable mRNA structure. Pseudoknots generally contain two stems, in which complementary nucleotides form double-stranded RNA, and two or three loops of unpaired nucleotides. Because the mRNA strand passes over and behind itself to form the stems, the whole structure looks like a small knot of the kind that would unravel if its two ends were pulled.
Unexpectedly, a computational analysis undertaken by Dinman and his colleagues reveals that the pseudoknot in the SARS-CoV frameshift signal contains three stems. The researchers provide further evidence for this novel structure by finding potential three-stemmed pseudoknots in the frameshift signals of other coronaviruses and by doing biochemical and structural NMR studies on the SARS-CoV signal. They also show that the SARS-CoV frameshift signal behaves like other viral frameshift signals in several frameshifting assays, and their mutagenesis studies indicate that specific sequences and structures within stem 2 of the pseudoknot are needed for efficient frameshifting.
The exact role of the extra stem in the SARS-CoV frameshifting signal remains to be determined, but the researchers speculate that it could help to regulate the exact ratio of polyprotein 1a to 1a/1b. The current results also suggest that the three stems may fold back on one another to from a complex globular RNA structure. The elucidation of this structure by high-resolution NMR, the researchers say, should facilitate the rational development of therapeutic agents designed to interfere with SARS-CoV programmed −1 ribosomal frameshifting and should also increase our understanding of how pseudoknots stimulate frameshifting.
| 0 | PMC1110910 | CC BY | 2021-01-05 08:28:14 | no | PLoS Biol. 2005 Jun 17; 3(6):e199 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030199 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030209SynopsisAnimal BehaviorNeuroscienceInsectsFly Movie Theater Reveals Secrets of How Insects See the World Synopsis6 2005 17 5 2005 17 5 2005 3 6 e209Copyright: © 2005 Public Library of Science.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.
Function of a Fly Motion-Sensitive Neuron Matches Eye Movements during Free Flight
==== Body
Seeing the world pass by as you're moving is a complex feat. In flies, images moving across the eye—so-called optic flow—provide sensory information to neurons to guide behaviors like avoiding obstacles and chasing down mates. This maneuverability, combined with a fast image processing speed and the ease of examining fly neurons, has made flies a popular model for dissecting how this occurs. But since much of what we know about how neurons process visual information comes from static situations—for example, placing a fly in a fixed position and presenting it with moving images—the question of how these processes naturally occur is largely unexplored.
In a new study, a German and a Dutch group headed by Martin Egelhaaf and Roland Kern and by Hans van Hateren, respectively, joined forces to take the analysis of optic flow in the blowfly one step closer to the natural situation. To do this, they used a “panoramic virtual reality stimulator” to show a fly in the laboratory what it would see while flying in nature. The authors combined this with measurements of the response of a motion-sensitive blowfly neuron, called the horizontal system equatorial (HSE) cell.
The traditional model holds that the HSE cell extracts information about the motion of a fly from optic flow. Previous work, which involved recording the responses of the HSE cell to simple visual stimuli (measured as a change of electrical potential of the neuron), suggested that HSE responds only to rotations of the visual world. However, the use of more natural visual stimuli suggests that some functions of the HSE cell may have been missed.
HSE neurons in the blowfly allow the fly to extract visual information about the depth structure of the world during flight
How HSE cells respond cannot be recorded in freely moving animals due to technical difficulties. To get around this problem, the Dutch group recorded the free flight of blowflies, including their characteristic head and body movements. The visual stimuli during this behavior were then reconstructed in computer-generated simulations and played back by the German group to flies in the panoramic virtual reality stimulator. Since this was done in the laboratory, the authors could record the responses of the HSE cell as the flies watched this movie.
Normal blowfly flight style involves a combination of saccadic (jerky) turns, where the head rotates at high velocity, and periods of forward motion accompanied by a constant gaze. During a playback of ten different versions of this behavior, the authors did not see a positive change in potential of the HSE cell during saccadic turns, as might be expected from previous conventional recordings. Instead, the HSE cell was depolarized by optic flow between saccades—when the fly's head was not rotating.
This surprising result suggests that blowflies may gather useful visual information about the world from translational (movement without rotation) optic flow—when their heads are not rotating and their gaze is fixed. In fact, the authors found that the blowfly's flight strategy allows information to be extracted from translational optic flow under situations where optic flow from rotation might otherwise dominate. Thus, HSE cells may not only encode information from dominant rotations of the fly itself, but allow the fly to extract “behaviorally relevant information” about the depth structure of the world. However, as the authors point out, it is not yet known whether the blowfly's nervous system can garner rotational and translation information from the combined output of HSE cells.
Nevertheless, through the use of a novel method to play back natural flight behaviors, the authors have been able to discern a new function for a well-studied motion-sensitive neuron—a function that appears to emerge from the blowfly's own behavior. Future experiments can now begin to explore how the fly uses the information generated by this new function.
| 0 | PMC1110911 | CC BY | 2021-01-05 08:21:23 | no | PLoS Biol. 2005 Jun 17; 3(6):e209 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030209 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030226SynopsisDevelopmentGenetics/Genomics/Gene TherapyNeuroscienceMus (Mouse)Migration and Fate Specification in the Ventral Striatum Synopsis6 2005 17 5 2005 17 5 2005 3 6 e226Copyright: © 2005 Public Library of Science.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.
Neuronal Migration and Ventral Subtype Identity in the Telencephalon Depend on SOX1
==== Body
Like other complex organs, the brain develops in stages, under tight control of multiple genetic signals. The key to understanding brain development, therefore, lies in understanding what these signals are, where they act, and what their effects are. The SOX family of transcription factors has emerged as a major group of developmental regulators in the brain, but their exact roles are not well known. In this issue, Vasso Episkopou and colleagues show that the mouse gene Sox1 is required for the differentiation and migration of neurons within a portion of the developing brain.
A major event in brain development is the emergence of the lateral ganglionic eminence (LGE), a bulbous protrusion whose neurons go on to form several vital brain structures, including the ventral striatum (VS), which is ultimately responsible for control of various aspects of motor, cognitive, and emotional functions. Several transcription factors have been shown to operate within the LGE precursors of the VS, but these are not present after these neurons have stopped dividing. To determine whether the protein SOX1 might be playing a role in these cells, the authors examined mice missing the SOX1 protein only in these cells. They found that the absence of the protein from these post-mitotic cells (cells after cell division) prevented normal development of two VS structures, the striatal bridges and the olfactory tubercles.
Sox1 gene expression in wild-type (left) and Sox1 mutant (right) mouse forebrain
It appeared that the defect in mice missing SOX1 from all cells was not in the proliferation of neuronal precursors, which appeared normal, but instead in their differentiation and migration. And while the previously identified factors played no role after the neurons had stopped dividing, continued expression of Sox1 in post-mitotic neurons was required for correct specification of cell identity. Overexpression of Sox1, on the other hand, did not increase the normal number of olfactory tubercle–specified neurons, or alter their migration, suggesting that regulation of these cells is under more complex layers of control.
These results, combined with previous work by the same authors showing a role for Sox1 in terminal differentiation of the mouse lens, suggest that Sox1 and other closely related family members may help specify and maintain final post-mitotic cell identity in a variety of tissue types.
| 0 | PMC1110912 | CC BY | 2021-01-05 08:21:24 | no | PLoS Biol. 2005 Jun 17; 3(6):e226 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030226 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030161Book Reviews/Science in the MediaBotanyZoologyEukaryotesExplorer Naturalists Book Review/Science in the MediaVega Fernando E 5 2005 17 5 2005 17 5 2005 3 5 e161Pick N ( 2004 ) The rarest of the rare: Stories behind the treasures at the Harvard Museum of Natural History. Introduction by E. O. Wilson; photographs by M. Sloan .
New York : HarperResource . 178 p (hardcover) 0-06-053718-3. US$22.95 Copyright: © 2005 Fernando E. Vega.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.Nancy Pick's recent book The Rarest of the Rare: Stories behind the Treasures at the Harvard Museum of Natural History brings a vast collection to life.
==== Body
The famous Russian writer Vladimir Nabokov once described writing as “a torture and a pastime” and contrasted it to “a long and exciting career as an obscure curator of lepidoptera in a great museum” [1]. For six years before moving to Cornell University as a professor of Russian literature, Nabokov worked as a research fellow at Harvard's Museum of Comparative Zoology, where he specialized in “blues” (Family Lycaenidae). The public face of this museum (as well as the Harvard University Herbaria and the Mineralogical and Geological Museum) is Harvard's Museum of Natural History, whose collection consists of more than 21 million specimens acquired over two centuries. In a fascinating new book, The Rarest of the Rare: Stories behind the Treasures at the Harvard Museum of Natural History, Nancy Pick describes highlights of this vast collection in a rarely found combination of seamless prose, outstanding photographs, history, legendary figures, and, of course, science.
Choosing among 21 million specimens appears a daunting task, and stripped of their historical significance, some of Pick's choices may seem, at first glance, somewhat pedestrian. It is not until we read her description of the specimen of a common sand dollar, that we learn it was collected by Charles Darwin in 1834 during his journeys as a naturalist on the Beagle. Darwin sent the specimen to an echinoid specialist in Switzerland named Louis Agassiz, who later moved to Harvard and obtained funding for the Museum of Natural History, which opened in 1859. A contemporary of Agassiz at Harvard was the famous botanist Asa Gray, who, in contrast to Agassiz, was a strong believer in evolution. One of the items in the collection is an 1857 letter from Darwin to Gray (a photograph of the letter is shown in the book) describing his thoughts on natural selection, two years before The Origin of Species was published. Another specimen that's uninteresting until we know its provenance is a turtle that Harvard College graduate Henry David Thoreau sent to Agassiz in 1847. The turtle was collected in the Walden Pond of Thoreau's classic book.
The birds of American history are well represented by a pair of pheasants that the Marquis de Lafayette sent to George Washington in 1786. The original home of these pheasants was in the first scientific museum in the country, founded by the painter Charles Wilson Peale in Philadelphia. When the museum closed in 1849, the specimens eventually found their way to Harvard. A more exotic specimen in the collection, the now extinct bird known as the common mamo (Drepanis pacifica), was collected by Captain James Cook in 1778 in Hawaii. The yellow feathers from this species were used for King Kamehameha's cloak, requiring over 80,000 mamos. In the book, we can also appreciate parts of a dodo skeleton, and an egg from the elephant bird (Aepyornis maximus), a now extinct flightless bird from Madagascar that grew to ten feet tall and laid the largest eggs known for a bird.
In the realm of insects, Pick has chosen to show us some fabulous specimens, including a 35-million-year-old fossil butterfly; the largest insect wing on record, belonging to a dragonfly-like creature collected in Oklahoma and having a wing span of 2.5 feet; and an unbelievable gynandromorphic morpho butterfly, with the wing colors and size of a male on one side, and those of a female on the other.
Pick also presents cases of murder and fraud, including the famous case of John W. Webster, a Professor of Chemistry at Harvard Medical College, who murdered George Parkman. The tale involves Webster's mineral collection and the purchase of a mastodon for the museum. Bearing witness to fraud in science is a painting by John James Audubon in the museum collection. In trying to prove he had depicted the common grouse (Bonasa umbellus) before his competitor, Alexander Wilson (known as the father of American ornithology), Audubon dated his chalk and watercolor illustration with the year 1805. The problem is that the watermark in the paper is dated 1810.
Perhaps the best known specimens at Harvard's Museum of Natural History are the Blaschka Glass Models of Plants, created by Leopold and Rudolf Blaschka in Germany. Originally commissioned for teaching botany by the first director of Harvard's Botanical Museum, an exclusive ten-year contract signed in 1890 extended into 50 years of work and resulted in 4,400 models so exquisite and realistic, it is hard to believe they are actually made of glass.
Other specimens presented in the book include the all-too-familiar regal lily (Lilium regale) collected in China and brought to the United States by Ernest H. Wilson in 1911; a trilobite collected in 1824 by Charles D. Wolcott, discoverer of the famous Burgess Shale fossil deposit in the Canadian Rocky Mountains; a female anglerfish (Linophryne bicornis) showing the parasitic male still attached; a woodpecker (Melanerpes lewis) collected by Meriwether Lewis and believed to be the only complete specimen from the Lewis and Clark expedition (1804–1806); and Richard Evan Schultes's hallucinogenic mushrooms.
The Rarest of the Rare brings forward a vivid image of the romance of an era when a large part of scientific research was conducted by “explorer naturalists,” a breed that has become nearly extinct, pushed aside by DNA sequencers, or should we say, “the barcoders of life.” It is hard to put this book away without thinking about all that remains to be discovered in the natural world—the organisms within organisms, the fossils, the remaining new species—because, fortunately, one thing is certain: there is no end to science.
Citation: Vega FE (2005) Explorer naturalists. PLoS Biol 3(5): e161.
Fernando E. Vega is coeditor (with Meredith Blackwell) of Insect-Fungal Associations: Ecology and Evolution, published by Oxford University Press (2005). He lives in Silver Spring, Maryland, United States of America. E-mail: [email protected]
==== Refs
Reference
Nabokov VV Strong opinions 1973 New York McGraw-Hill 335
| 0 | PMC1110913 | CC BY | 2021-01-05 08:21:23 | no | PLoS Biol. 2005 May 17; 3(5):e161 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030161 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1588497910.1371/journal.pbio.0030177Community PageCell BiologyGenetics/Genomics/Gene TherapyEukaryotesThe Epigenome Network of Excellence Community PageAkhtar Asifa Cavalli Giacomo 5 2005 17 5 2005 17 5 2005 3 5 e177Copyright: © 2005 Asifa Akhtar and Giacomo Cavalli.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 initiative funded by the European Union is building a collaborative network of established and younger research groups to tackle key questions in epigenetics.
==== Body
The term “epigenetics” was first proposed by Conrad Waddington to designate the study of the processes by which the genetic information of an organism, defined as genotype, interacts with the environment in order to produce its observed traits, defined as phenotype [1]. More recently, the term has been used to describe heritable changes in genome function that occur without a change in DNA sequence [2]. These two definitions are closer than they seem. In eukaryotic cells, genomic DNA is packaged by histones and non-histone proteins into a dynamic polymer defined as chromatin. Several enzymes can modify the architecture and the composition of chromatin, both locally and globally, and they can direct the inheritance of local chromatin structures through cell division [3,4]. Thus, an individual's cells all share the same linear sequence of DNA nucleotides—the genome—but different cell types are characterized by the presence of different chromatin flavors of this genome—the epigenomes—that specify the characteristic functions of each cell type and allow the maintenance of the memory of these functions through cell division. Indeed, a large set of genome regulatory processes involve epigenetic determination and inheritance.
Because of the vast complexity of epigenetic regulation, ranging from single genes to the folding of whole chromosomes in the nucleus, epigenetics involves an amazingly wide spectrum of techniques and approaches. These involve structural, molecular, and developmental biology, advanced imaging, genomics, proteomics, bioinformatics, and mathematical analysis. Epigenetic research is applied to different organisms, and it often includes comparative analysis from an evolutionary perspective. Researchers in this field must thus cross borders between different research areas in order to understand the impact of novel findings. This makes epigenetics a highly dynamic area of research, one that at the same time provides an ideal ground for collaborative and synergistic interactions.
With the support of the Sixth Framework Programme of the European Union, a consortium of 25 core laboratories and 26 associate laboratories initiated the Epigenome Network of Excellence (NoE) in 2004. This network will tackle key questions in epigenetic research, integrate young research teams in Europe, provide the scientific community with a large spectrum of resources, including Web-based information, and disseminate scientific knowledge to scientists, students, and the wider public. The Epigenome NoE coordinator, Thomas Jenuwein, says “Epigenetic research represents one of the new frontiers in modern biology with considerable impact on many basic mechanisms and on human biology and disease. With this NoE, the European Commission has allowed the affiliated network members to establish a coherent framework for promoting this exciting research area also to non-members and to interested colleagues outside the field. I am convinced that the enthusiastic spirit of the kick-off meeting at the Mendel Abbey in Brno (September 2004) has inspired many positive and long-lasting signals for a highly functional and integrative network all across Europe.”
The Epigenome NoE has four main aims. The first is to advance scientific discoveries via a strong joint programme of activities, of which a large part is dedicated to research carried out by member labs, alone or in collaboration. These projects range from chromatin modification to cell fate and disease and address some of the “big questions” in epigenetic research (Figure 1).
Figure 1 The Impact of Epigenetic Gene Control
Diverse biochemical modifications of DNA and histones, such as DNA methylation (indicated by small hexagons), histone methylation (large hexagons), acetylation (triangles), and phosphorylation (circles), occur in response to the environment and modulate chromatin structure. The organization of chromatin controls the access of many proteins, including transcription factors (ovals), to the DNA template and thus regulates gene expression. This epigenetic gene control has an impact on a variety of biological processes, with implications for agriculture and human biology and disease, including our understanding of stem cells, cancer, and aging.
The second aim is to integrate young researchers into the network via a programme called Newly Established Teams (NET). NET provides funding for research and integration in the joint activity programme to promising junior investigators. Newly established researchers often struggle to acquire lab funding, establish collaborations, or simply gather information on available resources in the field. The NET programme provides an ideal platform for such investigators by generating a collaborative atmosphere and opportunities for setting up joint projects and obtaining top-level biochemical and genomic reagents as well as access to state-of-the-art genomic and proteomic facilities. The more established NoE members also act as mentors for the NET participants. In a first wave of integration, twelve NET investigators were selected by an external advisory board (comprising seven United States scientists who are world leaders in epigenetic research) and participated in the kick-off meeting. A second call to integrate ten more NET members will be launched in 2006, advertised towards the end of this year through the Epigenome NoE Web site (http://www.epigenome-noe.net). This call will complete the NET programme for the five-year funding period of this NoE. If successful, this type of integrated effort may serve as an example for future large-scale collaborative projects. The network is gearing up to maintain its funding in the longer term, so that it can continue attracting future generations of top-level young European scientists to epigenetics. Jacques Remacle, EU Commission Scientific Officer, says “To foster research progress in epigenetics, the Epigenome NoE will have to maintain a dynamic structure involving also the participation and integration of newly established research teams. By its flexible nature, the Network of Excellence instrument allows the consortia to evolve over the time to meet new research challenges and/or to promote openness, expansion and durability. The new 12 NET [investigators] that are joining the Epigenome NoE following an open call (57 applications) and a transparent selection process are all of outstanding quality. I strongly believe that their participation and integration to the Epigenome NoE project will benefit epigenetic research in Europe. Following this successful example, the concept of open call for young investigators is now promoted by several NoEs in other research areas.”
A third aim of the network is to provide the scientific community with a large set of resources and information in epigenetic research, and to centralize access to the information already available from existing sources. The key to this aim is the Epigenome NoE Web site (http://www.epigenome-noe.net). During 2005, the site will grow in its role as a database of technology and resources available in individual labs and institutions. Any scientist external to the network will have the chance to establish collaborations with NoE members and thus to access these resources. Furthermore, the Web site is being augmented with searchable databases of experimental protocols, laboratory profiles, and relevant conferences and courses. Interactive Web forums for discussion of each resource and other hot topics in the field of epigenetics will be opened, and job opportunities will be posted regularly. This Web site is therefore expected to become a major source of information for epigenetic researchers.
A fourth aim of the Epigenome NoE is to have annual meetings and workshops that are organized around specific research topics, which will foster collaborations that are likely to be as important, if not more important, than the funding initiative itself. The NoE also aims at reaching the wider scientific community and the public, bringing the concept and the importance of epigenetics to their attention, and raising the level of awareness of the major impact of epigenetic research in important areas such as agriculture and medicine. This will be achieved by organizing public scientific events, by publishing material written in lay language that reviews specific subjects of particular interest, by setting up a Web information service on epigenetics and genomics, and by maintaining an updated “Frequently Asked Questions” Web site. An education page is already posted, which contains a large collection of Web teaching material on issues related to epigenetics. Educational materials developed in house will be added here in the future. The public science component in this NoE should enable us to communicate both the beauty of the underlying scientific questions as well as the significant medical and ethical relevance of epigenetics to the general public, thus bridging a gap between scientists and society. Genetics is becoming deeply rooted in the cultural foundations of modern societies. Epigenetics is equally important, and thanks to the activities of the Epigenome NoE, this relatively new branch of science may soon become familiar to a much wider audience.
The authors would like to thank S. Grünert and E. Webb for critical discussion during preparation of the manuscript.
Citation: Akhtar A, Cavalli G (2005) The epigenome network of excellence. PLoS Biol 3(5): e177.
Asifa Akhtar is at the European Molecular Biology Laboratory in Heidelberg, Germany. Giacomo Cavalli is at the Institute of Human Genetics, Centre National de la Recherche Scientifique, in Montpellier cedex, France. E-mail: [email protected] (AA), Giacomo. E-mail: [email protected] (GC)
Abbreviations
NETNewly Established Teams
NoENetwork of Excellence
==== Refs
References
Waddington CH The epigenotype Endeavour 1942 1 18 20
Holliday R Epigenetics: An overview Dev Genet 1994 15 453 457 7834903
Turner BM Decoding the nucleosome Cell 1993 75 5 8 8402900
Jenuwein T Allis CD Translating the histone code Science 2001 293 1074 1080 11498575
| 15884979 | PMC1110914 | CC BY | 2021-01-05 08:21:23 | no | PLoS Biol. 2005 May 17; 3(5):e177 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030177 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1588498010.1371/journal.pbio.0030178PrimerNeuroscienceRattus (Rat)Functional Implications of Sleep Development PrimerSiegel Jerome M 5 2005 17 5 2005 17 5 2005 3 5 e178Copyright: © 2005 Jerome M. Siegel.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 Neural Substrates of Infant Sleep in Rats
Infant Sleep: A Precursor to Adult Sleep?
Why do we sleep? The sleep patterns and mechanisms that occur throughout development may give us a clue.
==== Body
Although frazzled new parents may beg to differ, infants do sleep more than adults. This sleep pattern is seen in a wide variety of mammalian species, with some obvious selective advantages. Sleep is a time of reduced body and brain metabolic rate [1,2], allowing energy conservation, particularly if a warm place is available, as can be provided by a compliant parent or sibling. The sleeping, quiescent infant is also less likely to attract predators and is easier to transport. At the earliest ages, infants who have not yet opened their eyes and whose cortex is not yet developed have limited learning opportunities from interactions with the outside world: another reason for reduced waking.
But sleep comes in many forms. Evolutionary arguments may make sense for slow-wave deep sleep patterns at birth that are associated with a general shutdown of the brain, but may not provide such an obvious explanation for the relative predominance of rapid eye movement (REM) sleep. For example, in human neonates REM sleep constitutes approximately eight hours per day, or 50% of the total sleep time, whereas human adults devote less than two hours per day, or 20% of their seven to eight hours of sleep time, to REM sleep [3]. REM sleep is characterized by high brain metabolic and neuronal activity rates [4], reduced muscle tone, irregular and relatively automatic respiration uncoupled from its usual regulatory mechanisms [5], and diminished thermoregulation [6]. These properties seem maladaptive, which suggests that there must be some compensatory survival benefit for REM sleep to have persisted. Could REM sleep play a particularly important role in development?
Interesting evidence for this hypothesis has come from studying the effects of REM sleep deprivation on the development of the visual system. It is known that the occlusion of one eye during the maturation of visual connections that occurs after birth causes the open eye to acquire more central connections than the closed eye. This disproportionate representation seems to result from a difference of activity in the optic nerve between the open eye and closed eye [7]. Although early ideas that REM sleep was necessary for brain plasticity might suggest that REM sleep deprivation would prevent this reorganization, just the reverse occurs. REM sleep deprivation accelerates the shift of connections to favor the open eye [8,9]. Rather than facilitating change [10], REM sleep may therefore be a source of endogenous activity that tends to prevent altered sensory stimulation from causing abnormal connections to form. REM sleep may prevent the programmed cell death and the pruning of connections that occurs when critical synapses are not stimulated.
Another possible role for neonatal REM sleep might be in thermoregulation of the growing brain. It is known that nonREM sleep tends to cool the brain, reducing its thermoregulatory set point [11]. In contrast, REM sleep tends to heat certain brain regions [12]. The nonREM–REM alternation comprises a thermoregulatory oscillation.
It is often assumed that the amount of time spent in different sleep states is determined by processes controlled by the cerebral cortex. The emphasis on the cortical role in sleep may result more from the technical ease of recording electroencephalograms from the cortex than from persuasive functional evidence. At birth, cortical metabolism and neuronal firing are minimal [13], yet this is the time of greatest sleep. In adults, damage to the cortex produces little or no change in sleep, indicating that the signal for sleep does not originate in or at least does not require the cortex [14]. Animals with proportionally larger cortices do not have more REM or nonREM sleep time than animals with relatively little cortex [15]. The effects of long-term sleep deprivation have been shown to be largely autonomic in nature, including elevated body temperature, skin lesions, and increased food intake [16]. Such effects cannot be duplicated by any cortical lesions. However, many of these symptoms appear to be consistent with hypothalamic dysfunction [17,18].
Evolutionary evidence also suggests that the cortex may be a relatively recent participant in REM sleep. Plesiomorphic (primitive) mammals such as the egg-laying echidna and platypus have large amounts of REM-sleep-like activity in brainstem structures at birth [19,20]. The brainstem is the key region for REM sleep generation, being both necessary and sufficient for its occurrence [4]. However, the cortex of these animals scarcely changes activity during these states, showing slow-wave patterns during the REM sleep state. In this respect the sleep of placental mammals may represent ontogeny recapitulating phylogeny, since a reduction in electroencephalogram power is a late-developing component of REM sleep.
A prominent feature of REM sleep is the rapid eye movements and associated twitches that define the state. These are particularly marked and vigorous in neonates. It has been shown that twitches with some resemblance to REM sleep activity are present in the isolated spinal cord of neonates and diminish in the transected cord of older animals [13]. This has suggested to some that a primal phasic activity of the central nervous system transforms postnatally over an extended time period into the very different brainstem-generated pattern seen in adults. But in this issue of PLoS Biology, Karlsson et al. [21] show that this is not the case. In a set of technically demanding experiments, they demonstrate a remarkable similarity between sleep control mechanisms in the one-week-old rat and those in the adult cat, and by implication throughout the mammalian line.
By severing the connections to and from the forebrain (cerebral cortex and associated structures), Karlsson et al. were able to study sleep-related activity in the midbrain and brainstem. They described the rat homologs of the medullary neurons that induce the atonia seen in sleeping adult cats and narcoleptic dogs [22,23,24] (Figure 1). More rostrally, they identified neural activity in the region of the locus coeruleus that facilitates movement and report contrasting inhibitory activity in the adjacent subcoeruleus region, again paralleling studies in the cat [25,26,27]. They also found cells that appear to generate or at least contribute to the twitches of REM sleep.
Figure 1 Model of Some of the Major Systems Involved in Regulating Muscle Activity in REM Sleep Drawn on a Sagittal Section of the Brainstem
Cholinergic (ACh) neurons in the pons, which are under the inhibitory control of noradrenergic (NE) and serotonergic (5-HT) neurons, trigger REM sleep. They activate descending glutamatergic neurons, which in turn activate glycinergic and GABAergic neurons. Other glycinergic interneurons in the spinal cord are also activated by unknown descending inputs. The release of glycine and GABA inhibits motoneurons. The descending glutamatergic pathway also activates GABAergic interneurons, which inhibit noradrenergic and serotonergic neurons. The reduction in norepinephrine and serotonin release during REM sleep disfacilitates motoneurons. Descending glutamatergic neurons that connect directly to motoneurons produce phasic excitation during REM sleep. The net result of the action of this network is an absence of muscle tone in the “antigravity” muscles in REM sleep, interrupted by twitches (see text for references).
The similarities to the adult cat's REM sleep control mechanisms are so striking that what becomes interesting are the small differences that are reported. The locus coeruleus REM “sleep-off cells,” which are active in waking, reduce activity in nonREM sleep, and cease activity in REM sleep, appear to not have long-duration waveforms in the neonatal animals examined by Karlsson et al., unlike the case of the adult rat and cat [28,29]. Another difference is the apparent absence of the cessation of dorsal raphe (serotonin) unit discharge in REM sleep. Although the authors speculate that this is due to the absence of forebrain connections in their experimental preparation, it has been shown that forebrain mechanisms are not necessary for this cessation of raphe activity in adult cats [30]. However, identification of the narrow dorsal raphe nucleus is difficult even in adult cats, and it is certainly possible that these neurons were overlooked in the neonatal rat.
The upshot of these findings is a picture of a largely mature REM sleep generator mechanism at birth. The developmental progression of REM sleep signs, particularly the reduction in sleep duration and the development of the characteristic reduction in electroencephalogram voltage to a waking-like pattern in REM sleep, may result from the maturation of the targets of these brainstem systems, the modulation of these generator mechanisms by developing systems, or a relatively subtle maturing of connections within the REM sleep generator systems. This work pushes the probable organization of the REM sleep generator system in rats back to before one week of age, possibly to an in utero stage.
What does all this say about the function of REM sleep? Although we are left with the same initial speculations, the neonatal model provides a different perspective for approaching these functions. It is particularly useful to know that key elements of the REM sleep system are present in neonatal rats, since these animals are ideal subjects for in vitro studies of tissue slices [31,32]. It is not practical to perform in vitro experiments on the adult brainstem. However, there has always been some question as to whether studies of neonatal brainstems would be applicable to the question of adult REM sleep mechanisms. One can now imagine examining the metabolism and membrane characteristics of these critical cell groups as a means of gaining better insight into REM sleep function. However, as Karlsson et al.'s work demonstrates [21], most of the neurons of interest are not homogenously concentrated in any easily targeted region. Identifying the individual neurons of interest in vitro remains a challenge. This challenge will have to be surmounted in order to identify the control mechanism and better understand the function of REM sleep.
Research was supported by the Medical Research Service of the Department of Veterans Affairs and National Institutes of Health grants NS14610, HL41370, MH64109, and HL060296.
Citation: Siegel JM (2005) Functional implications of sleep development. PLoS Biol 3(5): e178.
Jerome Siegel is at the Veterans Administration Greater Los Angeles Health Care System in Sepulveda, California, and the Department of Psychiatry at the University of California, Los Angeles, California, United States of America. E-mail: [email protected]
Abbreviation
REMrapid eye movement
==== Refs
References
Nofzinger EA Buysse DJ Germain A Price JC Miewald JM Functional neuroimaging evidence for hyperarousal in insomnia Am J Psychiatry 2004 161 2126 2128 15514418
Maquet P Sleep function(s) and cerebral metabolism Behav Brain Res 1995 69 75 83 7546320
Roffwarg HP Muzio JN Dement WC Ontogenetic development of the human sleep-dream cycle Science 1966 152 604 619 17779492
Siegel JM Nottebohm F Kryger MH Roth T Dement WC Brainstem mechanisms generating REM sleep Principles and practice of sleep medicine 2000 Philadelphia W. B. Sanders 112 133
Baker T McGinty D Reversal of cardiopulmonary failure during active sleep in hypoxic kittens: Implications for sudden infant death Science 1977 199 419 421
Parmeggiani PL Azzaroni A Calasso M Systemic hemodynamic changes raising brain temperature in REM sleep Brain Res 2002 940 55 60 12020875
Wiesel TN Early explorations of the development and plasticity of the visual cortex: A personal view J Neurobiol 1999 41 7 9 10504186
Shaffery JP Roffwarg HP Speciale SG Marks GA Ponto-geniculo-occipital-wave suppression amplifies lateral geniculate nucleus cell-size changes in monocularly deprived kittens Brain Res Dev Brain Res 1999 114 109 119 10209248
Shaffery JP Oksenberg A Marks GA Speciale SG Mihailoff G REM sleep deprivation in monocularly occluded kittens reduces the size of cells in LGN monocular segment Sleep 1998 21 837 845 9871946
Siegel JM The REM sleep-memory consolidation hypothesis Science 2001 294 1058 1063 11691984
McGinty D Szymusiak R Keeping cool: A hypothesis about the mechanisms and functions of slow-wave sleep Trends Neurosci 1990 13 480 487 1703678
Wehr TA A brain-warming function for REM sleep Neurosci Biobehav Rev 1992 16 379 397 1528526
Corner MA McGinty DJ Drucker-Colin R Morrison A Parmeggiani PL Ontogeny of brain sleep mechanisms Brain mechanisms of sleep 1985 New York Raven Press 175 198
Villablanca J Marcus R Sleep-wakefulness, EEG and behavioral studies of chronic cats without neocortex and striatum: The ‘diencephalic’ cat Arch Ital Biol 1972 110 348 382 4349190
Zepelin H Siegel JM Tobler I Kryger MH Roth T Dement WC Mammalian sleep Principles and practice of sleep medicine 2005 Philadelphia Elsevier Saunders 91 100
Rechtschaffen A Bergmann BM Sleep deprivation in the rat: An update of the 1989 paper Sleep 2002 25 18 24 11833856
Everson CA Crowley WR Reductions in circulating anabolic hormones induced by sustained sleep deprivation in rats Am J Physiol Endocrinol Metab 2004 286 E1060 E1070 14871886
Everson CA Nowak TSJ Hypothalamic thyrotropin-releasing hormone mRNA responses to hypothyroxinemia induced by sleep deprivation Am J Physiol Endocrinol Metab 2002 283 E85 E93 12067847
Siegel JM Manger P Nienhuis R Fahringer HM Pettigrew J The echidna Tachyglossus aculeatus combines REM and nonREM aspects in a single sleep state: Implications for the evolution of sleep J Neurosci 1996 16 3500 3506 8627382
Siegel JM Manger PR Nienhuis R Fahringer HM Shalita T Sleep in the platypus Neuroscience 1999 91 391 400 10336087
Karlsson K.Æ Gall AJ Mohns EJ Seelke AMH Blumberg MS The neural substrates of infant sleep in rats PLoS Biol 2005 3 e143 10.1371/journal.pbio.0030143 15826218
Sakai K McGinty DJ Drucker-Colin R Morrison AR Parmeggiani PL Anatomical and physiological basis of paradoxical sleep Brain mechanisms of sleep 1985 New York Raven Press 111 138
Siegel JM Wheeler RL McGinty DJ Activity of medullary reticular formation neurons in the unrestrained cat during waking and sleep Brain Res 1979 179 49 60 228803
Siegel JM Nienhuis R Fahringer H Paul R Shiromani P Neuronal activity in narcolepsy: Identification of cataplexy related cells in the medial medulla Science 1991 262 1315 1318
Kodama T Lai YY Siegel JM Changes in inhibitory amino acid release linked to pontine-induced atonia: An in vivo microdialysis study J Neurosci 2003 23 1548 1554 12598643
Lai YY Kodama T Siegel JM Changes in monoamine release in the ventral horn and hypoglossal nucleus linked to pontine inhibition of muscle tone: An in vivo microdialysis study J Neurosci 2001 21 7384 7391 11549748
Mileykovskiy BY Kiyashchenko LI Kodama T Lai YY Siegel JM Activation of pontine and medullary motor inhibitory regions reduces discharge in neurons located in the locus coeruleus and the anatomical equivalent of the midbrain locomotor region J Neurosci 2000 20 8551 8558
Aston-Jones G Bloom FE Activity of norepinephrine-containing locus coeruleus neurons in behaving rats anticipates fluctuations in the sleep-waking cycle J Neurosci 1981 1 876 886 7346592
Hobson JA McCarley RW Wyzinski PW Sleep cycle oscillation: Reciprocal discharge by two brainstem neuronal groups Science 1975 189 55 58 1094539
Hoshino K Pompeiano O Selective discharge of pontine neurons during the postural atonia produced by an anticholinesterase in the decerebrate cat Arch Ital Biol 1976 144 244 277
van den Pol AN Ghosh PK Liu RJ Li Y Aghajanian GK Hypocretin (orexin) enhances neuron activity and cell synchrony in developing mouse GFP-expressing locus coeruleus J Physiol 2002 541 169 185 12015428
Greene RW Gerber U McCarley RW Cholinergic activation of medial pontine reticular formation neurons in vitro Brain Res 1989 476 154 159 2914210
| 15884980 | PMC1110915 | CC BY | 2021-01-05 08:21:23 | no | PLoS Biol. 2005 May 17; 3(5):e178 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030178 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1588497310.1371/journal.pbio.0030145FeatureEcologyEvolutionScience PolicyNoneSouth Africa—Serious about Biodiversity Science FeatureCherry Michael 5 2005 17 5 2005 17 5 2005 3 5 e145Copyright: © 2005 Michael Cherry.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.A new government act, the creation of several centres of excellence and an injection of funding all indicate that biodiversity science is thriving in South Africa.
==== Body
In 1772, Carolus Linnaeus wrote a letter, now oft-quoted, to Ryk Tulbagh, the Governor of the Cape—in which he envied Tulbagh's “sovereign control of that Paradise on Earth, the Cape of Good Hope, which the beneficent Creator has enriched with his choicest wonders”. Two and a half centuries later, South Africa's biodiversity remains a great source of interest to the scientific community—and for good reason (Box 1). Plant biodiversity, with over 20 000 different species, is in the foreground: South Africa, which comprises less than 1% of the world's land surface, contains 8% of its plant species. Perhaps less well known is that the country also contains 7% of all bird, mammal, and reptile species, and 15% of known coastal marine species.
Box 1. Biodiversity and the South African Economy
The extraordinary diversity of habitats found on the southern tip of the African continent includes three globally recognized biodiversity hotspots: the temperate Cape Floristic Region (see Figure 1), the arid Succulent Karoo, and the subtropical Maputaland-Pondoland-Albany area. On account of its early colonization and relative wealth, South Africa has good universities, museums, and herbaria, and reasonably well-run conservation agencies at both the provincial and national levels. But in a country whose history has been characterized by fighting over land, the 6.6% of its land surface with formal conservation status (in other words, protected by the state) lags behind the global mean of 11.5%. By contrast, 17% of its coastline is formally protected.
Protection is important for a number of reasons. A decade after the advent of democracy, the economy is booming at last, with the country currently experiencing the longest sustained period of growth in its history since the early 1960s. Rising levels of affluence have led to increased demand for housing, roads, and recreational facilities—all developments that affect biodiversity. The benefits that biodiversity brings to the economy are increasingly being realized, notably through ecotourism. Tourism is the fastest growing sector of the economy, having risen to 7% of GDP in 2003, from only 2% a decade previously. The virtual abandonment of agriculture subsidies has led to much marginal agricultural land—previously farmed essentially to generate subsidy—being converted to private nature reserves, used either for ecotourism or for hunting, and sometimes for both. Such land now comprises 13% of the country's surface—more than twice the area protected by the state.
There are also direct benefits associated with harvesting indigenous flora and fauna. Some are quantifiable, such as the fishing industry, worth just over half a billion US dollars last year. Others cannot be measured directly, but are no less important for that. For example, almost 20% of South Africa's plants, or 3 689 species, are used as traditional medicines, which still provide the first resource for primary health care to almost three-quarters of South Africa's population. The challenge of sustainable harvesting is difficult enough when yields are known, but even more daunting when they are undetermined.
The South African National Biodiversity Institute
South Africa's new Biodiversity Act, signed on September 1, 2004, expands the mandate of the National Botanical Institute (NBI) to include responsibilities relating to the full diversity of the country's fauna and flora; it is now known as the South African National Biodiversity Institute (SANBI) (Pretoria, South Africa). Previously responsible for eight national botanical gardens and three herbaria, as well as botanical research centres in Pretoria and at its largest garden at Kirstenbosch on the slopes of Table Mountain, it now additionally should influence the prospects of all collections of specimens; coordinate research on indigenous biodiversity and its sustainable use; advise conservation agencies and municipalities with regard to planning decisions relating to biodiversity; coordinate the control of invasive species; and monitor the effect of any genetically modified organisms released into the environment.
“There are very few developing countries which have the prospect of delivering jobs related to the conservation industry. South Africa has this prospect.”
Acting Chief Executive Officer Brian Huntley (Figure 2) admits openly that this is quite a brief. It's not difficult to see why it is the former NBI that has inherited this mantle, since it has become, over the past decade, by far the largest and most dynamic South African institution working in the biodiversity arena. Operating under the aegis of the Department of Environment Affairs, it was formed in 1989 through the amalgamation of what had previously been the National Botanical Gardens and the Botanical Research Institute. Currently supporting 680 staff, it has flourished particularly during Huntley's tenure, which has been characterized by an influx of externally funded projects, to the extent that external income —$18 million per annum—now exceeds the $16 million it receives from its parliamentary grant and from entrance fees paid by the million or so visitors to its gardens each year. Huntley is optimistic that this brief can succeed, although he concedes that in few countries does any single institution bear responsibility for research, information dissemination, and applications relating to biodiversity. But he believes that South Africa is a small enough country, with enough good intellectual capacity, for this model to work.
Figure 2 Brian Huntley, Acting Chief Executive of the South African National Biodiversity Institute (SANBI)
This view is echoed by Andrew Balmford of Cambridge University, who is spending a sabbatical at the Percy Fitzpatrick Institute for African Ornithology at the University of Cape Town. “While the obvious challenge is to link biodiversity conservation to development needs”, he says, “there are very few developing countries which have the prospect of delivering jobs related to the conservation industry. South Africa has this prospect, not only because it is unbelievably diverse, but because of international goodwill towards the country”.
Huntley's strategy will be to bring a sound scientific base to the enterprise, as he has already done with the NBI. There are several examples of this. One is the African Plants Initiative—being led by the SANBI, Kew Gardens in the United Kingdom (London), and the United States Missouri Botanical Garden (St. Louis, Missouri, United States)—whose aim is to create an electronic library of the type specimens of all African plants: an estimated 300 000 accessions of 60 000 species. This includes scanned pictures of each individual specimen, the quality of which, according to Huntley, “is as good as if one were examining the specimen through a standard dissecting microscope.” Another example involves placing the 2.5 million specimens in South Africa's herbaria on a computerized database, a task now 40% complete. A third example is the Southern African Botanical Diversity Network (Pretoria, South Africa), founded in 1996, which has, to date, trained 200 botanists in ten countries in the region.
By contrast, research on zoological diversity, traditionally the domain of the country's natural history museums, has lagged behind. The Iziko South African Museum in Cape Town, for example, one of the country's five national natural history museums, now has only seven research staff in natural history compared to the 16 it had in 1989. Why have they failed to capitalize on external funding in the way the NBI has done? One answer is that, unlike the three national herbaria, which all fell under the jurisdiction of the NBI, these five institutions have retained their institutional autonomy, and consequently have remained fragmented in their efforts. One, the South African Institute for Aquatic Biodiversity (Grahamstown, South Africa), is run by the National Research Foundation, while the other four are funded by grants from the Department of Arts and Culture, which has tended to view them as educational, rather than research, organizations. Huntley emphasizes that the SANBI does not aspire “to do what other organizations are already doing well.” With regard to natural history museums, he says that the first step will be to take the initiative in conducting a thorough review of their funding, and the “best practice of dealing with large and dispersed collections”.
Centres of Excellence in Biodiversity
Another related development is the recent announcement of the Department of Science and Technology that it will fund six centres of excellence nationally at South African universities, with effect from 2005. No fewer than three of these centres are focused on biodiversity: one at the Fitzpatrick Institute (Cape Town, South Africa), concerned with birds as models for understanding biodiversity processes; one at the University of Pretoria (Pretoria, South Africa), which will be concerned with pathogens on indigenous trees; and a third in the Faculty of Science at the University of Stellenbosch, which will focus on invasion biology. All of the centres are based at the host institution, but can disburse funds to collaborators elsewhere in the country.
“These centres of excellence...are a manifestation of the seriousness with which the South African government is taking science”
These centres of excellence, says Steven Chown (Figure 3), director of the Centre for Invasion Biology, “are a manifestation of the seriousness with which the South African government is taking science”. Others are more sceptical. “I don't think that in the biodiversity field research is optimally conducted by large groups, but by smaller groups of collaborators”, says David Ward of the School of Biological and Conservation Science at the University of KwaZulu-Natal (KwaZulu-Natal, South Africa). “Unlike fields like nuclear physics, in ecology costs are relatively low—large centres just incur additional administrative costs, without improving the quality of the science produced”, he adds. Rob Slotow, from the same school, feels that the centres have confined their collaborative efforts to junior colleagues outside their own institutions. “There is very little real inter-institutional collaboration taking place at a senior level, which is disappointing”, he says, as “the opportunity to kick-start a different level of funding for biodiversity research in the country—the aim of the centres-of-excellence concept—has been missed”.
Figure 3 Steven Chown, Director of the Centre for Invasion Biology
The research programme of the Fitzpatrick Institute is based on two interlinking themes: understanding and maintaining avian biodiversity. Tim Crowe will lead a group investigating the processes responsible for the origins of African biodiversity, which will investigate how the process of speciation in birds occupying disjunct distributions in habitats ranging from montane forest to desert may have been influenced by past biogeographic corridors that shifted with changing climates many millions of years ago.
Understanding how relationships between organisms and their environments influence the form and functioning of biological systems is the core focus of a second grouping of researchers. For example, Phil Hockey is studying life-history traits and movement patterns of African Black Oystercatchers (Figure 4A and 4B), where research to date indicates that migration in juveniles is facultative, responding to body condition. Oystercatcher populations are increasing, primarily as a consequence of a ban imposed several years ago on four-wheel-drive vehicles on beaches. These increasing populations provide a unique opportunity to test the hypothesis that migration in stable habitats evolves—initially in juveniles—in response to population density exceeding carrying capacity.
Figure 4 African Black Oystercatchers
(A) Portrait of an adult African Black Oystercatcher. (Photo: Philip Hockey)
(B) An African Black Oystercatcher chick with the numbered colour rings that are used to follow its survival and migratory movements over several years. (Photo: Doug Loewenthal)
The Fitzpatrick Institute's teaching efforts have been impressive; its master's course in conservation biology has produced close to 150 graduates from 15 different African countries since its inception in 1992. But are birds really good models for understanding changes in patterns of biodiversity? Many would argue that they are not, since their mobility allows them to respond to environmental changes by colonizing new areas with relative ease. The centre's director, Morné du Plessis, counters that “while birds are often not good indicators of environmental change, they are a group for which good baseline information exists, as well as being relatively easy to study.”
The centre for Tree Health Biotechnology forms part of the Forestry and Agriculture Biotechnology Institute at the University of Pretoria. The institute has to date focused largely on pathogens on trees used in commercial forestry, most of which are alien species, but according to its director, Mike Wingfield, the centre will be devoted specifically to studying pathogens on indigenous trees. But the two, he adds, are closely related. “Alien trees used in commercial forestry are often able to thrive because they are distanced from their natural pathogens”, he says, but “we are now observing natural pathogens of native trees switching hosts to alien species”. This happens usually when alien and native trees are reasonably closely related. Wingfield believes that an example is the fungus causing Cryphonectria canker, which he and his collaborators have recently shown, on the basis of DNA sequences, to occur on both the native waterberry tree, Syzgium cordatum, and on the alien Eucalyptus (widely used for forestry in South Africa), from which it was first reported in 1989.
Similarly, native trees, which are often of importance to local communities, could be at risk from pathogens imported on alien species. The kiaat tree Pterocarpus angolensis, for example, is widely used by wood-carvers, as well as in traditional medicines. Trees are reported to be dying, but it is unknown whether this is on account of pathogens, climate change, or changing fire regimens. There have also been sporadic reports of dying baobabs (Figure 5)—one of the icons of the African savannah—over the past 15 years. Wingfield says that “while at the present time, there is no clear evidence that an unknown fungus is killing baobabs, these reports should not go unheeded”. “Both kiaat and baobab deaths merit attention, which the centre should now be able to provide,” says Wingfield.
Figure 5 African Baobab
The African baobab, Adansonia digitata, can reach up to 10 meters in diameter and can live more than 1 000 years.
(Photo: Peter Jones)
The Centre for Invasion Biology is somewhat different from the other two centres in that it focuses on a specific question, namely how invasions affect biodiversity and ecosystem functioning. Of its annual budget of $1 million, five-sixths will come from the government, with the remainder being provided by the University of Stellenbosch. Chown believes that this is a bargain, considering the magnitude of the problem: the global cost of addressing biological invasions is estimated to be $1.4 trillion annually—about 5% of global GDP.
In South Africa, invasive plants are a particular problem. Apart from the threat they pose to indigenous diversity, they are a fire hazard in several ecosystems in which burning is part of the natural cycle—and perhaps most importantly, they are a huge drain on water in a country in which this is a scarce commodity. This has led to a programme—certainly the largest of its kind in any developing country—called Working for Water, in which unemployed persons have been hired to conduct alien clearances on a large scale. Chown's centre will provide policy inputs to the programme as part of a broad range of pure and applied research objectives.
The Centre for Invasion Biology will address both long-term studies of invasive organisms in different habitats and the outcomes of remediation programmes, which Chown views as large-scale ecological experiments whose effects need to be studied. “The Working for Water rehabilitation programme provides excellent opportunities for understanding relationships between changes in species richness and changes in ecosystem function, and how alien invasion and clearance affects both phenomena”, he says.
A second component will attempt to study invasions from the outset, as opposed to post hoc. Chown proposes to investigate concomitant climate and land-use changes in the Cedarberg mountains, a range 200 km north of Cape Town. The predominant land-use patterns of agriculture and ecotourism are changing rapidly in the area, which he predicts will be accompanied by changes in the extent and identity of invasive species. Additionally, climate-change models predict that this relatively arid part of the fynbos biome (the major vegetation type of the Cape Floristic Region) will be transformed within 50 years into a semi-desert system.
To what extent will these different ventures find a common purpose? There are some obvious links: Huntley sits on the board of the Fitzpatrick Institute, whose master's course in conservation biology has supplied many graduates to the NBI over the past 15 years. Chown sits on the board of the SANBI, together with representatives of the departments of Science and Technology, Agriculture and Environment Affairs, and David Mabunda, chief executive officer of South African National Parks. As the chief executive of the SANBI now exercises a huge degree of statutory influence over the nation's biodiversity, the answer to this question is closely related to that of who will replace Huntley, who is now 61. Huntley's tenure will be a hard act to follow, and the future of South Africa's biodiversity will lie largely in the hands of his successor.
Figure 1 Cape Flowers in August
(Photo: Peter Jones)
Where to Find Out More
South African National Biodiversity Institute (SANBI): http://www.nbi.ac.za/
Percy Fitzpatrick Institute of African Ornithology: http://www.fitzpatrick.uct.ac.za/
DST Centre of Excellence for Invasion Biology: http://academic.sun.ac.za/cib/
Centre of Excellence in Tree Health Biotechnology: http://fabinet.up.ac.za/CoE/
Working for Water Programme: http://www-dwaf.pwv.gov.za/wfw/
Citation: Cherry M (2005) South Africa—Serious about biodiversity science. PLoS Biol 3(5): e145.
Michael Cherry is an associate professor in the Department of Botany and Zoology at the University of Stellenbosch in Stellenbosch, South Africa. E-mail: [email protected]
Competing Interests: The author has declared that he has no competing interests.
Abbreviations
NBINational Botanical Institute
SANBISouth African National Biodiversity Institute
| 15884973 | PMC1110916 | CC BY | 2021-01-05 08:21:24 | no | PLoS Biol. 2005 May 17; 3(5):e145 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030145 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1588497610.1371/journal.pbio.0030164PrimerAnimal BehaviorNeuroscienceSongbirdThe Neural Basis of Birdsong PrimerNottebohm Fernando 5 2005 17 5 2005 17 5 2005 3 5 e164Copyright: © 2005 Fernando Nottebohm.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.
Vocal Experimentation in the Juvenile Songbird Requires a Basal Ganglia Circuit
To a Zebra Finch: How the Brain Cultivates Birdsong
Songbirds represent an excellent model system for understanding the neural mechanisms underlying learning.
==== Body
There is a tradition in biology of using specific animal models to study generalizable basic properties of a system. For example, the giant axon of squid was used for the pioneering work on nerve transmission; the fruit fly (Drosophila) has played a key role in researchers discovering the role of homeobox genes in embryogenesis; the sea slug (Aplysia) is used to study the molecular biology of learning; and the round worm (Caenorhabditis elegans) is used to study programmed cell death. Basic insights gained from these four systems apply widely to other multicellular animals. Here, I will review basic discoveries made by studying birdsong that have helped answer more general questions in vertebrate neuroscience.
Vocal Learning and the “Song System”
Oscine songbirds (e.g., zebra finches, canaries, and white-crowned sparrows) learn their song by imitating those of older members of their own species [1,2]. This is done by modifying vocal output until the auditory feedback it generates matches a memorized model [3]. In some birds vocal learning gives rise to easily discernible song dialects, which then act as local cultural traditions [4]. In most songbirds mastery of a song model takes many weeks. Song learning starts with a stage that has been likened to human infant babbling called “subsong,” during which highly variable, low-amplitude sounds are produced in a non-communicatory context, often while the juvenile seems to doze. The sounds of subsong provide the raw material from which imitations emerge. As these imitations become recognizable, they are referred to as “plastic song.” As the imitations are perfected, song becomes less and less variable. The stable song typical of adults is in place by the time the sexually mature bird is ready to start to defend a territory and woo a mate. Intriguingly, in birds as in human infants, the path of vocal change that culminates with imitation of a model can be very idiosyncratic, as if this were an exercise in problem solving for which there is no single solution [5].
The acquisition and production of learned song is made possible by a group of discrete brain nuclei and their connecting pathways, referred to as the “song system” [6,7], which has similarities in the three groups of birds—songbirds, parrots, and hummingbirds—that evolved learned song [8,9]. This system, described in considerable detail in oscine songbirds, has two main branches: the posterior descending pathway (PDP), necessary for both the acquisition and production of learned song, and the anterior forebrain pathway (AFP), necessary for acquisition only (see Figure 1). The high vocal center (HVC) is at the starting point of both these pathways, but the HVC cells that project to the PDP and AFP differ. In mammalian terms the PDP is homologous to a motor pathway that starts in the cerebral cortex and descends through the brain stem [6], while the AFP is homologous to a cortical pathway through the basal ganglia and thalamus [7,10,11].
Figure 1 The Song System of Songbirds
Nucleus HVC feeds information into two pathways that ultimately lead to the neurons in the tracheosyringeal half of the hypoglossal nucleus (nXIIts) that project to vocal muscles. HVC projects to nucleus RA directly (PDP), and indirectly via Area X, the dorsolateral anterior thalamic nucleus (DLM), and LMAN (AFP) in a manner that shares similarities with the mammalian pathway cortex→basal ganglia→thalamus→cortex.
Several of the telencephalic nuclei that participate in the production and acquisition of learned song are small in nestlings, before the onset of song development, and their volume, cell number, cell size, and connections grow during the subsequent weeks or months. As a result of these changes, many of the components of the circuits for the acquisition and production of learned song are formed and connected during the very period when song first develops (reviewed in [12]). Another peculiarity of this system is that the right and left sides of the brain can operate, to some extent, independently, each responsible for a different array of sounds. In birds such as the canary, the chaffinch, and the white-crowned sparrow a majority of the sounds of song are produced by the left syringeal half, under the control of left, uncrossed pathways. This phenomenon has been referred to as “left hypoglossal” or “left hemispheric” dominance [13].
Adult Variation, Neurogenesis, and Neuronal Replacement
The song system of birds is sexually dimorphic: it is better developed in males, which usually sing more and produce a more complex repertoire than females. For example, the nucleus HVC of canaries is three times larger in males than in females; in zebra finches it is eight times larger [14]. In seasonal singers such as the canary and song sparrow, song system nuclei such as HVC are significantly larger in the spring than in late summer, after breeding stops [15,16]. Cells in many of the song control nuclei are androgen and estrogen sensitive [17]. The nucleus HVC of adult female canaries treated with physiological doses of testosterone doubles in volume, and these birds start to sing in a male-like manner [18]. Initially such changes were thought to result solely from growth of dendritic trees and synapse formation [19], but subsequently it was found that new neurons were added, too. These new neurons, as in embryos, are born in the wall of the forebrain's lateral ventricle [20,21]. Interestingly, the addition of new neurons to nucleus HVC occurred also in male and female adult canaries that had received no hormonal treatment [22].
Earlier claims of neurogenesis in the adult mammalian brain [23,24] had met resistance [25]. Nucleus HVC yielded the first unambiguous example of adult neurogenesis, in that individual cells labeled with a cell birth marker provided neurophysiological recordings that were unmistakably neuronal [26]. We now know that the recruitment of new HVC neurons is part of a process of constant replacement [27,28]. This replacement is particularly active in canaries during seasonal changes in the song repertoire [29]. Male canaries develop a new song repertoire each year. New neurons are constantly added, as well, to many regions of the adult avian telencephalon, where they are probably involved in a variety of brain functions. There is no evidence of neuronal addition to other parts of the adult songbird brain [22]. We now know that adult neurogenesis and neuronal replacement are probably common to all vertebrates [30]. The song system of birds helped change the way in which we think of brain circuits and their potential for rejuvenation and repair. Just as important, the discovery of neuronal replacement has raised basic questions about the brain variables that set limits to learning [31].
Neurophysiology Offers Insights on the Mechanisms for Vocal Learning
From early on, the song system drew the attention of neurophysiologists. It was known that lesions of HVC and the robust nucleus of the arcopallium (RA) affected the organization of song differently, the former being more devastating than the latter [6]. Likewise, stimulation of HVC during song interrupted and reset the song program, something that did not happen if the stimulating electrode was in RA [32]. This hierarchical relation between HVC and RA was confirmed by recording from HVC and RA while the bird sang [33], but the manner in which the sounds of song were represented in HVC remained unclear. This issue was resolved by recording from individual HVC neurons that projected to nucleus RA. These neurons, it was shown, fired very sparsely and at narrowly defined times, each neuron firing always during the same six-millisecond window while the bird produced its single learned song [34]. The inference that these neurons—and the PDP of which they are part—carried the learned pattern of song seems inescapable. Since these are the very HVC neurons that are replaced when birds modify their song [31], it follows that the replacement cells learn their score.
But what, then, is the role of the AFP in song learning? It was known that the AFP was necessary for the acquisition but not for the production of learned song [35]. It was known, too, that the variable song typical of juvenile songbirds became very stereotyped after bilateral lesions of the lateral magnocellular nucleus of the nidopallium (LMAN) (part of the AFP), from which it was inferred that LMAN played a crucial role in fostering circuit plasticity necessary for learning [36]. But the mechanism for this effect remained unknown. Two recent independent studies now show how this effect comes about [37,38]. In this issue of PLoS Biology, Ölveczky and colleagues show that the LMAN neurons that project to RA fire in a quasi-random pattern when variable song is produced in juvenile birds. Thus, while the HVC→RA projection carries the learned song, the LMAN→RA projection carries the jitter that induces the variability in motor output necessary for the imitation of a model. This jitter, presumably, is imposed on the firing of the same RA neurons that receive the more orderly output from HVC. When the LMAN neurons are silent (or absent), the HVC→RA pathway produces a stereotyped pattern; when the LMAN→RA neurons are firing, song is more variable. This is a most elegant breakthrough. However, it is not clear exactly how this pathway functions in song learning. One possibility is that birds trying to imitate a model succeed by retaining, from the variability generated, those patterns that more closely approximate the model and discard the rest, thus, over a period of time, achieving a perfect imitation. A second possibility is that the variable mismatch between a model and the attempted imitation drives output modification, so that patterns that had not occurred before now first appear. Both mechanisms would depend on auditory feedback. It is the first time we are so close to a mechanism for vocal learning.
Open Questions
The stage is set for many more insights. An unsolved question is the extent to which the very pathways that produce learned song may also partake in the perception of song. On a different front, why is it that HVC→RA neurons are periodically replaced? It was thought that changes in dendritic configuration, dendritic spines, and synaptic number and efficiency provided all the plasticity needed to change circuit configuration and explain how new information was acquired and remembered. But if so, why replace whole neurons? Is it possible that dendritic and synaptic changes underlying learning are less easy to achieve in older neurons? If so, are older neurons replaced to reinstate a level of plasticity necessary for learning? Or might it be that in some cases the whole neuron, rather than the synapse, is the unit of learning? In this latter scenario, changes associated with learning would be committed as permanent gene expression changes akin to those characterizing cellular differentiation. Such a change would be a very stable way to encode learning, but it would have a major drawback: the more learning that occurred, the fewer neuronal pupils would remain. Thus, we are left to wonder whether neuronal replacement takes place to make up for the lost plasticity of aging neurons, or whether it takes place as part of a normal recycling of memory space and of the memories it holds.
During the early 1970s, before the song system was discovered, it was widely believed that the learning of any one skill had a wide representation in the vertebrate brain [39]. The discovery of discrete brain regions devoted to song learning and execution in the bird brain helped change that view. It was also widely believed that the brains of male and female vertebrates were virtually identical, with small allowances for the levels of circulating hormones. The song system changed that, too. And it was widely believed that the anatomy of adult brains was set, but we now know that the volume of brain structures can change seasonally and in response to blood hormone levels. Most importantly, it was widely held that though a straggling few neurons might still be added after birth to late-developing parts of the brain, brain cells, once lost, could not be replaced. Again, work on the song system changed the prevailing view. It may well be that our best understanding of how complex skills are acquired and how broken circuits can be fixed will come not from humans, or other primates, but from the way birds learn their song.
Citation: Nottebohm F (2005) The neural basis of birdsong. PLoS Biol 3(5): e164.
Fernando Nottebohm is a professor at Rockefeller University in New York, New York, United States of America, and head of the Rockefeller University's Laboratory of Animal Behavior, and Director of the Rockefeller University Field Research Center for Ethology and Ecology in Millbrook, New York. E-mail: [email protected]
Abbreviations
AFPanterior forebrain pathway
HVChigh vocal center
LMANlateral magnocellular nucleus of the nidopallium
PDPposterior descending pathway
RArobust nucleus of the arcopallium
==== Refs
References
Thorpe WH The learning of song patterns by birds, with special reference to the song of the chaffinch, Fringilla coelebs
Ibis 1958 100 535 570
Marler P A comparative approach to vocal learning: Song learning in white-crowned sparrows J Comp Physiol Psychol 1970 71 1 25
Konishi M The role of auditory feedback in the control of vocalization in the white-crowned sparrow Z Tierpsychol 1965 22 770 783 5874921
Marler P Tamura M Culturally transmitted patterns of vocal behavior in sparrows Science 1964 146 1483 1486 14208581
Liu WC Gardner TJ Nottebohm F Juvenile zebra finches can use multiple strategies to learn the same song Proc Natl Acad Sci U S A 2004 101 18177 18182 15608063
Nottebohm F Stokes TM Leonard CM Central control of song in the canary, Serinus canaria
J Comp Neurol 1976 165 457 486 1262540
Vates GE Vicario DS Nottebohm F Reafferent thalamo-“cortical” loops in the song system of oscine songbirds J Comp Neurol 1997 380 275 290 9100137
Paton JA Manogue KR Nottebohm F Bilateral organization of the vocal control pathway in the budgerigar, Melopsittacus undulates
J Neurosci 1981 1 1279 1288 6171631
Jarvis ED Ribeiro S da Silva ML Vielliard Ventura D Behaviorally driven gene expression reveals song nuclei in hummingbird brain Nature 2000 406 628 632 10949303
Bottjer SW Johnson F Circuits, hormones, and learning: Vocal behavior in songbirds J Neurobiol 1997 33 602 618 9369462
Luo M Perkel DJ A GABAergic, strongly inhibitory projection to a thalamic nucleus in the zebra finch song system J Neurosci 1999 19 6700 6711 10414999
Nottebohm F Hauser MD Konishi M The anatomy and timing of vocal learning in birds The design of animal communication 1999 Cambridge (Massachusetts) MIT Press 63 110
Nottebohm F Nottebohm F Harnad S Asymmetries in neural control of vocalization in the canary Lateralization in the nervous system 1977 New York Academic Press 23 44
Nottebohm F Arnold AP Sexual dimorphism in vocal control areas of the songbird brain Science 1976 194 211 213 959852
Nottebohm F A brain for all seasons: Cyclical anatomical changes in song control nuclei of the canary brain Science 1981 214 1368 1370 7313697
Brenowitz EA Nalls B Wingfield JC Kroodsma DE Seasonal changes in avian song nuclei without seasonal changes in song repertoire J Neurosci 1991 11 1367 1374 2027052
Arnold AP Nottebohm F Pfaff DW Hormone concentrating cells in vocal control and other areas of the brain of the zebra finch (Poephila guttata ) J Comp Neurol 1976 165 487 512 1262541
Nottebohm F Testosterone triggers growth of brain vocal control nuclei in adult female canaries Brain Res 1980 192 89 107 7378793
DeVoogd TJ Nottebohm F Gonadal hormones induce dendritic growth in the adult brain Science 1981 214 202 204 7280692
Goldman SA Nottebohm F Neuronal production, migration and differentiation in a vocal control nucleus of the adult female canary brain Proc Natl Acad Sci U S A 1983 80 2390 2394 6572982
Alvarez-Buylla A Nottebohm F Migration of young neurons in adult avian brain Nature 1988 335 353 354 3419503
Nottebohm F Neuronal replacement in adulthood Ann N Y Acad Sci 1985 457 143 161 3913361
Altman J Autoradiographic investigation of cell proliferation in the brains of rats and cats Anat Rec 1963 145 573 591 14012334
Kaplan MS Hinds JW Neurogenesis in the adult rat. Electron microscopic analysis of light radioautographs Science 1977 197 1092 1094 887941
Rakic P Limits of neurogenesis in primates Science 1985 227 154 156 17843070
Paton JA Nottebohm F Neurons generated in adult brain are recruited into functional circuits Science 1984 225 1046 1048 6474166
Kirn JR Nottebohm F Direct evidence for loss and replacement of projection neurons in adult canary brain J Neurosci 1993 13 1654 1663 8385206
Scharff C Kirn J Macklis Grossman J Nottebohm F Targeted neuronal death affects neuronal replacement and vocal behavior in adult songbirds Neuron 2000 25 481 492 10719901
Kirn JR O'Loughlin B Kasparian S Nottebohm F Cell death and neuronal recruitment in the high vocal center of adult male canaries are temporally related to changes in song Proc Natl Acad Sci U S A 1994 19 7844 7848
Gross CG Neurogenesis in the adult brain: Death of a dogma Nat Rev 2000 1 67 72
Nottebohm F Why are some neurons replaced in adult brain? J Neurosci 2002 22 624 628 11826090
Vu ET Mazurek MK Kuo YC Identification of a forebrain motor programming network for the learned song of zebra finches J Neurosci 1994 14 6924 6934 7965088
Yu AC Margoliash D Temporal hierarchical control of singing in birds Science 1996 273 1871 1875 8791594
Hahnloser RHR Kozhevnikov AA Fee MS An ultra-sparse code underlies the generation of neural sequences in a songbird Nature 2002 419 65 70 12214232
Bottjer SW Miesner EA Arnold AP Forebrain lesions disrupt development but not maintenance of song in passerine birds Science 1984 224 901 903 6719123
Scharff C Nottebohm F A comparative study of the behavioral deficits following lesions of various parts of the zebra finch song system: Implications for vocal learning J Neurosci 1991 11 2896 2913 1880555
Kao HM Doupe AJ Brainard MS Contributions of an avian basal ganglia-forebrain circuit to real-time modulation of song Nature 2005 433 638 643 15703748
Ölveczky BP Andalman AS Fee MS Vocal experimentation in the juvenile songbird requires a basal ganglia circuit PLoS Biol 2005 3 e153 10.1371/journal.pbio.0030153 15826219
Lashley KS In search of the engram In: Physiological mechanisms in animal behaviour, Symp Soc Exp Biol IV 1950 Cambridge Cambridge University Press 454 482
| 15884976 | PMC1110917 | CC BY | 2021-01-05 08:21:23 | no | PLoS Biol. 2005 May 17; 3(5):e164 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030164 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1588498110.1371/journal.pbio.0030182PrimerEcologyEvolutionGenetics/Genomics/Gene TherapyMicrobiologyVirologyVirusesEubacteriaThe Third Age of Phage PrimerMann Nicholas H 5 2005 17 5 2005 17 5 2005 3 5 e182Copyright: © 2005 Nicholas H. Mann.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.
Three Prochlorococcus Cyanophage Genomes: Signature Features and Ecological Interpretations
Genomes Offer Ecological Clues to Viruses That Target Ubiquitous Ocean Bacteria
The third age of phage has begun with the recognition that phages may be key to the great planetary biogeochemical cycles and represent the greatest potential genetic resource in the biosphere
==== Body
So, naturalists observe, a flea Has smaller fleas that on him prey; And these have smaller still to bite 'em; And so proceed ad infinitum.—Jonathan Swift
If Jonathan Swift was wrong in envisioning his infinite series of parasites, the “ultimate flea” could be a DNA sequence whose sole biological property is to ensure its own reproduction. But close to the diminutive end of this parasitic spectrum, there are the phages-viruses that infect bacteria. Phages were discovered early in the 20th century, and, at first, interest focussed on the potential of phages as therapeutic tools in the fight against bacterial infectious diseases. The advent of antibiotics was an influential factor in sidelining this ambition. Then for many years the study of phages underpinned the development of modern molecular biology, but that too became passé. The third age of phage has begun only recently with the growing recognition that phages may be major players in the great planetary biogeochemical cycles [1] and also may represent the greatest potential genetic resource in the biosphere.
Phage Abundance and Diversity
This resurgence of interest in phages may have begun in part with the discovery of the astonishing abundance of viruses in aquatic environments; typically on the order of 107 viruses per millilitre of seawater. Electron microscopy of water samples first suggested the extent of this abundance in the world's oceans [2], which was soon confirmed by faster and more accurate methods for the estimation of virus abundance in seawater based on epifluorescence microscopy (Figure 1) and flow cytometry. It is now widely accepted that phages with a very distinctive morphology, the so-called tailed phages (Figure 2), which dominate the marine virus population, represent the most abundant biological entities on the planet, and total phage abundance in the biosphere has been estimated at 1030 or more [3].
Figure 1 An Image from an Epifluorescence Microscope of Seawater Stained with the Dye SYBR Green to Reveal the Bacterial Cells and Smaller Viral Particles
(Image: Jed Fuhrman)
Figure 2 A Transmission Electron Microscope Image of the Synechococcus Phage S-PM2
(Image: Hans-Wolfgang Ackermann)
It is only very recently, however, through the metagenomic analysis of uncultured marine virus communities, that we have begun gaining an idea of their biodiversity in the marine environment [4]. Approximately 65% of the sequences obtained in a 2002 study did not have homologues in the nucleotide databases, suggesting that marine viral diversity is largely unsampled. Where database homologues were detected, the most common hits were to viruses, including all the major families of tailed phages with double-stranded DNA genomes, and algal viruses. An estimate of the complexity of the virus community structure can be derived by asking how frequently the same sequences are obtained, i.e., how frequently the same individual genome is sampled. On this basis, uncultured phage communities are amongst the most diverse ever analysed, with between 3,000 and 7,000 viral types in a 200-litre sample of seawater [4]. Furthermore, phages can move between very different environments, with very similar phage genes being found in marine, freshwater, sediment, and terrestrial samples [5,6]. This is in keeping with the idea that all tailed phages, regardless of the natural environment of their host, have access to a common gene pool [7].
Phages Drive Bacterial Diversity
There is also growing evidence that viral activity is a driving force in maintaining genetic diversity amongst the bacterial community and profoundly influences ecosystem functioning [8].The total population of prokaryotes (bacteria and archaea) within the water column of the oceans is collectively known as bacterioplankton and constitutes over 90% of the total biological carbon in the oceans [1]. In the surface layers of the oceans, bacterioplankton populations may be on the order of 106 to 107 cells per millilitre of seawater, and the phage population is typically one order of magnitude larger (reviewed in [9]).
A major component of the total marine bacterial population in the upper illuminated layers of the oceans is marine unicellular cyanobacteria of the genera Synechococcus and Prochlorococcus. These cyanobacteria make a substantial contribution to the overall productivity of the oceans, both in terms of carbon dioxide fixation and oxygen production. Obviously, phage infection of this extremely important group of organisms will affect the way in which organic carbon is cycled in the oceans. Phages infecting Synechococcus strains have been studied since 1993, and the majority of isolates are myoviruses, phages with a double-stranded DNA genome and a long contractile tail (for review see [10]). There is considerable genetic diversity among the Synechococcus hosts, and this is reflected in the fact that individual phages only infect certain Synechococcus hosts. An analysis of phages infecting Prochlorococcus strains revealed a similar pattern of host range variation, with some phages being highly strain-specific and other phages having a broad host range. In fact, some phages were able to infect both Synechococcus and Prochlorococcus strains [11]. The study by Sullivan et al. [11] yielded one particularly interesting observation: phages that infect strains of Prochlorococcus adapted to high light were all viruses with very short tails and very limited host ranges, whilst phages infecting low-light-adapted strains predominantly had broad host ranges and long contractile tails. The evolutionary pressures that have led to this situation are unclear, as are the ecological implications.
Genome Studies and Insights into Host-Phage Interactions
The study of marine phage genomes is a comparatively recent activity, and only a few genomes have been completed (for review see [12]). However, in this issue of PLoS Biology, Sullivan et al. [13] describe the analysis of the genomes of three phages-a podovirus and two myoviruses-capable of infecting Prochlorococcus strains. The two myoviruses have large genomes, with one possessing more than 327 potential protein-encoding genes. They are, therefore, genetically complex entities capable of a considerable repertoire of “behaviours”. A key feature of all three genomes is that, in addition to containing typical phage genes, they encode genes specifying proteins whose closest homologues are found in cyanobacteria. Some of these genes are thought to represent “signature” cyanophage genes. Are these genes that have been accidentally acquired from the host and confer no fitness benefit on the phage, or do they indicate important features of how the phage interacts with its host?
Consider, for example, the psbA gene that is found in all three phages [14] and has also been found in a Synechococcus phage [15]. The D1 protein encoded by psbA is a central component of photosystem II, which is responsible for the water-splitting reaction that produces oxygen during photosynthesis. The D1 protein is subject to damage during photosynthesis, and the damaged protein needs to be removed from the reaction centre and replaced by undamaged D1 in a continuous repair cycle in order for photosynthesis to continue. In an uninfected cell the D1 protein is encoded by the cell's genome. However, during infection a common phage strategy is to switch off host gene expression, which could impair photosynthesis and thereby deplete the energy required for viral replication. The provision of a viral D1 protein would permit the repair cycle to continue until the cell lysed to release the phage progeny. Thus, intriguingly, a proportion of the oxygen in the atmosphere may actually be produced by a viral form of photosynthesis.
Another example of cyanophage signature genes are those associated with nutrient acquisition. The central regions of the oceans are nutrient poor, and phosphorus is often likely to be present only in growth-limiting amounts. Both of the Prochlorococcus myovirus genomes encode proteins associated with phosphate transport (PstS) or induction in response to phosphate starvation (PhoH) [13].
The Next Stage for Phage
Thus, the analysis of the genomes of the phages of marine Synechococus and Prochlorococcus is providing novel insights into phage-host interactions in oceanic low-nutrient environments and considerably extending the paradigms derived from the studies of phages infecting heterotrophic bacteria. However, what genomic studies will not tell us is the proportion of infected cells in the oceans and therefore the true impact of these Swiftian fleas on ocean processes. The answer will require the techniques of molecular ecology, but phage genomics will surely provide the underpinning for such investigations.
Citation: Mann NH (2005) The third age of phage. PLoS Biol 3(5): e182.
Nicholas H. Mann is in the Department of Biological Sciences, University of Warwick, Coventry, United Kingdom. E-mail: [email protected]
==== Refs
References
Wilhelm SW Suttle CA Viruses and nutrient cycles in the sea-Viruses play critical roles in the structure and function of aquatic food webs Bioscience 1999 49 781 788
Bergh O Borsheim KY Bratbak G Heldal M High abundance of viruses found in aquatic environments Nature 1989 340 467 468 2755508
Chibani-Chennoufi S Bruttin A Dillmann ML Brussow H Phage-host interaction: An ecological perspective J Bacteriol 2004 186 3677 3686 15175280
Breitbart M Salamon P Andresen B Mahaffy JM Segall AM Genomic analysis of uncultured marine viral communities Proc Natl Acad Sci U S A 2002 99 14250 14255 12384570
Breitbart M Miyake JH Rohwer F Global distribution of nearly identical phage-encoded DNA sequences FEMS Microbiol Lett 2004 236 249 256 15251204
Short CM Suttle CA Nearly identical bacteriophage structural gene sequences are widely distributed in both marine and freshwater environments Appl Environ Microbiol 2005 71 480 486 15640224
Hendrix RW Smith MCM Burns RN Ford ME Hatfull GF Evolutionary relationships among diverse bacteriophages and prophages: All the world's a phage Proc Natl Acad Sci U S A 1999 96 2192 2197 10051617
Weinbauer MG Rassoulzadegan F Are viruses driving microbial diversification and diversity? Environ Microbiol 2004 6 1 11 14686936
Wommack KE Colwell RR Virioplankton: Viruses in aquatic ecosystems Microbiol Mol Biol Rev 2000 64 69 114 10704475
Mann NH Phages of the marine cyanobacterial picophytoplankton FEMS Microbiol Rev 2003 27 17 34 12697340
Sullivan MB Waterbury JB Chisholm SW Cyanophages infecting the oceanic cyanobacterium Prochlorococcus
Nature 2003 424 1047 1051 12944965
Paul JH Sullivan MB Segall AM Rohwer F Marine phage genomics Comp Biochem Physiol B Biochem Mol Biol 2002 133 463 476 12470812
Sullivan MB Coleman ML Weigele P Rohwer F Chisholm SW Three Prochlorococcus cyanophage genomes: Signature features and ecological interpretations PLoS Biol 2005 3 e144 15828858
Lindell D Sullivan MB Johnson ZI Tolonen AC Rohwer F Transfer of photosynthesis genes to and from Prochlorococcus viruses Proc Natl Acad Sci U S A 2004 101 11013 11018 15256601
Mann NH Cook A Millard A Bailey S Clokie M Marine ecosystems: Bacterial photosynthesis genes in a virus Nature 2003 424 741
| 15884981 | PMC1110918 | CC BY | 2021-01-05 08:21:23 | no | PLoS Biol. 2005 May 17; 3(5):e182 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030182 | oa_comm |
==== Front
AIDS Res TherAIDS Research and Therapy1742-6405BioMed Central London 1742-6405-2-31586013010.1186/1742-6405-2-3HypothesisClade, Country and Region-specific HIV-1 Vaccines: Are they necessary? Slobod Karen S [email protected] Chris [email protected] Scott A [email protected] John [email protected] Xiaoyan [email protected] Sherri [email protected] Bart G [email protected] Amy [email protected] Pamela J [email protected] Brita [email protected] Robert [email protected] Mattia [email protected] Julia L [email protected] Department of Infectious Diseases, St Jude Children's Research Hospital, 332 N. Lauderdale, Memphis, TN 38105 USA2 Department of Immunology, St Jude Children's Research Hospital, 332 N. Lauderdale, Memphis, TN 38105 USA3 Department of Pediatrics, College of Medicine, 899 Madison Ave., University of Tennessee, Memphis, TN 38163 USA4 Department of Pathology, College of Medicine, 899 Madison Ave., University of Tennessee, Memphis, TN 38163 USA5 Department of Microbiology and Immunology, University of Melbourne, Vic 3010, Australia6 Department of Clinical and Biological Sciences, University of Insubria, Varese, 21100, Italy2005 28 4 2005 2 3 3 1 4 2005 28 4 2005 Copyright © 2005 Slobod et al; licensee BioMed Central Ltd.2005Slobod 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.
Today, scientists are often encouraged to custom-design vaccines based on a particular country or clade. Here, we review the scientific literature and then suggest that the overwhelming endeavor to produce a unique vaccine for every world region or virus subtype may not be necessary.
==== Body
Clade, country or region-specific vaccines
It is generally agreed that HIV-1 arose decades ago by transfer of virus from chimps to humans [1]. The subsequent travel of infected persons and the continued practice of high-risk behaviors fostered virus transmission to virtually every world region. Once HIV-1 awareness was heightened and HIV-1 sequencing projects were implemented, regional similarities of viral sequences, presumably a consequence of the founder effect, became evident. Clade designations (e.g. clade A, B, C) were then used as a means to categorize viruses based on genetic sequence; thus such clade designations also tended to cluster viruses according to geographical location. Today, due to continuous virus transmission, mutation and recombination, the demarcation of HIV-1 subtypes has become increasingly blurred, and the categorization of viruses by clade is increasingly difficult [2-5]. Nonetheless scientists are currently encouraged to custom-design vaccines based on a particular country or clade [6-11]. To this end, a single viral sequence may be selected, possibly based on a formula of ancestry or consensus, to represent all other viruses in the targeted category.
Designing vaccines in this way prompts careful consideration: must a unique vaccine be prepared to represent every clade, country or region of the world? If so, how will this be accomplished and for which country should first vaccines be produced? Who will decide? The complexity of such an undertaking and the many difficulties that attend it encourage a second look at the strategy. Review of the scientific literature may provide reassurance that the seemingly unachievable endeavor to custom-produce a vaccine for every clade, country or region may not be necessary.
Do immune responses discriminate between clades?
While differences in encoded protein sequence may permit discrimination between certain HIV-1 subtypes, successful vaccine development requires that viral proteins elicit protective immune responses, regardless of sequence. It has long been known that clades, as defined by genetic sequence, do not correspond to immunotypes, as defined by mutually exclusive immune responses [12-14]. Both B- and T-cells elicited by a virus from one clade may recognize viruses from other clades. This cross-clade responsiveness is explained by the fact that the B- and T-cells recognize precise epitopes rather than the overall sequence similarity of viruses. Antibody binding depends on three-dimensional structure, and the molecular structures bound by antibodies can occur on proteins that differ widely in primary sequence. T-cells recognize peptides in association with Class I or Class II MHC molecules, but like B-cells, T-cells can cross-react with non-identical targets. Conversely, two viruses may have 99% sequence similarity, yet a particular neutralizing antibody or T-cell receptor may discriminate between them. This discrimination may be due to a single amino acid change within the receptor contact site or in a sequence that alters epitope display [15,16]. Thus it is the detail of epitope and epitope context, not overall sequence similarity that defines lymphocyte specificity.
Cross-clade protection is achieved by priming the immune system with diverse viral sequences from a single clade
The issues described above suggest that although a single-component vaccine may not be sufficient to target any clade, a cocktail vaccine, designed to represent the natural diversity of HIV-1, may be sufficient to target all clades. The latter point is supported by studies of HIV-1-infected humans and SIV-infected macaques. Although infected subjects cannot clear endogenous virus (due to its sequestration in "privileged" sites, hidden from the immune system), most individuals are resistant to super-infection [17-22]. This protection likely arises as the result of many successive rounds of endogenous viral mutation in the infected host. Each time an immune response is elicited in the periphery of an infected subject, new virus mutants appear [23,24]. The new viruses, by definition, have altered T- and B-cell determinants, allowing escape from the established antibodies and T-cell receptors. Following several rounds of immune response and virus escape, the B- and T-cells are primed to recognize a broad spectrum of determinants [25]. Thus, superinfections are rare, even in subjects likely to have been serially exposed to viruses from different clades. The rare double infections in humans (explaining the origin of virus recombinants [4]) are perhaps a consequence of (i) drug regimens which block the natural evolution of virus in the infected subject, (ii) repeated HIV-1 exposures prior to maturation of the adaptive immune response, and/or (iii) disease-related immunodeficiency.
The fact that a mature immune response to HIV-1 cannot clear sequestered virus, but can prevent super-infection emphasizes the importance of priming the system preemptively. Similar considerations pertain to the design of vaccines against human herpesviruses (e.g. VZV and EBV), as these viruses provoke both lifelong infections and long-term protective immunity to superinfection. As with the successful VZV vaccine [26], an effective HIV-1 vaccine should be administered before virus exposure, infection and sequestration.
Could a cocktail vaccine ever be large enough to prevent HIV-1 infections?
Perhaps careful vaccine formulation will preclude the need for assembly of enormous cocktails. Consideration that envelope structure is constrained by function suggests that the formulation of an effective envelope-based vaccine is feasible. The virus envelope must bind target cells to mediate infection, and only a few target cell receptor molecules (e.g. CD4, CCR5, CXCR4), have been described. Therefore, the number of discrete envelope shapes that maintain full cell-binding potential and function is likely to be limited [27]. Because the virus envelope is the target of both neutralizing antibodies and T cells, the strengths of both arms of the immune system may be harnessed by an envelope-based vaccine cocktail [28-30]. Diverse proteins need not be cross-inhibitory. In fact, type-specific immune responses have been recognized toward a single envelope construct represented as only 1% of a mixed vaccine [31]. Cocktail vaccines are effective in controlling other diverse pathogens (e.g. pneumococcus, poliovirus), despite early doubts about their prospect of success [32].
Clade, Country or Region-specific HIV Vaccines may not be necessary
The assembly of envelope cocktail vaccines will probably be necessary to represent the natural diversity of HIV-1, even within a single clade. Careful vaccine design may reveal a cocktail formulation able to prevent virus infections in every world region, and to overcome the political and financial dilemmas associated with the production of clade, country or region-specific vaccines.
Acknowledgements
This work was supported in part by NIH NIAID P01-AI45142, NCI Cancer Center Support Core Grant P30-CA21765, the Mitchell Fund, the Federated Department Stores, the James B. Pendleton Charitable Trust and the American Lebanese Syrian associated Charities (ALSAC).
==== Refs
Perrin L Kaiser L Yerly S Travel and the spread of HIV-1 genetic variants Lancet Infect Dis 2003 3 22 27 12505029 10.1016/S1473-3099(03)00484-5
Delwart EL Orton S Parekh B Dobbs T Clark K Busch MP Two percent of HIV-positive U.S. blood donors are infected with non-subtype B strains AIDS Res Hum Retroviruses 2003 19 1065 1070 14709241 10.1089/088922203771881149
Anderson JP Rodrigo AG Learn GH Madan A Delahunty C Coon M Testing the hypothesis of a recombinant origin of human immunodeficiency virus type 1 subtype E J Virol 2000 74 10752 10765 11044120 10.1128/JVI.74.22.10752-10765.2000
McClutchan FE Carr JK Murphy D Piyasirisilp S Gao F Hahn B Precise mapping of recombination breakpoints suggests a common parent of two BC recombinant HIV type 1 strains circulating in China AIDS Res Hum Retroviruses 2002 18 1135 1140 12402948 10.1089/088922202320567879
Thomson MM Perez-Alvarez L Najera R Molecular epidemiology of HIV-1 genetic forms and its significance for vaccine development and therapy Lancet Infect Dis 2002 2 461 471 12150845 10.1016/S1473-3099(02)00343-2
Williamson C Morris L Maughan MF Ping LH Dryga SA Thomas R Characterization and selection of HIV-1 subtype C isolates for use in vaccine development AIDS Res Hum Retroviruses 2003 19 133 144 12639249 10.1089/088922203762688649
Gaschen B Taylor J Yusim K Foley B Gao F Lang D Diversity considerations in HIV-1 vaccine selection Science 2002 296 2354 2360 12089434 10.1126/science.1070441
Hanke T Barnfield C Wee EG Agren L Samuel RV Larke N Construction and immunogenicity in a prime-boost regimen of a Semliki Forest virus-vectored experimental HIV clade A vaccine J Gen Virol 2003 84 361 368 12560568 10.1099/vir.0.18738-0
Novitsky V Smith UR Gilbert P McLane MF Chigwedere P Williamson C Human immunodeficiency virus type 1 subtype C molecular phylogeny: consensus sequence for an AIDS vaccine design? J Virol 2002 76 5435 5451 11991972 10.1128/JVI.76.11.5435-5451.2002
Hanke T McMichael AJ Design and construction of an experimental HIV-1 vaccine for a year-2000 clinical trial in Kenya Nature Medicine 2000 6 951 955 10973301 10.1038/79626
Agwale SM Zeh C Robbins KE Odama L Saekhou A Edubio A Molecular surveillance of HIV-1 field strains in Nigeria in preparation for vaccine trials Vaccine 2002 20 2139
Cao H Mani I Vincent R Mugerwa R Mugyenyi P Kanki P Cellular immunity to human immunodeficiency virus type 1 (HIV-1) clades: relevance to HIV-1 vaccine trials in Uganda J Infect Dis 2000 182 1350 1356 11023458 10.1086/315868
Moore JP Cao Y Leu J Qin L Korber B Ho DD Inter- and intraclade neutralization of human immunodeficiency virus type 1: genetic clades do not correspond to neutralization serotypes but partially correspond to gp120 antigenic serotypes J Virol 1996 70 427 444 8523556
Ferrantelli F Kitabwalla M Rasmussen RA Cao C Chou TC Katinger H Potent cross-group neutralization of primary human immunodeficiency virus isolates with monoclonal antibodies – implications for acquired immunodeficiency syndrome vaccine J Infect Dis 2004 189 71 74 14702155 10.1086/420833
D'Costa S Slobod KS Webster RG White SW Hurwitz JL Structural features of HIV envelope defined by antibody escape mutant analysis AIDS Res Hum Retroviruses 2001 17 1205 1209 11522189 10.1089/088922201316912808
Zhan X Slobod KS Surman S Brown SA Lockey TD Coleclough C Limited breadth of a T-helper cell response to a human immunodeficiency virus envelope protein J Virol 2003 77 4231 4236 12634380 10.1128/JVI.77.7.4231-4236.2003
Gonzales MJ Delwart E Rhee SY Tsui R Zolopa AR Taylor J Lack of detectable human immunodeficiency virus type 1 superinfection during 1072 person-years of observation J Infect Dis 2003 188 397 405 12870121 10.1086/376534
Tsui R Herring BL Barbour JD Grant RM Bacchetti P Kral A Human immunodeficiency virus type 1 superinfection was not detected following 215 years of injection drug user exposure J Virol 2004 78 94 103 14671091 10.1128/JVI.78.1.94-103.2004
Altfeld M Allen TM Yu XG Johnston MN Agrawal D Korber BT HIV-1 superinfection despite broad CD8+ T-cell responses containing replication of the primary virus Nature 2002 420 434 439 12459786 10.1038/nature01200
Daniel MD Kirchhoff F Czajak SC Sehgal PK Desrosiers RC Protective effects of a live attenuated SIV vaccine with a deletion in the nef gene Science 1992 258 1938 1941 1470917
Burns DPW Desrosiers RC Envelope sequence variation, neutralizing antibodies and primate lentivirus persistence Curr Top Microbiol Immunol 1994 188 185 219 7523031
Titti F Sernicola L Geraci A Panzini G Di Fabio S Belli R Live attenuated simian immunodeficiency virus prevents super-infection by cloned SIVmac251 in cynomolgus monkeys J Gen Virol 1997 78 2529 2539 9349474
Wrin T Crawford L Sawyer L Weber P Sheppard HW Hanson CV Neutralizing antibody responses to autologous and heterologous isolates of human immunodeficiency virus J Acquir Immune Defic Syndr 1994 7 211 219 8106963
Richman DD Wrin T Little SJ Petropoulos CJ Rapid evolution of the neutralizing antibody response to HIV type 1 infection Proc Natl Acad Sci U S A 2003 100 4144 4149 12644702 10.1073/pnas.0630530100
Rencher SD Slobod KS Dawson D Lockey TD Hurwitz JL Does the key to a successful HIV vaccine lie among the envelope sequences of infected individuals? Aids Res Hum Retroviruses 1995 11 1131 1133 8554911
Seward JF Watson BM Peterson CL Mascola L Pelosi JW Zhang JX Varicella disease after introduction of varicella vaccine in the United States, 1995–2000 JAMA 2002 287 606 611 11829699 10.1001/jama.287.5.606
Nyambi PN Nadas A Mbah HA Burda S Williams C Gorny MK Immunoreactivity of intact virions of human immunodeficiency virus type 1 (HIV-1) reveals the existence of fewer HIV-1 immunotypes than genotypes J Virol 2000 74 10670 10680 11044111 10.1128/JVI.74.22.10670-10680.2000
Slobod KS Lockey TD Howlett N Srinivas RV Rencher SD Freiden PJ Subcutaneous administration of a recombinant vaccinia virus vaccine expressing multiple envelopes of HIV-1 Eur J Clin Microbiol Infect Dis 2004 23 106 110 14735404 10.1007/s10096-003-1075-3
Hurwitz JL Slobod KS Lockey TD Wang S Chou T-HW Lu S Application of the polyvalent approach to HIV-1 vaccine development Current Drug Targets-Infectious Disorders 2005
Stambas J Brown SA Gutierrez A Sealy R Yue W Jones B Long lived multi-isotype anti-HIV antibody responses following a prime-double boost immunization strategy Vaccine 2005 23 2454 2464 15752831 10.1016/j.vaccine.2004.10.030
Zhan X Slobod KS Surman S Brown SA Coleclough C Hurwitz JL Minor components of a multi-envelope HIV vaccine are recognized by type-specific T-helper cells Vaccine 2004 22 1206 1213 15003649 10.1016/j.vaccine.2003.09.028
Burnet FM Poliomyelitis in the light of recent experimental work Health Bulletin Department of Health-Victoria, Australia 1945
| 15860130 | PMC1112584 | CC BY | 2021-01-04 16:38:33 | no | AIDS Res Ther. 2005 Apr 28; 2:3 | utf-8 | AIDS Res Ther | 2,005 | 10.1186/1742-6405-2-3 | oa_comm |
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-381582630410.1186/1471-2407-5-38Research ArticleMyeloid antigens in childhood lymphoblastic leukemia:clinical data point to regulation of CD66c distinct from other myeloid antigens Kalina Tomas [email protected] Martina [email protected] Ester [email protected] Jozef [email protected] Jan [email protected] Jan [email protected] Ondrej [email protected] Department of Immunology, Charles University 2nd Medical School, Prague, Czech Republic2 Department of Pediatric Hematology and Oncology, Charles University 2nd Medical School, Prague, Czech Republic3 CLIP – Childhood Leukemia Investigation Prague Czech Republic2005 12 4 2005 5 38 38 16 11 2004 12 4 2005 Copyright © 2005 Kalina 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
Aberrant expression of myeloid antigens (MyAgs) on acute lymphoblastic leukemia (ALL) cells is a well-documented phenomenon, although its regulating mechanisms are unclear. MyAgs in ALL are interpreted e.g. as hallmarks of early differentiation stage and/or lineage indecisiveness. Granulocytic marker CD66c – Carcinoembryonic antigen-related cell adhesion molecule 6 (CEACAM6) is aberrantly expressed on ALL with strong correlation to genotype (negative in TEL/AML1 and MLL/AF4, positive in BCR/ABL and hyperdiploid cases).
Methods
In a cohort of 365 consecutively diagnosed Czech B-precursor ALL patients, we analyze distribution of MyAg+ cases and mutual relationship among CD13, CD15, CD33, CD65 and CD66c. The most frequent MyAg (CD66c) is studied further regarding its stability from diagnosis to relapse, prognostic significance and regulation of surface expression. For the latter, flow cytometry, Western blot and quantitative RT-PCR on sorted cells is used.
Results
We show CD66c is expressed in 43% patients, which is more frequent than other MyAgs studied. In addition, CD66c expression negatively correlates with CD13 (p < 0.0001), CD33 (p = 0.002) and/or CD65 (p = 0.029). Our data show that different myeloid antigens often differ in biological importance, which may be obscured by combining them into "MyAg positive ALL". We show that unlike other MyAgs, CD66c expression is not shifted from the onset of ALL to relapse (n = 39, time to relapse 0.3–5.3 years). Although opposite has previously been suggested, we show that CEACAM6 transcription is invariably followed by surface expression (by quantitative RT-PCR on sorted cells) and that malignant cells containing CD66c in cytoplasm without surface expression are not found by flow cytometry nor by Western blot in vivo. We report no prognostic significance of CD66c, globally or separately in genotype subsets of B-precursor ALL, nor an association with known risk factors (n = 254).
Conclusion
In contrast to general notion we show that different MyAgs in lymphoblastic leukemia represent different biological circumstances. We chose the most frequent and tightly genotype-associated MyAg CD66c to show its stabile expression in patients from diagnosis to relapse, which differs from what is known on the other MyAgs. Surface expression of CD66c is regulated at the gene transcription level, in contrast to previous reports.
==== Body
Background
Although expression of surface markers in acute lymphoblastic leukemia (ALL) parallels that of normal hematopoietic precursors, several markers of myeloid lineage are found on ALL lymphoblasts. This phenomenon is referred to as "aberrant expression". The issue of the regulatory mechanisms that allow it has been addressed repeatedly throughout the recent 40 years [1,2]. Although several hypotheses stressing either possible lineage indecisiveness or genetic misprogramming have been raised, the phenomenon is still not fully understood. We and others have shown that the myeloid antigen CD66c is very frequently aberrantly expressed in B-precursor ALL, however, a large study showing its frequency in the light of other myeloid antigens has been missing. CD66c expression was found on cases of childhood and adult ALL in strong correlation with nonrandom genetic changes (BCR/ABL positivity [3], hyperdiploidy and TEL/AML1 negativity [4], reviewed in [5]).
CD66c (CEACAM6, previously called Nonspecific cross-reacting antigen, NCA 90/50 and KOR-SA3544 antigen) is a member of the carcinoembryonic antigen family. This heavily glycosylated molecule consists of two constant Ig-like domains and one variable Ig-like domain and it is anchored to the membrane via its glycosylphosphatidylinositol (GPI). Within the hematopoietic system, CD66c expression is limited to granulocytes and its precursors [3,6], where it serves homotypic and heterotypic adhesion [7], Ca2+ mediated signaling [8] and is markedly upregulated from intracellular stores after activation [9].
It is also found in epithelia of various organs [7]. Upregulation of CD66c is an early molecular event in transformation leading to colorectal tumors [10]. It was also confirmed to inhibit anoikis (apoptotic response induced in normal cells by inadequate or inappropriate adhesion to substrate) in the in vitro model of carcinoma of colon [11] and specific silencing of this gene led to decreased metastatic potential in pancreatic adenocarcinoma [12].
Surprisingly, Sugita et al [13] reported intracellular presence of CD66c in all leukemic cell lines examined, regardless of surface presence or absence, with a different antigen distribution in cytoplasm that determined surface expression. They speculated that presence of an undisclosed transporter would target this molecule to granules and for surface expression, whereas surface CD66cneg cell lines lack this transporter. This intriguing hypothesis prompted us to test whether transcription of CEACAM6 gene and/or intracellular CD66c expression is always followed by surface expression.
Uniqueness of aberrant expression of CD66c on malignant lymphoblast is exploited for diagnosis of ALL and follow-up of a minimal residual disease (MRD) using flow cytometry [14,15]. To use a marker for a MRD assessment a critical question must be addressed, whether the aberrant expression is a stable property of the malignant clone or whether it can be subject to immunophenotype shift.
In the present study we set out to address the frequency of CD66c molecule expression in childhood ALL, the regulation of CD66c expression from gene transcription to cytoplasmic and surface expression, and we follow immunophenotype stability from diagnosis to relapse. We also discuss relevance of CD66c for prognosis prediction.
Methods
Patients
The cohort of all Czech children (<18 years) diagnosed with B-precursor ALL investigated in our reference laboratory from 1.5.1997 to 23.7.2004 was used for current study (n = 381). Informed consent was obtained from patients and/or their guardians. The presence of TEL/AML1, BCR/ABL and MLL/AF4 fusion genes was detected by two-round nested PCR, hyperdiploidy was assessed using DNA index flow cytometric measurement as described previously [4]. Patients' genotype and corresponding surface CD66c expression is shown in Figure 1 (genotype available in 98% of patients). For intracellular staining and FACS sorting, only samples with enough material were selected.
Cell lines
Surface CD66c negative cell lines with typical translocation found in childhood ALL: TEL/AML1pos (REH) was kindly provided by R. Pieters (University Hospital Rotterdam), MLL/AF4pos (RS4;11) translocation and with no fusion (NALM-6) were obtained from German Cell Line collection (DSMZ, Braunschweig, Germany)
Flow Cytometry
Flow cytometry immunophenotyping of bone marrow (BM) aspirates was performed in at diagnosis and at relapse. Routine immunophenotypic classification using panel of monoclonal antibodies (moAbs) was performed as described previously [4]. Briefly, BM samples were stained with 2-, 3- and 4-color combinations of moAbs for 15 min in darkness, erythrocytes were lysed with NH4Cl-containing lysing solution for 15 min, washed and data were acquired using single FACS Calibur instrument throughout the study (BD Biosciences, San Jose, CA, USA) flow cytometer. Anti-CD66c (CEACAM6) moAb used in all diagnostic and relapse measurements in this study was clone KOR-SA3544 directly labeled to FITC (Immunotech, Marseille, France). Intracellular staining was performed using Fix & Perm kit (Caltag, Burlingame, CA, USA) according to manufacturer's protocol. Acquired data was analyzed with Cell Quest (BD Biosciences) or Flow Jo (Tree Star, Ashland, OR, USA) software, lymphoblast gate was drawn based on optical scatter and CD19pos blast population was selected for further analysis.
Value of 20% was chosen as a threshold of positivity as recommended by EGIL [16]. For robust prognostic significance testing, other threshold values were also tested as indicated in results.
Cross-blocking of CD66c moAbs
Bone marrow samples of CD66c positive blasts were stained with anti-CD66c moAb clone 9A6 (Genovac, Freiburg, Germany) moAb for 15 min, erythrocytes were lysed with NH4Cl-containing lysing solution for 15 min, washed and sample was incubated with anti-CD66c moAb KOR-SA3544 PE moAb conjugate.
Western blot
Samples containing 5 × 106 cells were lysed for 30 min at 4°C in 100 μl lysis buffer containing 20 mM Tris-HCl (pH 8.2), 100 mM NaCl, 50 mM NaF, 10 mM EDTA, 10 mM pyrophosphate (Na4P2O7) and Complete Mini EDTA-Free (protease inhibitor cocktail tablets, Roche Diagnostics, Mannheim, Germany). Debris was sedimented by centrifugation for 3 min at 13000 rpm, 0°C. Supernatants were mixed with 100 μl 2× Laemmli's SDS-polyacrylamide gel electrophoresis (PAGE) sample loading buffer, and heated for 5 min at 100°C. Proteins were fractionated by SDS-PAGE on 12.5% gels and electrophoretically transferred to PVDF membranes (Bio-Rad, Hercules, CA, USA). Membranes were blocked for 1 h in PBS (pH 7.4) containing 0.5% Tween-20 and 5% nonfat dried milk. Blots were then incubated for 1 h at room temperature with anti-KOR-SA3544 (Immunotech, Marseille, France) or anti-beta-actin (Sigma-Aldrich, Saint Louis, MO, USA) moAbs and then developed using goat anti-mouse IgG (H+L)-HRP conjugate (Bio-Rad). Immunoreactive material was then revealed by enhanced chemiluminescence (ECL, Amersham, Little Chalfont Buckinghamshire, UK) according to the manufacturer's instructions.
Isolation of RNA and Real-Time Quantitative PCR analysis (RQ-PCR)
For RQ- PCR analysis, leukemic blasts were FACS sorted using sorting option on FACS Calibur or on FACS Aria instrument (1.1 × 104 - 4.7 × 105 cells from one patient). Isolation of RNA from FACS-sorted cells was performed using Trizol-reagent (Gibco BRL, Carlsbad, CA, USA) according to manufacturer's instructions [17]. Complementary DNA was prepared using M-MLV Reverse Transcriptase (Gibco) according to manufacturer instructions. Glycogen (Gibco) 250 μg /mL was added when initial cell number was lower than 105. Quality of cDNA was verified by PCR on beta-2-microglobulin (B2M) housekeeping gene.
RQ-PCR was performed in the LightCycler™ rapid thermal cycler system (Roche Diagnostic GmbH, Mannheim, Germany), according to manufacturer's instructions, using SYBR green intercalating dye. CEACAM6 specific primers 3'-CGCCTTTGTACCAGCTGTAA and 5'-GCATGTCCCCTGGAAGGA designed by Baranov [18] were used for CEACAM6 amplification and B2M specific primers 3'-GATGCTGCTTACATGTCTCG 5'-CCAGCAGAGAATGGAAAGTC [19]were used for total cDNA quantification.
PCR amplification was carried out in 1× reaction buffer (20 mmol/L Tris-HCl, pH 8.4; 50 mmol/L KCl); and 2.0 mmol MgCl2 containing 200 μmol/L of each dNTP, 0.2 μmol/L of each primer, 5 μg bovine serum albumin per reaction, and 1 U of Platinum Taq DNA polymerase (all from Gibco) in a final reaction volume of 20 μL. For each PCR reaction, 2 μL of cDNA template and 2 μl of SYBR Green 5 × 10-4 (FMC BioProducts, Rockland, MA, USA) fluorescent dye was included. The cycling conditions were 2.0 minutes at 95°C followed by 45 cycles of denaturation at 94°C for 5 seconds, annealing at 59°C for 30 seconds, and extension at 72°C for 15 seconds. CEACAM6 and B2M gene were amplified separately from the same cDNA, and all experiments were performed in duplicate. Melting curve analysis was performed after each run; in case of peak melting temperature shift, PCR products were verified on agarose gel electrophoresis.
Normalized CEACAM6 Expression (CEACAM6n)
Amplification and calibration curves were generated by using affiliated software (LightCycler 3 data-analysis software; version 3.5.28; Idaho Technology Inc., Salt Lake City, UT, USA). A calibration curve for the B2M and CEACAM6 housekeeping gene was generated using the series of 10× diluted cDNA from peripheral blood granulocytes as a standard for both reactions. Crossing point (Cp) value was calculated with LightCycler 3 software using second derivative maximum method. CEACAM6n value is relative and represents a ratio of CEACAM6 to B2M (CEACAM6n = CEACAM6/ B2M). Standard cDNA from granulocytes was assigned CEACAM6n value of 1, the same aliquot of granulocytes cDNA was used throughout of study.
Statistics
Statistical evaluation was done with Statview software, (SAS Institute Inc, NC, USA). We used Fisher's exact test, regression coefficient, Mann-Whitney test and Logrank (Mantel-Cox) test as described in text.
Results
Frequency of CD66c and myeloid antigen (MyAg) expression
We selected 365 patient's samples obtained at diagnosis of B-precursor ALL with available information on the expression of MyAg CD13, CD15, CD33, CD65 and CD66c. This subcohort represents 96% of all B-precursor ALL diagnosed in the study period. The CD66c molecule was expressed on 43% cases (Table 1, cases with >20% positive blasts were considered positive). For the fraction of positive cells and correlation with genotype see [5], of note, 29% of patients expressed CD66c on more then 50% blasts. Comparison with other MyAg showed that CD66c is more frequently expressed. Coexpression of CD66c with other MyAg was not a usual finding (Table 1, Figure 2). Expression of CD13, CD33 and CD65 tended to be non-random (mutually exclusive) with CD66c (Table 1). Coexpression of CD66c with any 2 of the other MyAg was found in fewer than 4 cases in each combination. Interestingly, mutual relationship of other MyAg was random, with the exception of CD13 and CD33 coexpression (p < 0.0001) and CD15 and CD65 coexpression (p = 0.0002). The analysis was performed also at different cutoff values (10, 30 and 50 %; data not shown). The same or less significant correlations were also observed at different cutoff values.
Cross-blocking of KOR-SA3544 clone with 9A6 clone
The moAb clone KOR-SA3544 was not included in Human Leukocyte Differentiation Antigens workshop, but was characterized by Sugita et al [13]. To prevent ambiguous interpretation of our data we extended characterization of KOR-SA3544 clone of CD66c moAb by blocking experiments on CD66cpos blasts. Pretreatment of cells with workshop-typed clone 9A6 moAb completely blocked binding of KOR-SA3544 clone in all 9 leukemic specimens and in granulocytes (data not shown).
Cytoplasmic presence of CD66c in ALL blasts
We have studied surface and cytoplasmic expression of CD66c in 20 ALL diagnostic samples by flow cytometry. In contrast to findings of Sugita et al [13], we have detected CD66c exclusively in all 8 surface positive cases. None of the 12 surface negative cases stained in cytoplasm (Figure 3). The probable cause of the opposite finding in several cases (lower percentage after permeabilization than on surface) is a higher background after permeabilization (isotypic control mean fluorescence intensity was 4.3 ± 2.0 and 9.7 ± 3.7 for surface and permeabilized staining, respectively), which covers borderline events.
Transcription of CEACAM6 gene
To extend the above findings, we used Real-Time Quantitative Reverse Transcription-PCR (RQ-RT-PCR) to quantitatively assess presence of specific CEACAM6 mRNA. We FACS-sorted CD19posCD66cneg or CD19posCD66cpos blast cells for RQ-RT-PCR analysis. We didn't find significant amount of CEACAM6 transcript in surface CD66cneglymphoblasts, whereas CD66cpos cells contained CEACAM6. When CD66cneg and positive fraction was FACS-sorted of heterogeneous specimens (lymphoblasts partly positive for CD66c) the level of CEACAM6 was observed higher in CD66cneg cells and lower in CD66cpos cells as compared to uniform populations (Figure 4). In one specimen (ALL patient with Down syndrome), CEACAM6 wasn't increased in CD66cpos fraction.
Western blot
We further question the intracellular CD66c positivity in surface CD66c negative cell lines. We performed Western blot as described by Sugita et al. [13] on REH (TEL/AML1pos) and RS4;11 (MLL/AF4pos) cell lines and found no CD66c protein (Figure 5). Furthermore we found NALM-6 (surface CD66cneg, no translocation) cell line negative. Two BCR/ABL and four hyperdiploid (all surface CD66cpos) diagnostic samples used as positive controls were positive, with the similarly narrow band contrasting to broad band detected in granulocytes (Figure 5), suggesting different glycosylation in keeping with report by Sugita.
Stability of surface expression from diagnosis to relapse
All relapsed patients up till 12/2003 with available information on CD66c expression at diagnosis and at relapse were used to assess stability of CD66c expression. Comparison of CD66c expression in 39 cases of relapsed childhood ALL cases to their immunophenotype at diagnosis revealed that both negativity and positivity of this antigen was retained from diagnosis to relapse (Figure 6; median time to relapse 2.5y min 0.3y, max 5.3y). Although the quantitative levels of CD66c expression differed in some patients (median difference 0.0%, standard deviation 21%), no case of CD66c complete loss or gain was found in our cohort.
Prognostic significance of CD66c expression
Only B-precursor ALL patients treated on the same ALL BFM 95 treatment protocol [20] (n = 254) were evaluated for prognostic impact. The prognosis did not differ for cases with either CD66cpos blasts exceeding either 20% (Figure 7) or any other cutoff value tested (5%, 10% and 50%, data not shown).
Next, we asked whether CD66c expression correlated with the risk factors used in ALL BFM-95 protocol for stratification into risk groups [21]. No difference in relapse free survival (RFS) was noted when analyzed separately for each risk group or higher and lower initial leukocytosis (cutoff value: 2 × 104 cells per ml), age group or response to prednisone (groups as in Table 2).
When analyzed with respect to a genotype, we found no prognostic value of CD66c in any defined group (BCR/ABLpos, TEL/AML1pos, hyperdiploid ALL and none of the above-mentioned genetic changes, Figure 7 and Table 2). In contrast to the study by Hanenberg et al [22], there was no correlation between initial leukocytosis and CD66c in our cohort (Table 2).
Discussion
Our data on childhood B-precursor ALL show that CD66c is more frequently expressed than the myeloid antigens included in the standard immunophenotyping panels for ALL. To our knowledge, CD66c is the most frequent myeloid marker in childhood ALL. This, together with the tight correlations between CD66c and genotype [5], makes CD66c a pertinent object of research on aberrant expression regulation.
In line with the data from Sugita, we confirm the specificity of KOR-SA3544 clone moAb for CD66c by CEACAM6 mRNA detection and by cross-blocking of KOR-SA3544 binding by representative 9A6 clone, that suggests a spatial proximity of the two epitopes recognized. Furthermore we show that all CD66cpos ALL specimens show a similar extent of glycosylation as cell lines analyzed by Sugita, which differs from the extent of glycosylation in granulocytes.
Since there is a strong correlation of ALL genotype and CD66c expression, we hypothesized that surface CD66c expression would be controlled by gene transcription rather than by targeting to surface from intracellular stores as proposed by Sugita [13]. In accordance with this, both intracellular staining and Western blot failed to identify cytoplasmic CD66c protein in any surface CD66cneg cells. Down the same line, no CEACAM6 transcript was detected in surface CD66cneg lymphoblasts. Overall our data suggest that transcription is the checkpoint that leads to surface expression, rather then the former model, which proposed that all malignant lymphoblasts generate the CD66c molecule but only some of them target it for the cell membrane.
Interestingly, importance of this molecule was shown in a model of colorectal carcinoma where transfection with CEACAM6 inhibited anoikis (10), high CEACAM6 predicted high risk patients with resectable colorectal cancer (9) and CEACAM6 gene silencing decreased resistance to anoikis in vitro leading to inhibition of metastatic ability in mouse model (11). Although the function of CEACAM6 in ALL blasts is still unknown, this molecule's function has been recently associated with pathogenesis of other types of cancer in man [10-12,23,24]. Study of anti-CEACAM6 immunotoxin-based therapy in mouse model of pancreatic carcinoma was published recently [25].
So far, prognostic significance of expression of myeloid antigens CD13, CD14, CD33, CD65w, CD11b and CD15 has been studied with conflicting results (summarized in [26]). As determined in our large cohort of patients treated on ALL BFM 95 protocol, no prognostic significance of CD66c could be revealed in general, nor when we analyzed separate risk groups or TEL/AML1pos, BCR/ABLpos, hyperdiploid and other B-precursor ALL cases separately. Furthermore, instability of aberrant expression was reported for most myeloid markers (CD13, CD14, CD15, CD33 and CD65).
Stability of expression is a major concern of flow cytometric studies of MRD. In present, use of multiple CD markers is widely recommended to prevent MRD underestimation due to the immunophenotype shift (discussed in [15,27]). In current study we show for the first time that CD66c expression stays qualitatively stable from diagnosis to relapse in all relapsed cases studied. This finding, together with high frequency of CD66cpos cases, supports inclusion of CD66c into a moAbs panels for MRD detection in patients positive for this CD marker at diagnosis. However, anecdotal downregulation of CD66c expression during chemotherapy has been observed [15], but has not been methodically studied yet. Any temporary downregulation might lead to falsely lower values of MRD measurement – thus, it would be worthwhile to disclose whether this phenomenon occurs regularly at certain points of chemotherapy.
Mutual exclusiveness of MyAg expression as well as different stability of CD66c compared to other MyAgs [28] challenges the general practice of prognostic evaluation of MyAgpos ALL cases as a group [26] and favors individual evaluation of contribution/regulation of each MyAg for blast cell.
Conclusion
CD66c presents some of the tightest associations with ALL genotype. Although our findings indicate that CD66c is unlikely to gain a practical importance as a prognosis predictor, there are several reasons to focus on it in diagnostic and MRD studies. CD66c, apparently the most frequently expressed aberrant antigen in childhood ALL, is very useful in discriminating leukemic blasts from non-malignant cells. Aberrant expression remains a puzzling phenomenon that warrants further investigation. If it is confirmed by techniques sensitive enough that the so called "aberrant markers" are truly not expressed on any subtle population of lymphoid precursors, there will be an opportunity to find new targets for specific ALL therapy (e.g. monoclonal antibodies against differently glycosylated form of CD66c) that will spare the non-leukemic precursors, thus reducing the treatment toxicity.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
TK performed flow cytometry, cell sorting, RQ-RT-PCR study and drafted the manuscript, MV carried out the Western blot study, EM acquired and analyzed patients flow cytometry data and performed the statistical analysis, JM designed and assisted to the RQ-RT-PCR study, JT designed RT-PCR, did the genotype detection and critically discussed the manuscript, JS contributed to the study design and organization and OH conceived of the study, analyzed data and drafted 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
This work was supported by the Grant Agency of Charles University #44/2001 and #65/2004, IGA #7430-3 and MSM0021620813. Superb technical assistance of J. Ridoskova, K. Pospisilova, L. Gondorcinova, P. Hanusova, K. Muzikova and M. Kalinova as well as the collaboration of all Czech Pediatric Hematology (CPH) centers (data manager A. Vrzalova, leaders: B. Blazek (Ostrava), Z. Cerna (Plzen), Y. Jabali (Ceske Budejovice), V. Mihal (Olomouc), D. Prochazkova (Usti nad Labem), J. Stary (Praha), J. Sterba (Brno), J. Hak and K. Tousovska (Hradec Kralove)) is highly appreciated. V. Horejsi is acknowledged for consulting in molecular immunology. We thank to F. Grunert for providing us with a sample of 9A6 clone of CD66c.
Figures and Tables
Figure 1 Correlation of ALL genotype categories and percentage of CD66c positivity. Median percentage of CD66cpos blasts is listed below each genotype group. Data of consecutive unselected patients with BCP ALL (n = 373) are shown.
Figure 2 Graphical illustration of myeloid antigen positivity in childhood B-precursor ALL. For each antigen, positive cases are represented by a colored form. The areas of the forms roughly correspond to the frequency of positive cases (observed numbers of patients are marked in red) while the shapes are constructed to illustrate the respective coexpressions. An arbitrary cutoff value of 20% is used for all antigens. The CD66c positivity correlates with negativity of any of the following: CD33 (p = 0.002), CD13 (p < 0.0001) and CD65 (p = 0.029). There was a significant correlation between CD33 and CD13 positivity (p < 0.0001) and between CD15 and CD65 positivity (p = 0.0002) whereas the positivity of no other two antigens of the ones shown correlated significantly with each other. Total number of B-precursor cases illustrated is 365.
Figure 3 Relationship of surface and cytoplasmic expression of CD66c. Percentage of surface expression of CD66c in ALL blasts is plotted against cytoplasmic expression (after cell membrane permeabilization). Samples of 20 patients at ALL diagnosis are shown, 12 CD66c negative and 8 CD66c positive. Regression coefficient R2 = 0.927
Figure 4 Transcription of CEACAM6 versus surface CD66c expression on sorted cells. FACSsorted CD66c surface negative (CD66cneg) or positive (CD66cpos) ALL lymphoblasts, five patients with heterogeneous CD66c expression were sorted into both CD66c negative and CD66c positive fraction (lines connect sorted fractions from the same specimen). Mann-Whitney test was used to compare groups (n = 32). CEACAM6n value is normalized to beta-2-microglobulin (see Methods).
Figure 5 Western blot of granulocytes, ALL samples of CD66c positive cases and surface CD66cneg cell lines with TEL/AML1pos (REH), MLL/AF4pos (RS4;11) translocation and with no fusion (NALM-6).
Figure 6 Stability of CD66c from diagnosis to relapse. Each circle represents one patient (n = 39). Percentage of CD66cpos blasts at diagnosis is plotted against percentage of CD66cpos blasts at relapse. Regression line with 95% confidence R2 = 0.755
Figure 7 Relapse free survival of cases with CD66c pos (blue line) or CD66cneg(red line) B-precursor ALL. Unselected consecutive patients treated on ALL BFM95 protocol (median follow up 3.64 years). Since surface CD66c associates with genotype, separate analyses for distinct genotype subgroups are shown.
Table 1 Frequency of CD66c and myeloid antigen expression. Cases with >20% blasts are regarded positive, coexpression of CD66c and other MyAg is tested by Fisher's exact test.
Molecule No of cases (total = 365) Proportion [%] Coexpression with CD66c
CD66c 156 43
CD33 85 23
CD15 72 20
CD13 57 16
CD65 14 3.8
CD66c and CD33 21 5.8 mutually exclusive p = 0.002
CD66c and CD15 30 8.2 random NS
CD66c and CD13 9 2.5 mutually exclusive P < 0.0001
CD66c and CD65 2 0.55 mutually exclusive p = 0.029
Table 2 Correlation between risk factors and CD66c expression. The distribution of CD66cpos and CD66cneg cases (cutoff 20%) is shown. In addition, no difference was observed in the RFS of the risk-defined subsets based on the CD66c expression (log-rank test p-value > 0.05 in all analyses). Only patients treated by a single ALL BFM-95 protocol are shown here (n = 254).
CD66cpos cases CD66cneg cases p-value (chi-square)
All patients 109 145 N/A
Prednisone poor responder 9 12 n.s.
Prednisone good responder 100 133
Initial leukocytosis = > 20 × 109/L 28 44 n.s.
Initial leukocytosis < 20 × 109/L 81 101
TEL/AML1 2 77 P < 0.0001
BCR/ABL 7 1
MLL/AF4 0 1
Hyperdiploid 55 7
Other genotype (not TEL/AML1, BCR/ABL, MLL/AF4 or hyperdiploidy) 45 59
Age 1–5 59 88 n.s.
Age >5 50 57
Standard risk group 40 58 n.s.
Intermediate risk group 54 72
High risk group 15 15
==== Refs
Markert CL Neoplasia: a disease of cell differentiation Cancer Res 1968 28 1908 1914 5676741
Greaves MF Differentiation-linked leukemogenesis in lymphocytes Science 1986 234 697 704 3535067
Mori T Sugita K Suzuki T Okazaki T Manabe A Hosoya R Mizutani S Kinoshita A Nakazawa S A novel monoclonal antibody, KOR-SA3544 which reacts to Philadelphia chromosome-positive acute lymphoblastic leukemia cells with high sensitivity Leukemia 1995 9 1233 1239 7543176
Hrusak O Trka J Zuna J Houskova J Bartunkova J Stary J Aberrant expression of KOR-SA3544 antigen in childhood acute lymphoblastic leukemia predicts TEL-AML1 negativity. The Pediatric Hematology Working Group in the Czech Republic Leukemia 1998 12 1064 1070 9665191 10.1038/sj.leu.2401072
Hrusak O Porwit-MacDonald A Antigen expression patterns reflecting genotype of acute leukemias Leukemia 2002 16 1233 1258 12094248 10.1038/sj.leu.2402504
Boccuni P Di Noto R Lo Pardo C Villa MR Ferrara F Rotoli B Del Vecchio L CD66c antigen expression is myeloid restricted in normal bone marrow but is a common feature of CD10+ early-B-cell malignancies Tissue Antigens 1998 52 1 8 9714468
Kishimoto T Kikutani H von dem Borne AEGK Goyert SM Mason DY Miyasaka M Moretta L Okumura K Shaw S Springer TA Sugamura K Zola H Leukocyte Typing VI 1997 New York, London, Garland Publishing Inc 1342
Klein ML McGhee SA Baranian J Stevens L Hefta SA Role of nonspecific cross-reacting antigen, a CD66 cluster antigen, in activation of human granulocytes Infect Immun 1996 64 4574 4579 8890209
Skubitz KM Campbell KD Skubitz AP CD66a, CD66b, CD66c, and CD66d each independently stimulate neutrophils J Leukoc Biol 1996 60 106 117 8699114
Jantscheff P Terracciano L Lowy A Glatz-Krieger K Grunert F Micheel B Brummer J Laffer U Metzger U Herrmann R Rochlitz C Expression of CEACAM6 in resectable colorectal cancer: a factor of independent prognostic significance J Clin Oncol 2003 21 3638 3646 14512395 10.1200/JCO.2003.55.135
Ordonez C Screaton RA Ilantzis C Stanners CP Human carcinoembryonic antigen functions as a general inhibitor of anoikis Cancer Res 2000 60 3419 3424 10910050
Duxbury MS Ito H Zinner MJ Ashley SW Whang EE CEACAM6 gene silencing impairs anoikis resistance and in vivo metastatic ability of pancreatic adenocarcinoma cells Oncogene 2004 23 465 473 14724575 10.1038/sj.onc.1207036
Sugita K Mori T Yokota S Kuroki M Koyama TO Inukai T Iijima K Goi K Tezuka T Kojika S Shiraishi K Nakamura M Miyamoto N Karakida N Kagami K Nakazawa S The KOR-SA3544 antigen predominantly expressed on the surface of Philadelphia chromosome-positive acute lymphoblastic leukemia cells is nonspecific cross-reacting antigen-50/90 (CD66c) and invariably expressed in cytoplasm of human leukemia cells Leukemia 1999 13 779 785 10374883 10.1038/sj/leu/2401408
Campana D Coustan-Smith E Detection of minimal residual disease in acute leukemia by flow cytometry Cytometry 1999 38 139 152 10440852 10.1002/(SICI)1097-0320(19990815)38:4<139::AID-CYTO1>3.0.CO;2-H
Campana D Coustan-Smith E Advances in the immunological monitoring of childhood acute lymphoblastic leukaemia Best Pract Res Clin Haematol 2002 15 1 19 11987913 10.1053/beha.2002.0182
Bene MC Castoldi G Knapp W Ludwig WD Matutes E Orfao A van't Veer MB Proposals for the immunological classification of acute leukemias. European Group for the Immunological Characterization of Leukemias (EGIL) Leukemia 1995 9 1783 1786 7564526
Chomczynski P Sacchi N Single-step method of RNA isolation by acid guanidinium thiocyanate-phenol-chloroform extraction Anal Biochem 1987 162 156 159 2440339 10.1016/0003-2697(87)90021-2
Baranov V Yeung MM Hammarstrom S Expression of carcinoembryonic antigen and nonspecific cross-reacting 50-kDa antigen in human normal and cancerous colon mucosa: comparative ultrastructural study with monoclonal antibodies Cancer Res 1994 54 3305 3314 8205554
Madzo J Zuna J Muzikova K Kalinova M Krejci O Hrusak O Otova B Stary J Trka J Slower molecular response to treatment predicts poor outcome in patients with TEL/AML1 positive acute lymphoblastic leukemia: prospective real-time quantitative reverse transcriptase-polymerase chain reaction study Cancer 2003 97 105 113 12491511 10.1002/cncr.11043
Muller HJ Beier R Loning L Blutters-Sawatzki R Dorffel W Maass E Muller-Weihrich S Scheel-Walter HG Scherer F Stahnke K Schrappe M Horn A Lumkemann K Boos J Pharmacokinetics of native Escherichia coli asparaginase (Asparaginase medac) and hypersensitivity reactions in ALL-BFM 95 reinduction treatment Br J Haematol 2001 114 794 799 11564065 10.1046/j.1365-2141.2001.03009.x
Dworzak MN Froschl G Printz D Mann G Potschger U Muhlegger N Fritsch G Gadner H Prognostic significance and modalities of flow cytometric minimal residual disease detection in childhood acute lymphoblastic leukemia Blood 2002 99 1952 1958 11877265 10.1182/blood.V99.6.1952
Hanenberg H Baumann M Quentin I Nagel G Grosse-Wilde H von Kleist S Gobel U Burdach S Grunert F Expression of the CEA gene family members NCA-50/90 and NCA-160 (CD66) in childhood acute lymphoblastic leukemias (ALLs) and in cell lines of B-cell origin Leukemia 1994 8 2127 2133 7808000
Scholzel S Zimmermann W Schwarzkopf G Grunert F Rogaczewski B Thompson J Carcinoembryonic Antigen Family Members CEACAM6 and CEACAM7 Are Differentially Expressed in Normal Tissues and Oppositely Deregulated in Hyperplastic Colorectal Polyps and Early Adenomas Am J Pathol 2000 156 595 605 10666389
Duxbury MS Ito H Benoit E Zinner MJ Ashley SW Whang EE Overexpression of CEACAM6 promotes insulin-like growth factor I-induced pancreatic adenocarcinoma cellular invasiveness Oncogene 2004 23 5834 5842 15208677 10.1038/sj.onc.1207775
Duxbury MS Ito H Ashley SW Whang EE CEACAM6 as a novel target for indirect type 1 immunotoxin-based therapy in pancreatic adenocarcinoma Biochem Biophys Res Commun 2004 317 837 843 15081416 10.1016/j.bbrc.2004.03.128
Putti MC Rondelli R Cocito MG Arico M Sainati L Conter V Guglielmi C Cantu-Rajnoldi A Consolini R Pession A Zanesco L Masera G Biondi A Basso G Expression of Myeloid Markers Lacks Prognostic Impact in Children Treated for Acute Lymphoblastic Leukemia: Italian Experience in AIEOP-ALL 88-91 Studies Blood 1998 92 795 801 9680347
San Miguel JF Ciudad J Vidriales MB Orfao A Lucio P Porwit-MacDonald A Gaipa G van Wering E van Dongen JJ Immunophenotypical detection of minimal residual disease in acute leukemia Crit Rev Oncol Hematol 1999 32 175 185 10633847
Mejstrikova E Kalina T Trka J Stary J Hrusak O Correlation of CD33 with poorer prognosis in childhood ALL implicates a potential of anti-CD33 frontline therapy Leukemia 2005 in press 15830012 10.1038/sj.leu.2403737
| 15826304 | PMC1112585 | CC BY | 2021-01-04 16:03:05 | no | BMC Cancer. 2005 Apr 12; 5:38 | utf-8 | BMC Cancer | 2,005 | 10.1186/1471-2407-5-38 | oa_comm |
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-431586013410.1186/1471-2407-5-43Research ArticleA novel duplication polymorphism in the FANCA promoter and its association with breast and ovarian cancer Thompson Ella [email protected] Rebecca L [email protected] Sally-Anne [email protected] Diana M [email protected] Ian G [email protected] Alexander [email protected] Department of Pathology, Peter MacCallum Cancer Centre, Locked Bag 1 A'Beckett St, Melbourne, Victoria 8006, Australia2 University of Adelaide Department of Medicine and Department of Haematology/Oncology, The Queen Elizabeth Hospital, Adelaide, South Australia 5011, Australia3 Wessex Clinical Genetics Service, Princess Anne Hospital, Southampton, SO16 5YA, UK4 Centre for Genomics and Predictive Medicine, Peter MacCallum Cancer Centre, Locked Bag 1, A'Beckett St, Melbourne, Victoria, 8006, Australia5 Department of Pathology, University of Melbourne, Parkville Victoria, 3002, Australia2005 29 4 2005 5 43 43 15 9 2004 29 4 2005 Copyright © 2005 Thompson 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 FANCA gene is one of the genes in which mutations lead to Fanconi anaemia, a rare autosomal recessive disorder characterised by congenital abnormalities, bone marrow failure, and predisposition to malignancy. FANCA is also a potential breast and ovarian cancer susceptibility gene. A novel allele was identified which has a tandem duplication of a 13 base pair sequence in the promoter region.
Methods
We screened germline DNA from 352 breast cancer patients, 390 ovarian cancer patients and 256 normal controls to determine if the presence of either of these two alleles was associated with an increased risk of breast or ovarian cancer.
Results
The duplication allele had a frequency of 0.34 in the normal controls. There was a non-significant decrease in the frequency of the duplication allele in breast cancer patients. The frequency of the duplication allele was significantly decreased in ovarian cancer patients. However, when malignant and benign tumours were considered separately, the decrease was only significant in benign tumours.
Conclusion
The allele with the tandem duplication does not appear to modify breast cancer risk but may act as a low penetrance protective allele for ovarian cancer.
==== Body
Background
Fanconi anaemia is a rare autosomal recessive disorder characterised by congenital abnormalities, progressive bone marrow failure, and predisposition to acute myelogenous leukemia and other malignancies. At the cellular level, the disease is characterised by an inability to repair cross-linked DNA [1]. There are multiple genes in which mutations can give rise to Fanconi anaemia.
The FANCA gene is defective in more than 65% of Fanconi anaemia cases. FANCA and five other Fanconi anaemia genes code for components of a complex that is required for the ubiquitination of FANCD2 in response to DNA damage. Ubiquitinated FANCD2 is targeted to nuclear foci of DNA repair proteins including BRCA1 and RAD51 (reviewed in [2]). The complex also interacts directly with the FANCD1 protein [3]), now known to be the product of the breast and ovarian cancer predisposition gene, BRCA2 [4]. As individuals heterozygous for BRCA2 mutations have a high lifetime risk of acquiring breast and ovarian cancer, it is likely that alterations in other Fanconi anaemia genes might be associated with an increased risk of breast and ovarian cancer.
In human breast and ovarian cancers, recurrent loss of heterozygosity has been shown to occur on the long arm of chromosome 16 [5,6] and is localised to 16q24.3 [7-9]. This suggests the presence of one or more tumour suppressor genes in this region. However, no recurrent tumour-specific mutations in any of the 16q24.3 candidate tumour suppressor genes assessed for mutations have been reported in breast tumours (e.g. [10-12]).
FANCA localises to 16q24.3 [13] and is a plausible tumour suppressor gene candidate because of its role in the repair of DNA damage. Cleton-Jansen et al [14] did not detect any mutations in 19 cases of breast cancer with 16q24.3 loss of heterozygosity and concluded that FANCA was not the tumour suppressor gene underlying 16q24.3 loss of heterozygosity. Nevertheless, FANCA remains an attractive candidate as either a cancer predisposition gene or a target of genetic or epigenetic inactivation in sporadic tumours.
While screening the FANCA promoter region by single strand conformation analysis, we identified a polymorphism in the FANCA promoter region. As promoter polymorphisms can alter the transcription or regulation of a gene, we sought to determine whether one of these two alleles might be associated with an altered risk of developing breast or ovarian cancer.
Methods
Subjects
All breast cancer cases were systematically ascertained through breast clinics in the Wessex region of southern England as described previously [15,16] These cases were selected on the basis of an age at onset under 40 years, a family history of breast/ovarian cancer (defined as two or more cases of breast/ovarian cancer in a first or second degree female relative) or bilateral breast cancer irrespective of family history or age at onset. Family histories were verified as far as possible from medical records and death certificates. Blood was taken from all recruits who consented to molecular analysis for breast cancer predisposition genes. The age range of the breast cancer participants was 19–76 with a mean age of 40 years
Details of the ovarian tumour cases have been described previously [17,18]. Briefly, cases of ovarian tumours were ascertained from women undergoing primary surgery in hospitals from southern England between 1993–1998. A specialist gynaecological pathologist confirmed the histological diagnosis for each tumour. A total of 390 ovarian tumours were included in the study consisting of 313 malignant (127 serous, 82 endometrioid, 42 mucinous, 13 clear cell, 49 undifferentiated adenocarcinomas), 15 borderline tumours (11 mucinous, 4 serous) and 62 benign tumours (18 fibromas, 26 serous, 18 mucinous). The age range of the ovarian cancer patients was 23–90 with a mean of 62 years.
The controls represent the population from which the cases arose and consisted of 256 Caucasian female volunteers who were either patients attending for non-neoplastic disease conditions or staff at the Princess Anne Hospital, Southampton. The age of the controls ranged from 18 – 84 with a mean age of 39 years. Both control and cancer groups were drawn from the same geographical area. Epidemiological data such as reproductive factors, oral contraceptive use, smoking and obesity were not available for any of the cancer or control groups. We have obtained approval for this research from the appropriate ethics committees and are in compliance with the Helsinki Declaration.
Molecular genetic analysis
DNA for genotyping was prepared from the peripheral blood samples of patients and controls. The region containing the duplication was amplified using the primers 5'CCAAACGCAAAAACTACCTCACCG3' and 5'CGCTGCCTTCCTATTGGCTGC3'. Fifty ng of DNA was used in a 25 μl reaction using 0.5 U HotStarTaq (Qiagen, Hilden, Germany), 200 nM of each of the primers, 800 μM total nucleotides in the buffer supplied by the manufacturer (Qiagen) and 1.5x Q solution (Qiagen). Cycling conditions were 15 min at 95°C, 11 cycles of 95°C/45 s, 60°-50°C/45 s (decreasing by 1°C per cycle), 72°C/45 s, followed by 34 cycles of 95°C/45 s, 50°C/45 s, 72°C/45 s and finally 10 min at 72°C. The PCR primers amplified a 151 base pair product for allele 1 and a 164 base pair product for allele 2. The products were separated by electrophoresis on 3% agarose gels. PCR products from the three genotypes were sequenced on the ABI Prism 3100 Genetic Analyser (Applied Biosystems, Foster City, CA) using BigDye Terminator v3.1 chemistry (Applied Biosystems).
Statistical analysis
The Hardy-Weinberg equilibrium was assessed by the standard methods. The data were considered using models assuming dominant inheritance (i.e., women with one or two duplication alleles had the same relative hazard), codominant inheritance (i.e., the relative hazard differed between women with one duplication allele compared with those with two duplication alleles), or recessive inheritance (i.e., only women with two duplication alleles were at increased risk). For all analyses, the control subjects were treated as a single group without stratification. Fisher's Exact test using Instat 3.01 software (Graphpad Software) was used to calculate the significance (p value) and odds ratio (OR) with a 95% confidence interval. All statistical calculations were two-sided and p values were considered statistically significant when less than 0.05.
Results
Identification of the FANCA promoter polymorphism
We identified a polymorphism in the FANCA promoter region using single strand conformation analysis. Variant bands were purified by PCR amplification from a band stab and sequenced in both directions directly from the PCR product. Comparison of the sequences obtained for each band and the sequence of the FANCA promoter (Genbank Accession AC005360) showed that the variation in the patterns was due to the presence of either a single or duplicated 13 base pair sequence.
The 13 base pair sequence is located at -98 to -110 bases upstream of the beginning of transcription as defined by the NCBI reference sequence for FANCA (NM 000135.1 23-Dec 2003). The translation start site is 32 base pairs further downstream (Figure 1). Allele 1 has a single copy of the sequence GGCCACGACGCAA. Allele 2 has two tandemly arranged copies of this sequence. Interestingly, the sequence immediately downstream of the 13 base pair sequence, GGCCtCGACctgA shows considerable homology to the 13 base pair sequence (divergent nucleotides in lower case).
Case control study of the FANCA promoter polymorphism
We designed a PCR assay in which both alleles are amplified simultaneously using a set of primers flanking the polymorphism (Figure 2). The frequency of the polymorphism was determined in breast cancer patients, ovarian cancer patients and controls to assess if the presence of either allele was associated with a predisposition to breast or ovarian cancer.
Table 1 shows the distribution of genotypes in the breast and ovarian cancer patients and the controls. The distribution of the genotypes within each of the groups did not deviate significantly from those expected under Hardy-Weinberg equilibrium.
The frequency of the duplication allele (allele 2) in the breast cancer patients was not significantly different from the controls (0.32 versus 0.34, p = 0.53). The distribution of genotypes in the breast cancer cases was similar to the controls. The genotype distribution was similar when stratifying the breast cancers according to the selection criteria although the frequency of the 12 and 22 genotypes showed a non-significant elevation among the family history group.
The frequency of the duplication allele in the ovarian cancer patients was 0.29 which was significantly different from the controls (p = 0.048). In addition, the ovarian cancer patients had a nominally significant decrease in the number of 12 and 22 genotype carriers compared to the controls (p = 0.05) suggesting that the duplication may be protective for the disease with an odds ratio of 0.72 (95% CI, 0.53–0.99). All subgroups of the ovarian cancer patients showed a reduced frequency of 12 and 22 genotypes, with the exception of patients with borderline tumours and clear cell carcinomas. However, the number of cases in these latter categories was very small. Interestingly, the 62 patients with benign tumours showed a highly significant decrease in the 12 and 22 genotype frequency (p = .007) with an odds ratio of 0.46 (95% CI 0.26–0.81).
Discussion
The identification of BRCA2 as a member (FANCD1) of the Fanconi anaemia group of genes has raised the possibility that variation in other Fanconi anaemia genes may predispose to breast or ovarian cancer. The novel 13 base pair duplication allele identified in this study is of interest as even single base changes in promoter sequences can alter regulation of gene expression and contribute to tumorigenesis, particularly if this affects a transcription factor binding site (e.g. [19,20]). As both alleles of the FANCA polymorphism are common, it is unlikely that either allele would represent a high penetrance predisposition allele. We therefore undertook a case control study to determine whether either allele might be responsible for a more modest increase in the rate of breast and or ovarian cancer.
Genotyping of the promoter polymorphism in high risk breast cancer patients showing features of genetic predisposition revealed no significant difference in the allele or genotype distribution compared to the normal controls. The study had 80% power to detect an odds ratio of ≥ 1.6 for carriers heterozygous for the duplication and an odds ratio ≥ 1.9 for carriers homozygous for the duplication. As we could not adjust for known breast cancer risk factors, we cannot exclude the possibility that confounding factors may have led to a type II error. However, confounding due to differences in ethnicity is unlikely as both the cases and controls were residents of the Southampton area, which has a predominantly Anglo-Saxon population. Nevertheless, studies of breast cancer cases selected based on other clinical characteristics, such as postmenopausal onset, may be warranted.
Among the ovarian tumour cases, there was a significant decrease in the frequency of the combined 12 and 22 genotypes, suggesting that allele 2 may protect against ovarian cancer. The trend was particularly evident among the benign tumours (p = .0007). However, it is possible that this association represents a type I error due to confounding factors such as ethnicity and the influence of known risk factors such as oral contraceptive use. Consequently, it will be important to replicate our findings in larger population-based case-control studies.
Conclusion
We have identified a novel promoter polymorphism in FANCA. It is unknown what biological effect the duplication might have but it may alter the basal rate of transcription or the regulation of transcription. As FANCA is likely to function as a tumour suppressor, it is plausible that allelic variants with altered activity may modify cancer risk. This pilot study has provided some evidence of an association with ovarian cancer. Further studies to verify this association are warranted as well as studies involving predisposition to other cancers.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
ET performed the genotyping of the breast and ovarian cancer patients and controls. RD identified the polymorphism and did the initial breast cancer patient genotyping. ET and RD contributed equally to this work. SS was involved in planning through the course of the project and helped prepare the manuscript. DE collected the patients used in this study. IC provided the DNA samples, participated in shaping the study and analysed the results. AD conceived the study, supervised the research and prepared and revised the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Figures and Tables
Figure 1 Sequence of FANCA promoter allele in the polymorphic region. Allele 1 has a single copy of a 13 base pair sequence. This has been defined as GGCCACGACGCAA in accordance with current nomenclature for mutations [21] although it could theoretically be GCCACGACGCAAG or CCACGACGCAAGG. Allele 2 has two tandemly arranged copies of this sequence. The 13 base pair sequence is located at -98 to -110 bases upstream of the beginning of transcription as defined by the NCBI reference sequence NM_000135.1. Exon 1 is in bold and the coding sequence in capitals.
Figure 2 Genotyping the FANCA promoter polymorphism by PCR. Allele 1 amplifies as a band of 151 base pair, allele 2 as a band of 164 base pair. The 3 genotypes are readily distinguishable on a 3 % agarose gel run at 100 V for 1 hour. Homozygous samples for both allele 1 and allele 2 are shown as are heterozygous samples which have both bands.
Table 1 The distribution of FANCA promoter polymorphism genotypes in breast and ovarian cancer cases and controls
Study Group No. 11 12 22 12 + 22
n (%) n (%) n (%) n (%) p valuea ORb (95% CI)
Control 256 112 (43.8) 115 (44.9) 29 (11.3) 144 (56.2)
All Breast Cancer 352 164 (46.6) 150 (42.6) 38 (10.8) 188 (53.4) 0.51 0.89 (0.64–1.23)
Diagnosis ≤ 40 years 203 104 (51.2) 79 (38.9) 20 (9.9) 99 (48.8) 0.13 0.74 (0.51–1.07
Family History 105 39 (37.1) 53 (50.5) 13 (12.4) 66 (62.9) 0.29 1.32 (0.83–2.10)
Bilateral Disease 44 21 (47.7) 18 (40.9) 5 (11.4) 23 (52.3) 0.63 0.85 (0.45–1.62)
All Ovarian Tumours 390 202 (51.8) 153 (39.2) 35 (9.0) 188 (48.2) 0.05 0.72 (0.53–0.99)
All Malignant 313 157 (50.2) 126 (40.2) 30 (9.6) 156 (49.8) 0.13 0.77 (0.55–1.08)
Serous 127 62 (48.8) 52 (40.9) 13 (10.2) 65 (51.1) 0.38 0.81 (0.53–1.25)
Endometrioid 82 44 (53.6) 29 (35.4) 9 (11.0) 38 (46.4) 0.13 0.67 (0.41–1.11)
Mucinous 42 23 (54.8) 16 (38.1) 3 (7.1) 19 (45.2) 0.24 0.64 (0.33–1.24)
Clear Cell 13 4 (30.8) 8 (61.5) 1 (7.7) 9 (69.2) 0.40 1.75 (0.52–5.83)
Adenocarcinoma 49 24 (49.0) 21 (42.8) 4 (8.2) 25 (51.0) 0.53 0.81 (0.44–1.49)
Borderline 15 6 (40.0) 8 (53.3) 1 (6.7) 9 (60.0) 1.00 1.16 (0.40–3.37)
Benign 62 39 (62.9) 19 (30.6) 4 (6.4) 23 (37.0) 0.007 0.46 (0.26–0.81)
Allele 1 corresponds to the single copy allele, allele 2 corresponds to the duplication allele.
aFisher's exact test (two-sided) for the combined 12/22 genotype frequency using the 11 homozygotes as reference.
bThe odds ratio (OR) and 95% confidence intervals (CI) are shown in parentheses.
==== Refs
Fujiwara Y Tatsumi M Sasaki M Cross-link repair in human cells and its possible defect in Fanconi's anemia cells J Mol Biol 1977 113 635 649 894713 10.1016/0022-2836(77)90227-3
D'Andrea AD Grompe M The Fanconi anaemia/BRCA pathway Nat Rev Cancer 2003 3 23 34 12509764 10.1038/nrc970
Hussain S Witt E Huber PA Medhurst AL Ashworth A Mathew CG Direct interaction of the Fanconi anaemia protein FANCG with BRCA2/FANCD1 Hum Mol Genet 2003 12 2503 2510 12915460 10.1093/hmg/ddg266
Howlett NG Taniguchi T Olson S Cox B Waisfisz Q De Die-Smulders C Persky N Grompe M Joenje H Pals G Ikeda H Fox EA D'Andrea AD Biallelic inactivation of BRCA2 in Fanconi anemia Science 2002 297 606 609 12065746 10.1126/science.1073834
Dutrillaux B Gerbault-Seureau M Zafrani B Characterization of chromosomal anomalies in human breast cancer. A comparison of 30 paradiploid cases with few chromosome changes Cancer Genet Cytogenet 1990 49 203 217 2170003 10.1016/0165-4608(90)90143-X
Sato T Tanigami A Yamakawa K Akiyama F Kasumi F Sakamoto G Nakamura Y Allelotype of breast cancer: cumulative allele losses promote tumor progression in primary breast cancer Cancer Res 1990 50 7184 9 1977515
Radford DM Fair KL Phillips NJ Ritter JH Steinbrueck T Holt MS Donis-Keller H Allelotyping of ductal carcinoma in situ of the breast: deletion of loci on 8p, 13q, 16q, 17p and 17q Cancer Res 1995 55 3399 3405 7614479
Chen T Sahin A Aldaz CM Deletion map of chromosome 16q in ductal carcinoma in situ of the breast: refining a putative tumour suppressor gene region Cancer Res 1996 56 5605 5609 8971163
Launonen V Mannermaa A Stenback F Kosma VM Puistola U Huusko P Anttila M Bloigu R Saarikoski S Kauppila A Winqvist R Loss of heterozygosity at chromosomes 3, 6, 8, 11, 16, and 17 in ovarian cancer: correlation to clinicopathological variables Cancer Genet Cytogenet 2000 122 49 54 11104033 10.1016/S0165-4608(00)00279-X
Moerland E Breuning MH Cornelisse CJ Cleton-Jansen AM Exclusion of BBC1 and CMAR as candidate breast tumour-suppressor genes Br J Cancer 1997 76 1550 1553 9413939
Whitmore SA Settasatian C Crawford J Lower KM McCallum B Seshadri R Cornelisse CJ Moerland EW Cleton-Jansen AM Tipping AJ Mathew CG Savnio M Savoia A Verlander P Auerbach AD Van Berkel C Pronk JC Doggett NA Callen DF Characterization and screening for mutations of the growth arrest-specific 11 (GAS11) and C16orf3 genes at 16q24.3 in breast cancer Genomics 1998 52 325 331 9790751 10.1006/geno.1998.5457
Kochetkova M McKenzie OL Bais AJ Martin JM Secker GA Seshadri R Powell JA Hinze SJ Gardner AE Spendlove HE O'Callaghan NJ Cleton-Jansen AM Cornelisse C Whitmore SA Crawford J Kremmidiotis G Sutherland GR Callen DF CBFA2T3 (MTG16) is a putative breast tumor suppressor gene from the breast cancer loss of heterozygosity region at 16q24.3 Cancer Res 2002 62 4599 4604 12183414
Pronk JC Gibson RA Savoia A Wijker M Morgan NV Melchionda S Ford D Temtamy S Ortega JJ Jansen S Localisation of the Fanconi anaemia complementation group A gene to chromosome 16q24.3 Nat Genet 1995 1 338 340 10.1038/ng1195-338
Cleton-Jansen AM Moerland EW Pronk JC van Berkel CG Apostolou S Crawford J Savoia A Auerbach AD Mathew CG Callen DF Cornelisse CJ Mutation analysis of the Fanconi anaemia A gene in breast tumors with loss of heterozygosity at 16q24.3 Br J Cancer 1999 79 1049 1052 10098735 10.1038/sj.bjc.6690168
Eccles D Marlow A Royle G Collins A Morton NE Genetic epidemiology of early onset breast cancer J Med Genet 1994 31 944 949 7891376
Eccles DM Englefield P Soulby MA Campbell IG BRCA1 mutations in southern England Br J Cancer 1998 77 2199 2203 9649133
Manolitsas TP Englefield P Eccles DM Campbell IG No association of a 306-bp insertion polymorphism in the progesterone receptor gene with ovarian and breast cancer Brit J Cancer 1997 75 1398 1399 9155067
Morland SJ Jiang X Hitchcock A Thomas EJ Campbell IG Mutation of galactose-1-phosphate uridyl transferase and its association with ovarian cancer and endometriosis Int J Cancer 1998 77 825 827 9714048 10.1002/(SICI)1097-0215(19980911)77:6<825::AID-IJC4>3.0.CO;2-W
Zhu Y Spitz MR Lei L Mills GB Wu X A single nucleotide polymorphism in the matrix metalloproteinase-1 promoter enhances lung cancer susceptibility Cancer Res 2001 61 7825 782 11691799
Bond GL Hu W Bond EE Robins H Lutzker SG Arva NC Bargonetti J Bartel F Taubert H Wuerl P Onel K Yip L Hwang SJ Strong LC Lozano G Levine AJ A single nucleotide polymorphism in the MDM2 promoter attenuates the p53 tumor suppressor pathway and accelerates tumor formation in humans Cell 2004 119 591 602 15550242 10.1016/j.cell.2004.11.022
| 15860134 | PMC1112586 | CC BY | 2021-01-04 16:03:03 | no | BMC Cancer. 2005 Apr 29; 5:43 | utf-8 | BMC Cancer | 2,005 | 10.1186/1471-2407-5-43 | oa_comm |
==== Front
BMC Dev BiolBMC Developmental Biology1471-213XBioMed Central London 1471-213X-5-81584016510.1186/1471-213X-5-8Research ArticleC. elegans serine-threonine kinase KIN-29 modulates TGFβ signaling and regulates body size formation Maduzia Lisa L [email protected] Andrew F [email protected] Huang [email protected] Xia [email protected] Lena J [email protected] Cole M [email protected] Stephen [email protected] Xin-Hua [email protected] Richard W [email protected] Waksman Institute, Department of Molecular Biology and Biochemistry, and Cancer Institute of New Jersey, Rutgers University, Piscataway, NJ 08854-8020, USA2 Michael E. DeBakey Department of Surgery, Baylor College of Medicine, Houston, TX 77030, USA3 Department of Molecular and Cellular Biology, Baylor College of Medicine, Houston, TX 77030, USA2005 19 4 2005 5 8 8 19 11 2004 19 4 2005 Copyright © 2005 Maduzial et al; licensee BioMed Central Ltd.2005Maduzial 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 C. elegans there are two well-defined TGFβ-like signaling pathways. The Sma/Mab pathway affects body size morphogenesis, male tail development and spicule formation while the Daf pathway regulates entry into and exit out of the dauer state. To identify additional factors that modulate TGFβ signaling in the Sma/Mab pathway, we have undertaken a genetic screen for small animals and have identified kin-29.
Results
kin-29 encodes a protein with a cytoplasmic serine-threonine kinase and a novel C-terminal domain. The kinase domain is a distantly related member of the EMK (ELKL motif kinase) family, which interacts with microtubules. We show that the serine-threonine kinase domain has in vitro activity. kin-29 mutations result in small animals, but do not affect male tail morphology as do several of the Sma/Mab signal transducers. Adult worms are smaller than the wild-type, but also develop more slowly. Rescue by kin-29 is achieved by expression in neurons or in the hypodermis. Interaction with the dauer pathway is observed in double mutant combinations, which have been seen with Sma/Mab pathway mutants. We show that kin-29 is epistatic to the ligand dbl-1, and lies upstream of the Sma/Mab pathway target gene, lon-1.
Conclusion
kin-29 is a new modulator of the Sma/Mab pathway. It functions in neurons and in the hypodermis to regulate body size, but does not affect all TGFβ outputs, such as tail morphogenesis.
==== Body
Background
The transforming growth factor β (TGFβ) superfamily is involved in many developmental decisions from primitive animals such as Cnidaria and sponges to higher animals [1-4]. The core of the signaling pathway has been elucidated in the last few years and reveals a rather simple signaling cascade. These ligands transmit the TGFβ signal by binding transmembrane receptor serine-threonine kinases (RSKs). Once ligand is bound, the type II RSK phosphorylates the type I RSK in a cytoplasmic region rich in glycine and serine residues (GS domain). Phosphorylation activates the type I RSK and enables it to phosphorylate downstream mediators referred to as the Smads. Once the receptor-regulated Smads (R-Smads) are phosphorylated, they are able to physically interact with another subset of Smads identified as the common Smads (Co-Smads) and translocate to the nucleus where they affect target gene transcription [3,5-7].
Of the five TGFβ-like ligands in C. elegans, dbl-1 (dpp and BMP-like) and daf-7 (dauer formation abnormal) are the best characterized. dbl-1 transmits the Sma/Mab (small/male tail abnormal) pathway signal while daf-7 regulates formation of dauer, an alternative life stage entered in response to low food or high population density [8-10]. These two pathways share a common type II RSK, daf-4. daf-4 animals are small, exhibit fused male tail sensory rays and constitutively form dauer larvae [11,12]. Mutants of all other known components of the dauer pathway are either dauer constitutive (daf-c) like daf-7 or dauer defective (daf-d) [3,13]. Based on the Sma/Mab phenotypes of daf-4 mutants, sma-2, sma-3, sma-4 and sma-6 were identified and cloned. sma-2, sma-3 and sma-4 encode Smads while sma-6 encodes a type I RSK [12,14]. The Sma/Mab signal is transmitted upon binding of the ligand, DBL-1, to the type II and type I RSKs, DAF-4 and SMA-6 respectively. Once stimulated, SMA-6 activates the Smads, SMA-2, -3 and -4, causing them to affect target gene transcription.
Although the core TGFβ pathway is known, additional components that may further refine signaling remain to be identified. To address this issue, we previously conducted a genetic screen for Sma animals and isolated several new mutants, including kin-29 (also known as sma-11) [15]. We find that kin-29 is able to suppress the long mutant phenotype generated by animals over-expressing the ligand dbl-1. Additionally, we observe that lon-1, a Sma/Mab pathway target gene whose product shows homology to proteins of the cysteine-rich secretory protein (CRISP) family, is up-regulated in kin-29(lf) mutant animals in a similar manner to that seen in a sma-6 null mutant background [16,17]. kin-29 mutant animals are also developmentally delayed and this defect is partially suppressed by loss of lon-1 function. These data suggest that kin-29 genetically interacts with Sma/Mab pathway signaling downstream of dbl-1 but upstream of lon-1.
Several of the Sma/Mab pathway components have been shown to function in the hypodermis to regulate body size morphogenesis. sma-3, sma-6 and lon-1, when specifically expressed in the hypodermis, have been shown to rescue the body size defects associated with each of these loss-of-function mutations [15-18]. The Sma/Mab ligand DBL-1 is primarily expressed in neuronal tissues [8]. It is likely that DBL-1 is secreted from these tissues and targets the hypodermis in order to regulate body size formation. We find that tissue-specific expression of kin-29(+) in the hypodermis rescues the small body size phenotype of kin-29(lf) animals. In addition, we find that kin-29(+), when expressed in the same tissues as dbl-1, also rescues the small body size phenotype of kin-29(lf) animals. Therefore, kin-29 can function in both hypodermal and neuronal tissues with known Sma/Mab pathway components to regulate body size morphogenesis.
In order to understand how kin-29 functions in Sma/Mab pathway signaling, we undertook the molecular characterization of kin-29. It is encoded by F58H12.1, which has an N-terminal kinase domain and a novel C-terminal region. Its kinase domain makes it a distant member of the EMK kinase family, which modulates microtubule organization. kin-29 has a role in olfaction [19], suggesting that the ability to sense environmental signals influences body size regulation.
Results
kin-29 encodes a serine threonine kinase
Mutations in members of the Sma/Mab TGFβ-like signaling pathway result in animals that are phenotypically smaller than wild-type. These Sma/Mab mutants are approximately 70% the size of their wild-type counterparts [8,9,11,12,14]. Based on this small body size phenotype, we set out to isolate additional loci which when mutated resulted in small animals. From an F2 EMS screen of N2 wild-type animals, we identified kin-29(wk61) [15]. kin-29(lf) animals are small, like known Sma/Mab pathway components (Table 1). However, unlike the known pathway components, kin-29 does not possess male tail ray fusions or crumpled spicules, suggesting that kin-29 is involved in regulation of body size morphogenesis but not male tail development.
Table 1 Body size measurements of kin-29 alleles
Genotype* Perimeter (mm)** %N2 n
N2 2.60 ± 0.14 45
sma-6(wk7) 1.85 ± 0.20 71% 46
kin-29(wk61) 1.97 ± 0.19 76% 41
kin-29(oy38) 2.20 ± 0.18 85% 42
kin-29(oy39) 1.98 ± 0.22 76% 38
lon-1(wk50) 2.72 ± 0.19 105% 48
lon-1(wk50);kin-29(wk61) 2.49 ± 0.17 96% 35
* All animals were measured 48 hours after L4.
** Data are means ± std.
In comparison to N2, all animals are significantly different in size (p < 0.0001).
kin-29 was mapped to linkage group X between unc-2 and fax-1. Appropriate YACs, cosmids, and DNA fragments were used to rescue the gene. The longest cDNA available, y293c7 (Y. Kohara, National Institute of Genetics), spanning this open reading frame was obtained. Based on the ORF sequenced from y293c7, a 10 kb region of genomic DNA containing kin-29 was fused in frame with GFP. This construct, kin-29p::kin-29 gfp, was then injected into kin-29 mutant animals and conferred rescue (Table 3).
Table 3 Rescue of kin-29(wk61) by promoter fusion constructs
Genotype* Perimeter (mm)** %N2 n
kin-29(wk61) 2.06 ± 0.13 79% 36
kin-29(wk61);kin-29p::kin-29:gfp 2.60 ± 0.12 99% 33
kin-29(wk61);elt-3p::kin-29:gfp (hypodermal) 2.41 ± 0.19 92% 35
kin-29(wk61);dbl-1p::kin-29:gfp (neuronal) 2.47 ± 0.18 94% 35
* All animals were measured 48 hours after L4.
** Data are means ± std.
In comparison to kin-29(wk61), all animals are significantly different in size from control (p < 0.0001).
We searched for molecular lesions in kin-29(wk61). Genomic DNA spanning the entire coding region of kin-29 was isolated from kin-29(wk61) animals, sequenced and compared to sequence obtained from EST y293c7. Sequence analysis reveals kin-29 to consist of 16 exons that encode a protein of 822 amino acids in length (Fig. 1). A mutation found in the eighth exon of kin-29(wk61) changes a single nucleotide from cytosine to thymine. This change results in a premature termination codon and a truncated protein of 273 amino acids. While this work was in progress, kin-29 was cloned as a modifier of olfactory gene expression [19]. Two alleles from that study, kin-29(oy38) and kin-29(oy39), result from a 526 bp deletion, which is replaced by sequence found on LG X, and a missense mutation in the kinase domain, respectively [19].
Figure 1 kin-29 encodes a serine-threonine kinase. Schematic of kin-29 exon/intron structure including 16 exons. The shaded region at the N-terminus consists of the kinase domain with the Q-to-stop mutation of kin-29(wk61) shown. The function of the C-terminus has not yet been determined but it may act as a regulator of kinase activity.
kin-29 encodes a predicted serine-threonine kinase. Within its kinase domain, KIN-29 is homologous to members of the ELKL motif kinase (EMK) family and salt-induced kinase family (~66% identity)(Fig. 2A,2B). Members of the EMK family include C. elegans PAR-1, Drosophila PAR-1, S. pombe KIN-1, and mammalian MARK (microtubule-affinity-regulating kinase) [20-23]. EMK family members have been shown to affect cell polarity as well as microtubule stability [20,22,23]. The kinase domain of KIN-29 also shows significant homology salt-induced kinases [24,25]. Salt induced kinases (SIK) were cloned from subtractive libraries derived from genes expressed in the adrenal glands after high salt diets in rat. The biological function of these kinases is not clear. The C-terminal domain of KIN-29 is more divergent and shows little homology with domains in other kinases.
Figure 2 Molecular analysis of the kinase domain. (A) Sequence alignment between the kinase domains of KIN-29 (~amino acids 16–267), rat salt-induced kinase (rSalt-Ind Kinase), Drosophila CG4290 (a salt-induced kinase member), and mouse EMK (mEMK). Identical matches in three of the four sequences are indicated by white letters. (B) Dendrogram showing the relationship between the kinase domain of KIN-29 (~amino acids 16–267), and several additional kinases.
KIN-29 encodes a functional kinase
In order to assess whether KIN-29 acts as a functional kinase, 293T cells were transfected with either C-terminal FLAG-tagged full length kin-29 or various constructs, which truncate the carboxy terminal region of the protein. As controls, both the kinase active and kinase inactive mammalian TGFβ type II RSK were also transfected. Lysates were immuniprecipitated with anti-Flag antibody and in vitro kinase assays performed. We observe that full length KIN-29 is capable of autophophorylation (Fig. 3). However, when we truncate the C-terminal domain, we find that the kinase domain, along with the ubiquitin-associated domain (UBA) or the kinase domain (with or without lysine 45 changed to arginine) are no longer capable of autophosphorylation. Lysine 45 is a conserved residue essential for catalytic function in kinases. This indicates that the C-terminal domain is required for autophosphorylation. The C-terminal domain could be required for kinase activity or it may simply be the substrate for autophosphorylation. Lanjuin and colleagues have previously shown that animals possessing a mutation (oy39) in the kinase domain have a small body size [19], indicating that the kinase domain is required for proper body size.
Figure 3 KIN-29 is a functional kinase. 293T cells were transfected with C-terminal flag tagged kin-29 constructs (top panel). KIN-29-KU contains amino acids 1–354, which includes the kinase domain and UBA domain. KIN-29K contains amino acids 1–300 which includes only the kinase domain and KIN-29K(K45R) contains amino acids 1–300 with a point mutation at position 45 that changes a lysine to an arginine. FLAG-tagged proteins were immunoprecipitated using anti-Flag antibody and in vitro kinase assays performed. Full length KIN-29 is capable of autophosphorylation similar to mammalian Tβ RII (TGFβ type II RSK) (top panel). Truncating the C-terminal domain of KIN-29 prevents autophosphorylation.
Placement of kin-29 in the Sma/Mab pathway
Epistasis between kin-29 and dbl-1 or lon-1 was examined in order to determine the relationship between kin-29 and Sma/Mab pathway signaling. Double mutant analysis between several of the known pathway components and lon-1 results in animals that are long (Lon) [16,17], making it the most downstream gene in the pathway. We examined lon-1(wk50); kin-29(wk61) double mutants to determine whether kin-29 can be placed in a similar position in the pathway as the current Sma/Mab components. We find that double mutants are longer than the single mutants of kin-29(wk61), suggesting that lon-1 suppresses the Sma phenotype of kin-29 (Table 1).
Next, we examined the relationship between dbl-1 and kin-29. Over-expression of dbl-1 results in Lon animals, suggesting more ligand causes an increase in the TGFβ signal output. When dbl-1 over-expressing animals are crossed into sma-2, sma-3, sma-4, sma-6 or daf-4 mutant backgrounds, the Lon phenotype is suppressed [8] and the animals are Sma. This places the type I receptor and the Smads downstream of the ligand, dbl-1. When dbl-1 is over-expressed in a kin-29(wk61) background, the animals are also Sma (Table 2). Additionally, using a weak allele of sma-6, we generated a sma-6(e1482)unc4(e120); kin-29(wk61) double mutant and examined its body size. We find that these animals are similar in size to that observed for sma-6(e1482)unc4(e120) mutants alone suggesting that sma-6 and kin-29 may not function in parallel pathways (Table 2). This indicates that kin-29 behaves in a manner consistent with known Sma/Mab pathway signaling molecules and is likely to function within this signaling cascade.
Table 2 kin-29 suppresses the dbl-1 over-expression phenotype
Genotype* Perimeter (mm)** %N2 n
kin-29(wk61)1, 2) 1.90 ± 0.15 73% 42
ctIs40 [pTG96(sur-5::gfp)]dbl-1(+)2 2.76 ± 0.12 106% 42
kin-29(wk61); ctIs40 [pTG96(sur-5::gfp)]1 1.84 ± 0.20 71% 51
sma-6(e1482) unc-4(e120)3 1.47 ± 0.13 57% 37
sma-6(e1482) unc-4(e120);kin-29(wk61)3 1.48 ± 0.17 57% 38
* All animals were measured 48 hours after L4.
** Data are means ± std.
1-Animals are not significantly different from each other (p > 0.05).
2-Animals are significantly different from each other (p < 0.0001)).
3-Animals are not significantly different from each other (p > 0.05).
Since lon-1 is genetically downstream of the Sma/Mab pathway signaling, we examined whether lon-1 mRNA levels are altered in kin-29 mutants. We have previously shown that lon-1 mRNA levels are up-regulated in sma-6(wk7) mutants and down-regulated in animals that over-express dbl-1 [16,17]. To test whether lon-1 mRNA is regulated in kin-29 animals, we examined lon-1 mRNA levels in a kin-29(wk61) background (Fig. 4). kin-29(wk61) animals show an increase in the expression level of the lon-1 transcript. This increase is comparable to that previously observed in sma-6(wk7) mutant animals [16].
Figure 4 mRNA levels of the Sma/Mab target gene, lon-1, are negatively regulated in kin-29(wk61) animals. Northern blot showing lon-1 mRNA expression observed in mixed stage populations of N2, sma-6(wk7), sma-3(wk30), and kin-29(wk61) animals. Sma/Mab pathway mutants sma-6(wk7) and sma-3(wk30) show an up-regulation of the lon-1 transcript (lanes 2 and 3). Similarly, kin-29(wk61) also shows an increase in lon-1 mRNA (lane 4). Elongation factor-2 (eft-2) was used to control for amounts of RNA loaded per lane. Levels of mRNA were quantitated using a phosphorimager and IQMacv1.2 software. See Materials and Methods for details on relative transcript level calculations.
kin-29 has been shown to affect the expression of a subset of olfactory receptor genes [19]. Several olfactory receptors expressed in AWB, ASH and ASK sensory neurons were either reduced or up-regulated in the kin-29 mutant background. Given that kin-29 affects mRNA levels of these olfactory receptors, we asked whether kin-29 alters the mRNA expression levels of the Sma/Mab components. We examined sma-6 mRNA expression (the type I receptor) in a kin-29 background and find no changes in expression levels of sma-6 mRNA. Because kin-29 and dbl-1 expression patterns overlap, we examined mRNA expression levels of dbl-1 in a kin-29 mutant background. We find that kin-29 does not affect dbl-1 mRNA expression levels (data not shown).
kin-29 expression is diverse and dynamic
In efforts to elucidate the function of kin-29 in TGFβ signaling, we examined its expression pattern. A construct consisting of the kin-29 promoter and coding region fused in frame to gfp was injected into kin-29(wk61) animals. As described above, this construct was able to rescue the small body size phenotype of kin-29(wk61) to wild-type (Table 3). Upon examination of the expression pattern, we observe KIN-29 to be localized to various tissue types (Fig. 5). Most notably, KIN-29 is seen in several neuronal cells in the head and tail throughout the course of development. Several of the sensory neurons found in the head express KIN-29, including ASH, AFD and ASI [19]. Additional neuronal staining is observed in both CAN cells and the ventral nerve cord (Fig. 5A). We find expression both in pharyngeal and body wall muscle (Fig. 5B). During the L1, L3 and L4 stages, we see expression throughout the intestine both in the nuclei and to a lesser extent in the cytoplasm (Fig. 5C) and in cells in the tail (Fig. 5D). This intestinal expression is rarely seen in later stages of development. Occasionally, expression is seen in vulval muscles as well.
Figure 5 Expression patterns of kin-29p:: kin-29:gfp rescuing construct in wild-type animals. Animals shown are L4 stage photographed at 63x. kin-29 promoter fusion constructs are expressed throughout development: (A) in the CAN neuron, (B) in body wall muscle, (C) in the intestine, and (D) in cells in the tail.
Hypodermal and neuronal expression of kin-29 rescues the small body size phenotype
In order to determine where kin-29 activity is required, we examined the body size of kin-29(lf) animals transformed with constructs expressing kin-29 in specific tissues. elt-3 and rol-6 promoters drive expression in the hypodermis, while dbl-1 drives expression in a subset of neurons. All three promoters were fused to kin-29 genomic DNA sequences. Each of these constructs was injected into kin-29(wk61) animals and transgenic strains were analyzed for body size. Several of the Sma/Mab pathway components, sma-3, sma-6, and lon-1, when specifically expressed in hypodermal tissues, rescue the body size defects associated with loss-of-function mutations in each of these genes [16-18,26]. Using the rol-6 and elt-3 promoters to drive hypodermal expression of the genomic region of kin-29 results in rescue of the small body size phenotype of kin-29(wk61). Since KIN-29 expression overlaps that of the Sma/Mab pathway ligand DBL-1 in the amphid neurons, ventral nerve cord, CAN cells and body wall muscle, we reasoned that KIN-29 and DBL-1 may function together in the same tissues to regulate body size morphogenesis [8,9]. We find that kin-29 under the control of the dbl-1 promoter rescues the small body size phenotype of kin-29(wk61) animals (Table 3). These data suggest that KIN-29 functions in neuronal and hypodermal tissues to regulate body size morphogenesis (Table 3 and data not shown).
kin-29 mutants are small, have delayed growth rates, and reduced brood sizes
The growth properties of Sma/Mab animals differ from other small animals. For example, mutants of the spectrin gene, sma-1, which have been shown to affect embryonic elongation but not thought to be involved in TGFβ signaling, are approximately 50% the size of wild-type animals at hatching [27]. This is in contrast to the body size of L1 animals mutant in known Sma/Mab pathway components. For example, sma-3, sma-6, dbl-1, and lon-1 are indistinguishable in length from wild-type L1 animals at hatching [26]. This suggests that the Sma/Mab pathway components are defective in post-embryonic rather than embryonic stages of development. Sma/Mab pathway mutant animals grow at a slower rate as development progresses through the later larval stages into adulthood [26]. There is no defining switch during development that regulates body growth. We tested whether kin-29 mutations cause body size defects in a similar manner to Sma/Mab pathway mutations or whether kin-29 possessed embryonic defects. We examined the body size of kin-29 mutant animals in comparison to sma-6(wk7) and N2 animals at hatching and then at 24 hour intervals to 96 hours. We find that all three alleles of kin-29 are similar in length at the L1 stage to N2 animals. This is also what we observe for sma-6(wk7) which suggests that kin-29 delays growth post-embryonically, as do Sma/Mab pathway components (Fig. 6). The Sma body size of kin-29 is therefore due to a delay in development in later larval stages.
Figure 6 The small body size phenotype of kin-29animals is a result of defects in postembryonic development. N2, sma-6(wk7), kin-29(wk61), kin-29(oy38) and kin-29(oy39) were hatched and synchronized as L1 animals. L1 animals were measured at time zero and then at 24-hour time points spanning a 96 hour period. kin-29 animals are developmentally delayed and over time, kin-29(lf) animals never reach a wild-type body size. Perimeter measurements for at least 22 animals were averaged at each time point. Error bars represent standard deviation values. Values for N2 and kin-29 mutants are significantly different (p < 0.001).
In addition, we find that kin-29 grows more slowly than N2 and Sma/Mab pathway mutants do. Animals hatched and grown at 20°C were scored based on their developmental stage after 72 hours. We find that 99% of wild-type animals are adults at this time point, while only 2% of kin-29(wk61) animals are adults (Table 4). Lanjuin and colleagues report a similar observation; 98% of wild-type animals hatched and grown at 25°C for 3 days were adults in comparison to approximately 24% of kin-29(oy38) animals [19]. We asked if lon-1(lf) could suppress the developmental delay characteristic of kin-29(wk61) animals (Table 4). lon-1(wk50) mutants on their own show a slight delay in development, but which is distinguishable from the Sma/Mab mutants. In the double mutant lon-1(wk50);kin-29(wk61), we find that the developmental defect of kin-29(wk61) can be partially suppressed by lon-1(wk50). This result is consistent with our conclusion that lon-1 is genetically downstream of kin-29.
Table 4 lon-1 partially suppresses the developmental defect of kin-29(wk61)
% Adult animals % Adults 4 animals
Genotype 20°C 20°C
N2 99 (185) 99 (185)
lon-1(wk50) 64 (245) 80 (245)
kin-29(wk61) 2 (475) 43 (475)
lon-1(wk50);kin-29(wk61) 40 (202) 63 (202)
Number of animals scored is shown in parentheses.
We observed that Sma/Mab pathway mutants have a reduced brood size. In addition to the developmental defects, kin-29(wk61) also has a reduced brood size (Table 5). Like sma-6(lf) and lon-1(lf), kin-29(wk61) shows a brood size approximately 30% the size of that seen in wild-type animals. We find that sma-6(wk7) and lon-1(wk50) along with kin-29(oy38) and kin-29(oy39) have a reduction in brood size as well. Although brood size is affected, embryonic survival rate appears to be normal.
Table 5 Brood size analysis of kin-29 alleles
Genotype % of wild-type brood size
N2 100 (270)
sma-6(wk7) 64 (172)
lon-1(wk50) 81 (219)
kin-29(wk61) 32 (86)
kin-29(oy38) 81 (218)
kin-29(oy39) 80 (217)
Number of eggs scored for each genotype is shown in parentheses.
kin-29 affects dauer pathway signaling
Several components of the Sma/Mab pathway have been shown to genetically interact with members of the dauer pathway [9,14]. The dauer-constitutive (Daf-c) phenotype of the type I receptor daf-1 is enhanced by mutations in sma-6. At 15°C, daf-1 mutant strains exhibit a very weak dauer-constitutive phenotype. However, sma-6(wk7); daf-1(m40) mutants show a 50% increase in the number of dauered animals at 15°C [14]. In addition, double mutants between the ligand daf-7(e1372) and either dbl-1(kk3) or sma-2(e502) also have been shown to enhance the weak Daf-c phenotype of daf-7(e1372) at 20°C [9]. These data suggest that there is some crosstalk between the Sma/Mab pathway and the TGFβ-like daf-7 dauer pathway.
Based on these findings, we examined the effects of the kin-29 alleles on dauer formation. Double homozygotes were made between daf-7(e1372) and each of the three alleles of kin-29. Genetic interactions were analyzed at 15°C, 20°C and 25°C. For comparison, daf-7(e1372) mutants raised at 25°C show almost 100% dauered animals compared to no dauered animals at 15°C or 20°C. At 15°C and 20°C, kin-29(oy38) is able to enhance dauer formation of daf-7(e1372) similar to the enhancement observed in sma-6(wk7); daf-7(e1372) mutant animals (Fig. 7A,B). We also see that kin-29(wk61) shows a weak enhancement of dauer formation while the missense mutant kin-29(oy39) shows no genetic interaction at all. These results are consistent with genetic interactions previously observed between Sma/Mab and dauer pathway components. However, at 25°C, we find that kin-29 can also suppress the constitutive dauer formation of daf-7(e1372). kin-29(wk61) and kin-29(oy39) are able to suppress the Daf-c defects of daf-7(e1372) while kin-29(oy38) does not (Fig. 7C).
Figure 7 Interaction between kin-29 and the daf-7 TGFβ-like pathway. N2 or dauered animals are not seen at 15°C or 20°C (A, B). daf-7(e1372) mutant animals form constitutive dauers at 25°C (C). kin-29 mutants can both suppress and enhance the dauer constitutive phenotype of daf-7(e1372). At 25°C, daf-7(e1372); kin-29(wk61) and daf-7(e1372); kin-29(oy39) mutants show suppression of the Daf-c phenotype while daf-7(e1372); kin-29(oy38) mutants do not (C). At 15°C and 20°C, daf-7(e1372); kin-29(oy38) mutants show an enhancement in dauer formation similar to that observed in sma-6(wk7); daf-7(e1372) animals(A, B). Additionally, daf-7(e1372); kin-29(wk61) mutants show a mild enhancement of the Daf-c daf-7 phenotype (A, B).
Discussion
KIN-29 functions in hypodermal and neuronal tissues to regulate body size
Expression of the Sma/Mab pathway components in the hypodermis is sufficient to rescue the body size defects seen in mutants. Specific expression of sma-3, sma-6 and lon-1 in the hypodermal tissues has been shown to restore body length in these respective mutant animals [16-18,26]. This implies that body size is regulated largely via hypodermal function. Our work presented here further supports that C. elegans body size is regulated in hypodermal tissues. When the genomic region of kin-29 is specifically expressed in the hypodermis, under the control of the elt-3 and rol-6 promoters, we see that the small body size phenotype of kin-29(wk61) is partially rescued. Although KIN-29 functions in the hypodermis to regulate body size morphogenesis, we do not see KIN-29::GFP, under the control of its own promoter, expressed in the hypodermal tissues, suggesting that KIN-29 expression levels are relatively low in these tissues.
DBL-1 is expressed primarily in neurons, which includes several amphid and pharyngeal neurons, ventral nerve cord, and CAN cells [8,17]. Since KIN-29 expression closely parallels that seen for DBL-1, we also examined whether the dbl-1 promoter driving kin-29 genomic sequences is capable of rescuing the body size defect of kin-29(wk61). When kin-29 is expressed in the same tissues as dbl-1, we observe partial rescue of the small body size defect seen in kin-29(lf) animals. It has been demonstrated that kin-29 under the control of the unc-14 and odr-4 promoters is able to rescue the body size defect of kin-29 mutant animals [19]. unc-14 is expressed in all neuronal cells while odr-4 is expressed in a subset of the sensory neurons including the AFD neurons where DBL-1 and KIN-29 are also expressed [28,29]. This suggests that neuronal expression of KIN-29 is also sufficient to regulate body size morphogenesis. Determining how this occurs will require further study.
kin-29 is a tissue specific factor that affects the Sma/Mab pathway
Mutations in the ligand dbl-1, the receptors sma-6 and daf-4, and Smads sma-2, sma-3 and sma-4, result in animals that are approximately 70% the size of wild-type animals [8,9,11,12,14]. Additional defects are seen as male tail ray fusions and crumpled spicules. However, the negatively regulated Sma/Mab target gene lon-1, suppresses the small body size phenotype of sma-2, sma-3, sma-4 and sma-6 but not the male tail defects observed in each of these loss-of-function mutants [16,17]. This implies that Sma/Mab pathway signaling may branch downstream of the Smads to regulate a subset of genes that control body size morphogenesis while others specifically affect male tail development. We find that kin-29(lf) animals do not posses ray fusions or crumpled spicules and may exert its effects upstream of this branch point in the signaling pathway. In addition, we see that lon-1 is genetically downstream of kin-29 and that kin-29 suppresses the Lon phenotype associated with over expression of dbl-1. Taken together, this data suggests that kin-29 may function in tissues with Sma/Mab pathway components to regulate body size but not male tail formation.
The EMK kinase family and kin-29
Members of the EMK family have been shown to affect cell polarity and microtubule stability. Mammalian MARK phosphorylates microtubule associated proteins and has been shown to destabilize microtubules when over expressed in CHO cells [23]. Drosophila PAR-1 influences the cytoskeletal organization of the oocyte [21]. In wild-type Drosophila oocytes, the microtubules are arranged in an anterior to posterior gradient with no microtubules observed at the most posterior region of the oocyte. Microtubules in Drosophila par-1 mutants, however, are organized around the cortex of the oocyte. In this reorganization, microtubules are now observed in the most posterior region of the oocyte. In addition, posterior localization of Drosophila oskar is dependent on microtubule polarity. oskar is mislocalized in dpar-1 mutants, further supporting the involvement of dpar-1 in regulating microtubule dynamics. C. elegans PAR-1 regulates the early asymmetrical cell divisions of the embryo but has not been shown to have any affects on the microtubule network [20]. KIN-29 only shows homology to the EMK family members within its N-terminal kinase domain, indicating that KIN-29 is a more distantly related member of the EMK family. However, lack of homology between the C-termini of KIN-29 and EMK family proteins suggests that KIN-29 activity may diverge from that observed for members of this family.
How does kin-29 function?
We have observed that KIN-29 functions in both neuronal and hypodermal tissues. How kin-29 functions in each of these tissues to regulate body size morphogenesis is unclear. Since kin-29 encodes a kinase it might act to regulate the activities of a variety of molecules that affect Sma/Mab pathway signaling. Recently, it has been shown that several olfactory receptors are misexpressed in kin-29(lf) animals, suggesting that KIN-29 may regulate proper expression levels of various genes [19]. One model is that KIN-29 phosphorylates a transcription factor and/or co-factor, which leads to the transcriptional mis-regulation of some component important for Sma/Mab pathway signaling. We have examined the expression levels of the Sma/Mab ligand dbl-1 and the type I RSK sma-6 and do not see any alteration in their levels of expression. However, this does not rule out that other genes that impinge on pathway signaling might be affected at the transcriptional level in neurons and hypodermal cells. Alternatively, KIN-29 may function in microtubule dynamics as described above [21,23]. kin-29 might therefore influence microtubule (MT) organization in both neuronal and hypodermal tissues and affect Sma/Mab pathway signaling in each of these cell types.
Dauer interactions show that kin-29(lf) mutants may not sense external cues properly
We have shown that KIN-29 helps to promote dauer formation at 25°C and to suppress dauer formation at 15°C. ttx-3, a LIM homeobox gene, shows a similar genetic interaction with daf-7 [30]. Like kin-29, ttx-3 single mutants do not affect dauer formation. daf-7(e1372); ttx-3(ks5) double mutant animals show an enhanced Daf-c phenotype of daf-7 at 15°C while suppressing it at 25°C. In wild-type animals, high temperatures contribute to dauer formation, while lower temperatures suppress dauer formation. ttx-3 decouples both hot and cold inputs from the dauer pathway, and kin-29 may act similarly [30]. tph-1, a tryptophan hydroxylase involved in the synthesis of serotonin, has been shown to form 10–15 % dauers in the presence of food and this defect is not dependent on temperature which implicates serotonergic signaling in modulating temperature sensitive dauer arrest [31]. tph-1 is also able to enhance the constitutive dauer phenotype of daf-7 mutants at 15°C similar to the enhancement observed for kin-29 [31]. Sze and colleagues did not examine the affects of tph-1 mutants on daf-7 at 25°C. They did show, however, that over-expression of tph-1 in a daf-7 background at 25°C suppresses the Daf-c phenotype of daf-7. This is opposite to what we observe for kin-29. Although there are some similarities between tph-1 and kin-29, tph-1(mg280); kin-29(oy38) double mutants are synthetic-daf at 20°C and 25°C suggesting that kin-29 may not function in a linear pathway with tph-1 but rather parallel to tph-1 [19]. tph-1 has also been shown to regulate the expression of daf-7 while kin-29 does not, suggesting that kin-29 functions downstream or in parallel to daf-7 production to affect TGFβ signaling outputs [19,31]. kin-29 may partly influence dauer formation through a serotonin mediated pathway and a non-serotonin mediated pathway such as the Sma/Mab pathway. The Sma/Mab pathway has been shown to influence dauer formation in combination with the TGFβ-like daf-7 pathway [9,14].
Starved animals are smaller than animals grown with abundant food supplies, indicating that environmental conditions influence body size morphogenesis [13]. Drosophila S6 kinase mutants are smaller in body size due to decreased cell size, which is similar to the body size defect observed in nutrient deprived flies [32,33]. It is thought that S6 kinase alters cell growth in response to nutrients and growth factors by regulating the efficiency of the translational apparatus [34]. Recently, it has been shown that a deletion found within the C. elegans homolog of S6 kinase (sv31) also results in a reduced body size in the adult stage (J. Friberg and S, Tuck, personal communication). S6 kinase (sv31) and kin-29(lf) animals also show other similar phenotypes, including reduced brood size and slow growth defects. In addition, fat accumulation is also observed in S6 kinase (sv31) animals similar to that observed in dauered animals. Previously, it has been demonstrated that animals that show pheromone hypersensitivity are unable to sense food or temperature signals properly [35-39]. Recently, kin-29 mutants have been shown to be hypersensitive to pheromone [19]. In addition, kin-29 mutants also possess hyperforaging activity in the presence of abundant food supplies [19]. Hyperforaging is normally observed in animals that have been deprived of food. These defects suggest that kin-29(lf) mutant animals may not sense food or temperature signals properly and this may influence body size regulation. Taken together, this evidence supports an environmental role in regulation of body size. kin-29 may function to transmit these environmental cues to the Sma/Mab TGFβ signaling pathway, thereby affecting proper body size morphogenesis.
Conclusion
In this study, kin-29 was identified in a genetic screen designed to identify modifiers of body size in C. elegans. Mutants in kin-29 result in small animals, and we show that kin-29 affects the dbl-1 signaling pathway in C. elegans. kin-29 also modifies phenotypes from a second TGFβ pathway in C. elegans, the dauer pathway. Further, we show that KIN-29 does contain kinase activity, and that it is capable of phosphorylating itself. KIN-29 functions in neurons and in the hypodermis to control aspects of body size.
Methods
Strains
C. elegans strains were grown using standard methods [40]. kin-29(wk61) was used for body size rescue experiments. N2, sma-6(wk7), kin-29(wk61), kin-29(oy38) and kin-29(oy39) were used in generating growth curves [19]. Interactions between kin-29 and the dauer pathway were examined using daf-7(e1372). Total RNA used for northern blot analysis was isolated from N2, sma-6(wk7), sma-3(wk30), and kin-29(wk61) animals. Epistasis was determined using kin-29(wk61), lon-1(wk50), dbl-1 over expressing strain ctIs40 [pTG96 (sur-5::gfp)], and lon-1(wk50); kin-29(wk61) mutant animals.
Isolation of kin-29(wk61)
kin-29(wk61) was generated from an EMS F2 screen designed to isolate small body size mutants [15]. kin-29(oy38) and kin-29(oy39) were obtained from P. Sengupta [19].
Cloning of kin-29
kin-29(wk61) was mapped to a small region on the X chromosome between unc-2 and fax-1 using genetic markers and deficiencies. The YAC clone Y76F7 from this interval was purified from total yeast DNA using pulse field gel electrophoresis. Injection of YAC Y76F7 into kin-29(wk61) rescued the small body size phenotype. Next DNA from cosmids contained within the region of Y76F7 was isolated and transgenic lines were generated. Cosmid F58H12, with one predicted open reading frame (ORF), conferred rescue. ESTs spanning this region were obtained from the C. elegans cDNA project (Y. Kohara, National Institute of Genetics). The longest EST, y293c7, was sequenced and shows minor differences from the Genefinder prediction. A 10 kb genomic region fused in frame to GFP, which contained the corresponding sequence from y293c7, was generated by PCR as described below. This kin-29p::kin-29:gfp fusion construct rescued the small body size of kin-29(wk61). Genomic DNA from homozygous kin-29(wk61) animals was sequenced. Two independent PCR amplifications were generated for each of the three regions spanning the kin-29 coding region using the following primer sets: CGCTGCGGCCGCTTCAGGCGCCGCCACACCAA/ CGCCGCTGCAGCCGCCGGCAACGAGAATGTA; CGCTGCGGCCGCCCAAGCCAACGTTGCAGGTA/ CGCCGCTGCAGGATAACATGCTCCACTGGCTA; CGCTGCGGCCGCCACCGCACGGGCTAGATATT/ CGCCGCTGCAGCCATTCACTCCGAGCTCCAG. Each PCR product was digested with Not I and Pst I, subcloned into pBluescript SK+, and sequenced.
GFP fusion and tissue-specific expression constructs
kin-29p::kin-29 gfp contains the 10 kb genomic region of kin-29 fused in frame to gfp. This construct was generated using the primers CGCGCTGCAGCAGACCATGGACGT GTTTTAATG and CCGGGGATCCTCCGAGCTCCAGCTTGGATCA, digesting with Pst I and BamH I, and inserting the PCR product into the promoterless vector pPD95.75. kin-29p::gfp was generated by cloning 1.4 kb upstream of the predicted kin-29 ATG into the Hind III and Xba I sites of the GFP insertion vector pPD95.69 (A. Fire, Stanford University). The 1.4 kb piece was generated by PCR using primers CCGGAAGCTTCAGACCATGGACGTGTTTTAATG and CGCGTCTAGATGCAGTGTTGGTGTGGCGGC. Fluorescent GFP expression patterns were examined in larval and adult animals using a Zeiss compound microscope.
kin-29 genomic DNA was ectopically expressed in specific tissues using the promoters rol-6, elt-3 and dbl-1 [8,41,42]. The rol-6 and elt-3 promoters express in the hypodermis, while the dbl-1 promoter expresses primarily in neuronal tissues. PCR fragments containing the elt-3, rol-6 and dbl-1 promoters were generated using the primer sets CCGGAAGCTTGTGACACGTTGTTTCACGGTCAT/ CCGGCTGCAGGAAGTTTGAAATACCAGGTAGCCGA, CCGGCTGCAGCTTCGTATTAGATCTCAGCAGC/ CGCGCGTCGACAGTTAGATCTAAAGATATATCCAG, and CCGGCTGCAGCCCGGAAATCACGACCAAATGGGTC/ CGCGCGTCGACAGTTGAGTTGGGCGCATCAGGCAG respectively. elt-3 PCR products were digested with Hind III and Pst I while rol-6 and dbl-1 products were digested with Pst I and Sal I. 7.7 kb PCR fragments comprising kin-29 genomic DNA were generated using the primer sets CCCGGGTCGACATGGCTGCGCCACGGCGGCGTAT/CCGGGGATCCTC CGAGCTCCAGCTTGGATCA and CGCGCTGCAGCAGACCATGGACGTGTT TTAATG/ CCGGGGATCCTCCGAGCTCCAGCTTGGATCA and digested with either Sal I and Bam HI or Pst I and Bam HI respectively. Fragments were inserted in frame into the promoterless gfp insertion vector pPD95.75. All constructs were injected into kin-29(wk61) and transformants were analyzed for body size rescue.
Analysis of body size, brood size and growth rates
For body size measurements, animals were photographed 48 hours after the L4 stage using a Nikon SMZ-U dissecting microscope set at 3.5× magnification and software from Strata Video Shop (Strata Inc.). Screen dimensions were 680 X 460 pixels. Perimeter analysis was done using Image Pro Plus (Mediacybernetics).
For brood size analyses, single L4 animals were picked to individual plates. Every 12 hours, animals were transferred to new plates to continue egg laying. All eggs were counted. 24 hours later, hatchlings were scored.
For growth rate analyses, animals were synchronized. Gravid animals were treated with a hypochlorite/NaOH solution in order to isolate eggs. The eggs were allowed to hatch in M9 for at least 24 hours. Approximately 30 L1 animals were placed onto plates seeded with OP50. Animals were initially measured at the L1 stage (time zero) and then at 24 hour intervals thereafter. The final time point was taken 96 hours after the L1 stage. Images were obtained and perimeter analyses were performed as described above.
Genetic interactions with daf-7(e1372)
Double mutants were generated between daf-7(e1372) and the following mutants: sma-6(wk7), kin-29(wk61), kin-29(oy38), and kin-29(oy39). Gravid animals were placed onto plates well seeded with OP50 and allowed to lay eggs at room temperature. Animals were removed from plates after approximately 30 – 50 eggs were laid. Eggs were allowed to hatch at 15°C, 20°C and 25°C. The number of dauered animals was counted and graphed.
Northern blot analysis
Total RNA from L4 animals was isolated from N2, sma-6(wk7), sma-3(wk30) and kin-29(wk61) as described previously (previously described in [16]. Equal amounts (20–30 μg) of total RNA were loaded per lane onto a 1.2% agarose/6.6% formaldehyde gel and resolved by electrophoresis. Samples were transferred to nitrocellulose (Osmonics Inc.) and baked at 80°C for 2 hours. The lon-1 and dbl-1 probes were generated by digesting both the lon-1 cDNA B1.11 and the dbl-1 cDNA with Eco RI. The sma-6 probe was generated by PCR using the primer set: GCCGCCTCGAGATGAACATCACCTTTATATTTATTCTC/ GCCGCGGATCCTTAAGATTGATTGGTGGCTGAC. Elongation factor-2 and α-tubulin were used as controls to indicate the amount of total RNA loaded per lane. Before probes were added, the nitrocellulose blots were prehybridized with 1 mM EDTA, 0.5 M NaPO4, pH7.2, 7% SDS, and 1% BSA fraction V (Sigma) for at least 30 minutes. Probes were labeled using the Prime It II kit (Stratagene), added to the prehybridization solution, and incubated overnight at 65°C. Blots were washed (1 mM Na2EDTA, 40 mM NaPO4, pH 7.2, and 1% SDS) at least three times at 65°C for 15 minutes. Each blot was placed onto a phosphorimager screen for at least 48 hours and analyzed using a Molecular Dynamics Phosphorimager (Amersham Pharmacia Biotech) and IQMacv1.2 software. For each band, intensity levels were corrected for background and normalized according to the loading control (eft-2 mRNA). Relative transcript levels of the mutants were normalized to the intensity ratio of lon-1/eft-2 of N2.
Kinase assays
C-terminally tagged Flag kin-29 constructs were generated in the mammalian vector pRK5. KIN-29-KU contains amino acids 1–354 which includes the kinase domain and UBA domain. KIN-29K contains amino acids 1–300 which includes only the kinase domain and KIN-29K(K45R) contains amino acids 1–300 with a point mutation at position 45 that changes a lysine to an arginine. Cell transfection, immunoprecipitation and kinase assays were carried out as previously described [41]. Human 293T cells at 30% confluency were transfected with each construct (2 μg in 100 mm plates) using LipofectAMINE (Life Technologies, Inc.). Forty eight h after transfection, cells were lysed in the lysis buffer (25 mM Tris-HCl, 300 mM NaCl, and 1% Triton X-100). Lysates were immunoprecipitated using anti-Flag antibody M2 (Sigma), washed 3 times in the same buffer and a final wash in the kinase buffer (10 mM HEPES-KOH, pH 7.5, 5 mM MgCl2, and 5 mM CaCl2). For in vitro kinase assays, the immunoprecipitated protein samples were divided into two aliquots. One aliquot was analyzed by anti-Flag western blotting. The second aliquot was subjected to a kinase autophosphorylation assay at room temperature for 30 min in the kinase buffer containing 5 μCi of 32P-ATP (5000 μCi/mmol). The reaction was terminated by adding an equal volume of 2 × SDS sample buffer (80 mM Tris, pH 6.8, 3.2% SDS, 16% glycerol, 200 mM dithiothreitol, 0.02% bromphenol blue), then subjected to SDS-PAGE and visualized by autoradiography.
Authors' contributions
L.L.M. performed the majority of experiments, including the mapping and molecular cloning of kin-29. S.C initiated the early phases of mapping kin-29. A.F.R. analyzed brood size. C.M.Z. and A.F.R. assisted with generating transgenic nematode lines. H.W. and L.C. provided assistance in construct design and scoring of genetic experiments. X.L. and X-H.F. performed the kinase assays. R. W. P. implemented and supervised the project and R.W.P. and L.L.M. prepared the manuscript.
Acknowledgements
We thank A. Lanjiun and P. Sengupta for exchanging information and strains prior to publication and for useful discussions, and Y. Kohara for cDNA clones. We are grateful to T. Gumienny and C. Savage-Dunn and for useful comments on the manuscript, and lab members for useful scientific discussions. We also thank J. Friberg and S. Tuck for communicating unpublished work regarding S6 kinase. This work was supported by a grant from the NIH to R.W.P. L.L.M. was supported by the American Association of University Women Dissertation Fellowship and by the New Jersey Commission on Cancer Research. C.M.Z. was supported by the New Jersey Commission on Cancer Research and A.F.R. was supported by a Charles and Joanna Busch Predoctoral Fellowship and Biotechnology Training Grant from the NIH. Some mutant nematode strains were obtained from the Caenorhabditis Genetics Center, which is supported by the NIH National Center for Research Resources (NCRR).
==== Refs
Massagué J Blain SW Lo RS TGFβ signaling in growth control, cancer, and heritable disorders Cell 2000 103 295 309 11057902 10.1016/S0092-8674(00)00121-5
Suga H Ono K Miyata T Multiple TGF-β receptor related genes in sponge and ancient gene duplications before the parazoan-eumetazoan split FEBS Lett 1999 453 346 350 10405173 10.1016/S0014-5793(99)00749-8
Patterson GI Padgett RW TGF β-related pathways. Roles in Caenorhabditis elegans development Trends Genet 2000 16 27 33 10637628 10.1016/S0168-9525(99)01916-2
Finnerty JR Pang K Burton P Paulson D Martindale MQ Origins of bilateral symmetry: Hox and Dpp expression in a sea anemone Science 2004 304 1335 1337 15131263 10.1126/science.1091946
Whitman M Smads and early developmental signaling by the TGFβ superfamily Genes Dev 1998 12 2445 2462 9716398
Massagué J TGFβ Signal Transduction Annu Rev Biochem 1998 67 753 791 9759503 10.1146/annurev.biochem.67.1.753
Heldin CH Miyazono K ten Dijke P TGF-β signalling from cell membrane to nucleus through SMAD proteins Nature 1997 390 465 471 9393997 10.1038/37284
Suzuki Y Yandell MD Roy PJ Krishna S Savage-Dunn C Ross RM Padgett RW Wood WB A BMP homolog acts as a dose-dependent regulator of body size and male tail patterning in Caenorhabditis elegans Development 1999 126 241 250 9847238
Morita K Chow KL Ueno N Regulation of body length and male tail ray pattern formation of Caenorhabditis elegans by a member of the TGF-β family Development 1999 126 1337 1347 10021351
Ren P Lim CS Johnsen R Albert PS Pilgrim D Riddle DL Control of C. elegans larval development by neuronal expression of a TGF-β homolog Science 1996 274 1389 1391 8910282 10.1126/science.274.5291.1389
Estevez M Attisano L Wrana JL Albert PS Massagué J Riddle DL The daf-4 gene encodes a bone morphogenetic protein receptor controlling C. elegans dauer larva development Nature 1993 365 644 649 8413626 10.1038/365644a0
Savage C Das P Finelli AL Townsend SR Sun CY Baird SE Padgett RW Caenorhabditis elegans genes sma-2, sma-3, and sma-4 define a conserved family of transforming growth factor β pathway components Proc Natl Acad Sci U S A 1996 93 790 794 8570636 10.1073/pnas.93.2.790
Riddle DL Albert PS Riddle DL, Blumenthal T, Meyer BJ, Priess JR Genetic and environmental regulation of dauer larva development C elegans II 1997 Cold Spring Harbor Press 739 768
Krishna S Maduzia LL Padgett RW Specificity of TGFβ signaling is conferred by distinct type I receptors and their associated SMAD proteins in Caenorhabditis elegans Development 1999 126 251 260 9847239
Savage-Dunn C Maduzia LL Zimmerman CM Roberts AF Cohen S Tokarz R Padgett RW Genetic screen for small body size mutants in C. elegans reveals many TGFβ pathway components Genesis 2003 35 239 247 12717735 10.1002/gene.10184
Maduzia LL Gumienny TL Zimmerman CM Wang H Shetgiri P Krishna S Roberts AF Padgett RW lon-1 regulates Caenorhabditis elegans body size downstream of the dbl-1 TGFβ signaling pathway Dev Biol 2002 246 418 428 12051826 10.1006/dbio.2002.0662
Morita K Flemming AJ Sugihara Y Mochii M Suzuki Y Yoshida S Wood WB Kohara Y Leroi AM Ueno N A Caenorhabditis elegans TGF-β, DBL-1 controls the expression of LON-1, a PR-related protein, that regulates polyploidization and body length EMBO J 2002 21 1063 1073 11867534 10.1093/emboj/21.5.1063
Yoshida S Morita K Mochii M Ueno N Hypodermal expression of Caenorhabditis elegans TGF-β type I receptor SMA-6 is essential for the growth and maintenance of body length Dev Biol 2001 240 32 45 11784045 10.1006/dbio.2001.0443
Lanjuin A Sengupta P Regulation of chemosensory receptor expression and sensory signaling by the KIN-29 Ser/Thr kinase Neuron 2002 33 369 381 11832225 10.1016/S0896-6273(02)00572-X
Guo S Kemphues KJ par-1, a gene required for establishing polarity in C. elegans embryos, encodes a putative Ser/Thr kinase that is asymmetrically distributed Cell 1995 81 611 620 7758115 10.1016/0092-8674(95)90082-9
Shulman JM Benton R St Johnston D The Drosophila homolog of C. elegans PAR-1 organizes the oocyte cytoskeleton and directs oskar mRNA localization to the posterior pole Cell 2000 101 377 388 10830165 10.1016/S0092-8674(00)80848-X
Levin DE Bishop JM A putative protein kinase gene (kin1+) is important for growth polarity in Schizosaccharomyces pombe Proc Natl Acad Sci U S A 1990 87 8272 8276 2236039
Drewes G Ebneth A Preuss U Mandelkow EM Mandelkow E MARK, a novel family of protein kinases that phosphorylate microtubule- associated proteins and trigger microtubule disruption Cell 1997 89 297 308 9108484 10.1016/S0092-8674(00)80208-1
Horike N Takemori H Katoh Y Doi J Min L Asano T Sun XJ Yamamoto H Kasayama S Muraoka M Nonaka Y Okamoto M Adipose-specific expression, phosphorylation of Ser794 in insulin receptor substrate-1, and activation in diabetic animals of salt-inducible kinase-2 J Biol Chem 2003 278 18440 18447 12624099 10.1074/jbc.M211770200
Wang Z Takemori H Halder SK Nonaka Y Okamoto M Cloning of a novel kinase (SIK) of the SNF1/AMPK family from high salt diet-treated rat adrenal FEBS Lett 1999 453 135 139 10403390 10.1016/S0014-5793(99)00708-5
Savage-Dunn C Tokarz R Wang H Cohen S Giannikas C Padgett RW SMA-3 smad has specific and critical functions in DBL-1/SMA-6 TGFβ- related signaling Dev Biol 2000 223 70 76 10864461 10.1006/dbio.2000.9713
McKeown C Praitis V Austin J sma-1 encodes a bH-spectrin homolog required for Caenorhabditis elegans morphogenesis Development 1998 125 2087 2098 9570773
Dwyer ND Troemel ER Sengupta P Bargmann CI Odorant receptor localization to olfactory cilia is mediated by ODR-4, a novel membrane-associated protein Cell 1998 93 455 466 9590179 10.1016/S0092-8674(00)81173-3
Ogura K Shirakawa M Barnes TM Hekimi S Ohshima Y The UNC-14 protein required for axonal elongation and guidance in Caenorhabditis elegans interacts with the serine/threonine kinase UNC- 51 Genes Dev 1997 11 1801 1811 9242488
Hobert O Mori I Yamashita Y Honda H Ohshima Y Liu Y Ruvkun G Regulation of interneuron function in the C. elegans thermoregulatory pathway by the ttx-3 LIM homeobox gene Neuron 1997 19 345 357 9292724 10.1016/S0896-6273(00)80944-7
Sze JY Victor M Loer C Shi Y Ruvkun G Food and metabolic signalling defects in a Caenorhabditis elegans serotonin-synthesis mutant Nature 2000 403 560 564 10676966 10.1038/35000609
Montagne J Stewart MJ Stocker H Hafen E Kozma SC Thomas G Drosophila S6 kinase: a regulator of cell size Science 1999 285 2126 2129 10497130 10.1126/science.285.5436.2126
Robertson FW The ecological genetics of growth in Drosophila IV. The genetic correlation between the duration of the larval period and body size in relation to the larval diet Genetical Research 1963 4 74 92
Potter CJ Xu T Mechanisms of size control Curr Opin Genet Dev 2001 11 279 286 11377964 10.1016/S0959-437X(00)00191-X
Ailion M Thomas JH Dauer formation induced by high temperatures in Caenorhabditis elegans Genetics 2000 156 1047 1067 11063684
Daniels SA Ailion M Thomas JH Sengupta P egl-4 acts through a transforming growth factor-β/SMAD pathway in Caenorhabditis elegans to regulate multiple neuronal circuits in response to sensory cues Genetics 2000 156 123 141 10978280
Golden JW Riddle DL A pheromone-induced developmental switch in Caenorhabditis elegans: temperature-sensitive mutants reveal a wild-type temperature-dependent process Proc Natl Acad Sci USA 1984 81 819 823 6583682
Golden JW Riddle DL The Caenorhabditis elegans dauer larva: developmental effects of pheromone, food, and temperature Dev Biol 1984 102 368 378 6706004 10.1016/0012-1606(84)90201-X
Thomas JH Birnby DA Vowels JJ Evidence for parallel processing of sensory information controlling dauer formation in Caenorhabditis elegans Genetics 1993 134 1105 1117 8375650
Brenner S The genetics of Caenorhabditis elegans Genetics 1974 77 71 94 4366476
Gilleard JS Shafi Y Barry JD McGhee JD ELT-3: A Caenorhabditis elegans GATA factor expressed in the embryonic epidermis during morphogenesis Dev Biol 1999 208 265 280 10191044 10.1006/dbio.1999.9202
Mello CC Kramer JM Stinchcomb DT Ambros V Efficient gene transfer in C. elegans: extrachromosomal maintenance and integration of transforming sequences EMBO J 1991 10 3959 3970 1935914
| 15840165 | PMC1112587 | CC BY | 2021-01-04 16:40:17 | no | BMC Dev Biol. 2005 Apr 19; 5:8 | utf-8 | BMC Dev Biol | 2,005 | 10.1186/1471-213X-5-8 | oa_comm |
==== Front
BMC Evol BiolBMC Evolutionary Biology1471-2148BioMed Central London 1471-2148-5-281583109510.1186/1471-2148-5-28Research ArticleSubfunctionalization of duplicated genes as a transition state to neofunctionalization Rastogi Shruti [email protected] David A [email protected] Computational Biology Unit, BCCS, University of Bergen, 5020 Bergen, Norway2005 14 4 2005 5 28 28 5 1 2005 14 4 2005 Copyright © 2005 Rastogi and Liberles; licensee BioMed Central Ltd.2005Rastogi and Liberles; 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
Gene duplication has been suggested to be an important process in the generation of evolutionary novelty. Neofunctionalization, as an adaptive process where one copy mutates into a function that was not present in the pre-duplication gene, is one mechanism that can lead to the retention of both copies. More recently, subfunctionalization, as a neutral process where the two copies partition the ancestral function, has been proposed as an alternative mechanism driving duplicate gene retention in organisms with small effective population sizes. The relative importance of these two processes is unclear.
Results
A set of lattice model genes that fold and bind to two peptide ligands with overlapping binding pockets, but not a third ligand present in the cell was designed. Each gene was duplicated in a model haploid species with a small constant population size and no recombination. One set of models allowed subfunctionalization of binding events following duplication, while another set did not allow subfunctionalization. Modeling under such conditions suggests that subfunctionalization plays an important role, but as a transition state to neofunctionalization rather than as a terminal fate of duplicated genes. There is no apparent selective pressure to maintain redundancy.
Conclusion
Subfunctionalization results in an increase in the preservation of duplicated gene copies, including those that are neofunctionalized, but never represents a substantial fraction of duplicate gene copies at any evolutionary time point and ultimately leads to neofunctionalization of those preserved copies. This conclusion also may reflect changes in gene function after duplication with time in real genomes.
==== Body
Background
A number of mechanisms can generate duplicate copies of genes, ranging from single gene duplications to regional and whole genome duplications [1-3]. Large increases in gene number have been coupled to increases in organismal complexity and radiative divergence at several points in the history of metazoans including during the chordate/vertebrate transition and during the teleost fish divergence [1,4,5].
Metazoans differ from prokaryotes in their much smaller effective population sizes, where theory predicts that neutral stochastic processes will be relatively more important than adaptive processes in the expected case that adaptive mutations are rarer than nearly neutral mutations [6]. Large scale analyses, based upon the ratio of nonsynonymous to synonymous nucleotide substitution rates [7] or MacDonald-Kreitman statistics [8] have indicated small to intermediate degrees of positive selection (adaptive substitutions) in mammals, but these clearly do not represent the majority of substitutions. In such studies, it appears to be specific positions in protein-encoding genes, rather than the genes as a whole that are under positive selection [7]. Even examining substitution as a neutral walk through sequence in a folded protein (ignoring positive selection) has shown such a process to have fairly complex dynamics [9]. From this, it is relevant to examine population genomic phenomena, like the fates of duplicated genes, in the context of physical models of proteins. Further, it is not possible to systematically identify fates of real genes (subfunctionalization to the exclusion of neofunctionalization or vice-versa), so modeling under increasingly realistic conditions is likely to be the best way to understand evolutionary mechanisms.
Pseudogenization or nonfunctionalization is a purely neutral process that ultimately eliminates one of the duplicated copies as a functional gene and is the most common fate. Subfunctionalization, is an alternative neutral process that leads to an increase in organismal gene number for genes or functions that show modularity (one representative type of modularity is modeled here, but other types are also possible). Neofunctionalization is an alternative process leading to an increase in organismal gene number, but dependent upon rarer adaptive mutations. Neofunctionalization can include the evolution of a completely new binding capability (as modeled here) or modification/improvement of existing binding capabilities under positive selection after removal of pleiotropic constraint. These alternative fates are presented in the context of a lattice model in Figure 1.
Figure 1 Eight different fates are possible in the two simulations, nonfunctionalization (pseduogenization) without cellular death, subfunctionalization, pleiotropic neofunctionalization plus either nonfunctionalization of the other copy or subfunctionalization, non-pleiotropic neofunctionalization, non-pleiotropic neofunctionalization also involving subfunctionalization, redundancy, and cellular death. Some of the fates have additional combinations of activity that are not represented in this figure. It should also be noted that some of the characterizations overlap. For example, pleiotropic neofunctionalization occurs in combination with nonfunctionalization, subfunctionalization, or is redundant.
Lattices are models of folded proteins in square or cubic shapes (a cubic lattice was employed here). The folding of a lattice is dictated by the contacts from amino acids that are not adjacent in the primary sequence (these contacts are present in the folded and unfolded states). Because lattices are small and the folding rules are simple, they can be used for evolving populations of proteins to study their structural properties.
Lattice models have previously been used to make important predictions about the behavior of proteins in evolutionary contexts, including their metastability [10] and the evolvability of new folds [11]. Lattices that bind to peptides [12] and small hydrophobic molecules [13] have been described and the latter used to show that subfunctionalization can lead to an increase in duplicate gene retention rates. Here, model genes that fold into lattices and bind peptides were duplicated, with neofunctionalization and subfunctionalization (simulation A) or just neofunctionalization (simulation B) as possible events that would preserve duplicated copies in a genome, with nonfunctionalization (pseudogenization) as an alternative fate (see Figure 1 and Table 1). The relative levels of duplicate gene preservation and the importance of both neofunctionalization and subfunctionalization were assessed.
Table 1 The initial amino acid sequences including binding sites and folding energies (RT units) are shown for the 10 proteins. Additionally, the three ligands for each protein are shown with their initial binding energies (also RT units) at the two sites (A in italics and B in bold). It should be noted that the two binding sites are adjacent in three dimensions and overlap by one amino acid.
Sequences Conf. Eng. Binding Site A Ligand A Ligand Eng. at A Binding Site B Ligand B Ligand Eng. at B Ligand C Ligand Eng. at A Ligand Eng. at B
1 MSKTAQKRLLKELQQLIKDSPPGIVAGPKSEN
NIFIWDCLIQGPPDTPYADGVFNAKLEFPKDY -0.78 RPAV YRGM -0.84 RKIK YSMD -1.42 YLEG +0.59 +0.40
2 MSTPARRRLMRDFKRMKEDAPPGVSASPLPDN
VMVWNAMIIGPADTPYEDGTFRLLLEFDEEYP -0.73 EALI EIIL -1.07 ERTP EERQ -1.54 EKEP +0.79 +0.58
3 MTTSKERHSVSKRLQQELRTLLMSGDPGITAF
PDGDNLFKWVATLDGPKDTVYESLKYKLTLEF -0.68 TLLS SILL -0.51 TNDP SDEW -0.81 STYL +0.61 +0.23
4 MNMSGIALSRLAQERKAWRKDHPFGFVAVPTK
NPDGTMNLMNWECAIPGKKGTPWEGGLFKLRM -0.88 DLTW GFHW -0.83 DGMN GTFE -1.22 GEEC +0.29 +0.60
5 MSTPARKRLMRDFKRLQQDPPAGISGAPQDNN
IMLWNAVIFGPDDTPWDGGTFKLSLQFSEDYP -0.66 DNSG GTDD -0.86 DNAV GTWF -0.64 GIFH +0.80 +0.53
6 MIVPYNLPLPGGVVPRMLITILGTVKPNANRI
ALDFQRGNDVAFHFNPRFNENNRRVIVCNTKL -1.00 RVKP YLEE -1.61 RQVV YKII -1.13 YQRD +0.78 +0.92
7 MTEENSKSEALLDIPMLEQYLELVGPKLITDG
LAVFEKMMPGYVSVLESNLTAQDKKGIVEEGH -0.96 IGSE ATEY -0.84 ILGT AIPG -1.00 AEFF +0.59 +0.90
8 MEAVIKVISSACKTYCGKTSPSKKEIGAMLSL
LQKEGLLMSPSDLYSPGSWDPITAALSQRAMI -0.64 QMSK YWQY -1.22 QSQS YEPG -1.04 YRFF +0.84 +0.75
9 MDEPPADGALKRAEELKTQANDYFKAKDYENA
IKFYSQAIELNPSNAIYYGNRSLAYLRTECYG -0.37 YAKL EITF -0.71 YDED EKKT -1.92 EMVR +0.94 +1.32
10 MKSRRWFHPNITGVEAENLLLTRGVDGSFLAR
PSKSNPGDLTLSVRRNGAVTHIKIQNTGDYYD -1.45 TERP ASER -1.47 TYRS AMDD -1.25 ALIL +1.18 +0.82
Results and discussion
A set of 10 stably folded lattices was designed to each bind to 2 different ligands at overlapping sites. A third ligand was present in the cell, but did not bind at either site at the start of the simulation. The lattice was duplicated in a constant population of 1000 cells, where those cells that bound the third ligand were 5% more likely to appear in the next generation (a selection coefficient of 5% is arbitrary, but only serves as a scaler of the results). In each generation, 10% of molecules became nonfunctional at random through transcriptional knock-out. The fitness function required molecules to fold and genomes to have binding capabilities for the first two ligands. Cells were selected under the constraint that the first and second ligands needed to be bound, but could be bound by either molecule (subfunctionalization possible) in simulation A. The second simulation (simulation B) tightened this constraint and required both ligands to be bound to the same molecule (subfunctionalization not possible). In simulation B, only neofunctionalization is possible as a mechanism to preserve both copies non-redundantly. Neofunctionalization can occur through two mechanisms, a pleiotropic mechanism where the third ligand binds at a site that is also capable of binding one of the other two ligands and a non-pleiotropic mechanism, where the third ligand binds to an inactive site. The average values of each fate (from 10 different lattices) in each of the two simulations are shown in Tables 2, 3.
Table 2 For simulation A the final average (over 10 different peptides each repeated 10 times) frequency of each fate across generations is reported. Error bars are reported as the standard error of the mean.
A
Generations Nonfunc. Sub Neo-P Neo-NP Neo-P+Sub Neo-NP+Sub Redundant
1 40.27 ± 1.10 0.01 ± 0.01 8.71 ± 0.90 0.23 ± 0.08 0.00 ± 0.00 0.00 ± 0.00 949.11 ± 1.37
20 360.89 ± 5.67 1.80 ± 0.24 127.72 ± 0.59 10.22 ± 0.88 0.37 ± 0.10 0.11 ± 0.04 469.49 ± 9.90
40 421.28 ± 6.83 4.41 ± 0.50 217.26 ± 13.95 18.54 ± 1.49 3.63 ± 0.75 1.55 ± 0.31 294.25 ± 10.85
60 410.64 ± 7.58 5.95 ± 0.53 294.73 ± 14.91 25.71 ± 1.99 7.57 ± 1.10 2.86 ± 0.48 209.70 ± 9.83
80 381.06 ± 7.72 7.47 ± 0.59 356.11 ± 14.76 29.03 ± 2.09 14.37 ± 1.67 5.21 ± 0.75 161.78 ± 9.00
100 352.85 ± 8.02 8.02 ± 0.61 402.22 ± 14.33 30.66 ± 2.33 24.05 ± 2.37 7.46 ± 0.95 128.69 ± 7.76
120 329.56 ± 7.83 9.32 ± 0.79 432.93 ± 14.01 32.30 ± 2.73 31.85 ± 2.95 8.14 ± 1.08 109.06 ± 6.94
140 312.37 ± 6.97 10.15 ± 0.93 450.22 ± 13.10 30.22 ± 2.69 43.83 ± 3.24 12.14 ± 1.49 93.30 ± 5.98
160 291.73 ± 7.02 10.33 ± 0.83 473.25 ± 12.97 29.06 ± 2.69 49.20 ± 3.77 13.92 ± 1.72 83.90 ± 5.23
180 284.48 ± 6.47 11.58 ± 0.84 478.46 ± 12.60 27.93 ± 2.51 57.99 ± 4.42 14.78 ± 2.03 76.08 ± 4.81
200 274.61 ± 6.18 12.51 ± 1.00 481.93 ± 12.78 27.62 ± 2.68 68.17 ± 4.70 16.09 ± 2.27 71.22 ± 4.35
Table 3 For simulation B the final average (over 10 different peptides each repeated 10 times) frequency of each fate across generations is reported. Error bars are reported as the standard error of the mean.
Generations Nonfunc. Neo-P Neo-NP Redundant
1 60.78 ± 2.06 8.85 ± 0.88 0.17 ± 0.06 928.63 ± 2.03
20 495.42 ± 9.68 128.62 ± 10.05 10.41 ± 0.91 336.46 ± 8.71
40 550.78 ± 11.61 222.18 ± 14.30 18.90 ± 1.49 168.86 ± 7.35
60 525.49 ± 12.97 303.64 ± 15.17 25.79 ± 1.86 102.54 ± 5.14
80 486.51 ± 12.83 369.06 ± 14.85 29.59 ± 2.21 69.92 ± 3.64
100 451.13 ± 12.54 418.99 ± 14.29 30.86 ± 2.37 52.86 ± 2.55
120 420.75 ± 12.05 457.10 ± 13.89 31.39 ± 2.45 44.07 ± 1.90
140 397.89 ± 11.73 488.39 ± 13.43 28.02 ± 2.25 38.97 ± 1.69
160 382.95 ± 11.24 509.07 ± 12.82 26.47 ± 2.24 34.40 ± 1.54
180 377.63 ± 10.28 518.26 ± 11.70 24.83 ± 2.24 31.15 ± 1.50
200 370.23 ± 9.59 529.04 ± 10.87 22.63 ± 2.10 29.79 ± 1.21
While initially, both models generate similar levels of neofunctionalization, with time model A begins to show significantly more neofunctionalization. In model A, the total number of subfunctionalized genes, including those that have also neofunctionalized increased initially, but then reached a plateau. These results are shown in Figures 2 (neofunctionalization), 3 (subfunctionalization), and 4 (nonfunctionalization, including those that have also neofunctionalized on the other copy). It is clear that allowing subfunctionalization results in a greater retention rate of duplicate genes with less nonfunctionalization, although subfunctionalization without neofunctionalization never accounts for a large fraction of the duplicate genes at any point in evolutionary time (total terminal preservation of both duplicates is shown in Figure 5). Figure 5 indicates that the retention profile is completely different when subfunctionalization occurs compared to when it does not. It is also clear in these simulations that there is not a strong selective pressure to retain robustness through redundancy, as seen in Figure 6.
Figure 2 The total frequency of cells with neofunctionalized molecules (capable of utilizing ligand C) is shown for simulation A in green and simulation B in red. With time, the total level of neofunctionalization in simulation A surpasses that of simulation B.
Figure 3 The frequency of subfunctionalization without neofunctionalization is shown for simulation A, where subfunctionalization is allowed. Subfunctionalization alone never represents a large fraction of cellular genomic fates.
Figure 4 The total rate of nonfunctionalization is shown for simulation A in green and simulation B in red. Simulation B shows a much higher rate of nonfunctionalization at all evolutionary times.
Figure 5 The total rate of duplicate copy retention without redundancy through either neofunctionalization or subfunctionalization is shown for simulation A in green and for simulation B in red. Allowing subfunctionalization to occur results in a different retention profile, with a much higher rate of duplicate copy retention.
Figure 6 The frequency of redundant copies decreases with time in both simulation A (green) and simulation B (red). This occurs faster in simulation B, due to the faster rate of nonfunctionalization.
The role of subfunctionalization as a transition is based upon increasing the mutational space accessible to duplicates to neofunctionalize with removal of selective constraint at a binding site. This walk will differ for different lattices (and for different real proteins), modulating the importance of the effect of subfunctionalization. The rate of neofunctionalization in the absence of gene duplication (the emergence of new function in orthologs) is also related to the accessibility of this pleiotropic walk, but is expected to be even slower than that of neofunctionalization in the absence of subfunctionalization.
While it is not possible to systematically analyze duplicated fates and classify duplicated proteins as neofunctionalized, subfunctionalized, or redundant, this study has implications for our understanding of the role of duplication in the evolution of genomes. Protein segments that have lost function but are stably maintained in an expressed form will drift through sequence space until they achieve a function that makes them the targets of selection. Looking back to the origin of chordates, there is little doubt that gene duplication and the evolution of new function (as evidenced by annotation) went hand in hand. However, it may be that subfunctionalization initially played an important role in preserving copies that subsequently neofunctionalized over the past hundreds of millions of years.
Conclusion
Subfunctionalization has previously been shown to increase the retention rate of duplicate genes using a similar approach [13]. However, when neofunctionalization is included as a possible fate for duplicate genes, subfunctionalization is still important in short time frames after duplication. However, with increasing time, subfunctionalization decreases in importance and its role seems to be to preserve duplicate copies for eventual neofunctionalization, a role as a transition state. Subfunctionalization can still play an important role with larger finite population sizes, but the importance of neofunctionalization as a terminal fate becomes even more dramatic with increasing population size.
Methods
Lattice model for protein sequences
We considered a simplified model of evolving proteins. Our model consisted of a chain of 64 random (codon derived) amino acid monomers on a three dimensional 4 × 4 × 4 cubic lattice, simulating a folded protein. Gene sequences were selected randomly and lattices folded as below. Two adjacent binding sites were randomly selected on each lattice. Three peptides were then designed: two that bound specifically at each site and a third that was bound by neither site, as shown by the binding energies in Table 1.
Lattice folding and selection
Each amino acid was embedded at a single lattice point with distinct amino acids correspond to a distinct lattice point. All amino acids were considered to be of uniform size connected with covalent bonds of uniform lengths. A protein fold corresponded to a self avoiding walk over the embedding. The walking algorithm tracks the sites visited to avoid visiting them again. A contact was assumed to exist between two residues if they were not adjacently covalently connected but were on adjacent lattice points. The energy of the protein in a particular conformation was calculated according to the formula,
where γ(Ai, Aj) is the contact potential between residue type Ai at position i and residue type Aj at position j, and Uij is equal to one if residues i and j are not adjacent in sequence but are on adjacent lattice sites, and zero otherwise. The value of γ(Ai, Aj) is obtained from the symmetric interaction matrix given by Miyazawa & Jernigan [14].
Evolution of lattice proteins
We have simulated two evolution models. Model A, corresponding to the evolution of a set of ten protein sequences (shown in Table 1) evolving to the alternative fates shown in Figure 1 and Model B corresponding to the evolution of the same set of proteins evolving without allowing subfunctionalization. Cells that did not bind ligands A and B (Model A) and ligands A and B in the same molecule (Model B) died. All molecules also needed to fold to be active.
In each model, we considered 1000 haploid cells that did not recombine between copies (independently for all ten gene duplicate pairs), with a protein molecule evolving according to a Poisson distribution with an average of 1 DNA mutation per gene per generation after the duplication event with a transition to transversion ratio of 2. After every generation, 10% of genes were knocked out at random to simulate mutations to transcriptional regulatory sequences and the cells were subsequently divided into the different fates shown in Figure 1. The next generation of cells was picked randomly from the living cells of the previous generation to keep a constant population size, with a 5% selective advantage to the neofunctionalized cells according to the Wright-Fisher selection model [15]. At the start of each generational round, each cell was viable with two gene copies to bind each set of ligands, as shown in Table 1.
Authors' contributions
SR wrote all programs and carried out the simulations and analysis. DAL conceived of the study, supervised its execution, and wrote the manuscript.
Acknowledgements
We are grateful to Matthew Betts, Inge Jonassen, Rein Aasland, and Jessica Liberles for careful reading of this manuscript. We thank FUGE, the Norwegian Functional Genomics Platform for financially supporting this research.
==== Refs
Ohno S Evolution by Gene Duplication 1970 New York: Springer-Verlag
Lynch M O'Hely M Walsh B Force A The probability of preservation of a newly arisen gene duplicate Genetics 2001 159 1789 1804 11779815
Eichler EE Sankoff D Structural dynamics of eukaryotic chromosome evolution Science 2003 301 793 797 12907789 10.1126/science.1086132
McLysaght A Hokamp K Wolfe KH Extensive genomic duplication during early chordate evolution Nat Genet 2002 31 200 204 12032567 10.1038/ng884
Gu X Wang Y Gu J Age distribution of human gene families shows significant roles of both large- and small-scale duplications in vertebrate evolution Nat Genet 2002 31 205 209 12032571 10.1038/ng902
Lynch M Conery JS On the origin of genome complexity Science 2003 302 1401 1404 14631042 10.1126/science.1089370
Liberles DA Schreiber DR Govindarajan S Chamberlin SG Benner SA The Adaptive Evolution Database (TAED) Genome Biology 2001 2 research0028.1 0028.6 11532212 10.1186/gb-2001-2-8-research0028
Fay JC Wyckoff GJ Wu CI Positive and negative selection on the human genome Genetics 2001 158 1227 1234 11454770
Bastolla U Porto M Roman HE Vendruscolo M Statistical properties of neutral evolution J Mol Evol 2003 57 S103 S119 15008407 10.1007/s00239-003-0013-4
Taverna DM Goldstein RA Why are proteins marginally stable? Proteins 2002 46 105 109 11746707 10.1002/prot.10016
Tiana G Shakhnovich BE Dokholyan NV Shakhnovich EI Imprint of evolution on protein structures Proc Natl Acad Sci, USA 2004 101 2846 2851 14970345 10.1073/pnas.0306638101
Williams PD Pollock DD Goldstein RA Evolution of functionality in lattice proteins J Mol Graph Model 2001 19 150 156 11381526 10.1016/S1093-3263(00)00125-X
Braun FN Liberles DA Retention of enzyme gene duplicates by subfunctionalization Int J Biol Macromol 2003 33 19 22 14599579 10.1016/S0141-8130(03)00059-X
Miyazawa S Jernigan RL Estimation of effective interresidue contact energies from protein crystal structures- Quasi-chemical approximation Macromol 1985 18 534 552 10.1021/ma00145a039
Nei M Molecular Evolutionary Genetics 1987 New York: Columbia University Press
| 15831095 | PMC1112588 | CC BY | 2021-01-04 16:37:17 | no | BMC Evol Biol. 2005 Apr 14; 5:28 | utf-8 | BMC Evol Biol | 2,005 | 10.1186/1471-2148-5-28 | oa_comm |
==== Front
BMC Fam PractBMC Family Practice1471-2296BioMed Central London 1471-2296-6-171584515010.1186/1471-2296-6-17Research ArticleFrequency of alcohol use and obesity in community medicine patients Rohrer James E [email protected] Barbara M [email protected] Anne [email protected] Anthony [email protected] Health Services Research, Department of Family and Community Medicine, Texas Tech University Health Sciences Center, Amarillo, Texas, United States2 Department of Psychiatry, Texas Tech University Health Sciences Center, Amarillo, Texas, USA3 Preventive Medicine, Department of Family and Community Medicine, Texas Tech University Health Sciences Center, Lubbock, Texas, USA2005 22 4 2005 6 17 17 3 1 2005 22 4 2005 Copyright © 2005 Rohrer et al; licensee BioMed Central Ltd.2005Rohrer 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
Obesity is an important public health problem. However, the effects of alcohol use on the risk for obesity have not been thoroughly explored. This study focuses on how frequency of alcohol use is related to the risk of obesity in a community medicine clinic population.
Methods
This study used a cross-sectional survey to test the hypothesis that obesity (BMI > 30) is associated with alcohol use. The convenience sample was drawn from three clinics that primarily serve low-income populations. Independent variables included frequency of alcohol use, frequency of binge drinking, demographic characteristics, health behaviors and health status.
Results
In comparison to non-drinkers, people who consumed alcohol 3 or more days per month had lower odds of being obese (Adjusted Odds Ratio = .49, p < .04). As expected, there was a significant association between watching eight or more hours of television per day and obesity (AOR = 2.34, p < .01).
Conclusion
More frequent drinking and less television time are independently associated with reduced odds of obesity in this sample of community medicine patients. Additional research is needed to isolate casual mechanisms.
==== Body
Background
Overweight and obesity are subjects of much public health scrutiny and concern. Obesity is significantly correlated with many chronic diseases, including diabetes, cardiovascular disease, some cancers, and gallbladder disease. The Centers for Disease Control and Prevention (CDC) estimates that in 2003 healthcare expenditures attributable to obesity reached $75 billion [1].
Weight control involves a complex interaction of physical, genetic, psycho-social and environmental factors. Addressing the problem requires a solid understanding of the risk factors and how they can be influenced. Epidemiologists have studied the health consequences of obesity [2-4], and several studies have explored the risk factors for obesity [5-13]. The health behaviors that determine obesity, such as eating too much and exercising too little, are related to other health behaviors such as cigarette smoking and alcohol consumption. However, no consensus has emerged as to casual connections among these behaviors.
Investigators from the National Institutes of Health analyzed the 2001 National Health Interview Survey to estimate the prevalence of multiple behavioral risk factors for chronic diseases [14]. These investigators found that 17 percent of adults had three or more of the following risk factors: cigarette smoking, risky drinking of alcoholic beverages, physical inactivity, and overweight. Less than ten percent of Americans had zero risk factors. About 73 percent had only one or two risk factors, proving that unhealthy behaviors do not always co-occur. We note that smoking illustrates the complexity of these relationships since it is a risk factor for most chronic diseases but is inversely related to obesity.
The purpose of this study was to investigate the importance of alcohol consumption in predicting self-reported obesity in a sample of primary care patients visiting community medicine clinics. The analysis adjusts for the impact of other health behaviors.
Methods
This study used a cross-sectional survey to test the hypothesis that obesity (BMI > 30) is associated with alcohol consumption in a community medicine population. This study was approved by the Institutional Review Board of Texas Tech Health Sciences Center (Amarillo). The convenience sample was drawn from three clinics that serve low-income populations. Participation was voluntary. Pregnant women and persons under age 18 were excluded from participation. Of 1471 surveys distributed to patients and accompanying adult family members, completed forms were received from 793 (54%). Return rates varied by clinic. Height and weight responses were complete for 747 subjects.
Clinic 1 is a university-based family medicine clinic providing a full range of primary care services to cross-generational clients. It is staffed by family medicine physicians and residents. Daily census was approximately 85 clients daily, of which, less than 5 % were non-English speaking. Clinic personnel distributed 500 survey forms over an eight week period with an 80.8% return rate. Clinic 2 serves women and children, providing obstetrical, well care (including immunizations), and acute care services to a targeted high-risk, low socioeconomic sub-population. It is staffed by pediatric and OB-GYN physicians and residents. Approximately 30% of the clinic clients do not speak English. A total of 471 surveys were distributed over a period of 18 weeks with a return of 37.6%. Both the large number of obstetrical patients (ineligible for survey) and percentage of non-English speaking clients contributed to the low return rate. Clinic 3 provides primary care services to a population of indigent adults meeting residential and income screening requirements. It is staffed by internal medicine physicians and residents. A total of 500 surveys were distributed in this clinic over a ten week period with a return of 42.6%.
The dependent variable was BMI calculated from self-reported height and weight data. Of 793 completed surveys, a BMI could be calculated for 747. The continuous BMI variable was collapsed into two classifications: Non-Obese (BMI <30) and Obese (BMI ≥ 30).
Alcohol use was measured by both frequency and intensity. Frequency of alcohol consumption was measured as days per month (grouped into none, 1–2 and 3 or more). Intensity of alcohol consumption as measured by frequency of bingeing (times per month when drank five or more drinks, grouped as none, 1–2 and 3 or more). Other independent variables included gender, age, race/ethnicity (non-Hispanic White, Hispanic, Black, other), marital status (married vs single), educational level (less than high school, high school, some college, college graduate), number of persons living in the home (just me, one additional person, two or more), number of places to walk (none, one place, two or more), worry about having enough food (yes vs no), frequent mental distress (less than 15 days per month versus 15 or more), hours spent watching TV daily (less than 3, 3–7, or 8 or more), number of cigarettes per day (none, 1–20, more than 20), days per week of impaired activities of daily living (none vs one or more), days per week of exercise (none, 1–3, 4 or more) and health confidence. Health confidence was measured by the respondent's confidence that he or she can take of her own health (strongly agree, agree, not sure, disagree, strongly disagree).
Initial bivariate analysis using Pearson χ2 was used to identify associations between obesity and the independent variables. Variables with p values less than 0.1 in the bivariate tests were retained for inclusion in a multivariate logistic regression model. Some cases were dropped due to missing data, resulting in a final sample of 594 in the regression model.
Results
Of the 747 calculated BMI measurements, 515 (65%) fell into the overweight and obese categories. This is very similar to the BRFSS 2002 overweight/obese percentage of 63% for the state of Texas as a whole. Approximately, 40% of the sample BMI measurements were equal or greater than 30 (obese range). About 70% of respondents were non-drinkers and about 90% never binged.
In the univariate analyses, there was no statistically significant difference in obesity by gender, race/ethnicity, educational level, marital status or family size. Days per month of alcohol use was associated with obesity (p = .001), as was intensity (p = .01). Both bingers and daily drinkers were less likely to be obese. Age was significantly associated with obesity, as was number of places to walk, worry about having enough food, health confidence, frequent mental distress, hours watching TV, not smoking cigarettes, frequent mental distress, days with impaired ADLs, and days per week of exercise (see Tables 1 and 2).
Table 1 Percent Obese by Socioeconomic Characteristics of Sample (N = 747)
Overall Not Obese Obese p-value
% (#) % %
Gender% 0.734
Male 19.46 (144) 61.11 38.89
Female 80.45 (596) 59.56 40.44
Age group% 0.004
Under 35 36.44 (270) 66.67 33.33
35–45 19.57 (145) 56.55 43.45
46–55 18.89 (140) 48.57 51.43
56–65 12.42 (92) 56.52 43.48
Over 65 12.69 (94) 65.96 34.04
Race/ethnicity% 0.367
non-Hispanic White 67.40 (490) 61.22 38.78
Hispanic 23.93 (174) 56.90 43.10
Black 7.29 (53) 50.94 49.06
Other 1.38 (10) 70.00 30.00
Marital status% 0.145
Married 52.94 (234) 62.57 37.43
Single 47.46 (140) 57.30 42.70
Educational status% 0.148
< high school degree 8.67 (35) 54.69 45.31
high school degree 36.86 (272) 64.71 35.29
Some college 38.48 (284) 55.99 44.01
college graduate 15.99 (118) 61.86 38.14
Number of persons living in the home% 0.956
Just me 16.21 (119) 60.50 39.50
One additional person 26.29 (193) 60.62 39.38
Two or more 57.49 (422) 59.48 40.52
Number of places to walk% 0.023
None 29.91 (210) 52.38 47.62
One place 15.95 (112) 60.71 39.29
Two or more 54.13 (380) 63.95 36.05
Worried about having enough food% 0.033
Yes 26.91 (194) 53.09 46.91
No 73.09 (527) 61.86 38.14
Table 2 Percent Obese by Health and Behavioral Characteristics of Sample (N = 747)
Overall Not Obese Obese p-value
% (#) % %
Usually can take care of own health% 0.020
Strongly agree 20.08 (142) 69.72 30.28
Agree 35.79 (253) 62.06 37.94
Not sure 13.30 (94) 53.19 46.81
Disagree 22.21 (157) 52.87 47.13
Strongly disagree 8.63 (61) 55.74 44.26
Frequent mental distress% 0.001
Less than 15 days/month 72.50 (514) 63.81 36.19
15 days or more 27.50 (195) 50.26 49.74
Days with impaired ADL% 0.027
None 58.92 (416) 63.46 36.54
One or more 41.08 (290) 55.17 44.83
Hours spent watching TV daily% 0.001
Less than 3 39.41 (256) 65.63 34.38
3 to 7 hours 50.00 (327) 54.74 45.26
8 or more 10.86 (71) 42.25 57.75
Number of cigarettes per day% 0.008
None 71.72 (535) 56.26 43.74
1 to 20 20.11 (150) 68.67 31.33
More than 20 8.18 (61) 68.85 31.15
Drink days per month% 0.001
None 69.45 (516) 55.62 44.38
1 to 2 16.15 (120) 65.83 34.17
3 or more 14.14 (107) 73.83 26.17
5+ Drinks/day per month % 0.010
None 86.35 (639) 58.22 41.78
1 to 2 7.85 (58) 62.07 37.93
3 or more 5.81 (43) 81.40 18.60
Days/Week Exercise% 0.005
None 31.65 (226) 57.96 42.04
1 to 3 45.10 (322) 54.04 45.96
4 or more 23.25 (166) 69.28 30.72
In the multivariate analysis, having places to walk, days with impaired ADLs, and being worried about food, and days of exercise per week were not found to be independently related to obesity (Table 3). Obesity was inversely related to cigarette smoking. Specifically, in comparison to non-smokers, persons who smoked 1–20 cigarettes per day had significantly lower adjusted odds of being obese (AOR = .56) and persons who smoked more than 20 cigarettes per day did as well (AOR = .33).
Table 3 Multiple logistic regression analysis* of obesity (BMI greater than or equal to 30) in community medicine patients (N = 594)
Odds Ratio (95% C.I.) p-value
Days/Week Exercise
None Reference
One to three 1.22 (0.81–1.86) 0.341
Four or more 0.61 (0.37–1.01) 0.054
Usually can take care of own health
Strongly Agree Reference
Agree 1.32 (0.79–2.19) 0.286
Not Sure 2.01 (1.05–3.82) 0.034
Disagree 1.54 (0.88–2.71) 0.133
Strongly disagree 1.14 (0.54–2.41) 0.722
Frequent mental distress
Less than 15 days Reference
15 days or more 1.56 (1.01–2.40) 0.043
Hours spent watching TV daily
Less than 3 Reference
3 to 7 hours 1.44 (0.97–2.12) 0.068
8 or more 2.34 (1.23–4.46) 0.009
Number of cigarettes per day
None Reference
1 to 20 0.56 (0.35–0.89) 0.014
More than 20 0.33 (0.16–0.68) 0.003
Drink days/month
None Reference
1 to 2 0.61 (0.35–1.05) 0.074
3 or more 0.49 (0.25–0.96) 0.037
5+ Drinks/day per month
None Reference
1 to 2 1.13 (0.52–2.46) 0.764
3 or more 0.91 (0.33–2.54) 0.856
*Adjusted for age, places to walk, worried about having enough food, days with impaired activities of daily living.
In comparison to non-drinkers, people who consumed alcohol 3 or more days per month had lower odds of being obese (AOR = .49, p = .037). Binge drinking was not significantly related to obesity. However, we note that only 43 respondents binged three or more times per month. Therefore, power may not have been adequate to test this hypothesis.
As expected, there was a strong association between watching eight or more hours of television per day and obesity (AOR = 2.34, p<.01).
Persons who were not sure about the statement "I have usually been able to take care of my own health without the help of a doctor or a nurse" had increased odds of being obese (adjusted odds ratio = 2.01, p = 0.034). There was a higher degree of poor mental health as measured by frequent mental distress among obese respondents (AOR = 1.56, p = 0.043).
Discussion
Recent epidemiological studies of alcohol use and health have measured alcohol use in a variety of ways. Some measured frequency of alcohol use while others measured intensity (drinks per occasion or total drinks) and some measure alcohol use both ways. Researchers also measured obesity in several different ways. These have included body mass index, waist circumference, and hip circumference.
The relationship between episodic heavy alcohol use and overall self-rated health was examined by Okosun et al [15]. Episodic heavy drinking was defined as consumption of 5 or more drinks at one occasion (for men) or 4 or more drinks (for women). This kind of bingeing was a significant risk factor for poor self-rated health. Okosun et al did not report any findings relating to total drinks per week or number of drinking episodes.
Paschall, Freisthler and Lipton, who studied alcohol use and depression among young adults created seven categories of alcohol use based on drinks per occasion rather than frequency of drinking [16]. They defined moderate drinking as no more than 1–2 drinks per occasion. Their analysis of data from the National Longitudinal Study of Adolescent Health showed that moderate drinkers were less likely to have depressive symptoms than either heavy drinkers or abstainers. Since depression and alcohol use both are related to obesity, this is an important finding.
Vadstrup et al studied waist circumference in a large population survey of Danish adults [10]. They measured alcohol use with total number of drinks per week but counted neither the number of drinking days nor the number of drinks per occasion. Vadstrup et al found the smallest waist circumference in persons who consumed 1–7 drinks per week.
In another Danish population-based study, Tolstrup et al studied both frequency of drinking and total alcohol intake [11]. They found obesity to be positively correlated with total drinks consumed but inversely correlated with frequency of drinking. Obesity was measured by waist circumference and hip circumference.
Finally, we note that Breslow and Smothers found a positive and linear relationship between BMI and quantity of alcohol, while drinking frequency was positively related to BMI [12]. The lowest BMI was found among persons who drank small quantities frequently. This study was limited to persons who never smoked, so the results may not be generalizable.
Our study differs from previous research in that it uses a sample drawn from a primary care population that is largely low-income. We employed BMI to measure obesity and we measured alcohol use with drinking frequency as well as intensity. Our measure of intensity (binge episodes) was answered too infrequently to allow for firm conclusions. With this design, we found that persons who drink more often were less likely to be obese. This finding corroborates the results reported by Breslow and Smothers. Since our study appears to be the first to focus on drinking frequency in a low-income primary care population, we think the results are useful to investigators who study the epidemiology of obesity.
Drinking frequency is inversely related to obesity in this sample of primary care patients who visited community medicine clinics. While it is unlikely that alcohol consumption has a direct beneficial effect on obesity, we suspect that a report of no alcohol consumption may be a proxy for some other, unknown variable that increases risk of obesity. One hypothesis is that persons who do not consume any alcohol may be following a religious prohibition that is associated with other behaviors that relate to obesity, such as group dining involving high-calorie meals. Another possible explanation is that a subgroup of persons who refrain from all alcohol but have frequent mental distress may be using food as a method for coping with stress.
Our finding about the relationship between drinking and lower risk of obesity clearly should be replicated in other studies before attributing it to any theory or generalizing it to other populations. Nevertheless, we note that earlier studies have reported that non-drinkers are at increased risk of obesity and the National Institute on Alcohol Abuse and Alcoholism (NIAAA) has found some benefits from moderate use of alcohol [8,13,14,17-19]. The NIAAA reported that the relationship between moderate alcohol consumption and weight gain, BMI, or obesity remains inconclusive. At the same time, there is some protective effect of moderate alcohol consumption on two major sequelae of obesity, i.e., metabolic syndrome and diabetes. And, of course, numerous studies have consistently found that coronary heart disease deaths are less likely for persons who consume moderate amounts of alcohol [17].
Our findings should not lead researchers or clinicians to conclude that high levels of alcohol consumption are healthy. In addition to the risk of alcoholism, persons who consume large amounts of alcohol are at increased risk for depressive symptoms. Since many of the subjects in our sample already have frequent mental distress, promoting moderate use of alcohol in this population would represent a misguided effort to combat obesity with adverse health ramifications analogous to recommending cigarette smoking as a weight-loss strategy.
Our study confirmed the known relationship between poor mental health and obesity. In 2001, the Behavioral Risk Factor Surveillance System (BRFSS) survey found that 10% of adults reported frequent mental distress (poor mental health ≥ 15 days per month). Persons reporting frequent mental distress (FMD) were more likely to have adverse, but modifiable, health behaviors including smoking, drinking, physical inactivity, and obesity [19]. In a study examining BRFSS data aggregated across several years, 12.3% of women reported frequent mental distress and this group was more likely to consume cigarettes and alcohol [20]. Hence, persons with frequent mental distress are more likely to have physical problems that require treatment in primary care settings and engage in behaviors that make treatment more difficult. Identifying and treating the underlying causes of mental distress is important from an economic, medical, and psychological perspective [21-25].
In our study, persons with obesity were more likely to report frequent mental distress than persons who were not obese. "Frequent mental distress" is not diagnostic of any particular psychiatric illness, such as depression. Nevertheless, the increased self report of mental distress among patients with obesity suggests that psychopathology, potentially treatable, may be unrecognized in a group with several identifiable risk factors for depression.
In July 2003, the President's New Freedom Commission published recommendations calling for enhanced efforts to screen for mental disorders in primary care settings [26]. Depression screening in primary care settings has also been recommended by the US Preventive Task Force (USPTF) [27]. Obese patients represent a readily identifiable subgroup of patients that can clearly benefit from the use of office-based screening tools that are brief, inexpensive, and easy to administer. Despite the availability of a number of instruments available to screen for mental disorders in primary care settings, such screening often does not occur [28-30]. Although the reasons why screening has not been widely implemented are unclear, they include time constraints, lack of reimbursement, stigma (on the part of both the patient and the physician), and financial barriers to effective treatments (e.g., psychotherapy and antidepressant agents).
Finally, we note that health confidence, believing that one can control one's own health, appears to have a protective effect against obesity. Earlier research has shown that health confidence, sometimes called medical skepticism, is related to better self-rated health [31].
There are several limitations to this study. The use of self-reported data for height and weight may result in underestimation of true BMI. Additionally, the study population was intentionally chosen to over sample high-risk populations and may not be representative of the general community population. Epidemiological studies have demonstrated an association between physical and psychiatric disorders. Using a convenience sample of ambulatory care patients and their families may over represent the portion of persons with poor mental health. Lastly, this survey was taken from a rather homogeneous group that was relatively young. Only 13% of persons in this study were over 65 years old. The results of this study may not be applicable to all groups, especially geriatric patients.
Conclusion
Our findings demonstrated an association between obesity and frequency of alcohol consumption. This result is reminiscent of the discovery that moderate alcohol consumption was related to reduced risk of heart disease [17]. An accumulation of evidence on this point led some physicians to recommend consumption of one or two drinks a day to their patients. If other researchers find occasional drinking is related to reduced risk of obesity, a similar change in medical advice could occur.
The mechanism by which alcohol use might reduce obesity is not immediately clear. Some investigators have observed that consumption of sugars is greater among non-drinkers [32]. It is possible that alcohol is an appetite suppressant for regular drinkers who do not binge.
While the association between frequent mental distress and obesity is not new or surprising, recognition of that relationship may provide an easy, effective, and important screening mechanism to detect depression in primary care settings. We believe that obesity, particularly obesity in women, may identify a high risk group that can be targeted for depression screening in primary care settings. Replication of these results by other investigators is important. In particular, establishing a direct association between frequent mental distress and depression among obese women will be important in improving the health care in this group of patients.
Obesity seems to be a phenomenon of middle age. This suggests a cohort effect, probably related to community changes in energy balance in the past few decades (e.g. increasing eating quantity and calorie density, and decreasing physical activity. We also note that obesity in this sample seems to be associated with a constellation of personal characteristics: non-smoking, TV watching, mentally distressed, and alcohol abstaining. While it is tempting to try to "explain" each of these characteristics individually (e.g. smoking is known to reduce body weight), it might be more productive to look at them as a personality pattern. For example, if this pattern were interpreted as describing passive people with little self-efficacy, then it might be more effective to try changing their environments than their volitional behavior. Research projects are needed that sort out causal mechanisms and test 'health protection' approaches to weight control: i.e., experimental interventions targeted at changing the environments of the middle-aged person who may be unable to change her or his own behavior.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
JR planned the study, organized the survey and wrote the introduction, discussion and conclusions. BR wrote sections relating to mental health and AW wrote sections relating to family practice. AD analyzed the data and wrote the first draft of the methods and results.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Finkelstein EA Fiebeldorn IC Wang G State-level estimates of annual medical expenditures attributable to obesity Obesity Research 2004 12 18 24 14742838
Ford ES Moriarty DG Zack MM Mokdad AH Chapman DP Self-Reported Body Mass Index and Health-Related Quality of Life: Findings from the Behavioral Risk Factor Surveillance System Obesity Research 2001 9 21 31 11346664
Basset MT Perl S Obesity: The Public Health Challenge for Our Time Am J Public Health 2004 94 1477 15359488
Must A Spadano J Coakley EH Field AE Colditz G Dietz WH The disease burden associated with overweight and obesity JAMA 1999 282 1523 1529 10546691 10.1001/jama.282.16.1523
Tavani A Negri E La Vecchia C Determinants of body mass index: a study from northern Italy Int J Obes Relat Metab Disord 1994 18 497 502 7920876
Lopez R Urban Sprawl and Risk for Being Overweight or Obese Am J Public Health 2004 94 1574 1579 15333317
Rohrer JE Rohland BM Psychosocial risk factors for obesity in a family planning clinic BMC Fam Pract 2004 5 20 15380026 10.1186/1471-2296-5-1
Kaplan MS Huguet N Newsom JT McFarland BH Lindsay J Prevalence and Correlates of Overweight and Obesity Among Older Adults: Findings From the Canadian National Population Health Survey J Gerontol A Biol Sci Med Sci 2003 58 1018 1030 14630884
Haas JS Lee LB Kaplan CP Sonneborn D Phillips KA Liang SY The Association of Race, Socioeconomic Status and Health Insurance Status With the Prevalence of Overweight Among Children and Adolescents Am J Public Health 2003 93 2105 2110 14652342
Vadstrup ES Petersen L Sorensen TIA Gronbaek M Waist circumference in relation to history of amount and type of alcohol: results from the Copenhagen City Heart Study Int J Obes Relat Metab Disord 2003 27 238 346 12587005 10.1038/sj.ijo.802203
Tolstrup JS Heitmann BL Tjonneland AM Overvad OK Sorensen TIA Gronbaek MN The relation between drinking pattern and body mass index and waist and hip circumference Int J Obes Relat Metab Disord 2005 29 490 497 15672114
Breslow RA Smothers BA Drinking patterns and body mass index in never smokers. National Health Interview Survey, 1997–2001 Am J Epidemiol 2005 161 368 376 15692081 10.1093/aje/kwi061
Smothers B Bertolucci D Alcohol consumption and health-promoting behaviors in a U.S. household sample: leisure-time physical activity J Stud Alcohol 2001 62 467 476 11513224
Fine LJ Philogene S Gramling R Coups EJ Sinha S Prevalence of Multiple Chronic Disease Risk Factors; 2001 National Health Interview Survey Am J Prev Med 2004 27 18 24 15275670 10.1016/j.amepre.2004.04.017
Okosun IS Seale JP Daniel JB Eriksen MP Poor health is associated with episodic heavy alcohol use: evidence from a National Survey Pubic Health 2005 119 509 517 10.1016/j.puhe.2004.08.016
Paschall MJ Freisthler B Lipton RI Moderate alcohol use and depression in young adults: findings from a national longitudinal study Am J Public Health 2005 95 453 457 15727976 10.2105/AJPH.2003.030700
National Institute on Alcohol Abuse and Alcoholism State of the Science Report on the Effects of Moderate Drinking National Institutes of Health, Department of Health and Human Services December 19, 2003
Flechner-Mors M Biesalski HK Jenkinson CP Adler G Ditschuneit HH Effects of moderate consumption of white wine on weight loss in overweight and obese subjects Int J Obesity 2004 28 1420 1426 10.1038/sj.ijo.0802786
Strine TW Balluz L Chapman DP Morarity DG Owens M Mokdad AH Risk Behaviors and Healthcare Coverage Among Adults by Frequent Mental Distress Status, 2001 Am J Prev Med 2004 26 213 216 15026100 10.1016/j.amepre.2003.11.002
Ahluwalia IB Mack KA Mokdad A Mental and physical distress and high-risk behaviors among reproductive-age women Obstetrics Gynecology 2004 104 477 483 15339756
Simon GE Fleck M Lucas R Bushnell DM Prevalence and predictors of depression treatment in an international primary care study Am J Psychiatry 2004 161 1626 34 15337653 10.1176/appi.ajp.161.9.1626
Wells K Quality of Care for Primary Care Patients With Depression in Managed Care Arch Fam Med 1999 8 529 536 10575393 10.1001/archfami.8.6.529
Thompson D Hylan T McMullen W Romeis M Buesching D Oster G Predictors of a Medical-Offset Among Patients Receiving Antidepressant Therapy Am J Psychiatry 1998 155 824 827 9619157
Lyness JM King DA Conwell Y Duberstein PR Self-Rated Health, Depression, and One-Year Health Outcomes in Older Primary Care Patients Am J Geriatr Psychiatry 2004 12 110 113 14729566 10.1176/appi.ajgp.12.1.110
Rohrer JE Medical care usage and self-rated mental health BMC Public Health 2004 4 3 15070417 10.1186/1471-2458-4-3
U.S. Department of Health and Human Services Mental Health: A Report of the Surgeon General Rockville, MD: US Department of Health and Human Services, Substance Abuse and Mental Health Services Administration, Center For Mental Health Services, National Institutes of Health, National Institute of Mental Health 1999
U.S Preventive Services Task Force Guide to Clinical Preventive Services 1996 2 Washington DC: Office of Disease Prevention and Health Promotion
Henkel V Mergl R Coyne JC Kohnen R Moller HJ Hegerl U Screening for depression in primary care: will one or two items suffice? Eur Arch Psychiatry Clin Neurosci 2004 254 215 23 15309389
McAlpine DD Wilson AR Screening for depression in primary care: what do we still need to know? Depress Anxiety 2004 19 137 45 15129415 10.1002/da.20000
Jarjoura D Polen A Baum E Kropp D Hetrick S Rutecki G Effectiveness of screening and treatment for depression in ambulatory indigent patients J Gen Intern Med 2004 19 78 84 14748864 10.1111/j.1525-1497.2004.21249.x
Rohrer JE Borders TF Healthy Skepticism Prev Med 2004 39 1234 1237 15539061 10.1016/j.ypmed.2004.04.038
Colditz GA Giovannucci E Rimm EB Stampfer MJ Rosner B Speizer FE Gordis E Willett WC Alcohol intake in relation to diet and obesity in women and men Am J Clin Nutr 1991 54 49 55 2058587
| 15845150 | PMC1112589 | CC BY | 2021-01-04 16:29:13 | no | BMC Fam Pract. 2005 Apr 22; 6:17 | utf-8 | BMC Fam Pract | 2,005 | 10.1186/1471-2296-6-17 | oa_comm |
==== Front
BMC GenomicsBMC Genomics1471-2164BioMed Central London 1471-2164-6-551583677910.1186/1471-2164-6-55Research ArticleExpression analysis of secreted and cell surface genes of five transformed human cell lines and derivative xenograft tumors Stull Robert A [email protected] Roya [email protected] Scot [email protected] Steve [email protected] Rachel [email protected] Yan [email protected] Cheryl [email protected] Arie [email protected] Daniel J [email protected] PPD Discovery, 1505 O'Brien Street, Menlo Park, California 94025, USA2 Piedmont Research Center, Morrisville, North Carolina 27560, USA2005 18 4 2005 6 55 55 20 9 2004 18 4 2005 Copyright © 2005 Stull et al; licensee BioMed Central Ltd.2005Stull 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
Since the early stages of tumorigenesis involve adhesion, escape from immune surveillance, vascularization and angiogenesis, we devised a strategy to study the expression profiles of all publicly known and putative secreted and cell surface genes. We designed a custom oligonucleotide microarray containing probes for 3531 secreted and cell surface genes to study 5 diverse human transformed cell lines and their derivative xenograft tumors. The origins of these human cell lines were lung (A549), breast (MDA MB-231), colon (HCT-116), ovarian (SK-OV-3) and prostate (PC3) carcinomas.
Results
Three different analyses were performed: (1) A PCA-based linear discriminant analysis identified a 54 gene profile characteristic of all tumors, (2) Application of MANOVA (Pcorr < .05) to tumor data revealed a larger set of 149 differentially expressed genes. (3) After MANOVA was performed on data from individual tumors, a comparison of differential genes amongst all tumor types revealed 12 common differential genes. Seven of the 12 genes were identified by all three analytical methods. These included late angiogenic, morphogenic and extracellular matrix genes such as ANGPTL4, COL1A1, GP2, GPR57, LAMB3, PCDHB9 and PTGER3. The differential expression of ANGPTL4 and COL1A1 and other genes was confirmed by quantitative PCR.
Conclusion
Overall, a comparison of the three analyses revealed an expression pattern indicative of late angiogenic processes. These results show that a xenograft model using multiple cell lines of diverse tissue origin can identify common tumorigenic cell surface or secreted molecules that may be important biomarker and therapeutic discoveries.
==== Body
Background
The process of tumorigenesis has long been recognized to depend upon complex interactions of a tumor with its non-transformed tissue environment [1]. Beyond transformation and increased proliferation, many pathways are activated both in the growing tumor and its environment to culminate in an established solid tumor. For example, adhesive pathways are activated to enable transformed cells to aggregate and form a microtumor. Subsequently, microtumors must avoid destruction by the immune system and elicit vasculature formation for continued growth [2,3]. In support of these events, cell-matrix adhesion proteins, cell surface antigens, angiogenic factors and modulatory agents have been found differentially expressed in several experimental models of tumorigenesis [4-6] and in tumor biopsy samples relative to control tissues [7,8]. Experimental models with established tumorigenic human cell lines have compared the gene expression profiles between the cultured parental cells and after implantation into immune-deficient murine hosts [6]. In this study, we examined this problem with a more focused approach with respect to the transcripts as well as a broader survey by examining multiple tumor sources in order to identify differential genes common to multiple solid tumors in a xenograft model of tumorigenesis.
To recapitulate the attachment and growth of a micro- or metastatic tumor, our experimental tumorigenesis model examined human xenograft tumors in nude mice. It is believed that primary or metastatic microtumors about 1 mm3 in size are metastable; they are either (i) resolved by the immune system, (ii) remain in a steady-state with balanced proliferation and apoptosis or (iii) undergo aggressive growth as long as a vasculature is developed to provide nutrients to the growing mass [9]. Since the end-point of the xenograft assay is the formation of a solid tumor, genes supporting vasculogenesis and angiogenesis are likely differentially expressed relative to the parental cell lines that were adapted to culture in vitro. However, the extent of vascularization to support an established tumor will vary according to the tumor type and tissue environment as a result of variable levels of proteases, receptors or regulators of pericyte and/or endothelial migration, proliferation, and differentiation [3,10]. Additionally, some tumors such as early grade astrocytomas can leverage existing normal brain blood vessels without substantial vasculogenesis for subsequent angiogenic sprouting of new vessels from preexisting vessels [11]. Further, vascularization depends upon a tuned interaction in the tissue microenvironment between endothelial cells and pericytes [12,13]. Vascularization of solid tumors may also be heterogeneous with a rapidly growing margin surrounding a hypoxic core following regression of co-opted vessels that supported early tumor growth [10]. Complicating this picture is the potential for Vascular mimicry' where breast tumor derived cells express endothelial markers and may serve as rudimentary channels [14].
Many angiogenesis studies have used cultured primary vascular endothelial cells and shown the significant roles of VEGF, FGF, PDGF, chemokines and cell-matrix adhesion proteins [3,15,16]. These assays for endothelial cell migration include the chorioallantoic membrane [17], matrigel migration assays [18] or 3D-collagen assays [19]. However, the limits of studying the angiogenic process with established endothelial cells in vitro have been recognized. Tumorigenesis involves both heterophilic and homophilic cellular communication and adhesion between not only endothelial cells but also pericytes and smooth muscle cells; hence other cell surface proteins and secreted factors are absent from such assays [3].
A search for tumorigenic genes common to tumors of diverse origin should be as broad as possible and hence should not be limited to a single tumor type or tissue source. In order to find common tumorigenic genes regardless of tissue origin, we chose to study a panel of 5 adenocarcinoma cell lines from breast, colon, and lung, ovarian and prostate tumors. These cell lines reproducibly yield solid tumors in a standard xenograft assay in immuno-compromised mice [20-22]. While there may be individual differences in capillary branching or density between tumor types, the xenograft assay requires vascular development to support solid tumor formation in a relatively avascular subcutaneous site.
Since the early tumorigenic events largely rely upon secreted factors, cell surface receptors or integral membrane proteins, we devised a strategy to employ a custom microarray to focus on the expression of genes chosen on the basis of their cellular localization. Hence, we implemented an experimental microarray strategy with high replication and coverage of all possible secreted and cell surface proteins. Also, focusing on all known and predicted cell surface and secreted genes allowed us to design more intra-chip replicates for improved data reliability. While prioritizing on the 'Function' category of the Gene Ontology [23], the range of 'Biological Processes' covered by the gene selection remained broad. In contrast to early concerns that a sub-selection of genes might result in a systemic bias, relatively small numbers of genes were found to be common to all xenograft tumors due to the robust experimental design and statistical analysis.
Results
We developed a custom 60-mer oligonucleotide microarray to focus on an ontologically restricted set of secreted and cell surface genes for higher data reliability using a matrix design with intra-chip replicates in addition to replicate chips. Due to the limits of the Gene Ontology classification, multiple strategies had to be used to derive a relatively complete collection of secreted and cell surface genes. For example, some proteins have multiple localization sites on the basis of newer experimental evidence absent from curated databases; e.g. SORCS3, HDGF. For proteins with multiple cellular localizations, the literature (PubMed, NCBI) was the annotation source for finding other secreted and cell surface proteins. Finally, other putative secreted and transmembrane-encoding genes and exons were analyzed from hypothetical predictions from the UCSC Human Genome. Redundant genes were removed by a combination of blastn/blastp comparisons and manual curation, but many putative membrane-encoding exons of potential proteins were included. A final tally of 3531 genes was composed of 1057 secreted genes, 1338 G-protein coupled receptor (GPCR) genes with the remainder classified as various integral membrane proteins and cell surface proteins. An ontological view of the custom chip's content is shown in Fig. 1. Finally, in consideration of potential global changes of a selected set of genes, numerous positive and negative controls were included in the array design; including genes characteristic of some tumors (e.g. the estrogen receptor for a subset of breast tumors) and many 'housekeeping' transcripts (e.g. β-actin) commonly used to normalize quantitative PCR studies. However, co-hybridizing all samples with a reference cDNA derived from a mixture of 10 human cell lines enabled 'normalization' with respect to feature, chip, and dye for the MANOVA analysis. This strategy minimizes the potential concern for a skewed normalization by a sub-selected gene population or possible differential behavior of the included 'housekeeping' genes in the xenograft tumors.
Figure 1 Gene ontology of custom chip probes. The ontological classification of 3531 cell surface or secreted genes was extracted from the Gene Ontology at the third level. Genes lacking GO annotations at this level were derived from level 2.
Identification of characteristic tumor-specific genes by all tumor data or individual tumor types by multivariate analysis of variation
We performed several multivariate analyses of the microarray data to find characteristic tumorigenic genes. The MAANOVA tools [24] were chosen for their sensitivity and robustness in measuring differential expression versus previous T-test and log-ratio methods using thresholds for induction or suppression. This was particularly important in these studies that used a relatively complex design with on-chip and inter-chip probe replication, multiple tumor samples and tumor types, dye-swap and a common reference RNA sample for all hybridizations. Thus, this strategy helps avoid any systematic bias from using a chip containing probes for only secreted and cell surface genes. We developed a custom database [25] to allow dynamic re-grouping of data to facilitate multiple analytical models such as all tumor data or individual tumor types and their parental cell lines.
Initially, we identified the differentially expressed genes in all tumors relative to all parental cells regardless of tissue origin. Hence compared all the xenograft tumor data to all the parental cell line data without regard to tumor type. Similarly, both the tumor and parental cell line data were compared to the all reference cDNA hybridization data. These data were analyzed by both principal components analysis (PCA) and multivariate analysis of variance (MANOVA).
Principal components analysis
To visualize all tumor and parental cell data and assess overall quality, we subjected the entire dataset to principal components analysis. As shown in Fig. 2, a discrete segmentation of the data into 3 major aggregates corresponding to xenografts (circles), parental cell lines ("X's") and the universal reference cDNA (solid dots) can be identified. The third principal component shown by the vertical Y-axis provided the best separation between parental cell data and the xenograft tumor data, Fig. 2.
Figure 2 Principal components analysis of array data. The mean expression values of all samples from all arrays were analyzed by principal components analysis. The first 3 principal components of the analysis are shown from the best vantage point to show separation of the three classes. Open circles represent the parental cell lines, "X" denotes the various xenograft tumors, and the small solid dots are the reference cDNA sample (derived from the Universal RNA) co-hybridized with all experimental samples. The cell lines corresponding to the various tissue sources of the parental cell lines were: Ovary, SKOV3; Prostate, PC3; Breast, MDA MB-231; Colon, HCT116; and Lung, A549.
Linear discriminant analysis
In order to identify a profile characteristic of xenograft tumors where the combination of multiple genes might be more predictive than any single gene, we performed a linear discriminant analysis. Hence, we iteratively 'trimmed' versions of the third principal component since it had the highest correlation to sample type. The 'trimmed' list of coefficients were tested to determine their accuracy in assigning samples to either the tumor or cell line categories. This analysis retained 70 of the largest coefficients of the third principal component and represents a simple linear discriminant (LD) of 70 probes that corresponds to 54 genes. The profile of 70 probes fairly accurately distinguishes between the two sample types of parental cell lines and xenograft tumors, Fig. 3A. In 'leave-one-out' testing where each of the 99 samples was removed in separate analyses, this method generated a profile that was 79.8% accurate in predicting a xenograft tumor. The same method applied to 1000 label-permuted datasets never exceeded 65% accuracy with a median and minimum accuracy of 49% and 39.3% respectively. This suggests that the gene profile generated by our analysis can distinguish between the xenograft data and the cell line data in a verifiable manner.
Figure 3 Genes identified by linear discriminant analysis. The top 70 PCA coefficients along the third principal component were selected. Panel A: Plot of linear discriminant profile of 70 probes that distinguish xenograft tumors from parental cell lines. Positive values in orange indicate "Xenograft tumor" while negative values in blue indicate "Parental Cell line". The y-axis shows either numbered tumor (left) or parental cell (right) samples and the x-axis is an arbitrarily scaled output reflecting the accuracy in assigning a sample as a xenograft tumor or parental cell line. The numbered tumors were grouped according to tissue type as indicated by C for colon (HCT116), B for breast (MDA MB-231), L for lung (A549), P for prostate (PC3) and O for ovary (SKOV-3). Panel B: Graphical representation of the LD-p54 genes expression profiles. For genes with multiple probes, the highest value is shown. Classified by a non-redundant filtering of the Gene Ontology biological process terms, the genes are shown with a color scale representing relative fold induction to all parental cell line data. The left-most color column designated by 'X' is the average ratio, while the remaining five columns correspond to Colon (HCT116), Breast (MDA MB-231), Lung (A549), Prostate (PC3) and Ovarian (SKOV-3) carcinoma xenografts respectively.
Ontological classification of genes identified by a linear discriminant
The 54-gene profile derived from the linear discriminant (LD-p54) was distributed amongst numerous biological processes using the Gene Ontology classification terms, Table 1. Many genes were classified in multiple biological process categories as a result of their biological complexity; e.g. fibronectin (FN1) is classified into 8 biological processes including cell motility, response to stress, cell communication, response to external stimuli, extracellular matrix structural constituent, protein binding and glycosaminoglycan binding. Other genes are involved with cell adhesion or extracellular matrix, cellular growth or the regulation of cellular proliferation, various membrane proteins with known or inferred functions, transporters or channels, and proteases or protease inhibitors. A non-redundant ontological classification of the genes identified by the linear discriminant is shown with a graphical representation of their behavior across all tumor types, Fig 3B. Since the linear discriminant analysis uses a weighted sum, not all of the identified genes behaved consistently across all xenograft tumors; e.g. CD164 or COL4A1, Fig 3B.
Table 1 Gene ontology classification of 54 genes identified by a linear discriminant.
GO Process Genes
GO:0006928 cell motility HAS1 TSPAN-3 FN1 IL8
GO:0006950 response to stress CXCL2 CXCL1 SEPP1 FN1 SPP1 IL8
GO:0007154 cell communication MAGP2 LTBP1 PTGER3 COL4A1 COL12A1 IGFBP3 GPR48 CXCL2 PCDHB9 COL5A1 TNC FZD1 CD164 CHODL CXCL1 HAS1 LAMB3 GPR57 EFNA1 FN1 LAMB1 SPP1 GPR23 GPR44 PRSS11 RAP2B INHBB NPY1R ESR1 IL8 KITLG
GO:0007397 histogenesis and organogenesis KITLG
GO:0007599 hemostasis TFPI2
GO:0007631 feeding behavior NPY1R
GO:0008151 cell growth and/or maintenance FSTL1 NOV IGFBP3 RBP4 MGC2376 CD164 CXCL1 TSPAN-3 SLC11A3 SLC16A8 PLEC1 KTN1 SPP1 COL5A2 PRSS11 INHBB IGFBP7 ESR1 IL8 KITLG
GO:0008152 metabolism PTGER3 KLK13 HAS1 SEPP1 TLL1 PRSS11 MMP7 INHBB RNASE4 ESR1
GO:0008219 cell death PTGER3 SPP1
GO:0009605 response to external stimulus RBP4 CXCL2 CD164 CXCL1 SEPP1 FN1 SPP1 GPR44 INHBB IL8
GO:0009653 morphogenesis ANGPTL4 COL12A1 PCDHB9 CXCL1 LAMB3 TSPAN-3 SPP1 COL1A1 TLL1 INHBB IL8
GO:0009791 post-embryonic development INHBB
GO:0016265 death PTGER3 SPP1
GO:0019058 viral infectious cycle IL8
GO:0030154 cell differentiation SPP1 INHBB
GO:0042698 menstrual cycle INHBB
GO:0046849 bone remodeling SPP1
GO:0046903 secretion INHBB
NA not known CD63 FLJ20559 GP2 AB065858
A level 3 annotation of the biological process Gene Ontology terms was applied to the list. Due to biological complexity, a gene can occur in more than one category.
Analysis of variation of all xenograft data
The expression data was also subjected to ANOVA using all xenograft and parental cell line data. In this analysis, the type of tumor or parental line was ignored. This analysis identified 156 probes representing 149 differentially regulated genes at the 99.9% confidence level, Table 2. The range of induction or suppression of this set of genes (ANOVA-p149) was 6-fold induction and 5-fold suppression. Twenty-nine of the 54 genes found by the above linear discriminant analysis were found in the list of 149 ANOVA-qualified probes. An ontological clustering of the ANOVA-p149 genes revealed patterns of proteases and protease inhibitors, cell-matrix adhesion genes, receptors, ion channels, various ligands including chemokines and interleukins, additional angiogenic genes and several genes of unknown function, Tables 3 and 4 show the major ontological groups.
Table 2 Differentially expressed genes from three analyses. ANOVA of xenograft data vs parental cell lines found 149 differential genes (designated ' Ap'), Linear discriminant analysis found 54 genes (designated 'LD') and ANOVA of individual xenograft tumors yielded a consensus of 12 genes (designated 'Ai'). For each gene, its presence is denoted by '1' and its absence noted by '0'. The maximum MANOVA Pvalue is reported along with the aggregate ratio (designated by 'R'). For genes with multiple independent probes, the probe reporting the maximum Pvalue is shown. Seven genes common to all three lists are in bold text.
Ap LD Ai Gene Pval R
1 1 1 LAMB3 0.001 1.9
1 1 1 ANGPTL4 0.001 2.1
1 1 1 COL1A1 0.001 3.6
1 1 1 PCDHB9 0.001 4.0
1 1 1 GPR57 0.001 5.7
1 1 1 GP2 0.001 5.7
1 1 1 PTGER3 0.001 6.4
1 1 0 KITLG 0.001 0.4
1 1 0 RAP2B 0.001 0.4
1 1 0 COL5A1 0.237 1.0
1 1 0 SEPP1 0.054 1.0
1 1 0 CXCL1 0.3 1.1
1 1 0 TNC 0.001 1.3
1 1 0 LTBP1 0.009 1.3
1 1 0 PRSS11 0.001 1.3
1 1 0 FN1 0.008 1.4
1 1 0 FZD1 0.019 1.4
1 1 0 SPP1 1 1.5
1 1 0 IGFBP7 0.001 1.7
1 1 0 RNASE4 0.008 1.9
1 1 0 CHODL 0.003 2.1
1 1 0 NOV 0.003 2.2
1 1 0 COL12A1 0.001 2.2
1 1 0 MAGP2 0.001 2.6
1 1 0 GPR23 0.574 3.0
1 1 0 TLL1 0.001 3.2
1 1 0 GPR44 0.069 3.6
1 1 0 MGC2376 0.001 4.7
1 1 0 NPY1R 0.183 5.2
1 0 1 EMP3 0.004 0.5
1 0 1 HLA-A 0.001 0.6
1 0 1 GNAO1 0.001 2.5
1 0 0 CCR5 0.001 0.2
1 0 0 C20orf52 0.001 0.4
1 0 0 SORCS3 0.001 0.4
1 0 0 PF4 0.005 0.4
1 0 0 SPINK2 0.001 0.4
1 0 0 IGSF6 0.008 0.4
1 0 0 GPR110 0.001 0.5
1 0 0 OR1J5 0.001 0.5
1 0 0 BGLAP 0.001 0.5
1 0 0 GALR2 0.001 0.5
1 0 0 HCN2 0.001 0.5
1 0 0 CD81 0.001 0.5
1 0 0 OGFR 0.001 0.5
1 0 0 GPR6 0.001 0.5
1 0 0 OMP 0.001 0.5
1 0 0 CMA1 0.001 0.5
1 0 0 DKFZP564DO 0.001 0.6
1 0 0 CHRM1 0.001 0.6
1 0 0 PYY 0.001 0.6
1 0 0 FGF19 0.004 0.6
1 0 0 AGTR2 0.047 0.6
1 0 0 SSTR3 0.001 0.6
1 0 0 TMPO 0.001 0.6
1 0 0 TAS2R16 0.003 0.6
1 0 0 ADORA2B 0.003 0.6
1 0 0 GPR10 0.001 0.6
1 0 0 ADCYAP1R1 0.001 0.6
1 0 0 OR1F10 0.001 0.6
1 0 0 HDGF 0.001 0.6
1 0 0 CD151 0.001 0.6
1 0 0 PDAP1 0.001 0.7
1 0 0 A1BG 0.001 0.7
1 0 0 LIPF 0.001 0.7
1 0 0 PBEF 0.001 0.7
1 0 0 ART-4 0.034 0.7
1 0 0 C1QTNF3 0.029 0.7
1 0 0 SLC39A4 0.022 0.7
1 0 0 IFNGR2 0.001 0.8
1 0 0 ENT3 0.001 0.8
1 0 0 SERPINC1 0.001 0.8
1 0 0 NRP1 0.006 0.8
1 0 0 CACNA1H 0.011 0.8
1 0 0 CD44 0.001 0.8
1 0 0 STC2 0.018 0.8
1 0 0 DLK1 0.064 0.8
1 0 0 F2R 0.388 0.8
1 0 0 EMP2 0.001 0.8
1 0 0 HBE1 0.003 0.8
1 0 0 BSG 0.003 0.8
1 0 0 GPR80 0.001 0.8
1 0 0 APOB48R 0.016 0.8
1 0 0 AMELY 0.001 0.8
1 0 0 IL26 0.006 0.8
1 0 0 TRPM5 0.001 0.8
1 0 0 ENSA 0.001 0.8
1 0 0 OR1F1 0.001 0.8
1 0 0 GP3ST 0.001 0.8
1 0 0 BDNF 0.001 0.9
1 0 0 PLXN3 0.005 0.9
1 0 0 APMCF1 0.134 0.9
1 0 0 SCAMP1 0.001 0.9
1 0 0 PALMD 0.001 0.9
1 0 0 MMP8 0.02 0.9
1 0 0 MFAP3 0.004 0.9
1 0 0 SPAG11 0.001 0.9
1 0 0 A2M 0.031 0.9
1 0 0 NET-2 0.092 0.9
1 0 0 CXCL11 0.001 1.0
1 0 0 KLRB1 0.003 1.0
1 0 0 TF 0.988 1.0
1 0 0 COL14A1 0.001 1.0
1 0 0 IL7 0.002 1.1
1 0 0 COL9A1 0.001 1.1
1 0 0 CCR4 0.001 1.1
1 0 0 FPR1 0.034 1.1
1 0 0 FAP 0.001 1.2
1 0 0 OPCML 0.001 1.2
1 0 0 GPR145 0.001 1.2
1 0 0 GFRA3 0.001 1.2
1 0 0 EDN3 0.001 1.2
1 0 0 IL12B 0.043 1.3
1 0 0 CXCR4 0.026 1.3
1 0 0 PCSK5 0.427 1.3
1 0 0 NID2 0.168 1.3
1 0 0 ITGA4 0.73 1.3
1 0 0 KIAA1870 0.016 1.3
1 0 0 FBLN5 0.001 1.4
1 0 0 TRPV2 0.001 1.4
1 0 0 FGF23 0.119 1.4
1 0 0 TEM5 0.001 1.4
1 0 0 CR1 0.008 1.4
1 0 0 GPA33 0.001 1.4
1 0 0 CLCA4 0.001 1.4
1 0 0 TIMP3 0.006 1.4
1 0 0 MMP10 0.001 1.4
1 0 0 FUT8 0.197 1.4
1 0 0 FIBL-6 0.001 1.4
1 0 0 V1RL1 0.001 1.5
1 0 0 EBI2 0.003 1.5
1 0 0 ADAM28 0.001 1.5
1 0 0 GPLD1 0.008 1.5
1 0 0 CP 0.003 1.5
1 0 0 EPHA3 0.003 1.5
1 0 0 KLK11 0.001 1.6
1 0 0 OR7A17 0.001 1.6
1 0 0 IFI27 0.001 1.7
1 0 0 RNASE6 0.003 1.7
1 0 0 SELPLG 0.001 1.7
1 0 0 CST7 0.092 1.7
1 0 0 LEC3 0.001 1.7
1 0 0 TSHR 0.001 2.1
1 0 0 MC2R 0.001 2.1
1 0 0 SV2 0.001 2.1
1 0 0 SERPINA4 0.001 2.1
1 0 0 ANGPT2 0.003 2.2
1 0 0 LOC84664 0.008 2.3
1 0 0 RNASE1 0.001 2.9
0 1 1 HAS1 1 0.3
0 1 0 SLC16A8 1 0.4
0 1 0 CD164 1 1.0
0 1 0 FSTL1 1 1.0
0 1 0 IL8 1 1.0
0 1 0 KTN1 1 1.0
0 1 0 RBP4 1 1.1
0 1 0 COL5A2 1 1.1
0 1 0 TSPAN-3 1 1.1
0 1 0 CD63 1 1.1
0 1 0 IGFBP3 1 1.1
0 1 0 PLEC1 1 1.1
0 1 0 CXCL2 1 1.2
0 1 0 GPR48 1 1.2
0 1 0 FLJ20559 1 1.2
0 1 0 LAMB1 1 1.3
0 1 0 COL4A1 0.994 1.3
0 1 0 TFPI2 1 1.4
0 1 0 ESR1 0.996 1.5
0 1 0 SLC11A3 0.999 1.6
0 1 0 EFNA1 1 1.6
0 1 0 KLK13 1 2.5
0 1 0 AB065858 1 3.1
0 1 0 MMP7 0.987 3.4
0 1 0 INHBB 1 3.5
0 0 1 PI3 1 0.4
Table 3 Biological process classification of 175 genes derived from three analyses. The 149 genes derived from the ANOVA analysis of xenograft versus parental cell line data, the 54 genes identified by the linear discriminant analysis and the 12 genes derived from the intersect of ANOVA of individual tumors are shown. Gene Ontology terms were extracted at level 3 for the Unigene gene names. Not shown are genes multiply annotated into additional singular categories or genes absent from the Gene Ontology. Percentages were calculated from a total of 317 classifications into 31 Biological Process terms.
GO Process % Genes
GO:0007154 cell communication 26.8% HDGF PTGER3 CD44 TAS2R16 GALR2 IGFBP3 COL5A1 OR1F1 SORCS3 FZD1 LAMB3 SELPLG GFRA3 IL26 CXCR4 PDAP1 SSTR3 ENSA CD151 COL9A1 OPCML GPR145 GPR44 EPHA3 TNC GPR80 HAS1 BGLAP EFNA1 EBI2 EDN3 TSHR F2R PRSS11 NRP1 OMP MC2R INHBB OR7A17 IL8 AGTR2 GPR48 CHODL CXCL1 OR1F10 CHRM1 GPR10 GPR57 NID2 GPR6 LAMB1 CCR5 SPP1 ADCYAP1R1 CXCL11 PCSK5 GPR23 RAP2B IFNGR2 IL7 COL12A1 PYY CXCL2 PCDHB9 GPR110 CD164 PBEF KLRB1 FN1 BSG IGSF6 FBLN5 STC2 ANGPT2 ADORA2B PF4 IL12B FPR1 GPLD1 CCR4 NPY1R ESR1 GNAO1 ITGA4 KITLG
GO:0008151 cell growth and/or maintenance 1. 55% HDGF IGFBP3 SORCS3 TSPAN-3 CXCR4 PDAP1 SSTR3 ENSA TRPV2 SLC39A4 TF EMP3 HBE1 CACNA1H TSHR F2R COL5A2 PRSS11 NRP1 INHBB CLCA4 IGFBP7 IL8 RBP4 MGC2376 CXCL1 CHRM1 A2M CP CD81 CCR5 KTN1 SPP1 SCAMPI OGFR IL7 TRPM5 NOV PYY CD164 PBEF SLC16A8 PLEC1 ANGPT2 HCN2 IL12B EMP2 ESR1 KITLG
GO:0009605 response to external stimulus 12.6% TAS2R16 OR1F1 IL26 CXCR4 ENSA TRPV2 GPR44 EBI2 EDN3 F2R OMP RNASE6 INHBB OR7A17 IL8 HLA-A RBP4 CXCL1 TIMP3 SEPP1 CD81 CCR5 SPP1 CXCL11 IFNGR2 IL7 CST7 CXCL2 CD164 CR1 KLRB1 FN1 IGSF6 STC2 ADORA2B PF4 IL12B FPR1 CCR4 IFI27
GO:0008152 metabolism 7.9% PTGER3 LIPF MMP7 EPHA3 KLK11 HAS1 CMA1 F2R MMP10 FAP PRSS11 RNASE6 INHBB MMP8 CHRM1 SEPP1 CD81 PCSK5 RNASE4 KLK13 ADAM28 FIBL-6 TLL1 IL12B ESR1
GO:0009653 morphogenesis 7.9% BDNF LAMB3 TSPAN-3 GFRA3 CXCR4 CACNA1H BGLAP F2R NRP1 INHBB IL8 AMELY CXCL1 CHRM1 CCR5 SPP1 COL1A1 COL12A1 PCDHB9 FGF19 ANGPT2 TLL1 PF4 IL12B GNAO1
GO:0006950 response to stress 6.0% IL26 CXCR4 F2R IL8 CXCL1 SEPP1 CCR5 SPP1 CXCL11 IFNGR2 IL7 CXCL2 CR1 KLRB1 FN1 ADORA2B IL12B FPR1 CCR4
GO:0006928 cell motility 3.5% GALR2 TSPAN-3 HAS1 CACNA1H F2R NRP1 IL8 PYY FN1 FPR1 GNAO1
GO:0008219 cell death 2.5% PTGER3 CXCR4 SSTR3 EMP3 F2R AGTR2 SPP1 EMP2
GO:0016265 death 2.5% PTGER3 CXCR4 SSTR3 EMP3 F2R AGTR2 SPP1 EMP2
GO:0030154 cell differentiation 1.9% BGLAP INHBB SPP1 FGF23 PF4 IL12B
GO:0007397 histogenesis and organogenesis 1.6% CXCR4 COL9A1 NRP1 IL7 KITLG
GO:0007599 hemostasis 1.3% SERPINC1 TFPI2 F2R PF4
GO:0000003 reproduction 0.9% SPAG11 ADCYAP1R1 ADAM28
GO:0007631 feeding behavior 0.9% GALR2 PYY NPY1R
GO:0009405 pathogenesis 0.9% CXCR4 EDN3 TSHR
GO:0008015 circulation 0.9% CACNA1H EDN3 AGTR2
GO:0046849 bone remodeling 0.9% BGLAP AMELY SPP1
GO:0007586 digestion 0.6% GALR2 PYY
GO:0019098 reproductive behavior 0.3% PI3
GO:0030198 extracellular matrix organization and biogenesis 0.3% COL14A1
Table 4 Molecular function classification of 175 genes derived from three analyses. As in Table 3, the gene names from three analyses were annotated according to the Gene Ontology Molecular Function categories. Not shown are genes multiply annotated into additional singular categories or genes absent from the Gene Ontology. Percentages were calculated from a total of 251 gene classifications into 52 Molecular Function terms.
GO Function % Genes
GO:0004872 receptor activity 20.3% PTGER3 CD44 TAS2R16 GALR2 OR1F1 SORCS3 FZD1 GFRA3 CXCR4 APOB48R SSTR3 OPCML TRPV2 GPR145 GPR44 EPHA3 GPA33 TNC GPR80 EBI2 TSHR F2R NRP1 MC2R OR7A17 AGTR2 HLA-A GPR48 OR1F10 CHRM1 GPR10 GPR57 GPR6 CCR5 ADCYAP1R1 GPR23 IFNGR2 OGFR APMCF1 GPR110 CR1 KLRB1 IGSF6 ADORA2B IL12B GPLD1 FPR1 CCR4 NPY1R ESR1 ITGA4
GO:0005102 receptor binding 10.8% HDGF BDNF SELPLG GFRA3 IL26 ENSA EFNA1 EDN3 F2R INHBB IL8 CXCL1 SPP1 CXCL11 IL7 NOV PYY CXCL2 FGF23 PBEF FBLN5 FGF19 STC2 ANGPT2 PF4 IL12B KITLG
GO:0016787 hydrolase activity 7.6% RNASE1 LIPF MMP7 KLK11 CMA1 MMP10 FAP PRSS11 RNASE6 MMP8 PCSK5 RAP2B RNASE4 KLK13 ADAM28 FIBL-6 TLL1 GPLD1 GNAO1
GO:0005515 protein binding 7.6% CD44 TMPO IGFBP3 CXCR4 LTBP1 PRSS11 INHBB IGFBP7 PI3 MGC2376 A2M NID2 CCR5 IFNGR2 SERPINA4 NOV PLEC1 FN1 CCR4
GO:0046872 metal ion binding 6.0% FSTL1 MMP7 TF CACNA1H BGLAP MMP10 LTBP1 MMP8 CP NID2 PCDHB9 ADAM28 FIBL-6 FBLN5 TLL1
GO:0042277 peptide binding 6.0% GALR2 SORCS3 CXCR4 SSTR3 OPCML GPR44 F2R MC2R AGTR2 GPR10 CCR5 OGFR FPR1 CCR4 NPY1R
GO:0005201 extracellular matrix structural constituent 4.8% COL5A1 COL4A1 COL9A1 COL14A1 MAGP2 TFPI2 COL5A2 AMELY COL1A1 COL12A1 FN1 MFAP3
GO:0004857 enzyme inhibitor activity 3.6% SPINK2 SERPINC1 TFPI2 PI3 AGTR2 A2M TIMP3 SERPINA4 CST7
GO:0015267 channel/pore class transporter activity 2.8% TRPV2 CACNA1H CLCA4 MGC2376 CHRM1 TRPM5 HCN2
GO:0005539 glycosaminoglycan binding 2.8% HDGF FSTL1 CD44 COL5A1 SERPINC1 FN1 PF4
GO:0003676 nucleic acid binding 2.4% TMPO RNASE1 APOB48R RNASE6 RNASE4 ESR1
GO:0000166 nucleotide binding 2.4% EPHA3 FLJ20559 RAP2B APMCF1 HCN2 GNAO1
GO:0016740 transferase activity 2.4% EPHA3 HAS1 GP3ST FLJ20559 NRP1 FUT8
GO:0004895 cell adhesion receptor activity 1.6% CD44 TNC GPLD1 ITGA4
GO:0015075 ion transporter activity 1.6% TRPV2 SLC39A4 CP SLC16A8
GO:0016301 kinase activity 1.2% EPHA3 FLJ20559 NRP1
GO:0030246 carbohydrate binding 0.8% CHODL KLRB1
GO:0005386 carrier activity 0.8% A2M SLC16A8
GO:0005180 peptide hormone 0.8% PYY STC2
GO:0008565 protein transporter activity 0.8% SORCS3 SCAMPI
GO:0008147 structural constituent of bone 0.8% BGLAP COL1A1
GO:0003800 antiviral response protein activity 0.4% IFNGR2
GO:0008189 apoptosis inhibitor activity 0.4% SPP1
GO:0015457 auxiliary transport protein activity 0.4% ENSA
Verification of selected genes by quantitative PCR analysis
The differential expression of selected genes was confirmed by quantitative real-time PCR using the same RNA samples subjected to microarray hybridization. The vast majority of the genes tested by PCR validated the array analysis, Fig. 5. In some instances, discrepancies in fold-induction can be explained by methodological differences since the array data were all normalized to the co-hybridized universal-RNA sample, while the PCR data were normalized to a β-actin probe (data not shown). Differential expression of ANGPTL4, GP2, GNAO1, CCR4, FGF23, SPP1 and COL1A1 were qualitatively consistent in both the PCR and array analyses. However, two of the down-regulated genes identified by the array analysis, both G-protein coupled receptors, were found by PCR to be elevated, albeit with large variability; GPR10 was induced 281-fold SD = 469 and GPR110 induced 50-fold SD = 105. Of the two down-regulated genes examined by quantitative PCR, CD81 was consistent in both assays, while CD44 was measured by PCR as unchanged or minimally induced yet array analysis indicated CD44 was suppressed. However, the aggregate 2-fold CD44 induction as measured by quantitative PCR is the threshold of what is considered significantly distinguishable from unchanged. Finally, while we did not perform PCR with species-specific probes for every gene present in the ANOVA-p149 list, we were able to confirm differential expression of several human genes from mouse genes such as the osteopontin genes, Fig. 5. While this analysis does not rule out the possibility of partial contamination of the array results by some weak cross-hybridization, to guard against this possibility we carefully designed probes to be species-specific under the stringent hybridization conditions used in this study.
Figure 5 Quantitative PCR analysis of selected genes. Two tumors of each tumor type were analyzed by quantitative PCR. The measured fold change relative to cell line was determined. RNA amounts per well being normalized by betaactin signal. In general <2-fold changes are not significant. Hence a call of 1.5 fold down may not actually differ from 1.5 up. Specific tumor types are indicated by the first initial followed by the tumor number: i.e. C1 = colon tumor #1, O1 = ovary tumor #1, L1 = lung tumor #1, B1 = breast tumor #1, P1 = prostate tumor #1.
ANOVA analysis of individual tumor types
To accommodate the possibility that tumor type was an important contributor to differential gene behavior, we performed a third analysis by examining the intersection between the differential genes of each individual tumor type. For this restrictive analysis, we simply examined each tumor type relative to its parental cell line by ANOVA. Approximately 91–312 genes were differentially expressed at 99.9% confidence for each cell line: SKOV-3, 125 differential genes; MDA, 312 differential genes; HCT116, 124 differential genes; A549, 159 differential genes; and PC3, 91 differential genes (data not shown). Twelve genes were found in common amongst these separately analyzed tumor types, ANGPLT4, COL1A1, epithelial membrane protein 3 (EMP3), GNAO1, glycoprotein 2 (GP2), GPR57, HAS1, HLAA, laminin beta 3 (LAMB3), PCDHB9, protease inhibitor 3 (PI3), and PTGER3, Table 2.
Comparison of multiple analyses
In a typical analysis of multivariate data, a particular method is often chosen as a filter for subsequent analyses. In this study, due to the high statistical reliability imparted by the high replicate probe count (n = 18 to 30) enabled by the custom array design, we chose to compare the results of 3 different approaches to the intact dataset but modeled as either all data or individual tumor types. An estimate of the statistical significance of the overlap in differentially expressed genes common to the three analytical methods gave a Pvalue of < 1 × 10-6 as described in the legend to Fig. 6. As shown in Fig. 6, seven of the twelve differential genes found amongst individual tumor ANOVA analyses were common to the linear discriminant gene profile (LD-p54): ANGPLT4, COL1A1, GP2, GPR57, LAMB3, PCDHB9, and PTGER3.
Figure 6 Overlap of differentially expressed genes identified by three analyses: ANOVA-p149 = 149 genes derived from the ANOVA analysis of all data, LD-p54 = linear discriminant list of 54 genes from all data, and ANOVA-i12 = twelve genes resulting from a comparison of differentially expressed genes from the ANOVA analysis of individual tumors compared to parental cell lines. An estimate for the statistical significance for the overlap of differentially expressed genes by the 3 analytical methods was estimated by calculating the product of individual probabilities for the results of each analytical method applied to 3531 genes. The null hypothesis in this case is that each method's "call" as to a given gene's differential expression is independent of the call made by the other two methods. Thus if p1, p2, and p3 represent the chance that each method calls a given gene as differentially expressed (easily estimated as number of genes called/ number of total genes), the chance that all three methods do so is simply pAll = p1*p2*p3 = (54/3531)*(149/3531)*(12/3531) = 2.193e-5. Under our null hypothesis, the total number of genes called by all three methods k will follow a binomial distribution with parameters p = pAll, n = 3531 where P(k = L) ~ Bin(pAll, N). Standard calculation techniques allow us to calculate a p-value for this; i.e. p = P(k > = K) – the chance under the null hypothesis we see as much or more overlap than was actually observed. For our data, we thus have p = P(k> = 7) < 1E-6. Thus, if the methods identified random noise as differential expression, they would be very unlikely to produce the overlap observed, thus supporting the statistical significance of the results. The heat maps indicate relative fold-induction or suppression in a linear color-encoded scale shown at the bottom. Mean ratios are indicated by X, C = colon, B = breast, L = lung, P = prostate, O = ovary.
Real-time PCR analysis generally confirmed either induction or suppression in multiple tumor samples but with higher induction ratios; e.g. from Fig. 5, the level of ANGPTL4 was measured by PCR as induced 19 to 453 fold with a average fold induction of 185 SD = 170 for 10 tumors (2 of each type). The aggregate induction of ANGPTL4 in the array analysis was 2.09 fold (Pcorr < 2e-9). Similarly, COL1A1 was measured by PCR as induced in most tumors with an average 9.8-fold (SD = 9.1) versus a 3.64-fold induction found by microarray analysis. Finally, in ovarian and prostate tumors, angiopoietin 2 (ANGPT2) measured by PCR was elevated 6-fold (data not shown) versus the 2.2-fold induction found by microarray analysis.
Discussion
Overall, the pathways represented by the differential genes in xenograft tumors support a model for late angiogenic expression patterns. In light of the collection of xenografts after 28–29 days post-implantation, is not surprising to find patterns of differential gene expression that reflect a portion of the tumorigenic process rather than a preponderance of early transforming events. This premise is largely supported by the abundance of extracellular matrix, cell adhesion and angiopoetic genes common to the three analyses.
Ten of the 12 induced genes identified by the ANOVA of xenografts were either well-characterized functions or biological roles, particularly angiogenesis (ANGPTL4), morphogenesis (LAMB3, COL1A1, PCDHB9, or cellular mobility or communication (HAS1, PTGER3, PCDHB9, and LAMB3). The role of extracellular matrix genes in tumor growth has been previously noted [7,8]. Interestingly, five of the extracellular matrix genes from the linear discriminant analysis were collagens (COL1A1, COL4A1, COL5A1, COL5A2 and COL12A1) and four of these collagens (COL1A1, COL4A1, COL5A1, and COL5A2) have been previously found induced in primary renal cell carcinomas (4.8, 5.0, 3.25 and 3.6 fold respectively [26]. COL1A1 has also been found induced in most breast carcinomas [27,28], and a subset of ovarian and colon carcinomas [28].
Consistent with an overall pattern of late-stage angiogenesis in xenograft tumors, ANGPTL4 was found consistently induced relative to the parental cell lines by all analyses. ANGPTL4 originally was described as an induced target of peroxisome proliferator-activated receptor gamma that is involved in glucose homeostasis and differentiation of adipose tissue [29]. Subsequently ANGPTL4 was shown to possess angiogenic activity in the chick allochorionic migration assay [30]. More recently, ANGPTL4 was shown to bind and inhibit lipoprotein lipase [31], a function consistent with the cachexia induced by tumors, where a reduction of fatty acid incorporation into fat cells serves the energy needs of the tumor rather than the host. ANGPTL4's angiogenic action has been reported to be independent of VEGF in a renal carcinoma model [30]. Similarly to previous observations of induced angiopoietins in primary renal cell carcinomas (ANGPT2 8.18-fold induced and ANGPTL4 18–32-fold induced [26], we found both ANGPTL4 induction (2.09 fold, Pcorr < 2e-9), and ANGPT2 induction (2.23-fold Pcorr < .005).
Other post-VEGF angiogenic pathways
The role of other elevated angiogenic genes downstream of VEGF bears discussion. The induction of the prostaglandin E receptor 3 (PTGER3- 6.4-fold, Pcorr < .001) is of interest since prostaglandins can induce VEGFA production [32,33] via a hypoxia-induced pathway [34]. Coincident with these observations, IGFBP7 which was differential by ANOVA and linear discriminant analysis, modulates IGF mitogenic activity [35]. IGFBP7 also stimulates prostacyclin synthesis [36] perhaps to take advantage of our observed 6-fold increased PTGER3 expression. Similarly, a human-specific probe for TEM5, a marker of tumor endothelial angiogenesis [37], was also found mildly increased (1.37-fold Pcorr < .001) possibly as a result of vasculogenic mimicry [14,38].
Other factors such as FGF can play an angiogenic role. One FGF isoform was found significantly differential in some tumor combinations; FGF7 was elevated in colon and prostate xenograft tumors (1.5-fold, Pcorr < 8.7e-6 and 3.7-fold, Pcorr < 7.5e-7) respectively but 2-fold suppressed in ovarian tumors (Pcorr <.006), Fig. 4. FGF7 was previously shown to stimulate the growth of endothelial cells of small but not large vessels in the rat cornea [39] and hence supports the notion of vascular remodeling versus vasculogenesis. That differential expression of this gene was found only in some tumor combinations is consistent with the concept that each type of tumor will display individual differences in the balance angiogenic activators and inhibitors, yet the end physiological result, increased tumor vascularization, is the same [3]. Finally, as noted above, genes that help destabilize or remodel vessels such as ANGPT2 and ANGPTL4 were induced, consistent with an overall pattern of late-stage angiogenesis.
Figure 4 Comparison of differential expression of genes in parental cells versus reference cDNA synthesized from universal RNA (left) and all tumors versus parental cell lines (right). Genes differentially expressed in the parental cells relative to the reference cDNA were analyzed by a 2-way ANOVA (Pcorr < .001). A subset of the differentially expressed genes is shown. The corresponding cognate tumors with differential expression at a 99.9% confidence level by ANOVA analysis of tumors vs parental cell line data are shown. The heat maps indicate relative fold-induction or suppression in a linear color-encoded scale shown at the bottom. Mean ratios are indicated by X, C = colon, B = breast, L = lung, P = prostate, O = ovary.
Linking angiogenic pathways to neuropeptide signaling pathways
Additional support for the late, post-VEGF angiogenic pattern of gene expression in xenografts froms from the observed 5-fold induction of NPY1R by both ANOVA and linear discriminant analyses. NPY1R has been reported to play a role downstream of VEGF in vasoconstriction [40] and capillary sprouting and differentiation [41]. Consistent with the observation of NPY1R induction, the potent effect of ligand neuropeptide (NPY) upon angiogenesis was shown to yield branching vasodilated structures distinct from those generated by VEGF [17]. Similarly, neuropeptide Y has been reported to trigger angiogenesis via the NPY2 receptor in ischemic muscle of mice [41]. Interestingly, neuropilin 1 (NRP1) which can act as a co-receptor with VEGFR2 [42] was found suppressed (1.31-fold, Pcorr < .006) while other VEGF receptor levels were not significantly altered. Finally, previous expression profile studies have found NPY1R to be substantially induced in many breast, prostate and pancreatic carcinomas [28].
Additionally, two other differential genes involved in neuropeptide signaling were observed: melanocortin-2 receptor (MC2R)and SORCS3/neurotensin receptor gene. Both MC2R and the SORCS3 were found differentially expressed by ANOVA. MC2R is a GPCR that binds the ACTH peptide while SORCS3 is a homolog of the rat sortilin gene with VPS10 domains characteristic to neuropeptide-binding proteins [43-45]. ACTH has been found to increase angiogenesis of cultured endothelial cells in a 3D-collagen assay [19] and other neuropeptides have been implicated in stimulating VEGF in prostate cancer cells [46].
Conclusion
In this study we compared the expression profiles of secreted and cell surface genes from five different tissue sources. Multiple tumors were derived from each parental cell line to examine the potential for tumor heterogeneity arising from the primary isolate, but we found relatively consistent behavior within any tumor group. However, we also found tumor-specific genes for each tumor type while identifying a profile of genes shared amongst all tumor types by multiple analytical approaches. Overall, our results comprise a foundation of commonly regulated tumorigenic genes across tissues such as fundamental angiogenic inducers and regulators. Given the diverse and complex expression behavior of primary human tumors from any single tissue source [27,28], in the future it will be necessary to examine several established lines from many histologically similar primary tumors as well as different tumor types from the same tissue. Similarly, it will be important to compare the effect of orthotopic implantation sites to the subcutaneous injection site in these preliminary studies. To resolve xenograft micro-heterogeneity, microarray analysis of micro-dissected xenograft or primary tumors can be used. Micro-dissection will also allow the assessment of potential vasculogenic mimicry by aggressive tumor cells that can express endothelial genes [38]. Additionally, the xenograft model can be more readily extended to monitor time-dependent expression profile changes in the development of tumors. Such results can be used in combination or as a filter with other biomarker technologies such as tissue arrays [47] or mass spectroscopy [48] to fully characterize clinical specimens for diagnostic or prognostic purposes. By identifying genes known to participate in angiogenesis and tumorigenesis, our work establishes a baseline to evaluate and compare the full spectrum of gene profile changes in xenografts and clinical specimens. Hence, time and tissue-specific gene and protein profiles may be useful for the discovery of both biomarkers and new therapeutic strategies.
Methods
Custom array design
A two-stage strategy was employed to design the custom oligonucleotide microarray chip. First, for the known secreted and cell surface proteins, we performed keyword filtering of the gene descriptions and annotations of curated public databases such as SwissProt/Trembl [49], the Gene Ontology tables [23], the UCSC Human Genome assembly (hg13, NCBI Build 31 [50]), the GPCR database [51] and public gene tables from technical supply vendors (Affymetrix, Agilent and Illumina). Some of the keywords used were "secreted", "trans-membrane", "glycosylated" and "olfactory". Redundancies and false positives were removed by manual curation.
In order to accommodate continued optimization of a custom chip design, we chose a chip platform that met several criteria: it must allow rapid changes to the master template even for small production batches, possess relative high density, exhibit strong signal-to-noise properties and have high reproducibility (CV < 10%). Hence, a custom oligonucleotide microarray chip (Agilent, Palo Alto, CA) was designed using the curated collection of secreted and cell surface proteins with human-specific 60-mer probes derived from the 3' 1500 nt region of each mRNA sequence. The custom chip was designed with a matrix of technical probe replicates and multiple probes for some genes; e.g. 2 or 3 probes with 1, 3 or 5 copies each per array represented some genes. All probes were curated by elimination of sequences with unfavorable Tm properties, predicted secondary structure or homo-polymer regions. Finally, Blastn [52] analysis was used to confirm human specificity by comparison to mouse sequences.
Cell lines and mice
All cell lines (A549, MDA MB-231, HCT-116, SK-OV3, and PC3) were obtained from the ATCC (Manassas, VA). Xenograft tumors were generated from each parental cell line by either implantation of cells or passage of a fragment from a primary tumor (Piedmont Research Center, Morrisville, NC). For the A549, MDA MB-231 and SKOV-3 lines, le-7 cells grown with 10% fetal calf serum in Dulbecco's modified Eagle's medium at 37°C in 5% C02 were implanted subcutaneously into the flank of between 8 and 10 BalbC (Harlan Labs, Indianapolis, IN) mice. Between 50 and 75% of the mice yielded a palpable primary xenograft tumor. For the HCT116 and PC3 xenograft tumors, 1 mm3 tumor fragments between 103–110 mg were excised from a primary xenograft tumor and passaged into secondary mice for the HCT-116 and PC3 xenograft tumors. For PC3 tumors, 8 male mice were implanted with fragments; otherwise recipient mice were female. The number of tumors processed for hybridization were 5 for SK-OV-3, 5 for PC3, 4 for MDA MB-231, 3 for HCT-116 and 5 for A549.
RNA preparation
For the parental cell lines, total RNA was harvested from 4 x 106 cells using a High Pure RNA isolation kit (Roche Applied Science, Indianapolis, IN) according to manufacturer's instructions. Tumors were excised 22–29 days post-implantation under accredited procedures (Piedmont Research Center, Morrisville, NC), snap-frozen in liquid nitrogen and stored at -80°C until use. Total RNA was prepared from frozen specimens by 24 hr immersion at -80°C in RNAlater-ICE (Ambion, Austin, TX) to 'transition' solid tumors for subsequent homogenization by grinding with a liquid nitrogen-chilled mortar/pestle, followed by resuspension in Trizol (Sigma-Aldrich, E. St. Louis, Mo) and sonication to complete the tissue disruption. Total RNA was extracted using Phase-lock gels (Brinkmann, Westbury, NY), ethanol precipitated, resuspended in RNase-free water, and aliquoted prior to use. Quality control of the total RNA was facilitated by the use of a microcapillary electrophoresis system (Agilent 2100 Bioanlyzer; Agilent Technologies, Palo Alto, CA).
Experimental design and array hybridization
To identify cell surface genes that are consistently differentially regulated amongst the derivative tumors, multiple tumor specimens and their parental source cell lines were hybridized to the custom chips. All biological specimens were co-hybridized with a reference cDNA synthesized from mRNA that is mixture of 10 human established cell lines (Universal RNA; Stratagene, Carlsbad, CA). For each array, amino-allyl labeled single-stranded cDNA was synthesized from 10 (g of sample total RNA and from 10 ug universal RNA using the Agilent Fluorescent Direct Label Kit according to manufacturer's instructions, except that a dNTP mix containing 5-[3-Aminoallyl]-2'-deoxyuridine 5'-triphosphate (AA-dUTP; Sigma-Aldrich) was used (final concentration: 100 (M dATP, dCTP, dGTP; 50 (M dTTP, AA-dUTP). Amino-allyl labeled cDNA was purified using QIAquick PCR columns (Qiagen, Valencia CA) and coupled to either N-hydroxysuccinimidyl-esterified Cy3 or Cy5 dyes (Cy-Dye mono-functional NHS ester; Amersham, Piscataway NJ). Dye-conjugated cDNAs were purified from free dye using the CyScribe GFX purification kit (Amersham). Targets were hybridized to the microarray for 16 hrs at 60°C using an Agilent In Situ Hybridization Kit per manufacture's instructions, washed 10 min in 6× SSC, 0.005% Triton X-102 at 22°C, 0.1× SSC, 0.005% Triton X-102 for 10 min at 4°C, dried under a stream of nitrogen, and scanned with an Agilent Microarray Scanner. Hybridization signals were extracted with Agilent Feature Extraction Software version 7.1, which yielded the median of all pixel intensities for each feature. Since two identical arrays of 8500 features were printed on each chip, a complete dye-swap comparison could be performed per chip. For example, on the left array, a Cy3-labeled biological specimen was co-hybridized with Cy5-labeled cDNA made from universal RNA. For the cognate dye-swap experiment on the right array, a Cy-5 labeled biological specimen was co-hybridized with Cy3-labeled cDNA made from universal RNA. No tumor samples were mixed with any other tumors.
To enable identification of differentially expressed genes with higher statistical reliability, we performed a matrix of hybridizations. The hybridization matrix follows: for the 5 SK-OV-3, A549 and PC3 tumor specimens, 3 of the tumor samples were hybridized to 2 chips each (hence 4 arrays per tumor sample) while 2 tumor samples were hybridized to a 1 chip each (hence 2 arrays for each of these tumors). For the 4 MDA MB-231 tumor specimens, 3 of the tumor samples were hybridized to 2 chips each and 1 tumor was hybridized to a single chip of 2 arrays. For the 3 HCT-116 tumor specimens, all 3 tumors were hybridized to 2 chips each (4 arrays each). For the parental cell lines, HCT-116 cells were hybridized to 6 chips (12 arrays) while the other cell lines were hybridized to 2 chips each (4 arrays). Since most probes were present minimally in triplicate on each array, whenever a tumor sample was hybridized to 2 chips n = (3*4) = 12 per probe. However, since dye-swap hybridizations were routinely performed, n = 6 for the Cy3 and Cy5 signals respectively.
Quantitative PCR
Real-time (RT-) PCR analysis of selected RNA transcripts was performed using either a GeneAmp 5700 Sequence Detection System or an ABI PRISM 7900HT Sequence Detection System with SyBr green chemistry (Applied Biosystems, Foster City, CA). The cDNA produced by reverse transcribing the equivalent of 10 ng of total RNA was loaded per RT-PCR reaction. The following primers pairs were used: beta actin (ACTB) CCTGGCACCCAGCACAAT, GCCGATCCACACGGAGTACT; Human osteopontin (HSPP) AGCAAAATGAAAGAGAACATGAAATG, TTCAACCAATAAACTGAGAAAGAAGC; murine osteopontin (mSpp) ATTTTGGGCTCTTAGCTTAGTCTGTT, GGTTACAACGGTGTTTGCATGA; angiopoietin-like 4 (ANGPTL4)ATGTGGCCGTTCCCTGC, TCTTCTCTGTCCACAAGTTTCCAG; chemokine (C-C motif) receptor 4 (CCR4)ATTCCTGAGCCAGTGTCAGGAG, CTGTCTTTCCACTGTGGGTGTAAG; fibroblast growth factor 23 (FGF23)GGCAAAGCCAAAATAGCTCC, CTGCCACATGACGAGGGATAT; G protein, alpha activating activity polypeptide O (GNAO1) CTAGTCTTTGGGAAACGGGTTGT, AAATCCAACACGGCAAAGGA; glycoprotein 2 (GP2) GCTTTCCACTCCAATTCACACA, CCTGGCCTTGATTCTGTTAATACC; collagen, type I, alpha 1 (COL1A1)TCCCCAGCTGTCTTATGGCT, CAGCACGGAAATTCCTCC; G protein-coupled receptor 10 (GPR10)CATGCTCGAGTCATCAGCCA, TTTCACTGCCCCCTTTGTGT; G protein-coupled receptor 110 (GPR110)AAGCTCTGGAGGCCGACTG, GGCCTTGTCATCCCGACTC; (CD44)TACAGCATCTCTCGGACGGAG, GGTGCTATTGAAAGCCTTGCA; (CD81)CCCTAAGTGACCCGGACACTT, CGTTATATACACAGGCGGTGATG. The identity of each amplicon was confirmed by melting curve analysis at the end of the RT-PCR run.
Array analysis
While the array vendor's feature extraction software 'processed' the hybridization signal to correct for image intensity, background and minor spatial artifacts, chip-to-chip comparisons such as 'reference' versus 'experimental' sample were handled by a custom database [25] built in MySQL [53] with a web interface served by Apache [54]. The database allows the control of experimental design and specification of comparisons and analyses to be performed. Some calculations, like T-Tests and ratios, can be performed in the database or its interface layer, but MATLAB (Mathworks, Natick, MA) was used for ANOVA and principal components analysis (PCA).
For identification of differentially expressed genes, we used the MAANOVA package [55] an implementation of ANOVA for microarray analysis [24]. Array data were loaded into the database and minimally pre-processed for use with this package: where replicate features of the same probe existed in the array design, an arithmetic mean was calculated to yield a single expression level for each probe for each chip. Each tumor or cell line sample was hybridized to 3 separate chips. All signals were Log2 transformed prior to subsequent analyses. These data were used to fit a linear model with factors Gene, Array, Array × Gene, Dye, Dye × Gene, and Sample × Gene. This last attribute is the quantity used for analysis, representing the differential expression of a given gene under a given experimental condition, with the other factors serving to normalize the data. In order to identify differential expression these residuals were analyzed with three statistical tests: a standard ANOVA F-test and two minor variations. A probe had to pass these three tests, generally at 99.9% significance, in order to be called as differentially expressed. A permutation analysis and one-step multiple comparisons correction were applied in conjunction with these tests. It should be noted that since three tests are applied, three P-values result, and when single P-values are listed; the maximum of the three P-values is reported. Finally, since all samples were co-hybridized with cDNAs made from a universal RNA sample, for comparisons of differential gene behavior, approximate 'ratios' were calculated by dividing the paired individual tumor/universal RNA ratio by the paired parental cell/universal RNA ratio.
Ontology annotation
Unigene Gene names were classified by the consistent terms of the Gene Ontology(tm) consortium [23] and the fatiGO interface to the Gene Ontology [56].
Authors' contributions
RAS helped design and implement the experimental strategy by developing many protocols, carried out many of the hybridization experiments and analyzed the PCR data. RT carried out the sample preparation, labelings, microarray and PCR experiments. SK performed the statistical analysis and assisted database design. SO designed and built the microarray database and LIMs. RH and YL helped curate, annotate and design the custom microarray chip design. CN carried out the xenograft studies. AA conceived of the experimental design. DJC designed & managed the experimental strategy, helped curate the gene lists and wrote the manuscript with input from co-authors.
Acknowledgements
This work was completed in collaboration with Surromed, Mountain View, CA. We also wish to acknowledge the helpful comments of Beth Hollister of the Piedmont Research Center, Morrisville, NC.
==== Refs
Paget S The distribution of secondary growths in cancer of the breast Lancet 1889 1 571 3 10.1016/S0140-6736(00)49915-0
Khong HT Restifo NP Natural selection of tumor variants in the generation of "tumor escape" phenotypes Nat Immunol 2002 3 999 1005 12407407 10.1038/ni1102-999
Bergers G Benjamin LE Tumorigenesis and the angiogenic switch Nat Rev Cancer 2003 3 401 10 12778130 10.1038/nrc1093
Glinsky GV Krones-Herzig A Glinskii AB Gebauer G Microarray analysis of xenograft-derived cancer cell lines representing multiple experimental models of human prostate cancer Mol Carcinog 2003 37 209 21 12891630 10.1002/mc.10139
Pedersen N Mortensen S Sorensen SB Pedersen MW Rieneck K Bovin LF Poulsen HS Transcriptional gene expression profiling of small cell lung cancer cells Cancer Res 2003 63 1943 53 12702587
Creighton C Kuick R Misek DE Rickman DS Brichory FM Rouillard JM Omenn GS Hanash S Profiling of pathway-specific changes in gene expression following growth of human cancer cell lines transplanted into mice Genome Biol 2003 4 R46 12844362 10.1186/gb-2003-4-7-r46
Perou CM Molecular portraits of human breast tumours Nature 2000 406 747 52 10963602 10.1038/35021093
Dhanasekaran SM Barrette TR Ghosh D Shah R Varambally S Kurachi K Pienta KJ Rubin MA Chinnaiyan AM Delineation of prognostic biomarkers in prostate cancer Nature 2001 412 822 6 11518967 10.1038/35090585
Fidler IJ The pathogenesis of cancer metastasis: the "seed and soil" hypothesis revisted Nature Reviews 2003 3 453 8 12778135 10.1038/nrc1098
Holash J Wiegand SJ Yancopoulos GD New model of tumor angiogenesis: dynamic balance between vessel regression and growth mediated by angiopoietins and VEGF Oncogene 1999 18 5356 62 10498889 10.1038/sj.onc.1203035
Vajkoczy P Farhadi M Gaumann A Heidenreich R Erber R Wunder A Tonn JC Menger MD Breier G Microtumor growth initiates angiogenic sprouting with simultaneous expression of VEGF, VEGF receptor-2, and angiopoietin-2 J Clin Invest 2002 109 777 85 11901186 10.1172/JCI200214105
Benjamin LE Hemo I Keshet E A plasticity window for blood vessel remodelling is defined by pericyte coverage of the preformed endothelial network and is regulated by PDGF-B and VEGF Development 1998 125 1591 8 9521897
Gerhardt H Betsholtz C Endothelial-pericyte interactions in angiogenesis Cell Tissue Res 2003 314 15 23 12883993 10.1007/s00441-003-0745-x
Shirakawa K Wakasugi H Heike Y Watanabe I Yamada S Saito K Konishi F Vasculogenic mimicry and pseudo-comedo formation in breast cancer Int J Cancer 2002 99 821 8 12115483 10.1002/ijc.10423
Aonuma M Iwahana M Nakayama Y Hirotani K Hattori C Murakami K Shibuya M Tanaka NG Tumorigenicity depends on angiogenic potential of tumor cells: dominant role of vascular endothelial growth factor and/or fibroblast growth factors produced by tumor cells Angiogenesis 1998 2 57 66 14517376 10.1023/A:1009054410624
Hattori K Vascular endothelial growth factor and angiopoietin-1 stimulate postnatal hematopoiesis by recruitment of vasculogenic and hematopoietic cells J Exp Med 2001 193 1005 14 11342585 10.1084/jem.193.9.1005
Ekstrand AJ Cao R Bjorndahl M Nystrom S Jonsson-Rylander AC Hassani H Hallberg B Nordlander M Cao Y Deletion of neuropeptide Y (NPY) 2 receptor in mice results in blockage of NPY-induced angiogenesis and delayed wound healing Proc Natl Acad Sci USA 2003 100 6033 8 12730369 10.1073/pnas.1135965100
Maeshima Y Colorado PC Torre A Holthaus KA Grunkemeyer JA Ericksen MB Hopfer H Xiao Y Stillman IE Kalluri R Distinct antitumor properties of a type IV collagen domain derived from basement membrane J Biol Chem 2000 275 21340 8 10766752 10.1074/jbc.M001956200
Mallet C Feraud O Ouengue-Mbele G Gaillard I Sappay N Vittet D Vilgrain I Differential expression of VEGF receptors in adrenal atrophy induced by dexamethasone: a protective role of ACTH Am J Physiol Endocrinol Metab 2003 284 E156 67 12485811
Giard DJ Aaronson SA Todaro GJ Arnstein P Kersey JH Dosik H Parks WP In vitro cultivation of human tumors: establishment of cell lines derived from a series of solid tumors J Natl Cancer Inst 1973 51 1417 23 4357758
Cailleau R Young R Olive M Reeves WJ Breast tumor cell lines from pleural effusions J Natl Cancer Inst 1974 53 661 74 4412247
Kaighn ME Narayan KS Ohnuki Y Lechner JF Jones LW Establishment and characterization of a human prostatic carcinoma cell line (PC-3) Invest Urol 1979 17 16 23 447482
Gene Ontology
Kerr MK Churchill GA Experimental design for gene expression microarrays Biostatistics 2001 2 183 201 12933549 10.1093/biostatistics/2.2.183
Osborn S Kennedy S Chin D A flexible pipeline for experimental design processing and analysis of microarray data IEEE Comp Systems Bioinformatics 2003 IEEE Computer Society, Los Alamitos, CA 411 412
Lenburg ME Liou LS Gerry NP Frampton GM Cohen HT Christman MF Previously unidentified changes in renal cell carcinoma gene expression identified by parametric analysis of microarray data BMC Cancer 2003 3 31 14641932 10.1186/1471-2407-3-31
Perou CM Distinctive gene expression patterns in human mammary epithelial cells and breast cancers Proc Natl Acad Sci USA 1999 96 9212 7 10430922 10.1073/pnas.96.16.9212
Su AI Molecular classification of human carcinomas by use of gene expression signatures Cancer Res 2001 61 7388 93 11606367
Yoon JC Chickering TW Rosen ED Dussault B Qin Y Soukas A Friedman JM Holmes WE Spiegelman BM Peroxisome proliferator-activated receptor gamma target gene encoding a novel angiopoietin-related protein associated with adipose differentiation Molec Cell Biol 2000 20 5343 5349 10866690 10.1128/MCB.20.14.5343-5349.2000
Le Jan S Angiopoietin-like 4 is a proangiogenic factor produced during ischemia and in conventional renal cell carcinoma Am J Pathol 2003 162 1521 8 12707035
Yoshida K Shimizugawa T Ono M Furukawa H Angiopoietin-like protein 4 is a potent hyperlipidemia-inducing factor in mice and inhibitor of lipoprotein lipase J Lipid Res 2002 43 1770 2 12401877 10.1194/jlr.C200010-JLR200
Harada S Nagy JA Sullivan KA Thomas KA Endo N Rodan GA Rodan SB Induction of vascular endothelial growth factor expression by prostaglandin E2 and El in osteoblasts J Clin Invest 1994 93 2490 6 8200985
Gallo O Franchi A Magnelli L Sardi I Vannacci A Boddi V Chiarugi V Masini E Cyclooxygenase-2 pathway correlates with VEGF expression in head and neck cancer Implications for tumor angiogenesis and metastasis Neoplasia 2001 3 53 61 11326316 10.1038/sj.neo.7900127
Fukuda R Kelly B Semenza GL Vascular endothelial growth factor gene expression in colon cancer cells exposed to prostaglandin E2 is mediated by hypoxia-inducible factor 1 Cancer Res 2003 63 2330 4 12727858
Oh Y Nagalla SR Yamanaka Y Kim HS Wilson E Rosenfeld RG Synthesis and characterization of insulin-like growth factor-binding protein (IGFBP)-7 Recombinant human mac25 protein specifically binds IGF-I and -II J Biol Chem 1996 271 30322 5 8939990 10.1074/jbc.271.48.30322
Yamauchi T Umeda F Masakado M Isaji M Mizushima S Nawata H Purification and molecular cloning of prostacyclin-stimulating factor from serum-free conditioned medium of human diploid fibroblast cells Biochem J 1994 303 591 8 7980422
Carson-Walter EB Watkins DN Nanda A Vogelstein B Kinzler KW St Croix G Cell surface tumor endothelial markers are conserved in mice and humans Cancer Res 2001 61 6649 55 11559528
Hendrix MJ Seftor EA Hess AR Seftor RE Molecular plasticity of human melanoma cells Oncogene 2003 22 3070 5 12789282 10.1038/sj.onc.1206447
Gillis P Savla U Volpert OV Jimenez B Waters CM Panos RJ Bouck NP Keratinocyte growth factor induces angiogenesis and protects endothelial barrier function J Cell Sci 1999 112 2049 57 10341222
Zukowska-Grojec Z Dayao EK Karwatowska-Prokopczuk E Hauser GJ Doods HN Stress-induced mesenteric vasoconstriction in rats is mediated by neuropeptide Y Y1 receptors Am J Physiol 1996 270 H796 800 8779858
Lee EW Neuropeptide Y induces ischemic angiogenesis and restores function of ischemic skeletal muscles J Clin Invest 2003 111 1853 62 12813021 10.1172/JCI200316929
Soker S Takashima S Miao HQ Neufeld G Klagsbrun M Neuropilin-1 is expressed by endothelial and tumor cells as an isoform-specific receptor for vascular endothelial growth factor Cell 1998 92 735 45 9529250 10.1016/S0092-8674(00)81402-6
Hampe W Rezgaoui M Hermans-Borgmeyer I Schaller HC The genes for the human VPS10 domain-containing receptors are large and contain many small exons Hum Genet 2001 08 529 36 11499680 10.1007/s004390100504
Lintzel J Franke I Riedel IB Schaller HC Hampe W Characterization of the VPS10 domain of SorLA/LR11 as binding site for the neuropeptide HA Biol Chem 2002 383 1727 33 12530537 10.1515/BC.2002.193
Vincent JP Mazella J Kitabgi P Neurotensin and neurotensin receptors Trends Pharmacol Sci 1999 20 302 9 10390649 10.1016/S0165-6147(99)01357-7
Levine L Lucci JA Pazdrak B Cheng JZ Guo YS Townsend CM Hellmich MR Bombesin stimulates nuclear factor kappa B activation and expression of proangiogenic factors in prostate cancer cells Cancer Res 2003 63 3495 502 12839933
Hoos A Urist MJ Stojadinovic A Mastorides S Dudas ME Leung DH Kuo D Brennan MF Lewis JJ Cordon-Cardo C Validation of tissue microarrays for immunohistochemical profiling of cancer specimens using the example of human fibroblastic tumors Am J Pathol 2001 158 1245 51 11290542
Petricoin EF Use of proteomic patterns in serum to identify ovarian cancer Lancet 2002 359 572 7 11867112 10.1016/S0140-6736(02)07746-2
Swiss-Prot
UCSC Genome
GPCR Database
NIH Blast
MySQL
Apache
Maanova Tools
FatiGO
| 15836779 | PMC1112590 | CC BY | 2021-01-04 16:39:53 | no | BMC Genomics. 2005 Apr 18; 6:55 | utf-8 | BMC Genomics | 2,005 | 10.1186/1471-2164-6-55 | oa_comm |
==== Front
BMC GenomicsBMC Genomics1471-2164BioMed Central London 1471-2164-6-581585422310.1186/1471-2164-6-58Research ArticleInvestigating hookworm genomes by comparative analysis of two Ancylostoma species Mitreva Makedonka [email protected] James P [email protected] Prema [email protected] John [email protected] John [email protected] Mike [email protected] Todd [email protected] Jian [email protected] Jason E [email protected] Wadim [email protected] Sandra W [email protected] Robert H [email protected] Richard K [email protected] Genome Sequencing Center, Department of Genetics, Washington University School of Medicine, St. Louis, MO 63108, USA2 Divergence Inc., St. Louis, MO 63141, USA3 College of Veterinary Medicine, Department of Molecular Biomedical Sciences, North Carolina State University, Raleigh, NC 27606, USA4 Department of Microbiology and Tropical Medicine, George Washington University Medical Center, Washington, DC 20037, USA5 Department of Molecular Genetics and Microbiology, Duke University, Durham, NC 27710, USA6 Department of Infectious Diseases, Microbiology and Parasitology, Faculty of Veterinary Medicine, Warsaw Agricultural University, Warszawa, Poland7 School of Biology, University of Leeds, LEEDS LS2 9JT, UK8 Department of Genome Sciences, University of Washington, Seattle, WA 98195, USA2005 26 4 2005 6 58 58 14 12 2004 26 4 2005 Copyright © 2005 Mitreva et al; licensee BioMed Central Ltd.2005Mitreva 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
Hookworms, infecting over one billion people, are the mostly closely related major human parasites to the model nematode Caenorhabditis elegans. Applying genomics techniques to these species, we analyzed 3,840 and 3,149 genes from Ancylostoma caninum and A. ceylanicum.
Results
Transcripts originated from libraries representing infective L3 larva, stimulated L3, arrested L3, and adults. Most genes are represented in single stages including abundant transcripts like hsp-20 in infective L3 and vit-3 in adults. Over 80% of the genes have homologs in C. elegans, and nearly 30% of these were with observable RNA interference phenotypes. Homologies were identified to nematode-specific and clade V specific gene families. To study the evolution of hookworm genes, 574 A. caninum / A. ceylanicum orthologs were identified, all of which were found to be under purifying selection with distribution ratios of nonsynonymous to synonymous amino acid substitutions similar to that reported for C. elegans / C. briggsae orthologs. The phylogenetic distance between A. caninum and A. ceylanicum is almost identical to that for C. elegans / C. briggsae.
Conclusion
The genes discovered should substantially accelerate research toward better understanding of the parasites' basic biology as well as new therapies including vaccines and novel anthelmintics.
==== Body
Background
Comparative sequence analysis is an approach proven to aid in recognition of genes and defining of their function, especially when comparing genomes of close evolutionary distance. In addition, when partial genomes are placed in a context of a well-studied and fully sequenced model organism they can greatly facilitate the understanding of the less studied organisms' biology.
Hookworms are blood-feeding nematodes that infect one billion people causing iron deficiency anemia and retarded physical and cognitive development in children [1]. The two major species infecting humans are Necator americanus and Ancylostoma duodenale. The closely related hookworm species of canids, Ancylostoma caninum, and canines and felines, A. ceylanicum, are minor parasites of humans, but are important as laboratory models for hookworm infection and disease. Other hookworms infect raccoons, sheep, seals and a variety of other mammals [2]
Adult (Ad) hookworms inhabit the small intestine and produce eggs that pass in the feces and hatch in the soil. The first stage larva (L1) feeds on bacteria and molts twice to form the non-feeding, infective third stage (iL3). iL3 enters the host by penetrating the skin, or orally in the case of Ancylostoma species, molts twice, and matures to Ad in the small intestine. A. duodenale and A. caninum L3s can also infect a host, temporarily abort maturation and enter an arrested state (hypobiosis) within the host's somatic tissues [3], reactivating in response to host physiological changes such as pregnancy [4].
Current hookworm control strategies are limited to de-worming of infected people using anthelmintic drugs. However, rapid re-infection in endemic areas and the lack of sterile immunity necessitates repeated treatments and can in turn result in resistance. Additionally, tissue-arrested stages are relatively resilient to the effects of anthelmintics [5]. The Human Hookworm Vaccine Initiative is beginning clinical trials of a larval hookworm antigen, ASP-2, from N. americanus, as a vaccine antigen [6]. There is a critical need for further research to identify new vaccine and drug targets as well as to better understand hookworm biology. Lack of sequence information has been a major hindrance to hookworm molecular studies. High throughput sequencing of expressed sequence tags (ESTs; sequences derived from randomly selected cDNA clones) has proven a cost-effective tool for discovering genes [7]. Because the hookworm superfamily (Ancylostomatoidea) falls within nematode Clade V [8,9], which also contains the well-studied model nematode Caenorhabditis elegans [10], predictions may be made and tested based on their close relatedness. Previous genome-based characterization of hookworms has been limited to sampling of few hundred ESTs [11] and molecular studies of individual genes of interest (eg. [12]; reviewed in [13]). EST approaches have also been initiated for other Strongylid parasites including Haemonchus contortus [14,15] and Nippostrongylus brasiliensis [16].
In this report we describe the comparative analyses of almost 20,000 ESTs from 7 different cDNA libraries representing pre-parasitic and parasitic larval through adult stages of the hookworms A. caninum and A. ceylanicum. The dataset defined nearly 7,000 hookworm genes, including a number of putative developmentally expressed genes and candidates for further study as drug target or vaccine components.
Results
Nearly 20,000 Ancylostoma derived ESTs were submitted to GenBank between 1999 and 2003 [see Additional file 1]. For simplicity, the results and analysis described are presented in the same order beginning with A. caninum and followed by A. ceylanicum, except where specified.
EST acquisition and NemaGene organization
ESTs originated from 7 cDNA libraries, representing three and two life-cycle stages respectively (Table 1). Clustering, implemented to reduce data redundancy and improve sequence quality and length, grouped ESTs into contigs which were further organized into clusters (Table 1), providing a non-redundant catalog of represented genes. ESTs within a contig derive from nearly identical transcripts while contigs within a cluster may arise from splice isoforms, alleles, or closely related paralogs [17]. Fifty-one potentially chimeric ESTs were discarded. Clusters ranged in size from a single EST to 203 and 323 for A. caninum and A. ceylanicum respectively (Figure 1). Most clusters for each species (72% and 55%) have ten or fewer ESTs. GC content for coding sequences was similar in the two species (44% and 48%) and consistent with other Clade V nematodes like C. elegans and C. briggsae [18].
Table 1 Ancylostoma libraries sequenced and their properties
Nematode stage (vector or SL1 based) ESTs Submitted Nucleotides (million) Mean read length (bp) StDev
A. caninum Infective L3 (UniZap) 5,679 2,632 358 109
Tissue arrested L3 (SL1) 820 0,318 344 151
Serum stimulated L3 (pAMP) 2,832 1,273 441 150
Overall 9,331 4,223 381 137
Contigs – 5,484 (4,020 clusters) 502 168
A. ceylanicum Infective L3 (λZAP II) 3,359 2,021 500 127
Infective L3 (SL1) 3,306 1,550 400 143
Adult (M1 SL1) 629 0,319 460 134
Adult (M2 SL1) 480 0,255 467 131
Adult (λZAP II) 2,817 1,646 500 139
Overall 10,591 5,791 465 135
Contigs – 4,953 (3,369 clusters) 572 179
Figure 1 Ancylostoma NemaGene v2.0 clustering showing the distribution of ESTs by cluster size. For example, there are three A. caninum cluster of size 22 containing a sum of 66 ESTs and there were eight A. ceylanicum clusters of size 22 containing a sum of 176 ESTs. Cluster size (x-axis) is shown to scale for 1–75 members, with the size of larger clusters indicated.
The number of clusters may overestimate gene discovery, as one gene may be represented by multiple non-overlapping clusters (fragmentation). By using C. elegans as a reference genome (19,522 genes; [10]) and discounting for fragmentation calculated as 4.5% and 6.5 % respectively [17], the estimated gene numbers were reduced to 3,840 for A. caninum and 3,149 for A. ceylanicum giving a gene discovery rate of 41% (3,840 × 100/9,283) and 30% (3,149 × 100/10,588). These numbers also indicate 20% and 16% representation of all genes for each species respectively. The number of genes in common for more stage-specific Ancylostoma libraries analysed was as low as 9% and 11% (Figure 2). This may reflect the EST sample size or stage-specific expression, as will be discussed. In either case, the results clearly show the advantage gained for gene discovery in Ancylostoma by including diverse life stages in the analysis.
Figure 2 Venn diagram of A. caninum (A) and A. ceylanicum (B) clusters, based on stage of origin of each cluster's EST members. The majority of clusters are represented by only one stage in this investigation, though greater depth of sampling would likely increase representation by multiple stages.
Functional classification based on Gene Ontology and KEGG assignments
Thirty-four percent of A. caninum and 54% of A. ceylanicum clusters align to InterPro domains and 21% and 36% map to Gene Ontology (GO) respectively. Following this same pattern, A. caninum also had fewer BLAST matches (see below). Seven of the ten most abundantly represented InterPro domains were common to both species (Table 2). GO representation is shown by biological process, cellular component, and molecular function (Table 3). Among the most common GO categories are protein metabolism (GO:0019538) and catalytic activity (GO:0003824).
Table 2 Most abundantly represented protein domains in A. caninum and A. ceylanicum datasets
Species InterPro ID Clusters # Domain descriptor
A. caninum
IPR001230 141 Prenyl group, CAAX box, attachment site
IPR001687 66 ATP/GTP-binding site motif A (P-loop)
IPR000694 62 Proline-rich region
IPR001472 51 Bipartite nuclear localization signal
IPR000345 29 Cytochrome c heme-binding site
IPR001283 25 Allergen V5/Tpx-1 related
IPR002048 24 Calcium-binding EF-hand
IPR000504 21 RNA-binding region RNP-1 (RNA recognition motif)
IPR000719 19 Protein kinase
IPR007087 17 Zn-finger, C2H2 type
A. ceylanicum
IPR000694 214 Proline-rich region
IPR001230 153 Prenyl group, CAAX box, attachment site
IPR001687 125 ATP/GTP-binding site motif A (P-loop)
IPR000345 50 Cytochrome c heme-binding site
IPR006209 48 EGF-like domain
IPR000504 37 RNA-binding region RNP-1 (RNA recognition motif)
IPR001283 34 Allergen V5/Tpx-1 related
IPR001472 32 Bipartite nuclear localization signal
IPR000169 32 Eukaryotic thiol (cysteine) protease
IPR001534 27 Transthyretin-like
Table 3 GO mappings for A. caninum and A. ceylanicum clusters
A. caninum A. ceylanicum
Categories and subcategories Representation % Representation of total Representation % Representation of total
biological process
cellular process 192 4.80 238 4.36
cell communication 62 1.55 66 1.21
cell motility 1 0.03 0 0.00
cell death 1 0.03 1 0.02
cell growth and/or maintaince 140 3.50 180 3.30
transport 119 2.98 153 2.80
cell organization and biogenesis 21 0.53 32 0.59
cell proliferation 4 0.10 6 0.11
cell homeostasis 1 0.03 4 0.07
physiological process 579 14.48 785 14.38
response to endogenous stimulus 1 0.03 5 0.09
response to external stimulus 12 0.30 16 0.29
response to stress 8 0.20 14 0.26
death 1 0.03 1 0.02
metabolism 466 11.65 653 11.96
hemostasis 1 0.03 0 0.00
homeostasis 3 0.08 4 0.07
secretion 0 0.00 1 0.02
development 11 0.28 18 0.33
cellular component
cell 327 8.18 385 7.05
intracellular 212 5.30 277 5.08
cytoplasm 166 4.15 181 3.32
nucleus 45 1.13 84 1.54
ribonucleoprotein complex 102 2.55 102 1.87
respiratory chain complex 4 0.10 5 0.09
chromosome 8 0.20 13 0.24
thylakoid 0 0.00 1 0.02
membrane 146 3.65 146 2.67
extracellular 33 0.83 50 0.92
Unlocalized 1 0.03 7 0.13
molecular function
binding 323 8.08 521 9.55
carbohydrate binding 6 0.15 20 0.37
lipid binding 6 0.15 12 0.22
metal ion binding 56 1.40 72 1.32
nucleic acid binding 109 2.73 201 3.68
nucleotide binding 133 3.33 214 3.92
protein binding 11 0.28 22 0.40
apoptosis regulator activity 1 0.03 1 0.02
chaperone activity 5 0.13 10 0.18
cell adhesion molecule activity 2 0.05 1 0.02
catalytic activity 293 7.33 445 8.15
enzyme regulator activity 24 0.60 35 0.64
molecular function unknown 38 0.95 46 0.84
motor activity 3 0.08 12 0.22
signal transducer activity 55 1.38 61 1.12
structural molecule activity 107 2.68 127 2.33
transcription regulator activity 19 0.48 33 0.60
translation regulator activity 14 0.35 17 0.31
transporter activity 128 3.20 180 3.30
Within Ancylostoma spp. clusters that had extracellular mappings, 70% and 56% respectively were in the category of Allergen V5/Tpx-1 proteins (IPR001283) related to the secreted venom proteins from hymenopteran insects (Table 2). The Ancylostoma secreted proteins (ASPs) belong to this large gene family [19], members of which been shown to play roles in host-parasite interactions for both mammalian [20,21] and plant parasitic nematodes [22], and to induce protective responses [6]. ASP-1 is one of the major proteins secreted by serum-stimulated A. caninum iL3 [12]. In addition, four A. ceylanicum clusters were classified in extracellular matrix (GO:0005578) as tissue inhibitor of metalloprotease (TIMP) domain proteins. A TIMP homolog is reported as the most abundant protein in adult hookworm excretory/secretory products and may inhibit host metalloproteases [23].
Ten % and 15% unique clusters for A. caninum and A. ceylanicum respectively, mapped to 89 metabolic pathways grouped in 11 categories (Table 4). Complete listings and graphical representations of the KEGG mappings are available at . Pathways well represented by both species include glycolysis/gluconeogenesis, citrate cycle, oxidative phosphorylation and fatty acid biosynthesis and metabolism. KEGG analysis (Table 4) also suggests specific biochemical differences among Ancylostoma stages. For example, while serum stimulated L3-specific clusters make up to 27% of all AC clusters, they account for 40% of all KEGG pathway mappings. In contrast, iL3-specific clusters that account for 55% of all AC clusters make-up only 38% of KEGG pathway mappings. It is unclear whether the predominance of enzyme mappings from the ssL3 stage versus iL3 stage is indicative of greater metabolic activity, greater metabolic complexity, differences in library construction methods, or other differences.
Table 4 Kegg Biochemical pathway mappings for A. caninum and A. ceylanicum clusters
AC Clusters per library AE Clusters per library Total # of enzymes in KEGG
KEGG CATEGORY REPRESENTEDa Clb iL3 taL3 ssL3 Mixed Enzc Clb iL3 Ad Mixed Enzc
1. Carbohydrate metabolism
1.1 Glycolysis / Gluconeogenesis 23 13 0 5 5 22 25 10 11 4 23 40
1.2 Citrate cycle (TCA cycle) 17 8 0 4 5 16 15 4 8 3 15 23
1.3 Pentose phosphate pathway 9 3 0 4 2 9 12 5 3 4 8 34
1.4 Pentose and glucuronate interconversions 8 3 0 3 2 9 9 5 4 0 8 53
1.5 Fructose and mannose metabolism 14 6 0 4 4 15 20 7 11 2 15 61
1.6 Galactose metabolism 10 4 0 4 2 8 12 6 5 1 12 37
1.7 Ascorbate and aldarate metabolism 7 4 0 2 1 4 5 5 0 0 4 29
1.8 Pyruvate metabolism 25 11 0 11 3 23 27 8 17 2 26 67
1.9 Glyoxylate and dicarboxylate metabolism 13 6 0 5 2 14 9 1 5 3 17 58
1.10 Propanoate metabolism 22 7 1 10 4 20 25 11 8 6 22 46
1.11 Butanoate metabolism 22 9 1 7 5 23 29 14 14 1 26 52
1.12 C5-Branched dibasic acid metabolism 4 3 0 1 0 2 2 1 0 1 1 20
1.13 Inositol metabolism 6 2 0 1 3 4 7 2 3 2 4 5
2. Energy metabolism
2.1 Oxidative phosphorylation 24 7 0 6 11 11 33 10 14 9 13 14
2.2 ATP synthesis 8 2 0 3 3 1 11 4 3 4 1 1
2.4 Carbon fixation 11 3 0 3 5 11 11 3 5 3 13 23
2.5 Reductive carboxylate cycle (CO2 fixation) 12 7 0 1 4 8 9 2 4 3 7 13
2.6 Methane metabolism 6 4 0 0 2 5 6 0 4 2 6 26
2.7 Nitrogen metabolism 11 2 0 5 4 14 12 5 5 2 15 64
2.8 Sulfur metabolism 5 1 0 1 3 9 6 3 1 2 9 30
3. Lipid metabolism
3.1 Fatty acid biosynthesis (path 1) 6 2 0 3 1 11 7 3 3 1 6 14
3.2 Fatty acid biosynthesis (path 2) 8 2 0 5 1 6 6 3 2 1 5 8
3.3 Fatty acid metabolism 14 6 1 6 1 17 21 13 7 1 16 28
3.4 Synthesis and degradation of ketone bodies 2 0 0 1 1 2 8 4 3 1 3 6
3.5 Sterol biosynthesis 4 1 1 2 0 4 4 2 2 0 9 35
3.6 Bile acid biosynthesis 11 7 1 2 1 11 11 6 4 1 10 27
3.8 Androgen and estrogen metabolism 7 5 0 2 0 9 7 4 3 0 8 26
4. Nucleotide metabolism
4.1 Purine metabolism 27 11 0 11 5 28 32 14 11 7 32 99
4.2 Pyrimidine metabolism 16 8 1 5 2 15 22 9 12 1 22 61
4.3 Nucleotide sugars metabolism 6 4 0 0 2 3 4 2 2 0 4 30
5. Amino acid metabolism
5.1 Glutamate metabolism 11 3 0 5 3 14 16 8 7 1 18 36
5.2 Alanine and aspartate metabolism 14 1 0 8 5 15 14 5 6 3 15 38
5.3 Glycine, serine and threonine metabolism 19 7 0 9 3 14 21 8 10 3 24 56
5.4 Methionine metabolism 6 1 0 4 1 9 6 5 1 0 8 24
5.5 Cysteine metabolism 8 2 0 2 4 11 7 5 2 0 9 23
5.6 Valine, leucine and isoleucine degradation 16 3 1 8 4 16 22 11 6 5 18 32
5.7 Valine, leucine and isoleucine biosynthesis 7 1 0 4 2 7 9 6 2 1 8 15
5.8 Lysine biosynthesis 11 1 0 6 4 10 9 4 3 2 8 31
5.9 Lysine degradation 19 8 0 8 3 14 19 12 6 1 17 47
5.10 Arginine and proline metabolism 18 4 0 8 6 20 22 11 7 4 20 71
5.11 Histidine metabolism 10 4 0 4 2 8 10 5 4 1 8 39
5.12 Tyrosine metabolism 18 8 0 7 3 19 19 11 5 3 19 67
5.13 Phenylalanine metabolism 13 5 0 3 5 11 14 7 6 1 12 40
5.14 Tryptophan metabolism 17 8 0 8 1 15 22 16 4 2 18 61
5.15 Phenylalanine, tyrosine and tryptophan biosynthesis 5 1 0 2 2 6 4 2 0 2 7 31
5.16 Urea cycle and metabolism of amino groups 10 1 0 5 4 14 10 2 6 2 11 35
6. Metabolism of other amino acids
6.1 beta-Alanine metabolism 14 4 1 7 2 13 11 8 1 2 10 32
6.2 Taurine and hypotaurine metabolism 1 0 0 0 1 1 2 2 0 0 3 14
6.3 Aminophosphonate metabolism 4 0 0 3 1 3 5 2 3 0 5 15
6.4 Selenoamino acid metabolism 7 0 0 4 3 12 12 6 4 2 15 22
6.5 Cyanoamino acid metabolism 2 1 0 1 0 1 7 5 1 1 6 19
6.6 D-Glutamine and D-glutamate metabolism 2 0 0 1 1 2 2 1 0 1 2 12
6.7 D-Arginine and D-ornithine metabolism 3 1 0 0 2 2 3 0 2 1 2 10
6.9 Glutathione metabolism 5 1 0 0 4 4 9 5 3 1 6 27
7. Metabolism of complex carbohydrates
7.1 Starch and sucrose metabolism 18 2 0 13 3 18 20 13 6 1 20 75
7.2 N-Glycans biosynthesis 7 4 0 1 2 7 7 2 4 1 9 27
7.3 O-Glycans biosynthesis 3 1 0 1 1 2 6 5 1 0 3 8
7.5 Aminosugars metabolism 6 3 0 2 1 6 10 5 5 0 10 39
7.8 Glycosaminoglycan degradation 1 1 0 0 0 1 1 1 0 0 1 13
7.9 Chondroitin / Heparan sulfate biosynthesis 5 3 0 1 1 4 6 2 4 0 4 18
7.10 Keratan sulfate biosynthesis 1 0 0 1 0 1 2 1 1 0 1 6
8. Metabolism og complex lipids
8.1 Glycerolipid metabolism 24 10 0 9 5 22 25 10 13 2 22 80
8.3 Inositol phosphate metabolism 8 4 0 4 0 4 8 5 2 1 3 25
8.4 Sphingophospholipid biosynthesis 1 1 0 0 0 1 2 1 1 0 2 8
8.5 Phospholipid degradation 3 2 0 0 1 3 1 1 0 0 1 11
8.6 Sphingoglycolipid metabolism 11 1 1 9 0 7 10 8 2 0 4 20
8.9 Globoside metabolism 2 1 0 1 0 2 2 1 1 0 1 12
8.11 Prostaglandin and leukotriene metabolism 8 2 0 1 5 8 7 4 3 0 6 19
9. Metabolism of cofactors and vitamins
9.2 Riboflavin metabolism 4 1 0 3 0 2 2 2 0 0 2 13
9.3 Vitamin B6 metabolism 6 4 0 1 1 3 6 4 1 1 5 23
9.4 Nicotinate and nicotinamide metabolism 11 2 0 7 2 7 15 8 7 0 7 32
9.5 Pantothenate and CoA biosynthesis 8 2 0 4 2 9 8 6 2 0 7 27
9.7 Folate biosynthesis 5 2 1 0 2 5 6 1 4 1 5 25
9.8 One carbon pool by folate 5 3 0 2 0 10 7 3 3 1 8 24
9.10 Porphyrin and chlorophyll metabolism 18 4 0 10 4 12 27 15 8 4 13 56
9.11 Ubiquinone biosynthesis 19 10 0 5 4 13 28 11 14 3 14 22
10. Biosynthesis of secondary metabolites
10.1 Terpenoid biosynthesis 0 0 0 0 0 0 2 0 2 0 4 12
10.3 Flavonoids, stilbene and lignin biosynthesis 6 3 0 1 2 7 8 5 3 0 7 39
10.4 Alkaloid biosynthesis I 5 2 0 3 0 6 3 3 0 0 5 36
10.8 Streptomycin biosynthesis 2 0 0 2 0 3 4 2 1 1 4 14
10.9 Erythromycin biosynthesis 2 0 0 2 0 3 3 2 1 0 3 6
11. Biodegradation of xenobiotics
11.4 Nitrobenzene degradation 4 1 0 3 0 5 4 2 1 1 3 17
11.9 Tetrachloroethene degradation 6 4 0 1 1 3 2 2 0 0 3 5
11.10 Styrene degradation 4 2 0 2 0 3 6 5 0 1 5 18
11.1 gamma-Hexachlorocyclohexane degradation 6 3 0 3 0 5 5 3 2 0 4 12
11.1 Fluorene degradation 3 1 0 2 0 4 2 2 0 0 2 13
11.2 Benzoate degradation via CoA ligation 22 8 1 10 3 18 25 14 10 1 18 38
11.2 Benzoate degradation via hydroxylation 7 2 0 5 0 7 5 4 1 0 5 45
aA. caninum – 839 multiple and 786 unique mappings; A. ceylanicum – 957 multiple and 840 unique mappings. b Cluster, c Enzymes
Homologs in other organisms, orthologs within Ancylostoma spp. and estimates of selective pressure
Within A. ceylanicum clusters, 83% had homology to proteins from other organisms as compared to only 66% for A. caninum (Figure 3). To investigate why contigs from closely related species would show a difference in identified homologies, we compared sequence lengths and the open reading frame (ORF) lengths of contigs with and without homologies in both species. EST lengths and contig lengths, respectively, were shorter for A. caninum (410 and 549 nucleotides) than for A. ceylanicum (490 and 637 nucleotides). The differences were even more striking for ORFs (Figure 4). Hence, A. caninum contigs very likely identify fewer homologs because these sequences are shorter, contain smaller ORFs, and probably include more 3' UTR versus the superior quality dataset from A. ceylanicum. Most likely, differences in library construction and sampling rather than intrinsic differences between the species explain this discrepancy. Accounting for such differences is important as it keeps analysis focused upon interesting features of the dataset related to the organisms' biology rather than artifactual differences arising from data collection.
Figure 3 Venn diagram showing distribution of A. caninum (A) and A. ceylanicum (B) cluster BLAST matches by database. Amino acid level homologies (≥ e-05) were identified to non-Ancylostoma sequences for 65.8% (2,646/4,020) of A. caninum and 83.1% (2,801/3,369) of A. ceylanicum clusters. Databases used are: for C. elegans, Wormpep v.97 and mitochondrial protein sequences; for other nematodes, all GenBank nucleotide data for nematodes except C. elegans and Ancylostoma; for non-nematodes, nrGenBank (3/20/2003) with all nematode sequences removed.
Figure 4 Distribution of A. caninum and A. ceylanicum contigs with and without database amino acid level homology by size of the longest predicted open reading frame (ORF).
The distribution of identified homologs (Figure 3) was consistent with earlier observations [17]. Besides C. elegans, one of the more informative nematode datasets for this study is a collection of 4,780 ESTs from the human hookworm Necator americanus to which homologies were commonly found (42% and 38% respectively). Within Ancylostoma itself homologies were common with 34% of total A. caninum clusters matching the A. ceylanicum dataset and 44% of total A. ceylanicum clusters matching A. caninum. Searching for putative orthologs between all A. caninum and A. ceylanicum contigs resulted in 1,304 reciprocal best TBLASTX hits. The ortholog pair members were very similar in GC composition (46% and 47%) and the average length of alignment was 327 bp. All ortholog pairs (574) were under purifying selection (dN/dS < 1; Figure 5) and the average dS was 0.65 ± 0.83 and dN was 0.11 ± 0.2. The average dN/dS ratio (~0.17) is higher than that reported for C. briggsae/C. elegans (~0.06; [18]), and closer to that for mouse/human (0.115; [24]), indicating that the levels of purifying selection are somewhat different. In addition, to examine if this purifying selection is more frequently detected in genes with essential function, we cross-referenced the C. elegans genes matched by Ancylostoma orthologs with a list of C. elegans genes with available RNA interference (RNAi) information (; eg. [25]). Of the 67% of the orthologous genes matching C. elegans genes with available RNAi data, 45% had an observable phenotype after transcript knock-down. A vast majority of the observed phenotypes were severe (88% sterility and embryonic lethality). Ancylostoma orthologs matching C. elegans genes that had phenotypes showed a somewhat lower dN/dS ratio than those matching genes that remained wild type after RNAi treatment, though the difference was not statistically significant at P < 0.05 (0.09 vs. 0.14; sign. diff. at P < 0.2).
Figure 5 Distribution of dN/dS ratios among Ancylostoma ortholog pairs. dN and dS are the rates of nonsynonymous and synonymous amino acid substitutions, respectively.
In a 4-way comparison of orthologs in C. elegans, C. briggsae, A. caninum, and A. ceylanicum, the phylogenetic distance between A. caninum and A. ceylanicum is almost identical to that between C. elegans and C. briggsae and the distance between the two genera is just over four times the within genera distance (Figure 6). Average branch lengths for the set of 452 orthologous proteins did not show a significant difference in relative rate of molecular change. Maximum likelihood trees [26] were constructed for each 4-way ortholog and relative branch lengths compared between Ancylostoma spp. for both the protein and nucleotide sequences. For protein sequences, 109 trees had equal branch lengths for the hookworm species while 175 trees had longer A. caninum branches and 168 had longer A. ceylanicum branches. To look for differences between genes in the Ancylostoma species, we constructed a distribution of branch length differences between each of the sister species pairs in our tree. Because some genes may be rapidly evolving in all nematode lineages we evaluated a subset of the trees where the difference in branch lengths between C. briggsae and C. elegans were less than one standard deviation from the mean but which had significantly different branch lengths in the two Ancylostoma species. This final dataset had 23 genes with significantly longer branch lengths in A. ceylanicum and 16 in A. caninum (1 SD, P < 0.05). However the set did not show any significant bias towards either species (p < 0.34, sign test). This suggests there is no significant rate difference in protein evolution between the two hookworms, although some of these genes are relatively rapidly evolving. Repeating the analysis for nucleotide sequences we find marginally significant differences (p < 0.08; sign test).
Figure 6 Relative distance based upon protein maximum likelihood. A. ceylanicum to A. cananum distance is similar to C. elegans to C. briggsae distance. Ancylostoma to Caenorhabditis distance for any species is 4.3X the Ancylostoma to Ancylostoma distance and 4.1X the Caenorhabditis to Caenorhabditis distance. The length of each line segment is proportional to the calculated branch length between the species.
Using the C. elegans genome to interpret hookworm sequences
As expected, the C. elegans genome provides the best source of information for interpreting hookworm sequences as a majority of A. caninum and A. ceylanicum clusters with BLAST homologies outside Ancylostoma had C. elegans homologs (Figure 3; 25 most conserved nematode genes between each Ancylostoma species and C. elegans are available online; [see Additional file 2]). Furthermore, C. elegans orthologs of hookworm genes with available RNAi or other data provide information that may be relevant to understanding the role of the parasite genes. Of all the Ancylostoma clusters with C. elegans homology, 97% and 92% matched C. elegans genes with available RNA interference knock-down information , and in turn 33% and 37% of these C. elegans genes produce RNAi phenotypes (versus a rate of only ~15% phenotypes for all C. elegans genes). Phenotype classification [see Additional file 3] showed that C. elegans genes with expressed Ancylostoma homologs were somewhat more likely to have severe phenotypes [17]. Hence, certain genes in the Ancylostoma datasets encode proteins if disabled may disrupt survival of the parasite. Some examples include abundant clusters (AC00023.cl, AE01104.cl; Table 5 and Table 6). A group of particular interest is proteins that are required for nematode survival and lack strong homologies outside of the phyla (nematode-specific), since these targets could provide for nematode control without toxicities to the host or other non-target organisms. Of the Ancylostoma nematode-specific clusters (Figure 3), 85 and 91 respectively had C. elegans matches with RNAi phenotypes. Among these, AC04398.cl and AE00474.cl matched hypothetical protein F42A8.1 (1e-65, 2e-76 respectively), a gene without a mammalian homolog yet likely involved in multiple developmental processes based on observed mutant phenotypes [25]. Homologs are found in at least 13 nematode species to date including free-living species (3), and parasites of mammals (8) and plants (1). Further analysis will identify additional genes which warrant detailed investigation.
Table 5 The most abundantly represented transcripts in the A. caninum cDNA libraries
non-redundant GenBank
A. caninum cluster id ESTs per cluster ESTs from library Accession C. elegans gene
iL3 taL3 ssL3 Best identity descriptor SW / TRa E-value Wormpep97b
AC00932.cl 203 186 0 17 Ancylostoma duodenale cytochrome oxidase subunit I CAD10437 2e – 304 C06G4.2b
AC00471.cl 120 2 0 118 C. elegans CMP-sialic acid transporter O02345 2e – 83 ZK896.9b
AC00048.cl 114 100 11 3 C. elegans hsp-12.6 alpha-B-crystallin Q20165 6e – 40 F38E11.2b
AC01032.cl 104 104 0 0 Ancylostoma duodenale cytochrome oxidase subunit II AAL50814 1e – 09 -
AC01031.cl 93 90 0 3 Ancylostoma duodenale cytochrome oxidase subunit III CAD10435 2e – 143 -
AC00807.cl 69 4 2 63 Necator americanus ancylostoma secreted protein 1 precursor AAD13340 3e – 28 F33A8.2
AC00205.cl 65 51 1 13 Ancylostoma duodenale COX2, cytochrome c oxidase subunit II NP_579953 4e – 125 F26E4.12
AC00967.cl 54 9 0 45 Ancylostoma ceylanicum cathepsin D-like aspartic protease AAO22152 8e – 174 R12H7.2
AC00134.cl 44 44 0 0 C. elegans putative protein, nematode specific NP_497272 9e – 53 K02F3.9b
AC01029.cl 39 37 2 0 C. elegans stress associated endoplasmic reticulum protein NP_510604 9e – 37 F59F4.2b
AC00137.cl 38 37 0 1 C. elegans RNA recognition motif CAB03222 4e – 35 R06C1.4b
AC00023.cl 38 36 0 2 C. elegans rpl-2 Ribosomal Proteins L2 Q9XVF7 3e – 148 B0250.1b
AC00060.cl 36 36 0 0 Ancylostoma caninum secreted protein ASP-2 precursor AAC35986 2e – 134 F11C7.3b
AC01400.cl 32 1 0 31 C. elegans ham-2 zinc finger protein NP_508781 1e – 20 C07A12.1b
AC00976.cl 32 29 0 3 Tetrahymena pigmentosa metallothionein MT-2 AAL87687 6e – 12 K11G9.6
AC00193.cl 31 7 0 24 Pisum sativum putative senescence-associated protein BAB33421 2e – 61 F58H1.7
AC00931.cl 29 6 0 23 novel - - -
AC00079.cl 28 25 2 1 Ostertagia ostertagi unknown protein AAC08432 8e – 06 -
AC00980.cl 27 11 0 16 C. elegans Glycerol kinase AAA79749 8e – 70 R11F4.1b
AC00971.cl 26 9 0 17 C. elegans rpl-1 Ribosomal Protein Large subunit NP_491061 5e – 116 Y71F9AL.13ab
AC01023.cl 25 22 2 1 Ostertagia ostertagi putative ES protein F7 CAD20464 9e – 87 F02A9.2
AC00913.cl 25 6 17 2 C. elegans ribosomal protein L37 O62388 2e – 50 W01D2.1b
AC02930.cl 24 0 0 24 C. elegans calponin-like protein NP_504712 4e – 119 T25F10.6b
AC00252.cl 24 12 12 0 C. elegans rpl-43 CAB54440 4e – 49 Y48B6A.2b
AC01020.cl 23 20 3 0 C. elegans rps-15 Ribosomal Protein Small subunit RPS-15 NP_492384 1e – 88 F36A2.6b
a SW/TR is Swiss-prot and TrEMBL Proteinknowledgebase .
b C. elegans homolog has higher probability match than the best GenBank descriptor.
Table 6 The most abundantly represented transcripts in the A. ceylanicum cDNA libraries
non-redundant GenBank
A. ceylanicum cluster id ESTs Per cluster ESTs from Library Accession C. elegans gene
iL3 Ad Best identity descriptor SW / TRa E-value Wormpep97b
AE00908.cl 323 320 3 C. elegans stress associated endoplasmic reticulum protein NP_510604 9e – 37 F59F4.2b
AE00787.cl 205 205 0 C. elegans hsp-12.6 alpha-B-crystallin Q20165 7e – 39 F38E11.2b
AE01104.cl 155 155 0 C. elegans microsomal signal peptidase 25 kDa subunit Q9XWW1 8e – 80 Y37D8A.10b
AE00463.cl 119 118 1 C. elegans dlc-1 dynein light chain (10.3 kD) NP_498422 9e – 56 T26A5.9b
AE00121.cl 110 0 110 C. elegans vit-3 Vitellogenin 3 precursor NP_508613 7e – 121 F59D8.1b
AE00890.cl 84 84 0 C. elegans spp-4 SaPosin-like Protein family NP_509237 5e – 16 T08A9.8b
AE00065.cl 84 84 0 C. elegans putative endoplasmic reticulum protein NP_508656 1e – 35 F47B7.1b
AE00360.cl 74 74 0 novel - - -
AE00048.cl 72 72 0 C. elegans rpl-29 60S ribosomal protein L29 NP_502671 6e – 25 B0513.3b
AE00003.cl 70 3 67 novel - - -
AE01410.cl 63 60 3 Ostertagia ostertagi putative ES protein F7 CAD20464 3e – 86 F02A9.2
AE00056.cl 62 56 6 C. elegans hypothetical protein AAK77617 5e – 39 M01H9.3ab
AE00464.cl 61 61 0 novel - - -
AE00227.cl 59 0 59 Zea mays extensin-like protein S49915 2e – 31 ZK84.1
AE00746.cl 54 0 54 C. elegans protein contains chitin binding peritrophin-A domain AAA19083 3e – 54 B0280.5b
AE00750.cl 50 36 14 C. elegans far-7 fatty Acid/Retinol binding protein NP_493708 7e – 45 K01A2.2ab
AE00072.cl 47 0 47 Beta vulgaris chitinase S51939 6e – 12 C34D4.11
AE00591.cl 45 45 0 C. elegans hypothetical protein AAF99918 7e – 26 F29B9.11b
AE01221.cl 44 44 0 Volvox carteri hydroxyproline-rich glycoprotein DZ-HRGP CAB62280 1e – 29 Y59A8B.19
AE01407.cl 41 41 0 C. elegans elt-3 GATA-binding transcription factor like CAA93510 5e – 18 K02B9.4b
AE00033.cl 41 0 41 Nippostrongylus brasiliensis hsp-20 Nbhsp20 CAA50655 8e – 56 T27E4.3
AE01361.cl 40 39 1 C. elegans ICD-1 inhibitor of cell death AAA68776 1e – 75 C56C10.8b
AE00322.cl 39 37 2 C. elegans hypothetical protein CAB54416 7e – 28 Y38E10A.24b
AE00536.cl 36 34 2 Homo sapiens unnamed protein product BAB71316 4e – 134 F25B5.4a
AE00503.cl 35 13 22 C. elegans eft-3 elongation factor 1-alpha NP_498520 3e – 283 F31E3.5b
a SW/TR is Swiss-prot and TrEMBL Proteinknowledgebase .
b C. elegans homolog has higher probability match than the best GenBank descriptor.
Repeating the analysis in Stein et al. [18] indicates that 6–7% of C. elegans and C. briggsae proteins are candidate "orphans", lacking homologs outside of the species. We examined whether these genes are truly orphans that have arisen in a Caenorhabditis sub-lineage or are instead genes present in an ancestral nematode that have been lost or evolved beyond recognition in one species. Of candidate orphan proteins, ten from C. elegans (Table 7) and 27 from C. briggsae [see Additional file 4] match A. caninum and/or A. ceylanicum clusters, with three and eight, respectively, having matches in both species. Most of the C. elegans orphans are hypothetical proteins of unknown function though some had functional information from InterPro domains (R10E9.3) or mutant phenotypes (ZK686.1). Therefore, at least a portion of the genes identified in either C. elegans or C. briggsae as "orphans" are actually ancestral nematode genes with homologs found in other clade V species and further clade V sequencing will likely reveal more such cases.
Table 7 C. elegans candidate orphans (1,358 out of 21,437) matching Ancylostoma clusters
C. elegans genea Descriptor Ancylostoma cluster idb ESTs per cluster E-value C. elegans gene length (aa) Matched region length (%) %ID
F31E8.1 Hypothetical protein AC05087.cl 1 1e – 07 249 14.5 45
F57B10.14 Hypothetical protein AE02023.cl 2 3e – 20 56 75.0 69
R10E9.3 Contains Cytochrome C heme-binding site AC02329.cl 1 2e – 10 149 79.2 32
AE00556.cl 2 5e – 19 149 97.3 31
T07A9.13 Putative nuclear encoded protein AE02236.cl 1 4e – 31 111 91.9 49
Y35H6.1 Hypothetical protein AE03902.cl 1 6e – 23 161 47.2 48
Y41C4A.3 Hypothetical protein AE00269.cl 14 1e – 05 162 49.4 37
Y54G2A.27 Hypothetical protein AC04390.cl 1 2e – 05 229 14.4 38
AE01938.cl 9 8e – 07 229 11.4 48
ZC487.3 Hypothetical protein AC04655.cl 1 2e – 08 79 81.0 38
ZK686.1 Nuclear transition protein AC00867.cl 16 2e – 07 44 68.2 62
AE01651.cl 3 3e – 07 44 68.2 62
ZK84.5 Hypothetical protein AC00410.cl 3 6e – 11 84 70.2 46
a Of 21,437 proteins 1,358 were candidate orphans
b AC, Ancylostoma caninum; AE, A. ceylanicum
Abundant transcripts expressed in Ancylostoma species
The 25 most abundantly represented clusters account for 14% and 19% of ESTs for A. caninum and A. ceylanicum respectively. The representation of the abundant transcripts varied from shared to stage-specific (Table 5 and Table 6). Hookworm developmental stages differ in habitat, morphology and behavior, hence highly represented gene transcripts may identify functions that are important to the parasites at various stages. Differences in gene expression between A. ceylanicum stages have been demonstrated previously for several genes [27,28]. Our comparison of iL3 and adult suggests additional examples (see Discussion). In fact, only 9% of the A. ceylanicum clusters are shared between iL3 and adult (Figure 2) and of the 25 largest clusters, 23 were biased toward one of the developmental stages (Table 6). While representation in EST collections generally correlates with source expression level [29], artifacts can occur [30,31]. Differences in expression are most likely to be accurate when comparing the most abundant transcripts in each stage. Therefore, while follow-up work is needed to confirm expression levels, examination of ESTs provides a list of candidates for various expression profiles.
Discussion
Overview
We have taken a genomics approach to the study of hookworm species, key parasites of humans and domestic animals that are related to the model nematode Caenorhabditis elegans. Nearly 20,000 ESTs from Ancylostoma caninum and A. ceylanicum identified approximately 7,000 genes including over 1,300 likely orthologs represented in both species. Close to 900 genes encode putative enzymes involved in 88 metabolic pathways. Over 3,100 genes contain recognizable protein domains many of which have been categorized in the Gene Ontology hierarchy. 4,600 genes have homologs in C. elegans including numerous nematode-specific genes and hundreds with observable RNAi phenotypes. ESTs originated from libraries representing infective L3 larva, stimulated L3, tissue arrested L3, and adults resulting in an improved rate of gene discovery and allowing the identification of transcripts abundant in various stages.
Gene expression in iL3 and dauers
Infective L3 (iL3) are developmentally-arrested, non-feeding pre-parasitic stages, which when triggered by the infection process and host-specific signals reactivate, molt and complete development. A similar stage in C. elegans is called the dauer larva. In Ancylostoma species host factors such as dog serum stimulate feeding and an activation response in serum stimulated L3 (ssL3) [32] that approximates the transition to parasitism in the host [33]. A. caninum tissue-arrested L3 (taL3) recovered from infected mice are a distinct population that potentially share properties with the arrested iL3. Developmentally arrested, non-feeding larvae would be expected to be dependent on stored energy reserves and lipid metabolic pathways; accordingly, the KEGG biochemical pathway mappings show a substantive number of clusters for fatty acid metabolism especially with the A. ceylanicum iL3 clusters (Table 4).
C. elegans microarray experiments identified 540 dauer-enriched genes along with genes involved in dauer-recovery [34]. C. elegans SAGE experiments identified 358 candidate dauer-specific genes [35]. Genes shown to be abundantly expressed in C. elegans dauers include a variety of genes that may play roles in extended survival including heat shock protein encoding genes like hsp-12.6 and daf-21 (Hsp90), ctl-1 (cytosolic catalase), sod-3 (superoxide dismutase), and hil-1 and hil-3 (Histone H1's). A number of genes identified both in Ancylostoma L3s and C. elegans dauers are discussed below.
Heat-shock Proteins
hsp-12.6, a member of the hsp-20 family, was one of the most highly represented clusters in A. caninum iL3 and taL3 as well as A. ceylanicum iL3 (Table 5 and Table 6). Among a Strongyloides stercoralis EST collection, the gene is also found in iL3 but not L1 [17]. C. elegans hsp-12.6 is upregulated in dauer and starved L1s [34,36] and is a transcriptional target of the FOXO transcription factor DAF-16 [37]. Unlike other HSPs, C. elegans hsp-12.6 is not stress-induced and does not prevent aggregation of unfolded proteins, suggesting a novel role. AE00033.cl, found exclusively in adult ESTs, encodes an ortholog of the Nippostrongylus brasiliensis HSP-20 protein. Nb-hsp-20 is more similar to the HSP-16 group of the HSP-20 family of small HSPs in C. elegans, is also expressed in the adult [38], and is not stress regulated, suggesting that it may function as an adult version of hsp-12.6.
Candidate stress-response proteins
A. caninum iL3 showed abundant clusters encoding homologs of the mitochondrial cytochrome oxidase subunits I, II, III and a stress associated endoplasmic reticulum protein not seen in ssL3 (Table 5). One A. ceylanicum iL3 abundant cluster encoded a ribosome-associated membrane 4 protein (RAMP4) involved in ER protein translocation [39] which is over-expressed in hypoxia and suppresses degradation of ER membrane proteins [40]. A homolog of C. elegans spp-4 was also expressed at high levels in A. ceylanicum iL3. spp-4 encodes an amoebapore, a member of the saposin-like protein superfamily that kill bacteria by forming membrane ion channels [41]. Amoebapore proteins are one of a number of putative stress response proteins regulated by DAF-16 in C. elegans [37,42]. These proteins, also including lysozyme and thaumatin, may provide a defense against worm pathogens and contribute to dauer longevity [43]. Hookworm free-living stages are also soil dwelling microbiverous organisms exposed to soil pathogens, so it is possible that spp-4 plays an antibacterial role in A. ceylanicum.
Gene expression in ssL3, adults, and multiple stages
In contrast to iL3, A. caninum ssL3 showed a CMP-sialic acid transporter, cathepsin D-like aspartic protease, calponin-like protein, and ham-2 zinc finger protein among the abundant transcripts. While the significance of these molecules is unknown, upregulation of an aspartic protease during the transition to parasitism and tissue penetration/migration is consistent with its role in degradation of serum proteins and collagens [44].
Abundant adult-specific clusters are likely to be involved in reproduction. For example, A. ceylanicum (Table 6) encodes an ortholog of the C. elegans VIT-3 protein, a lipid binding protein and major yolk component [45]. VIT-3 is expressed exclusively in the C. elegans adult hermaphrodite intestine, secreted, and taken up by oocytes. Two clusters encode genes involved in metabolism of chitin, an important constituent of the nematode eggshell [46]. One encodes a protein similar to C. elegans protein C34D4.11, and shows some similarity to a beet chitinase; the other encodes an ortholog of C. elegans B0280.5, a protein required for early embryonic development [47]. B0280.5 mRNA is expressed specifically in the adult hermaphrodite germ line and is a target of GLD-1, an RNA binding protein required for oocyte meiotic cell cycle progression [48].
ASP's
While there are differences in the cluster profiles among Ancylostoma stages, there are shared transcripts as well. For example, the Ancylostoma secreted protein ASP-1 and ASP-like cDNAs are present in abundance in both A. caninum iL3 and ssL3. The secretion of ASP-1 protein by ssL3s was noted as a marker of the transition to parasitism [12]. These results support conclusions made by Wang and Kim [34] that arrested larvae are transcriptionally prepared for dauer exit and upon receipt of appropriate stimulatory signals, exit from the arrested state is accompanied by a burst of translational activity in addition to further transcriptional activity. In contrast to ASP-1, ESTs for ASP-2 were exclusively detected in A. caninum iL3.
FAR Proteins
Two of the most abundant A. ceylanicum transcripts encode fatty acid/retinol binding (FAR) proteins (Table 6). FAR proteins are novel fatty acid and retinol binding proteins described in nematodes including A. caninum, other Strongylida, filarial, and plant parasitic species [49-51]. In C. elegans 8 FAR members are divided into 3 groups. All the parasitic nematodes FARs described to date are most similar to the C. elegans A group containing Ce-far-1, -2, and -6. Seven A. ceylanicum clusters encode FAR proteins. Four of which (9 ESTs) were found in the adult cDNA library; clusters AE00748.cl and AE03203.cl were nearly identical to Ac-far-1 and Ac-far-2 (98% and 99% nucleotide identity) whereas cluster AE02490.cl showed the highest similarity to Ce-far-1 and AE01700.cl to Ac-far-2. The iL3 specific clusters AE01410.cl (60/63 ESTs from iL3) and AE03983.cl (2/2 ESTs from iL3) were both most closely related to a FAR protein from Ostertagia ostertagi [52], and more distantly to Ac-far-1 and -2. Therefore, as seen in other parasitic nematodes, most A. ceylanicum FAR proteins are related to C. elegans group A FAR proteins. However, the A. ceylanicum cluster AE00750 was most similar to group C FAR protein Ce-far-7. Group C proteins differ from the other FARs in important ways including lacking an N-terminal signal peptide (suggesting an intracellular location), containing several cysteines, and failing to bind DAUDA [53]. AE00750.cl represents the first report of a FAR-7 like protein from a parasitic nematode. The function of FAR proteins is unknown but may represent a lipid acquisition system in which released FARs bind to lipids followed by uptake of the complex by a specific receptor mediated process. Retinoids are required for nematode growth and development, but are not synthesized by the worms. In parasitic nematodes, release of FAR proteins may also modify local inflammatory and immunological responses by delivering or sequestering biochemically important lipids [54].
Conclusion
The application of genomic approaches to hookworms has resulted in more than a 100-fold increase in available sequence data from Ancylostoma species thereby allowing an initial bioinformatic analysis of transcripts from these important parasites and establishing a foundation for the eventual completion of a hookworm genome. Semi-automated informatic approaches that are now being applied to all nematode sequences [55] allow uniform comparisons across many genomes and provide databases for further exploration. Transcripts in A. caninum and A. ceylanicum include clear candidates for stage specific expression representing the very different biological processes underway in different points of the lifecycle. The availability of the C. elegans and C. briggsae genomes has allowed highly informative comparisons to the two hookworm species showing extensive overlap in gene complements, including genes demonstrated to be essential in C. elegans and numerous genes specific to nematodes. As the most closely related major human pathogen to C. elegans, hookworms provide an attractive near-term application for using a model organism to better understand and eventually control a key disease-causing species. Beyond categorization of hookworm genes, clear research avenues are available to apply this information to improved methods for hookworm control including anthelmintic and vaccine development, diagnostics, population studies, as well as better understanding of fundamental aspects of hookworm biology, such as host immune system modulation.
Methods
Nematode extraction
A Shanghai strain of A. caninum was maintained in beagles as described [56]. Infective L3 (iL3) were recovered from 7–10 day old coprocultures using a modified Baermann technique, washed clean of debris with BU buffer (50 mM Na2PO4/22 mM KH2PO4/70 mM NaCl, pH 6.8; [57]), and snap-frozen by immersion in liquid N2. Frozen larvae were stored at -80°C until used for library construction. Serum stimulated larvae (ssL3) were generated by incubating iL3s harvested from a North Carolina strain of A. caninum in 5% normal dog serum for 20–24 h at 37°C, 5% carbon dioxide. Tissue-arrested L3 larvae (taL3) were recovered from BALB/c mice infected with 1,000–1,500 iL3 (North Carolina strain) and euthanized at 10–14 days post-infection [58].
A Warsaw strain of A. ceylanicum was maintained in Syrian hamsters as described [59], and L3 recovered, washed, and frozen as above. For the recovery of adult stage A. ceylanicum, infected hamsters with patent infections were sacrificed and the small intestine removed. The intestine was cut into 3 sections, opened longitudinally, and hung in 50 ml centrifuge tubes containing phosphate buffered saline at 37°C for 2–3 hrs. Following incubation, adult worms were recovered from the sediment, washed free of debris, and snap-frozen as above. All animals were housed and treated in accordance with institutional care and use committee guidelines.
Preparation of A. caninum staged RNA and cDNA libraries
Pulverization for the ssL3 and taL3 was performed using an Alloy Tool Steel Set (Fisher Scientific International). Total RNA from adult and larval parasites was prepared using TRIzol Reagent (GibcoBRL, Life Technologies or Invitrogen, Carlsbad, CA).
SMART based serum stimulated L3 library – Library construction was based on the SMART cDNA library construction system (Clontech Laboratories; [60]). mRNA was extracted from 10 μg of total RNA using a Dynabeads mRNA Purification kit (Dynal Biotech) with some modifications. First strand synthesis was performed with the mRNA bound to the oligo-dT of the Dynabeads using Superscript II RT (Invitrogen, Life Technologies) and the primer smart T7_3G_5. Concatemers were digested with Not I on the bead. Second strand synthesis was performed with the smartT7_5 and the smartCDSII primer. Amplification of the cDNA was performed with the smartT7_5 and smartT7_3 primers with cycling parameters of 95°C for 5 minutes, seven cycles of 95°C for 5 seconds, 60°C for 5 seconds and 68°C for 6 minutes followed by a 4 minute extension at 68°C. Following amplification, the cDNA was purified using the High Pure PCR Product Purification Kit (Roche). The final 5 cycles of PCR introduced the deoxy-UMP primers needed for cloning into the pAMP1 vector (Invitrogen, Life Technologies). cDNA fragments >1 kb were size selected on a 0.8% TAE agarose gel and cloned into the pAMP1 vector following the CloneAMP pAMP1 System (Invitrogen, Life Technologies). The ligation mix introduced into E. coli DH10B chemically competent cells (GibcoBRL, Life Technologies) resulted in 4.36 × 105 primary transformants.
SL1-PCR-based tissue arrested L3 library – mRNA was extracted from 2 μg of total RNA using a Dynabeads mRNA Purification kit (Dynal Biotech) and eluted with 10 μl 10 mM Tris-HCl. First strand synthesis was performed using linker primer (GAGAGAGAGAGAGAGAGAGAACTAGTCTCGAGTTTTTTTTTTTTT) and Superscript II RT (Invitrogen, Life Technologies). Amplification with Taq Polymerase used the SL1 (GGGTTTAATTACCCAAGTTTGA) and Xhop (GAGAGAGAACTAGTCTCGA) primers and 5 μl of the first strand reaction. Cycling parameters were 95°C for 5 minutes, 30 cycles of 95°C for 1 minute, 47°C for 1 minute, 72°C for 3 minutes followed by 5 minutes at 72°C. cDNA fragments >1 kb were size selected on a 0.8% TAE agarose gel and cloned into the pCRII-TOPO vector following the TOPO TA protocol (Invitrogen). The ligation mix was introduced into E. coli DH10B chemically competent cells (GibcoBRL, Life Technologies).
Hawdon infective L3 library – Frozen A. caninum L3 pellets were ground to powder on a mortar pre-chilled with liquid nitrogen. Total RNA was isolated from the powder using Trizol reagent (Invitrogen, Carlsbad, CA). Poly (A)+ RNA was isolated from total RNA using the Oligotex mRNA isolation kit (Qiagen, Chatsworth, CA). Approximately 5 μg of mRNA was used to construct a directional cDNA library in Lambda ZAP II (Stratagene, La Jolla, CA) as previously described [61]. pBluescript phagemid were mass excised prior to sequencing. The library had >95% recombinants, and insert size varied from 700–3,000 bp. The library was amplified once (106 pfu).
Preparation of A. ceylanicum staged RNA and cDNA libraries
Hawdon adult and infective L3 A. ceylanicum libraries – Total RNA and poly (A)+ mRNA were isolated from the appropriate A. ceylanicum life-cycle stage using Trizol reagent and the Oligotex mRNA isolation kit as described above. Approximately 5 μg of mRNA was used to construct directional cDNA libraries in Lambda ZAP II (Stratagene, La Jolla, CA) as previously described [61]. Both libraries had 99% recombinants with inserts ranging from 500–2500 bp (average 1500 bp), and each underwent one round of amplification (106 pfu). Inserts were mass excised as described above.
SL1-PCR-based infective L3 and adult libraries – cDNA was PCR amplified, using SL1-EcoRI primer on the 5' end and oligo(dT)-XhoI on 3' end, gel fractionated [62], and non-directionally cloned into pCR-TOPO-XL (Invitrogen), following XL-Topo TA cloning protocol. The cDNA inserts were excised with EcoRI.
Sequencing and clustering
Sequencing, EST processing and clustering were performed as described [17]. Information for clone requests and sequence trace files are available at . The completed cluster assemblies, NemaGene Ancylostoma caninum v 2.0 and A. ceylanicum v 2.0, were used as the basis for all subsequent analyses and are available for searching and acquisition by FTP at . "Fragmentation", defined as the representation of a single gene by multiple non-overlapping clusters, was estimated by examining Ancylostoma clusters with homology to C. elegans [17]. Overall representation of Ancylostoma genes is based on a theoretical gene number of 21,437, comparable to C. elegans wormpep97.
Analysis and functional assignments
Homology assignments – WU-BLAST sequence comparisons [63,64] were performed using A. caninum and A. ceylanicum contig consensus sequences which were further organized into clusters. Consensus sequences were used to search multiple databases, including the non-redundant GenBank (3/20/2003) and Wormpep v.97 C. elegans (Wellcome Trust Sanger Institute, unpublished) protein databases. Internally constructed databases using intersections of data from Genbank, allowed examination of sequences in specific phylogenetic distributions. Homologies were reported for E (expect) value scores of ≥ 1e -05.
To identify cases where Ancylostoma homologs in C. elegans have been surveyed for knock-down phenotypes using RNA interference, Wormpep BLAST matches were cross-referenced to a list of 17,042 C. elegans genes with available RNAi information (20th February 2005) . For each Ancylostoma cluster, only the best C. elegans match was considered.
Functional classification – Clusters were assigned putative functional categorization using two methods. First, InterProScan v.3.1 was used to search contig translations versus InterPro domains (11/08/02) [65,66]. Using InterPro, clusters were mapped to the three organizing principles of the Gene Ontology (GO_200211_assocdb.sql) [67]. Mappings are stored by MySQL database, displayed using AmiGo (11/25/02) , and are available at . Second, clusters were assigned by enzyme commission number to metabolic pathways using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (2/24/2004)[68]. All matches better than 1e-10 were taken into consideration.
Orthologs and dN/dS ratio – A. caninum / A. ceylanicum orthologs were determined by reciprocal best TBLASTX match using a threshold of E value ≥ 10-5. In addition, the ORFs accepted to be the correct translation were required to have the best C. elegans gene match in the same frame as the TBLASTX matches. Only continuous alignments longer than 30 amino acids were accepted. 'Suboptimal alignment program' (Jason Stajich, unpublished), scripted using tools in Bioperl [69] and utilizing 'yn00' from PAML [70], calculated the synonymous (dS) and non-synonymous substitutions (dN) per ortholog pair. A 4-way orthologs were assigned by using SSEARCH [71,72] to find the best C. elegans and C. briggsae homolog for each ortholog pair of A. caninum and A. ceylanicum. Orthologs were assigned if both sequences agreed on the best hits. Multiple sequence alignments were performed with MUSCLE [73]. Trees were built using the programs 'protml' and 'nucml' for protein and nucleotide sequences respectively [26]. An exhaustive search was used first to enumerate the possible topologies and then -R rearrangement search was used to identify the most likely branch lengths and bootstrap values. Only genes for which well supported topologies where A. caninum and A. ceylanicum appeared as sisters were used in subsequent analysis. Tree branch lengths were parsed and processed with Perl scripts written using modules from the Bioperl package and statistical tests were applied with the R package [74].
List of abbreviations used
Ad, adult parasite stage; BLAST, basic local alignment search tool; dN, non-synonymous substitutions; dS, synonymous substitutions; EST, expression sequence tag; GO, gene ontology; iL3, infective third larval stage; KEGG, Kyoto encyclopedia of genes and genomes; PCR, polymerase chain reaction; ssL3, serum-stimulated L3; taL3, tissue-arrested L3.
Authors' contributions
MM, JPM, SWC, RKW, and RHW conceived and designed the research plan and participated in all aspects of data collection and analysis. MM, JPM, MD, TW, JX, and JES analyzed and interpreted the data. PA, JH, and WK contributed material and constructed cDNA libraries. MM, JPM, PA, and JH drafted the manuscript. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
Accession numbers.
Click here for file
Additional File 2
Most conserved nematode genes between A. caninum and C. elegans.
Click here for file
Additional File 3
Classification of C. elegans RNAi phenotypes for genes with A. caninum and A. ceylanicum homologs.
Click here for file
Additional File 4
C. briggsae candidate orphans matching Ancylostoma clusters.
Click here for file
Acknowledgements
Ancylostoma EST sequencing at Washington University was supported by NIH-NIAID research grant AI 46593 to RHW. The authors would like to thank B. Chiapelli, C. Murphy, D. Pape, Bin Zhan, Adriana Magalska and E. Janecka for technical assistance, Reshad Dobardzic, Halina Wedrychowicz and Jerzy Behnke, for providing larval strains, and Peter Hotez and the Human Hookworm Initiative from the Sabin Vaccine Institute for their support. JPM was supported by a Helen Hay Whitney/Merck Fellowship. JES is supported by an NSF pre-doctoral fellowship. JPM is employee and equity holder of Divergence Inc; this research was not company funded.
==== Refs
de Silva NR Brooker S Hotez PJ Montresor A Engels D Savioli L Soil-transmitted helminth infections: updating the global picture Trends Parasitol 2003 12 547 551 14642761 10.1016/j.pt.2003.10.002
Anderson RC Nematode Parasites of Vertebrates, Their Development and Transmission 2000 New York: CABI Publishing
Schad GA Schad GA, Warren KS Hypobiosis and related phenomena in hookworm infection Hookworm Disease Current Status and New Directions 1990 London: Taylor and Francis 71 88
Stone WM Girardeau MR Transmammary passage of Ancylostoma caninum larvae in dogs J Parasitol 1968 54 426 429 5761146
Arasu P Anthelmintic efficacy against tissue-arrested larvae of Ancylostoma caninum in murine hosts J Parasitol 1998 84 1263 1267 9920326
Hotez PJ Zhan B Bethony JM Loukas A Williamson A Goud GN Hawdon JM Dobardzic A Dobardzic R Ghosh K Progress in the development of a recombinant vaccine for human hookworm disease: the Human Hookworm Vaccine Initiative Int J Parasitol 2003 33 1245 1258 13678639 10.1016/S0020-7519(03)00158-9
Boguski MS Tolstoshev CM Bassett DEJ Gene discovery in dbEST Science 1994 265 1993 1994 8091218
Blaxter ML De Ley P Garey JR Liu LX Scheldeman P Vierstraete A Vanfleteren JR Mackey LY Dorris M Frisse LM A molecular evolutionary framework for the phylum Nematoda Nature 1998 392 71 75 9510248 10.1038/32160
Blaxter M Genes and genomes of Necator americanus and related hookworms Int J Parasitol 2000 30 347 355 10731559 10.1016/S0020-7519(99)00198-8
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
Daub J Loukas A Pritchard DI Blaxter M A survey of genes expressed in adults of the human hookworm, Necator americanus Parasitology 2000 120 171 184 10726278 10.1017/S0031182099005375
Hawdon JM Jones BF Hoffman DR Hotez PJ Cloning and characterization of Ancylostoma-secreted protein. A novel protein associated with the transition to parasitism by infective hookworm larvae J Biol Chem 1996 271 6672 6678 8636085 10.1074/jbc.271.12.6672
Gasser RB Newton SE Genomic and genetic research on bursate nematodes: significance, implications and prospects Int J Parasitol 2000 30 509 534 10731573 10.1016/S0020-7519(00)00021-7
Hoekstra R Visser A Otsen M Tibben J Lenstra JA Roos MH EST sequencing of the parasitic nematode Haemonchus contortus suggests a shift in gene expression during transition to the parasitic stages Mol Biochem Parasitol 2000 110 53 68 10989145 10.1016/S0166-6851(00)00255-3
Jasmer DP Dautova Mitreva M McCarter JP mRNA sequences for Haemonchus contortus intestinal cathepsin B-like cysteine proteases display an extreme in abundance and diversity compared with other adult mammalian parasitic nematodes Mol Biochem Parasitol 2004 137 297 305 15383300 10.1016/j.molbiopara.2004.06.010
Harcus YM Parkinson J Fernandez C Daub J Selkirk ME ML B Maizels RM Signal sequence analysis of expressed sequence tags from the nematode Nippostrongylus brasiliensis and the evolution of secreted proteins in parasites Genome Biol 2004 5 R39 15186490 10.1186/gb-2004-5-6-r39
Mitreva M McCarter JP Martin J Dante M Wylie T Chiapelli B Pape D Clifton SW Nutman TB Waterston RH Comparative Genomics of Gene Expression in the Parasitic and Free-living Nematodes Strongyloides stercoralis and Caenorhabditis elegans Genome Res 2004 14 209 220 14762059 10.1101/gr.1524804
Stein LD Bao Z Blasiar D Blumenthal T Brent MR Chen N Chinwalla A Clarke L Clee C Coghlan A The Genome Sequence of Caenorhabditis briggsae: A Platform for Comparative Genomics PLoS Biol 2003 1 E45 14624247 10.1371/journal.pbio.0000045
Bin Z Hawdon J Qiang S Hainan R Huiqing Q Wei H Shu-Hua X Tiehua L Xing G Zheng F Hotez P Ancylostoma secreted protein 1 (ASP-1) homologues in human hookworms Mol Biochem Parasitol 1999 98 143 149 10029316 10.1016/S0166-6851(98)00157-1
Hotez PJ Hawdon JM Cappello M Jones BF Pritchard DI Molecular pathobiology of hookworm infection Infect Agents Dis 1995 4 71 75 7613730
Tawe W Pearlman E Unnasch TR Lustigman S Angiogenic activity of Onchocerca volvulus recombinant proteins similar to vespid venom antigen 5 Mol Biochem Parasitol 2000 109 91 99 10960168 10.1016/S0166-6851(00)00231-0
Ding X Shields J Allen R Hussey RS Molecular cloning and characterisation of a venom allergen AG5-like cDNA from Meloidogyne incognita Int J Parasitol 2001 30 77 81 10675748 10.1016/S0020-7519(99)00165-4
Zhan B Badamchian M Meihua B Ashcom J Feng J Hawdon J Shuhua X Hotez PJ Molecular cloning and purification of Ac-TMP, a developmentally regulated putative tissue inhibitor of metalloprotease released in relative abundance by adult Ancylostoma hookworms Am J Trop Med Hyg 2002 66 238 244 12139214
Waterston RH Lindblad-Toh K Birney E Rogers J Abril JF Agarwal P Agarwala R Ainscough R Alexandersson M An P Initial sequencing and comparative analysis of the mouse genome Nature 2002 420 520 562 12466850 10.1038/nature01262
Kamath RS Fraser AG Dong Y Poulin G Durbin R Gotta M Kanapin A Le Bot N Moreno S Sohrmann M Systematic functional analysis of the Caenorhabditis elegans genome using RNAi Nature 2003 421 231 237 12529635 10.1038/nature01278
Adachi J Hasegawa M MOLPHY version 23: Programs for Molecular Phylogenetics based on Maximum Likelihood 1996 Tokyo: Institute of Statistical Mathematics
Hawdon JM Narasimhan S Hotez PJ Ancylostoma secreted protein 2: cloning and characterization of a second member of a family of nematode secreted proteins from Ancylostoma caninum Mol Biochem Parasitol 1999 99 149 165 10340481 10.1016/S0166-6851(99)00011-0
Zhan B Hotez PJ Wang Y Hawdon JM A developmentally regulated metalloprotease secreted by host-stimulated Ancylostoma caninum third-stage infective larvae is a member of the astacin family of proteases Mol Biochem Parasitol 2002 120 291 296 11897134 10.1016/S0166-6851(01)00453-4
Audic S Claverie JM The Significance of Digital Gene Expression Profiles Genome Res 1997 7 986 995 9331369
Mitreva M Jasmer DP Appleton J Martin J Dante M Wylie T Clifton SW Waterston RH McCarter JP Gene discovery in the adenophorean nematode Trichinella spiralis: an analysis of transcription from three life cycle stages Mol Biochem Parasitol 2004 137 277 291 15383298 10.1016/j.molbiopara.2004.05.015
Munoz ET Bogarad LD Deem MW Microarray and EST database estimates of mRNA expression levels differ: the protein length versus expression curve for C. elegans BMC Genomics 2004 5 30 15134588 10.1186/1471-2164-5-30
Arasu P Kwak D Developmental arrest and pregnancy-induced transmammary transmission of Ancylostoma caninum larvae in the murine model J Parasitol 1999 85 779 784 10577710 10.1007/s004360050631
Brand A Hawdon JM Phosphoinositide-3-OH-kinase inhibitor LY294002 prevents activation of Ancylostoma caninum and Ancylostoma ceylanicum third-stage infective larvae Int J Parasitol 2004 34 909 914 15217729 10.1016/j.ijpara.2004.04.003
Wang J Kim SK Global analysis of dauer gene expression in Caenorhabditis elegans Development 2003 130 1621 1634 12620986 10.1242/dev.00363
Jones SJ Riddle DL Pouzyrev AT Velculescu VE Hillier L Eddy SR Stricklin SL Baillie DL Waterston R Marra MA Changes in Gene Expression Associated with Developmental Arrest and Longevity in Caenorhabditis elegans Genome Res 2001 11 1346 1352 11483575 10.1101/gr.184401
Leroux MR Ma BJ Batelier G Melki R Candido EP Unique structural features of a novel class of small heat shock proteins J Biol Chem 1997 272 12847 12853 9139746 10.1074/jbc.272.19.12847
Murphy CT McCarroll SA Bargmann CI Fraser A Kamath RS Ahringer J Li H Kenyon C Genes that act downstream of DAF-16 to influence the lifespan of Caenorhabditis elegans. [see comment] Nature 2003 424 277 283 12845331 10.1038/nature01789
Tweedie S Grigg ME Ingram L Selkirk ME The expression of a small heat shock protein homologue is developmentally regulated in Nippostrongylus brasiliensis Mol Biochem Parasitol 1993 61 149 153 8259127 10.1016/0166-6851(93)90168-W
Yamaguchi A Hori O Stern DM Hartmann E Ogawa S Tohyama M Stress-associated endoplasmic reticulum protein 1 (SERP1)/Ribosome-associated membrane protein 4 (RAMP4) stabilizes membrane proteins during stress and facilitates subsequent glycosylation J Cell Biol 1999 147 1195 1204 10601334 10.1083/jcb.147.6.1195
Wang L Dobberstein B Oligomeric complexes involved in translocation of proteins across the membrane of the endoplasmic reticulum FEBS Letters 1999 457 316 322 10471800 10.1016/S0014-5793(99)01075-3
Leippe M Andra J Nickel R Tannich E Muller-Eberhard HJ Amoebapores, a family of membranolytic peptides from cytoplasmic granules of Entamoeba histolytica: isolation, primary structure, and pore formation in bacterial cytoplasmic membranes Mol Microbiol 1994 14 895 904 7715451
Nicholas HR Hodgkin J Responses to infection and possible recognition strategies in the innate immune system of Caenorhabditis elegans Mol Immunology 2004 41 479 493 10.1016/j.molimm.2004.03.037
Garsin DA Villanueva JM Begun J Kim DH Sifri CD Calderwood SB Ruvkun G Ausubel FM Long-lived C.elegans daf-2 mutants are resistant to bacterial pathogens Science 2003 300 1921 12817143 10.1126/science.1080147
Williamson AL Brindley PJ Loukas A Hookworm cathepsin D aspartic proteases: contributing roles in the host-specific degradation of serum proteins and skin macromolecules Parasitology 2003 126 179 185 12636356 10.1017/S0031182002002706
Spieth J Blumenthal T The Caenorhabditis elegans vitellogenin gene family includes a gene encoding a distantly related protein Mol Cellular Biology 1985 5 2495 2501
Spindler KD Spindler-Barth M Londershausen M Chitin metabolism: a target for drugs against parasites Parasitol Res 1990 76 283 288 2186405 10.1007/BF00928180
Lee MH Schedl T Identification of in vivo mRNA targets of GLD-1, a maxi-KH motif containing protein required for C. elegans germ cell development Genes Dev 2001 15 2408 2420 11562350 10.1101/gad.915901
Hansen D Wilson-Berry L Dang T Schedl T Control of the proliferation versus meiotic development decision in the C. elegans germline through regulation of GLD-1 protein accumulation Development 2004 131 93 104 14660440 10.1242/dev.00916
Basavaraju S Zhan B Kennedy MW Liu Y Hawdon J Hotez PJ Ac-FAR-1, a 20 kDa fatty acid- and retinol-binding protein secreted by adult Ancylostoma caninum hookworms: gene transcription pattern, ligand binding properties and structural characterisation Mol Biochem Parasitol 2003 126 63 71 12554085 10.1016/S0166-6851(02)00253-0
Garofalo A Klager SL Rowlinson MC Nirmalan N Klion A Allen JE Kennedy MW Bradley JE The FAR proteins of filarial nematodes: secretion, glycosylation and lipid binding characteristics Mol Biochem Parasitol 2002 122 161 170 12106870 10.1016/S0166-6851(02)00097-X
Prior A Jones JT Blok VC Beauchamp J McDermott L Cooper A Kennedy MW A surface-associated retinol- and fatty acid-binding protein (Gp-FAR-1) from the potato cyst nematode Globodera pallida: lipid binding activities, structural analysis and expression pattern Biochem J 2001 356 387 394 11368765 10.1042/0264-6021:3560387
Vercauteren I Geldhof P Peelaers I Claerebout E Berx G Vercruysse J Identification of excretory-secretory products of larval and adult Ostertagia ostertagi by immunoscreening of cDNA libraries Mol Biochem Parasitol 2003 126 201 208 12615319 10.1016/S0166-6851(02)00274-8
Garofalo A Rowlinson MC Amambua NA Hughes JM Kelly SM Price NC Cooper A Watson DG Kennedy MW Bradley JE The FAR protein family of the nematode Caenorhabditis elegans. Differential lipid binding properties, structural characteristics, and developmental regulation J Biol Chem 2003 278 8065 8074 12502713 10.1074/jbc.M206278200
Bradley JE Nirmalan N Klager SL Faulkner H Kennedy MW River blindness: a role for parasite retinoid-binding proteins in the generation of pathology? Trends Parasitol 2001 17 471 475 11587960 10.1016/S1471-4922(01)02036-0
Parkinson J Mitreva M Whitton C Thomson M Daub J Martin J Hall N Barrell B Waterston RH McCarter JP Blaxter M A transcriptomic analysis of the phylum Nematoda Nature Genetics 2004 36 1259 1267 15543149 10.1038/ng1472
Hawdon JM Datu BJ The second messenger cyclic GMP mediates activation in Ancylostoma caninum infective larvae Int J Parasitol 2003 33 787 793 12865078 10.1016/S0020-7519(03)00088-2
Hawdon JM Schad GA Long-term storage of hookworm infective larvae in buffered saline solution maintains larval responsiveness to host signals J Helm Soc Wash 1991 58 140 142
Arasu P In vitro reactivation of Ancylostoma caninum tissue-arrested third-stage larvae by transforming growth factor-beta J Parasitol 2001 87 733 738 11534634
Hawdon JM Volk SW Pritchard DI Schad GA Resumption of feeding in vitro by hookworm third-stage larvae: a comparative study J Parasitol 1992 78 1036 1040 1491295
Ciavatta V Cairney J Isolation of full-length cDNA clones using SMART cDNA and a biotin-streptavidin bead system Biotechniques 2000 29 444 446 10997256
Zhan B Hawdon J Shan Q Ren H Qiang H Xiao SH Li TH Feng Z Hotez P Construction and analysis of cDNA library of Necator americanus third stage larvae J Parasitol Paras Dis 2000 18 26 28
Osoegawa K Woon PY Zhao B Frengen E Tateno M Catanese JJ de Jong PJ An improved approach for construction of bacterial artificial chromosome libraries Genomics 1998 15 1 8 9740665 10.1006/geno.1998.5423
Altschul SF Gish W Miller W Myers EW Lipman DJ Basic local alignment search tool J Mol Biol 1990 215 403 410 2231712 10.1006/jmbi.1990.9999
Gish W 1996–2002
Apweiler R Attwood TK Bairoch A Bateman A Birney E Biswas M Bucher P Cerutti L Corpet F Croning MD The InterPro database, an integrated documentation resource for protein families, domains and functional sites Nucleic Acids Res 2001 29 37 40 11125043 10.1093/nar/29.1.37
Zdobnov EM Apweiler R InterProScan – an integration platform for the signature-recognition methods in InterPro Bioinformatics 2001 17 847 848 11590104 10.1093/bioinformatics/17.9.847
The Gene Ontology Consortium Gene ontology: tool for the unification of biology Nat Genet 2000 25 25 29 10802651 10.1038/75556
Kanehisa M Goto S KEGG: kyoto encyclopedia of genes and genomes Nucleic Acids Res 2000 28 27 30 10592173 10.1093/nar/28.1.27
Stajich JE Block D Boulez K Brenner SE Chervitz SA Dagdigian C Fuellen G Gilbert JG Korf I Lapp H The Bioperl toolkit: Perl modules for the life sciences Genome Res 2002 12 1611 1618 12368254 10.1101/gr.361602
Yang Z PAML: A program package for phylogenetic analysis by maximum likelhood Comput Appl Biosci 1997 13 555 556 9367129
Pearson WR Searching protein sequence libraries: comparison of the sensitivity and selectivity of the Smith-Waterman and FASTA algorithms Genomics 1991 11 635 650 1774068 10.1016/0888-7543(91)90071-L
Smith TF Waterman MS Identification of common molecular subsequences J Mol Biol 1981 147 195 197 7265238 10.1016/0022-2836(81)90087-5
Edgar RC MUSCLE: multiple sequence alignment with high accuracy and high throughput Nucleic Acids Res 2004 32 1792 1797 15034147 10.1093/nar/gkh340
R Development Core team R: A language and environment for statistical computing 2004
| 15854223 | PMC1112591 | CC BY | 2021-01-04 16:39:53 | no | BMC Genomics. 2005 Apr 26; 6:58 | utf-8 | BMC Genomics | 2,005 | 10.1186/1471-2164-6-58 | oa_comm |
==== Front
BMC GenomicsBMC Genomics1471-2164BioMed Central London 1471-2164-6-601586971010.1186/1471-2164-6-60SoftwarePathwayVoyager: pathway mapping using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database Altermann Eric [email protected] Todd R [email protected] Department of Food Science, Box 7624, North Carolina State University, Raleigh, NC 27695-7624, USA2005 3 5 2005 6 60 60 21 12 2004 3 5 2005 Copyright © 2005 Altermann and Klaenhammer; licensee BioMed Central Ltd.2005Altermann and Klaenhammer; 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
Equally important and challenging as genome annotation, is the subsequent classification of predicted genes into their respective pathways. The Kyoto Encyclopedia of Genes and Genomes (KEGG) represents a database consisting of known genes and their respective biochemical functionalities. Although accessible online, analyses of multiple genes are time consuming and are not suitable for analyzing data sets that are proprietary.
Results
Presented here is a new software solution that utilizes the KEGG online database for pathway mapping of partial and whole prokaryotic genomes. PathwayVoyager retrieves user-defined subsets of the KEGG database and stores the data as local, blast-formatted databases. Previously selected datasets can be re-used, reducing run-time significantly. Whole or partial genomes can be automatically analyzed using NCBI's BlastP algorithm and ORFs with similarities below the user-defined threshold will be marked on pathway maps. Multiple gene hits are sorted by similarity. Since no sequence information is transmitted over the Internet, PathwayVoyager is an ideal solution for pathway mapping and reconstruction of confidential DNA sequence data.
Conclusion
PathwayVoyager represents an alternative approach to many already existing, more complex pathway reconstructions software solutions. This software does not require any dedicated hardware or software and is flexible and straightforward to use. It is ideally suited for environments where analyses on variable datasets are desired.
==== Body
Background
The ongoing sequencing of complete genomes of prokaryotes and eukaryotes reveals a tremendous amount of uncharted data. In prokaryotic genomes, roughly 25 to 30 percent of the predicted ORFeome remain functionally unknown with many Open Reading Frames (ORFs) only showing similarities to conserved hypothetical ORFs of other organisms. However, a significant number of the predicted ORFs do show similarities to functionally classified genes with defined roles in the complex network of metabolic pathways. Previously, the most common approach to determine the functionality of a gene and its gene product was to experimentally determine phenotypic changes upon inactivation or overexpression of the gene. This is the most effective approach for determining the roles of single genes, but it is unfeasible to investigate complete ORFeomes with over 2000 ORFs. The Kyoto Encyclopedia of Genes and Genomes (KEGG) represents an ambitious and successful attempt to assign known enzymes into known biochemical pathways and is updated on a regular basis [1-3]. The database is represented by a web-based browser and a multitude of different analyses are possible. Genes can be analyzed using online Blast algorithms and if significant Blast similarities resulted in the assignment of a defined enzyme class (i.e. EC classification), these genes can be marked in corresponding KEGG pathways. However, analysis of larger numbers of genes using this manual approach is tedious, inflexible, and time consuming. Since un-encrypted data transmission is used for most public servers, remote Blast analyses with confidential sequence queries is not desirable. In addition to this web-browser based approach, the KEGG database can also be accessed directly via an application programming interface (API) and the underlying databases can be downloaded for local uses. Third party-software most commonly use the KEGG database for gene-classifications, often in combination with whole genome annotation efforts [4], or utilizes the database content for reference gene sets used in further experiments [5]. Other remote software solutions like DAVID [6] integrate various databases and experimental results to allow for extensive query-based data mining on given gene lists. Similarly, Pathway Tools utilizes dedicated server and database backbones to realize a sophisticated environment. New genomes must be manually integrated into a Pathway/Genome Database (PGDB), which in turn sets the basis for more complex queries and analyses. However, the significant resources required to implement genome information for Pathway Tools might not be readily available. PathFinder [7] and BioMiner [8] use a different approach in that they utilize some of the data the KEGG database provides and then approach pathway reconstruction using software specific algorithms. However, these solutions tend to be complex, thoroughly web-based, and often utilize the whole KEGG databases without selective options. Furthermore, most of these solutions require specific data formats that are compatible with the respective application for initial data parsing and entry. For example, although PathFinder can parse common EMBL input files, only genes with an EC-number tag can initially be integrated into the database system, implying a sophisticated level of existing genome annotation. Although this approach can be used to generally classify genes, it lacks the necessary flexibility to compare specific pathways of interest in single organisms or groups of organisms to a selected set of genes. In addition, these complex algorithms are not always needed and more simplistic and faster solutions with less hardware and software requirements would be preferable due to their ease of use and flexibility.
Closely-related groups of organisms may differ in certain key elements that define specific strain/species related differences. Comparing these organisms or groups of organisms with each other should highlight these differences and reveal specific properties or lead to new genetic targets for pathway engineering. Therefore, it may be desirable to choose only subsets of organisms and pathways from the overall KEGG database content.
PathwayVoyager was developed to overcome most of these obstacles. The software resembles a tool to analyze an unlimited number of protein sequences against a user-selected subset of the KEGG database using NCBI's BlastP algorithm and subsequently places them into their proper pathway positions. Results are displayed in colored pathway maps and hits can easily be analyzed using the graphical interface. This tool reflects a different approach to pathway mapping, in that it provides a simplistic and flexible approach with few prerequisites. No dedicated hardware (i.e. background server) or software (i.e. relational database backbones) are necessary to analyze given datasets. A standard PC with the Windows operating system is sufficient to operate PathwayVoyager. In contrast to more complex tools, no underlying protein annotation is necessary and plain protein sequences in FASTA format can be used as query templates. This approach is ideal for draft phase genomes and ongoing annotation efforts in completed genomes where the emphasis lies on the establishment and verification of gene annotation and an initial assessment of metabolic capabilities. The resulting main advantage of PathwayVoyager is its speed and economy for initial pathway mapping. Also, the resulting data can easily be accessed on different locations by transferring the generated flatfile database to the respective computers. Once the research objective shifts to comparative and predictive pathway analyses, other tools like DAVID or PathwayTools become more advantageous. PathwayVoyager fills a niche for environments with limited hardware and software resources that still require a significant and meaningful way to perform small and large scale pathway mapping projects from varying data sources.
Implementation
PathwayVoyager is written completely in Perl/Tk and requires the Perl interpreter . No further Perl modules are required. However, two external distributions are required, namely the NCBI Blast distribution , and the SOAP::Lite client to utilize the KEGG API. Perl/Tk provides the interpreter for PathwayVoyager, and the SOAP::Lite client facilitates interaction with the KEGG API. The standalone Blast distribution is used to generate Blast compatible databases and to perform the local Blast analyses. For data analysis and browsing, the Perl/Tk interpreter is the only pre-requisite for PathwayVoyager. The standalone Blast distribution, and the SOAP::Lite client can be omitted and no internet connection is necessary.
The software was developed to optimally complement the GAMOLA annotation suite [9] but accepts any protein sequence in FASTA format. The sequential numbering of ORFs in GAMOLA annotated genomes is reflected by gene-name tags in the generated Genbank files. Extracting protein sequences into FASTA files preserves this numbering scheme and is subsequently presented in the browser module of PathwayVoyager. This permits fast and efficient ORF-tracking throughout the genome and often provides preliminary identification of gene clusters. PathwayVoyager operates as a stand alone software solution without the need of additional database backbone systems. The use of PathwayVoyager and the KEGG database system implies the agreement to the license terms specified for KEGG at .
The general procedure used by PathwayVoyager to map genes is illustrated in Figure 1. User defined organism and pathway subsets of the KEGG database at the time of analysis result in the online retrieval of a defined set of protein sequences via the KEGG API (API methods: list_organisms; list_pathways). This dataset is converted into a BlastP compatible database using "formatdb" provided by NCBI. The query protein sequences in FASTA format are then subjected to a BlastP analysis, utilizing "blastall", also part of the standalone Blast distribution. Blast results for each query are formatted and saved into a flatfile database, linking each query to KEGG EC-number entries. KEGG map points are then requested for each selected pathway (API methods: get_enzymes_by_pathway, get_genes_by_pathway,) and stored in an internal tabular format. Subsequently, the Blast flatfile database is analyzed and Blast hits above a user-defined e-value threshold are discarded. The remaining significant Blast hits are then parsed to group query-sequences to KEGG map points for each selected pathway. Each KEGG entry in a given pathway is assigned to a list of significant Blast hits. Each hit displays the respective query ORFs and its best Blast hit based on the selected KEGG database subset, and the corresponding Blast scores and e-values. Based on this query-assignment, tagged KEGG pathway graphs are requested online using the KEGG API (API method: mark_pathway_by_objects). Resulting GIF maps and ASCII group assignment data files are stored in a flatfile database and can be visualized using the graphical browser module. Previous analyses can be accessed directly from the browser without the need to re-analyze the query sequences.
Figure 1 Pathway Voyager mapping procedure. The analysis and mapping procedure of PathwayVoyager is shown in a flowchart diagram. Manual selection of organisms and pathways present in the KEGG database, at the time of analysis, results in the retrieval of a specific set of protein sequences that are subsequently reformatted into a BlastP database. Protein query sequences are then used as templates for local BlastP analyses. Results are subjected to a user-defined threshold and subsequently parsed to retrieve tagged KEGG pathway maps. Pathway graphs and parsed BlastP results are stored as a flatfile database and can be displayed using the graphical browser. Opposite double arrows represent Internet-access using the KEGG API. Symbols used are according to general flowchart conventions.
The graphical user interface was designed to be self-explanatory and easy to use. After the initial pathway setup, no further installation steps are necessary. Although PathwayVoyager requires an internet connection in order to retrieve data from the KEGG database, all analyzes involving the provided gene sets are performed locally and no sensitive data are transmitted. This eliminates one of the major security concerns when working with confidential data and permits the real-time use of the KEGG database system. PathwayVoyager does not require any dedicated hardware and has been tested on a standard PC and the Windows platform. Linux versions of Perl/Tk, the SOAP::Lite client, and the standalone Blast distribution are freely available and would allow PathwayVoyager to operate under a Linux environment, as well. For certain selectable pathways (e.g. Ribosomal reference pathway) KEGG does not yet support organism independent marking. For practical reasons, no hits will be displayed for these pathways.
Results and discussion
Upon start, the main window displays two list-boxes and the currently active buttons (Figure 2, A, B, and 2C, respectively). Changing the default values is possible at this stage through the "Setup" function and has to be done before initializing the analysis. The "Start" button (Figure 2, C) commences the analysis and all organisms currently present in the KEGG database will be displayed in a listbox (Figure 2, A). Single and multiple selections are possible and will be confirmed with the "Retrieve Organisms" button. The second listbox (Figure 2, B) will then be automatically populated with KEGG pathways. After selecting the desired pathways to be analyzed, the "Retrieve Pathways" button confirms the selection and starts the KEGG pathway mapping. Both, the organisms and the pathways present in the KEGG database at the time of the analysis can be selected independently from each other, with the exception of organism-specific pathways (i.e. ABC transporter or two-component regulatory systems). This guarantees use of the most flexible solution for selective comparative analyses between groups of organisms. By selecting all organisms and pathways, the given gene set can be compared against the complete KEGG database.
Figure 2 PathwayVoyager main window. Screenshot of the graphical user interface. (A) and (B) indicate listboxes showing the organisms that can be individually selected and pathways represented in KEGG at the time of analysis. Button for exiting the software, redo the analysis, and accessing the setup are located beneath both boxes. Section (C) resembles the step-by-step design to start the analysis, retrieve the selected organisms and pathways, and subsequently obtain the respective protein sequences. In manual mode, the KEGG analysis and pathway retrieval can be activated by the user. Existing KEGG analyses can be reviewed by accessing the lowest button. On-the-fly options, for re-using previous organism-pathway selections and respective BlastP results, are located in the top region of section (C).
The user-selected organism and pathway combination is shown in a separate pop-up window (not shown). The current status of the KEGG pathway mapping is also shown in a separate log-window (not shown). In general, the right panel (Figure 2, C) harbors the user-guide interface and was designed to lead the user through the analyses in a step-by-step approach. By default, the organism and pathway confirmation automatically initializes the KEGG pathway mapping. If only the selected and retrieved protein sequences are required, or a manual start for pathway mapping is desired, the setup module allows the configuration for manual mode. KEGG pathway mapping can then be initiated with the "Submit to KEGG" button. Selected pathway/organism combinations are saved as an ASCII text file. The retrieved protein sequences are stored into a separate ASCII file and a Blast-compatible database is generated. For future analyses, the pathway/organism selection and the respective database can be re-used with different query protein sequence sets.
The possibility to re-use previous selections dramatically reduces the time needed to complete KEGG analyses, as retrieval of individual protein sequences from KEGG is omitted. In addition, Blast results obtained with the given query set can also be re-used. This shortens the run time further, enabling rapid mappings and analyses of pathways with varying relaxed or stringent threshold values.
The provided gene set will then be compared to the local database generated from the selected organism-pathway protein sequence combination using the BlastP algorithm. Blast hits featuring an e-value below the user-selectable threshold will be used to generate the marked KEGG pathway requests. Pathway maps are saved as GIF files and the URL for the respective KEGG pathway map including the corresponding BlastP results are stored separately in text files.
Results are displayed in a separate window. Figure 3 illustrates pathway mapping for the Glycolysis/Gluconeogenesis pathway (KEGG pathway code: 00010) using the ORFeome of L. acidophilus NCFM [10] as query set and the complete KEGG database as template.
Figure 3 Interactive KEGG Pathway display. The screenshot illustrates KEGG pathway mapping for the glycolysis/gluconeogenesis pathway using the predicted ORFeome of the GAMOLA annotated L. acidophilus NCFM genome as query template. The e-value threshold was set to 1e-10. Section (A) shows all pathways used for this analysis. Section (B) lists the marked entries. Section (C) represents the graphical representation of a selected pathway. Elements exhibiting BlastP hits below the selected threshold e-value are marked as red boxes. Section (D) shows the corresponding BlastP results, comprising the respective query protein designation, the corresponding KEGG hit, its amino-acid length, the BlastP-score, and e-value. Entries are sorted by ascending e-values.
In general, previously selected pathways are displayed by either their KEGG pathway code or full name. Alternative analyses can be displayed by changing the default mapping directory, using the "Directory" function (Figure 3, A). The selected pathway will then be graphically displayed and BlastP hits below the specified threshold are indicated as red boxes, bearing the respective EC numbers (Figure 3, C). Each marked element is shown by its EC-number code, numerically sorted, in a listbox (Figure 3, B). Upon selection of an entry, all BlastP hits below the threshold are sorted by ascending e-values and displayed accordingly (Figure 3, D). This workflow allows for a quick pathway mapping throughout a given gene set and those potentially involved in multiple pathways can be easily identified and analyzed.
In the example shown, the conversion of glyceronephosphate to glyceraldehyde-3-phosphate is mediated by a triosephosphate isomerase (EC 5.3.1.1). Selecting this entry from the EC entry list (Figure 3, B), highlights all query hits found in L. acidophilus below the defined threshold (Figure 3, D). Two entries below an e-value of 1e-120 were found, namely ORFs Lba699 (e-value: 1e-127) and Lba700 (e-value: 1e-131). Both entries show significant similarities to triosephosphate isomerases. Further analyses showed that the conversion of glyceraldehyde-3-phosphate to glycerate-1,3-bisphosphate and to glycerate-3-phosphate is mediated by Lba698 (EC 1.2.1.12, e-value 1e-176) and Lba699 (EC 2.7.2.3, e-value 0), respectively. The ambiguity found for EC 5.3.1.1 could be resolved and, consequently, the genome annotation was updated accordingly. More detailed analyses revealed the presence of the complete pathway for uptake and conversion of glucose into pyruvate and L-lactate. A more detailed analysis of the complete metabolic pathway reconstruction of L. acidophilus NCFM using PathwayVoyager is described elsewhere [10].
PathwayVoyager does not evaluate or extrapolate the displayed hits and the quality and significance of the results depend on the current content of the KEGG database. As with every predictive software, results should be carefully analyzed and seen in their genetic context to evaluate activities and potential substrate specifity-variances carried out by homologous enzymes. Results from previous analyses can be displayed by selecting the "View existing KEGG pathways" option in the PathwayVoyager main window (Figure 2, C).
Run times for PathwayVoyager may vary, depending on the number of selected pathways and organisms. Analysis of a complete genome of ~2,000 open reading frames (ORFs) using the complete KEGG database can be carried out in less than 36 h.
Conclusion
PathwayVoyager differs significantly in its approach from other software solutions for pathway reconstructions that already exist. In contrast to the often highly complex and specific algorithms, PathwayVoyager represents a more straight-forward approach, and doesn't require substantial resources on the users' side. Relying on the Blast algorithm and the ambitious KEGG database, PathwayVoyager utilizes widely accepted resources to analyze and map data. Despite the uncomplicated approach, evidential data can rapidly be obtained and easily analyzed during genome analyses. PathwayVoyager represents an effective pathway mapping tool for large or confidential data sets.
Availability and requirements
• Project name: Biological Pathway Mapping
• Project home page: none
• Operating system(s): Platform independent
• Programming language: PERL
• Other requirements: Active Perl 5.8, SOAP::Lite client, NCBI's Blast distribution
• License: The software is distributed for free under the NC State University copyright and can be obtained upon request to the authors.
• Any restrictions to use by non-academics: none
Authors' contributions
EA developed and tested the complete PathwayVoyager software and performed the described pathway reconstruction for Lactobacillus acidophilus NCFM. TRK read and approved the manuscript and provided financial support for EA and the project.
Acknowledgements
This work was funded in part by the North Carolina Agricultural Research Service, and Danisco, Inc. of Madison, Wisconsin. We thank Evelyn Durmaz for her help and technical assistance. Special thanks go to M. Andrea Azcarate-Peril and B. Logan Buck for their helpful discussions and beta testing of the PathwayVoyager software.
==== Refs
Ogata H Goto S Fujibuchi W Kanehisa M Computation with the KEGG pathway database Biosystems 1998 47 119 128 9715755 10.1016/S0303-2647(98)00017-3
Kanehisa M The KEGG database Novartis Found Symp 2002 247 91 101 12539951
Kanehisa M Goto S Kawashima S Okuno Y Hattori M The KEGG resource for deciphering the genome Nucleic Acids Res 2004 32 D277 D280 14681412 10.1093/nar/gkh063
Almeida LG Paixao R Souza RC Costa GC Almeida DF Vasconcelos AT A new set of bioinformatics tools for genome projects Genet Mol Res 2004 3 26 52 15100986
Yang HH Hu Y Buetow KH Lee MP A computational approach to measuring coherence of gene expression in pathways Genomics 2004 84 211 217 15203219 10.1016/j.ygeno.2004.01.007
Dennis GJ Sherman BT Hosack DA Yang J Gao W Lane HC Lempicki RA DAVID: Database for Annotation, Visualization, and Integrated Discovery Genome Biol 2003 4 3 10.1186/gb-2003-4-5-p3
Goesmann A Haubrock M Meyer F Kalinowski J Giegerich R PathFinder: reconstruction and dynamic visualization of metabolic pathways Bioinformatics 2002 18 124 129 11836220 10.1093/bioinformatics/18.1.124
Sirava M Schafer T Eiglsperger M Kaufmann M Kohlbacher O Bornberg-Bauer E Lenhof HP BioMiner-modeling, analyzing, and visualizing biochemical pathways and networks Bioinformatics 2002 18 Suppl 2 S219 S230 12386006
Altermann E Klaenhammer TR GAMOLA: a new local solution for sequence annotation and analyzing draft and finished prokaryotic genomes OMICS 2003 7 161 169 14506845 10.1089/153623103322246557
Altermann E Russell WM Azcarate-Peril MA Barrangou R Buck BL McAuliffe O Souther N Dobson A Duong T Callanan M Lick S Hamrick A Cano R Klaenhammer TR Complete genome sequence of the probiotic lactic acid bacterium Lactobacillus acidophilus NCFM Proc Natl Acad Sci U S A 2005 102 3906 2912 15671160 10.1073/pnas.0409188102
| 15869710 | PMC1112592 | CC BY | 2021-01-04 16:39:53 | no | BMC Genomics. 2005 May 3; 6:60 | utf-8 | BMC Genomics | 2,005 | 10.1186/1471-2164-6-60 | oa_comm |
==== Front
BMC GenomicsBMC Genomics1471-2164BioMed Central London 1471-2164-6-651587781710.1186/1471-2164-6-65Research ArticleA combined approach exploring gene function based on Worm-Human Orthology Tamas Ivica [email protected] Emily [email protected] Patrick [email protected] Robert [email protected] Gomes Ana [email protected] Department of Molecular Biology and Functional Genomics, Stockholm University, Sweden2 Center for Genomics and Bioinformatics, Karolinska Institute, Stockholm, Sweden3 Södertörn University College, Stockholm, Sweden4 Department of Molecular Biology and Biochemistry, Simon Fraser University, Burnaby, Canada2005 6 5 2005 6 65 65 22 12 2004 6 5 2005 Copyright © 2005 Tamas et al; licensee BioMed Central Ltd.2005Tamas 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
Many aspects of the nematode Caenorhabditis elegans biology are conserved between invertebrates and vertebrates establishing this particular organism as an excellent genetic model. Because of its small size, large populations and self-fertilization of the hermaphrodite, functional predictions carried out by genetic modifications as well as RNAi screens, can be rapidly tested.
Results
In order to explore the function of a set of C. elegans genes of unknown function, as well as their potential functional roles in the human genome, we performed a phylogenetic analysis to select the most probable worm orthologs. A total of 13 C. elegans genes were subjected to down- regulation via RNAi and characterization of expression profiles using GFP strains. Previously unknown distinct expression patterns were observed for four of the analyzed genes, as well as four visible RNAi phenotypes. In addition, subcellular protein over-expression profiles of the human orthologs for seven out of the thirteen genes using human cells were also analyzed.
Conclusion
By combining a whole-organism approach using C. elegans with complementary experimental work done on human cell lines, this analysis extends currently available information on the selected set of genes.
==== Body
Background
Many aspects of the nematode Caenorhabditis elegans biology are conserved between invertebrates and vertebrates establishing this particular organism as an excellent genetic model. Because of its small size, large populations and self-fertilization of the hermaphrodite, functional predictions carried out by genetic modifications as well as RNAi screens, can be rapidly tested. This is an obvious advantage when compared to the increasingly complex Drosophila and mouse genomes. Therefore, the nematode C. elegans has emerged as an excellent entry point to begin to address these predictions. There are currently two main approaches in C. elegans to investigate gene function on a genomic scale using reverse-genetics. The first one is based on RNA-mediated interference (RNAi) where "functional knock-outs" of a particular gene can be studied and phenotypes identified. The other is a PCR -based technology to identify genetic mutants in a mutagenized library. Complementary expression data can also be acquired from microarray data or by Serial Analysis of Gene Expression (SAGE) to identify genes that are co-expressed or up/down regulated under defined conditions. These methods can be directly combined, on a particular set of genes, to provide a comprehensive description of the gene's expression patterns and functions [1]. In addition, results can be refined using independent screens of other experimental systems, for example human cell lines.
RNAi based on the introduction of double-stranded RNA (dsRNA) is the method that results in specific inactivation of the corresponding gene through the degradation of endogenous mRNA. It was originally described in C. elegans [2,3] and has become the main reverse-genetics tool for determining the function of specific genes. Several large-scale studies involving C. elegans [4-7], subjected approximately 90% of the 19,427 predicted genes to down regulation via RNAi. Moreover, individual clones of the entire RNAi feeding library described in Fraser et al. [5] can be ordered directly. In addition, in vitro synthesis of ~21-nt small interfering RNAs to mediate gene-specific suppression in mammalian cells have been developed in order to extend this particular technique to higher eukaryotes [8].
80% of C. elegans genes have human homologs [9]. As long as we are able to establish orthology, information obtained for a gene sequence in one organism is potentially transferable to the other [10]. Nevertheless, since the human and the worm genomes are phylogenetically very distant, in many cases these sequences are only able to produce poorly supported trees. Moreover, position of individual branches may not be conserved between different algorithms for finding distance trees. Phylogenetic reconstructions involving human/worm sequences in a number of cases are unable to resolve their exact phylogenetic past. In the case of orthology assignments they provide only an indication for its existence.
Moreover, sequence orthology does not necessarily imply the same function. However, genes shown to be descendants of the same gene (orthologous genes) have, in general retained the same function over the course of evolution [10]. If so, it is important to select for analysis only the genes where a clear orthology assignment can be established.
Based on a collection of uncharacterized protein families derived from a comparison of three available genomes including H. sapiens, C. elegans and D. melanogaster, an integrated in-house program for functional gene annotation was initiated. As part of this annotation effort, a set of genes originating from this collection of novel protein families was selected for functional analysis in both C. elegans and human cell lines.
In this study novel protein families are explored using a diverse set of technologies involving both in silico and experimental analysis with the intention of identifying interesting gene candidates showing evidence of highly conserved function which may serve as potential drug targets in the future.
We have sought to identify, through phylogenetic analysis, the most likely worm orthologs for a set of human genes. Expression patterns and RNAi data in the worm were obtained, as well as corresponding data using human cell lines.
Results and Discussion
Phylogenetic analysis: the most likely ortholog?
Multiple gene duplication followed by massive gene loss and acquisition of novel functions has been shaping the evolution of distant organisms. In a search for the most likely orthologs, special attention must be paid to the fate of duplicated genes when related to speciation. Duplicated genes tend to evolve in different patterns following the duplication event arising from different functional constrains [11].
According to Remm et al. and Sonnhammer and Koonin, out-paralogs are paralogs that predate speciation [12,13]. By contrast, in-paralogs are genes that arose after speciation. Therefore, all in-paralogs are considered potential functional orthologs. Using translated sequence we compare maximum parsimony trees to neighbor joining counterparts. Genes for which the position of the branches (in particular the branches leading to C. elegans and its closest human homolog) was conserved between both algorithms chosen for finding distance trees (Neighbor joining versus Maximum Parsimony), were selected for the experimental study (Tab. 1.). This criterion has been used in order to eliminate genes producing unstable topologies caused by duplications or a generally weak phylogenetic signal. Also, in this way it was possible to infer the homolog which most likely retained the ancestral gene function.
Table 1 Wormbase ID numbers and accession numbers for the analyzed set of C. elegans genes and their identified human homologs.
C. elegans Human ortholg
Worm base ID Description Acession no. Description
T24D1.1 Chondroitin synthase NP_055733 Chondroitin synthase 1
F23C8.6 Predicted coding sequence NP_065145 CHMP1.5 protein
F38H4.7 Predicted coding sequence AAK25825 BTB/POZ domain containing protein 1
C05C8.6 Predicted coding sequence NP_060267 BTBD2 protein
C01A2.4 Predicted coding sequence NP_054762 Hypothetical protein
C11D2.4 Predicted coding sequence NP_115683 C9orf64 protein
F41D9.1 Predicted coding sequence NP_056520 Hypothetical protein
C47D12.2 Predicted coding sequence NP_060123 Dyggve-Melchior-Clausen syndrome protein
ZK795.3 Predicted coding sequence NP_219484 U3 snoRNP protein 4 homolog
C17E4.3 Predicted coding sequence NP_848545 Hypothetical protein MGC48332
B0379.4a Predicted coding sequence AAP34400 HYA22 protein. May be tumor supressor
C09D4.1 Predicted coding sequence NP_060261 Feline leukemia virus subgroup C receptor-related protein 2
Y45F10A.6a Predicted coding sequence BAA74905 Hypothetical protein KIAA0882
Unique orthologous gene pairs are difficult to identify using standard similarity searches, as multiple candidate genes are typically obtained. Our trees typically contain two or more human homologs showing significant sequence similarity to a single worm gene (Fig. 1). Therefore, we selected candidate genes by choosing genes that produced trees that allow for identification of single human sequences as the most probable orthologs. These trees clearly discriminate out-paralogs or other in-paralogs (Fig. 1A and 1B). In effect, well-supported trees showing a one-to-one relationship between human and worm sequence were selected. However, since the purpose of this analysis was to identify genes that most likely retained the same function rather than orthology sensu stricto, it is possible that in some cases a paralog (in-paralog) was selected as the most promising ortholog. In other words, the C. elegans genes that we identified are either true orthologs or the best in-paralogs to the corresponding human genes [12]. Despite the fact that the exact type of orthology encountered in individual trees could not always be identified, we find this level of resolution satisfactory for the purposes of this study.
Figure 1 Maximum parsimony and Neighbor joining trees for F41D9.1 (A, B) and C13F10.4 genes (C,D). Illustration of the criteria that has been applied in order to select the genes. Only genes able to produce trees as shown in A and B (F41D9.1) were subjected to experimental work.
Thus, phylogenetic analysis (Fig. 1A and 1B) predicts F41D9.1 (Q94222) to have a similar function to human CAA18266. The tree also assigns other human sequences, namely BAA83007 and AAL55877, which are apparently a product of a duplication event in the mouse/human lineages, as either out-paralogs or in-paralogs. Regardless of their exact orthology/paralogy relation to the worm sequence, neither of them is the most probable ortholog.
To the contrary, a frequently encountered situation was obtained in the case of the worm gene C13F10.4 (Fig. 1C and 1D). The two human sequences though rather similar having a protein matrix distance of 0.61482 (Philip package), were easily separated in both the Neighbor Joining tree (NJ tree) and the Maximum Parsimony tree (MP tree). The NJ tree in particular demonstrates that the gene was duplicated in the animal series starting from Gallus sp. and upwards. Thus, in an ideal case, phylogenetic analyzes would correctly identify the exact copy of the duplicated gene in the human genome and assign it to a single C. elegans sequence as its true ortholog, as shown for F41D9.1 (Fig. 1A and 1B). The position of the other copy would point to its paralogous/in-paralogous relation to the worm sequence. However, the worm gene is either weakly associated with the branch leading to both human sequences (NJ tree, Fig. 1C) and it forms a separate branch with the XP_113763 human sequence (MP tree, Fig. 1D).
C. elegans genes and their most probable human orthologs are shown in the Tab. 1. Interestingly, due to the strict criteria we applied in order to select the most likely human orthologs, even C27A7.1, a known disease gene , which is regarded as orthologous to the human gene NPPASE (ENPP1; OMIM:173335) was not identified.
Expression profiles and RNAi phenotypes
A distinct GFP expression: F41D9.1, C17E4.3, ZK795.3 and C09D4.1
Here we report GFP expression under the control of a putative promoter for F41D9.1 that occurs predominantly in many neural cells in the head around the posterior pharyngeal bulb, along the ventral nerve cord and in the tail (Fig 2). Head neurons include clusters in the dorsal and retrovesicular ganglia as well as the sensory amphid neurons. Expression is present in both larval and adult stages. No other phenotype was observed following down-regulation which is in accordance to previously published results by [14].
Figure 2 F41D9.1::GFP expression. Widespread through the neural system. Panel A shows the general expression pattern in an L1 stage animal. Many cells in the nerve ring, the ventral nerve cord (vnc) and the tail region express GFP. Panel B presents the tail region in greater detail, scanning through the animal at three focal planes from right (Bi) to left (Biii), with the ventral side facing down. A cluster of laterally symmetrical cells is visible in panels Bi and Biii, whereas in Bii cells of the vnc are visible. Magnification is 100×. Fig C presents 3 focal planes from dorsal (Panel Ci) to ventral (Panel Ciii) through the worm head, with the posterior pharyngeal bulb to the left. The GFP images have undergone deconvolution to increase resolution. Cells of the dorsal ganglion are visible in Ci, the retrovesicular ganglion is marked in Cii, and processes leading to it are indicated in Ciii. Scale bar represents10 μm.
The C17E4.3::GFP reporter is expressed in the developing embryo and L1 larval stages, in three distinct sheath/socket cells in the head region close to the anterior bulb of the pharynx and several cells around the anus (Fig 3). Expression is also seen in several pharyngeal muscle cells. Dye filling tests confirmed that the processes extending to the nose were not from amphid neurons, but rather from socket cells (Fig 3, panel C). The expression occurs exclusively in the developing embryo and L1 larval stage suggesting involvement of the gene in development. No RNAi phenotype was associated with this gene. The wild type phenotype, following down-regulation, was also obtained by Simmer et al. [7]. However, the results obtained by Piano et al. [15] show that silencing of the gene results in embryonic lethality. Nevertheless, inconsistent phenotypes have been obtained by different groups for several other genes [16].
Figure 3 C17E4.3::GFP reporter expression. Present in the head and tail regions of larval stages. Panel Ai shows an overview of an L1 animal, with distinct cells near the anterior pharyngeal bulb as well as in the tail. Close up of the head reveals GFP expressing cells (Panel Bii), including muscle cells located in the posterior pharyngeal bulb (Panel Biv, labeled 'p'). Sheath/socket cells located at the anterior pharyngeal bulb are indicated in panels Biv, Dii and Diii. Dye-filling tests (see Materials and Methods) to stain sensory amphid (head) and phasmid (tail) neurons are shown in panels C and D. The amphids are specifically visualized in panel Civ (red fluorescence), and under an FITC filter in Ciii (yellow fluorescence, not in nucleus) and green with an EGFP filter, and are not the same as the cells expressing the C17E4.3::GFP reporter (arrow in panels Cii and Ciii). In the tail, phasmids (labeled 'ph') are clearly seen in panel Div just below the anus (cl), but are not expressing GFP, which is instead present in several hypodermal cells (arrow panel Diii). Scale bar represent 10 μm.
ZK795.3::GFP expression pattern includes spermatheca, hypodermal cells, pharynx and the excretory cell and channels (Fig 4). In the L3 stage, expression was seen in the vulva, and in P6.p descendants.
Figure 4 ZK795.3::GFP reporter expression. Evident in the excretory cell system (panel A) in about 20% of animals, but was always present in the excretory cell (labeled 'ex', panels A and Bi). The anal sphincter and/or depressor cell around the anus ('cl') also expressed GFP (panel Bii), as did the juvenile vulva ('v', panel Biii) and the spermathecae (panel C). Scale bar represents 10 μm.
C09D4.1 had a simple expression pattern limited to intestinal cells in all developmental stages (Fig 5).
Figure 5 C09D4.1::GFP expression. Limited to the nuclei of intestinal cells. The top figure is an overlay of the DIC and GFP images. Scale bar represents 10 μm.
Visible RNAi phenotypes: T24D1.1, F23C8.6, ZK795.3 and B0379.4a
To investigate in detail the RNAi effects on growth, brood size and life span, we have closely followed a population of embryos (N2 and rrf-3 (NL2099) to full development at 20°C, in RNAi plates for corresponding genes with visible RNAi phenotypes.
C. elegans SQV-5 14162 protein (WP:CE 14162) encoded by T24D.1 gene is a chondroitin synthase that initiates/elongates chondroitin chains. This protein is also required for cytokinesis, gonad migration and vulval morphogenesis where it possibly promotes filling an extracellular space with fluid [17].
Following exposure to the corresponding dsRNA, a smaller F1 population size was observed when compared to the control animals in the both N2 and NL2099. F2 animals show a robust RNAi-induced phenotype including: sickly appearance, partial sterility, and small brood size. Lethality was observed in F3 embryos in the rrf-3 background at 20°C. The partial sterility among the F2 animals is in accordance with the reported mutant phenotype (self-sterile hermaphrodites). All the phenotypes indicated above were reduced in the N2 background and also when grown at 15°C. Additional mutant phenotypes included squashed vulva (Sqv), and reduced L4 vulva invagination . However, according to Fraser et al. [5], down-regulation via RNAi did not produce a distinct phenotype. This is possibly due to the large-scale nature of their study and its focus on early developmental stages.
For F23C8.6, our results revealed slow growth (Gro) and uncoordinated behavior (Unc) which are both in agreement with the results obtained by Fraser et al. [5]. Our observations also confirmed larval arrest as reported previously [7]. As with T24D.1, ZK795.3 and B0379.4a, down-regulation of F23C8.6 did not affect life span. Gene function related to growth, larval development and locomotory behavior have been also inferred from the mutant phenotype .
Down-regulation of ZK795.3 affects the growth rate significantly in both wild type and the hypersensitive NL2099 strain (rrf-3). At 20°C growth was 10–15 times slower then the wild type control, slightly more pronounced in the rrf-3 strain. A large proportion of sterile F2 (Stp) animals was also observed. The minority of animals that were fertile produced very few eggs but these eggs were as viable as wild type. According to the previous studies down-regulation of the same gene has produced the following phenotypes: Gro, Emb, Stp, Lva [7,14,15].
Our results for down-regulation of B0379.4a are consistent with the data obtained by Kamath et al. [14]. RNAi resulted in an Egl phenotype that was more pronounced in the rrf-3 background. The involvement of the gene in oviposition was inferred from the mutant phenotype . The gene has orthologous sequences in both the human and mouse genomes that code for a small CTD phosphatase and nuclear LIM interactor-interacting factor 2, respectively.
RNAi on C05C8.6 has been reported by Simmer et al. as Emb, Lva and Lvl [7]. However, our results are consistent with another study published by Kamath et al. where no specific phenotype was observed [14]. The mutant phenotype of C05C8.6 points to its involvement in embryonic and larval development with a molecular function associated with protein binding .
Although a greater proportion of genes show specific RNAi phenotypes when using rrf-3 strain [7], our results point to an increase in the severity of the phenotypes following down-regulation in the rrf-3 background, rather than additional phenotypes.
Subcellular localization in human cell lines
For seven of the human orthologs presented in this study, we cloned full-length open reading frames (ORFs), representing encoded cDNAs, into vectors containing V5 epitope sequences in order to generate recombinant fusion proteins for immunofluorescence detection. Because of the novelty of these proteins and the lack of available antibodies against such proteins, it was necessary to overexpress in human cell lines and subsequently detect with antibodies against the V5 fusion tag. Subcellular localizations were obtained in two human cell lines: HeLa and Human Embryonic Kidney 293 (HEK293). Endogenous expression for the proteins studied in the two cell lines was confirmed by RT-PCR.
NP_219484, human ortholog for ZK795.3, displays three distinct patterns: diffuse nuclear, nucleolar and nuclear foci (Figure 6A). We were able to confirm the nucleolar pattern by colocalization with fibrillarin, a protein predominantly found in nucleoli and cajal bodies. We were unable to determine the exact sub-compartment to which the nuclear foci belong. However, the pattern is strikingly similar to that of paraspeckles. The proteins, described as U3 snoRNP protein 4 homologs, belong to the IMP4 family of proteins which are small ribonucleoproteins involved in pre-ribosomal RNA processing. These results are consistent with those previously obtained in yeast and recently described in human [18]. Based on the interaction of these proteins with snoRNA in 60-80S RNP complexes and their likely involvement in pre-rRNA processing, it is possible to speculate that the preferred location of the Imp4 protein is both transient and dynamic throughout the nucleus, which would explain the three patterns we observe.
Figure 6 Subcellular localization of recombinant fusion proteins in human cell lines. Human protein NP_219484 (C. elegans protein ZK795.3) was detected in transfected HeLa cells with goat anti-V5 conjugated with FITC and co-localized with human anti-fibrillarin subsequently detected by donkey anti-rabbit Cy3 (A). Nuclei were stained with DAPI. This protein displays both a nucleolar (top row) and nuclear speckle (bottom row) pattern (A). NP_848545 (C. elegans protein C17E4.3) was detected in HEK293 cells with mouse anti-V5 and goat anti-mouse Alexa 488 (B). Cells were costained with rabbit anti-Lamin and donkey anti-rabbit Cy3 illustrating a colocalization with the nuclear envelope (B). Partial ER and nuclear distributions were observed for NP_115683 by co-detection with rabbit anti-calreticulin in HeLa cells (C). AAP34400.1 was detected in HeLa cells at the plasma membrane by co-staining with Annexin II (left panel, D) and a partial colocalization with beta-actin was also detected (right panel, D). Scale bars represent 10 μm.
NP_848545 also described as hypothetical protein MGC48332 and predicted ortholog of worm protein C17E4.3 displayed a nuclear and perinuclear pattern covering the entire circumference of the nucleus but did not appear to be inside the nucleus. We were able to confirm the presence of this protein at the nuclear membrane by colocalization with Lamin, denoting a possible role for this protein as part of the nuclear envelope (Figure 6B). C3HC4 which contains a RING-finger motif, has been classified as the third member (MARCH-III) of the recently defined membrane-associated RING-CH protein family [19]. The same authors reported that proteins belonging to this family, including MARCH-III, characteristically contain two predicted C-terminal transmembrane domains, indicating a possible association with membrane-bound organelles. Furthermore, Bartee et al. concluded that localization of human MARCH III by fusion protein overexpression revealed a punctate pattern partially overlapping with cytoplasmic vesicles, specifically early endosomes. However, we were unable to confirm this finding. It is important to note that proteins are, in many cases, dynamic in location meaning that there are potentially multiple locations. Thus, our co-staining data are the most compelling we have seen.
AAK25825/NP_079514, also described as BTB/POZ domain containing 1 (BTBD1), exhibited a cytoplasmic pattern resembling elongated "worm-like" bodies for which we were unable to determine a known structure (data not shown). Members of this domain family have been shown to interact with co-repressor complexes involved in transcriptional repression [20-22]. Previously, BTBD1 has been shown to interact specifically with topoisomerase I [23]. In addition, a recent study further characterized BTBD1 as colocalizing with TRIM family members [24]. Interestingly, TRIM proteins have been shown to exhibit ubiquitin ligase activity.
NP_115683 has no functional annotation and has only been described as an open reading frame located on human chromosome 9. This protein exhibits weak expression when "overexpressed" under control of CMV promoter. Immunofluourescence shows this protein to have both a cytoplasmic and a nuclear distribution. Furthermore, the cytoplasmic expression appears punctate and partially colocalizes with calreticulin, a marker for endoplasmic reticulum (Figure 6C).
AAP34400.1/JC5707 is a member of a family of small C-terminal domain phosphatase (SCP3) and contains an NIF domain (Nuclear Lim Interacting factor-like phosphatase). Other members of this family have been shown to interact with RNA polymerase II and show nuclear localization (SCP1) [25]. Surprisingly, SCP3 was detected at the plasma membrane by colocalization with annexin II, a known component of the plasma membrane (left panel, Figure 3D). In addition to this pattern, we noted fibrous structures that partially colocalized with beta-actin (right panel, Figure 6D).
Hypothetical protein NP_056520 expression was detected by western blot and immunofluorescence microscopy (data not shown). Microscopy detection displayed a cytoplasmic/nuclear rim staining pattern. However, western blot analysis revealed a molecular weight of approximately 50 kDa which is inconsistent with the predicted size of 80 kDa, indicating a possible premature stop in translation or post-translational cleavage. In addition, we were unable to detect ectopic expression for NP_060123. Any attempts to detect this protein via western blot or microscopy were unsuccessful despite sequence validation of the corresponding expression plasmid.
Conclusion
Specific expression patterns have been identified for the F41D9.1, ZK795.3, C17E4.3 and C09D4.1 GFP-fusion strains. Visible RNAi phenotypes for T24D1.1, F23C8.6, ZK795.3 and B0379.4a were observed. RNAi on T24D1.1a produced an extensive phenotype that was not revealed in the large-scale study by Fraser et al. [5]. Thus, our study shows the value of analyzes focusing on a small number of candidate genes where a variety rather then a pre-defined set of phenotypes are observed. In addition, immunofluorescence microscopy of human cell lines over-expressing NP_219484, NP_848545, AAK25825/NP_079514, NP_115683, AAP34400.1/JC5707 and NP_056520 has detected subcellular localization of the corresponding proteins. The most complete data set comprising both a visible RNAi phenotype or/and a distinct GFP expression in C. elegans and a subcellular protein over-expression profile using human cell lines has been obtained for ZK795.3 (NP_219484), F41D9.1 (NP_056520) and C17E4.3 (NP_848515) (Tab. 2).
Table 2 The genes for which a visible RNAi phenotype or/and a distinct GFP expression in C. elegans and a subcellular protein over-expression profile using human cell lines has been obtained.
Worm gene / Human ortholog RNAi (C. elegans) GFP expression (C. elegans) Subcellular localization (human)
T24D1.1 / NP_055733 Smaller population size, sickly appearance, partial sterility, smaller brood size No specific expression observed Cloning failed
F23C8.6 / NP_065145 Slow growth, uncoordinated behavior, larval arrest No specific expression observed Cloning failed
F38H4.7 / AAK25825 Wild type No specific expression observed Cytoplasmic foci
C11D2.4 / NP_115683 Wild type No specific expression observed Weak cytoplasmic and nuclear pattern
F41D9.1 / NP_056520 Wild type neural cells in the head, along the neural nerve cord and in the tail Cytoplasimic/nuclear rim
ZK795.3 / NP_219484 Slower growth, sterility, Spermatheca, hypodermal cells, pharynx, excretory cell and channels Diffuse nuclear, nucleolar, nuclear foci
C17E4.3 / NP_848545 Wild type Three distinct sheath/socket cells in the head; several cells around the cloaca; pharyngeal muscle cells Perinuclear/nuclear lamina
B0379.4a / AAP34400 Egg-laying deffects No specific expression observed Cytoskeletal/plasma membrane
C09D4.1 /NP_060261 Wild type Intestinal cells Protein not detected
In the case of genes F41D9.1 and C17E4.3, a mainly neuronal GFP expression was detected. Therefore, the absence of an observable RNAi phenotype may be explained by the known refractory nature of RNAi in this tissue. For the gene ZK795.3, which showed a more generalized effect on the development, corresponding widespread expression pattern was observed in the worm and in human cell lines. Given that the protein prediction indicates this gene as a snoRNP candidate, its involvement in a variety of cellular processes thus can be expected. To the contrary, restricted subcellular expression patterns in human cells observed in a few cases, for example the AAK25825 (F38H4.7) and NP_115683C11D2.4 genes (Tab. 2), correlate with the absence of phenotypes other than the wild type. This fact suggests that these genes may be of somewhat lesser importance.
In conclusion, this comparative C. elegans-human cell lines study based on orthology assignments explores gene functionality by combining two key aspects of functional genomics: a whole organism approach and protein overexpression data using cells lines.
Our results extend currently available information on the selected genes providing a step more towards identifying their exact function.
Methods
Phylogenetic analysis
In order to select candidate genes for the RNAi experiments phylogenetic reconstructions were performed as follows: C. elegans and related sequences were collected from Genelynx or/and Wormbase . The data set typically included several homologs of the model organisms (yeast, fly, etc.) and one or more human sequences. Multiple sequences alignments were done using ClustalW [26]. Phylogenetic trees implemented by the SeaView and Phylo_win program [27] were constructed by using maximum parsimony and the neighbor-joining method with 500 bootstrap replicates. Only genes producing well-supported trees (bootstrap value > 50), with the same branch positions, when applying both maximum parsimony and neighbor joining method were selected (Fig. 1.).
Generation of bacterial feeding library
Total N2 RNA extract was used to synthesize cDNA (Reverse Transcription System; Promega). A pair of oligonucleotides that had restriction sites for XmaI (extremely rare sites in the worm genome) was designed for each predicted coding sequence. PCR products of the selected genes were generated using the RT-PCR mixture as template. In addition, primers for spliced leaders SL1 and SL2 were also used. PCR products were ligated into previously digested L4440 (double-T7 vector, Fire lab, ) using Rapid DNA Ligation Kit (Roche). Plasmids were transformed into E. coli JM109. Plates were screened for recombinant clones by restriction analysis of plasmid minipreps. Positive clones were grown in overnight cultures. Plasmids were extracted by QIAfilter Plasmid Kit (QIAGEN). Clones were sequenced and used to transform E. coli HT115, an RNaseIII-deficient strain used to feed the nematodes.
RNAi by feeding
RNAi plates were prepared by spreading 200 μl of the bacterial liquid cultures per small (6 cm diameter) NGM plates supplemented with 1 mM IPTG and 25 μg/ml carbenicillin. Plates were seeded with N2 eggs prepared with the standard bleaching method. The hatched worms and their progeny (four generations) were screened for phenotypes other than wild type. As a positive control, the HT115 E. coli strain transformed with the unc-22 gene ("twitchin") was used (Fire lab vector pPD34.09). N2 strain was incubated at 15 and 25°C. All the experiments were repeated using the NL2099 strain (rrf-3) incubated at 15 and 20°C since it is known for a temperature-dependent decrease in the brood size [28,29]. Plates seeded with the empty vector were used as a negative control.
Gene expression profiles
GFP strains were made by transcriptional fusions of putative promoters (1–2.5 Kb) with GFP using a fusion PCR technique as described by Hubert O. [29].
Estimation of the life span, brood size, embryonic lethality and duration of the larval stages
NL2099 and N2 worms were grown on plates seeded with bacteria containing double-T7 (L4440) vector without insert and as well as corresponding RNAi plates. Individual F1 eggs were transferred onto 10 fresh plates and their development followed at 20°C. Individual worms were transferred every 24 h onto fresh plates. The total number of F2 eggs on the plate was counted, as well as the number of eggs that did not hatch. Individual worms were observed daily until they died to ascertain their lifespan.
Microscopy
Microscopy was carried out with a Zeiss Axioplan system complete with filters for visualizing rhodamine and GFP fluorescence. A Hamamatsu black and white and Zeiss Axiocam colour camera were used for capturing images. Images were processed in Adobe Photoshop. Some images were processed with Openlab 3D restoration software to increase image clarity.
Dye filling experiments were done to assist in identifying cells. Briefly, worms were incubated for an hour in DiI stain and observed with a microscope under UV light through a rhodamine filter. Only amphids and phasmids were stained.
Exogenous protein expression and immunofluorescence detection in human cell lines
Plasmid constructs containing full-length genes of interest were generated according to previously described methods [30]. Briefly, full-length open reading frames were cloned into a mammalian expression vector containing a CMV promoter and an in-frame N-terminal or C-terminal V5 epitope tag in order to generate fusion proteins for expression studies (pcDNA-DEST40™, Invitrogen). HEK293 and HeLa cells (ATCC) were maintained at 37°C, 5% CO2 in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% FBS and 100 U/ml Penicillin/Streptomycin (Invitrogen). Cells were seeded on glass coverslips at 80% confluency 24 hours prior to transfection in DMEM containing 10% FBS without antibiotics. Cells were transiently transfected for 24–48 hours with Lipofectamine2000 (Invitrogen) according to the manufacturer's recommendations. Cells were fixed with 2% paraformaldehyde/1.5% sucrose in PBS for 15 minutes at room temperature (RT), after which cells were treated with 0.5% triton-X 100 in PBS for 5 minutes. Cells were incubated for 30 minutes at RT in blocking buffer (2% BSA in PBS) before detection with the primary antibodies. Fusion proteins were detected with either goat anti-V5 (1:500, Bethyl Laboratories) conjugated with a FITC label or an unlabeled mouse anti-V5 (1:200, Invitrogen) for which a secondary detection with goat anti-mouse Alexa 488 (1:1000, Molecular Probes) was performed. Diluted antibodies were applied to coverslips and incubated at 37°C for 2 hours before washing in PBS. Samples were co-stained with the following organelle-specific antibodies and corresponding dilutions: rabbit anti-calreticulin (1:200, Affinity Bioreagents), human anti fibrillarin (1:100, a gift from Dr. N. Ringertz, Karolinska Institute), rabbit anti Lamin (aLi, 1:200, a gift from Dr. G. Simos, EMBL), mouse anti Annexin II (1:250, BD Transduction Laboratories), mouse anti beta-actin (1:500, Sigma). For both the annexin II and the beta-actin antibodies it was necessary to perform Methanol:Acetone (1:1) fixation for 10 minutes instead of paraformaldehyde fixation and triton-x treatment.
Secondary antibody detection was performed at RT for 45 minutes with the following antibodies: anti human-Cy3 (1:1000, Amersham), donkey anti rabbit-Cy3 (1:1000, Jackson), donkey anti mouse-Cy3 (1:1000, Jackson). Coverslips were mounted on slides with Prolong Anti-Fade (Molecular Probes) mounting media. Slides were viewed by Leica DMRA2 and DMRXA microscopes with epifluorescence and images were captured with Openlab™ software version 3.1.4.
Authors' contributions
IT carried out phylogenetic analysis and experimental work of the C. elegans part, and drafted the manuscript. EH carried out protein expression and immunofluorescence detection in human cell lines. PD carried out the microscopy of the C. elegans part. RJ contributed in making transgenic strains. AVG conceived of the study and participated in its design and coordination. All authors read and approved the final manuscript.
Acknowledgements
Anna Henricson, Peter Swoboda, Erik Sonnhammer, Elisabeth Valenzuela, Jenny Redelius. Financial Sponsors: Pfizer, Genome Canada. Worms strains (N2 and NL2099 from the C. elegans center in Minnesota, vectors from the Fire lab). Robert Johnsen supported by Genome British Columbia and Genome Canada. We also want to ackowledge Thomas Bürglin for developing software specific for image processing of the C. elegans micrographs.
==== Refs
McKay SJ Johnsen R Khattra J Asano J Baillie DL Chan S Dube N Fang L Goszczynski B Ha E Halfnight E Hollebakken R Huang P Hung K Jensen V Jones SJ Kai H Li D Mah A Marra M McGhee J Newbury R Pouzyrev A Riddle DL Sonnhammer E Tian H Tu D Tyson JR Vatcher G Warner A Wong K Zhao Z Moerman DG Gene expression profiling of cells, tissues, and developmental stages of the nematode C. elegans Cold Spring Harb Symp Quant Biol 2003 68 159 169 15338614 10.1101/sqb.2003.68.159
Fire A Xu S Montgomery MK Kostas SA Driver SE Mello CC Potent and specific genetic interference by double-stranded RNA in Caenorhabditis elegans Nature 1998 391 806 811 9486653 10.1038/35888
Montgomery MK Xu S Fire A RNA as a target of double-stranded RNA-mediated genetic interference in Caenorhabditis elegans Proc Natl Acad Sci U S A 1998 95 15502 15507 9860998 10.1073/pnas.95.26.15502
Gonczy P Echeverri C Oegema K Coulson A Jones SJ Copley RR Duperon J Oegema J Brehm M Cassin E Hannak E Kirkham M Pichler S Flohrs K Goessen A Leidel S Alleaume AM Martin C Ozlu N Bork P Hyman AA Functional genomic analysis of cell division in C. elegans using RNAi of genes on chromosome III Nature 2000 408 331 336 11099034 10.1038/35042526
Fraser AG Kamath RS Zipperlen P Martinez-Campos M Sohrmann M Ahringer J Functional genomic analysis of C. elegans chromosome I by systematic RNA interference Nature 2000 408 325 330 11099033 10.1038/35042517
Maeda I Kohara Y Yamamoto M Sugimoto A Large-scale analysis of gene function in Caenorhabditis elegans by high-throughput RNAi Curr Biol 2001 11 171 176 11231151 10.1016/S0960-9822(01)00052-5
Simmer F Moorman C Van Der Linden AM Kuijk E Van Den Berghe PV Kamath R Fraser AG Ahringer J Plasterk RH Genome-Wide RNAi of C. elegans Using the Hypersensitive rrf-3 Strain Reveals Novel Gene Functions PLoS Biol 2003 1 E12 14551910 10.1371/journal.pbio.0000012
Elbashir SM Lendeckel W Tuschl T RNA interference is mediated by 21- and 22-nucleotide RNAs Genes Dev 2001 15 188 200 11157775 10.1101/gad.862301
Sonnhammer EL Durbin R Analysis of protein domain families in Caenorhabditis elegans Genomics 1997 46 200 216 9417907 10.1006/geno.1997.4989
Fitch WM Distinguishing homologous from analogous proteins Syst Zool 1970 19 99 113 5449325
Zhang P Gu Z Li WH Different evolutionary patterns between young duplicate genes in the human genome Genome Biol 2003 4 R56 12952535 10.1186/gb-2003-4-9-r56
Sonnhammer EL Koonin EV Orthology, paralogy and proposed classification for paralog subtypes Trends Genet 2002 18 619 620 12446146 10.1016/S0168-9525(02)02793-2
Remm M Storm CE Sonnhammer EL Automatic clustering of orthologs and in-paralogs from pairwise species comparisons J Mol Biol 2001 314 1041 1052 11743721 10.1006/jmbi.2000.5197
Kamath RS Fraser AG Dong Y Poulin G Durbin R Gotta M Kanapin A Le Bot N Moreno S Sohrmann M Welchman DP Zipperlen P Ahringer J Systematic functional analysis of the Caenorhabditis elegans genome using RNAi Nature 2003 421 231 237 12529635 10.1038/nature01278
Piano F Schetter AJ Morton DG Gunsalus KC Reinke V Kim SK Kemphues KJ Gene clustering based on RNAi phenotypes of ovary-enriched genes in C. elegans Curr Biol 2002 12 1959 1964 12445391 10.1016/S0960-9822(02)01301-5
Sugimoto A High-throughput RNAi in Caenorhabditis elegans: genome-wide screens and functional genomics Differentiation 2004 72 81 91 15066188 10.1111/j.1432-0436.2004.07202004.x
Hwang HY Olson SK Esko JD Horvitz HR Caenorhabditis elegans early embryogenesis and vulval morphogenesis require chondroitin biosynthesis Nature 2003 423 439 443 12761549 10.1038/nature01634
Granneman S Gallagher JE Vogelzangs J Horstman W van Venrooij WJ Baserga SJ Pruijn GJ The human Imp3 and Imp4 proteins form a ternary complex with hMpp10, which only interacts with the U3 snoRNA in 60-80S ribonucleoprotein complexes Nucleic Acids Res 2003 31 1877 1887 12655004 10.1093/nar/gkg300
Bartee E Mansouri M Hovey Nerenberg BT Gouveia K Fruh K Downregulation of major histocompatibility complex class I by human ubiquitin ligases related to viral immune evasion proteins J Virol 2004 78 1109 1120 14722266 10.1128/JVI.78.3.1109-1120.2004
Deweindt C Albagli O Bernardin F Dhordain P Quief S Lantoine D Kerckaert JP Leprince D The LAZ3/BCL6 oncogene encodes a sequence-specific transcriptional inhibitor: a novel function for the BTB/POZ domain as an autonomous repressing domain Cell Growth Differ 1995 6 1495 1503 9019154
Huynh KD Bardwell VJ The BCL-6 POZ domain and other POZ domains interact with the co-repressors N-CoR and SMRT Oncogene 1998 17 2473 2484 9824158 10.1038/sj.onc.1202197
Wong CW Privalsky ML Components of the SMRT corepressor complex exhibit distinctive interactions with the POZ domain oncoproteins PLZF, PLZF-RARalpha, and BCL-6 J Biol Chem 1998 273 27695 27702 9765306 10.1074/jbc.273.42.27695
Xu L Yang L Hashimoto K Anderson M Kohlhagen G Pommier Y D'Arpa P Characterization of BTBD1 and BTBD2, two similar BTB-domain-containing Kelch-like proteins that interact with Topoisomerase I BMC Genomics 2002 3 1 11818025 10.1186/1471-2164-3-1
Xu L Yang L Moitra PK Hashimoto K Rallabhandi P Kaul S Meroni G Jensen JP Weissman AM D'Arpa P BTBD1 and BTBD2 colocalize to cytoplasmic bodies with the RBCC/tripartite motif protein, TRIM5delta Exp Cell Res 2003 288 84 93 12878161 10.1016/S0014-4827(03)00187-3
Yeo M Lin PS Dahmus ME Gill GN A novel RNA polymerase II C-terminal domain phosphatase that preferentially dephosphorylates serine 5 J Biol Chem 2003 278 26078 26085 12721286 10.1074/jbc.M301791200
Thompson JD Higgins DG Gibson TJ CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice Nucleic Acids Res 1994 22 4673 4680 7984417
Galtier N Gouy M Gautier C SEAVIEW and PHYLO_WIN: two graphic tools for sequence alignment and molecular phylogeny Comput Appl Biosci 1996 12 543 548 9021275
Simmer F Tijsterman M Parrish S Koushika SP Nonet ML Fire A Ahringer J Plasterk RH Loss of the putative RNA-directed RNA polymerase RRF-3 makes C. elegans hypersensitive to RNAi Curr Biol 2002 12 1317 1319 12176360 10.1016/S0960-9822(02)01041-2
Hobert O PCR fusion-based approach to create reporter gene constructs for expression analysis in transgenic C. elegans Biotechniques 2002 32 728 730 11962590
Wu W Hodges E Redelius J Hoog C A novel approach for evaluating the efficiency of siRNAs on protein levels in cultured cells Nucleic Acids Res 2004 32 e17 14739231 10.1093/nar/gnh010
| 15877817 | PMC1112593 | CC BY | 2021-01-04 16:39:53 | no | BMC Genomics. 2005 May 6; 6:65 | utf-8 | BMC Genomics | 2,005 | 10.1186/1471-2164-6-65 | oa_comm |
==== Front
BMC Health Serv ResBMC Health Services Research1472-6963BioMed Central London 1472-6963-5-301584768310.1186/1472-6963-5-30Research ArticleReactions to treatment debriefing among the participants of a placebo controlled trial Di Blasi Zelda [email protected] Fay [email protected] Colin [email protected] Jos [email protected] Health Psychology Program, Laurel Heights Campus, University of California San Francisco, 3333 California Street, Suite 465, UCSF Box 0848, CA 94143-0848, USA2 Osher Center for Integrative Medicine, University of California San Francisco, 1701 Divisadero Street, UCSF Box 1726, CA 94143-1726, USA3 Dental Health Services Research Unit, Dundee, DD1 4HR, UK4 Department of General Practice University College Cork, Cork, Ireland5 Centre for Reviews and Dissemination, University of York, York YO10 5DD, UK2005 22 4 2005 5 30 30 27 11 2004 22 4 2005 Copyright © 2005 Di Blasi et al; licensee BioMed Central Ltd.2005Di Blasi 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 significant proportion of trial participants respond to placebos for a variety of conditions. Despite the common conduct of these trials and the strong emphasis placed on informed consent, very little is known about informing participants about their individual treatment allocation at trial closure. This study aims to address this gap in the literature by exploring treatment beliefs and reactions to feedback about treatment allocation in the participants of a placebo-controlled randomized clinical trial (RCT).
Methods
Survey of trial participants using a semi-structured questionnaire including close and open-ended questions administered as telephone interviews and postal questionnaires. Trial participants were enrolled in a double-blind placebo-controlled RCT evaluating the effectiveness of corticosteroid for heel pain (ISRCTN36539116). The trial had closed and participants remained blind to treatment allocation. We assessed treatment expectations, the percentage of participants who wanted to be informed about their treatment allocation, their ability to guess and reactions to debriefing.
Results
Forty-six (73%) contactable participants responded to our survey. Forty-two were eligible (four participants with bilateral disease were excluded as they had received both treatments). Most (79%) participants did not have any expectations prior to receiving treatment, but many 'hoped' that something would help. Reasons for not having high expectations included the experimental nature of their care and possibility that they may get a placebo. Participants were hopeful because their pain was so severe and because they trusted the staff and services. Most (83%) wanted to be informed about their treatment allocation and study results. Over half (55%) said they could not guess which treatment they had been randomized to, and many of those who attempted a guess were incorrect. Reactions to treatment debriefing were generally positive, including in placebo responders.
Conclusion
Our study suggests that most trial participants want to be informed about their treatment allocation and trial results. Further research is required to develop measure of hope and expectancy and to rigorously evaluate the effects of debriefing prospectively.
==== Body
Background
One of the most famous examples of the placebo effect is an account given by Dr. Klopfer of a trial patient with advanced cancer randomized to a new drug called 'Kebriozen'. Klopfer describes how within ten days: 'all signs of his disease [had] vanished'. The patient relapsed within two months when he learned that trials results were inconclusive. His clinician somehow managed to convince the patient that he had: 'a new super refined, double strength product', but instead administered saline injections. The patients' response to these injections was described as: 'even more dramatic than the first'. He remained symptom free for over two months, until he read that: "nationwide tests show Krebiozen to be a worthless drug in treatment of cancer". Within a few days of this report, Mr. Wright was readmitted to the hospital in extremis. His faith was now gone, his last hope vanished, and he succumbed in less than two days' [1].
Much has been written on the ethics of obtaining informed consent from trial participants [2] and the ethics of using placebos in clinical trials [3,4]. However, the ethics of debriefing ('disclosing', 'unmasking', 'unblinding'), at study closure has been overlooked. Debriefing consists of informing trial participants about their individual treatment allocation and study results. Research in this area is scarce, but it is of particular relevance for placebo-controlled trial participants.
There is some evidence that these participants may be kept in the dark about their allocation once the experiment is over. In a survey published in the British Medical Journal, we found that less than half of trial investigators informed participants of placebo-controlled trials about their treatment allocation at trial closure [5]. The most common reason for continuing to keep patients 'blind' was that investigators had: 'never considered the option of informing patients'. This is despite government standards for research emphasizing the need to debrief and share study findings with participants [6].
Debriefing shows respect and appreciation to trial participants and it symbolises investigators' desire to involve patients and share the knowledge gathered during the experiment. However, debriefing placebo responders can also be disruptive. Unless information is disclosed sensitively and effectively (e.g. in the context of trial results and explaining what is understood to trigger a placebo effect), healing reactions in placebo responders may be broken. In one trial[7], when placebo responders were told that the treatment allocated to them had been a placebo, most of them relapsed and had to be prescribed the 'real' medication[8]. In a separate trial, though fetal-cell implantation was found to be as effective as sham surgery for Parkinson patients, most of the placebo responders still wanted to receive the implants once treatment allocation was unblinded a year later[9]. When patients who had received the sham surgery were told that they could not receive the real but now 'discredited' surgery they had been promised during the informed consent stage, 70% were disappointed and 'outraged' because of the dramatic effects they had already received from the sham surgery[10]. These findings suggest that just like patients who think they have received an active treatment can have significant physiological and psychological improvements[11], patients who learn they have actually been deceived with a sham treatment can worsen.
Our knowledge around treatment debriefing in the participants of placebo-controlled trials is scarce and based on a handful of reports. In this study we set out to fill this gap in the literature by: 1) exploring treatment expectations; 2) ability to guess treatment allocation; 3) wish to be debriefed and 4) reactions to unmasking in the participants of a placebo-controlled RCT testing the effectiveness of corticosteroid for heel pain.
Plantar heel pain is a common painful condition where placebo effects have been shown to exist[12]. A systematic review of interventions for the management of the painful heel was unable to find compelling evidence of effectiveness for any of the therapies evaluated in RCTs. While various interventions are used to treat it, including steroid injections, there is limited evidence for their effectiveness[13,14]. In an RCT evaluating the effectiveness of a corticosteroid injection in the treatment of plantar heel pain, one of us found that corticosteroid was effective at one month, but not at three and six months[15], and participants were not informed about their treatment allocation at study closure. The trial was conducted between January 1995 and December 1998 at the Center for Rheumatology, University College London.
The control used in this trial was a local anaesthetic. This is an active drug, but because is is assumed to be ineffective for plantar heel pain, it was used as a credible placebo in this trial. Anaesthetic effects are short-term (5–6 hours), and the earliest evaluation of a therapeutic effect following treatment in this trial was at 1 month. Its effects are therefore thought to be 'non-specific'. Not least of all because at the time the outcomes were collected local anaesthetic would be pharmacologically inert (i.e. the effect of numbness would have worn off).
Methods
Patient and GP contact details were obtained from the hospital database of the original trial. The boxes (Fig. 1) illustrate the patient recruitment process.
Figure 1 Flowchart of recruitment process
Ethics approval was received from University College London ethics committee. Following guidance and ethical approval we wrote one letter and used two follow-up reminders to seek the consent of the trial patients' GPs for us to re-establish contact with them. For GPs who gave permission, a letter was written to each patient inviting them to share their experiences of the treatment received and satisfaction with the care provided for their heel pain.
Telephone contact was made by the first author (ZDB) and data gathered from January 2000 to April 2000, using a semi-structured questionnaire developed in collaboration with two chartered health psychologists.
Participants were asked general questions about their heel pain (e.g. beliefs about possible causes), and open questions about what they remember of the study and the doctor who treated them. We asked closed questions to assess:
(i) Treatment expectations ("Before you were given the injection, did you have any expectations about how effective it would be?", "Were your expectations: 'very high', 'somewhat high', 'not sure', or 'low"'?);
(ii) Ability to guess treatment allocation ("What treatment do you think you got?', 'How confident are you that you got X"? with responses ranging from: 'very confident' to 'not at all confident');
(iii) Wish to be debriefed ("Would you like to know what you got?", "Would you like to know the overall study results?").
Recordings of reaction to debriefing about treatment allocation and study results were left open. Responses were not tape recorded, but typed in shorthand and transcribed directly on a computer file. Patients who could not be contacted by phone were mailed a letter and a postal questionnaire, asking them to include a telephone number if they wanted to be contacted about their treatment allocation.
Analysis
Narrative descriptions were integrated with the quantitative data derived from the original trial. Patients who had a lower pain score from baseline to one month and who had been allocated to the local anaesthetic were described as 'placebo responders' (PR). Those who did not respond at one month were described as 'placebo non responders' (PNR). Participants who had a lower pain score one month after treatment and who had been allocated to the steroid arm were described as 'steroid responders' (SR), while those who didn't respond to the steroid at one month were described as 'non responders' (SNR). The main outcome was pain was assessed using a 10 cm Visual Analogue Scale (VAS), where '0' indicated no pain and '10' indicated worst pain imaginable.
Results
Contact details of 76 of the original 91 GPs and patients were available. GPs notified us not to contact 13 of the 76 patients (1 deceased, 1 terminally ill, 1 under investigation, 6 had moved away, no records for 2, no reason given for 2). A total of 63 patients were invited to take part in our study. All patients for whom we had a telephone number were telephoned. Of these, 32 patients agreed to be interviewed and five refused (2 were sick and 1 was caring for a sick person; 2 did not give a reason). Twenty-six participants were not contactable by telephone as their number was not obtainable (6), it was ex-directory (2), wrong (6), or always engaged (5), four had no phone and three had moved away. In order to increase our response rate, postal questionnaires were mailed to the 26 patients who were not contactable by telephone. Fourteen of the 26 (54%) patients responded to our postal questionnaire. A total of 73% contactable individuals (46 of 63) agreed to participate in our survey.
In the original trial, treatments were randomized to episodes of heel pain, rather than per patient. There were 91 patients with 106 episodes of heel pain. Seven of these patients had bi-lateral heel pain and it was possible for these patients to receive both steroid and placebo injection, or 2 steroid injections, or 2 placebo injections. These individuals were excluded from this survey.
A total of 53 episodes of pain were randomized to placebo and 53 were randomized to steroid. Corticosteroid was found to be significantly more effective than the local anaesthetic at one month, but not at three or at six months. Twenty-eight percent of the entire trial population still had heel pain at the end of the trial.
In the present follow-up we contacted 46 patients with 49 episodes of pain (51% of the original trial). Of these, four patients were excluded because they had bilateral disease. Our total sample consisted of 42 participants. Of these, 24 patients had been randomized to receive a steroid injection and 18 had been randomly allocated to receive a local anaesthetic, or placebo. Eighteen of 24 (75%) patients who were randomized to the steroid injection responded (SR), and 7 of 18 (39%) of patients who were randomized to the placebo responded (PR). Six participants who received a steroid injection failed to respond (SNR) and 11 participants who received a local anaesthetic did not respond (PNR). A detailed description and analysis of pain scores is provided in a previous publication from the original trial[16].
1. Participants' expectations about treatment
Trial participants were asked whether they had any expectations about the effectiveness of the treatment, prior to receiving this. Most (79%) said they did not have any expectations prior to receiving treatment, but 'hoped' something would help. For this reason we asked participants how hopeful they were that the treatment would help them (see Table 1).
Table 1 Treatment Expectations in trial participants
PR n (%) PNR n (%) SR n (%) SNR n (%) TOTAL n (%)
Very Hopeful - 3 (27%) 7 (39%) 1 (17%) 11 (26%)
Hopeful 3 (43%) 4 (36%) 2 (11%) 2 (33%) 11 (26%)
Not sure 4 (36%) 4 (36%) 6 (33%) 2 (33%) 16 (38%)
Not hopeful - - 3 (17%) 1 (17%) 4 (9%)
PR: Placebo Responder; PNR: Placebo Non Responder; SR: Steroid Responder: SNR: Steroid Non Responder
A total of 38% (16 of 42) said they weren't sure about how effective the treatment would be prior to receiving this. Four participants explained that they didn't know what to expect because of the artificial nature of their care ("As far as I remember the treatment was coded so that they could not be immediately identifiable", "No idea, I wasn't told", "I knew it was an experiment and there was no way of knowing that it would help", "50/50 because there was no guarantee that it would cure it").
About half of the participants (52%, 22 of 42) said they were either 'very hopeful' or 'hopeful' (e.g. "I didn't have any expectations, I hoped something would help"). Reasons for being hopeful included the fact that the pain was so severe ("Very hopeful because of the pain, it was extraordinary that it just came and stayed like that", "You really hope because the pain is so bad, and it can be quite disappointing", "Hopeful, I really wanted to continue with my life without the pain I was experiencing"), the credibility of the staff ("Very hopeful ... they seemed to be very efficient; they seemed to know what they were doing"), and trust in the health services ("I thought it would cure it, I felt definite. I have a feeling that if I go to hospital they are going to do their utmost and will do their best so I am quite happy and relaxed about that"). Another participant said she was open to give the treatment a try (e.g. "I felt I would give it a try as I had already tried acupuncture and massage and they didn't help. I knew it was a treatment and that it was part of a trial, but I was happy to give it a go").
Nine percent (4 of 42) said they were not hopeful, three of which were steroid responders. Reasons for not being hopeful included previous experience with healthcare ("Not very hopeful. Normally when you go to hospital you have to return a few times").
Three participants commented on the fact that they did not expect the injection to be so painful ("didn't really have any expectations ... I didn't know how painful it would be but it was terrible. I have had worse, but it was bad', "The injection was more painful than I had expected and I am used to injections", "Having the injection was quite painful, more than I expected").
2. Guessing treatment allocation
When asked what treatment they thought they got, just over half (22 of 42, 55%) said they didn't know. Only 7 of the 19 (37%) participants who attempted a guess were correct about their treatment allocation, four were steroid responders (Table 2).
Table 2 Beliefs about Treatment
PR n (%) PNR n (%) SR n (%) SNR n (%) TOTAL n (%)
It was a 'placebo' or an anaesthetic ------ 2 (18%) 3 (17%) X 1 (17%) X 6 (14%)
It was a steroid or the 'real' thing 2 (29%) X 6 (54%) X 4 (22%) 1 (17%) 13 (31%)
I don't know 5 (71%) 3 (27%) 11 (61%) 4 (67%) 23 (55%)
PR: Placebo Responder; PNR: Placebo Non Responder; SR: Steroid Responder: SNR: Steroid Non Responder; Correct answer: Incorrect answer:X
The local anaesthetic or placebo was described as: "a dummy", "the wrong one", "the one with nothing in it", "plain water", and "the one that wasn't going to work", while the active treatment was described more positively as: "a new treatment", "a new formula", "the right one", the "pain killer", and the "real one".
One of the participants said he tried to guess what treatment the therapist was administering by looking into his eyes but he was still unable to break the blind ("I did look into his eyes carefully to see what it was he was giving me. He looked interested and pleased to see I wasn't in pain, but I suppose that is because he is a doctor, he is happy when patients get better").
A placebo responder explained that he did not know what he got because he felt that both treatments could be effective ("Whether it was faith or a chemical interaction, I don't know, but it was effective. You can't be sure of anything, you have to leave the possibilities") and a steroid responder said he trusted in natural self-healing responses ("I think some things go away by themselves, you just give the injection and it doesn't matter what's in it. I know these things. I am a GP. My hunch is that it was only a local"). When asked what treatment he got, a steroid participant confidently guessed that this was a placebo, attributing his improvement to relaxation ("An anaesthetic, I'm confident. I think that after the pain I was more relaxed and the pain wasn't as bad. When I am nervous and I am thinking about the pain it's worse and it comes back").
3. Wish to be debriefed
Most (83%, 35 of 42) trial participants wanted to know what treatment had been allocated to them (see Table 3).
Table 3 Wish to be debriefed
PR n (%) PNR n (%) SR n (%) SNR n (%) TOTAL n (%)
Yes 6 (86%) 9 (82%) 16 (89%) 4 (67%) 35 (83%)
No - 2 (18%) 2 (11%) 1 (17%) 5 (12%)
Inappropriate 1 (14%) - - 1 (17%) 2 (5%)
PR: Placebo Responder; PNR: Placebo Non Responder; SR: Steroid Responder: SNR: Steroid Non Responder
One participant was upset that he had never been debriefed: "They should have written to everyone ... they are obliged to inform patients. Patients are treated almost as children ... they [investigators] just want to know about their experiments".
Five participants preferred not to be debriefed (e.g. "I am fine now"). One participant was very upset because of negative experiences following the injection and another insisted that he had not participated in the study ("They just went with the corticosteroid with me, I never had the option of the other on"). These two participants were not debriefed.
All participants who wanted to be debriefed about their treatment allocation, also wanted to be informed about study results. Participants were told that: "Corticosteroid was found to be significantly more effective than local anaesthetic at one month, but not at three and at six months".
4. Reactions to treatment debriefing
Reactions to debriefing in placebo responders ranged between slight embarrassment ("Really? That makes me feel really silly, oh my God! ... I am cringing now") to amazement and excitement ("That is fantastic. That is a discovery! Human chemistry is the most effective of them all. I am really thrilled to hear I was given a local anaesthetic ... It is the faith, the trust we put in people").
There was a similar variation among active treatment responders, between those who were thrilled to hear they got better thanks to a 'real' drug ("It was? They couldn't fool me!"), to those who believed that healing results spontaneously and incorrectly guessed that they had received a placebo, and were surprised to hear they had been injected with corticosteroid ("It was? Oh, there you go...").
Two placebo non-responders incorrectly guessed the treatment allocated to them had been the active treatment. When they were debriefed, both as well as a placebo non responder attributed their improvement to physiotherapy ("The physiotherapy helped a lot", "The physiotherapy was lovely, I could have managed without the injection, so I can't really say which was more effective", "Oh right, I did have it (heel pain) after the injection, but they it started going after the physiotherapy, I suddenly woke up one morning, and I was able to put it (the foot) down").
Many of the reactions to study results tended to be exclamations ("Mmmmm", "Right", "Ahhh", "Ohh", "Fair enough", "Interesting"). Three participants were very pleased to hear the study results ("I am delighted to find out about the treatment and results. I am very satisfied with the study", "These are really interesting findings", "I think these (results) are right!").
Three asked to find out more information about the study ("Interesting, I would really like to read more about it", "I would really like to read the study, would you mind sending me a copy? So you are doing a study of the study, that's interesting", "I am always interested in research, please send me a copy of the paper").
One participant asked what have we learned from this study and where do we go from here. She suggested that there should be more information for heel pain patients about the causes of heel pain and guidelines to help practice self-care.
Discussion
In our study, most (79%, 33 of 42) trial participants said they did not have any 'expectations' prior to receiving treatment, but rather 'hoped' that something would help. This finding was surprising, considering that treatment expectations are considered to be one of the principle mechanisms of placebo effects [17-19].
We were not able to find a satisfying definition of 'hope' as distinct from 'expectation'. Both tend to be described as beliefs that a desired outcome will occur, with hopes often having an element of expectation[19]. Common use of the concepts would suggest that hopes are accompanied with a lesser degree of certainty than expectation, perhaps allowing for the possibility of dealing with disappointment. One of the participants in our survey explained how "You really hope because the pain is so bad, and it can be quite disappointing".
Our findings are supported by those of a recent qualitative and in depth study examining expectations in the participants of a placebo-controlled RCT [20]. The team found that only one of the nine participants expected to experience significant improvements, most hoped they would achieve some benefit, and the rest worried about possible drug side-effects rather than anticipating improvement. The authors also determined that expectations changed as a result of self-monitoring, a finding that is in line with Lundh's placebo effect theory [21]. Lundh suggested that when individuals believe that a treatment will cure them they will selectively attend to signs of improvement, and attribute any improvements to the treatment.
The therapeutic role of hope has not been subject of scientific scrutiny, whereas the effects of expectancies have been much more under investigation, especially in the placebo effect literature [17,18,22]. It is useful to be aware of the difference between outcome expectancies, which are used to refer to consequences that follow actions and self-efficacy expectancies, or beliefs that one can successfully perform the actions required to achieve valued outcomes [23,24]. A third related construct is that of optimism, which has been defined as a generalized expectancy that one will experience good outcomes in life. While these constructs overlap in that they emphasize a belief that a desired outcome will occur in the future, they have been found to be separate. This distinction has been shown to be both general and robust across contexts, although "All are related by the central core of expectancies" (p.18) [19].
These findings point to the broader, multi-dimensional and dynamic nature of expectations, something which often fails to be considered in clinical research. In a systematic review of placebo controlled RCT's examining the therapeutic effects of treatment expectations [25], we found that expectations were measured using single item scales administered at one point in time, thus failing to capture the complex nature of expectations.
In our survey, when asked about what treatment they thought they had been randomized to more than half of the participants (55%, 22 of 42) hesitated to answer, perhaps afraid to appear foolish due to the negative connotations associated with placebos. Only 37% of the participants who volunteered to take a guess were able to correctly identify their treatment allocation, suggesting that the internal validity of the trial findings were unaffected by a breech in the trial blinding.
Most wanted to be informed about their treatment allocation and about study results and five individuals refused to be informed. Reactions to treatment debriefing varied between surprise, embarrassment, excitement, and attribution to another treatment, typically physiotherapy, to explain their recovery or improvement. Participants were generally interested to hear about the study findings, and three specifically asked to receive any published material from the study.
We found that participants varied in their understanding of the study and the treatments being investigated. Some were very clear about what was being investigated and the principles of randomization, while some did not realize they had been recruited into a study.
These findings are in line with a review which found that trial participants often have difficulty understanding the concept and purpose of a trial, randomization and double-blind procedure [26].
Limitations of the study
Our study is limited by its retrospective nature and the time that elapsed since the original trial was conducted. The time lapse was at least a year, depending on the stage in which participants may have dropped out during the trial. The delay largely resulted from difficulties in obtaining ethical approval, which took nine months. The time lapse and the elderly age of these patients may have been affected patients' recall. Although many trial participants could not remember the names of the treatments being investigated, most were able to distinguish between a pharmacologically active and a non-specific treatment which they described in their own words as placebo (e.g. 'the one with nothing in it') or the wrong treatment.
The trial is the largest RCT examining the effectiveness of corticosteroid for heel pain to date. Despite this, the trial population we were able to follow up was relatively small. Furthermore, 43% of the participants interviewed reported that their heel pain was recurrent and for almost 20% it was continuous. The cyclical nature of heel pain further limits our ability to extrapolate on the relationship between cognitions and outcomes.
Reactions to debriefing were recorded immediately after the information was communicated. While this method identified immediate response to debriefing, research would benefit from monitoring more long-term effects and examining debriefing at different time points.
Recommendations
Further research is required to better understand beliefs around placebo effects, to develop sensitive ways to inform placebo responders about their treatment allocation and to evaluate the effects of debriefing. This research should combine qualitative research alongside a prospective longitudinal or randomized trial design, using objective outcomes and large clinical samples. Research in this area would increase participant involvement, improve trial methodology and would further our understanding around the therapeutic role of thoughts, feelings and patient-practitioner interactions in RCT's [27].
We recommend that at the informed consent stage investigators ask eligible trial participants whether they would like to be informed about trial results and treatment allocation. It is important to try and avoid creating feelings of embarrassment, mistrust, or disillusionment and to prevent damaging a healing response. For this reason the distinction between 'placebo' and 'placebo effect' should to be made at informed consent, so that participants being enrolled in placebo-controlled trials do not focus on the possibility of receiving a 'dummy' or 'sham' pill, but are informed about the effect that derives from feelings and beliefs, the characteristics of the setting, and the effects of health care interactions [28]. Highlighting this distinction at the informed consent stage should help with debriefing at trial closure.
Conclusion
Our findings emphasize the importance of involving trial participants at study closure, by sharing study results and individual treatment allocation, while respecting individuals who do not wish to be debriefed. We recommend that this information is communicated in a sensitive manner, perhaps discussing placebo effects within a broader framework by considering the effects of spontaneous recovery, treatment beliefs and health care interactions.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
Zelda Di Blasi conceived, conducted and reported the work, and will take responsibility for the integrity of the data and the accuracy of data analysis. All authors gave input to the design of the study and commented on drafts of the paper.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We thank John Weinman and Hilary Bekker in the design of the questionnaire and advice in the conduct of this study, and David Reilly, Ted Katpchuk and Musetta Joyce for helpful comments and suggestions. We are grateful to the Medical Research Council Health Service Research Council for a studentship supporting the first author, and to the Chief Scientist Office in funding the second author.
==== Refs
Klopfer B Psychological variables in human cancer Journal of Projective Techniques 1957 21 331 340 13492241
Faden RR Beauchamp TL A history and theory of informed consent. 1986 Oxford, Oxford University Press
Michels KB Rothman KJ Update on unethical use of placebos in randomised trials Bioethics 2003 17 188 204 12812185 10.1111/1467-8519.00332
Rothman KJ Michels KB The continuing unethical use of placebo controls N Engl J Med 1994 331 394 398 8028622 10.1056/NEJM199408113310611
Di Blasi Z Kaptchuk TJ Weinman J Kleijnen J Informing participants of allocation to placebo at trial closure: postal survey BMJ 2002 325 1329 12468480 10.1136/bmj.325.7376.1329
D.o.H. Research Governance Framework for Health and Social Care 2001 1 41
Leuchter AF Cook IA Witte EA Morgan M Abrams M Changes in brain function of depressed subjects during treatment with placebo Am J Psychiatry 2002 159 122 129 11772700 10.1176/appi.ajp.159.1.122
Wrolstad J Depression Patients Placebo Effect News Factor Network 2002
Husten L Fetal-cell-implantation trial yields mixed results Lancet 1999 353 1501 1501 10.1016/S0140-6736(99)00077-X
Macklin R The ethical problems with sham surgery in clinical research N Engl J Med 1999 341 992 996 10498498 10.1056/NEJM199909233411312
Stoessl AJ de la Fuente-Fernandez R Willing oneself better on placebo--effective in its own right Lancet 2004 364 227 228 15262083 10.1016/S0140-6736(04)16689-0
Crawford F Thomson C Interventions for treating plantar heel pain Cochrane Database Syst Rev 2003 CD000416 12917892
Atkins D Crawford F Edwards J Lambert M A systematic review of treatments for the painful heel Rheumatology (Oxford) 1999 38 968 973 10534547
Crawford F Plantar heel pain and fasciitis Clin Evid 2003 1327 1338
Crawford F Atkins D Young P Edwards J Steroid injection for heel pain: evidence of short-term effectiveness. A randomized controlled trial Rheumatology 1999 38 974 977 10534548 10.1093/rheumatology/38.10.974
Crow R Gage H Hampson S Hart J Kimber A Thomas H The role of expectancies in the placebo effect and their use in the delivery of health care: a systematic review. Health Technology Assessment 1999 3 1 96 10448203
Kirsch I Sapirstein G Listening to Prozac but Hearing Placebo: A Meta-Analysis of Antidepressant Medication Prevention & Treatment 1998 1 0002a
Magaletta PR Oliver JM The hope construct, will, and ways: their relations with self-efficacy, optimism, and general well-being J Clin Psychol 1999 55 539 551 10392785 10.1002/(SICI)1097-4679(199905)55:5<539::AID-JCLP2>3.0.CO;2-G
Stone DA Kerr CE Jacobson E Conboy LA Kaptchuk TJ Patient expectations in placebo-controlled randomized clinical trials Journal of Evaluation in Clinical Practice 2005 11 77 84 15660541 10.1111/j.1365-2753.2004.00512.x
Lundh LG Placebo, belief, and health. A cognitive-emotional model Scand J Psychol 1987 28 128 143 3317812
Kirsch I Conditioning, expectancy, and the placebo effect: comment on Stewart-Williams and Podd (2004) Psychol Bull 2004 130 341 3; discussion 344-5 14979776 10.1037/0033-2909.130.2.341
Bandura A Adams NE Beyer J Cognitive processes mediating behavioral change J Pers Soc Psychol 1977 35 125 139 15093 10.1037//0022-3514.35.3.125
Bandura A Self-efficacy: toward a unifying theory of behavioral change Psychol Rev 1977 84 191 215 847061 10.1037//0033-295X.84.2.191
Di Blasi Z Harkness E Ernst E Georgiou A Kleijnen J Influence of context effects on health outcomes: a systematic review Lancet 2001 357 757 762 11253970 10.1016/S0140-6736(00)04169-6
Edwards A Elwyn G Involving patients in decision making and communicating risk: a longitudinal evaluation of doctors' attitudes and confidence during a randomized trial J Eval Clin Pract 2004 10 431 437 15304143 10.1111/j.1365-2753.2004.00502.x
Di Blasi Z Reilly D Placebos in medicine: medical paradoxes need disentangling BMJ 2005 330 45 15626817 10.1136/bmj.330.7481.45-a
Kaptchuk TJ The placebo effect in alternative medicine: can the performance of a healing ritual have clinical significance? Ann Intern Med 2002 136 817 825 12044130
| 15847683 | PMC1112594 | CC BY | 2021-01-04 16:31:50 | no | BMC Health Serv Res. 2005 Apr 22; 5:30 | utf-8 | BMC Health Serv Res | 2,005 | 10.1186/1472-6963-5-30 | oa_comm |
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-5-301587173510.1186/1471-2334-5-30Case ReportParadise – not without its plagues: Overwhelming Blastomycosis pneumonia after visit to lakeside cottages in Northeastern Ontario Parmar Malvinder S [email protected] Division of Clinical Sciences, Northern Ontario School of Medicine Laurentian & Lakehead Universities, Sudbury & Thunder Bay, Ontario, Canada2 Department of Medicine, University of Ottawa, Ottawa, Ontario, Canada2005 4 5 2005 5 30 30 3 1 2005 4 5 2005 Copyright © 2005 Parmar; licensee BioMed Central Ltd.2005Parmar; 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
Visiting lakes and cottages is a common leisure activity during summer among most Canadians and paradise for some. Various leisure activities are involved during these visits, including cleaning and 'airing' the cottage after long-winters, activities at the lakes and dock building etc, exposing the Canadians to moist soil and decaying woods – a source of white or tan mould – Blastomyces dermatitidis that may cause a flu-like illness to severe pneumonia that often remains a diagnostic challenge and results in delay in diagnosis and appropriate treatment thereby increasing associated morbidity and mortality.
Case Presentations
Five cases of overwhelming acute blastomycosis pneumonia are presented. Four of the five patients presented within few weeks of their visit to the cottages and surrounding lakes and all were initially treated as "community acquired pneumonia" that resulted in delay in diagnosis and poor outcome in the first patient. The first case, however, taught an important lesson that led to high-index of suspicion in the others with early diagnosis and improved outcomes. Interestingly, all patients were obese and had a shorter incubation period and severe clinical course. The possible mechanism for early and severe disease in association with obesity is speculated and literature is reviewed.
Conclusion
High-index of suspicion is important in the early diagnosis and appropriate management acute blastomycosis pneumonia to improve associated morbidity and mortality.
==== Body
Background
During the few summer months in Canada visit to cottages and lakes is a frequent leisure activity. Access to a second home – camp, cabin, chalet or cottage, whatever we call it – has a strong cultural significance for Canadians. The cottage, for many Canadians, is a paradise where extended family and friends gather together, where there is time for leisure and contact with nature.
Pulmonary blastomycosis can be difficult to diagnose and only 18% of patients were correctly suspected to have blastomycosis in an endemic area and are often initially misdiagnosed and treated as community acquired pneumonia, malignant tumor or tuberculosis resulting in unnecessary surgery and treatment delays[1]. Experience with five cases of blastomycosis pneumonia from Northeastern Ontario is presented emphasizing the high-index of suspicion for early diagnosis and improving outcome. The patients presented with overwhelming acute blastomycosis pneumonia within few weeks after visiting lakes and cottages. These were initially treated as community acquired pneumonia and a lesson learnt from first case is emphasized and a high-index of suspicion by the author led to early diagnosis in others. Interestingly all patients were obese and had severe disease. A possible association with obesity and severity of disease is observed and mechanism speculated.
Case presentations
The demographics, month of presentation, time to illness and diagnosis after exposure and the outcomes are summarized in table 1. A brief history on all with a detailed history of case 3 is presented.
Table 1 A summary of 5 cases
Month of presentation to hospital Age (years) Gender Weight (Pounds) # of days symptoms started after visit to cottage or lake # of days to diagnosis after onset of symptoms/after hospitalization Outcome
July 30 Male 340 7–10 40/22 Died
July 34 Female 270 7 20/7 Recovered
July 36 Female 200 10 36/12 Recovered
July 36 Male 230 6 19/1 Recovered
February 57 Female >350 Not clear 30/3 Died
Case 1
In July 1998, a 30-year old man with morbid obesity [weight 340 pounds] presented to emergency department with 3-weeks history of 'flu-like' symptoms with fever, chills, night sweats and cough with yellowish phlegm that started after building a dock at his cottage. Chest x-ray revealed bibasilar infiltrates. He was admitted with bilateral pneumonia and was started on intravenous penicillin and oral clarithromycin after obtaining blood and sputum cultures. Three days later he developed rash and penicillin was discontinued and antibiotics were changed to intravenous ceftriaxone and erythromycin. He remained febrile for next 7 days. Sputum Gram stain showed many pus cells but cultures remained negative. His condition continued to deteriorate. Local bronchoscopist was unavailable to perform bronchoscopy and arrangements were made to transfer him to another hospital. Before transfer, his condition deteriorated and was intubated. Tracheal secretions were aspirated and sent for cultures – including fungal and tubercular. The night of transfer, he developed cardio-respiratory arrest and couldn't be resuscitated. Next day, the wet preparation on the tracheal secretions revealed thick walled budding yeast consistent with blastomycosis. Autopsy showed severe pulmonary disease with solidification of both lungs and cultures confirmed the diagnosis.
Case 2
In July 2000, a 34-year old obese woman [270 pounds] presented to the emergency department with pleuritic chest pain associated with chills and night sweats about a week after she was camping at the lakeside cottage. She denied symptoms of cough or phlegm. She was diagnosed with musculoskeletal pain and discharged home on analgesics. Five days later she presented with ongoing symptoms of fever, chills with dry cough and right lower chest pain. A chest X-ray revealed an infiltrate in the right lower lobe. She was diagnosed with community acquired right lower lobe pneumonia and sent home on oral clarithromycin. She presented 4-days later with ongoing symptoms to the emergency department and the chest x-ray now showed worsening pneumonia. She was admitted to the hospital and started on intravenous ampicillin and ceftriaxone, and a ventilation-perfusion scan showed a matched perfusion defect in right lower lobe. She continued to feel weak with night sweats and chest X-ray showed worsening of infiltrate. A medical consult was requested. At this time, during consultation it was noted that her symptoms started after a short stay at the cottage that reminded me of the previous case, and the possibility of acute blastomycosis pneumonia was raised. Her white blood cell count remained slightly elevated at 13.0 but erythrocyte sedimentation rate was markedly elevated at 112. As she didn't have productive cough, a bronchoscopy was recommended that couldn't be performed locally [local bronchoscopist was away] and she was referred to a tertiary care center where bronchio-aleveolar lavage showed budding yeast and cultures confirmed growth of blastomycosis. She was treated initially with intravenous amphotericin B and later switched to oral itraconazole for a year and made full recovery.
Case 3
In July of 2002, a 36-year old obese woman [weight 200 pounds, height 5 feet 1 inch] with history of type 2 diabetes for 5 years, presented to the emergency department with 5 days history of fever, chills and cough with yellowish phlegm and sharp pain in her right lower chest that aggravated with deep breathing. Physical examination was unremarkable. She was afebrile and lungs were clear. Homen's sign was negative. Initial laboratory data showed slightly elevated d-dimer of 0.374 ug/mL [normal <0.25 ug/mL]. A chest X-ray (figure 1) showed right lower lobe infiltrate. She was diagnosed with community acquired pneumonia and sent home on oral clarithromycin 500 mg BID, cefurox 500 mg BID for 7 days and acetaminophen as required.
Figure 1 Chest X-ray (PA view) at initial presentation showing early right lower lobe consolidation.
She continued to have symptoms of fever, chills, night sweats with productive cough and dyspnea and presented to the hospital five days later. She denied hemoptysis. Physical examination revealed a temperature of 39 degree Celsius, mild tachycardia with heart rate of 104 beats per minute, normal blood pressure of 125/74 mmHg and respiratory rate of 18. Pulse oxymetry revealed oxygen saturation of 96% on room air. There was decreased air entry with bronchial breath sounds in right lower chest. A white cell count was slightly elevated at 13.5, with normal hemoglobin of 128 g/L and d-dimer between 0.25–0.50 ug/mL. A second chest X-ray (figure 2) showed dense infiltrate in right lower lobe. She was admitted with the diagnosis of community-acquired pneumonia and started on intravenous levofloxacin 500 mg a day.
Figure 2 Chest X-ray (PA view), 5 days later, showing progression of disease with infiltrates in right and left lower lobes.
She had a positive Mantoux test in 1976. There was no history of travel save for visiting her sister's cottage at the local lake a weak before where she slept at the gazebo for two nights. There were no pets at home. She smoked one pack per day and entertained social drinks.
She continued to have temperature up to 39 degrees Celsius, productive cough and shortness of breath. Sputum Gram stain showed 4+ neutrophils and 1+ normal flora and blood cultures remained negative. She received intravenous levofloxacin and clindamycin for one week. Bronchoscopy performed on day 6 was unremarkable save for inflammatory changes in lower lungs and samples were collected for bacterial and tuberculosis cultures [in retrospect, the specimen was not sent for fungal cultures]. On 7th day because of ongoing symptoms, the antibiotics were changed to intravenous imipenem-cilastatin sodium 1 gm every 12 hour and she has had a CT scan of chest (figure 3) that showed bibasilar consolidation with sparing of the apices. The white cell count fluctuated between 13.6 to 20.2 [normal 4.0–11.0 × 109 /L]. Hemoglobin decreased to 97 g/L. BUN, serum creatinine, electrolytes, AST, ALT, GGT and ALP were normal. Urinalysis was negative. Blood, urine and sputum cultures remained negative. Sputum gram stain showed 3-4+ neutrophils without organisms. ESR on day 8 was elevated at 112 mm/hr [normal 0–15]. Antinuclear antibody was negative. Arterial blood gas on day 8 showed pH of 7.46, pCO2 of 35, pO2 of 59 and oxygen saturation of 91% on room air. She was started on supplemental oxygen by nasal prongs. Bronchoscopic specimen cultures were negative save for pending TB cultures.
Figure 3 A section of CT scan of chest performed 8 days after hospitalization showing bilateral consolidation of lungs, mainly of lower lung fields.
After five days of intravenous imipenem-cilastatin sodium therapy, there was no improvement in her clinical condition. Another medical consult was requested, when the author got involved in her care. She had a temperature of 39 C, pulse rate of 104 and blood pressure of 114/70 mmHg. There was no lymphadenopathy and heart sounds were normal. Her chest examination showed decreased breath sounds at both bases with egophony in right lower chest. There were no pleural rubs. There was no clubbing or cyanosis. The rest of the examination was unremarkable save for an obese benign abdomen.
Chest X-ray (figure 4) now showed further deterioration of the bilateral infiltrates with nodular appearance and the radiologist remarked, "The lesions are suspicious of metastases." Sputum cytology showed a large number of acute inflammatory cells without malignant cells. The possibility of vasculitis was entertained because of markedly elevated ESR and antineutrophilic cytoplasmic antibody (ANCA) was negative.
Figure 4 Chest X-ray (PA view) – after 8 days of intravenous antibiotic therapy, showing further worsening of bilateral lower lung disease with nodular pattern – raising suspicion of metastatic disease.
The history of sleeping at the cottage for two nights, a week before her illness started; progressive bilateral pneumonia with ongoing high-grade fever, chills, and night sweats despite sufficient antibiotic coverage and sputum Gram stain showing a large number of neutrophils without organisms (culture negative) raised the author's suspicion for the possibility of fungal process such as acute blastomycosis pneumonia and the Public Health Laboratory (PHL) was specifically asked to perform wet preparation for blastomycosis on the sputum. The PHL laboratory confirmed the presence of 'round thick walled budding yeast like cells (figure 5) suggestive of Blastomyces dermatitidis and this was confirmed on fungal cultures. She was promptly started on intravenous amphotericin B and responded well and later switched to oral itraconazole 400 mg twice a day for 6 months and then 200 mg a day for another 6 months. She recovered fully.
Figure 5 Wet preparation of sputum [25% NaOH with 5% Glycerol as the mounting medium, 40× magnification] showing budding yeast [blastocyst]
Case 4
Same July of 2002, while case #3 was still-in hospital, a 42-year old man [weight 230 pounds] presented to a peripheral community hospital with symptoms of fever, chills and malaise about 7–10 days after a trip to the local beach. He was diagnosed with community acquired left lower lobe pneumonia and treated with three different kind of antibiotics. He remained febrile with temperature of 39–40 degree Celsius and 10-days after his admission, he was transferred to our hospital and again the past experience and a recent case of similar nature, although from a different region, reminded of the similar process and he underwent bronchoscopy and the wet preparation showed budding yeast and cultures were confirmatory. He was promptly started on intravenous amphotericin B and later switched to oral itraconazole and made full recovery.
Case 5
In February of 2004, a 57-year old woman with morbid obesity [weight over 350 pounds] with history of hypertension and type 2 diabetes presented with one month history of fever, night sweats and cough and was admitted to the hospital with right sided community acquired pneumonia [figure 6] and started on intravenous levaquin. She had history of gall stones and had mild upper abdominal pain. Next morning, while undergoing an ultrasound examination she became unresponsive and suffered cardio-respiratory arrest. She was successfully resuscitated, but remained hypotensive and required inotropic support and admitted to intensive care unit under my care. Because of shock and dense lobar pneumonia, the possibility of severe pneumococcal pneumonia was entertained and intravenous penicillin was added. She had copious amount of thick pus from the endotracheal tube and because of thick pus, possibility of acute blastomycosis pneumonia was entertained although it was -36 degree Celsius outside, and specimen was sent to Public Health Laboratory to perform a wet preparation to confirm or rule out the possibility blastomycosis. Although blastomycosis was suspected but empiric therapy with amphotericin B was not started, as there was no clear cut history of exposure and she presented in the middle of winters when the outside temperature was -30 to -40 degrees Celsius. She was difficult to ventilate and remained unstable during the night. Next morning she arrested and could not be resuscitated after 2-hours of resuscitation. At noon on that day, Public Health Laboratory confirmed the diagnosis of acute blastomycosis pneumonia.
Figure 6 Chest X-ray (PA view) at initial presentation showing consolidation within the right mid and both lower lobes.
Discussion
Blastomycosis is a relatively rare but important and lethal disease and is caused by a thermal dimorphic fungus – Blastomyces dermatitidis – that exists in mycelial form in the environment and develops into yeast form in the host [2-4]. Clinical spectrum is variable and range from asymptomatic infection to pyogranulomatous inflammation with fulminant hypoxic respiratory failure or acquired respiratory distress syndrome (ARDS) to extrapulmonary manifestations[5,6]. Often acute pulmonary disease mimics a bacterial pneumonia or cause respiratory distress syndrome and pose a diagnostic challenge and results in delay in diagnosis, incorrect treatment and poor outcome.[1,5,7]
The incidence of blastomycosis remains largely unknown as it is not a reportable disease in Canada and was removed from the list of reportable diseases in Ontario in 1990[8]. It is an uncommon, though regularly seen, life threatening disease in endemic areas. In North America the disease is concentrated along the Mississippi and Ohio River basins, but also extends into Northern Wisconsin, Minnesota and the Canadian provinces bordering the Great Lakes[2,5]. In Ontario, it is endemic especially in Northwestern Ontario, north and west of Lakes Superior and Huron and in neighboring boreal Manitoba[9]. In the endemic areas the disease is predominantly encountered in rural or wilderness areas where the causative agent likely colonizes the soil or plant litter in riparian sites[10]. Outbreak studies have implicated, building of a hunting lodge, proximity to a construction site, raccoon hunting, exposure to a beaver lodge, and activities by riverbanks as sources of exposure. [9,11-13] A retrospective review from 1990 to 1998 in Northwestern Ontario revealed an incidence of 117 per 100,000 population with much higher rate of 404 per 100,000 population in aboriginals[14]. Public Health Laboratory in Timmins covers most of the Northeastern Ontario and identifies an average of 5–6 isolates per year [personal communication, PHL, Timmins]. Three of the patients [#1, 2 and 5] were from Timmins and one [#3] about 200 Km south and the other [#4] about 100 km North of Timmins (figure 7).
Figure 7 Map of Northeastern Ontario, showing the location of exposure [adapted from , accessed April 2, 2005].© Queens Printer for Ontario, 2002. Adapted and reproduced with permission
Four of the five patients presented in July compared to the reported occurrence between September and January. The incubation period was much shorter than the reported period of 21 to 106 days with an average of 45 days[10]. The reason for short incubation period is unclear but it is interesting that all patients were significantly obese and two had diabetes. Diabetes increases the risk for fungal infections and may be present in 22% of patients[1]. Whether obesity altered the course and severity of disease requires further study but it is known that obesity can alter pulmonary function by its adverse effects on respiratory mechanics, resistance within the respiratory system, respiratory muscle function and lung volumes[15] and thereby might cause decreased clearance of inhaled spores resulting in severe disease with shorter incubation period. Baik et al noted a two-fold increase in the risk of community acquired pneumonia both in men and women who gained over 40 pounds from their adulthood weight and a direct association with increasing BMI in women[16]. Obesity is also shown to impair T and B cell function [17,18] that may be mediated in part by consequences of obesity, such as hyperglycemia and insulin resistance[19]. It is possible that non-obese people might have had such exposure and were either able to clear the inhaled spores or developed a milder form of disease indistinct from flu-like illness.
Laboratory diagnosis
The role of sputum culture in the management of community-acquired pneumonia is controversial and only 14.4% of patients yielded good-quality sputum in a recent study[20] and is not routinely recommended[21]. However, the yield from respiratory secretions [wet preparation and cultures] is high [80–100%] but is underutilized.[22] Wet preparation [10% potassium hydroxide] examination of respiratory secretions showing budding yeast is the gold standard initial test and cultures are confirmatory but may take up to 5 weeks. The first case taught an important lesson and re-enforced the old dictum – "presence of a large number of neutrophils on the Gram stain indicates that the sputum is of good quality". When no organisms are seen or cultured from this quality specimen, it should raise the suspicion for non-bacterial (fungal, mycobacterial, etc.) causes. This lesson was helpful, especially to me, in making a timely diagnosis in other four patients, where the patients were treated initially with different antibiotics for "pneumonia" before my involvement. The diagnosis was suspected at my initial encounters and confirmed within 24–72 hours in all.
Lesson: Sputum full of neutrophils indicates 'a good quality specimen' and with negative cultures should alert the physician to look for non-bacterial causes.
Radiographic features
The radiographic features of blastomycosis are highly variable and no one pattern exists[23], making differential diagnosis from other bacterial, fungal and neoplastic disease difficult. Often initial radiographic presentation tends to be localized airspace disease characterized by patchy and confluent airspace opacities with indistinct borders, in a segmental, subsegmental, or non-segmental distribution[3]. Due to these radiographic features and clinical presentation of fever, chills and productive cough, the initial diagnosis of blastomycosis is often overlooked in favor of a diagnosis of bacterial pneumonia. In chronic form, when presents as a focal mass it mimics bronchogenic carcinoma prompting additional tests such as needle biopsy and even a lobectomy[1,5].
Treatment
Treatment of blastomycosis depends on the clinical presentation and presence or absence of extra-pulmonary manifestations. Patients with mild pulmonary disease whose symptoms are resolving at time of diagnosis require observation. All other patients with symptomatic pulmonary disease or extrapulmonary disease require antifungal therapy. amphotericin B is recommended for patients with severe acute pulmonary disease [as in the cases presented], with associated ARDS[5], immunocompromised hosts and in those with central nervous system involvement. Oral azoles – Itraconazole or fluconazole are recommended for mild to moderate acute, subacute or chronic pulmonary and disseminated forms and after intravenous amphotericin B therapy in severe cases. A 6 to12 months of therapy with antifungal agents is often required.
Prognosis
The prognosis for most patients with mild to moderate pulmonary disease with or without dissemination is good provided that antifungal therapy is initiated promptly after diagnosis[23]. Mortality is about 10% in patients with severe pulmonary disease without ARDS but reaches close to 90% in patients with associated ARDS.[24]
Prevention
There are no specific prevention strategies. However, activities that bring individuals closer to rotting wood or moist soil near water are associated with a greater risk and a high-index of suspicion especially in endemic areas with early and appropriate testing is important. Soil testing, even in endemic areas, is neither cost-effective nor reliable and is not recommended.
Conclusion
Visit to cottages and lakes, is a paradise for most Canadians where they spend most of their summers relaxing and participating in outdoor activities that may expose them to Blastomyces dermatitidis. The five cases presented here are not to illustrate the drawbacks of these activities but to increase the awareness of blastomycosis to the public, visitors to endemic areas and healthcare providers. It is important for the patients to provide their history of stay at the cottages and lakes and contact with wet soil and decaying woods and for the physicians to be vigilant and consider this diagnosis early and request appropriate testing to reduce associated morbidity and mortality.
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
Lemos LB Baliga M Guo M Blastomycosis: The great pretender can also be an opportunist. Initial clinical diagnosis and underlying diseases in 123 patients Ann Diagn Pathol 2002 6 194 203 12089732 10.1053/adpa.2002.34575
Goldman M Johnson PC Sarosi GA Fungal pneumonias: The Endemic Mycoses Clinics in Chest Med 1999 20 507 19
Kuzo RS Goodman LR Blastomycosis Seminars in Roentgenol 1996 31 45 51
Bradsher RW Histoplasmosis and Blastomycosis Clinical Infectious Diseases 1996 22 S102 111 8722836
Davis SF Sarosi GA Epidemiologic and clinical features of pulmonary blastomycosis Seminars in Respiratory Infections 1997 12 206 211 9313292
Lester RS DeKoven JG Kane J Simor AE Krajden S Summerbell RC Novel cases of blastomycosis acquired in Toronto CMAJ 2000 163 1309 12 11107469
Meyer KC McManus EJ Maki DG Overwhelming pulmonary blastomycosis associated with the adult respiratory distress syndrome N Engl J Med 1993 329 1231 1236 8413389 10.1056/NEJM199310213291704
Public Health Branch Summary of reportable diseases 1990 Toronto: Communicable Disease Control, Ontario Ministry of Health 1991
Kane J Righter J Krajden S Lester RS Blastomycosis: a new endemic focus in Canada CMAJ 1983 129 728 31 6616383
DiSalvo AF Al-Doory Y, DiSalvo AF The ecology of Blastomyces dermititidis Blastomycosis 1992 New York. Plenum Medical Books 43 73
Klein BS Vergermont JM DiSalvo AF Kaufman L Davis JP Two outbreaks of blastomycosis along rivers in Wisconsin: isolation of Blastomyces dermatitidis from riverbank soil and evidence of transmission along waterway Am Rev Respir Dis 1987 136 1333 38 3688635
Lowry PW Keso KY McFarland LM Blastomycosis in Washington, Parish, Louisiana 1976–1985 Am J Epidemiol 1989 130 151 59 2787106
Armstrong CW Jenkins SR Kaufman L Kerkering TM Rouse BS Miller GB Jr Common-source outbreak of blastomycosis in hunters and their dogs J Infect Dis 1987 155 568 70 3805778
An outbreak of human blastomycosis The epidemiology of blastomycosis in the Kenora catchment region of Ontario, Canada Canada Communicable Disease Report 26 15 may 2000 [, accessed 12 October 2004]
Koenig SM Pulmonary complications of obesity Am J Med Sci 2001 321 249 79 11307867 10.1097/00000441-200104000-00006
Baik I Curhan GC Rimm EB Bendich A Willett WC Fawzi WW A prospective study of age and lifestyle factors in relation to community acquired pneumonia in US men and women Arch Intern Med 2000 160 3082 88 11074737 10.1001/archinte.160.20.3082
Kuczmarski RJ Flegal KM Campbell SM Johnson CL Increasing prevalence of overweight among US adults JAMA 1994 272 205 11 8022039 10.1001/jama.272.3.205
Pi-Synyer FS Medical hazards of obesity Ann Intern Med 1993 119 655 60 8363192
Stallone DD The influence of obesity and its treatment of immune system Nutr Rev 1994 52 37 50 8183468
Garcia-Vazquez E Marcos MA Mensa J de Roux A Puig J Font C Francisco G Torres A Assessment of the usefulness of sputum culture for diagnosis of community-acquired pneumonia using the PORT predictive scoring system Arch Intern Med 2004 164 1807 11 15364677 10.1001/archinte.164.16.1807
Niederman MS Mandell LA Anzueto A Bass JB Broughton WA Campbell GD Dean N File T Fine MJ Gross PA Martinez F Marrie TJ Plouffe JF Ramirez J Sarosi GA Torres A Wilson R Yu VL American Thoracic Society Guidelines for the management of adults with community-acquired pneumonia. Diagnosis, assessment of severity, antimicrobial therapy, and prevention Am J Respir Crit Care Med 2001 163 1730 54 11401897
Martynowicz MA Prakash UBS Pulmonary blastomycosis – An appraisal of diagnostic techniques Chest 2002 121 768 773 11888958 10.1378/chest.121.3.768
Patel RG Patel B Petrini MF Carter RR 3rdGriffith J Clinical presentation, radiographic findings, and diagnostic methods of pulmonary blastomycosis: A review of 100 consecutive cases South Med J 1999 92 289 295 10094269
Vasquez JE Mehta JB Agrawal R Sarubbi FA Blastomycosis in northeast Tennessee Chest 1998 114 436 443 9726727
| 15871735 | PMC1112595 | CC BY | 2021-01-04 16:28:15 | no | BMC Infect Dis. 2005 May 4; 5:30 | utf-8 | BMC Infect Dis | 2,005 | 10.1186/1471-2334-5-30 | oa_comm |
==== Front
BMC Int Health Hum RightsBMC International Health and Human Rights1472-698XBioMed Central London 1472-698X-5-41585048010.1186/1472-698X-5-4Research ArticleGood governance and good health: The role of societal structures in the human immunodeficiency virus pandemic Menon-Johansson Anatole S [email protected] John Hunter Clinic, St. Stephen's Centre, Chelsea & Westminster Hospital, London, SW10 9NH, UK2005 25 4 2005 5 4 4 19 12 2004 25 4 2005 Copyright © 2005 Menon-Johansson; licensee BioMed Central Ltd.2005Menon-Johansson; 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
Only governments sensitive to the demands of their citizens appropriately respond to needs of their nation. Based on Professor Amartya Sen's analysis of the link between famine and democracy, the following null hypothesis was tested: "Human Immunodeficiency Virus (HIV) prevalence is not associated with governance".
Methods
Governance has been divided by a recent World Bank paper into six dimensions. These include Voice and Accountability, Political Stability and Absence of Violence, Government Effectiveness, Regulatory Quality, Rule of Law and the Control of Corruption. The 2002 adult HIV prevalence estimates were obtained from UNAIDS. Additional health and economic variables were collected from multiple sources to illustrate the development needs of countries.
Results
The null hypothesis was rejected for each dimension of governance for all 149 countries with UNAIDS HIV prevalence estimates. When these nations were divided into three groups, the median (range) HIV prevalence estimates remained constant at 0.7% (0.05 – 33.7%) and 0.75% (0.05% – 33.4%) for the lower and middle mean governance groups respectively despite improvements in other health and economic indices. The median HIV prevalence estimates in the higher mean governance group was 0.2% (0.05 – 38.8%).
Conclusion
HIV prevalence is significantly associated with poor governance. International public health programs need to address societal structures in order to create strong foundations upon which effective healthcare interventions can be implemented.
==== Body
Background
It has been argued that famine only occurs in nations that are immune to the political will of their people [1]. Political freedom in famine free countries is additionally coupled, albeit unevenly, to other freedoms such as education, health, the control of family size and the ability to seek employment.
Relatively recently, global institutions such as the World Health Organization (WHO) and the World Bank have made the link between macroeconomics and health [2,3]. Analysis of poverty around the world highlights those countries that are 'very unlikely' to meet the World Bank groups' millennium development goals (MDGs) [4]. These MDGs include the combat of HIV/AIDS, malaria and other diseases, improvement in maternal health, achievement of universal primary education, promotion of gender equality and empowerment of women, reduction in child mortality and the eradication of extreme poverty and hunger.
Some of the shared societal structures underpinning economic growth and health are the absence of violence, government effectiveness, the rule of law, lack of corruption and the ability to select a government. Even though all of these are clearly desirable the relative weight of each societal structure necessary for a strong nation state is debatable [5]. The risk of infectious disease is determined not only by pathogens and the response of the patient but also by powerful societal forces that override individual knowledge and choice [6]. Paul Farmer has coined the phrase 'structural violence' that reflects the limit of life choices, particularly of women, by racism, sexism, political violence, and grinding poverty.
The 2004 World Health Report discusses the challenges of tackling the HIV pandemic [7]. In the African continent, HIV is implicated in poor economic performance and falling gross domestic product (GDP). Within this document it describes the wide range of international support garnered to meet this challenge. However, even though the requirement of local and national government co-operation is stressed within this document, it does not elaborate on the massive heterogeneity inherent within this mandatory component.
In order to investigate the strength of the relationship between the quality of societal structures and the HIV pandemic, World Bank and UNAIDS sources were used to test the null hypothesis: "HIV prevalence is not associated with governance".
Methods
A recent World Bank paper entitled Governance Matters III collated governance indicators for 199 countries / regions [8]. Governance in this document has been broken down into six dimensions that are defined in Table 1. Using these definitions, this research collected data for each country from 18 sources that are listed in Table 2. Governance data were then aggregated for each country and plotted along a continuum. Only the 2002 Governance data has been used in this paper. This dataset is available in a spreadsheet format from the World Bank website [9].
Table 1 Definitions of governance dimensions
Voice and accountability
"how those in authority are selected and replaced"
Political Stability
"perceptions of the likelihood that the government in power will be destabilized or overthrown by possibly unconstitutional and/or violent means, including domestic violence and terrorism"
Government Effectiveness
"we combine into a single grouping responses on the quality of public service provision, the quality of the bureaucracy, the competence of civil servants, the independence of the civil service from political pressures, and the credibility of the government's commitment to policies"
Regulatory Quality
"includes measures of the incidence of market-unfriendly policies such as price controls or inadequate bank supervision, as well as perceptions of the burdens imposed by excessive regulation in areas such as foreign trade and business development"
Rule of Law
"several indicators which measure the extent to which agents have confidence in and abide by the rules of society. These include perceptions of the incidence of crime, the effectiveness and predictability of the judiciary, and the enforceability of contracts"
Corruption
"measures perceptions of corruption, conventionally defined as the exercise of public power for private gain. Despite this straightforward focus, the particular aspect of corruption measured by the various sources differs somewhat, ranging from the frequency of "additional payments to get things done," to the effects of corruption on the business environment, to measuring "grand corruption" in the political arena or in the tendency of elite forms to engage in "state capture" "
Table 2 The sources used for governance data
Afrobarometer
Business Environment Risk Intelligence
Columbia University
The Economist Intelligence Unit
European Bank for Reconstruction and Development
Freedom House
Gallup International
Global Insight's DRI/McGraw-Hill
Heritage Foundation / Wall Street Journal
Institute for Management and Development
Latinobarometro
Political Risk Services
PriceWaterhouseCoopers
Reporters Without Borders
State Department / Amnesty international
World Bank
World Economic Forum
World Markets Research Center
The 2002 HIV prevalence estimates were obtained for each country. HIV prevalence is the percentage of adults aged between 15 and 49 years of age infected with HIV. One hundred and forty nine of the 199 countries / regions cited by the World Bank paper had published UNAIDS 2002 HIV prevalence estimates. Those countries / regions excluded from analysis due to the lack of HIV prevalence data are listed in Table 3. Those countries given a UNAIDS HIV prevalence estimate of < 0.1% were given a set value of 0.05%.
Table 3 Lists those countries that do not have HIV prevalence estimates for 2002
Countries where UNAIDS commissioned a report but no HIV prevalence figure was published Mean Governance Ranking position (1 – 199) Countries / Regions where UNAIDS did not commission a report Mean Governance Ranking position (1 – 199)
Afganistan 4 Andorra 180
Albania 70 Antigua and Barbuda 146
Brunei 139 Bermuda 170
Comoros 52 Cape Verde 119
Djibouti 50 Cayman Islands 181
Gabon 86 East Timor 44
Guinea 29 French Guiana 151
Kuwait 128 Grenada 136
Lebanon 74 Kiribati 100
Liberia 6 Liechtenstein 183
Mauritania 103 Macao 138
Myanmar 5 Marshall Islands 110
North Korea 13 Martinique 157
Niger 60 Micronesia 91
Paraguay 27 Monaco 161
Qatar 135 Nauru 156
Saudi Arabia 105 Puerto Rico 166
Seychelles 126 Samoa 131
Syria 58 San Marino 171
Tunisia 114 Soa Tome and Principe 102
United Arab Emerites 152 Solomon Islands 43
St. Kitt's and Nevis 127
St. Lucia 129
St. Vincents and the Grenadines 132
Taiwan 160
Tonga 66
Tuvalu 172
Vanuatu 87
West Bank 26
In addition to separate analysis of each governance dimension, an average governance figure was obtained based on the assumption that each governance dimension was of equal importance. The null hypothesis was tested by measuring association between ranked governance and HIV prevalence data across the whole spectrum of countries (Kendall tau test, two tailed). Statistical analysis was performed using SPSS version 12.
Other health and economic data have been included in the Tables to illustrate the development needs of countries. The most recent maternal mortality data from the World Health Organization (WHO) data are available from the year 2000 [10]. The mean maternal mortality ratio (MMR) is the number of mothers who die per 100,000 live births. The number of physicians available per 100,000 inhabitants from 1990–2003 were obtained from United Nations [11]. Access to improved drinking water in 2002, expressed as a percentage, was obtained from WHO / UNICEF [12]. Life expectancy and the GDP per capita in US dollars corrected for purchaser power parity (GDP-PPP) were obtained from the Central Intelligence Agency World Factbook 2002 [13]. The GINI index data (1994–2001) from the United Nations are quoted to give an indication of the equity of income and resource distribution for each country [14]. A value of zero on the GINI index indicates fully equitable distribution of income and resources.
The relative investment by governments in health, education and the military is expressed as a ratio. This ratio is calculated from the percentage of GDP spent on health and education divided by the percentage of GDP spent on health, education and the military. Education expenditure from 2001 and government expenditure on health from 1999–2003 were obtained from the Human Development Report 2004 whilst military expenditure as a percentage of GDP for 2002 were obtained from the International Institute for Strategic Studies [15].
Results
There were fifty distinct HIV prevalence rankings from the 149 countries with UNAIDS HIV prevalence estimates in 2002. Botswana had the highest HIV prevalence estimates (38.8%) in the world that year whilst the majority of countries were placed within the lowest ranking, where HIV prevalence estimates were reported by UNAIDS to be < 0.1% (written as 0.05%). The distribution of HIV prevalence estimates by mean governance ranking is shown in Figure 1.
Figure 1 A scatter graph of Mean Governance Ranking and HIV prevalence for 149 countries.
Non-parametric analysis of association between governance and HIV prevalence is shown in Table 4. The negative correlations indicate that HIV prevalence falls as the governance improves for each governance dimension and mean governance. The three most influential dimensions of governance were government effectiveness, the rule of law and corruption. All correlations were significant thus rejecting the null hypothesis.
Table 4 HIV prevalence correlations for each governance dimension and mean governance
Governance dimension Correlation coefficient (N = 149) p value
Voice & accountability -0.123 0.032
Political Stability and Absence of Violence -0.164 0.004
Government Effectiveness -0.204 0.000
Regulatory Quality -0.157 0.006
Rule of Law -0.194 0.001
Corruption -0.184 0.001
Mean Governance -0.170 0.003
The dataset of 149 nations has been divided into three groups that represent the lowest (n = 50), middle (n = 50) and the highest (n = 49) governance ranking positions for each governance dimension and mean governance in Tables 5, 6 and 7 respectively. When these nations were divided into three groups, the median (range) HIV prevalence estimates remained constant at 0.7% (0.05 – 33.7%) and 0.75% (0.05% – 33.4%) for the lower and middle mean governance groups respectively despite improvements in other health and economic indices. The median HIV prevalence estimates in the higher mean governance group was 0.2% (0.05 – 38.8%).
Table 5 HIV, health and development data for the lowest governance ranking group.
Country
Median (Range)
N = 50 2002 HIV prevalence (%) MMR (maternal deaths / 100,000 live births) in 2000 Physicians (per 100,000), 1990–2003 Improved drinking water (%) in 2002 Life Expectancy (years) in 2002 GDP-PPP ($ per capita) in 2002 GINI index 1994 – 2001 Ratio Health + Education / Health + Education + Military Spending in 2002
Voice & Accountability 1.3 (0.05 – 33.7) 525 (16 – 2000) 25 (1 – 596) 72 (22 – 100) 55.4 (37.1 – 76.4) 1621 (498 – 15650) 37.2 (26.8 – 62.9) 0.62 (0.14 – 0.91)
Political Stability 0.45 (0.05 – 33.7) 340 (7 – 2000) 27 (1 – 463) 77 (22 – 100) 63.1 (37.1 – 78.7) 2201 (498 – 18558) 37.8 (26.8 – 62.9) 0.62 (0.20 – 0.92)
Government Effectiveness 1.65 (0.05 – 33.7) 570 (7 – 2000) 23.5 (1 – 463) 71 (22 – 100) 54.5 (37.1 – 75.6) 1622 (498 – 8663) 44.1 (26.8 – 62.9) 0.69 (0.20 – 0.91)
Regulatory Quality 1.0 (0.05 – 33.7) 525 (7 – 2000) 25 (1 – 596) 73 (22 – 100) 56.6 (36.4 – 76.4) 1651 (498 – 12732) 38.2 (26.8 – 62.9) 0.62 (0.17 – 0.90)
Rule of Law 1 (0.05 – 33.7) 415 (7 – 2000) 25 (1 – 596) 74 (29 – 100) 60.4 (36.4 – 76.4) 1771 (498 – 12732) 39.6 (26.8 – 62.9) 0.68 (0.20 – 0.91)
Corruption 1.0 (0.05 – 33.7) 435 (7 – 2000) 26 (1 – 463) 75 (29 – 100) 60.8 (36.4 – 75.6) 1963 (498 – 12732) 39.3 (26.8 – 62.9) 0.70 (0.20 – 0.91)
Mean governance 0.7 (0.05–33.7) 505 (7 – 2000) 25 (1 – 596) 73 (22 – 100) 59.2 (37.1 – 76.4) 1681 (498 – 12732) 37.2 (26.8 – 62.9) 0.64 (0.14 – 0.91)
Table 6 HIV, health and development data for the middle governance ranking group.
Country
Median (Range)
N = 50 2002 HIV prevalence (%) MMR (maternal deaths / 100,000 live births) in 2000 Physicians (per 100,000), 1990–2003 Improved drinking water (%) in 2002 Life Expectancy (years) in 2002 GDP-PPP ($ per capita) in 2002 GINI index 1994 – 2001 Ratio Health + Education / Health + Education + Military Spending in 2002
Voice & Accountability 0.45 (0.05 – 31) 130 (5 – 1500) 87 (2 – 463) 84 (39 – 100) 68.5 (36.4 – 79.6) 3383 (693 – 25102) 44 (28.2 – 70.7) 0.77 (0.35 – 0.94)
Political Stability 0.95 (0.05 – 33.4) 180 (10 – 1800) 68 (2 – 596) 85 (34 – 100) 65.4 (36.4 – 77.2) 3544 (675 – 35831) 44.0 (29.0 – 70.7) 0.80 (0.14 – 0.95)
Government Effectiveness 0.4 (0.05 – 33.4) 130 (2 – 1500) 90 (2 – 596) 85 (41 – 100) 69.1 (36.4 – 77.5) 3596 (675 – 12732) 40.0 (25.8 – 70.7) 0.77 (0.13 – 0.94)
Regulatory Quality 0.5 (0.05 – 33.4) 150 (5 – 1800) 85 (4 – 420) 84 (34 – 100) 66.9 (37.1 – 77.5) 3481 (770 – 10338) 44.6 (28.2 – 70.7) 0.79 (0.33 – 0.96)
Rule of Law 0.45 (0.05 – 31.0) 145 (2 – 1800) 76 (3 – 344) 84 (22 – 100) 68.0 (37.1 – 77.5) 3384 (595 – 10212) 42.2 (25.8 – 60.7) 0.77 (0.14 – 0.95)
Corruption 0.45 (0.05 – 33.4) 145 (5 – 1400) 68 (2 – 596) 83 (22 – 100) 68.5 (38.6 – 77.5) 3418 (595 – 9575) 43.0 (28.9 – 70.7) 0.77 (0.14 – 0.95)
Mean governance 0.75 (0.05 – 33.4) 160 (5 – 1800) 75 (2 – 420) 83 (41 – 100) 66.9 (36.4 – 77.5) 3418 (693 – 9575) 44.6 (28.2 – 70.7) 0.79 (0.17 – 0.95)
Table 7 HIV, health and development data for the highest governance ranking.
Country
Median (Range)
N = 49 2002 HIV prevalence (%) MMR (maternal deaths / 100,000 live births) in 2000 Physicians (per 100,000), 1990–2003 Improved drinking water (%) in 2002 Life Expectancy (years) in 2002 GDP-PPP ($ per capita) in 2002 GINI index 1994 – 2001 Ratio Health + Education / Health + Education + Military Spending in 2002
Voice & Accountability 0.2 (0.05 – 38.8) 10 (0 – 730) 287 (25 – 607) 100 (24 – 100) 76.8 (37.1 – 80.8) 15961 (3085 – 35894) 34.2 (24.4 – 63) 0.86 (0.20 – 1.00)
Political Stability 0.2 (0.05 – 38.8) 10 (0 – 850) 278 (5 – 607) 100 (24 – 100) 76.0 (37.1 – 80.8) 15108 (1001 – 35894) 32.7 (24.4 – 63.0) 0.85 (0.20 – 1.00)
Government Effectiveness 0.2 (0.05 – 38.8) 10 (0 – 420) 269 (5 – 607) 100 (24 – 100) 76.9 (37.1 – 80.8) 17122 (1122 – 35894) 35.3 (24.4 – 63.0) 0.85 (0.20 – 1.00)
Regulatory Quality 0.2 (0.05 – 38.8) 10 (0 – 730) 279 (25 – 607) 100 (24 – 100) 76.9 (37.1 – 80.8) 17122 (1911 – 35894) 33.1 (24.4 – 63.0) 0.85 (0.20 – 1.00)
Rule of Law 0.2 (0.05 – 38.8) 10 (0 – 730) 269 (29 – 607) 100 (24 – 100) 76.9 (37.1 – 80.8) 17122 (1911 – 35894) 35.3 (24.4 – 70.7) 0.85 (0.20 – 1.00)
Corruption 0.2 (0.05 – 38.8) 10 (0 – 730) 269 (5 – 607) 100 (24 – 100) 76.9 (37.1 – 80.8) 17122 (1122 – 35894) 35.3 (24.4 – 63.0) 0.85 (0.20 – 1.00)
Mean governance 0.2 (0.05 – 38.8) 10 (0 – 730) 279 (29 – 607) 100 (24 – 100) 76.9 (37.1 – 80.8) 17122 (1911 – 35894) 34.2 (24.4 – 63) 0.85 (0.20 – 1.00)
Discussion
It is possible to divide those nations affected by HIV / AIDS into three groups that approximate to governance ranking. The higher governance group is characterized by significant wealth and effective healthcare systems. The main challenges for these countries consists of the provision of sexual health services, health care access to marginalized groups, continuation of education and research into new and improved prevention and treatment strategies.
The HIV prevalence is generally low in higher governance group however this figure conceals differences found within specific population groups. For example in the USA, HIV prevalence amongst African American women is almost twenty three times that in whites [16]. Whilst in the UK, the prevalence of HIV amongst men who have sex with men (MSM) within London in 2001 was 100 times the national average [17]. The disparity in HIV prevalence amongst 'at risk' groups in the UK and US highlight the general difficulty of using the UNAIDS country HIV prevalence estimates. The quality of surveillance methods has been discussed and graded by UNAIDS surveillance teams, and it is clear that some HIV prevalence estimates are inaccurate [18].
Most sentinel surveillance methods use antenatal screening due to the ease of patient access and the benefits of provision of anti-retroviral treatment (ART) to prevent mother to child transmission. This surveillance strategy, though easier to implement, does not sample high risk groups such as MSM, intravenous drug users and commercial sex workers and thus underestimates the true HIV prevalence figure for the country. Within the 2002 UNAIDS HIV prevalence estimates used in this analysis there are at least four grades of surveillance. As highlighted in Table 3 some countries / regions were not included in formal UNAIDS surveillance and then there are countries where a UNAIDS report was commissioned yet no HIV prevalence estimate was provided. Reasons were not given as to why certain countries did not have a report commissioned. One fifth of the countries with UNAIDS reports quote a HIV prevalence estimate less than 0.1% and yet other health and economic indices would predict that this is an optimistic figure. Finally, there are those countries that report a HIV prevalence estimate greater than 0.1% which is complicated to varying degrees by the observer bias described above. Additional file 1 tabulates the UNAIDS HIV prevalence estimates for each of the 149 countries included in this analysis.
The governance dataset by Kaufmann et al in 2002 is the first global assessment of societal structures. These authors point out the variability inherent in collecting subjective material and, like UNAIDS, state the need to improve the quality of the data collected in subsequent analyses. Ideally sources would be able to freely report on each nation however, the extent of data available decreases in those nations with poorer governance. The relatively large margins of error within the governance data make direct cross-country comparisons difficult to interpret. Due to the variability of both the governance and HIV prevalence dataset the whole spectrum of data was chosen and subjected to non-parametric ranking analysis. The null hypothesis 'HIV prevalence is not associated with governance' is rejected for each dimension of governance with variations in the relative importance of different governance dimensions. Previously, Fareed Zakaria [19] has argued that democracy is less important in the development of a strong nation than the rule of law, corruption and political stability. The correlation coefficient of the voice and accountability dimension of governance with HIV prevalence was the lowest in this analysis somewhat supporting this contention.
Those countries in the lowest governance ranking group of governance are defined by poverty, ineffective health care systems, elevated HIV prevalence and significant international debt. The elevated HIV prevalence in many of these vulnerable countries was predicted more than a decade ago following the analysis of health, economic and human rights data [20]. Historically, international support has focused on short-term 'vertical' disease control strategies to tackle healthcare problems [21]. Long-term, 'horizontal' capacity building strategies are vital if HIV / AIDS is going to be effective managed in nations with limited healthcare infrastructure [22]. It has been shown in a number of resource poor settings that the provision of voluntary counselling and testing (VCT) for HIV is facilitated by the provision of free primary care services and ART [23]. The provision of effective primary care support to pregnant women is the most effective way to provide VCT services for HIV and thereby identify HIV positive mothers, prevent mother to child transmission and facilitate VCT of their partner(s). Like surveillance, this strategy though relatively effective fails to test and treat vulnerable 'high risk' groups within the population.
The poor are those most at risk of infectious disease. The role of poverty as a risk factor for disease has been clear for over 300 years [24]. Health and wealth are inextricably linked. All who become chronically ill enter a negative cycle of limited horizons. Indeed, what is true for the individual is equally true for the nation state. The effect of HIV on economic under performance and negative growth is testament to this. It is vital that essential healthcare is free, so that those that catch treatable infectious diseases are allowed to live. Encouragingly there are a few positive examples in resource poor countries, such as Uganda, Senegal and Cuba, where leadership, good communication and support of civil society have made a difference in their respective HIV epidemics. There are however many countries within this group of vulnerable nations that need the bulk of international healthcare and financial institution commitment in order to address their devastating healthcare challenges.
One example is Nigeria which is the most populous nation in Africa that has been a democracy since 1999. Despite its vast resource wealth, this country has suffered from repeated religious and ethnic conflicts that have compromised its development [25]. Only recently has the civilian government made HIV / AIDS its top priority and initiated some selective treatment programs. This change followed pressure from civil society and the military, in this latter group HIV prevalence is estimated to be 20% [26]. In contrast, South Africa is 115 mean governance ranking points higher than Nigeria, has had universal franchise for a decade and the per capita wealth is almost ten times greater than Nigeria yet the HIV prevalence is four times greater in South Africa. Some of the multiple factors that help explain the HIV prevalence in South Africa include: an earlier HIV epidemic, migration from high HIV prevalence neighbours, the violence and inequality of the Apartheid era and government inaction over the last decade. The government of South Africa has recently responded to national and international civil society pressure and has promised to provide ART free to all patients with advanced HIV disease [27].
Cuba and Haiti are islands with a similar population size and GDP-PPP per capita yet the HIV prevalence estimates are 0.05% and 6.1% respectively. HIV is thought to have entered Haiti from the USA via the sex trade in the early 1980s. The main exposure risk for Cuban nationals was from military and healthcare worker interaction with sub-Saharan Africa. Cuba was one of the first countries in the Americas to launch a nationwide HIV policy to contain transmission and care for those people living with HIV / AIDS [28]. Healthcare in Cuba is provided free to its citizens by the state and there is strong political commitment supporting health as well as national and international HIV / AIDS action. In contrast, there have been 33 coups in Haiti in the last two centuries of independence. Political instability in addition to other governance factors have been attributed to the lack of development of a responsive healthcare system [29].
As governance improves fewer women die in childbirth, more physicians exist per population, there is better access to improved water and life expectancy is longer. In addition with improvements in governance there is more GDP-PPP per capita, more equitable distribution of income and greater investment in health and education compared to the military. Interestingly, the median HIV prevalence estimates does not change between the lower and middle third groups of mean governance ranking despite a step-wise improvement in all other indices. The majority of middle ranking economies can be found within the middle governance group. Many of these countries have large populations where the HIV epidemics are set to explode. Three nations in this group who have the economic and technological power to halt their respective epidemics are Russia, China and India.
Russia has over 3 million intravenous drug users and relatively expensive ART that help to fuel the HIV epidemic [30]. The collapse of the USSR produced significant strain on the health of the people [31]. Life expectancy in Russia fell 9 years following its transition to a market economy and there has been a significant rise in 'social diseases' of Tuberculosis (TB), HIV and Hepatitis. Intravenous drug use accounts for approximately 80% of those infected with HIV however recently a new phase of the epidemic has developed that is driven by sexual transmission [32]. It is only since 2003 that there has been an increase in leadership and commitment at higher political levels to combat HIV and AIDS.
UNAIDS reported in 2002 that the number of overall infections in China increased 30% since 1998, with over 1 million people infected with HIV [33]. It is feared that China may soon experience an explosive and widespread HIV epidemic. Intravenous drug use and the sharing of contaminated needles in the south and north-west of China was one mechanism of initial transmission the other was unsafe practices among paid blood donors. Unsafe blood collections in the 1990s led to the appearance of HIV and subsequent AIDS deaths in Chinas central provinces. In response to this the Chinese authorities have recently announced that they are providing free ART in central provinces [34].
The first main focus of HIV in India was Mumbai where there is a large commercial sex work industry and the HIV prevalence reported amongst these workers is 50%. It is expected that HIV will become the largest cause of adult mortality in India in the coming decade. Despite the government making HIV its national topmost priority, any attempt to address the problem is hampered by its fractured health care infrastructure, poor literacy figures and widespread poverty [35]. At the end of 2003 the Indian government began providing free ART in eight government hospitals with the plan to expand it to a total of 25 centres [36].
The aim of this paper was to attempt to dissect out the role of governance in the HIV pandemic. It is not possible yet to determine if the relationship seen represents correlation or causation. Even though this first analysis alludes to causation, for those 149 countries with UNAIDS HIV prevalence data, the relationship will become clearer over time when it is possible to compare nations that appear similar today. Currently Brazil and India have equivalent overall governance and HIV prevalence estimates at 0.8% and 0.7% respectively. However when other health and economic indices are examined it is clear that India invest less than Brazil in health and education, has one quarter the number of physicians and double the MMR. The GDP-PPP per capita is three times greater in Brazil but it is more equitably distributed in India which is likely to contribute to equivalent life expectancy seen in both countries. India and Brazil are the main producers of generic ART. However Brazil, unlike India, has consistently provided strong political support for HIV / AIDS patients after the end of the military dictatorship in 1990. In 1996, the Brazilian government guaranteed by national law the permanent allocation of financial resources and universal access to care, including ART [37]. The current disparities between India and Brazils HIV treatment policy predicts that the Indian epidemic will progress more rapidly and is likely to impact on its development. All ten countries selected for discussion are summarised in Table 8.
Table 8 Ten selected countries through which the relationship of governance and HIV prevalence is discussed.
Country Mean Governance Ranking 2002 HIV prevalence (%) MMR (maternal deaths / 100,000 live births) in 2000 Physicians (per 100,000), 1990–2003 Improved drinking water (%) in 2002 Life Expectancy (years) in 2002 GDP-PPP ($ per capita) in 2002 GINI index 1994 – 2001 Ratio Health + Education / Health + Education + Military Spending in 2002
Haiti 7 6.1 680 25 71 49.3 1824 na 0.73
Nigeria 15 5.8 800 27 60 51.1 924 50.6 0.31b
Cuba 53 0.05 16 596 91 76.4 1717 na 0.79
Russia 65 0.9 45 420 96 67.3 7699 45.6 0.59
China 84 0.1 28 164 77 71.6 3535 40.3 0.33b
India 95 0.8 430 51 86 62.8 2136 37.8 0.65
Brazil 113 0.7 260 206 89 63.2 6477 60.7 0.77
South Africa 130 20.1 230 25 87 48.0 8466 59.3 0.85
United States 179 0.6 11 279 100 77.2 35831 40.8 0.78
United Kingdom 187 0.1 8 164 100 77.8 22801 36.0 0.82
b = no education expenditure data
na = not available
HIV / AIDS control in Russia, China and India will only be possible if they follow the example set by Brazil. International institutions need to support national civil society groups within these nations to focus the attention and resources of their respective governments for progressive healthcare changes. The global plan to stop TB outlines the possibilities and challenges that will be faced treating chronic illness, such as HIV [38]. It is pertinent that international health and financial institutions work together to influence change so that robust healthcare networks and responsive government are developed in order to apply best healthcare and economic practice.
The WHO goal of three million HIV positive persons being on ART by 2005 would be readily met if civil society in resource rich countries was able to precipitate progressive societal changes. Health is a fundamental human right, consequently each global institution, organization and citizen needs to work towards stable and progressive societal structures that can facilitate the provision of healthcare 'access for all'. The current HIV pandemic represents collective inaction and indifference towards global health. The promotion of good governance is a necessary step to enable national civil society to engineer long-term healthcare changes to deal with HIV / AIDS and future healthcare challenges.
Conclusion
Using World Bank governance data and UNAIDS HIV prevalence estimates for 2002 this paper tests the hypothesis 'HIV prevalence is not associated with governance'. Additional health and economic indices are used to highlight the development needs for each country. The accuracy of both governance and HIV prevalence estimates are discussed and some country comparisons are made. HIV prevalence is significantly associated with poor governance. International public health programs need to address societal structures in order to create strong foundations upon which effective healthcare interventions can be implemented.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
ASM-J designed the study, performed statistical analysis and wrote the manuscript. The author 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
Submitted Governance and Health additional file. Tabulates 149 countries by their governance ranking, HIV prevalence, health and economic data.
Click here for file
Acknowledgements
I would like to thank everybody who provided feedback and support for this work.
==== Refs
Sen A Development as Freedom Chapter 7: Famine and other crises 2001 Oxford University Press ISBN: 0192893300
World Health Organisation Macroeconomics and Health: Investing in Health for Economic Development 2001
United Nations Development Programme Human Development Report 2003. Millennium development goals: A compact among nations to end human poverty Oxford University Press
The World Bank Group Millennium development goals
Easterly W The Elusive Quest for Growth 2002 MIT press ISBN: 0-262-55042
Farmer P Infections and Inequalities 1999 Berkeley, University of California Press ISBN: 0-520-22913-4
World Health Organisation World Health Report 2004. Changing History
Kaufmann D Kraay A Mastruzzi M Governance Matters III: Governance Indicators for 1996–2002 World Bank Policy Research Working Paper 3106 2003
World Bank Group Governance Matters III: Governance Indicators for 1996–2002
WHO UNICEF and UNFPA Maternal Mortality in 2000: Estimates developed by WHO, UNICEF and UNFPA
United Nations Development programme Human Development report 2004
WHO WHO/UNICEF Joint Monitoring Programme for Water Supply and Sanitation; Meeting the MDG drinking water and sanitation target: a mid-term assessment of progress, 2004 ISBN 92 4 156278 1
Central Intelligence Agency The World Factbook 2004
United Nations Development Programme Inequality in Income or consumption Human Development Indicators 2003
International Institute for Strategic Studies The Military Balance 2004–2005 2004 Oxford University Press ISSN 0459-7222
Centre for Disease Control HIV/AIDS among African Americans 2002
Health Protection Agency HIV and other Sexually Transmitted Infections in the United Kingdom in 2003 Annual Report 2004
Garcia-Calleja JM Zaniewski E Ghys PD Stanecki K Walker N A global analysis in trends in the quality of HIV sero-surveillance Sex Transm Infect 2004 80 i25 i30 15249696 10.1136/sti.2004.010298
Zakaria F The Future of Freedom 2003 W.W.Norton & company, ISBN: 0-393-04764-4
Mann JM Tarantola DJ Netter TW AIDS in the World Chapter 14: Assessing vulnerability to HIV Infection and AIDS 1992 Harvard University press ISBN0-674-01265-8
Tan DHS Upshur REG Ford N Global plagues and the Global Fund: Challenges in the fight against HIV, TB and malaria BMC International Health and Human Rights 2003 3
Scalway T Missing the Message? 20 years of learning from HIV / AIDS Panos ISBN 1 87067065 5
Partners in Health The PIH Guide to the Community-Based Treatment of HIV in Resource-Poor Settings XV International AIDS Conference, Bangkok Edition Partners in Health 2004 ISBN: 0-9744222-1-5
Dale P Sir WP of Romsey LVTAS Group 1987 ISBN 0-906921-05-8
UNAIDS National Response Brief on Nigeria 2002
Elbe S HIV / AIDS and the changing landscape of war in Africa International Security 2002 27 159 177 10.1162/016228802760987851
Plus News SOUTH AFRICA: Chronology of HIV/AIDS treatment plan, August 2003 to April 2004 The HIV / AIDS news service 1st April, 2004
Farmer P The Pathologies of Power 2003 University of California Press ISBN: 0520235509
Farmer P Political Violence and Public Health in Haiti NEJM 2004 350 1483 1486 15071121 10.1056/NEJMp048081
UNAIDS Report on the global AIDS epidemic 2004
Garrett L Betrayal of Trust: The Collapse of Global Public Health 2001 Oxford University Press ISBN: 0198509952
UNAIDS National Response Brief on Russia 2002
UNAIDS National Response Brief on China 2002
Xinhau News Agency Guangdong to offer free HIV / AIDS treatment 31st May, 2004
Editorial Political neglect in India's health The Lancet 2004 363 1565 15145624 10.1016/S0140-6736(04)16233-8
National AIDS Control Organization Anti-retroviral Treatment: A new initiative 2003
UNAIDS National Response Brief on Brazil 2002
World Health Organisation Progess Report on the global plan to Stop Tuberculosis 2004
| 15850480 | PMC1112596 | CC BY | 2021-01-04 16:29:56 | no | BMC Int Health Hum Rights. 2005 Apr 25; 5:4 | utf-8 | BMC Int Health Hum Rights | 2,005 | 10.1186/1472-698X-5-4 | oa_comm |
==== Front
BMC Med EducBMC Medical Education1472-6920BioMed Central London 1472-6920-5-121585048210.1186/1472-6920-5-12Research ArticleBeing uninformed on informed consent: a pilot survey of medical education faculty Mavis Brian E [email protected] Rebecca C [email protected] Office of Medical Education Research and Development, A202 East Fee Hall, Michigan State University, East Lansing, Michigan, 48824-1316, USA2005 25 4 2005 5 12 12 28 10 2004 25 4 2005 Copyright © 2005 Mavis and Henry; licensee BioMed Central Ltd.2005Mavis and Henry; 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
This paper describes a pilot survey of faculty involved in medical education. The questionnaire focuses on their understanding of IRB policies at their institution, specifically in relation to the use of student assessment and curriculum evaluation information for scholarship.
Methods
An anonymous survey was distributed to medical educators in a variety of venues. Two brief scenarios of typical student assessment or curriculum evaluation activities were presented and respondents were asked to indicate their likely course of action related to IRB approval. The questionnaire also asked respondents about their knowledge of institutional policies related to IRB approval.
Results
A total of 121 completed surveys were obtained; 59 (50%) respondents identified themselves as from community-based medical schools. For the first scenario, 78 respondents (66%) would have contact with the IRB; this increased to 97 respondents (82%) for the second scenario. For both scenarios, contact with the IRB was less likely among respondents from research-intensive institutions. Sixty respondents (55%) were unsure if their institutions had policies addressing evaluation data used for scholarship. Fifty respondents (41%) indicated no prior discussions at their institutions regarding IRB requirements.
Conclusion
Many faculty members are unaware of IRB policies at their medical schools related to the use of medical student information. To the extent that policies are in place, they are highly variable across schools suggesting little standardization in faculty understanding and/or institutional implementation. Principles to guide faculty decision-making are provided.
==== Body
Background
The Liaison Committee for Medical Education (LCME) expects medical schools to implement strategies for student assessment and curriculum evaluation that facilitate educational program management and improvement. In addition there is an expectation that faculty involved in these activities will reflect on what is learned and participate in the community of scholars to share this knowledge [1]. In practice, this has created confusion among faculty members involved in educational evaluation. Many faculty members are not clear under what circumstances institutional review board (IRB) approval is necessary. In some cases they conduct educational evaluations solely for institutional program improvement, while in other situations similar activities are undertaken to produce generalizable knowledge of interest to others [2,3]. As frequently occurs, student assessment and program evaluation activities intended for internal program monitoring or improvement yield outcomes that can lead to generalizable knowledge appropriate for scholarly publication, further blurring the distinction between evaluation and research [4]. A case in point were the recent allegations against the American Association of Medical Colleges, claiming that medical students completing the Graduation Questionnaire (GQ) were acting as subjects in a research study and human subjects protections were not followed. While many of the allegations were not upheld, the AAMC did agree to submit the GQ to an IRB for review [5].
There has been increasing debate among medical educators about the need to submit proposals for scholarship arising from student assessment and curriculum evaluation activities for IRB review. Here, the use of the word "scholarship" is used deliberately as it is not always synonymous with research, where there are clear criteria governing IRB review. The Common Rule defines research as "the systematic investigation, including research development, testing and evaluation, designed to develop generalizable knowledge [6]." Scholarship based upon quality improvement and program evaluation initiatives can fall into that uncertain territory that sometimes is equated with research while at other times is pursued for the single purpose of improving programs. Defining the point at which an internally directed program inquiry becomes publicly shared scholarship can vary across disciplines and/or the institution where the decision is rendered. Casarett and colleagues have published a more in depth discussion of criteria that might help distinguish research from other initiatives to document or improve program quality [7]. Our informal observation is that many medical educators set out to conduct evaluations to improve educational programs but after the evaluation is completed, the possibility of publishing or presenting the findings emerges, clouding the intent of the original activity. To add to the definitional murkiness, the determination of whether an initiative is research or some other activity is made locally by the institutional IRB, leading to variability across institutions.
In the context of evaluating educational programs, the question of human subjects protections for medical students is not new; it was posed over 20 years ago by Christakis [8]. Recent literature in this area suggests there is great variability in the extent to which medical schools address this issue [4]. Originally created to protect against abuses of human subjects in federally funded clinical research, 45 CFR 46 [9] is increasingly being applied more broadly to include research in the social sciences and education. Because medical education has historically used the curriculum and its related educational components as a laboratory for studying the teaching and learning process, ultimately with the goal of improving professional competency, the distinction between improvement and research becomes significant [4]. This tradition is consistent with Boyer [10], who urged educators to consider a broader view of scholarship, targeting the curriculum and the classroom as a source of inquiry. Faculty have been encouraged – if not mandated – to wear multiple hats [1,4], first as educators who participate in on-going teaching and evaluation, and second as scholars who uncover interesting observations and findings leading to more generalizable knowledge disseminated via professional meetings and publications. Given that accreditation and outcome assessment increasingly rely on learner performance data and that faculty are urged to publish from these efforts, the issue of using routine student assessment and program evaluation data for research remains salient. Until recently few medical education journals required evidence of IRB approval for manuscripts accepted for publication [4].
This study describes a pilot survey of faculty involved in medical education, and focuses on their understanding of IRB policies at their institutions related to the use of student information derived from assessment and evaluation activities for scholarship.
Methods
We developed a brief questionnaire that described two short scenarios of typical student assessment or curriculum evaluation activities. For both cases, respondents were asked to choose from among several likely courses of action related to IRB approval: (a) submit an IRB application, (b) talk with the IRB chair, or (c) submit a conference abstract without IRB review.
Case study 1
Your department has responsibility for the on-going evaluation of the clinical skills curriculum for preclinical medical students. In reviewing students' test scores from the course multiple-choice exams and faculty performance ratings of students, you identify some interesting relationships. In discussing these findings with the course director, you both agree that they have educational significance and decide to submit an abstract based on these data for the next regional medical education conference.
Case study 2
Your department has responsibility for the on-going evaluation of the clinical skills curriculum for preclinical medical students. The clinical skills course coordinator inquired about the relationship between student performance in the first and second year of the clinical skills curriculum. Of particular interest is the bottom 20% of students based on faculty performance ratings. To answer this question, first and second year videotaped interviews for the bottom 20% of the class were reviewed by three faculty members, who rated the performance using standard checklists and rating scales. After reviewing the analyses of the data, the course director and faculty raters decided to submit an abstract for a regional medical education conference presentation.
The second section of the questionnaire asked respondents about their knowledge of institutional policies related to IRB approval. The final questionnaire item asked respondents to indicate if they were from a community-based medical school, research-intensive medical school or some other medical school/health professions program. This study was reviewed and approved by the Michigan State University Committee for Research Involving Human Subjects.
The anonymous survey was administered in person to 121 medical educators in a variety of different venues from Fall 2001 through Fall 2003 (Table 1). All of the fellowship programs included in this study focused on developing skills necessary for careers in medical education research.
Table 1 Recruitment venues for survey respondents
Group Date Number % of Total Sample
Association of American Medical Colleges conference workshop Nov. 2001 5 4%
Central Group on Educational Affairs conference session March 2002 19 16%
Michigan State University primary care faculty development fellowship program March 2002 19 16%
OBGYN faculty development fellowship March 2002 20 17%
Surgery educational research fellowship April 2002 9 7%
Michigan State University primary care faculty development fellowship program Sept. 2002 19 16%
Surgery educational research fellowship April 2003 11 9%
Michigan State University primary care faculty development fellowship program Sept. 2003 19 16%
This was a descriptive study: the results were analyzed in terms of frequencies of responses for each of the questionnaire items. In addition, the open-ended responses were reviewed for themes, which were categorized for presentation in terms of percentages.
Results
A total of 121 completed questionnaires were received. The exact response rate is unknown for the conference-based sessions due to incomplete records, but is greater than 90% for the fellowship sessions. Since not all parts of the questionnaire were completed, the specific sample sizes for each item are presented in the tables. The only identifier was the type of institution represented: 59 (50%) from community-based medical schools, 50 (42%) from research-intensive medical schools and 10 (8%) designated as another type of institution.
Case studies
When presented with the first case study, respondents were fairly equally divided among the three courses of action provided; approximately two-thirds of the respondents would submit an IRB application or talk with the IRB chair (Table 2). For both cases, a small number of respondents indicated that they were unsure as to their likely course of action. Twenty-one respondents from research-intensive medical schools (43%) indicated that they would submit the abstract without involving the IRB compared to thirteen respondents (22%) from community-based medical schools (Chi-Sq = 6.34, df = 3, p = 0.09, Odds Ratio = 2.60).
Table 2 Responses to case study scenarios for research intensive and community-based medical schools
Response to Case Scenario
Submit IRB application Talk with IRB chair Submit abstract without IRB review Don't know
Case 1: χ2 = 6.34
All respondents (N = 119)1 35 (29%) 43 (36%) 39 (32%) 2 (2%)
Research-intensive (N = 49) 14 (29%) 14 (29%) 21 (43%) 0 (0%)
Community-based (N = 58) 21 (36%) 22 (38%) 13 (22%) 2 (3%)
Case 2: χ2 = 13.25 **
All respondents (N = 118)1 56 (47%) 41 (35%) 19 (16%) 2 (2%)
Research-intensive (N = 49) 20 (41%) 14 (29%) 15 (31%) 0 (0%)
Community-based (N = 57) 30 (53%) 22 (39%) 3 (5%) 2 (4%)
1: Includes respondents who classified their institution as "other"
* p < .05; ** p < .01
For the second case study, ninety-seven (82%) respondents indicated involvement of the IRB. Though there was more IRB consultation overall for this case study, respondents from research intensive institutions more often reported (31% vs. 5%) that they would submit the abstract without IRB consultation than subjects from community-based schools (Chi-Sq = 13.25, df = 3, p = 0.004, OR = 7.94).
Knowledge about institutional policies
Sixty-one respondents (52%) indicated that their IRB was university-based, compared to forty-six (39%) who interact with IRBs through their medical center (Table 3). This dichotomy is important insofar as it distinguishes between a centralized university IRB addressing all human subjects concerns and institutions where there are multiple specialized IRBs. Many of the applications to university-based IRBs focus on the protection of students as human subjects, and IRB members are likely to have more experience with educational research protocols. Historically, IRBs within medical centers have focused on the protection of patients as human subjects, both as clinical research participants and more recently with the introduction of quality improvement initiatives.
Table 3 Knowledge of institutional policies for research intensive and community-based medical schools
Respondent Group Test Statistic
All Respondents1 Research Intensive Community-Based
For medical education studies that require IRB review, to which IRB would you submit your application? (N = 117)
University 61 52% 29 58% 27 47% χ2 = 3.66
Medical center 46 39% 20 40% 24 42%
Other 9 8% 1 2% 5 9%
Don't know 1 1% 0 0% 1 2%
Does your institution have stated policies on the use of existing educational evaluation data for faculty scholarship? (N = 110)
Yes 30 27% 7 16% 11 20% χ2 = 1.71
No 20 18% 15 33% 12 22%
Unsure 60 55% 23 51% 32 58%
Which best describes procedures in place at your medical school for obtaining consent from students to use their performance data and test scores for educational research & scholarship? (N = 115)
Students can decline consent 24 21% 6 12% 3 5% χ2 = 2.75
Matriculation conditional on consent 11 10% 8 16% 14 25%
There are no procedures 50 43% 21 43% 26 46%
Don't know 30 26% 14 29% 13 23%
Have you participated in discussions with others at your institution about IRB requirements for using evaluation data for faculty scholarship? (N = 119)
Faculty in your department 53 45% 23 46% 24 41% χ2 = 0.31
Faculty in other departments 24 20% 13 26% 11 19% χ2 = 0.85
College faculty meetings 9 8% 6 12% 0 0% χ2 = 7.49**
Dean, administrator, etc. 19 16% 12 24% 5 9% χ2 = 4.96*
Other 17 14% 6 12% 9 15% χ2 = 0.24
No discussions reported 50 41% 21 42% 25 42% χ2 = 0.002
1: Includes respondents who classified their institution as "other"
* p < .05; ** p < .01
Sixty respondents (55%) were unsure if their institutions had policies in place addressing the use of educational evaluation data for scholarly dissemination. Thirty respondents (27%) indicated that their institutions did have a policy in place. Only eleven respondents (10%) reported that their institution made matriculation contingent on students providing consent to have their academic information used for faculty research. Twenty-four respondents (21%) indicated that students can decline to give consent for participation in faculty scholarship. There were no differences in the responses for participants from research-intensive institutions compared to community-based medical schools.
Twenty-four of the participants (20%) provided written comments elaborating their responses about institutional policies; some respondents made multiple comments so the number of comments exceeds the number of respondents. Six of this subgroup (30%) reported simply that an IRB application was required, while some indicated the specific level of review as exempt (N = 7; 35%) or expedited (N = 4; 20%). Three respondents (15%) specified that the proposal was exempt only if the data were anonymous, and one respondent (5%) indicated that students must provide consent. One respondent (5%) wrote that there was no explicit policy but faculty were advised to consult with the IRB chair, while two respondents (10%) distinguished between evaluation for program improvement versus generalizable knowledge leading to scholarship. Another faculty member (5%) replied that secretarial staff took care of matters related to the IRB. Finally, one respondent (5%) confessed that he or she had never heard of this issue prior to involvement in the fellowship program.
Discussions with colleagues
Respondents were asked if they had participated in discussions with others at their institutions related to IRB approval for the use of student evaluation data for faculty scholarship. A significant proportion of respondents (N = 50, 41%) did not report any prior discussions with others at their institution regarding IRB requirements; only one respondent (0.8%) reported participating in discussions with all five of the groups listed. Overall, when faculty reported engaging in such discussions, it was most likely to have occurred with faculty members within their own (N = 53, 45%) or other (N = 24, 20%) departments. This was consistent across respondent subgroups. More respondents from research intensive institutions reported discussions at college level faculty meetings (Chi-Sq = 7.49, df = 1, p = 0.006) and with institutional administrators (Chi-Sq = 4.96, df = 1, p = 0.03, OR = 3.41) than respondents from community-based medical schools.
Discussion
There has been increasing interest among medical educators about the need for human subjects protections for faculty scholarship derived from student assessment and curriculum evaluations activities. Tomkowiak and Gunderson [1] recently mused how many medical educators were aware that scholarship derived from the evaluation of a curricular innovation could be considered research subject to federal human research standards and governance? The results of this pilot survey suggest that many faculty members are unaware of relevant policies at their home institutions; to the extent that policies for human subjects review and approval are in place, faculty understanding and reporting of these policies are highly variable across institutions. The implication is that many faculty active in educational research and evaluation lack an understanding of their institutions policies regarding the use of students or other learners as research subjects. In rare cases, faculty delegate this responsibility to staff members. The variability with which institutions have addressed this issue, or how faculty have acted on these concerns, makes discourse about the human subjects concerns difficult for those involved in medical education as a profession. It suggests a lack of standards and standardization with respect to IRB review of educational research of a magnitude less common in other fields of scientific inquiry.
In practice, IRB standards are subject to local interpretation when institutional procedures are established. The data from the two case studies presented suggest that while a majority of faculty members would minimally seek IRB consultation for educational research, respondents from research-intensive institutions have been less likely to involve the IRB. Respondents from research-intensive institutions were more likely to engage in discussions about this issue. This could mean that there is a better shared understanding of the policies and practices at the specific institution that might not require application for IRB review for educational research. Conversely, it could signal a continuing legacy of research practices among faculty that is independent of institutional policy as well as Office of Human Research Protections (OHRP) guidelines, which indicate that the IRB not the investigator determines whether a study qualifies for exempt or expedited review [11]. Additional information distinguishing between exempt, expedited and full review procedures can be found at the OHRP website [11]. In some cases, respondents reported that they would contact the IRB chair, presumably for consultation. This demonstrates critical information seeking behavior and such consultation would likely yield a recommendation from the IRB chair about the need to apply for IRB review. Nonetheless, consultation with the IRB chair cannot be construed as equivalent to seeking IRB review since the faculty member's ultimate course of action is unknown. These findings highlight the importance of institutional culture in shaping faculty practices related to research and scholarship.
This study focuses on educational research and evaluation with students as human subjects. Much of the information used in such investigations, such as grades, test scores and performance ratings, are natural byproducts of students in their role as learners. These investigations appear to be minimal risk and are typically considered exempt by IRBs when the information is low-risk, presented in aggregate and used by faculty members responsible for the specific curricular component. However, the complexity of educating medical students presents many opportunities for collecting a wide range of personal information, linking student information across datasets, studying subgroups of students and in some instances, providing faculty with access to student information to which they might not otherwise have access. It is likely that students are not aware of the extent to which this occurs within their institution or the circumstances and protections surrounding such occurrences. In situations such as these, IRB review can weigh the risks and benefits to assure adequate protections for students. Some might argue that IRB review is necessary even when the information is used for internal program monitoring and quality control [7]. Involvement of the IRB in educational evaluations does not have an impact on students' obligation to complete the requirements of their curriculum, but can have an impact on whether or not this information can be used for other purposes such as conference presentations or publications.
This study was designed as an exploratory descriptive investigation, and as such is limited as a pilot sample based on a small number of respondents. However we have attempted to sample across a variety of cohorts of faculty actively involved in medical education. Compared with other medical school faculty one could hypothesize that they should be better informed about IRB policies as their professional emphasis is in medical education and the products of their work would likely involve students as subjects. Faculty have the right and responsibility to collect information, including student assessments, to improve instruction, curricula and educational outcomes. However, this right does not extend to the use of this information for publication or scholarship without student consent or IRB waiver of consent.
Because the questionnaire was anonymous little is known about the respondents in terms of their academic background and training or experience in medical education. In addition, the findings are derived for faculty reports of institutional policy rather than a review of actual institutional policies. However, even the preliminary comparisons of individuals from different types of institutions show significant differences suggestive of variability within the profession. A more systematic sample of subjects and a questionnaire that includes more respondent information is needed to provide a more comprehensive perspective faculty knowledge and practice related to human subjects concerns for educational evaluation and research.
Conclusion
Many faculty members lack understanding about their institutions policies for IRB review of educational evaluation and other quality improvement strategies. Professional organizations and journals in medical education could assist by developing position statements and clear expectations about protecting students when their performance and survey data are used in publications and presentations.
Given this inconsistent understanding of institutional policies, what might assist in educating faculty about when to seek IRB review? Some have suggested a series of questions that might guide us on when educational research may be exempt [12]. Central to these questions is:
1. When conducting an evaluation, faculty should consider whether or not the activity is research (is it designed to contribute to generalizable knowledge?).
2. If the activity is research, do the learners meet the criteria for human subjects?
3. If the activity is human subject research, the faculty member should seek IRB consultation regarding whether or not the research activity is exempt.
4. If research was not an original intent of the evaluation, but the faculty member later determines that the results of the evaluation might contribute to generalizable knowledge, the above principles become applicable.
For any evaluation the Office of Human Research Protections (OHRP) recommends that the local IRB or another independent institutional authority be consulted in making these determinations [11].
Perhaps a more fundamental concern is for those faculty members who never seek IRB review or consultation in the first place. Our advice to all faculty involved in activities that could be construed as medical education research is to assume that any medical students or others involved in the educational process, such as residents, faculty or standardized patients, might well meet the criteria to be considered human subjects and appropriate consultation should be sought.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
BM contributed to the design of the study, analysis of the data and preparation of the first draft of the manuscript. RH contributed to the design of the study, data collection and revisions to the draft manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
None.
==== Refs
Tomkowiak JM Gunderson AJ To IRB or not to IRB? Acad Med 2004 79 628 632 15234912 10.1097/00001888-200407000-00004
Morrison J Prideaux D Ethics approval for research in medical education Med Educ 2001 35 1008 11703634 10.1046/j.1365-2923.2001.01076.x
DuBois JM When is informed consent appropriate in educational research? Regulatory and ethical issues IRB: Ethics and Human Research 2002 24 1 8 14509289
Roberts LW Geppert C Connor R Nguyen K Warner TD An invitation for medical educators to focus on ethical and policy issues in research and scholarly practice Acad Med 2001 76 876 885 11553501
Borror K Human research subject protections under federalwide assurance (FWA) 1666 activities involving the Graduation Questionnaire (GQ)
Protection of Human Subjects, 56 Federal Register 28003 (1991) [codified at 45 CFR 46]. United States
Casarett D Karlawish JHT Sugarman J Determining when quality improvement initiatives should be considered research: Proposed criteria and potential implications JAMA 2000 283 2275 2280 10807388 10.1001/jama.283.17.2275
Christakis N Do medical student research subjects need special protection? IRB: Ethics and Human Research 1985 7 1 4 11649647
Henry RC Wright DE When do medical students become human subjects of research? The case of program evaluation Acad Med 2001 76 871 875 11553500
Boyer E Scholarship Reconsidered: Priorities of the Professoriate 1990 Princeton NJ: Carnegie Foundation for the Advancement of Teaching
Office for Human Research Protections (OHRP); Department of Health and Human Services
Institutional Review Board Decision Tree University of Arizona
| 15850482 | PMC1112597 | CC BY | 2021-01-04 16:30:57 | no | BMC Med Educ. 2005 Apr 25; 5:12 | utf-8 | BMC Med Educ | 2,005 | 10.1186/1472-6920-5-12 | oa_comm |
==== Front
BMC Musculoskelet DisordBMC Musculoskeletal Disorders1471-2474BioMed Central London 1471-2474-6-221587174210.1186/1471-2474-6-22Research ArticleThe factor validity of the Western Ontario Rotator Cuff Index Wessel Jean [email protected] Helen [email protected] Yasmin [email protected] Richard [email protected] School of Rehabilitation Science, McMaster University, Hamilton, Canada2 Department of Rehabilitation, University of Toronto, Orthopaedic and Arthritic Institute, Sunnybrook & Women's College Health Sciences Centre, Toronto, Canada3 Student in Medical Programme, University College Cork, Cork, Ireland4 Department of Surgery, University of Toronto, Orthopaedic and Arthritic Institute, Sunnybrook & Women's College Health Sciences Centre, Toronto, Canada2005 4 5 2005 6 22 22 17 1 2005 4 5 2005 Copyright © 2005 Wessel et al; licensee BioMed Central Ltd.2005Wessel 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 Western Ontario Rotator Cuff Index (WORC) is a self-report questionnaire developed specifically to evaluate disability in persons with pathology of the rotator cuff of the shoulder. The authors created items in 5 categories based on a model of quality of life, but never validated this structure. The purpose of this study was to examine the validity of the original 5-domain model of the WORC by performing factor analysis.
Methods
Three hundred twenty nine subjects (age, mean: 52, SD: 12) were tested prior to undergoing surgery for rotator cuff pathologies. They completed the WORC, a self-report questionnaire, which has 21 items on the effect of the rotator cuff problem on symptoms, activities and emotions. Statistical calculations included correlations between items, Cronbach's alpha of the total scale and subscales, and principal component factor analysis with oblique rotation.
Results
Correlations ranged from .09 to .70 between all the items, from .29 to .70 between items within a subscale, and from .53 to .72 between subscale scores. Cronbach's alpha was .93 for the total scale, and .72 to .82 for the subscales. The factor analysis produced 3 factors that explained 57% of the variance. The first factor included symptoms and emotional items, the second included strength items and the third included daily activities.
Conclusion
The results of this study did not support the 5-domain model of the WORC.
==== Body
Background
The Western Ontario Rotator Cuff Index is a recent self-report questionnaire that was designed to measure "health-related quality of life" in persons with injuries and conditions of the rotator cuff of the shoulder. Kirkley et al [1] felt the measure should represent the impact of the condition on health as defined by the World Health Organization – "a state of complete physical, mental and social well-being". They, therefore, included items in 5 domains in the questionnaire: 1) pain and physical symptoms, 2) sports and recreation, 3) work, 4) lifestyle, and 5) emotions. The authors followed a systematic, clinimetric method of generating and reducing the items. This resulted in 21 items that respondents answered on visual analogue scales (VAS) with anchors such as no pain/difficulty and extreme pain/difficulty.
Items for the WORC were derived from published health status scales, functional measures of the shoulder, discussions with healthcare professionals, and interviews with 30 patients from a registry of 150 with rotator cuff pathology. Both professionals and patients were asked to identify ways in which the shoulder condition affected quality of life in general, and the 5 domains in particular. The 30 patients interviewed included males and females, aged 30–76, with different degrees of rotator cuff pathology from tendinitis to massive tears.
An original list of 321 items was reduced to 76 by the investigators eliminating duplicated, incomprehensible or ambiguous items. A random selection of 100 patients from the same registry were then asked to indicate whether they experienced each of the items, and to rate the importance of the symptom/disability to their overall shoulder functioning. A frequency importance product was calculated for each item and the 50 items with the highest values were correlated with each other. For every pair of items with coefficients greater than 0.6, one of the items was eliminated, resulting in the final 21 questions. It is not clear whether this criterion applied to items across domains because the only example provided included 3 items from the same domain.
In that same paper [1], the authors reported an ICC for reliability of .96 when they tested subjects over a 2-week period and omitted those who reported any change on a global rating scale. The ICCs for the subscales ranged from .54 (4-item work) to .91 (6-item physical symptoms).
Construct validity has been tested by the original authors [1] and others [2,3] by examining the correlation of the WORC with other shoulder measures (Constant, American Shoulder and Elbow Surgeons Standardized Shoulder Assessment Form [ASES], Disabilities of the Arm, Shoulder and Hand [DASH], University of California Los Angeles [UCLA], Simple Shoulder Test [SST]). The correlations of the WORC total score with the other instruments have ranged from .48 to .91, with generally higher correlations with instruments that have disability items similar to those in the WORC (see Table 1). The correlations of the change scores were in a similar range (.44 to .85).
Table 1 Correlations of the WORC with other shoulder measures
Article Scores Constant ASES DASH UCLA SST
Kirley et al [1] Cross-sectional .57 .63 .48
Change .44 .76 .66 .72
Holtby & Razjmou [3] Cross-sectional .61–.66 .73
Change .70–.77 .85
Getahun et al [2] Cross-sectional .88 .91
Two studies [3,4] have examined the responsiveness of the WORC and other shoulder measures by calculating the standardized response mean (SRM) in patients who have been measured before and after surgery (see Table 2). It should be noted that the SRM of the WORC was not noticeably different from the comparative measures (Constant, SST and DASH) in the same study. Holtby and Razmjou [3] had lower overall SRMs than MacDermid et al [4] who included only the responders in their calculations. MacDermid et al [4] also reported the SRM for the subscales of the WORC. The values ranged from 1.2 for the work subscale to 1.8 for the lifestyle subscale.
Table 2 Standardized response means for the WORC and other shoulder measures
WORC Constant (absolute) Constant (relative) ASES DASH SST
Holtby & Razjmou [3] 1.3 1.4 1.3 .9
MacDermid et al [4] 2.1 1.8 2.0
When Kirkley et al [1] developed the WORC, they argued for the use of "disease-specific" measures to evaluate orthopedic treatment because they are more responsive than global health measures. However, they set out to develop, not only measures specific to the shoulder, but instruments specific to conditions of the shoulder. Now there exist Western Ontario tools for the measurement of disability in shoulder instability (WOSI) [5] osteoarthritis [6] and rotator cuff conditions [1].
The results provided above, however, suggest that generic measures of the shoulder may perform as well as condition-specific measures. The WORC was highly correlated with both the DASH and the SST [2], and had a standardized response mean similar to these two instruments [4]. Therefore, it may not be necessary to have a tool that is specific to a particular condition in the shoulder. Moreover, the WORC is more time consuming to complete and to score, and may not be as attractive as the other scales for use in a clinical setting.
One advantage of the WORC may be its comprehensiveness. It was designed to tap 5 domains of health and may provide information that is unavailable in the other measures. However, the subscales have not been studied in detail, nor has there ever been a confirmation that the WORC items fall into the 5 domains. The purpose of the present study was to examine the validity of the original 5-domain model of the WORC by performing factor analysis.
Methods
Subjects
The data were drawn from a database that included all patients who were to undergo arthroscopic acromioplasty for surgical management of impingement or rotator cuff pathology of the shoulder (Table 3) in a tertiary level hospital in Toronto, Canada between October 2000 and July 2004. Complete data were available on 329 (196 males, 133 females) out of a total of 334 patients. All patients subsequently underwent arthroscopic acromioplasty with or without rotator cuff repair. A number of patients had superior labral anterior and posterior (SLAP) lesions that required surgical repair.
Design
The data of this study were prospectively collected. All patients were sent a number of questionnaires that included the WORC, 3–4 weeks before surgery via mail. Just prior to surgery, the patients were then seen by a physical therapist who performed some physical tests (not reported in this study), and checked that the questionnaires were completed. The data extracted for this study included demographics (Table 3) and the scores on each of the individual items of the WORC questionnaire.
Table 3 Subject characteristics
Variable Mean or Frequency* Percent
Age (years) 52.2 (23–81)
Gender (female, male) 133, 196 40, 60
Duration of symptoms
≤ 1 year 109 33
1–2 years 151 46
>2–5 years 3 1
>5 years 53 16
missing data 13 4
Dominant side
Right 294 89
Left 26 8
Ambidextrous: 9 3
Affected side
Right 202 61
Left 127 39
General health status
Good 184 56
Diabetes 27 8
Chronic illness 60 18
Other 47 14
Missing data 11 3
Surgery
Acromioplasty 329 100
Resection of distal clavicle 107 32
Rotator cuff repair 96 29
SLAP repair 32 10
*Mean and range of age are provided. The remaining values are frequencies
WORC measure
As indicated previously, the WORC is a 21-item questionnaire examining the impact of rotator cuff pathology on "quality of life". Subjects answer each question on a 100 mm visual analogue scale and the higher numbers indicate worse pain or difficulty. The questions in each of the theoretical domains are presented together. The WORC total is obtained by adding the scores on all the items. The subscale scores are totals of the item scores in that domain.
The WORC questionnaire has been published in full [1]. However the 1998 copyright version obtained from the authors and used in this study varies slightly from the published version. The minor differences are noted in Table 4.
Table 4 Differences between copyright and published versions of WORC
Item Published version [1] 1998 copyright version used in present study
PS 4 How much stiffness do you... How much stiffness or lack of range of motion do you...
PS 5 How much do you experience clicking... How much are you bothered by clicking...
PS 6 How much discomfort do you experience in your neck... How much discomfort do you experience in the muscles of your neck...
SR 8 SR 9
SR 9 SR 10
SR 10 SR 8
W 12 How much...above your head How much...above your shoulder
W 14 How much...objects from the ground or below shoulder level How much...objects at or below shoulder level
W 21 How worried...occupation or work? How worried...occupation?
PS physical symptoms
SR sports and recreation
W work
Data analyses
Descriptive statistics were calculated for the 21 items, for the subscale scores and for the total WORC score. Correlations between the items and between the subscales were examined with Pearson Product Moment Correlations. A Cronbach's alpha was calculated to determine the internal consistency of the total score and the subscale scores.
Principal component analysis was the extraction method used for the factor analysis. Only factors with eigenvalues greater than 1 were considered. The Kaiser-Meyer-Olkin Measure and Bartlett's Test of Sphericity were performed to determine whether the data were suitable for factor analysis. [7] Because all the subscales were correlated, an oblique rotation method (SPSS direct oblimin option, SPSS version 11.0.1, SPSS Inc) was used. An item was considered to be loaded on a factor if its pattern matrix coefficient was .5 or greater. We also noted those items that loaded between .4 and .5 but had no higher loading on another factor.
Results
The descriptive statistics, alpha coefficients and inter-item correlations are outlined in Table 5. The correlations between items ranged from .09 to .70, with the lowest correlations being between the emotion items and two of the sports/recreation items. The correlations between items within a subscale varied between .29 and .70. The correlation between subscale scores varied between .53 and .72. Internal consistency of the subscales was .72 to .82 (Cronbach's alpha). The Cronbach's alpha for the total scale was .93.
Table 5 Descriptive statistics for WORC item and subscale scores
Item* Item mean Item SD Subscale mean (SD) Cronbach's alpha Range of inter-item correlations within subscale
PS1 66 25.5 382 (121.7) .81 .32 – .60
PS2 64 27.3
PS3 71 24.9
PS4 68 24.9
PS5 57 33.7
PS6 56 32.2
SR7 69 28.5 302 (81.8) .72 .29 – .55
SR8 86 21.6
SR9 83 27.6
SR10 64 32.2
W11 65 25.7 296 (78.5) .78 .38 – .64
W12 85 19.1
W13 77 24.4
W14 69 30.8
LS15 68 29.0 255 (96.3) .82 .42 – .68
LS16 56 34.3
LS17 75 27.3
LS18 55 28.4
E19 77 27.1 196 (84.8) .80 .52 – .70
E20 58 34.6
E21 62 37.7
PS physical symptoms
SR sports and recreation
W work
* See Table 6 for explanation of items
The data met the criteria for factor analysis. As can be seen from Table 6, the factor analysis revealed 3 factors that explained 57% of the variance. The factors converged in 19 iterations. Factor 1 included all the emotional items and some symptoms not related to specific tasks (shoulder clicking, neck discomfort, and affect on fitness). Three additional items loaded between .4 and .5. They were all questions about pain. Two of the sports items (ability to throw hard/far, and difficulty with push-ups) loaded on factor 2, with the weakness item loading between .4 and .5. The third factor included several items that asked about difficulty performing specific activities. The Cronbach's alpha values for the three factors were: .87 (9 items), .67 (3 items) and .89 (8 items) respectively.
Table 6 Pattern matrix following oblique rotation [listed by items loading on a given factor]
Factors
1 2 3
Item Emotions & symptoms Disability – strength activities Disability – daily activities
PS5 Shoulder clicking, grinding, crunching .64 .15 <.01
PS6 Neck discomfort .52 <.01 -.16
SR7 Affect fitness level .58 .45 <.01
E19 How much frustration .73 <.01 <.01
E20 How depressed .77 -.13 -.11
E21 How worried about effect on occupation .81 <.01 <.01
PS1 Sharp pain .47 <.01 -.29
PS2 Constant, nagging pain .49 <.01 -.34
LS15 Difficulty sleeping .45 -.21 -.45
SR8 Difficulty with push-ups .14 .85 .11
SR9 Affect ability to throw -.14 .71 -.17
PS3 How much weakness .23 .41 -.28
PS4 How much stiffness <.01 .25 -.56
SR10 Difficulty with contact with shoulder .27 .12 -.53
W11 Difficulty with daily house/yard activities .30 .15 -.56
W12 Difficulty working above shoulder <.01 .33 -.55
W14 Difficulty lifting heavy objects <.01 <.01 -.63
LS16 Difficulty styling hair <.01 -.14 -.88
LS18 Difficulty dressing/undressing <.01 <.01 -.82
LS17 Difficulty roughhousing .20 .31 -.46
W13 How much use of uninvolved arm <.01 .36 -.38
Factor loadings > 0.5 are in bold.
Factor loadings between 0.4 and 0.5 are in italics if that item did not load higher on another factor
E emotions
LS lifestyle
PS physical symptoms
SR sports and recreation
W work
To see the factor loadings with items listed by the domains of the original scale, see additional file 1: Pattern Matrix by domains.doc.
Discussion
The main purpose of this study was to determine whether the WORC items fell into 5 domains as proposed by the creators of the scale. Although some of the items grouped together as hypothesized, the factor analysis did not support the 5-domain construct of the WORC.
The factor analysis revealed 3 factors, not 5. The 3 factors appear to be: 1) symptoms and emotions, 2) strenuous shoulder tasks, and 3) difficulty with daily tasks. Based on the groupings, it appears that symptoms of pain are associated with emotions, and lack of range of motion or stiffness with difficulty with daily activities. The symptom of "weakness' was associated with two very specific shoulder tasks – throwing hard and push-ups. Based on the mean values for these two items (S8, S9), they were likely the most difficult tasks as well. Thus it is not surprising that "weakness" was associated with difficulty with these activities.
Although factor analysis has not been previously performed on the WORC, other authors have reported a mix of symptoms, disability and social/emotional items within factors derived from other shoulder measures. Veehof et al [8] noted that all 30 items of the DASH loaded positively on the first factor following principal component factor analysis. Only 3 loaded less than .50. The DASH has questions on physical function, symptoms and social/role function. Similarly, Roddey et al [9] reported that both the pain and disability items of the SPADI loaded on one factor (.613 to .905). On the other hand, two factors were derived from the Simple Shoulder Test (SST) [9], which was designed to measure one construct, functional ability in activities of daily living. All of these results suggest that patients with shoulder problems may not differentiate disability and symptoms, and that such theoretical groupings of items are not appropriate.
This lack of separation of pain and disability has been seen in measures of the lower limb as well. Kennedy et al [10] found that the items of the Western Ontario and McMaster Osteoarthritis Index (WOMAC) factored out on type of activity rather than pain or difficulty. The authors [10] felt their results might be due to the similarity of the questions in the two domains. For example, 'pain with sitting or lying' is in the pain subscale, and 'difficulty with lying in bed' and 'difficulty with sitting' are both in the physical function subscale. There is no such duplication in the WORC items, and yet, in the present study, there was at least one symptom question, and one "difficulty" question in each factor. Thus, it may be that individuals do not inherently separate symptoms and functional ability in musculoskeletal conditions, no matter how the questions are worded.
In their systematic review of shoulder measures, Bot et al [11] considered a measure to have good internal consistency when its structure was explored by factor analysis, and Cronbach's alpha for each separate factor was .70 to .90. Two of the three subscales derived from the factor analysis met this criterion. The middle factor/subscale, with only 3 items, did not meet the .70 criterion. However, the Cronbach's alpha increased to .70 when the weakness item, which loaded only .41, was removed. The other two factor/subscales had alpha coefficients higher than the original subscales.
The WORC was originally developed and tested on a heterogeneous group of patients that likely had a wider range of disease severity than the pre-surgical patients used in the present study. Because Kirkley et al [1] did not present any descriptive statistics for the total or subscale scores of the WORC, the actual range of disability of the subjects in the two studies can not be directly compared. Even so, it is possible, that the results might have been closer to the 5-domain model proposed by Kirkley et al [1] if the subjects were similar in the two studies. However, one would expect a robust measure to have similar properties when used on all types of patients for which it was intended. Additional research should be conducted to confirm the subscales found in the present study, to examine their properties and determine the value of their use in the clinical or research setting.
Conclusion
The results of this study indicate that the WORC has 3 factors, which explain 57% of the variance. All factors include both 'function' and 'symptom' questions. The three items from the original emotional scale were the only ones that grouped together, but that factor also included items from 3 other subscales. The results of this factorial analysis do not support the 5-domain structure proposed by the creators of the WORC. Based on the results of the present study and on previous work conducted on the WORC, there does not appear to be a significant advantage to using this condition-specific questionnaire over some other well-established measures for the shoulder.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
JW proposed the study, developed the research protocol, and wrote the first draft of the paper. HR was responsible for subject selection, data collection and management of the database. YM was involved in the review of the literature, data input and initial analysis of data. RH was involved in patient recruitment and providing clinical and surgical diagnosis. All authors were involved in the preparation of the manuscript, and read and approved the final version.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
Word file showing the factor loadings grouped by the theoretical domains
Click here for file
Acknowledgements
The authors wish to thank Terry Leeke, data management consultant at the Research Facilitation Office of the Centre for Research In Women's Health for development of the shoulder database, from which data for this study were extracted.
==== Refs
Kirkley A Alvarez C Griffin S The development and evaluation of a disease-specific quality-of-life questionnaire for disorders of the rotator cuff: The Western Ontario Rotator Cuff Index Clin J Sport Med 2003 13 84 92 12629425 10.1097/00042752-200303000-00004
Getahun TY MacDermid JC Patterson SD Concurrent validity of patient rating scales in assessment of outcome after rotator cuff repair J Musculoskelet Res 2000 4 119 127 10.1142/S021895770000015X
Holtby R Razmjou H Measurement properties of a rotator cuff outcome measure in patients undergoing shoulder surgery: A preliminary report J Shoulder Elbow Surg
MacDermid JC Drosdowech D Faber K Responsiveness of self-report scales in patients recovering from rotator cuff surgery J Shoulder Elbow Surg
Kirkley A Griffin S McLintock H Ng L The development and evaluation of a disease-specific quality of life measurement tool for shoulder instability. The Western Ontario Shoulder Instability Index (WOSI) Am J Sports Med 1998 26 764 772 9850776
Lo IK Griffin S Kirkley A The development of a disease-specific quality of life measurement tool for osteoarthritis of the shoulder: The Western Ontario Osteoarthritis of the Shoulder (WOOS) Index Osteoarthritis Cartilage 2001 9 771 778 11795997 10.1053/joca.2001.0474
Norusis MJ Advanced statistics SPSS/PC+ 1986 Chicago: SPSS Inc
Veehof MM Sleegers EJ van Veldhoven NH Schuurman AH Meeteren NL Psychometric qualities of the Dutch language version of the Disabilities of the Arm, Shoulder, and Hand Questionnaire (DASH-DLV) J Hand Ther 2004 15 347 354 12449349
Roddey TS Olson SL Cook KF Gartsman GM Hanten W Comparison of the University of California-Los Angeles Shoulder Scale and the Simple Shoulder Test with the shoulder pain and disability index: single-administration reliability and validity Phys Ther 2000 80 759 768 10911414
Kennedy D Stratford PW Pagura SM Wessel J Gollish JD Woodhouse LJ Exploring the factorial validity and clinical interpretability of the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) Physiother Can 2003 55 160 168
Bot SD Terwee CB van der Windt DA Bouter LM Dekker J de Vet HC Clinimetric evaluation of shoulder disability questionnaires: a systematic review of the literature Ann Rheum Dis 2004 63 335 341 15020324 10.1136/ard.2003.007724
| 15871742 | PMC1112598 | CC BY | 2021-01-04 16:32:03 | no | BMC Musculoskelet Disord. 2005 May 4; 6:22 | utf-8 | BMC Musculoskelet Disord | 2,005 | 10.1186/1471-2474-6-22 | oa_comm |
==== Front
BMC NeurosciBMC Neuroscience1471-2202BioMed Central London 1471-2202-6-281583313710.1186/1471-2202-6-28Research ArticleTranscriptome analysis in primary neural stem cells using a tag cDNA amplification method Sievertzon Maria [email protected] Valtteri [email protected] Alex [email protected] Konstantinos [email protected] Rikard [email protected]öm Lilian [email protected]én Jonas [email protected] Joakim [email protected] Royal Institute of Technology, AlbaNova University Center, KTH Genome Center, Department of Biotechnology, S-106 91 Stockholm, Sweden2 NeuroNova AB, S-114 33 Stockholm, Sweden3 Department of cell- and molecular biology, Medical Nobel Institute, Karolinska Institute, S-171 77 Stockholm, Sweden2005 15 4 2005 6 28 28 12 1 2005 15 4 2005 Copyright © 2005 Sievertzon et al; licensee BioMed Central Ltd.2005Sievertzon 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
Neural stem cells (NSCs) can be isolated from the adult mammalian brain and expanded in culture, in the form of cellular aggregates called neurospheres. Neurospheres provide an in vitro model for studying NSC behaviour and give information on the factors and mechanisms that govern their proliferation and differentiation. They are also a promising source for cell replacement therapies of the central nervous system. Neurospheres are complex structures consisting of several cell types of varying degrees of differentiation. One way of characterising neurospheres is to analyse their gene expression profiles. The value of such studies is however uncertain since they are heterogeneous structures and different populations of neurospheres may vary significantly in their gene expression.
Results
To address this issue, we have used cDNA microarrays and a recently reported tag cDNA amplification method to analyse the gene expression profiles of neurospheres originating from separate isolations of the lateral ventricle wall of adult mice and passaged to varying degrees. Separate isolations as well as consecutive passages yield a high variability in gene expression while parallel cultures yield the lowest variability.
Conclusions
We demonstrate a low technical amplification variability using the employed amplification strategy and conclude that neurospheres from the same isolation and passage are sufficiently similar to be used for comparative gene expression analysis.
==== Body
Background
The most frequently used method to analyse scarce RNA samples is to employ RNA amplification technology [1,2], enabling analysis of the full length transcripts. We have recently reported on an alternative transcriptome amplification method that minimises differences in transcript length in the amplification step [3,4]. This method is based on fragmentation of the mRNA (cDNA) population followed by isolation of a unique, short and representative 3'end tag of each transcript prior to amplification by PCR. Here we have evaluated and applied the methodology on neural stem cells (NSCs).
NSCs can be isolated from the fetal or adult mammalian brain and grown in vitro in the presence of growth factors to form floating aggregates of cells denoted neurospheres [5-7]. A neurosphere is derived from one clonally expanded NSC or progenitor cell [8]. As the original NSC or progenitor cell proliferates the new cells adhere to each other, eventually forming a neurosphere. Every neural stem cell in a neurosphere has the potential to differentiate towards a neuronal or a glial lineage depending on the internal neurosphere milieu and external signals. Neurospheres are thus complex structures consisting of many cell types that can have varying degrees of differentiation commitment, but that are all derived from the same clonally expanded cell. Neurospheres have extensive cell-cell contacts and a dense extracellular matrix. When plated onto solid support in combination with growth factor withdrawal the cells start to differentiate into all neural cell types (neurons, astrocytes and oligodendrocytes)[9]. In vitro expanded neural stem cells may therefore serve as an in vitro model of neurogenesis. The similarities between the in vivo and in vitro processes of neurogenesis are not well established although some characteristics are expected to be conserved [10] and therefore challenging a cell in vitro will unveil some of its developmental properties and potentials. By subjecting neurospheres to different microenvironments (e.g. through the addition or withdrawal of drugs or factors) it is possible to uncover factors and mechanisms important for proliferation or differentiation into certain cell lineages, for example neurons of a particular type [11,12]. Furthermore, NSCs expanded as neurospheres also hold the promise of becoming an important source of cells for cell replacement therapies of different neurological diseases [13,14].
Due to the great scientific interest in NSCs and the promise of their clinical use we decided to investigate NSCs from a gene expression perspective. An important aspect was to investigate if neurosphere heterogeneity [8] is reflected in their transcriptome. Neurosphere populations from different levels of technical and biological replication were analysed by taking advantage of microarrays with 5159 spotted mouse cDNA clones, in combination with a highly sensitive amplification method. We compared neurospheres cultured under identical conditions but in separate culture flasks, as well as from different passages and from parallel isolations. The results are discussed from the perspective of differences in the number and extent of differentially expressed genes.
Results
Different sources of neurospheres were used to investigate the extent of heterogeneity between neurosphere populations at the gene expression level. To facilitate a broad transcript analysis of this relatively scarce material a recently developed amplification methodology [3,4] was used (Figure 1A) in combination with microarray technology. In brief, the approach involves biotinylation of the 3'-end of the cDNA using a biotinylated oligo(dT) primer in the first-strand cDNA synthesis reaction. The cDNA is randomly fragmented by sonication into 50–500 bp fragments. The 3'-ends (denoted 3'-end signature tags), representing the most unique part of most transcripts, are isolated by binding to streptavidin-coated beads. Linkers are ligated onto the 3'-end signature tags, which are subsequently cleaved off the beads and finally amplified using PCR. This generates a smear of random-sized fragments (data not shown) which is labelled by asymmetric PCR and then hybridised to microarrays. Here we used a mouse microarray comprising of 5159 mouse cDNA clones, printed in duplicate.
Figure 1 Experimental approach; A) Schematic overview of the utilised amplification protocol. For details see text. B) Experimental design. NSCs were isolated from the lateral ventricular region of brains from three pools of mice (three isolations) and grown as neurospheres. Sample G was induced to differentiate by withdrawing the growth factors from the culture medium, plating on solid support and adding serum. RNA was isolated from different passages as indicated and used for subsequent microarray hybridisations. Blue arrows represent duplicate hybridisations, arrowhead represents labelling with Cy5 and arrow tail represents labelling with Cy3.
Differential expression was determined in a series of microarray experiments, as outlined in Figure 1B. Neurosphere cultures were initiated from cells dissociated from three pools of adult lateral ventricle wall tissue dissected from either 3 or 10 mice, as three identical but separate isolations. Primary neurospheres were passaged one or two times and harvested three to four days after passage. Average neurosphere size was deemed a more critical factor than the length of the incubation time, hence some neurosphere cultures were incubated for one day longer than others to obtain uniformity in neurosphere size between cultures at the key times of passaging and harvesting. When passaged twice the neurospheres were split into two or three equivalent cultures. This allowed us to measure the variability in gene expression levels between different isolations, as well as between passages and between parallel cultures. In order to estimate the technical noise, self-to-self hybridisations were performed using RNA from one of the cultures. To confirm that we were able to detect differential gene expression cells in one of the parallel cultures were induced to differentiate into neurons, astrocytes and oligodendrocytes by withdrawing the growth factors from the culture medium, plating on solid support and adding serum (in this work referred to as differentiated cells). The nomenclature of the samples is given in Figure 1B. Seven different comparisons were made; A1-A2 (technical replicate), CI-CII and CII-CIII (culture replicates), B-CI (different passages), A2-CIII and CII-FI (different isolations) and F-G (neurospheres vs. differentiated cells). The use of short-term passaged neurospheres limits the number of cells that can be generated. Consequently the amount of RNA that can be isolated is below that normally used in labelling reactions for microarray hybridisations (approximately 10 μg total RNA without amplification). After mRNA isolation and cDNA synthesis we therefore chose to amplify the obtained material using the method described above. Two replicate and two dye-swap hybridisations were performed for each comparison, adding up to four hybridisations for each comparison in total.
The microarray data was filtered (for details see Methods) and print-tip lowess normalised. Differentially expressed genes were identified using an empirical Bayes moderated t-test and by calculation of the associated p-values [15,16]. In the t-test the contribution of within-array replicate features was taken into account [17] and the genes were ranked according to the probability of differential expression (B-value; depending on both the fold change and the variability over the four hybridisations). Higher B-value indicates higher probability of differential expression. Genes were defined as differentially expressed if the fdr-adjusted p-value was < 0.001 (corresponding to an approximative B-value > 0.3). In Figure 2 the B-value distribution for each comparison is shown. The figure shows no differentially expressed (DE) genes in the technical replicate (using the amplification strategy), a higher number of DE genes in neurospheres cultured in parallel, an even higher number of DE genes in neurospheres from different isolations and passages, and the highest number of DE genes in neurospheres vs. differentiated cells. Note the high number of DE genes in neurospheres from the same isolation but subsequent passages (B-CI), indicating that the neurospheres may change character over time as they grow and proliferate in vitro.
Figure 2 B-value distribution for each of the comparisons; The B-value is calculated through empirical Bayes statistics and scores the genes according to their probability of differential expression. Higher B-value means higher probability of differential expression. NS = neurosphere, DC = differentiated cells.
To further investigate the variability in gene expression between the different neurosphere samples we visualised the data using a series of plots displayed in Figure 3. In panel A) the average A-value (1/2log2(sample X intensity * sample Y intensity)) for each gene is plotted against the corresponding M-value (log2(sample X intensity / sample Y intensity)). The plots show that after filtration and normalisation there is no intensity bias in the distribution of DE genes. They also show that the M-values for the technical replicates (A1-A2) are collected and centred close to zero (corresponding to a ratio of 1), whereas the spreading of the M-values are higher for culture replicates (CI-CII and CII-CIII), passage replicates (B-CI) and isolation replicates (A2-CIII and CII-F). The highest spreading of M-values can be seen for the neurosphere vs. differentiated cells hybridisations (F-G), where many genes have M-values between +/-1 and +/-3 (corresponding to a fold change of 2 to 8). In panel B) the B-value for each gene is plotted against the corresponding M-value. By definition genes with a high M-value will obtain a higher B-value, which gives the plots the characteristic volcano shape. Also here it is clear that the technical replicates have no statistically significant DE genes (with high B-values), whereas culture, passage and isolation replicates have several genes with high B-values, and the neurosphere vs. differentiated cells comparison clearly has the highest number of DE genes. This is further visualised in Figure 3, panel C), where the average of the signal intensity for each gene and sample is plotted against the average signal intensity for that gene in the other sample. Once again the Pearson correlation for the two samples is highest for the technical replicates (r = 0.99), lower for the culture replicates (r = 0.98 and r = 0.98 respectively), passage replicate (r = 0.94) and isolation replicates (r = 0.95 and r = 0.96 respectively) and lowest for the neurosphere vs. differentiated cells (r = 0.85).
Figure 3 Graphs displaying the variability of the data at different levels of replication; In all graphs one dot represents one gene. Panel A) shows MA-plots for each comparison. The x-axis represents the intensity of the feature (A = 1/2log2(Cy5*Cy3)). The y-axis represents the magnitude of differential expression of the gene (M = log2(Cy5/Cy3)), calculated after filtration and normalisation of the data. Dotted lines are drawn at M-values 1 and -1, i.e. at a 2-fold difference in signal intensity between the compared samples. Panel B) shows volcano plots for each sample. The x-axis shows the M-value for each gene and the y-axis the corresponding B-value (calculated by empirical Bayes moderated t-test) for that gene. Panel C) shows scatter plots for each comparison. The x-axis displays the average signal intensity for one sample and the y-axis the average signal intensity for the other sample. Also shown are the values of the Pearson correlation coefficient (r) and the coefficient of determination (R2).
The number of differentially expressed genes in each comparison is summarised in Figure 4. Genes with a false discovery rate (fdr) adjusted p-value < 0.001, giving less than one false positive per 1000 genes, are included. This further demonstrates that the lowest number of DE genes is in the technical replicates and the highest number in the comparison of the neurospheres vs. differentiated cells (748 genes). Again, the most noteworthy result is that neurospheres of different passages (B-CI) have a surprisingly high number of DE genes (383 genes) compared to the other comparisons. To further explore the magnitude of differential expression of these genes (with p < 0.001) a table of their distribution over fold change was made (Table 1). As expected, a high number of genes have high fold changes for the passage and isolation replicates and the neurosphere vs. differentiated cells comparison, whereas fewer genes are within the higher fold change ranges for the culture replicates.
Table 1 Distribution of differentially expressed genes over fold change.
M interval Fold change A1-A2 CI-CII CII-CIII B-CI A2-CIII CII-F F-G
+/-(0–0.5) 0–1.4 17 16
+/-(0.5–1) 1.4–2 14 51 260 7 147 455
+/-(1–1.5) 2–2.8 12 26 108 20 14 190
+/-(1.5–2) 2.8–4 1 4 14 5 2 52
+/-(2–2.5) 4–5.6 1 1 19
+/-(>2.5) >5.6 1 16
Total p < 0.001 (fdr) 0 27 82 383 32 181 748
Genes with p < 0.001, calculated by empirical Bayes moderated t-test and false discovery rate adjustment, are included. (M = log2(Cy5/Cy3)).
Figure 4 The number of differentially expressed genes in each comparison; Genes with p < 0.001, calculated by empirical Bayes moderated t-test and false discovery rate adjustment, are included. NS = neurosphere, DC = differentiated cells.
To investigate whether the transcript level differences in the two culture-to-culture comparisons are consistent or random events a Venn diagram was created in Figure 5, displaying the number of shared and unique DE genes in the CI-CII, CII-CIII and F-G comparisons. Five genes out of 27 (CI-CII) and 82 (CII-CIII) overlap, equivalent to 19% and 6% respectively. The vast majority of DE genes are thus not shared between the two comparisons, indicating non-systematic changes in gene expression. When compared to the neurosphere vs. differentiated cells gene list only one of the five genes is in common, further demonstrating random differences between cultures.
Figure 5 Overlap of differentially expressed genes; Compared are the results from the parallel cultures (CI-CII and CII-CIII) and the neurosphere vs. differentiated cells comparison (F-G). Genes with p < 0.001, calculated by empirical Bayes moderated t-test and false discovery rate adjustment, are included.
A list of the DE genes found in the neurosphere vs. differentiated cells comparison (F-G) is provided as an additional data file 1: Differentially expressed genes in neurosphere vs. differentiated cells comparison (the complete results for all comparisons are available in ArrayExpress using experiment accession number E-MEXP-297). The genes that show four-fold or greater fold change (M ≥ |2|) and adjusted p-values < 0.001 are shown in Table 2. This demonstrates that genes found in the F-G comparison are genes that are expected to be involved in the differentiation of neurospheres. For example, included are several myelin related genes such as myelin-associated oligodendrocytic basic protein (Mobp), myelin basic protein (Mbp) and myelin-associated glycoprotein (Mag), all of which are up-regulated in the differentiated sample. Also, there are some genes related to transmitter substances and their signaling; gamma-aminobutyric acid (GABA-A) receptor, subunit beta 1 (Gabrb1)and guanine nucleotide binding protein, alpha o (Gnao1), which is involved in dopamine signaling [18]. Distal-less homeobox 1 (Dlx1), is also widely expressed in the brain and is involved in brain development and neural differentiation [19-21].
Table 2 Genes differentially expressed in the neurosphere vs. differentiated cells comparison (F-G).
Genbank Acc.no. Unigene ClusterID GeneID Gene Name M B
CX240827 Mm.200608 12759 Clusterin -3,503 22,292
CX241145 Mm.289645 17263 GTL2, imprinted maternally expressed untranslated mRNA -3,774 21,684
CX238761 Mm.354720 14681 Guanine nucleotide binding protein, alpha o -3,112 21,128
CX241318 Mm.289645 17263 GTL2, imprinted maternally expressed untranslated mRNA -4,429 18,116
CX235874 Mm.40461 17433 Myelin-associated oligodendrocytic basic protein -2,961 18,854
CX236810 Mm.210815 20411 Sorbin and SH3 domain containing 1 -2,287 17,999
CX241478 Mm.25874 77976 RIKEN cDNA B230104P22 gene -3,435 17,660
CX241468 Mm.4543 13390 Distal-less homeobox 1 -2,843 16,192
CX240901 Mm.271178 67166 ADP-ribosylation factor-like 10C -2,606 15,924
CX238347 Mm.252063 17196 Myelin basic protein -2,622 15,450
CX240173 Mm.291442 20692 Secreted acidic cysteine rich glycoprotein -2,008 14,973
CX240709 Mm.30035 107747 Formyltetrahydrofolate dehydrogenase 2,065 14,714
CX238313 Mm.206505 21858 Tissue inhibitor of metalloproteinase 2 -2,669 14,533
CX236462 Mm.34126 74617 Serine carboxypeptidase 1 -2,544 13,912
CX242347 Mm.271770 433496 Similar to Myl9 protein -2,860 13,549
CX243886 Mm.289645 17263 GTL2, imprinted maternally expressed untranslated mRNA -2,213 13,547
CX235484 Mm.178246 216848 Chromodomain helicase DNA binding protein 3 -2,313 12,737
CX241157 Mm.194225 70397 RIKEN cDNA 1110020A09 gene -2,045 12,600
CX243757 Mm.260601 242521 Kelch-like 9 (Drosophila) -2,848 12,607
CX237579 Mm.40461 17433 Myelin-associated oligodendrocytic basic protein -2,000 11,957
CX242456 Mm.291442 20692 Secreted acidic cysteine rich glycoprotein -2,835 11,842
CX238326 Mm.1426 51791 Regulator of G-protein signaling 14 -2,193 11,543
CX239776 Mm.252063 17196 Myelin basic protein -2,606 11,414
CX235557 Mm.241355 17136 Myelin-associated glycoprotein -2,220 11,257
CX239390 Mm.213204 80906 Kv channel-interacting protein 2 -2,126 10,915
CX239520 Mm.21549 80888 Heat shock 27 kDa protein 8 -2,104 10,657
CX238589 Mm.226704 14400 gamma-aminobutyric acid (GABA-A) receptor, subunit beta 1 -2,039 10,744
CX242889 Mm.29358 67971 RIKEN cDNA 2700055K07 gene -2,032 10,827
CX235601 Mm.4857 12323 Calcium/calmodulin-dependent protein kinase II, beta 2,418 10,025
CX242440 Mm.121920 20660 Sortilin-related receptor, LDLR class A repeats-containing -2,180 8,923
CX240196 Mm.181959 13653 Early growth response 1 2,469 8,456
CX242602 #N/A -2,692 8,087
CX240394 Mm.18830 77569 RIKEN cDNA 3732412D22 gene -2,078 7,216
CX237453 Mm.181959 13653 Early growth response 1 2,059 6,141
CX244387 Mm.211275 76441 Dishevelled associated activator of morphogenesis 2 -2,347 4,185
CX236987 Mm.329668 20743 Spectrin beta 3 -2,014 2,723
Genes with p < 0.001, calculated by empirical Bayes moderated t-test and false discovery rate adjustment, and fold change ≥ 4 (-2 ≥ M ≥ 2), are included. Genes with M-value < 0 are up-regulated in differentiated cells, genes with M-value > 0 are up-regulated in neurospheres. M = log2(Cy5/Cy3), B = B-value calculated by empirical Bayes moderated t-test. Higher B-values mean higher probability for differential expression. For a full list of differentially expressed genes in the F-G comparison see additional data file 1: Differentially expressed genes in neurosphere vs. differentiated cells comparison.
To understand the biological significance of the overall changes in gene expression, DE genes in the neurosphere vs. differentiated cells comparison (F-G) were categorised according to their gene ontology theme annotation. Genes with an adjusted p < 0.001 were included in the analysis and the probability of a theme being over-represented in the data was calculated using Jackknife Fisher's exact probability test implemented in the EASE software [22]. Table 3 shows the most highly represented gene ontology themes (of category biological function). Themes that showed enrichment in differentiated cells include neurogenesis, synaptic transmission, cell-cell signalling and development. In contrast, themes that showed enrichment in neurospheres include electron transport and mitotic cell cycle.
Table 3 The most highly represented gene ontology themes in the neurosphere vs. differentiated cells comparison (F-G).
Biological process no of genes on array no of DE genes in F-G enriched in NS enriched in DC Fisher Exact
cell adhesion 56 17 6 (35%) 11 (65%) 0.00939
lipid metabolism 71 20 8 (40%) 12 (60%) 0.0123
neurogenesis 35 12 2 (17%) 10 (83%) 0.0101
mitotic cell cycle 30 10 7 (70%) 3 (30%) 0.0225
synaptic transmission 35 11 2 (18%) 9 (82%) 0.0266
development 166 37 8 (22%) 29 (78%) 0.0429
neuromuscular physiol. process 37 11 2 (18%) 9 (82%) 0.0396
transmission of nerve impulse 37 11 2 (18%) 9 (82%) 0.0396
electron transport 38 11 8 (73%) 3 (27%) 0.0476
endocytosis 24 8 2 (25%) 6 (75%) 0.0399
organismal movement 43 12 3 (25%) 9 (75%) 0.0511
siderochrome transport 34 10 4 (40%) 6 (60%) 0.0522
alcohol metabolism 39 11 4 (36%) 7 (64%) 0.0565
neurophysiological process 49 13 3 (23%) 10 (77%) 0.0626
organismal physiological process 86 20 5 (25%) 15 (75%) 0.0842
energy pathways 36 10 7 (70%) 3 (30%) 0.0739
cell-cell signaling 41 11 2 (18%) 9 (82%) 0.0777
Organogenesis 89 20 4 (20%) 16 (80%) 0.112
Analysis is restricted to themes of category biological function. Genes with p < 0.001, calculated by empirical Bayes moderated t-test and false discovery rate adjustment were included in the analysis. Shown is the total number of genes found in the respective theme, as well as the number of those genes that are enriched in neurospheres (NS) and differentiated cells (DC) respectively. Jackknife Fisher's exact probability test was used to identify over-represented themes in the data. Corresponding p-values are listed.
Discussion
This study has taken advantage of a recent template amplification method to study neurospheres at the level of transcription. RNA from different isolations, cultures and passages was isolated, amplified and analysed by microarrays. The comparison was performed by analysis of the number of differentially expressed genes for the different conditions. The results show excellent performance of the amplification protocol. No differentially expressed genes were found in the technical replicates indicating that methodological noise in all comparisons should be considered minor.
Fluctuations of transcript levels in different populations of neurospheres
The array results for the different neurosphere conditions were much more divergent than the technical replications and we observe a varying degree of heterogeneity among the different neurosphere populations, obtained from different isolations of adult mouse lateral ventricle wall tissue, from different passages and from parallel cultures. The results show that there is a large variation in gene expression between neurospheres from different isolations as well as between neurospheres from the same isolation but from different passages. Neurospheres have previously been shown to gain altered properties through extensive, long-term passaging (more than 10 passages) [23]. Short-term passaged neurospheres have been considered rather stable, with unaltered multipotency and capacity for self-renewal [24]. Here we have shown that already between passages one and two neurospheres show altered gene expression with up to 383 DE genes (p < 0.001). Whether this is due to different properties of the parental, clonally expanded cell(s) giving rise to the neurospheres in each passage or some other reason needs to be further investigated.
Parallel culturing of neurospheres from the same isolation and the same number of passages, grown in identical conditions, show fewer DE genes (up to 82 genes, p < 0.001) than neurospheres compared between passages (383 genes, p < 0.001). Furthermore, when neurospheres are induced to differentiate and compared to undifferentiated neurospheres cultured in parallel, the number of genes DE as well as the magnitude of the M-values are clearly higher (748 genes, p < 0.001). These data indicate that an extended 3–4 day culturing, per se, is sufficient to induce changes in gene expression, but with careful experimental design and an appropriate number of biological replicates neurospheres cultured in parallel, from the same isolation and passage, may be used to study for example the effect of exposure to different microenvironments on gene expression.
The gene expression heterogeneity of neurospheres may be related to a number of different factors such as the age of the animal from which they were isolated, neurosphere size and the identity of the first clonally expanded cell [25]. Our results are also confirmed by observations by Suslov and co-workers that examined the expression of 16 transcripts from single neurospheres of different sizes [8]. The obtained information was used to cluster the individual neurospheres according to similar gene expression pattern. It revealed an inter-clonal heterogeneity that might reflect the maturity and developmental commitment of the parental clonogenic cell, as well as the size of the neurosphere and its time in culture. In another study it was shown that populations of neurospheres from different regions of the brain as well as from different species differ in properties such as growth rate, neuronal production and cell morphology [26].
Genes expressed in neurospheres
The different neurosphere populations show heterogeneity in their expression profiles, yet many of the genes expressed are representative of a neurosphere transcript signature. As described earlier, neurospheres consist of several cell types of varying degrees of differentiation, a dense extracellular matrix and extensive cell-cell contacts. Electron-microscopy studies of rat fetal striatum EGF-expanded neurospheres [27], have shown that they consist of two types of cells, electron-dense and electron-lucent cells, both of which could be either healthy, apoptotic or necrotic. These neurosphere cells also demonstrated an expression of the cell adhesion molecules E- and N-cadherin (Cdh1 and Cdh2), α- and β-catenin (Catna1, LOC297357 and RGD:70487) and growth factor receptors for epidermal growth factor (Egfr) and fibroblast growth factor (Fgfr1), as well as fibroblast growth factor 2 (Fgf2). Also neurospheres from adult human brain have been characterised, revealing the same type of heterogeneous, complex structure [28,7]. These express a variety of different markers, such as nestin (NES; neural stem/progenitor and immature glial marker), vimentin (VIM; immature glia), glial fibrillary acidic protein (GFAP; astrocytes), β-III-tubulin (TUBB3; neuronal marker) and cell adhesion molecule L1 (L1CAM; neuronal marker), proteolipid protein 1 (PLP1; oligodendrocytes), B-cell CLL/lymphoma 2 (BCL2; anti-apoptotic), paired box gene 6 (PAX6; a developmentally regulated gene) and tenascin C (TNC; extracellular matrix protein). In our study the genes related to these phenotypes and markers are expressed at similar levels in all neurosphere replicates (CI-CII, CII-CIII, B-CI, A2-CIII and CII-F). For example we observe many genes involved in apoptosis; Bcl2-associated X protein (Bax), Bcl2-associated athanogene 1 (Bag1), cytochrome c-1 (Cyc1), death associated protein 3 (Dap3), programmed cell death 6 interacting protein (Pdcd6ip) and phosphoprotein enriched in astrocytes 15 (Pea15). Expressed are also α-E-catenin (Catna1), β-catenin (Catnb) and fibroblast growth factor 3 (Fgfr3), and other neurosphere markers such as nestin (Nes), glial fibrillary acidic protein (Gfap), β-III-tubulin (Tubb3) and proteolipid protein 1 (Plp1) (The complete data set is available in ArrayExpress using experiment accession number E-MEXP-297).
The list of DE genes in the neurosphere vs. differentiated cells comparison (Table 2 and additional data file 1: Differentially expressed genes in neurosphere vs. differentiated cells comparison) as well as an overview of the corresponding gene ontology classification (Table 3) also demonstrates the anticipated differences between neurospheres and differentiated neurospheres.
Summary
The genes observed to be differentially expressed in identical but parallel cultures appear to be random, shown by the low overlap in DE genes between the two parallel culture comparisons (Figure 5). The number of erroneously identified DE genes, due to biological fluctuations, could hence be lowered by increasing the number of biological replicates. Hereby random differences will be removed and true DE genes can be selected by statistical means. It should be noted that the random differences mainly correspond to small fold changes as compared to the larger changes in the neurospheres vs. differentiated cells. Reliable differences in gene expression could therefore be obtained and studied without increasing the number of replicates if a higher cut-off for DE genes, such as fold change > 2, was chosen.
Conclusions
We have shown that the tag cDNA amplification method is well suited for the analysis of neurospheres, demonstrating low technical variability. Furthermore we have demonstrated large differences between passages of neurospheres, but less variability between parallel cultures. The described variability appears to be random and the underlying cause(s) needs further investigations. The neurosphere variability can be addressed by increasing the number of biological replicates and careful experimental design, which will facilitate future use of neurospheres as a tool to study gene expression changes involved in neurogenesis.
Methods
Adult mouse neural stem cell culture
Three adult mouse neural stem cell cultures were initiated, the first originating from tissue isolated from ten mice (Culture 1) while the second (Culture 2) and third (Culture 3) cultures originated from three mice each. For each culture, identical dissection, dissociation and culture protocols were used. Briefly, the lateral wall of the lateral ventricle of 5–6-week-old mice was enzymatically dissociated in 0.8 mg/ml hyaluronidase and 0.5 mg/ml trypsin in Dulbecco's modified Eagle medium (DMEM) containing 4.5 mg/ml glucose and 80 U/ml DNase at 37°C for 20 min. The cells were gently triturated and mixed with three volumes of neurosphere medium (DMEM/F12, B27 supplement, 12.5 mM HEPES pH7.4) containing 20 ng/ml EGF, 100 U/ml penicillin and 100 μg/ml streptomycin. After passing through a 70-μm strainer, the cells were pelleted at 160 × g for 5 min. The supernatant was subsequently removed and the cells resuspended in neurosphere medium supplemented as above, plated in uncoated culture dishes and incubated at 37°C. Neurospheres were ready to be split 7–8 days after plating.
To split neurosphere cultures, neurospheres were collected by centrifugation at 160 × g for 5 min. The neurospheres were resuspended in 0.5 ml Trypsin/EDTA in HBSS (1x), incubated at 37°C for 2 min and triturated gently to aid dissociation. Following a further three-min incubation at 37°C and trituration, 3 volumes of ice-cold neurosphere medium containing EGF were added. The cells were pelleted at 220 × g for 4 min, resuspended in fresh neurosphere medium supplemented with 20 ng/ml EGF.
From Cultures 1, 2 & 3, dissociated cells were plated and grown in neurosphere medium supplemented with EGF for a further 3–4 days by which time secondary neurospheres had developed. The secondary neuropheres originating from Culture 1 were harvested for mRNA isolation (Sample A). Approximately a quarter of the secondary neurospheres originating from Culture 2 were also taken for mRNA isolation (Sample B), while the remainder were dissociated and replated in three equal fractions (100,000 cells / well (6 well plate)), cultured in neurosphere medium supplemented with EGF for 3 days, and harvested for mRNA isolation (Samples CI, CII, CIII). Secondary neurospheres originating from Culture 3 were dissociated and divided into two fractions. The first fraction was replated (100,000 cells / well (6 well plate)) and cultured identically to the cells generating Samples CI, CII & CIII. After 3 days, the cells were harvested for mRNA isolation (Sample F). The second fraction was replated in neurosphere medium supplemented with 1% fetal calf serum (FCS) onto poly-D-lysine plates to which the cells adhered. After incubating, overnight FCS concentration was reduced to 0.5%, and the cells cultured a further 2 days before centrifugation and subsequent mRNA isolation (Sample G, differentiated cells). All experiments were approved by the Karolinska Institute Ethical Committee.
cDNA synthesis
Messenger RNA was isolated using Dynabeads® mRNA DIRECT™ Kit from Dynal (Dynal A.S., Norway), according to the manufacturer's instructions. First- and RNaseH dependent second-strand cDNA synthesis (SuperScript Choice System for cDNA Synthesis) was performed according to the manufacturer's instructions (Invitrogen, CA, USA) using 45 pmol biotinylated NotI-oligo(dT) primer (5'-biotin-GAGGTGCCAACCGCGGCCGC (T)15-3'). The cDNA was phenol-chloroform extracted and ethanol precipitated and the pellet was dissolved in 40 μl of 1 × TE (10 mM Tris-HCl, 1 mM EDTA). Excess NotI-oligo(dT) primer was removed by Chromaspinn TE-100 column (Clontech, CA, USA).
Amplification of 3'-end signature tags
The cDNA was fragmented and amplified according to a protocol previously described [3,4]. Shortly, fragmentation of the cDNA was performed in 40 μl 1 × TE using an inverted sonication probe, using 16 × 10 sec pulses at 90% effect (Sonifier® B-12, Branson Sonic Power Company, CT, USA). Biotinylated 3'-end signature tags from the fragmented cDNA population were isolated onto 20 μl of paramagnetic streptavidin-coated beads (10 mg/ml) (Dynal A.S.) in 40 μl sample plus 40 μl Binding/Washing buffer (2 M NaCl, 0.1% Tween 20 in 1 × TE, pH 7.7) at 37°C for one hour with rotation. The immobilised signature tags were end repaired using 1.5 U T4 DNA polymerase (New England BioLabs, MA, USA) in a 30-μl reaction volume at 12°C for 20 minutes according to the supplier's recommendations. Blunt-end adapters (Sima18: 5'-GGATCCGCGGTG-3'; Sima19: 5'-TCTCCAGCCTCTCACCGCGGATCC-3') were pre-annealed and ligated onto the immobilised repaired 3'-end signature tags using a solution comprising 1.1 nmol adapter, ligase buffer (66 mM Tris-HCl, pH 7.6, 5 mM MgCl2, 5 mM DTT, 50 μg/ml BSA), 0.2 mM ATP, 1200 U T4 DNA ligase (New England BioLabs) in a final volume of 60 μl. Ligation was performed overnight at room temperature with constant rotation to keep beads in suspension. The signature tags were released from the magnetic beads by restriction with NotI (New England BioLabs) for 2 hours in a volume of 60 μl while keeping the beads in suspension. Five micro litres of the eluate containing the 3'-end signature tags was used as template in a subsequent PCR. The PCR was performed in 100 μl containing 200 μM of each dNTP, 0.75 μM Sima19, 0.75 μM NotI-oligo(dT) primer, 65 mM Tris-HCl pH 8.8, 4 mM MgCl2, 16 mM (NH4)2SO4, 0.5 μM BSA and 3 U AmpliTaq DNA polymerase (Perkin Elmer, MA, USA). Cycling was performed according to the following procedure, initial incubation at 72°C for 3 min, followed by addition of Taq DNA polymerase and subsequent cycling: 72°C for 20 min, 95°C for 1 min, 45°C for 5 min, 72°C for 15 min, followed by four cycles (95°C for 1 min, 50°C for 1 min, 72°C for 15 min), and 13 cycles (as previously optimised) (95°C for 1 min, 50°C for 1 min, 72°C for 2 min).
Target labelling and microarray hybridisation
The 3'-end signature tags were purified using QIAquick® PCR purification kit (Qiagen, Germany). Direct labelling was performed using Cy3-dCTP or Cy5-dCTP (Perkin Elmer, MA, USA) in a linear, asymmetric PCR. The reaction was performed in a 50-μl labelling mix containing 100–200 ng purified 3'-end signature tags, 80 μM dATP, dGTP and dTTP, 20 μM dCTP, 5 μM Sima19 primer, 2 mM MgCl2, 1 × PCR Buffer II (Applied Biosystems, Ca, USA), 3 U AmpliTaq Gold® (Applied Biosystems) and 60 pM Cy3-dCTP or Cy5-dCTP. The labelling mix was cycled as follows: 95°C for 12 min, then 20 cycles (95°C for 30 sec, 50°C for 30 sec, 72°C for 10 min). Excess primer and nucleotides were removed using QIAquick® PCR purification kit (Qiagen). The eluted labelling products were speed vacuumed until dry, then dissolved in 55 μl hybridisation buffer (24% formamide, 5 × SSC and 0.1% SDS) (20 × SSC contains 3 M NaCl and 0.3 M Na3citrate × 2H2O). Cy3 and Cy5 labellings were blended and mixed with 25 μg human Cot-1 DNA (Invitrogen) and 50 μg polyA DNA (Operon Biotechnologies GmbH, Germany). The arrays (ArrayExpress accession number E-MEXP-297, submission in progress) [29] contained 5169 probes originating from a lateral ventricle wall cDNA library (clone library "Mus Musculus Lateral Ventricle Wall C57BL/6 adult") and a set of control features all printed in duplicate. Details regaring the array manufacturing are available through ArrayExpress. Briefly, probes were generated through PCR amplification and subsequently purified using Multiscreen-384 filter plates (Millipore). Purified products in 50% DMSO were printed onto GAPS-II slides (Corning Inc) using the QArray arrayer (Genetix) and attached using 250 mJ UV-light (Stratalinker). The arrays were first prehybridised for 30 min in a 42°C prehybridisation solution (1% BSA, 5 × SSC, 0.1% SDS), then washed in water and isopropanol and dried through centrifugation. The sample was denatured in 95°C for 3 min, then applied to the array and incubated in a hybridisation chamber in 42°C for 18 hours. After hybridisation the arrays were washed in three successive wash buffers with increasing stringency: (1) 1 × SSC and 0.2% SDS, 42°C, (2) 0.1 × SSC and 0.2% SDS, room temperature, (3) 0.1 × SSC, room temperature. All wash steps were made on a shaking table for 4 min. After the last step the array was immediately centrifuged in a slide centrifuge and kept in the dark until scanning. Scanning was performed using the GMS 418 Array Scanner from Genetic MicroSystems (Affymetrix Inc, CA, USA).
Image and data analysis
All image and data analysis steps were conducted in GenePix Pro 4.1.1.4 (Axon Instruments Inc, CA, USA) or R [30]. The analysis in R was carried out using Bioconductor [31], LIMMA [16], aroma [32] and the kth-package [33]. The analysis was conducted according to the following workflow. (1) Image tiff-files were created by scanning the microarrays with the GMS 418 Array Scanner. (2) Feature identity and foreground/background intensities were extracted from the tiff files using GenePix Pro 4.1.1.4. (3) GenePix result files were imported into R and gene expression measurements were obtained for each feature by subtracting the median of the local background from the median of the foreground signal. (4) A filter was used to identify and correct for features that had one channel (Cy3 or Cy5) below the background or at zero and one channel stronger than the background. The signal in the weaker channel was for these spots set to one plus the intensity of the local background. Features with both channels below the background or at zero were removed from the data set. (5) A second filter was used to remove features that were saturated in both channels. (6) A third filter was used to remove features with abnormal size (below 110 and above 230 μm in diameter). (7) A fourth filter was used to remove features where both signals had more than 70% of the pixels in the feature below the local background signal plus two standard deviations. (8) The last filter was used to remove features that were flagged as not found by GenePix. (9) Filtered data was normalised separately for each individual block on the slide using a robust local regression, print-tip lowess normalisation [34]. (10) An empirical Bayes moderated t-test [15-17] was used to rank the genes according to evidence of differential expression. The obtained p-values were adjusted for multiple testing using the false discovery rate adjustment [35] implemented in R. A p-value of less than 0.001 was considered significant and the associated gene termed differentially expressed (DE). The experimental design included reciprocal dye label assignments. These were swapped prior to the moderated t-test so that in each comparison the genes in the sample with an abbreviation that comes earlier in alphabetical order (e.g. B in B vs. CI) have positive M-values if they have a higher expression level.
List of abbreviations
NSC; neural stem cell
DE; differentially expressed
NS; neurosphere
DC; differentiated cells
fdr; false discovery rate
Authors' contributions
MS participated in the design of the study, drafted the manuscript, coordinated and carried out microarray experiments as well as performed data processing and data analysis. VW coordinated and carried out the manufacturing of the microarrays, performed data analysis and statistical analysis as well as assisted with the manuscript. AM participated in the design of the study, cultured cells and assisted with the manuscript. KM dissected the lateral ventricle wall tissue, isolated mRNA and coordinated the cDNA library construction for the array manufacturing, as well as assisted in writing of the manuscript. RE participated in the clone selection for the microarrays. LW conceived of the study and participated in its design and coordination. JF conceived of the study, participated in its design and coordination and assisted with the manuscript. JL conceived of the study, participated in its design and coordination of the study and helped to draft the manuscript, principle investigator. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
Genes differentially expressed in the neurosphere vs. differentiated cells comparison (F-G). Genes with p < 0.001, calculated by empirical Bayes moderated t-test and false discovery rate adjustment, are included. Genes with M-value < 0 are up-regulated in differentiated cells, genes with M-value > 0 are up-regulated in neurospheres. M = log2(Cy5/Cy3); A = 1/2log2(Cy5*Cy3); p-value = unadjusted p-value; fdr adjusted p-value = false discovery rate adjusted p-value; B = B-value calculated by empirical Bayes moderated t-test. Higher B-values mean higher probability for differential expression.
Click here for file
Acknowledgements
We thank Anna Westring, Peter Nilsson and Cecilia Williams for valuable assistance and comments. This work was supported by grants from the Knut and Alice Wallenberg Foundation, the Wallenberg Consortium North, the Swedish Cancer Foundation and the Swedish Scientific Research Council.
==== Refs
Van Gelder RN von Zastrow ME Yool A Dement WC Barchas JD Eberwine JH Amplified RNA synthesized from limited quantities of heterogeneous cDNA Proc Natl Acad Sci U S A 1990 87 1663 1667 1689846
Eberwine J Yeh H Miyashiro K Cao Y Nair S Finnell R Zettel M Coleman P Analysis of gene expression in single live neurons Proc Natl Acad Sci U S A 1992 89 3010 3014 1557406
Hertzberg M Sievertzon M Aspeborg H Nilsson P Sandberg G Lundeberg J cDNA microarray analysis of small plant tissue samples using a cDNA tag target amplification protocol Plant J 2001 25 585 591 11309148 10.1046/j.1365-313x.2001.00972.x
Sievertzon M Agaton C Nilsson P Lundeberg J Amplification of mRNA populations by a cDNA tag strategy Biotechniques 2004 36 253 259 14989090
Reynolds BA Weiss S Generation of neurons and astrocytes from isolated cells of the adult mammalian central nervous system Science 1992 255 1707 1710 1553558
Gottlieb DI Large-scale sources of neural stem cells Annu Rev Neurosci 2002 25 381 407 12052914 10.1146/annurev.neuro.25.112701.142904
Johansson CB Svensson M Wallstedt L Janson AM Frisen J Neural stem cells in the adult human brain Exp Cell Res 1999 253 733 736 10585297 10.1006/excr.1999.4678
Suslov ON Kukekov VG Ignatova TN Steindler DA Neural stem cell heterogeneity demonstrated by molecular phenotyping of clonal neurospheres Proc Natl Acad Sci U S A 2002 99 14506 14511 12381788 10.1073/pnas.212525299
Johansson CB Momma S Clarke DL Risling M Lendahl U Frisen J Identification of a neural stem cell in the adult mammalian central nervous system Cell 1999 96 25 34 9989494 10.1016/S0092-8674(00)80956-3
Qian X Shen Q Goderie SK He W Capela A Davis AA Temple S Timing of CNS cell generation: a programmed sequence of neuron and glial cell production from isolated murine cortical stem cells Neuron 2000 28 69 80 11086984 10.1016/S0896-6273(00)00086-6
Galli R Pagano SF Gritti A Vescovi AL Regulation of neuronal differentiation in human CNS stem cell progeny by leukemia inhibitory factor Dev Neurosci 2000 22 86 95 10657701 10.1159/000017430
Caldwell MA He X Wilkie N Pollack S Marshall G Wafford KA Svendsen CN Growth factors regulate the survival and fate of cells derived from human neurospheres Nat Biotechnol 2001 19 475 479 11329020 10.1038/88158
Lindvall O Kokaia Z Martinez-Serrano A Stem cell therapy for human neurodegenerative disorders-how to make it work Nat Med 2004 10 Suppl S42 50 15272269 10.1038/nm1064
Uchida N Buck DW He D Reitsma MJ Masek M Phan TV Tsukamoto AS Gage FH Weissman IL Direct isolation of human central nervous system stem cells Proc Natl Acad Sci U S A 2000 97 14720 14725 11121071 10.1073/pnas.97.26.14720
Lönnstedt I Speed T Replicated microarray data Statistica Sinica 2002 12 31 46
Smyth G Linear models and empirical Bayes methods for assessing differential expression in microarray experiments Statistical Applications in Genetics and Molecular Biology 2004 3 Article 3 16646809
Smyth GK Michaud J Scott H The use of within-array duplicate spots for assessing differential expression in microarray experiments Submitted 2003
Jiang M Spicher K Boulay G Wang Y Birnbaumer L Most central nervous system D2 dopamine receptors are coupled to their effectors by Go Proc Natl Acad Sci U S A 2001 98 3577 3582 11248120 10.1073/pnas.051632598
Andrews GL Yun K Rubenstein JL Mastick GS Dlx transcription factors regulate differentiation of dopaminergic neurons of the ventral thalamus Mol Cell Neurosci 2003 23 107 120 12799141 10.1016/S1044-7431(03)00016-2
Saino-Saito S Berlin R Baker H Dlx-1 and Dlx-2 expression in the adult mouse brain: relationship to dopaminergic phenotypic regulation J Comp Neurol 2003 461 18 30 12722102 10.1002/cne.10611
Yun K Fischman S Johnson J Hrabe de Angelis M Weinmaster G Rubenstein JL Modulation of the notch signaling by Mash1 and Dlx1/2 regulates sequential specification and differentiation of progenitor cell types in the subcortical telencephalon Development 2002 129 5029 5040 12397111
Hosack DA Dennis GJ Sherman BT Lane HC Lempicki RA Identifying biological themes within lists of genes with EASE Genome Biol 2003 4 R70 14519205 10.1186/gb-2003-4-10-r70
Morshead CM Benveniste P Iscove NN van der Kooy D Hematopoietic competence is a rare property of neural stem cells that may depend on genetic and epigenetic alterations Nat Med 2002 8 268 273 11875498 10.1038/nm0302-268
Reynolds BA Weiss S Clonal and population analyses demonstrate that an EGF-responsive mammalian embryonic CNS precursor is a stem cell Dev Biol 1996 175 1 13 8608856 10.1006/dbio.1996.0090
Bez A Corsini E Curti D Biggiogera M Colombo A Nicosia RF Pagano SF Parati EA Neurosphere and neurosphere-forming cells: morphological and ultrastructural characterization Brain Res 2003 993 18 29 14642827 10.1016/j.brainres.2003.08.061
Ostenfeld T Joly E Tai YT Peters A Caldwell M Jauniaux E Svendsen CN Regional specification of rodent and human neurospheres Brain Res Dev Brain Res 2002 134 43 55 11947936 10.1016/S0165-3806(01)00291-7
Lobo MV Alonso FJ Redondo C Lopez-Toledano MA Caso E Herranz AS Paino CL Reimers D Bazan E Cellular characterization of epidermal growth factor-expanded free-floating neurospheres J Histochem Cytochem 2003 51 89 103 12502758
Kukekov VG Laywell ED Suslov O Davies K Scheffler B Thomas LB O'Brien TF Kusakabe M Steindler DA Multipotent stem/progenitor cells with similar properties arise from two neurogenic regions of adult human brain Exp Neurol 1999 156 333 344 10328940 10.1006/exnr.1999.7028
ArrayExpress—a public repository for microarray gene expression data at the EBI
R Development Core Team R: A language and environment for statistical computing
Gentleman R Carey V Bates D Bolstad B Dettling M Dudoit S Ellis B Gautier L Ge Y Gentry J Hornik K Hothorn T Huber W Iacus S Irizarry R Leisch F Li C Maechler M Rossini A Sawitzki G Smith C Smyth G Tierney L Yang J Zhang J Bioconductor: open software development for computational biology and bioinformatics Genome Biology 2004 5 R80 15461798 10.1186/gb-2004-5-10-r80
Bengtsson H aroma - An R Object-oriented Microarray Analysis environment
Wirta V Lindberg J Gry Björklund M The kth-package for microarray data analysis 2004
Yang YH Dudoit S Luu P Lin DM Peng V Ngai J Speed TP Normalization for cDNA microarray data: a robust composite method addressing single and multiple slide systematic variation Nucleic Acids Res 2002 30 e15 11842121 10.1093/nar/30.4.e15
Benjamini Y Hochberg Y Controlling the false discovery rate: a practical and powerful approach to multiple testing Journal of the Royal Statistical Society Series B 1995 57 289 300
| 15833137 | PMC1112599 | CC BY | 2021-01-04 16:39:09 | no | BMC Neurosci. 2005 Apr 15; 6:28 | utf-8 | BMC Neurosci | 2,005 | 10.1186/1471-2202-6-28 | oa_comm |
==== Front
BMC PediatrBMC Pediatrics1471-2431BioMed Central London 1471-2431-5-61585048410.1186/1471-2431-5-6Research ArticleIdentification of children who may benefit from self-hypnosis at a pediatric pulmonary center Anbar Ran D [email protected] Susan C [email protected] Department of Pediatrics, University Hospital, State University of New York Upstate Medical University, Syracuse, NY, USA2005 25 4 2005 5 6 6 15 12 2004 25 4 2005 Copyright © 2005 Anbar and Geisler; licensee BioMed Central Ltd.2005Anbar and Geisler; 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
Emotional difficulties can trigger respiratory symptoms. Thus, children presenting with respiratory complaints may benefit from a psychological intervention. The purpose of this study was to define the proportion of patients referred to a Pediatric Pulmonary Center who may benefit from instruction in self-hypnosis, as a psychological intervention.
Methods
A retrospective chart review was conducted for all newly referred patients to the SUNY Upstate Medical University Pediatric Pulmonary Center during an 18 month period beginning January 1, 2000. Patients were offered hypnosis if they presented with symptoms or signs suggestive of psychological difficulties. Hypnosis was taught in one or two 15–45 minute sessions by a pediatric pulmonologist.
Results
Of 725 new referrals, 424 were 0–5 years old, 193 were 6–11 years old, and 108 were 12–18 years old. Diagnoses of anxiety, habit cough, or vocal cord dysfunction accounted for 1% of the 0–5 year olds, 20% of the 6–11 year olds, and 31% of the 12–18 year olds. Hypnotherapy was offered to 1% of 0–5 year olds, 36% of 6–11 year olds, and 55% of 12–18 year olds. Of 81 patients who received instruction in self-hypnosis for anxiety, cough, chest pain, dyspnea, or inspiratory difficulties, 75% returned for follow-up, and among the returning patients 95% reported improvement or resolution of their symptoms.
Conclusion
A large number of patients referred to a Pediatric Pulmonary Center appeared to benefit from instruction in self-hypnosis, which can be taught easily as a psychological intervention.
anxietyasthmahabit coughhypnosisvocal cord dysfunction
==== Body
Background
Emotional difficulties can trigger respiratory symptoms such as dyspnea [1]. Further, psychological issues can arise as a result of patients' reactions to pulmonary disease, e.g., depression with end-stage cystic fibrosis [2]. In some patients, a vicious cycle ensues as pulmonary disease leads to psychological difficulties, which in turn trigger further symptoms that can be misinterpreted as arising from the pulmonary disease [3]. Thus, a patient with severe asthma can develop anxiety as a result of the life-threatening nature of the disease. Such stress can cause dyspnea, which might be treated inappropriately with therapy for asthma rather than anxiety [4]. In a study by Ortega et al. [5], 49% of children with asthma from cohorts in New Haven, Atlanta, NewYork, and Puerto Rico were identified through the Diagnostic Interview Schedule for Children as having an anxiety disorder. Further, a history of psychogenic stressors or psychiatric disorder often is identified in pediatric patients with functional respiratory disorders, such as sighing dyspnea, habit cough, and vocal cord dysfunction [6]. For example, in a literature review, 52% of patients with vocal cord dysfunction were diagnosed as having a conversion disorder [7]. This evidence suggests that children presenting with respiratory complaints may benefit from a psychological intervention.
Previously, we reported the outcome of hypnotherapy offered at our Pediatric Pulmonary Center [1,4,8-10]. Reported rates of improvement following hypnotherapy ranged from 86% of patients with anxiety [8], 90% of patients with habit cough [10], 91% of patients with vocal cord dysfunction [8], and 100% of patients with chronic dyspnea, who had normal lung function at rest [9].
The purpose of the current study was to define the proportion of all patients referred to our Pediatric Pulmonary Center who might receive benefit from instruction in self-hypnosis as a psychological intervention.
Methods
A retrospective chart review was undertaken for all patients newly referred to the SUNY Upstate Medical University Pediatric Pulmonary Center during the 18 months beginning January 1, 2000. Most of these referrals were from primary care providers. Information collected included age, gender, referral diagnosis, whether and for what purpose they were offered hypnotherapy or any other psychological intervention, the results of the intervention, and diagnosis in 2003 at the time of the data collection, or at the time of discharge from the Center if this occurred before 2003. Assessment of intervention effectiveness was based on the patients' subjective reports, except in the cases of patients with habit cough, or stridor associated with vocal cord dysfunction, whose symptoms were observed to have resolved during a visit at our Center.
Patients were evaluated for their respiratory complaints by a thorough review of their history, physical examination, and laboratory investigations, including pulmonary function testing, radiological investigations, and blood studies. Patients were offered hypnosis if they presented with symptoms or signs suggestive of psychological difficulties, such as those listed in Table 1. Formal testing for psychological disorders or hypnotizability was not utilized. Those who expressed interest in hypnotherapy were instructed in self-hypnosis techniques by a pediatric pulmonologist. Hypnosis was taught in one or two 15–45 minute sessions, as described previously [8]. Patients who required psychological intervention other than hypnotherapy were to be referred to a child psychiatrist.
Table 1 Symptoms and signs suggestive of psychological difficulties*
Respiratory symptoms
Difficulty with inspiration
Disruptive cough
Dyspnea despite normal lung function
Hyperventilation
Inspiratory noise (e.g., stridor, gasping, rasping, or squeak)
Localization of breathing problem to the neck or upper chest
Sighing
Other symptoms
Anxious appearance
Dizziness
Feeling something is stuck in the throat
Palpitations
Paresthesias
Shakiness
Symptom characteristics
Absence during sleep or when patient is distracted
Associated with a particular location or activity
Emotional response to symptoms
Emotional trigger of symptoms
Exposure to traumatic life event
Incomplete response to medications
* Adapted from references 1, 4, 8–10.
As this study involved a retrospective chart review without identification of patients, exemption was given from review by the Institutional Review Board.
Results
Of the 725 newly referred patients (424 were 0–5 years, 193 were 6–11 years, and 108 were 12–18 years old), 133 (18%) were offered hypnotherapy. No patients required referral to a psychiatrist. Patients offered hypnotherapy tended to be older: Hypnotherapy was offered to 1% of 0–5 year olds, 36% of 6–11 year olds, and 55% of 12–18 year old patients. Table 2 lists the main reasons for the hypnotherapy.
Table 2 Reasons for offering hypnotherapy
n = 133
Percent of patients
Respiratory symptoms
Cough 19
Chest pain 8
Dyspnea 25
Inspiratory difficulties 8
Other reasons
Altering palatability of medications 8
Anxiety 9
Headaches 6
Insomnia 4
Relaxation 11
Other 14
Anxiety, habit cough, and vocal cord dysfunction were the three diagnoses made at our Center in this study population that are recognized commonly as having major psychological components [5,6]. These diagnoses accounted for 1% of the 0–5 year olds, 20% of the 6–11 year olds, and 31% of the 12–18 year olds.
A referral diagnosis of asthma accounted for 28% of 0–5 year olds, 50% of the 6–11 year olds, and 53% of the 12–18 year olds. Among these patients, anxiety, habit cough, and vocal cord dysfunction were diagnosed in 4% of 0–5 year olds, 21% of the 6–11 year olds, and 26% of the 12–18 year old patients.
One hundred sixteen of the 133 patients (87%) offered hypnotherapy agreed to receive instruction in self-hypnosis. Eighty-one of the patients received such instruction for anxiety or the respiratory symptoms listed in Table 2. Among these 81, 75% returned for follow-up, and among the returning patients 95% reported improvement or resolution of their symptoms, as previously described in detail for many of these patients [1,8-10].
Discussion
This study demonstrates that a large number of patients referred to a Pediatric Pulmonary Center may benefit from instruction in self-hypnosis. As reported elsewhere, many of the patients in this report, e.g., those with anxiety, habit cough or vocal cord dysfunction, failed to improve prior to introduction of hypnosis [1,8-10]. Therefore, it is likely that hypnosis was important for their recovery. It is possible that the time spent with the pulmonologist, or the reassurance received regarding the absence of physiologic disease were the critical parts of the intervention, as opposed to the hypnotherapy. Even if this were the case, the findings in this report underscore that a significant number of patients can respond to a therapeutic interaction that addresses their psychological needs.
Given our finding that a large number of referred patients may benefit from psychological intervention, we believe that health care providers should familiarize themselves with symptoms and signs that may indicate psychological difficulties (see Table 1). Further, providers should identify mechanisms by which patients' psychological issues can be addressed appropriately. Provision of instruction in self-hypnosis techniques can allow for a rapid, effective intervention that patients often accept readily [8-11]. Hypnosis should not be offered in situations where it might aggravate existing emotional problems, or the problem might be treated more effectively by another method [11]. Interested clinicians can receive training in self-hypnosis techniques through workshops sponsored by the American Society of Clinical Hypnosis, the Society for Clinical and Experimental Hypnosis, or the Society for Developmental and Behavioral Pediatrics [11].
The proportion of patients with psychological difficulties contributing to respiratory symptoms at our tertiary-care Center may be greater than that in a general pediatric practice because patients who respond well to medical therapy are less likely to be referred. Further, the proportion also may be different when compared to other Pediatric Pulmonary Centers. For example, if the socio-economic mix of referred patients is different between Centers, the proportion of patients with physiologic disease may vary between Centers as a result of different levels of environmental exposures and adherence to prescribed therapy. Also, the type of patient referred to our Center may be biased as a result of our recognized interest in the treatment of symptoms with a possible psychological basis. On the other hand, it is possible that we would have diagnosed more patients with psychological difficulties had we used formal psychological testing [5,11]. The relatively high number of patients lost to follow-up in this report is attributable to the study population that was derived from a clinical practice. Thus, controlled studies with close follow-up are needed to help better define the utility of self-hypnosis.
Psychological intervention in the comprehensive management of pediatric patients also is likely to be of benefit for a large number of patients in other pediatric practice settings including general pediatrics [11], and sub-specialty centers, including gastroenterology (e.g., for functional abdominal pain) [12], nephrology (e.g., for enuresis and dysfunctional voiding) [11], neurology (e.g., for headaches) [11], and surgery (e.g., for promotion of recovery) [11].
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
RA is the pediatric pulmonologist described in this report. He conceived the study and wrote the manuscript. SG collected and analyzed the data, and revised the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Anbar RD Stressors associated with dyspnea in childhood: patients' insights and a case report Am J Clin Hypn 2004 47 93 101 15554462
Bennett DS Depression among children with chronic medical problems: a meta-analysis J Pediatr Psychol 1994 19 149 169 8051600
Mrazek DA Psychiatric complications of pediatric asthma Ann Allergy 1992 69 285 290 1416262
Anbar RD Self-hypnosis for anxiety associated with severe asthma: a case report BMC Pediatrics 2003 3 7 12875663 10.1186/1471-2431-3-7
Ortega AN Huertas SE Canino G Ramirez R Rubio-Stipec M Childhood asthma, chronic illness, and psychiatric disorders J Nerv Ment Dis 2002 190 275 281 12011605 10.1097/00005053-200205000-00001
Butani L Oconnell EJ Functional respiratory disorders Ann Allergy Asthma Immunol 1997 79 91 101 9291412
Lacy TJ McManis SE Psychogenic stridor Gen Hosp Psychiatry 1994 16 213 223 8063089 10.1016/0163-8343(94)90104-X
Anbar RD Hypnosis in pediatrics: applications at a pediatric pulmonary center BMC Pediatrics 2002 2 11 12460456 10.1186/1471-2431-2-11
Anbar RD Self-hypnosis for management of chronic dyspnea in pediatric patients Pediatrics 2001 107
Anbar RD Hall HR Childhood habit cough treated with self-hypnosis J Pediatr 2004 144 213 217 14760264 10.1016/j.jpeds.2003.10.041
Olness K Kohen DP Hypnosis and Hypnotherapy with Children 1996 3 New York: The Guilford Press
Anbar RD Self-hypnosis for the treatment of functional abdominal pain in childhood Clin Pediatr (Phila) 2001 40 447 451 11516052
| 15850484 | PMC1112600 | CC BY | 2021-01-04 16:31:06 | no | BMC Pediatr. 2005 Apr 25; 5:6 | utf-8 | BMC Pediatr | 2,005 | 10.1186/1471-2431-5-6 | oa_comm |
==== Front
BMC Public HealthBMC Public Health1471-2458BioMed Central London 1471-2458-5-401584017310.1186/1471-2458-5-40Research ArticleWas an increase in cocaine use among injecting drug users in New South Wales, Australia, accompanied by an increase in violent crime? Degenhardt Louisa [email protected] Carolyn [email protected] Wayne [email protected] Elizabeth [email protected] Stuart [email protected] National Drug and Alcohol Research Centre, University of New South Wales, Sydney Australia2 National Centre in HIV Epidemiology & Clinical Research, University of New South Wales, Sydney Australia3 Office of Public Policy and Ethics, Institute for Molecular Bioscience, University of Queensland, St Lucia Australia2005 19 4 2005 5 40 40 20 8 2004 19 4 2005 Copyright © 2005 Degenhardt et al; licensee BioMed Central Ltd.2005Degenhardt 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 sharp reduction in heroin supply in Australia in 2001 was followed by a large but transient increase in cocaine use among injecting drug users (IDU) in Sydney. This paper assesses whether the increase in cocaine use among IDU was accompanied by increased rates of violent crime as occurred in the United States in the 1980s. Specifically, the paper aims to examine the impact of increased cocaine use among Sydney IDU upon police incidents of robbery with a weapon, assault and homicide.
Methods
Data on cocaine use among IDU was obtained from the Illicit Drug Reporting System (IDRS). Monthly NSW Police incident data on arrests for cocaine possession/use, robbery offences, homicides, and assaults, were obtained from the Bureau of Crime Statistics and Research. Time series analysis was conducted on the police data series where possible. Semi-structured interviews were conducted with representatives from law enforcement and health agencies about the impacts of cocaine use on crime and policing.
Results
There was a significant increase in cocaine use and cocaine possession offences in the months immediately following the reduction in heroin supply. There was also a significant increase in incidents of robbery where weapons were involved. There were no increases in offences involving firearms, homicides or reported assaults.
Conclusion
The increased use of cocaine among injecting drug users following the heroin shortage led to increases in violent crime. Other States and territories that also experienced a heroin shortage but did not show any increases in cocaine use did not report any increase in violent crimes. The violent crimes committed did not involve guns, most likely because of its stringent gun laws, in contrast to the experience of American cities that have experienced high rates of cocaine use and violent crime.
==== Body
Background
In the United States in the mid and late 1980s, there was a cocaine epidemic fuelled by the use of "crack" cocaine [1-4]. At the same time, increases were noted in violent crime [5]. New York experienced a particularly notable increase in the extent of violent crime in the city [4,6,7]. These violent crimes often involved firearms and led to an increased homicide rate [5]. More recently, increases in the availability and use of crack in the United Kingdom (UK) [8,9] have also been accompanied by increases in violent crime, also involving firearms [10].
Goldstein [11] proposed a tripartite model to explain why crime may be related to drug use. First, the psychopharmacology of the drug may increase the users' likelihood of acting in a violent manner. Second, violent crimes may be committed by users to finance expensive drug habits. Third, crime may be related to the distribution and sale of drugs, particularly with respect to distributors' need to protect market share. Previous analyses of the US crack cocaine epidemic suggested that all three of these factors may have been involved in the crack cocaine epidemic, and in the escalation of violent crime in the US [4-7]. Suggestive evidence has been collected that the increase in crime may have been related to the relative youth of drug market participants [5]; the profitability of crack cocaine distribution and hence disputes over market share and distribution points [6]; and some contribution from the pharmacological effects of sustained crack cocaine use among low level user-dealers [7]. The relatively easy availability of firearms at that time has been argued to be a large contributor to the increases in homicide observed in the US during the period [5].
In contrast to the US, Australia has had relatively little notable cocaine use among problematic drug users [12-14]. In the later part of the 20th century harms related to cocaine use have historically been low across the country [15,16]. This may be related to the high cost and relatively low availability of the drug in street based drug markets [12,16]; low rates of cocaine injecting or crack smoking among regular IDU [12,16]; and the purported concentration of use (because of its high costs) among smaller, advantaged social groups or commercial sex workers (who may have greater disposable income) [17-19].
In New South Wales (NSW) in the late 1990s, heroin was the drug most frequently reported by regular IDU as their drug of injection and choice. In early 2001, there were reports of a dramatic decline in the availability of heroin in Sydney, NSW [20,21]. This was confirmed by the 2001 Illicit Drug Use Reporting System (IDRS), Australia's strategic early warning system. The IDRS observed an overall reduction in the availability and street level purity of heroin, and an increase in heroin price for all major heroin markets that began in early 2001 and was sustained for much of that year [22,23].
Following this reduction in heroin supply, regular IDU reported less frequent heroin use, and more frequent cocaine use [24,25]. The availability of cocaine powder had also increased [24], although there was no evidence of the emergence of crack cocaine. An examination of changes in drug distribution in NSW suggested that those involved in street level and mid level heroin distribution began distributing cocaine when heroin became less available [26].
This sudden reduction in heroin use and increase in powder cocaine use provided a unique opportunity to conduct a natural experiment into the relationship between powder cocaine use and violent crime. We examined if changes in the nature or extent of violent crime in NSW following evidence of increased availability and use of powder cocaine were similar to those observed in New York when crack cocaine availability increased. Specifically, this paper aimed to do the following:
1. Examine changes in cocaine use in NSW from 2001 (see [24,26,27]);
2. Examine potential changes in rates of violent crime at this time;
3. Examine the extent to which these changes in cocaine use and crime were related.
Methods
Data used in the study
Semi-structured interviews with heroin users
Heroin users were recruited via advertisements placed in opioid pharmacotherapy clinics. They had to have (a) recent experience of the drug market and (b) to have commenced pharmacotherapy either between August and December 2000 (pre-shortage) or between February and April 2001 (during shortage). Fifty three users were interviewed in total, approximately half entering treatment in each time period. Users were surveyed on a range of issues including their involvement in and experience of drug markets prior to and during the heroin shortage [28].
Semi-structured interviews with key informants (KI)
Selection of key informants was based on one or more of the following:
• the extent of their contact with the illicit drug market;
• their level of knowledge of the illicit drug market and illicit drug users;
• the focus of their position (e.g. direct/indirect, operational/policy); and
• the length of time the key informant had held the position, particularly their ability to comment on changes over time, pre to post heroin shortage.
Law enforcement
The NSW Police Service comprises three levels of command: State, Region and Local Area Command (LAC). Key informants were selected across all three levels and across the four LAC responsible for policing the three Sydney open air drug markets, two region commands in which these LAC were located and a range of squads within the State Command (including squads focused on organised crime groups and drug crime).
A total of 22 law enforcement key informants were interviewed for this study, 20 of whom were sworn officers. Seven were state level personnel, 2 regional personnel and 13 LAC personnel. Seven interviewees held the position of Commander of their squad, 7 were managers of their unit, and 8 held general duty or operational positions (the latter includes 2 civilians – an analyst and a pharmacist).
Health
A total of 49 health KI were recruited for this study, including 5 from NSW State organisations. The remainder were recruited in the Sydney drug markets of Kings Cross (n = 16), Cabramatta (n = 14) and Redfern (n = 15). The roles of these KI were as follows: drug health (n = 25), community health (n = 3), community welfare (n = 5), emergency health (n = 1), indigenous health (n = 2), mental health (n = 2), prenatal health (n = 4), primary health care (n = 2) and youth services (n = 5).
NSW police incident data
NSW Police record all police activity in a centralised database known as the Computerised Operational Policing System (COPS). This information can be analysed at the level of 'event' or 'incident'. An event is a record created in COPS whenever police attend a criminal or non-criminal activity. An event includes the incidents that comprise it (what happened, where, who was involved) and the actions taken by police in response to the event. This information is not reliant on a charge having been laid (but offender details on gender and age may not be provided if the offender is not arrested). Information from this 'real time' dataset is downloaded at regular intervals for analysis by the NSW Bureau of Crime Statistics and Research (BOCSAR). The following incident types were used in the current study: cocaine possession/use, robbery with a firearm, robbery with a weapon (not a firearm), robbery without a weapon, homicide, assault, and weapons offences.
Time series analysis
The indicator data series were analysed using an ARIMA model time series. The heroin shortage was represented in these models in the following three ways as: 1) a permanent effect (step); 2) a brief effect (pulse); or 3) a change in slope. Analyses dated the onset of the heroin shortage from January 2001, in accordance with the findings of other research on the course of the event [23]. Intervention models were fitted using SAS v 8.2. Intervention ARIMA models can require estimation of many parameters, and some of the data series lacked clearly definable responses at the point of the heroin shortage (e.g. Figure 4). In order to avoid large probability of type I error, analysis of data series which showed no evidence of a response to the heroin shortage on visual inspection were analysed by examination of crosscorrelation functions only. If the crosscorrelation functions for these series showed no clear evidence of an effect due to the shortage no further modelling was conducted on these series and the conclusion of no noticeable effect due to the heroin shortage was drawn.
Figure 4 Incidents of homicide in NSW, 1997–2002
Results
Trends in cocaine use
Clear increases were observed in the use of cocaine among regular injecting drug users in 2001 (Figure 1). This was true whether IDU were asked about their use of cocaine in the previous day, the number of days used in the past 6 months, or whether it was the last drug they had injected. This increase did not persist, however, with the proportion reporting cocaine use decreasing in 2002 and further in 2003 (Figure 1).
Figure 1 Proportion of IDU reporting cocaine use in the past six months, daily use, and use on the day preceding interview, 1996–2003
Figure 2 shows the number of incidents recorded for cocaine possession/use in NSW. This peaked at 64 in March 2001 and remained high throughout the year but declined in 2002. The modelled series (Figure 2) showed that while police incidents for cocaine possession or use were at a steady level prior to the reduction in heroin supply, they increased significantly over the six months following the reduced heroin supply before returning to the levels seen prior to the heroin shortage. The maximum increase of 207% occurred 2 months after the shortage began (March 2001).
Figure 2 Incidents of cocaine possession/use in NSW, 1997–2002
Trends in robbery offences
Figure 3 shows the number of robbery offences in NSW for the period January 1997 to December 2002. The onset of the reduction in heroin supply was associated with a 33% (p < 0.0001) increase in the incidence of robbery without a weapon. The trend seen in incidents of robbery with a weapon other than a firearm followed a similar pattern. In contrast, there was no apparent effect of the heroin shortage upon the series robbery with a firearm.
Figure 3 Incidents of robbery offences in NSW 1997–2002
Results of the time series analysis were consistent with the qualitative information collected in KI interviews. An increase in the incidence of robberies was the single most commonly reported change in criminal activity. KI consistently attributed it to a combination of the behavioural effects of cocaine and the need to increase criminal activity to fund the higher cost of using cocaine.
So, more likely to commit crime or for violence to be included in the crime and that seems to be likely given that people would be more desperate and presumably also, you know, if they are using cocaine in associated with their crimes, more reckless and more aggressive, abusive, volatile. (Health/Welfare KI)
KI reports described thefts 'gone wrong' in which excessive force and crude weapons were opportunistically used. Some drug users also reported attempting ill-planned armed robberies and being caught in the act by police
KIs reported that the type of crime engaged in by individual users during the heroin shortage was "out of character", and that users were less careful in the commission of crime. Overall, the tone of the offences changed: KIs and users in all markets reported that drug-related crime became more desperate, violent and impulsive. KIs reported that users stepped up their involvement in crime, moving from non-violent acquisitive crime (i.e. theft) to violent acquisitive crime (i.e. robbery).
The behavioural effects of cocaine meant that the execution of a theft often became more violent than intended or than was typical for that offender.
"The whole nature of the offences changed. There was no change in so far as people were still doing property offences, stealing – it's all the same. But people weren't getting enough money so they'd turn to violent offences. But not just that, because of the amount of cocaine they were using, it was just making them angry." (law enforcement KI)
Trends in homicide and assault
Figures 4 and 5 show the incidence of homicide and assault offences between 1997 and 2002. There did not appear to be a change in either time series around the time when cocaine use increased. Apart from the general increase in violence commonly reported by KIs, there were no reports of any changes in the incidence of homicides and assaults.
Figure 5 Incidents of assault in NSW, 1997–2002
Trends in weapons offences
Figure 6 shows the number of weapons offences in NSW. This offence category includes charges relating to the illegal possession, sale and discharge of firearms and offences relating to explosive/dangerous articles or threats. There was no change in the incidence of weapons offences at the time of the shortage (Jan–Apr 2001), either at a state or local level. It should be noted that the sharp increase in the series at the beginning of 1999 reflects a change in the legislation that gave NSW Police the power to conduct knife searches.
Figure 6 Incidents of weapons offences, NSW 1997–2002
KIs typically mentioned the increased use of weapons in acquisitive crimes. Some thought that organised crime groups were involved in the distribution of firearms as part of their criminal repertoire, but these activities were not linked to either the reduction in heroin supply or the increase in cocaine availability and use.
Discussion
This study has found a clear increase in the rate of violent crime concurrent with an increase in cocaine use. As has been shown elsewhere, there were marked increases in the use of cocaine among regular injecting drug users in major drug markets in Sydney [24,27]. There was also evidence from NSW Police records that this increase was observed at a State level for cocaine possession/use offences [26]. These findings were supported by data on the number of calls of concern to NSW telephone help lines about cocaine [27], and increased reports of cocaine as the last drug injected by NSP attendees [27],. The consistency of these changes suggests that there was a definite shift in drug use patterns in the IDU community from heroin to cocaine injecting.
These increases in cocaine use were accompanied by increased rates of violent crime. Consistent with the model proposed by Goldstein [11], interviews with KI of all types suggested that increases in violent acquisitive crime was related to both the psychopharmacological effects of heavy cocaine use, and also to the increased financial costs of users' drug use. KI reports were also obtained of violent crime occurring among those involved in cocaine distribution, but these could not be evaluated using police data.
Comparable research in other Australian States revealed little, if any, change in cocaine use among similar populations of IDU [29,30]. Furthermore, there was no significant increase in violent crime in these States. The absence of any increase provides further support for the argument that the increase in cocaine use among this disadvantaged group in NSW was causally related to the change in violent crime in that State. An increase in the rate of non-violent acquisitive crime in NSW [26] provided further evidence to support the notion that part of the increase in violent acquisitive crimes may have been related to the increased costs of drug users' habits following increases in the price of heroin, the previously dominant drug and the relatively higher cost of the cocaine that some switched to using after the onset of the heroin shortage.
Perhaps the most interesting difference between the experience in Sydney and New York with increasing cocaine use was the lack of any increase in the number of gun related incidents in Sydney, compared to a dramatic increase in such incidents that occurred in New York. There was an increase in the number of incidents of robberies involving weapons, but these did not involve guns. In New York, by contrast, many crimes involved guns and the homicide rate involving firearms increased markedly.
In New York and Sydney, organised crime groups have access to, and are involved in, the sale of illicit firearms. However, firearms have only limited availability for personal use in Australia [31], whereas they were relatively easily available in the United States at the height of the crack cocaine epidemic [5]. It seems possible that this availability leads to two sources of firearms in the US: licit sources, and illicit sources, whereby illicit sources may comprise the diversion of legally registered firearms as well as the large scale distribution of illegal firearms. In Australia, however, there are limited legal sources of handguns [31]. This suggests that maintaining such stringent controls upon firearms may have assisted in maintaining a low rate of firearm offences [32], even in the face of increased cocaine use among criminally involved IDU that increased the risk of violent incidents.
It is necessary to consider the potential influence of the forms of cocaine used in Australia, compared to the US and UK. In Australia, there is little or no use of "crack" cocaine in the country [12,33]; in contrast, the concerns related to cocaine use in the UK and US have largely centred on harms related to apparent epidemics of "crack" cocaine use. However, the current study has found an association between the increased use and availability of cocaine powder and violence, suggesting that the increased availability and use of cocaine powder may have a similar impact on violent crime as crack cocaine.
Limitations
This paper is subject to the flaws that beset all natural experiments, in that it is not possible to guarantee that the intervention being studied was the only event that affected cocaine use and/or violent crime in the time period. However, similar research on rates of crime conducted in the same time period in both Victoria and South Australia provided a control series. These two states were geographically isolated from NSW and both experienced a heroin shortage but neither experienced any increase in cocaine use among IDU.
Although it might be possible that some other event interfered in NSW drug markets at the same time as the heroin shortage, such possibilities were examined in a process of extensive crosschecking through KIs, consultation with stakeholders and analysis of other data sources in the wider project from which this study is drawn. No plausible alternative explanations remained [34,35].
Another limitation of the analysis concerns the relative simplicity of the analyses we have conducted. Ideally, it would be of interest to model trends over time in violent acquisitive crime that include not only cocaine use, but also other factors such as risk of arrest for robbery, the proportion of those committing robbery offences who were imprisoned, and the rate of unemployment in the community. As a result of reduced heroin availability, police reported having greater resources (in terms of available personnel) to target drug dealers involved in distributing drugs other than heroin, but it was consistently reported by police to be the case that those involved in low-level heroin distribution switched to cocaine distribution [26].
The increase in availability and use of cocaine was relatively short-lived, which was driven largely by a lack of cocaine available to sustain the 2001 levels [16]. Likewise, the increase in violent crime was also short-lived, thereby adding to the case for it being a causal factor in the increased rate of violent offences. It is unknown what the consequences would have been if the increased cocaine supply and use had persisted for a much longer period of time.
This natural experiment provided a unique opportunity to identify the effects of a sudden increase in cocaine use in a major Australian drug market, and to investigate previous findings regarding the role of cocaine use in violent criminal activity. Given the extensive attempts to eliminate other causes of the increase in violent crime and the existence of a partial control group, it seems reasonable to conclude that a transient increase in cocaine use among IDU in New South Wales produced a transient increase in violent crime.
Conclusion
Increases in cocaine use in NSW were accompanied by increases in violent crime as were observed in New York in the 1980s. However, these violent crimes did not involve the use of firearms, providing some supporting for the value of stringent gun control laws in reducing access to guns by criminals [32].
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
LD & CD conceived of the study, and participated in its design and coordination and helped to draft the manuscript. WH drafted the manuscript. EC carried out the data collection and drafted the manuscript. SG performed the statistical analysis. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This research was part of a larger project funded by the National Drug Law Enforcement Research Fund. The authors wish to thank Ms Linette Collins, Ms Amy Gibson, Dr Libby Topp and Professor Peter Reuter for their input to the project, and staff at NSW Police and Dr Don Weatherburn and colleagues at the Bureau of Crime Statistics and Research for providing data and assisting with interpretation.
==== Refs
Golub A Johnson B A recent decline in cocaine use among youthful arrestees in Manhatten, 1987 through 1993 American Journal of Public Health 1994 84 1250 1254 8059880
Agar M The story of crack: Towards a theory of illicit drug trends Addiction Research and Theory 2003 11 3 29 10.1080/1606635021000059042
Golub A Johnson B Crack's Decline: Some Surprises Across U.S. Cities National Institute of Justice Research in Brief 1997 Washington, DC, U.S. Department of Justice 16
Bowling B The rise and fall of New York murder: zero tolerance or crack's decline? Br J Criminol 1999 39 531 554 10.1093/bjc/39.4.531
Blumstein A Rivara F Rosenfeld R The rise and decline of homicide - and why Annual Reviews of Public Health 2000 21 505 541 10.1146/annurev.publhealth.21.1.505
Johnson B Golub A Dunlap E Blumstein A and Wallman J The rise and decline of hard drugs, drug markets, and violence in inner-city New York The crime drop in America 2000 New York, Cambridge University Press 164 206
McBride DC Swartz JA Drugs and violence in the age of crack cocaine Weisheit, Ralph A (Ed) (1990) Drugs, crime and the criminal justice system Academy of Criminal Justice Sciences monograph series (pp 141-169) v, 405pp Academy of Criminal Justice Sciences monograph series. 1990
George S Has the cocaine epidemic arrived in the UK? Forensic Science International 2004 143 187 190 15240042 10.1016/j.forsciint.2004.03.028
Gossop M Marsden J Stewart D Kidd T Changes in use of crack cocaine after drug misuse treatment: 4-5 year follow-up results from the National Treatment Outcome Research Study (NTORS). Drug & Alcohol Dependence 2002 66 21 28 11850132 10.1016/S0376-8716(01)00178-8
Home Office Tackling crack: A national plan 2004 London, Home Office
Goldstein P The drugs/violence nexus: A tripartite conceptual framework Journal of Drug Issues 1985 14 493 506
Darke S Kaye S Topp L Cocaine use in New South Wales, Australia, 1996-2000: 5 year monitoring of trends in price, purity, availability and use from the Illicit Drug Reporting System Drug and Alcohol Dependence 2002 67 81 88 12062781 10.1016/S0376-8716(02)00019-4
Hall W Hando J Burrows DFBAMAM Patterns of illicit psychostimulant use in Australia Psychostimulant Use in Australia 1993 Canberra, Australian Government Publishing Service
Hando J Flaherty B Rutter S An Australian profile on the use of cocaine. Addiction 1997 92 173 182 9158229 10.1046/j.1360-0443.1997.9221736.x
Degenhardt L Barker B Investigating trends in cocaine and methamphetamine mentions in accidental drug-induced deaths in Australia 1997-2002 2003 Sydney, National Drug and Alcohol Research Centre, University of New South Wales.
Breen C Degenhardt L Roxburgh A Bruno R Fetherston J Fischer J Jenkinson R Moon C Kinner S Ward J Weekley J Australian Drug Trends 2003: Findings from the Illicit Drug Reporting System (IDRS). NDARC Monograph No. 51 2004 Sydney, National Drug and Alcohol Research Centre, University of NSW
Homel P Flaherty B Reilly C Hall W Carless J The drug market position of cocaine among young adults in Sydney British Journal of Addiction 1990 85 891 897 2397316
Hall W Carless J Homel P Flaherty B Reilly C The characteristics of cocaine users among young adults in Sydney Medical Journal of Australia 1991 155 11 14 2067428
Mooseburger R Plant A Pierce J The drug use behaviour of cocaine users Drug and Alcohol Review 1990 9 251 258 16840147
Weatherburn D Jones C Freeman K Makkai T Supply control and harm reduction: Lessons from the Australian heroin 'drought' Addiction 2003 98 83 91 12492758 10.1046/j.1360-0443.2003.00248.x
Day C Topp L Rouen D Darke S Hall W Dolan K Changes in heroin availability in Sydney, Australia in early 2001 Addiction 2003 98 93 95 12492759 10.1046/j.1360-0443.2003.00266.x
Topp L Kaye S Bruno R Longo M Williams P O'Reilly B Fry C Rose G Darke S Australian Drug Trends 2001: Findings from the Illicit Drug Reporting System (IDRS). NDARC Monograph No. 48 2002 Sydney, National Drug and Alcohol Research Centre.
Day C Degenhardt L Hall W NSW heroin markets: Documenting the heroin shortage Drug Alcohol Rev
Roxburgh A Degenhardt L Breen C Changes in patterns of drug use among injecting drug users following a reduction in the availability of heroin in New South Wales, Australia Drug and Alcohol Review
Topp L Day C Degenhardt L Changes in patterns of drug injection concurrent with a sustained reduction in the availability of heroin in Australia Drug Alcohol Depend 2003 70 275 286 12757965 10.1016/S0376-8716(03)00013-9
Degenhardt L Conroy E Gilmour S Collins L The effect of a reduction in heroin supply in Australia upon drug distribution and acquisitive crime Brit J Criminol
Degenhardt L Day C Degenhardt L and Day C Changes in patterns of drug use The course and consequences of the heroin shortage in New South Wales 2004 Adelaide, Australasian Centre for Policing Research
Day C Gibson A Collins L Degenhardt L Dietze P Research methodologies used to obtain retrospective self-reports of the impact of changes in heroin supply on regular users in and out of treatment. NDARC Technical Report No. 176 2004 Sydney, National Drug and Alcohol Research Centre, University of New South Wales
Dietze P Miller P Clemens S Matthews S Gilmour S Collins L The course and consequences of the heroin shortage in Victoria. NDLERF Monograph No. 6 2004 Adelaide, Australasian Centre for Policing Research
Harrison A Christie P Longo M Pointer S Ali R The course and consequences of the heroin shortage in South Australia. NDLERF Monograph No. 5 2004 Adelaide, Australasian Centre for Policing Research
Fitzgerald J Briscoe S Weatherburn D Firearms and violent crime in New South Wales. BOCSAR Crime and Justice Bulletin No. 57 2001 Sydney, NSW Bureau of Crime Statistics and Research
Kaplan M Geling O Firearm suicides and homicides in the United States: Regional variations and patterns of gun ownership Social Science and Medicine 1998 46 1227 1233 9572612 10.1016/S0277-9536(97)10051-X
Breen C Degenhardt L Roxburgh A Bruno R Duquemin A Fetherston J Fischer J Jenkinson R Kinner S Longo M Rushforth C Australian Drug Trends 2002: Findings from the Illicit Drug Reporting System (IDRS). NDARC Monograph No. 50 2003 Sydney, National Drug and Alcohol Research Centre, University of NSW
Degenhardt L Day C Hall W The causes, course and consequences of the heroin shortage in Australia. NDLERF Monograph No. 3 2004 Adelaide, Australasian Centre for Policing Research
Degenhardt L Day C The course and consequences of the heroin shortage in New South Wales. NDLERF Monograph No. 4 2004 Adelaide, Australasian Centre for Policing Research
| 15840173 | PMC1112601 | CC BY | 2021-01-04 16:28:56 | no | BMC Public Health. 2005 Apr 19; 5:40 | utf-8 | BMC Public Health | 2,005 | 10.1186/1471-2458-5-40 | oa_comm |
==== Front
BMC Plant BiolBMC Plant Biology1471-2229BioMed Central London 1471-2229-5-61583678810.1186/1471-2229-5-6Research ArticleMolecular phylogeny and evolution of alcohol dehydrogenase (Adh) genes in legumes Fukuda Tatsuya [email protected] Jun [email protected] Toru [email protected] In-Ja [email protected] Takuro [email protected] Toshinori [email protected] Akira [email protected] Toshiaki [email protected] Masayuki [email protected] Graduate School of Life Sciences, Tohoku University, Sendai 980-8577, Japan2 Graduate School of Life Sciences, Tohoku University, Sendai 980-8578, Japan3 Graduate School of Science, Hiroshima University, Hiroshima 739-8526, Japan2005 18 4 2005 5 6 6 13 11 2004 18 4 2005 Copyright © 2005 Fukuda 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
Nuclear genes determine the vast range of phenotypes that are responsible for the adaptive abilities of organisms in nature. Nevertheless, the evolutionary processes that generate the structures and functions of nuclear genes are only now be coming understood. The aim of our study is to isolate the alcohol dehydrogenase (Adh) genes in two distantly related legumes, and use these sequences to examine the molecular evolutionary history of this nuclear gene.
Results
We isolated the expressed Adh genes from two species of legumes, Sophora flavescens Ait. and Wisteria floribunda DC., by a RT-PCR based approach and found a new Adh locus in addition to homologues of the Adh genes found previously in legumes. To examine the evolution of these genes, we compared the species and gene trees and found gene duplication of the Adh loci in the legumes occurred as an ancient event.
Conclusion
This is the first report revealing that some legume species have at least two Adh gene loci belonging to separate clades. Phylogenetic analyses suggest that these genes resulted from relatively ancient duplication events.
==== Body
Background
The alcohol dehydrogenase (Adh) genes encode a glycolytic enzyme and have been characterized at the molecular level in a wide range of flowering plants [1-3] as well as in Pinus banksiana, a conifer species [4]. The ADH enzyme is essential for anaerobic metabolism [5-7]. In both Arabidopsis thaliana and maize, oxygen stress and cold stress induces transcription from the Adh promoters; in addition, dehydration induces Adh transcription in A. thaliana [5-7]. Flowering plant species generally possess two or three isozymes [8], although A. thaliana has a single Adh locus [9].
The Adh genes in Arabidopsis thaliana [10], Arabidopsis gemmifera [3] and Leavenwortia [11] in Brassicaceae, cottons [2], and grasses [12-15] have been subjected to molecular evolutionary studies. However, the broader evolutionary histories of the Adh genes in the angiosperms remain unclear since few studies have investigated the evolution of the Adh genes in a wide range of angiosperms. Recently, Small and Wendel [2] suggested that some Adh gene duplications may have predated the origin of each of the flowering plant families. However, the details of the gene duplications and deletions experienced by the Adh genes of most groups of the angiosperms remain unclear. Additional studies are needed to understand the evolutionary history of the Adh genes in various plant groups.
In the legume family (Fabaceae), the Adh genes have only been investigated in crop species such as Glycine max and Pisum sativum. The purpose of these studies was to determine the ADH structures and functions rather than to explore the evolutionary processes of the Adh genes [e.g., [16,17]], although these studies suggested that these legume species contained only a single Adh gene locus [16,17]. Previous phylogenetic analyses of the Adh genes from various flowering plants have revealed that all of the Adh genes in legume plants characterised to date constitute a monophyletic group [1,2]. In contrast, the Adh genes in Rosaceae, a family that is closely related to the Fabaceae [18,19], appear in two separate lineages of the gene tree, suggesting that a gene duplication event had occurred before the Rosaceae evolved [2]. Although these observations hint that the legume family may actually bear other Adh gene copies, this has not yet been investigated. Consequently, it remains unclear whether Adh gene duplication occurred during the evolution of the legume family.
Here, we report the isolation of Adh genes from two quite disparate legume species. We found that both of these species contain another Adh gene locus in addition to the locus that was identified in legume species previously. We also investigated the molecular evolutionary history of the Adh genes in this family to gain further understanding of the evolutionary dynamics of nuclear gene families.
Results
Isolation of the Adh genes in legume plants
Two Adh sequences were isolated from each of the two legume species examined in this study. The Adh genes isolated from Sophora flavescens Ait. were denoted SfADH1 and SfADH2 while the isolates from Wisteria floribunda DC. were denoted as WfADH1 and WfADH2. For these genes, 708 bp were sequenced. As shown in Fig. 1, this resulted in a predicted amino acid sequence consisting of 236 residues. The sequences determined in this study have been submitted to the DDBJ / EMBL / GenBank nucleotide sequence databases (Table 1). At the amino acid level, the homology among the Adh genes in the legume plants ranged from 70.7% to 91.8%.
Phylogenetic analyses
We conducted phylogenetic analyses of the Adh genes using seven sequences from Pinus banksiana (Pinaceae) as outgroups [4]. To determine the phylogenetic position of the legume Adh genes isolated in this study, we subjected their sequences to ML analysis by employing a data set including the previously published Adh gene family sequences from various phylogenetic groups [e.g., [1,2]]. Our resulting Adh gene tree roughly consisted of two monophyletic groups that we denoted "Clade I" and "Clade II" (Fig. 2). Clade I contains only Adh genes from dicots, while Clade II contains Adh genes from both dicots and monocots. The legume Adh genes isolated in this study appeared in two separate clusters, one in Clade I and the other in Clade II (Fig. 2). For convenience, we call these clusters " Legume-clade I" and " Legume-clade II". Legume-clade I contained the SfADH1 and WfADH1 sequences as well as previously published Adh genes sequences from the legumes Glycine max, Pisum sativum, Phaseolus actifolius and Trifolium repens (Fig. 2). Legume-clade II consisted of only the SfADH2 and WfADH2 sequences and was located far from the other legume Adh sequences (Fig. 2). None of the other legume Adh sequences that have been published previously fell into Legume-clade II. However, the Adh gene in Pyrus communis (Rosaceae), which belongs to the family that is closely related to the Fabaceae [e.g. [18,19]], occurred at the sister position to Legume-clade II.
GeneTree analysis using the Adh gene sequences suggested that the legume Adh genes were duplicated before and after the angiosperms diversified (Fig. 3). This indicates that the Adh genes in Clade II have undergone more duplication events than those in Clade I (Fig. 3).
Discussion
Molecular phylogeny of the Adh sequences in legume plants
Although a previous study detected a monophyletic group of Adh genes in legumes [1], we found additional legume Adh genes that were related more distantly to the previously detected legume Adh genes. This is the first report showing that there are two Adh lineages in legume plants, each of which belongs to quite separate clades denoted as Legume-clade I and II, which themselves fall into distinct clades denoted as Clade I and II (Fig. 2). Notably, the Adh genes belonging to Legume-clade I are closely related to the Arabidopsis thaliana gene in Clade I (Fig. 2). Arabidopsis thaliana has a single Adh locus and transcription from its promoter increases under cold and oxygen stress [5-7]. Thus, the legume Adh genes in Legume-clade I may have similar functions to that of the A. thaliana gene. Our study also revealed that the legume Adh genes belonging to Legume-clade II form a sister group to the Adh gene isolated from Pyrus communis in Rosaceae (Fig. 2), which is a closely related family to the Fabaceae in the angiosperm phylogeny [20,21].
The function of the Adh gene in maize is also similar to that of A. thaliana [5-7]. Thus, our phylogenetic result suggests that function is the plesiomorphic character of the Adh gene family (Fig. 2). On the other hand, Clade II consists of many genes of both monocots and dicots, suggesting that the functions of the Adh genes in this clade may be more diversified due to the accumulation of many mutations during the course of angiosperm diversification that alter the primary structure of the ADH proteins. However, our phylogenetic analyses failed to indicate whether the genes in the Legume-clade II are orthologues or paralogues of the Adh gene in maize (Fig. 2). Thus, the function of the Adh genes in Legume-clade II remains unclear.
Gene duplication of Adh genes in legume plants
This study revealed the complicated evolution of the Adh gene family that occurred during the course of plant diversification. In our study, the phylogenic tree resulting from GeneTree analysis showed that some Adh genes in flowering plants evolved in complex manner that included several duplication events (Fig. 3). Duplication events in Adh genes have also been detected in other plant groups at various evolutionary levels. For example, Sang et al. [22] showed that diploid species of Paeonia (Paeoniaceae) had two or three Adh sequences and that repeated duplication or deletion events occurred after the diversification of this genus. Small and Wendel [2] analyzed Adh genes in Gossypium (Malvaceae) in great detail and found that these Adh sequences (denoted as GrADHA, GrADHB, GrADHC, GrADHD, and GrADHE) had experienced duplication events both before and after the divergence in Gossypium. Consistent with this, our GeneTree analysis revealed that in legumes, duplication of Adh genes occurred before the legume diverged, since the two quite distinct legumes Wisteria floribunda and Sophora flavescens have paralogous genes in each of two clades (Fig. 3), although all previously known Adh genes in legume plants such as Glycine max, Pisum sativum and Phaseolus actifolius belong only to Legume-clade I.
Why were additional Adh loci not found in other legumes? It is possible that the expression of the Legume-clade II Adh genes in Glycine max, Pisum sativum and Phaseolus actifolius Adh genes is limited to a specific developmental period or organ. Further analysis of Adh mRNA expression during various developmental phases and in different organs of these plants, such as roots, stems and fruits, may reveal the presence of an additional Adh gene in these species. Another possibility is that orthologues of the Legume-clade II Adh gene in the previously examined species have lost their function. Additional investigations throughout the legume family are needed to test this hypothesis.
Conclusion
Duplicated genes arise frequently in eukaryotic genomes through local events that generate tandem duplications, large-scale events that duplicate chromosomal regions or entire chromosomes, or genome-wide events that result in complete genome duplication [23]. Indeed, the existence of multigene families is evidence of the repeated gene duplication that has occurred over the history of life. One of the examples of the comprehensive analysis of gene duplication events in plants is the study of the MADS-box gene family. This gene family, which plays a central role in the morphogenesis of plant reproductive organs such as ovules and flowers, had experienced duplication events before the origin of angiosperms [24]. Moreover, some specific functions were gained through duplication events that took place after the diversification of flowering plants [24]. Thus, gene duplication has long been recognized as an important mechanism for the creation of new gene functions [25-27]. It is likely that each of the Adh genes in the legumes that were identified in the present study would have been subjected to different selective pressures over a long period. To determine whether this resulted in new functions, functional analyses of the legume Adh genes in each clade will have to be performed in the future.
Methods
Plant materials
In this study, we chose Sophora flavescens Ait. and Wisteria floribunda DC. from the legume family (Fabaceae). They belong to different subfamilies or tribes in the traditional classification [28]. They also fall into different phylogenetic groups in the phylogenetic tree constructed using legume rbcL sequences [29,30]. We also used tissues from Antirrhinum majus L. (Scrophulariaceae) and Trillium camtschatcense Ker-Gawl. (Trilliaceae). Flowers and leaf tissues were collected from the experimental garden of Tohoku University and native individuals of these species in the field. Vouchers for all species used in this study are listed in Table 2 and have been deposited in the Herbarium, Graduate School of Science, Tohoku University (TUS).
Isolation of RNA
Total mRNA was isolated according to the modified protocol of Hong et al. [31]. Thus, 3 g of flowers and leaf tissues were homogenized for 2 min with 3 volumes of detergent buffer containing 10 mM Tris-HCl (pH8.8), 50 mM NaCl, 6% (w/v) p-aminosalicylic acid, 2% (w/v) triisopropylnaphtalensulfonic acid, and 6% (v/v) 1-butanol. The homogenates were extracted three times with an equal volume of phenol/chloroform/isoamyl alcohol (25:24:1, v/v/v) with vigorous shaking. The final aqueous phase was collected and the total RNA was precipitated with ethanol and 3 M sodium acetate on ice for 1 hr. The total RNA was then treated with Oligotex-dT30 (TAKARA, Japan) to purify the poly(A) RNA.
Cloning and sequence analysis
Single-stranded cDNA was synthesized by priming with the random 9-mer or the oligo-dT adaptor primer (TAKARA). The cDNA was amplified by PCR in a 50 μL reaction volume containing approximately 50-ng total DNA, 10-mmol/L Tris-HCl buffer (pH 8.3) with 50-mmol/L KCl and 1.5-mmol/L MgCl2, 0.2-mmol/L of each dNTP, 1.25 units Taq DNA polymerase (TAKARA) and 0.5-μmol/L of each primer. The primers used have been published previously and are denoted as ADH-F1, ADH-R1 and ADH-R2 [22]. A degenerate primer was also used (LADH-1F1: 5'-ATATTTGGTCAYGAAGCTGG-3'). This primer was designed on the basis of the conserved region of Adh, which was determined by comparing the published sequences of Adh [22]. We carried out PCR with the following thermocycle protocol: (94°C, 2 min) × 1 cycle; (94°C; 30 sec, 50°C; 30 sec, 72°C; 120 sec) × 45 cycles; (72°C; 15 min) × 1 cycle. After the amplification, the reaction mixtures were subjected to electrophoresis in 1.5% low-melting-temperature agarose gels and the amplified products were purified. The purified PCR products were then cloned using the TA cloning kit (Invitrogen). Plasmids containing the cloned fragments were isolated by the alkali method and digested with EcoRI. Plasmids containing fragments less than 1.5 kb in size were selected and sequenced using the Thermo Sequence II dye terminator cycle sequencing premix kit (Amersham Pharmasia Biotech) or the BigDye Terminator cycle sequencing premix kit (Applied Biosystems) with the Model 373A or 310 automated sequencer (Applied Biosystems) according to the manufacturer's instructions.
Phylogenetic analysis
The sequences of the Adh genes used in this study were obtained from the GenBank/EMBL/DDBJ database (Table 1). The predicted amino acid sequences were aligned using CLUSTAL X [32] based on the GONNET protein weight matrix. The phylogenetic relationships between the genes were analyzed using the maximum-likelihood (ML) method. For the ML analyses, we used the PROTML program of PHYLIP version 3.6 [33]. We employed the JTT model of amino acid substitution. All indels were counted as missing. We performed ten random sequence addition searches using the J option and global branch swapping using the G option to isolate the ML tree with the best log-likelihood. In addition, we performed bootstrap analysis with 100 replications.
To infer the evolutionary events affecting the Adh genes, an analysis using GeneTree ver. 1.3 [34] was conducted, as described by Fukuda et al. [35]. The fully-resolved species tree used in the analysis was constructed on the basis of the previously published rbcL sequences in chloroplast DNA; the tree is considered to indicate the evolutionary relationships of the plants from which the Adh genes studied in this study were isolated [28]. The ML tree with the highest log-likelihood was used for the gene tree. Both gene duplications and losses were considered to reconcile the gene tree with the species tree. Gene lineages that do not coalesce on each branch of the species tree were counted as deep coalescence [36].
Authors' contributions
TF carried out the molecular genetic studies, IS and TN participated in the sequence alignment, and IT and TO drafted the manuscript. JY participated in the design of the study and performed the phylogenetic analysis. AK, TK and MM conceived of the study, participated in its design and coordination, and helped to draft the manuscript. All authors read and approved the final manuscript.
Acknowledgements
We thank Mr. K. Sato, Mr. H. Tokairin and Dr. M. Ohara for helping to provide and culture the plants. We are also grateful to Messers. H. Yamaji, N. Sasamoto, K. Yoshida, Y. Uyama, S. Matsumura, S. Horie, T. Yamashiro, K. Saito, M. Kitame, A. Shiro, P.-Y. Yun, Y. Mashiko, H. Ashizawa, M. Nakada, R. Shinohara, M. Komatsu, S.-Y. Kim and M. Hirai for help and advice. This work was supported, in part, by a Grant-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology, Japan and in part by the Sasakawa Scientific Research Grant from The Japan Science Society.
Figures and Tables
Figure 1 Alignment of the predicted amino acid sequences from selected Adh gene representatives. The lines indicate the groups that correspond to those in Figure 2.
Figure 2 The phylogenetic tree based on Adh gene sequences obtained by the maximum-likelihood method. The log-likelihood of the best ML tree is -3981.05. The numbers below the branches are the bootstrap values of 50% or more support. The Adh genes from legumes roughly fall into two monophyletic clades that we denoted as Clade I and Clade II.
Figure 3 Part of the simplest reconciled tree that has the lowest number of duplication and deletion events. The reconciled tree involves 24 gene duplications and 44 gene losses for a total cost of 68, and requires 40 deep coalescenses. The solid boxes indicate gene duplications that were inferred on the basis of mismatches between the gene tree and the species tree. The open boxes indicate those duplications that required multiple copies of Adh genes within the same species. The gray lines indicate the lineages that are presumed to be lost after the duplications or were not found in our analysis.
Table 1 List of accession numbers used in this study. List of taxa and source of this study.
Phylogenetic group
Family
Species ADH locus Acc. No. References
Gymnosperms
Pinaceae
Pinus banksiana PbADHC1 U48366 Perry and Furnier 1996
PbADHC2 U48373 Perry and Furnier 1996
PbADHC3 U48368 Perry and Furnier 1996
PbADHC4 U48369 Perry and Furnier 1996
PbADHC5 U48370 Perry and Furnier 1996
PbADHC6 U48371 Perry and Furnier 1996
PbADHC7 U48372 Perry and Furnier 1996
Angiosperms
Dicotyledon
Rosid
Paeoniaceae
Paeonia humulis PhADH1 AF126226 Sang and Zhang 1999
PhADH2 AF126232 Sang and Zhang 1999
Paeonia lutea PaelADH1 AF009042 Sang et al. 1997
PaelADH2 AF009057 Sang et al. 1997
Brassicaceae
Arabidopsis thaliana AtADH D63464 Miyashita et al. 1996
Arabidopsis lyrata AlADH AJ251284 Savolainen et al. 2000
Arabis hirsuta AhADH AB015502 Miyashita et al. 1998
Rosaceae
Fragaria x ananassa FaADH X15588 Wolyn and Jelenkovic 1990
Pyrus communis PcADH3 AF031899 Chervin et al. 1999
PcADH4 AF031900 Chervin et al. 1999
Fabaceae
Glycine max GmADH1 AF079058 Preiszner et al. unpubl.
GmADH2 AF079499 Preiszner et al. unpubl
Phaseolus acutifolius PaADH Z23171 Garvin et al. 1994
Pisum sativum PisaADH X06281 Llewellyn et al. 1987
Sophora flavescens SfADH1 AB191335 This study
SfADH2 AB191336 This study
Wisteria floribunda WfADH1 AB191337 This study
WfADH2 AB191338 This study
Malvaceae
Gossypium raimondii GrADHA AF182116 Small and Wendel 2000
GrADHB AF226635 Small and Wendel 2000
GrADHC AF036568 Small and Wendel 2000
GrADHD AF250203 Small and Wendel 2000
GrADHE AF250208 Small and Wendel 2000
Asterid
Scrophulariaceae
Antirrhinum majus AmADH AB191334 This study
Solanaceae
Solanum tuberosum StADH1 M25154 Matton et al. unpubl.
StADH2 M25153 Matton et al. unpubl.
StADH3 M25152 Matton et al. unpubl.
Asteraceae
Lactuca sativa LaADH D44449 Toyomasu et al. 1995
Monocotyledon
Trilliaceae
Trillium camtschatcense TrcADH AB191339 This study
Arecaceae
Washingtonia robusta WrADHA U65973 Morton et al. 1996
WrADHB U65972 Morton et al. 1996
Poaceae
Hordeum vulgare HvADH1 X07774 Good et al. 1988
HvADH2 X12733 Trick et al. 1988
Oryza sativa OsADH1 X16296 Xie and Wu 1989
OsADH2 X16297 Xie and Wu 1989
Pennisetum glaucum PgADH X16547 Bui et al. 1990
Zea mays ZmADH1 M32984 Osterman and Dennis 1989
ZmADH2 X02915 Dennis et al. 1985
Table 2 List of taxa from which Adh was isolated in this study and source of plant materials. List of taxa that Adh was isolated in this study and source of plant materials.
Taxa Family Locality Collecters
Sophora flavescens Fabaceae Japan: Miyagi Pref., Sendai Fukuda 99081
Wisteria floribunda Fabaceae Japan: Miyagi Pref., Shiroishi Fukuda and Yoshida 99051
Antirrhinum majus Scrophulariaceae cultivated in Tohoku University Fukuda 98031
Trillium camtschatcense Trilliaceae Japan: Hokkaido Pref., Hidaka, Shizunai Fukuda and Yokoyama 0404301
==== Refs
Clegg MT Cummings MP Durbin ML The evolution of plant nuclear genes Proc Natl Acad Sci USA 1997 94 7791 7798 9223265 10.1073/pnas.94.15.7791
Small RL Wendel JF Copy number lability and evolutionary dynamics of the Adh gene family in diploid and tetraploid cotton (Gossypium) Genetics 2000 155 1913 1926 10924485
Miyashita NT DNA variation in the 5' upstream region of the Adh locus of the wild plants Arabidopsis thaliana and Arabis gemmifera Mol Biol Evol 2001 18 164 171 11158375
Perry DJ Furnier GR Pinus banksiana has at least seven expressed alcohol dehydrogenase genes in two linked groups Proc Natl Acad Sci USA 1996 93 13020 13023 8917537 10.1073/pnas.93.23.13020
Dolferus R Bruxelles GD Dennis ES Peacock WJ Regulation of the Arabidopsis Adh gene by anaerobic and other environmental stress Ann Bot 1994 74 301 308 10.1006/anbo.1994.1121
Dolferus R Jacob M Peacock WJ Dennis ES Differential interactions of promoter elements in stress responses of the Arabidopsis Adh gene Plant Physiol 1994 105 1075 1087 7972489 10.1104/pp.105.4.1075
Dolferus R Osterman JC Peacock WJ Dennis ES Cloning of the Arabidopsis and rice formaldehyde dehydrogenase genes: implications for the origin of plant ADH enzymes Genetics 1997 146 1131 1141 9215914
Gottlieb LD Conservation and duplication of isozymes in plants Science 1982 216 373 380
Chang G Meyerowitz EM Molecular cloning and DNA sequence of the Arabidopsis thaliana alcohol dehydrogenase gene Proc Natl Acad Sci USA 1986 83 1408 1412 2937058
Innan H Tajima F Terauchi R Miyashita NT Intragenic recombination in the Adh locus of the wild plant Arabidopsis thaliana Genetics 1996 143 1761 1770 8844162
Charlesworth D Liu F Zhang L The evolution of the alcohol dehydrogenase gene family in plants of the genus Leavenworthia (Brassicaceae): loss of introns, and an intronless gene Mol Biol Evol 1998 15 552 559 9580984
Gaut BS Morton BR McCaig BC Clegg MT Substitution rate comparisons between grasses and palms: synonymous rate differences at the nuclear gene Adh parallel rate differences at the plastid gene rbcL Proc Natl Acad Sci USA 1996 93 10274 10279 8816790 10.1073/pnas.93.19.10274
Morton BR Gaut BS Clegg MT Evolution of alcohol dehydrogenase genes in the Palm and Grass family Proc Natl Acad Sci USA 1996 93 11735 11739 8876206 10.1073/pnas.93.21.11735
Cummings MP Clegg MT Nucleotide sequence diversity at the alcohol dehydrogenase 1 locus in wild barley (Hordeum vulgare ssp. spontaneum): an evaluation of the background selection hypothesis Proc Natl Acad Sci USA 1998 95 5637 3642 9576936 10.1073/pnas.95.10.5637
Gaut BS Peek AS Morton BR Clegg MT Patterns of generic diversification within the Adh gene family in the grasses (Poaceae) Mol Biol Evol 1999 16 1086 1097 10474904
Llewellyn DJ Finnegan EJ Ellis JG Dennis ES Peacock WJ Structure and expression of an alcohol dehydrogenase 1 gene from Pisum sativum J Mol Biol 1987 195 115 123 3309331
Garvin DF Weeden NF Doyle JJ The reduced stability of a plant alcohol dehydrogenase is due to the substitution of serine for a highly conserved phenylalanine residue Plant Mol Biol 1994 26 643 655 7948919 10.1007/BF00013750
Chase MW Soltis DE Olmstead RG Morgan D Les DH Mishler BD Duvall MR Price RA Hills HG Qiu YL Kron KA Rettig JH Conti E Palmer JD Manhart JR Sytsma KJ Michaeks HJ Kress WJ Karol KG Clark WD Hedroen M Gaut BS Jansen RK Kim KJ Wimpee CF Smith JF Furnier GR Strauss SH Xiang QY Plunkett GM Soltis PS Swensen SM Williams SE gadek PA Quinn CJ Eguiarte LE Golenberg E Learn GM JrGraham SW Barrett SC Dayanandan S Albert VA Phylogenetics of seed plants: an analysis of nucleotide sequences from the plastid gene rbcL Ann Missouri Bot Gard 1993 80 528 580
APG (The Angiosperm Phylogeny Group) An ordinal classification for the families of the flowering plants Ann Missouri Bot Gard 1998 85 531 533
Mathews S Donoghue DJ The root of angiosperm phylogeny inferred from duplicate phytochrome genes Science 1999 286 947 950 10542147 10.1126/science.286.5441.947
Qiu YL Lee J Bernasconi-Quadroni F Soltis DE Soltis PS Zain M Zimmer EA Chen Z Savolainen V Chase MW The earliest angiosperms: evidence from mitochondrial, plastid and nuclear genomes Nature 1999 402 404 407 10586879 10.1038/46536
Sang T Donoghue MJ Zhang D Evolution of alcohol dehydrogenase genes in paeonies (Paeonia): phylogenetic relationships of putative nonhybrid species Mol Biol Evol 1997 14 994 1007 9335140
Dujon B Sherman D Fischer G Durrens P Casaregola S Lafontaine I De Montigny J Marck C Neuveglise C Talla E Goffard N Frangeul L Aigle M Anthouard V Babour A Barbe V Barnay S Blanchin S Beckerich JM Beyne E Bleykasten C Boisrame A Boyer J Cattolico L Confanioleri F De Daruvar A Despons L Fabre E Fairhead C Ferry-Dumazet H Groppi A Hantraye F Hennequin C Jauniaux N Joyet P Kachouri R Kerrest A Koszul R Lemaire M Lesur I Ma L Muller H Nicaud JM Nikolski M Oztas S Ozier-Kalogeropoulos O Pellenz S Potier S Richard GF Straub ML Suleau A Swennen D Tekaia F Wesolowski-Louvel M Westhof E Wirth B Zeniou-Meyer M Zivanovic I Bolotin-Fukuhara M Thierry A Bouchier C Caudron B Scarpelli C Gaillardin C Weissenbach J Wincker P Souciet JL Genome evolution in yeasts Nature 2004 430 35 44 15229592 10.1038/nature02579
Theissen G Becker A Di Rosa A Kanno A Kim JT Münster T Winter KU Saedler H A short history of MADS-box genes in plants Plant Mol Biol 2000 42 115 149 10688133 10.1023/A:1006332105728
Wagner A The fate of duplicated genes: loss or diversification? BioEssays 1998 20 785 788 10200118 10.1002/(SICI)1521-1878(199810)20:10<785::AID-BIES2>3.0.CO;2-M
Lynch M Conery JS The evolutionary fate and consequences of duplicate genes Science 2000 290 1151 1155 11073452 10.1126/science.290.5494.1151
Wagner A Birth and death of duplicated genes in completely sequenced eukaryotes Trends Genet 2001 17 237 239 11335019 10.1016/S0168-9525(01)02243-0
Polhill RM Bisby FA, Buckingham J, Harborne JB Classification of the Leguminosae Phytochemical dictionary of Leguminosae New York, Chapman and Hall 35 57
Doyle JJ Doyle JL Ballenger JA Dickson EE Kajita T Ohashi H A phylogeny of the chloroplast gene rbcL in the Leguminosae: Taxonomic correlation and insights into the evolution of nodulation Am J Bot 1997 84 541 554
Kajita T Ohashi H Tateishi Y Bailey CD Doyle JJ RbcL and legume phylogeny, with particular reference to Phaseoleae, Millettieae, and allies Syst Bot 2001 26 515 536
Hong JC Nagao RT Key JL Characterization and sequence analysis of a developmentally regulated putative cell wall protein gene isolated from soybeen J Biol Chem 1987 262 8367 8376 3036811
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 24 4876 4882 10.1093/nar/25.24.4876
Felsenstein J PHYLIP: Phylogeny Inference Package, version 3.6 (alpha) University of Washington, Seattle, USA 2000
Page RDM Cotton JC Sankoff D, Nadeau J GeneTree: a tool for exploring gene family evolution Comparative genomics: empirical and analytical approaches to gene order dynamics, map alignment, and the evolution of gene families Dordrecht, Kluwer Academic Publishers 525 536
Fukuda T Yokoyama J Maki M Molecular evolution of cycloidea-like genes in Fabaceae J Mol Evol 2003 57 588 59 14738317 10.1007/s00239-003-2498-2
Maddison WP Gene trees in species trees Syst Biol 1997 46 523 536
| 15836788 | PMC1112602 | CC BY | 2021-01-04 16:03:52 | no | BMC Plant Biol. 2005 Apr 18; 5:6 | utf-8 | BMC Plant Biol | 2,005 | 10.1186/1471-2229-5-6 | oa_comm |
==== Front
BMC SurgBMC Surgery1471-2482BioMed Central London 1471-2482-5-81583110410.1186/1471-2482-5-8Research ArticleEvaluation of POSSUM scoring system in patients with gastric cancer undergoing D2-gastrectomy Bollschweiler Elfriede [email protected] Thomas [email protected] Stefan P [email protected] Arnulf H [email protected] Department of Visceral- and Vascular Surgery University of Cologne, Joseph-Stelzmann Str. 9, 50931 Köln, Germany2005 15 4 2005 5 8 8 30 8 2004 15 4 2005 Copyright © 2005 Bollschweiler et al; licensee BioMed Central Ltd.2005Bollschweiler 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
Risk adjustment and stratification play an important role in quality assurance and in clinical research. The Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity (POSSUM) is a patient risk prediction model based on 12 patient characteristics and 6 characteristics of the surgery performed. However, because the POSSUM was developed for quality assessment in general surgical units, its performance within specific subgroups still requires evaluation.
The aim of the present study was to assess the accuracy of POSSUM in predicting mortality and morbidity in patients with gastric cancer undergoing D2-gastrectomy.
Methods
137 patients with gastric cancer undergoing gastrectomy were included in this study. Detailed, standardized risk assessments and thorough documentation of the post-operative courses were performed prospectively, and the POSSUM scores were then calculated.
Results
The 30- and 90- day mortality rates were 3.6% (n = 5) and 5.8% (n = 8), respectively. 65.7% (n = 90) of patients had normal postoperative courses without major complications, 14.6% (n = 20) had moderate and 13.9% (n = 19) had severe complications. The number of mortalities predicted by the POSSUM-Mortality Risk Score (R1) was double the actual number of mortalities occurring in the median and high-risk groups, and was more than eight times the actual number of mortalities occurring in the low-risk group (R1 < 20%). However, the calculated R1 predicted rather well in terms of severe morbidity or post-operative death in each risk group: in predicted low risk patients the actual occurrence rate (AR) of severe morbidity or post-operative death was 14%, for predicted medium risk patients the AR was 23%, and for predicted high risk patients the AR was 50% (p < 0.05). The POSSUM-Morbidity Risk Score (R2) overestimated the risk of morbidity.
Conclusion
The POSSUM Score may be beneficial and can be used for assessment of the peri- and post-operative courses of patients with gastric carcinoma undergoing D2-gastrectomy. However, none of the scores examined here are useful for preoperative prediction of postoperative course.
==== Body
Background
Risk adjustment and stratification play an important role in quality assurance and are indispensable tools used in clinical research. The Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity (POSSUM) is a patient risk prediction model based on 12 patient characteristics and 6 characteristics of the surgery performed [1]. However, because the POSSUM model was developed for quality assessment purposes in general surgical units, in order to implement it for specific subgroups of patients, its performance within such subgroups needs to be evaluated. The results of two prospective studies [2,3] showed higher mortality and morbidity rates after D2-lymphadenectomy (LAD) than after D1-lymphadenectomy for patients with gastric cancer, although there is apparently no difference in the long-term prognoses for patients after the two procedures [4,5]. However, for certain patient subgroups, radical lymph node (LN) dissection does improve prognosis. For this reason, it would be greatly beneficial to calculate the risks of morbidity and mortality for each patient preoperatively. In addition, for prognostic studies it would be useful to be able to stratify patients according to their risk factors.
In some countries, the American Society of Anesthesiologists' (ASA) classification is widely used to provide quality assurance of surgical procedures [6,7]. The ASA-score is easy to use, but the classification is not precise [8], and it does not consider the severity of surgical insult. POSSUM has been used to make comparisons between different vascular [9,10] and colorectal [11] surgical units, and to compare individual surgeons' performance within a single unit [12,13].
The aim of the present study was to assess the accuracy of the POSSUM instrument to predict mortality and morbidity in patients with gastric cancer undergoing D2-gastrectomy.
Methods
Patients
All patients with gastric cancer undergoing total gastrectomy (n = 123) or subtotal gastrectomy (n = 16) at the Department of Surgery, University of Cologne between January 1, 1997 and December 31, 2001 were included in this study. Preoperatively, all patients underwent esophago-gastro-duodenoscopy with biopsies and histopathologic examination. In addition, endosonography of the stomach was performed to stage the depth of tumour infiltration (T-category), and CT-scans were done to look for evidence of metastases.
Surgical procedure and extent of lymphadenectomy
In all cases an en bloc resection of the stomach with extended D2-lymphadenectomy was performed. The lymph node dissection included compartments I and II.
Compartment I comprises all lymph node groups along the lesser curvature (No.s 1,3 and 5) and the greater curvature (No.s 2, 4, and 6) of the stomach. Compartment II comprises lymph node stations 7 to 12 according to the General Rules for the Gastric Cancer Study in Surgery and Pathology [14]. Type II (cardia) and type III (subcardial) adenocarcinomas of the gastroesophageal junction (using the Siewert/Hölscher classification system) [15] were treated with a trans-hiatal extended gastrectomy including D2-lymphadenectomy and lymph node dissection of the lower mediastinum. In the cases of subtotal gastrectomy (n = 16), only lymph node stations 3 to 6 (compartment I) and lymph node stations 7 to 12 (compartment II) were resected en bloc. Sampling of compartments III and IV nodes (No. 13 to 16) was optional. The surgeon divided the en bloc resected tissue containing lymph nodes into separate stations and assigned numbers to these stations according to the Japanese classification system [14]. Splenectomy was performed (n = 35) in cases of proximal gastric carcinoma (types II and III) and in cases of metastatic infiltration of the splenic hilar nodes (No. 10), but not as a general rule [16].
Risk assessment
The selection of patients for surgery was based on the surgeons' "end of the bed assessment" backed by a detailed risk analysis described elsewhere [17,18]. This risk analysis has been evaluated in a prospective study. All data were routinely available. Ninety-two per cent of the operations were performed by four surgeons specializing in upper gastrointestinal surgery. Two patients were excluded from the study owing to incomplete data despite extensive tracking of case notes. The remaining 137 patients were scored retrospectively using the POSSUM-score, and the predicted risk of morbidity and death was calculated for each patient according to the following previously described logistic regression equations [1]: log e [R1/(1 - R1)] = - 7.04 + (0.13 × physiological score) + (0.16 × operative severity score) where R1 = risk of death, and log e [R2/(1 - R2)] = - 5.91 + (0.16 × physiological score) + (0.19 × operative severity score) where R2 = risk of morbidity.
Because the equations for R1 and R2 require information about the operative insult severity, and this data was not available preoperatively, we also calculated the physiological score (PPS) and in addition the V-POSSUM [19], which uses only the physiological score: log e [R3/(1 - R3)] = - 6.0386 + (0.1539 × physiological score), where R3 = risk of death.
The postoperative course was defined as (corresponds to McPeek Index 4 – 6 [20]):
1 = normal course of disease: Patient had no significant surgical or general postoperative complications.
2 = moderately favorable course of disease: Patient had postoperative complications, but the complications were treatable with appropriate therapy.
3 = poor course of disease: Patient had multiple complications that were difficult to treat with any kind of therapy.
4 = Died as a consequence of surgery (90-day mortality).
Definition of postoperative morbidity
Pulmonary complications: Emphysema, pneumothorax, acute pneumonia, aspiration.
Cardiac complications: Cardiovascular collapse, cardiac decompensation, bradycardia, myocardial infarction, hypertensive or hypotensive cardiovascular crisis.
Cerebral complications: Cerebral infarction, cerebral edema, organic brain syndrome.
Renal complications: Renal failure, renal bleeding, urinary tract infection.
In computing post-operative mortality, deaths occurring in and outside of the hospital were not differentiated [21]. Complications were documented using a detailed questionnaire. The severity of the post-operative course was evaluated by the treating physician while the patient was undergoing intensive care. Evaluations were based on the overall clinical impression and did not necessarily depend on the precise number of complications.
The observed and predicted operative mortality rates were compared using frequency tables. Model performance was evaluated with the Hosmer-Lemeshow 2 statistic (HL), which is a measure of calibration or goodness of fit [22]. Calibration refers to the ability of the model to assign correct outcome probabilities to patients, i.e. whether the model-estimated probability of mortality for patients with particular risk factors agrees with the actual observed mortality rate. To obtain this statistic, the estimated probability of death for each patient was computed based on the model and then stratified into different groups. The numbers of predicted and observed outcomes for each group were then evaluated statistically. Higher values of the HL statistic represent poorer model calibration.
Statistical analysis was two-sided using a significance level of 5 per cent. The Chi-Square test was calculated using the Yates correction. All calculations were performed using the computer software package SPSS © version 11 for Windows (SPSS, Chicago, Illinois, USA).
Graphical presentation of results was done with SigmaPlot Version 8.0.
Results
The epidemiologic data of the 137 patients are shown in table 1. The mean number of resected lymph nodes was 37.7. The number of metastatic lymph nodes was an average of 7.6 (min: 0, max: 48). The median postoperative stay in the Intensive Care Unit was 1 day (min 1, max 30), and the median postoperative stay in the hospital was 17 days (min: 8 max 103).
Table 1 Clinico-patho logic data of 137 patients with gastric cancer and gastrectomy with D2-lym phadenectomy.
median age 65 y range (38 – 85 y)
gender: m:f 3:2
pT-categ ory n = 137 %
pT1 28 20, 4
pT2 46 33, 6
pT3 52 38, 0
pT4 11 8,0
pN-categ ory n = 137 %
pN0 47 34, 3
pN+ 90 65, 7
pM0 n = 108 78, 8 %
UICC-stage n = 137 %
Stage IA 20 14, 6
Stage IB 23 16, 8
Stage II 27 19, 6
Stage IIIA 15 11, 0
Stage IIIB 9 6, 6
Stage IV 43 31, 4
Location of tumour n = 137 %
Upper third 67 48, 9
Middle third 29 21, 2
Distal third 41 29, 9
Postoperative course
The 30-day mortality rate was 3.6% (n = 5) and the 90-day mortality rate was 5.8% (n = 8). 65.7% (n = 90) of patients had normal postoperative courses without major complications. 14.6% (n = 20) had a medium postoperative course and 13.9% (n = 19) experienced severe complications during the postoperative period. The list of surgical and systemic complications is shown in table 2.
Table 2 List of surgical and systemic complications in 137 patients with gastrectomy and D2-lymphadenectomy. Multiple complications are possible.
total Postoperative Course
1 2 3 4
Surgical complications n % n n n n
Relaparotomy 5 3.6 - 1 2 2
Perforation 2 1.5 - 1 - 1
Anastomic leakage 7 5.1 1 3 2 1
secondary 1 0.7 - 1 - -
haemorrhage
Ileus 2 1.4 - - 1 1
Pancreatitis 5 3.6 - - 3 2
Peritonitis 5 3.6 - 1 1 3
Wound infection 8 5.8 2 1 2 3
Abscess 3 2.2 - 1 1 1
Pancreatic fistula 13 9.4 1 8 4 -
Catheter infection 5 3.6 - 2 3 -
Systemic complications
Reanimation 6 4.4 - - 2 4
Sepsis 7 5.1 - 1 2 4
pulmonary compl. 22 16.0 2 7 9 4
cardiac compl. 23 16.8 2 4 10 7
renal compl. 5 3.6 1 1 1 2
cerebral compl. 6 4.4 1 2 2 1
Other compl. 10 7.3 6 2 2 0
POSSUM-Score
Calculation of the POSSUM-Mortality Risk (R1) is shown in table 3. The number of mortalities predicted by the calculated R1 value was double the actual number of mortalities occurring in the median and high-risk groups, and was more than eight times the actual number of mortalities occurring in the low-risk group (R1 < 20%). The calculated R1 was much better as an estimate of severe morbidity or post-operative death, however, with predicted values matching the number of observed cases. For patients in the low-risk group (predicted risk 0 – 20%), the actual rate of severe morbidity or post-operative death was 14%; for patients in the median risk group (predicted risk 21 – 40%), the actual rate observed was 23%; and for patients in the high-risk group (predicted risk > 40%), the actual rate of severe morbidity or post-operative death was 50%. The observed rates of morbidity and mortality differed significantly between the three groups (p < 0.05).
Table 3 Results of POSSUM Mortality calculation (R1) compared to observed mortality (outcome 4) and to the rate of severe morbidity and death (outcome 3 +4) in 137 patients with gastrectomy and D2-lymphadenectomy.
Possum Mortality Risk (R1) patients expected death observed death outcome = 4 Observed morbidity and death outcome = 3+4
(%) n n n % n %
0–10 21 1 0 0 3 14,3
38675 51 7 1 2,0 7 13,7
low risk 0 – 20 72 1 1, 0 10 13, 9
21–30 38 9 3 7,8 9 23,7
31–40 19 6 2 10,5 4 21,1
med risk 21–40 57 5 8, 8 13 22, 8
41–60 4 1 1 25,0 1 25,0
71–100 4 2 1 25,0 3 75,0
high risk > 40 8 2 25, 0 4 50, 0
total 137 29 8 27
The POSSUM Morbidity equation (R2) predicted nearly twice as many cases of mild or severe morbidity (including death) than were actually observed (table 4). Only for patients with a very low risk (R2 < 40%) or very high-risk (R2 > 90%), predictions were good. However, for patients with a calculated R2 less than or equal to 60% (low-risk group), 19.1% (9 of 47) actually developed complications or died after D2-resection, and for patients with higher calculated R2 values >60% (high-risk group), 42.2% (38 of 90) actually did so. The observed morbidity and mortality (for outcomes 2–4 defined above) differed significantly between these low and high-risk groups (p < 0.01). There was no significant difference measured for outcomes 3 and 4, but the test lacked sufficient power.
Table 4 Results of POSSUM morbidity calculation (R2) compared to morbidity rate (outcome 2, 3 and 4) of 137 patients with gastrectomy and D2-lymphadenectomy.
POSSUM Risk of Morbidity R2 Patients Expected morbidity observed severe morbidity and mortality Outcome (3 + 4) observed morbidity and mortality Outcome (2 – 4)
% n n n % n %
0 – 30 2 0 0 0 0 0
31 – 40 12 3,9 3 25,0 4 25,0
41 – 50 14 6,6 1 7,1 2 14,3
50 – 60 19 10,6 2 10,5 3 15,8
low risk 0 – 60 47 6 12, 8 9 19, 1
61 – 70 21 14,9 2 9,5 8 38,1
71 – 80 35 25,7 7 20,0 14 40,0
81 – 90 28 22,4 7 25,0 10 35,7
91–100 6 5,7 5 83,3 6 100,0
high risk > 60 90 21 23, 3 38 43, 3
Total 137 27 47
However, for cases of pre-operative predictions of post-operative course using the POSSUM score, only physiologic criteria and not operative data are available for calculations. Therefore, we used the POSSUM Physiological Score (PPS) for outcome prediction. To assess the predictive value of the PPS, we used logistic regression analysis. Worse outcomes appeared to occur more frequently in patients with higher PPS scores, however this correlation was not statistically significant. The correlation between patient age, PPS-score, and mortality is shown in figure 1. There was no significant correlation between predicted risk and actual post-operative course using the calculated V-POSSUM score (data is not shown).
Figure 1 Correlation between age of the patients, POSSUM Physiological Score (PPS) and the postoperative course of 137 patients with gastrectomy and D2-lymphadenectomy: M = case with postoperative mortality, size of circles shows the number of cases with the same PPS.
Discussion
For patients with gastric cancer, especially in advanced stages, extensive lymphadenectomy (LAD) can improve prognosis. The results of the Dutch prospective randomized study show comparable long-term outcomes for patients undergoing either D1- or D2-lymphadenectomy [23]. However, this study also demonstrated an increased rate of morbidity and mortality for patients undergoing D2-LAD versus D1-LAD [4]. Therefore, it is important to know pre-operatively which patients are more likely to benefit from the more radical operation. Furthermore, in order to compare study results for outcomes, it is necessary to stratify the investigated patients according to risk profiles.
Patients with gastric carcinoma are usually older than 60 years of age and have corresponding concomitant medical problems, which may significantly influence the post-operative course. For example in our study, 6 of the 8 post-operative mortalities occurred in patients between 70 and 80 years of age. The risks for such older patients could be only partially evaluated using the POSSUM Physiologic Score (PPS). The lack of significant correlation between higher PPS scores and higher risk for older patients may be due to the small number of participants in this study.
There are various established assessment systems designed to assess the gravity of pre-existing illnesses. The ASA-Classification Score, for example, is used most often for this purpose in surgical and anesthesia settings [6,7]. This score was developed by anesthesiologists to consider the risks of anesthetic procedure. It functions well as a fast assessment of patients, and should ascertain whether a life-threatening condition exists or whether peri-operative problems can be expected. However, the ASA-score is less suitable to determine whether a patient will develop serious complications as a result of the magnitude of the operation performed [8]. In our study, half of the patients were classified as ASA II, and half as ASA III. This risk stratification correlated only weakly with the patients' actual post-operative course.
The POSSUM Score has been evaluated in numerous studies [1,10-13,24]. The main objective in these studies was to ascertain whether this score is suitable to evaluate the Case-Mix [1,10,13,19]. It has also been used to assess surgeon-dependent risk factors [12,13]. A number of studies have evaluated the applicability of the POSSUM Score for particular medical conditions, and therefore a number of varieties have emerged, i.e. the P-Possum-Score [25] or the V-POSSUM-Score [19]. As shown in the foregoing results, the POSSUM Mortality Index (R1) calculated the probability of mortality for patients with gastric carcinoma undergoing D2-LAD at two to three times the actual rates occurring in our patient population. Although this index was developed to predict the 30-day mortality rate, we applied it to the 90-day mortality rate as well. Because of the low mortality rate in our patient cohort, the difference (3.6% – 5.1%) was irrelevant. An overestimate of mortality risk using this index has also been found in other studies, i.e. for patients with esophageal carcinoma, where a high post-operative mortality is expected [26,27].
Despite these discouraging results, there was a significant correlation of the predicted risk of severe morbidity (including mortality) with the actual incidence using the R1 index. This correlation was evident in all three of the classified risk levels, where the observed outcomes agreed completely with the predicted risk for severe morbidity. In light of these results, the R1 classification appears to be suitable for risk stratification purposes in clinical studies.
The calculated POSSUM-Morbidity Index (R2) overestimated the risk of developing post-operative complications in this study. There is an acceptable correlation between predicted values and observed rates of morbidity when using the R2 to evaluate both very low risk patients and very high-risk patients. In the other categories, however, the risk was overestimated two to three-fold.
The POSSUM Score is used to predict the post-operative course of patients, using both the pre-operative assessment of the severity of pre-existing concomitant medical conditions as well as information gathered during the peri-operative period i.e. severity of surgical insult, intra-operative blood loss, etc. However, peri-operative information was not available pre-operatively; the decision regarding magnitude of gastrectomy had not yet been made. Therefore, our investigation focused on whether application of the POSSUM Physiological Score would be sufficient to make an accurate pre-operative prediction. The V-POSSUM Score, which was previously evaluated primarily for the assessment of patients with vascular disease [24], was ineffective for predicting post-operative course in our patient cohort. Using this score we showed, admittedly, that patients with high PPS scores also had higher risks of complications or post-operative mortality, but the correlation was not significant enough to give a valid pre-operative risk assessment.
There are still more instruments that may be used for risk assessment of patients with concomitant medical problems. For example, the Charlson Comorbidity Index (CCI) was developed particularly to address this issue [28,29]. This assessment focuses primarily on long-term outcomes. However, some studies have shown that the CCI is not suitable for the prediction of post-operative course [8]. Another assessment instrument, the APACHE II Score, was developed for patients in the Intensive Care Unit to predict the courses of patients there [30,31]. Unfortunately, the APACHE Score is irrelevant in our patient cohort, where some of the assessed parameters do not exist pre-operatively, or change post-operatively.
Conclusion
This study shows that the POSSUM Mortality Score (R1) is a suitable instrument to risk stratify patients with gastric carcinoma undergoing D2-LAD for the development of severe post-operative complications (including post-operative mortality), based on pre-existing or concomitant medical problems. The POSSUM Morbidity Index (R2) is particularly suitable for risk assessment if the target parameters include moderate to severe complications. When using this instrument, however, the overestimation of risk must be considered. For our purposes, none of the instruments (i.e. PPS, V-POSSUM) examined for pre-operative risk assessment were effective models. Finally, in our study we showed that accurate documentation of standardized risk scores is possible under routine conditions and that the necessary parameters for a second score like the POSSUM Score could be formulated.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
EB conceived the study, performed the statistical analysis and drafted the manuscript.
TL documented the risk factors and complications
SM participated in the design and coordination of the study
AH participated in the design of the study and carried out surgery
All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Copeland G Jones D Walters M POSSUM: a scoring system for surgical audit Br J Surg 1991 78 355 360 2021856
Bonenkamp J Songun I Hermans J Sasako M Welvaart K Plukker J van EP Obertop H Gouma D Taat C Randomised comparison of morbidity after D1 and D2 dissection for gastric cancer in 996 Dutch patients Lancet 1995 345 745 748 7891484 10.1016/S0140-6736(95)90637-1
Cuschieri A Fayers P Fielding J Craven J Bancewicz J Joypaul V Cook P Postoperative morbidity and mortality after D1 and D2 resections for gastric cancer: preliminary results of the MRC randomised controlled surgical trial. The Surgical Cooperative Group Lancet 1996 347 995 999 8606613 10.1016/S0140-6736(96)90144-0
Bonenkamp J Hermans J Sasako M van dVC Extended lymph-node dissection for gastric cancer. Dutch Gastric Cancer Group N Engl J Med 1999 340 908 914 10089184 10.1056/NEJM199903253401202
Cuschieri A Weeden S Fielding J Bancewicz J Craven J Joypaul V Sydes M Fayers P Patient survival after D1 and D2 resections for gastric cancer: long-term results of the MRC randomized surgical trial. Surgical Co-operative Group Br J Cancer 1999 79 1522 1530 10188901 10.1038/sj.bjc.6690243
Lorenz W Menke H Heintz D Junginger T Dick W Rothmund M [Analysis of risk in the perioperative period] Zentralbl Chir 1990 115 129 133 2184628
Wolters U Wolf T Stutzer H Schroder T ASA classification and perioperative variables as predictors of postoperative outcome Br J Anaesth 1996 77 217 222 8881629
Lubke T Monig S Schneider P Holscher A Bollschweiler E [Does Charlson-comorbidity index correlate with short-term outcome in patients with gastric cancer?] Zentralbl Chir 2003 128 970 976 14669119 10.1055/s-2003-44805
Prytherch D Whiteley M Higgins B Weaver P Prout W Powell S POSSUM and Portsmouth POSSUM for predicting mortality. Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity Br J Surg 1998 85 1217 1220 9752863 10.1046/j.1365-2168.1998.00840.x
Wijesinghe L Mahmood T Scott D Berridge D Kent P Kester R Comparison of POSSUM and the Portsmouth predictor equation for predicting death following vascular surgery Br J Surg 1998 85 209 212 9501818 10.1046/j.1365-2168.1998.00709.x
Tekkis P Kessaris N Kocher H Poloniecki J Lyttle J Windsor A Evaluation of POSSUM and P-POSSUM scoring systems in patients undergoing colorectal surgery Br J Surg 2003 90 340 345 12594670 10.1002/bjs.4037
Tekkis P Kocher H Bentley A Cullen P South L Trotter G Ellul J Operative mortality rates among surgeons: comparison of POSSUM and p-POSSUM scoring systems in gastrointestinal surgery Dis Colon Rectum 2000 43 1528 1532 discusssion 1532-1524. 11089587
Sagar P Hartley M MacFie J Taylor B Copeland G Comparison of individual surgeon's performance. Risk-adjusted analysis with POSSUM scoring system Dis Colon Rectum 1996 39 654 658 8646952
Organisation JGC Japanese Classification of Gastric Carcinoma Gastric Cancer 1998 2 Tokyo 10 24 11957040
Siewert J Holscher A Becker K Goessner W Cardia Cancer: attempt at a therapeutically relevant classification Chirurg 1987 58 25 29 3829805
Monig S Collet P Baldus S Schmackpfeffer K Schroder W Thiele J Dienes H Holscher A Splenectomy in proximal gastric cancer: frequency of lymph node metastasis to the splenic hilus J Surg Oncol 2001 76 89 92 11223832 10.1002/1096-9098(200102)76:2<89::AID-JSO1016>3.0.CO;2-I
Bollschweiler E Schroder W Holscher A Siewert J Preoperative risk analysis in patients with adenocarcinoma or squamous cell carcinoma of the oesophagus Br J Surg 2000 87 1106 1110 10931059 10.1046/j.1365-2168.2000.01474.x
Bollschweiler E Scoringsysteme und Qualitätsmanagement Qualitätsmangement in der Medizin 2002 18 ecomed
Neary W Heather B Earnshaw J The Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity (POSSUM) Br J Surg 2003 90 157 165 12555290 10.1002/bjs.4041
McPeek B Gasko M Mosteller F Measuring outcome from anaesthesia Theor Surg 1986 1 2 9
Bollschweiler E Siewert J Lorenz W Ohmann C Selbmann H [Recording the postoperative course. What data are necessary? Mortality or fatality, during hospitalization, within 30 and 90 days?] Langenbecks Arch Chir 1992 377 378 384 1479863 10.1007/BF00574778
Hosmer D Lemeshow S Applied Logisitic Regression 2000 2 New York: John Wiley
Hartgrink H van dVC Putter H Bonenkamp J Klein KE Songun I Welvaart K van KJ Meijer S Plukker J Extended lymph node dissection for gastric cancer: who may benefit? Final results of the randomized Dutch gastric cancer group trial J Clin Oncol 2004 22 2069 2077 15082726 10.1200/JCO.2004.08.026
Prytherch D Sutton G Boyle J Portsmouth POSSUM models for abdominal aortic aneurysm surgery Br J Surg 2001 88 958 963 11442527 10.1046/j.0007-1323.2001.01820.x
Whiteley M Prytherch D Higgins B Weaver P Prout W An evaluation of the POSSUM surgical scoring system Br J Surg 1996 83 812 815 8696749
Zafirellis K Fountoulakis A Dolan K Dexter S Martin I Sue-Ling H Evaluation of POSSUM in patients with oesophageal cancer undergoing resection Br J Surg 2002 89 1150 1155 12190681 10.1046/j.1365-2168.2002.02179.x
Tekkis P McCulloch P Poloniecki J Prytherch D Kessaris N Steger A Risk-adjusted prediction of operative mortality in oesophagogastric surgery with OPOSSUM Br J Surg 2004 91 288 295 14991628 10.1002/bjs.4414
Charlson M Pompei P Ales K MacKenzie C A new method of classifying prognostic comorbidity in longitudinal studies: development and validation J Chronic Dis 1987 40 373 383 3558716 10.1016/0021-9681(87)90171-8
van Doorn C Bogardus S Williams C Concato J Towle V Inouye S Risk adjustment for older hospitalized persons: a comparison of two methods of data collection for the Charlson index J Clin Epidemiol 2001 54 694 701 11438410 10.1016/S0895-4356(00)00367-X
Lazarides M Arvanitis D Drista H Staramos D Dayantas J POSSUM and APACHE II scores do not predict the outcome of ruptured infrarenal aortic aneurysms Ann Vasc Surg 1997 11 155 158 9181770 10.1007/s100169900026
Knaus W Wagner D Draper E Zimmerman J Bergner M Bastos P Sirio C Murphy D Lotring T Damiano A The APACHE III prognostic system. Risk prediction of hospital mortality for critically ill hospitalized adults Chest 1991 100 1619 1636 1959406
| 15831104 | PMC1112603 | CC BY | 2021-01-04 16:28:04 | no | BMC Surg. 2005 Apr 15; 5:8 | utf-8 | BMC Surg | 2,005 | 10.1186/1471-2482-5-8 | oa_comm |
==== Front
BMC Womens HealthBMC Women's Health1472-6874BioMed Central London 1472-6874-5-51586562710.1186/1472-6874-5-5Research ArticleThe risk of menstrual abnormalities after tubal sterilization: a case control study shobeiri Mehri Jafari [email protected] Simin [email protected] Department of Obstetrics and Gynecology, Al-zahra Hospital, Tabriz University of Medical Sciences, South Artesh Ave., Tabriz, Iran2 Department of Anesthesia, Al-zahra Hospital, Tabriz University of Medical Sciences, South Artesh Ave., Tabriz, Iran2005 2 5 2005 5 5 5 22 10 2004 2 5 2005 Copyright © 2005 shobeiri and AtashKhoii; licensee BioMed Central Ltd.2005shobeiri and AtashKhoii; 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
Tubal sterilization is the method of family planning most commonly used. The existence of the post-tubal-ligation syndrome of menstrual abnormalities has been the subject of debate for decades.
Methods
In a cross-sectional study, 112 women with the history of Pomeroy type of tubal ligation achieved by minilaparatomy as the case group and 288 women with no previous tubal ligation as the control group were assessed for menstrual abnormalities.
Results
Menstrual abnormalities were not significantly different between the case and control groups (p = 0.824). The abnormal uterine bleeding frequency differences in two different age groups (30–39 and 40–45 years old) were statistically significant (p = 0.0176).
Conclusion
Tubal sterilization does not cause menstrual irregularities.
==== Body
Background
Tubal sterilization is the most commonly used method of family planning. In 1990 the corresponding percentage of married women in reproductive age who used sterilization was 22% in developing countries and the corresponding percentage in developed countries was 11%. These women represented 44% and 18% of all contraceptive users in developing and developed countries, respectively. Questions regarding the existence of a post tubal ligation syndrome of menstrual abnormalities continue. Questions arose initially when Williams and colleagues reported in 1951 that sterilized women had a higher than expected occurrence of menorrhagia and metrorrhagia [1]. After that the existence of a post-tubal-ligation syndrome of menstrual abnormalities has been debated for decades [2]. Many authors have investigated the sequelae of female sterilization [2-9]. Increased premenstrual distress, heavier and more prolonged menstrual bleeding, and increased dysmenorrhea have been reported [3]. However, failure to control for use of oral contraceptives, age, obesity, parity, interval since sterilization, or type of sterilization may have effects on the results of these studies [1,3]. Because of the importance of this debate, we compared the occurrence of menstrual abnormalities in women with and without a prior history of tubal ligation.
Methods
This cross sectional case control study has been carried out on 500 women at Al-zahra hospital during 1999 to 2001 to assess the effect of tubal sterilization on the menstrual cycle. 260 women with abnormal uterine bleeding referred for diagnostic curettage, and 240 healthy women under the coverage of the hospital family planning center were selected randomly, and all were assessed for tubal ligation.
All women aged 30 to 46 were selected from a low-income urban population, with body weight between 50 to 90 kg. In the abnormal uterine bleeding group, those who had intrauterine device (IUD), leiomyoma on sonography, uterine size of greater than 9 cm or suffered from medical disorders were excluded from the study. Of 260 patients with menstrual irregularities, 30 subjects were excluded from the study. From the remaining 230 subjects, assessed for tubal sterilization, 87 patients had tubal ligation. Of 240 healthy women assessed for tubal ligation, 95 had previous tubal ligation. Totally 182 subjects with previous tubal ligation (case) and 288 subjects with no history of previous tubal ligation (control) were compared for abnormal uterine bleeding. Those subjects in the case group who had menstrual abnormalities, IUD, medical disorders or were on hormonal contraception, during the first year prior to the sterilization were excluded from the study. Those who were at least 30 and at most 40 years of age by the time of tubal ligation and had Pomeroy type of interval tubal ligation via minilaparatomy were included the study. Finally, considering the exclusion and inclusion criterias, 112 subjects remained in the case group and 288 with no tubal ligation in the control group were evaluated for menstrual abnormalities. Information on demographic, obstetrics, medical and menstrual bleeding pattern of all subjects were obtained. Women were asked about the duration and amount of bleeding, and length of cycle (number of days from the beginning of one menstrual period to the beginning of the next one). A menstrual interval of 21 to 35 days was considered normal. A menstrual interval shorter than 21 days was defined as polymenorrhea. Duration of flow of 7 days or less was considered normal. A patient's self-described history of normal or heavy blood loss was indicative of the amount of flow. Regularly timed heavy bleeding and duration of flow greater than 7 days were considered menorrhagia and hypermenorrhea respectively. Excessive and prolonged bleeding that occurred irregularly was defined as menometrorrhagia.
Data was analyzed by the SPSS statistical software (version, 12) and compared with the chi-square test. P values of 0.05 or less were considered as statistically significant.
Results
By considering the exclusion and inclusion criterias, 112 tubal ligated (case) and 288 non-tubal ligated subjects (control) were evaluated for menstrual abnormalities. Of 112 subjects in the case group, 57 (50.8%) had menstrual abnormalities. The corresponding figure in the control group was 143, accounting for 49.6% of the studied subjects in this group. The results of chi-square analysis, indicate that there was no significant difference in the menstrual abnormalities between two groups, χ2 = 0.050, p = 0.824.
The highest frequency of the menstrual abnormalities in the case group was 54.3% for the group aged between 30–39 while in the control group this value was 65% for those aged 40–45. There was significant difference in the menstrual abnormalities frequency of two groups by different age groups, χ2 = 9.06, p = 0.0176 (Table 1).
Table 1 Demographic and obstetrical information of subjects with menstrual irregularities in case and control groups
(57) case group (143) control group df pv
No. (%) No. (%)
Age groups(year):
30–39 31(54.3) 50(34.96) 1 0.0176819
40–45 26(45.6) 93(65.1)
Parity groups (No.):
2–4 8(14) 31(22) 2 0.3819013
5–7 28(50) 58(40)
>7 21(36) 54(38)
Type of abnormal bleeding is given in Figure 1. The most common type of menstrual changes in case and control groups was polymenorrhea (35%) and menorrhagy (30%) respectively. The differences were not significant, χ2 = 6.93 p = 0.2260.
Figure 1 Comparison of the menstrual irregularities type of case and control groups
The frequency onset of abnormal bleeding after sterilization in the case group was 61% during the first year, 34% in 2–5 years after sterilization and 5% over 5 years.
The menstrual abnormality frequency distributions by different parity groups in the case and control groups are shown in Table 1. The most common menstrual abnormalities frequency which belonged to parity of 5–7, was 50% and 40% in case and control groups respectively. The parity differences between two groups were not significant, χ2 = 1.93 p = 0.3819.
The most common histologic findings in case and control groups were proliferative endometrium 31.6% and anovulatory cycle (28.7%) respectively. There was no significant difference in the histologic finding of two groups, χ2 = 5.351, p = 0.253.
Discussion
There are some factors other than sterilization per se that may have influences on post sterilization menstrual changes. Two such factors are the use of oral contraceptives and IUD. The women who use oral contraceptive may have some menstrual changes after sterilization attributable solely to cessation of oral contraceptive use. In order to exclude the interventional effect of IUD and oral contraceptive, we included patients who did not use them during one year before sterilization. Since the type of tubal sterilization may have effects on study results, we included only Pomeroy type of interval sterilization by minilaparatomy. The results are similar to those of Gentile et al[3]., Bernard et al[4], Peterson et al[2], who showed no significant changes in menstrual cycle characteristics in women with or without tubal ligation. Concerning the demographic information including the socioeconomic status among the case and control groups, all participants were of a low-income population. In the unadjusted analysis, when the sterilized groups were compared to the control group, slight but not statistically significant changes were noted in menstrual indices. The results are similar to those of Peterson et al[2], Bhiwandiwala et al[11], who showed no menstrual pattern changes following sterilization.
Although we had excluded patients who were on hormonal contraceptives and had IUD, we found that most of the menstrual changes occurred at first year of sterilization (61%). After first year of sterilization the menstrual changes decreased to 34% in 2–4 years and 5% after 5 or more years of sterilization. The results are similar to those of Parsanezhad et al[12], who found that almost all menstrual changes occurred between 6 and 24 months after sterilization. Thus it may be concluded that sterilization related menstrual changes during the first years of sterilization may occur due to some psychological reaction to tubal ligation. DeStefano et al[13], in their long term follow up of sterilized women found an increased risk of menstrual abnormalities even after a long period of 49 to 87 months after sterilization. These late menstrual changes are difficult to explain, because it is not easy to postulate a physiologic mechanism that would take more than 4 years to develop and adversely affect menstrual cycles. Our results are dissimilar to those of Kasonde and Bonnar[14] who did not find any increased menstrual blood loss up to 6–12 months after sterilization. It seems that different results of these studies may because Kasonde and Bonnar objectively measured blood loss whereas this study relied on subjective self reported amounts of blood loss.
Shy et al[15] believe that menstrual changes effect of sterilization depends on age at the time of sterilization. Women who undergo sterilization between 20 and 29 years of age have more menstrual irregularities than women who undergo the procedure after age 30. In order to exclude this factor, we included only the patients with at least 30 and at most 40 years of age by the time of sterilization. The results show that the most common age group of menstrual irregularities is 30–39 years (Table 1). These results are similar to those of Wilcox et al[16] and Shy et al[15] who found that sterilization at younger ages leads to more menstrual irregularities than sterilization at older ages.
Conclusion
Women who have undergone a Pomeroy type of tubal ligation have no more menstrual abnormalities than those without tubal ligation. Sterilization at younger ages has more affect on menstrual irregularities than sterilization at older ages.
It seems that more frequency of menstrual changes at first year of sterilization is due to other factors such as psychiatric problems. Further studies on psychiatric changes of sterilization are mandatory to evaluate its effects on immediate post sterilization menstrual irregularities.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
MJSH is the principle investigator and was involved in planning, coordinating the research, sample collection, handling, writing and editing of the manuscript, SA was involved in coordinating and supervising data entry and analysis and was involved in planning, coordination of the research at the Hospital family planning center, and reviewing of the paper. Both authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors would like to thank Dr. Kamran Sedagat for statistical advice, Oshaghi for providing type and print of this article and all reviewers for useful suggestions.
==== Refs
Peterson HerbertB Pollack ArmyE Warshaw JeffreyS John A Rock, Howard W Jones, III Tubal sterilization Te Lindes's Operative Gynecology 2003 9 Philadelphia: Lippincott Williams & Wilkins 537 56
Peterson HB Jeng G Folger SG Hillis SA Marchbanks PA Wilco The risk of menstrual abnormalities after tubal sterilization N Engl J Med 2000 343 1681 7 11106717 10.1056/NEJM200012073432303
Gentile GP Kaufman SC Helbig DW Is there any evidence for a post-tubal sterilization syndrome? Fertil Steril 1998 69 179 186 9496325 10.1016/S0015-0282(97)00229-X
Harlow BernardL Stacey A Missmer Daniel W Gramer Barbieri RobertL Dose tubal sterilization influence the subsequent risk of menorrhagia or dymenorrhea? Fertil Steril 2002 77 754 60 11937129 10.1016/S0015-0282(01)03253-8
Shain RN Miller WB Mitchell GW Holden AE Rosenthat M Menstrual pattern change 1 year after sterilization: resultes of a controlled, prospective study Fertil Steril 1989 52 192 203 2753169
Cattanach JF Milan BJ Post-tubal sterilization problems corrected with ovarian steroidogenesis Contraception 1988 38 541 50 3197418 10.1016/0010-7824(88)90157-6
Buytaert P Viaene P laparoscopic tubal sterilization: Postoperative follow-up a late gynecological complaints Eur J Obstet Gynecol Reprod Biol 1980 10 119 24 6444903
Cattanach J Oestrogen deficiency after tubal ligation Lancet 1985 1 847 9 2858712 10.1016/S0140-6736(85)92209-3
Stubblefield PhillipG Larry J Copeland Contraception Text book of Gynecology 2000 2 Philadelphia: W.B. Saunders Company 287 328
Carmona F Cristobal P casamitjana R Balasch J Effect of tubal sterilization on ovarian follicular reserve P, and function Am J Obstet Gynecol 2003 189 447 52 14520216 10.1067/S0002-9378(03)00487-3
Bhiwandiwala Mumford SD Feldblum PJ Menstrual pattern changes following laparoscopic sterilization with different occlusion techniques. A review of 10004 cases Am J Obstet Gynecol 1983 145 684 94 6219585
Parsanezhad ME Alborzi SA Namavar Jahromi B Menstrual abnormalities and pain after five tubal sterilization methods: a randomized controlled trial IJMS 2003 28 57 61
DeStefano F Huezeo CM Peterson HB Rubin GL Layde PM Ory HW Menstrual changes two years after tubal sterilization Obstet Gynecol 1983 62 673 81 6633993
Kasonde JM Bonnar J Effect of sterilization on menstrual blood loss Br J Obstet Gynaecol 1976 83 572 5 952786
Shy KK Stergachis A Grothaus LG Wanger EH Hecht J Anderson G Am J Obstet Gynecol 1992 166 1698 705 discussion 1705-6 1615977
Wilcox LS Martinez-Schnell B Peterson HB Ware JH Hughes JM Am J Epidemiol 1992 135 1368 81 1510083
| 15865627 | PMC1112604 | CC BY | 2021-01-04 16:30:36 | no | BMC Womens Health. 2005 May 2; 5:5 | utf-8 | BMC Womens Health | 2,005 | 10.1186/1472-6874-5-5 | oa_comm |
==== Front
Biomed Eng OnlineBioMedical Engineering OnLine1475-925XBioMed Central London 1475-925X-4-291586213110.1186/1475-925X-4-29ResearchA portable near infrared spectroscopy system for bedside monitoring of newborn brain Bozkurt Alper [email protected] Arye [email protected] Harel [email protected] Banu [email protected] School of Biomedical Engineering, Science and Health Systems, Drexel University, 3141 Chestnut Street Philadelphia, Pennsylvania 19104, USA2 St.Peter's University Hospital, 254 Easton Ave, New Brunswick, New Jersey 08801, USA2005 29 4 2005 4 29 29 1 3 2005 29 4 2005 Copyright © 2005 Bozkurt et al; licensee BioMed Central Ltd.2005Bozkurt 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
Newborns with critical health conditions are monitored in neonatal intensive care units (NICU). In NICU, one of the most important problems that they face is the risk of brain injury. There is a need for continuous monitoring of newborn's brain function to prevent any potential brain injury. This type of monitoring should not interfere with intensive care of the newborn. Therefore, it should be non-invasive and portable.
Methods
In this paper, a low-cost, battery operated, dual wavelength, continuous wave near infrared spectroscopy system for continuous bedside hemodynamic monitoring of neonatal brain is presented. The system has been designed to optimize SNR by optimizing the wavelength-multiplexing parameters with special emphasis on safety issues concerning burn injuries. SNR improvement by utilizing the entire dynamic range has been satisfied with modifications in analog circuitry.
Results and Conclusion
As a result, a shot-limited SNR of 67 dB has been achieved for 10 Hz temporal resolution. The system can operate more than 30 hours without recharging when an off-the-shelf 1850 mAh-7.2 V battery is used. Laboratory tests with optical phantoms and preliminary data recorded in NICU demonstrate the potential of the system as a reliable clinical tool to be employed in the bedside regional monitoring of newborn brain metabolism under intensive care.
==== Body
I. Background
When light in the near infrared (NIR) range of the spectrum is shone through the scalp, injected photons follow various paths inside the head. Some of these photons are absorbed by different layers of the tissue such as skin, skull and brain. Others exit the head after following the so-called "banana" pattern due to scattering effect of the tissue [1]. Backscattered photons can be detected by means of appropriate optical apparatus. When the absorption spectrum of light is analyzed, it is seen that the main absorbers in the NIR range are blood chromophores of oxygenated and deoxygenated hemoglobin (HbO2 and Hb, respectively). Water and lipid are relatively transparent to NIR light. Therefore, changes in the amplitude of backscattered light can be interpreted as changes in blood chromophore concentrations. The procedure of estimating blood chromophore concentrations by means of near infrared light is called Near Infrared Spectroscopy (NIRS)[1]. Blood chromophore information can be used to estimate blood volume and tissue oxygenation which are indications of hemodynamic activity.
Different approaches can be used to implement NIRS such as time resolved, frequency domain and continuous wave techniques [2]. Among these methods, continuous wave (cw) NIRS is the most practical one, where light with constant amplitude is injected to tissue and amplitude decay of the light intensity due to absorption is analyzed. Changes in light amplitude are used to calculate changes in concentrations of blood chromophores. Due to its practicality cwNIRS systems allow for bedside or home monitoring of blood chromophores for extended periods.
Pulse oximetry, a NIR light based technique similar to NIRS, is being widely used in current clinical practice. The aim of pulse oximetry is to detect arterial blood saturation. CwNIRS further expands the application window of NIR light by providing information about blood dynamics in capillaries. Although, the arterial saturation obtained via pulse oximeter can only provide global information about the clinical state of the patient, capillary blood dynamics studied with cwNIRS is capable of supplying local tissue oxygenation information which can be used for various clinical purposes.
One of the important and promising application areas of cwNIRS technique is clinical neonatology. CwNIRS systems are non-invasive and low-power systems. They allow real time measurements without removing newborns from nursery units, thereby not interfering with intensive care. This is especially important for the vulnerable preterm neonate population. Some clinical situations, where neonatal cwNIRS can be employed, are asphyxia, hypoglycemia, apnea, endotracheal suctioning, aminophylline administration, indomethacin and exogenous surfactant administrations [3]. By spanning such a large application space, cwNIRS has a strong potential to be an inevitable part of clinical neonatalogy by monitoring cerebral hemodynamics of pre- and full-term newborns. Until now, a number of NIRS systems have been reported by various groups to assess oxygenation related changes on infant brain [4-7]. Various clinical studies have been run by using these systems [8-13].
In this paper, we present and discuss design issues of a portable, battery-operated, low-cost, low-noise, fast cwNIRS system which has been designed for bedside cerebral hemodynamic monitoring of newborns in various clinical studies. To our knowledge, none of the reported NIRS systems provide these features taken together, which is a crucial factor in transferring the technology to NICU for continuous bedside clinical monitoring, particularly, of regional brain metabolism. Moreover, the system provides access to raw photon data coming from detectors for further off-line signal processing. There are a number of open biomedical signal processing questions such as removing physiological noise and motion artifacts from NIRS data as well as hemodynamic detection of evoked events which are still waiting to be explored by biomedical signal processing experts.
As a preliminary evaluation, we present pilot results demonstrating that the system is capable of monitoring changes in HbO2 and Hb concentrations in Intralipid solution and in infant brain.
II. Methods
A. Principles of cwNIRS Measurement Device
The change in the amount of blood chromophores in the tissue can be predicted by means of Modified Beer Lambert Law as explained below [14,15].
When the NIR light source and detector are located as in fig 1, the detector receives backscattered photons. Some of the injected photons are lost as a result of scattering and absorption due to different structures in the tissue. The attenuation of light between the source and detector can be formulated as follows:
Figure 1 Propagation of photons between source and detector and parameters for Modified Beer Lambert Law
where Iin is incident light, Iout is the detected light and ODλ is the optical density for wavelength λ. Therefore, ODλ can be defined as attenuation in intensity of light as a function of wavelength λ. This attenuation is the superposition of absorption (Aλ) and scattering (Sλ) of light with wavelength λ. Thus,
Blood chromophores HbO2 and Hb are the main absorbers of light in the NIR region of the spectra. Therefore, absorption of light can be formulated as:
where εi,λ is the specific extinction coefficient of blood chromophore for wavelength λ, Ci is the concentration of blood chromophores and Lλ is the pathlength of light at λ. Pathlength can be expressed in terms of source detector separation as:
Lλ = d·DPFλ (4)
Where d is the distance between the light source and the detector and DPFλ is the differential pathlength factor. The differential pathlength factor is the correction in the mean photon pathlength for scattering and defined as [16]:
where μa,λ is the absorption coefficient and μs',λ is the reduced scattering coefficient at wavelength λ. Here, the reduced scattering coefficient is used instead of scattering coefficient due to the anisotropy of the scattering.
When scattering, therefore the differential pathlength factor, is assumed to be constant, two successive measurements yield the differential OD value of:
where the effect of scattering is cancelled.
Since each chromophore has a specific extinction coefficient and differential pathlength factor, measurement with two wavelengths leads to:
where
and
and are defined as and .
Therefore,
Equation 9 provides a transformation from light output change to change in blood chromophore concentrations. By using blood chromophore concentrations we define two parameters, namely,
OXY = ΔCHbO2 - ΔCHb (10)
and
BV = ΔCHbO2 + ΔCHb (11)
OXY and BV are estimates proportional to, respectively, oxygenation and blood volume changes in the tissue due to hemodynamic activation.
Calculation of parameters
Absorption and reduced scattering coefficients cannot be measured with cwNIRS method, therefore, are calculated by using the values given in the literature[17,18]. An assumption of 85% saturation and 100 μM total hemoglobin concentration[19] results in the absorption and reduced scattering coefficients of .
By using these coefficients in equation 5, DPF values are calculated as 5.0055 and 4.6564 for 730 nm and 850 nm, respectively. These values are consistent with the measurements in the literature[20]. On the other hand, is obtained from the values given in the literature[15] as:
By inserting into equation 8, the relationship between light intensity change and blood chromophore concentration is derived as:
B. cwNIRS Instrumentation
a. Description of system design
The cwNIRS system consists of three parts: probe, control circuit and processing unit (fig 2). The probe constitutes the interface between the control system and the subject. It holds the light source and detector in an appropriate geometry. Operation of the light source and detectors are manipulated by the control circuit which can be subdivided as transmitter and receiver. Transmitter and receiver are controlled by the computer software for coherent detection of two wavelengths. The computer also stores and displays received light information after applying necessary signal processing schemes.
Figure 2 System containing computer, control box and probe
The design of the system can be explained through the use of four subtitles: transmitter, probe, receiver and computer processing. The role of each part is displayed in the detailed block diagram of the system (fig 3)
Figure 3 Block diagram of the cwNIRS system
1. Transmitter
The transmitter part of the control circuit is composed of an adjustable LED driver and wavelength selector. The purpose of the adjustable LED driver is to regulate the light output of the LED in order to compensate for absorption differences in various tissue types. For subjects with lighter skin color, sufficient amount of output can be achieved by applying lower power light whereas larger intensity of light is necessary for darker skin color. This is because of the relative amount of melanin in skin. Since melanin concentration is constant during the measurement, it does not affect the results related to concentration change. However, darker skin color causes a decrease in signal to noise ratio (SNR) by absorbing more light. Therefore, the same level of SNR can be achieved by adjusting the intensity of NIR light.
Light intensity can be changed by varying the current passing through the light source. The user determines the amount of current and enters it into the user interface. Then, the software adjusts the current by changing the resistance of a digital potentiometer in driver circuit. The range of current that can pass through the light source is from 0 to 100 mA. Values larger than 100 mA have potential to damage the light source; therefore, the software warns the user about these values through the user-interface and does not initiate the operation. The device also has an indicator light on the electronics box which turns on to warn the user when a current larger than 100 mA passes through the light source due to an undesired short circuit.
As described in the previous section, two different wavelengths are required to resolve two types of blood chromophores. We employed these two wavelengths together by means of time multiplexing. This is implemented by a multiplexer IC which is controlled by the software. At the same time, data from corresponding detector channel were registered to satisfy coherent detection. The multiplexer was not directly connected to the light source, since current in the order of 10 to 100 mA passes through the source. Relatively higher RON resistance of the multiplexer causes a large voltage drop and necessary voltage may not be supplied to light source to turn it on. Instead, we connected multiplexer to operate an analog switch with a very low resistance value to turn on/off the light source. This permits the use of conventional camcorder batteries with low voltage values to power the system.
During each cycle, after time multiplexing two wavelengths, we allow an idle period where no wavelength is turned on. The reason for this idle period is that NIR light source is a semiconductor junction and heats up during the operation. Experimentally, it has been shown that an idle time helps light source to cool down to safe ranges (see system performance part). Moreover, when NIR light is shone through the tissue, detector readings are composed of ambient light penetrating through the tissue and the offset of the electronic components in addition to optical information coming from the brain. These offset values can be determined by recording detector output when NIR wavelengths are turned off. During each cycle, detector reading for idle period is used to correct the mentioned offset values in order to increase the accuracy of the readings. Furthermore, detector output during dark period can be used to monitor the optical isolation of detectors from direct incidence of the ambient light. If offset values are larger than expected, the probe is reattached to the skin for better coupling and isolation.
2. Receiver
The receiver part transfers light information obtained from detectors to the computer after amplifying and analog filtering. The aim of the amplification part is to bring the signal level close to the top of the dynamic range of the analog to digital converter. This operation minimizes the error that occurs during quantization process. The amplification process is initiated by the user. The user hits a button on the user interface after obtaining a satisfactory probe-skin coupling during calibration process. The software reads the light output coming from the brain and adjusts the value of a digital potentiometer in the gain amplifier such that the signal value is amplified to the full dynamic range. A quantization SNR of 75 dB is guaranteed as the result of closed-loop gain adjustment procedure where quantization resolution is 12 bit. The gain value is also displayed on the user-interface. Typical photon sampling frequency is 60 kHz and wavelength multiplexing frequency is 10 Hz.
Before digital conversion, the signal is low-pass filtered with cutoff frequency of 1.5 kHz to avoid aliasing during the sampling process.
3. Probe
The probe is the most critical part of the system design since it establishes the interface between electronics and the subject. It holds the light source and two detectors in an appropriate geometry.
Detectors are positioned on opposite sides of the source with a source-detector distance of 2 cm as seen in fig 4. This geometry allows monitoring of two different locations of the brain, simultaneously. Approximate penetration depth is 1 cm for this configuration [2]. Alternatively, detectors can be located on the same side with different source-detector distances for multi-depth measurements. By monitoring different depths, absolute saturation of the brain tissue can be detected.
Figure 4 The first prototype of the "flexible" probe holding low power LED and detectors
In our research, we prefer light emitting diode (LED) as the source of the NIR light instead of using laser. Higher light intensity levels can be utilized with LED, since it is a non-coherent and non-collimated light source. Power consumption of LED is minimal with respect to the laser source, hence the system can be battery operated. Laser light source also suffers from extreme heating of the semiconductor junction. Optical fibers are required to carry laser light to tissue.
Therefore, more instrument space is needed to provide cooling. Furthermore, laser source needs additional precautions due to its injury potential on the eye. It may not be appropriate for use in neonatal intensive care units for vulnerable population of newborns. As a result, practical and portable systems are quite difficult and inappropriate to build with laser source.
Single package LEDs which contain 2–3 wavelengths together are available in the market (e.g. Epitex, Inc). We selected a source where 730 and 850 nm are enclosed in a TO18 package. Crosstalk between these two wavelengths is less significant with respect to other wavelength pairs[21]. This combination of wavelengths is appropriate for our purpose of resolving two chromophore concentrations. LED can emit output power between 3 and 17 mW where the typical power used is 9 mW. The spectrum purity of this LED in the NIR region is around 30 nm. Although it is wider with respect to diode lasers (5–6 nm), 30 nm gives enough sharpness to resolve two wavelengths since absorption spectra is relatively flat around source wavelengths.
Monolithic photodiode/preamplifier integrated circuits housed in a clear 8-pin DIP package (OPT101, Burr-Brown®, Co.) were used to detect the light coming from the brain. Active sensing area of this detector is 2.29 × 2.29 mm2. Feedback resistance of 1 Mohm provides a transimpedance (V/A) gain of 106 and bandwidth more than 100 Hz.
The most challenging aspect of the design has been the coupling of optical elements (optodes) to the scalp of the subject. We have used flexible circuit technology to fit the optodes to the curved surface of the subject's head. The source and detectors attached to the flexible circuit board are embedded into medically graded foam (fig 4). This prevents relatively sharp corners of optodes to come in contact with the scalp of the subject. Moreover, it blocks the ambient light which tends to saturate the photodiodes.
We developed various methods to affix the flexible probe to the surface of the head. The first method is to use double sided medically graded sticky tape with one side attached to the foam and other side attached to the scalp (Adchem®). This provides a satisfactory and stable optode-scalp coupling. Another method uses a medically graded self-sticky silicon material (Implantech, Inc) instead of sticky tape and the foam; optodes can easily be embedded in this silicon material. In yet another method, we used medically graded foam again. In this case, stable optode-scalp coupling is provided by inserting the foam under a baby hat as in fig 5. This baby hat is generally utilized in NICU to hold incubator tubes running to newborn's nose and mouth. The performances of these methods were similar in terms of ease-of-use and coupling efficiency.
Figure 5 Baby hat method used for optode-scalp coupling (see "Acknowledgment")
In order to avoid the effect of ambient light further and to increase the SNR, we attached an NIR filter (Edmund Optics®) with cut-off wavelength of 700 nm to the surface of the photodiode. This introduced an insulation layer between photodiodes and scalp, in addition to rejecting non-NIR wavelengths.
An 1850 mAh Li-Ion camcorder battery with 7.2 V voltage supply was used to power the transmitter, receiver and the probe. A charging system was also added to the box to charge the battery without taking it out of the box.
4. Processing Unit
A data acquisition card with 10 V dynamic range and 12-bit resolution (DAQcard 1200, National Instruments™) has been used to convert analog information to digital where sampling rates up to 100 kHz can be achieved. 5 digital output and 2 analog input channels of the data acquisition card have been sufficient for the entire operation of the system.
The system hardware is software (visual C++®) operated: the custom software manages time multiplexing operation between source and detector pairs, the coherent detection from multiple sensors and incorporation of real time digital signal processing algorithms. The user interface software displays changes in blood volume, oxygenation and, concentrations of Hb and HbO2 by implementing Beer-Lambert Law in real time. In order to monitor optode-scalp coupling during the operation, the user interface also displays raw voltage data coming from control box. Especially, the offset value read during the idle period of the light source is an important parameter demonstrating the ambient light leakage. Data is recorded in a file for further off-line processing.
Typical time multiplexed voltage output of the detectors can be represented as seen in fig 6. Each square pulse corresponds to detected photons when a particular wavelength is turned on. This signal can be interpreted as a low-pass signal multiplied by a train of square pulses, which corresponds to convolution with sinc functions in frequency domain. Thus, processing of time multiplexed signal is not straightforward. However, we can get rid of time multiplexing scheme by applying finite impulse response (FIR) filtering followed by decimation. As we convolve FIR filter with detector output, we only record the result of the convolution when all FIR filter coefficients coincides with the samples of the square pulse as in fig 6. To have a sharper cut-off, filter order is selected to be equal to the number of samples that each square pulse contains. We designed an FIR filter with 10 Hz cut-off frequency by using Dolph-Chebyshev window function with a ripple factor of 40. The operation of FIR filtering and decimation are applied real time by the control software.
Figure 6 Summary of the filtering operation
After decimation process, an additional FIR low-pass filter is applied to the signal to clean various artifacts. The NIRS signal that corresponds to hemodynamic activity of the brain has a very slow time course[15]. The expected frequency range of the hemodynamic activity is 0.06–0.16 Hz interval. On the other hand, hemodynamic activity signal is contaminated by other factors such as respiration, arterial pulsations and motion artifacts. Fortunately, all these artifacts are in higher frequency ranges with respect to hemodynamic activity. An online FIR filter is used to clean the frequencies above 0.16 Hz which is out of region of interest. Unfiltered data is recorded in parallel to filtered data for further offline-processing.
III. Results
A. System Performance
The performance of the device was evaluated by means of initial laboratory tests on phantom. First, shot limited signal to noise ratio (SNR) was measured. Shot noise is due to the random arrival of photons in the photodiode. Since the cause of shot noise is the quantum nature of light, it is impossible to remove it. To assess the shot noise limited SNR, we used an optical phantom (μa = 0.01 and μs' = 1.00 mm-1), which is free of motion artifact and physiological noise.
We measured SNR for various values of time multiplexing parameters: duty ratio and pulse repetition rates. SNR depends on these factors in two ways. First, duty cycle and pulse repetition rate determine the number of collected data points. More number of data points for each pulse is supposed to provide an increased SNR since the performance of the FIR filter is improved. On the other hand, duty cycle and pulse repetition rate affect the heating of the semiconductor junction of the LED. Longer duty cycle and pulse repetition rate cause an elevation in the temperature of the semiconductor junction. Temperature increase adds an extra noise to the generated light output, thereby, significantly limiting the SNR. Furthermore, this is an undesired condition since elevated temperature is a risk for the safety of the device on subject population. Therefore, we face a trade-off between filter performance and temperature rise of the semiconductor junction for duty ratio and pulse repetition rate to determine SNR.
The experimental results of temperature increase and SNR values for different duty ratios and pulse repetition rates are presented in fig 7 and 8. 60 kHz of sampling rate has been used during all measurements and temperature value at the 30th minute of the operation has been taken when temperature asymptotically converges to a stable value. Increases in both pulse repetition rate and duty cycle caused a temperature increase in the semiconductor junction. The expected rise in SNR due to increased number of samples is suppressed by extra noise in the light output of the LED as result of temperature increase. So, overall SNR of the device decreased.
Figure 7 Temperature increase & SNR vs. pulse repetition rate (duty ratio = 12.5%)
Figure 8 Temperature increase & SNR vs. duty ratio (pulse repetition rate = 10 Hz)
As a result of this analysis, operation duty cycle and pulse repetition rate were selected to be 12.5% and 10 Hz, respectively. This results in a relatively small amount of temperature increase whereas a SNR of 67 dB was obtained. Such high values of SNR could be achieved as a result of powering the entire system including the control box and processing unit by battery. An off-the-shelf 1850 mAh Lithium-Ion battery operates the system for more than 30 hours.
Temperature increase is an important factor to evaluate the safety of the system. NIR light is a non-ionizing type of light and its only known potential hazard is thermal injury of the skin. There are two factors that contribute to the temperature increase of the skin: radiated heat due to NIR light absorption by the skin and conducted heat due to heating of semiconductor junctions inside the LED, since LED is attached to the skin. Our previous study[22] demonstrates that temperature increase due to radiated heat is around 0.5–1°C, whereas, temperature increase due to conducted heat can be as high as 10°C. On the other hand, the heating of the semiconductor junction can be controlled by adjusting the pulse repetition rate and duty cycle, therefore, the effect of conductive heat can be minimized. When above proposed pulse repetition rate and duty cycle are used, the combined temperature increase of the skin is around 1.5°C and has no harmful effect to the subject under normal conditions. Therefore, the risk of the system for any potential burn injury is minimal.
Stray light rejection has found to be larger than 99% under normal illumination levels. This has been increased even more with the use of NIR filter. Inter-channel crosstalk between detectors was measured to be less than 0.1%.
B. Preliminary Evaluation
a. Liquid phantom experiment
A dynamic liquid phantom simulating the optical properties of the tissue was used in order to test the efficacy of the system. The basis of the phantom was formed by a scattering solution of Intralipid (Liposyn III) and phosphate buffered saline with pH = 7.4. 27.2 ml of intralipid was added to 1000 ml of water in order to obtain an overall reduced scattering coefficient of 0.8 mm-1 (730 nm). The solution was placed in a cylindrical beaker. A magnetic stirring rod was also placed in the beaker and the phantom was stirred during the course of the experiment. After the basis was formed, red blood cells obtained from healthy human blood were added to scattering solution to achieve a volume fraction of 1.5% and total hemoglobin concentration of 26 μM. This is a typical value for normal physiological conditions with an assumption of 4.0% blood volume and 40% hematocrit. The hemoglobin saturation was measured to be 85%.
In order to induce deoxygenation on the liquid phantom, 4 g of dry Bakers yeast was added to the solution. The temperature of the phantom was maintained at 37°C to keep the yeast active. This condition was maintained over a course of about 20–30 min. Deoxygenation was observed until hemoglobin saturation reached a steady state at 25%. After deoxygenation of yeast-intralipid solution reached steady state, oxygenation was simulated again by delivering extra oxygen to the phantom from an oxygen tank. Oxygen supply was maintained until a steady state level of oxygenation was obtained. Steady state hemoglobin saturation was 89%.
During this procedure, two sets of data were collected by attaching the cwNIRS probe to the side of the beaker. The first measurement was performed to observe the effect of HbO2 and Hb addition to the intralipid solution. Since there were no hemoglobin molecules in the intralipid solution, this measurement provided the absolute concentrations of added HbO2 and Hb as seen in figure 9. In this figure, the ratio of HbO2 and Hb is consistent with the measured saturation value of 85%.
Figure 9 Changes in hemoglobin concentration during red blood cell addition
In the second measurement, our aim was to observe deoxygenation as a result of yeast addition. Here, the baseline was taken as the steady state intralipid-hemoglobin solution without yeast. Yeast activation related changes in HbO2 and Hb concentrations can be seen in figure 10 in addition to BV and OXY curves. When yeast was added to the solution, deoxygenation was due to the consumption of oxygen by yeast which reached steady state within 15 minutes. Steady state values are consistent with the measured saturation of 25%. Re-oxygenation, when extra oxygen was supplied, is displayed in the same figure. During the procedure of deoxygenation and oxygenation, as expected, blood volume did not change significantly.
Figure 10 Changes in hemoglobin concentration (1), OXY and BV values (2) during the yeast test
b. Monitoring of newborn subjects
The validity of the system was tested on a newborn in NICU of St. Peter's University Hospital (NJ) by monitoring oxygenated and deoxygenated hemoglobin concentrations on the temporal region of the brain as a response to a standard auditory brainstem response (ABR) based hearing screening test. A hearing screening test, which is mandatory by law in many states, is applied to newborns to screen for hearing problems. In our experimental protocol, we used hearing screening test to create a controlled auditory stimulation in the temporal region of the infant brain. Local temporal activation has been observed as a response to auditory stimuli in fMRI studies in the literature[23].
All studies were carried out under an IRB approved protocol and informed consent was obtained prior to studies. In this study, a five day old 37 week estimated gestational age male infant underwent an ABR based hearing screening test while he was being monitored by the cwNIRS system. In an ABR based hearing screening test, EEG electrodes are attached with adhesive to the baby's scalp. While the baby sleeps, clicking sounds are applied through tiny earphones in the baby's ears. The test measures the brain's electrical activity in response to the sounds and displays the result as "passed" or "failed". The noninvasive procedure takes only a few minutes.
In our study, the hearing screening test was repeated twice. During the first test, the probe was placed on the temporal region of the brain which corresponds to the auditory cortex. In the second test, the probe was located on the forehead of the newborn for the control measurement (fig. 5). Baseline data were collected for 20 seconds before each test. After baseline, a hearing screening device from Natus®, Inc. (ALGO 3®) was used for auditory stimulation by applying 34 clicks per second until the infant passed the test. Both tests lasted approximately 2 minutes. The ALGO 3® hearing screening system displayed the results as "passed" for both tests which confirmed that we created auditory stimuli successfully. The entire procedure was repeated twice to validate the self-consistency of the data.
Average changes in the hemoglobin concentrations in addition to OXY and BV signals during stimulation with respect to baseline period are presented in figure 11. Error bars indicate the results of two successive experiments. An increase in both oxygenated and deoxygenated hemoglobin concentrations was observed in the temporal (test) region during auditory stimulation. Observed increase in blood volume was a result of blood rush to the local tissue. On the other hand, average deoxyhemoglobin concentration change was bigger than oxyhemoglobin concentration which was an indication of oxygen consumption as a result of the local activity. In the forehead (control) probe, only small fluctuations around the baseline were recorded which were uncorrelated with the stimulation. Observed changes are consistent with other NIRS based neonatal measurements in the literature [10,11].
Figure 11 Average changes in hemodynamic parameters in response to hearing screening test
IV. Conclusion
We present the design of a low-cost, battery operated continuous wave NIRS system that is intended for continuous bedside brain hemodynamic monitoring of newborns in neonatal intensive care units. Such a system can potentially assist clinicians in assessing functional changes in cerebral oxygenation and blood volume without removing the baby from the NICU.
Design parameters were defined and optimized for a safe and effective system performance. The most critical design issue was the trade-off between the temperature of the LED and the performance of the digital filter. SNR of 67 dB was obtained for a temporal resolution of 10 Hz. The system can be operated for 30 hours with an off-the-shelf 1850 mAh Li-Ion battery. Stray light rejection was satisfactory and inter-channel crosstalk of the channels was less than 0.1%.
Preliminary experiments performed both in the laboratory and in a clinical setting suggest that the system can be used to track functional changes of blood volume and oxygenation. This system can be used for bedside monitoring of neonates undergoing various clinical studies such as apnea, asphyxia, hypoglycemia, endotracheal suctioning, surfactant, aminophylline and indomethacin administrations.
Authors' contributions
AB and HR carried out the in-vivo and in-vitro experiments and AB drafted the manuscript. BO and AR conceived of the study, and participated in its design and coordination. All authors read and approved the final manuscript
Acknowledgements
We acknowledge, with thanks, Dr. Kambiz Pourrezaei, Frank Kepics, and Gunay Yurtsever for their helpful discussions on technical issues. We also extend our appreciations to Dr. Britton Chance and his team for allowing us to use their laboratory equipment and for helpful comments on the manuscript. Written and signed consent was obtained from parents of the newborn to publish figure 5 as a part of our article. This work has been sponsored in part by funds from the Defense Advanced Research Projects Agency (DARPA) Augmented Cognition Program and the Office of Naval Research (ONR), under agreement numbers N00014-02-1-0524 and N00014-01-1-0986.
==== Refs
Villringer A Chance B Non-invasive optical spectroscopy and imaging of human brain function Trends in Neurosciences 1997 20 435 42 9347608 10.1016/S0166-2236(97)01132-6
Strangman G Boas DA Sutton JP Non-invasive neuroimaging using near-infrared light Biological Psychiatry 2002 52 679 93 12372658 10.1016/S0006-3223(02)01550-0
Skov L Brun NC Greisen G Neonatal Intensive Care: Obvious Yet Difficult Area for Cerebral Near Infrared Spectroscopy Journal of Biomedical Optics 1997 2 7 14
Wyatt JS Cope M Delpy DT Wray S Reynolds EOR Quantification of cerebral oxygenation and haemodynamics in sick newborn infants by near infrared spectrophotometry Lancet 1986 8515 1063 1066 2877225 10.1016/S0140-6736(86)90467-8
Chance B Anday E Nioka S Zhou S Hong L Worden K Li C Murray T Ovetsky Y Pidikiti D Thomas R A novel method for fast imaging of brain function, non-invasively, with light Optics Express 1998 2 411 423
Siegel AM Marota JJA Boas DA Design and evaluation of a continuous-wave diffuse optical tomography system Optics Express 1999 4 287 298
Schmidt FEW Fry ME Hillman EMC Hebden JC Delpy DT A 32-channel time-resolved instrument for medical optical tomography Review of Scientific Instruments 2000 71 256 265 10.1063/1.1150191
du Plessis AJ Near-infrared spectroscopy for the in vivo study of cerebral hemodynamics and oxygenation Current Opinion in Pediatrics 1995 7 632 639 8776012
Meek JH Firbank M Elwell CE Atkinson J Braddick O Wyatt JS Regional hemodynamic responses to visual stimulation in awake infants Pediatric Research 1998 43 840 843 9621996
Sakatani K Chen S Lichty W Zuo H Wang Y Cerebral blood oxygenation changes induced by auditory stimulation in newborn infants measured by near infrared spectroscopy Early Human Development 1999 55 229 236 10463787 10.1016/S0378-3782(99)00019-5
Zaramella P Freato F Amigoni A Salvadori S Marangoni P Suppjei A Schiavo B Chiandetti L Brain auditory activation measured by near-infrared spectroscopy (NIRS) in neonates Pediatric Research 2001 49 213 219 11158516
Taga G Asakawa K Hirasawa K Konishi Y Hemodynamic responses to visual stimulation in occipital and frontal cortex of newborn infants: a near-infrared optical topography study Early Hum Dev 2003 75 S203 10 14693406 10.1016/j.earlhumdev.2003.08.023
Hintz SR Benaron DA Siegel AM Zourabian A Stevenson DK Boas DA Bedside functional imaging of the premature infant brain during passive motor activation Journal of Perinatal Medicine 2001 29 335 343 11565203 10.1515/JPM.2001.048
Cope M Delpy DT System for long-term measurement of cerebral blood flow and tissue oxygenation on newborn infants by infrared transillumination Med Biol Eng Comput 1998 26 289 294 2855531
Cope M The Development of a Near-Infrared Spectroscopy System and Its Application for Noninvasive Monitoring of Cerebral Blood and Tissue Oxygenation in the Newborn Infant 1991 University College London, London
Boas DA Franceschini MA Dunn AK Strangman G Frostig RD "Noninvasive Imaging of Cerebral Activation with Diffuse Optical Tomography." In Vivo Optical Imaging of Brain Function 2002 Boca Raton: CRC Press 193 221
Wray S Cope M Delpy DT Wyatt JS Reynolds EO Characterization of the near infrared absorption spectra of cytochrome aa3 and haemoglobin for the non-invasive monitoring of cerebral oxygenation Biochim Biophys Acta 1988 933 184 92 2831976
Van der Zee P Cope M Arridge SR Experimentally measured optical pathlengths for the adult's head, calf and forearm and the head of the newborn infant as a function of interoptode spacing Adv Exp Med Biol 1992 316 143 153 1288074
Hazinski MF Darovic GO Pediatric Evaluation and Monitoring Considerations Hemodynamic Monitoring: Invasive and Noninvasive Clinical Application 2002 Philadelphia, PA: W.B.Saunders 471 514
Benaron DA Kurth CD Steven JM Delivoria-Papadopoulos M Chance B Transcranial optical path length in infants by near-infrared phase-shift spectroscopy J Clin Monit 1995 11 109 17 7760083 10.1007/BF01617732
Yamashita Y Maki A Koizumi H Wavelength dependence of the precision of noninvasive optical measurement of oxy-, deoxy-, and total-hemoglobin concentration Med Phys 2001 28 1108 14 11439480
Bozkurt A Onaral B Safety assessment of near infrared light emitting diodes for diffuse optical measurements BioMedical Engineering OnLine 2004 3 9 15035670 10.1186/1475-925X-3-9
Anderson AW Marois R Colson ER Peterson BS Duncan CC Ehrenkranz RA Schneider KC Gore JC Ment LR Neonatal auditory activation detected by functional magnetic resonance imaging Magn Reson Imaging 2001 19 1 5 11295339 10.1016/S0730-725X(00)00231-9
| 15862131 | PMC1112605 | CC BY | 2021-01-04 16:37:37 | no | Biomed Eng Online. 2005 Apr 29; 4:29 | utf-8 | Biomed Eng Online | 2,005 | 10.1186/1475-925X-4-29 | oa_comm |
==== Front
Biomed Eng OnlineBioMedical Engineering OnLine1475-925XBioMed Central London 1475-925X-4-301586971210.1186/1475-925X-4-30Book ReviewReview of "Fractals and Chaos: The Mandelbrot Set and Beyond", by B. Mandelbrot Diaspro Alberto [email protected] Laboratory for Advanced Microscopy, Bioimaging and Spectroscopy (LAMBS), MicroScoBio Research Center, IFOM, Department of Physics, University of Genoa, Via Dodecaneso, 33 16146 Genoa Italy2005 3 5 2005 4 30 30 >Mandelbrot BB .
Fractals and Chaos. The Mandelbrot Set and Beyond .
New York: Springer . 2004 . 308 pages, ISBN 0-387-20158-0 . 29 3 2005 3 5 2005 Copyright © 2005 Diaspro; licensee BioMed Central Ltd.2005Diaspro; 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
Benoit Mandelbrot has produced a comprehensive, well-presented review of essential topics related to Mandelbrot set theory and applications. The last part of the title "The Mandelbrot set and beyond" fully describes its potential allowing the reader to navigate through pictures, hard-to-find early papers and important and effective chapters on the historical background. All chapters are assembled in a way that the overall mix becomes a very well integrated source of know-how and knowledge bringing the readers into the Mandelbrot set world. The spirit of the book is well summarized in a sentence on page 34: "When seeking new insights, I look, look, look, and play with many pictures. (One picture is never enough)." It is certainly true that in the last twenty years, mathematics has changed so deeply that to younger persons some chapter's stories might be simply incredible (p.36), as well, one should admit that after Mandelbrot's sets, initially describing trees, coastlines' shapes or allowing measuring the length of the Britain coast, and after the seminal book on "The Fractal Geometry of Nature" our way of looking at the world changed. Mandelbrot wrote: "Why is geometry often described as 'cold' and 'dry'? One reason lies in its inability to describe the shape of a cloud, a mountain, a coastline or a tree. Clouds are not spheres, mountains are not cones, coastlines are not circles, and bark is not smooth, nor does lightning travel in a straight line". I think our vision of the world, from the atom to the higher length scales, is still changing using those concepts clearly illustrated in the current Mandelbrot's book. Selected notes and papers make this book unique within the several books published on this topic. It is clear the touch of the author under all aspects: a touch of pure genius.
There are five main topics dominating the book, namely: Quadratic iteration and its Mandelbrot set – Quadratic Julia and Mandelbrot sets; Nonquadratic iterations – Nonquadratic rational dynamics; Kleinian groups' limit set – Iterated nonlinear function systems and the fractal limit sets of Kleinian groups; Multifractal invariant measures – Exponentially vanishing multifractal measures; Background and History. Cumulative bibliography is impressive and well done. It is clearly pointed out, following the pathway through the book, how fractal geometry played an important role in offering a quantitative tool in several areas. Circumstances and facts are put together also to bring important lessons for young scientists. The author made a serious and effective effort to realize a book that contains more than history, more than mathematics... it is a sort of ideal book for stimulating new ideas, new concepts, and new discoveries. So far, it is an excellent book also for supporting courses at University, PhD and Post doc level. Moreover, it is indispensable for scientists not only as a lesson of a pathway in science but also as an important source for science of tomorrow. This is a valuable reference source to researchers from these and related areas including bioengineering, biophysics, nanobiosciences and, of course, applied mathematics.
| 0 | PMC1112606 | CC BY | 2021-01-04 16:37:37 | no | Biomed Eng Online. 2005 May 3; 4:30 | utf-8 | Biomed Eng Online | 2,005 | 10.1186/1475-925X-4-30 | oa_comm |
==== Front
Cell Commun SignalCell communication and signaling : CCS1478-811XBioMed Central London 1478-811X-3-61583311110.1186/1478-811X-3-6ResearchComparable response of ccn1 with ccn2 genes upon arthritis: An in vitro evaluation with a human chondrocytic cell line stimulated by a set of cytokines Moritani Norifumi H [email protected] Satoshi [email protected] Toshio [email protected] Masaharu [email protected] Department of Biochemistry and Molecular Dentistry, Okayama University Graduate School of Medicine and Dentistry, Okayama 700-8525, Japan2 Department of Oral and Maxillofacial Reconstructive Surgery, Okayama University Graduate School of Medicine and Dentistry, Okayama, Japan2005 15 4 2005 3 6 6 29 3 2005 15 4 2005 Copyright © 2005 Moritani et al; licensee BioMed Central Ltd.2005Moritani 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 chondrosarcoma-derived HCS-2/8 has been known to be an excellent model of human articular chondrocytes. By mimicking the arthritic conditions through the treatment of HCS-2/8 cells with cytokines, we estimated the gene expression response of ccn1 and ccn2 during the course of joint inflammation in vitro.
Results
In order to mimic the initiation of inflammation, HCS-2/8 cells were treated with tumor necrosis factor (TNF)-α. To induce pro-inflammatory or reparative responses, TGF-β was employed. Effects of an anti-inflammatory glucocorticoid were also evaluated. After stimulation, expression levels of ccn1 and ccn2 were quantitatively analyzed. Surprisingly, not only ccn2, but also ccn1 expression was repressed upon TNF-α stimulation, whereas both mRNAs were uniformly induced by transforming growth factor (TGF)-β and a glucocorticoid.
Conclusion
These results describing the same response during the course of inflammation suggest similar and co-operative roles of these 2 ccn family members in the course of arthritis.
==== Body
Background
Osteoarthritis (OA) and rheumatoid arthritis (RA) are most common orthopaedic complexities among aged individuals [1]. Although they may be etiologically distinct each other, both are characterized by progressive destruction of articular cartilage, which is mainly caused by inflammatory stresses [1]. The most major problem in the therapeutics of these joint diseases is the difficulty in regenerating damaged articular cartilage. It is widely recognized that even a defect of 2 mm in diameter in normal articular cartilage may not be repaired naturally [2,3]. Therefore, cartilage regeneration is one of the most preferred subjects for the investigators in medical research field.
In our previous report, we for the first time clarified that CCN2/connective tissue growth factor (CTGF) was capable of regenerating full-thickness defect in articular cartilage and promoted the repair of damaged cartilage in an OA model in vivo [2]. Also, expression of ccn2 in OA cases was reported [3]. As such, it is now quite clear that CCN2 is involved in inflammatory response and repair process of articular cartilage during arthritis. Therefore, it is of critical importance to investigate the regulation of ccn2 expression along with the course of joint inflammation.
The CCN family is a novel group of proteins that act as multiple mediators among a variety of extracellular signaling molecules. This family of secreted proteins consists of 6 members: Cyr61 (CCN1), CTGF (CCN2), Nov (CCN3), Elm-1/WISP-1 (CCN4), rCop-1/WISP-2/CTGF-L (CCN5), and WISP-3 (CCN6) [4-9]. These structurally conserved proteins share 4 modules with sequence similarities to insulin-like growth factor-binding proteins, von Willebrand factor type C module, thrombospondin type 1 repeat, and growth factor cysteine knots, respectively [4-9]. CCN2/CTGF is a classical member of CCN family and shares significantly functional similarities with another member, CCN1/Cyr61. Both proteins share –45% amino acid sequence identity [10], bind heparin, are associated with the ECM [11], and exhibit remarkable functional versatility [12,13]. CCN1 and CCN2 can stimulate chemotaxis and promote proliferation of endothelial cells and fibroblasts in culture, induce neovascularization in vivo, and promote chondrogenic differentiation, the last-mentioned action being consistent with their expression in prechondrogenic mesenchyme during embryogenesis [5-7,13-15]. Considering such biological involvement of CCN1 in mesenchymal tissue formation, active roles of CCN1 in cartilage regeneration and repair may not be disregarded as well.
Therefore, in our present study, we assessed the gene regulation profile of CCN1 as well as CCN2 in a human chondrocytic cell line, HCS-2/8, upon the stimulation mimicking the course of chronic inflammation. This particular cell line was selected, since HCS-2/8 was established from a chondrosarcoma and has best retained a variety of mature chondrocytic phenotypes [16-18]. Using this in vitro model, we found CCN1 may also be involved in the inflammatory response in joints and may be useful in cartilage regenerative therapeutics, as described in CCN2.
Results
Strict and discriminating quantification of ccn1 and ccn2 mRNAs by real-time-RT-PCR
Since ccn1 and ccn2 are members of the same gene family, it is critically important to examine the specificity of each quantification system in order to avoid possible cross-recognition [4]. First, RT-PCR products of the initial amplification under our protocol were separated by 2 % agarose gel electrophoresis. All of the amplified LightCycler-PCR products showed single bands of the expected lengths (i.e., 219 bp for CCN1, 153 bp for CCN2, and 101 bp for β-actin; Fig. 1B). Second, we further verified the identity of each PCR product by direct nucleotide sequencing. As a result, sequence analysis of the PCR products showed 100 % homology to the published sequences. After these initial examinations, specificity of all of the products was confirmed each time by melting curve analysis via LightCycler Software 3.39 (Roche). Melting curves were analyzed for all PCR runs. Continuous fluorescence monitoring while slowly elevating the temperature resulted in a sudden decrease in SYBR green I (Roche) fluorescence intensity, when denaturation of the PCR product occurred. Thus, the melting curve analysis revealed a specific pattern for each target. Figure 1C shows an example of a melting curve analysis, indicating no nonspecific PCR product. As such, accurate and distinctive quantification of ccn1 and ccn2 was deemed warranted.
Figure 1 Distinctive quantification of ccn1 and ccn2 mRNA by real-time RT-PCR. A. Primers used for real-time PCR and the structures of human ccn1 and ccn2 mRNAs. Schematic representations are illustrated in reference to the modular structure of human CCN1 and CCN2 (stippled boxes). The small open circle and AAAAA at the left and right ends denote the 5'-cap structure and poly-A tail, respectively. Names, locations for recognition, nucleotide sequence of the primers, and the expected sizes of the PCR products are given. Abbreviations: IGFBP, insulin-like growth factor binding module; VWC, von Willebrand factor type C module; TSP1, thrombospondin type 1 repeat; CT, C-terminal module. B. The CCN1 (219 bp), CCN2 (153 bp), and β-actin (101 bp) PCR products amplified by LightCycler were analyzed by agarose electrophoresis. C. Melting curve analysis of the RT- PCR products of ccn1 and ccn2. Melting curves were acquired after 45 cycles of amplification. Melting curve pattern is displayed on the left panel, where fluorescent intensity (F1) from SYBR green is plotted against temperature. Melting peak pattern can be found on the right panel, in which the decrement of F1 is plotted against temperature.
Repressive response of both ccn1 and ccn2 genes upon inflammatory provocation by TNF-α
TNF-α is one of the best-known inflammatory cytokines and is involved in a number of inflammatory diseases including arthritis [1]. Therefore, we evaluated the effect of TNF-α on the expression of the ccn1 and ccn2 genes in the chondrocytic HCS-2/8 cells. HCS-2/8 cells were treated with 10 nM TNF-α for 12 – 24 hours. By this stimulation, ccn2 mRNA levels were repressed 0.6 (0.64) to 0.4 (0.41) fold, and ccn1 mRNA levels, 0.9 (0.89) to 0.7 (0.66) fold, by 24 h (Fig. 2A). The result that both genes were uniformly downregulated upon TNF-α stimulation strongly suggests similar or complementary functions of these gene products in the middle of inflammatory process.
Figure 2 Coordinative repression and induction of ccn1 and ccn2 transcripts in response to inflammatory TNF-α (A) and pro-inflammatory TGF-β (B) in chondrocytic HCS-2/8 cells. Results of mRNA quantification by real-time RT-PCR analysis are displayed. Plotted values represent relative mRNA copy numbers versus those at time 0. Each copy number was computed by standardizing the raw data against those of β-actin. Open circles, ccn1; solid circles, ccn2. Mean values and standard deviations (SDs) of at least 2 experiments are shown.
Induction of ccn1 and ccn2 by pro-inflammatory and regenerative signal in chondrocytic HCS-2/8 cells
Tissue repair and regeneration are the last step of inflammation, in which damaged tissue is initially filled with fibroblasts surrounded by a vast amount of extracellular matrices (ECM). However, in general, inflammatory stimulation and ECM formation occur in parallel along the course of chronic inflammation. Through such process, TGF-β and CCN2 is known to critically regulate ECM deposition from a variety of mesenchymal cells. Molecular regulation of ccn2 by TGF-β has been relatively investigated [19,20]. Nevertheless, little is known concerning the interplay between ccn1 and TGF-β in cartilage. As such, we evaluated the response of ccn1 gene in chondrocytic HCS-2/8 cells to TGF-β, in comparison with ccn2. Treatment with 10 ng/ml TGF-β induced ccn2 mRNA up to 1.5 to 1.9 fold; and ccn1 mRNA, by 1.5 fold up to 24 h (Fig. 2B). Additionally, we found that these mRNAs were better induced by 50 ng/ml TGF-β than by 10 ng/ml (data not shown). These findings suggest coordinative functions of ccn1 and ccn2 in repair and chronic inflammation of articular cartilage.
Anti-inflammatory treatment and expression of ccn1 and ccn2 genes
Glucocorticoids are known to possess strong anti-inflammatory effects and also to be involved in the control of cartilage metabolism [21]. Particularly, the effectiveness of glucocorticoids on RA symptom is so prominent that it is frequently applied clinically, despite its adverse systemic effects. It is already known that ccn2 gene expression is induced by glucocorticoids in chondrocytic cells as well as in fibroblasts [22,23]. In contrast, molecular interaction between glucocorticoids and ccn1 gene in chondrocytes has not been investigated. Thus, we performed Northern blot analysis, as well as the real-time RT-PCR quantification to estimate the effects of glucocorticoids on ccn1 and ccn2 gene expression. When HCS-2/8 cells were treated with 50 nM dexamethasone for 2.5 – 5 hours and then examined by real-time PCR, ccn2 mRNA was induced by 1.8 to 2.3 fold, and ccn1 mRNA was similarly induced by 1.5 to 1.9 fold up to 5 h (Fig. 3A). The results obtained by Northern blot analysis were quite similar to the PCR ones (Fig 3B).
Figure 3 Comparable response of ccn1 and ccn2 to anti-inflammatory dexamethasone in HCS-2/8 cells. A. Expression levels of ccn1 and ccn2 transcripts examined by quantitative real-time PCR. Plotted values represent relative mRNA copy numbers, as described in the legend to Figure 2. Open and solid circles denote ccn1 and ccn2, respectively. B. Control experiments with estrogen, showing no significant alteration in gene expression either in ccn1, or ccn2. These data are mean values of 2 experimental results shown with error bars (SD). C. Northern blot analysis of ccn1 and ccn2 gene expression in HCS-2/8 cells in response to dexamethasone treatment. The ccn1 and ccn2 mRNAs (upper and middle panels) and total RNAs stained with 0.02 % methylene blue (lower panel) are shown. : 28S, 28S ribosomal RNA; 18S, 18S ribosomal RNA.
As a control experiment, we carried out similar analysis with 17β-estradiol, which is another steroid hormone and is also one of the regulators to maintain the homeostasis of connective tissue. However, treatment of HCS-2/8 cells with 10 nM estrogen for 1 – 24 hours resulted in no significant change in ccn1 and ccn2 mRNA expression levels. In fact, no effects were observed even up to the concentration of 100 nM (data not shown).
Discussion
In this study, we comparatively analyzed the expression profiles of ccn1 and ccn2, while simulating the course of arthritis; i.e., inflammation, tissue regeneration and anti-inflammatory treatment, utilizing a human chondrocytic HCS-2/8. In advance to the evaluation, we established a real-time PCR quantification method by using a LightCycler system. In view of the data provided for sensitivity, linearity, and reproducibility, this assay system accurately allowed the discriminating quantification of these mRNAs from the same gene family. The total reliability of this system was evident, as represented by the facts that every specific primer set produced distinct and specific PCR products (Fig. 1) and quantitative results were comparable to those of Northern blot analysis (Fig. 3).
The involvement of CCN2 in arthritic diseases has been indicated. In fact, expression of ccn2 gene in clinical cases was reported [3]. Also in an experimental OA model, distinct induction of ccn2 gene expression was observed. These findings are regarded as a regenerative response of damaged cartilage, since exogenously applied CCN2 was proven to be effective in cartilage regeneration. In rat models, CCN2 captured in collagen hydrogel to allow gradual release of this factor efficiently promoted the regeneration of full-thickness cartilage defect and experimentally induced OA cartilage [2]. Therefore, expression profile of ccn2 in chondrocytes along the time course of inflammation is thought to represent the proper gene regulation to provide a regenerative molecule during arthritis, and thus itself is worth investigated. More interestingly, ccn1 gene expression exactly followed the fluctuation pattern of ccn2 gene expression upon any kind of stimulation tested. These results clearly indicate that not only CCN2, but also CCN1 may be provided as a regenerative molecule in arthritis. This hypothesis is supported by a number of their functional similarities. Indeed, these factors are associated with the ECM, stimulate chemotaxis and promote proliferation of endothelial cells and fibroblasts, and promote neovascularization and chondrogenic differentiation [7-15,24]. Considering such similarities and the concomitant fluctuation of gene expression upon inflammation together, CCN1 is expected to be one of the useful molecular tools to promote cartilage regeneration. In order to examine this hypothesis, it is necessary to evaluate the regenerative potential of CCN1 protein in damaged articular cartilage, as was examined with CCN2. In vivo evaluation of the expression of ccn1 upon OA and RA and the effects of CCN1 protein on cartilage regeneration is currently in progress. Since all of the CCN family members are thought to be mediators of multiple signaling molecules, therapeutic utility of another member, such as CCN3/NOV is also expected and obviously, need to be explored.
Conclusion
In vitro simulation of arthritis with a human chondrocytic cell line revealed the same response pattern of ccn1 as that of ccn2, which is known as a regenerative mediator in cartilage repair. Together with similar functionalities of CCN1 and CCN2 in mesenchymal tissues, these results suggest possible utility of CCN1 in regenerative therapy of damaged mesenchymal tissues.
Methods
Materials
TNF-α and TGF-β1 were purchased from Promega (Madison, WI, USA). Dexamethasone and estrogen (17β-estradiol) were purchased from Sigma (St. Louis, MO,USA).
Cell culture
HCS-2/8 cells, a chondrocytic cell line derived from a well-differentiated type of human chondrosarcoma [19], were maintained in Dulbecco's modified Eagle's medium (D-MEM) supplemented with 10 % fetal bovine serum (FBS) under an atmosphere of humidified air containing 5 % CO2. In the experiments in which the effect of estrogen was studied, the medium was replaced with phenol red-free DMEM and 2 mM glutamine (Nissui Pharmaceutical Co. Ltd., Tokyo, Japan) containing 2 % charcoal-treated FBS, after the HCS-2/8 cells had become subconfluent. In the experiments with TGF-β1, dexamethasone, and estrogen, the cells were harvested after the treatment of subconfluent cells (simulating growing phase upon regeneration) with each factor for the desired time periods. In the TNF-α experiment, the cells were harvested after the treatment of confluent cells (simulating quiescent phase before inflammatory damage) with the factor for the desired time periods.
RNA extraction and reverse transcription (RT)
Total RNA was extracted from HCS-2/8 cells by the acid guanidinium phenol-chloroform method previously described [25]. Reverse transcription by avian myelosarcoma virus reverse transcriptase was carried out by using a commercially available kit (Takara Shuzo, Tokyo, Japan) and 1.0 μg of total RNA. Then, the samples were diluted by 20-fold with RNase-free H2O for subsequent quantification.
Quantitative real-time PCR amplification
On the basis of the published cDNA sequences of CCN2/CTGF (GenBank accession no. NM_001901) and CCN1/Cyr61 (no. AF031385), specific primers were designed for each. Their nucleotide sequences are displayed in Fig. 1. Real-time quantitative PCR was performed with a LightCycler system (Roche Molecular Biochemicals, Mannheim, Germany). For each assay, reaction mixtures containing 2.0 μl of a cDNA pool, 1.0 μl of LC DNA Master SYBR Green I mixture (Roche), 50 ng of the primers, and 0.8 μl of 25 mM MgCl2 were prepared on ice. After the reaction mixtures had been loaded into glass capillary tubes, amplification was performed under the following cycling conditions: initial denaturation at 95°C for 10 min, followed by 45 cycles of denaturation at 95°C for 15 sec, annealing at 55°C for 10 sec, and extension at 72°C for 10 sec. The temperature transition rate was set at 20°C/ sec. The fluorescence representing double-strand DNA formation was measured in single-acquisition mode at 72°C after each cycle. For each sample, the cDNA copy numbers of the target and an internal control (β-actin) genes were determined based on calibration curves (see below). The relative amount of the target cDNA was then computed by dividing the copy number by that of the internal control to obtain a normalized value. Separate calibration curves for ccn1, ccn2, and β-actin were prepared with serially diluted plasmid DNAs containing the target sequences, which were amplified and evaluated simultaneously in each assay. To distinguish specific signal from non-specific products, melting curve analysis was performed after each amplification cycle. Samples were maintained at 63°C for 10 sec, and then the temperature was gradually increased to 95°C at a rate of 0.1°C/sec, while the signals were monitored with a step-acquisition mode, as described previously [26]. The real-time PCR analysis condition was optimised to a level without non-specific amplification under an acceptable PCR efficiency (a slope ranging from -2.9 to -4.5). To verify the melting curve results, we analyzed representative PCR samples in 2.0 % agarose gels, purified and directly sequenced them from both directions by an automated DNA sequencer (ABI prism 310 Genetic Analyzer; Applied Biosystems, Foster City, CA, USA).
Northern blot analysis
Ten micrograms of total RNA was electrophoresed in formaldehyde agarose gel and subsequently blotted onto a nylon membrane. For hybridization probes, CCN2/CTGF and CCN1/Cyr61 cDNA fragments were prepared by RT-PCR with pairs of specific primers. Primers specific for CCN2/CTGF [27] and CCN1/Cyr61 [28] were described previously. These PCR products were radiolabeled and used for hybridization as described earlier [20].
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
NHM performed molecular biological studies and prepared most part of data. SK arranged the constitution of the entire work and drafted the manuscript. TS participated in the design of the work. MT participated in coordination and drafting the manuscript and provided general support.
Acknowledgements
This work was supported in part by grants from the program Grants-in-Aid for Scientific Research (C) (to S.K.) and (S) (to M.T.) and Exploratory Research (to M.T.) from the Ministry of Education, Culture, Sports, Science and Technology of Japan; the Naito Foundation (to M.T.); the Nakatomi Health Science Foundation (to S.K., M.T.); Ryobi-teien Memorial Foundation (to S.K.); the Foundation for Growth Science in Japan (to M.T.); and the Sumitomo Foundation (to M.T.). The authors wish to thank Drs. Tohru Nakanishi, Takashi Nishida, Takanori Eguchi, Seiji Kondo, Hideyuki Doi, and Kyouji Nakao for helpful suggestions and discussions; Mr. Shosai Moritani and Ms. Hiroko Moritani for overall assistance; and Ms. Yuki Nonami for secretarial help..
==== Refs
Reines BP Is rheumatoid arthritis premature osteoarthritis with fetal-like healing? Autoimmun Rev 2004 3 305 311 15246027 10.1016/j.autrev.2003.11.002
Nishida T Kubota S Kojima S Kuboki T Nakao K Kushibiki T Tabata Y Takigawa M Regeneration of defects in articular cartilage in rat knee joints by CCN2 (connective tissue growth factor) J Bone Miner Res 2004 19 1308 1319 15231019
Omoto S Nishida K Yamaai Y Shibahara M Nishida T Doi T Asahara H Nakanishi T Inoue H Takigawa M Expression and localization of connective tissue growth factor (CTGF/Hcs24/CCN2) in osteoarthritic cartilage Osteoarthritis Cartilage 2004 12 771 778 15450526 10.1016/j.joca.2004.06.009
Bork P The modular architecture of a new family of growth regulators related to connective tissue growth FEBS Lett 1993 327 125 130 7687569 10.1016/0014-5793(93)80155-N
Brigstock DR The connective tissue growth factor/cysteine-rich 61/nephroblastoma overexpressed (CCN) family Endocr Rev 1999 20 189 206 10204117 10.1210/er.20.2.189
Lau LF Lam SC The CCN family of angiogenic regulators: the integrin connection Exp Cell Res 1999 248 44 57 10094812 10.1006/excr.1999.4456
Takigawa M Nakanishi T Kubota S Nishida T The role of CTGF/Hcs24/ecogenin in skeletal growth control J Cell Physiol 2003 194 256 266 12548546 10.1002/jcp.10206
Takigawa M CTGF/Hcs24 as a multifunctional growth factor for fibroblasts, chondrocytes and vascular endothelial cells Drug News Perspect 2003 16 11 21 12682668 10.1358/dnp.2003.16.1.829302
Perbal B CCN proteins: multifunctional signalling regulators Lancet 2004 363 62 4 14723997 10.1016/S0140-6736(03)15172-0
Brunner A Chinn J Neubauer M Purchio AF Identification of a gene family regulated by transforming growth factor-beta DNA Cell Biol 1991 10 293 300 2029337
Kubota S Eguchi T Shimo T Nishida T Hattori T Kondo S Nakanishi T Takigawa M Novel mode of processing and secretion of connective tissue growth factor/ecogenin (CTGF/Hcs24) in chondrocytic HCS-2/8 cells Bone 2001 29 155 161 11502477 10.1016/S8756-3282(01)00492-6
O'Brien TP Yang GP Sanders L Lau LF Expression of cyr61, a growth factor-inducible immediate-early gene Mol Cell Biol 1990 10 3569 77 2355916
O'Brien TP Lau LF Expression of the growth factor-inducible immediate early gene cyr61 correlates with chondrogenesis during mouse embryonic development Cell Growth Differ 1992 3 645 654 1419914
Ryseck RP Macdonald-Bravo H Mattei MG Bravo R Structure, mapping, and expression of fisp-12, a growth factor-inducible gene encoding a secreted cysteine-rich protein Cell Growth Differ 1991 2 225 33 1888698
Wong M Kireeva ML Kolesnikova TV Lau LF Cyr61, product of a growth factor-inducible immediate-early gene, regulates chondrogenesis in mouse limb bud mesenchymal cells Dev Biol 1997 192 492 508 9441684 10.1006/dbio.1997.8766
Takigawa M Tajima K Pan HO Enomoto M Kinoshita A Suzuki F Takano Y Mori Y Establishment of a clonal human chondrosarcoma cell line with cartilage phenotypes Cancer Res 1989 49 3996 4002 2736538
Takigawa M Pan HO Kinoshita A Tajima K Takano Y Establishment from a human chondrosarcoma of a new immortal cell line with high tumorigenicity in vivo, which is able to form proteoglycan-rich cartilage-like nodules and to respond to insulin in vitro Int J Cancer 1991 48 717 725 2071232
Moritani NH Kubota S Eguchi T Fukunaga T Yamashiro T Takano-Yamamoto T Tahara H Ohyama K Sugahara T Takigawa M Interaction of AP-1 and the ctgf gene: a possible driver of chondrocyte hypertrophy in growth cartilage J Bone Miner Metab 2003 21 205 10 12811624
Mori T Kawara S Shinozaki M Hayashi N Kakinuma T Igarashi A Takigawa M Nakanishi T Takehara K Role and interaction of connective tissue growth factor with transforming growth factor-beta in persistent fibrosis: A mouse fibrosis model J Cell Physiol 1999 181 153 159 10457363 10.1002/(SICI)1097-4652(199910)181:1<153::AID-JCP16>3.0.CO;2-K
Nakanishi T Kimura Y Tamura T Ichikawa H Yamaai Y Sugimoto T Takigawa M Cloning of a mRNA preferentially expressed in chondrocytes by differential display-PCR from a human chondrocytic cell line that is identical with connective tissue growth factor (CTGF) mRNA Biochem Biophys Res Commun 1997 234 206 210 9168990 10.1006/bbrc.1997.6528
Takano T Takigawa M Suzuki F Stimulation by glucocorticoids of the differentiated phenotype of chondrocytes and the proliferation of rabbit costal chondrocytes in culture J Biochem (Tokyo) 1985 97 1093 1100 4030717
Dammeier J Beer HD Brauchle M Werner S Dexamethasone is a novel potent inducer of connective tissue growth factor expression. Implications for glucocorticoid therapy J Biol Chem 1998 273 18185 18190 9660779 10.1074/jbc.273.29.18185
Kubota S Moritani NH Kawaki H Mimura H Minato M Takigawa M Transcriptional induction of connective tissue growth factor/hypertrophic chondrocyte-specific 24 gene by dexamethasone in human chondrocytic cells Bone 2003 33 694 702 14555275 10.1016/S8756-3282(03)00227-8
Nakanishi T Nishida T Shimo T Kobayashi K Kubo T Tamatani T Tezuka K Takigawa M Effects of CTGF/Hcs24, a product of a hypertrophic chondrocyte-specific gene, on the proliferation and differentiation of chondrocytes in culture Endocrinology 2000 141 264 73 10614647 10.1210/en.141.1.264
Moritani NH Kubota S Nishida T Kawaki H Kondo S Sugahara T Takigawa M Suppressive effect of overexpressed connective tissue growth factor on tumor cell growth in a human oral squamous cell carcinoma-derived cell line Cancer Lett 2003 192 205 214 12668285 10.1016/S0304-3835(02)00718-8
Woo TH Patel BK Cinco M Smythe LD Symonds ML Norris MA Dohnt MF Real-time homogeneous assay of rapid cycle polymerase chain reaction product for identification of Leptonema illini Anal Biochem 1998 259 112 117 9606151 10.1006/abio.1997.2532
Moritani NH Kubota S Nishida T Kawaki H Kondo S Sugahara T Takigawa M Suppressive effect of overexpressed connective tissue growth factor on tumor cell growth in a human oral squamous cell carcinoma-derived cell line Cancer Lett 2003 192 205 214 12668285 10.1016/S0304-3835(02)00718-8
Babic AM Kireeva ML Kolesnikova TV Lau LF CYR61, a product of a growth factor-inducible immediate early gene, promotes angiogenesis and tumor growth Proc Natl Acad Sci USA 1998 95 6355 6360 9600969 10.1073/pnas.95.11.6355
| 15833111 | PMC1112607 | CC BY | 2021-01-04 16:39:15 | no | Cell Commun Signal. 2005 Apr 15; 3:6 | utf-8 | Cell Commun Signal | 2,005 | 10.1186/1478-811X-3-6 | oa_comm |
==== Front
Comp HepatolComparative Hepatology1476-5926BioMed Central London 1476-5926-4-51587173910.1186/1476-5926-4-5Case ReportHypervascular nodule in a fibrotic liver overloaded with iron: identification of a premalignant area with preserved liver architecture Sá Cunha António [email protected] Jean-Frédéric [email protected] Hervé [email protected] Ledinghen Victor [email protected] Charles [email protected] Paulette [email protected] Fédération d'hépato-gastroentérologie CHU Bordeaux, GREF Inserm E362, Université Bordeaux 2, France2005 4 5 2005 4 5 5 21 12 2004 4 5 2005 Copyright © 2005 Sá Cunha et al; licensee BioMed Central Ltd.2005Sá Cunha 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 presence of a hypervascular nodule in a patient with cirrhosis is highly suggestive of a hepatocellular carcinoma.
Case presentation
A 55 year old man with idiopathic refractory anaemia was addressed for the cure of a recently appeared 3.3 cm hypervascular liver nodule. The nodule was not visible on the resected fresh specimen, but a paler zone was seen after formalin fixation. The surrounding liver was fibrotic (METAVIR score F3) and overloaded with iron. However, the paler zone, thought to be the nodule, had in fact a normal architecture, was less fibrotic, and contained some "portal tract-like structures" (but with arteries only); moreover, this paler area was devoid of iron, contained less glycogen and was characterized by foci of clear hepatocytes.
Conclusion
In spite of the absence of architectural distortion, and a normal proliferative index, the possibility of premalignancy or malignancy should be considered in this type of hypervascular and hyposiderotic nodule, occurring in the context of an iron overloaded liver.
==== Body
Background
The presence of a hypervascular nodule in a patient with liver disease is highly suggestive of a hepatocellar carcinoma (HCC) [1]. Increased iron stores in patients with HCC developed on a non-cirrhotic liver is well documented [2-5]; iron stores are seldom depleted at the time of the discovery of the HCC [6]. Few cases of premalignant nodules associated with HCC have also been reported under these circumstances [7,8]. In a fibrotic liver overloaded with iron, we report a case of a hypervascular and hyposiderotic nodule with premalignant features, but with a normal architecture.
Case presentation
General data
A 55 year old man with idiopathic refractory anaemia was addressed to our Unit for the cure of a recently appeared 3.3 cm hypervascular liver nodule in segment II (November, 2003). Physical examination was normal, including BMI. Liver function tests were as follows: ASAT = 53 IU/L (Normal = 40); ALAT = 58 IU/L (N = 50); Billirubin = 44 μmol/L (N = 17); PT = 70% (N = 70–100); V = 65% (N = 70–100); RBC = 2.9 × 106 cells/μl; Ht = 22.5% and Hb = 7.3 gm/dl; WBC = 4.6 × 103 cells/μl; and platelets = 210 × 109 /l. Ferritinemia was 1891 ng/l (N < 300), transferrin saturation was 100% (N < 40), iron concentration was 290 μmol/g (assessed by MRI, N < 36). AFP in blood was within the normal range. The patient, of Italian origin, was C282 Y -/-, H63D +/-, and S65C -/-, with no family history of iron overload. Markers for viral and autoimmune diseases were negative. Blood glucose was normal. He used to smoke 30 cigarettes per day; but had stopped for the last 8 years. He drank alcohol only occasionally. The treatment of his refractory anaemia consisted in blood transfusion (total of 10 packs), Desferoxamine, and Deferiprone.
Imaging results
Ultrasound examination showed a hypoechogenic ovoid nodule (3.2 × 1.9 cm) in segment II. MRI showed a hypointense non-tumoral liver on T1- and T2-weighted images due to iron overload. In comparison, the nodule was hyperintense on T1- and T2-weighted images. After gadolinium injection, the nodule was hyperintense on T1-weigthed images and remained so in the portal phase (Fig. 1).
Figure 1 Hyperintensity of the nodule on T1- and T2-weighted images (left), which remains so after gadolinium injection (right), on T1-weigthed image in the portal phase.
Liver pathology
On December 2003, a left lobe hepatectomy was performed laparoscopically. Follow-up was uneventful. The resected specimen was carefully sliced but no nodule was found on the fresh specimen, and in the expected area. However, after formalin fixation, a 2 cm in diameter paler area was identified (Fig. 2, arrow). All slices were routinely processed. The following stainings were performed: H&E, trichrome, Perls, reticulin, PAS, and several immunostains (CD34, cytokeratins 7 and 19, CRBP1 and α-SMA; the latter for identification of quiescent and/or activated hepatic stellate cells [9]). The liver was fibrotic (METAVIR score F3) (Figs. 3a, 4a) with an iron overload 3 + (according to Searle score), mainly in hepatocytes of zones 1 and 2 (Fig. 5a). Liver iron concentration was 286 μmol/g (n < 36), and the iron concentration / age ratio was 5.1. Small foci of clear cells devoid of iron were also observed (not shown).
Figure 2 Formalin fixed specimen: a flat and slightly clearer area is visible in the expected zone (arrow).
The paler zone, poorly limited from the adjacent parenchyma, was strikingly different. The architecture was preserved but the area was far less fibrotic (METAVIR score F1; Figs. 3b, 4b), with less iron (Fig. 5b), less glycogen (Fig. 6a), and with foci of clear hepatocytes (Figs. 6a, 6b). In these clear areas, hepatocytes were slightly bigger and occasionally displayed in two cell-thick plates. In these areas, as elsewhere, reticulin network, as well as Ki-67 (Mib-1) and CD34 were normal (not shown). One of the most striking findings was the presence of different types of portal tracts: some were normal (Fig. 7), whereas others contained mainly ductules (Fig. 8) and others arteries (Fig. 9). Regarding the number of CRBP1 and α-SMA positive cells, no obvious differences were seen between the fibrotic and non-fibrotic parts of the liver. CRBP 1 positive cells seemed to contain few lipid droplets.
Figure 3 Septal fibrosis in non tumoral liver (a), contrasting with absence of fibrosis in the nodule (b). Masson's trichrome.
Figure 4 Septal fibrosis in non tumoral liver (a), contrasting with absence of fibrosis in the nodule (b). Reticulin staining.
Figure 5 Iron overload in non tumoral liver (a), contrasting with less iron in the nodule (b) Perls staining.
Figure 6 PAS staining: (a) foci of clear hepatocytes (arrow) close to the border between the non tumoral PAS positive zone, on the left side, and the PAS negative nodule on the right side of the photograph; (b) a clear focus in the nodule.
Figure 7 Normal portal tract in the nodule. H&E staining.
Figure 8 On the left side, ductular reaction around a portal tract in the nodule. On the right side, in another portal tract the bile duct is visible but not the portal vein and the artery. CK7 immunostaining.
Figure 9 An unpaired artery in the nodule. α-SMA immunostaining.
Discussion
The mechanism accounting for the major hepatic iron overload is possibly multifactorial, including refractory sideroblastic anemia and blood transfusion, although the patient received only a limited number (<10) of blood transfusions. An associated hereditary iron overload such as transferrin receptor 2 haemochromatosis in this Italian patient cannot be ruled out.
Premalignant lesions have previously been described in iron overloaded patients in the absence of cirrhosis, although these lesions were discovered in the clinical context of a HCC [7,8]. To our knowledge, this is the first reported case of a premalignant area mimicking by imaging a HCC, but exhibiting microscopically a still well-preserved architecture, in an otherwise fibrotic liver.
The hyperarterialized nodule did not correspond to a macroscopically visible nodule, but rather to an ill-defined area with preserved architectural organization. However, this area was considered as premalignant based on the following arguments: arterial hypervascularization with isolated arteries in the parenchyma [3]; loss of iron [10], and of glycogen; and presence of clear hepatocytes foci [11]. The diagnosis of a focal nodular hyperplasia (FNH)-like nodule as described in cirrhosis [12-14], particularly related to alcohol, seems unlikely due to the loss of iron and to the presence of clear hepatocytes foci. Nonetheless, that diagnosis cannot be ruled out, and should always be kept in mind, especially if a liver transplantation is foreseen. Recently, it has been reported that coexisting iron overload could significantly worsen the course of FNH [15]. Unfortunately, and because no frozen material of the lesion (which was not visible) was available, in this case no specific molecular studies could be carried out to settle that important issue.
Our case strengthens previous observations [7,8] showing that malignancy can overrun cirrhosis in iron overload (Fig. 10). A minor degree (stage) of fibrosis in areas devoid of iron could be a direct consequence of iron loss (less toxicity) and/or related to the malignant process [16,17] in its early phase (as denoted by absence of cellular disorganization, and negativity of the MIB-1 immunostaining). It was concluded that this patient needs a strict surveillance because he may be at risk of recurrence. Indeed, clear hepatocytes foci devoid of iron were also observed outside the nodule [10].
Figure 10 Pathways leading to HCC in iron overload: (a) classical pathway; (b) alternate pathway (rarely observed).
Conclusion
The possibility of premalignancy or malignancy should be considered even in the absence of cirrhosis, when a nodule is observed in a patient with past or present liver iron overload.
List of abbreviations used
AFP – α-fetoprotein; ALAT – alanine aminotransferase; ASAT – aspartate aminotransferase; BMI – body-mass index; CRBP1 – cellular retinol-binding protein 1; FNH – focal nodular hyperplasia; Hb – hemoglobin; Ht – hematocrit; HCC – hepatocellar carcinoma; MRI – magnetic resonance imaging; PAS – periodic acid Schiff; PT – protrombine time; RBC – red blood cells; α-SMA – α-smooth muscle actin; WBC – white blood cells.
Authors' contributions
A Sá Cunha performed the surgery. JF Blanc collected the references and contributed to the writing. H Trillaud reviewed the MRI. V De Ledinghen, hepatologist, was in charge of the patient. C Balabaud wrote the paper. P Bioulac-Sage interpreted the liver histology and contributed to the writing. All authors read and approved the final manuscript.
==== Refs
Taouli B Losada M Holland A Krinsky G Magnetic resonance imaging of hepatocellular carcinoma Gastroenterology 2004 127 S144 52 15508078
Turlin B Juguet F Moirand R Le Quilleuc D Loreal O Campion JP Launois B Ramee MP Brissot P Deugnier Y Increased liver iron stores in patients with hepatocellular carcinoma developed on a noncirrhotic liver Hepatology 1995 22 446 450 7635411 10.1016/0270-9139(95)90564-2
Kowdley KV Iron, hemochromatosis, and hepatocellular carcinoma Gastroenterology 2004 127 S79 86 15508107
Bioulac-Sage P Le Bail BL Winnock M Balabaud C Bemard C Blanc JF Saric J Occurrence of hepatocellular carcinoma in nonfibrotic livers Hepatology 2000 32 1411 1412 11186870
Blanc JF De Ledinghen V Bernard PH de Verneuil H Winnock M Le Bail B Carles J Saric J Balabaud C Bioulac-Sage P Increased incidence of HFE C282Y mutations in patients with iron overload and hepatocellular carcinoma developed in non-cirrhotic liver J Hepatol 2000 32 805 811 10845668 10.1016/S0168-8278(00)80250-0
Pellise M Gonzalez-Abraldes J Navasa M Miquel R Bruguera M [Hepatocellular carcinoma in a patient with hereditary hemochromatosis without cirrhosis] Gastroenterol Hepatol 2001 24 132 134 11261224
Blanc JF De Ledinghen V Trimoulet P Le Bail B Bernard PH Saric J Balabaud C Bioulac-Sage P Premalignant lesions and hepatocellular carcinoma in a non-cirrhotic alcoholic patient with iron overload and normal transferrin saturation J Hepatol 1999 30 325 329 10068114 10.1016/S0168-8278(99)80080-4
Attia A Blanc JF Saric J Balabaud C Bioulac-Sage P [Premalignant lesions and hepatocellular carcinoma on non cirrhotic liver overloaded with iron] Gastroenterol Clin Biol 2000 24 955 959 11084432
Lepreux S Bioulac-Sage P Gabbiani G Sapin V Housset C Rosenbaum J Balabaud C Desmouliere A Cellular retinol-binding protein-1 expression in normal and fibrotic/cirrhotic human liver: different patterns of expression in hepatic stellate cells and (myo)fibroblast subpopulations J Hepatol 2004 40 774 780 15094224 10.1016/j.jhep.2004.01.008
Deugnier YM Charalambous P Le Quilleuc D Turlin B Searle J Brissot P Powell LW Halliday JW Preneoplastic significance of hepatic iron-free foci in genetic hemochromatosis: a study of 185 patients Hepatology 1993 18 1363 1369 7902316 10.1016/0270-9139(93)90225-C
Su Q Bannasch P Relevance of hepatic preneoplasia for human hepatocarcinogenesis Toxicol Pathol 2003 31 126 133 12597457 10.1080/01926230309732
Libbrecht L Bielen D Verslype C Vanbeckevoort D Pirenne J Nevens F Desmet V Roskams T Focal lesions in cirrhotic explant livers: pathological evaluation and accuracy of pretransplantation imaging examinations Liver Transpl 2002 8 749 761 12200773 10.1053/jlts.2002.34922
Nakashima O Kurogi M Yamaguchi R Miyaaki H Fujimoto M Yano H Kumabe T Hayabuchi N Hisatomi J Sata M Kojiro M Unique hypervascular nodules in alcoholic liver cirrhosis: identical to focal nodular hyperplasia-like nodules? J Hepatol 2004 41 992 998 15582133 10.1016/j.jhep.2004.08.014
Quaglia A Tibballs J Grasso A Prasad N Nozza P Davies SE Burroughs AK Watkinson A Dhillon AP Focal nodular hyperplasia-like areas in cirrhosis Histopathology 2003 42 14 21 12493020 10.1046/j.1365-2559.2003.01550.x
Hohler T Lohse AW Schirmacher P Progressive focal nodular hyperplasia of the liver in a patient with genetic hemochromatosis--growth promotion by iron overload? Dig Dis Sci 2000 45 587 590 10749337 10.1023/A:1005413728101
Theise ND Lapook JD Thung SN A macroregenerative nodule containing multiple foci of hepatocellular carcinoma in a noncirrhotic liver Hepatology 1993 17 993 996 8390398 10.1016/0270-9139(93)90112-Z
Park YN Yang CP Cubukcu O Thung SN Theise ND Hepatic stellate cell activation in dysplastic nodules: evidence for an alternate hypothesis concerning human hepatocarcinogenesis Liver 1997 17 271 274 9455731
| 15871739 | PMC1112608 | CC BY | 2021-01-04 16:38:24 | no | Comp Hepatol. 2005 May 4; 4:5 | utf-8 | Comp Hepatol | 2,005 | 10.1186/1476-5926-4-5 | oa_comm |
==== Front
Health Qual Life OutcomesHealth and Quality of Life Outcomes1477-7525BioMed Central London 1477-7525-3-251583109310.1186/1477-7525-3-25Research+Psychometric evaluation of the MacDQoL individualised measure of the impact of macular degeneration on quality of life Mitchell Jan [email protected] James S [email protected] Alison [email protected] Stephen J [email protected] Carolyn V [email protected] Timothy [email protected] Martin [email protected] Winfried [email protected] Clare [email protected] Department of Psychology, Royal Holloway, University of London, Egham, Surrey, TW20 0EX, UK2 Neurosciences Research Institute, Aston University, Birmingham, B4 7ET, UK3 Hospital for Tropical Diseases, Capper Street, London WC1E 6AU, UK4 Eye Department, Queen's Medical Centre, Derby Road, Nottingham, NG7 2UH, UK2005 14 4 2005 3 25 25 6 10 2004 14 4 2005 Copyright © 2005 Mitchell et al; licensee BioMed Central Ltd.2005Mitchell 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 MacDQoL is an individualised measure of the impact of macular degeneration (MD) on quality of life (QoL). There is preliminary evidence of its psychometric properties and sensitivity to severity of MD. The aim of this study was to carry out further psychometric evaluation with a larger sample and investigate the measure's sensitivity to MD severity.
Methods
Patients with MD (n = 156: 99 women, 57 men, mean age 79 ± 13 years), recruited from eye clinics (one NHS, one private) completed the MacDQoL by telephone interview and later underwent a clinic vision assessment including near and distance visual acuity (VA), comfortable near VA, contrast sensitivity, colour recognition, recovery from glare and presence or absence of distortion or scotoma in the central 10° of the visual field.
Results
The completion rate for the MacDQoL items was 99.8%. Of the 26 items, three were dropped from the measure due to redundancy. A fourth was retained in the questionnaire but excluded when computing the scale score. Principal components analysis and Cronbach's alpha (0.944) supported combining the remaining 22 items in a single scale. Lower MacDQoL scores, indicating more negative impact of MD on QoL, were associated with poorer distance VA (better eye r = -0.431 p < 0.001; worse eye r = -0.350 p < 0.001; binocular vision r = -0.419 p < 0.001) and near VA (better eye r = -0.326 p < 0.001; worse eye r = -0.226 p < 0.001; binocular vision r = -0.326 p < 0.001). Poorer MacDQoL scores were associated with poorer contrast sensitivity (better eye r = 0.392 p < 0.001; binocular vision r = 0.423 p < 0.001), poorer colour recognition (r = 0.417 p < 0.001) and poorer comfortable near VA (r = -0.283, p < 0.001). The MacDQoL differentiated between those with and without binocular scotoma (U = 1244 p < 0.001).
Conclusion
The MacDQoL 22-item scale has excellent internal consistency reliability and a single-factor structure. The measure is acceptable to respondents and the generic QoL item, MD-specific QoL item and average weighted impact score are related to several measures of vision. The MacDQoL demonstrates that MD has considerable negative impact on many aspects of QoL, particularly independence, leisure activities, dealing with personal affairs and mobility. The measure may be valuable for use in clinical trials and routine clinical care.
==== Body
Background
Macular degeneration (MD) is a chronic, progressive eye condition that mainly affects people over the age of 50 years. It is the leading cause of blindness among those of European descent over the age of 60 years [1]. Recently it was estimated that, in the UK, between 182,000 and 300,000 people are blind or partially sighted because of MD [2]. For the majority there is no treatment and, where treatment is available, it does not cure the condition but instead slows or halts its progress for an indeterminate period [3]. People with MD lose their central vision and this precludes daily activities requiring fine vision such as reading, driving, watching TV and recognising faces. Peripheral vision is usually retained. MD can impair efficiency in performing most daily activities and may compromise the ability to live an independent life. The psychological impact of the condition can be devastating [4,5]. An ageing population means that the prevalence of MD is likely to increase [3].
New treatments for MD are being developed, as are rehabilitation programmes. Quality of life (QoL) is increasingly required as an outcome measure in clinical trials and an appropriate instrument is necessary. There has been little consensus about the definition and measurement of QoL in ophthalmology, just as in other areas of medicine [6]. Measures of health status, functional status and psychological well-being have all been used and described as QoL measures, but the interpretation of data used in this way can be misleading [7]. Some researchers into the impact of vision impairment on QoL have used health status measures such as the SF-36 [8] or the Sickness Impact Profile [9], but these have not proved informative [10,11], as many of the aspects of 'health' investigated in generic measures are unlikely to be affected by MD. Others have measured functional status (e.g. activities of daily living) [12], referring to it as QoL. Measures of health status and functional status do not correlate well with visual acuity (VA). Self-reported visual function, investigated using measures such as the NEI-VFQ [13] or the Activities of Daily Vision Scale [14] is moderately associated with VA. While such instruments can provide valuable information about functional impairment caused by vision loss, they do not measure the impact on QoL. One useful way of measuring the impact of an eye condition on QoL is to consider the importance to individuals of the aspects of life investigated in the questionnaire as well perceptions of the impact of their eye condition on each aspect. The principle of including participants' ratings of the importance of domains to their QoL (by ranking the domains) has been adopted in some generic QoL measures including the SEIQoL [15] and the Patient Generated Index [16].
The MacDQoL is an individualised measure of the impact of MD on QoL, based on the design of the Audit of Diabetes Dependent Quality of Life (ADDQoL) [17], which is increasingly used [18-20]. The questionnaire begins with two overview items, measuring: a) present QoL. (In general, my present quality of life is:), scored from +3 (excellent), through 0 (neither good nor bad) to -3 (extremely bad), b) MD-specific QoL (If I did not have MD, my quality of life would be:), scored from -3 (very much better) through 0 (the same) to +1 (worse). The 26 domain-specific items in the MacDQoL were developed from focus group meetings with people who have MD and with reference to the literature and to psychologists experienced in this field (Table 2) [21]. Each has questions asking about both the impact of MD on that aspect of life and the importance of the aspect of life to QoL. The paper version is designed for completion by visually impaired people. Figures 1 and 2 show the presentation in the questionnaire of the two overview items and one domain-specific item, with the scores for each response option shown. For the domain-specific items, impact scores (from -3 to +1) are multiplied by importance scores (from 0 to 3) to give a weighted impact score for each domain of between -9 and +3. The use of impact and importance scores enables an estimation of the impact of MD on an individual's QoL, not merely on function. For example, MD may adversely affect the time it takes an individual to do things, but if time taken is not important to his/her quality of life there will be no negative impact on QoL. Conversely, a small impact on a domain such as family life may lead to a considerable diminution of QoL if family life is very important to a person. Some domains have a 'not applicable' option (indicated by *, Table 2). A final item asks the respondent whether MD affects his/her life in any ways not already covered by the questionnaire, with a space to write a response for people who reply 'yes'. The measure has face and content validity and preliminary evidence of internal consistency reliability and sensitivity to differences in vision status (registered as blind, partially-sighted or not registered) has been reported previously [21]. Other work has shown preliminary evidence of reproducibility using self-completion in a sample of 61 people with MD [22]. The correlation between scores at time one and time 2 (mean interval 39 days) was 0.9 and there was no difference between AWI scores at times one and two (t = 1.2, p > 0.05).
Figure 1 MacDQoL present QoL and MD-specific overview items with scores shown.
Figure 2 MacDQoL domain-specific item with scores shown.
The research reported here formed the first part of a longitudinal study to carry out further evaluation of the MacDQoL.
Previous research has indicated that completion of vision-related questionnaires by pen and paper (self-completion) and by interview may not yield equivalent results [23]. This is also the case for the MacDQoL [22]. We anticipated that a substantial proportion of participants in this study would be unable to self-complete the MacDQoL because of their visual impairment and it was decided to complete the measure by telephone interview for all participants.
Methods
Participants
Potential participants were identified from the clinic lists (NHS and private) of a consultant ophthalmologist (WA). Patients were considered suitable if they had age-related MD, treated or untreated, in one or both eyes. They were excluded for any of the following:
• cataracts that were considered sufficiently severe to impair vision
• glaucoma
• diabetic retinopathy sufficiently severe to impair vision
• degenerative myopia
• any macular condition other than age-related MD
• one non-functioning eye for reasons other than age-related MD
• unable to understand and speak English
Procedure
Patients who met the inclusion criteria were contacted, initially by telephone, by an ophthalmic nurse known to all the patients. She told patients about the research, reading from a prepared script, and invited them to participate. Those willing to take part were given an appointment for a vision assessment at the hospital. Written information was despatched within three days of the telephone conversation. A member of the research team (JM) telephoned soon after and agreed the time of a telephone interview, which was carried out by a psychologist (CM or JM) not more than 14 days prior to the vision assessment appointment. During the interviews participants completed:
• MacDQoL
• demographic items
• other vision-related questions followed the MacDQoL and the demographic items. These will be reported fully elsewhere.
Interviewers were not informed of the clinical characteristics of the individual participants
Responses to questions were entered into a computerised on-screen questionnaire using SPSS Data Entry Builder [24]. The data were automatically stored as an SPSS data file.
Vision assessments, carried out by optometrists (SA, JW, MR) included:
• distance visual acuity, using Bailey-Lovie logMAR charts with Early Treatment for Diabetic Retinopathy Study (ETRDS) protocol [25] for monocular and binocular vision
• near visual acuity (MNREAD charts with ETDRS protocol) for monocular and binocular vision [26]
• comfortable visual acuity for monocular and binocular vision. This was computed from time taken to read script of different sizes of print (MNREAD charts with ETDRS protocol). The time taken to read each line was recorded. When the time to read a line increased substantially, this showed that it was no longer 'comfortable' to read that size print and smaller prints [26].
• contrast sensitivity, using Pelli-Robson charts [27] for monocular and binocular vision
• presence of distortion or a scotoma in central 10 degrees of vision (Amsler grid with concentric circles) for monocular and binocular vision. Patients fixated the central spot and identified the presence of distorted or missing grid lines in their peripheral field [28].
• colour vision (PV-16 colour vision test for visually impaired people) for binocular vision only. This consisted of a number of coloured blocks that the participant was asked to arrange in the order of the spectrum and is an enlarged version of the D-15 colour vision test [29].
• recovery from glare (Eger stressometer glare test) for binocular vision only. This test recorded the number of seconds taken to be able to read the smallest readable print again, after a brief flash of light [30].
The optometrists who carried out the vision assessments were not provided with participants' questionnaire responses.
These data were entered manually into Excel and transferred to SPSS.
Ethical approval was obtained from the Nottingham Research Ethics Committee.
Statistical methods
SPSS 9.0 [31] was used. The range of responses was examined to ascertain the need for the full range of response options and the 'not applicable' options. The effect of incorporating impact and importance ratings on the rank order of domains was investigated.
Fourteen of the 26 MacDQoL domain-specific items had a non-normal distribution. Since reliability and factor analyses are parametric procedures, measures were taken to normalise the data using transformations. Principal components analyses were carried out on both raw and transformed data.
Factor structure
Principal components analysis was carried out to identify possible subscales within the MacDQoL. To allow for data from the maximum number of participants to be used in the psychometric analyses, principal components analysis and internal consistency reliability analyses were conducted twice: first with missing data due to items being not applicable recoded as zero and participants with missing data being deleted listwise; secondly with 'not applicable' responses treated as missing data and pairwise deletion being used to minimise loss of data.
Internal consistency reliability
Cronbach's alpha coefficient of internal consistency reliability of was calculated. The higher the alpha, the stronger the internal consistency reliability, indicating that all items are measuring aspects of the same underlying construct. Corrected item-total correlations were carried out to investigate the strength of individual items' associations with the construct.
Redundancy
Redundancy of items was investigated by examining correlations between items. The distributions of the scores of the items were examined and Wilcoxon signed rank tests were carried out to compare scores of items of similar content. Principal components analysis and Cronbach's alpha were repeated after removal of redundant items.
Construct validity
Construct validity is established by examining predicted relationships between the questionnaire scores and other clinical or psychological variables. Spearman's correlations and Mann Whitney tests were used to investigate the relationship between the MacDQoL overview items and the average weighted impact score (AWI) with twelve measures of vision (see Table 6).
It was hypothesised that the MD-Specific QoL overview item and the AWI would be correlated with better eye and binocular distance visual acuity (VA), better eye and binocular near VA, better eye and binocular contrast sensitivity, binocular colour recognition, binocular comfortable reading speed and binocular presence or absence of scotoma and distortion, with greater visual impairment being associated with greater impact of MD on QoL. Since it does not focus specifically on the impact of MD on QoL, it was also hypothesised that the present QoL overview item would be correlated with these variables, but less strongly than the MD-Specific QoL overview item and the AWI.
Results
Participants
Of the 223 people telephoned by the research nurse, 38 people (17%) declined to take part (mean age of those who declined = 79.8 ± 13 years, 47% women, 53% men). Reasons for non-participation included being too ill, having too far to travel to the hospital or being unable to make suitable travel arrangements, having no one to accompany them to the vision assessment, being unavailable on the vision assessment dates and not being interested in taking part in the research. Twenty-nine people (69% women, 31% men, mean age 82.6 years) who agreed initially to take part subsequently changed their minds, or did not attend the vision assessments for other reasons. Of these, five completed the telephone interview before deciding not to participate further.
The mean age of the 156 participants was 78.96 years (s.d. 6.64, median 79.76, range 52.47 to 91.61). The mean age at leaving full time education was 15.28 years (s.d. 2.21, median 14 years, minimum 12 years, maximum 27 years). Other demographic data are reported in Table 1.
Table 1 Patient characteristics: Sex, marital status, number of eyes affected by MD, type of MD, whether both eyes diagnosed at same time, registration status.
Demographic and clinical data N (valid %)
Sex women 99 (63.5)
men 57 (36.5)
Marital status married or living with partner 74 (47.4)
widowed 68 (43.6)
divorced or separated 8 (5.1)
single 6 (3.6)
Number of eyes affected by MD one 6 (3.8)
two 150 (96.2)
Type of MD wet only 90 (57.7)
dry only 19 (12.2)
wet and dry 42 (26.9)
wet and type MD in 2nd eye not specified 4 (2.6)
type of MD not specified 1 (0.6)
Both eyes diagnosed at same time yes 46 (32.6)
no 95 (67.4)
missing 15
Registration status blind 8 (5.4)
partially sighted 67 (45.6)
not registered 72 (46.2)
missing 9
Table 2 Frequencies of impact and importance scores for domains of the MacDQoL
Item Impact score frequencies Importance score frequencies
-3 -2 -1 0 1 3 2 1 0
household tasks 46 56 26 28 0 55 76 20 5
personal affairs 65 41 22 28 0 75 57 17 6
shopping 67 43 24 22 0 47 76 23 10
*work 1 2 0 0 0 1 1 1 0
*personal relationship 10 12 12 47 0 58 19 3 0
*family life 27 35 24 60 3 108 36 4 1
friends and social 33 43 26 54 0 63 71 14 8
physical appearance 22 22 41 71 0 67 57 26 6
do physically 44 50 37 25 0 80 62 12 2
get out and about 62 34 30 30 0 92 50 12 2
*long journeys 28 37 13 24 1 27 41 26 10
*holidays 38 33 23 27 0 38 48 27 8
leisure activities 97 37 16 13 0 64 65 21 6
hobbies 68 46 19 22 0 63 64 23 6
self-confidence 41 55 29 31 0 80 58 12 6
motivation 31 48 31 45 1 51 64 33 8
people's reaction 12 25 27 91 0 48 66 27 14
society's reaction 14 25 26 88 0 28 61 43 22
future 43 56 27 30 0 52 60 33 11
financial situation 11 13 12 119 1 38 79 30 9
independence 71 37 27 21 0 97 45 10 4
do for others 56 46 23 31 0 66 65 21 4
mishaps 40 34 43 39 0 69 60 21 6
enjoy meals 31 33 26 66 0 49 72 26 9
time taken 41 49 32 33 0 30 56 45 24
enjoy nature 48 47 23 38 0 56 62 26 12
* indicates a 'not applicable option
Clinical data are reported in Table 1. Only six (3.8%) people had just one eye affected by MD. Ninety people (57.7%) had wet MD in both eyes.
The MacDQoL: range of responses
The completion rate for the MacDQoL items was 99.8%
The full range of scoring options for impact of MD (-3 to +1) was used in four domains (Table 2). All scoring options except +1 (indicating a positive impact of MD on QoL) were used in all other domains except work, where only -2 and -3 were used. The most negatively impacted domain in the MacDQoL was work (-2.33), although this domain was applicable to only three respondents (Table 3). Among the least impacted domains were finances (-0.45) and people's reaction (-0.73) (Table 3).
Table 3 MacDQoL domain-specific items in descending order of impact; mean impact scores, mean importance scores and positions of domains in rank order of weighted impact
Domains in descending order of impact score (n) Mean impact score (s.d.) Mean importance rating (s.d.) Rank order of weighted impact
1 work (3) -2.33 (1.08) 2 (1) 3
2 leisure activities (155) -2.31 (0.96) 2.2 (0.81) 2
3 hobbies (156) -2.03 (1.07) 2.18 (0.82) 4
4 independence (156) -2.01 (1.08) 2.51 (0.73) 1
5 shopping (156) -1.99 (1.07) 2.03 (0.84) 9
6 personal affairs (156) -1.92 (1.13) 2.3 (0.82) 5
7 get out and about (156) -1.82 (1.16) 2.49 (0.7) 6
8 do for others (156) -1.81 (1.13) 2.24 (0.78) 7
9 household tasks (156) -1.77 (1.06) 2.16 (0.77) 12
10 do physically (156) -1.72 (1.04) 2.41 (0.69) 8
11 future (156) -1.72 (1.07) 1.98 (0.91) 13
12 self-confidence (156) -1.68 (1.07) 2.36 (0.79) 11
13 holidays (121) -1.68 (1.14) 1.96 (0.9) 14
14 nature (156) -1.67 (1.15) 2.04 (0.91) 10
15 long journeys (103) -1.65 (1.14) 1.82 (0.93) 17
16 time taken (156) -1.63 (1.09) 1.59 (0.97) 19
17 mishaps (156) -1.48 (1.13) 2.23 (0.83) 15
18 motivation (156) -1.40 (1.12) 2.01 (0.86) 20
19 friends and social (156) -1.35 (1.16) 2.21 (0.81) 18
20 enjoy meals (156) -1.19 (1.18) 2.03 (0.85) 21
21 family life (149) -1.15 (1.2) 2.68 (0.56) 16
22 physical appearance (156) -1.14 (1.08) 2.19 (0.85) 22
23 personal relationship (81) -0.81 (1.10) 2.69 (0.54) 23
24 society's reaction (153) -0.77 (1.03) 1.62 (0.94) 25
25 people's reaction (155) -0.73 (1.00) 1.95 (0.92) 24
26 financial situation (156) -0.45 (0.92) 1.94 (0.82) 26
The full range of importance ratings (0 – 3) was used in 24 of the 26 domains (Table 2). Mean importance ratings ranged from 2.69 (personal relationship) to 1.59 (time taken) (Table 3). The rank order of domains in order of impact score changed when impact scores were multiplied by importance to give weighted impact scores, with only three domains remaining in the same position after weighting by importance (Table 3). Positions in the rank order of mean values changed by between zero and five places. Changes for individual respondents were even more substantial. Figure 3 shows the weighted impact scores of each domain. The greatest negative impact was reported for independence (-5.29) followed by leisure and work. The least impacted domain was finances (-1.02).
Figure 3 Mean weighted impact scores of MacDQoL domains.
Five items had not applicable (N/A) options (Table 2). The greatest use of the N/A option was for work (n = 153, 98%), followed by personal relationship (n = 75, 48%). Only seven (4.5%) people reported that family life was N/A.
Transforming the data
Data for some MacDQoL domains were not normally distributed. Average weighted impact scores were transformed using first log and then reflect and log transformations. It was not possible to achieve normality for all domains using either transformation, though the size of the sample will protect against the problems of non-normality. Principal components analyses using transformed data produced results that were very similar to those using untransformed data. The results reported here were obtained using untransformed data.
Structure of 26-item MacDQoL
(a) Not applicable items scored as zero
Principal components analysis with varimax rotation produced five components with Eigenvalues greater than 1. Eleven items loaded >0.4 on the first factor, including items relating to activities, such as household tasks, personal affairs, getting out and about, hobbies and do things for others. Eight items loaded on the second factor, including several relating to self-consciousness, such as appearance, people's reaction and mishaps. Finances loaded on factors 2 and 5 and leisure and hobbies loaded on both factors 1 and 3. In a forced single-factor analysis, all items loaded > 0.4 except work and finances.
(b) Not applicable items scored as missing, using pairwise deletion
Work was removed from the analysis because it was applicable for only three people. Principal components analysis with varimax rotation seeking Eigenvalues >1 revealed a 4-factor structure. Seven items double-loaded and the factor structure and the factors were not conceptually distinct. A forced single-factor analysis showed loadings very similar to the one with N/A scored as 0 except that personal relationship loaded 0.662 with N/A scored missing compared with 0.419 with N/A scored as zero. Finances still loaded < 0.4.
(c) Removal of items
A priority was to shorten the questionnaire to reduce the demand on respondents. Three pairs of items were investigated to establish whether there was any redundancy: People's reaction and society's reaction; leisure and hobbies; holidays and long journeys.
The items society's reaction and people's reaction were originally both included to establish which one was easier to understand. The telephone interviewers found that participants hesitated less over people's reaction and sometimes had difficulty differentiating between the two items. The item scores were highly correlated with each other (r = 0.692, p <0.001), more so than with any other items. The distributions of impact and importance scores were similar for the two items. To control for familywise error, a Bonferroni correction was applied (p < 0.016 accepted). There was no difference in weighted impact scores between the two items (median people's reaction = 0 [range 0 to -9]; median society's reaction = 0 [range 0 to -9], p > 0.05). People's reaction is easier to translate into other languages and this is an important consideration if the measure is to be used in international trials. Finally, evidence from semi-structured interviews in the UK and Germany during the development of a similar measure for use in diabetic retinopathy (RetDQoL) [32] supported the inclusion of people's reaction rather than society's reaction on grounds of ease of comprehension. So ciety's reaction was therefore dropped and people's reaction retained.
The items leisure activities and hobbies and interests were highly correlated with each other (r = 0.711, p < 0.001). Distribution of scores was similar for the two items. A Wilcoxon signed ranks test showed no significant difference after applying the Bonferroni correction (median leisure activities = -6 [range 0 to -9], median hobbies and interests = -6 [range 0 to -9]; Z = -2.33, p = 0.02; p < 0.016 accepted). The telephone interviewers noted that people often talked about hobbies and other interests such as embroidery and playing musical instruments when considering the leisure item. The understanding of these two items appeared to overlap and retaining both may lead to artificial inflation of the AWI. Therefore only one item was retained and reworded to specify leisure activities as well as hobbies. For the purposes of analysing the present data the mean of the two weighted impact scores was calculated for each participant (hobbies and leisure = [hobbieswi + leisurewi]/2).
The items long journeys and holidays were highly correlated with each other (r = 0.692, p <0.001). The patterns of distribution of the scores were similar for both items. There was no difference in weighted impact scores for the two items (median long journeys = -3 [range 3 to- -9], median holidays = -3 [range 0 to -9]; Z = -1.82, p > 0.05). Fewer scores were lost to the N/A option with holidays than with long journeys. During telephone interviews, the earlier item, long journeys, elicited comments about holidays, and respondents often considered the two activities to be part of the same event, since most people were retired and so work-related travel was not a consideration. To keep both items may lead to artificial inflation of the AWI, so holidays was retained and long journeys removed.
(d) Structure of the 23-item MacDQoL
Further principal components analyses were carried out following the removal of the three items. An unforced analysis with varimax rotation yielded four factors. The first factor still contained predominantly activity items together with confidence. The remaining three factors could not be labelled coherently. Appearance did not load on to any factor >0.4. In a forced single-factor analysis, all items except work and finances loaded > 0.42.
Work was removed and the analyses re-run, with N/A items scored as zero. Again, principal components analysis yielded four factors (Table 4). Six items double-loaded and one of the factors was not conceptually distinct. In a forced one-factor analysis of the 22 items, all items loaded >0.42, except finances, although the loading of this now approached 0.4 (0.356)(Table 4). The item work was applicable to only three people, but those for whom it was applicable reported a high negative impact. It was decided that work should remain in the questionnaire, but be scored as a separate item. The weighted impact score of finances was the lowest of all remaining 22 items, at -1.02. However, some negative impact of MD on finances was reported by 35 (23%) of participants and only nine (5.8%) people thought it was not at all important. It was decided to retain finances, not only because of the relevance to a minority in the present UK sample but also because this aspect of life is likely to be more impacted in people from countries where there is greater financial hardship for people with vision loss due to MD.
Table 4 Unforced principal components analysis with varimax rotation after removal of items and forced one-factor analysis with N/A items scored as zero (items loading at > 0.4 in bold).
Item Rotated Component Matrix
Four factor solution
(variance explained = 64.3%) Single factor solution
(variance explained = 49%)
Factor 1 Factor 2 Factor 3 Factor 4 Factor 1
household tasks 0.734 0.098 0.136 0.312 0.6869
personal affairs 0.754 0.254 0.146 0.099 0.7264
shopping 0.747 0.042 0.317 0.219 0.7145
personal relationship 0.210 -0.062 0.132 0.778 0.4219
family 0.179 0.411 0.199 0.629 0.6127
friends and social 0.252 0.604 0.095 0.541 0.6961
appearance 0.342 0.210 0.443 0.181 0.5821
do physically 0.587 0.473 0.269 0.023 0.7659
get out and about 0.699 0.324 0.252 0.125 0.7809
holidays 0.199 0.793 0.188 0.050 0.6582
hobbies/leisure 0.589 0.549 0.081 0.091 0.7484
self-confidence 0.495 0.282 0.259 0.266 0.6691
motivation 0.387 0.444 0.310 0.291 0.7150
peoples reaction 0.265 0.259 0.702 -0.071 0.5935
future 0.189 0.284 0.533 0.162 0.5541
finances 0.048 -0.070 0.747 0.148 0.3558
independence 0.716 0.444 0.173 0.158 0.8345
do for others 0.679 0.322 0.227 0.099 0.7491
mishaps 0.387 0.651 0.419 0.176 0.7906
enjoy meals 0.375 0.349 0.538 0.160 0.7141
time taken 0.321 0.460 0.466 0.204 0.7208
nature 0.305 0.752 0.150 0.127 0.7122
Reliability of the 22-item MacDQoL AWI scale score
Internal consistency reliability of the shortened, 22-item scale was investigated, first with N/A items scored as zero (N = 151). Cronbach's alpha coefficient of internal consistency reliability was 0.944. When the analysis was repeated with N/A items scored as missing (N = 62), alpha was 0.946. In both cases only finances detracted from the reliability, reducing it negligibly, by 0.012 in each case. The pattern of results was similar for both methods of dealing with N/A items. Table 5 shows the reliability analysis with N/A scored as zero.
Table 5 Reliability of the 22-item MacDQoL scale, with N/A items scored as zero (Cronbach's alpha = 0.9440)
MacDQoL
item Scale mean if
item deleted Scale variance if
item deleted Corrected
item-total
correlation Alpha if
item deleted
household tasks -71.92 1890.0 0.65 0.9414
personal affairs -71.19 1855.3 0.68 0.9408
shopping -71.53 1876.8 0.68 0.9410
personal relationship -74.74 1967.5 0.39 0.9446
family -72.93 1879.3 0.58 0.9426
friends and social -72.80 1875.0 0.66 0.9412
appearance -73.65 1915.5 0.54 0.9428
do physically -71.57 1858.6 0.72 0.9402
get out and about -71.32 1838.4 0.74 0.9399
holidays -73.16 1882.9 0.61 0.9419
hobbies/leisure -70.87 1874.7 0.71 0.9405
confidence -71.66 1879.3 0.63 0.9417
motivation -72.83 1876.9 0.68 0.9409
people's reaction -74.16 1918.8 0.55 0.9427
future -72.03 1909.2 0.52 0.9433
financial situation -74.89 1987.8 0.33 0.9452
independence -70.57 1822.7 0.80 0.9390
do for others -71.45 1847.3 0.71 0.9404
mishaps -72.31 1839.5 0.76 0.9397
meals -73.30 1882.1 0.68 0.9410
time -72.88 1870.7 0.69 0.9408
nature -72.08 1866.9 0.67 0.9410
Table 6 Mean scores for MacDQoL variables and vision measures. For distance VA and near VA, larger numbers indicate poorer vision. For contrast sensitivity, larger numbers indicate greater sensitivity. Larger numbers indicate poorer colour recognition and comfortable VA. For the glare test, larger numbers indicate a longer recovery time.
Variable Mean s.d. Median
MacDQoL present QoL overview 0.90 1.13 1
MacDQoL MD-specific overview -2.13 0.89 -2
MacDQoL AWI -3.57 2.14 -3.7
Distance VA (logMAR) better eye 0.42 0.46 0.39
worse eye 1.23 0.95 0.95
binocular 0.39 0.46 0.37
Near VA better eye 0.45 0.43 0.3
worse eye 1.09 0.58 1.1
binocular 0.42 0.41 0.30
Contrast sensitivity better eye 1.01 0.77 1.05
worse eye 0.43 0.49 0.15
binocular 1.01 0.44 1.05
Comfortable VA 0.50 0.35 0.4
Colour recognition (errors) 21.60 7.74 21.9
Glare test recovery (seconds) 11.02 8.94 8.5
Missing data
The AWI score can be computed despite some missing data. Missing data for up to half the items can be tolerated without Cronbach's alpha falling below 0.8. The AWI score can be calculated from the items for which responses have been given providing at least 11 items have complete responses.
Correlation between MacDQoL AWI and overview items
Mean scores of the MacDQoL overview items and AWI scores are shown in Table 6. Spearman's r correlations indicated that the AWI score was, as expected, more highly correlated with the MD-specific QoL overview item (r = 0.58, N = 156, p < 0.001) than with the present QoL item (r = 0.47, N = 156, p < 0.001).
Construct validity
Construct validity of the MacDQoL was investigated by examining relationships between the two overview items and AWI scores and the twelve measures of vision taken at the visual assessments. Since the MD-specific overview item and several of the vision measures yielded non-normal data, non-parametric tests were used.
Mean scores of the vision measures for better and worse eye and binocular vision are shown in Table 6. Spearman's correlations were used to investigate relationships between these and the three MacDQoL variables (Table 7). To control for the possibility of familywise error with 36 correlations, a Bonferroni correction was applied (p < 0.00138 accepted). Twenty-nine of the 36 correlations indicated associations of poorer QoL with worse vision, with p-values of <0.05. Twenty of these associations were still significant after correcting for familywise error (p < 0.00138). As expected, in most cases, the AWI score correlated with vision measures more strongly than did the two overview items. For near VA, distance VA and contrast sensitivity, the strongest correlations were with better-eye scores as predicted. Binocular measures showed similar relationships and worse eye measures showed poorer and less consistent associations with the MacDQoL variables Table (7). Comfortable VA and colour recognition were not associated with present QoL and comfortable VA was not associated with the MD-specific QoL overview item. None of the three MacDQoL variables was associated with recovery from glare Table (7), neither were relevant individual items, such as holidays or get out and about.
Table 7 Correlations (Spearman's r) between MacDQoL outcome variables and vision measures (*remains significant after Bonferroni correction).
Present QoL p-value MD-specific QoL p-value AWI p-value
Distance VA better eye -0.301 <0.001* -0.310 <0.001* -0.431 <0.001*
Near VA better eye -0.327 <0.001* -0.192 0.017 -0.326 <0.001*
Contrast sensitivity better eye 0.200 0.012 0.300 0.001* 0.392 <0.001*
Colour vision binocular -0.204 0.011 -0.291 <0.001* -0.417 <0.001*
Comfortable VA binocular -0.207 0.012 -0.121 >0.05 -0.283 <0.001*
Glare test binocular -0.069 >0.05 -0.010 >0.05 0.022 >0.05
Frequencies of reported scotomas and distortion are given in Table 8. Mann Whitney tests were carried out to compare MacDQoL scores in those who did and did not report binocular distortion or scotomas within 10° of vision. A Bonferroni correction was applied (six tests, p <0.0083 accepted). None of the MacDQoL scores distinguished between those who did and did not report distortion, but compared with those who did not have binocular scotomas, those who did reported poorer present QoL (means [s.d.]: yes = 0.56 [1.21], no = 1.00 [1.09], U = 1728, p = 0.037), poorer MD-specific QoL (means [s.d.]: yes = -2.44 [0.79], no = -2.03 [0.88], U = 1607, p = 0.007) and lower AWI scores (means [s.d.]: yes = -4.73 [2.04], no = -3.10 [2.02], U = 1244, p < 0.001) The MD-specific QoL overview item and the AWI score comparisons remained significant after applying the Bonferroni correction.
Table 8 Frequencies (and valid %) of reported scotomas and distortion in each eye and with binocular vision
Scotoma Distortion
Yes (valid%) No (valid%) Yes (valid%) No (valid%)
Right eye 80 (52.3) 73 (47.7) 68 (44.4) 85 (55.6)
Left eye 75 (48.7) 79 (51.3) 50 (32.7) 103 (67.3)
Binocular 39 (25.7) 113 (74.3) 58 (37.9) 95 (62.1)
Open-ended question
In response to the final, open-ended question, 'Does MD affect your quality of life in any ways that have not been covered by the questionnaire?', 56 people answered 'yes'. Those people stated one or more ways in which MD affected their QoL. In most cases, the statements were covered by items in the MacDQoL. Sixteen mentioned reading, 11 hobbies, 6 getting out and about and 7 mentioned driving specifically. Seven people mentioned not being able to recognise people, which may be related to friends and social life or to people's reaction, but it may not be fully encompassed by either item. Five people said they were frustrated by MD. Frustration might be caused by many aspects of living with MD, including items in the MacDQoL, such as time taken, mishaps and losing things, household tasks, personal affairs among others. There was no clear case for needing additional items.
Non-attenders
Five people completed the interview but subsequently did not attend the vision assessment. Mean MacDQoL scores (and s.d.s) for those people were: present QoL = 1.00 (1.22); MD-specific QoL overview = -2.4 (0.89); AWI = -3.13 (3.1). There were no significant differences in the MacDQoL scores between attenders and non-attenders (p's > 0.05).
Discussion
A total of 156 people completed both the telephone interview and the vision assessment. This was 70% of those initially approached, representing a good response rate, particularly for this elderly population. The excellent completion rate of MacDQoL items (99.85%) far exceeds the 75% obtained with utility measures [33] and it indicates that the MacDQoL is a questionnaire that is acceptable to respondents.
The wide individual variation in the ratings of impact and of importance in the MacDQoL confirms that an individualised measure is needed. Weighting impact scores by importance ratings further refines the validity and investigative qualities of the measure. The fact that only three item means remained in the same rank order of impact once importance ratings had been incorporated shows that incorporating importance scores has a noticeable effect on QoL domain scores even at the level of group means and individual scores are markedly affected by weighting with importance scores.
The high reported negative impact on MD-specific domains such as independence, personal affairs and do for others suggests that the condition-specific measure will be more sensitive than a generic QoL measure, as it investigates aspects of life that are particularly impacted by MD and these are not included in many, if any, generic measures.
The 26-item MacDQoL was a long questionnaire and the removal of three redundant items will reduce the burden involved in its completion. Their meaning is encompassed in items retained in the questionnaire. A fourth item, finances, also appeared to be a candidate for removal, with a low impact rating, the lowest weighted impact score and a small reduction to the internal consistency of the scale. Also, it did not load well in the forced single-factor analysis. However, at the time of the study, the currently favoured treatment for focal wet MD, photodynamic therapy, was not available free of charge through the National Health Service and, for those who elected to have the treatment, the financial burden was considerable. The mean weighted impact score for finances masked considerable individual variation, suggesting it would be inappropriate to remove the item. If the MacDQoL is used in countries where payment for treatment is also the norm, the finances item will be salient. MD can also affect finances in other ways, with extra costs being incurred for work such as dressmaking, housework and house maintenance, which people with MD may have undertaken themselves when they had good eyesight. For some, there may be an improvement in finances due to an entitlement to disability allowances for severely visually impaired people. In the present study one person reported that his financial situation would be worse if he did not have MD.
Our preference was for a single factor, since a single score is easier to use in both research and clinical contexts. Principal components analysis was carried out to investigate the possibility of a strong multi-factor structure. Factors should be not only mathematically but also conceptually distinct and they should form logical rather than apparently arbitrary groups. There was some evidence of logical grouping in the analyses but it was not convincing for all items. In addition, when items were removed during the item reduction process, the factor structure did not withstand these changes, indicating that the factor structure was not stable. The forced single-factor analysis showed that all items except work and finances loaded well together and demonstrated a good single-factor structure, supporting the use of an overall average weighted impact score. The single factor structure improved further with the removal of work from the scale. Since the work item was applicable to so few people and there was little variability in the scores, this had an adverse effect on the cohesiveness of the scale. However, the item is likely to show high impact and importance for those who do work, and so it is important to retain the item in the questionnaire for scoring separately. The factor structure of the MacDQoL will be revisited at a later date using longitudinal data to ensure that sensitivity to change over time is not better measured using subscales.
The Cronbach's alpha of 0.944 for the 22-item scale indicates high internal consistency reliability. An alpha of at least 0.8 is regarded as adequate for group comparisons but for clinical work, with individual patients, an alpha of 0.9 is regarded as a minimum [34]. Together with the single-factor analysis, the reliability coefficient offered considerable support for combining the MacDQoL items in a single scale. Item-total correlations were also encouraging, ranging from 0.33 to 0.80. The alpha-if-item deleted figures showed that items are similar in their effect on alpha and these data did not offer clear evidence for the exclusion of any particular items. The high Cronbach's alpha, however, suggested that other items could be removed without detriment to the scale properties. It may be useful to consider further the weighted item scores and assess the impact of removing those with low weighted impact. Nevertheless, as seen with reference to the finances item, there may be good reasons for retaining some items even though their weighted impact scores are low.
Correlations between the MacDQoL AWI and the two overview items were moderate, with a higher correlation between the AWI and the MD-specific QoL item, as expected. The magnitude of the correlation, 0.58, which is markedly less than the 0.7 required to indicate minimum equivalence [35], indicates that the MD-specific QoL item is no substitute for the AWI score.
Investigations of the relationships between the MacDQoL scores and the scores on vision measures suggested that the questionnaire has construct validity since, before Bonferroni correction, 29 of the 36 associations investigated were significant (p < 0.05), and 20 of these remained significant after correction for familywise error. The associations that remained significant were, for the large part, those that were expected to show the strongest relationships. Overall, measures of better eye and binocular vision were more strongly associated with the MacDQoL variables than measures of worse eye vision. This is to be expected, since visual ability is largely a function of the better eye and binocular function will be mainly dependent on function in the better eye. The MacDQoL demonstrates that MD has considerable negative impact on many aspects of QoL, particularly independence, leisure activities, ability to deal with personal affairs and mobility. The more severe the visual impairment due to MD, the greater is the negative impact of the condition on QoL.
The AWI score showed stronger correlations with vision measures than the MD-specific QoL overview item. The AWI is a combination of scores from domains that participants are specifically asked to consider, and the variable thus offers a more systematic assessment of the impact of MD on QoL than does the overview item. We would expect it to show a stronger association with the vision measures than the overview item. Nevertheless, the MD-specific QoL overview item may be sufficiently sensitive to be considered for use alone, for example, for audit purposes.
The present QoL item was less strongly associated with vision measures than were the AWI score and, to a lesser extent, the MD-specific QoL overview item and this finding was anticipated since, in assessing present QoL, individuals consider many factors other than the impact of MD on QoL. The present QoL item did show significant associations with a number of measures (particularly measures of binocular vision), even after the Bonferroni correction, and this demonstrates the extent of the damage done to QoL by vision loss resulting from MD.
The only vision test with which the MacDQoL scores did not show any relationship was the Eger glare test. The usefulness of the Eger stressometer in assessing the effect of glare on people with MD has yet to be established but the present data suggest little if any impact on QoL of glare as measured by this new method of assessment.
Bradley et al [17] noted that items with an N/A option presented challenges when carrying out psychometric evaluation of QoL measures. The procedure employed here for dealing with missing data caused by N/A options was used by Bradley et al [17] in order to retain sufficient data to carry out the psychometric analyses and to make best use of the available data. Both in the earlier study [17] and in the present study, using zero to replace N/A, with listwise deletion and treating N/A as missing with pairwise deletion yielded similar results. In subsequent data analysis, however, items that were N/A were excluded from the weighted mean scores. If there were no N/A options available, participants would be likely to use 'the same' responses to the impact scale and score zero for any item that was not relevant to them and this would artificially lower the AWI score. The N/A option is a feature of this and other measures based on the ADDQoL that, in addition to weighting items by importance, makes the instruments individualised measures.
The open-ended question, which asked if there were any ways in which MD affected QoL that were not covered by the questionnaire, solicited 56 responses. However, in the majority of cases, people commented on aspects of life that had in fact been covered by the questionnaire. People often made comments at this point to emphasize the things that were most important to them, such as reading and driving. Five people mentioned frustration. Frustration has not been stated explicitly in any MacDQoL items, but it is implicit in most of them. Seven people mentioned not recognising people and, whereas there are a number of items that refer to interacting with others, this problem may not adequately be addressed by any of them. It will be monitored in future work.
There was no difference in the MacDQoL scores of those who did and did not attend the vision assessment. The large discrepancy in the number of people in each group would make a significant difference unlikely, but most of those who did not attend were either ill on the day or unable to travel due to bad weather. It is not surprising, therefore, that the MacDQoL scores are similar.
The MacDQoL was originally designed for self-completion. Due to the anticipated severity of visual impairment of some people in this sample the questionnaire was administered by telephone. A mixture of completion methods was decided against since previous research has shown that people report less negative impact of MD on QoL during telephone interviews than when self-completing the MacDQoL [22]. Using a single administration method ensured that real differences in QoL due to severity of MD were not masked by biases caused by methodology.
The sample participating in this study may differ from the MD population as a whole in several ways. All but six people had MD in both eyes. This probably reflects the fact that participants were selected from an ophthalmologist's clinic records. In the general population, many cases of dry MD in one eye alone remain undiagnosed, or do not get referred to a specialist, as there is currently no treatment available. The full range of severity of MD, wet and dry, was represented in the sample ensuring that the suitability of the MacDQoL was assessed for representatives of the MD population as a whole. The AWI score may be relatively high due to the sample generally having more severe MD than in the MD population as a whole, but this would not affect the psychometric properties of the questionnaire in any way or its usefulness for people with milder MD.
Participants completed the MacDQoL by telephone interview and the methodology precluded seven people who were originally selected to be invited to participate but who did not have a telephone. A number of people who were also originally selected to be invited to participate were not contactable because contact details, including telephone numbers, were not up to date. It was not possible to ascertain whether those people had moved, no longer had a telephone or were deceased. Hearing impairment would also have precluded people from participating, but no one who was approached gave hearing impairment as a reason for not wishing to take part.
Conclusion
The MacDQoL individualised measure of the impact of MD on quality of life has been shown to have good psychometric properties. By inviting participants to rate both the impact of MD on domains of life and the importance of those domains to QoL and by providing 'not applicable' options it allows for a more individualised investigation of the impact of MD on QoL than is possible with visual function questionnaires. Excellent completion rates attest to the acceptability of the measure. The MacDQoL has been shown to have good face and construct validity with expected associations with visual function, particularly when assessed binocularly or with the better eye. The measure demonstrates that MD has a considerable negative impact on many aspects of life and on quality of life per se. The MacDQoL is now ready for use in clinical trials, routine clinical care and the evaluation of service provision.
Authors' contributions
JM participated in the design of the study and in coordination of the research, carried out telephone interviews, performed statistical analysis and drafted the manuscript. JW participated in the design of the study, carried out vision assessments and prepared the vision assessment data for analysis. AW participated in the design of the study, led the writing of the protocol and application for ethical approval and participated in coordination of the research. SJA participated in the design of the study, carried out vision assessments and, with JW, prepared the clinical data for analysis. CM carried out telephone interviews and contributed to the selection of redundant items. Tf participated in the design of the study and advised on clinical matters in the selection of participants. MR participated in the design of the study and carried out vision assessments. WA participated in the design of the study, selected participants from his clinics and oversaw the recruitment of participants and the vision assessment clinics. CB, the lead investigator, conceived of the study, participated in its design and oversaw progress of the work. All authors read and approved the final manuscript.
Copyright of MacDQoL questionnaire
For access to and permission to use the MacDQoL questionnaire, contact the copyright holder, Clare Bradley PhD, Professor of Health Psychology, Health Psychology Research, Royal Holloway, University of London, Egham, Surrey, TW20 0EX: [email protected].
Acknowledgements
Alcon Research Ltd provided funding for the research. Our thanks go to the research nurse, Kate Willbond, for her part in recruiting patients and managing the vision assessment clinics. Our thanks also go to all the study participants.
==== Refs
Evans J Wormald R Is the incidence of registrable age-related macular degeneration increasing? British Journal of Ophthalmology 1996 80 9 14 8664242
Fletcher A Donoghue M Owen C Low vision services for people with age-related macular degeneration in the UK: A review of service need and provision 2001 Denbigh, Macular Disease Society
Chopdar A Chakravarthy U Dinesh V Age related macular degeneration B M J 2003 326 485 488 12609947 10.1016/S0022-2836(02)01426-2
Williams RA Brody BL Thomas RG Kaplan RM Brown SI The psychosocial impact of macular degeneration Arch Ophthalmol 1998 116 514 520 9565052
Mitchell J Bradley P Anderson SJ ffytche T Bradley C Perceived quality of health care in macular disease: a survey of members of the Macular Disease Society Br J Ophthalmol 2002 86 777 781 12084749 10.1136/bjo.86.7.777
Bowling A Measuring Health 1998 2nd Buckingham, Open University Press
Bradley C Importance of differentiating health status from quality of life. Lancet 2001 357 7 8 11197385 10.1016/S0140-6736(00)03562-5
Ware JE Snow K Kosinski M Gaandek B SF-36 health survey: Manual and interpretation guide 1993 Boston, MA, The Health Institute, New England Medical Center
Bergner M Walker SR and Rosser RM Development, testing and use of the Sickness Impact Profile Quality of Life Assessment and application 1993 Lancaster, MIT press
Mangione CM Gutierrez PR Lowe G Orav EJ Seddon JM Influence of age-related maculopathy on visual functioning and health-related quality of life American Journal of Ophthalmology 1999 128 45 53 10482093 10.1016/S0002-9394(99)00169-5
Mills RP Correlation of quality of life with clinical symptoms and signs at the time of glaucoma diagnosis. Trans American Ophthalmological Society 1998 XCVI 753 812
Rubin GS Roche KB Prasada-Rao P Fried LP Visual impairment and disability in older adults. Optometry and Vision Science 1994 71 750 760 7898882
Mangione CM Lee PP Pitts J Gutierrez P Berry S Hays RD Psychometric properties of the National Eye Institute Visual Function Questionnaire (NEI-VFQ). NEI-VFQ Field Test Investigators Archives of Ophthalmology 1998 116 1496 1504 9823352
Mangione CM Phillips RS Seddon JM MG L EF C R D L G Development of the 'Activities of Daily Vision Scale': a measure of functional health status. Medical Care 1992 30 1111 1126 1453816
McGee HM O'Boyle CA Hickey AM O'Malley K Joyce CRB Assessing the quality of life of the individual: the SEIQoL with a healthy and a gastroenterology unit population Psychological Medicine 1991 21 749 759 1946863
Ruta DA Garratt AM Russell IT Patient centred assessment of quality of life for patients with four common conditions Quality in Health Care 1999 8 22 29 10557665
Bradley C Todd C Gorton T Symonds E Martin A Plowright R The development of an individualized questionnaire measure of perceived impact of diabetes on quality of life: the ADDQoL. Quality of Life Research 1999 8 79 91 10457741 10.1023/A:1026485130100
Kinmonth AL Woodcock A Griffin S Spiegel N Campbell MJ Randomised control trial of patient-centred care in general practice: impact on current well-being and future disease risk British Medical Journal 1998 317 1202 1208 9794859
Bradley C Speight J Patient perceptions of diabetes and diabetes therapy: assessing quality of life Diabetes Metabolism and Research Reviews 2002 18 S64 S69 10.1002/dmrr.279
DAFNE Study Group Training inflexible, intensive insulin management to enable dietary freedom in people with type 1 diabetes: the dose adjustment for normal eating (DAFNE) randomised controlled trial B M J 2002 325 746 749 12364302
Mitchell J Bradley C Design of an individualised measure of the impact of macular disease on quality of life: the MacDQoL Quality of Life Research 2004 13 1163 1175 15287282 10.1023/B:QURE.0000031348.51292.4a
Mitchell J Woodcock A Bradley C Comparison between telephone interview and self-completion of the MacDQoL Quality of Life Research 2004 13 1548 abstract 10.1023/B:QURE.0000031348.51292.4a
Wolffsohn JS Cochrane AL Watt NA Implementation methods for vision-related quality of life questionnaires. British Journal of Ophthalmology 2000 84 1035 1040 10966961 10.1136/bjo.84.9.1035
SPSS SPSS Data Entry Builder 3.0 user's guide 2001 Chicago, SPSS Inc.
Ferris FL Kassoff A Bresnick GH Bailey I New visual acuity charts for clinical research American Journal of Ophthalmology 1982 94 91 96 7091289
Ahn SJ Legge GE Luebker A Printed cards for measuring low-vision reading speed. Vision Research 1995 35 1939 1944 7660599 10.1016/0042-6989(94)00294-V
Pelli DG Robson JG Wilkins AJ The design of a new chart to measure contrast sensitivity. Clinical Vision Science 1988 2 187 199
Engler CB Sander B Koefoed P Larsen M Vinding T Lundandersen H Interferon alpha-2A treatment of patients with subfoveal neovascular macular degeneration - a pilot investigation. Acta Ophthalmologica 1993 71 27 31 7682747
Atchison DA Bowman KJ Vingrys AJ Quantitave scoring methods for D15 panel tests in the diagnosis of congenital colour-vision deficiencies. Optometry and Vision Science 1991 68 41 48 2023715
Bartlett H Davies LN Eperjesi F Reliability, normative data, and the effect of age-related macular disease on the Eger Macular Stressometer. Ophthalmic Physiol Opt 2004 24 594 599 15491488 10.1111/j.1475-1313.2004.00245.x
Statistics Package for the Social Sciences. 1998 9.,0 Chicago, SPSS Inc
Woodcock A Bradley C Plowright R Ffytche T Kennedy-Martin T Hirsch A The influence of diabetic retinopathy on quality of life. Interviews to guide the design of a condition-specific, individualised questionnaire: the RetDQoL. Patient Educ Couns 2004 53 365 383 15186876 10.1016/j.pec.2003.10.007
Mitchell J Bradley C Measuring quality of life in macular disease: what use are utilities? Quality of Life Research 2003 12 865
Nunnally JC Bernstein IH Psychometric theory: third edition 1994 New York, McGraw-Hill
Tabachnick BG Fidell LS Using multivariate statistics 1996 3 New york, Harper Collins
| 15831093 | PMC1112609 | CC BY | 2021-01-04 16:38:14 | no | Health Qual Life Outcomes. 2005 Apr 14; 3:25 | utf-8 | Health Qual Life Outcomes | 2,005 | 10.1186/1477-7525-3-25 | oa_comm |
==== Front
J Circadian RhythmsJournal of Circadian Rhythms1740-3391BioMed Central London 1740-3391-3-61586013110.1186/1740-3391-3-6Short PaperCentral fatigue and nycthemeral change of serum tryptophan and serotonin in the athletic horse Piccione Giuseppe [email protected] Anna [email protected] Francesco [email protected] Maurizio [email protected] Giovanni [email protected] Dipartimento di Morfologia, Biochimica, Fisiologia e Produzioni Animali, Facoltà di Medicina Veterinaria, Università degli Studi di Messina, 98168 Messina, Italy2005 28 4 2005 3 6 6 23 4 2005 28 4 2005 Copyright © 2005 Piccione et al; licensee BioMed Central Ltd.2005Piccione 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 serotonergic system is associated with numerous brain functions, including the resetting of the mammalian circadian clock. The synthesis and metabolism of 5-HT in the brain increases in response to exercise and is correlated with high levels of blood-borne tryptophan (TRP). The present investigation was aimed at testing the existence of a daily rhythm of TRP and 5-HT in the blood of athletic horses.
Methods
Blood samples from 5 Thoroughbred mares were collected at 4-hour intervals for 48 hours (starting at 08:00 hours on day 1 and finishing at 4:00 on day 2) via an intravenous cannula inserted into the jugular vein. Tryptophan and serotonin concentrations were assessed by HPLC. Data analysis was conducted by one-way repeated measures analysis of variance (ANOVA) and by the single cosinor method.
Results
ANOVA showed a highly significant influence of time both on tryptophan and on serotonin, in all horses, on either day, with p values < 0.0001. Cosinor analysis identified the periodic parameters and their acrophases (expressed in hours) during the 2 days of monitoring. Both parameters studied showed evening acrophases.
Conclusion
The results showed that serotonin and tryptophan blood levels undergo nycthemeral variation with typical evening acrophases. These results enhance the understanding of the athlete horse's chronoperformance and facilitate the establishment of training programs that take into account the nycthemeral pattern of aminoacids deeply involved in the onset of central fatigue.
==== Body
Background
Fatigue is an important factor affecting exercise and sporting performances. It is defined physiologically as the inability to maintain power output, [14] and the organism uses it as a defence mechanism to avoid irreversible damage due to excessive exertion. Fatigue is a complex multifactorial element with peripheral and central components. Central fatigue develops in the central nervous system and involves brain serotonin (5-HT) level [15]. The serotonergic system is associated with numerous brain functions that can positively or negatively affect endurance [6]. Accordingly, the synthesis and metabolism of 5-HT in the brain increases in response to exercise [4]. Furthermore, the rise of brain serotonin concentration is associated with markers of central fatigue such as decreased motivation, lethargy, tiredness and loss of motor coordination [6].
Increase of 5-HT synthesis in the brain is correlated with high levels of blood-borne tryptophan (TRP), the amino acid precursor to serotonin. The rate limiting step in the synthesis of 5-HT is the transport of TRP across the blood-brain barrier into the brain [9].
Many behavioral and physiological processes display 24-hour rhythms that are controlled by the circadian clock mechanism. The internal circadian clock is a molecular time-keeping system that generates a biological rhythm, regulating diverse physiological processes [18]. Serotonin (5-HT) is an important neurotransmitter and plays important roles in many physiological functions, including the operation of the mammalian circadian clock [12,17]. 5-HT is synthesized from the amino acid tryptophan hydroxylase (TPH) and aromatic L-amino acid decarboxylase [2,11]. 5-HT is a metabolic precursor of melatonin in the pineal gland and is believed to be involved in the control of sleep and in clock resetting [10].
In view of the above, and taking into account that the modulation of the physiological onset of fatigue acts positively on exercise adaptation, a good athlete's goal is to delay the occurrence of fatigue in order to maintain high performance standards. This can be achieved through a specific and continuous training programme [3]. Consequently, training and sporting activity in the horse should be accomplished in the predictably favourable phase of the day.
The present investigation was aimed at testing the existence of a daily rhythm of TRP and 5-HT in the plasma of horses. For this purpose, plasma samples were collected across 48 h from horses exposed to natural photoperiodic conditions and subjected to regular feeding and training schedules.
Materials and methods
Five Thoroughbred mares, 8 years old, were used. For 30 days prior to the study, the animals underwent the same pattern of daily activity. They were housed in individual stalls under a natural photoperiod (sunrise at 06:06, sunset at 18:49) and natural indoor temperature (19–21°C). All the horses were feed traditional rations, based hay and a mix of cereals (oats and barley), it was provided three times daily (at 08:00, 12:00 and 17:00). Water was available ad libitum. The horses were trained from 15:00 to 16:00. Training included walking, trotting, galloping and obstacle jumping.
Blood samples were collected at four-hour intervals over a 48-hour period (starting at 08:00 hours on day 1 and finishing at 4:00 on day 2) via an intravenous cannula inserted into the jugular vein. Blood samples were immediately centrifuged for 10 min at 3000 rpm with a standardized procedure and stored at + 4°C for a maximum of 24 h. Individual serum samples were deproteinized with 5-sulfosalicylic acid (5-SSA), centrifuged for 10 min at 3000 rpm, and immediately processed. On the filtered supernatant, the concentrations of tryptophan and serotonin were assessed by high-performance liquid chromatography (HPLC).
All the results were expressed as mean ± SD. One-way repeated measures analysis of variance (ANOVA) was used to determine significant difference. Probabilities < 0.05 were considered statistically significant. In addition, we applied a trigonometric statistical model to the average values of each time series, so as to describe the periodic phenomenon analytically, by individuating the main rhythmic parameters according to the single cosinor procedure [13]: Mesor (Midline Estimating Statistic of Rhythm), expressed in the same conventional unit of the relative parameter, with the confidence interval (C.I.) at 95%, Amplitude (A), expressed in the same unit as the relative Mesor, and Acrophase (Φ), expressed in hours with 95% confidence intervals.
Results
The results obtained during the experimental period indicate the existence of daily rhythms of tryptophan and serotonin serum concentration in the horse, as shown in Figures 1 and 2. ANOVA showed a highly significant effect of time on serum concentration of tryptophan and serotonin, in either day, as follows: tryptophan F(11,44) = 38.41, p < 0.0001; serotonin F(11,44) = 64.21, p < 0.0001. The application of the periodic model and the statistical analysis of the cosinor procedure enabled us to define the periodic parameters and their acrophases (expressed in hours) during the 2 days of monitoring. Both parameters studied showed nocturnal acrophases, as follows: tryptophan at 18:45 in the first day and at 18:16 in the second day; serotonin at 19:00 in the first day and 18:24 in the second day.
Figure 1 Daily rhythm of tryptophan blood level in the horse. Each time point represents the mean value ± SD. Φ represents the acrophase. Black and white stripes at the bottom of the graphic represent dark and light duration of the natural photoperiod.
Figure 2 Daily rhythm of serotonin blood level in the horse. Each time point represents the mean value ± SD. Φ represents the acrophase. Black and white stripes at the bottom of the graphic represent dark and light duration of the natural photoperiod.
Conclusion
The results obtained in this study outline a nycthemeral pattern regarding blood levels of serotonin and tryptophan. For both parameters, acrophases occurred during the evening hours at the onset of the dark phase of the experimental dark/light cycle. This suggest that photoperiod affects the timing of the investigated parameters since animals exposed to an autumnal photoperiod showed nocturnal acrophase (tryptophan at 00:40 and serotonin at 00:28) [1].
Tryptophan acrophase occurred 30 minutes earlier than serotonin acrophase. This is consistent with the role of tripthophan hydroxilation in the control of 5-HT biosynthesis. However, various patterns of daily variation of 5-HT and 5-hydroxyindoleacetic acid (5-HIAA) were observed in rats, suggesting that the nycthemeral factors affecting serotonin metabolism can be different among brain areas [16].
It has long been known that nutritional status can alter brain neurochemistry, especially that involving carbohydrates and serotonin [5,8,19]. It has been hypothesized that tryptophan infusion may increase fTrp (free-tryptophan) and the fTrp-to-BCAA (branched-chain amino acids) ratio in plasma at the same time as it decreases treadmill endurance in horses [7]. Thus central fatigue may limit endurance capacity in horses and, by manipulating fTrp and BCAA, equine exercise capacity might be altered predictably [7]. Tryptophan infusion results are consistent with the central fatigue hypothesis that an increased plasma fTrp concentration is related to the early onset of fatigue during prolonged exercise [15]. Therefore, it is likely that exercise performed at the time of the acrophase of the tryptophan rhythm (18:45, 18:16) affects the onset of physiological fatigue, thus turning on the body's exercise adaptation mechanisms in order to maintain better physical performance.
Authors' contributions
GP- Designed the study and conducted statistical analysis.
AA- Conducted bibliographic research.
FF- Carried out the data collection procedure.
MP- Carried out the data collection procedure.
GC- Supervised the data collection procedures and conducted bibliographic research.
All authors read and approved the final manuscript.
==== Refs
Assenza A Arcigli A Piccione G Velis A Bergero D Caola G Daily rhythms of blood serum concentrations of some neutral amino acids and serotonin in the horse: a preliminary study Proceedings of the symposium "Biological rhythms in livestock" Messina, Italy 95 99 14 October 2002
Borjigin J Wang MM Snyder SH Diurnal variation in mRNA encoding serotonin N-acetyltrasferase in pineal gland Nature 1995 378 783 785 8524412 10.1038/378783a0
Caola G Fisiologia dell'esercizio fisico del cavallo 2001 Bologna: Calderini Edagricole
Chaouloff F Physical exercise and brain monoamines: a review Acta Physiol Scand 1989 137 1 13 2678895
Curzon G Filippini GA Brain tryptophan: normal and disturbed control Recent advances in tryptophan research 1996 New York: Plenum Press 27 34
Davis JM Alderson NL Welsh RS Serotonin and central nervous system fatigue: nutritional considerations Am J Clin Nutr 2000 72 573S 578S 10919962
Farris JW Hincheliff KW McKeever KH Lamb BR Thompson DL Effect of tryptophan and of glucose on exercise capacity of horses J Appl Physiol 1998 85 807 816 9729551
Fernstrom MH Fernstrom JD Brain tryptophan concentration and serotonin synthesis remain responsive to food consumption after the ingestion of sequential meals Am J Clin Nutr 1995 61 312 319 7840068
Fernstrom JD Aromatic amino acids and monoamine synthesis in the central nervous system: influence of the diet J Nutr Biochem 1990 1 508 517 15539167 10.1016/0955-2863(90)90033-H
Ganguly S Coon SL Klein DC Control of melatonin synthesis in the mammalian pineal gland: the critical role of serotonin acetylation Cell Tissue Res 2002 309 127 137 12111543 10.1007/s00441-002-0579-y
Jèquier E Robinson DS Lovenberg W Sjoerdsma A Further studies on tryptophan hydroxylase in rat brainstem and beef pineal Biochem Pharmacol 1969 18 1071 1081 5789774 10.1016/0006-2952(69)90111-7
Lovenberg W Jèquier E Sjoerdsma A Tryptophan hydroxylation: measurement in pineal gland, brainstem, and carcinoid tumor Science 1967 155 217 219 6015530
Nelson W Tong YL Lee JK Halberg F Methods for cosinorrhythmometry Chronobiologia 1979 6 305 23 548245
Newsholm EA Application of principles of metabolic control to the problem of metabolic limitations in sprinting, middle-distance, and marathon running Int J Sports Med 1986 7 66 70 3017875
Newsholme EA Acworth IN Blomstrad E Benzi G Amino acids, brain neurotransmitters and a functional link between muscle and brain that is important in sustained exercise Advances in Myochemistry 1987 London: John Libby Eurotext 127 138
Newsholme EA Blomstrand E Hassmen P Ekblom B Physical and mental fatigue: do changes in plasma amino acids play a role? Biochem Soc Trans 1991 19 358 362 1679730
Poncet L Denoroy L Jouvet M Daily variations in vivo tryptophan hydroxylation and in the contents of serotonin and 5-hydroxyindoleacetic acid in discrete brain areas of the rat J Neural Transm Gen Sect 1993 92 137 150 7690229 10.1007/BF01244873
Prosser RA Serotonin phase-shifts the mouse suprachiasmatic circadian clock in vitro Brain Res 2003 966 110 115 12646314 10.1016/S0006-8993(02)04206-3
Reppert SM Weaver DR Molecular analysis of mammalian circadian rhythms Ann Rev Physiol 2001 63 647 76 11181971 10.1146/annurev.physiol.63.1.647
Wurtman RJ Wurtman JJ Carbohydrates and depression Sci Am 1989 68 75 2642626
| 15860131 | PMC1112610 | CC BY | 2021-01-04 16:39:14 | no | J Circadian Rhythms. 2005 Apr 28; 3:6 | utf-8 | J Circadian Rhythms | 2,005 | 10.1186/1740-3391-3-6 | oa_comm |
==== Front
J Inflamm (Lond)Journal of Inflammation (London, England)1476-9255BioMed Central London 1476-9255-2-31584017610.1186/1476-9255-2-3ResearchDexamethasone inhibits IL-9 production by human T cells Holz Lauren E [email protected] Kristoffer P [email protected] Snick Jacques [email protected] Francoise [email protected] William A [email protected] Garvan Institute of Medical Research, 384 Victoria St, Darlinghurst, NSW 2010, Australia2 Centre for Immunology, St. Vincent's Hospital, University of NSW, NSW 2052, Australia3 Ludwig Institute of Cancer Research, Brussels Branch and the Experimental Medicine Unit, Universite de Louvain, B-1200 Brussels, Belgium4 St Vincent's Clinical School, University of NSW, NSW 2052, Australia2005 20 4 2005 2 3 3 3 12 2004 20 4 2005 Copyright © 2005 Holz et al; licensee BioMed Central Ltd.2005Holz 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
Interleukin 9 (IL-9) is produced by activated CD4+ T cells. Its effects include stimulation of mucus production, enhanced mast cell proliferation, enhanced eosinophil function, and IgE production. These effects are consistent with a role in allergic diseases. Glucocorticoids have potent anti-inflammatory effects, including suppression of cytokine synthesis, and are widely used in the treatment of allergic conditions.
Methods
We examined the effect of the glucocorticoid dexamethasone (Dex) on IL-9 mRNA expression and protein secretion with real-time RT-PCR and ELISA. Peripheral blood mononuclear cells (PBMC) were prepared from human volunteers and activated with OKT3. CD4+ T cells were purified from PBMC and activated with OKT3 plus PMA.
Results
IL-9 mRNA abundance and protein secretion were both markedly reduced following treatment of activated PBMC with Dex. mRNA levels were reduced to 0.7% of control values and protein secretion was reduced to 2.8% of controls. In CD4+ T cells, Dex reduced protein secretion to a similar extent. The IC50 value of Dex on mRNA expression was 4 nM.
Conclusion
These results indicate that IL-9 production is very markedly inhibited by Dex. The findings raise the possibility that the beneficial effects of glucocorticoids in the treatment of allergic diseases are in part mediated by inhibition of IL-9 production.
==== Body
Background
CD4+ T cells of the T helper 2 (Th2) type have been implicated as major contributors to the pathology of allergic asthma [1]. Th2 cells produce the cytokines IL-4, IL-5, IL-9 and IL-13. IL-9, which was first identified as a T cell growth factor [2], has multiple effects consistent with a role in allergic inflammation. IL-9 acts on the pulmonary epithelium to induce production of mucus [3] and chemokines [4]. It enhances eosinophil function via induction of the IL-5 receptor [5]. IL-9 induces immunoglobulin synthesis of all isotypes, especially IgE [6]. Mast cell numbers are elevated in the lung by IL-9 [7].
There is evidence in clinical studies for an association between IL-9 and allergic asthma. In bronchial biopsies, the cells expressing IL-9, which were predominantly T cells, were increased in patients with allergic asthma, and this was associated with bronchial hyper-reactivity [8,9]. An association between IL-9 expressing cells and eosinophilia has also been described [10]. In allergic asthma patients, IL-9 in the bronchoalveolar fluid was increased after segmental allergen challenge [11]. Among a range of cytokines produced by in vitro stimulated PBMC, IL-9 was found to have the best correlation with allergic reactivity as measured by skin prick tests [12].
Several animal studies have investigated the role of IL-9 in allergic asthma. In transgenic mice with elevated pulmonary expression of IL-9, there was increased influx of inflammatory cells to the lungs, increased mucus production and increased mast cell numbers [13]. In two separate mouse model studies of allergen-induced asthma, administration of neutralising anti-IL-9 antibodies reduced eosinophilia, BHR, airway damage and IgE [14,15]. In a model of parasitic infection with a Th2 response, IL-9 knockout mice displayed markedly reduced goblet cell hyperplasia and mastocytosis [16]. However, in a model of allergic asthma, airway hyperreactivity, eosinophilia and goblet cell hyperplasia were not impaired in IL-9 knock-out mice [17]. Despite the findings in knock-out mice, overall the evidence from animal models is consistent with clinical evidence that IL-9 may have a role in allergic asthma.
Glucocorticoids (GC) are a major component of the treatment of asthma and other allergic disorders. GC bind to cytoplasmic glucocorticoid receptors (GR) and GC/GR complexes translocate to the cell nucleus where they stimulate or inhibit the transcription of a large number of genes. The anti-inflammatory effects of GC have been associated with inhibition of transcription of numerous cytokines [18]. GC markedly reduce gene transcription of the Th2 cytokines IL-4 [19], IL-5 [20] and IL-13 [21] as well as inhibiting the production of many other cytokines including IL-2 [22], GM-CSF [23] and interferon gamma (IFN-γ) [24]. By contrast, GC induce the expression of certain cytokines, including IL-10 [25], IL-1 receptor antagonist (IL-1Ra) [26] and transforming growth factor-beta [27], and GC do not affect expression of M-CSF [23].
Given the extensive evidence indicating IL-9 may be a candidate cytokine in the pathogenesis of allergic diseases, further research into the regulation of IL-9 production is warranted. Because glucocorticoids are effective in the treatment of allergic diseases, it is important to understand their effects on genes that are potentially relevant to the pathogenesis of these diseases. Therefore we have investigated the effect of the synthetic glucocorticoid dexamethasone (Dex) on IL-9 production.
Methods
Cell culture
Peripheral blood was donated by healthy volunteers from the Garvan Institute of Medical Research and the Centre for Immunology. The procedures were approved by the Human Research Ethics Committee, St Vincent's Hospital, Sydney and are in compliance with the Helsinki Declaration. Peripheral blood mononuclear cells (PBMC) were isolated by ficoll-based density centrifugation. Cells were resuspended in complete medium consisting of RPMI 1640 medium (JRH Biosciences, Lenexa, KS, USA) supplemented with 10% v/v heat-inactivated foetal bovine serum (FBS) (CSL Ltd, Parkville, Australia), 2 mM L-glutamine, 20 mM HEPES buffer, 100 U/mL penicillin and 100 μg/mL streptomycin (all from Invitrogen, Carlsbad, CA, USA). Cell counts and viabilities were determined by trypan blue exclusion in a haemocytometer. Viability was always greater than 95%.
PBMC were adjusted to 1 × 106 cells/mL and were incubated at 37°C in 5% CO2 for the activation period. PBMC were treated with 100 ng/mL OKT3 (diluted in PBS) or with a corresponding volume of PBS. OKT3, a kind gift of Janssen-Cilag, Sydney, Australia, causes T cell activation by binding to the T-cell specific surface molecule CD3. Cells were treated with Dex (Sigma, Castle Hill, Australia) or with a corresponding volume of PBS. Dex was diluted in PBS and added immediately after OKT3.
In some experiments, CD4+ T cells were purified by incubating PBMC in complete medium for 90 minutes at 37°C and 5% CO2 to deplete adherent cells. The non-adherent cells were then centrifuged and resuspended in MACS Buffer (0.5% FBS and 2 mM EDTA in PBS) and MACS human CD4+ micro beads (Miltenyi Biotec, Auburn, California, USA) according to the manufacturer's instructions. After incubation, the cells were washed and CD4+ cells were then isolated by a MACS LS Column placed in a MACS Separator according to the manufacturer's instructions (Miltenyi).
Small aliquots of the CD4+ cells were analysed by flow cytometry. Cells were stained with anti-CD3 FITC and anti-CD4 PE antibodies and analysed on a FACSCalibur using CellQuest software (all BD Biosciences, San Jose, CA). At least 98% of the cells expressed CD3 and CD4. CD4+ cells were cultured as above except that they were stimulated with a combination of 8 ng/mL PMA (Sigma) and plate-bound OKT3. OKT3 was bound to 12-well plates by addition of 10 μg/mL of OKT3 in PBS at 4°C overnight. The antibody solution was removed immediately prior to addition of the cells.
RT-PCR
After culture for 24 hours, cells were centrifuged at 440 g for 5 min. Total RNA was extracted by Trizol (Invitrogen) according to the manufacturer's instructions. RNA was dissolved in DEPC-treated water and stored at -70°C until required. RNA concentration was determined by spectrophotometry. 2 μg of total RNA was heated to 65°C for 5 min, cooled for 2–3 min on ice, and reverse transcribed by avian myeloblastosis virus reverse transcriptase (AMV-RT), with 1 μM oligo (dT)15 primer (Roche, Castle Hill, Australia), 20U AMV-RT enzyme (Roche), 1 mM dNTP (Roche), AMV-RT buffer (50 mM Tris-HCl, 8 mM MgCl2, 30 mM KCl, 1 mM dithiothreitol) (Roche) and DEPC-water in a 20 μL volume at 42°C for 1 hour. Tubes were heated to 65°C for 5 min and stored at -20°C until required.
For IL-9 PCR, in a 20 μL reaction mixture, 1 μL cDNA was amplified by Platinum Quantitative PCR Supermix UDG (1.5 U Platinum Taq Polymerase, 20 mM Tris-HCl, 50 mM KCl, 3 mM MgCl2, 200 μM dGTP, 200 μM dATP, 200 μM dCTP, 200 μM dUTP, 1U Uracil DNA glycosylase (UDG)) (Invitrogen), Milli-Q water, 0.4 μM of the forward and reverse primers and Taqman probe (Geneworks, Rundle Mall, Adelaide, Australia) with sequences 5'CCTGGACATCAACTTCCTCATC3', 5'CATGGCTGTTCACAGGAAAA3' and 5'FAM-CTCTGACAACTGCACCAGA-TAMRA3', respectively. PCR was performed with a Rotorgene 3000 real-time PCR machine (Corbett Research, Mortlake, Sydney, Australia). No template controls (NTC) with water instead of cDNA were included in all experiments. The reaction conditions for the IL-9 real-time PCR were 95°C for 3 min followed by 40 cycles of 95°C for 15 sec then 60°C for 60 sec. Forward and reverse primers were designed to bind to different exons so that any genomic DNA amplification could be distinguished from cDNA.
The PCR amplification efficiency was determined in every experiment by serial four-fold dilutions of the activated sample containing no Dex. These diluted samples and all the undiluted samples were analysed again in duplicate by real-time PCR under the same conditions. The amplification efficiency was determined by plotting the mean threshold cycle (Ct) value of the diluted samples against the log of the dilution. IL-9 amplification efficiencies ranged from 1.63 to 1.99. The actual amplification efficiencies were then used to determine the ratios of samples treated with and without Dex.
A β-actin PCR was also performed on each sample. 1 μL cDNA was amplified in 25 μL in PCR buffer (10 mM Tris-HCl, 1.5 mM MgCl2, 50 mM KCl) (Roche), 0.25 mM dNTP (Roche), 1 X SybGr (Molecular Probes, Eugene, OR, USA), 0.75 U Taq polymerase (Roche), 2 mM MgCl2 and 0.32 μM of forward and reverse β-actin primer (Geneworks) with sequences 5'CCAACTGGGACGACATG3' and 5'CAGGGATAGCACAGCCT3' respectively [20]. Samples were amplified by 94°C for 2 min followed by 30 cycles of 94°C for 15 sec, 56°C for 20 sec and 72°C for 20 sec. To confirm the identity of PCR products, all products were size-fractionated by agarose gel electrophoresis, and products with apparent mobility consistent with the expected size (277 bp for IL-9 and 203 bp for β-actin) were detected.
ELISA assays
ELISA assays were used to determine the IL-4, IL-9 and IFN-γ concentration in the culture supernatants. The IL-9 reagents (capture antibody, standard, and detection antibody) have been described previously [28], whereas the IL-4 and IFN-γ kits were purchased from BD Biosciences. 384 well flat bottom MAXISorp plates (Nunc, Roskilde, Denmark) were used. In the IL-9 ELISA the capture antibody, mh9a4, was diluted in a coating buffer (20 mM glycine, 30 mM NaCl, pH 9.2) at a concentration of 5 μg/mL. After overnight incubation at 4°C and washing with 0.05% Tween-20 in PBS, the plate was blocked with the assay diluent, 1% (w/v) BSA in PBS, incubated at 37°C for at least 2 hours and washed again. Before a final overnight incubation at 4°C the samples and standards were prepared in assay diluent, and loaded into the wells in triplicate. The standards were prepared in two-fold dilutions from 500 pg/mL to 3.9 pg/mL. After washing, detection antibody mh9a3-biotin was added in a 1:2000 dilution for 2 hours at 37°C. The plates were washed and streptavidin-horseradish peroxidase conjugate was added (DakoCytomation, Glostrup, Denmark) 1:500 in assay diluent, and plates were incubated at room temperature for 30 min. The IL-4 and IFN-γ ELISA assays were performed according to the manufacturer's instructions. The lower limits of detection were 3.9–15.6 pg/mL for IL-9, 7.8 pg/mL for IL-4 and 3.9 pg/mL for IFN-γ. When results with and without Dex were presented as percentages, if a sample was undetectable in the ELISA, the lower limit of detection of the assay was used in the calculation.
All assays were washed, loaded with TMB Substrate solution (BD Biosciences) in a 1:1 mixture of TMB substrate A and B, and incubated at room temperature in the dark for 30–45 minutes before the reaction was stopped with 2 M H2SO4. Absorbance was measured by a Spectra Image reader using X-read Plus software (both Tecan, Maennedorf, Switzerland).
Statistics
Samples were compared with the Wilcoxon signed rank test (Statview Software 5.0, Abacus Concepts, Berkeley, California, USA). A p value of <0.05 was considered significant.
Results
Dexamethasone reduces IL-9 mRNA abundance
In preliminary experiments, real-time RT-PCR revealed that OKT3 was a highly effective stimulus of IL-9 expression in PBMC, as previously reported [2]. IL-9 mRNA was induced from 4 to 48 h after activation (Fig. 1A), and 24 h was chosen as a suitable time for detection of mRNA in subsequent experiments. The effect of Dex on IL-9 mRNA abundance in PBMC was examined in 13 healthy individuals by real-time RT-PCR. PBMC were cultured with OKT3, with or without 10-6 M Dex. In all samples treated with OKT3 without Dex, IL-9 mRNA expression was readily detected. Addition of Dex to cultures stimulated by OKT3 was followed by a marked reduction in IL-9 mRNA abundance. All samples treated with Dex had much higher Ct values than those without Dex (Table 1). Statistical analysis revealed a highly significant effect of Dex (p < 0.01). All RT-PCR products were subjected to gel electrophoresis and the results were consistent with the real-time data. In the samples activated with OKT3, a single strong band was detected with apparent mobility consistent with the predicted fragment size of 277 bp (Fig. 1B). After treatment with OKT3 and Dex, a very faint band of the same mobility was detected, and no bands were detected in the unactivated samples. Except for some very low molecular size material, there was no evidence of any other band apart from the 277 bp band. The cDNA samples were also assessed for the housekeeping gene β-actin by real-time RT-PCR, and Dex had no significant effect. In activated cells, Ct values for β-actin were 15.5 ± 2.7 (SD) for samples given Dex, compared with 16.5 ± 4.6 for samples not given Dex. The findings with β-actin indicate that Dex did not cause a generalized reduction of gene expression.
Table 1 Effect of Dex on IL-9 mRNA in 13 different individuals.
Expt Ct no Dex Ct with Dex Amplification Efficiency % IL-9 in Dex vs no Dex
1 22.6 34.3 1.74 0.15
2 21.1 32.1 1.76 0.20
3 24.0 35.0 1.91 0.08
4 21.6 30.8 1.85 0.34
5 27.0 35.8 1.70 0.91
6 23.1 34.6 1.79 0.12
7 20.9 37.8 1.85 0.03
8 24.5 33.7 1.85 0.34
9 22.0 30.3 1.63 1.72
10 24.5 30.2 1.79 3.57
11 19.7 32.2 1.70 0.13
12 24.0 30.7 1.99 1.00
13 23.2 34.6 1.79 0.13
PBMC were activated with or without 10-6 M Dex, and Ct values for IL-9 were determined. The amplification efficiencies were measured for each sample, and were applied to the Ct differences between the Dex and no Dex samples to determine the proportion of IL-9 in samples treated with and without Dex.
Figure 1 IL-9 RT-PCR. A. Time course. PBMC were incubated for various times with OKT3, RNA was extracted, IL-9 real time RT-PCR was performed, and the mean threshold cycle (Ct) was determined. The data shown are from an experiment on one representative individual. The values are means of duplicate determinations. B. Gel electrophoresis. PBMC were incubated for 24 h with or without OKT3 and with or without Dex (10-6 M). RNA was extracted and IL-9 RT-PCR performed for 40 cycles. For each condition, duplicate PCRs were performed on cDNA from one representative individual. Products were analysed in a 2% agarose gel. The left lane contains HaeIII cut ΦX174 molecular size markers (Roche); the arrow indicates the position of the 281/271 bp markers.
The relative change in IL-9 mRNA expression produced by Dex was ascertained by calculating the difference in Ct values between the activated and activated + Dex samples (Table 1). This difference was then corrected for the amplification efficiency of samples from each individual PBMC donor. Amplification efficiency was determined by serial dilution of each of the samples activated and not treated with Dex. The percentage of IL-9 transcription in the Dex-treated samples compared to controls ranged from 0.03% to 3.57% with a mean of 0.67% and a median of 0.20%.
Concentration-response studies
The effectiveness of Dex was assessed by comparing IL-9 transcription in samples not treated with Dex to samples treated with 10-6 M to 10-11M Dex. Mean Ct values of duplicate samples were determined, and in each individual the mean Ct value of the sample not treated with Dex was given a figure of 100%. PCR was then performed on serial dilutions of the samples not treated with Dex to correct for amplification efficiency as described in the Methods. Dex inhibited IL-9 transcription in PBMC activated with OKT3 in a concentration dependent manner in four different individuals. The average percentage value for each Dex concentration is plotted in Figure 2. 10-7M Dex was almost as inhibitory as 10-6 M Dex, and 10-8 M Dex reduced IL-9 transcript abundance to 20% of control levels. At lower concentrations of Dex, transcription increased towards control levels. In 2 of 4 experiments, the samples treated with 10-10 M Dex had a higher level of transcription than control samples, contributing to the slightly higher average IL-9 expression level at 10-10 M Dex compared with no Dex (Fig. 2). The concentration of Dex that inhibited 50% of IL-9 transcription in activated PBMC, the IC50, was calculated to be 10-8.4 M or 4 nM.
Figure 2 Concentration-response effect of Dex on IL-9 mRNA in activated PBMC. Cells were incubated with OKT3 and the stated concentration of Dex. 24 hours later, RNA was extracted and real time RT-PCR for IL-9 was performed. Data were corrected for amplification efficiency as described in Methods. Each sample was measured in duplicate. The results are expressed as the % of the response in cells not treated with Dex. The data are the mean ± SEM of four different individuals.
Dex inhibits IL-9 protein secretion
After activation with OKT3, IL-9 secretion was readily detected by sandwich ELISA. Supernatants were harvested at various times after activation, and IL-9 was measured in triplicate. The amount of IL-9 after 72 h of culture was defined as 100%. IL-9 was not detected at 0 h. At 24 h, the IL-9 level was 16 ± 1 % (mean ± SD) and at 48 h it was 85 ± 3 %. Thus IL-9 levels had almost peaked by 48 h, and supernatants were harvested at this time in subsequent experiments. PBMC from 11 different donors were treated with or without OKT3 and with or without 10-6 M Dex throughout the culture period. In all samples stimulated with OKT3, there were high levels of IL-9 secretion in the absence of Dex, and IL-9 concentrations were in the range of 207–2,526 pg/mL. IL-9 secretion was very markedly reduced after treatment with Dex (p < 0.005). In the Dex-treated samples, secretion was only 2.8 % ± 2.5% (SD) of control values (Fig. 3). In 8 of the 11 samples treated with OKT3 and Dex, the IL-9 concentration was below the lower limit of detection of the assay. In most of the cultures not treated with OKT3, IL-9 could not be detected. It was detected at very low levels in 3 samples in the absence of Dex and in 1 sample in the presence of Dex.
Figure 3 Effect of Dex on IL-9 secretion by PBMC. Cells from 11 different individuals were treated with OKT3 and with or without 10-6 M Dex. Culture supernatants were harvested 48 hours later and measured for IL-9 by sandwich ELISA. Data represent the mean ± SD of triplicate determinations.
In six of the 11 samples, the culture supernatants were also tested for IFN-γ and IL-4. In activated cells treated with Dex, the IFN-γ and IL-4 concentrations were always above the lower detection limit of the assays. Dex significantly reduced the concentrations of both cytokines (p < 0.05 in both cases). The effect of Dex on IFN-γ secretion was similar to that on IL-9. Activated cells treated with Dex secreted 2.4 ± 2.1 % as much IFN-γ as control activated cells. By contrast, Dex had substantially less inhibitory effect on IL-4 secretion. The Dex-treated cells secreted 31.4 ± 14.1 (SD) % as much IL-4 compared with control activated cells.
CD4+ T cells were purified from 7 individuals, to determine whether Dex was acting directly on these cells. In cells not activated with OKT3, IL-9 was detected at very low levels in 4 samples without Dex and in 2 samples in the presence of Dex. Activated cells not treated with Dex secreted IL-9 in the range 222–1,939 pg/mL. As with PBMC, Dex markedly inhibited IL-9 secretion in activated cells (Fig. 4) (p < 0.02). Samples treated with 10-6 M Dex secreted only 2.9 % ± 2.5% (SD) as much IL-9 as control samples. In activated cells treated with Dex, IL-9 was below the detection limit of the assay in 3 of 7 cultures in these experiments.
Figure 4 Effect of Dex on IL-9 secretion by CD4+ T cells. Cells from 7 different individuals were treated with OKT3 and with or without 10-6 M Dex. Culture supernatants were harvested 48 hours later and measured for IL-9 by sandwich ELISA. Data represent the mean ± SD of triplicate determinations.
Discussion
The study demonstrates that Dex is an efficient pharmaceutical agent for inhibition of IL-9 production. In activated PBMC, Dex reduced IL-9 secretion to a mean of 2.8% of control levels, whereas in the case of mRNA, the corresponding value was 0.7 %. The difference between these 2 percentage values may have arisen because in the real-time PCR analysis, it was always possible to determine a value for IL-9 mRNA in the Dex-treated samples, whereas in the ELISA assay, the corresponding samples were usually undetectable. In the latter samples, the lower limit of detection of the assay was used to calculate percentages, which may have over-estimated the IL-9 concentration in the Dex-treated samples.
To determine if the inhibitory effect was specific for helper T cells, experiments were also carried out with purified CD4+ cells. These populations contained at least 98% CD3+CD4+ cells, making it very likely that the observed effects directly involve helper T cells. The data indicate that CD4+ T cells produce substantial amounts of IL-9, although the possibility that other cells in PBMC also produce IL-9 has not been excluded. Dex markedly reduced IL-9 secretion in CD4+ T cells, and the data are most consistent with a direct effect of Dex on CD4+ T cells.
Dex was found to inhibit the synthesis of IL-9 mRNA in PBMC in a concentration dependent manner. Marked inhibition of IL-9 transcription was observed with Dex concentrations as low as 10-8 M, and Dex had an IC50 value of 4 nM. Similar Dex concentration response curves have been observed with IL-2 [22] and IL-5 [20] expression in T cells, as well as IL-4 and IL-5 in mast cells [29]. ICAM-1 expression [30] as well as prostaglandin synthesis and release in alveolar tissue [31] have also been found to have similar responses to a range of concentrations of Dex. IC50 values for Dex have been obtained for ICAM-1 expression of <1 nM [30], COX activity of 1–10 nM [31], IL-11 expression of 1 nM [32] and IL-5 expression in T cells of 1 nM [20]. In mast cells, Dex had an IC50 value of 1.6 nM on IL-5 expression indicating that the sensitivity of T cells and mast cells to Dex is similar for Th2 cytokines [29]. These findings, taken together, suggest that Dex may be inhibiting similar pathways involved in regulation of expression of a variety of different genes in T cells and mast cells.
Glucocorticoids can mediate effects on transcription in two ways. After translocation of the GC/GR to the nucleus, the GR can bind directly to glucocorticoid response element (GRE) sequences in the promoter regions of target genes. The expression of many genes is stimulated in this fashion. However there is limited evidence for GRE involved in inhibition of gene expression. Alternatively, GC act indirectly by GR binding to transcription factors so as to prevent them from interacting with DNA. Previous studies have found that the two mechanisms are mediated by different concentrations of Dex. The inhibitory effect of Dex on collagenase expression was found to be mediated by interaction between GC/GR and the transcription factor AP-1 [33]. In the absence of GC, AP-1 binds to the promoter of the collagenase gene to stimulate transcription, whereas in the presence of GC, binding between GC/GR and AP-1 prevents the latter from associating with DNA, so that transcription is inhibited. Half maximal repression of collagenase expression was reached with 1.5 nM Dex, whereas half-maximal induction of gene expression via GRE binding required 10 nM or greater [33]. We found Dex to have an IC50 value of 4 nM, consistent with an indirect effect via interference with transcription factor(s).
Among possible transcription factors, NF-AT is a likely candidate. In the case of the IL-5 promoter, we observed that Dex inhibited binding to the NF-AT site but not to the GATA-3 site [34]. The IL-9 promoter contains binding sequences for NF-AT [35], and the transcription of other cytokines including IL-2 [36] and IL-4 [37] involves NF-AT. IL-4, IL-5 and IL-9 all reside within the Th2 gene cluster on human chromosome 5 [38] raising the possibility that they may have similar regulatory mechanisms. Other factors which may be involved include AP-1, NF-κB and CREB, which have DNA binding sites in the IL-9 promoter [35] and which can be inhibited by glucocorticoids [36,39,40].
Expression of IL-9 by T cells may depend on the effects of other cytokines produced after activation [41]. This is consistent with the delayed induction of IL-9 mRNA, which did not peak until 24 h after activation (Fig. 1A). It is therefore possible that the effect of Dex on IL-9 production may be a consequence of its inhibitory effect on cytokines produced earlier after T cell activation. Dex inhibited the production of the key Th1 cytokine IFN-γ to a similar extent to IL-9 (Table 2). In other experiments on PBMC, we observed that 10-6 M Dex reduced the secretion of IL-5 to 0.8 % of control PHA activated cells, and that of IL-13 to 6.2 % of controls (n = 6 for IL-5 and IL-13) (M. Irvine & W. A. Sewell, unpublished observations). However, not all Th2 cytokines are as markedly inhibited by Dex, because IL-4 was only inhibited to 31% of control levels (Table 2). The relative resistance of IL-4 to the inhibitory effects of Dex may explain an unexpected effect of Dex in enhancing the development of Th2 cells [42]; these findings could be explained by more efficient suppression by Dex of IFN-γ than IL-4, leaving sufficient IL-4 to favour differentiation of T cells into Th2 cells.
Table 2 Effect of Dex on IFN-γ, IL-4 and IL-9 secretion.
Cytokine OKT3 range OKT3 plus Dex range % cytokine in Dex vs no Dex
IFN-γ (ng/mL) 11–56 0.15–1.1 2.4 ± 2.1
IL-4 (pg/mL) 23–81 12–22 31 ± 14
§ IL-9 (pg/mL) 234–781 * undetectable 4.3 ± 2.9
PBMC from 6 different individuals were activated with OKT3 and treated with or without 10-6 M Dex. Cytokine concentration was measured in triplicate. For each individual, the % cytokine secretion in Dex versus no Dex was determined, and the Table shows the mean ± SD of these values. § The IL-9 data are for these 6 individuals only; the results are not significantly different from the results for all 11 individuals shown in Fig. 3. * For IL-9, all the Dex treated samples were below the lower limit of detection of the assay which was 7.8–15.8 pg/mL. The latter figures were used to calculate the % cytokine figure.
Conclusion
IL-9 mRNA expression and protein secretion were very markedly inhibited by Dex. The findings suggest that the beneficial effects of glucocorticoids in the treatment of allergic diseases may, in part, be mediated by inhibition of IL-9 production. Glucocorticoids are a mainstay in the treatment of allergic asthma and other allergic diseases, but their usefulness is limited by side effects. Drugs that inhibit effector cytokines, but lack the side effects of glucocorticoids, would potentially be very useful in the treatment of allergy. Our findings suggest that, when such novel drugs are evaluated, their effects on IL-9 should be taken into consideration.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
LEH performed the RT-PCR experiments. KPJ performed the ELISA experiments. LEH and KPJ drafted the manuscript. JvS prepared the anti-IL-9 antibodies and revised the manuscript. FC prepared the anti-IL-9 antibodies. WAS conceived of the project, supervised its design and coordination, and revised the manuscript. All authors read and approved the final manuscript.
Acknowledgements
The work was supported by a grant from the St Vincent's Hospital Research Committee.
==== Refs
Robinson DS Hamid Q Ying S Tsicopoulos A Barkans J Bentley AM Corrigan C Durham SR Kay AB Predominant TH2-like bronchoalveolar T-lymphocyte population in atopic asthma N Engl J Med 1992 326 298 304 1530827
Renauld JC Goethals A Houssiau F Merz H Van Roost E Van Snick J Human P40/IL-9. Expression in activated CD4+ T cells, genomic organization, and comparison with the mouse gene J Immunol 1990 144 4235 4241 1971295
Longphre M Li D Gallup M Drori E Ordonez CL Redman T Wenzel S Bice DE Fahy JV Basbaum C Allergen-induced IL-9 directly stimulates mucin transcription in respiratory epithelial cells J Clin Invest 1999 104 1375 1382 10562299
Dong Q Louahed J Vink A Sullivan CD Messler CJ Zhou Y Haczku A Huaux F Arras M Holroyd KJ Renauld JC Levitt RC Nicolaides NC IL-9 induces chemokine expression in lung epithelial cells and baseline airway eosinophilia in transgenic mice Eur J Immunol 1999 29 2130 2139 10427975 10.1002/(SICI)1521-4141(199907)29:07<2130::AID-IMMU2130>3.0.CO;2-S
Gounni AS Gregory B Nutku E Aris F Latifa K Minshall E North J Tavernier J Levit R Nicolaides N Robinson D Hamid Q Interleukin-9 enhances interleukin-5 receptor expression, differentiation, and survival of human eosinophils Blood 2000 96 2163 2171 10979962
Dugas B Renauld JC Pene J Bonnefoy JY Peti-Frere C Braquet P Bousquet J Van Snick J Mencia-Huerta JM Interleukin-9 potentiates the interleukin-4-induced immunoglobulin (IgG, IgM and IgE) production by normal human B lymphocytes Eur J Immunol 1993 23 1687 1692 7686859
Godfraind C Louahed J Faulkner H Vink A Warnier G Grencis R Renauld JC Intraepithelial infiltration by mast cells with both connective tissue-type and mucosal-type characteristics in gut, trachea, and kidneys of IL-9 transgenic mice J Immunol 1998 160 3989 3996 9558107
Shimbara A Christodoulopoulos P Soussi-Gounni A Olivenstein R Nakamura Y Levitt RC Nicolaides NC Holroyd KJ Tsicopoulos A Lafitte JJ Wallaert B Hamid QA IL-9 and its receptor in allergic and nonallergic lung disease: increased expression in asthma J Allergy Clin Immunol 2000 105 108 115 10629460 10.1016/S0091-6749(00)90185-4
Tsicopoulos A Shimbara A de Nadai P Aldewachi O Lamblin C Lassalle P Walls AF Senechal S Levitt RC Darras J Hamid Q Wallaert B Involvement of IL-9 in the bronchial phenotype of patients with nasal polyposis J Allergy Clin Immunol 2004 113 462 469 15007348 10.1016/j.jaci.2003.12.009
Ying S Meng Q Kay AB Robinson DS Elevated expression of interleukin-9 mRNA in the bronchial mucosa of atopic asthmatics and allergen-induced cutaneous late-phase reaction: relationships to eosinophils, mast cells and T lymphocytes Clin Exp Allergy 2002 32 866 871 12047433 10.1046/j.1365-2222.2002.01376.x
Erpenbeck VJ Hohlfeld JM Volkmann B Hagenberg A Geldmacher H Braun A Krug N Segmental allergen challenge in patients with atopic asthma leads to increased IL-9 expression in bronchoalveolar lavage fluid lymphocytes J Allergy Clin Immunol 2003 111 1319 1327 12789235 10.1067/mai.2003.1485
Macaubas C Sly PD Burton P Tiller K Yabuhara A Holt BJ Smallacombe TB Kendall G Jenmalm MC Holt PG Regulation of T-helper cell responses to inhalant allergen during early childhood Clin Exp Allergy 1999 29 1223 1231 10469031 10.1046/j.1365-2222.1999.00654.x
Temann UA Ray P Flavell RA Pulmonary overexpression of IL-9 induces Th2 cytokine expression, leading to immune pathology J Clin Invest 2002 109 29 39 11781348 10.1172/JCI200213696
Kung TT Luo B Crawley Y Garlisi CG Devito K Minnicozzi M Egan RW Kreutner W Chapman RW Effect of anti-mIL-9 antibody on the development of pulmonary inflammation and airway hyperresponsiveness in allergic mice Am J Respir Cell Mol Biol 2001 25 600 605 11713102
Cheng G Arima M Honda K Hirata H Eda F Yoshida N Fukushima F Ishii Y Fukuda T Anti-interleukin-9 antibody treatment inhibits airway inflammation and hyperreactivity in mouse asthma model Am J Respir Crit Care Med 2002 166 409 416 12153980 10.1164/rccm.2105079
Townsend MJ Fallon PG Matthews DJ Smith P Jolin HE McKenzie AN IL-9-deficient mice establish fundamental roles for IL-9 in pulmonary mastocytosis and goblet cell hyperplasia but not T cell development Immunity 2000 13 573 583 11070175 10.1016/S1074-7613(00)00056-X
McMillan SJ Bishop B Townsend MJ McKenzie AN Lloyd CM The absence of interleukin 9 does not affect the development of allergen-induced pulmonary inflammation nor airway hyperreactivity J Exp Med 2002 195 51 57 11781365 10.1084/jem.20011732
Umland SP Schleimer RP Johnston SL Review of the molecular and cellular mechanisms of action of glucocorticoids for use in asthma Pulm Pharmacol Ther 2002 15 35 50 11969362 10.1006/pupt.2001.0312
Robinson D Hamid Q Ying S Bentley A Assoufi B Durham S Kay AB Prednisolone treatment in asthma is associated with modulation of bronchoalveolar lavage cell interleukin-4, interleukin-5, and interferon-gamma cytokine gene expression Am Rev Respir Dis 1993 148 401 406 8342904
Rolfe FG Hughes JM Armour CL Sewell WA Inhibition of interleukin-5 gene expression by dexamethasone Immunology 1992 77 494 499 1493921
Naseer T Minshall EM Leung DY Laberge S Ernst P Martin RJ Hamid Q Expression of IL-12 and IL-13 mRNA in asthma and their modulation in response to steroid therapy Am J Respir Crit Care Med 1997 155 845 851 9117015
Boumpas DT Anastassiou ED Older SA Tsokos GC Nelson DL Balow JE Dexamethasone inhibits human interleukin 2 but not interleukin 2 receptor gene expression in vitro at the level of nuclear transcription J Clin Invest 1991 87 1739 1747 2022743
Tobler A Meier R Seitz M Dewald B Baggiolini M Fey MF Glucocorticoids downregulate gene expression of GM-CSF, NAP-1/IL-8, and IL-6, but not of M-CSF in human fibroblasts Blood 1992 79 45 51 1370208
Umland SP Nahrebne DK Razac S Beavis A Pennline KJ Egan RW Billah MM The inhibitory effects of topically active glucocorticoids on IL-4, IL-5, and interferon-gamma production by cultured primary CD4+ T cells J Allergy Clin Immunol 1997 100 511 519 9338546
John M Lim S Seybold J Jose P Robichaud A O'Connor B Barnes PJ Chung KF Inhaled corticosteroids increase interleukin-10 but reduce macrophage inflammatory protein-1alpha, granulocyte-macrophage colony-stimulating factor, and interferon-gamma release from alveolar macrophages in asthma Am J Respir Crit Care Med 1998 157 256 262 9445307
Levine SJ Benfield T Shelhamer JH Corticosteroids induce intracellular interleukin-1 receptor antagonist type I expression by a human airway epithelial cell line Am J Respir Cell Mol Biol 1996 15 245 251 8703481
Batuman OA Ferrero A Cupp C Jimenez SA Khalili K Differential regulation of transforming growth factor beta-1 gene expression by glucocorticoids in human T and glial cells J Immunol 1995 155 4397 4405 7594600
Jenmalm MC Van Snick J Cormont F Salman B Allergen-induced Th1 and Th2 cytokine secretion in relation to specific allergen sensitization and atopic symptoms in children Clin Exp Allergy 2001 31 1528 1535 11678852 10.1046/j.1365-2222.2001.01190.x
Sewell WA Scurr LL Orphanides H Kinder S Ludowyke RI Induction of interleukin-4 and interleukin-5 expression in mast cells is inhibited by glucocorticoids Clin Diagn Lab Immunol 1998 5 18 23 9455873
Cronstein BN Kimmel SC Levin RI Martiniuk F Weissmann G A mechanism for the antiinflammatory effects of corticosteroids: the glucocorticoid receptor regulates leukocyte adhesion to endothelial cells and expression of endothelial-leukocyte adhesion molecule 1 and intercellular adhesion molecule 1 Proc Natl Acad Sci U S A 1992 89 9991 9995 1279685
Newton R Seybold J Kuitert LM Bergmann M Barnes PJ Repression of cyclooxygenase-2 and prostaglandin E2 release by dexamethasone occurs by transcriptional and post-transcriptional mechanisms involving loss of polyadenylated mRNA J Biol Chem 1998 273 32312 32321 9822711 10.1074/jbc.273.48.32312
Wang J Zhu Z Nolfo R Elias JA Dexamethasone regulation of lung epithelial cell and fibroblast interleukin-11 production Am J Physiol 1999 276 L175 185 9887070
Jonat C Rahmsdorf HJ Park KK Cato AC Gebel S Ponta H Herrlich P Antitumor promotion and antiinflammation: down-modulation of AP-1 (Fos/Jun) activity by glucocorticoid hormone Cell 1990 62 1189 1204 2169351 10.1016/0092-8674(90)90395-U
Quan A McCall MN Sewell WA Dexamethasone inhibits the binding of nuclear factors to the IL-5 promoter in human CD4 T cells J Allergy Clin Immunol 2001 108 340 348 11544451 10.1067/mai.2001.118512
Zhu YX Kang LY Luo W Li CC Yang L Yang YC Multiple transcription factors are required for activation of human interleukin 9 gene in T cells J Biol Chem 1996 271 15815 15822 8663174 10.1074/jbc.271.26.15815
Paliogianni F Raptis A Ahuja SS Najjar SM Boumpas DT Negative transcriptional regulation of human interleukin 2 (IL-2) gene by glucocorticoids through interference with nuclear transcription factors AP-1 and NF-AT J Clin Invest 1993 91 1481 1489 8473495
Chen R Burke TF Cumberland JE Brummet M Beck LA Casolaro V Georas SN Glucocorticoids inhibit calcium- and calcineurin-dependent activation of the human IL-4 promoter J Immunol 2000 164 825 832 10623828
Renauld JC New insights into the role of cytokines in asthma J Clin Pathol 2001 54 577 589 11477111 10.1136/jcp.54.8.577
Ray A Prefontaine KE Physical association and functional antagonism between the p65 subunit of transcription factor NF-kappa B and the glucocorticoid receptor Proc Natl Acad Sci U S A 1994 91 752 756 8290595
Imai E Miner JN Mitchell JA Yamamoto KR Granner DK Glucocorticoid receptor-cAMP response element-binding protein interaction and the response of the phosphoenolpyruvate carboxykinase gene to glucocorticoids J Biol Chem 1993 268 5353 5356 8449898
Houssiau FA Schandene L Stevens M Cambiaso C Goldman M van Snick J Renauld JC A cascade of cytokines is responsible for IL-9 expression in human T cells. Involvement of IL-2, IL-4, and IL-10 J Immunol 1995 154 2624 2630 7876537
Ramirez F Glucocorticoids induce a Th2 response in vitro Dev Immunol 1998 6 233 243 9814597
| 15840176 | PMC1112611 | CC BY | 2021-01-04 16:36:23 | no | J Inflamm (Lond). 2005 Apr 20; 2:3 | utf-8 | J Inflamm (Lond) | 2,005 | 10.1186/1476-9255-2-3 | oa_comm |
==== Front
J NeuroinflammationJournal of Neuroinflammation1742-2094BioMed Central London 1742-2094-2-101581397010.1186/1742-2094-2-10ResearchAnti-inflammatory therapy by ibudilast, a phosphodiesterase inhibitor, in demyelination of twitcher, a genetic demyelination model Kagitani-Shimono Kuriko [email protected] Ikuko [email protected] Yasushi [email protected] Kinuko [email protected] Keiichi [email protected] Yoshihiro [email protected] Masako [email protected] Department of Developmental Medicine (Pediatrics), Osaka University Graduate School of Medicine, 2-2 Yamadaoka, Suita, Osaka 565-0871, Japan2 Department of Molecular Behavioral Biology, Osaka Bioscience Institute, 6-2-4, Furuedai, Suita, Osaka, 565-0874, Japan3 Department of Pathology and Laboratory Medicine, University of North Carolina at Chapel Hill, 919 Brinkhous-Bullitt Bldg, CB7525 Chapel Hill, NC, 27599-7525, USA2005 6 4 2005 2 10 10 4 12 2004 6 4 2005 Copyright © 2005 Kagitani-Shimono et al; licensee BioMed Central Ltd.2005Kagitani-Shimono 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
Twitcher mouse (twi/twi) is an authentic murine model of Krabbe's disease. Accumulation of psychosine, resulting in apoptosis of oligodendrocytes and subsequent demyelination, is a cardinal event to the pathogenesis of this disease. Moreover, recruitment of inflammatory cells plays a significant role in the pathological process in the twi/twi central and peripheral nervous systems. In this study, we investigated the 1) the relationship between tumor necrosis factor-α (TNFα), pro-inflammatory cytokine, and the progression of this disease and 2) effect of the anti-inflammatory therapy by ibudilast, a phosphodiesterase inhibitor.
Methods
We quantified the expression level of TNFα and TNF-receptor mRNA in twi/twi using semi-quantitative RT-PCR. The relationship between TNFα expression, apoptosis of oligodendrocytes and demyelination was studied with immunohistochemistry and TUNEL method. We then treated twi/twi with a daily intraperitoneal injection of ibudilast (10 mg/kg), which suppress TNFα production in the brain.
Results
We found that TNFα-immunoreactive microglia/macrophages appeared in the twi/twi brain and that the mRNA levels of TNFα and TNF-receptor 1 was increased with the progression of demyelination. The distribution profile of TNFα-immunoreactive microglia/macrophages overlapped that of TUNEL-positive oligodendrocytes in the twi/twi brain. When twi/twi was treated with ibudilast from PND30, the number of oligodendrocytes undergoing apoptosis was markedly reduced and demyelination was milder. Obvious improvement of clinical symptom was noted in two of five. The failure of constant clinical improvement by ibudilast may result from hepatotoxicity and/or the inhibition of proliferation of NG2-positive oligodendrocyte precursors.
Conclusion
We conclude that anti-inflammatory therapy by a phosphodiesterase inhibitor can be considered as a novel alternative therapy for Krabbe's disease.
==== Body
Background
The twitcher mouse (C57BL/6J-GALCtwi; twi/twi) is a model of human globoid cell leukodystrophy (Krabbe's disease), a disorder caused by an inherited deficiency of the lysosomal enzyme galactosylceramidase [1-3]. Twi/twi shows the symptoms of cerebellar dysfunction such as action tremor and ataxia around postnatal day (PND) 25, progressive weight loss after PND 35, and cranial and peripheral nerve palsy, eventually leading to death around PND 45 [4,5]. Obvious demyelination is recognized after PND 30 in the central nervous system (CNS). Cliniconeuropathological similarities of this model and the human disease make this murine model useful for investigations of pathogenesis as well as for therapeutic approaches [6]. The pathological physiology of twi/twi shares many common features with that of multiple sclerosis (MS), an autoimmune demyelinating disease, including the expression of major histocompatibility complex (MHC) molecules in the CNS [7-9], activation of resident microglia, recruitment of blood-borne macrophages [10], and the strong expression of pro-inflammatory cytokines such as TNFα and interleukin (IL)-6 in the demyelinating focus [10,11]. Therefore, this murine model is useful for investigating the pathomechanism of demyelination and devising therapeutic approaches to the neuroinflammation in general.
We previously showed that demyelination of twi/twi was strongly associated with apoptosis of oligodendrocytes (OLs) [12]. TNFα is the most potent inducer of apoptosis of OLs among many cytokines in vitro [13]. Additionally, in twi/twi brains, TNFα was reported to be increased in demyelinating regions [11] and expression of TNFα and other immune-related molecules were down-regulated in the pathologically improved regions [10].
Phosphodiesterase inhibitors increase the intracellular cAMP levels and reduce the inflammatory cytokines such as TNFα in vitro [14]. Ibudilast, a non-selective phosphodiesterase inhibitor, was reported to reduce demyelination in experimental allergic encephalomyelitis (EAE) and to suppress TNFα production by microglia in vitro [15,16].
In this study we found that 1) the expression of TNFα and its receptor TNF-R1 was associated with demyelination and that 2) ibudilast could reduce demyelination and alleviate the progression of disease and suppress TNFα production in twitcher brain. These results were consistent with the hypothesis that TNFα signaling enhances apoptosis of OLs and demyelination in twi/twi, and suggested that suppression of inflammation may provide new therapeutic approaches to demyelinating diseases.
Methods
Animals
All animal experiments were performed according to the Guidelines for the Protection of Experimental Animals issued by the Japanese Government, the US National Institutes of Health, and the Society for Neuroscience. Heterozygous breeder pairs of twitcher (twi/+) were originally purchased from Jackson Laboratory (Bar Harbor, ME). Twi/twi and normal age-matched siblings (+/+) were identified by genotyping with genomic DNA extracted from the clipped tails by use of a Puregene DNA Isolation Kit (Gentra Systems, Minneapolis, MN). Genotyping was performed as previously reported [17].
Materials
The following primary antibodies were used: phycoerythrin (PE)-conjugated anti-TNFα (1:50; PharMingen, San Diego, CA), mouse monoclonal anti-myelin basic protein (MBP) antibody (1:200; Sternberger Monoclonals Incorporated, Lutherville, MA), rabbit polyclonal anti-rat-pi-form of glutathione-S-transferase (pi-GST) antibody (1:1000; MBL, Nagoya, Japan), rabbit polyclonal anti-cow glial fibrillary acidic protein (GFAP) antibody (prediluted; DAKO, Glostrup, Denmark), biotinylated Ricinus communis-agglutinin-1 (RCA-1) (50 μg/ml; Vector Laboratories, Burlingame, CA), and rabbit polyclonal NG2 chondroitin sulfate proteoglycan (NG2) antibody (1:200; Chemicon International Inc., Temecula, CA). Biotinylated Ricinus communis-agglutinin-1 (RCA-1) (50 μg/ml) was purchased from Vector Laboratories (Burlingane, CA).
Tissue preparation
Brains from twi/twi and +/+ mice killed at PND 20, 30, and 40 (n = 3 for each timing period) were immunostained for TNFα. The mice were perfused with cold physiological saline under deep inhalation anesthesia with sevoflurane, and the isolated brains were quickly frozen in liquid nitrogen. For routine histochemical staining, mice (n = 3 for each groups) were perfused with physiological saline, followed by 4% paraformaldehyde in 0.1 M phosphate buffer (PB, pH 7.4). The brain was removed, postfixed and embedded in paraffin blocks. Luxol fast blue (LFB)-periodic acid Schiff (PAS) staining was performed on the paraffin sections of twi/twi and +/+ at PND 40 for evaluation of neuropathology.
For the determination of mRNA levels, groups of twi/twi and +/+ (n = 3 each timing period) were killed at PND 20, 30, and 40 under appropriate anesthesia. The brains were then removed, divided into the cerebrum and cerebellum/brain stem, and quickly frozen in liquid nitrogen.
Immunocytochemistry
Frozen sections were fixed at 4°C in acetone and incubated with PE-conjugated rat anti-mouse TNFα antibody for 48 h. For double labeling with RCA-1 and anti-TNFα, TNFα-stained sections were reacted with biotinylated RCA-1 for 30 min at room temperature, and then with avidin-D-fluorescein isothiocyanate isomer (avidin-FITC; Vector Laboratories), diluted 1:1000 with PBS, for 30 min. For NG2 immunostaining, after blocking with 0.3% Triton-X100 for 1 h, frozen sections were incubated with anti-NG2 antibody for 12 h at 4°C, and incubated with Alexa 488-conjugated anti-rabbit IgG (H+L) (1:400; Molecular Probes, Inc., Eugene, OR) for 2 h.
Paraffin sections were used for immunostaining for MBP and pi-GST, and terminal deoxynucleotidyltransferase (TdT)-mediated dUTP nick end labeling (TUNEL). For immunocytochemistry, sections on glass slides were incubated serially with mouse anti-MBP or rabbit anti-pi-GST antibody, biotinylated goat anti-mouse or anti-rabbit immunoglobulins (Vector Laboratories), and avidin-biotin complex by using an ABC elite kit (ABC; Vector Laboratories). Immunoreactions were visualized by immersing the slides in a 0.03% H2O2 solution in 50 mM Tris-HCl (pH 7.6) containing 0.05% diaminobenzidine tetrahydrochloride (DAB) and 0.25% nickel ammonium sulfate. Twi/twi and +/+ at PND 40 were subjected to TUNEL staining. Nuclei with DNA fragmentation were detected by using an in situ apoptosis detection kit (Takara Biomedicals, Osaka, Japan). Briefly, after pretreatment with 0.1% trypsin for 15 min at 37°C, sections were reacted with TdT, dNTPs, and FITC-labeled dUTP for 90 min at 37°C, followed by horseradish peroxidase (HRP)-conjugated anti-FITC antibody overnight at 4°C. The immunoproduct was visualized with the same protocol described above.
To identify the type of TUNEL-positive cells, we combined the staining for pi-GST, GFAP and RCA-1 with the TUNEL procedure. After TUNEL staining, sections were incubated with PBS containing 0.3% TritonX-100 and 10% normal goat serum for 30 min and then with rabbit anti rat-pi GST antibody, rabbit anti-cow GFAP antibody or biotinylated RCA-1 at 4°C overnight. The procedures were basically the same as described above except for the use of ABC-alkaline phosphatase and naphthol AS-BI phosphate coupled with hexazotized new fuchsin (Merck, Darmstadt, Germany) as a chromogen.
Quantification of the level of TNFα-mRNA
Total RNA was isolated from the quick-frozen brains with Isogen (Nippon gene, Toyama, Japan). The random 9-mers-primed cDNA was prepared with an RNA-LA-PCR Kit (Takara Shuzo, Kyoto, Japan) and 2 μg of total RNA.
A LightCycler PCR and detection system (Roche Diagnosis, Mannheim, Germany) was used for the amplification and quantification of mRNA for TNFα, TNFR1, TNFR2 and glycerol aldehyde-3-phosphate dehydrogenase (G3PDH) as previously described [18]. G3PDH served as an internal control. The sequence-specific primers used were as follow: TNFα forward primer: 5'-AGTGACAAGCCTGTAGCCCACG-3', TNFα reverse primer: 5'-TTTCTCCTGGTATGAGATAGC-3', TNFR1 forward primer: 5'-CTAAACAGCAGAACCGAGTGT-3', TNFR1 reverse primer: 5'-AGATACGTAGAGTGTCCTTGG-3', TNFR2 forward primer: 5'-ATAAAGCCACACCCACAACCT-3', TNFR2 reverse primer: 5'-CATCTCCCTGCCACTCACAA-3', G3PDH forward primer: 5'-TGAACGGGAAGCTCACTGG-3', and G3PDH reverse primer: 5'-TCCACCACCCTGTTGCTGTA-3'. The constructs, used to create a standard curve, were made by cloning each amplified fragment into the Hind III site of a pGEM vector (Promega, Madison, WI). The number of copies was calculated by plotting a dilution series on this standard curve in each PCR experiment. For amplification detection, the LightCycler DNA Master Hybridization Probes Kit was used. Quantification of TNFα mRNA was performed by conducting 50 cycles of repeated denaturation (1 s at 89°C), annealing (5 s at 58°C), and enzymatic chain extension (10 s at 72°C). The PCR amplification conditions for G3PDH were 40 cycles of repeated denaturation (1 s at 87°C), annealing (5 s at 57°C), and enzymatic chain extension (10 s at 72°C). Quantification of TNFR1 and TNFR2 mRNAs was made by using 50 cycles of repeated denaturation (1 s at 89°C), annealing (5 s at 58°C), and enzymatic chain extension (10 s at 72°C). Duplicated PCR products were evaluated by melting curve analysis.
Administration of Ibudilast
Ibudilast was a generous gift from Kyorin Pharmaceutical Co. Ltd. (Tokyo, Japan). After dissolved to a concentration of 1 mg/ml in physiological saline containing 10% v/v of polyoxyethylene hydrogenated castor oil 60 (HCO60), ibudilast (10 mg/kg) was injected intraperitoneally daily into three twi/twi from PND 15 to PND 40, and five twi/twi from PND 30 to PND 45. For controls, the same volume of HCO 60 was injected into two twi/twi from PND 15 to PND 40 and four twi/twi from PND 30 to PND 45. The density of TUNEL-positive cells in the demyelinating lesion in twi/twi, treated from PND 30 to PND 45 was calculated by using MacSCOPE software (Mitani Co, Fukui, Japan). Two independent neuropathologists examined the LFB-PAS-stained coronal sections (four sections per mouse) at the level of the optic chiasm and at the cerebellopontine angles containing the paraflocculus in a double-blind manner and scored the severity of demyelination from 0 to 5. 0: no demyelination, 1: slight demyelination, 2: less than 25% of the areas are occupied by a demyelination focus, 3: 25% ~ 50% of the areas occupied, 4: 50 ~ 75% of the areas occupied, 5: more than 75% of the areas occupied. The scores were average of two examiners' evaluations.
In situ hybridization for TNFα
The cDNA probe for TNFα comprised a 268-bp PCR fragment (forward primer; 5'-GATGGGTTGTACCTTGTCTACTCC-3' and reverse primer; 5'-CTAAGTACTTGGGCAGATTGACCT-3') from the mouse TNFα, and was subcloned into a pGEM-T Easy vector (Promega, Madison, Wisconsin). In situ hybridization was carried out by using manual capillary action technology with a Microprobe staining system (Fisher Scientific International, Hampton, NH) as previously described [19,20]. First, brain sections (10 μm) were deparaffinized with Auto Dewaxer (Research Genetics, Huntsville, AL). The sections were rinsed in Auto Alcohol, Universal Buffer, and Immuno/DNA buffer (Research Genetics). Predigestion by proteinase K (15 μg/ml; Sigma-Aldrich, St Louis, MO) was performed to increase the tissue penetration of the probe. After this digestion, the tissue sections were treated with Immuno/DNA buffer. The DIG-labeled cRNA probe was diluted to 0.5 μg/ml with Brigati probe diluent (Research Genetics), 50% deionized formamide, and 50% dextran sulfate. The probe solution was heated at 90°C to denature the cRNA structures and applied to the slides. The hybridization of tissue and probe was done at 50°C for three hours. After hybridization, the slides were washed in 2 × SSC containing nonionic detergent. The detection of the DIG-labeled RNA was performed by using the Genius DNA labeling and detection kit (Roche Diagnostics). For counterstaining, neutral red was applied.
Statistical analysis
Student's t test was performed by using Stat View software (SAS Institute, Cary, NC). p< 0.05 was considered as significant.
Results
Levels of TNFα and TNFR1 are increased in the twitcher cerebellum
The level of TNFα mRNA was the same in both cerebellum and cerebrum of the +/+ at any age examined. In the cerebrum, the level of TNFα-mRNA in twi/twi was almost the same as that in +/+ until PND 30, however, it increased to become approximately 15 times higher at PND 40 than that of +/+. In the cerebellum, there was no difference in the TNFα mRNA level between twi/twi and +/+ at PND 20, however, its level increased significantly in twi/twi after PND 30, becoming 40 times higher in twi/twi than +/+ at PND 40 (Fig. 1A).
Figure 1 TNFα and its receptors increased as demyelination proceeded. A-B: Quantification of mRNA for TNFα (A) and its receptors (B). The copies of mRNA for TNFα have increased in twi/twi (■) after PND 30, especially in the cerebellum, when compared with those in +/+ (▴). Those for TNFR1 in the cerebellum have increased in twi/twi after PND 30. The copies of mRNA for TNFR2 have increased in twi/twi only after PND 40, when compared with those for +/+, but the difference was not significant (B). Bar represents mean ± SE. * p < 0.01. C-F: TNFα immunostaining in the cerebellum. There are no TNFα-positive cells in the cerebellum of twi/twi mice at PND 20 (C). Immunoreactive cells for TNFα are progressively increased in number in the twi/twi cerebellar white matter between PND 30 (D) and PND 40 (E). In contrast, there are no TNFα positive cells in +/+ brains at any ages examined (F). Tw and W represent twi/twi and wild-type mice, respectively. The data represent mean ± SE. IG: internal granular layer, CWM: cerebellar white matter. Scale bar = 50 μm.
Next, we investigated the levels of TNFR1 and TNFR2. In the +/+ cerebellum, the level of TNFR1 mRNA was constant throughout all the ages examined, whereas in the twi/twi cerebellum, it significantly increased with the progression of demyelination, becoming 50 times higher than that in +/+ at PND 40. In contrast, mRNA for TNFR2 increased in twi/twi only after PND 40, when compared with that for +/+ (Fig. 1B).
Immunocytochemical analysis revealed that TNFα-immunoreactive cells were not recognized at PND 20 (Fig. 1C) in twi/twi. However, many TNFα-immunoreactive cells were found in the cerebral white matter, brain stem and cerebellar white matter (CWM) at PND 30 (Fig. 1D) and 40 (Fig. 1E). On the other hand, TNFα-immunoreactive cells were not detected anywhere in the +/+brain even at PND 40 (Fig. 1F). These data were compatible with the data of the quantitative RT-PCR.
TNFα expression is increased in microglia/macrophages within demyelinating lesions in twi/twi
The morphological characteristics of TNFα-positive cells were an irregular cellular contour and lack of delicate processes, reminiscent of ameboid microglia/macrophages. Furthermore, TNFα-positive cells were positive for RCA-1, a marker for macrophage (arrows in Fig. 2A), but negative for pi-GST, a marker for OLs, or GFAP, a marker for astrocytes (data not shown), confirming those cells to be microglia/macrophages. In the twi/twi brain, both TNFα-positive cells and TUNEL-positive cells were most abundant in the CWM (Fig. 2B, C) and in the spinal trigeminal tract (sp5) in the superior midbrain (Fig. 2E, F). The majority of TUNEL-positive cells were also positive for pi-GST (arrowheads in Fig. 2C, F, I), identifying them as OLs (inset in Fig. 2C). These lesions of the cerebellum were most severely demyelinated judged by MBP immunostaining (Fig. 2D, G). In contrast, in the corpus callosum, where demyelination was milder than in the cerebellum, only a few TNFα-positive cells were detected (Fig. 2H – J).
Figure 2 TNFα is expressed in activated microglia/macrophages in the regions where many apoptotic OLs are recognized with severe demyelination. A: Double labeling of TNFα and RCA-1 of the twi/twi cerebrum at PND 40. Arrows indicate microglia/macrophages, which are double positive for TNFα and RCA-1. B-J : In twi/twi at PND 40, there are many TNFα-positive cells (B, E) as well as many TUNEL-positive cells (C, F) in the CWM and sp5, where severe demyelination was present as judged from the results of MBP immunostaining (D, G). These apoptotic cells are immunostained with pi-GST, identified to be OLs (inset in C). In the corpus callosum (cc), there are only a few TNFα-positive cells (H) and TUNEL-positive cells (I), where demyelination was milder than in the cerebellum (J). Asterisks and double asterisks represent the same region in the serial sections. Scale bars = 50 μm (B-J), 10 μm (inset in "C").
Administration of phosphodiesterase inhibitor ameliorates demyelination and the clinical symptoms
To investigate whether the inflammatory response in microglia/macrophages contributes to the demyelination in twi/twi, we administered a phosphodiesterase inhibitor, ibudilast, to twi/twi. Two out of five twi/twi treated from PND 30 revealed strikingly milder clinical symptoms (Fig. 3A). Even at PND 45, two of ibudilast-treated twi/twi from PND 30 could move smoothly despite mild hindlimb paralysis, and showed less severe tremor and ataxia than vehicle-treated twi/twi. These mice were bigger than vehicle-treated twi/twi, as they had less weight loss (Fig. 3B). In contrast, ibudilast-treated twi/twi from PND 15 showed neither apparent clinical improvement nor elongation of lifespan, however, their body weights were heavier than those of vehicle-treated twi/twi.
Figure 3 A: Two twi/twi at PND 44, one ibudilast-treated and other vehicle-treated from PND 30. The ibudilast-treated twi/twi is much bigger and can walk faster and reach the feedbox, in spite of mild paralysis and spasticity in lower limbs. In contrast, the vehicle-treated twi/twi can no longer walk nor feed itself. In addition, the ibudilast-treated twi/twi has much milder tremor than the vehicle-treated twi/twi. B: The change of body weight (g) of ibudilast- and vehicle-treated twi/twi. Both twi/twi treated with ibudilast or vehicle from PND 15 (●: ibudilast-treated twi/twi, ○: vehicle-treated twi/twi) showed less weight gain compared with those treated from PND 30 (■: ibudilast-treated twi/twi, □: vehicle-treated twi/twi), and no prolongation of the life span. However, ibudilast-treated twi/twi showed less body weight loss than vehicle-treated twi/twi. N = 3 and 2 in ibudilast- and vehicle-treated twi/twi from PND 15. The ibudilast-treated twi/twi from PND 30 were bigger and showed milder clinical detrerioration. N = 5 and 4 in ibudilast- and vehicle-treated twi/twi from PND 30. The data represent mean ± SE.
The signal for TNFα mRNA obtained by in situ hybridization was recognized in the cells with small nuclei in the CWM and sp5 of vehicle-treated twi/twi (inset in Fig. 4A), corresponding to the presence of TNFα-immunoreactivity in the microglia. This signal was significantly reduced in the ibudilast-treated twi/twi (Fig. 4B, D). The number of TUNEL-positive cells was decreased in the CWM in ibudilast-treated twi/twi (Fig. 4F, H) compared with that of the vehicle-treated mice (Fig. 4E, G). TUNEL-positive cells were decreased in other regions such as the 8th nerve (8 n) and sp5 in ibudilast-treated twi/twi than in vehicle-treated mice (Fig. 5, the upper bar graph).
Figure 4 Suppression of TNF mRNA expression is accompanied by inhibition of apoptosis and subsequent milder demyelination in ibudilast-treated twi/twi at PND45. A, B, E, F, I, J: CWM, C, D, G, H, K, L: sp5. A-D: In situ hybridization of TNFα mRNA in vehicle-treated twi/twi (A, C) and ibudilast-treated twi/twi (B, D). Whereas vehicle-treated twi/twi show abundant signals in CWM (A) and sp5 (C), TNFα mRNA signals are remarkably reduced in the ibudilast-treated twi/twi (B, D). Inset in "A" shows TNF-α mRNA-positive microglia. E-H: TUNEL staining of vehicle-treated twi/twi (E, G) and ibudilast-treated twi/twi (F, H). Ibudilast-treated twi/twi shows fewer TUNEL-positive cells than are seen in vehicle-treated twi/twi. Arrowheads indicate TUNEL-positive cells. I-L: LFB-PAS staining of vehicle-treated twi/twi (I, K) and ibudilast-treated twi/twi (J, L). In the ibudilast-treated twi/twi, CWM and sp5 show much milder demyelination than in vehicle-treated twi/twi. Scale bar = 100 μm (I-L), 50 μm (A-H), 10 μm (inset in "A").
Figure 5 Ibudilast-treated twi/twi show pathological improvement. Population of TUNEL-positive cells and neuropathological scores of LFB-PAS in ibudilast- (closed-boxed; N = 4) or vehicle-treated (hatched; N = 3) twi/twi. In CWM, 8 n, and sp5 of the ibudilast-treated twi/twi, the number of TUNEL-positive cells is decreased to half of those in the vehicle-treated twi/twi. They also recognized significantly milder demyelination in LFB-PAS stain. 8 n: the 8th nerve. *p < 0.01, **p < 0.05. The error bars represented standard deviations.
LFB-PAS staining revealed that the demyelination was remarkably suppressed in the ibudilast-treated mice from PND 30 (Fig. 4J, L) compared with the vehicle-treated ones (Fig. 4I, K), as shown in the score of demyelination (Fig. 5, lower bar graph). From these lines of evidence, we concluded that the demyelination and clinical symptoms were reduced with inhibition of TNFα in twi/twi.
Ibudilast treatment decreased NG2-positive OL progenitors
To evaluate the effect of ibudilast to the OL progenitors, frozen sections were stained with anti-NG2 antibody. In contrast to the vehicle-treated twi/twi, ibudilast-treated twi/twi showed fewer NG2-positive OL progenitors (Fig. 6), suggesting that incomplete clinical improvement may result from the insufficient remyelination in ibudilast-treated twi/twi.
Figure 6 Ibudilast surpresses proliferation of NG2-positive OL progenitors. A: Vehicle-treated twi/twi shows many NG2-positive OL progenitors. B: Ibudilast-treated twi/twi shows decreased number of NG2-positive OL progenitors. Allows: NG2-positive OL progenitors labeled with Alexa 488. Scale bar = 50 μm
Discussion
Our results suggested that secondary inflammation via TNFα produced in microglia/macrophages remarkably enhances the apoptosis of OLs and aggravates the demyelination due to the metabolic defect in twi/twi. These are consistent with previous reports showing that TNFα induces apoptosis of OLs in vitro [21,22], and that TNFα is upregulated in macrophages and globoid cells in twi/twi [11].
TNFα is a well-established pro-inflammatory mediator of immune process, and is essential to the maintenance of CNS homeostasis. However, its overexpression leads to the development of chronic CNS inflammation and degeneration [23]. We previously observed emergence of TNFα-expressing cells with progression of demyelination and the number of those cells declined following bone marrow transplantation with prolonged survival in twi/twi [10]. TNFα was expressed by infiltrating blood mononuclear cells, and its expression was well correlated with the extent of demyelination in another genetic demyelinating disease, X-linked adrenoleukodystrophy[24], and in the MS [25]. TNFα-transgenic mice showed more severe demyelination and macrophage infiltration in EAE, a mouse model for MS [26]. Of two TNFRs, TNFR1 was reported to mediate the pathogenetic effects of TNFα, such as inflammation, cytotoxicity, and apoptosis of OLs in EAE [13,27-29]. Our study showed that TNFR1 was dominant from the early demyelinating stage and that demyelination and OL apoptosis was alleviated by the suppression of TNFα in ibudilast-treated twi/twi. These lines of evidence suggested that the stimulation of TNFR1 was associated with apoptosis of OLs and demyelination in twi/twi. Therefore, we believe that TNFα/TNFR1-mediated secondary inflammation is involved in the progression of pathology in varieties of demyelinating diseases.
In this study, we selected ibudilast as an immunomodulatory agent which also suppressed the production of other inflammatory mediators, such as nitric oxide (NO), IFN-γ, and IL-6, and enhanced the production of the inhibitory cytokine, IL-10, and neurotrophic factors, including nerve growth factor (NGF), glia-derived neurotrophic factor (GDNF) and neurotrophin (NT-4) [30]. Since inducible nitric oxide (iNOS) and IL-6 were strongly upregulated in twi/twi and Krabbe's disease [10,11,31], the positive effect of ibudilast may be also associated with suppression of iNOS and IL-6, and enhancement of inhibitory cytokines and neurotrophic factors. However, taking into account that TNFα is the most potent cytotoxic cytokine, and that signals for TNFα mRNA were remarkably suppressed in the areas of severe demyelination in ibudilast-treated twi/twi, the effect of ibudilast may be mediated, at leaset in part, by the suppression of TNFα expression.
Several different types of anti-TNFα therapy have been recently reported. For example, TNF-receptor-p55-immunoglobulin fusion protein was reported to suppress demyelination in EAE [32,33], whereas it showed no significant efficacy in MS patients [34,35]. Infliximab and etanercept, used as anti-TNFα agents for rheumatoid arthritis and Crohn's disease, are rather reported to induce demyelination [36,37]. In contrast to the poor outcomes of these direct TNFα suppression, interferon (IFN) β [38,39] and glatiramer acetate (GA) [40,41] have been widely approved as effective immunomodulatory treatments for MS. TNFα production was significantly reduced in monocytes from patients treated by GA [42], which acts primarily as an antigen for T lymphocytes. Furthermore, MS patients who received administration of IFNβ revealed decreased mRNA for TNFα [43] and an increase in serum TNFRs, of which TNFR2 may play a protective role for myelin [44].
The clinical symptoms were improved in only two ibudilast-treated twi/twi, whereas the demyelination was milder in all of the treated twi/twi. In the ibudilast-treated twi/twi without clinical improvement, the number of NG2-immunoreactive OL progenitors was decreased, compared with that in vehicle-treated twi/twi. Lack of TNFα has been reported to result in a significant delay of remyelination in a cuprizone-induced demyelination model, due to a reduced number of proliferating OL progenitors [45], since the signal transduction of TNFα via p75 TNF receptor 2 (TNFR2) is known to induce proliferation of OL progenitors [27,28]. Therefore, TNFα stimulation may be involved not only in the apoptotic signal pathway mediated by TNFR1, but may also play a regenerative role via activation of TNFR2 [46]. Earlier treatment with ibudilast from PND 15 showed less apparent clinical effect compared with that from PND30, probably due to the following two reasons: daily intraperitoneal injection itself could be too invasive for younger twi/twi to gain weight and/or TNFR2-stimulated proliferation of OLs in this period of active myelination is profoundly inhibited by the reduced TNFα production. These lines of evidence suggested that TNFα inhibitor should be used for a limited period of time or in a TNFR1-specific manner.
The cytotoxicity of ibudilast may be another explanation for the failure of clinical improvement in some cases: when we administered a high dosage (20 mg/kg) of ibudilast to twi/twi, it induced vacuolar degeneration of hepatocytes and the mice died of the hepatic failure (data not shown). When ibudilast was directly administered by an intraventricular injection to avoid systemic adverse effect, periventricular tissues were extensively damaged by this chemical. These results indicate that other drugs with less cytotoxicity are necessary to improve the symptoms of twi/twi and other demyelination diseases.
From these lines of evidence, we propose that anti-inflammatory therapy by a phosphodiesterase inhibitor during an appropriate period, may be a reliable supportive treatment for Krabbe's disease for which there is no effective treatment except bone marrow transplantation [6,23,47-49].
Conclusion
These results suggest that the suppression of inflammation by a phosphodiesterase inhibitor could be a novel therapy in genetic demyelination.
List of abbreviations
twitcher mouse (twi/twi)
tumor necrosis factor-α (TNFα)
postnatal day (PND)
central nervous system (CNS)
multiple sclerosis (MS)
major histocompatibility complex (MHC)
interleukin (IL)
oligodendrocytes (OLs)
experimental allergic encephalomyelitis (EAE)
phycoerythrin (PE)
myelin basic protein (MBP)
pi-form of glutathione-S-transferase (pi-GST)
glial fibrillary acidic protein (GFAP)
Ricinus communis-agglutinin-1 (RCA-1)
phosphate buffer (PB)
fluorescein isothiocyanate isomer (FITC)
terminal deoxynucleotidyltransferase (TdT)-mediated dUTP nick end labeling (TUNEL)
diaminobenzidine tetrahydrochloride (DAB)
horseradish peroxidase (HRP)
Luxol fast blue (LFB)-periodic acid Schiff (PAS)
glycerol aldehyde-3-phosphate dehydrogenase (G3PDH)
cerebellar white matter (CWM)
interferon (IFN)
glatiramer acetate (GA)
nitric oxide (NO)
nerve growth factor (NGF)
glia-derived neurotrophic factor (GDNF)
neurotrophin (NT)
inducible nitric oxide synthase (iNOS)
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
KKS was responsible for the majority of the experimental studies, and for writing the manuscript. IM and YF contributed to technical tutorship and the editing of the manuscript. KS and KO contributed to editing of the manuscript. MT and YU contributed to the conception, interpretation of results and the writing and editing of the manuscript. All authors read and approved the final manuscript.
Acknowledgements
This study was supported by funding from the following; the Ministry of Education, Culture, Sports, Science and Technology of Japan (Grant-in-aid for Exploratory Research, No.15659246; MT), the Osaka Medical Research Foundation for Incurable Diseases (M.T.), National Institutes of Health USPHS (NS-24453 and HD-03110; K.S), Takeda Science Foundation (Y.U.), Mitsubishi Foundation (Y.U.), Japan Foundation for Applied Enzymology (Y.U.), Japan Aerospace Exploration Agency (Y.U.), and Osaka City (Y.U.).
==== Refs
Duchen LW Eicher EM Jacobs JM Scaravilli F Teixeira F Hereditary leucodystrophy in the mouse: the new mutant twitcher Brain 1980 103 695 710 7417782
Kobayashi T Yamanaka T Jacobs JM Teixeira F Suzuki K The Twitcher mouse: an enzymatically authentic model of human globoid cell leukodystrophy (Krabbe disease) Brain Res 1980 202 479 483 7437911 10.1016/0006-8993(80)90159-6
Suzuki K Suzuki K The twitcher mouse. A model of human globoid cell leukodystrophy (krabbe's disease) Am J Pathol 1983 111 394 397 6859223
Taniike M Suzuki K Spacio-temporal progression of demyelination in twitcher mouse: with clinico-pathological correlation Acta Neuropathol (Berl) 1994 88 228 236 7528964
Suzuki K Taniike M Murine model of genetic demyelinating disease: the twitcher mouse Microsc Res Tech 1995 32 204 214 8527855 10.1002/jemt.1070320304
Suzuki K Hoogerbrugge PM Poorthuis BJ Bekkum DW Suzuki K The twitcher mouse. Central nervous system pathology after bone marrow transplantation Lab Invest 1988 58 302 309 3279262
Ohno M Komiyama A Martin PM Suzuki K MHC class II antigen expression and T-cell infiltration in the demyelinating CNS and PNS of the twitcher mouse Brain Res 1993 625 186 196 8275302 10.1016/0006-8993(93)91058-Z
Matsushima GK Taniike M Glimcher LH Grusby MJ Frelinger JA Suzuki K Ting JP Absence of MHC class II molecules reduces CNS demyelination, microglial/macrophage infiltration, and twitching in murine globoid cell leukodystrophy Cell 1994 78 645 656 8069913 10.1016/0092-8674(94)90529-0
Taniike M Marcus JR Nishigaki T Fujita N Popko B Suzuki K Suzuki K Suppressed UDP-galactose: ceramide galactosyltransferase and myelin protein mRNA in twitcher mouse brain J Neurosci Res 1998 51 536 540 9514207 10.1002/(SICI)1097-4547(19980215)51:4<536::AID-JNR13>3.0.CO;2-1
Wu YP McMahon EJ Matsuda J Suzuki K Matsushima GK Suzuki K Expression of immune-related molecules is downregulated in twitcher mice following bone marrow transplantation J Neuropathol Exp Neurol 2001 60 1062 1074 11706936
LeVine SM Brown DC IL-6 and TNFalpha expression in brains of twitcher, quaking and normal mice J Neuroimmunol 1997 73 47 56 9058758 10.1016/S0165-5728(96)00166-X
Taniike M Mohri I Eguchi N Irikura D Urade Y Okada S Suzuki K An apoptotic depletion of oligodendrocytes in the twitcher, a murine model of globoid cell leukodystrophy J Neuropathol Exp Neurol 1999 58 644 653 10374755
Akassoglou K Bauer J Kassiotis G Pasparakis M Lassmann H Kollias G Probert L Oligodendrocyte apoptosis and primary demyelination induced by local TNF/p55TNF receptor signaling in the central nervous system of transgenic mice: models for multiple sclerosis with primary oligodendrogliopathy Am J Pathol 1998 153 801 813 9736029
Taguchi I Oka K Kitamura K Sugiura M Oku A Matsumoto M Protection by a cyclic AMP-specific phosphodiesterase inhibitor, rolipram, and dibutyryl cyclic AMP against Propionibacterium acnes and lipopolysaccharide-induced mouse hepatitis Inflamm Res 1999 48 380 385 10450787 10.1007/s000110050475
Suzumura A Ito A Yoshikawa M Sawada M Ibudilast suppresses TNFalpha production by glial cells functioning mainly as type III phosphodiesterase inhibitor in the CNS Brain Res 1999 837 203 212 10434004 10.1016/S0006-8993(99)01666-2
Fujimoto T Sakoda S Fujimura H Yanagihara T Ibudilast, a phosphodiesterase inhibitor, ameliorates experimental autoimmune encephalomyelitis in Dark August rats J Neuroimmunol 1999 95 35 42 10229113 10.1016/S0165-5728(98)00251-3
Sakai N Inui K Tatsumi N Fukushima H Nishigaki T Taniike M Nishimoto J Tsukamoto H Yanagihara I Ozono K Okada S Molecular cloning and expression of cDNA for murine galactocerebrosidase and mutation analysis of the twitcher mouse, a model of Krabbe's disease J Neurochem 1995 66 1118 1124 8769874
Loeffler J Henke N Hebart H Schmidt D Hagmeyer L Schumacher U Einsele H Quantification of fungal DNA by using fluoresence energy transfer and the light cycler system J Clin Microbiol 2000 38 586 590 10655350
Wang JY Montone KT A rapid simple in situ hybridization method for herpes simplex virus employing a synthetic biotin-labeled oligonucleotide probe: a comparison with immunohistochemical methods for HSV detection J Clin Lab Anal 1994 8 105 115 8189321
Unger ER Brigati DJ Colorimetric in-situ hybridization in clinical virology: development of automated technology Curr Top Microbiol Immunol 1989 143 21 31 2670458
Selmaj KW Raine CS Tumor necrosis factor mediates myelin and oligodendrocyte damage in vitro Ann Neurol 1988 23 339 346 3132891 10.1002/ana.410230405
Zajicek JP Wing M Scolding NJ Compston DA Interactions between oligodendrocytes and microglia. A major role for complement and tumour necrosis factor in oligodendrocyte adherence and killing Brain 1992 115 1611 1631 1486453
Probert L Akassoglou K Kassiotis G Pasparakis M Alexopoulou L Kollias G TNF-alpha transgenic and knockout models of CNS inflammation and degeneration J Neuroimmunol 1997 72 137 141 9042105 10.1016/S0165-5728(96)00184-1
Lannuzel A Aubourg P Tardieu M Excessive production of tumour necrosis factor alpha by peripheral blood mononuclear cells in X-linked adrenoleukodystrophy Eur J Paediatr Neurol 1998 2 27 32 10726843 10.1016/1090-3798(98)01002-7
Hofman FM Hinton DR Johnson K Merrill JE Tumor necrosis factor identified in multiple sclerosis brain J Exp Med 1989 170 607 612 2754393 10.1084/jem.170.2.607
Taupin V Renno T Bourbonniere L Peterson AC Rodriguez M Owens T Increased severity of experimental autoimmune encephalomyelitis, chronic macrophage/microglial reactivity, and demyelination in transgenic mice producing tumor necrosis factor-alpha in the central nervous system Eur J Immunol 1997 27 905 913 9130643
Bachmann R Eugster HP Frei K Fontana A Lassmann H Impairment of TNF-receptor-1 signaling but not fas signaling diminishes T-cell apoptosis in myelin oligodendrocyte glycoprotein peptide-induced chronic demyelinating autoimmune encephalomyelitis in mice Am J Pathol 1999 154 1417 1422 10329594
Eugster HP Frei K Bachmann R Bluethmann H Lassmann H Fontana A Severity of symptoms and demyelination in MOG-induced EAE depends on TNFR1 Eur J Immunol 1999 29 626 632 10064079 10.1002/(SICI)1521-4141(199902)29:02<626::AID-IMMU626>3.3.CO;2-1
Probert L Eugster HP Akassoglou K Bauer J Frei K Lassmann H Fontana A TNFR1 signalling is critical for the development of demyelination and the limitation of T-cell responses during immune-mediated CNS disease Brain 2000 123 2005 2019 11004118 10.1093/brain/123.10.2005
Mizuno T Butler D Kurotani T Komatsu Y Kawanokuchi J Kato H Mitsuma N Suzumura A Neuroprotective role of phosphodiesterase inhibitor ibudilast on neuronal cell death induced by activated microglia Neuropharmacology 2004 46 404 411 14975696 10.1016/j.neuropharm.2003.09.009
Giri S Jatana M Rattan R Won JS Singh I Singh AK Galactosylsphingosine (psychosine)-induced expression of cytokine-mediated inducible nitric oxide synthases via AP-1 and C/EBP: implications for Krabbe disease FASEB J 2002 16 661 672 11978730 10.1096/fj.01-0798com
Baker D Butler D Scallon BJ O'Neill JK Turk JL Feldmann M Control of established experimental allergic encephalomyelitis by inhibition of tumor necrosis factor (TNF) activity within the central nervous system using monoclonal antibodies and TNF receptor-immunoglobulin fusion proteins Eur J Immunol 1994 24 2040 2048 8088324
Selmaj KW Raine CS Experimental autoimmune encephalomyelitis: immunotherapy with anti-tumor necrosis factor antibodies and soluble tumor necrosis factor receptors Neurology 1995 45 S44 S49 7783912
Liu J Marino MW Wong G Grail D Dunn A Bettadapura J Slavin AJ Old L Bernard CC TNF is a potent anti-inflammatory cytokine in autoimmune-mediated demyelination Nat Med 1998 4 78 83 9427610 10.1038/nm0198-078
TNF neutralization in MS: results of a randomized, placebo-controlled multicenter study. The Lenercept Multiple Sclerosis Study Group and The University of British Columbia MS/MRI Analysis Group. Neurology 1999 53 457 465 10449104
Mohan N Edwards ET Cupps TR Oliverio PJ Sandberg G Crayton H Richert JR Siegel JN Demyelination occurring during anti-tumor necrosis factor alpha therapy for inflammatory arthritides Arthritis Rheum 2001 44 2862 2869 11762947 10.1002/1529-0131(200112)44:12<2862::AID-ART474>3.0.CO;2-W
Thomas CW JrWeinshenker BG Sandborn WJ Demyelination during anti-tumor necrosis factor alpha therapy with infliximab for Crohn's disease Inflamm Bowel Dis 2004 10 28 31 15058523 10.1097/00054725-200401000-00004
Panitch HS Early treatment trials with interferon beta in multiple sclerosis Mult Scler 1995 S17 21 9345392
Placebo-controlled multicentre randomised trial of interferon beta-1b in treatment of secondary progressive multiple sclerosis. European Study Group on interferon beta-1b in secondary progressive MS. Lancet 1998 352 1491 1497 9820296 10.1016/S0140-6736(98)10039-9
McDonald WI New treatments for multiple sclerosis BMJ 1995 310 345 346 7866205
Wolinsky JS Copolymer 1: a most reasonable alternative therapy for early relapsing-remitting multiple sclerosis with mild disability Neurology 1995 45 1245 1247 7617175
Weber MS Starck M Wagenpfeil S Meinl E Hohlfeld R Farina C Multiple sclerosis: glatiramer acetate inhibits monocyte reactivity in vitro and in vivo Brain 2004 127 1370 1378 15090474 10.1093/brain/awh163
Gayo A Mozo L Suarez A Tunon A Lahoz C Gutierrez C Interferon beta-1b treatment modulates TNFalpha and IFNgamma spontaneous gene expression in MS Neurology 1999 52 1764 1770 10371521
Laske C Oschmann P Tofighi J Kuehne SB Diehl H Bregenzer T Kraus J Bauer R Chatzimanolis N Kern A Traupe H Kaps M Induction of sTNF-R1 and sTNF-R2 by interferon beta-1b in correlation with clinical and MRI activity Acta Neurol Scand 2001 103 105 113 11227128 10.1034/j.1600-0404.2001.103002105.x
Arnett HA Mason J Marino M Suzuki K Matsushima GK Ting JP TNF alpha promotes proliferation of oligodendrocyte progenitors and remyelination Nature Neuroscience 2001 4 1116 1122 11600888 10.1038/nn738
Hu WH Johnson H Shu HB Activation of NF-kappaB by FADD, Casper, and caspase-8 J Biol Chem 2000 275 10838 10844 10753878 10.1074/jbc.275.15.10838
Yeager AM Brennan S Tiffany C Moser HW Santos GW Prolonged survival and remyelination after hematopoietic cell transplantation in the twitcher mouse Science 1984 225 1052 1054 6382609
Biswas S Pinson DM Bronshteyn IG LeVine SM IL-6 deficiency allows for enhanced therapeutic value after bone marrow transplantation across a minor histocompatibility barrier in the twitcher (globoid cell leukodystrophy) mouse. J Neurosci Res 2001 15 298 307 11494365 10.1002/jnr.1154
Kaye EM Lysosmal Storage Diseases Curr Treat Options Neurol 2001 3 249 256 11282040
| 15813970 | PMC1112612 | CC BY | 2021-01-04 16:38:21 | no | J Neuroinflammation. 2005 Apr 6; 2:10 | utf-8 | J Neuroinflammation | 2,005 | 10.1186/1742-2094-2-10 | oa_comm |
==== Front
Lipids Health DisLipids in Health and Disease1476-511XBioMed Central London 1476-511X-4-81582631910.1186/1476-511X-4-8ResearchLipoprotein docosapentaenoic acid is associated with serum matrix metalloproteinase-9 concentration Solakivi Tiina [email protected] Olli [email protected] Anne [email protected] Mari [email protected]äki Anne [email protected]äki Terho [email protected]öyhtyä Matti [email protected] Hannu [email protected] Seppo T [email protected] Department of Medical Biochemistry, Medical School, University of Tampere, Tampere, Finland2 Institute of Medical Technology, University of Tampere, Tampere, Finland3 Department of Clinical Chemistry, Tampere University Hospital, Tampere, Finland4 Laboratory of Atherosclerosis Genetics, Department of Clinical Chemistry, Medical School, University of Tampere, Tampere, Finland5 Medix Biochemica, Kauniainen, Finland2005 13 4 2005 4 8 8 24 3 2005 13 4 2005 Copyright © 2005 Solakivi et al; licensee BioMed Central Ltd.2005Solakivi 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
Polyunsaturated fatty acids (PUFA) are thought to play important roles in inflammation. The n-3 series is considered as anti-inflammatory, and some studies have reported increased plasma n-3 polyunsaturated fatty acid pattern in chronic inflammatory conditions. In this study we sought to clarify relationships of the levels of arachidonic acid and the polyunsaturated n-3 fatty acid compositions of isolated LDL, HDL2 and HDL3 particles with matrix metalloproteinase-9 (MMP-9), a marker of inflammation.
Results
The subjects were divided into two groups: those with lower and those with higher than the median serum MMP-9 concentration. In all lipoprotein fractions, the mean percentage of docosapentaenoic acid (C22:5n-3) was higher in the group of subjects with higher MMP-9 level than in those with lower serum MMP-9 concentration (P < 0.01 for all). Likewise, the ratio of docosapentaenoic acid to arachidonic acid (C20:4n-6) was higher in the subjects with higher MMP-9 compared with the lower MMP-9 group (P < 0.001 for all).
Conclusion
So far, the evidence for an anti-inflammatory role of the n-3 PUFA has come from dietary interventions. Our results were obtained from a free-living population and indicate that there is a positive correlation between n-3 docosapentaenoic acid and MMP-9. What had triggered the rise in MMP-9 is not known, since serum level of MMP-9 is raised in many inflammatory conditions. These findings may indicate an increased biosynthesis of n-3 polyunsaturated fatty acids in subclinical inflammation.
==== Body
Background
As a consequence of arterial inflammation, serum MMP-9 is increased in primary arteritis [1,2] and in severe CAD [3]. In addition to these processes, elevated serum MMP-9 has been reported in cancer [4], asthma [5] and rheumatoid arthritis [6]. Thus, in inflammation in general there is a rise of serum MMP-9 concentration.
Polyunsaturated fatty acids (PUFA) may play an important role in inflammation. The two classes of polyunsaturated fatty acids (PUFA), the n-6 and n-3 series have opposing physiological functions. Arachidonic acid (20:4n-6), a metabolite of linoleic acid (18:2n-6), is the substrate of cyclooxygenases and lipoxygenases in the production of potent inflammatory eicosanoids. The polyunsaturated fatty acids of the n-3 series (eicosapentaenoic acid 20:5n-3, docosapentaenoic acid 22:5n-3, and docosahexaenoic acid 22:6n-3) in turn inhibit the synthesis of these mediators and produce eicosanoids with much weaker effects [7,8]. As outlined above, the n-3 series is considered as anti-inflammatory.
Low rates of CHD were found in Greenland Eskimos who are exposed to a diet rich in n-3 fatty acid fish oil [9]. Such diets have also been suggested to reduce the risk of inflammatory bowel disease [10]. Interestingly, some studies have reported increased plasma n-3 polyunsaturated fatty acid pattern in chronic inflammatory bowel diseases, such as Crohn's disease and ulcerative colitis [11,12]. In fact, significantly higher levels of docosapentaenoic acid (22:5n-3) and lower levels of arachidonic acid (20:4n-6) have been reported in serum of patients with long-standing Crohn's disease in comparison with controls [13]. Moreover, ileal and colonic fatty acids profiles in these patients show a substantial increase of the highly polyunsaturated fatty acids [14]. Patients with Crohn's disease have also increased percentage of n-3 PUFA in peripheral blood monocytes [15]. Rats with experimental ulcerative colitis have increased n-3 fatty acids in colon mucosa [16]. These findings were suggested to indicate an increased biosynthesis of polyunsaturated fatty acids in inflammation due to increased activity in the desaturase/elongation enzymes by hitherto unknown mechanisms. With more severe inflammation the essential fatty acids that are precursors to the polyunsaturated fatty acids are reduced, as is observed in patients with rheumatoid arthritis [17] and active colitis ulcerosa [11].
In this study we sought to clarify relationships of the levels of arachidonic acid and the polyunsaturated n-3 fatty acid compositions of isolated LDL, HDL2 and HDL3 particles with MMP-9, a marker of inflammation.
Results
The subjects were divided into two groups: those with higher and those with lower MMP-9 levels than the median of 43 μg/l. The characteristics of the groups and the arachidonic acid, eicosapentaenoic acid, docosapentaenoic acid and docosahexaenoic acid compositions of LDL, HDL2 and HDL3 are shown in Tables I and II, respectively. The background characteristics of the two groups were very similar. The gender distribution was the same in both groups. Mean values for age, BMI, total cholesterol, triacylglycerol HDL or LDL cholesterol did not differ between the two groups. Thus, these background variables were not associated with serum MMP-9 concentrations.
Table 1 Background characteristics of the MMP-9 subgroupsa
Group Ib n = 29 Group IIc n = 29 Significanced (p-value)
Sex (female/male) 16/13 16/13 1.000000
MMP9 (μg/l) 32.8 ± 5.1 53.8 ± 8.4 0.000000
Age (years) 38.6 ± 10.4 39.7 ± 11.1 0.70
BMI (kg/m2) 24.9 ± 4.0 23.9 ± 3.1 0.32
Cholesterol (mmol/l) 5.44 ± 0.81 5.48 ± 1.13 0.90
Triacylglycerol (mmol/l) 1.26 ± 0.66 1.44 ± 1.26 0.51
HDL-cholesterol (mmol/l) 1.65 ± 0.34 1.64 ± 0.42 0.90
LDL-cholesterol (mmol/l) 3.22 ± 0.77 3.25 ± 1.11 0.90
ApoA-I (g/l) 1.61 ± 0.21 1.56 ± 0.21 0.47
ApoB (g/l) 0.90 ± 0.20 0.89 ± 0.26 0.76
fB-Glucose (mmol/l) 4.5 ± 0.6 4.4 ± 0.5 0.36
aMeans ± SD are given; b MMP-9 <43 ug/l; c MMP-9 >43 ug/l; dMann-Whitney U-test
Table 2 Percentages of PUFA in total fatty acids from separated lipoproteins in the MMP-9 subgroupsa
PUFAb Group Ic n = 29 Group IId n = 29 Significancee (p-value)
Arachidonic Acid
LDL-particles 5.61 ± 1.11 5.25 ± 0.87 0.25
HDL2-particles 7.02 ± 1.24 6.70 ± 1.18 0.43
HDL3-particles 7.86 ± 1.46 7.42 ± 1.13 0.28
Eicosapentaenoic Acid
LDL-particles 1.13 ± 0.39 1.25 ± 0.74 0.99
HDL2-particles 1.24 ± 0.40 1.37 ± 0.78 0.98
HDL3-particles 1.38 ± 0.43 1.50 ± 0.87 0.86
Docosapentaenoic Acid
LDL-particles 0.39 ± 0.08 0.47 ± 0.09 0.0015
HDL2-particles 0.63 ± 0.14 0.74 ± 0.14 0.0057
HDL3-particles 0.69 ± 0.15 0.80 ± 0.15 0.0080
Docosahexaenoic Acid
LDL-particles 2.13 ± 0.41 2.05 ± 0.62 0.35
HDL2-particles 3.49 ± 0.63 3.27 ± 0.92 0.16
HDL3-particles 3.66 ± 0.70 3.47 ± 0.97 0.21
aMeans ± SD are given; bPUFA, polyunsaturated fatty acids; cMMP-9 <43 ug/l;
dMMP-9 >43 ug/l; eMann-Whitney U-test
When the polyunsaturated fatty acids were examined, there was no difference between the groups in the proportions of arachidonic acid, eicosapentaenoic acid, and docosahexaenoic acid in the lipoprotein fractions. However, the mean percentage of docosapentaenoic acid in LDL, HDL2 and HDL3 was significantly higher in the group of subjects with the higher MMP-9 level in comparison with those with the lower serum MMP-9 concentration (Table 2). Furthermore, the ratio of docosapentaenoic acid to arachidonic acid in LDL, HDL2 and HDL3 was higher in subjects with higher MMP-9 than in those with lower MMP-9 (p < 0.01 for all, data not shown).
Discussion
Our result of a positive association between MMP-9 and the percentage of docosapentaenoic acid in isolated lipids of HDL and LDL may reflect increased metabolism of long chain fatty acids in subclinical inflammation, analogous to the situation seen in chronic non-active inflammatory bowel disease [12-15]. Our subjects were subjectively healthy, and the cause of variation of serum MMP-9 concentration remains unknown.
Docosapentenoic acid is one of the three major n-3 long chain polyunsaturated fatty acids in fish and marine oils. In addition to diet, eicosapentaenoic acid, docosapentaenoic acid, and docosahexaenoic acid are obtained by synthesis in the body from α-linolenic acid (18:3n-3) through a series of desaturation and elongation reactions. There is a possible explanation why increased synthesis of n-3 fatty acids could lead to increase in docosapentaenoic acid and not docosahexaenoic acid. It seems that docosapentaenoic acid is not easily further metabolized to docosahexaenoic acid in the human body because the pathway requires a rate-limiting Δ-6 desaturase reaction, chain elongation and translocation of the resulting 24:6n-3 to peroxisomes for oxidation [19]. N-3 fatty acids are often treated as a unity, and the effects of docosapentaenoic acid are rarely compared with those of eicosapentaenoic acid, and docosahexaenoic acid. Therefore the functions of docosapentaenoic acid are more poorly known.
Dietary supplementation with n-3 fatty acids has thought to be of benefit in the management of inflammatory diseases, because several studies have shown that n-3 fatty acids can inhibit the synthesis and release of proinflammatory cytokines such as tumor necrosis factor alpha and interleukin-1-β that are produced during inflammation [20]. These cytokines increase MMP-9 synthesis and activation by inflammatory cells [21]. Therefore, it is possible that a chronic inflammatory condition through increased utilization of anti-inflammatory n-3 fatty acids leads to increased activity of the biosynthesis of n-3 fatty acids. These changes would be seen in all lipoprotein classes through exchange of lipids.
Conclusion
In conclusion, our results indicate that there is a positive correlation between n-3 docosapentaenoic acid and MMP-9, a potential marker of inflammation. What had triggered the rise in MMP-9 is not known. Together with previous studies, our findings suggest an increased biosynthesis of n-3 polyunsaturated fatty acids in subclinical inflammation.
Methods
Subjects
A total of 59 subjectively healthy 20 to 60 year-old women (n = 32) and men (n = 27) were recruited amongst the personnel and students of Medical School of the University of Tampere and Tampere University Hospital. All participants filled a questionnaire, where emphasis was given to their health status (diseases and use of medication) in addition to health related behavior (smoking, use of alcohol and vitamins). All participants gave their written consents to the study. The study protocol was approved by the ethics committee of the Tampere University Hospital.
Laboratory methods
Blood samples were taken from the antecubital vein into suitable tubes (Vacuette, Greiner) using minimal stasis after a 12-hour fast while the subjects were seated (after a 15-min rest). Samples for the isolation of lipoproteins were taken into EDTA-containing tubes, immediately placed in ice, and plasma was separated after centrifugation (Heraeus, 2000 × g, + 4°C). EDTA-plasmas were supplemented with sucrose (0.6% w/v final concentration). All samples were kept frozen at -70°C until analyzed. Fasting blood glucose concentration was determined from capillary blood using Hemocue Glucose Analyzer (Hemocue, Ängelholm Sweden).
Plasma cholesterol, HDL cholesterol, triacylglycerol, apoA-I and apoB concentrations were measured with Cobas Integra 700 automatic analyzer using reagents and calibrators as recommended by the manufacturer (Roche Diagnostic, Basel, Schwitzerland). LDL cholesterol was calculated according to Friedewald.
For the accurate assessment of serum MMP-9, aliquots of sera were removed and stored at -70°C in a freezer that was not in daily use until analysis. Quantification of immunoreactive MMP-9 was carried out by enzyme-linked immunosorbent assay (ELISA) (Diabor Ltd, Oulu, Finland). ELISAs were performed on 96-well microtiter plates using standard protocols. Recombinant MMP-9 was used as standard. The microtiter plate was coated with the monoclonal antibody (code GE-213). The bound proteins from serum and standards were detected with a secondary polyclonal antibody produced in chicken against MMP-9. A peroxidase-labeled anti-chicken-IgG (Chemicon, USA) was used for detection of the bound secondary antibody. O-phenylenediamine (OPD) was used to visualize the peroxidase label. The color formation was measured at 450 nm (Anhos 2000 microplate reader) and calculations were done using a Multicalc program (Wallac, Turku, Finland). The monoclonal antibody recognized both the free MMP-9 and that bound to its inhibitor, tissue inhibitor of metalloproteinases-1 (TIMP-1) [2,18].
Lipoproteins were fractionated by isopycnic density gradient ultracentifugation. Two ml of plasma was mixed with 4.0 ml of d 1.35 g/l NaCl/KBr solution in a 14 × 95 mm tube (Beckman, Palo Alto, USA) and then successively overlayered with 4.5 ml of a d 1.006 salt solution and 1.0 ml of distilled water. For centrifugation a Beckman SW40 Ti rotor, at 36000 rpm for 40 hours in a Beckman L60 centrifuge at 10°C was used. After ultracentrifugation the contents of the tubes were fractionated with an Isco gradient fractionator (Model 640, Lincoln, USA). The 280 nm absorbance of the effluent was continuously monitored with an Isco UA-5 absorbance detector. The different lipoproteins were well separated by the resulting slightly curving salt gradient. The fractions belonging to LDL, HDL2 and HDL3 were pooled on the basis of the absorbance curve.
The total fatty acid compositions of the ultracentrifugally isolated LDL, HDL2 and HDL3 particles were analyzed by gas-liquid chromatography. Lipids were extracted with chloroform/methanol, partitioned and the chloroform phase was dried under N2. The lipids were then hydrolyzed and transesterified with H2SO4 in dry methanol at 85°C for 2 h under N2. Following the addition of water, methyl esters of the fatty acids were extracted with petroleum ether and analyzed in a Shimadzu GC-14A gas chromatograph (Shimadzu Corporation, Kyoto, Japan) with a flame ionization detector using a Supelco SP 2560 capillary column (100 m, 0.25 mm I.D., 0.20 μm film thickness). The carrier gas was helium. The column temperature was 180°C for 15 min, then programmed to increase at 3°C/min to 230°C and held for 40 min. The individual fatty acids were identified with the aid of a standard mixtures of methyl esters (Lipid standards 189-15 and 189-17, Sigma). The areas were measured with a Shimadzu C-R4A Chromatopac Integrator and the results expressed as percentages of the sum of all fatty acids from 14:0 to 22:6n-3. As a control sample we used a pool of isolated HDL that was suitably diluted and kept frozen at -70°C. The inter assay coefficient of variation for the percentage of different fatty acids ranged from 0.3 to 4.4 %.
Statistical analysis
Results are expressed as means ± standard deviation. Plasma triacylglycerol concentrations were used as their logarithms but reported as original results. Comparisons were conducted by Mann-Whitney U-test. Univariate associations between variables were analyzed using Spearman's correlation coefficients. The Statistica for Windows (version 5.1) software package (Statsoft Inc., Oklahoma, USA) was used for statistical analysis.
Authors' contributions
TS and OJ conceived of the study, performed the statistical analyses and wrote the initial manuscript. MP, AS and AK carried out the laboratory analyses. MH designed the ELISA. TL, HJ and STN participated in the study design and coordination and helped to draft the manuscript. All authors read and approved the final manuscript.
Acknowledgements
The authors thank Marita Koli, Nina Peltonen, Marja Jousimies and Marjo Virkki for skilful laboratory assistance. The study was supported by the Medical Research Fund of Tampere University Hospital, The Finnish Foundation of Cardiovascular Research and the Finnish Association of Clinical Biochemists
==== Refs
Sorbi D French DL Nuovo GJ Kew RR Arbeit LA Gruber BL Elevated levels of 92-kD type IV collagenase (matrix metalloproteinase 9) in giant cell arteritis Arthritis Rheum 1996 39 1747 1753 8843867
Nikkari ST Hoyhtya M Isola J Nikkari T Macrophages contain 92-kD gelatinase (MMP-9) at the site of degenerated internal elastic lamina in temporal arteritis Am J Pathol 1996 149 1427 33 8909231
Kalela A Koivu TA Sisto T Kanervisto J Hoyhtya M Sillanaukee P Lehtimaki T Nikkari ST Serum matrix metalloproteinase-9 concentration in angiographically assessed coronary artery disease Scand J Clin Lab Invest 2002 62 337 42 12387578 10.1080/00365510260296483
Sonnante AM Correale M Linsalata M Di Leo A Guerra V Circulating levels of matrix metalloproteinase-9 in patients with colorectal cancer Scand J Gastroenterol 2000 35 671 72 10912671 10.1080/003655200750023679
Belleguic C Corbel M Germain N Lena H Boichot E Delaval PH Lagente V Increased release of matrix metalloproteinase-9 in the plasma of acute severe asthmatic patients Clin Exp Allergy 2002 32 217 23 11929485 10.1046/j.1365-2222.2002.01219.x
Ahrens D Koch AE Pope RM Stein-Picarella M Niedbala MJ Expression of matrix metalloproteinase 9 (96-kd gelatinase b) in human rheumatoid arthritis Arthritis Rheum 1996 39 1576 87 8814070
Mohrhauer H Christiansen K Gan MV Deubig M Holman RT Chain elongation of linoleic acid and its inhibition by other fatty acids in vitro J Biol Chem 1967 242 4507 14 4383633
Needleman P Raz A Minkes MS Ferrendelli JA Sprecher H Triene prostaglandins: prostacyclin and thromboxane biosynthesis and unique biological properties Proc Natl Acad Sci USA 1979 76 944 48 218223
Dyerberg J Bang HO A hypothesis on the development of acute myocardial infarction in greenlanders Scand J Clin Labs Invest 1982 7 13
Belluzzi A N-3 fatty acids for the treatment of inflammatory bowel diseases Proc Nutr Soc 2002 61 391 95 12296296 10.1079/PNS2002171
Esteve-Comas M Ramirez M Fernandez-Banares F Abad-Lacruz A Gil A Cabre E Gonzalez-Huix F Moreno J Humbert P Guilera M Plasma polyunsaturated fatty acid pattern in active inflammatory bowel disease Gut 33 1365 69 1446861
Esteve-Comas M Nunez MC Fernandez-Banares F Abad-Lacruz A Gil A Cabre E Gonzalez-Huix F Bertran X Gassull MA Abnormal plasma polyunsaturated fatty acid pattern in non-active inflammatory bowel disease Gut 1993 34 1370 73 8244103
Geerling BJ v Houwelingen AC Badart-Smook A Stockbrugger RW Brummer RJ Fat intake and fatty acid profile in plasma phospholipids and adipose tissue in patients with crohn's disease, compared with controls Am J Gastroenterol 1999 94 410 17 10022638 10.1111/j.1572-0241.1999.869_a.x
Buhner S Nagel E Korber J Vogelsang H Linn T Pichlmayr R Ileal and colonic fatty acid profiles in patients with active crohn's disease Gut 1994 35 1424 8 7959199
Trebble TM Arden NK Wootton SA Mullee MA Calder PC Burdge GC Fine DR Stroud MA Peripheral blood mononuclear cell fatty acid composition and inflammatory mediator production in adult Crohn's disease Clin Nutr 2004 23 647 55 15297102 10.1016/j.clnu.2003.10.017
Nieto N Giron MD Suarez MD Gil A Changes in plasma and colonic mucosa fatty acid profiles in rats with ulcerative colitis induced by trinitrobenzene sulfonic acid Dig Dis Sci 1998 43 2688 95 9881501 10.1023/A:1026607428716
Jacobsson L Lindgarde F Manthorpe R Akesson B Correlation of fatty acid composition of adipose tissue lipids and serum phosphatidylcholine and serum concentrations of micronutrients with disease duration in rheumatoid arthritis Ann Rheum Dis 1990 49 901 5 1701623
Kalela A Koivu TA Sisto T Kanervisto J Hoyhtya M Sillanaukee P Lehtimaki T Nikkari ST Serum matrix metalloproteinase-9 concentration in angiographically assessed coronary artery disease Scand J Clin Lab Invest 2002 62 337 42 12387578 10.1080/00365510260296483
Sprecher H Metabolism of highly unsaturated n-3 and n-6 fatty acids Biochim Biophys Acta 2000 1486 219 31 10903473
Endres S Ghorbani R Kelley VE Georgilis K Lonnemann G van der Meer JW Cannon JG Rogers TS Klempner MS Weber PC The effect of dietary supplementation with n-3 polyunsaturated fatty acids on the synthesis of interleukin-1 and tumor necrosis factor by mononuclear cells N Engl J Med 1989 320 265 715 2783477
Wahl LM Corcoran ML Regulation of monocyte/macrophage metalloproteinase production by cytokines J Periodontol 1993 64 467 73 8391076
| 15826319 | PMC1112613 | CC BY | 2021-01-04 16:39:19 | no | Lipids Health Dis. 2005 Apr 13; 4:8 | utf-8 | Lipids Health Dis | 2,005 | 10.1186/1476-511X-4-8 | oa_comm |
==== Front
Malar JMalaria Journal1475-2875BioMed Central London 1475-2875-4-211585751210.1186/1475-2875-4-21ResearchExpression of Plasmodium falciparum erythrocyte membrane protein 1 in experimentally infected humans Lavstsen Thomas [email protected] Pamela [email protected] Cornelus C [email protected] Ali [email protected] Anja TR [email protected] Robert [email protected] Lars [email protected] Thor G [email protected] Trine [email protected] Centre for Medical Parasitology at Institute for Medical Microbiology and Immunology, University of Copenhagen, Panum Institute 24-2, Blegdamsvej 3, 2200 Copenhagen N, Denmark2 Centre for Medical Parasitology at Department of Infectious Diseases, Copenhagen University Hospital (Rigshospitalet), Copenhagen, Denmark3 Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands2005 27 4 2005 4 21 21 9 2 2005 27 4 2005 Copyright © 2005 Lavstsen et al; licensee BioMed Central Ltd.2005Lavstsen 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
Parasites causing severe malaria in non-immune patients express a restricted subset of variant surface antigens (VSA), which are better recognized by immune sera than VSA expressed during non-severe disease in semi-immune individuals. The most prominent VSA are the var gene-encoded Plasmodium falciparum erythrocyte membrane protein 1 (PfEMP1) family, which is expressed on the surface of infected erythrocytes where it mediates binding to endothelial receptors. Thus, severe malaria may be caused by parasites expressing PfEMP1 variants that afford parasites optimal sequestration in immunologically naïve individuals and high effective multiplication rates.
Methods
var gene transcription was analysed using real time PCR and PfEMP1 expression by western blots as well as immune plasma recognition of parasite cultures established from non-immune volunteers shortly after infection with NF54 sporozoites.
Results
In cultures representing the first generation of parasites after hepatic release, all var genes were transcribed, but GroupA var genes were transcribed at the lowest levels. In cultures established from second or third generation blood stage parasites of volunteers with high in vivo parasite multiplication rates, the var gene transcription pattern differed markedly from the transcription pattern of the cultures representing first generation parasites. This indicated that parasites expressing specific var genes, mainly belonging to group A and B, had expanded more effectively in vivo compared to parasites expressing other var genes. The differential expression of PfEMP1 was confirmed at the protein level by immunoblot analysis. In addition, serological typing showed that immune sera more often recognized second and third generation parasites than first generation parasites.
Conclusion
In conclusion, the results presented here support the hypothesis that parasites causing severe malaria express a subset of PfEMP1, which bestows high parasite growth rates in individuals with limited pre-existing immunity.
==== Body
Background
Plasmodium falciparum-encoded variant surface antigens (VSA) are expressed on the surface of infected erythrocytes (IE) and mediate binding to a range of endothelial cell receptors [1]. Endothelial adhesion contributes to the particular virulence of the P. falciparum and most likely has evolved as a mechanism to avoid parasite clearance in the spleen [2-4]. Individuals living in areas of intense parasite transmission develop immunity towards severe malaria early in life [5]. Parasites causing severe malaria in young children with limited pre-existing immunity tend to express a limited, relatively conserved subset of VSA (VSASM) that is more often and better recognized by antibodies from most parasite-exposed individuals than the larger and more diverse VSAUM subset expressed by parasites causing uncomplicated malaria [6-8]. It thus appears that expression of VSASM confers a selective advantage in non-immune individuals, perhaps by allowing particularly efficacious endothelial sequestration and consequently high effective growth rates. The best characterized VSA are the var gene-encoded P. falciparum erythrocyte membrane protein 1 (PfEMP1) family [9-11]. Each haploid parasite genome contains 50–60 var genes, of which the 59 var genes annotated in the fully sequenced P. falciparum clone 3D7 can be divided into three major groups, A, B and C, based on sequence analysis [12,13]. The functional relevance of this grouping is supported by the parallel differences in CD36-binding characteristics of PfEMP1 CIDR1α domains. Thus, GroupA CIDR1α domains do not bind CD36, whereas CIDR1α domains encoded by GroupB and GroupC var genes do [14]. The 3D7 PfEMP1 repertoire may well represent the VSAUM -VSASM spectrum observed in field isolates, and recent findings point to GroupA as encoding VSASM-type PfEMP1 molecules in patient isolates [13,15-17]. Unfortunately, little is known about var gene expression in vivo, and studies have been frustrated by the difficulties in detecting and quantifying expression in parasites with unknown var gene repertoires. This difficulty has been overcome by taking advantage of the knowledge of the var gene repertoire in 3D7 and analyzing var gene expression in NF54 parasites (the parental line of 3D7) isolated from non-immune individuals experimentally infected by mosquito challenge. Immediately upon release from the liver, the parasites appeared to transcribe all var genes, with GroupA genes being the least transcribed. However, within one or two parasite generations this pattern changed, in particular in those parasites exhibiting the fastest in vivo growth rates. Here, only a few genes dominated the var transcript population. The data indicate that PfEMP1-determined differences in growth rates shape the expressed PfEMP1 repertoire, and that some PfEMP1 variants confer high effective parasite multiplication rates in non-immune individuals.
Materials and methods
Malaria parasites
Parasites were isolated from Dutch volunteers exposed to mosquitoes infected with P. falciparum isolate NF54[18] as part of ongoing studies of experimental P. falciparum infections. On day 0, ten non-immune volunteers were subjected to two or five infectious bites. Chloroquine treatment was initiated on the first day a thick smear was positive. Parasite cultures were established from 400 microlitres of packed blood cells drawn on days 8, 9, and 10 and parasites were cultured in vitro for 27 or 33 days (Figure 1) to obtain sufficient parasite material for DNA, RNA and protein analysis. The parasites were cultured in 0 Rh+ erythrocytes as described [19], with the addition of 2% non-immune human serum to the culture media. Long-term in vitro 3D7 cultures expressing VSAUM-type antigens or selected in vitro to express VSASM-type antigens were used as controls[20].
Figure 1 A: Parasite densities in the six volunteers from whom parasite isolates were established. Note that parasitaemia scales are different. Closed circles indicate time points where parasite density was determined by PCR. Time points where blood samples were cultured successfully are underlined. A cross indicates time of chloroquine treatment. B: Representation of parasite generation and stage composition of a P. falciparum infection after liver release. Parasites were first detected on day 6,33. Estimates of parasite release from the liver (~24 hours, over all volunteers), duration of circulating stages (1,18 days) and adhesive stages (0,64 days) were taken from reference 24. Time of blood sampling is framed. When sampling blood the early circulating stages are isolated, i.e. blood drawn on day 8 predominantly contains first-generation parasites after liver release.
DNA, RNA, cDNA and quantitative real-time PCR
The development of parasitaemia was monitored by quantitative real-time PCR as described [21]. DNA, RNA and cDNA for var gene transcription analysis were prepared from synchronized parasite cultures as described [17]. Quantitative real-time PCR was performed using a Rotorgene thermal cycler system (Corbett Research, Motlake, Australia). Real-time PCR-optimized and gene-specific primers for each of all full-length var genes and a pseudogene in the 3D7 isogenic NF54 P. falciparum genome were those described in [22], except for PFI1830c. Real time primers for this gene were forward 5'ACAACAATTTCGCAAGCAAG 3', reverse 5'TTCCTCTGCCTCCTCTTCAT 3'. Standard curves for the estimation of product-related fluorescent bias and amplification efficiencies were generated for all primer pairs. For 15 primer sets, standard curves were generated both from dilution series of genomic DNA and from cloned gene fragments [17]. As the two approaches led to identical standard curves (not shown), standard curves for the remaining genes were determined from genomic DNA only. The standard curves were linear across a range of seven logs of DNA concentrations (R = 0.9779 to 0.9995) with amplification efficiencies between 90 and 101%. Standard curves were used for primer bias corrections in calculations of absolute transcript levels. The detection limit of the system was ≥ 20 copies. The housekeeping genes seryl-tRNA synthetase and fructose-bisphosphate aldolase have uniform transcription profiles throughout the parasite life cycle[22], and were used as endogenous controls. Differences in var transcript distribution between samples were calculated by the ΔΔ CT method using the endogenous controls for normalization.
Immunoblot analysis and flow cytometry
Immunoblot analysis was performed as previously described [17]. Flow cytometryand plasma from malaria-exposed children and adults were used to classify the VSA expressed by parasites isolated on days 8, 9 and 10 as previously described[7,20,23]
Results
Experimental infections and establishment of parasite culture lines
Successful infections were established in eight of the 10 volunteers. Despite low parasitaemias, a total of 13 parasite culture lines from blood collected from six of the volunteers at three time points after infection were established (Figure 1A). The highest parasite growth rates were observed in volunteers 1 and 2 with peak parasitaemias of 13,832 parasites/ml and 4,637 parasites/ml, respectively, whereas peak parasitaemias in all other volunteers were below 1,400 parasites/ml. As recognized from previous similar human vaccination trials [21], parasitaemias fluctuated in a distinct manner, probably reflecting liver release, sequestration of trophozoite/schizonts and release of new generations of merozoites from schizonts. Parasites obtained on day 8 were predominantly first-generation blood parasites, assuming that parasites were released from the liver between days 6,33 and 7,33[24] (Figure 1B). Similarly, day9 parasites were assumed to represent second-generation parasites, and day10 parasites a mixture of second and third generation parasites.
Uniform transcription of var genes in first-generation asexual parasites
The pattern of var gene transcription was surprisingly consistent in all the six parasite lines obtained on day8 from six of the volunteers (Figure 2). Transcripts of all var genes could be detected in ring-stage cultures, and most were transcribed at roughly similar levels. Interestingly, nine of the 10 lowest transcribed genes belonged to var subgroup A or B/A, which have previously been associated with severe malaria [13]. In agreement with previous studies [17] all var genes were transcribed at markedly lower levels in thetrophozoite/schizont-stage compared to ring-stage parasites (data not shown). The pseudo-gene PFE1640w (var1) behaved differently and was expressed at similar levels by late stage parasites, comprising approximately 20% of the total number of var gene transcripts in trophozoite-stage parasites (data not shown).
Figure 2 Var gene transcription profile of NF54 ring-stage parasite cultures established on day 8 from six volunteers. The mean transcription levels ± 1 SD of primer bias-corrected and normalized values of the six cultures are shown relative to the overall mean var transcription level. Var gene name and group are indicated.
Marked changes in the var gene transcription patterns from the first to the second and third generations of asexual parasites
The var gene transcription patterns of the five isolates obtained on day9 and the two obtained at day10 were different from those of the day8 isolates, in particular in the isolates obtained from volunteers 1 and 2 in whom high parasite growth was observed (Figure 1 and 3). Ring-stage parasites from volunteer 1 showed remarkably large transcriptional changes for five var genes. Four (PF11_0008, PFD1235w, MAL6P1.1, MAL7P1.55) were transcribed at much higher (>20-fold) and one gene (PFA0015c) at markedly lower levels (12-fold) in the day10 isolate compared to the day8 isolate. The estimation of absolute copy numbers (not shown) revealed that PF11_0008, MAL6P1.1 and MAL7P1.55 were the three highest transcribed var genes in the day10 isolate, comprising approximately 22%, 40% and 8% of all var transcripts, respectively. The abovementioned five genes also showed the most pronounced changes in gene transcription (10–20 fold) when trophozoite stage parasites from days 8 and 10 were compared. The most prominent change among ring-stage parasites from volunteer 2 was the 15-fold increased transcription of PFD0020c, which was the highest (~8%) transcribed var gene in the day 9 isolate (Figure 3). Only minor changes in var gene transcription patterns between days 8 and 9/10 were observed in the parasites isolated from volunteer 4, 6 and 10.
Figure 3 Fold difference in var gene transcription between NF54 ring-stage parasites isolated from the same volunteer on different days. Genes are sorted by gene groups as defined in [13]. Note that the fold-change scale for volunteer 1 is different from the other panels. Vertical dashed lines mark an arbitrarily defined two-fold cut off value for biologically significant changes in var gene transcription. Experiments with volunteer 1 were repeated three times and results are shown as means ± SD.
Overall, none of the var genes had a consistently altered transcript proportion in all volunteers. However, transcription of seven genes (PFD0005w, PFD0020c, PFD1235w, MAL6P1.1, PF10_0406, PF11_0007 and PFL0005w) increased in more than one volunteer, whereas transcription of three genes (PFA0015c, MAL6P1.4 and PFE1640w) decreased in more than one volunteer (Figure 3). In general, analysis of ring-stage and trophozoite/schizont-stage parasites yielded similar results.
Expected changes in PfEMP1 expression between first, second and third generation asexual parasites were confirmed by western blot analysis
To investigate PfEMP1 translation in the 13 isolates, western blot analysis were performed on protein extractions of trophozoite/schizont-stage cultures using rabbit and murine antisera raised against the conserved acidic terminal segment (ATS) and against DBL5δ of PFD1235w, respectively (Figure 4). The analysis of parasites from volunteer 1 revealed differential expression from day8 to day10 (Figure 4A). Thus, a high molecular weight band of around 400 kDa was seen only in the day 10 culture when using the ATS-specific antibody (Figure 4A). This band size corresponds to the expected size of PFD1235w, which also exhibited a large increase in transcription from day8 to day10 among parasites from this donor, and its identity was confirmed using the PFD1235w DBL5δ antibody (Figure 4B). PFD1235w could not be detected in any of the other cultures, although small increases in transcription were found in several of the cultures (not shown). Two bands of 350 kDa and 260 kDa were particularly intense in the day10 culture of volunteer 1 (Figure 4A). This finding correlates with the expected sizes and elevated transcription of PF11_0008 (345 kDa), MAL6P1.1 and MAL7P1.55 (both ~258 kDa). Transient bands at around 230 kDa and 310 kDa appeared to emerge on day 9 and disappear on day 10 in the cultures of volunteer 1. No obvious identity could be assigned to these proteins. In the remaining cultures, differential PfEMP1 expression detected by the ATS antibody was found in volunteers 6 and 10. In both cases a band of 310 kDa was observed in day10 and 9 parasites, respectively. The best candidate gene for this band is PFD0005c, which is predicted to have the observed size and was found to be more highly transcribed in the these cultures. All day 8 cultures were very similar and all appeared to dominantly express PfEMP1s around 260 kDa, which corresponds to the expected sizes of the highest transcribed genes in these cultures (figure 2).
Figure 4 PfEMP1 expression in trophozoite-stage cultures of unselected 3D7 (3D7UM), 3D7 selected for expression of VSASM-type IE surface antigens (3D7SM) and of NF54 established from six volunteers on different days after infection. A: Western blot using antibodies (αATS) targeting the acidic terminal segment (ATS), which is conserved between most PfEMP1 types. Black arrows indicate changes in protein expression between isolates of the same volunteer. B: Western blot identifying the αATS-detected 400 kDa band in the day 10 culture of volunteer 1 as PFD1235w/VAR4 using an αDBL5δ antibody. VAR4 is expressed on the surface of the 3D7SM line selected for high immune serum recognition [17].
Second and third generation parasites express VSA that are recognized more frequently by immune plasma than first generation parasites
The VSA phenotype of parasites can be classified relative to each other and in the spectrum between VSASM and VSAUM, depending on the proportion of individuals from an endemic area that possess antibodies to the expressed VSA [7,20]. To establish their VSA phenotype, the serological recognition of the isolated NF54 parasites was tested using a panel of plasma obtained from children and adults living in Coastal Ghana (Figure 5). All day 8 parasite isolates, representing the first generation of asexual blood-stage parasites, expressed VSA that were recognized by IgG in plasma samples from only a minority of the children and from about half of the adults (Figure 5). These parasites were also less well recognized than the standard 3D7 VSAUM line, which dominantly expresses PfEMP1 encoded by GroupC var genes. The seven lines isolated on days 9 or 10 all expressed VSA that were more frequently recognized than the corresponding day8 line (P = 0.01, Wilcoxon signed-ranked test). This trend was particularly clear for the day9/10 lines from volunteers 1, 2 and 10, which were recognized by all of the plasma samples from adults and from the majority of the children, similarly to the recognition of 3D7 selected for expression of VSASM-type IE surface antigens (Figure 5). However, quantitatively the serum recognition of all isolates was lower than that of the 3D7SM line (data not shown). The 3D7 VSASM line has been selected for VSASM expression using IgG from semi-immune children and dominantly expresses the product of PFD1235w (VAR4) on the surface on infected erythrocytes. Interestingly, parasites from volunteers 1, 2 and 10 had an increased transcription of a particular var gene (PFD1235w; var4) (Figure 2) as does 3D7 in response to selection for VSASM expression[17].
Figure 5 Plasma recognition profiles of trophozoite-stage cultures of NF54 established from six volunteers on different days after infection. Profiles of unselected 3D7 (UM) and 3D7 selected in vitro for expression of VSASM-type IE surface antigens (SM) are shown for comparison. Recognition was measured by flow cytometry using IgG from Ghanaian adults and children (see Materials and Methods). Filled boxes indicate mean FITC-fluorescenceindex (MFI)abovea cut-off defined by the mean + 2 standard deviations of 8 Danish control plasma.
Discussion
The inter- and intra-clonal variability of the var genes have frustrated attempts to investigate the roles of the encoded PfEMP1 proteins in pathogenesis and protection. The most common strategy has been to quantify var transcription by counting the frequency of unique sequence tags, amplified by degenerate primers targeting semi-conserved blocks of DBL domains. This has been used to study phenotypically distinct laboratory lines [10,25-29]and parasite strains isolated from patients with defined clinical outcomes [16,30-32] The only previous study of var gene transcription in experimentally infected humans also applied this strategy [33]. However, it requires the sequencing of a large number of clones for statistical significance and is inherently susceptible to primer bias. Together, this makes data interpretation difficult. To overcome these difficulties, sensitive and gene-specific tools were used to analyse in detail the pattern and dynamics of var gene expression in non-immune volunteers infected with a parasite with a known var gene repertoire. The necessity for in vitro expansion of parasites from blood samples with submicroscopic parasitaemias makes this approach susceptible to two separate types of bias. Firstly, cultures were established from a relatively small number of parasites and the transcription profiles of these parasites may not represent the profile of the entire in vivo population. However, in most cultures the founder population was between 100 and 4,000 parasites, and only two cultures, which did not exhibit biased transcription patterns (day 8 culture of volunteers 1 and 4) were established from less than 100 parasites. Secondly, P. falciparum has been reported to switch var gene expression at variable rates in vitro [34,35]. Hence, the var gene expression in the cultures at the time of transcription analysis may not reflect the expression profile in vivo at the time of blood collection. The transcription profiles of the cultures isolated on day 8 were similar and different to the profiles of the parasites isolated on days 9 and 10. This, and the fact that the differential var gene transcription, translation and serological recognition of PfEMP1 correlated with parasite growth, however, indicates that in vitro switching did not invalidate the analyses.
No var gene was dominantly transcribed in the day 8 cultures. Instead, a relatively large group of genes belonging to var groups B and C were expressed at almost similar levels and interestingly, nine of the 10 lowest transcribed genes belonged to group A or B/A. These data imply that the PfEMP1 expression pattern at the beginning of the infection is broad and anticipatory. In addition, the consistent increase in recognition by immune sera with time of infection and the apparent association between high growth rates and differential expression of a few group A and B genes from first to second and third generation imply that it is the host environment that modulates the PfEMP1 expression. This is a plausible scenario because of the large and immediate differences in survival fitness that are likely to be imposed on the asexual parasites by the physiology and pre-existing immunity of the host. While most – or all – of the PfEMP1 variants that can be expressed by a given parasite will be exposed to the immune system according to this model, the majority of the variants are likely to be present too briefly to induce a significant immune response. Crucially, survival fitness differences depend on which PfEMP1 variants are being expressed and necessitate a parasite response much faster than which can be achieved by switching to advantageous var genes. Although differences in switching rates can contribute to the pattern of var gene expression [35] and may well be responsible for the differences in transcription observed on day 8 (Figure 2), differences in survival rates may be far more important in focusing and ordering PfEMP1 expression in vivo. In non-immune individuals this process would be expected to focus expression on the restricted and relatively conserved subset of VSA (VSASM) associated with severe disease in patients with little pre-existing immunity [13,17] It has previously been documented that in vitro selection of 3D7 for acquisition of the VSASM phenotype is associated with expression of a subset of var genes [[15]. Strikingly, four of the five marked differentially transcribed genes (PF11_0008, PFD1235w, MAL7P1.55 and PFA0015c) in the volunteer 1 cultures, were among the few genes differentially transcribed genes upon selection for the 3D7SM phenotype, in which there was selection for expression of PF11_0008, PFD1235w and MAL7P1.55 and against expression of PFA0015.
In the study of Peters et al [33] the clone frequency strategy was used to investigate var transcription profiles of parasites isolated from two volunteers on day 12 and 13 after infection with 3D7 by eight or nine infectious mosquito bites. One transcript, PF11_0007, belonging to var group B, comprised half of the 39 and 41 sequences cloned from the two volunteers respectively. In total 10 and nine different var tags were found and only one belonged to var group A. The parasites were predicted to be 3rd or 4th generation and the parasitaemias in the two volunteers were 18,000 and 212,000 parasites/ml. Thus, these profiles are best compared with that of the day 10 isolate of volunteer 1 presented in this study. Taking into account the potential ambiguities in the interpretation of the results presented by Peters et al, we believe that the two data sets could reflect similar dynamics in the human host.
Conclusion
In conclusion, the data – in combination with earlier findings – suggest that PfEMP1 expression is determined and ordered mainly by host physiology and immunity, and that this will cause infections in non-immune individuals to be dominated by VSASM-type variants such as those encoded by Group A and B var genes.
Abbreviations
PfEMP1 Plasmodium falciparum erythrocyte membrane protein 1
VSAUM variant surface antigens associated with uncomplicated malaria
VSASM variant surface antigens associated with severe malaria
Authors' contributions
TL carried out the transcription analysis, took part in the parasite sampling and culturing and prepared the manuscript. PM performed the western blot analysis. CCH ran the human vaccination trials. TS managed the sample collection, culturing of parasites as well as performed and the flow cytometry analysis. TL, CCH, TGT and TS conceived the design of the study. All authors helped to draft and approved the final manuscript.
Acknowledgements
We thank Liselotte Wolters for performing QRT-PCR and Petra Schneider and the team at Clinical Center for Malaria Studies, University Medical Center Nijmegen, The Netherlands for sampling parasites from volunteers. Maiken Christensen and Kirsten Pihl are thanked for excellent technical assistance. The PlasmoDB database has been a valuable resource for this work and the database developers and researchers who have made their data available here are thanked. The study received financial support from The Danish Medical Research Council (SSVF grant no. 63686), and the Commission of the European Communities (grant no. QLK2-CT-2002-01197, EUROMALVAC). PM and AS are supported by the Gates Malaria Partnership Ph.D. scholarships.
==== Refs
Kyes S Horrocks P Newbold C Antigenic variation at the infected red cell surface in malaria Annu Rev Microbiol 2001 55 673 707 11544371 10.1146/annurev.micro.55.1.673
David PH Hommel M Miller LH Udeinya IJ Oligino LD Parasite sequestration in Plasmodium falciparum malaria: spleen and antibody modulation of cytoadherence of infected erythrocytes Proc Natl Acad Sci U S A 1983 80 5075 5079 6348780
Howard RJ Barnwell JW Roles of surface antigens on malaria-infected red blood cells in evasion of immunity Contemp Top Immunobiol 1984 12 127 191 6365440
Berendt AR Ferguson DJP Newbold CI Sequestration in Plasmodium falciparum malaria: sticky cells and sticky problems Parasitol Today 1990 6 247 254 15463355 10.1016/0169-4758(90)90184-6
Gupta S Snow RW Donnelly CA Marsh K Newbold C Immunity to non-cerebral severe malaria is acquired after one or two infections Nature Med 1999 5 340 343 10086393 10.1038/6560
Bull PC Kortok M Kai O Ndungu F Ross A Lowe BS Newbold CI Marsh K Plasmodium falciparum -infected erythrocytes: agglutination by diverse Kenyan plasma is associated with severe disease and young host age J Infect Dis 2000 182 252 259 10882604 10.1086/315652
Nielsen M Staalsoe T Kurtzhals J Goka B Dodoo D Alifrangis M Theander TG Akanmori BD Hviid L Plasmodium falciparum variant surface antigen expression varies between isolates causing severe and non-severe malaria and is modified by acquired immunity J Immunol 2002 168 3444 3450 11907103
Nielsen MA Vestergaard LS Lusingu J Kurtzhals JA Giha HA Grevstad B Goka BQ Lemnge MM Jensen JB Akanmori BD Theander TG Staalsoe T Hviid L Geographical and temporal conservation of antibody recognition of Plasmodium falciparum variant surface antigens Infect Immun 2004 72 3531 3535 15155661 10.1128/IAI.72.6.3531-3535.2004
Baruch D Pasloske BL Singh H Bi X Ma X Feldman M Taraschi TF Howard RJ Cloning the P. falciparum gene encoding PfEMP1, a malarial variant antigen and adherence receptor on the surface of parasitized human erythrocytes Cell 1995 82 77 87 7541722 10.1016/0092-8674(95)90054-3
Smith J Chitnis C Craig A Roberts D Hudson-Taylor D Peterson D Pinches R Newbold CI Miller LH Switches in expression of Plasmodium falciparum var genes correlate with changes in antigenic and cytoadherent phenotypes of infected erythrocytes Cell 1995 82 101 110 7606775 10.1016/0092-8674(95)90056-X
Su X Heatwole V Wertheimer S Guinet F Herrfeldt J Peterson D Ravetch JA Wellems TE The large diverse gene family var encodes proteins involved in cytoadherence and antigenic variation of Plasmodium falciparum -infected erythrocytes Cell 1995 82 89 100 7606788 10.1016/0092-8674(95)90055-1
Gardner MJ Hall N Fung E White O Berriman M Hyman RW Carlton JM Pain A Nelson KE Bowman S Paulsen IT James K Eisen JA Rutherford K Salzberg SL Craig A Kyes S Chan MS Nene V Shallom SJ Suh B Peterson J Angiuoli S Pertea M Allen J Selengut J Haft D Mather MW Vaidya AB Martin DM Fairlamb AH Fraunholz MJ Roos DS Ralph SA McFadden GI Cummings LM Subramanian GM Mungall C Venter JC Carucci DJ Hoffman SL Newbold C Davis RW Fraser CM Barrell B Genome sequence of the human malaria parasite Plasmodium falciparum Nature 2002 419 498 511 12368864 10.1038/nature01097
Lavstsen T Salanti A Jensen ATR Arnot DE Theander TG Sub-grouping of Plasmodium falciparum 3D7 var genes based on sequence analysis of coding and non-coding regions Malar J 2003 2 27 14565852 10.1186/1475-2875-2-27
Robinson BA Welch TL Smith JD Widespread functional specialization of Plasmodium falciparum erythrocyte membrane protein 1 family members to bind CD36 analysed across a parasite genome Mol Microbiol 2003 47 1265 1278 12603733 10.1046/j.1365-2958.2003.03378.x
Bian Z Wang G Tian X Fan J Expression of Plasmodium falciparum -infected erythrocyte membrane protein from cerebral malaria patients Zhongguo Ji Sheng Chong Xue Yu Ji Sheng Chong Bing Za Zhi 1999 17 359 362 12563735
Kirchgatter K Portillo HA Association of severe noncerebral Plasmodium falciparum malaria in Brazil with expressed PfEMP1 DBL1 alpha sequences lacking cysteine residues Mol Med 2002 8 16 23 11984002
Jensen AT Magistrado P Sharp S Joergensen L Lavstsen T Chiucchiuini A Salanti A Vestergaard LS Lusingu JP Hermsen R Sauerwein R Christensen J Nielsen MA Hviid L Sutherland C Staalsoe T Theander TG Plasmodium falciparum associated with severe childhood malaria preferentially expresses PfEMP1 encoded by group A var genes J Exp Med 2004 199 1179 1190 15123742 10.1084/jem.20040274
Delemarre BJ Van der Kaay HJ Malaria tropica op natuurlijke wijze verkregen in Nederland Ned Tijdschr Geneeskd 1979 123 1981 1982 390409
Cranmer SL Magowan C Liang J Coppel RL Cooke BM An alternative to serum for cultivation of Plasmodium falciparum in vitro Trans R Soc Trop Med Hyg 1997 91 363 365 9231219 10.1016/S0035-9203(97)90110-3
Staalsoe T Nielsen MA Vestergaard LS Jensen AT Theander TG Hviid L In vitro selection of Plasmodium falciparum 3D7 for expression of variant surface antigens associated with severe malaria in African children Parasite Immunol 2003 25 421 427 14651589 10.1111/j.1365-3024.2003.00652.x
Hermsen CC Telgt DS Linders EH van de Locht LA Eling WM Mensink EJ Sauerwein RW Detection of Plasmodium falciparum malaria parasites in vivo by real-time quantitative PCR Mol Biochem Parasitol 2001 118 247 251 11738714 10.1016/S0166-6851(01)00379-6
Salanti A Staalsoe T Lavstsen T Jensen A Sowa M Arnot D Hviid L Theander TG Selective upregulation of a single distinctly structured var gene in chondroitin sulphate A-adhering Plasmodium falciparum involved in pregnancy-associated malaria Molecular Microbiology 2003 49 179 1911 12823820 10.1046/j.1365-2958.2003.03570.x
Staalsoe T Giha HA Dodoo D Theander TG Hviid L Detection of antibodies to variant antigens on Plasmodium falciparum -infected erythrocytes by flow cytometry Cytometry 1999 35 329 336 10213198 10.1002/(SICI)1097-0320(19990401)35:4<329::AID-CYTO5>3.0.CO;2-Y
Hermsen CC de Vlas SJ Van Gemert GJ Telgt DS Verhage DF Sauerwein RW Testing vaccines in human experimental malaria: statistical analysis of parasitaemia measured by a quantitative real-time polymerase chain reaction Am J Trop Med Hyg 2004 71 196 201 15306710
Taylor HM Kyes SA Harris D Kriek N Newbold CI A study of var gene transcription in vitro using universal var gene primers Mol Biochem Parasitol 2000 105 13 23 10613695 10.1016/S0166-6851(99)00159-0
Noviyanti R Brown GV Wickham ME Duffy MF Cowman AF Reeder JC Multiple var gene transcripts are expressed in Plasmodium falciparum infected erythrocytes selected for adhesion Mol Biochem Parasitol 2001 114 227 237 11378202 10.1016/S0166-6851(01)00266-3
Duffy MF Brown GV Basuki W Krejany EO Noviyanti R Cowman AF Reeder JC Transcription of multiple var genes by individual, trophozoite-stage Plasmodium falciparum cells expressing a chondroitin sulphate A binding phenotype Mol Microbiol 2002 43 1285 1293 11918813 10.1046/j.1365-2958.2002.02822.x
Fernandez V Chen Q Sundstrom A Scherf A Hagblom P Wahlgren M Mosaic-like transcription of var genes in single Plasmodium falciparum parasites Mol Biochem Parasitol 2002 121 195 203 12034453 10.1016/S0166-6851(02)00038-5
Winter G Chen Q Flick K Kremsner P Fernandez V Wahlgren M The 3D7var5.2 (var COMMON) type var gene family is commonly expressed in non-placental Plasmodium falciparum malaria Mol Biochem Parasitol 2003 127 179 191 12672527 10.1016/S0166-6851(03)00004-5
Afonso NP Wunderlich G Shugiro TM d'Arc Neves CJ Jose MM Scherf A Pereira-da-Silva LH Plasmodium falciparum : analysis of transcribed var gene sequences in natural isolates from the Brazilian Amazon region Exp Parasitol 2002 101 111 120 12427465 10.1016/S0014-4894(02)00107-8
Fried M Duffy PE Two DBLgamma subtypes are commonly expressed by placental isolates of Plasmodium falciparum Mol Biochem Parasitol 2002 122 201 210 12106874 10.1016/S0166-6851(02)00103-2
Kaestli M Cortes A Lagog M Ott M Beck HP Longitudinal assessment of Plasmodium falciparum var gene transcription in naturally infected asymptomatic children in Papua New Guinea J Infect Dis 2004 189 1942 1951 15122533 10.1086/383250
Peters J Fowler E Gatton M Chen N Saul A Cheng Q High diversity and rapid changeover of expressed var genes during the acute phase of Plasmodium falciparum infections in human volunteers Proc Natl Acad Sci U S A 2002 99 10689 10694 12142467 10.1073/pnas.162349899
Roberts DJ Craig AG Berendt AR Pinches R Nash G Marsh K Newbold CI Rapid switching to multiple antigenic and adhesive phenotypes in malaria Nature 1992 357 689 692 1614515 10.1038/357689a0
Horrocks P Pinches R Christodoulou Z Kyes SA Newbold CI Variable var transition rates underlie antigenic variation in malaria Proc Natl Acad Sci U S A 2004 101 11129 11134 15256597 10.1073/pnas.0402347101
| 15857512 | PMC1112614 | CC BY | 2021-01-04 16:37:31 | no | Malar J. 2005 Apr 27; 4:21 | utf-8 | Malar J | 2,005 | 10.1186/1475-2875-4-21 | oa_comm |
==== Front
Respir ResRespiratory Research1465-99211465-993XBioMed Central London 1465-9921-6-301581997810.1186/1465-9921-6-30ResearchQuantification of collagen and proteoglycan deposition in a murine model of airway remodelling Reinhardt Alistair K [email protected] Stephen E [email protected] Geoffrey J [email protected] Robin J [email protected] Centre for Respiratory Research, University College London, Rayne Building, 5 University Street, London WC1E 6JJ, UK2005 8 4 2005 6 1 30 30 27 9 2004 8 4 2005 Copyright © 2005 Reinhardt et al; licensee BioMed Central Ltd.2005Reinhardt 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
Sub-epithelial extracellular matrix deposition is a feature of asthmatic airway remodelling associated with severity of disease, decline in lung function and airway hyperresponsiveness. The composition of, and mechanisms leading to, this increase in subepithelial matrix, and its importance in the pathogenesis of asthma are unclear. This is partly due to limitations of the current models and techniques to assess airway remodelling.
Methods
In this study we used a modified murine model of ovalbumin sensitisation and challenge to reproduce features of airway remodelling, including a sustained increase in sub-epithelial matrix deposition. In addition, we have established techniques to accurately and specifically measure changes in sub-epithelial matrix deposition, using histochemical and immunohistochemical staining in conjunction with digital image analysis, and applied these to the measurement of collagen and proteoglycans.
Results
24 hours after final ovalbumin challenge, changes similar to those associated with acute asthma were observed, including inflammatory cell infiltration, epithelial cell shedding and goblet cell hyperplasia. Effects were restricted to the bronchial and peribronchial regions with parenchymal lung of ovalbumin sensitised and challenged mice appearing histologically normal. By 12 days, the acute inflammatory changes had largely resolved and increased sub-epithelial staining for collagen and proteoglycans was observed. Quantitative digital image analysis confirmed the increased deposition of sub-epithelial collagen (33%, p < 0.01) and proteoglycans (32%, p < 0.05), including decorin (66%, p < 0.01). In addition, the increase in sub-epithelial collagen deposition was maintained for at least 28 days (48%, p < 0.001).
Conclusion
This animal model reproduces many of the features of airway remodelling found in asthma and allows accurate and reproducible measurement of sub-epithelial extra-cellular matrix deposition. As far as we are aware, this is the first demonstration of increased sub-epithelial proteoglycan deposition in an animal model of airway remodelling. This model will be useful for measurement of other matrix components, as well as for assessment of the molecular mechanisms contributing to, and agents to modulate airway remodelling.
==== Body
Background
Asthma is a chronic disease with increasing prevalence characterised by persistent bronchial hyperreactivity [1]. There is evidence of epithelial cell damage, inflammatory cell infiltration, together with hypertrophy and hyperplasia of goblet cells, submucosal glands and airway smooth muscle. Additionally, there is increased deposition of extracellular matrix molecules including collagen types I [2], III [3-5] and V [3,5-7], fibronectin [2], tenascin [5,8] and the proteoglycans, lumican, biglycan, versican and decorin [9,10] in the sub-epithelial lamina reticularis. This leads to a 2-3-fold increase in the thickness of the lamina reticularis [11,12] that has been correlated with increased fibroblast/myofibroblast number [11].
The importance of increased airway extracellular matrix deposition in the pathogenesis of asthma is uncertain. However, increasing evidence suggests it may play a significant role. Severity of asthma and decline in FEV1 correlate with sub-epithelial fibrosis [5,13,14]. Airway hyperresponsiveness correlates with bronchial wall thickening [15], sub-epithelial fibrosis [5,7,13], airway wall proteoglycan immunoreactivity [9] and airway fibroblast proteoglycan production [16]. In addition, mathematical modelling suggests that thickening of the sub-epithelial layer will result in an altered folding pattern of the airway with fewer circumferential folds and an increased tendency to airway obstruction [17]. The mechanisms responsible for increased sub-epithelial extracellular matrix deposition in asthma are unknown, although they are thought to involve a complex interaction between cells and mediators in the airway wall [18,19]. Furthermore, the effects of current treatments on this aspect of the pathology are equivocal [6,20-24].
Further studies are therefore required to determine the precise mechanisms involved in the sub-epithelial deposition of extracellular matrix proteins, its importance in relation to airway function and the effects of existing and new therapeutic agents on the process. Mechanistic studies of airway remodelling are difficult in man. Consequently, there is a need for good, reproducible animal models of airway remodelling and techniques to assess these parameters, preferably in mice, which would allow study of genetically-modified strains. Increased sub-epithelial deposition of collagen has been demonstrated in mouse models [25-28], but there is little or no information on other extra-cellular matrix proteins. The sensitivity of mice to airway remodelling is variable [29,30] and protocols to achieve sustained airway remodelling can take several months [28]. In addition, the methods used to assess sub-epithelial matrix deposition are either qualitative [31] or, where quantitative methods have been used, they are often subjective, laborious [25,32], or not specific to measurement of airway changes [33].
In this study we have used C57Bl6/SV129 mice, a strain commonly used in the generation of genetically-modified animals, to develop a model which exhibits many characteristics of asthmatic airways and produces a rapid, reproducible and persistent increase in sub-epithelial extracellular matrix. In conjunction with this we developed accurate and sensitive histological methods using computer-assisted image analysis to specifically quantitate changes in sub-epithelial collagen and proteoglycan deposition.
Methods
Ovalbumin sensitisation and challenge
All animal studies were undertaken with appropriate local ethical and government regulatory approval. SV129/C57BL/6 mice were bred at University College London from breeding pairs obtained from the Jackson Laboratory, Bar Harbor, ME. Two to three month old mice were sensitised by intra-peritoneal injection of 10 μg chicken ovalbumin (grade V, Sigma, Poole, U.K.) in 0.1 ml normal saline or saline alone on two occasions 10 days apart. Twenty-one days after the second sensitisation, mice were challenged with six daily intratracheal instillations of 400 μg ovalbumin in 50 μl normal saline or saline alone. Briefly, mice were anaesthetised using an intra-peritoneal injection of 0.1 ml Saffan™ anaesthetic (9 mg alfaxalone and 3 mg alfadolone acetate/ml, Schering-Plough Animal Health, Welwyn Garden City, U.K.). An intra-venous cannula (22G/25 mm, BOC Ohmeda AB, Helsingborg, Sweden) was introduced through the mouth into the trachea and ovalbumin was instilled using a Hamilton syringe passed through the lumen of the cannula. Sham challenged mice received normal saline. Mice were killed in groups either one or twelve days after the final challenge to assess acute airway changes and airway remodelling respectively. Animals were injected intra-peritoneally with 0.1 ml Euthatal (200 mg pentobarbitone sodium/ml, Rhône Mérieux Ltd., Harlow, U.K.). A longitudinal ventral incision was made in the abdomen and the major vessels were sectioned. Lungs were removed and either wax embedded or frozen as described below.
Lung tissue preparation for histochemical and immunohistochemical staining of paraffin wax sections
Insufflation of the bronchial tree was performed using tracheal instillation of 4% paraformaldehyde at 20 cm 4% paraformaldehyde. The trachea was then ligated and the thoracic contents were removed. Lungs were fixed in 4% paraformaldehyde at 4°C overnight and then transferred to 15% sucrose in PBS at 4°C overnight. They were rinsed in 50% ethanol/H2O and stored in 70% ethanol/H2O prior to further dehydration and processing using an automated vacuum tissue processor (Leica TP1050, Leica Microsystems (U.K.) Ltd., Milton Keynes, U.K.). Lungs were embedded in paraffin wax using an embedding unit (Blockmaster III, Raymond A. Lamb, London, U.K.). Transverse 3 μm sections were cut for histochemical staining and 5 μm sections for immunohistochemistry. After sacrifice, care was taken to perform instillation at a constant pressure of paraformaldehyde. Lungs from each experiment were processed together in precisely the same way and sections were cut and stained at the same time. Automated staining was also pivotal to the uniform production of stained sections for image analysis.
Lung tissue preparation for proteoglycan staining of frozen sections
Insufflation of the bronchial tree was performed using tracheal instillation of a 1 : 3 mixture of OCT embedding matrix (Cell Path plc, Newtown, U.K.) : PBS at a pressure of 25 cm of this mixture. Due to the increased viscosity of this mixture, compared with paraformaldehyde, a higher pressure was required to attain the same degree of inflation. The trachea was ligated and thoracic contents were removed. Lungs were transferred to 50 ml polypropylene tubes containing OCT embedding matrix and the tubes were placed into a container of iso-pentane, chilled using liquid nitrogen. Frozen sections were cut at a thickness of 7 μm, placed on polylysine-coated slides and stored at -20°C before staining.
Histochemical staining
Paraffin wax embedded sections (3 μm) were stained with haematoxylin and eosin to demonstrate tissue architecture and inflammatory cell infiltration, and para-amino salicylic acid/Alcian blue to demonstrate mucous cells. A modified Martius Scarlet Blue (MSB) trichrome stain was employed for the localisation and quantification of collagen using computerised image analysis.
Immunohistochemistry
Immunohistochemistry for type III collagen and decorin was performed on paraffin embedded sections (5 μm). Sections were dewaxed, rehydrated and antigen retrieval achieved by either proteinase K digestion (10 μg/ml) for 10 minutes at room temperature for type III collagen or by microwaving sections in 10 mM citrate buffer, pH6, for 10 min for decorin. Sections were washed in PBS and endogenous peroxidase activity was blocked with 3% hydrogen peroxide (Sigma, Poole, U.K.) for 30 min at RT. Sections were then washed in PBS and incubated with 4% (v/v) goat serum (DakoCytomation, Ely, U.K.) for 20 min at RT to block non-specific binding sites. Excess serum was removed by blotting and the sections were incubated with either a 1 : 200 dilution of rabbit anti-human type III collagen antibody (Chemicon, Chandlers Ford, UK) or a 1 : 80 000 dilution of Decorin antibody, LF-113 [34] (kindly provided by Dr L.W. Fisher, National Institutes of Health, Bethesda, MA. U.S.A.), overnight at 4°C. The sections were washed and incubated with 0.5% (v/v) goat anti-rabbit biotin-labelled secondary antibody (DakoCytomation, Ely, U.K.) for 60 min at RT. Further washes in PBS were performed before the sections were incubated with 0.5% (v/v) streptavidin (DakoCytomation, Ely, U.K.) for 30 min at RT. The streptavidin was removed by three washes in TBS and the sections were treated with DAB for 10 min. They were then washed in tap water, counterstained in Mayer's haematoxylin, differentiated in acid alcohol, dehydrated and mounted.
Proteoglycan Staining
Frozen sections (7 μm) were stained with Cupromeronic Blue™ (Seikagaku Corporation, Tokyo, Japan) using critical electrolyte concentration (CEC) methodology [38]. Cupromeronic Blue is an intensely coloured and electron-dense cationic dye used for the detection of sulphated polyanions including proteoglycans [36]. Sections were rinsed for 5 min in distilled water and placed into the staining solution overnight. This consisted of 0.05% Cupromeronic Blue in 25 mM sodium acetate buffer (pH 5.8) containing 2.5% glutaraldehyde and 250 mM magnesium chloride. Different CEC were trialled before 250 mM of magnesium chloride was selected because it provided optimal staining. This CEC is appropriate for detection of small leucine-rich proteoglycans such as decorin (J.E. Scott, personal communication). The next day slides were washed in 0.025 M sodium acetate buffer (pH 5.8) containing 250 mM magnesium chloride. They were dehydrated to xylene and coverslips were applied.
Measurement of airway sub-epithelial matrix components using image analysis
Sections were examined by light microscopy using a x10 objective. Airways were selected using the following pre-defined criteria. Suitable airways were: complete, of an appropriate size to be contained within a high power field, not attached to other airways and cut in a plane perpendicular to their length (the minimum internal diameter : maximum internal diameter ratio was more than 0.5 in all cases). All suitable airways in each section were analysed. Images were digitised using a digital video camera (JVC KY-F55B, Imaging Associates, Thame, U.K.) with a resolution of 768 × 576 (vertical × horizontal) pixels. Pixel size was converted into micrometers and image analysis was performed using image analysis software (Zeiss KS300 Release 3.0, Imaging Associates, Thame, U.K.). Thresholding using pre-defined RGB criteria for airspace was performed. This allowed lumen airspace to be differentiated from airway wall. Airway lumen perimeter was then measured for each suitable airway. Thresholding using predefined RGB criteria for extra-cellular matrix constituents was performed. This produces a superimposed digitised image of both airway and parenchymal matrix. Next, a 'scrapping' procedure was completed in which pixels not adjoining at least ten other pixels were deleted. This provided better definition of airway matrix as links between airway and parenchymal matrix were removed. In this way, lumen perimeter and matrix values were obtained for each airway. Results were expressed as area of sub-epithelial matrix/unit airway perimeter (μm2/μm).
Statistical analysis
Data are presented as means ± SEM. Statistical evaluations were performed using ANOVA or unpaired t-tests for single group comparisons. A p-value of less than 0.05 was considered significant.
Results
Many of the histological changes associated with airway remodelling in asthma are reproduced in this murine model following ovalbumin sensitisation and intra-tracheal challenge. Characteristic differences were seen one and twelve days after final challenge.
Histological changes seen 24 hours after the final challenge
Airway changes on the day after the final challenge are shown in figure 1. H&E-stained sections from saline sensitised/saline challenged mice demonstrated that the epithelial layer was generally intact and epithelial cells appeared regular in shape (figure 1A). Inflammatory cell infiltration of the airway wall was absent. Very occasional mucus-secreting goblet cells were demonstrated using PAS/Alcian Blue staining (figure 1E). Lung sections from saline sensitised/ovalbumin challenged and ovalbumin sensitised/saline challenged mice were not distinguishable from those of saline sensitised/challenged mice (data not shown). H&E staining of sections from ovalbumin sensitised/challenged mice revealed widespread disruption of the epithelial layer (figure 1B). Epithelial cells were irregular in shape, some were pyknotic and some were shed into the airway lumen. A large number of mucus-secreting goblet cells were demonstrated using PAS/Alcian Blue staining (figure 1F). Airway walls were infiltrated with inflammatory cells that were predominantly eosinophilic but also included neutrophils, lymphocytes and macrophages (figure 1B). Margination and diapedesis of inflammatory cells was seen in adjacent blood vessels (figure 1B). Inflammatory cell infiltration was restricted to the bronchial and peribronchial regions. The lung parenchyma appeared to be unaffected by ovalbumin sensitisation and challenge, and was histologically similar to control sections (figure 1C, D).
Figure 1 Histological appearance of murine lung 24 hours following final ovalbumin challenge. H&E staining of (A) saline sensitised/challenged and (B) ovalbumin sensitised/challenged airways at 24 hours after final challenge. Control airways did not show any inflammatory cell infiltration. A mixed inflammatory cell infiltrate can be seen around ovalbumin sensitised/challenged airways, particularly in those adjacent to blood vessels where margination and diapedesis of cells are seen. H&E staining of typical areas of parenchyma from (C) saline sensitised/challenged and (D) ovalbumin sensitised/challenged mice are similar, demonstrating the lack of interstitial inflammation produced by this model. PAS/Alcian Blue staining of (E) saline sensitised/challenged and (F) ovalbumin sensitised/challenged airways at 24 hours after final challenge. Most of the epithelial cells are replaced by blue/purple staining goblet cells in ovalbumin sensitised/challenged airways. Scale bar represents 50 μm.
Histological changes seen 12 days after the final challenge
Epithelial changes
Inflammatory changes were no longer apparent in ovalbumin sensitised/challenged airways and were not seen in any of the control groups (figure 2). The epithelial layer was generally intact and epithelial cells appeared regular in shape. Very occasional mucus-secreting goblet cells were demonstrated using PAS/Alcian Blue staining (data not shown). All three control groups had similar appearances (figure 2A and 2C).
Figure 2 Increased sub-epithelial collagen deposition in the airways of ovalbumin sensitised and challenged mice. MSB staining of representative airways from each treatment group at 12 days after final challenge. (A) saline sensitisation/saline challenge (B) saline sensitisation/ovalbumin challenge (C) ovalbumin sensitisation/saline challenge (D) ovalbumin sensitisation/ovalbumin challenge. Increased blue-staining sub-epithelial collagen is seen in the ovalbumin sensitised/challenged airway compared with the three controls, which have similar appearances. Scale bar represents 50 μm.
Increased sub-epithelial collagen deposition
In control sections, MSB staining showed a diffuse band of blue-stained collagen surrounding the airways beneath the epithelial layer (figure 2A2B2C). In smaller airways the band of collagen sometimes appeared discontinuous. However, in the airways of ovalbumin sensitised and challenged mice, the staining appeared more intense, the band of staining was thickened compared with controls and usually appeared continuous around all sizes of airway (compare figure 2D with figures 2A2B2C). Quantitative image analysis showed there were no significant differences in the area of sub-epithelial collagen staining between the control groups (figure 3A). Values for control groups were 6.31 ± 0.32 μm2/μm for saline sensitised/challenged mice, 5.63 ± 0.27 μm2/μm for saline sensitised/ovalbumin challenged mice and 5.65 ± 0.26 μm2/μm for ovalbumin sensitised/saline challenged mice. Ovalbumin sensitisation and challenge produced a significant increase in sub-epithelial collagen/unit lumen perimeter compared with control groups (7.77 ± 0.37 μm2/μm, p < 0.01). Compared with the mean of the controls, there was a 33% increase in sub-epithelial collagen in ovalbumin sensitised/challenged animals (p < 0.01).
Figure 3 The effect of different sensitisation/challenge combinations on sub-epithelial collagen deposition. The area of sub-epithelial collagen/unit lumen perimeter was quantitated in mouse airways using image analysis of MSB stained sections. (A) Mean values for all suitable airways in each experimental group were compared. The numbers of mice/total airways analysed per group were: saline sensitised/saline challenged (NS/NS) 8/89, saline sensitised/ovalbumin challenged (NS/Ova) 9/95, ovalbumin sensitised/saline challenged (Ova/NS) 7/88 and ovalbumin sensitised/ovalbumin challenged (Ova/Ova) 11/147. (B) Mean values for all suitable larger (perimeter>1000 μm) airways in the same mice were also compared. The numbers of airways analysed were NS/NS 48, NS/Ova 30, Ova/NS 22 and Ova/Ova 39. * p < 0.01 compared with the mean of the controls.
In these studies the area of sub-epithelial collagen staining was expressed with respect to length of airway lumen perimeter. Therefore differences could reflect changes in lumen perimeter as well as collagen deposition. However, we found no significant difference in lumen perimeter between control groups (991 ± 40 μm) and the ovalbumin sensitised/ovalbumin challenged group (1066 ± 55 μm) indicating that the changes observed were directly attributable to an increase in sub-epithelial collagen deposition.
To determine whether ovalbumin sensitisation and challenge had a differential effect on sub-epithelial collagen depending on airway size, airway data were sub-divided into lumen perimeters of greater or less than 1000 μm (figure 3B). In airways with lumen perimeters greater than 1000 μm, values for control groups were 5.84 ± 0.42 μm2/μm for saline sensitised/challenged mice, 7.00 ± 0.61 μm2/μm for saline sensitised/ovalbumin challenged mice and 6.36 ± 0.48 μm2/μm for ovalbumin sensitised/saline challenged mice. There were no significant differences between these groups. Sub-epithelial collagen/unit lumen perimeter was 9.73 ± 0.73 μm2/μm in ovalbumin sensitised/challenged mice, representing a 53% increase in collagen compared with the mean of the controls (p < 0.01). Ovalbumin sensitisation and challenge did not appear to increase the area of sub-epithelial collagen compared with saline sensitisation and challenge in airways with a lumen perimeter less than 1000 μm. Values for control groups were 6.63 ± 0.55 μm2/μm for saline sensitised/challenged mice (number of airways analysed, n = 43), 5.03 ± 0.25 μm2/μm for saline sensitised/ovalbumin challenged mice (n = 66) and 5.19 ± 0.31 μm2/μm for ovalbumin sensitised/saline challenged mice (n = 50). Sub-epithelial collagen/unit lumen perimeter was 6.58 ± 0.40 μm2/μm in ovalbumin sensitised/challenged mice (n = 88).
In a further experiment sub-epithelial collagen was measured 28 days after final challenge in order to assess whether the increase in collagen persisted over time. Saline sensitised/saline challenged mice were compared with ovalbumin sensitised/ovalbumin challenged mice. Only one control group was used as no significant differences had been shown between control groups in the previous experiment. Sub-epithelial collagen/unit lumen perimeter was 3.69 ± 0.22 μm2/μm in saline sensitised/saline challenged mice and 5.46 ± 0.42 μm2/μm in ovalbumin sensitised/challenged mice, a 48% increase in collagen (p < 0.001). In airways with lumen perimeters greater than 1000 μm, values for saline sensitised/challenged mice were 4.43 ± 0.35 μm2/μm (n = 34) and 6.92 ± 0.79 μm2/μm (n = 21) in ovalbumin sensitised/challenged mice, corresponding to a 56% increase in collagen (p < 0.002) and similar to that observed at 12 days. In contrast to the 12-day data, in airways with lumen perimeters less than 1000 μm, values for saline sensitised/challenged mice were 2.88 ± 0.16 μm2/μm (n = 31) and 4.58 ± 0.43 μm2/μm (n = 35) in ovalbuminsensitised/challenged mice, corresponding to a 59% increase in collagen (p < 0.001). The increase in sub-epithelial collagen is therefore maintained for at least four weeks.
Sections from two groups at 12 days were compared using the same techniques by an independent observer. Sub-epithelial collagen/unit lumen perimeter was 4.16 ± 0.21 μm2/μm in saline sensitised/saline challenged mice and 5.65 ± 0.24 μm2/μm in ovalbumin sensitised/challenged mice. This represents an increase of 36% (p < 0.001) and is of a similar magnitude to that obtained by the first observer.
To confirm that the changes in MSB staining reflected changes in sub-epithelial collagen deposition we immunostained sections for type III collagen and quantified the area of sub-epithelial staining. Type III collagen showed a similar localisation pattern to that of the blue staining of MSB (figure 4). The area of sub-epithelial type III collagen staining in saline sensitised/saline challenged mice was 2.26 ± 0.14 μm2/μm (n = 8) and 2.92 ± 0.16 μm2/μm (n = 11) in ovalbumin sensitised/challenged mice (p < 0.01). This increase of approximately 30% was similar to the increase observed in sections from the same groups stained with MSB.
Figure 4 The effect of ovalbumin sensitisation and challenge on airway sub-epithelial deposition of type III collagen. Immunostaining of representative saline sensitised/challenged (A)and ovalbumin sensitised/challenged (B) airways for type III collagen 12 days after final challenge. In control lungs staining was localised to the alveolar septa, airway and vessel walls. Airway sub-epithelial staining was increased following ovalbumin sensitisation/ challenge. Scale bar represents 50 μm.
Increased sub-epithelial proteoglycan deposition
Cupromeronic Blue staining demonstrated a reticular pattern of proteoglycan deposition beneath the epithelial layer (figure 5A). Staining was usually continuous around the airway circumference and was also seen in the lung parenchyma to a lesser degree. Increased staining was seen in ovalbumin sensitised/ovalbumin challenged mice compared with controls (figure 5A and 5B). Sub-epithelial proteoglycan/unit lumen perimeter was 4.13 ± 0.44 μm2/μm in control mice and 5.46 ± 0.39 μm2/μm in ovalbumin sensitised/challenged mice, corresponding to a 32% increase in proteoglycans (p < 0.05) (figure 6).
Figure 5 The effect of ovalbumin sensitisation and challenge on airway sub-epithelial proteoglycan and decorin deposition. Cupromeronic Blue staining of representative airways from (A) control and (B) ovalbumin sensitised/challenged airways 12 days after final challenge. The reticular pattern of staining is seen particularly around the airway circumference in a distribution corresponding to the sub-epithelial layer. Staining is greater in ovalbumin sensitised/challenged airways. Decorin immunostaining of representative airways from (C) control and (D) ovalbumin sensitised/challenged airways at the same time point. Immunostaining is concentrated in the walls of airways and vessels and is not seen in the lung parenchyma. Immunostaining is greater in ovalbumin sensitised/challenged airways and vessels. Scale bars represent 50 μm.
Figure 6 The effect of ovalbumin sensitisation/challenge on sub-epithelial total proteoglycans and decorin. (A) The area of sub-epithelial proteoglycan/unit lumen perimeter was quantitated using image analysis of Cupromeronic Blue stained sections. The numbers of mice/total airways analysed per group were: Control 9/32 and Ova/Ova 10/52. (B) The area of sub-epithelial decorin/unit lumen perimeter was quantitated using decorin immunostaining. The numbers of mice/total airways analysed per group were: Control 7/40 and Ova/Ova 10/70. * p < 0.05, ** p < 0.01.
Increased sub-epithelial decorin deposition
In control airways staining for decorin was weak with stronger staining associated with blood vessels (figure 5C). In contrast, in the airways of ovalbumin sensitised/challenged mice decorin immunohistochemistry demonstrated a well-localised, intense band of staining beneath the epithelium (figure 5D). Immunostaining was not seen elsewhere in lung sections. Sub-epithelial decorin/unit lumen perimeter was 1.66 ± 0.27 μm2/μm in control mice and 2.76 ± 0.28 μm2/μm in ovalbumin sensitised/challenged mice. This equates with an increase of 66% in decorin in ovalbumin sensitised/challenged mice (p < 0.01) (figure 6).
Discussion
Using a modified ovalbumin sensitisation/challenge protocol in a mouse strain often used for the generation of transgenics and knockouts, many of the features of airway remodelling have been reproduced. At 24 hours after final challenge, epithelial changes similar to those found in acute asthma are seen. There is epithelial cell loss, mucous cell metaplasia and inflammatory cell infiltration of the airway wall. This model, in contrast to several others published in the literature, has the advantage of not producing any obvious parenchymal changes. 12 days after final challenge, the acute inflammatory changes have largely resolved and increased deposition of sub-epithelial collagen and proteoglycans, including decorin, are demonstrated.
The current protocol is based on the widely used ovalbumin sensitisation and challenge method [25,26,37,38] using two intra-peritoneal sensitising injections and six daily intra-tracheal challenges of ovalbumin. The doses of ovalbumin for sensitisation and challenge fall within the ranges used in previous studies but the timings have been optimised to produce reproducible airway remodelling with increased and persistent deposition of extra-cellular matrix proteins including collagens and proteoglycans. Challenges were given by direct intra-tracheal instillation rather than by aerosol as this may more accurately control the dose of ovalbumin received by each mouse. Intra-tracheal instillation rather than aerosol inhalation may also have contributed to the reduction in parenchymal effects seen in this study.
We found similar increases in collagen deposition 12 and 28 days after final challenge when we analysed all airways together. However, when airways were sub-divided into groups with lumen perimeters of greater or less than 1000 μm the difference remained significant for larger airways at both time points but was only significant in smaller airways at the later time. A similar study in rats also found significant changes in smaller airways only at later times [39]. The reason for this is unclear but could be due to remodelling occurring more slowly in smaller airways, although further studies would be required to confirm this. The studies of Palmans and co-workers in rats categorised airway size according to basement membrane length and divided airways into groups with basement membrane length ≤1000 μm, >1000 μm and ≤2000 μm, and >2000 μm. In the present study with mice, airway size was based on the lumen perimeter as this could be measured more accurately with the stains used. Both basement membrane length and lumen perimeter measurements have previously been validated as markers of airway size that are independent of lung volume [40]. In our study very few airways had a perimeter of greater than 2000 μm. Therefore, airways were grouped as ≤1000 μm or >1000 μm.
The number of challenges varies dramatically between different published protocols used to induce airway remodelling. Two opposing theoretical considerations exist. In favour of a greater number of challenges is the, essentially intuitive, belief that remodelling only results from chronic, repetitive injury and repair [26,39]. In favour of fewer challenges is the risk of immunological tolerance developing as more challenges are given. This is not just a hypothetical risk as it has been demonstrated that prolonged exposure of mice to intra-nasal or aerosolised ovalbumin results in peripheral CD4(+) T-cell unresponsiveness, reduced ovalbumin-specific IgE production and suppression of airway inflammation [41,42]. We therefore used the minimum number of challenges that produced a significant and reproducible increase in sub-epithelial matrix.
Allergic airways disease is generally thought of as a Th2 driven response and most previous models of ovalbumin induced airway remodelling in mice have been developed in the BALB/c strain which has a strong Th2 response. The C57BL/6 and SV129 background strains do not generally exhibit strong Th2 responses [43] and may therefore not be ideal in some respects. However, the C57BL/6 strain has previously been used in studies of ovalbumin-induced airway remodelling [44] and the SV129 strain exhibits similar fibroproliferative responses to those of C57BL/6 in other models of fibrogenesis. Whilst we have not assessed the Th1/Th2 balance in the current studies, the fact that significantly increased and sustained deposition of sub-epithelial extracellular matrix proteins was observed, suggests that a strong Th2 response may not be critical to the development of airway remodelling. Further studies would be required to confirm this.
The sensitivity of different mouse strains is a considerable problem when modelling the various pathologic features of asthma including airway remodelling [29,30]. This has lead to the development of a multitude of ovalbumin sensitisation and challenge protocols in order to optimise the modelling of particular features of asthma in different mouse strains. The protocol developed in this study has used mixed strain mice, which are commonly used in the generation of genetically engineered mice and is commercially available. Ideally, genetically engineered mice should be backcrossed to a pure strain. However, due to constraints of time and cost, this is often not achieved. We therefore believe the model described here is readily reproducible and will be useful for studies utilising genetically engineered mice to investigate mechanisms involved in airway remodelling.
An important feature of this study is the use of digital image analysis for the quantification of sub-epithelial extra-cellular matrix deposition. Previous studies in mice have used a variety of histological methods including projection of airway images onto grids and point-counting [25], measurement of airway wall thickness with a microscope eyepiece reticle [32]and measurement of immunostaining intensity from black and white photomicrographs [31]. Other studies have attempted to quantitate airway extra-cellular matrix deposition using different techniques. These include measurement of hydroxyproline in whole lung [30,33] or in micro-dissected airways [45], and measurement of bronchoalveolar lavage fluid (BALF) cellular fibronectin content [37]. Significant disadvantages exist with each of these methods. Measurements derived from whole lung or BALF cannot be airway specific and this problem will be exacerbated in models that induce significant parenchymal changes. Accurate micro-dissection of airways as small as those found in the mouse is extremely difficult, particularly when we are concerned with a layer as thin and variable as the lamina reticularis. Computerised image analysis of stained sections has a number of advantages. It is reproducible between observers and allows accurate determination of extra-cellular matrix in the sub-epithelial layer of each airway. Anatomically, murine airways are frequently associated with other airways and blood vessels. Use of computerised image analysis allows individual airways to be isolated from surrounding structures. This facilitates selection of airway from non-airway matrix. The technique is also highly adaptable, allowing use of different stains to quantify different extra-cellular matrices.
The increase in sub-epithelial collagen in ovalbumin sensitised/challenged mice that we found using six consecutive challenges over six days is similar to that observed in other acute [25] and more chronic (3–12 weeks) [26,28,30,33] ovalbumin challenge studies. For example Blyth and co-workers observed a 55% increase in sub-epithelial reticulin staining after 6 ovalbumin challenges, three days apart, which was maintained for 50 days [25]. Similarly, Temelkovski and co-workers described a 55% increase in the thickness of the sub-epithelial layer after three ovalbumin challenges/week for 4 weeks, with further increases if challenges were continued for up to 8 weeks [26]. Furthermore, these changes are of a similar order of magnitude to those observed in the airways of asthmatics.
To further confirm that the sub-epithelial area of blue staining measured from MSB stained sections truly reflected an increase in collagen deposition we also immunostained sections for type III collagen. The percentage increases in the area of staining obtained for type III collagen and MSB staining following ovalbumin sensitisation/challenge were very similar. However, the values for type III collagen were about one third of those obtained from the MSB sections. This most likely reflects the difference between staining specifically for type III collagen and MSB, which is thought to stain all fibrillar collagens. Types I and III collagens represent greater than 95% of total lung collagen with type III collagen contributing approximately 30% of this total [46]. The proportional area of staining for type III collagen compared with the total blue MSB stained area is therefore consistent with the distribution of collagen types in the lung.
Previous studies examining increased sub-epithelial extra-cellular matrix deposition in murine models of airway remodelling have measured changes in collagen [28,32,45,48] and BAL cellular fibronectin [37,49]. Similarly, airway collagen and fibronectin have been measured in a rat model [39]. A unique feature of this study, therefore, is its application to measurement of increased sub-epithelial proteoglycans, including decorin.
The proteoglycans lumican, biglycan, versican, hyaluronan and decorin have been identified in biopsy studies of normal and asthmatic human airways. Lumican, biglycan, and versican immunostaining is predominantly found in the sub-epithelial layer of control airways [10]. Versican and hyaluronan are also localised around and internal to airway smooth muscle bundles [50]. Data on decorin immunostaining in controls are conflicting. Minimal airway decorin was demonstrated in one study [10], whereas extensive sub-epithelial staining was apparent in another [51]. In this study we show a similar sub-epithelial localisation of proteoglycan staining in mouse airways. We also show that decorin is localised to airway sub-epithelial extra-cellular matrix in mice as described previously for human [51], bovine [52] and rat [53] airways, and is increased following ovalbumin sensitisation and challenge. Since this part of the study used only saline sensitised/saline challenged animals as a control we cannot exclude the possibility that increased proteoglycan deposition is due to non-specific effects of instilling large amounts of ovalbumin into the airways. However, this seems unlikely given the demonstrated lack of effect of ovalbumin on collagen deposition in the saline sensitised/ovalbumin challenged mice (figures 2 and 3) and the evidence for increased sub-epithelial proteoglycan deposition in asthmatic airways.
Lumican, biglycan and versican are increased in the sub-epithelial layer of asthmatic airways compared with controls [10,24]. In our study, Cupromeronic Blue staining of proteoglycans was increased in ovalbumin sensitised/challenged mice with a similar sub-epithelial localisation to that observed in asthmatic airways, suggesting this is a good model of the changes observed in humans [51].
The finding of increased proteoglycans in the sub-epithelial layer of the airways of asthmatics and ovalbumin sensitised/challenged mice is of particular interest. Proteoglycans have an important role in determining organisation of collagen [54]. Decorin, particularly, is an important regulator of collagen fibril assembly [55,56]. Proteoglycans bind growth factors relevant to extra-cellular matrix deposition in the sub-epithelial layer. Perlecan binds specifically to fibroblast growth factor-7 [57]. A number of proteoglycans bind transforming growth factor-β, including decorin, biglycan and fibromodulin [58]. The increased immunostaining for decorin observed in ovalbumin sensitised/challenged mice is consistent with its sub-epithelial co-localisation with increased extra-cellular matrix, and TGF-β which has previously been shown to be increased [28]. The increase in sub-epithelial proteoglycans may be an important factor in the development of sub-epithelial airway remodelling.
Conclusion
This animal model reproduces many of the features of airway remodelling found in asthma and allows accurate and reproducible measurement of sub-epithelial extra-cellular matrix deposition. As far as we are aware this is the first demonstration of increased sub-epithelial proteoglycans in an animal model of airway remodelling. Proteoglycans and collagen are increased by a similar magnitude to that demonstrated in asthmatics, providing a useful model of this aspect of the pathology. The properties of these matrix components suggest that they will have direct and indirect effects on airway function and the regulation of airway remodelling. The model will be invaluable in the assessment of the molecular mechanisms contributing to sub-epithelial airway remodelling and of agents to modulate such remodelling.
Authors' contributions
AKR played a major role in the design of the study, acquisition, analysis and interpretation of data and drafting the manuscript. SEB contributed to the acquisition of data and development of the image analysis methods. GJL contributed to the interpretation of data. RJM 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 Grants from the Wellcome Trust (grant No. 060109) and Asthma UK.
==== Refs
ISAAC. The International Study of Asthma and Allergies in Childhood (ISAAC) Steering Committee Worldwide variation in prevalence of symptoms of asthma, allergic rhinoconjunctivitis, and atopic eczema Lancet 1998 351 1225 1232 9643741 10.1016/S0140-6736(97)07302-9
Roche WR Beasley R Williams JH Holgate ST Subepithelial fibrosis in the bronchi of asthmatics Lancet 1989 1 520 524 2466184 10.1016/S0140-6736(89)90067-6
Wilson JW Li X The measurement of reticular basement membrane and submucosal collagen in the asthmatic airway Clin Exp Allergy 1997 27 363 371 9146928 10.1046/j.1365-2222.1997.600864.x
Chu HW Halliday JL Martin RJ Leung DY Szefler SJ Wenzel SE Collagen deposition in large airways may not differentiate severe asthma from milder forms of the disease Am J Respir Crit Care Med 1998 158 1936 1944 9847289
Hoshino M Nakamura Y Sim J Shimojo J Isogai S Bronchial subepithelial fibrosis and expression of matrix metalloproteinase-9 in asthmatic airway inflammation J Allergy Clin Immunol 1998 102 783 788 9819295
Jeffery PK Godfrey RW Adelroth E Nelson F Rogers A Johansson SA Effects of treatment on airway inflammation and thickening of basement membrane reticular collagen in asthma. A quantitative light and electron microscopic study Am Rev Respir Dis 1992 145 890 899 1554218
Cho SH Seo JY Choi DC Yoon HJ Cho YJ Min KU Lee GK Seo JW Kim YY Pathological changes according to the severity of asthma Clin Exp Allergy 1996 26 1210 1219 8911709
Laitinen A Altraja A Kampe M Linden M Virtanen I Laitinen LA Tenascin is increased in airway basement membrane of asthmatics and decreased by an inhaled steroid Am J Respir Crit Care Med 1997 156 951 958 9310019
Huang J Olivenstein R Taha R Hamid Q Ludwig M Enhanced proteoglycan deposition in the airway wall of atopic asthmatics Am J Respir Crit Care Med 1999 160 725 729 10430752
Roberts CR Is asthma a fibrotic disease? Chest 1995 107 111S 117S 7874987
Brewster CE Howarth PH Djukanovic R Wilson J Holgate ST Roche WR Myofibroblasts and subepithelial fibrosis in bronchial asthma Am J Respir Cell Mol Biol 1990 3 507 511 2223105
Jeffery PK Wardlaw AJ Nelson FC Collin JV Kay AB Bronchial biopsies in asthma: An ultrastructural, quantitative study and correlation with hyperreactivity Am Rev Respir Dis 1989 140 1745 1753 2690708
Chetta A Foresi A Del Donno M Bertorelli G Pesci A Olivieri D Airways remodeling is a distinctive feature of asthma and is related to severity of disease Chest 1997 111 852 857 9106559
Minshall EM Leung DY Martin RJ Song YL Cameron L Ernst P Hamid Q Eosinophil-associated TGF-beta1 mRNA expression and airways fibrosis in bronchial asthma Am J Respir Cell Mol Biol 1997 17 326 333 9308919
Boulet L Belanger M Carrier G Airway responsiveness and bronchial-wall thickness in asthma with or without fixed airflow obstruction Am J Respir Crit Care Med 1995 152 865 871 7663797
Westergren-Thorsson G Chakir J Lafreniere-Allard MJ Boulet LP Tremblay GM Correlation between airway responsiveness and proteoglycan production by bronchial fibroblasts from normal and asthmatic subjects Int J Biochem Cell Biol 2002 34 1256 1267 12127576 10.1016/S1357-2725(02)00058-4
Wiggs BR Hrousis CA Drazen JM Kamm RD On the mechanism of mucosal folding in normal and asthmatic airways J Appl Physiol 1997 83 1814 1821 9390950
Akers IA McAnulty RJ Laurent GJ Page CP, Banner KH Spina D Fibroblasts and myofibroblasts Cellular mechanisms in airways inflammation 2000 Basel: Birkhauser Verlag 159 198
McAnulty RJ Laurent GJ Barnes PB, Drazen JD Rennard SR Thomson NT Fibroblasts Asthma and COPD: Basic mechanisms and clinical management 2002 London: Academic Press 139 144
Olivieri D Chetta A Del Donno M Bertorelli G Casalini A Pesci A Testi R Foresi A Effect of short-term treatment with low-dose inhaled fluticasone propionate on airway inflammation and remodeling in mild asthma: a placebo-controlled study Am J Respir Crit Care Med 1997 155 1864 1871 9196087
Hoshino M Nakamura Y Sim JJ Yamashiro Y Uchida K Hosaka K Isogai S Inhaled corticosteroid reduced lamina reticularis of the basement membrane by modulation of insulin-like growth factor (IGF)-I expression in bronchial asthma Clin Exp Allergy 1998 28 568 577 9645593 10.1046/j.1365-2222.1998.00277.x
Hoshino M Takahashi M Takai Y Sim J Inhaled corticosteroids decrease subepithelial collagen deposition by modulation of the balance between matrix metalloproteinase-9 and tissue inhibitor of metalloproteinase-1 expression in asthma J Allergy Clin Immunol 1999 104 356 363 10452757
Boulet LP Turcotte H Laviolette M Naud F Bernier MC Martel S Chakir J Airway hyperresponsiveness, inflammation, and subepithelial collagen deposition in recently diagnosed versus long-standing mild asthma. Influence of inhaled corticosteroids Am J Respir Crit Care Med 2000 162 1308 1313 11029336
Flood-Page P Menzies-Gow A Phipps S Ying S Wangoo A Ludwig MS Barnes N Robinson D Kay AB Anti-IL-5 treatment reduces deposition of ECM proteins in the bronchial subepithelial basement membrane of mild atopic asthmatics J Clin Invest 2003 112 1029 1036 14523040 10.1172/JCI200317974
Blyth DI Pedrick MS Savage TJ Hessel EM Fattah D Lung inflammation and epithelial changes in a murine model of atopic asthma Am J Respir Cell Mol Biol 1996 14 425 438 8624247
Temelkovski J Hogan SP Shepherd DP Foster PS Kumar RK An improved murine model of asthma: selective airway inflammation, epithelial lesions and increased methacholine responsiveness following chronic exposure to aerosolised allergen Thorax 1998 53 849 856 10193371
Kumar RK Foster PS Murine model of chronic human asthma Immunol Cell Biol 2001 79 141 144 11264707 10.1046/j.1440-1711.2001.00981.x
Cho JY Miller M Baek KJ Han JW Nayar J Lee SY McElwain K McElwain S Friedman S Broide DH Inhibition of airway remodeling in IL-5-deficient mice J Clin Invest 2004 113 551 560 14966564 10.1172/JCI200419133
Kumar RK Foster PS Modeling allergic asthma in mice: pitfalls and opportunities Am J Respir Cell Mol Biol 2002 27 267 272 12204888
Shinagawa K Kojima M Mouse model of airway remodeling: strain differences Am J Respir Crit Care Med 2003 168 959 967 12857720 10.1164/rccm.200210-1188OC
Kumagai K Ohno I Imai K Nawata J Hayashi K Okada S Senoo H Hattori T Shirato K The involvement of matrix metalloproteinases in basement membrane injury in a murine model of acute allergic airway inflammation Clin Exp Allergy 2002 32 1527 1534 12372136 10.1046/j.1365-2745.2002.01491.x
Tang W Geba GP Zheng T Ray P Homer RJ Kuhn C IIIFlavell RA Elias JA Targeted expression of IL-11 in the murine airway causes lymphocytic inflammation, bronchial remodeling, and airways obstruction J Clin Invest 1996 98 2845 2853 8981933
Tanaka H Masuda T Tokuoka S Komai M Nagao K Takahashi Y Nagai H The effect of allergen-induced airway inflammation on airway remodeling in a murine model of allergic asthma Inflamm Res 2001 50 616 624 11822788
Schaefer L Hausser H Altenburger M Ugorcakova J August C Fisher LW Schaefer RM Kresse H Decorin, biglycan and their endocytosis receptor in rat renal cortex Kidney Int 1998 54 1529 1541 9844129 10.1046/j.1523-1755.1998.00149.x
Scott JE Histochemistry of Alcian blue. 3. The molecular biological basis of staining by Alcian blue 8GX and analogous phthalocyanins Histochemie 1972 32 191 212 4117063 10.1007/BF00306028
Scott JE Proteoglycan histochemistry-a valuable tool for connective tissue biochemists Coll Relat Res 1985 5 541 575 2421971
Trifilieff A Fujitani Y Coyle AJ Kopf M Bertrand C IL-5 deficiency abolishes aspects of airway remodelling in a murine model of lung inflammation Clin Exp Allergy 2001 31 934 942 11422160 10.1046/j.1365-2222.2001.01084.x
Henderson WR JrTang LO Chu SJ Tsao SM Chiang GK Jones F Jonas M Pae C Wang H Chi EY A role for cysteinyl leukotrienes in airway remodeling in a mouse asthma model Am J Respir Crit Care Med 2002 165 108 116 11779739
Palmans E Kips JC Pauwels RA Prolonged allergen exposure induces structural airway changes in sensitized rats Am J Respir Crit Care Med 2000 161 627 635 10673209
James AL Pare PD Hogg JC Effects of lung volume, bronchoconstriction, and cigarette smoke on morphometric airway dimensions J Appl Physiol 1988 64 913 919 3366748 10.1063/1.341893
Tsitoura DC Blumenthal RL Berry G Dekruyff RH Umetsu DT Mechanisms preventing allergen-induced airways hyperreactivity: role of tolerance and immune deviation J Allergy Clin Immunol 2000 106 239 246 10932065 10.1067/mai.2000.108429
Sakai K Yokoyama A Kohno N Hamada H Hiwada K Prolonged antigen exposure ameliorates airway inflammation but not remodeling in a mouse model of bronchial asthma Int Arch Allergy Immunol 2001 126 126 134 11729350 10.1159/000049503
Charles PC Weber KS Cipriani B Brosnan CF Cytokine, chemokine and chemokine receptor mRNA expression in different strains of normal mice: implications for establishment of a Th1/Th2 bias J Neuroimmunol 1999 100 64 73 10695717 10.1016/S0165-5728(99)00189-7
Tsuchiya T Nishimura Y Nishiuma T Kotani Y Funada Y Yoshimura S Yokoyama M Airway remodelling of murine chronic antigen exposure model J Asthma 2003 40 935 944 14736094 10.1081/JAS-120024594
Kenyon NJ Ward RW McGrew G Last JA TGF-beta1 causes airway fibrosis and increased collagen I and III mRNA in mice Thorax 2003 58 772 777 12947136 10.1136/thorax.58.9.772
Laurent GJ McAnulty RJ Corrin B Cockerill P Biochemical and histological changes in pulmonary fibrosis induced in rabbits with intratracheal bleomycin Eur J Clin Invest 1981 11 441 448 6174345
Kumar RK Herbert C Yang M Koskinen AM Mckenzie AN Foster PS Role of interleukin-13 in eosinophil accumulation and airway remodelling in a mouse model of chronic asthma Clin Exp Allergy 2002 32 1104 1111 12100061 10.1046/j.1365-2222.2002.01420.x
Masuda T Tanaka H Komai M Nagao K Ishizaki M Kajiwara D Nagai H Mast cells play a partial role in allergen-induced subepithelial fibrosis in a murine model of allergic asthma Clin Exp Allergy 2003 33 705 713 12752602 10.1046/j.1365-2222.2003.01588.x
Trifilieff A El Hashim A Bertrand C Time course of inflammatory and remodeling events in a murine model of asthma: effect of steroid treatment Am J Physiol Lung Cell Mol Physiol 2000 279 L1120 L1128 11076802
Pare PD Roberts CR Bai TR Wiggs BJ The functional consequences of airway remodeling in asthma Monaldi Arch Chest Dis 1997 52 589 596 9550873
Redington AE Roche WR Holgate ST Howarth PH Co-localization of immunoreactive transforming growth factor-beta 1 and decorin in bronchial biopsies from asthmatic and normal subjects J Pathol 1998 186 410 415 10209491 10.1002/(SICI)1096-9896(199812)186:4<410::AID-PATH198>3.0.CO;2-9
Brahimi-Horn MC Deudon E Paul A Mergey M Mailleau C Basbaum C Dohrman A Capeau J Identification of decorin proteoglycan in bovine tracheal serous cells in culture and localization of decorin mRNA in situ Eur J Cell Biol 1994 64 271 280 7813515
Wang Y Sakamoto K Khosla J Sannes PL Detection of chondroitin sulfates and decorin in developing fetal and neonatal rat lung Am J Physiol Lung Cell Mol Physiol 2002 282 L484 L490 11839542
Scott JE Structure and function in extracellular matrices depend on interactions between anionic glycosaminoglycans Pathol Biol (Paris) 2001 49 284 289 11428163
Vogel KG Paulsson M Heinegard D Specific inhibition of type I and type II collagen fibrillogenesis by the small proteoglycan of tendon Biochem J 1984 223 587 597 6439184
Danielson KG Baribault H Holmes DF Graham H Kadler KE Iozzo RV Targeted disruption of decorin leads to abnormal collagen fibril morphology and skin fragility J Cell Biol 1997 136 729 743 9024701 10.1083/jcb.136.3.729
Mongiat M Taylor K Otto J Aho S Uitto J Whitelock JM Iozzo RV The protein core of the proteoglycan perlecan binds specifically to fibroblast growth factor-7 J Biol Chem 2000 275 7095 7100 10702276 10.1074/jbc.275.10.7095
Hildebrand A Romaris M Rasmussen LM Heinegard D Twardzik DR Border WA Ruoslahti E Interaction of the small interstitial proteoglycans biglycan, decorin and fibromodulin with transforming growth factor beta Biochem J 1994 302 527 534 8093006
| 15819978 | PMC1112615 | CC BY | 2021-01-04 16:36:26 | no | Respir Res. 2005 Apr 8; 6(1):30 | utf-8 | Respir Res | 2,005 | 10.1186/1465-9921-6-30 | oa_comm |
==== Front
Respir ResRespiratory Research1465-99211465-993XBioMed Central London 1465-9921-6-391585048610.1186/1465-9921-6-39ResearchValidation of a brief telephone battery for neurocognitive assessment of patients with pulmonary arterial hypertension Taichman Darren B [email protected] Jason [email protected] Rosette [email protected] Jennifer [email protected] Joanne [email protected] Sandra [email protected] John [email protected] Harold I [email protected] C Gregory [email protected] Ramona O [email protected] Pulmonary, Allergy and Critical Care Division, University of Pennsylvania School of Medicine, Philadelphia, PA, USA2 Physical Medicine and Rehabilitation, University of Pennsylvania School of Medicine, Philadelphia, PA, USA3 Department of Medicine, Pulmonary and Critical Care Divisions, University of Utah and LDS Hospital, Salt Lake City, Utah, USA4 Psychology Department and Neuroscience Center, Brigham Young University, Provo, Utah, USA2005 25 4 2005 6 1 39 39 26 1 2005 25 4 2005 Copyright © 2005 Taichman et al; licensee BioMed Central Ltd.2005Taichman 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 effects of pulmonary arterial hypertension on brain function are not understood, despite patients' frequent complaints of cognitive difficulties. Using clinical instruments normally administered during standard in-person assessment of neurocognitive function in adults, we assembled a battery of tests designed for administration over the telephone. The purpose was to improve patient participation, facilitate repeated test administration, and reduce the cost of research on the neuropsychological consequences of acute and chronic cardiorespiratory diseases. We undertook this study to validate telephone administration of the tests.
Methods
23 adults with pulmonary arterial hypertension underwent neurocognitive assessment using both standard in-person and telephone test administration, and the results of the two methods compared using interclass correlations.
Results
For most of the tests in the battery, scores from the telephone assessment correlated strongly with those obtained by in-person administration of the same tests. Interclass correlations between 0.5 and 0.8 were observed for tests that assessed attention, memory, concentration/working memory, reasoning, and language/crystallized intelligence (p ≤ 0.05 for each). Interclass correlations for the Hayling Sentence Completion test of executive function approached significance (p = 0.09). All telephone tests were completed within one hour.
Conclusion
Administration of this neurocognitive test battery by telephone should facilitate assessment of neuropsychological deficits among patients with pulmonary arterial hypertension living across broad geographical areas, and may be useful for monitoring changes in neurocognitive function in response to PAH-specific therapy or disease progression.
==== Body
Introduction
Pulmonary arterial hypertension (PAH) is a devastating disease characterized by progressive shortness of breath and the eventual development of life-threatening heart failure [1-4]. While its effects on cardiovascular function have been well documented, little is known about effects of this disease on other organ systems, notably the brain. Patients frequently complain of changes in memory, concentration and judgment in association with the development of cardiopulmonary symptoms[5]. Objective assessments of cognitive function, however, have not been performed.
The "gold standard" for comprehensive assessment of neurocognitive function is a comprehensive battery of individually validated tests that are administered in-person by an experienced interviewer. Comprehensive, standard testing often takes four hours or longer and may require two or more separately scheduled sessions. The experience can be stressful and fatiguing, particularly for chronically ill or physically disabled patients. The stress of additional travel to a testing site adds to the burden. Consequently, many patients decline to participate in clinical research that uses formal in-person neurocognitive testing, especially if the protocol includes repeated testing over time [6,7].
To address these concerns, we developed a focused battery of neurocognitive tests for administration over the telephone. The battery is comprised of individually validated neurocognitive test instruments appropriate for administration to adults with cardiopulmonary disease who are fluent in English and capable of communicating verbally. Telephone administration of the test battery has been evaluated for feasibility and validity in survivors of the acute respiratory distress syndrome [8]. The present study establishes the validity of telephone administration of the test battery by comparing the results of telephone testing against "gold standard" in-person administration in adult ambulatory patients with moderately to severely symptomatic PAH.
Methods
Study Population
Consecutive patients diagnosed with pulmonary arterial hypertension according to standard criteria [2,9] were prospectively recruited for neurocognitive testing from the Pulmonary Hypertension Clinic at LDS Hospital. Written informed consent was obtained from the patients for both the in-person and telephone neuropsychological testing. Inclusion criteria were age 18 years or older and the ability to give informed consent. Exclusion criteria included non-fluency in English, a history of major psychiatric illness (e.g. schizophrenia, schizoaffective disorder, bipolar disorder, psychoses requiring medication and/or hospitalization, and major depression requiring hospitalization), known learning disability, prior traumatic brain injury, diagnosis of dementia, cerebral vascular accident, neurologic disorder (e.g. multiple sclerosis, Huntington's Disease, etc.), prior cardiac surgery, or current alcohol or drug abuse.
Patients were recruited for neurocognitive assessment from a group who had consented to in-person testing. Sixty-seven patients were screened. Seventeen patients declined, two were excluded due to non-fluency in English, and two were medically unstable at the time of evaluation and died during the recruitment period. Of the 46 patients who consented to in-person neurocognitive testing, 25 consented to additional testing by telephone.
Patient demographic, medical and laboratory data were collected for all enrolled patients. This study was approved by the University of Pennsylvania and LDS Hospital Institutional Review Boards and conformed to institutional and federal guidelines for the protection of human subjects.
Neurocognitive Assessment
An interdisciplinary team of neuropsychology, traumatic brain injury, rehabilitation medicine, and pulmonary disease specialists selected a battery of standardized neurocognitive tests amenable to both in-person and telephone administration. Tests were also chosen on the basis of established sensitivity in detecting impairment in patients with cardiopulmonary disorders and concomitant hypoxemia [10-13]. The cognitive domains assessed and the tests included in the battery are listed in Table 1. All neurocognitive tests included in the battery have been empirically validated and standardized [14-17] with established reliability, internal and external validity[12,14,16,18,19]. The neurocognitive tests were administered in a random sequence to minimize order effects. However, as a delay is required between the Wechsler Memory Scale-III Logical Memory I and II tests, Logical memory I (immediate recall) was the first and Logical Memory II (delay recall) the last test administered in each session.
Table 1 Neurocognitive Battery for Telephone Administration
COGNITIVE DOMAIN TEST INSTRUMENT
Attention WMS-III: Digits Forward
Concentration/Working Memory WMS-III: Digits Backward
WMS-III: Letter-Number Sequencing
Executive Function Hayling Sentence Completion Test
Reasoning WAIS-III: Similarities
Language / Crystallized Intelligence WAIS-III Vocabulary
Memory WMS-III: Logical Memory I & II
The in-person assessment was carried out in a private office at LDS Hospital. The identical tests were administered subsequently by telephone at a prearranged time when patients were at home and free from distraction. To minimize potential learning effects, telephone testing was performed at least 2.5 months following in-person assessment, except with one patient who was tested 57 days after in-person evaluation. A Ph.D. neuropsychologist (ROH) administered the in-person tests and a neuropsychology doctoral student administered the telephone tests with no knowledge of the results of the previous in-person testing. During both the in-person and telephone tests, patients were instructed not to write down information and to answer questions without assistance. The in-person and telephone assessments were both conducted in single sessions, and each took 35 to 45 minutes to complete.
All neuropsychological tests were scored according to the published guidelines. Each test yields a raw score that was converted into a scaled score (mean = 10; SD = 3), which was used for statistical analyses, except for Logical Memory where the raw scores are used.
Statistical Analysis
Descriptive statistics were carried out for demographic and medical data. The neuropsychological test scores from the in-person administration were compared to telephone test scores using interclass correlations. To facilitate interpretation of significant correlations (p ≤ 0.05) and because traditional significance levels for correlations coefficients are influenced by factors such as group size, range of scores, and multiple comparisons, we used the following conservative classification: fair correlation with coefficients between 0.21 and 0.40; moderate correlation 0.41 to 0.60; substantial correlation 0.61 to 0.80; and almost perfect correlation 0.81 to 1.00 [20].
To assess potential learning effects, systematic differences between first and second administrations for each of the tests were assessed using paired sample t-tests. The differences between the in-person and telephone test results are expressed as standardized effects sizes (T2-T1 differences divided by T1 standard deviation) [21].
Results
Twenty-five patients with pulmonary arterial hypertension were enrolled for neurocognitive evaluation using both in-person and telephone testing. All 25 patients completed in-person testing. Telephone testing could not be completed on one subject due to a non-functioning telephone line, and one subject died of progressive right heart failure. All of the remaining 23 subjects completed both the in-person and telephone assessments and were included in the validation group. Eighty-three percent (n = 19) of these subjects were women. The mean ± SD age was 49.7 ± 13.9 years (range 20 to 60 years) and the mean education level was 13.6 ± 3.0 years (range 6 to 20 years). The mean number of days between in-person and telephone testing was 121.6 (range 57 to 200 days). The etiology of PAH was: idiopathic ("primary") PAH in ten patients (43%), associated with anorexigen use in six (26%), collagen vascular disease in four (17%), congenital heart disease in two (9%) and one with portopulmonary hypertension (4%). The mean (± SD) right atrial pressure was 5.1 ± 1.8 mmHg, mean pulmonary artery pressure 52.1 ± 16.9 and pulmonary capillary wedge pressure 12.1 ± 6.2. The mean cardiac output was 5.1 ± 1.8 L/min. Demographic and medical data are shown in Table 2.
Table 2 Demographic and Medical Data
Mean ± SD Range
Gender (% female) (n = 25) 83%
Education (years) 13.6 ± 3.0 6 to 20
Age (years) 49.7 ± 13.9 20 to 69
Time Since Diagnosis (years) 1.8 ± 1.5 0.8 to 5.3
PaO2 mmHg 62.6 ± 13.5 38 to 97
Most recent 6 minute walk (meters) 455 ± 132 227 to 877
New York Heart Functional Class (N)
Class 1 0
Class 2 3
Class 3 20
Class 4 0
Supplemental Oxygen (N)
2 Liters per minute 7
3 Liters per minute 8
4 Liters per minute 4
The results of the telephone and in-person neurocognitive assessments are shown in Table 3. The correlation coefficients for the comparison between in-person and telephone testing are presented in Table 4. Interclass correlation coefficients of at least 0.54 (p < 0.05 to < 0.0001) were found for the agreement of telephone and in-person scores on tests assessing the cognitive domains of attention, memory, concentration / working memory, reasoning, and language / crystallized intelligence. An almost perfect correlation was observed in the assessment of reasoning (Similarities). Substantial correlations were found for the Digit Span, Similarities, and Vocabulary tests (.61 to .80) and moderate correlations (.41 to .60) were found for each Logical Memory immediate and delay recall. A moderate correlation (0.56) was seen with the in-person and telephone administration of the Digit-Span-backward (concentration / working memory). Only a fair correlation (0.28; p= 0.09) was found between the in-person and telephone administration of the Hayling Sentence Completion test of executive function. For Letter-Number Sequence test (concentration/working memory) scores were not correlated for the in-person and telephone tests.
Table 3 In-person and telephone neuropsychological test scores.
Test Mean Median SD Range
Number-letter Sequencing
In-person 9.3 10 2.6 5 to 14
Telephone 9.3 9 2.5 5 to 15
Logical Memory
Immediate Recall†
In-person 24.4 23 5.9 14 to 33
Telephone 27.0 26 8.2 17 to 44
Delay Recall†
In-person 18.8 18 5.4 10 to 30
Telephone 22.0 23 7.5 6 to 34
Digit Span
In-person 11.6 11 3.2 6 to 18
Telephone 9.3 9 2.4 5 to 15
Hayling Sentence Completion Test
In-person 5.7 6 1.1 4 to 8
Telephone 6.6 6 0.89 6 to 10
Similarities
In-person 11.0 12 2.9 4 to 16
Telephone 10.9 10 3.1 5 to 16
Vocabulary
In-person 10.6 11 2.6 6 to 16
Telephone 11.5 11 2.6 7 to 17
All values are scaled scores (mean = 10, standard deviation = 3) except † = raw scores.
Logical memory and Number-letter sequencing are from the WMS-III; digit span, similarities and vocabulary are from the WAIS-R.
Table 4 Reliability of the in-person and telephone neuropsychological test scores.
Neuropsychological Test Learning Effect (Improvement T1 to T2 expressed in SDs) Intraclass Correlation 95% C.I. p
Number-letter Sequencing .78 (0 to 3.1) .15 -.28 to .52 0.25
Logical Memory
Immediate Recall† 1.3 (0 to 5.0) .55 -.20 to .72 0.05
Delay Recall† 1.2 (.18 to 3.4) .54 -.24 to .82 0.05
Digit Span
Forward† .74 .41 to .90 0.0002
Backward† .56 -.24 to .84 0.05
Both .83 (0 to 1.9) .63 .29 to .83 0.0006
Hayling Sentence
Completion Test .98 (0 to 3.8) .28 -.14 to .61 0.09
Similarities .65 (0 to 2.0) .82 .62 to .92 <0.0001
Vocabulary .53 (0 to 1.5) .68 .38 to .85 0.001
All scores shown are scaled scores (mean = 10, standard deviation = 3), except † = raw scores. Logical memory and Number-letter sequencing are from the WMS-III; digit span, similarities and vocabulary are from the WAIS-R.
Stability over time was greatest for the Similarities and Vocabulary tests. The effects of learning showed that test scores tended to increase between the in-person and telephone tests, with the most improvement for verbal memory (e.g. logical memory immediate and delayed recall).
Discussion
We found the battery of well-established neurocognitive tests to be amenable to administration by telephone and valid for the identification of neurocognitive deficits in patients with PAH. Testing was readily completed in a single, 30–60 minute session, and required neither specialized testing facilities nor travel by physically debilitated patients spread across a broad geographic area.
Our study was designed to validate the administration by telephone of a battery of neurocognitive tests against the in-person ("gold standard") performance of these same assessments. Each of the tests in the battery has been previously validated for the identification of neurocognitive deficits in various populations, including those with cardiopulmonary disease-. As such, our subjects' scores during in-person testing served as matched controls for comparison with the results obtained upon application of these same tests over the telephone.
The scores from telephone and in-person assessments correlated strongly for the majority of tests. Overall, the strengths of the correlations with in-person testing found here are comparable to those that we reported previously for the same test battery applied to ARDS survivors [8] and the correlations reported for other telephone neurocognitive test batteries [21-25]. Two items in the test battery did not correlate as well as the others. The interclass correlations for the Hayling Sentence Completion test only approached significance (p = 0.09). The in-person and telephone administrations of the Letter-Number-Sequencing component of the WMS III (a test of concentration/working memory) did not correlate well. Some subjects appeared to have difficulty discriminating phonetically similar sounds (e.g. the letters 'm' and 'n') when presented during telephone sessions; visual cues may have alleviated such issues at in-person sessions. In contrast, another component of the WMS III (Digits Backward) evaluating the same cognitive domain (concentration/working memory) had substantial correlations.
Although the correlations we found between in-person testing and subsequent telephone administration of the same test battery were moderate or higher, they were not perfect. The effects of learning or practice suggest that test scores increase between the in-person and telephone tests, with the most improvement in verbal memory. Thus, the improvement in test scores on the telephone administration of the test likely reflects practice effects. An alternative explanation for the tendency of subjects to perform better on the telephone test battery may be environmental factors. For example subjects scored higher on certain tasks when assessed at home as compared to similar tasks performed in a clinic setting [26]. Improved orientation to time and place have been found when patients were tested in their own residence [22]. Further, patients report less anxiety and prefer telephone testing compared to in-person evaluation [27]. In addition to the pragmatic advantages, telephone testing may provide a better assessment of patients' cognitive function in their normal environment. Finally, it is possible that neurocognitive function improved during the interval between the two test sessions (mean 122 days). The reason for a potential improvement in neurocognitive performance will be important in future studies that use repeated test administration to determine the effect of drugs for PAH or other interventions on neurocognitive function.
An important limitation of our study was our inability to reverse the order of administration (in-person and telephone) [28]. Our subjects were enrolled in another ongoing study, which required in-person assessments prior to enrollment in this study of telephone testing [29]. An alternative would be to repeat in-person and telephone assessments in random order following the initial interview. Such additional testing, however, might have increased the potential for learning or practice affects, and the further time and travel commitments for patients likely impacting study participation. Future studies should counterbalance the order of in-person and telephone administration. Due to pragmatic limitations the time interval between in-person and telephone testing was somewhat longer than the minimum time necessary to minimize recall and learning effects.
Telephone testing has been used successfully in the assessment of neruocognitive impairment in other patient populations. Further, the ease in application and relative low cost of telephone testing have enabled assessments in several large clinical studies of cognitive function: screening of 4,932 elderly patients for Alzheimer's Disease using the modified Mini-Mental State Examination for telephone administration [30]; cognitive function in 4,023 patients with cardiovascular risk factors [31]; 466 patients with coronary artery bypass graft surgery [32], and in a self-referred ARDS patient group [8]. In addition to neurocognitive testing, telephone-based assessments have provided accurate determinations of quality of life, medication usage, 24-hour physical activity and dietary recall [33-35].
Cognitive function has not been studied in pulmonary arterial hypertension, despite the frequent reports of problems with memory and concentration [5]. Cognitive impairments are important complications of other chronic and life-threatening illnesses, and are associated with a significantly worse prognosis [36-38]. Cognitive impairments can also profoundly reduce quality of life [39-42]. Therapies that improve physical function in PAH may have important (but as yet unknown) effects on neurocognitive function, positive or negative. For these reasons, research on neurocognitive function is warranted. The brief neurocognitive telephone test battery is valid for assessment of cognitive function in this population and provides the means to pursue further, larger studies to assess the frequency and risk factors of cognitive sequelae in patients with PAH.
Conclusion
This study has demonstrated that scores on a battery of neurocognitive tests obtained by telephone administration correlated well with in-person testing in patients with pulmonary arterial hypertension. The strong correlations observed are comparable to previous studies that assessed in-person and telephone versions of neurocognitive tests. With minor modification, the telephone neuropsychological test battery described here provides an economical and reliable method for assessing cognitive function in patients with pulmonary arterial hypertension.
List of Abbreviations Used
PAH: Pulmonary Arterial Hypertension
SD: Standard Deviation
ARDS: Acute Respiratory Distress Syndrome
WMD: Wechsler Memory Scale
WAIS: Wechsler Adult Intelligence Scale
Competing Interests
The author(s) declare that they have no competing interests.
Authors' Contributions
DBT, JHF, HIP and ROH designed and SK coordinated this study. RP and JC designed the neurocognitive testing battery. JM, JW and ROH performed and interpreted the neuropsychological assessments and ROH the statistical analysis. CGE recruited patients and assessed results. The manuscript was written by DBT, JHF and ROH and has been read and approved by all authors.
Acknowledgements
These studies were supported by a Development Partner's Junior Faculty Award from GSK Pharmaceuticals (DBT). ROH is a recipient of grant support from the National Institute of Mental Health (R01 MH065406-01A1).
==== Refs
Rubin LJ Primary pulmonary hypertension N Engl J Med 1997 336 2 111 7 8988890 10.1056/NEJM199701093360207
Barst RJ Diagnosis and differential assessment of pulmonary arterial hypertension J Am Coll Cardiol 2004 43 12 Suppl S 40S 47S 15194177 10.1016/j.jacc.2004.02.032
Rich S Primary pulmonary hypertension. A national prospective study Ann Intern Med 1987 107 2 216 23 3605900
D'Alonzo GE Survival in patients with primary pulmonary hypertension. Results from a national prospective registry Ann Intern Med 1991 115 5 343 9 1863023
Hopkins RO Cognitive dysfunction in patients with pulmonary arterial hypertension Am J Respir Cell Mol Biol 2003 167 7 A273
Rothenhausler HB The relationship between cognitive performance and employment and health status in long-term survivors of the acute respiratory distress syndrome: results of an exploratory study Gen Hosp Psychiatry 2001 23 2 90 6 11313077 10.1016/S0163-8343(01)00123-2
Jackson JC Six-month neuropsychological outcome of medical intensive care unit patients Crit Care Med 2003 31 4 1226 34 12682497 10.1097/01.CCM.0000059996.30263.94
Christie J Long-term cognitive, mood, and quality of life impairments in a select population of ARDS survivors from an internet-based ARDS support center Am J Respir Cell Mol Biol 2002 165 8 A220
McGoon M Screening, early detection, and diagnosis of pulmonary arterial hypertension: ACCP evidence-based clinical practice guidelines Chest 2004 126 1 Suppl 14S 34S 15249493 10.1378/chest.126.1_suppl.14S
Gale SD Hopkins RO Effects of hypoxia on the brain: neuroimaging and neuropsychological findings following carbon monoxide poisoning and obstructive sleep apnea J Int Neuropsychol Soc 2004 10 1 60 71 14751008 10.1017/S1355617704101082
Gale SD Hopkins RO Weaver LK Walker JM Bigler ED Cloward TV Hippocampal atrophy following sleep apnea and carbon monoxide poisoning: simialrities and differences J Int Neuropsychol Soc 2000 6 154
Hopkins RO Neuropsychological sequelae and impaired health status in survivors of severe acute respiratory distress syndrome Am J Respir Crit Care Med 1999 160 1 50 6 10390379
Weaver LK Hyperbaric oxygen for acute carbon monoxide poisoning N Engl J Med 2002 347 14 1057 67 12362006 10.1056/NEJMoa013121
Wechsler D Wechsler Memory Scale 1997 3 San Antonio: The Psychological Corporation
Wechsler D Wechsler Adult Intelligence Scale 1997 San Antonio: The Psychological Corporation
Kiernan RJ The Neurobehavioral Cognitive Status Examination: a brief but quantitative approach to cognitive assessment Ann Intern Med 1987 107 4 481 5 3631786
Justice AC Covinsky KE Berlin JA Assessing the generalizability of prognostic information Ann Intern Med 1999 130 6 515 24 10075620
Engelhart C Eisenstein N Meininger J Psychometric properties of the neurobehavioral cognitive status exam Clin Neuropsychol 1994 8 4 405 415
Mitrushina M Abara J Blumenfeld A Aspects of validity and reliability of the Neurobehavioral Cognitive Status Examination (NCSE) in assessment of psychiatric patients J Psychiatr Res 1994 28 1 85 95 8064643 10.1016/0022-3956(94)90037-X
Kukull WA Interrater reliability of Alzheimer's disease diagnosis Neurology 1990 40 2 257 60 2300244
Prince MJMA Richards N Quarishi 5 Horn I The development and initial validation of a telephone-administered cognitive test battery (TACT) International journal of methods in psychiatric research 1999 8 49 57
Roccaforte WH Validation of a telephone version of the mini-mental state examination J Am Geriatr Soc 1992 40 7 697 702 1607586
Ferrucci L Is the telephone interview for cognitive status a valid alternative in persons who cannot be evaluated by the Mini Mental State Examination? Aging (Milano) Ferrucci 1998 10 4 332 8
Desmond DW Tatemichi TK Hanzawa L The telephone interview for cognitive status (TICS): reliability and validity in stroke sample International journal of geriatric psychiatry 1994 9 803 807 10.1002/gps.930091006
Debanne SM Validation of a Telephone Cognitive Assessment Battery J Am Geriatr Soc 1997 45 11 1352 9 9361661
Ward HW Cognitive function testing in comprehensive geriatric assessment. A comparison of cognitive test performance in residential and clinic settings J Am Geriatr Soc 1997 38 10 1088 92 2229861
Kent J Plomin R Testing specific cognitive abilities by telephone and mail Intelligence 1987 11 391 400 10.1016/0160-2896(87)90019-5
Robins LN Reflections on testing validity of psychiatric interviews Arch Gen Psychiatry 1985 42 918 924 3899050
White J Relationship between cognitive and Emotional Function, asnd Quality of Life in Patients with Pulmoanry Arterial Hypertension (PAH) Am J Respir Crit Care Med 2004 169 7 A174
Breitner JC APOE-epsilon4 count predicts age when prevalence of AD increases, then declines: the Cache County Study Neurology 1999 53 2 321 31 10430421
Garrett KD The relationship between cardiovascular risk factors and cognitive decline J Int Neuropsychol Soc 2003 9 243
Potter GG The effects of coronary artery bypass graft on cognitive status change among elderly male twins J Int Neuropsychol Soc 2003 9 243 244
Glasso KD Relative validity of multiple telephone versus face-to-face 24-hour dietary recalls Ann Epidemiol 1994 4 332 336 7921324
Matthews CE Evaluation of computerized 24 hour physical activity recall Med Sci Sports Exerc 2002 34 suppl S 236
Matthews CE Comparing physical activity assessment methods in the Seasonal Variation of Blood Cholesterol Study Med Sci Sports Exerc 2000 32 5 976 84 10795789 10.1097/00005768-200005000-00015
Cohen RA Neurocognitive functioning and improvement in quality of life following participation in cardiac rehabilitation Am J Cardiol 1999 83 9 1374 8 10235098 10.1016/S0002-9149(99)00103-4
Andersen K Cognitive impairment and mortality among nonagenarians: the Danish 1905 cohort survey Dement Geriatr Cogn Disord 2002 13 3 156 63 11893837 10.1159/000048647
Zuccala G The effects of cognitive impairment on mortality among hospitalized patients with heart failure Am J Med 2003 115 2 97 103 12893394 10.1016/S0002-9343(03)00264-X
Tozzi V Neurocognitive performance and quality of life in patients with HIV infection AIDS Res Hum Retroviruses 2003 19 8 643 52 13678465 10.1089/088922203322280856
Tozzi V Neurocognitive impairment influences quality of life in HIV-infected patients receiving HAART Int J STD AIDS 2004 15 4 254 9 15075020 10.1258/095646204773557794
Weitzner MA Psychosocial and neuropsychiatric aspects of patients with primary brain tumors Cancer Invest 1999 17 4 285 91 discussion 296–7 10225009
Harder H Cognitive functioning and quality of life in long-term adult survivors of bone marrow transplantation Cancer 2002 95 1 183 92 12115332 10.1002/cncr.10627
| 15850486 | PMC1112616 | CC BY | 2021-01-04 16:36:26 | no | Respir Res. 2005 Apr 25; 6(1):39 | utf-8 | Respir Res | 2,005 | 10.1186/1465-9921-6-39 | oa_comm |
==== Front
Virol JVirology Journal1743-422XBioMed Central London 1743-422X-2-391584273110.1186/1743-422X-2-39ResearchEBV latent membrane protein 1 abundance correlates with patient age but not with metastatic behavior in north African nasopharyngeal carcinomas Khabir Abdelmajid [email protected] Hela [email protected] Sandrine [email protected]é Mathieu [email protected] Jamel [email protected] Mounir [email protected] Tahia [email protected] Jaap [email protected] Rachid [email protected] Pierre [email protected] Laboratoire d'Anatomie et de Cytologie Pathologiques, Hôpital Universitaire Habib Bourguiba, 3029 Sfax, Tunisia2 Laboratoire de Bactériologie-Virologie, Hôpital Universitaire Habib Bourguiba, 3029 Sfax, Tunisia3 UMR 8126 CNRS/IGR, Institut Gustave Roussy, 94805 Villejuif Cedex, France4 Département de Santé Publique, Institut Gustave Roussy, 94805 Villejuif Cedex, France5 Service de Radiothérapie, Hôpital Universitaire Habib Bourguiba, 3029 Sfax, Tunisia6 Service de Chimiothérapie, Hôpital Universitaire Habib Bourguiba, 3029 Sfax, Tunisia7 Dept of Pathology, Free University Hospital, De Boelelaan 1117, 1081 HV Amsterdam, The Netherlands8 Laboratoire Privé de Pathologie, Cité-Jardin, 3029 Sfax, Tunisia2005 20 4 2005 2 39 39 4 4 2005 20 4 2005 Copyright © 2005 Khabir et al; licensee BioMed Central Ltd.2005Khabir 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
Undifferentiated nasopharyngeal carcinomas are rare in a majority of countries but they occur at a high incidence in South China and to a lesser extent in North Africa. They are constantly associated with the Epstein-Barr virus (EBV) regardless of patient geographic origin. In North Africa, the distribution of NPC cases according to patient age is bi-modal with a large group of patients being around 50 years old (80%) and a smaller group below 25 years old. We and others have previously shown that the juvenile form of NPC has distinct biological characteristics including a low amount of p53 and Bcl2 in the tumor tissue and a low level of anti-EBV IgG and IgA in the peripheral blood.
Results
To get more insight on potential oncogenic mechanisms specific of these two forms, LMP1 abundance was assessed in 82 NPC patients of both groups, using immuno-histochemistry and semi-quantitative evaluation of tissue staining. Serum levels of anti-EBV antibodies were simultaneously assessed. For LMP1 staining, we used the S12 antibody which has proven to be more sensitive than the common anti-LMP1 CS1-4 for analysis of tissue sections. In all NPC biopsies, at least a small fraction of cells was positively stained by S12. LMP1 abundance was strongly correlated to patient age, with higher amounts of the viral protein detected in specimens of the juvenile form. In contrast, LMP1 abundance was not correlated to the presence of lymph node or visceral metastases, nor to the risk of metastatic recurrence. It was also independent of the level of circulating anti-EBV antibodies.
Conclusion
The high amount of LMP1 recorded in tumors from young patients confirms that the juvenile form of NPC has specific features regarding not only cellular but also viral gene expression.
==== Body
Background
Nasopharyngeal carcinoma has a highly variable incidence depending on the geographic area [1]. It is rare in most countries including Europe and North America [1]. Very high incidence foci are located in South China (as much as 25 per 100,000-year). In addition, there are large areas of intermediate incidence including several countries of North Africa (Tunisia, Algeria and Morocco) and South-East Asia (Vietnam, Indonesia)(between 3 and 8 per 100,000-year). The vast majority of NPCs are undifferentiated (WHO type II and III). They are constantly associated with EBV except for a few cases of differentiated forms (WHO I) occuring in non-endemic areas, often related to tobacco and alcohol consumption [2].
EBV-infection of epithelial cells often results in the production of EBV particles; virus-cell interactions are peculiar in NPC cells where EBV-infection is mainly latent [3]. The full length viral genome is contained in the nuclei of all malignant cells which generally contain several copies of EBV DNA in the form of circular extra-chromosomal elements or episomes. Most viral genes – especially genes involved in the productive viral cycle – are silent, in a very large majority of tumor cells. Only a few viral genes compatible with EBV latency are consistently transcribed in NPC. These genes encode small untranslated RNAs called EBER 1 and 2 (Epstein-Barr encoded RNA) and a nuclear protein called EBNA1 (Epstein-Barr nuclear antigen 1) detected in all NPC biopsies and visualized in the majority of malignant cells. Another EBV protein called LMP1 (Latent membrane protein 1) is frequently detected in NPC biopsies but with wide variations between individual tumors. According to numerous reports from various parts of the world, there are about 50 to 60 % NPC biopsies where LMP1 can be visualized in a majority of malignant cells using conventional immuno-histo-chemistry [4-7]. Recent reports have shown that other EBV proteins – LMP2 and the BARF1 protein – are often expressed in NPC biopsies, probably also with wide quantitative variations but this remains to be substantiated [8,9]. All these viral products EBERs, EBNA1, LMP1, LMP2 and BARF1 (BamH1 A open Reading Frame 1) have oncogenic activity in experimental systems and are suspected to contribute to the malignant phenotype of NPC cells [3,9].
Another aspect of EBV association with NPC is the presence of aberrant levels of circulating antibodies directed against viral proteins, in particular against EBNA1 and lytic cycle antigens, such as EA (early antigen) and VCA (Viral Capsid Antigen) but with low antibody levels against LMP1 [10-13]. Although viral lytic cycle proteins are usually not detected in malignant cells there is a relationship between the tumor mass and the concentration of anti-VCA and EA in the blood. A likely explanation of this paradox could be that a very small fraction of malignant cells entering the lytic productive cycle is sufficient to trigger and sustain antibody response although these cells are not easily detected on tissue sections [14].
While in South China, most NPC patients are between 40 and 60 years old, in North Africa, the distribution of NPC according to age is bi-modal. Beside the main peak of incidence around 50 (80% cases), there is a secondary peak between the age of 10 and 25 (20% cases). Previous reports have shown that the juvenile forms of NPC have some specific clinical features, sometimes reminiscent of malignant lymphomas [15,16]. For example, young NPC patients have a higher rate of lymph node metastases than adult patients and they are subjected to earlier recurrences. On the other hand, there is a good presumption that young NPC patients are cured when the complete remission last more than one year [15]. We and others have previously reported that the juvenile and adult forms of NPC have distinct biological characteristics. P53 and Bcl2 are more abundantly expressed in the adult forms whereas c-kit is more frequently detected in the juvenile form [16-18]. There are also reports showing that anti-VCA and EA antibodies are less abundant in the juvenile form suggesting a lower rate of escape from viral latency in tumors from youg patients [13,19]. LMP1 whose expression is highly variable in NPC specimens is suspected to play a role not only in oncogenesis but also in the maintenance of latency [20]. Therefore the aim of this study was to combine investigations on LMP1 expression with assessment of anti-VCA and EA antibodies in the two age groups of North African NPCs. We have found that LMP1 is expressed at a higher level in the juvenile form of NPC. However there is no direct relationship between LMP1 abundance and a low level of circulating anti-VCA and EA antibodies.
Results
Patients and tumor specimens
Primary NPC biopsy samples were collected with informed consent from 82 patients, prior to any treatment, in the Sfax University Hospital, between January 1993 and December 1999. The ages ranged from 10 to 77 years (mean age: 43 years). Twenty two (27%) patients were less than thirty years old. The clinical stage of the disease was determined according to the TNM classification of the AJCC/UICC (1997). Five (6%) patients were at stage II, twenty (25%) patients were at stage III and fifty seven (69%) were at stage IV. NPC histological type was determined on tissue sections according to the World Health Organisation (WHO) classification, resulting in the following distribution : 1/82 keratinising squamous cell carcinoma (SCC, WHO type 1, 1.2%), 52/82 non-keratinizing carcinoma (NKC, WHO type 2, 63%) and 29/82 undifferentiated carcinomas (UC, WHO type 3, 35%). All patients were treated by irradiation of the nasopharynx and/or cervical lymph nodes. Fifty one (62%) were first treated by induction chemotherapy. The follow-up period which was the time between the last day of radiation therapy and either the day of death or the date of the last examination varied from 1 to 116 months.
LMP1 expression in tumor cells and correlations with clinical data
Immunohistochemistry using the anti-LMP1 antibody S12 resulted in highly heterogenous staining between tumors from different patients. It was assessed using a scoring system based on the percentage of positive cells and the intensity of staining. Scores of LMP1 varied from 2 to 12 with a mean of 7.6 (+/- 2.6 SD)(Fig. 1 and Table 1). LMP1 staining was also highly heterogeneous within the tumor tissue for each single patient. Both types of heterogeneity did not simply result from the presence of the EBV-negative infiltrating lymphocytes. There were true variations in the amount of LMP1 staining visible in malignant cells, from one patient to another and within a given tumor. We found no NPC specimens with complete absence of S12 staining. Even when staining was minimal, a fraction of cells were nevertheless LMP1-positive with moderate intensity, thus resulting in a score of 2. In contrast, we found a complete absence of staining on sections of lung or laryngeal carcinomas used as negative controls, resulting in a minimal score of 0 (Fig. 1 and data not shown). In the NPC sections with minimal LMP1 staining, we found no specific features of the rare LMP1-positive malignant cells, in terms of cell morphology or relationship with tumor vessels, lymphoid infiltrate or foci of necrosis.
Figure 1 LMP1 immunostaining on tissue sections of NPC samples. A. Intense and diffuse LMP1 expression in an NPC biopsy from a 47 year old patient (score 12, 400X) B. Intense LMP1 expression in a limited area in an NPC biopsy from a 17 year old patient (score 7, 600X) C. Moderate and diffuse LMP1 expression in an NPC biopsy from a 44 year patient (score 8, 400X) D. Absence of LMP1 expression in a lung carcinoma biopsy (score 0, 600 X)
Table 1 Variations of the LMP1 score according to clinical and histo-pathological data
Number of Specimens Mean Score (SDa) pb
Sex
Male 57 7.5 (2.6) p = 0.47
Female 25 7.9 (2.8)
Age
< 30 22 9.0 (2.4) p = 0.004
≥ 30 60 7.1 (2.6)
Histological typec
SCC 1 7.1 (3.1) p = 0.42
NKC 52 8.0 (2.3)
UC 29 7.6 (2.7)
TNMd
T2 + T3 40 7.2 (2.6) p = 0.18
T4 42 8.0 (2.7)
N0 21 7.4 (2.9) p = 0.67
N+ 61 7.7 (2.6)
M0 73 7.8 (2.5) p = 0.83
M+ 9 7.6 (2.7)
Metastatic relapse
+ 24 7.5 (2.8) p = 0.94
- 47 7.6 (2.6)
NAe 11
a SD: standard deviation b Based on the Student t-test cHistological type : SCC : squamous cell carcinoma, NKC : non-keratinizing carcinoma, UC : undifferentiated carcinoma. d Clinical staging: primary tumor extension classified T2, T3 or T4 according to AJCC/UICC (1997); regional lymph node extension classified N0 in the absence of clinical or radiological evidence of lymph node invasion at the initial workup, N+ in the other cases; metastatic status defined as M0 in the absence of clinical or radiological evidence of distant metastasis at the initial workup, M+ in the other cases (synchronous metastases). eNA : not applicable.
We attempted to find relationships of the LMP1 score with various clinical parameters. We found a highly significant influence of patient age on LMP1 score (p = 0.004)(Table 1). In contrast, we found no relationships with lymph node or extra-nodal metastases at initial examination neither with the occurrence of a metastatic relapse. There was also no relationship with the WHO histological type (Table 1).
Lack of correlations between LMP1 expression and levels of serum anti-EBV antibodies
As previously reported in other studies, the serum profile of anti-EBV antibodies was not identical in the two age groups of NPC patients [13,21]. Serum levels of anti-VCA and EA IgG were significantly lower in the juvenile form whereas the anti-EA and VCA IgA were undetectable (<10) in majority of young patients (Table 2). Because LMP1 is known to antagonize entry in the lytic cycle in some experimental models we hypothesized that LMP1 might block production of EA and VCA in NPC cells and therefore prevent an increase of circulating antibodies directed to these viral proteins [20]. With this in mind we attempted to find an inverse relationship between the levels of anti-VCA and -EA IgG and IgA on one hand and the level of LMP1 expression in the tumor tissue on the other hand. Using univariate analysis, a significant inverse relationship was found only between the level of LMP1 expression and the level of serum anti-EA IgA (Table 3; p = 0.012). However, this result was not confirmed by multivariate analysis including patient age and title of anti-EA IgA as co-variables. In other words, both LMP1 amounts in the tumor tissue and titles of serum anti-EA IgA are strongly influenced by patient age but there is no direct link between these 2 parameters.
Table 2 Variations of anti-VCA and -EA Ig titles according to patient ages
Age category (patient number)
EBV-antibody titles < 30 years (n = 21) ≥ 30 years (n = 47) pa
IgG VCA < 320 7 (33,3%) 4 (8,5%) 0,03
IgG EA < 40 12 (57,1%) 6 (12,8%) 2.7 × 10-4
IgA VCA < 10 15 (75%) 7 (14,9%) < 10-5
IgA EA < 10 16 (76,2%) 12 (25,5%) <10-3
a Based on the Fisher exact test
Table 3 Variations of the LMP1 score according to serum levels of anti-EBV antibodies
Number of Specimens Mean Score (SDa) pb
IgA VCA title
< 10 22 8.4 (2.9) p = 0.10
≥ 10 45 7.2 (2.7)
NDc 15
IgA EA title
< 10 28 8.6 (2.8) p = 0.012
≥ 10 40 6.9 (2.5)
NDc 14
a SD: standard deviation b Based on the Student t-test c ND : not determined
Discussion
Heterogeneity in LMP1 expression in NPC biopsies has been noticed since early studies based on Western blotting. LMP1 amounts can vary from traces only detectable after long exposure of the immunoblots to high levels comparable to those found in EBV-transformed B-lymphocytes [22,23]. For this reason, the rate of NPC specimens recorded as LMP1-postive is highly dependent on the sensitivity of the method used for its detection. For example when using RT-PCR with one round of PCR amplification, LMP1 products are detected in only a fraction of NPC biopsies; in contrast, the percentage of positive samples is often close to 100% when making a second round of PCR using nested primers [24,25]. The same applies to investigations by immuno-histo-chemistry (IHC). According to a recent report by Dietz et al., the percentage of LMP1-postive NPCs markedly increases when using a tyramid-enhancement process instead of conventional tissue staining [26].
In contrast to our study, all previous articles reporting LMP1 detection in NPCs by conventional IHC have recorded a fraction of about 40% specimens as LMP1-negative tumors [4-7]. In most cases, these groups of LMP1-negative tumors were in fact made of 2 categories : specimens with complete absence of LMP1-positive cells and specimens with a percentage of stained cells below an arbitrary threshold of 5 or 10%. In our study, we have found no biopsy completely devoid of LMP1-positive cells. This is probably due to the fact that we have used the S12 antibody which is more sensitive in staining of tissue sections than the CS1-4 antibody from Dako [27]. Hence, to our knowledge, CS1-4 was used in all previous investigations of LMP1 expression in NPC biopsies [4-7]. In addition, we have chosen not to consider any threshold of minimal LMP1 expression; LMP1 staining has been scored even when the protein was visible in a very small fraction of malignant NPC cells.
A large series of studies performed in vitro have produced an impressive amount of data suggesting that LMP1 can induce various phenotypic changes consistent with a metastatic behavior. For example in transfected cells, LMP1 can induce the production of the c-Met receptor and of the metallo-protease MMP9 as well as the down-regulation of the E-cadherin [5,28,29]. In this context, it is surprising to find no relationship between LMP1 score and the presence of lymph node or visceral metastases at initial examination or the risk of metastatic recurrence. In this regard, our data are in contrast with two previous reports showing a relationship between LMP1 expression and the frequency of metastases [5,30]. However more recently, Jeon et al. have found a relationship between LMP1 expression and MMP9 expression but not between LMP1 and the presence of metastases [6]. Investigations of LMP1 expression on novel prospective series of NPC patients using the S12 monoclonal antibody might be useful to solve these discrepancies.
Conclusion
The most striking finding of this study is the observation of a higher level of LMP1 expression in the juvenile form of NPC. It provides clear evidence that this clinical form has specific biological features not only in terms of cellular gene expression but also in terms of latent viral gene expression. From previous studies it was known that anti-VCA and EA IgG and IgA were at a low level in the juvenile form by contrast with the adult form of NPC [13,19]. This observation was confirmed by our own data. However, we found no direct relationship between LMP1 expression and a low level of anti-VCA and EA IgG and IgA. In futures studies, it will be important to investigate in both age-groups of NPCs the status of other EBV-proteins which are suspected to be expressed in this malignancy with a rather heterogenous pattern, for example LMP2A, LMP2B and the BARF1 protein [8,9]. Another issue will be to investigate the anti-LMP1 immune response in the juvenile form of NPCs for example the status of circulating anti-LMP1 antibodies [11].
Methods
Pathological diagnosis and immunohistochemical staining of LMP1
All tumor specimens were fixed in Bouin's fixative (75 % saturated picric acid, 25 % formalin, 5% glacial acetic acid) and paraffin-embedded for ligth microscopy and immunohistochemistry. The diagnosis was based on morphological examination after Hematoxylin and Eosin staining. It was further assisted by immuno-staining of Leucocyte Common Antigen and cytokeratin in 29 cases, in order to facilitate the differential diagnosis with a malignant lymphoma or a sarcoma. Tumor sections from all 82 NPC patients were stained with the anti-LMP1 S12 monoclonal antibody. In addition, two squamous carcinomas of the larynx and one squamous lung carcinoma were also stained with S12 and used as negative controls. Five μm sections attached on silanized slides were de-waxed in xylene, rehydrated in graded ethanol, covered with 10 mM citrate buffer (pH 6) and heated in a microwave oven for two consecutive 10 minute periods, at 500 W. They were then incubated for 15 to 30 minutes with the purified primary antibody S12 (0.5 to 1 μg/ml)[27,31]. Primary antibody binding was visualized with biotin-labelled secondary antibodies and a streptavidin-peroxidase complexe using di-aminobenzidine as a chromogenic substrate (LSAB system, Dako).
Scoring method
Immuno-staining was scored on the basis of the approximate percentage of positive tumor cells and the relative immunostaining intensity. Sections from each biopsies were read and scored independently by two pathologists (AK and RJ) who were blinded to the patient clinical data. Five consecutive microscope fields were analyzed. The differences in scores between the two observers were resolved at a conference microscopy (AK, RJ and TB). The following grading system was adopted to score the number of positive tumor cells: 0, none seen in the section; 1, presence of positive cells even rare but not exceeding 25%; 2, 26 to 50% positive cells; 3, 51 to 75%; and 4, 76 to 100%. Immuno-staining intensity was rated as follows: 0, none; 1, weak; 2, moderate; and 3, intense. When the staining intensity was heterogeneous, each component of the tumor were scored independently and the results were summed. For example, when a specimen contained 50% of the tumor cells with moderate intensity (2 × 2 = 4), 25% of tumor cells with intense immunostaining (1 × 3 = 3), and 25% of cells with weak intensity (1 × 1 = 1), the score was 4 +3 +1 = 8. The maximal possible score was twelve.
Serological analysis
Serum samples were collected from 68 out of the 82 patients at initial diagnosis. IgG and IgA antibodies to EBV EA and VCA were titrated by indirect immunofluorescence on Raji and P3HR1 cells, respectively [13,32].
Statistical analysis
LMP1 immunostaining scoring results were expressed as means (standard deviation, SD) and compared using the Student t-test. Variations of anti-EBV antibody titles according to patient age were assessed using the Fisher exact test. To assess relationships between LMP1 score, age and anti-EBV antibody titles, multivariate analysis was carried out using a linear multiple regression (Sas software, version 8, SAS Institute Inc, Cary, NC, USA). All tests were bilateral with a 5% level.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
AK, RJ and TB made pathological diagnosis, immunohistochemistry and scoring of immunostaining, HK carried out assessment of serum EBV-antibodies, PB and SR participated in the design and coordination of the study and helped to draft the manuscript, MR performed the statistical analysis, JD and MF gathered clinical data, JM purified the S12 antibody and set up conditions for its use in immunohistochemistry. All authors read and approved the final manuscript.
Acknowledgements
This study was supported by a cooperative grant from the French CNRS and Tunisian DGRST (n° 17963) and by a grant from the "Comité du Cher" of the French "Ligue Nationale contre le Cancer".
==== Refs
Busson P Keryer C Ooka T Corbex M EBV-associated nasopharyngeal carcinomas: from epidemiology to virus-targeting strategies Trends Microbiol 2004 12 356 360 15276610 10.1016/j.tim.2004.06.005
Nicholls JM Agathanggelou A Fung K Zeng X Niedobitek G The association of squamous cell carcinomas of the nasopharynx with Epstein-Barr virus shows geographical variation reminiscent of Burkitt's lymphoma J Pathol 1997 183 164 168 9390028 10.1002/(SICI)1096-9896(199710)183:2<164::AID-PATH919>3.3.CO;2-A
Raab-Traub N Epstein-Barr virus in the pathogenesis of NPC Semin Cancer Biol 2002 12 431 441 12450729 10.1016/S1044579X0200086X
Niedobitek G Fahraeus R Herbst H Latza U Ferszt A Klein G Stein H The Epstein-Barr virus encoded membrane protein (LMP) induces phenotypic changes in epithelial cells Virchows Arch B Cell Pathol Incl Mol Pathol 1992 62 55 59 1352076
Horikawa T Sheen TS Takeshita H Sato H Furukawa M Yoshizaki T Induction of c-Met proto-oncogene by Epstein-Barr virus latent membrane protein-1 and the correlation with cervical lymph node metastasis of nasopharyngeal carcinoma Am J Pathol 2001 159 27 33 11438450
Jeon YK Lee BY Kim JE Lee SS Kim CW Molecular characterization of Epstein-Barr virus and oncoprotein expression in nasopharyngeal carcinoma in Korea Head Neck 2004 26 573 583 15229899 10.1002/hed.10370
Shao JY Ernberg I Biberfeld P Heiden T Zeng YX Hu LF Epstein-Barr virus LMP1 status in relation to apoptosis, p53 expression and leucocyte infiltration in nasopharyngeal carcinoma Anticancer Res 2004 24 2309 2318 15330177
Heussinger N Buttner M Ott G Brachtel E Pilch BZ Kremmer E Niedobitek G Expression of the Epstein-Barr virus (EBV)-encoded latent membrane protein 2A (LMP2A) in EBV-associated nasopharyngeal carcinoma J Pathol 2004 203 696 699 15141385 10.1002/path.1569
Seto E Yang L Middeldorp J Sheen TS Chen JY Fukayama M Eizuru Y Ooka T Takada K Epstein-Barr virus (EBV)-encoded BARF1 gene is expressed in nasopharyngeal carcinoma and EBV-associated gastric carcinoma tissues in the absence of lytic gene expression J Med Virol 2005 76 82 88 15778977 10.1002/jmv.20327
Meij P Vervoort MB Aarbiou J van Dissel P Brink A Bloemena E Meijer CJ Middeldorp JM Restricted low-level human antibody responses against Epstein-Barr virus (EBV)-encoded latent membrane protein 1 in a subgroup of patients with EBV-associated diseases J Infect Dis 1999 179 1108 1115 10191211 10.1086/314704
Xu J Ahmad A D'Addario M Knafo L Jones JF Prasad U Dolcetti R Vaccher E Menezes J Analysis and significance of anti-latent membrane protein-1 antibodies in the sera of patients with EBV-associated diseases J Immunol 2000 164 2815 2822 10679125
Fachiroh J Schouten T Hariwiyanto B Paramita DK Harijadi A Haryana SM Ng MH Middeldorp JM Molecular diversity of Epstein-Barr virus IgG and IgA antibody responses in nasopharyngeal carcinoma: a comparison of Indonesian, Chinese, and European subjects J Infect Dis 2004 190 53 62 15195243 10.1086/421245
Karray H Ayadi W Fki L Hammami A Daoud J Drira MM Frikha M Jlidi R Middeldorp JM Comparison of three different serological techniques for primary diagnosis and monitoring of nasopharyngeal carcinoma in two age groups from Tunisia J Med Virol 2005 75 593 602 15714486 10.1002/jmv.20310
Zhang JX Chen HL Zong YS Chan KH Nicholls J Middeldorp JM Sham JS Griffin BE Ng MH Epstein-Barr virus expression within keratinizing nasopharyngeal carcinoma J Med Virol 1998 55 227 233 9624611 10.1002/(SICI)1096-9071(199807)55:3<227::AID-JMV8>3.0.CO;2-3
Daoud J Toumi N Bouaziz M Ghorbel A Jlidi R Drira MM Frikha M Nasopharyngeal carcinoma in childhood and adolescence: analysis of a series of 32 patients treated with combined chemotherapy and radiotherapy Eur J Cancer 2003 39 2349 2354 14556927 10.1016/S0959-8049(03)00512-4
Khabir A Ghorbel A Daoud J Frikha M Drira MM Laplanche A Busson P Jlidi R Similar BCL-X but different BCL-2 levels in the two age groups of north African nasopharyngeal carcinomas Cancer Detect Prev 2003 27 250 255 12893071 10.1016/S0361-090X(03)00098-9
Khabir A Sellami A Sakka M Ghorbel AM Daoud J Frikha M Drira MM Busson P Jlidi R Contrasted frequencies of p53 accumulation in the two age groups of North African nasopharyngeal carcinomas Clin Cancer Res 2000 6 3932 3936 11051240
Bar-Sela G Kuten A Ben-Eliezer S Gov-Ari E Ben-Izhak O Expression of HER2 and C-KIT in nasopharyngeal carcinoma: implications for a new therapeutic approach Mod Pathol 2003 16 1035 1040 14559987 10.1097/01.MP.0000089778.48167.91
Sbih-Lammali F Clausse B Ardila-Osorio H Guerry R Talbot M Havouis S Ferradini L Bosq J Tursz T Busson P Control of apoptosis in Epstein Barr virus-positive nasopharyngeal carcinoma cells: opposite effects of CD95 and CD40 stimulation Cancer Res 1999 59 924 930 10029086
Prince S Keating S Fielding C Brennan P Floettmann E Rowe M Latent membrane protein 1 inhibits Epstein-Barr virus lytic cycle induction and progress via different mechanisms J Virol 2003 77 5000 5007 12663807 10.1128/JVI.77.8.5000-5007.2003
Sbih-Lammali F Djennaoui D Belaoui H Bouguermouh A Decaussin G Ooka T Transcriptional expression of Epstein-Barr virus genes and proto-oncogenes in north African nasopharyngeal carcinoma J Med Virol 1996 49 7 14 8732865 10.1002/(SICI)1096-9071(199605)49:1<7::AID-JMV2>3.0.CO;2-A
Young LS Dawson CW Clark D Rupani H Busson P Tursz T Johnson A Rickinson AB Epstein-Barr virus gene expression in nasopharyngeal carcinoma J Gen Virol 1988 69 ( Pt 5) 1051 1065 2836550
Fahraeus R Fu HL Ernberg I Finke J Rowe M Klein G Falk K Nilsson E Yadav M Busson P Expression of Epstein-Barr virus-encoded proteins in nasopharyngeal carcinoma Int J Cancer 1988 42 329 338 2843473
Chen F Hu LF Ernberg I Klein G Winberg G Coupled transcription of Epstein-Barr virus latent membrane protein (LMP)-1 and LMP-2B genes in nasopharyngeal carcinomas J Gen Virol 1995 76 ( Pt 1) 131 138 7844522
Brooks L Yao QY Rickinson AB Young LS Epstein-Barr virus latent gene transcription in nasopharyngeal carcinoma cells: coexpression of EBNA1, LMP1, and LMP2 transcripts J Virol 1992 66 2689 2697 1313894
Dietz A Logothetis CA Helbig M Flechtenmacher C Rudat V Dollner R Wallner F Bosch FX Prognostic Impact of EBV-Related LMP-1, Histologic Type, and Environmental Factors in Nasopharyngeal Carcinoma in a German Population Onkologie 2004 27 345 350 15347888 10.1159/000079086
Jiwa NM Oudejans JJ Dukers DF Vos W Horstman A van der Valk P Middledorp JM Walboomers JM Meijer CJ Immunohistochemical demonstration of different latent membrane protein-1 epitopes of Epstein-Barr virus in lymphoproliferative diseases J Clin Pathol 1995 48 438 442 7543118
Tsai CN Tsai CL Tse KP Chang HY Chang YS The Epstein-Barr virus oncogene product, latent membrane protein 1, induces the downregulation of E-cadherin gene expression via activation of DNA methyltransferases Proc Natl Acad Sci U S A 2002 99 10084 10089 12110730 10.1073/pnas.152059399
Takeshita H Yoshizaki T Miller WE Sato H Furukawa M Pagano JS Raab-Traub N Matrix metalloproteinase 9 expression is induced by Epstein-Barr virus latent membrane protein 1 C-terminal activation regions 1 and 2 J Virol 1999 73 5548 5555 10364303
Hu LF Chen F Zhen QF Zhang YW Luo Y Zheng X Winberg G Ernberg I Klein G Differences in the growth pattern and clinical course of EBV-LMP1 expressing and non-expressing nasopharyngeal carcinomas Eur J Cancer 1995 31A 658 660 7640034 10.1016/0959-8049(94)00468-K
Mann KP Staunton D Thorley-Lawson DA Epstein-Barr virus-encoded protein found in plasma membranes of transformed cells J Virol 1985 55 710 720 2991591
Henle W Henle G Zajac BA Pearson G Waubke R Scriba M Differential reactivity of human serums with early antigens induced by Epstein-Barr virus Science 1970 169 188 190 4316788
| 15842731 | PMC1112617 | CC BY | 2021-01-04 16:38:59 | no | Virol J. 2005 Apr 20; 2:39 | utf-8 | Virol J | 2,005 | 10.1186/1743-422X-2-39 | oa_comm |
==== Front
Virol JVirology Journal1743-422XBioMed Central London 1743-422X-2-401584514510.1186/1743-422X-2-40MethodologyStereophysicochemical variability plots highlight conserved antigenic areas in Flaviviruses Schein Catherine H [email protected] Bin [email protected] Werner [email protected] Sealy Center for Structural Biology, Department of Human Biology, Chemistry and Genetics, University of Texas Medical Branch at Galveston, TX, USA2 Sealy Center for Vaccine Development, Department of Human Biology, Chemistry and Genetics, University of Texas Medical Branch at Galveston, TX, USA3 Department of Microbiology and Immunology, University of Texas Medical Branch at Galveston, TX, USA2005 21 4 2005 2 40 40 21 3 2005 21 4 2005 Copyright © 2005 Schein et al; licensee BioMed Central Ltd.2005Schein 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
Flaviviruses, which include Dengue (DV) and West Nile (WN), mutate in response to immune system pressure. Identifying escape mutants, variant progeny that replicate in the presence of neutralizing antibodies, is a common way to identify functionally important residues of viral proteins. However, the mutations typically occur at variable positions on the viral surface that are not essential for viral replication. Methods are needed to determine the true targets of the neutralizing antibodies.
Results
Stereophysicochemical variability plots (SVPs), 3-D images of protein structures colored according to variability, as determined by our PCPMer program, were used to visualize residues conserved in their physical chemical properties (PCPs) near escape mutant positions. The analysis showed 1) that escape mutations in the flavivirus envelope protein are variable residues by our criteria and 2) two escape mutants found at the same position in many flaviviruses sit above clusters of conserved residues from different regions of the linear sequence. Conservation patterns in T-cell epitopes in the NS3- protease suggest a similar mechanism of immune system evasion.
Conclusion
The SVPs add another dimension to structurally defining the binding sites of neutralizing antibodies. They provide a useful aid for determining antigenically important regions and designing vaccines.
==== Body
Background
Flaviviruses, +-strand RNA viruses that cause diseases such as yellow fever (YF), Japanese encephalitis (JE), West Nile (WN), tick-borne encephalitis (TBE) and Dengue fever (DV), are endemic in many parts of the world. While some flaviviruses have relatively stable sequences, others are extremely variable. For example, some have suggested the term "quasispecies" for DV, as several different virus sequences could be isolated from the same blood sample [1,2]. The many asymptomatic human and animal carriers of these viruses represent an enormous reservoir for the development of new strains[3,4]. Continuous mutation at positions that are non-essential for replication allows flaviviruses to evade or confuse the immune system. This contributes to the development of fatal infections, such as Dengue hemorrhagic fever (DHF) [5,6]. To be effective, vaccines must induce efficient T-cell [7,8] and neutralizing antibody responses to functionally essential areas of the viral proteins[9].
Previous efforts to identify residues in flaviviruses that are essential for function have used escape mutants, viral progeny that survive in the presence of neutralizing antibodies to the virus [10-16]. However, while escape variants may have altered phenotypes[10,14], they do not prevent the replication of the virus, implying that the mutations are in residues not essential for function[17]. Here, we present a method that can be used to interpret escape mutations in a different way, by detecting conserved residues that are "cloaked" by these variable positions. These invariant residues are more likely to be the important targets of neutralizing antibodies the escape mutants, which typically occur at variable positions.
The method depends on our PCPMer program for analyzing variability, according to physicochemical properties of the amino acids, in sequence alignments. We have shown that the position specific variability data generated by the program, when coupled with structural analysis, can be used to identify areas that are important for function in families of related proteins [18-22]. Here, as a paradigm for the use of the method in vaccine design, we applied this method to the analysis of escape mutants of flaviviruses.
We used the PCPMer program to define areas conserved in physical chemical properties (PCP-motifs) of DV proteins of known structure. We then color coded the structures according to PCP-variability, and marked the position of known escape mutants and viral epitopes. The data divided the protein surface into a variable face, where all the escape mutants mapped, and a more conserved face. These areas were consistent with those previously defined by experimental methods [23-26]. We find that the escape mutants found in the same position in several different flaviviruses lie above highly conserved, known functional areas of the viral proteins, such as the receptor binding site, and disulfide bonded residues. These cloaked residues are more likely to be the true target for a neutralizing antibody.
Results
Defining PCP-motifs of DV proteins with PCPMer
A "PCP-motif" is an area in a group of related proteins with conserved physical chemical properties (PCPs). We have shown in previous work that PCP-motifs correspond to functional areas of proteins and can be used to identify functional homologues in sequence databases ([27,21]). The PCP motifs for two DV proteins of known structure, the Envelope and the serine protease domain of the non-structural protein NS3 are shown in Tables 1 and 2. For convenience in this paper, the motifs are given as areas of the Dengue virus protein sequence, rather than as the matrix of numbers relating to the conserved properties at each position that is their actual description (see methods).
Table 1 PCP-Motifs identified for the flavivirus Envelope proteins, using the sequence of DV-2env to indicate the sequence location and representative sequence. PCPMer parameters were: Gap cutoff of 2, length cutoff of 5 and the relative entropy range between 1 and 2.5 with a step of 0.1.
Motif No. PCP-motifs
1 9 RDFVEGVSG 17*i
2 24 VLEHGSCVTTMAKNKPTLD 42
3 54 ATLRKYCIEA 63
4 74 CPTQGEP 80F
5 98 DRGWGNGCGLFGKGG 112F
6 116 CAMFTC 121
7 133 ENLEYTV 139
8 151 VGNDT 155
9 159 GKEVKITPQSS 169
10 175 LTGYGTVTMEC 185
11 197 VLLQMK 202
12 209 HRQWFLD 215
13 240 FKNPHAKKQDV 250F
14 281 GHLKCRLRMDKLQLKGMSYSMC 302
15 314 ETQHGT 319
16 332 PCKIPF 337
17 349 GRLITVNP 356
18 368 EAEPPFGD 375**
19 391 WFKKGSSIGQ 400
20 416 GDTAWDFGSLGG 427
21 431 SIGKALHQVFGAI 443
22 448 FSGVSW 453
23 459 IGVIITWIGMNSR 471
24 475 LSVSLVLVGVVTLYL 489
*residues 9–12 are at the C-terminal end of the CD8-T-cell epitope mapped for yellow fever [46]
**a Y to H mutation in TBE virus just after this motif is attenuating for neurovirulence [44]
F Motifs that form the "fusion tip area" of domain II. Mutation of the analogous residue in TBE to the bold and underlined H in motif 13 prevents fusion [32].
i Residues equivalent to those that form a salt bridge in the TBE-envelope are bold (R9/E368), those in the interface of the trimer of the TBE envelope protein are underlined [31].
Table 2 PCP-motifs identified for the flavivirus NS3 proteases, using the sequence of DV-2 NS3 to indicate location and representative sequence. The catalytic residues (H51, D75, S135) are shadowed; residues in the substrate interaction pocket [28] are bold, and areas that are part of known T-cell epitopes are underlined.
Motif No. PCP-motifs
1 2 GVLWDVPSP 10
2 29 GILGYSQIGAG 39
3 43 EGTFHTMWHVTRGA 56
4 73 KKDLISYGGGW 83*
5 95 VQVLALEPG 103
6 133 GTSGSP 138**
7 148 GLYGNG 153
8 159 GAYVSAIAQ 167
*C-terminal end, cytotoxic T-cell epitope AA 71–79 [40].
**N-terminal end of 11 mer T-cell activating peptide [5].
The PCP-motifs include all the known functional areas of the proteins, according to previous experimental results, and indicate areas that are most probably responsible for the activities that are common to all the flaviviruses. For example, the motifs of the NS3 protease (Table 2) include all the catalytic amino acids and all but one of the residues that interact with a peptide substrate analogue in a crystal structure of the complex[28].
Mapping PCP-motifs of the DV-Env protein defines a conserved face and a fusion tip region
Mapping the motifs of Table 1 on the 3D structure, determined by X-ray crystallography, of the DV-2 envelope protein (DV residues 281– 674)[29,30](Figure 2a) shows that they map primarily to one, conserved face of the molecule. Many of these sequences are involved in interdomain and trimer interactions of the envelope protein from TBE[31]. The plot reveals that three of the motifs occur near one another at the end of Domain II. These are the previously defined "fusion peptide" and two other loops that are as much as 140 amino acids away in the linear sequence of the protein (the three areas are marked by F in Table 1). This suggests that the whole tip of the protein is involved in fusion. We note that mutation in one of these loops in TBE (at the absolutely conserved H shaded in Table 1) does indeed effect viral fusion[32].
Figure 1 The relative specific entropy (SE) function of PCPMer (Bin Zhou et al., in preparation) defines motifs even in alignments where the sequence conservation varies locally. The top of the figure shows a section of the sequence alignment for the NS3 protein. The next section shows the PCPMer output, indicating the motifs in the NS3 protease according to the sequence of DV-2 as a function of the specific entropy level (numbers to the left). PCPMer parameters were: Gap cutoff of 2, length cutoff of 5, relative entropy range between 1 and 2.5 with a step of 0.2. Note the conserved sequences around the active site residues (bold letters) of the protease are followed by variable regions that retain conservation in one of the five physical chemical property vectors. The output is colored to reflect the degree of conservation at each position.
Figure 2 Variability analysis of the envelope protein of DV-2 and illustration of how escape mutants mark cloaked conserved residues. a) PCP-motifs (blue) common to all flavivirus envelope proteins are mapped on the structure of DV-2 Env (PDB file 1OAN; the start and end residues are numbered). Note the high conservation of the fusion peptide (arrow) and two loop regions adjacent to it from other areas of the molecule. b) Stereochemical variability plot (SVP) of the DV2-Env (PDB file 1OKE), showing the per residue variation across the Flaviriridiae. Known escape mutants of DV-2 and DV-3 [10, 26] are labeled and the residue names are colored according to their variability. The boxed residues are intermediate in the conservation scale (white). c and d) Surface plots of the SVP shown in figure 2B, showing the conserved (overall blue, c) face, where the motifs of conserved areas map. The variable face (d, mostly red), which matches the orientation of the molecule where the escape mutants map to.
Visualizing clusters of conserved residues from different sequence areas with SVPs
Alternatively, the conservation of each amino acid, represented by the specific entropy, [33,34], as described in more detail in Figure 1 and Methods, can be mapped onto a protein's 3-D structure, by coloring each amino acid. The higher the specific entropy, the more conserved the position. These 3-D plots, which we refer to as "stereophysicochemical variability plots" (Figure 2, 3, 4, 5) can be used to find conserved areas of the protein, distant in the sequence, that are close together in 3D space.
Figure 3 a) Local surface plot of the DV-Env SVP around residue 124 (which is highly variable but has been colored green here for clarity), illustrating how the residue forms part of a patch of variable (red) residues b) Removing part of the surface reveals how I124 lies above the highly conserved residues Cys60-121 (disulfide bonded) and Tyr59 that are distant in the sequence of the protein.
Figure 4 a) Local surface plot of the area around two escape mutant positions in domain III. Residues 307 and 311 are highly variable and have been colored green here for visibility. The faint blue area on the surface near residue 307 comes from a highly conserved aromatic residue, Phe306, which lies under the variable residue and forms a cluster with another conserved residue, Tyr326 (b).
Figure 5 The conserved essential residues in two serotype specific T-cell epitopes of the NS3 protease are followed by variable residues that will affect MHC binding. The crystal structure of the NS3 protease of DV-2 is colored according to sequence variability across the flaviviruses(see figure 1 and 2 for details). The catalytic triad sidechains are shown in neon and labeled. The T-cell epitopes (right) around the catalytic residues Asp75 (residues 71–79), and Ser135 (133–143), are shown as space filling and color coded to reflect variability, except that the 100% conserved catalytic residues are both black.
The SVP can be used to define the amino acid profile of antibody binding sites, which have been localized by peptide mapping or escape mutants. An SVP for the DV-env structure (Figure 2b–d), with the specific entropy of each residue shown by color, shows the two faces, one conserved (where the PCP-motifs map; 2c) and one variable (2b). This is also in accord with another analysis of conservation in this protein, where only the identical residues in these 14 flaviviruses were plotted[35]. The SVP indicates that most of the residues on the "inner" face of the envelope protein (Figure 2c) are conserved in at least one PCP-vector, even if the sequence at these positions varies.
Escape mutants of the DV-Env protein are in variable positions near conserved residues
Known escape mutants of DV[10,26] map to the variable face of the SVP (Figure 2b, 2d) and are generally in highly variable positions. This is consistent with a previous sequence analysis of mutants of tick-borne flaviviruses [36]. Those in the more conserved positions show a limited range of alteration in the progeny escape virus. For example, Residue 112, which is at the border of the fusion peptide motif, is the most conserved of the mutants. This residue is either an S or G in all the flavivirus sequences, with one escape mutant described as "S112G"[26].
Lifting the cloak: the hidden essential residues
To illustrate how the method can be used to define the target of neutralizing antibodies, the areas around two escape mutants in the Env protein of DV, both in positions where escape mutations occur in many different flaviviruses, were analyzed.
Example 1: Residue 124 in domain 1
Type specific neutralizing antibodies that bind near this position have been found for four other flaviviruses[26], suggesting that the area cloaked by this residue constitutes an area essential for function. Surviving progeny with mutations at this position either conserve the residues hydrophobicity (YF-17D: Met125Ile), or convert it to a hydrophilic residue (DV: I124N, JE: I126T, MVE: A126E, TBE: A123K).
Zooming in on this region (figure 3a, detail) reveals that Ile124 lies above the PCP- motif 54–63 in the folded structure. The side chains of residues around it cloak two cysteines that are disulfide bonded, Cys60 and Cys121, and the highly conserved residue Tyr 59 (figure 3b). This suggests that converting Ile124 and the residues that surround it in the 3D-structure to small hydrophilic residues should enhance the immunogenicity of the areas below that must be blocked to obtain neutralization by the antibodies.
Example 2: Residue 307 in domain 3
Two DV escape mutants positions, 307 and 311, are part of a variable surface in domain III (figure 4a). A mutation at position 307 has also been observed in other flaviviruses. For example, a Lys307Glu mutation has been implicated in attenuating West Nile neurovirulence[13]. Further, this mutation and another that is close to it structurally, at position 330, block the binding of several antibodies that neutralize WN replication[17]. Attenuated Tick borne encephalitis virus was obtained by mutating residues near this position, which is considered to be an important site for receptor binding[37]. All this indicates that while residue 307 can vary, some residue near it must be essential and conserved.
According to our analysis, Residue 307 and the variable residues near it cloak two highly conserved aromatic residues, Phe 306 and Tyr 326, that overlap each other (Figure 4b). Similar analysis (not shown) of the NMR structure of WN-env in this area points to the equivalent residues being cloaked by two escape mutants (at positions 307 and 330 in WN), which both block neutralization by three different monoclonal antibodies[17]. We suggest that these aromatic residues contribute to the epitopes detected by the antibodies, and that the antibody prevents wild type virus replication by blocking their conformational change during receptor binding. In our alignment of the Flaviviruses (supplementary data), only the 17D vaccine strain of Yellow fever varies at these positions. Mutation studies are now underway to determine whether these residues play a role in attenuation.
T-cell epitopes in NS3 protease: variable residues alter binding to peptides containing essential conserved positions
A similar pattern can be seen with T-cell epitopes, which are also important determinants of the immune response [38,39]. Two dominant T-cell epitopes have been identified in the NS3 of DV, the protease that cleaves the polyprotein at several positions, at residues 71–79[40] and 133–143[5]. Both of these epitopes contain amino acids that are essential for protease activity, D75 and S135 respectively. In the SVP representation of the crystal structure [41] of the DV-2 NS3 protease domain (Figure 5), both the conserved essential amino acids are surrounded by variable residues which will alter binding to T-cell antigens. This could be the basis for reduced binding to the immune surveillance (HLA alleles) and clearance mechanisms in infected individuals. For the first epitope, the determinant for cytotoxic T-cell activation by DV-2 or DV-3 serotype is in one variable amino acid, D71[40], which is S71 in DV-3 serotypes (Figure 1). This sort of analysis can aid in choosing vaccine strains, on the basis of how their sequences conform to the known binding signatures for major MHC alleles, while retaining viral replication in cell culture.
Discussion
While residue conservation has long been recognized as a way to detect important areas of viral proteins[35,42,43], the new tools presented here for distinguishing conservation with respect to the PCPs of the amino acid side chains provide a rapid way to interpret an ensemble of sequence and escape mutant data. Plotting per residue conservation of physical chemical properties in the form of SVPs permits one to rapidly detect which residues are most likely to contribute to the epitope face (Figure 2b,d), and which are most important for the function of the virus (Figure 2a,c, Table 1 and 2). As we show, the mathematical methods are robust to sample size (in this case, PCPmer analysis with 8 or 14 dependable virus sequences gave similar plots, as described in Methods), and the specific entropy criterion is a useful measure of the importance of residues. For example, the PCP-motifs in the NS3 protease domain contain all amino acids known to be important for function and substrate binding (Table 2 and Figure 5).
Applying the PCPMer decomposition methods to the flavivirus family aided in interpreting experimental results and suggested site specific alterations that can be tested for vaccine design. PCPMer motifs, combined with structural analysis, are a rapid way to identify functionally important areas of proteins[21,22]. Colored SVPs supply a fourth dimension, variability, to a crystal or model structure, that is a valuable aid in interpreting experimental data. Mapping known escape mutants on SVPs shows they occur in areas of high variability. While the altered residues in escape mutants may have deleterious effects under some growth conditions[10,14,36,44], they are confined to residues that are not essential for replication. The masked, conserved residues below them are more likely to be important. For example, the escape mutants at positions 124 and 307 (Figure 3 and 4) mark the site of antibody binding, but the SVP suggests that conserved amino acids, close in space but 20–60 residues away in the linear sequence, are more important for neutralization. Similar principles may also apply to T-cell epitopes, according to the NS3 example (Figure 5). The SVP can further be used to suggest other amino acids in a composite site could be altered to better direct antibodies to an essential area of the protein.
An effective immune response must be generated against the conserved regions of the viral proteins, and not be diverted by the variable cloak around them. Our analysis of the whole range of flaviviruses indicates that physicochemical properties remain constant even in stretches of residues that would appear by other measures to be variable. The SVP methodology provides a novel way to select mutants or design a virus so as to enhance the accessibility of the conserved residues that are normally cloaked.
Conclusion
We have shown that PCP motifs and SVPs provide a rapid method to obtain information from viral sequence data. Once identified, these areas can be used to design vaccine candidate recombinant viruses or individual proteins that will more efficiently stimulate an effective immune response to these essential areas. We anticipate that the PCPmer program and related visualization tools will be a routine method in the future for analyzing sequence data of variable virus sequences.
Methods
Variability analysis
A large set (237) of physicochemical properties of amino acid side chains were reduced to five descriptors (E1-E5) by multidimensional scaling [34]. The 5 descriptors summarize all known quantitative properties that differentiate the sidechains, including among others the hydrophobicity (defined in numerous ways), amino acid size, tendency to occur in secondary structures, charge, binding to various affinity chromatography columns. These descriptors offer an alternative to the commonly used scoring matrices, such as the PAM series and Gonnet, which are based on statistical analysis of amino acid substitutions, to determine areas of residue conservation in proteins. Our program suite, PCPMer , defines areas of conservation in aligned protein sequences according to the values of the five vectors at each position in a sequence alignment [27,34]. The user specified values dictate allowed gaps, minimum length, and entropy range for the motifs. To determine the information content of the pattern of residue properties in a column of the multiple alignment, MOTIFMAKER determines a "specific entropy" value of the component Ei at position k relative to the expected random distribution [27] :
The term "entropy" is used here as it is in information science, as a measure of the uncertainty of a given event [45]. The relative entropy is thus the observed conservation of the physical chemical property vectors (b = 1–5) of residues in a column relative to that which would be expected if the position varied randomly. In this case, a high specific entropy indicates that the conservation in a column of a multiple alignment, according to a given physical property vector, is significantly greater than chance.
Relative specific entropy
Even in variable areas of sequence alignments, there may be a pattern of conservation in one of the vectors that underlies the amino acid sequence diversity. A relative specific entropy scale can be used to determine motifs in alignments where the variability depends on position in the sequence. Alternatively, a sliding entropy scale can be used to define particularly conserved regions in alignments that are generally more homogeneous in character. Figure 1 illustrates the usefulness of this feature. The top rows show a section of the alignment of NS3 proteins from 8 flaviviruses. The lower section shows the output of PCPMer for this area, according to which residues in the top (marker) sequence would be part of a motif at each specific entropy level. The program automatically takes the most highly conserved areas in each section of the alignment to be motifs.
PCP-motif definition
The user can choose the minimum length of motifs ("length cutoff"), the maximum number of variable positions between two conserved ones ("Gap cutoff") and the specific entropy range for defining motifs in PCPMer. For the sake of simplicity in this paper, the motifs in Figure 1 and Tables 1 and 2 are given according to the top sequence in the alignment from which they were derived. However, the actual definition of the PCP-motifs is a series of numerical matrices, that define the type and degree of conservation of the physical chemical properties of each column in the original sequence alignment. These matrices can be used to automatically scan sequence databases, using the MOTIFMINER program, to identify proteins that contain sequences similar to the PCP-motifs defined for the initial set of proteins [27]. In this work, we have chosen to identify PCP-motifs as highly conserved areas of the viral proteins whose conservation would indicate an important functional or structural role.
Stereophysicochemical variability plots
The plots of Figures 2 and 3 were drawn with MOLMOL, from the PDB coordinate files of the crystal structures, using a macro that colors each residue position according to the specific entropy values determined by PCPMer. The median of a histogram of the number of residues at each specific entropy level was defined as the midpoint in the color scheme. The highest specific entropy for any residue was set to blue and the lowest to red.
Flavivirus alignment
The whole genome sequences of 8 flaviviruses that included representatives of each DV serotype were downloaded from GENBANK and the areas for the envelope and NS3 protein selected (alignments are provided as supplementary data). The sequences were aligned with CLUSTALW using a GONNET matrix and standard set conditions. This alignment was used for the initial analysis of the envelope protein and the analysis of the NS3 protease shown in Figures 1 and 3. To test the robustness of the method, a second alignment of 14 flavivirus sequences (Supplementary data 1; used for Figures 2, 3, 4) for the envelope protein was generated and the analysis was repeated. There was little or no difference in the positional variability or the specific entropy calculations with the larger number of sequences, but the ends of two of the motifs were slightly different.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
Catherine H. Schein: Sequence decomposition, interpretation of results; alignment design and data analysis
Bin Zhou: chief programmer of the PCPMer program, prepared SVPs, sequence analysis
Werner Braun: design of the PCPMer method
Supplementary Material
Additional File 1
Multiple sequence alignment (CLUSTAL W (1.82)) of 14 flavirus envelope sequences. Escape mutant positions are bold, the conserved residues in domains I (Y59, C60, C121) and III (F306, Y326) that are near escape mutants common to several flaviviruses are in red. Corresponding residues that are variant in the yellow fever 17D strain are in bold and underlined.
Click here for file
Acknowledgements
This work was supported by grants from the Sealy Center for Vaccine Development (462855), the Department of Energy (DE-FG-00ER63041), and the National Institutes of Health (R21AI055746-01). We thank Robert Davey for suggestions during the development of the SVP method.
==== Refs
Holmes EC Moya A Is the quasispecies concept relevant to RNA viruses? J Virol 2002 76 460 465 11739715
Wang WK Lin SR Lee CM King CC Chang SC Dengue type 3 virus in plasma is a population of closely related genomes: quasispecies. J Virol 2002 76 4662 4665 11932434
Holmes EC Twiddy SS The origin, emergence and evolutionary genetics of dengue virus. Infect Genet Evol 2003 3 19 28 12797969
Halstead SB Deen J The future of dengue vaccines. LANCET 2002 360 1243 1245 12401270
Mongkolsapaya J Dejnirattisai W Xu X Vasanawathana S Tangthawornchaikul N Chairunsri A Sawasdivorn S Duangchinda T Dong T Rowland-Jones S Yenchitsomanus P McMichael A Malasit P Screaton G Original antigenic sin and apoptosis in the pathogenesis of dengue hemorrhagic fever. Nature Med 2003 9 921 927 12808447
Loke H Bethell DB Phuong CX Dung M Schneider J White NJ Day NP Farrar J Hill AV Strong HLA Class I-restricted T cell responses in dengue hemorrhagic fever: a double-edged sword? J Infect Dis 2001 184 1369 1373 11709777
Sbai H Mehta A DeGroot AS Use of T-cell epitopes for vaccine design Current Drug Targets-Infectious Disorders 2001 1 303 313 12455403
Muellbacher A Lobigs M Lee E Immunobiology of mosquito-borne encephalitic flaviviruses Advances in Virus Research 2003 60 87 120 14689692
Sun W Edelman R Kanesa-Thasan N Eckels KH Putnak JR King AD Houng HS Tang D Scherer JM Hoke CH Innis BL Vaccination of human volunteers with monovalent and tetravalent live-attenuated dengue vaccine candidates. Am J Trop Med Hyg 2003 69 24 31 14740952
Serafin IL Aaskov JG Identification of epitopes on the envelope (E) protein of dengue 2 and dengue 3 viruses using monoclonal antibodies. Archives of Virology 2001 146 2469 2479 11811694
Strauss EG Stec DS Schmaljohn AL Strauss JH Identification of antigenically important domains in the glycoproteins of Sindbis virus by analysis of antibody escape variants J Virol 1991 65 4654 4664 1714515
Friedrich T Dodds EJ Yant LJ Vojnov L Rudersdorf R Cullen C Evans DT Desrosiers RC Mothe BR Sidney J Sette A Kunstman K Wolinsky S Piatak M Lifson J Hughes AL Wilson N O'Connor DH Watkins DI Reversion of CTL escape-variant immunodeficiency viruses in vivo. Nat Med 2004 10 275 281 14966520
Chambers TJ Halevy M Nestorowicz A Rice CM Lustig S West Nile virus envelope proteins: nucleotide sequence analysis of strains differing in mouse neuroinvasiveness. Journal of General Virology 1998 79 2375 2380 9780042
Holzmann H Stiasny K Ecker M Kunz C Heinz FX Characterization of monoclonal antibody-escape mutants of tick-borne encephalitis virus with reduced neuroinvasiveness in mice J Gen Virol 1997 78 31 37 9010282
Beasley DW Aaskov JG Epitopes on the dengue 1 virus envelope protein recognized by neutralizing IgM monoclonal antibodies. Virology 2001 279 447 458 11162801
Beasley DW Barrett AD Identification of neutralizing epitopes within structural domain III of the West Nile virus envelope protein. J Virol 2002 76 13097 13100 12438639
Volk D Beasley DW Kallick DA Holbrook MR Barrett AD Gorenstein DG Solution structure and antibody binding studies of the envelope protein domain III from the New York strain of West Nile virus. J Biol Chem 2004 279 38755 38761 15190071
Izumi T Schein CH Oezguen N Feng Y Braun W Effects of backbone contacts 3' to the abasic site on the cleavage and the product binding by human apurinic/apyrimidinic endonuclease (APE1). Biochemistry 2004 43 684 689 14730972
Murtazina D Puchkaev AV Schein CH Oezguen N Braun W Nanavati A Pikuleva IA Membrane protein interactions contribute to efficient 27-hydroxylation of cholesterol by mitochondrial cytochrome P450 27A1. J Biol Chem 2002 277 37582 37589 12124390
Schein CH The shape of the messenger: using protein structural information to design novel cytokine-based therapeutics. Curr Pharm Des 2002 8 2113 2129 12369857
Schein CH Zhou B Oezguen N Venkatarajan SM Braun W Molego based definition of the architecture and specificity of metal binding sites Proteins 2005 58 200 210 15505785
Schein CH Özgün N Izumi T Braun W Total sequence decomposition distinguishes functional modules, "molegos" in apurinic/apyrimidinic endonucleases BMC-Bioinformatics 2002 3 37 12445335
Rey FA Heinz FX Mandl C Kunz C Harrison SC The envelope glycoprotein from tick-borne encephalitis virus at 2 A resolution Nature 1995 375 291 298 7753193
Guirakhoo F Heinz FX Kunz C Epitope model of tick-borne encephalitis virus envelope glycoprotein E: analysis of structural properties, role of carbohydrate side chain, and conformational changes occurring at acidic pH. Virology 1989 169 90 99 2466373
Mandl CW Guirakhoo F Holzmann H Heinz FX Kunz C Antigenic structure of the flavivirus envelope protein E at the molecular level, using tick-borne encephalitis virus as a model J Virol 1989 63 564 571 2463377
Roehrig JT Antigenic structure of flavivirus proteins Advances in Virus Research 2003 59 141 175 14696329
Venkatarajan SM Schein CH Braun W Identifying property based sequence motifs in protein families and superfamilies: application to APE. Bioinformatics 2003 19 1381 1390 12874050
Murthy HMK Clum S Padmanabhan R Dengue Virus NS3 Serine Protease. Crystal structure and insights into interaction of the active site with substrates by molecular modeling and structural analysis of mutational effects. J Biol Chem 1999 274 5573 5580 10026173
Modis Y Ogata S Clements D Harrison SC A ligand-binding pocket in the dengue virus envelope glycoprotein. Proc Natl Acad Sci 2003 100 6986 6991 12759475
Modis Y Ogata S Clements D Harrison SC Structure of the dengue virus envelope protein after membrane fusion. Nature 2004 427 313 319 14737159
Bressanelli S Stiasny K Allison SL Stura EA Duquerroy S Lescar J Heinz FX Rey FA Structure of a flavivirus envelope glycoprotein in its low-pH-induced membrane fusion conformation. EMBO J 2004 23 728 738 14963486
Chanel-Vos C Kielian M A conserved histidine in the ij loop of the Semliki Forest virus E1 protein plays an important role in membrane fusion. J Virol 2004 78 13543 13552 15564465
Hänggi G Braun W Pattern recognition and self-correcting distance geometry calculations applied to myohemerythrin. FEBS Lett 1994 344 147 153 8187874
Venkatarajan MS Braun W New quantitative descriptors of amino acids based on multidimensional scaling of a large number of physical-chemical properties. J Mol Mod 2001 7 445 453
Seligman SJ Bucher DJ The importance of being outer: consequences of the distinction between the outer and inner surfaces of flavivirus glycoprotein E Trends in Microbiology 2003 11 108 110 12648938
Gritsun TS Holmes EC Gould EA Analysis of flavivirus envelope proteins reveals variable domains that reflect their antigenicity and may determine their pathogenesis Virus Res 1995 35 307 321 7785318
Mandl CW Allison SL Holzmann H Meixner T Heinz FX Attenuation of Tick-Borne Encephalitis Virus by Structure-Based Site-Specific Mutagenesis of a Putative Flavivirus Receptor Binding Site JOURNAL OF VIROLOGY 2000 74 9601–9609 11000232
Mathew A Kurane I Rothman AL Zeng LL Brinton MA Ennis AA Dominant recognition by human CD8+ cytotoxic T lymphocytes of dengue virus nonstructural proteins NS3 and NS1.2a. J Clin Invest 1996 98 1684 1691 8833919
Butrapet S Huang CY Pierro DJ Bhamarapravati N Gubler DJ Kinney RM Attenuation markers of a candidate dengue type 2 vaccine virus, strain 16681 (PDK-53), are defined by mutations in the 5' noncoding region and nonstructural proteins 1 and 3. J Virol 2000 74 3011 3019 10708415
Zivny J DeFronzo M Jarry W Jameson J Cruz J Ennis FA Rothman AL Partial agonist effect influences the CTL response to a heterologous dengue virus serotype J Immunol 1999 163 2754 2760 10453018
Murthy HMK Judge K DeLucas L Padmanabhan R Crystal structure of dengue virus NS3 protease in complex with a Bowman-Birk inhibitor: implications for flaviviral polyprotein processing and drug design. J Mol Biol 2000 301 759 767 10966782
Chang GJ Gubler DJ and Kuno G Molecular biology of dengue viruses. Dengue and dengue hemorrhagic fever 1997 1st Wallingford, Oxfordshire, CAB International 175 198
Rico-Hesse R Microevolution and virulence of dengue viruses. Adv Virus Res 2003 59 315 341 14696333
Holzmann H Heinz FX Mandl CW Guirakhoo F Kunz C A single amino acid substitution in envelope protein E of Tick-borne encephalitis virus leads to attenuation in the mouse model J Virol 1990 64 5156 5159 2398538
Durbin R Eddy S Krogh A Mitchison G Biological Sequence Analysis: Probabilistic models of proteins and nucleic acids 2000 Cambridge U.K., Cambridge University Press
van der Most RG Harrington LE Giuggio V Mahar PL Ahmed R Yellow fever virus 17D envelope and NS3 proteins are major targets of the antiviral T cell response in mice Virology 2002 296 117 124 12036323
| 15845145 | PMC1112618 | CC BY | 2021-01-04 16:38:58 | no | Virol J. 2005 Apr 21; 2:40 | utf-8 | Virol J | 2,005 | 10.1186/1743-422X-2-40 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1001583314210.1186/1471-2105-6-100Research ArticleFiltering high-throughput protein-protein interaction data using a combination of genomic features Patil Ashwini [email protected] Haruki [email protected] Institute for Protein Research, Osaka University, 3-2 Yamadaoka, Suita, Osaka 565-0871, Japan2 Department of Biology, Graduate School of Science, Osaka University, 1-1 Machikaneyama-cho, Toyonaka, Osaka 560-0043, Japan2005 18 4 2005 6 100 100 21 12 2004 18 4 2005 Copyright © 2005 Patil and Nakamura; 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
Protein-protein interaction data used in the creation or prediction of molecular networks is usually obtained from large scale or high-throughput experiments. This experimental data is liable to contain a large number of spurious interactions. Hence, there is a need to validate the interactions and filter out the incorrect data before using them in prediction studies.
Results
In this study, we use a combination of 3 genomic features – structurally known interacting Pfam domains, Gene Ontology annotations and sequence homology – as a means to assign reliability to the protein-protein interactions in Saccharomyces cerevisiae determined by high-throughput experiments. Using Bayesian network approaches, we show that protein-protein interactions from high-throughput data supported by one or more genomic features have a higher likelihood ratio and hence are more likely to be real interactions. Our method has a high sensitivity (90%) and good specificity (63%). We show that 56% of the interactions from high-throughput experiments in Saccharomyces cerevisiae have high reliability. We use the method to estimate the number of true interactions in the high-throughput protein-protein interaction data sets in Caenorhabditis elegans, Drosophila melanogaster and Homo sapiens to be 27%, 18% and 68% respectively. Our results are available for searching and downloading at .
Conclusion
A combination of genomic features that include sequence, structure and annotation information is a good predictor of true interactions in large and noisy high-throughput data sets. The method has a very high sensitivity and good specificity and can be used to assign a likelihood ratio, corresponding to the reliability, to each interaction.
==== Body
Background
Protein-protein interactions in various organisms are increasingly becoming the focus of study in the identification of cellular functions of proteins. Though small scale experiments have contributed significantly to our knowledge of protein-protein interactions, the bulk of the data is available from high-throughput methods like yeast two hybrid (Y2H) and mass spectrometry of coimmunoprecipitated complexes (Co-IP) [1]. Such data is currently available for H. pylori [2], S. cerevisiae (baker's yeast) [3-6], C. elegans [7], D. melanogaster [8] and H. sapiens [9]. However, protein-protein interaction data obtained from high-throughput experiments is thought to have a large number of false positives i.e. interactions that are spurious or biologically irrelevant, and do not occur in the cell [10]. This fraction is estimated to be as high as 50% in yeast [1,11]. Since the false positives are unknown, there is no consensus on which interactions from these data sets should be used in prediction studies. Studies that use all the interactions run the risk of predicting spurious ones [12], while those that completely ignore the high-throughput data are limited by the amount of data from small scale experiments[13]. For some high-throughput studies, the authors specify the reliable interactions as 'high confidence' or 'core' interactions [4,7,8] which have fewer false positives but which do not take into account those low confidence interactions which are known to be true. Hence it is important to quantify the reliability of these interactions and identify the true positives i.e. interactions that actually occur in the cell.
Several methods have been previously used to identify true interactions from high-throughput experimental data in yeast. Sequence homology was used by Deane et al. [14] in the form of a paralogous verification method (PVM) whereby an interaction in yeast is judged to be true if the concerned proteins have paralogs that interact as well. But these results are limited by the number of proteins that have known paralogs. They also used similarity in gene expression profiles to identify true positives [14]. Structurally known interactions were used by Edwards et al., who compared experimental interactions found in RNA polymerase II, Arp2/3 complex and the proteasome with those observed in the 3D structures [15]. Though this method has a high reliability, it is limited by the number of structures available in Protein Data Bank (PDB) [16]. von Mering et al. found the interactions in yeast that are observed in more than one high-throughput experiment to estimate the fraction of true positives [1]. The results obtained were surprisingly small in this case due to inherent biases in different experimental methods. Database annotations have been used by Sprinzak et al. in the form of co-localization data of the interacting proteins and their cellular role to estimate the number of true positive interactions in yeast [11]. However, not all model organisms have well annotated genomes. Interaction network topology is another means of identifying true interactions. Saito et al. used an interaction generality measure (IG2), based on the topological properties of the interaction network, to assess the reliability of an interaction [17]. Bader et al. used screening statistics and network topology to quantify the confidence of each interaction [18]. Though these methods have a high specificity (low false positive rate), they have low sensitivity (low true positive rate), since the number of proteins with more than one interaction partners is relatively few.
Since none of the methods give a good performance (high sensitivity and specificity) by themselves, it follows that a combination of methods would perform better. Jansen et al. have shown that a combination of genomic features results in a more accurate prediction of the yeast protein interaction network [19]. In this study, we use a similar approach of combining various genomic features using naïve Bayesian networks to predict the true interactions in high-throughput data sets.
In selecting the genomic features to be used in our model, we decided to combine sequence, structure and database annotation information about the interaction. Sequence information was incorporated through homologous interactions. We used our Homologous Interactions (HINT) database [20] to obtain homologs for all high-throughput interactions [21]. Structure information was incorporated in the form of interacting Pfam domains [22] found in the PDB. We used the 3did database to obtain a list of such Pfam domains [23]. Database annotation information was used in the form of Gene Ontology (GO) terms used to describe the interacting proteins [24].
We computed the reliability of each feature using likelihood ratios and combined their evidence using naïve Bayesian networks in order to predict the true interactions from high-throughput data sets. Bayes' rule provides a good method to estimate posterior odds of an event in the presence of prior evidence [25]. Bayesian approaches have also been used frequently in the past to calculate the reliability or to assign probabilities to protein-protein interactions [15,19,26].
In this study, we show that an interaction can be judged to be true if either or all of the following are true:
1. the interacting proteins have homologs that interact,
2. the interacting proteins each have a Pfam domain found to interact with the other in PDB and,
3. the interacting proteins have at least one identical GO annotation.
We used protein-protein interaction data from the Database of Interacting Proteins (DIP) [27] (July 2004 release) and IntAct [28] (September 2004 release). We prove our hypothesis first in yeast by estimating likelihood ratios for high-throughput interactions based on the number of known true positives and false positives using Bayesian network approaches. Based on these results, we estimate the number of true positives in the high-throughput data sets of S. cerevisiae, C. elegans, D. melanogaster and H. sapiens. The results can be searched at and downloaded from our website [29].
Results
Calculating the reliability of each genomic feature
We used protein-protein interactions from high-throughput data sets for yeast as our test set to calculate the reliability of each genomic feature (see Methods). Of these 12,674 interactions, we chose a set of 3464 interactions as our gold standard – 1479 as the positive gold standard and 1985 interactions as the negative gold standard (see Methods). Our goal was to maximize the interactions identified in the gold positive set (high sensitivity) and at the same time minimize the number of interactions identified in the gold negative set (high specificity).
In these interactions, we identified all those that had homologous interactions. The true positives (TP) were those interactions with homologs that were in the gold positive set, and the false positives (FP) were those that were in the gold negative set. Using these values, we calculated the likelihood ratio (L) for the genomic feature of 'homologous interactions' (see Methods). Similarly, we calculated the likelihood ratios for the other two genomic features – interacting proteins with at least one identical GO annotation and interacting proteins having one of 2 Pfam domains known to interact in PDB. We also calculated the likelihood ratio for the absence of genomic features.
Figure 1 shows the likelihood ratios calculated. Likelihood ratio (L) expresses the reliability of each genomic feature. An L > 1 indicates the ability of the genomic feature to identify more true positives than false positives. As seen in Figure 1, all the genomic features have L values greater than 1. The absence of any genomic feature to support the interaction results in L < 1. This indicates that in the absence of any support from the selected genomic features, the interaction is more likely to be a false one. Interacting Pfam domains in the interacting proteins gives the highest L showing that interactions with evidence from structural data have the highest reliability. This is followed by the L values of similar GO annotations for interacting proteins and the presence of homologous interactions respectively.
Using naïve Bayesian Networks to combine the evidence of genomic features
We used naïve Bayesian networks to combine the evidence of each genomic feature for a particular interaction. Since naïve Bayesian networks require that the genomic features be conditionally independent of each other, we calculated the Pearson's correlation coefficient for a pair of genomic features to ascertain their independence (see Methods). We then combined the evidence of each interaction by simply multiplying the L values of each genomic feature found for the interaction. Thus, to each interaction in the gold set of 3464 interactions, we assigned an L value based on the genomic features it had. An L value greater than 1 represents higher posterior odds of an interaction being true than prior odds (see Methods). Hence all interactions with an L value greater than 1 were predicted as true. Table 4 shows the L values obtained for each possible combination of genomic features supporting an interaction. For instance, an interaction that is supported by the presence of all 3 genomic features has the highest L value, thus having the highest probability of being true.
Assessing the accuracy of the predictions in yeast
To assess the accuracy of our method, we identified the number of predicted true interactions in the gold positive set and those in the gold negative set respectively. We conducted 10-fold cross-validation on the limited set of yeast high-throughput interactions to calculate the sensitivity and specificity of the method. Figure 2 shows the receiver operating characteristic (ROC) curve for our method. Each point on the ROC curve denotes the sensitivity and specificity obtained on the inclusion of interactions with a lower L value. A particular L value is associated with a specific combination of the 3 genomic features (Table 4). Thus, including the interactions supported by the presence of all 3 genomic features (d+g+h) in the results gives a sensitivity of 12.3% and a specificity of 99.4%. On further including interactions supported by interacting Pfam domains and similar GO annotations (d+g), the sensitivity rises to 14.5% and the specificity marginally decreases to 99.3%. As interactions supported by each individual feature or other combinations of features that have an L > 1, are included in the results, the sensitivity increases at the cost of specificity. Thus our method predicts interactions, which are supported by at least one of the 3 genomic features, to be true with a sensitivity of 89.7% and a specificity of 62.8%.
Predicting true interactions in all high-throughput data sets
We used our method to assign L values to all interactions in three other high-throughput data sets for C. elegans, D. melanogaster and H. sapiens [7-9]. We also assigned L values to the interactions in yeast high-throughput data sets [3-6] that were not part of the gold standard. We predicted all interactions with L > 1 as true interactions. Table 5 shows the distribution of the predicted true interactions across different L values for each species. Figure 3 shows the percentage of interactions predicted as true in the high-throughput data sets of each species.
Authors of the high-throughput data sets usually assign a confidence level to interactions. Those interactions that are either reconfirmed experimentally or have a high probability of being true based on some statistical method are deemed as high confidence with the rest being low confidence interactions. We tested the overlap between our predicted true interactions and the high and low confidence data sets given by the authors. As seen in Figure 4, more high-confidence interactions are predicted as true in all data sets, except in H. sapiens [9]. For instance, 52.8% of the high confidence interactions in yeast are predicted to be true by our method, as opposed to 27.9% of the low confidence interactions.
Some validated predictions
Figure 5 shows two instances where our method predicts low confidence interactions to be true. Figure 5A gives the interactions between the proteins ps, mub, bl and aret. These proteins have all been recently shown to co-regulate the alternative splicing of Dscam exon 4 in D. melanogaster [30]. Figure 5B shows the interactions between the Lsm proteins in the mRNA degradation process in yeast that were predicted to be true by our method. These interactions were later confirmed by similar ones in the human mRNA degradation process [31].
Discussion
We present here a method to identify the true interactions in high-throughput protein-protein interaction data sets using a combination of three genomic features. We used the likelihood ratio (L) to evaluate the accuracy and reliability of each genomic feature. We combined the evidence from each genomic feature using naïve Bayesian networks. Our method gives a sensitivity of 89.7% which is higher than any of the other methods used so far. Our method also has a good specificity at 62.9%. We chose the three genomic features to maximize the inclusion of all aspects of information about the interactions.
Structure information was incorporated through Pfam domains found to interact in PDB structures in the 3did database. As would be expected, this feature has the highest accuracy and reliability as shown by its high L value (Figure 1). As seen in the ROC curve (Figure 2), this genomic feature gives the lowest number of false positives (high specificity). However, the number of true positives (sensitivity) is limited by the small number of complex structures in PDB that can be used to identify interacting Pfam domains. The sensitivity will significantly improve as the number of structures in PDB increases.
Database annotations were included through the use of GO annotations of the interacting proteins. This feature shows the second highest reliability (Figure 1). It is also able to identify the maximum number of true positives. Indeed, Lin et al. have recently shown that GO annotations are the dominant contributors in predicting protein-protein interactions [32]. As the number of annotated proteins increases, this method promises to be useful in filtering interaction data.
Sequence information was included in the form of homologous interactions found using the HINT database. Homologous interactions do not give the reliability expected (Figure 1), perhaps because they are not limited to orthologous or paralogous interactions. However, it is the only feature that does not require any protein annotations and is useful in identifying true interactions of un-annotated or hypothetical proteins. Methods based on network topology [17,18] are also independent of protein annotations and would be a useful addition to the genomic features. However, we have not considered it in the current study.
Though evidence from each feature can independently predict an interaction to be true, a combination of 2 or more features performs better (Table 4). For instance, the combination of interacting Pfam domains and similar GO annotations (d+g) or interacting Pfam domains and homologous interactions (d+h), has a higher L value than either of the features independently. Both combinations increase the sensitivity without much compromise in the specificity. Similarly, a combination of similar GO annotations and homologous interactions (g+h), predicts an interaction to be true with a higher probability than each feature independently. This combination too adds to the sensitivity with only a slight decrease in the specificity. Surprisingly, evidence from interacting Pfam domains (d) performs better than that of the combination of the other two features (g+h), highlighting the importance of the incorporation of structural evidence.
Due to the absence of information about non-interacting proteins, we prepared our gold negative set from proteins that have different subcellular localizations. However, some interactions are transient with interacting proteins residing in the same sub-cellular compartment for only a small fraction of their life time. As a result, some of the interactions in the gold negative set are actually true. Thus, the specificity of our method is probably higher than 62.9%.
Using the evidence of the three genomic features, we predicted the number of true interactions in various high-throughput data sets. Our prediction of 56.3% true interactions in yeast high-throughput data sets is in conformance with the previous estimates of the number of false positives in these data sets [1,11]. However, yeast Y2H [3,4] and Co-IP data sets [5,6] show very different numbers of true positives independently – 37% and 73% respectively (data not shown). The D. melanogaster data set [8] shows a very low rate of true positives. One reason could be that this experiment was performed using most of the predicted transcripts in the D. melanogaster genome, including biologically irrelevant ones. The C. elegans data set [7] includes a higher percentage of true positives than D. melanogaster, perhaps because the experiment was performed only on a restricted set of predicted proteins related to multi-cellular functions. The H. sapiens data set [9] shows the highest number of true positives at 67.9%. This data set was obtained from a study that focused on the identification of putative protein complexes in the TNF-α/NF-κB signal transduction pathway using Co-IP [9]. There are two possible reasons for the high number of true positives. Firstly, the choice of proteins from a specific signal transduction pathway precludes many random interactions between proteins of unrelated functionality. Secondly, the Co-IP approach to the identification of protein complexes, and thus interactions, is known to have a low false positive rate of around 20% [1,9], in comparison to Y2H approaches. This is also reflected in the prediction of a larger number (73%) of true positive interactions in the Co-IP data sets of yeast[5,6] by our method.
We also studied the overlap of the predicted true interactions with the high-confidence and low-confidence interactions as given by the respective authors. Though the number of interactions predicted in high-confidence data sets is higher, 17–28% of the interactions in low-confidence data sets are also predicted to be true, except in the H. sapiens data set. This shows that some low-confidence interactions can be biologically relevant. When compared to other data sets, the predicted true interactions in H. sapiens data set show a much higher overlap (68.7%) with the low confidence interactions. This is because the high confidence data set given by Bouwmeester et al. primarily focuses on interactions, novel or otherwise, that are most likely to be a part of the signalling cascade triggered by TNF-α [9]. Hence, interactions of proteins that are also part of other systems, like the cell cycle, are not included in this high confidence data set. These include interactions of nucleasome assembly proteins and MCM proteins, among others. Other interactions which have been filtered out include those of frequently copurified proteins like the Heat Shock Proteins. In order to limit their interaction map to the TNF-α/NF-κB signal transduction pathway, the authors have chosen a very stringent statistical criterion to identify the interactions of proteins that are expressed well above their normal levels on being triggered by TNF-α [9]. As a result, the high confidence data set as given by Bouwmeester et al. forms only 10% of the total interactions identified in their study, while our method predicts a large number of low confidence interactions to be true.
We were also able to confirm several of the low confidence interactions, that were predicted as true, in literature using iHOP [33]. In fact, most of the interactions of the Lsm proteins, shown in Figure 5B, are found in iHOP. Several interactions from the human dataset are also found in iHOP and the Human Protein Reference Database [34]. Among others, these include the low confidence interactions of the C-Rel proto-oncogene with itself, NK-κB p105 subunit, NF-κB p100/p49 subunits, Heat shock cognate 71 kDa protein and NF-κB beta inhibitor. This further reiterates the biological relevance of a large number of low confidence interactions.
Conclusion
In this study, we show that a combination of genomic features that includes sequence, structure and annotation information, can be used to identify true interactions from high-throughput protein-protein interaction data sets. We use likelihood ratios to assess the reliability of each genomic feature and combine their evidence using naïve Bayesian networks. We provide a likelihood ratio for each predicted true interaction based on the evidence that supports it. Our method has a high sensitivity and a good specificity. The results of our study are available on our website [29]for search and download.
Methods
Yeast high-throughput data sets
Table 1 shows the number and type of interactions from the 4 yeast high-throughput data sets used. Data inferred from mass spectrometry of coimmunoprecipitated complexes (Co-IP) is converted to binary interactions using the spoke model (the spoke model has been previously shown to be more reliable than the matrix model [18,35]).
Gold standard data sets
The gold standard positive data set consisted of:
1. all physical interactions from MIPS [36] yeast two-hybrid data (excluding interactions from Uetz et al. [3] and Ito et al.[4]),
2. MIPS complexes data (excluding complexes from Gavin et al. [5] and Ho et al.[6]),
3. small-scale yeast-two hybrid experimental data from DIP and IntAct and,
4. interactions found in more than one high-throughput data sets.
Table 2 shows the number and type of interactions from each data set. For this study, the gold standard positives are limited to those found in the yeast high-throughput data sets i.e. 1479 interactions, instead of all possible gold standard positives. This is because the aim is to identify the true protein-protein interactions in high-throughput data setsin yeast, as opposed to predicting all true protein-protein interactions in yeast.
The gold standard negative data set is derived from protein localization data in yeast cells [37]. Proteins that do not exist in the same sub-cellular compartment are assumed to be non-interacting since the majority of the interactions occur between proteins in the same sub-cellular compartment [19,38,39]. As with the gold standard positive, the gold standard negative data set is also limited to those interactions found in the yeast high-throughput data sets i.e. 1985.
Genomic features
In order to predict true interactions, we identified those that have at least one of the following genomic features based on:
1. Homologous interactions – Using our HINT database [21], we identified all interactions from high-throughput data sets that had homologous interactions, including orthologous or paralogous interactions. An interaction is deemed as homologous to a given interaction when each of its interacting proteins has homologs that are found to interact in DIP or IntAct. Homologs of interacting proteins are identified by HINT using PSIBlast with 5 iterations and an E-value cut-off of 10-8.
2. GO annotations – Using data from the GO database [24], we identified all interactions from high-throughput data sets where the interacting proteins shared at least one GO term, since interacting proteins generally share a common function [39].
3. Interacting Pfam domains – We identified interactions in high-throughput data sets, where each of the interacting proteins had one of the two Pfam domains that were found to interact in PDB structures by the 3did database [23].
Correlation between genomic features
The correlation between each genomic feature was calculated using Pearson's correlation coefficient for 100 random interactions from the high-throughput data sets. The significance of each correlation coefficient was tested using a t-test with 98 degrees of freedom. Table 3 shows the correlation coefficients, t values and the probability. All the genomic features were found to be independent of each other.
Bayesian networks
Bayesian networks can be used to combine evidence from different sources and calculate the posterior odds of an event based on prior evidence [25]. The relation between the posterior odds and prior odds of finding a true interaction is given by Bayes' rule as follows:
Oposterior = L(g1, g2, g3,..., gN) Oprior, (1)
where g1, g2, g3,....., gN are genomic features of an interaction,
Oprior = prior odds of an interactions being true,
Oposterior = posterior odds of an interaction with N genomic features being true,
L(g1, g2, g3,..., gN) = likelihood ratio of an interaction with genomic features.
where P (true) = probability of an interaction being true.
where P(true|g1, g2, g3,..., gN) = probability of an interaction with N genomic features being true.
From equation (1), the likelihood ratio is,
,
where P (g1, g2, g3,..., gN |true) = probability of a true interaction having N genomic features.
If the N genomic features, g1, g2, g3,......, gN, are conditionally independent, then the resulting Bayesian network is called a naïve Bayesian network and its likelihood ratio can be given as the product of the likelihood ratios for each feature:
where T = all true interactions (gold standard positives),
F = all false interactions (gold standard negatives),
TPi = number of true interactions in the high-throughput data set with the ith feature
FPi = number of false interactions in the high-throughput data set with the ith feature
For any organism, L(g1, g2, g3,..., gN) > 1, results in Oposterior > Oprior. This is because, in equation (1), Oprior is a constant and depends on the number of interactions in any organism. Hence, Oposterior is directly proportional to L(g1, g2, g3,..., gN). Thus, the posterior odds of an interaction being true, if it has one or more genomic features, increases as L(g1, g2, g3,..., gN) increases i.e. larger the L(g1, g2, g3,..., gN), the higher are the odds of an interaction being true.
ROC curve analysis
A Receiver Operating Characteristic (ROC) curve is a graphical representation of the accuracy of a test and expresses the trade-off between the sensitivity and the specificity of the test [40]. Sensitivity of a test is defined as the ability to identify a true positive in a data set. Specificity is defined as the ability to identify a true negative in a data set.
where TP = number of true positives,
TN = number of true negatives,
FP = number of false positives,
T = total number of positives,
F = total number of negatives.
The ROC curve is plotted with the Sensitivity on the Y-axis and (1-Specificity) on the X-axis. The smooth ROC curve is plotted using JROCFIT [41].
Cross-validation
Since the training set (data set used to calculate the likelihood ratios) and the test set (data set used to calculate the sensitivity and specificity) are the same yeast high-throughput data set, we used 10-fold cross-validation to assess our predictions. We divided the positive and negative gold standards into 10 approximately equal sets. We used 9 of these to calculate likelihood ratios for each genomic feature. Then we identified the true positives and false positives in the remaining set using these likelihood ratios. We did this in turn, so that each of the 10 sets was a test set and the remaining 9 sets were training sets. We then summed the number of true positives and false positives across all the 10 test sets to obtain the Sensitivity and Specificity and plotted the ROC curve.
Authors' contributions
HN and AP conceived of the study. AP performed the data collection, data analysis, web site preparation and drafted the manuscript under the guidance and supervision of HN. All authors read and approved the final manuscript.
Acknowledgements
We would like to thank Dr. Daron Standley in PDBj, for useful discussions. This study has been supported by grant-in-aid from Institute for Bioinformatics Research and Development, Japan Science and Technology Agency and by grant-in-aid for Scientific Research on priority areas No. 12144206 from the Ministry of Education, Science, Sports and Culture of Japan.
Figures and Tables
Figure 1 Likelihood ratios for genomic features.
Figure 2 ROC curve for the combination of genomic features using 10-fold cross validations. The dotted line shows the empirical ROC curve, while the solid line shows the fitted ROC curve (obtained using JROCFIT). Each point on the ROC curve corresponds to sensitivity and specificity for one or a combination of more than one genomic features. d: interacting Pfam domains; g: similar GO annotations; h: homologous interactions; none: no genomic features. More than one genomic features are indicated by listing the features separated by a '+' sign.
Figure 3 Percentage of interactions predicted true across different high-throughput data sets.
Figure 4 Percentage of interactions predicted true in high and low confidence interactions across different high-throughput data sets.
Figure 5 Some low confidence interactions predicted to be true by our method and confirmed by other publications. The Likelihood ratio for each interaction is indicated. Interactions with a Likelihood ratio greater than 100 are shown with a solid line, while those with a Likelihood ratio less than 10 are shown with a dashed line. (A) Interactions between proteins co-regulating the alternative splicing of Dscam exon 4 in D. menalogaster. (B) Interactions between proteins in the Lsm1-7 complex in S. cerevisiae confirmed by similar interactions found in H. sapiens.
Table 1 Yeast high-throughput data sets
Data set Interactions Type
Uetz et al. [3] 1438 Y2H
Ito et al. [4] 4449 Y2H
Gavin et al. [5] 3757 Co-IP (spoke model)
Ho et al. [6] 3618 Co-IP (spoke model)
Total unique interactions 12674 Binary
Y2H: Yeast two-hybrid; Co-IP: Mass Spectrometry of coimmunoprecipitated complexes, converted to binary interactions using the spoke model.
Table 2 Sources of Gold Standard Positive yeast protein interaction data
Data set Interactions Type
MIPS interactions 574 Y2H
MIPS complexes 490 Co-IP (matrix model)
Small scale interactions from DIP and IntAct 110 Y2H
More than one high-throughput data sets 305 Y2H ([3, 4]) Co-IP (spoke model) [5, 6]
Total 1479 Binary
Y2H: Yeast two-hybrid; Co-IP: Mass Spectrometry of coimmunoprecipitated complexes, expanded by spoke or matrix model as indicated.
Table 3 Correlation coefficients of the genomic features for 100 random interactions
Genomic Features r t(98) p-value
Homologous Interactions – Similar GO annotations -0.12605 -1.2579 0.2401
Homologous Interactions – Interacting Pfam Domains 0.022501 0.222802 0.8826
Similar GO annotations – Interacting Pfam Domains -0.01817 -0.17988 0.2868
r: Pearson's correlation coefficient; t(98): t-test with 98 degrees of freedom; p-value: probability. Since the p-value for all t-tests is greater than the significance level of 0.05, the null hypothesis, that the genomic features are not correlated, is accepted.
Table 4 Likelihood ratio, sensitivity and specificity for the combination of different genomic features
Genomic Feature(s) Likelihood ratio (L) Sensitivity (%) Specificity (%)
d + g + h 170.052 12.3 99.4
d + g 66.031 14.5 99.3
d + h 50.463 14.7 99.2
d 19.595 14.8 99.2
g + h 8.678 44.1 94.0
g 3.370 86.7 74.3
h 2.575 89.7 62.9
none 0.163 100 0
d: interacting Pfam domains; g: similar GO annotations; h: homologous interactions. More than one genomic features are indicated by listing the features separated by a '+' sign.
Table 5 Number of interactions in different ranges of likelihood ratios for high-throughput data sets of various species
Likelihood ratio (L) H. sapiens D. melanogaster C. elegans S. cerevisiae
0 – 1 541 16655 2925 5534
1 – 10 733 3119 852 5810
10 – 100 362 367 139 824
100 – 1000 50 260 99 506
Total 1686 20401 4015 12674
All interactions with a Likelihood ratio > 1 are predicted as true.
==== Refs
von Mering C Krause R Snel B Cornell M Oliver SG Fields S Bork P Comparative assessment of large-scale data sets of protein-protein interactions Nature 2002 417 399 403 12000970 10.1038/nature750
Rain JC Selig L De Reuse H Battaglia V Reverdy C Simon S Lenzen G Petel F Wojcik J Schachter V Chemama Y Labigne A Legrain P The protein-protein interaction map of Helicobacter pylori Nature 2001 409 211 215 11196647 10.1038/35051615
Uetz P Giot L Cagney G Mansfield TA Judson RS Knight JR Lockshon D Narayan V Srinivasan M Pochart P Qureshi-Emili A Li Y Godwin B Conover D Kalbfleisch T Vijayadamodar G Yang M Johnston M Fields S Rothberg JM A comprehensive analysis of protein-protein interactions in Saccharomyces cerevisiae Nature 2000 403 623 627 10688190 10.1038/35001009
Ito T Chiba T Ozawa R Yoshida M Hattori M Sakaki Y A comprehensive two-hybrid analysis to explore the yeast protein interactome PNAS 2001 98 4569 4574 11283351 10.1073/pnas.061034498
Gavin AC Bosche M Krause R Grandi P Marzioch M Bauer A Schultz J Rick JM Michon AM Cruciat CM Remor M Hofert C Schelder M Brajenovic M Ruffner H Merino A Klein K Hudak M Dickson D Rudi T Gnau V Bauch A Bastuck S Huhse B Leutwein C Heurtier MA Copley RR Edelmann A Querfurth E Rybin V Drewes G Raida M Bouwmeester T Bork P Seraphin B Kuster B Neubauer G Superti-Furga G Functional organization of the yeast proteome by systematic analysis of protein complexes Nature 2002 415 141 147 11805826 10.1038/415141a
Ho Y Gruhler A Heilbut A Bader GD Moore L Adams SL Millar A Taylor P Bennett K Boutilier K Yang L Wolting C Donaldson I Schandorff S Shewnarane J Vo M Taggart J Goudreault M Muskat B Alfarano C Dewar D Lin Z Michalickova K Willems AR Sassi H Nielsen PA Rasmussen KJ Andersen JR Johansen LE Hansen LH Jespersen H Podtelejnikov A Nielsen E Crawford J Poulsen V Sorensen BD Matthiesen J Hendrickson RC Gleeson F Pawson T Moran MF Durocher D Mann M Hogue CWV Figeys D Tyers M Systematic identification of protein complexes in Saccharomyces cerevisiae by mass spectrometry Nature 2002 415 180 183 11805837 10.1038/415180a
Li S Armstrong CM Bertin N Ge H Milstein S Boxem M Vidalain PO Han JDJ Chesneau A Hao T Goldberg DS Li N Martinez M Rual JF Lamesch P Xu L Tewari M Wong SL Zhang LV Berriz GF Jacotot L Vaglio P Reboul J Hirozane-Kishikawa T Li Q Gabel HW Elewa A Baumgartner B Rose DJ Yu H Bosak S Sequerra R Fraser A Mango SE Saxton WM Strome S van den Heuvel S Piano F Vandenhaute J Sardet C Gerstein M Doucette-Stamm L Gunsalus KC Harper JW Cusick ME Roth FP Hill DE Vidal M A Map of the Interactome Network of the Metazoan C. elegans Science 2004 303 540 543 14704431 10.1126/science.1091403
Giot L Bader JS Brouwer C Chaudhuri A Kuang B Li Y Hao YL Ooi CE Godwin B Vitols E Vijayadamodar G Pochart P Machineni H Welsh M Kong Y Zerhusen B Malcolm R Varrone Z Collis A Minto M Burgess S McDaniel L Stimpson E Spriggs F Williams J Neurath K Ioime N Agee M Voss E Furtak K Renzulli R Aanensen N Carrolla S Bickelhaupt E Lazovatsky Y DaSilva A Zhong J Stanyon CA Finley RLJ White KP Braverman M Jarvie T Gold S Leach M Knight J Shimkets RA McKenna MP Chant J Rothberg JM A Protein Interaction Map of Drosophila melanogaster Science 2003 302 1727 1736 14605208 10.1126/science.1090289
Bouwmeester T Bauch A Ruffner H Angrand PO Bergamini G Croughton K Cruciat C Eberhard D Gagneur J Ghidelli S Hopf C Huhse B Mangano R Michon AM Schirle M Schlegl J Schwab M Stein MA Bauer A Casari G Drewes G Gavin AC Jackson DB Joberty G Neubauer G Rick J Kuster B Superti-Furga G A physical and functional map of the human TNF-[alpha]/NF-[kappa]B signal transduction pathway Nat Cell Biol 2004 6 97 105 14743216 10.1038/ncb1086
Bork P Jensen LJ von Mering C Ramani AK Lee I Marcotte EM Protein interaction networks from yeast to human Current Opinion in Structural Biology 2004 14 292 299 15193308 10.1016/j.sbi.2004.05.003
Sprinzak E Sattath S Margalit H How Reliable are Experimental Protein-Protein Interaction Data? Journal of Molecular Biology 2003 327 919 923 12662919 10.1016/S0022-2836(03)00239-0
Lehner B Fraser A A first-draft human protein-interaction map Genome Biology 2004 5 R63 15345047 10.1186/gb-2004-5-9-r63
Yu H Luscombe NM Lu HX Zhu X Xia Y Han JDJ Bertin N Chung S Vidal M Gerstein M Annotation Transfer Between Genomes: Protein-Protein Interologs and Protein-DNA Regulogs Genome Res 2004 14 1107 1118 15173116 10.1101/gr.1774904
Deane CM Salwinski L Xenarios I Eisenberg D Protein interactions: two methods for assessment of the reliability of high throughput observations Mol Cell Proteomics 2002 1 349 356 12118076 10.1074/mcp.M100037-MCP200
Edwards AM Kus B Jansen R Greenbaum D Greenblatt J Gerstein M Bridging structural biology and genomics: assessing protein interaction data with known complexes Trends Genet 2002 18 529 536 12350343 10.1016/S0168-9525(02)02763-4
Berman HM Westbrook J Feng Z Gilliland G Bhat TN Weissig H Shindyalov IN Bourne PE The Protein Data Bank Nucleic Acids Res 2000 28 235 242 10592235 10.1093/nar/28.1.235
Saito R Suzuki H Hayashizaki Y Construction of reliable protein-protein interaction networks with a new interaction generality measure Bioinformatics 2003 19 756 763 12691988 10.1093/bioinformatics/btg070
Bader JS Chaudhuri A Rothberg JM Chant J Gaining confidence in high-throughput protein interaction networks Nat Biotech 2004 22 78 85 14704708 10.1038/nbt924
Jansen R Yu H Greenbaum D Kluger Y Krogan NJ Chung S Emili A Snyder M Greenblatt JF Gerstein M A Bayesian Networks Approach for Predicting Protein-Protein Interactions from Genomic Data Science 2003 302 449 453 14564010 10.1126/science.1087361
Patil A Nakamura H HINT - Homologous Interactions Database
Patil A Nakamura H HINT - a database of annotated protein-protein interactions and their homologs BIOPHYSICS 2005 1 21 24 10.2142/biophysics.1.21
Bateman A Coin L Durbin R Finn RD Hollich V Griffiths-Jones S Khanna A Marshall M Moxon S Sonnhammer ELL Studholme DJ Yeats C Eddy SR The Pfam protein families database Nucl Acids Res 2004 32 D138 141 14681378 10.1093/nar/gkh121
Stein A Russell RB Aloy P 3did: interacting protein domains of known three-dimensional structure Nucleic Acids Res 2005 33 D413 417 15608228 10.1093/nar/gki037
Gene Ontology Consortium The Gene Ontology (GO) database and informatics resource Nucl Acids Res 2004 32 D258 261 14681407 10.1093/nar/gkh036
Pearl J Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference 1997 San Fransisco , Morgan Kaufmann Publishers, Inc.
Asthana S King OD Gibbons FD Roth FP Predicting Protein Complex Membership Using Probabilistic Network Reliability Genome Res 2004 14 1170 1175 15140827 10.1101/gr.2203804
Salwinski L Miller CS Smith AJ Pettit FK Bowie JU Eisenberg D The Database of Interacting Proteins: 2004 update Nucl Acids Res 2004 32 D449 451 14681454 10.1093/nar/gkh086
Hermjakob H Montecchi-Palazzi L Lewington C Mudali S Kerrien S Orchard S Vingron M Roechert B Roepstorff P Valencia A Margalit H Armstrong J Bairoch A Cesareni G Sherman D Apweiler R IntAct: an open source molecular interaction database Nucl Acids Res 2004 32 D452 455 14681455 10.1093/nar/gkh052
Patil A Nakamura H HitPredict
Park JW Parisky K Celotto AM Reenan RA Graveley BR Identification of alternative splicing regulators by RNA interference in Drosophila Proc Natl Acad Sci U S A 2004 101 15974 15979 15492211 10.1073/pnas.0407004101
Lehner B Sanderson CM A Protein Interaction Framework for Human mRNA Degradation Genome Res 2004 14 1315 1323 15231747 10.1101/gr.2122004
Lin N Wu B Jansen R Gerstein M Zhao H Information assessment on predicting protein-protein interactions BMC Bioinformatics 2004 5 154 15491499 10.1186/1471-2105-5-154
Hoffmann R Valencia A A gene network for navigating the literature Nat Genet 2004 36 664 15226743 10.1038/ng0704-664
Peri S Navarro JD Amanchy R Kristiansen TZ Jonnalagadda CK Surendranath V Niranjan V Muthusamy B Gandhi TKB Gronborg M Ibarrola N Deshpande N Shanker K Shivashankar HN Rashmi BP Ramya MA Zhao Z Chandrika KN Padma N Harsha HC Yatish AJ Kavitha MP Menezes M Choudhury DR Suresh S Ghosh N Saravana R Chandran S Krishna S Joy M Anand SK Madavan V Joseph A Wong GW Schiemann WP Constantinescu SN Huang L Khosravi-Far R Steen H Tewari M Ghaffari S Blobe GC Dang CV Garcia JGN Pevsner J Jensen ON Roepstorff P Deshpande KS Chinnaiyan AM Hamosh A Chakravarti A Pandey A Development of Human Protein Reference Database as an Initial Platform for Approaching Systems Biology in Humans Genome Res 2003 13 2363 2371 14525934 10.1101/gr.1680803
Bader GD Hogue CW Analyzing yeast protein-protein interaction data obtained from different sources Nat Biotechnol 2002 20 991 997 12355115 10.1038/nbt1002-991
Mewes HW Frishman D Guldener U Mannhaupt G Mayer K Mokrejs M Morgenstern B Munsterkotter M Rudd S Weil B MIPS: a database for genomes and protein sequences Nucleic Acids Res 2002 30 31 34 11752246 10.1093/nar/30.1.31
Huh WK Falvo JV Gerke LC Carroll AS Howson RW Weissman JS O'Shea EK Global analysis of protein localization in budding yeast Nature 2003 425 686 691 14562095 10.1038/nature02026
Jansen R Gerstein M Analyzing protein function on a genomic scale: the importance of gold-standard positives and negatives for network prediction Current Opinion in Microbiology 2004 7 535 545 15451510 10.1016/j.mib.2004.08.012
Schwikowski B Uetz P Fields S A network of protein-protein interactions in yeast Nat Biotechnol 2000 18 1257 1261 11101803 10.1038/82360
Zhou XH Obuchowski NA McClish DK Statistical Methods in Diagnostic Medicine 2002 New York , John Wiley & Sons 437
Eng J ROC analysis: web-based calculator for ROC curves
| 15833142 | PMC1127019 | CC BY | 2021-01-04 16:02:48 | no | BMC Bioinformatics. 2005 Apr 18; 6:100 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-100 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1061585047910.1186/1471-2105-6-106Methodology ArticleQuadratic regression analysis for gene discovery and pattern recognition for non-cyclic short time-course microarray experiments Liu Hua [email protected] Sergey [email protected] Aaron S [email protected] Thomas V [email protected] Marilyn L [email protected] Arnold J [email protected] Department of Statistics, University of Kentucky, Lexington, KY 40506, USA2 Department of Physiology, University of Kentucky, Lexington, KY 40536, USA3 Department of Anatomy and Neurobiology, University of Kentucky, Lexington, KY 40536, USA4 Sanders-Brown Center on Aging, University of Kentucky, Lexington, KY 40536, USA2005 25 4 2005 6 106 106 25 8 2004 25 4 2005 Copyright © 2005 Liu 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
Cluster analyses are used to analyze microarray time-course data for gene discovery and pattern recognition. However, in general, these methods do not take advantage of the fact that time is a continuous variable, and existing clustering methods often group biologically unrelated genes together.
Results
We propose a quadratic regression method for identification of differentially expressed genes and classification of genes based on their temporal expression profiles for non-cyclic short time-course microarray data. This method treats time as a continuous variable, therefore preserves actual time information. We applied this method to a microarray time-course study of gene expression at short time intervals following deafferentation of olfactory receptor neurons. Nine regression patterns have been identified and shown to fit gene expression profiles better than k-means clusters. EASE analysis identified over-represented functional groups in each regression pattern and each k-means cluster, which further demonstrated that the regression method provided more biologically meaningful classifications of gene expression profiles than the k-means clustering method. Comparison with Peddada et al.'s order-restricted inference method showed that our method provides a different perspective on the temporal gene profiles. Reliability study indicates that regression patterns have the highest reliabilities.
Conclusion
Our results demonstrate that the proposed quadratic regression method improves gene discovery and pattern recognition for non-cyclic short time-course microarray data. With a freely accessible Excel macro, investigators can readily apply this method to their microarray data.
==== Body
Background
Microarray time-course experiments allow researchers to explore the temporal expression profiles for thousands of genes simultaneously. The premise for pattern analysis is that genes sharing similar expression profiles might be functionally related or co-regulated [1]. Due to the large number of genes involved and the complexity of gene regulatory networks, clustering analyses are popular for analyzing microarray time-course data. Heuristic-based cluster analyses group genes based on distance measures; the most commonly used methods include hierarchical clustering [2], k-means clustering [3], self-organizing maps [4], and support vector machines [5]. Due to the lack of statistical properties of these heuristic-based clustering methods, statistical models, especially analysis of variance (ANOVA) models and mixed models are often implemented as a precursor to clustering to ensure the genes used for clustering are statistically meaningful [6,7]. Only genes identified to be significantly regulated by statistical models are used for further clustering. Fitting statistical models prior to clustering usually dramatically reduces the number of genes used for clustering, which in general will improve the performance of the clustering method. An alternative way of clustering is statistical model-based clustering methods, which assume that the data is from a mixture of probability distributions such as multivariate normal distributions and describe each cluster using a probabilistic model [8,9].
In microarray time-course studies, time dependency of gene expression levels is usually of primary interest. Since time can affect the gene expression levels, it is important to preserve time information in time-course data analysis. However, most methods for analyzing microarray time-course data treat time as a nominal variable rather than a continuous variable, and thus ignore the actual times at which these points were sampled. Peddada et al. (2003) proposed a method for gene selection and clustering using order-restricted inference, which preserves the ordering of time but treats time as nominal [1]. Recently, a number of algorithms treating time as a continuous variable have been introduced. Xu et al. (2002) applied a piecewise regression model to identify differentially expressed genes [10]. Both Luan and Li (2003) and Bar-Joseph et al (2003) proposed B-splines based approaches [11,12], which are appropriate for microarray data with relatively long time-course, but their application to short time-course data is questionable. New methods for analyzing short time-course microarray data are needed [13].
In this paper, we propose a model-based approach, step down quadratic regression, for gene identification and pattern recognition in non-cyclic short time-course microarray data. This approach takes into account time information because time is treated as a continuous variable. It is performed by initially fitting a quadratic regression model to each gene; a linear regression model will be fit to the gene if the quadratic term is determined to have no statistically significant relationship with time. Significance of gene differential expression and classification of gene expression patterns can be determined based on relevant F-statistics and least squares estimates. Major advantages of our approach are that it not only preserves the ordering of time but also utilizes the actual times at which they were sampled; it identifies differentially expressed genes and classifies these genes based on their temporal expression profiles; and the temporal expression patterns discovered are readily understandable and biologically meaningful. A free Excel macro for applying this method is available at [14]. The proposed quadratic regression method is applied to a microarray time-course study of olfactory receptor neurons [15]. Biologically meaningful temporal expression patterns have been obtained and shown to be more effective classifications than ANOVA-protected k-means clusters. Comparison with Peddada et al.'s order-restricted inference method [1] showed that our method provides a different and interesting insight into the temporal gene profiles. Reliabilities of the results from all 3 methods were assessed using a bootstrap method [16] and regression patterns were shown to have the highest reliabilities.
Results
Step-down quadratic regression
We propose a step-down quadratic regression method for gene discovery and pattern recognition for non-cyclic short time-course microarray experiment. The first step is to fit the following quadratic regression model to the jth gene:
yij = β0j + β1jx + β2jx2 + εij (1)
where yij denotes the expression of the jth gene at the ith replication, x denotes time, β0j is the mean expression of the jth gene at x = 0, β1j is the linear effect parameter of the jth gene, β2j is the quadratic effect parameter of the jth gene, and, εij is the random error associated with the expression of the jth gene at the ith replication and is assumed to be independently distributed normal with mean 0 and variance . Two levels of significance, α0 and α1, need to be pre-specified, where α0 to is recommended to be small to reduce the false positive rate in the gene discovery and α1 less stringent to control pattern classification. α0 could be chosen using various multiple testing p-value adjustment procedures, for example, False Discovery Rate (FDR) [17]. The temporal gene expression patterns can be determined as follows:
1. If overall model (1) p-value >α0, the jth gene is considered to have no significant differential expression over time. The expression pattern of the gene is "flat".
2. If overall model (1) p-value ≤ α0, the jth gene will be considered to have significant differential expression over time. The patterns are then determined based on the p-values obtained from F tests (Table 1).
a. If both p-value of quadratic effect ≤ α1 and p-value of linear effect ≤ α1, the jth gene is considered to be significant in both the quadratic and linear terms. The expression pattern of the gene is "quadratic-linear".
b. If p-value of quadratic effect ≤ α1 and p-value of linear effect >α1, the jth gene is considered to be significant only in the quadratic term. The expression pattern of the gene is "quadratic".
c. If p-value of quadratic effect >α1, the jth gene is considered to be non-significant in the quadratic term. The quadratic term will be dropped and a linear regression model will be fitted to the gene:
yij = β0j + β1jx + εij (2)
From fitting model (2),
• If p-value of linear effect ≤ α1, the jth gene is considered to be significant in the linear term. The expression pattern of the gene is "linear".
• If p-value of linear effect >α1, the jth gene is considered to be non-significant in the linear term. The expression pattern of the gene is "flat".
The four expression patterns described above can be furthered classified into 9 patterns according to the up/down regulation of the gene expression based on the least-squares estimates and and the predicted signals (Table 2). A flow chart for the above procedure is shown in Figure 1. This procedure can be easily applied using the Excel macro available at [14].
Application of the quadratic regression method
Normality test based on Shapiro-Wilk statistics [18] suggested that most of the 3834 present genes in the olfactory receptor neuron data do not have a significant departure from the normal distribution (Figure 2). Therefore the quadratic regression method with normality assumption was applied to the data of 3834 present genes (Figure 3), where α0 was chosen to be 0.01 and α1 to be 0.05. 798 genes were determined to have significant differential expression over time at level 0.01. Examples of 9 regression patterns are shown in Figure 4.
Comparison with Peddada et al.'s method
Peddada et al.'s method [1] was applied to the expression data of 3834 present genes with 8 pre-specified profiles: monotone increasing (MI); monotone decreasing (MD); 3 up-down profiles with maximum at the second, third, forth time point (UD2, UD3, UD4); and 3 down-up profiles with maximum at the second, third, forth time point (DU2, DU3, DU4). Based on 4000 bootstrap, 379 genes were classified into one of the 8 pre-specified profiles at significance level 0.01. This indicates that Peddada et al.'s method might be relatively more conservative than regression method by selecting much fewer genes at significance level 0.01. Comparisons of Peddada et al.'s profiles and regression patterns are listed in Table 3. We observe that the majority of genes in MI are in LU, similarly for MD and LD, UD2 and QLCD, and DU2 and QLVU. However, each of the Peddada et al.'s profiles contains a mixture of regression patterns, and vice versa. This is reasonable because even though both methods perform gene selection and classification, they are aimed at different aspects of the temporal profiles. For example, Peddada et al.'s MI profile contains regression patterns LU, QLCU and QLVU. Although the gene expression level is increasing monotonically over time, the regression method gives more information on how it is increased: constantly (LU, Figure 5a, Gdp2), increases faster then slower (QLCU, Figure 5b, Ccl2), or increases slower then faster (QLVU, Figure 5c, Prom1). Peddada et al.'s UD2 profile contains genes that are first up-regulated then down-regulated with maximum at the second time points, which could be classified as regression pattern QLCD in general (Figure 5d, Oazin), but it could also be classified as LD if the expression levels of all time points are close to a line (Figure 5e, Grik5); or classified as QC if the expression profile is close to quadratic (Figure 5f, Ubl1); or classified as QLCU if the expression levels of last 4 time points are much closer than those of the first time point. Similarly, Peddada et al.'s UD3 profile could be classified as regression patterns QC, QLCU, and QLCD (Figure 5g, Bub3; 5h, Fut9; 5i, Phgdh).
Comparison with ANOVA-protected k-means clustering
ANOVA-protected k-means clustering was applied to the expression signals of 3834 present genes. Out of 3834 present genes, 770 were identified to be differentially expressed over time by one way ANOVA (overall model p-value ≤ 0.01). These 770 genes were used for classification by k-means clustering with k = 9 and the distance measure being Pearson correlation coefficient (Table 4).
In order to make the regression patterns comparable with the k-means clusters, the quadratic regression method was applied to the 770 ANOVA significant genes. Table 4 shows the number of genes in common when comparing each regression pattern with each k-means cluster. An example of a good match between regression patterns and k-means clusters is the QLCD regression pattern and k-means cluster K1. However, in most cases, k-means clusters contain a mixture of regression patterns and the regression patterns are separated into different k-means clusters. For example, genes that have the LU regression pattern are split into 4 k-means clusters (Figure 6a, Bzrp; 6b, Aqp1; 6c, Prg; 6d, Hnrpl). The similarity of the temporal expression profiles in Figure 6 indicates that it might be more appropriate to classify these genes into the same group, which occurs using the proposed regression method. Examples in Figure 7 show that some k-means clusters are also mixtures of expression profiles in terms of the mean signals (green lines). For example, a down-up-down-up pattern (down-regulated at the second time point, up-regulated at the third time point, etc, in terms of mean signals) appeared in both k-means clusters K5 and K6 (green lines), but are identified to have QLVU regression pattern (Figure 7c, Clu; and 7d, D17H6S56E-5); similarly see Figure 7a and 7b (a, Sfpi1; and b, Anxa2). Once again, the regression method provides better classification. Figure 8 is an example of genes with similar expression patterns but different initial starting time of the differential expression (Figure 8a, Psmb6; 8b, Adora2b). Adora2b clearly starts differential expression later than Psmb6 (see the blue dots in Figure 8). After the initial starting point (first time point for Psmb6 and second time point for Adora2b), these two genes show similar upward regulation. These two genes were classified into the same regression group, but in different k-means clusters. Based on the above analysis, our regression method is demonstrated to be more appropriate for the classification of temporal gene expression profiles than k-means method.
EASE functional analysis on regression patterns and k-means clusters
To further explore the effectiveness of the regression method on gene classification, EASE (Expression Analysis Systematic Explorer) software was used to examine the potential relationship between the biological functions of the genes and their expression patterns [19]. EASE calculates EASE scores (Jackknife one-sided Fisher exact p-values) to identify over-represented gene categories within lists of genes. EASE analysis was applied to each of the 9 regression patterns and 9 k-means clusters that were obtained from the classification of 770 ANOVA significant genes (Table 4). The results are summarized (see Additional file 1), with part of the information shown in Tables 5 and 6. The EASE analysis demonstrates that the proposed regression method is more effective for gene classification than the k-means clustering method. Almost all of the regression patterns contain genes mainly from one biological process. For example, LU has 9 over-represented gene categories, 8 of which are involved in immune regulation (Table 5). The majority of the LU and QLVU gene categories are in the immune regulation category. This suggests that there exist multiple regulatory mechanisms within the immune regulation. The immune regulation in QLVU appears to be a more complex regulatory mechanism for the initial up-regulation of these genes due to the slow upward regulation at early time points of this regression pattern (Figure 5c). The EASE results for the k-means clusters shows that the over-represented gene categories of most k-means clusters are involved in more than one biological process, for example, k-means cluster K5 contains 9 over-represented gene categories, 3 involved in immune regulation, 2 involved in cell death, etc. Notice that the immune regulation category is represented in 4 k-means clusters, which suggests that the immune regulation category is more consolidated in regression patterns than in k-means clusters (Table 6). Also, by comparing EASE scores in Tables 5 and 6, one can see that the over-represented gene categories in the regression patterns have, in general, smaller EASE scores than those in the k-means clusters, which further indicates the greater effectiveness of the regression method in pattern classification.
Reliability analysis
Kerr and Churchill (2001) introduced a bootstrap technique to assess the stability of clustering results [16]. We applied the same idea here to assess the reliability of regression patterns, Peddada et al.'s profiles, and k-means clusters. All 3 pattern classification methods were performed on the expression data of 770 ANOVA significant genes to make the results comparable. The reliability curves show that regression patterns have the highest reliability, and k-means clusters have the lowest reliability (Figure 9). This suggests that the regression method provides relatively more stable pattern classifications.
Simulation study
We investigated the false positive rate (gene specific) of our method via a simulation study. The data were generated randomly from N(0,1), containing expression signals of 10000 "null" genes (no gene differentially expressed over time), with 5 time points and 3 replications per time point per gene. 50 of such data were generated. The regression approach was applied to each gene in each simulated data at α0 = 0.01 and the numbers of significant genes in each of the 50 data were obtained. The average proportion of significance (average false positive rate) is 1.01% with standard deviation 0.01%. This demonstrates that the false positive rate of the regression method is accurate because 1% of 10000 genes would be expected to be significant at 0.01 level by chance. The false positive rates of the regression patterns LU, LD, QC, QV are all approximately equal to 1/6 of the average false positives, and those of QLCU, QLCD, QLVU, and QLVD are all approximately equal to 1/12 of the average false positives.
Discussion
The proposed step-down quadratic regression method is an effective statistical approach for gene discovery and pattern recognition. It utilizes the actual time information, and provides biologically meaningful classification of temporal gene expression profiles. Furthermore, it does not require replication at each time point, which ANOVA-type methods do require. Also, this method can identify genes with subtle changes over time and therefore discover genes that might be undetectable by other methods, eg, ANOVA-type methods. However, there are several limitations to this method. Firstly, it is designed to fit time-course data with a small number of time points. We recommend this method when there are 4 to 10 time points in the data. For an experiment with more time points, spline-type methods [11,12] could be a possible choice; for an experiment with 2 or 3 time points, ANOVA-type method is recommended. Secondly, the 9 regression patterns are rather limited considering the complexity of gene regulatory networks. For example, certain proportion of genes show cubic, "M", and "W" shaped patterns in 211 regression FLAT genes which are ANOVA significant (Table 4). These patterns could be caused by random chance, but they could also be real patterns. Fitting a higher order polynomial regression model may discover these types of genes profiles. Theoretically, one could fit a 4th-order polynomial regression model to this data (the highest order of the polynomial one can fit is the number of time points minus one). The model with 4th-order polynomial will work similarly to connecting the mean at each time point, therefore will provide a good fit to the data with smallest R2 and minimum Mean Squared Error, compared with lower-order polynomials. However, the purpose of pattern analysis is to cluster the data instead of fitting models, so the quadratic fit is useful even though the goodness of fit may not be great. Also, the use of high-order polynomials (higher than the second-order) should be avoided if possible [20], particularly in cases such as this where the regression coefficients are used primarily for classification. Another issue is the transformation of the experimental time. Transformation should be considered when the sampling time is unequally spaced. The choice for the type of transformation (log-transformation, square-root transformation, etc) is not critical because the resulting pattern classification will in general not be impacted.
In the reliability curves, at 95% reliability, regression patterns, Peddada et al.'s profiles, and k-means clusters have 33%, 12%, and 0% of genes, respectively; and at 80% reliability, the percentage of genes are 55%, 32%, and 0%, respectively (Figure 9). Even though the regression patterns have the highest reliability, only 33% of genes have 95% reliabilities. We examined the overall model (1) p-values of 770 genes by the regression method and found that genes that have the smallest overall model (1) p-values all have 95% reliabilities. This suggests that we could reduce the level of significance α0 to increase the stability of regression patterns. α0 could be reduced using various multiple testing p-value adjustment procedures, for example, Westfall and Young's step down method [21], and False Discovery Rate (FDR) [17]. Application of the FDR method can be done as follows (assuming FDR is controlled at level of α): let p(1) <p(2) < … <p(m) be the ordered overall model (1) p-values, start from the largest p-value p(m), compare each p(i) with α *i/m; let k be the largest i that p(k) ≤ α *k/m, conclude p(1), …, p(k) to be significant.
Both our quadratic regression method and Peddada et al.'s method serve the same overall goal: gene selection and classification. Peddada et al.'s method provides more choices of temporal profiles than our method. While our regression method offers less choice of patterns, it may provide deeper insight into the gene expression profiles than Peddada et al.'s method. Our method distinguishes patterns with different rates of change and provides more information on the relative relationship among the expression levels of all time points. For example, specifying a profile of up-down with maximum at one time point does not provide much information on the relative relationships among other time points (Figure 5). A further refinement of Peddada et al.'s method may provide such information about the relationship of other time points besides the maximum/minimum. However, it is less likely to separate the patterns in Figure 5a, b, and 5c by their method. Another fact is that Peddada et al.'s method provides exactly the location of the maximum/minimum, whereas our method provides the neighborhood of the location of the maximum/minimum. Furthermore, their method is based on bootstrap, which is computationally intensive. The result of their method, for example, the reliability curves, might be improved by applying more bootstrap, which is 4000 in this paper due to the computational difficulties and time constraints. Moreover, their method depends on the ordering of time but not the actual time at which the samples were taken, whereas the regression method accounts for both.
K-means is an iterative clustering algorithm [22]. The first step of this method is to randomly assign the data points to the k clusters. Next, the distance to the center of each cluster is calculated for each data point, and the data point is moved into the closest cluster. This step will be repeated until no data point is moving from one cluster to another. In k-means, the number of clusters, k, needs to be pre-specified. Researchers usually choose several different k and find the one which has the most biologically meaningful clusters. There are methods of finding the "optimal" k, for example, Bayesian Information Criterion [23]. In this paper, k was arbitrarily chosen to be 9. Since the k-means clustering does not perform well (Table 4; Figures 6, 7, and 8), we investigated different choice of k based on the Bayesian Information Criterion and identified that the "optimal" k is 15. However, as we examined these 15 k-means clusters, the pattern classification does not seem to be improved, the same problem exists as with k = 9. For example, Prom1, Clu, and D17H6S56E-5 (Figure 5c, Figure 7c and 7d) all have similar temporal profiles and are all classified to be QLVU, but they were separated into 3 of the 15 k-means clusters. This could be related to the distance measure used (Pearson correlation coefficient). As we discovered, genes in the same cluster do not necessarily have higher correlation than genes in different clusters. For example, Sfpi1 and Anxa2 (Figure 7a and 7b) are highly correlated (Pearson correlation coefficient is 0.9934) and their expression patterns are similar, but they are in different k-means clusters. A possible reason might be that the time-course in olfactory receptor neuron data is too short for correlation to perform well. Even though there are a total of 15 observations for each gene, correlation calculations are based on the 5 mean signals, which could be too few to describe the relationship between temporal profiles. There is also concern about using correlation as the distance measure. A large correlation coefficient does not necessarily indicate two similarly shaped profiles, nor does a small correlation coefficient necessarily indicate differently shaped profiles [1].
A number of regression algorithms have been proposed recently, which treat time as a continuous variable. Several of them are based on cubic B-splines [11,12]. B-splines are defined as a linear combination of a set of basis polynomials. In order to fit cubic B-splines to time-course data, the entire duration of experimental time needs to be divided into several segments by "knots" (the point to separate segments), and each segment will be fit by cubic polynomial. The successful application of these methods to microarray time-course data depends heavily on having a relatively large numbers of time points. The B-spline based methods will not be effective when there are a small number of time points in the time-course experiment [13]. For a data with 5 time points, cubic B-spline type methods would not be appropriate because it is recommended that there should be at least 4 or 5 experimental time points in each segment [24]. Xu et al used a piecewise quadratic regression model to identify differentially expressed genes [10]. In their approach, expression levels at 0 hour and 2 hours after treatment are fit differently from the rest of time points after treatment. Although appropriate for their data, their method cannot be applied to the dataset used in this paper.
The quadratic regression method that we applied to the olfactory receptor neuron data relies on the normality assumption. This is supported by the result of the Shapiro-Wilk normality test, which indicates that most of the genes used for the analysis follow a normal distribution. This might be due to the fact that we removed genes that are called "A" (absent) by Affymetrix across all chips. "A" calls are often assigned to low expression signals, which tend to be non-normal in general. Therefore removing genes with a high proportion of "A" calls may reduce the possibility of violation of the normality assumption, which will then make the test based on distributional assumption more likely to be valid, and thus avoid computational intensive resampling procedures, for example, bootstrap and permutation. If desired, experimenters could also try various types of data transformation to make their data closer to normal when the data are shown to have large departure from normality. However, the log transformation performed on the olfactory receptor neuron data was not to reduce the possible non-normality, but solely to make a fair comparison of our regression method and k-means method because it is the default transformation in Genespring. When the normality assumption (εij ~ N(0,)) does not hold, the bootstrap method [25] can be used to avoid the distributional assumption. For an experiment with m genes, T time points, and r replications per time point, the bootstrap procedure can be performed in the following way: form the data into a matrix of m × rT, each column in the matrix contains expressions of m genes in one chip and each row contains rT expressions of one gene; randomly draw rT columns with replacement to form a bootstrap sample; apply step-down quadratic regression procedure to the bootstrap sample to obtain F statistics from F tests; repeat the above steps 1000 times to form a bootstrap F distribution for each gene; claim a gene to be significance at level of α if its observed F statistics is greater than the upper (α / 2)thpercentile or less than the lower (α / 2)th percentile of its bootstrap F distribution. One concern about using bootstrap here is that the bootstrap F distribution might be too discrete due to the small number of time points. However, the fact that we are bootstrapping both the explanatory and response variables mitigates this issue by using all the data points, not just the time points. Additionally, in a small simulation study, we observed that the bootstrap F distribution is rather smooth (result not shown).
Conclusion
The proposed step-down quadratic regression approach is shown to be effective for gene discovery and pattern recognition for non-cyclic short time-course microarray experiment. Major advantages of this method are that it preserves the actual time information, and provides a useful tool for gene identification and pattern recognition. The nine regression patterns, obtained when applied to the olfactory receptor neuron data, are shown to be more reasonable classifications compared to ANOVA-protected k-means clustering method. EASE analysis further showed that our regression patterns are more biologically meaningful than the k-means clusters. Comparison with Peddada et al.'s method showed that our method provides a different perspective on the temporal gene profiles. Reliability study indicates that regression patterns are most reliable. In conclusion, this method should improve gene discovery and pattern recognition for microarray time-course data. With the freely accessible Excel macro, investigators can readily apply this method to their research data.
Methods
ANOVA-protected k-means clustering
One-way ANOVA model yijk = μj + τk + εijk was fitted to each gene in SAS v9, where yijk denotes the gene expression level of the jth gene at the ith replication of the kth time point, μj denotes the overall mean signal of the jth gene, τk denotes the effect of kth time point, εijk denotes the random error associated with the ith replication at the kth time point of the jth gene and is assumed to be independently distributed normal with mean 0 and variance . Genes that have overall ANOVA model p-values ≤ α0 will be used for k-means clustering. K-means clustering was performed in Genespring V6.1 (Silicon Genetics. Redwood City, CA) with k = 9. The similarity measure was chosen to be Pearson correlation coefficient, which was calculated from vectors of length 5 containing mean signals of 3 replications at each of the 5 time points. 500 additional random clusters were tested and the best clusters were selected by the software.
EASE functional analysis
EASE software was used to identify the over-represented categories of genes [19]. Gene Ontology Biological Process was chosen as the categorization system in EASE analysis. A functional gene category with an EASE score of less than 0.05 is considered to be over-represented. The EASE software is available at: .
Data description
The data used here are from a study of olfactory receptor neurons [15]. The goal is to investigate the induction of gene regulation at short time intervals following deafferentation of olfactory receptor neurons by target ablation at 2, 8, 16, and 48 hrs compared with the sham control. Total RNA was isolated from 3 male littermate mice per time point. Following hybridization with Affymetrix GeneChips MGU74Av2, 3 chips per time point, the signals were generated by GeneChip Analysis Suite v5.0. The data was filtered before statistical tests were performed. First, 66 Affymetrix quality control probesets and 6432 expressed sequence tags were removed. Next, the absent call (A) provided by Affymetrix was considered. 2156 genes that are called "A" across all 15 chips were removed from the data. The remaining 3834 present genes were used for the regression analysis (Figure 3). The hybridization signals of these 3834 genes were log-transformed in Genespring. Because the time points in this experiment are not equally spaced, ln(t+1) transformation was performed to each of the 5 time points, where t stands for the time point.
List of abbreviations
ANOVA: analysis of variance.
LD: linear down regulated regression pattern.
LU: linear up regulated regression pattern.
QC: quadratic concave regulated regression pattern.
QV: quadratic convex regulated regression pattern.
QLCD: quadratic-linear concave down regulated regression pattern.
QLCU: quadratic-linear concave up regulated regression pattern.
QLVD: quadratic-linear convex down regulated regression pattern.
QLVU: quadratic-linear convex up regulated regression pattern.
MI: monotone increasing.
MD: monotone decreasing.
UD2/UD3/UD4: up-down with maximum at the second/third/forth time point.
DU2/DU3/DU4: down-up with maximum at the second/third/forth time point.
Authors' contributions
HL conducted the statistical analyses and drafted the manuscript. ST wrote the Excel macro. ASB and TVG conducted the EASE analysis. TVG and MLG provided the microarray data. AJS supervised the analysis. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
Over-represented gene categories in each of the regression patterns and k-means clusters from EASE functional analysis. "Ease Analysis.xls" contains 2 worksheets: worksheet "regression" contains the over-represented gene categories in each of the 9 regression patterns obtained from EASE functional analysis; worksheet "k-means" contains the over-represented gene categories in each of the 9 k-means clusters obtained from EASE functional analysis.
Click here for file
Acknowledgements
We wish to thank Christopher P. Saunders for his help on the statistical analysis, Dr. Kuey-Chu Chen for her help in the application of EASE analysis, Dr. Shyamal D. Peddada for his help in the application of his method, and referees for their thoughtful comments. This work is supported by NIH-AG-016824-23 (TVG), NIH-1P20RR16481-03 (AJS), and NSF-EPS-0132295 (AJS).
Figures and Tables
Figure 1 Flow chart of the quadratic regression method. The gene selection and pattern classification procedure of our quadratic regression method. yij is the expression level; x is time; β0j, β1j, and β2j are the parameters of intercept, linear effect, and quadratic effect, respectively; εij is the random error. Among the 9 regression patterns, FLAT stands for no statistically significant differential expression over time; LU stands for linear up; LD stands for linear down; QC stands for quadratic concave; QV stands for quadratic convex; QLCU stands for quadratic linear concave up; QLCD stands for quadratic linear concave down; QLVU stands for quadratic linear convex up; QLVD stands for quadratic linear convex down.
Figure 2 Histogram of the Shapiro-Wilk p-values for normality test. The Shapiro-Wilk statistic was applied to the olfactory receptor neuron data for normality test. The horizontal axis is the Shapiro-Wilk p-values, and the vertical axis is the corresponding percentages. This histogram indicates that most of the 3834 present genes do not have a significant departure from the normality.
Figure 3 Flow chart of the filtering steps and quadratic regression analysis on the olfactory receptor neuron data. Our regression method is applied to the olfactory receptor neuron data. At first, Affymetrix quality controls, expressed sequence tags, and genes which have "A" calls across all chips were removed from the analysis. Nine regression patterns were identified among 3834 remaining genes (colored in red). FLAT stands for no statistically significant differential expression over time detected by the regression method; LU stands for linear up regulated regression pattern; LD stands for linear down regulated regression pattern; QC stands for quadratic concave regulated regression pattern; QV stands for quadratic convex regulated regression pattern; QLCU stands for quadratic linear concave up regulated regression pattern; QLCD stands for quadratic linear concave down regulated regression pattern; QLVU stands for quadratic linear convex up regulated regression pattern; QLVD stands for quadratic linear convex down regulated regression pattern.
Figure 4 An illustration of the nine temporal expression patterns identified by the quadratic regression method. The horizontal axis is the log transformation of time. The vertical axis is the hybridization signals obtained from the microarrays. The blue dots are the hybridization signals. The red line or curve is the fitted regression pattern. FLAT stands for no statistically significant differential expression over time detected by the regression method; LU stands for linear up regulated regression pattern; LD stands for linear down regulated regression pattern; QC stands for quadratic concave regulated regression pattern; QV stands for quadratic convex regulated regression pattern; QLCU stands for quadratic linear concave up regulated regression pattern; QLCD stands for quadratic linear concave down regulated regression pattern; QLVU stands for quadratic linear convex up regulated regression pattern; QLVD stands for quadratic linear convex down regulated regression pattern. The corresponding gene symbols are: FLAT. Cldn11; LU. Gba; LD. Col6a3; QC. Rab18; QV. unknown; QLCU. Psmb6; QLCD. Hnrpa2b1; QLVU. Tyrobp; QLVD. Acvr2b.
Figure 5 Examples of genes in the comparison among regression patterns and Peddada et al.'s profiles. a. Gdp2; b. Ccl2; c. Prom1; d. Oazin; e. Grik5; f. Ubl1; g. Bub3; h. Fut9; i. Phgdh. The genes in a, b, and c all have Peddada et al.'s MI profile, but are in 3 different regression patterns LU, QLCU and QLVU, the difference among the temporal profiles of these genes is the rate of increase. The genes in d, e, and f all have Peddada et al.'s UD2 profile, but are in 3 different regression patterns QLCD, LD, and QC. The genes in g, h, and i all have Peddada et al.'s UD3 profile, but are in 3 different regression patterns QC, QLCU, and QLCD. The differences among d, e, f and among g, h, i are due to the relationship among all time points and with the maximum. The horizontal axis is the log transformation of time. The blue dots are the signals. The red line or curve is the fitted regression pattern.
Figure 6 Examples of genes with the same LU regression pattern but in different k-means clusters. a. Bzrp is an example from k-means cluster K2; b. Aqp1 is an example from k-means cluster K5; c. Prg is an example from k-means cluster K6; d. Hnrpl is an example from k-means cluster K8. These 4 genes are all identified to have the LU regression pattern, but in 4 different k-means clusters. The LU regression pattern is clearly a good fit to the temporal expression profiles of these 4 genes. The horizontal axis is the log transformation of time. The blue dots are the signals. The green line is the connection of the mean signal at each time point. The red line is the LU regression pattern.
Figure 7 Examples of genes with similar expression patterns in terms of mean signal and regression. a. Sfpi1 is an example from k-means cluster K2; b. Anxa2 is an example from k-means cluster K8; c. Clu is an example from k-means cluster K5; d. D17H6S56E-5 is an example from k-means cluster K6. a and b are examples of genes with the same up-down-up-up pattern (up-regulated at the second time point, down-regulated at the third time point, then up-regulated at the last two time points) in terms of mean transformed signals (green lines). They also have the same LU regression pattern, but are in different k-means clusters. c and d are examples of genes with the same down-up-down-up pattern in terms of mean transformed signals (green lines). They also have the same QLVU regression pattern, but are in different k-means clusters. Clearly, the regression method provides better classification of the temporal expression profiles of these genes than the k-means clustering method. The horizontal axis is the log transformation of time. The blue dots are the signals. The green line is the connection of the mean signal at each time point. The red line or curve is the fitted regression pattern.
Figure 8 Examples of genes with the same regression pattern but different onset of differential expression. a. Psmb6 is an example in k-means cluster K8; b. Adora2b is an example in k-means cluster K5. Adora2b clearly starts differential expression later than Psmb6. After the onset point (first time point for Psmb6 and second time point for Adora2b), these two genes show similar upward regulation. The regression method classifies these two genes into the same group (QLCU regression pattern), but k-means clustering method does not. The horizontal axis is the log transformation of time. The blue dots are the signals. The green line is the connection of the mean signal at each time point. The red curve is the QLCU regression pattern.
Figure 9 Reliability curves of regression patterns, Peddada et al.'s profiles, and k-means clusters. The horizontal axis is the reliability (percentage of agreement of the bootstrap results with the original result), and the vertical axis is the corresponding percentage of genes. The regression patterns show the highest reliability, and k-means clusters show the lowest reliability.
Table 1 Type I Sum of Squares used to construct F test for pattern determination.
Type I Sum of Squares Interpretations F tests
SS(linear) total variability in the experiment due to the linear effect of time
SS(quadratic | linear) total variability in the experiment due to the quadratic effect of time that is not contained in SS(linear)
SS(residual) SS(total) - SS(linear) - SS(quadratic | linear)
SS(total) is the total variability in the experiment; df1, df2, and df3 represent the degree of freedoms of SS(linear), SS(quadratic | linear), and SS(residual), respectively.
Table 2 Determination of gene temporal expression patterns by the proposed regression method.
Regression Patterns Sign of Sign of Predicted Signals
Linear up (LU) + N/A N/A
down (LD) - N/A N/A
Quadratic concave (QC) N/A - N/A
convex (QV) N/A + N/A
Quadratic-Linear concave up (QLCU) N/A -
concave down (QLCD) N/A -
convex up (QLVU) N/A +
convex down (QLVD) N/A +
"+" if the estimate of or is positive, "-" if the estimate of or is negative, "N/A" if not applicable, is the predicted signal at the first time point, and is the predicted signal at the last time point.
Table 3 Comparisons of regression patterns and Peddada et al.'s profiles obtained from 3834 present genes.
Regression patterns Peddada et al.'s profiles
MI (48) MD (16) UD2 (155) UD3 (25) UD4 (34) DU2 (62) DU3 (26) DU4 (13)
LU (228) 41 0 0 0 13 11 0 1
LD (123) 0 11 8 0 0 0 0 0
QC (69) 0 0 14 7 1 0 0 0
QV (15) 0 0 0 0 0 0 3 1
QLCU (20) 2 0 1 3 3 0 0 0
QLCD (214) 0 3 102 9 1 0 0 0
QLVU (125) 2 0 0 0 0 44 13 2
QLVD (4) 0 0 0 0 0 0 0 1
The numbers in the parenthesis represent the numbers of genes contained in regression patterns or Peddada et al.'s profiles obtained from the analyses on the data of 3834 genes.
Table 4 Comparisons of regression patterns and k-means clusters obtained from 770 ANOVA significant genes.
Regression Patterns K-means Clusters
K1 (163) K2 (126) K3 (107) K4 (42) K5 (81) K6 (95) K7 (41) K8 (64) K9 (51)
FLAT (211) 12 30 36 18 19 8 41 20 27
LU (165) 0 68 0 0 51 10 0 36 0
LD (72) 19 0 53 0 0 0 0 0 0
QC (43) 5 0 0 20 0 0 0 0 18
QV (8) 0 1 0 0 0 7 0 0 0
QLCU (13) 0 0 0 0 4 0 0 8 1
QLCD (151) 127 0 15 4 0 0 0 0 5
QLVU (104) 0 27 0 0 7 70 0 0 0
QLVD (3) 0 0 3 0 0 0 0 0 0
The numbers in the parenthesis represent numbers of genes contained in regression patterns or k-means clusters obtained from the analyses on the data of 770 genes.
Table 5 Over-represented gene categories in some regression patterns from EASE functional analysis.
Reg. Patterns Gene Category List Hits List Total Pop. Hits Pop. Total EASE Score
LD Cell adhesion 10 64 52 699 3.53E-02
LU Immune regulation
defense response 43 159 98 699 9.77E-07
response to biotic stimulus 44 159 104 699 2.38E-06
immune response 37 159 88 699 2.77E-05
response to external stimulus 50 159 142 699 1.79E-04
immune cell activation 5 159 6 699 2.58E-02
cell activation 5 159 6 699 2.58E-02
lymphocyte activation 5 159 6 699 2.58E-02
B-cell activation 4 159 4 699 3.79E-02
Other biological functions 18 159 43 699 7.52E-03
QLCD Coenzyme and prosthetic group metabolism 7 139 12 699 1.73E-02
QLCU Signaling
cyclic-nucleotide-mediated signaling 3 12 4 699 1.33E-03
second-messenger-mediated signaling 3 12 5 699 2.20E-03
G-protein signaling, coupled to cyclic nucleotide second messenger 2 12 3 699 4.65E-02
cAMP-mediated signaling 2 12 3 699 4.65E-02
Protein metabolism 7 12 173 699 3.16E-02
QLVU Immune regulation
response to pest/pathogen/parasite 21 98 61 699 5.80E-05
response to wounding 14 98 40 699 1.54E-03
inflammatory response 12 98 32 699 2.19E-03
innate immune response 12 98 32 699 2.19E-03
defense response 24 98 98 699 3.88E-03
response to biotic stimulus 25 98 104 699 3.98E-03
immune response 22 98 88 699 4.82E-03
response to stress 23 98 98 699 8.67E-03
response to chemical substance 7 98 18 699 2.80E-02
humoral defense mechanism (sensu Vertebrata) 6 98 14 699 3.32E-02
response to external stimulus 28 98 142 699 3.53E-02
Cell surface receptor linked signal transduction 18 98 80 699 3.64E-02
Cell-matrix adhesion 4 98 6 699 3.78E-02
"Reg. Patterns" stands for the regression patterns identified by the proposed regression method; in the "Gene Category" column, the gene categories are further summarized to broader categories (in bold); "Pop. Total" stands for the number of total input genes (770) that are contained in EASE database, the remaining 71 genes do not have a biological function identified by EASE; "Pop. Hits" stands for the number of genes in "Pop. Total" that are classified into each gene category; "List Total" stands for the number of genes in "Pop. Total" that are classified into each regression pattern; "List Hits" stands for the number of genes in "List Total" that are classified into each gene category.
Table 6 Over-represented gene categories in some k-means clusters from EASE functional analysis.
K-means Clusters Gene Category List Hits List Total Pop. Hits Pop. Total EASE Score
K2 Immune Regulation
immune response 23 121 88 699 3.02E-02
defense response 25 121 98 699 3.02E-02
response to biotic stimulus 26 121 104 699 3.38E-02
K4 Humoral immune regulation 5 37 24 699 3.01E-02
K5 Immune Regulation
immune response 18 80 88 699 1.25E-02
response to biotic stimulus 20 80 104 699 1.51E-02
defense response 19 80 98 699 1.72E-02
Cell death
apoptosis 9 80 34 699 2.91E-02
programmed cell death 9 80 35 699 3.43E-02
Ion Homeostasis
ion homeostasis 5 80 13 699 4.88E-02
cell ion homeostasis 5 80 13 699 4.88E-02
Embryogenesis and morphogenesis 4 80 6 699 2.16E-02
Other biological functions 12 80 43 699 5.39E-03
K6 Immune Regulation
innate immune response 14 87 32 699 3.03E-05
inflammatory response 14 87 32 699 3.03E-05
response to pest/pathogen/parasite 20 87 61 699 3.32E-05
response to wounding 15 87 40 699 1.04E-04
defense response 24 87 98 699 6.22E-04
response to biotic stimulus 24 87 104 699 1.57E-03
immune response 21 87 88 699 2.41E-03
response to stress 22 87 98 699 4.07E-03
response to chemical substance 7 87 18 699 1.59E-02
acute-phase response 4 87 6 699 2.73E-02
chemotaxis 6 87 16 699 3.65E-02
taxis 6 87 16 699 3.65E-02
response to external stimulus 25 87 142 699 4.55E-02
Regulation
regulation of biological process 11 87 39 699 1.45E-02
regulation of cellular process 11 87 39 699 1.45E-02
regulation of cell proliferation 9 87 30 699 2.25E-02
Cell surface receptor linked signal transduction 17 87 80 699 2.50E-02
In the "Gene Category" column, the gene categories are further summarized to broader categories (in bold); "Pop. Total" stands for the number of total input genes that are contained in EASE database; "Pop. Hits" stands for the number of genes in "Pop. Total" that are classified into each gene category; "List Total" stands for the number of genes in "Pop. Total" that are classified into each k-means cluster; "List Hits" stands for the number of genes in "List Total" that are classified into each gene category.
==== Refs
Peddada SD Lobenhofer EK Li L Afshari CA Weinberg CR Umbach DM Gene selection and clustering for time-course and dose-response microarray experiments using order-restricted inference Bioinformatics 2003 19 834 841 12724293 10.1093/bioinformatics/btg093
Eisen MB Spellman PT Brown PO Botstein D Cluster analysis and display of genome-wide expression patterns PNAS 1998 95 14863 14868 9843981 10.1073/pnas.95.25.14863
Tavazoie S Hughes JD Campbell MJ Cho RJ Church GM Systematic determination of genetic network architecture Nature Genet 1999 22 281 285 10391217 10.1038/10343
Tamayo P Slonim D Mesirov J Zhu Q Kitareewan S Dmitrovsky E Lander ES Golub TR Interpreting patterns of gene expression with self-organizing maps: Methods and application to hematopoietic differentiation PNAS 1999 96 2907 2912 10077610 10.1073/pnas.96.6.2907
Brown MPS Grundy WN Lin D Cristianini N Sugnet CW Furey TS Ares M JrHaussler D Knowledge-based analysis of microarray gene expression data by using support vector machines PNAS 2000 97 262 267 10618406 10.1073/pnas.97.1.262
Wolfinger RD Gibson G Wolfinger ED Bennett L Hamadeh H Bushel P Afshari C Paules RS Assessing Gene Significance from cDNA Microarray Expression Data via Mixed Models Journal of Computational Biology 2001 8 625 637 11747616 10.1089/106652701753307520
Park T Yi S-G Lee S Lee SY Yoo D-H Ahn J-I Lee Y-S Statistical tests for identifying differentially expressed genes in time-course microarray experiments Bioinformatics 2003 19 694 703 12691981 10.1093/bioinformatics/btg068
Yeung KY Fraley C Murua A Raftery AE Ruzzo WL Model-based clustering and data transformations for gene expression data Bioinformatics 2001 17 977 987 11673243 10.1093/bioinformatics/17.10.977
Pan W Lin J Le C Model-based cluster analysis of microarray gene-expression data Genome Biology 2002 3 research0009.0001 research0009.0008 11864371 10.1186/gb-2002-3-2-research0009
Xu XL Olson JM Zhao LP A regression-based method to identify differentially expressed genes in microarray time course studies and its application in an inducible Huntington's disease transgenic model Hum Mol Genet 2002 10 1977 1985 10.1093/hmg/11.17.1977
Luan Y Li H Clustering of time-course gene expression data using a mixed-effects model with B-splines Bioinformatics 2003 19 474 482 12611802 10.1093/bioinformatics/btg014
Bar-Joseph Z Gerber G Simon I Gifford DK Jaakkola TS Comparing the continuous representation of time-series expression profiles to identify differentially expressed genes PNAS 2003 100 10146 10151 12934016 10.1073/pnas.1732547100
Bar-Joseph Z Analyzing time series gene expression data Bioinformatics 2004 20 2493 2503 15130923 10.1093/bioinformatics/bth283
The Excel macro for the step-down quadratic regression method
Getchell TV Liu H Vaishnav RA Kwong K Stromberg AJ Getchell ML Temporal profiling of gene expression during neurogenesis and remodeling in the olfactory epithelium at short intervals after target ablation Journal of Neuroscience Research 2005 80 309 329 15795924 10.1002/jnr.20411
Kerr MK Churchill GA Bootstrapping cluster analysis: Assessing the reliability of conclusions from microarray experiments PNAS 2001 98 8961 8965 11470909 10.1073/pnas.161273698
Benjamini Y Hochberg Y Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing Journal of the Royal Statistical Society Series B (Methodological) 1995 57 289 300
SAS v9 online document Cary, NC, USA: SAS Institute Inc
Hosack D Dennis G Sherman B Lane H Lempicki R Identifying biological themes within lists of genes with EASE Genome Biology 2003 4 R70 14519205 10.1186/gb-2003-4-10-r70
Montgomery DC Peck EA Vining GG Introduction to linear regression analysis 2001 3 John Wiley &Sons, Inc
Westfall PH Young SS Resampling-Based Multiple Testing: Examples and Methods for p-Value Adjustment 1993 John Wiley &Sons, Inc
Hartigan JA Clustering Algorithms 1975 John Wiley &Sons, Inc
Schwarz G Estimating the dimension of a model Annals of Statistics 1978 6 461 464
Seber GA Lee AJ Linear regression analysis, second edition 2003 John Wiley &Sons, Inc
Efron B Tibshirani RJ An Introduction to the Bootstrap 1993 Chapman and Hall
| 15850479 | PMC1127068 | CC BY | 2021-01-04 16:35:47 | no | BMC Bioinformatics. 2005 Apr 25; 6:106 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-106 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1071585049110.1186/1471-2105-6-107Methodology ArticleRedefinition of Affymetrix probe sets by sequence overlap with cDNA microarray probes reduces cross-platform inconsistencies in cancer-associated gene expression measurements Carter Scott L [email protected] Aron C [email protected] Brigham H [email protected] Isaac S [email protected] Zoltan [email protected] Children's Hospital Informatics Program, Harvard Medical School, Boston, MA, 02115 USA2 Laboratory of Functional Genomics, Brigham and Women's Hospital, 65 Landsdowne Street, Cambridge, MA 02139, USA2005 25 4 2005 6 107 107 10 1 2005 25 4 2005 Copyright © 2005 Carter 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
Comparison of data produced on different microarray platforms often shows surprising discordance. It is not clear whether this discrepancy is caused by noisy data or by improper probe matching between platforms. We investigated whether the significant level of inconsistency between results produced by alternative gene expression microarray platforms could be reduced by stringent sequence matching of microarray probes. We mapped the short oligo probes of the Affymetrix platform onto cDNA clones of the Stanford microarray platform. Affymetrix probes were reassigned to redefined probe sets if they mapped to the same cDNA clone sequence, regardless of the original manufacturer-defined grouping. The NCI-60 gene expression profiles produced by Affymetrix HuFL platform were recalculated using these redefined probe sets and compared to previously published cDNA measurements of the same panel of RNA samples.
Results
The redefined probe sets displayed a substantially higher level of cross-platform consistency at the level of gene correlation, cell line correlation and unsupervised hierarchical clustering. The same strategy allowed an almost complete correspondence of breast cancer subtype classification between Affymetrix gene chip and cDNA microarray derived gene expression data, and gave an increased level of similarity between normal lung derived gene expression profiles using the two technologies. In total, two Affymetrix gene-chip platforms were remapped to three cDNA platforms in the various cross-platform analyses, resulting in improved concordance in each case.
Conclusion
We have shown that probes which target overlapping transcript sequence regions on cDNA microarrays and Affymetrix gene-chips exhibit a greater level of concordance than the corresponding Unigene or sequence matched features. This method will be useful for the integrated analysis of gene expression data generated by multiple disparate measurement platforms.
==== Body
Background
The first years of microarray analysis of human cancer samples produced several promising results, introducing complex gene expression profiles for diagnostics and predicting disease outcome [1]. However, initial enthusiasm was replaced by uncertainty when classifiers produced for the same type of diseases in various studies shared few if any of the same marker genes [2]. Although microarray results are often reproducible for a single platform, inconsistencies in sensitivity, cross hybridization, and splice variant specificity may render the transfer of results between microarray platforms problematic.
One of the difficulties in the cross platform comparison of microarray data is to ascertain that probes on the various platforms aimed at the same gene do in fact quantify the same mRNA transcript. The various strategies to match probes between different platforms can be constrained by the amount of information provided by the manufacturers of the given microarray. Initially, actual probe sequence information was not released; therefore, probe matching could be based only on gene identifiers such as the Unigene ID. This strategy is known to produce a significant number of incorrect pairings [3]. As partial or complete probe sequence information has become available, more accurate strategies can now be implemented.
In a recent study, we compared several Affymetrix platforms (for which probe sequence information was available) to the Agilent Human 1 cDNA microarray platform [4]. Probe sequence information was unavailable for the Agilent platform except for a 100 base lead sequence at one end of each cDNA probe. Using this information, we queried whether the Affymetrix probes and the 100 base lead sequence could be mapped to a single Unigene transcript. Unigene matched probes across the two platforms that failed this sequence mapping test showed a significantly lower expression correlation across the two microarray platforms [4]. However, the lack of complete cDNA sequence information precluded determination of the actual sequence overlap level with high certainty.
In contrast to the Agilent probes, short sequences from both the 5' and 3' ends are generally available for clones on Stanford cDNA microarrays. Using these sequences to infer the complete clone sequence, we show that the level of probe sequence overlap is highly related to the gene expression concordance between the Affymetrix and cDNA microarray platforms. Eliminating non-overlapping probes allowed us to extract more consistent results from cancer associated gene expression data produced by different platforms and in different institutions.
Results
Depending on availability or the set of genes to be quantified, large scale gene expression profiling studies have used different versions of chips of a given microarray platform. For the data sets analyzed in this study two types of Affymetrix chips were used: the HuFL oligo chips and the U95Av2 chips. These contain 20 and 16 oligo probes per probe set, respectively. For the cDNA microarray studies, the pool of actual clones shows a very high level of diversity between various studies. Therefore, the exact number of overlapping probes depended on both the specific generation of Affymetrix platform and the set of cDNA clones to which it was mapped. A summary of these data is listed in Table 1.
Comparison of cDNA and Affymetrix expression measurements
Because cDNA microarray measurements are typically reported as the log ratio of an experimental (Cy5) and control (Cy3) channel, direct comparison with single-channel Affymetrix data required that one of the two data sources be converted to a scale compatible with the other. Because the spot-size on robotically spotted cDNA microarrays can vary substantially, considering only the experimental channel would have given expression measurements prone to probe-quantity artifacts. On the other hand, without direct measurement on the Affymetrix platform of the control RNA used in the cDNA hybridization, it was impossible to replicate exactly the reference response level of each measurement feature.
We attempted to address this difficulty by assuming that the reference RNA batches chosen for each cDNA hybridization uniformly reflect the diversity of experimental transcript populations and therefore that the mean of a gene's measured expression level across all experiments may serve as a reference for the normalization of Affymetrix data (methods). We verified that the mean expression measured by each Affymetrix array did not vary substantially (max- min < 0.25).
Sequence-overlapping probes give greater cross-platform consistency for the NCI-60 panel
The NCI-60 cell line panel consists of sixty well characterized human tumor cell lines derived from patients with leukaemia, melanoma, and lung, colon, central nervous system, ovarian, renal, breast and prostate cancers. This cell line panel has been developed by the Developmental Therapeutics Program of the National Cancer Institute and routinely used to screen potential anticancer drugs [5].
The gene expression profiles of the NCI-60 cell line panel measured by cDNA microarray and by Affymetrix HuFL oligo chips constitutes a unique data source. To the best of our knowledge, it is the only publicly available dataset in which replicates of a large number of diverse RNA samples have been quantified by these two microarray platforms. Affymetrix microarray probe sets were classified based on their shared sequence identity across the two platforms.
Since the actual number of overlapping probes can be between 0 and 20, a large number of potential stratification schemes can be implemented. However, for a clear presentation of results we chose to compare the following classes representing different levels of shared identity: a) Affymetrix probe sets that share a Unigene ID with a cDNA clone. (termed Shared Unigene probes) b) Affymetrix probe sets containing probes that could be sequence-matched to the same transcript sequence as the cDNA clone, but for which no Affymetrix probe actually overlaps the cDNA clone sequence (termed Shared Transcript probes); c) Affymetrix probe sets with 1 to 10 probes sequence overlapping with the cDNA clone (termed Partially Overlapping probes); d) Affymetrix probe sets with 20 (i.e. all) probes sequence overlapping with the cDNA clone (termed Completely Overlapping probes); e) alt-CDF or "redefined probe sets" for which all probes across the entire array that matched to a given cDNA clone insert were used to define a new derivative probe set. This new probe set may contain only a subset (even a single probe) of an original probe set; in other cases probes across several original probe sets were joined into the new derivative probe set (fig 1). For "partially overlapping" and "completely overlapping" probes (classes c and d), the entire original probe set was used for calculating gene expression levels, whereas for the "redefined" probe sets (class e) only the sequence mapped probes were retained.
Figure 2 demonstrates the correlation between the Affymetrix and cDNA microarray measurements for the various types of matched probes across the two platforms. Increasing the number of overlapping Affymetrix probes ensures increased cross-platform consistency both for matched genes and matched cell-lines. Additionally, concordance was greatest when only sequence-overlapping probes were used by redefining probe sets, even though in some cases only a single Affymetrix probe was considered. Redefined probes and completely overlapping probes showed the highest concordance levels. (The cumulative correlation distributions showed little difference, however the former method allowed a 4-fold increase in the number of available genes.) These results imply that probes targeting identical transcript sequence regions give substantially stronger concordance than probes that target identical contiguous transcript molecules at different sequence regions. In order to further investigate the effect of direct sequence overlap we examined the performance of Affymetrix probe sets that can be sequence mapped to the same transcript molecule but show no actual overlap with the cDNA clone insert ("shared transcript" probes, class b). These probe sets showed the lowest correlation. This might be due to a number of factors including the presence of splice variants, the probes being subject to different cross-hybridization patterns, or incorrect clone sequence predictions.
Figure 2A also shows, however, that a significant number of probes matched by complete sequence overlap show rather poor correlation (around zero) across the two platforms. The same applies to redefined probe sets. Because we used Pearson correlation as our concordance metric, we expect genes for which the signal fluctuation is below the resolution of the measurement platform to have low levels of concordance, (since the corresponding correlations will be made between noise.) We investigated the effect of removing genes with low levels of variation across the cell-lines on the cross-platform concordance (Fig. 3). Specifically, we removed genes from the Affymetrix dataset with standard deviations below 0.388, (representing the 50th percentile of standard deviation in the full Unigene-mapped dataset.) We removed genes from the cDNA dataset with standard deviations below 0.265, (representing the 50th percentile of standard deviation in the full cDNA dataset.) Matched gene and cell-line concordance was then assessed as described using the genes remaining in both datasets (Fig. 3).
As expected, removing these genes substantially increased both gene and cell-line concordance (Fig. 3). This improvement was substantially greater than that obtained by filtering genes based on mean expression (data not shown). Specifically, the range of median gene correlation increased from approximately 0.2 – 0.4 to 0.4 – 0.6. Interestingly, filtering did not give a substantial improvement near the low end of the distribution, suggesting that some correlations of < 0.1 may be due to incorrect mappings or non-functional probes.
Finally, we noted that "complete overlap" matched pairs performed better than redefined probe sets after standard deviation filtering. This may be due to a number of factors, such as the potentially small number of probes interrogating a given transcript level (in some cases only a single probe.) Alternatively, the redefined probe sets may contain spurious probes in cases where a false-positive clone sequence prediction led to the combination of several Affymetrix-defined probe sets. In any case, the ~4-fold increase in the number of mapped genes available through redefined probe sets may offset the small reduction in concordance.
Highly correlated genes are expected to produce a more reproducible unsupervised classification of the cell lines than that derived from a larger pool of genes with less correlation. This can be evaluated in several ways. For example, the hierarchical classification trees derived from the Affymetrix gene chip and cDNA microarray based measurements can be visually compared. Improved reproducibility of classification is indicated by the fact that more cell lines show similar or identical classification on the two hierarchical trees (fig 4).
Encouraged by our initial success, we merged the Affymetrix and cDNA microarray based gene expression profiles and hierarchically clustered the composite data set. More consistent measurements of gene expression across the two platforms would result in a greater number of instances in which the measurements of the same cell-line cluster together. In addition, co-clustering of cell lines of similar origin also provides circumstantial evidence that the gene expression profiles accurately reflect a certain tumor subtype.
Indeed, hierarchical clustering of the combined datasets resulted in a greater number of matched cell-lines clustering together when only sequence-overlapping measurements were used (fig 5). The majority of matched cell lines are more correlated to one another than to any other cell line from either platform. This was not the case when the expression measurements were Unigene-matched (fig 5A).
We were somewhat disconcerted by the fact that some of the cell lines showed a completely different localization on the two hierarchical trees. For example, the colon cancer cell line HT-29 clusters together with other colon cancer cell lines on the cDNA microarray derived tree but it is placed in a different cluster on the Affymetrix gene chip based classification tree (fig 4). An obvious explanation for this discrepancy would be the failure of the Affymetrix gene chip based measurement. Since no replicates were produced for any of the measurements, there is no statistically sound way of evaluating the quality of any of the gene expression profiles except by some circumstantial measures. For example, most cell lines had cross-platform correlation coefficients larger than 0.2 (Fig 2B). HT-29 was the single outlier with correlation consistently near 0. We obtained an alternative measurement of the same cell line based on an HG-U133A Affymetrix gene chip (a generous gift of Avalon Pharmaceuticals Inc.) We extracted a gene expression profile using the "redefined probe sets" strategy. This gene expression vector produced a much higher correlation coefficient (0.208) with the corresponding cDNA microarray measurements.
Sequence overlapping measurements improve cross-platform classification of breast cancer subtypes
We were seeking further confirmation for our method using gene expression profiles derived from various human tissue samples. These data sets do not allow highly controlled side-by-side comparisons such as the above presented analysis using in vitro cell lines. Therefore, we needed to rely on "indirect" measures of cross-platform consistency, such as classification reproducibility.
Namely, we investigated whether sequence matching of probes would enable us to reproduce the classification of primary breast tumor derived gene expression profiles produced by different microarray platforms.
A breast-cancer subtype classifier was derived from a cohort of patients profiled on cDNA microarrays [1]. This classifier transferred to Affymetrix HuFL gene expression data [6] only to a limited extent [7]. Recently, we improved on those results by using only those Affymetrix and cDNA probes that could be mapped to the same transcript [4]. This earlier publication, however, did not involve the selective use of only those oligo probes that actually matched the cDNA clone. Here we introduced the use of "redefined probe sets" as described in the methods. This was coupled with an advanced normalization method, RMA [8], leading to a strong overall improvement over the original results of Sørlie et al [7] (fig 6). In particular, with two exceptions, all samples could be assigned to a breast cancer subtype defined by the cDNA microarray derived centroids. In addition, more than 70% of all samples clustered in their own well-defined clusters.
Furthermore, we compared the transfer of the cDNA-based classifier [7] to two additional cohorts of breast cancer samples profiled on Affymetrix HG-U95Av2 gene-chips [9,10], using both the 'shared Unigene' (fig 7A) and 'redefined probe sets' (fig 7B) to match measurements (see methods). Since true classes are usually not known a priori for novel cancer subtypes, we focused our attention on a subtype where gene expression profiles associated with an independent immunohistochemical marker: Her-2 / erbB2 status. Significantly, the classification based on 'redefined probe sets' contains a larger and more coherent ERBB2+ subtype cluster than that based on shared Unigene identifier. The validity of this cluster was substantiated by the immunohistochemical assessment of Her-2 status (available only for the Santorini cohort); all of the tested samples in this cluster stained positive for Her-2 amplification.
Sequence-overlapping measurements improve cross-platform similarity of normal lung samples
Finally, we evaluated our sequence-overlap probe set redefinition method on a third cDNA platform. In this case, we evaluated the cross-platform similarity of normal lung samples profiled on cDNA microarrays [11] and Affymetrix HG-U95Av2 gene chips [12]. These two independent data sets contain normal samples from different patients. However, a robust gene expression profile was detected in both studies for the normal lung tissue samples [11,12]. If this robust, normal gene expression profile is accurately measured by both microarray platforms, then a high Pearson correlation coefficient would be expected between the normal samples, independently from the microarray platform used for a given tissue sample. Therefore, we calculated the correlation coefficient between each possible pair of normal gene expression profiles across the two platforms. Two probe matching strategies, the Unigene and sequence-overlap based mappings were compared (fig 8). The significance of the observed increase in cross-platform correlation was assessed at p = 0.0002 (methods), further highlighting the advantage of using only sequence-overlapping measurements for cross-platform comparison.
Discussion
Despite the fact that all microarray technologies are based on the same basic principle of complementary hybridization, various probe selection strategies aim to achieve optimal probe performance given the technological constraints using fundamentally different strategies. In order to be able to plan long-term microarray based experimental strategies, end users have hoped either for a clearly superior technology to emerge, perhaps supported by a large number of independent validations, or for a high level of cross-platform consistency when the same type of RNA is expression profiled on different platforms. The latter being true would mitigate the risk of committing to a less accurate technology. Unfortunately, this hope has not been fulfilled yet. The limited number of independent validations published so far suggested a similar level of accuracy, or lack thereof, for the most widely used platforms [13-15], and the first cross platform comparison studies revealed an alarming level of inconsistency between platforms such as the cDNA microarray and the Affymetrix oligo chip [16]. This provided little guidance for prospective users on how to choose the technology best suited for their experiments.
Cross platform consistency is an imperfect tool with which to validate microarray platforms. Lack of consistency can be caused by the inferior performance of either one or both platforms, without clear indication of their relative merit. On the other hand, highly similar results across platforms could be simply caused by consistent cross-hybridization patterns without either platform measuring the true level of expression. Nevertheless, a high level of cross platform consistency is desirable. If both platforms perform accurate measurements then cross platform consistency will automatically follow. In other words, cross platform consistency is the sine qua non of accurate microarray measurements but by itself will not validate the technology.
Cross platform inconsistencies can be caused by at least two major factors: a) significant differences in noise structure between technologies; b) differential hybridization of homologous probes designed to measure the same gene on various platforms. It has been shown that the most consistent results across different versions of the Affymetrix DNA chips are provided by identical probes [17]. Probes with less or no sequence overlap, even if targeting the same gene at different locations, show substantially lower consistency. Therefore, sequence matching probes provides a strategy for dissecting the sources of cross platform inconsistency.
There are only a few publicly-available data sets that allow comprehensive cross platform comparison of a relatively large number of RNA samples with ample probe sequence information available. The most widely studied of these is the gene expression profiling of the NCI-60 cell line panel produced by the Affymetrix and cDNA microarray technologies [5,16,18-20].
These two data sets showed an alarming level of inconsistencies in an early study when microarray probes, due to the lack of available probe sequence information, were matched across platforms by Unigene IDs [16]. A higher level of consistency was achieved in a subsequent study following the release of probe sequence information by Affymetrix [18]. The authors found a higher level of cross platform consistency using only the subset of probe sets that could effectively be sequence mapped to the same Unigene entity as the corresponding cDNA clone. We obtained similar results in a more limited cross platform comparison study [4]. However, this strategy did not take into consideration whether the short individual oligo probes actually overlapped the corresponding cDNA clone insert. Therefore, portions of the matched Affymetrix probe-sets could have been measuring different regions or different splice variants of the target transcript probed by the cDNA clone. This was perhaps the reason that reproducing the clustering of the NCI-60 cell lines required the highly biased supervised filtering of all genes with a low level of consistency [18]. We introduced here a further improvement that allowed us to rely solely on sequence information and eliminated any further supervised filtering based on expression data. Our strategy relied on using expression signals from only those short individual oligo probes that could be physically mapped onto the corresponding cDNA clone insert. Furthermore, this grouping was done irrespective of the default manufacturer-defined probe sets, in some cases combining probes from several of them. This was much facilitated by a recently introduced elegant computational tool that allows the redefinition of an entire Affymetrix chip definition file within the framework of Bioconductor [21,22]. This strategy constitutes the highest level of sequence based stringency for matching Affymetrix probe sets with cDNA clones to date. Given the importance of correctly designed probes, it is not surprising that this method provides the highest level of cross platform consistency at different levels of the analysis. In addition to the higher levels of correlation, it also improved the transfer of classification results between breast cancer associated gene expression data produced by different microarray platforms.
Conclusion
We have shown that probes which target overlapping transcript sequence regions on cDNA microarrays and Affymetrix gene-chips exhibit a greater level of concordance than the corresponding Unigene or sequence matched features. Despite these promising results, we should remain aware of the limitations of this method. Microarray signals are a composite of three factors: 1) true signal from the targeted gene, 2) cross-hybridization with other genes, and 3) random noise. The stringent sequence matching applied in this paper increases the consistency of the first two factors across the platforms. However, it does not allow for an easy deconvolution i.e. whether the higher level of observed cross-platform consistency is due to measurement of only the true signal or to reproduction of the cross hybridization pattern. This determination will require further studies underway in our laboratory.
Finally, the assumption that reference mRNA batches used in cDNA hybridizations reflect the full level of diversity in a target experimental mRNA population is imperfect. Without access to measurements of this mRNA on the experimental platform of interest, it is impossible to replicate exactly the normalization inherent in a cDNA log ratio. It is therefore important that the origin of the reference mRNA sample be kept prominently in mind when considering the results of any cDNA microarray experiment.
Methods
Inference of cDNA probe sequences
For a given cDNA clone, all corresponding read sequences were extracted from dbEST [23]. When both 5' and 3' read sequences were available for a given clone, these sequences were BLASTed against the Acembly transcript database corresponding to human genome build hg16. The alignment results were used to construct a list of putative insert regions. If both clone read sequences had a high-quality (expectation value < 0.001) hit in the correct sense to a given transcript, the transcript region comprising both read sequences and the flanked region is predicted to be the clone sequence. Statistics for the mapping of each cDNA microarray platform are summarized in table 1.
Mapping of Affymetrix probes
For a given Affymetrix platform, all probe sequences as obtained from Affymetrix were matched against the Acembly transcript database. Only exact matches were retained. Based on these results, we determined the number of Affymetrix probes in each probe set that overlapped each predicted clone sequence.
In addition to assessing the extent of whole probe set-level overlap with the clone sequence, we also constructed alternative groupings of Affymetrix probes for each platform. These redefined probe sets comprised all Affymetrix probes that overlapped the corresponding cDNA clone, whether or not those probes were intended to be a single probe set by the manufacturer. In some cases, these probes spanned several of the probe sets as defined by Affymetrix (table 1). We then re-computed normalized expression values for the datasets using these redefined probe sets using the "altcdfenvs" package in Bioconductor [21,22]. Applying this strategy allowed us to use only those short oligo probes that overlapped the corresponding cDNA clone insert. The alternate probe mappings are available in a format compatible with the "altcdfenvs" package [see Additional file 1].
Normalization of Affymetrix data for comparison with cDNA microarray data
All raw Affymetrix probe-level measurements were first transformed into log expression measures using RMA [8]. These expression measurements were then converted into log ratios by subtracting the mean (log) expression from each measurement. In all cases, this process was performed for each sample with respect to its complete original cohort. This was done to minimize artifacts resulting from differences in RNA amplification, labeling, hybridization conditions, etc. cDNA log ratios for each gene were mean centered with respect to the original data set.
NCI-60 concordance
Normalized cDNA microarray expression data for the NCI-60 cell lines was obtained from a previous study [18]. The reference RNA batch for this study was derived from "12 highly diverse cell lines of the 60" [19]. Raw CEL files were obtained for the same cell lines run on the Affymetrix HuFL oligonucleotide expression platform [20] and normalized as described above.
In addition to sequence-overlap methods of matching measurements across the platforms, we also assessed the weaker criterion of matching probes by Unigene identifiers (build #175). Unigene clusters corresponding to each probe set were obtained from Affymetrix (annotation downloaded September 2004.) Clones on the cDNA microarray were assigned to a Unigene cluster if that cluster included an entry annotated as a read sequence for the clone's IMAGE identifier.
Concordance was assessed by computing the Pearson correlation coefficient between matched-pairs of both genes and cell-lines across the two platforms. Genes were excluded if more than 50 of the cDNA measurements for that gene were missing. We also computed the average-linkage Pearson correlation hierarchical clustering of the combined datasets using both the Unigene and sequence-overlap mappings.
Breast cancer classification
Previously described cDNA microarray expression measurements from a cohort of breast cancer patients were obtained for an 'intrinsic' gene set used to classify tumor subtypes [7]. The original reference RNA batch used for the cDNA study was derived from 11 different cultured cell lines [24]. The samples were grouped into classes corresponding to the five subtypes, and median centroids were calculated for each class as described [7]. Putative clone sequences for each clone on the microarray were determined as described above.
Raw Affymetrix HG-U95Av2 CEL files were obtained for 199 samples from two additional cohorts of breast cancer patients profiled in previous studies [9,10] and normalized as described above. Each sample was then assigned to the subtype corresponding to the median centroid for which it attained the greatest Pearson correlation level, or was designated "unclassified" if no correlation exceeded 0.1. The quality of the classification produced by both mappings was evaluated by computing the average-linkage Pearson correlation hierarchical clustering of the classified samples, based on the rationale that a more meaningful classification should correspond to more coherent sample-clusters consisting of each subtype.
Normal lung sample comparison
cDNA microarray data profiling of 5 normal lung samples was obtained from a previous study of lung cancer [11]. The original reference RNA batch used for the cDNA study was derived from 11 different cultured cell lines [24] (the same reference as used in the breast cancer experiment.) Affymetrix HG-U95Av2 CEL files were obtained from an additional lung cancer study [12], 17 of which corresponded to normal lung samples, and normalized as described above. Cross-platform Unigene and sequence-overlap based mappings were constructed as for the previous analyses. Genes were standard deviation filtered as described for the NCI-60 analysis (min cDNA SD = 0.608, min Affy SD = 0.271.) For each mapping, we calculated the Pearson correlation between each of the 5 × 17 cross-platform sample-pairs and compared the cumulative distributions (Fig 8). The significance of the observed improvement in the redefined probe set mapping was quantified using an exact one-sided Kolmogorov-Smirnov test.
Authors' contributions
SLC participated in conceiving the study, carried out most of the analyses and prepared the manuscript, ACE participated in conceiving the study, carried out parts of the analyses and prepared the manuscript, BHM participated in the experimental design, ISK participated in conceiving the study and preparing the manuscript, ZS originated and conceived the study and prepared the manuscript.
Supplementary Material
Additional File 1
ZIP of 4 files allowing the remapping Affymetrix probe-sets described in this manuscript. These files can be used with the "altcdfenvs" package in Bioconductor to implement the redefinition of probe-sets based on sequence-matching with each of the 4 cDNA datasets described.
Click here for file
Acknowledgements
We thank Meena Augustus, Jeffrey Strovel, and Reinhard Ebner of Avalon Pharmaceuticals for sharing supporting data in the form of raw microarray files. We thank Robert Gentleman for providing computational resources. ISK was supported in part by the National Library of Medicine through grant U54LM008748-01. Z.S. was supported in part by the National Institutes of Health through grants HL02-005 and 1PO1CA-092644-01.
Figures and Tables
Figure 1 Composition of redefined Affymetrix probe-sets based on overlap with cDNA clone insert sequence. Stacked histograms show the distribution of probe-set size for sets consisting of a single Affymetrix-defined probe-set (black) and for those comprised of probes originally grouped into separate probe-sets by Affymetrix (gray). A, NCI-60 10 k cDNA microarray to HuFL alternative CDF. B, Breast cancer 8 k cDNA microarray to HuFL alternative CDF. C, Breast cancer 8 k cDNA microarray to HG-U95Av2 alternative CDF. D, Lung cancer 22 k cDNA microarray to HG-U95Av2 alternative CDF.
Figure 2 Sequence-overlapping probes give greater cross-platform concordance for the NCI-60 panel. (A) Pearson correlation coefficient was calculated for each gene between its expression values measured on the Affymetrix Hu6800 platforms and its expression values measured on the Stanford cDNA microarray across sixty cell lines of the NCI-60 panel. The figure shows the cumulative distribution of the Pearson correlation coefficients for all genes analyzed. The five different curves reflect the level of cross-platform consistency of probe sets with various levels of overlap between the two microarray platforms. Matched gene measurements across the two platforms showed higher correlation when greater numbers of probes in the Affymetrix probe sets overlapped the insert region of the cDNA clone. The highest correlation was attained when only those Affymetrix probes overlapping the insert-sequence of a given cDNA clone were retained. Measurements for which the probes targeted the same transcript as the cDNA clone, but did not overlap the clone sequence, showed the lowest correlation. (B), Pearson correlation coefficient was calculated across all genes for each matched sample pair profiled by the Affymetrix Hu6800 platform and by the Stanford cDNA microarray. The figure shows the cumulative distribution of the Pearson correlation coefficients for the sixty cell lines of the NCI-60 panel. Matched cell-line measurements showed identical stratification of correlation levels by feature-matching criteria.
Figure 3 Effect of standard deviation filtering on cross-platform NCI-60 concordance. Genes are filtered removing those with low standard deviations across the 60 cell-lines (methods.) Matching features are determined and concordance assessed as in Figure 1.
Figure 4 Conserved clustering pattern of the NCI-60 cell lines profiled using cDNA microarray and Affymetrix gene chips. Data was normalized as described (methods). Average linkage Pearson correlation hierarchical clustering was computed for each dataset. Cell line names are colored according to cancer type.
Figure 5 Improved hierarchical clustering of combined NCI-60 cell-lines profiled by Affymetrix gene-chip and cDNA microarray by sequence-overlapping probe measurements. The gene expression profiles obtained for the sixty cell lines by the Affymetrix gene chips and the Stanford cDNA microarray platform were pooled after data transformation as described in the text. Gene expression data by the two different platforms were matched by either Unigene ID matching or by redefining the Affymetrix probe sets based on the sequence overlap criteria of the probes. The pooled gene expression profiles were subjected to average linkage hierarchical clustering. Matched cell-lines from the two platforms which cluster together are marked by red branches in the dendrogram. (A) Unigene-matched measurements tended to cluster the cell-lines by measurement platform, and produced only 28 instances of matched cell-lines clustering together. (B) Sequence-overlapping probe measurements produced more (43) instances of matched cell-lines from each platform clustering together.
Figure 6 Increased efficiency of breast cancer subtype classification transfer from cDNA microarray to Affymetrix HuFL gene-chip tumor-profiles by sequence-overlapping probe measurements. Tumor samples profiled on the Affymetrix platform were classified according to their correlation with the set of subtype median-centroids derived from cDNA microarray measurements (see methods). The classified samples were then hierarchically clustered using Pearson correlation and average-linkage agglomeration. Affymetrix measurements matched to cDNA centroids by sequence-overlap of probe features produced more coherent classifications than those obtained in the original transfer (Sørlie), specifically, more coherent Luminal A and ERBB2+ subtype clusters.
Figure 7 Increased efficiency of breast cancer subtype classification transfer from cDNA microarray to Affymetrix HG-U95Av2 gene-chip tumor-profiles by sequence-overlapping probe measurements. Tumor samples profiled on the Affymetrix platform were classified according to their correlation with the set of subtype median-centroids derived from cDNA microarray measurements (see methods). The classified samples were then hierarchically clustered using Pearson correlation and average-linkage agglomeration. (A), Affymetrix measurements matched to the cDNA centroids by Unigene identifier. (B), Affymetrix measurements matched to cDNA centroids by sequence-overlap of probe features produced more coherent classifications. In particular, the large ERbB2+ subtype cluster (upper left) is mostly absent from the unigene-based classification. The significance of this cluster is supported by the observation that all tumors in this cluster for which Her-2 amplification was assessed by immunohistochemistry were designated positive.
Figure 8 Increased cross-platform similarity of normal lung samples by sequence-overlapping probe measurements. Shown are the cumulative distributions of the 5 × 17 cross-platform sample correlations (see methods.) substantially greater similarity is observed when only sequence-overlapping probe measurements are retained (black curve.)
Table 1 Summary of mapping cDNA microarray features to probes on Affymetrix gene-chips.
NCI-60 – HuFL
Brc 8k – HuFL
Brc8k – U95Av2
Lung – U95Av2
Total cDNA clones 9707 8820 8820 22691
Clones with both reads sequenced 6222 7015 7015 18645
Clones with predicted insert region 4639 6354 6354 14813
Total Probe-sets defined 1765 2403 3103 4597
Total (perfect match) probes on Affymetrix platform 131541 131541 199084 199084
Total mapped probes 26347 37559 48250 70001
Probes mapped to multiple clones 904 3224 4019 26765
Number of probe-sets with > 1 Affymetrix "probe-set" represented 115 310 664 888
==== Refs
Sørlie T Perou CM Tibshirani R Aas T Geisler S Johnsen H Hastie T Eisen MB van de Rijn M Jeffrey SS Thorsen T Quist H Matese JC Brown PO Botstein D Eystein Lonning P Borresen-Dale AL Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications Proc Natl Acad Sci USA 2001 98 10869 10874 11553815 10.1073/pnas.191367098
Lossos IS Czerwinski DK Alizadeh AA Wechser MA Tibshirani R Botstein D Levy R Prediction of survival in diffuse large-B-cell lymphoma based on the expression of six genes N Engl J Med 2004 350 1828 1837 15115829 10.1056/NEJMoa032520
Watson A Mazumder A Stewart M Balasubramanian S Technology for microarray analysis of gene expression Curr Opin Biotechnol 1998 9 609 614 9889134 10.1016/S0958-1669(98)80138-9
Mecham BH Klus GT Strovel J Augustus M Byrne D Bozso P Wetmore DZ Mariani TJ Kohane IS Szallasi Z Sequence-matched probes produce increased cross-platform consistency and more reproducible biological results in microarray-based gene expression measurements Nucleic Acids Research 2004 32 e74 15161944 10.1093/nar/gnh071
Boyd MR Paull KD Some practical considerations and applications of the National Cancer Institute in vitro anticancer drug discovery screen Drug Dev Res 1995 34 91 109 10.1002/ddr.430340203
West M Blanchette C Dressman H Huang E Ishida S Spang R Zuzan H Olson JA JrMarks JR Nevins JR Predicting the clinical status of human breast cancer by using gene expression profiles Proc Natl Acad Sci USA 2001 98 11462 11467 11562467 10.1073/pnas.201162998
Sørlie T Tibshirani R Parker J Hastie T Marron JS Nobel A Deng S Johnsen H Pesich R Geisler S Demeter J Perou CM Lonning PE Brown PO Borresen-Dale AL Botstein D Repeated observation of breast tumor subtypes in independent gene expression data sets Proc Natl Acad Sci USA 2003 100 8418 8423 12829800 10.1073/pnas.0932692100
Irizarry RA Bolstad BM Collin F Cope LM Hobbs B Speed TP Summaries of Affymetrix GeneChip probe level data Nucleic Acids Res 2003 31 e15 12582260 10.1093/nar/gng015
Huang E Cheng SH Dressman H Pittman J Tsou MH Horng CF Bild A Iversen ES Liao M Chen CM West M Nevins JR Huang AT Gene expression predictors of breast cancer outcomes Lancet 2003 361 1590 1596 12747878 10.1016/S0140-6736(03)13308-9
Signoretti S Marcotullio L Richardson A Ramaswamy S Isaac B Rue M Monti F Loda M Pagano M Oncogenic role of the ubiquitin ligase subunit Skp2 in human breast cancer The Journal of Clinical Investigation 2002 110 633 641 12208864 10.1172/JCI200215795
Garber M Troyanskaya OG Schluens K Peterson S Thaesler Z Oacyna-Genglebach M van de Rijn M Rosen GD Perou CM Whyte RI Altman RB Brown PO Botstein D Peterson I Diversity of gene expression in adenocarcinoma of the lung Proc Natl Acad Sci USA 2001 98 13784 13789 11707590 10.1073/pnas.241500798
Bhattacharjee A Richards WG Staunton J Li C Monti S Vasa P Ladd C Beheshti J Bueno R Gillete M Loda M Weber G Sugarbaker D Meyerson M Classification of human lung carcinomas by mRNA expression profiling reveals distinct adenocarcinoma subclasses Proc Natl Acad Sci USA 2001 98 13790 13795 11707567 10.1073/pnas.191502998
Tan PK Downey TJ Spitznagel EL JrXu P Fu D Dimitrov DS Lempicki RA Raaka BM Cam MC Evaluation of gene expression measurements from commercial microarray platforms Nucleic Acids Res 2003 31 5676 5684 14500831 10.1093/nar/gkg763
Gold D Coombes K Medhane D Ramaswamy A Ju Z Strong L Koo JS Kapoor M. A comparative analysis of data generated using two different target preparation methods for hybridization to high-density oligonucleotide microarrays BMC Genomics 2004 5 2 14709180 10.1186/1471-2164-5-2
Yuen T Wurmbach E Pfeffer RL Ebersole BJ Sealfon SC Accuracy and calibration of commercial oligonucleotide and custom cDNA microarrays Nucleic Acids Res 2002 30 e48 12000853 10.1093/nar/30.10.e48
Kuo WP Jenssen TK Butte AJ Ohno-Machado L Kohane IS. Analysis of matched mRNA measurements from two different microarray technologies Bioinformatics 2002 18 405 412 11934739 10.1093/bioinformatics/18.3.405
Nimgaonkar A Sanoudou D Butte AJ Haslett JN Kunkel LM Beggs AH Kohane IS Reproducibility of gene expression across generations of Affymetrix microarrays BMC Bioinformatics 2003 27 12823866 10.1186/1471-2105-4-27
Lee JK Bussey KJ Gwadry FG Reinhold W Riddick G Pelletier SL Nishizuka S Szakacs G Anneraeu J Shankavavaram U Lababidi S Smith LH Gottesman MM Weinstein JN Comparing cDNA and oligonucleotide array daya: concordance of gene expression across platforms for the NCI-60 cancer cells Genome Biology 2003 4 R82 14659019 10.1186/gb-2003-4-12-r82
Scherf U Ross DT Waltham M Smith LH Lee JK Tanabe L Kohn KW Reinhold WC Myers TG Andrews DT Scudiero DA Eisen MB Sausville EA Pommier Y Botstein D Brown PO Weinstein JN A gene expression database for the molecular pharmacology of cancer Nat Genet 2000 24 236 244 10700175 10.1038/73439
Staunton JE Slonim DK Coller HA Tamayo P Angelo MJ Park J Scherf U Lee JK Reinhold WO Weinstein JN Mesirov JP Lander ES Golub TR Chemosensitivity prediction by transcriptional profiling Proc Natl Acad Sci USA 2001 98 10787 10792 11553813 10.1073/pnas.191368598
Gautier L Moller M Friis-Hansen L Knudsen S Alternative mapping of probes to genes for Affymetrix chips BMC Bioinformatics 2004 5 111 15310390 10.1186/1471-2105-5-111
Gentleman RC Carey VJ Bates DM Bolstad B Dettling M Dudoit S Ellis B Gautier L Ge Y Gentry J Hornik K Hothorn T Huber W Iacus S Irizarry R Leisch F Li C Maechler M Rossini AJ Sawitzki G Smith C Smyth G Tierney L Yang JY Zhang J Bioconductor: open software development for computational biology and bioinformatics Genome Biology 2004 5 R80 15461798 10.1186/gb-2004-5-10-r80
Boguski MS Lowe TM Tolstoshev CM dbEST – database for "expressed sequence tags" Nat Genet 1993 4 332 333 8401577 10.1038/ng0893-332
Perou CM Sørlie T Eisen MB van de Rijn M Jefferey SS Rees CA Pollack JR Ross DT Johnson H Akslen LA Fluge Ø Pergamenschikov A Williams C Zhu SX Lønning PE Børresen-Dale A Brown PO Botstein D Molecular portraits of human breast tumors Nature 2000 406 747 752 10963602 10.1038/35021093
| 15850491 | PMC1127107 | CC BY | 2021-01-04 16:02:50 | no | BMC Bioinformatics. 2005 Apr 25; 6:107 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-107 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1091585750510.1186/1471-2105-6-109Methodology ArticleSome statistical properties of regulatory DNA sequences, and their use in predicting regulatory regions in the Drosophila genome: the fluffy-tail test Abnizova Irina [email protected] Boekhorst Rene [email protected] Klaudia [email protected] Walter R [email protected] MRC Biostatistics Unit, Institute of Public Health, Robinson Way, Cambridge CB2 2SR, UK2 Computer Science Department, University of Hertfordshire, College Lane, AL10 92BA, Hatfield Campus, UK2005 27 4 2005 6 109 109 17 12 2004 27 4 2005 Copyright © 2005 Abnizova 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
This paper addresses the problem of recognising DNA cis-regulatory modules which are located far from genes. Experimental procedures for this are slow and costly, and computational methods are hard, because they lack positional information.
Results
We present a novel statistical method, the "fluffy-tail test", to recognise regulatory DNA. We exploit one of the basic informational properties of regulatory DNA: abundance of over-represented transcription factor binding site (TFBS) motifs, although we do not look for specific TFBS motifs, per se . Though overrepresentation of TFBS motifs in regulatory DNA has been intensively exploited by many algorithms, it is still a difficult problem to distinguish regulatory from other genomic DNA.
Conclusion
We show that, in the data used, our method is able to distinguish cis-regulatory modules by exploiting statistical differences between the probability distributions of similar words in regulatory and other DNA. The potential application of our method includes annotation of new genomic sequences and motif discovery.
==== Body
Background
The transcription rate of genes is dictated primarily by interactions between DNA-binding transcription factors. Comparatively short sequences (several hundred to several thousand base pairs, depending on thespecies) upstream or downstream of the transcription start site often play a major role in the regulation of gene expression. Specific sites within such regions are recognized by regulatory proteins (transcription factors), which act upon binding as transcriptional repressors or activators, controlling the rate of transcription. The identification of regulatory regions, which are generally composed of dense clusters of target transcription factor binding sites, forms an essential step in understanding the regulatory interactions that govern the spatial and temporal expression of individual genes (see for example [1,2]) and genetic regulatory networks, (see for example [3]).
Ultimately, this task is accomplished experimentally using techniques such as empirical deletion analysis, direct binding measurements, and co-precipitation of protein-DNA complexes. However, experimental verification is expensive and time consuming. Therefore, to address the growing volumes of available genomic sequence, a number of algorithms that identify putative cis-regulatory modules and transcription factor binding sites using evolutionary comparisons, whole-genome data, and known descriptions of transcription factor binding sites, have been successfully developed. Regulatory regions of higher eukaryotes can be subdivided into proximal regulatory units – promoters – which are located close to and upstream of the gene, and distal transcription regulatory units called enhancers or cis-regulatory modules. These may be located far upstream or downstream of the target gene, and are much more difficult to recognise. In our work we focus on recognition of enhancers.
Methods for recognising regulatory DNA may be divided into the following approaches:
1. Recognition of regulatory DNA regions based on description of known transcription factor binding sites (TFBS). This approach exploits the clustering of known, cooperatively-acting transcription factors (TFs). Extracting clustered recognition motifs is one of the most reliable techniques, but is limited to the recognition of similarly regulated cis-regulatory regions. Among the most popular representatives of search by known TFBS are [4-9].
2. Recognition of regulatory DNA based on phylogenetic foot-printing [10-14]. Methods of this type assume that regulatory regions are highly conserved in cross-genomic comparison, and conserved segments can be extracted from evolutionary related genomes. Performance of phylogenetic foot-printing depends on the evolutionary distance between given species and on the conservation level of individual genes. This is an actively progressing area, as more and more sequenced genomes appear. However, such an approach offers little information as to the specific function of the conserved sequences. Furthermore, it is still an open question as to how many genomes are sufficient for reliable extraction of regulatory regions.
3. Methods based on the difference of local nucleotide composition between regulatory and non regulatory DNA [15-18]. It is assumed that this difference is due to presence of multiple transcription signals, such as binding motifs for TFs in regulatory regions. The works [15-17] are based on constructing a global interpolated Markov model, applied to promoter recognition only.
In our method, we assume that the abundance of regulatory motifs within regulatory regions leaves a distinct "signature" in nucleotide composition, and that it is possible to capture this "signature" statistically. More specifically, we hypothesize that it takes the form of an over-representation of "similar words" (which are not simple repeats).
The approach of looking for over-occurrence of words has also been widely used in motif discovery, but this is not our aim here. This over-representation of similar words should appear as outliers in the right tail of the distribution of similar word lists of variable length. The "fluffy tail test", proposed in this paper, is designed to identify such outliers and is a useful technique when data from multiple genes and genomes are lacking. It may also be used as a complementary tool when such data are available.
Results
In this section, we first present our new statistical 'fluffy tail' test for measuring the overrepresentation of similar words, and then show its performance on experimentally verified sequence data.
Test bed
To demonstrate the power of our test, we need a positive, experimentally verified, training set of regulatory sequence data, and also negative training sets of non- regulatory sequence data. We use three test beds. The positive training set is a collection of 60 experimentally verified functional Drosophila melanogaster regulatory regions [18]. This set consists of cis-regulatory modules located far from gene coding sequences and transcription start sites. It contains many binding sites (and site clusters), best known of which are bicoid, hunchback, Kruppel, knirps and caudal, – the sites involved in the regulation of developmental genes. The total size of the positive training set comprises about 68 Kb of sequence data, and contains 58 clusters of the same type of TFBS (homotypic). The two negative training sets are: (i) 60 randomly picked Drosophila exons, and (ii) 60 randomly picked Drosophila non-coding, non-regulatory DNA sequences: we excluded exons and regions of length 1 KB upstream and downstream of genes, using the Ensembl Genome Browser [19]. Each training set contains 68 Kb of sequences in total.
Estimation of distributions of similar words
To construct the distribution of similar words, we first need to specify the length of words under consideration. We try to mimic the TF core, which is the less variable part of a binding motif. Because the core of TFBSs is relatively short (around 3–5 bp) we considered 5-mer words, allowing for 1 mismatch. However, our results also hold for words of length 4 through 12, allowing for 1 through 4 mismatches (see Supplementary Materials [see Additional files 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12]). Thus, for each 5-mer word in each of the 180 sequences (60 sequences in each training set) we computed the number n of similar words of the same length. Thus, each word is the "seed" of a list of similar words. Next, the number of (non-disjoint) lists containing n words is counted, where n = 1,2,3....
(See Methods section for further details). As an example, thehistogram of the distribution of similar 5-mer words is plotted in Figure 1. In this histogram, the Y axis represents the number of lists containing 1,2, ..., n words and the X axis shows the number n of similar words in the list.
From this plot it can be seen that most lists contain 10 to 40 words, but there are outliers: some very large lists form a long, "fluffy" tail. We call a list having the largest size the maximal similar word list (MSWL). If the original sequence is characterized by the presence of an unusually high number of over-represented words, we expect it to contain more long lists in comparison to a random sequence.
To sample such a random distribution we shuffled the given sequence of original data 50 times. For each randomisation we assessed the frequency distribution of similar words. Figure 2 shows a typical example of the distribution of similar words for one of the randomly shuffled sequences of the same (knirps) cis-regulatory module as in Figure 1. Compared with the distribution of the original data (Figure1), the randomised sequence in Figure 2 lacks a heavy, "fluffy" right tail. Figure 3 shows the difference between original and randomised similar word distributions in cumulative form. The difference between the two curves reflects the fluffy right tail of the original data.
In Figure 4, ten randomised sequences are plotted as dotted contours together with the histogram of the original regulatory knirps data (solid). The cumulative histogram for original (solid) and randomised (dotted) sequences is shown in Figure 4 (right). All dotted tails are shorter than the solid one, indicating the statistical significance of the solid tail.
Definition of the fluffiness coefficient F
To measure how strong the distribution of similar words of regulatory regions deviate from randomness, we introduce a "fluffiness" coefficient F:
)
w here L max,original is the number of words in the maximal similar word list (MSWL) in the original sequence, and σr are the mean and standard deviation of the MSWL size in each of r shuffled sequences. Here we call the sequence "random" if it is obtained from original sequence by shuffling it, preserving its single nucleotide composition. We will omit the subscript r for Fr later in the paper for simplicity.
One can regard F as measuring the difference between signal and noise, where the signal is taken from the original sequence, and the noise from the randomised sequences with the same composition and length. Thus, the fluffiness coefficient is normalised for the length and base composition of the sequence, because we compare each original sequence only with respect to shuffled sequences of the same length and composition. Thus one can compare the fluffiness F for sequences of different base composition and length.
Results for regulatory regions
Figure 5 shows the distribution of fluffiness coefficient F for regulatory, coding and non-coding non-regulatory (NCNR) DNA. In each sequence we generated r = 50 shuffled versions, in calculating F. One can see that F = 2 distinguishes regulatory DNA from other types of DNA. Thus, we use the value F = 2 as a threshold. A sequence with F>2 we declare to have a "fluffy" tail. Moreover, we found that for each regulatory region having F>2, all the randomised sequences had a shorter tail. This value F = 2 is sufficiently robust: if we vary our threshold a little around F = 2, we still get a fair separation.
Our choice of r = 50 shuffled versions for each sequence allows us to obtain reliable estimates for the fluffiness coefficient F and make the computational time reasonable. Table 1 shows that F is somewhat unstable for smaller r for the knirps regulatory region. However, for each choice of r, F clearly exceeds the threshold value 2, in this example. See Supplementary Materials for more detailed descriptions [see Additional files 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12].
Using the methodology described above, we found that 51 out of 60 regulatory regions (85%) analysed in our positive training set exhibit the significant "fluffy-tail" pattern (see Table 2). The non-detection of the remaining "non-fluffy" regulatory regions could perhaps be partly due to the limited power of experimental deletion analyses to correctly distinguish the boundaries of the cis-regulatory modules.
We calculated the distribution of F for our two negative and one positive training sets. The separation of regulatory DNA from coding and non-coding, non-regulatory DNA on the basis of fluffiness was quantified by estimating the distribution of the F coefficients. A Kruskal-Wallis test showed that these regions differ significantly in the magnitude of the fluffiness coefficient (H = 132.81, N = 180, df = 2, p = 0.00001), with exons and non-coding non-regulatory DNA having much lower F-values than regulatory regions (See Fig. 6).
We now turn to examine the location of similar words in the MSWL for a given sequence.
When the start positions of each of the words in the MSWL are plotted, they tend to be fairly uniformly scattered along the length of the sequence, as illustrated in Figure 7.
We now examine the relationship between the MSWL and predicted TFBS sites. We found significant enrichment of most MSWLs with the occurrences of TFBS in databases: when submitted to the Transfac and Jaspar TFBS databases, the "seed" words for MSWLs exhibited 10–20 fold enrichment with putative TFBS in comparison with all 5-mer words within the given regulatory region: thus, for the most part, these "seed" words turned out to be instances of known TFBS (results not shown here).
Results for exons
We repeated the fluffy tail test for randomly picked Drosophila exons, and found that the distribution of over-represented words of the original sequences did not differ statistically from those of their randomised versions (See Table 2). Note the absence of a "fluffy tail" in Figure 8 (left) and the lack of distinction in the cumulative distribution (Figure 8 right).
Thus we have established a statistical difference between exons and regulatory DNA. Next we compare regulatory DNA with non-coding non-regulatory DNA.
Results for non-coding, presumed non-regulatory DNA
The similar words distribution for non-coding non-regulatory DNA typically shows two patterns: (1) without significant tails, as for exons and (2) with significant tails (Figure 9) but in this case – and in contrast to the regulatory sequences – the spatial locations of over-represented words are typically clustered (Figure 11c).
To deal with this, we developed a measure of spatial clustering of similar words. We say that two words w 1 and w 2 belong to the same cluster, if their genomic start positions s 1 and s 2 satisfy |s 1 - s 2| ≤ m·k , where m is the word length, and k is a constant. We examined the following choices for k: 1; 1.5; 2; 2.5; 3.
The size of a cluster is defined as the number of words in the cluster. For each MSWL we computed the coefficient of variation (CV) in cluster sizes, where CV is the standard deviation divided by the mean cluster size. We used analysis of variance to test for difference in coefficients of variance among four types of functional DNA: exons, non-fluffy NCNR, fluffy NCNR and regulatory regions. The assumptions for ANOVA (homogeneity of variance (CV), no correlation between means and standard deviations of the samples) were satisfied. The results show a strongly significant difference between the four types: see Figure 10. Thus we can use the cluster size CV to distinguish fluffy NCNR from regulatory DNA. CVs for fluffy NCNR are almost always more than 1, for k from 1 to 3; and significantly different from CVs for regulatory DNA.
We found that large clusters of adjacent over-represented words in fluffy NCNR DNA disappear after repeat-masking [20], thus revealing their identity as non-perfect simple repeats (Figure 11: compare panels a,b,c with d,e,f).
For details about spatial clustering and illustration of coefficient of variation robustness to choice of k and m, see Supplementary Materials [see Additional files 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12].
Discussion
Our method allows us to distinguish regulatory DNA from other non-regulatory DNA. In effect, our method aggregates many small signals contained in the region, and makes an internal comparison with background, represented by shuffled sequences.
We would like to extend the application of our method to larger sets of experimentally verified regulatory regions, from Drosophila or any other species. Unfortunately, few experimentally (not computationally!) verified sets are available. We managed to extended our positive training set a little, including a few experimentally verified regulatory regions from human, chicken, sea urchin, fruit fly and yeast (see Supplementary Materials [see Additional files 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12]), but it is still not a lot.
We would also like to explore the correlation between the genomic positions of words in MSWL (most abundant words), and positions of known regulatory elements. This may allow us to utilise our method as a kind of motif discovery algorithm. Unfortunately, again, the lack of reliably annotated regulatory regions with regulatory elements makes this step difficult.
Phylogenetic foot-printing is an important and rapidly developing branch of motif discovery methodology. It would be very interesting to compare genomic positions of words in MSWL with conserved sequences from phylogenetic foot-printing analyses. This would reveal whether such words are conserved, and therefore of functional significance.
In a similar vein, we would like to compare the results of fluffiness analysis results across multiple species. We could then answer the question whether cross-species conserved regions have "fluffy" regulatory region properties, and thus infer their putative function.
We are keen to compare results of our fluffy-tail-analysis with the results of recognition methods based on description of known TFBS, such as in the works [6] and [4]. These authors [4] also analysed developmental genes of Drosophila melanogaster containing approximately the same clusters of transcription factors.
The work [18] is closely related to our study. However, it is likely that their method is unable to distinguish non-perfect simple tandem repeat sequences from truly regulatory DNA. We have implemented their method as far as we can understand it, and found out that their separation of positive (cis-regulatory modules) and negative (coding and non-coding non-regulatory DNA) training sets due to local words frequency seems to be less clear than our separation due to "fluffiness" coefficient F (see Figure 6).
There might be possible other regulatory mechanisms apart from TFBS binding. It may be in some specific cases that the 3D local structure of DNA in the nucleus (chromatin) is the principal factor of gene expression and modulating regulatory modules play little or no role [21]. Thus one of the next steps in our work will be the incorporation of nucleosome position information.
Conclusion
We present a novel statistical approach that allows regulatory DNA to be distinguished from coding and non-coding non-regulatory regions according to its "fluffiness" values. This method is based on the presence of unusually high number of short runs of over-represented scattered words in the given DNA sequence.
The performance of the method on experimentally verified sequence data shows that the method allows us to predict whether a sequence may be regulatory.
Methods
Description of fluffy tail test
The fluffy tail test essentially consists of the comparison of similar word distributions for the original sequence and for a number of shuffled versions of the original sequences. These shuffled sequences clearly have the same length and single nucleotide composition as the original one.
To construct a similar words distribution one can perform the following two steps:
(1). First, obtain the distribution of similar words for a given DNA stretch (as described in detail below under "Distribution of similar words"). Then randomise the original sequence many times, and obtain a distribution of similar words for each shuffled sequence. These randomised sequences represent the null model (or the background model). The distributions of similar words obtained for the randomised sequences are compared with the corresponding distribution for the original sequence. If there are no statistical differences, we conclude that the sequence probably is an exon (related results are in [22]) or a homogeneous non-coding non-regulatory region.
However, if the given sequence does contain many similar words, these will show up in its distribution as a longer right tail that may even have a second mode. Such "fluffy" tails are seldom found in the distributions of the shuffled sequences, therefore suggesting the sequence is not exonic or homogeneous non-coding, non-regulatory DNA.
(2). To rule out "fluffy" tails due to non perfect simple tandem repeats, we check whether a) the similar words are spatially clustered and b) if the tails disappear after repeat-masking the sequence (using the on-line tool available at [20]) then repeating procedure (1).
Distribution of similar words
We considered 5-mer words, allowing for 1 mismatch. However, our results also hold for words of length 4 through 12, allowing for 1 through 4 mismatches (see Supplementary materials [see Additional files 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12]). Thus, for each 5-mer word in each of the 180 sequences (60 sequences in each training set) we computed the number n of similar words of the same length. Each word is the "seed" for a list of similar words.
As an example, consider a stretch of DNA :
accgggtgtaaaccgacctgatacccggtcgcccggttttaac...
The first "seed" 5-word 'accgg' forms the following list of similar words:
accgg, accga, acctg, acccg, cccgg,
which we have underscored in the above sequence.
The second 5-word 'ccggg' forms another list of similar words:
ccggg, ccggt, ccggt
etc. The first 5-word has the longest list of similar words here. The lists may intersect: e.g. the list for the 'accga'- seed word contains some words from the 'accgg'-seed word list.
Authors' contributions
WRG contributed to development of methodology, KW did numerical comparison with other related methods, RtB statistically processed the data, IA contributed to development of methodology, collected the data and wrote the software. All authors read and approved the final manuscript
Supplementary Material
Additional File 1
Contains short introduction and notation for Supplementary Material
Click here for file
Additional File 2
Contains Supplementary Table1 with results of Fluffy-tail test and Coefficients of Variation for some more experimentally verified regulatory regions for other than Fruit fly species.
Click here for file
Additional File 3
Contains a visual example of F dependence on the number of randomisations r.
Click here for file
Additional File 4
Gives some more details about spatial clustering threshold
Click here for file
Additional File 5
Shows some examples for consistence of fluffiness for different word length, tables.
Click here for file
Additional File 6
Shows some examples for consistence of fluffiness for different word length in the histogram form
Click here for file
Additional File 7
Consistent fluffiness and coefficient of variation for spatial cluster size for some example sequences.
Click here for file
Additional File 8
Contains the Figures showing fluffiness and spatial clustering of similar words for NCNR 3L4 region.
Click here for file
Additional File 9
Contains the Figures showing fluffiness and spatial clustering of similar words for NCNR repeat-masked 3L4 region.
Click here for file
Additional File 10
Contains the Figures showing fluffiness and spatial clustering of similar words for knirps regulatory region
Click here for file
Additional File 11
Contains the Figures showing fluffiness and spatial clustering of similar words for abdominantA regulatory region.
Click here for file
Additional File 12
Contains the Figures showing fluffiness and spatial clustering of similar words for internal exon 2r4.
Click here for file
Acknowledgements
We would like to acknowledge Yvonne Edwards, Tanya Vavouri, Adam Woolfe, Krys Kelly, Gayle McEwen, Greg Elgar, Carlo Berzuini, Tom Nye, Lorenz Wernisch, and Kenneth Evans for valuable discussions and support.
Figures and Tables
Figure 1 Histogram of similar words for the knirps cis-regulatory module. An example of a distribution of similar 5-mer words for the knirps cis-regulatory module Drosophila melanogaster . Note that the sequence contains an exceptionally large number (37) of lists with an exceptionally large number (137) of similar words. The Y axis shows the number of lists, the X axis is for list size.
Figure 2 Histogram of similar words for the knirps cis-regulatory module, after shuffling. The frequency distribution of similar words for one randomly shuffled version of the knirps cis-regulatory region, Drosophila melanogaster . The Y axis shows the number of lists, the X axis is for list size.
Figure 3 Cumulative histograms. Cumulative histograms for the data in Figures 1 and 2: solid line: original data from Figure 1, dotted line: randomised data from Figure 2. The X axis shows the size of lists of similar words, the Y axis is the number of lists.
Figure 4 Fluffy-tailed knirps distribution. (Left) The distribution of the original regulatory knirps sequence: (solid line); the distribution of 10 randomised sequences (dotted lines). (Right) The same distributions in cumulative form. The X axis shows the size of lists of similar words, the Y axis is the number of lists.
Figure 5 Histograms for regulatory (green), coding (cyan) and NCNR (magenta) sequences. The word length is 5, mismatch is 1, r is 50. The X axis shows the fluffiness coefficient F, the Y axis is the number of sequences in the set with this F.
Figure 6 Separation of regulatory DNA. Separation of regulatory DNA (column 2) from coding (column 1) and non-coding, non-regulatory (column 3) due to the fluffiness coefficient F (Y-axis). Box-plot of the Fluffiness (Y-axis) index for the three functional regions.
Figure 7 Spatial distribution of similar words in MSW L. Fairly uniform spatial distribution of start locations for words in the MSWL (n = 137, see Fig.1) of the knirps cis- regulatory region of Drosophila melanogaster . The X axis shows the positions of each word start in the sequence, the Y axis is the rank of this position in the list.
Figure 8 Histogram for exon cg3201 3. Distribution of similar words for the exon cg3201 3 of Drosophil a (solid line) compared to the histograms of the randomly shuffled versions (dotted lines) in direct (left) and cumulative (right) forms. The X axis shows the size of lists of similar words, the Y axis is the number of lists.
Figure 9 Histogram for non-coding presumed non-regulatory sequence. Distribution of similar words for a non-coding, non-regulatory sequence, randomly picked from chromosome 3L has significant tail because of simple repeats. The X axis shows the size of lists of similar words, the Y axis is the number of lists.
Figure 10 Coefficient of variation in spatial cluster size for four types of DNA: exons (1), non-fluffy NCNR (2), fluffy NCNR (3), regulatory regions (4); Vertical bars denote 95% confidence intervals. The Y axis shows coefficient of variation, the X axis is for four DNA type. We calculated CV based on spatial clustering coefficient k = 1.
Figure 11 Non-coding presumed non-regulatory sequence before and after repeat-masking. For a non-coding, non-regulatory sequence, randomly picked from chromosome 3L. Panels (a,b,c) show results before repeat-masking; panels (d,e,f) show results after repeat-masking. Panels (a,d) show histograms of similar words (solid: original data; dotted: after random shuffling) as in Figure 1; panels (b,e) show the same data in cumulative form as in Figure 3; panels (c,f) show start locations of similar words as in Figure 7.
Table 1 Sensitivity of F to choice of r, the number of randomisations, for the knirps regulatory region.
r F σr
25 14.7 5.39
50 8.65 8.77
100 10.22 7.56
Table 2 "Fluffiness" predictions for three types of functional region, showing the number of fluffy (F>2) sequences, the number of non-fluffy (F<2) sequences and corresponding positive and negative prediction rates, for each type of the region.
Functional type Fluffy tails (F>2) No fluffy tails (F<2) Positive rate Negative rate
Regulatory regions 51 9 85 % 15 %
Exons 1 59 1.6 % 98.4 %
Non-coding presumed non- regulatory 10 50 16 % 84 %
==== Refs
Yuh C Bolouri H Davidson EH Genomic cis-regulatory logic: functional analysis and computational model of a sea urchin gene control system Science 1998 279 1896 902 9506933 10.1126/science.279.5358.1896
Yuh C Bolouri H Davidson EH Cis-regulatory logic in the endo 16 gene: switching from a specification to a differentiation mode of control Development 2001 128 617 29 11171388
Davidson EH Genomic Regulatory Systems 2001 Academic Press
Berman B Nibu Y Pfeiffer B Tomancak B Celniker S Rubin G Levine M Eisen M Exploiting TFBS clustering to identify CRM involved in pattern formation in Drosophila genome PNAS 2002 99 757 62 11805330 10.1073/pnas.231608898
Wagner A A computational genomics approach to the identification of gene networks Nucleic Acids Research 1997 25 3594 604 9278479 10.1093/nar/25.18.3594
Markstein M Markstein P Markstein V Levine MS Genome-wide analysis of clustered Dorsal binding sites identifies putative target genes in the Drosophila embryo Proc Natl Acad Sci U S A 2002 99 763 68 11752406 10.1073/pnas.012591199
Johansson O Alkema W Wasserman WW Lagergren J Identification of functional lists of transcription factor binding motifs in genome sequences: the MSCAN algorithm Bioinformatics 2003 19 I169 I176 12855453 10.1093/bioinformatics/btg1021
Lifanov AP Makeev VJ Nazina AG Papatsenko DA Homotypic regulatory lists in Drosophila Genome Res 2003 13 579 88 12670999 10.1101/gr.668403
Rajewsky N Vergassola M Gaul U Siggia ED Computational detection of genomic cis-regulatory modules applied to body patterning in the early Drosophila embryo BMC Bioinformatics 2002 3 30 8 12398796 10.1186/1471-2105-3-30
Duret L Bucher P Searching for regulatory elements in human non coding sequences Curr Opin Struct Biol 1997 7 399 406 9204283 10.1016/S0959-440X(97)80058-9
Blanchette M Schwikowski B Tompa M Algorithms for phylogenetic footprinting J Comput Bio 2002 2 11 23
Couronne O Poliakov A Bray N Ishkhanov T Ryaboy D Rubin E Pachter L Dubchak I Strategies and tools for whole-genome alignments Genome Res 2003 13 73 80 12529308 10.1101/gr.762503
Boffelli D McAuliffe J Ovcharenko D Lewis KD Ovcharenko I Pachter L Rubin EM Phylogenetic shadowing of primate sequences to find functional regions of the human genome Science 2002 299 1391 4 10.1126/science.1081331
Elnitski L Hardison RC Li J Yang S Kolbe D Eswara P Connor OMJ Schwartz S Miller W Chiaromonte F Distinguishing regulatory DNA from neutral sites Genome Res 2003 13 64 72 12529307 10.1101/gr.817703
Ohler U Harbeck S Niemann H Noth E Reese MG Interpolated Markov chains for eukaryotic promoter recognition Bioinformatics 1999 15 362 9 10366656 10.1093/bioinformatics/15.5.362
Ohler U Promoter prediction on a genomic scale-the Adh experience Genome Res 2000 10 539 42 10779494 10.1101/gr.10.4.539
Ohler U Niemann H Liao G Rubin GM Joint modelling of DNA sequence and physical properties to improve eukaryotic promoter recognition Bioinformatics 2001 17 S199 206 11473010
Nazina A Papatsenko D Statistical extraction of Drosophila cis-regulatory modules using exhaustive assessment of local word frequency BMC Bioinformatics 2003 4 65 78 14690551 10.1186/1471-2105-4-65
RepeatMasker
Ensembl Genome Browser
Audit B Vaillant C Arneodo A d'Aubenton-Carafa Y Thermes C Long-range correlations between DNA bending sites: relation to the structure and dynamics of nucleosomes J Mol Biol 2002 316 903 18 11884131 10.1006/jmbi.2001.5363
Orlov Y Potapov V Complexity: an internet resource for analysis of DNA sequence complexity Nucleic Acids Research 2004 32 W628 W633 on-line. 15215465
| 15857505 | PMC1127108 | CC BY | 2021-01-04 16:02:48 | no | BMC Bioinformatics. 2005 Apr 27; 6:109 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-109 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1121587781510.1186/1471-2105-6-112Research ArticleConfirmation of human protein interaction data by human expression data Hahn Andreas [email protected]ührer Jörg [email protected] Priti [email protected] Thomas [email protected] Max-Planck-Institute for Informatics, Stuhlsatzenhausweg 85, D-66123 Saarbrücken, Germany2005 6 5 2005 6 112 112 22 9 2004 6 5 2005 Copyright © 2005 Hahn 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
With microarray technology the expression of thousands of genes can be measured simultaneously. It is well known that the expression levels of genes of interacting proteins are correlated significantly more strongly in Saccharomyces cerevisiae than those of proteins that are not interacting. The objective of this work is to investigate whether this observation extends to the human genome.
Results
We investigated the quantitative relationship between expression levels of genes encoding interacting proteins and genes encoding random protein pairs. Therefore we studied 1369 interacting human protein pairs and human gene expression levels of 155 arrays. We were able to establish a statistically significantly higher correlation between the expression levels of genes whose proteins interact compared to random protein pairs. Additionally we were able to provide evidence that genes encoding proteins belonging to the same GO-class show correlated expression levels.
Conclusion
This finding is concurrent with the naive hypothesis that the scales of production of interacting proteins are linked because an efficient interaction demands that involved proteins are available to some degree. The goal of further research in this field will be to understand the biological mechanisms behind this observation.
==== Body
Background
Gene expression data [1-3] and protein interaction data [4] are two types of data produced in the bioinformatics field. We investigated whether human gene expression levels of interacting protein pairs show a higher degree of dependence than those of random protein pairs. To date, such studies have only been performed in lower organisms like S. cerevisiae [5,6], in a comparative study using bacteriophage T7 and S. cerevisiae [7], and in C. elegans [8].
The first global evidence that genes with similar expression profiles are likely to encode interacting proteins has been provided in a study on S. cerevisiae by Ge et al. [5]. They compared the probability of interaction between proteins encoded by genes that belong to common expression profiling clusters with the probability of interaction between proteins encoded by genes that belong to different clusters. They found that proteins from the intra-group genes are more than five times as likely to interact with each other as proteins from the inter-group genes.
Tornow et al. [6] used superparamagnetic clustering to integrate protein interaction and expression data from independent experiments in S. cerevisiae and revealed hypothetical functional protein modules. Grigoriev [7] demonstrated the similarity of expression patterns for a pair of genes and interaction of the proteins they encode for both the bacteriophage T7 and in S. cerevisiae. He found the mean correlation coefficients of gene expression profiles between interacting proteins to be significantly higher than those between random protein pairs.
Recently Li et al. [8] analysed the transcriptome and interactome data of C. elegans and discovered that the correlation is lower than expected from observations in yeast.
A study by Jansen et al. [9] links gene expression on a genomic scale with protein-protein interaction in S. cerevisiae. They showed that while the subunits of the permanent protein complexes do indeed share significant correlation in their RNA expression, the correlated expression is relatively poor in detecting transient interactions. In a comprehensive study about S. cerevisiae conducted by Kemmeren et al. [10], up to 71% of the biologically verified interactions could be validated with the gene co-expression approach. Integration of expression and interaction data is thus a way to improve the confidence of protein-protein interaction data generated by high-throughput technologies. Kemmeren et al. [11] see enormous challenges in large genomes (orders of magnitude larger than S. cerevisiae) because of poor annotation, non-standardised gene names, and more complex interactions with the environment.
Results
Expression levels of genes encoding interacting proteins are correlated more strongly
Using five publicly available human expression datasets (Table 1) and 1369 human interacting protein pairs we compared the correlation of expression of genes encoding interacting proteins (empirical distribution) with the correlation of random protein pairs (background distribution). Figure 1 shows that the distribution of empirical correlations is slightly shifted to the right compared to the distribution of correlations in the case of random protein pairs. This result implies that in our data interacting proteins are preferentially encoded by coregulated genes.
Using mutual information as a measure of dependence the shift observed in figure 1 has almost vanished (figure 2). The difference between empirical and background distribution of the medians is 0.04 in the correlation case and <0.01 in the mutual information case. This observation suggests that correlation as a measure of dependence is more suitable than mutual information when analyzing dependencies between expression levels of interacting proteins. In the Methods section we give a possible interpretation of this observation.
Using each dataset separately we tested the hypothesis that correlation between expression of genes encoding interacting protein pairs is not higher than correlation between expression of genes encoding random protein pairs. The detailed algorithms are given in the Methods section. For four out of five analysed datasets this hypothesis is rejected at a significance level α = 0.05. This means that in these four datasets the correlation of expression levels of genes which encode interacting proteins is statistically significantly higher than the correlation of expression levels in genes which encode random pairs of proteins.
Increased p-values by use of mutual information instead of correlation
Using mutual information instead of correlation as a measure of dependence between gene expression levels leads to increased p-values for each of the five datasets. Thus the significance results of the analysis with correlation as dependence measure do not hold when using mutual information as dependence measure. This may be caused by the fact that most dependencies between expression levels are linear or close to linear and not parabolic which would preferentially be discovered by the mutual information measure. Because the correlation coefficient seems to be the more appropriate measure of dependence for this analysis we do not discuss mutual information in the following.
Expression of genes involved in different biological processes
Assigning GO-classes to the 1369 protein interactions as described in the Methods section below, for each dataset we analysed the expression levels of genes encoding interacting protein pairs both belonging to the same GO-class. Each of the figures 3, 4, 5, 6, 7 contains the box-and-whisker plots and the p-values of the twenty GO-classes that yield the most significant results for the respective dataset and of the GO-class biological process. Our method is different from the methods used by the authors that generated the datasets that we analysed. Thus our results cannot be compared directly with theirs. However, we feel that it is still useful to point out the similarity of some of our findings with the observations of these authors.
In the following we refer to the different datasets by the number they have received in table 1.
Figure 3 shows that in dataset 1 mainly genes included in the GO-classes cell cycle, cell growth, and cell proliferation are highly correlated. All in all only few GO-classes show low p-values which, in principle, agrees with the observations of Chi et al. [12] that siRNA-mediated gene silencing leads to only small variations in the gene expression pattern.
In figure 4, describing our results of the second expression dataset of Higgins et al. [13], many genes included in immune response and inflammatory response like interleukins, chemokine receptors, and chemokine ligands are highly correlated. This is an indication that the expression of chemokines and chemokine receptors is spatially and temporally restricted not only in the developing human kidney as reported in Gröne et al. [14] but also in the fully developed kidney.
Figure 5 shows our results of the GO-expression analysis of the third dataset. In this dataset Pathan et al. [15] found genes that are involved in bacterial infection to be significantly upregulated in blood after exposure to meningococci. We found the expression of genes that are involved in the response to (pathogenic) bacteria to be highly correlated which is in concurrence with the findings of Pathan et al. [15].
Zhang et al. [16] analysed the changes in transcript abundance occurring during senescence in human fibroblasts, as compared with early passage proliferating cells or quiescent cells. Figure 6 shows the results of our analysis of their expression dataset. In agreement with their findings we observed a strong correlation of genes that relate to apoptosis and genes that relate to transcription, but in contrast to them, we could not find significant correlation of genes that are involved in the cell cycle regulation.
Zhao et al. [17] analysed the effects of methylseleninic acid on the transcriptional program of human prostate cancer cells. Corresponding to their observation of decreased expression of genes involved in all phases of the cell cycle lines that do not express androgen receptor protein we found significant correlation of genes involved in M phase, nuclear division, and mitosis. Figure 7 displays the correlations and p-values of the top twenty GO-classes we calculated for the fifth dataset. In consensus with the expectations genes encoding proteins that are involved in the cell cycle show the lowest p-values in our analysis.
Discussion
This study investigates the relationship between two biological phenomena – gene expression and protein-protein interaction in H. sapiens – based on experimental data available in public databases. The study was prompted by the fact that in yeast and other lower non-mammalian organisms correlation is observed between expression levels of genes encoding interacting proteins. We were able to obtain convincing evidence of correlation using the Pearson's correlation coefficient but could not confirm these results when taking the mutual information as a measure of dependence. Using information on the GO-class to which both proteins of an interacting protein pair belong, we were able to find significant correlations of expression levels mostly in accordance to existing knowledge.
The results of our investigation lend additional credibility to the protein-protein interaction data used.
Once more interaction data are available, an analysis of the type presented here should be repeated including information on domains and phenotypes. For instance, one of the remaining open questions is whether the correlation of expression vectors of genes encoding interacting proteins with certain compared to random combinations of domains is statistically significantly different.
Larger interaction datasets will also provide the opportunity to analyse the question if genes encoding interacting proteins are located on the same chromosome or even in close neighbourhood to each other more often than expected when assuming a random order. This problem has been addressed recently by Hurst et al. [18].
Conclusion
In this study we observed a statistically significant correlation between expression of genes encoding interacting proteins in H. sapiens. This finding points towards a biological mechanism which coregulates the expression of such genes. Additionally it confirms the relevance of using gene expression data and interaction data in human genome analysis.
Methods
Gene expression data
For our study we used public datasets from the Stanford Microarray Database (SMD) [19]. This database includes much actual expression data from the same (cDNA microarray) platform, which is an important prerequisite for a well-founded analysis [20]. Datasets were selected by the following criteria:
• At least 20 000 clones per array
• At least 20 arrays per dataset
• Equal sets of measured clones per dataset
• Publication not earlier than 2003.
The following datasets were included in our study: Chi et al. [12] (human kidney cells), Higgins et al. [13] (normal tissue of kidney), Pathan et al. [15] (infection of blood cells), Zhang et al. [16] (gene transcription occurring during replicative senescence in human fibroblasts and mammary epithelial cells), and Zhao et al. [17] (effects of methylseleninic acid on the transcriptional program of prostate cancer LNCaP (Lymph Node Carcinoma of the Prostate) cells). The number of arrays ranges from 21 to 42 and the number of measured clones from 31 736 to 43 196. As expression level we used the binary logarithm of the normalised ratio of gene signal (channel 2) and reference signal (channel 1).
Protein-protein interaction data
As protein interaction database we used DIP [21] listing protein pairs that are known to interact with each other, because DIP allows the user to select interactions based on their species of origin (e.g. human). Interaction here means that two amino acid chains were experimentally identified to bind to each other. In September 2004 the database comprised 1369 human protein-protein interaction pairs including 51 self-interactions. These self-interactions were excluded from the analysis because the corresponding gene expression levels (which are two identical vectors) always have correlation 1.
Matching gene and protein identifiers
In order to determine the expression levels of genes encoding proteins that interact we have to know which proteins are encoded by which genes. Thus we have to match gene identifiers with protein identifiers. Specifically, we matched UniGene cluster IDs [22] from the expression files of the SMD [19] with Swiss-Prot accession numbers [23] (e.g. sp:Q07812), with PIR accession numbers [24] (e.g. pir:A47538), and with NCBI sequence identification numbers [25] (e.g. gi:539664) of the DIP files. As 'translator' we used a file called 'Hs.data' from the NCBI website [26] which contains the mentioned identifiers and the corresponding UniGene cluster IDs. In order to limit runtime, we refrained from applying sophisticated selection methods [27] where multiple matching occurred, but considered the first hit at all times. By using this approach, for 87% of the interacting protein pairs the genes encoding these proteins can be determined. In cases where this procedure was not successful, we used information from the Harvester website [28]. By this combination of methods the proportion increases from 87% to 94%. For many of these protein interactions no expression data of the encoding genes are available. Depending on the number of genes measured in the five expression data sets for at least 43% (dataset 4) and for up to 72% (dataset 3) of the proteins the corresponding gene expression levels can be determined. For evaluating the amount of dependence between the expression levels of two genes encoding interacting proteins, the expression of both genes has to be measured. Disregarding self-interactions this is the case in at least 10% (dataset 4) and up to 47% (dataset 3) out of 1369 interactions.
Pearson's correlation coefficient
Let (X1, Y1), (X2, Y2),..., (Xn, Yn) be the n pairs of expression levels of two random variables X (expression of first protein) and Y (expression of second protein). We wish to measure the degree to which X and Y are linearly dependent as opposed to being independent. The correlation then is defined by
Mutual information
Mutual information measures the mutual dependence of two variables based on information theory.
Two random variables, X and Y, with probability distributions pX(x) and pY(y) and the joint distribution pXY(x, y) are statistically independent if
pXY(x, y) = pX(x)·pY(y). (2)
The mutual information
quantifies the degree of dependence of X and Y using the distance between the joint distribution and the distribution in case of total independence. The mutual information becomes large if X and Y contain the same information.
Calculating the degree of dependence between expression levels of genes encoding interacting proteins for all datasets
In our analysis we used expression vectors each containing the expression levels of one gene from all arrays of a dataset extracted from the SMD [19]. For each dataset we determined the correlation of vectors each containing the expression levels of genes encoding two interacting proteins. For each interaction the median of the resulting set of five (one for each dataset) correlation coefficients was calculated. We used a permutation approach (with 10 000 permutations) to compare the empirical correlation and mutual information with the corresponding background distributions. In each permutation step we held the expression levels X of one protein fixed and permuted the interaction partners encoded by genes with expression levels Y. Thus for each permutation we got a new interaction dataset with random protein pairs. For each of these datasets we calculated the correlation values and their median as before for the original dataset.
We repeated this procedure using mutual information as measure of dependence. The distributions of correlation and mutual information are shown in figure 1 and in figure 2, respectively.
Calculating p-values for each dataset
As before we determined the correlation of vectors for each dataset, each vector containing the expression levels of genes encoding two interacting proteins. We also calculated these values for the permuted datasets (1000 permutations). To get more specific results, we did not use the median of the correlations or mutual information values here but performed the permutation approach for each dataset separately.
Denote with nperm the number of permutations and with nhigh the number of correlation and mutual information values higher than those in the original dataset. Then the estimated p-values are given by
p = nhigh/nperm. (5)
The corresponding p-values are shown in table 2.
GO analysis in each dataset
To further elucidate dependencies between expression levels in the five datasets we analysed for each dataset if genes encoding proteins within different GO-classes representing biological processes have correlated expression levels. Therefore, using QuickGO [29] we determined for each GO-class describing a biological process which of the 1369 interacting pairs include proteins, both of which belong to the respective biological process. We used these sets of interacting protein pairs to find biological processes that include protein pairs encoded by genes with highly correlated gene expression levels. By the use of a permutation test we compared the correlations of protein pairs belonging to a certain GO-class with the correlations of protein pairs not belonging to that GO-class. Analogous to the case without differentiation between GO-classes we can apply equation (5) again to get a p-value for each GO-class in each dataset (figure 3, 4, 5, 6, 7). We did not perform a correction procedure for multiple testing because the tested GO-classes often include very similar or even identical sets of interactions. Essential in this analysis is the ranking of the GO-classes.
Authors' contributions
AH developed and implemented the method, ran the calculations and prepared a draft of the paper. The other authors contributed to the development of the method, the interpretation of the results, and the refinement of the paper.
Figures and Tables
Figure 1 Empirical and background distribution of correlation values. For each interaction pair and for each dataset we calculated the correlation of expression levels of genes encoding interacting protein pairs. The graph shows slightly higher correlation values in the datasets (empirical distribution) than the correlation in the case of random protein pairs (background distribution).
Figure 2 Distribution of mutual information. For each interaction pair and for each dataset we calculated the mutual information of expression levels of genes encoding interacting protein pairs. The graph shows empirical and background distribution to be very similar.
Figure 3 Correlations and p-values of the expression dataset from Chi. The diagram contains the contains the box-and-whisker plots and the p-values of the twenty GO-classes that yield the most significant results for the respective dataset and of the GO-class biological process. It shows for different GO-classes, how strongly the expression levels of genes that encode interacting proteins from this common GO-class are correlated. The GO-classes along the x-axis are ordered by the corresponding p-value. This p-value gives the probability to get the depicted correlation results using random interacting protein pairs from the respective GO-class. For comparison the GO-class 'biological process', which comprises all interaction pairs (except the self-interactions), has been added.
Figure 4 Correlations and p-values of the expression dataset from Higgins. Analogous to figure 3
Figure 5 Correlations and p-values of the expression dataset from Pathan. Analogous to figure 3
Figure 6 Correlations and p-values of the expression dataset from Zhang. Analogous to figure 3
Figure 7 Correlations and p-values of the expression dataset from Zhao. Analogous to figure 3
Table 1 Information on expression datasets. The study includes a total of 155 arrays from five datasets. Each dataset has been published not earlier than 2003 and includes at least 20 arrays and 30 000 spots.
Number of dataset Dataset Year Number of arrays Number of spots
1 Chi [12] 2003 27 43 196
2 Higgins [13] 2004 34 43 196
3 Pathan [15] 2004 42 37 632
4 Zhang [16] 2003 21 31 736
5 Zhao [17] 2004 31 43 196
Table 2 P-values using correlation and mutual information as measure of dependence. The p-values describe the probability of obtaining a higher correlation (third column) or mutual information (fourth column) than the one observed, assuming that the expression levels of genes encoding interacting proteins are independent. The values have been estimated using 1000 permutations.
Number of dataset Dataset p-value (correlation) p-value (mutual information)
1 Chi [12] 0.067 0.243
2 Higgins [13] 0.008 0.277
3 Pathan [15] 0.004 0.150
4 Zhang [16] 0.016 0.264
5 Zhao [17] 0.019 0.368
==== Refs
Brown PO Botstein D Exploring the new world of the genome with DNA microarrays Nat Genet 1999 21 33 7 9915498 10.1038/4462
Lockhart DJ Winzeler EA Genomics, gene expression and DNA arrays Nature 2000 405 827 36 10866209 10.1038/35015701
Young RA Biomedical discovery with DNA arrays Cell 2000 102 9 15 10929708 10.1016/S0092-8674(00)00005-2
Cho S Park SG Lee do H Park BC Protein-protein interaction networks: from interactions to networks J Biochem Mol Biol 2004 37 45 52 14761302
Ge H Liu Z Church GM Vidal M Correlation between transcriptome and interactome mapping data from Saccharomyces cerevisiae Nat Genet 2001 29 482 6 11694880 10.1038/ng776
Tornow S Mewes HW Functional modules by relating protein interaction networks and gene expression Nucleic Acids Res 2003 31 6283 9 14576317 10.1093/nar/gkg838
Grigoriev A A relationship between gene expression and protein interactions on the proteome scale: analysis of the bacteriophage T7 and the yeast Saccharomyces cerevisiae Nucleic Acids Res 2001 29 3513 9 11522820 10.1093/nar/29.17.3513
Li S Armstrong CM Bertin N Ge H Milstein S Boxem M Vidalain PO Han JD Chesneau A Hao T Goldberg DS Li N Martinez M Rual JF Lamesch P Xu L Tewari M Wong SL Zhang LV Berriz GF Jacotot L Vaglio P Reboul J Hirozane-Kishikawa T Li Q Gabel HW Elewa A Baumgartner B Rose DJ Yu H Bosak S Sequerra R Fraser A Mango SE Saxton WM Strome S Van Den Heuvel S Piano F Vandenhaute J Sardet C Gerstein M Doucette-Stamm L Gunsalus KC Harper JW Cusick ME Roth FP Hill DE Vidal M A map of the interactome network of the metazoan C. elegans Science 2004 303 540 3 14704431 10.1126/science.1091403
Jansen R Greenbaum D Gerstein M Relating whole-genome expression data with protein-protein interactions Genome Res 2002 12 37 46 11779829 10.1101/gr.205602
Kemmeren P van Berkum NL Vilo J Bijma T Donders R Brazma A Holstege FC Protein interaction verification and functional annotation by integrated analysis of genome-scale data Mol Cell 2002 9 1133 43 12049748 10.1016/S1097-2765(02)00531-2
Kemmeren P Holstege FC Integrating functional genomics data Biochem Soc Trans 2003 31 1484 7 14641095
Chi JT Chang HY Wang NN Chang DS Dunphy N Brown PO Genomewide view of gene silencing by small interfering RNAs Proc Natl Acad Sci U S A 2003 100 6343 6 12730368 10.1073/pnas.1037853100
Higgins JP Wang L Kambham N Montgomery K Mason V Vogelmann SU Lemley KV Brown PO Brooks JD van de Rijn M Gene expression in the normal adult human kidney assessed by complementary DNA microarray Mol Biol Cell 2004 15 649 56 14657249 10.1091/mbc.E03-06-0432
Grone HJ Cohen CD Grone E Schmidt C Kretzler M Schlondorff D Nelson PJ Spatial and temporally restricted expression of chemokines and chemokine receptors in the developing human kidney J Am Soc Nephrol 2002 13 957 67 11912255
Pathan N Hemingway CA Alizadeh AA Stephens AC Boldrick JC Oragui EE McCabe C Welch SB Whitney A O'Gara P Nadel S Relman DA Harding SE Levin M Role of interleukin 6 in myocardial dysfunction of meningococcal septic shock Lancet 2004 363 203 9 14738793 10.1016/S0140-6736(03)15326-3
Zhang H Pan KH Cohen SN Senescence-specific gene expression fingerprints reveal cell-type-dependent physical clustering of up-regulated chromosomal loci Proc Natl Acad Sci U S A 2003 100 3251 6 12626749 10.1073/pnas.2627983100
Zhao H Whitfield ML Xu T Botstein D Brooks JD Diverse effects of methylseleninic acid on the transcriptional program of human prostate cancer cells Mol Biol Cell 2004 15 506 19 14617803 10.1091/mbc.E03-07-0501
Hurst LD Pal C Lercher MJ The evolutionary dynamics of eukaryotic gene order Nat Rev Genet 2004 5 299 310 15131653 10.1038/nrg1319
Sherlock G Hernandez-Boussard T Kasarskis A Binkley G Matese JC Dwight SS Kaloper M Weng S Jin H Ball CA Eisen MB Spellman PT The Stanford Microarray Database Nucleic Acids Res 2001 29 152 5 5 11125075 10.1093/nar/29.1.152
Mah N Thelin A Lu T Nikolaus S Kuhbacher T Gurbuz Y Eickhoff H Kloppel G Lehrach H Mellgard B Costello CM Schreiber S A comparison of oligonucleotide and cDNA-based microarray systems Physiol Genomics 2004 16 361 70 14645736 10.1152/physiolgenomics.00080.2003
Salwinski L Miller CS Smith AJ Pettit FK Bowie JU Eisenberg D The Database of Interacting Proteins: 2004 update Nucleic Acids Res 2004 D449 51 14681454 10.1093/nar/gkh086
UniGene
The EMBL Nucleotide Sequence Database: User Manual Release
PIR FAQ
BioinformaticSequence Identifiers: GI number and Accession.Version
Index of ftp://ftp.ncbi.nih.gov/repository/UniGene/
Rahnenführer J Domingues FS Maydt J Lengauer T Calculating the statistical significance of changes in pathway activity from gene expression data Stat Appl Genet Mol Biol 2004 3 Article 16
Bioinformatic Harvester EMBL Heidelberg
QuickGO: GO Browser
| 15877815 | PMC1127109 | CC BY | 2021-01-04 16:02:50 | no | BMC Bioinformatics. 2005 May 6; 6:112 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-112 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-911581997610.1186/1471-2105-6-91SoftwareA restraint molecular dynamics and simulated annealing approach for protein homology modeling utilizing mean angles Möglich Andreas [email protected] Daniel [email protected] Till [email protected] Wolfram [email protected] Hans Robert [email protected] Institut für Biophysik und physikalische Biochemie, Universität Regensburg, Universitätsstr. 31, D-93053 Regensburg, Germany2 Department of Biophysical Chemistry, Biozentrum, University of Basel, Klingelbergstr. 70, CH-4056 Basel, Switzerland3 Institut für Organische Chemie und Biochemie, Technische Universität München, Lichtenbergstr. 4, D-85747 Garching, Germany4 Department of Lead Discovery, Boehringer Ingelheim Pharma GmbH, Birkendorfer Str. 65, D-88397 Biberach, Germany2005 8 4 2005 6 91 91 1 9 2004 8 4 2005 Copyright © 2005 Möglich 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 have developed the program PERMOL for semi-automated homology modeling of proteins. It is based on restrained molecular dynamics using a simulated annealing protocol in torsion angle space. As main restraints defining the optimal local geometry of the structure weighted mean dihedral angles and their standard deviations are used which are calculated with an algorithm described earlier by Döker et al. (1999, BBRC, 257, 348–350). The overall long-range contacts are established via a small number of distance restraints between atoms involved in hydrogen bonds and backbone atoms of conserved residues. Employing the restraints generated by PERMOL three-dimensional structures are obtained using standard molecular dynamics programs such as DYANA or CNS.
Results
To test this modeling approach it has been used for predicting the structure of the histidine-containing phosphocarrier protein HPr from E. coli and the structure of the human peroxisome proliferator activated receptor γ (Ppar γ). The divergence between the modeled HPr and the previously determined X-ray structure was comparable to the divergence between the X-ray structure and the published NMR structure. The modeled structure of Ppar γ was also very close to the previously solved X-ray structure with an RMSD of 0.262 nm for the backbone atoms.
Conclusion
In summary, we present a new method for homology modeling capable of producing high-quality structure models. An advantage of the method is that it can be used in combination with incomplete NMR data to obtain reasonable structure models in accordance with the experimental data.
==== Body
Background
Due to the enormous progress that has been made in genomics a large number of DNA sequences including many whole genomes have been published. The evaluation of these data must include the determination of the three-dimensional structures of the proteins encoded. Although the two experimental techniques capable of determining three-dimensional structures of proteins and other biomolecules at atomic resolution, namely nuclear magnetic resonance (NMR) and X-ray crystallography, have seen significant improvements the process of structure determination remains very time-consuming and difficult. Unless unexpected advances of these techniques will occur in future, it is obvious that for the majority of all the primary sequence data available three-dimensional structures cannot be obtained experimentally. Therefore, only computational approaches are capable of filling the gap between existing protein sequences and structures. Although considerable progress has been achieved in ab initio structural prediction strategies [1-3] they are in general still unreliable when atomic resolution is demanded. However, when structures of homologous proteins are available, the prediction of the three-dimensional structure of entire proteins and protein domains is rather successful. In light of the fact that the protein structures elucidated so far only show a remarkably limited number of folds it would be desirable to accelerate the structure determination process especially for proteins possessing a fold already known. According to the SCOP classification [4,5] (release 1.65, 1 August 2003) 20619 protein structures stored in the Protein Data Bank share only 800 different folds. Comparison of different proteins with similar amino acid sequences showed that they quite often display very similar tertiary structures [6-9].
In the past several different homology modeling approaches were published which range from strongly interactive methods (model building) to fully automated methods (for reviews see e. g. [10] and [11]). Generally the starting point in these approaches is a search in structure databases such as the Protein Data Bank [12] or CATH [13] for all protein structures that are related to the target sequence and then to select those 3D structures that will be used as templates. For searching the structural databases one can employ pairwise sequence-sequence comparisons using for example programs such as FASTA [14] and BLAST [15]. When increased search sensitivity or a larger number of homologs are demanded methods which are based on multiple sequence alignments prove to be particularly efficient. Such an algorithm is implemented in the program PSI-BLAST [16]. An alternative strategy for homolog identification relies on so-called threading methods, which predict whether the target sequence adopts any of the known 3D folds. Threading methods should be useful in cases when no sequences can be found which are clearly related to the target [17].
When a list of related protein structures has been obtained the appropriate templates have to be chosen from these. In this procedure usually factors such as high overall sequence similarity between target and template sequences, quality of the template structure and conditions under which the template structure was obtained are taken into account. Then the selected templates have to be optimally aligned with the target sequence. Since the search methods mentioned above are usually optimized for detecting remote homologs they are not optimal for target-template alignment. A program often used for the latter type of alignments is CLUSTALX [18], which is also used within PERMOL.
Using the template-target alignment a variety of methods has been published for 3D model building. The group of methods which were developed first and are still frequently used were modeling by rigid body assembly [19-21]. Another group of methods use segment matching [22-25]. In the third, most recent group of methods spatial restraints obtained from the template structures are used in distance geometry calculations or energy optimization procedures to obtain the target model [26-31].
The PERMOL approach described presently also uses spatial restraints but in contrast to most other programs mainly dihedral angle restraints as opposed to restraints derived from inter-atomic distances are employed. These restraints enter molecular dynamics calculations in torsion angle space. In the following we will describe this method in more detail and mark differences to existing programs that have been published before.
In ab initio molecular dynamics (MD) simulations in addition to the applied force field only information about the amino acid sequence of the protein in study enters the calculations. While for small molecules such methods show results that are in very good agreement with the experimental data they mostly fail for more complex molecules. On the other hand restrained molecular dynamics calculations based on simulated annealing protocols are routinely and successfully used for the determination of solution NMR structures – in that case strong experimental information is available. Especially effective with regard to computational effort are calculations in torsion angle space as implemented in the programs DYANA [32] and CNS [33].
In this contribution we propose a method which combines the well-developed torsion angle dynamics calculations of DYANA or CNS with structural information extracted from three-dimensional structures of homologous proteins. This information is translated into conformational restraints. Local structural restraints are obtained by a weighted average of the backbone dihedral angles using an algorithm proposed by Döker et al. [34] These averaged dihedral angle restraints are usually well preserved within the local secondary structure elements and therefore are especially well suited for the modeling of these. The program MODELLER [28] for example also uses dihedral angle restraints in an optimization procedure but expresses them as so-called probability density functions which are derived from structural features in several families of homologous proteins. Global structural restraints are obtained from distance relations between carefully selected atoms of amino acids well separated in the primary structure. In contrast to other programs the distance restraints are mainly used for the global arrangement of the secondary structure elements which are defined by the dihedral angle restraints. The efficient structure calculations performed with DYANA or CNS allow calculating a large number of structural models in a relatively short amount of time. From the resulting ensemble of structures the best in terms of the DYANA target function or total energy (CNS) can be selected for further analysis. As has also been shown in NMR spectroscopy, it is useful to describe the target structure by an ensemble of model structures.
It should be noted that the PERMOL approach described here is related to the method detailed by Zhang et al. [35], which uses a combination of torsion angle dynamics and dihedral angle and distance restraints to predict the fold of helical proteins. In contrast to PERMOL the program from Zhang et al. uses methods for secondary structure and contact prediction to derive spatial restraints.
To benchmark the PERMOL approach we used it to determine a homology structure for the histidine-containing phosphocarrier protein (HPr) from E. coli of which the structure has been solved experimentally both by NMR [36] (PDB entry: 1HDN) and X-ray crystallography [37] (1POH). The homology model was compared to the target structures and to a homology model calculated with the program MODELLER [28]. To also investigate the performance of PERMOL on larger proteins that contain substantial disordered loop regions the human peroxisome proliferator activated receptor γ (Ppar γ) was used as a test case. Its structure has been determined previously by X-ray crystallography [38] (3PRG).
Results
Theoretical considerations and general strategy
In standard NMR structure determination the principal physical model of a protein is represented by empirical potentials determining the general geometry. The fast optimization is obtained by a simulated annealing protocol and the correct conformations are selected from the generally accessible conformational space by the experimental restraints which are transformed into pseudo-potentials. In the approach used in PERMOL the experimental restraints are replaced by restraints derived from three-dimensional structures of homologous proteins. Local conformations are optimally encoded by the distribution of the corresponding torsion angles. The overall fold is determined by distance relations since even small errors in dihedral angles can add up to very large distance errors between amino acids that are separated by several positions in the sequence.
The use of a molecular dynamics and simulated annealing protocol for homology modeling allows to encode the features of the statistical distribution of a given parameter αi individually for each group of restraints. To this end not only the expectation values are calculated from the homologous structures j (j = 1,..,Ni) but also the upper and lower limits, and . It is still under discussion in the NMR community how exactly the upper and lower limits of restraints have to be defined but it is clear that they are related to the expected error of a given, individual parameter. A generally accepted definition is not available yet. In addition the form of the pseudo-energy function used in the calculations has to depend on the error distribution of the given parameter (see e. g. [39]).
The homology modeling procedure proposed here comprises the following steps: step 1, selection of data and sequence alignment, step 2, selection of restraints, and step 3, the restrained molecular dynamics simulation. These conceptually different steps in the calculation are reflected in the implementation of PERMOL in corresponding levels of the modeling procedure.
Level 1 – Selection of data and sequence alignment
Initially, one or several structures of homologous proteins are selected as templates. Their amino acid sequences are aligned to the sequence of the target protein using the program CLUSTALX [18]. The resulting alignment is written to a text file and can be edited by the user. Conserved amino acids are characterized and classified for manual or automated selection of restraints. Based on the degree of sequence conservation in the different proteins a homology score value vi is calculated for each residue. The score values vi range from 1.0 for a completely conserved residue to 0.1 for a residue, which in the template proteins has been replaced in a non-conservative manner, e.g. a hydrophobic residue replaced by a charged one.
Level 2 – Selection of restraints
For the calculation of dihedral angle restraints usually only Φ and Ψ angles are taken into account but the ω-angle can be included as well. Structural restraints are only derived from residues, which have been selected. Additional residues can be selected either manually or automatically based upon the score value vi. Expectation values and standard deviations are calculated as described in the 'Methods' section with set to 1/ when structures are found in the pdb-file k as it is often the case for NMR-structures. Upper and lower limits for the dihedral angle restraints can be calculated either as the mean value plus/minus multiples of the standard deviations, <αi> ± b* <si> with a user defined constant b, or as the mean angle plus/minus a constant value. An additional weighting of the individual restraints can be performed on the basis of the score value vi which modifies the force constant of the restraint i in the MD calculation.
By default, distance restraints are automatically computed between the NH atoms of completely conserved amino acids. Restraints can also be generated for additional amino acids and atom types by appropriate selection. For the generation of distance restraints similar options are possible as for dihedral angle restraints. In addition, an upper distance limit for the pairs of atoms to be considered can be defined.
Conserved hydrogen bonds can also be used to generate distance restraints between the atoms involved in forming the bond. The criteria for selecting hydrogen bonds in the homologous protein structures can be modified by the user. By default, only hydrogen bonds are considered for which the N-O distance does not exceed 0.24 nm and the angle between the NH-HN and the C = O bond vectors does not deviate by more than 35° from 180°. Again, different options are possible for the calculation of the upper and lower limits. Hydrogen bonds which occur only in a few structures or are assigned to more than one pair of atoms, e.g. due to deviations between the different homologous proteins used as templates, can be automatically removed by corresponding filter functions.
Level 3 – Restrained molecular dynamics simulation
The restraint files generated by PERMOL can be directly used by the molecular dynamics programs DYANA and CNS. Standard simulated annealing protocols are employed.
Modeling of HPr from E. coli and of human Ppar γ
To test the modeling approach described in this paper we determined a homology structure for the histidine-containing phosphocarrier protein (HPr) from E. coli. HPr is an integral part of the bacterial phosphoenolpyruvate dependent phosphotransferase system (PTS) which efficiently catalyses phosphorylation and the import of carbohydrates into prokaryotic cells [40]. HPr molecules from different organisms have been extensively studied and many 3D structures have been elucidated. In particular the structure of HPr from E. coli has been solved both by NMR [36] (PDB entry: 1HDN) and X-ray crystallography [37] (1POH) and is thus especially suited to test our modeling strategy (see Table 1).
Four previously determined HPr structures from four different organisms have been used as model structures (PDB codes 1PTF, 1QFR, 1QR5, and 2HID). An overview of these structures is given in Table 1. Only 21 % of the amino acid sequence is strictly conserved between the HPr proteins of E. coli, S. faecalis, E. faecalis, S. carnosus, and B. subtilis (18 out of 85 residues). Spatial restraints for the structure calculation were generated as detailed in the 'Methods' section. For the derivation of inter-atomic distance restraints only residues which are completely conserved or display conservative amino acid exchanges (e. g. one hydrophobic residue replaced by another one) were considered. Upper and lower limits for these distances were determined as the mean distance value plus or minus the standard deviation, respectively. Restraints for the backbone dihedral angles Φ and Ψ were calculated for all residues and have been weighted according to the homology score value vi. Upper and lower limits were determined as for the distance restraints. Hydrogen bonds were analyzed using the default parameter values. Distance restraints between the corresponding HN and O atoms were computed as the mean distance value plus or minus the standard deviation. A summary of these restraints is presented in Table 3.
Based on these restraints an ensemble of homology structures was computed using the molecular dynamics program DYANA [B32] with the standard simulated annealing protocol. Out of 200 structures calculated, the group of the ten structures with the lowest pseudo-energies was further analyzed. These ten models showed a good convergence with a RMSD value for the backbone atom positions of 0.041 nm (Fig. 1, Table 5). They displayed the well-known secondary structure elements common to all HPr molecules studied so far, comprising a four-stranded antiparallel β-sheet and three α-helices designated as helices a, b, and c. Analysis of the ensemble of these ten structures with PROCHECK-NMR [41] showed that all backbone dihedral angles fell into the most favored and additionally allowed regions of the Ramachandran plot (Table 5). Modeling experiments where the dihedral angle restraints have been partly or completely left out from the structure calculations of the model structures underlined their importance in defining the correct secondary structure and local conformations (see below).
In order to further test our modeling strategy we set out to derive a homology structure for the human peroxisome proliferator activated receptor γ (Ppar γ). Ppar γ is considerably larger than HPr and comprises about 280 amino acid residues. Further, it contains larger relatively unstructured loop regions and it is worthwhile to investigate how PERMOL performs here. In addition this molecule is of particular importance for us since we are currently in the process of experimentally solving its solution structure. Via a BLAST [16] search for the primary sequence of Ppar γ we identified several related proteins for which three-dimensional structures are available (Table 2), namely Ppar α [42-44] (PDB codes: 1K7L, 1KKQ, and 1I7G) and Ppar δ [45] (1GWX and 3GWX).
Model structures were calculated as detailed for HPr and out of 125 calculated structures the 16 structures with the lowest pseudo energies were further analyzed. A summary of the used restraints is given in Table 4. These sixteen models showed a good convergence with a RMSD value for the backbone atom positions of 0.135 nm (residues 206 – 477) (Fig. 2, Table 6). The secondary structure elements observed in the model structures agree well with the corresponding X-ray structure of the template protein, comprising a four-stranded antiparallel β-sheet and twelve α-helices (Fig. 2). Analysis of the ensemble of the selected sixteen structures with PROCHECK-NMR [41] showed that almost all backbone dihedral angles fell into the most favored and additionally allowed regions of the Ramachandran plot (Table 6).
Comparison to target structures
The ensemble of modeled HPr structures was compared to the target structure of HPr from E. coli which before had been elucidated using NMR spectroscopy (1HDN) and X-ray crystallography (1POH). For 1HDN a bundle of 30 structures was deposited in the protein database. As stated in the header of the coordinate file the first structure is closest to the ensemble average. As a consequence this structure was selected as the NMR target structure. A comparison between the modeled structure and the target NMR and X-ray structures is shown in Fig. 3. The homology model displayed the same global fold and distribution of secondary structure elements as both target structures. To quantify the agreement between the individual structures the root mean square deviations (RMSD) between the different structures were calculated for the backbone atom positions. While the RMSD between the two target structures 1HDN and 1POH amounted to 0.11 nm the comparison of the best modeled structure with the target NMR structure and the X-ray structure yielded RMSD values of 0.17 nm and 0.15 nm, respectively. Although the agreement between the modeled and the target structures was worse than the agreement between the two target structures, the RMSD values were of similar magnitude. Deviations between the homology model and the experimentally determined structures were mainly seen in the loop regions and in the orientation of helices a and b. Interestingly, these are also the regions that are least well defined in the X-ray and NMR structures and where these structures diverge most. In contrast, the core region of HPr and its overall fold are reproduced well in the homology model.
Further, we used R-factor analysis [46] to compare the modeled structure to the target structures. The quality of the protein backbone was specifically assessed by only taking into account spectral signals arising from backbone protons. Low R-factors of similar magnitude were obtained when comparing the modeled structure with either the NMR target structure (R-factor 0.093) or the X-ray target structure (0.076). Consistent with the RMSD values the R-factors also indicated that the homology structure more closely resembled the X-ray structure than the NMR structure. A slightly lower R-factor of 0.073 was obtained when comparing the two target structures with each other (Table 7).
For Ppar γ the best model structure in terms of pseudo-energy was compared to the target X-ray structure (3PRG). The agreement between the two structures was assessed by calculating the corresponding RMSD value for the backbone atoms, which amounted to 0.262 nm (Table 8). Note that the first five unstructured residues and the region between residues 262 and 274 which were missing in the X-ray target structure were not considered in this analysis. Deviations between the homology model and the X-ray structure were mainly seen in the loop regions and in the orientation of the helices preceding and following the unstructured region between residues 262 and 274. The agreement between model and X-ray structure was further analyzed by the calculation of pseudo NMR R-factors (Table 8). Although somewhat higher R-factors were obtained for Ppar γ than for HPr, the R-factor analysis still showed a reasonable agreement between model and X-ray structure.
Importance of torsion angles
In principle, torsion angles can completely define the 3D-structure of a protein when the general geometry of the amino acids is predefined. However, small errors of torsion angles in the backbone propagate and lead to large errors in the Cartesian space for amino acids remote in the sequence. Nevertheless, torsion angles are optimal predictors for local folding. Fig. 4 exemplifies the importance of the torsion angles for the structure predictions. As an example it shows a structure prediction (calculation) of HPr from S. faecalis from a rather small number of restraints created from the X-ray structure (1PTF) of the protein. Only 427 torsion angle restraints together with 41 hydrogen bond restraints can be sufficient to determine the various secondary structure elements together with the global fold of the molecule. Even the loop regions for which no hydrogen bond restraints are present adopt native-like conformations. Only the third α-helix is rotated away from the core of the protein since its orientation is solely defined by the angle restraints of residues 67–69.
Discussion
In this contribution we have presented a new program for homology modeling of protein structures. Using restraint molecular dynamics simulations together with spatial restraints derived from template structures we calculated homology structures of HPr from E. coli and of human Ppar γ. An advantage of the proposed method is the use of spatial restraints with individual upper and lower limits depending on the local structural conservation in the template structures. This becomes especially evident for the obtained bundle of Ppar γ model structures where one can easily distinguish between the mostly well-defined secondary structure elements and less ordered regions e.g. some of the larger loop regions.
At first glance it appears to be a disadvantage of the proposed method that not a unique, seemingly perfect structure is the result of the calculations as in the case of threading methods. However, the structure bundle produced by our approach gives an idea of the conformational subspace determined by the available experimental basis and the physical model. This is a safeguard against typical over-interpretations of model structures where data in badly predictable regions are used for the detailed interpretation of functional data or are used during the drug design process.
An additional advantage of the simulated annealing approach is that restraint violations are not treated explicitly but contribute to the overall "energy" which is minimized. In contrast to other methods in the approach used in PERMOL the mean torsion angles and their errors provide the main information. A few distance restraints are used to define the long-range relations which cannot be described sufficiently well by the local data. Accordingly, details of the selection of these restraints are not critical. Thus, the selection of pairwise restraints between all conserved residues seems to be plausible. The same is true for conserved hydrogen bonds. However, the PERMOL software also allows to define a custom selection of restraints and thus an adaptation to specific needs. As an example all hydrophobic contacts between amino acid residues observed in the template structures could be selected to serve as restraints. The automated calculation of individual weighting factors during the calculation of the expectation values and standard errors of the individual restraints would permit to introduce information about the local and global sequence conservation and the precision of the used structures. Currently, we are undertaking efforts to address this question. The high quality of the structure models generated with PERMOL illustrates that the same MD programs used for the determination of NMR structures can also be utilized for homology modeling. The programs and strategies developed for NMR structure determination have evolved to efficient optimizers even when only limited information (i. e. small number of structural restraints) is available. This has been recognized for example by Dominguez et al. [47] who use restrained molecular dynamics together with the ARIA protocol [48] for solving the docking problem. While in the case of NMR structure determination the restraints that enter the molecular dynamics simulation are derived from experimental observables like NOE cross-peaks, J-couplings, and residual dipolar couplings, in the case of homology modeling synthetic restraints are generated from previously determined structures of homologous template proteins. The use of standard MD programs and protocols also has a disadvantage since it is not possible to directly introduce properties in the calculation which are not provided for by the programs. An example would be the use of specific potential forms with multiple minima which describe the homology-derived information in more detail as it is done e. g. by MODELLER [28].
We compared the HPr homology structure we obtained with PERMOL to a structural model of HPr from E. coli calculated using MODELLER (version 6v2). When the same alignment file and template structures were used, homology models of similar quality were obtained with the two programs.
A specific advantage of the approach presented here is that it can be well used in the context of standard structure determination by NMR. The restraint files generated by PERMOL are editable and can be easily combined with other data and be adapted for use with different programs. As the same MD programs are used both for modeling with PERMOL and for NMR structure determination, incomplete experimental data can be conveniently combined with spatial restraints derived from homologous template proteins. The validity of the resulting structure models can be checked by calculating NMR R-factors [46]. Different force fields and annealing protocols which are available for the NMR MD programs can also be utilized for homology modeling. In this way recent advances like the structure refinement in explicit solvent [49,50] can be readily exploited to derive more accurate homology structures.
Conclusion
In summary, we have presented a new method for homology modeling capable of producing high-quality structure models. Compared to many other homology structure prediction programs it is based on a different philosophy since its aim is not to predict a unique best structure but a bundle of structures representing the locally different degrees of reliability of the structure prediction. Since the homology-derived restraints are mainly used to reduce the conformational space to be searched by the MD calculation, their relative importance for obtaining a correct homology model is expected to decrease in future time as the physical model employed in these calculations is improved. Another advantage of the approach described here is its flexibility, conveniently allowing several template structures to be included as sources of structural restraints. Furthermore, the PERMOL software permits to determine which kinds of structural restraints enter the molecular dynamics calculation in a controlled fashion. We demonstrated that the standard MD programs used in the course of structure determination by NMR can also be well utilized for the purpose of homology modeling. Prediction on the basis of averaged torsion angles is a powerful tool which efficiently makes use of the structural information available in the protein data base and leads to well-defined structures.
Recently, a homology model determined with PERMOL was used in the resonance assignment [51] and structure determination process of a mutant form of HPr from S. carnosus [52] and to obtain an initial estimate for the molecular alignment tensor describing the partial orientation of the HPr molecule in anisotropic solution [53,54]. PERMOL has also been integrated in the NMR structure determination package AUREMOL [39]. In this molecule-centered top-down approach one starts with a trial structure e.g. a homology model obtained by PERMOL that is iteratively refined until it fits the experimental data sufficiently as verified by the calculation of NMR R-factors.
Methods
Calculation of the restraints for simulated annealing
Structural information obtained from a set of homologous structures j (j = 1,..,Ni) must be expressed in form of restraints. The restraint of a parameter αi is usually defined by its expectation value and the upper and lower limits and , respectively. PERMOL offers several ways to calculate these quantities from the expectation values observed in the template proteins <αi> and the corresponding standard deviations si. For non-cyclic parameters <αi> and si can be simply calculated according to eqs. (1) and (2).
and
with the weighting factor for a given event i and the total number of events Ni. For cyclic parameters like dihedral angles, which are mainly used within PERMOL such a definition does not directly apply but can be extended by the approach described by Döker et al. (1999) [34].
Here, the origin of the coordinate system is shifted to fulfill the condition
and the standard deviation is calculated according to eq. (2). The expectation value is obtained by
The parameters determine the statistical weight of a given homology structure used to calculate a restraint. In principle, their value will depend on factors such as the local and global sequence conservation and the quality of a structure, e. g. when comparing X-ray and NMR-structures.
Implementation overview
In order to facilitate the determination of structural restraints for homology modeling the software package PERMOL was developed. PERMOL was written in Perl/Tk and has been tested with the operating systems SGI IRIX, Linux and Windows. The software and a detailed manual explaining its use can be obtained free of charge from the authors . Sequence alignment is done by using the program CLUSTALX [18]. Structure calculations are performed with output data files generated by PERMOL which can be imported by the molecular dynamics programs DYANA [32] and CNS [33]. Dihedral angles from different structures are averaged following the algorithm described by Döker et al. [34]. The typical computing time for setting up the restraint and parameter files for the MD-calculation is negligible using a modern PC. The calculation of the structures strongly depends on the MD-program used, the number of structures calculated and the actual simulated annealing protocol. In the examples presented here structures were calculated on a standard Linux-PC using the MD program DYANA. The corresponding calculation times for a single structure model were around 30 and 160 seconds for HPr and Ppar γ, respectively. Figures 1, 2, 3, and 4 have been prepared with MOLMOL and rendered with PovRay .
Validation of homology models
Modeled structures can be quantitatively compared to their respective target structures by calculating NMR R-factors according to [46]. Analogous to crystallography R-factors, NMR R-factors are used to quantify how well a three-dimensional structure accounts for the spectral signals occurring in an experimental NMR spectrum. Using an implementation of the complete relaxation matrix analysis (RELAX, [56,57]) artificial NMR spectra are calculated for the given three-dimensional structure and compared to the experimental spectra. R-factors quantify the deviations between the two types of spectra and are therefore a measure for the quality of the trial structure. In the case of perfectly matching spectra the R-factor adopts a value of 0. Analogous, R-factor analysis can also be employed to quantify the agreement between two protein structures. In that case artificial NMR spectra are calculated for both structures and are compared to each other.
The agreement between two structures can be further assessed by determining the root mean square deviations (RMSD) between the atom positions of the structures. The program MOLMOL [55] is used to fit the structures atop of each other and to calculate RMSD values. The stereo-chemical quality of the obtained models was validated using the program PROCHECK-NMR [41].
Abbreviations
HPr: histidine-containing phosphocarrier protein, MD: molecular dynamics, NMR: nuclear magnetic resonance, NOE: nuclear Overhauser effect, PDB: Protein Data Bank Brookhaven, Ppar γ: human peroxisome proliferator activated receptor γ, PTS: phosphoenolpyruvate carbohydrate phosphotransferase system, RMSD: root mean square deviation
Authors' contributions
TM, WG, and HRK conceived the project. DW and TM performed initial feasibility studies and refined the overall modeling strategy. AM wrote the PERMOL software and a manual. AM, DW, and WG calculated the homology structures. AM drafted the manuscript. WG and HRK coordinated the study and wrote the manuscript. All authors read and approved the final manuscript.
Acknowledgements
The authors thank Dipl. Phys. K. Brunner, Dr. R. Döker, Dr. W. Kremer and Dipl. Phys. J. Trenner for helpful discussions. Financial support by the European Commission (SPINE-project) is gratefully acknowledged.
Figures and Tables
Figure 1 Homology structures of HPr from E. coli determined by PERMOL. Ensemble of the 10 homology structures with the lowest pseudo-energy out of 200 structures calculated with DYANA. (left) A superimposition of the Cα atom traces is shown. (right) A cartoon representation of the mean structure of the 10 models is displayed.
Figure 2 Comparison of the model structure of Ppar γ from human with the corresponding X-ray structure. Overall good agreement between the bundle of final model structures (helices in red and yellow, β-strands in blue and loops in grey) and the X-ray structure (orange) is obtained. Deviations are mainly seen in larger loop regions, the unstructured N-terminus and at the C-terminal end.
Figure 3 Comparison of the model structure of HPr from E. coli with the corresponding X-ray and NMR structures. A comparison of the modeled HPr homology structure with the structures experimentally determined by NMR spectroscopy (1HDN) and X-ray crystallography (1POH). The structures are shown in the same orientation as in Fig. 1 with the radius of the backbone splines indicating the RMSD of the Cα atom positions in the respective structures. (A) Overall good agreement between the model structure (yellow) and the X-ray structure (blue) is obtained. Deviations are mainly seen in loop regions and in the orientation of helices a and b. RMSD values for the Cα atom positions of the X-ray structure 1POH have been derived from the crystallographic B-factors, fB, using the Debye-Waller equation where isotropic displacement from the mean atom positions was assumed. (B) Comparison of the model (yellow) and the NMR structure (red). Deviations are seen in the same regions as before. (C) X-ray (blue) and NMR (red) structures superimpose well. Interestingly, deviations between them are mainly observed in regions where the two structures also diverge from the homology model.
Figure 4 Importance of torsion angle restraints exemplified on HPr from Streptococcus faecalis. On the left hand side the model structure calculated with PERMOL using 427 torsion angle restraints and 41 hydrogen bonds is displayed, while on the right hand side the target X-ray structure 1PTF is shown. The RMSD value for the heavy atoms of the two structures is 0.328 nm. Restraints for torsion angles and hydrogen bonds were directly generated from the X-ray structure 1PTF.
Table 1 Statistics of PDB structure files used for HPr
PDB code Organism Method Resolution [nm]a Reference
1HDN E. coli NMR 0.20 [36]
1POH E. coli X-ray 0.20 [37]
1PTF S. faecalis X-ray 0.16 [58]
1QFR E. faecalis NMR 0.27 [59]
1QR5 S. carnosus NMR 0.28 [60]
2HID B. subtilis NMR 0.19 [61]
aThe equivalent resolution of the NMR structures was calculated using PROCHECK-NMR [41].
Table 2 Statistics of PDB structure files used for Ppar γ
PDB code Organism Method Resolution [nm] Reference
3PRG human X-ray 0.29 [38]
1K7L human X-ray 0.25 [42]
1KKQ human X-ray 0.30 [43]
1I7G human X-ray 0.22 [44]
1GWX human X-ray 0.25 [45]
3GWX human X-ray 0.24 [45]
Table 3 Restraints for molecular dynamics calculation for HPr
Type of restraint Number
inter-atomic distances 186
hydrogen bonds 50
backbone dihedral angles 164
Table 4 Restraints for molecular dynamics calculation for Ppar γ
Type of restraint Number
inter-atomic distances 1391
hydrogen bonds 153
backbone dihedral angles 528
Table 5 Structural statistics for HPr
RMSD values for the ten lowest-energy structures RMSD [nm]
backbone atoms Cα, C', N 0.041
heavy atoms 0.111
Residues in the Ramachandran plot Incidencea
most favored regions 87.2 %
additional allowed regions 12.8 %
generously allowed regions 0.0 %
disallowed regions 0.0 %
aThe dihedral angles have been analyzed using the program PROCHECK-NMR.
Table 6 Structural statistics or Ppar γ
RMSD values for the sixteen lowest-energy structures RMSD [nm]
backbone atoms Cα, C', N 0.135
heavy atoms 0.191
Residues in the Ramachandran plot Incidencea
most favored regions 84.1 %
additional allowed regions 14.3 %
generously allowed regions 1.4 %
disallowed regions 0.2 %
aThe dihedral angles have been analyzed using the program PROCHECK-NMR.
Table 7 Comparison between model structures and experimental structures for HPr
Structures Quantitiesa NMR target structure X-ray target structure
X-ray structure backbone RMSD [nm] 0.106 0
heavy atom RMSD [nm] 0.273 0
R-factor 0.073 0
best NMR structure backbone RMSD [nm] 0 0.106
heavy atom RMSD [nm] 0 0.273
R-factor 0 0.072
best model structure backbone RMSD [nm] 0.169 0.147
heavy atom RMSD [nm] 0.273 0.253
R-factor 0.093 0.076
model structure bundle backbone RMSD [nm] 0.178 0.154
heavy atom RMSD [nm] 0.277 0.258
R-factor 0.097 0.081
aBackbone RMSDs include NH, Cα, and C' atoms. Heavy atoms include all atoms except protons. RMSDs are pairwise RMSDs. R-factors are calculated using the R-factor R3 according to [46] including only signals arising from backbone protons.
Table 8 Comparison between model structures and experimental structures for Ppar γ
Structures Quantitiesa X-ray target structure
best model structure backbone RMSD [nm] 0.262
heavy atom RMSD [nm] 0.317
R-factor 0.260
model structure bundle backbone RMSD [nm] 0.299
heavy atom RMSD [nm] 0.355
R-factor 0.231
aBackbone RMSDs include NH, Cα, and C' atoms. Heavy atoms include all atoms except protons. RMSDs are pairwise RMSDs. R-factors are calculated using the R-factor R3 according to [46] including only signals arising from backbone protons.
==== Refs
Baker D A surprising simplicity to protein folding Nature 2000 405 39 42 10811210 10.1038/35011000
Bonneau R Baker D Ab initio protein structure prediction: progress and prospects Annu Rev Biophys Biomol Struct 2001 30 173 189 11340057 10.1146/annurev.biophys.30.1.173
Hardin C Pogorelov TV Luthey-Schulten Z Ab initio protein structure prediction Curr Opin Struct Biol 2002 12 176 181 11959494 10.1016/S0959-440X(02)00306-8
Murzin AG Brenner SE Hubbard TJP Chothia C SCOP: a structural classification of proteins database for the investigation of sequences and structures J Mol Biol 1995 247 536 540 7723011 10.1006/jmbi.1995.0159
Lo Conte L Brenner SE Hubbard TJP Chothia C Murzin AG SCOP database in 2002: refinements accommodate structural genomics Nucleic Acids Res 2002 30 264 267 11752311 10.1093/nar/30.1.264
Chothia C Lesk AM The relation between the divergence of sequence and structure in proteins EMBO J 1986 5 823 826 3709526
Sander C Schneider R Database of homology-derived protein structures and the structural meaning of sequence alignment Proteins 1991 9 56 68 2017436
Martin ACR MacArthur MW Thornton JM Assessment of comparative modeling in CASP2 Proteins 1997 Suppl 1 14 28 9485491 10.1002/(SICI)1097-0134(1997)1+<14::AID-PROT4>3.0.CO;2-O
Vitkup D Melamund E Moult J Sander C Completeness in structural genomics Nat Struct Biol 2001 8 559 566 11373627 10.1038/88640
Martin-Renom M Stuart RC Fiser A Sanchez R Melo F Sali A Comparitive Protein Structure Modeling of Genes and Genomes Annu Rev Biophys Biomol Struct 2000 29 291 325 10940251 10.1146/annurev.biophys.29.1.291
Al-Lazikani B Jung J Xiang Z Honig B Protein structure prediction Curr Opin Chem Biol 2001 5 51 56 11166648 10.1016/S1367-5931(00)00164-2
Westbrook J Feng Z Chen L Yang H Berman HM The Protein Data Bank and structural genomics Nucleic Acids Res 2003 31 489 491 12520059 10.1093/nar/gkg068
Orengo CA Bray JE Buchan DW Harrison A Lee D Pearl FM The CATH protein family database: A resource for structural and functional annotation of genomes Proteomics 2002 2 11 21 11788987 10.1002/1615-9861(200201)2:1<11::AID-PROT11>3.3.CO;2-K
Pearson WR Comparison of methods for searching protein sequence databases Protein Sci 1995 4 1145 1160 7549879
Altschul SF Gish W Miller W Myers EW Lipman DJ Basic local alignmnet search tool J Mol Biol 1990 215 403 410 2231712 10.1006/jmbi.1990.9999
Altschul SF Madden TL Schaffer AA Zhang J Zhang Z Miller W 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
David R Korenberg MJ Hunter IW 3D-1D threading methods for protein fold recognition Pharmacogenomics 2000 1 445 455 11257928 10.1517/14622416.1.4.445
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
Browne WJ North ACT Phillips DC Brew K Vanaman TC Hill RC A possible three-dimensional structure of bovine lactalbumin based on that of hen's egg-white lysosyme J Mol Biol 1969 42 65 86 5817651 10.1016/0022-2836(69)90487-2
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
Greer J Comparative modeling methods: application to the family of the mammalian serine proteases Proteins 1990 7 317 334 2381905
Jones TA Thirup S Using known substructures in protein model building and crystallography EMBO J 1986 5 819 822 3709525
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
Claessens M Van Cutsem C Lasters I Wodak S Modelling the polypeptide backbone with 'spare parts' from known protein structures Protein Eng 1989 2 335 345 2928296
Levitt M Accurate modeling of protein conformation by automatic segment matching J Mol Biol 1992 226 507 533 1640463 10.1016/0022-2836(92)90964-L
Havel TF Snow ME A new method for building protein conformations from sequence alignments with homologues of known structure J Mol Biol 1991 217 7 10.1016/0022-2836(91)90603-4
Srinivasan S March CJ Sudarsanam S An automated method for modeling proteins on known templates using distance geometry Prot Sci 1993 2 277 289
Sali A Blundell TL Comparitive protein modelling by satisfaction of spatial restraints J Mol Biol 1993 234 779 815 8254673 10.1006/jmbi.1993.1626
Brocklehurst SM Perham RN Prediction of the three-dimensional structures of the biotinylated domain from yeast pyruvate carboxylase and the of the lipoylated H-protein from the pea leaf glycine cleavage system: a new automated method for the prediction of protein tertiary structure Prot Sci 1993 2 626 639
Aszodi A Taylor WR Homology modelling by distance geometry Fold Des 1996 1 325 334 9080179
Kolinski A Betancourt MR Kihara D Rotkiewicz P Skolnick J Generalized comparitive modeling (GENECOMP): A combination of sequence comparison, threading, and lattice modeling for protein structure prediction and refinement Proteins 2001 44 133 149 11391776 10.1002/prot.1080
Güntert P Mumenthaler C Wüthrich K Torsion Angle Dynamics for NMR Structure Calculation with the New Program DYANA J Mol Biol 1997 273 283 298 9367762 10.1006/jmbi.1997.1284
Brünger AT Adams PD Clore GM DeLano WL Gros P Grossekunstleve RW Crystallography & NMR System: A New Software Suite for Macromolecular Structure Determination Acta Cryst 1998 D54 905 921
Döker R Maurer T Kremer W Neidig K-P Kalbitzer HR Determination of Mean and Standard Deviation of Dihedral Angles BBRC 1999 257 348 350 10198215
Zhang C Hou J Kim S-H Fold prediction of helical proteins using torsion angle dynamics and predicted restraints Proc Natl Acad Sci U S A 2002 99 3581 3585 11904420 10.1073/pnas.052003799
van Nuland NAJ Hangyi IW van Schaik RC Berendsen HJ van Gunsteren WF Scheek RM The High-resolution Structure of the Histidine-containing Phosphocarrier Protein HPr from Escherichia coli Determined by Restrained Molecular Dynamics from Nuclear Magnetic Resonance Nuclear Overhauser Effect Data J Mol Biol 1994 237 544 559 8158637 10.1006/jmbi.1994.1254
Jia Z Quail JW Waygood EB Delbaere LT The 2.0-A resolution structure of Escherichia coli histidine-containing phosphocarrier protein HPr. A redetermination J Biol Chem 1993 268 22490 22501 8226757
Uppenberg J Svensson C Jaki M Bertilsson G Jendeberg L Berkenstam A Crystal Structure of the Ligand Binding Domain of the Human Nuclear Receptor Ppargamma J Biol Chem 1998 273 31108 31112 9813012 10.1074/jbc.273.47.31108
Gronwald W Kalbitzer HR Automated structure determination of proteins by NMR spectroscopy Prog NMR Spectrosc 2004 44 33 96 10.1016/j.pnmrs.2003.12.002
Postma PW Lengeler JW Jacobson GR Phosphoenolpyruvate:carbohydrate phosphotransferase systems of bacteria Microbiol Rev 1993 57 543 594 8246840
Laskowski RA Rullmann JAC MacArthur MW Kaptein R Thornton JM AQUA and PROCHECK-NMR Programs for checking the quality of protein structures solved by NMR J Biomol NMR 1996 8 477 486 9008363 10.1007/BF00228148
Xu HE Lambert MH Montana VG Plunket KD Moore LB Collins JL Structural determinants of ligand binding selectivity between the peroxisome proliferator-activated receptors Proc Natl Acad Sci U S A 2001 98 13919 13924 11698662 10.1073/pnas.241410198
Xu HE Stanley TB Montana VG Lambert MH Shearer BG Cobb JE Structural basis for antagonist-mediated recruitment of nuclear co-repressors by PPARalpha Nature 2002 415 813 817 11845213
Cronet P Petersen JFW Folmer R Blomberg N Sjoblom K Karlsson U Structure of the PPARalpha and -gamma ligand binding domain in complex with AZ 242; ligand selectivity and agonist activation in the PPAR family Structure 2001 9 699 706 11587644 10.1016/S0969-2126(01)00634-7
Xu HE Lambert MH Montana VG Parks DJ Blanchard SG Brown PJ Molecular recognition of fatty acids by peroxisome proliferator-activated receptors Mol Cell 1999 3 397 403 10198642 10.1016/S1097-2765(00)80467-0
Gronwald W Kirchhofer R Gorler A Kremer W Ganslmeier B Neidig KP RFAC, a program for automated NMR R-factor estimation J Biomol NMR 2000 17 137 151 10921778 10.1023/A:1008360715569
Dominguez C Boelens R Bonvin AMJJ HADDOCK: A Protein-Protein Docking Approach Based on Biochemical or Biophysical Information J Am Chem Soc 2003 125 1731 1737 12580598 10.1021/ja026939x
Linge JP Habeck M Rieping W Nilges M ARIA: automated NOE assignment and NMR structure calculation Bioinformatics 2003 19 315 316 12538267 10.1093/bioinformatics/19.2.315
Linge JP Williams MA Spronk CAEM Bonvin AMJJ Nilges M Refinement of protein structures in explicit solvent Proteins 2003 50 496 506 12557191 10.1002/prot.10299
Nabuurs SB Nederveen AJ Vranken W Doreleijers JF Bonvin AMJJ Vuister GW DRESS: a Database of REfined Solution NMR Structures Proteins 2004 55 483 486 15103611 10.1002/prot.20118
Gröger C Möglich A Pons M Koch B Hengstenberg W Kalbitzer HR NMR-Spectroscopic Mapping of an Engineered Cavity in the I14A Mutant of HPr from Staphylococcus carnosus Using Xenon J Am Chem Soc 2003 125 8726 8727 12862458 10.1021/ja030113t
Möglich A Koch B Gronwald W Hengstenberg W Brunner E Kalbitzer HR Solution structure of the active-centre mutant I14A of the histidine-containing phosphocarrier protein from Staphlococcus carnosus Eur J Biochem 2004 271 4815 4824 15606769 10.1111/j.1432-1033.2004.04447.x
Tjandra NL Bax A Direct Measurement of Distances and Angles in Biomolecules by NMR in a Dilute Liquid Crystalline Medium Science 1997 278 1111 1114 9353189 10.1126/science.278.5340.1111
Brunner E Residual dipolar couplings in protein NMR Concepts Magn Reson 2001 13 238 259 10.1002/cmr.1012
Koradi R Billeter M Wüthrich K MOLMOL: a program for display and analysis of macromolecular structures J Mol Graphics 1996 14 51 55 10.1016/0263-7855(96)00009-4
Görler A Kalbitzer HR Relax, a Flexible Program for the Back Calculation of NOESY Spectra Based on Complete-Relaxation-Matrix Formalism J Magn Reson 1997 124 177 188 9169213 10.1006/jmre.1996.1033
Görler A Gronwald W Neidig KP Kalbitzer HR Computer assisted assignment of 13C or 15N edited 3D-NOESY-HSQC spectra using back calculated and experimental spectra J Magn Reson 1999 137 39 45 10053131 10.1006/jmre.1998.1614
Jia Z Vandonselaar M Hengstenberg W Quail JW Delbaere LT The 1.6 A structure of histidine-containing phosphotransfer protein HPr from Streptococcus faecalis J Mol Biol 1994 236 1341 1355 8126724 10.1016/0022-2836(94)90062-0
Maurer T Döker R Görler A Hengstenberg W Kalbitzer HR Three-dimensional structure of the histidine containing phosphocarrier protein (HPr) from Enterococcus faecalis in solution Eur J Biochem 2001 268 635 644 11168402 10.1046/j.1432-1327.2001.01916.x
Görler A Hengstenberg W Kravanja M Beneicke W Maurer T Kalbitzer HR Solution Structure of the Histidine-Containing Phosphocarrier Protein from Staphylococcus carnosus Appl Magn Reson 1999 17 465 480
Jones BE Rajagopal P Klevit RE Phosphorylation on histidine is accompanied by localized structural changes in the phosphocarrier protein, HPr from Bacillus subtilis Protein Sci 1997 6 2107 2119 9336834
| 15819976 | PMC1127110 | CC BY | 2021-01-04 16:02:48 | no | BMC Bioinformatics. 2005 Apr 8; 6:91 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-91 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-911581997610.1186/1471-2105-6-91SoftwareA restraint molecular dynamics and simulated annealing approach for protein homology modeling utilizing mean angles Möglich Andreas [email protected] Daniel [email protected] Till [email protected] Wolfram [email protected] Hans Robert [email protected] Institut für Biophysik und physikalische Biochemie, Universität Regensburg, Universitätsstr. 31, D-93053 Regensburg, Germany2 Department of Biophysical Chemistry, Biozentrum, University of Basel, Klingelbergstr. 70, CH-4056 Basel, Switzerland3 Institut für Organische Chemie und Biochemie, Technische Universität München, Lichtenbergstr. 4, D-85747 Garching, Germany4 Department of Lead Discovery, Boehringer Ingelheim Pharma GmbH, Birkendorfer Str. 65, D-88397 Biberach, Germany2005 8 4 2005 6 91 91 1 9 2004 8 4 2005 Copyright © 2005 Möglich 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 have developed the program PERMOL for semi-automated homology modeling of proteins. It is based on restrained molecular dynamics using a simulated annealing protocol in torsion angle space. As main restraints defining the optimal local geometry of the structure weighted mean dihedral angles and their standard deviations are used which are calculated with an algorithm described earlier by Döker et al. (1999, BBRC, 257, 348–350). The overall long-range contacts are established via a small number of distance restraints between atoms involved in hydrogen bonds and backbone atoms of conserved residues. Employing the restraints generated by PERMOL three-dimensional structures are obtained using standard molecular dynamics programs such as DYANA or CNS.
Results
To test this modeling approach it has been used for predicting the structure of the histidine-containing phosphocarrier protein HPr from E. coli and the structure of the human peroxisome proliferator activated receptor γ (Ppar γ). The divergence between the modeled HPr and the previously determined X-ray structure was comparable to the divergence between the X-ray structure and the published NMR structure. The modeled structure of Ppar γ was also very close to the previously solved X-ray structure with an RMSD of 0.262 nm for the backbone atoms.
Conclusion
In summary, we present a new method for homology modeling capable of producing high-quality structure models. An advantage of the method is that it can be used in combination with incomplete NMR data to obtain reasonable structure models in accordance with the experimental data.
==== Body
Background
Due to the enormous progress that has been made in genomics a large number of DNA sequences including many whole genomes have been published. The evaluation of these data must include the determination of the three-dimensional structures of the proteins encoded. Although the two experimental techniques capable of determining three-dimensional structures of proteins and other biomolecules at atomic resolution, namely nuclear magnetic resonance (NMR) and X-ray crystallography, have seen significant improvements the process of structure determination remains very time-consuming and difficult. Unless unexpected advances of these techniques will occur in future, it is obvious that for the majority of all the primary sequence data available three-dimensional structures cannot be obtained experimentally. Therefore, only computational approaches are capable of filling the gap between existing protein sequences and structures. Although considerable progress has been achieved in ab initio structural prediction strategies [1-3] they are in general still unreliable when atomic resolution is demanded. However, when structures of homologous proteins are available, the prediction of the three-dimensional structure of entire proteins and protein domains is rather successful. In light of the fact that the protein structures elucidated so far only show a remarkably limited number of folds it would be desirable to accelerate the structure determination process especially for proteins possessing a fold already known. According to the SCOP classification [4,5] (release 1.65, 1 August 2003) 20619 protein structures stored in the Protein Data Bank share only 800 different folds. Comparison of different proteins with similar amino acid sequences showed that they quite often display very similar tertiary structures [6-9].
In the past several different homology modeling approaches were published which range from strongly interactive methods (model building) to fully automated methods (for reviews see e. g. [10] and [11]). Generally the starting point in these approaches is a search in structure databases such as the Protein Data Bank [12] or CATH [13] for all protein structures that are related to the target sequence and then to select those 3D structures that will be used as templates. For searching the structural databases one can employ pairwise sequence-sequence comparisons using for example programs such as FASTA [14] and BLAST [15]. When increased search sensitivity or a larger number of homologs are demanded methods which are based on multiple sequence alignments prove to be particularly efficient. Such an algorithm is implemented in the program PSI-BLAST [16]. An alternative strategy for homolog identification relies on so-called threading methods, which predict whether the target sequence adopts any of the known 3D folds. Threading methods should be useful in cases when no sequences can be found which are clearly related to the target [17].
When a list of related protein structures has been obtained the appropriate templates have to be chosen from these. In this procedure usually factors such as high overall sequence similarity between target and template sequences, quality of the template structure and conditions under which the template structure was obtained are taken into account. Then the selected templates have to be optimally aligned with the target sequence. Since the search methods mentioned above are usually optimized for detecting remote homologs they are not optimal for target-template alignment. A program often used for the latter type of alignments is CLUSTALX [18], which is also used within PERMOL.
Using the template-target alignment a variety of methods has been published for 3D model building. The group of methods which were developed first and are still frequently used were modeling by rigid body assembly [19-21]. Another group of methods use segment matching [22-25]. In the third, most recent group of methods spatial restraints obtained from the template structures are used in distance geometry calculations or energy optimization procedures to obtain the target model [26-31].
The PERMOL approach described presently also uses spatial restraints but in contrast to most other programs mainly dihedral angle restraints as opposed to restraints derived from inter-atomic distances are employed. These restraints enter molecular dynamics calculations in torsion angle space. In the following we will describe this method in more detail and mark differences to existing programs that have been published before.
In ab initio molecular dynamics (MD) simulations in addition to the applied force field only information about the amino acid sequence of the protein in study enters the calculations. While for small molecules such methods show results that are in very good agreement with the experimental data they mostly fail for more complex molecules. On the other hand restrained molecular dynamics calculations based on simulated annealing protocols are routinely and successfully used for the determination of solution NMR structures – in that case strong experimental information is available. Especially effective with regard to computational effort are calculations in torsion angle space as implemented in the programs DYANA [32] and CNS [33].
In this contribution we propose a method which combines the well-developed torsion angle dynamics calculations of DYANA or CNS with structural information extracted from three-dimensional structures of homologous proteins. This information is translated into conformational restraints. Local structural restraints are obtained by a weighted average of the backbone dihedral angles using an algorithm proposed by Döker et al. [34] These averaged dihedral angle restraints are usually well preserved within the local secondary structure elements and therefore are especially well suited for the modeling of these. The program MODELLER [28] for example also uses dihedral angle restraints in an optimization procedure but expresses them as so-called probability density functions which are derived from structural features in several families of homologous proteins. Global structural restraints are obtained from distance relations between carefully selected atoms of amino acids well separated in the primary structure. In contrast to other programs the distance restraints are mainly used for the global arrangement of the secondary structure elements which are defined by the dihedral angle restraints. The efficient structure calculations performed with DYANA or CNS allow calculating a large number of structural models in a relatively short amount of time. From the resulting ensemble of structures the best in terms of the DYANA target function or total energy (CNS) can be selected for further analysis. As has also been shown in NMR spectroscopy, it is useful to describe the target structure by an ensemble of model structures.
It should be noted that the PERMOL approach described here is related to the method detailed by Zhang et al. [35], which uses a combination of torsion angle dynamics and dihedral angle and distance restraints to predict the fold of helical proteins. In contrast to PERMOL the program from Zhang et al. uses methods for secondary structure and contact prediction to derive spatial restraints.
To benchmark the PERMOL approach we used it to determine a homology structure for the histidine-containing phosphocarrier protein (HPr) from E. coli of which the structure has been solved experimentally both by NMR [36] (PDB entry: 1HDN) and X-ray crystallography [37] (1POH). The homology model was compared to the target structures and to a homology model calculated with the program MODELLER [28]. To also investigate the performance of PERMOL on larger proteins that contain substantial disordered loop regions the human peroxisome proliferator activated receptor γ (Ppar γ) was used as a test case. Its structure has been determined previously by X-ray crystallography [38] (3PRG).
Results
Theoretical considerations and general strategy
In standard NMR structure determination the principal physical model of a protein is represented by empirical potentials determining the general geometry. The fast optimization is obtained by a simulated annealing protocol and the correct conformations are selected from the generally accessible conformational space by the experimental restraints which are transformed into pseudo-potentials. In the approach used in PERMOL the experimental restraints are replaced by restraints derived from three-dimensional structures of homologous proteins. Local conformations are optimally encoded by the distribution of the corresponding torsion angles. The overall fold is determined by distance relations since even small errors in dihedral angles can add up to very large distance errors between amino acids that are separated by several positions in the sequence.
The use of a molecular dynamics and simulated annealing protocol for homology modeling allows to encode the features of the statistical distribution of a given parameter αi individually for each group of restraints. To this end not only the expectation values are calculated from the homologous structures j (j = 1,..,Ni) but also the upper and lower limits, and . It is still under discussion in the NMR community how exactly the upper and lower limits of restraints have to be defined but it is clear that they are related to the expected error of a given, individual parameter. A generally accepted definition is not available yet. In addition the form of the pseudo-energy function used in the calculations has to depend on the error distribution of the given parameter (see e. g. [39]).
The homology modeling procedure proposed here comprises the following steps: step 1, selection of data and sequence alignment, step 2, selection of restraints, and step 3, the restrained molecular dynamics simulation. These conceptually different steps in the calculation are reflected in the implementation of PERMOL in corresponding levels of the modeling procedure.
Level 1 – Selection of data and sequence alignment
Initially, one or several structures of homologous proteins are selected as templates. Their amino acid sequences are aligned to the sequence of the target protein using the program CLUSTALX [18]. The resulting alignment is written to a text file and can be edited by the user. Conserved amino acids are characterized and classified for manual or automated selection of restraints. Based on the degree of sequence conservation in the different proteins a homology score value vi is calculated for each residue. The score values vi range from 1.0 for a completely conserved residue to 0.1 for a residue, which in the template proteins has been replaced in a non-conservative manner, e.g. a hydrophobic residue replaced by a charged one.
Level 2 – Selection of restraints
For the calculation of dihedral angle restraints usually only Φ and Ψ angles are taken into account but the ω-angle can be included as well. Structural restraints are only derived from residues, which have been selected. Additional residues can be selected either manually or automatically based upon the score value vi. Expectation values and standard deviations are calculated as described in the 'Methods' section with set to 1/ when structures are found in the pdb-file k as it is often the case for NMR-structures. Upper and lower limits for the dihedral angle restraints can be calculated either as the mean value plus/minus multiples of the standard deviations, <αi> ± b* <si> with a user defined constant b, or as the mean angle plus/minus a constant value. An additional weighting of the individual restraints can be performed on the basis of the score value vi which modifies the force constant of the restraint i in the MD calculation.
By default, distance restraints are automatically computed between the NH atoms of completely conserved amino acids. Restraints can also be generated for additional amino acids and atom types by appropriate selection. For the generation of distance restraints similar options are possible as for dihedral angle restraints. In addition, an upper distance limit for the pairs of atoms to be considered can be defined.
Conserved hydrogen bonds can also be used to generate distance restraints between the atoms involved in forming the bond. The criteria for selecting hydrogen bonds in the homologous protein structures can be modified by the user. By default, only hydrogen bonds are considered for which the N-O distance does not exceed 0.24 nm and the angle between the NH-HN and the C = O bond vectors does not deviate by more than 35° from 180°. Again, different options are possible for the calculation of the upper and lower limits. Hydrogen bonds which occur only in a few structures or are assigned to more than one pair of atoms, e.g. due to deviations between the different homologous proteins used as templates, can be automatically removed by corresponding filter functions.
Level 3 – Restrained molecular dynamics simulation
The restraint files generated by PERMOL can be directly used by the molecular dynamics programs DYANA and CNS. Standard simulated annealing protocols are employed.
Modeling of HPr from E. coli and of human Ppar γ
To test the modeling approach described in this paper we determined a homology structure for the histidine-containing phosphocarrier protein (HPr) from E. coli. HPr is an integral part of the bacterial phosphoenolpyruvate dependent phosphotransferase system (PTS) which efficiently catalyses phosphorylation and the import of carbohydrates into prokaryotic cells [40]. HPr molecules from different organisms have been extensively studied and many 3D structures have been elucidated. In particular the structure of HPr from E. coli has been solved both by NMR [36] (PDB entry: 1HDN) and X-ray crystallography [37] (1POH) and is thus especially suited to test our modeling strategy (see Table 1).
Four previously determined HPr structures from four different organisms have been used as model structures (PDB codes 1PTF, 1QFR, 1QR5, and 2HID). An overview of these structures is given in Table 1. Only 21 % of the amino acid sequence is strictly conserved between the HPr proteins of E. coli, S. faecalis, E. faecalis, S. carnosus, and B. subtilis (18 out of 85 residues). Spatial restraints for the structure calculation were generated as detailed in the 'Methods' section. For the derivation of inter-atomic distance restraints only residues which are completely conserved or display conservative amino acid exchanges (e. g. one hydrophobic residue replaced by another one) were considered. Upper and lower limits for these distances were determined as the mean distance value plus or minus the standard deviation, respectively. Restraints for the backbone dihedral angles Φ and Ψ were calculated for all residues and have been weighted according to the homology score value vi. Upper and lower limits were determined as for the distance restraints. Hydrogen bonds were analyzed using the default parameter values. Distance restraints between the corresponding HN and O atoms were computed as the mean distance value plus or minus the standard deviation. A summary of these restraints is presented in Table 3.
Based on these restraints an ensemble of homology structures was computed using the molecular dynamics program DYANA [B32] with the standard simulated annealing protocol. Out of 200 structures calculated, the group of the ten structures with the lowest pseudo-energies was further analyzed. These ten models showed a good convergence with a RMSD value for the backbone atom positions of 0.041 nm (Fig. 1, Table 5). They displayed the well-known secondary structure elements common to all HPr molecules studied so far, comprising a four-stranded antiparallel β-sheet and three α-helices designated as helices a, b, and c. Analysis of the ensemble of these ten structures with PROCHECK-NMR [41] showed that all backbone dihedral angles fell into the most favored and additionally allowed regions of the Ramachandran plot (Table 5). Modeling experiments where the dihedral angle restraints have been partly or completely left out from the structure calculations of the model structures underlined their importance in defining the correct secondary structure and local conformations (see below).
In order to further test our modeling strategy we set out to derive a homology structure for the human peroxisome proliferator activated receptor γ (Ppar γ). Ppar γ is considerably larger than HPr and comprises about 280 amino acid residues. Further, it contains larger relatively unstructured loop regions and it is worthwhile to investigate how PERMOL performs here. In addition this molecule is of particular importance for us since we are currently in the process of experimentally solving its solution structure. Via a BLAST [16] search for the primary sequence of Ppar γ we identified several related proteins for which three-dimensional structures are available (Table 2), namely Ppar α [42-44] (PDB codes: 1K7L, 1KKQ, and 1I7G) and Ppar δ [45] (1GWX and 3GWX).
Model structures were calculated as detailed for HPr and out of 125 calculated structures the 16 structures with the lowest pseudo energies were further analyzed. A summary of the used restraints is given in Table 4. These sixteen models showed a good convergence with a RMSD value for the backbone atom positions of 0.135 nm (residues 206 – 477) (Fig. 2, Table 6). The secondary structure elements observed in the model structures agree well with the corresponding X-ray structure of the template protein, comprising a four-stranded antiparallel β-sheet and twelve α-helices (Fig. 2). Analysis of the ensemble of the selected sixteen structures with PROCHECK-NMR [41] showed that almost all backbone dihedral angles fell into the most favored and additionally allowed regions of the Ramachandran plot (Table 6).
Comparison to target structures
The ensemble of modeled HPr structures was compared to the target structure of HPr from E. coli which before had been elucidated using NMR spectroscopy (1HDN) and X-ray crystallography (1POH). For 1HDN a bundle of 30 structures was deposited in the protein database. As stated in the header of the coordinate file the first structure is closest to the ensemble average. As a consequence this structure was selected as the NMR target structure. A comparison between the modeled structure and the target NMR and X-ray structures is shown in Fig. 3. The homology model displayed the same global fold and distribution of secondary structure elements as both target structures. To quantify the agreement between the individual structures the root mean square deviations (RMSD) between the different structures were calculated for the backbone atom positions. While the RMSD between the two target structures 1HDN and 1POH amounted to 0.11 nm the comparison of the best modeled structure with the target NMR structure and the X-ray structure yielded RMSD values of 0.17 nm and 0.15 nm, respectively. Although the agreement between the modeled and the target structures was worse than the agreement between the two target structures, the RMSD values were of similar magnitude. Deviations between the homology model and the experimentally determined structures were mainly seen in the loop regions and in the orientation of helices a and b. Interestingly, these are also the regions that are least well defined in the X-ray and NMR structures and where these structures diverge most. In contrast, the core region of HPr and its overall fold are reproduced well in the homology model.
Further, we used R-factor analysis [46] to compare the modeled structure to the target structures. The quality of the protein backbone was specifically assessed by only taking into account spectral signals arising from backbone protons. Low R-factors of similar magnitude were obtained when comparing the modeled structure with either the NMR target structure (R-factor 0.093) or the X-ray target structure (0.076). Consistent with the RMSD values the R-factors also indicated that the homology structure more closely resembled the X-ray structure than the NMR structure. A slightly lower R-factor of 0.073 was obtained when comparing the two target structures with each other (Table 7).
For Ppar γ the best model structure in terms of pseudo-energy was compared to the target X-ray structure (3PRG). The agreement between the two structures was assessed by calculating the corresponding RMSD value for the backbone atoms, which amounted to 0.262 nm (Table 8). Note that the first five unstructured residues and the region between residues 262 and 274 which were missing in the X-ray target structure were not considered in this analysis. Deviations between the homology model and the X-ray structure were mainly seen in the loop regions and in the orientation of the helices preceding and following the unstructured region between residues 262 and 274. The agreement between model and X-ray structure was further analyzed by the calculation of pseudo NMR R-factors (Table 8). Although somewhat higher R-factors were obtained for Ppar γ than for HPr, the R-factor analysis still showed a reasonable agreement between model and X-ray structure.
Importance of torsion angles
In principle, torsion angles can completely define the 3D-structure of a protein when the general geometry of the amino acids is predefined. However, small errors of torsion angles in the backbone propagate and lead to large errors in the Cartesian space for amino acids remote in the sequence. Nevertheless, torsion angles are optimal predictors for local folding. Fig. 4 exemplifies the importance of the torsion angles for the structure predictions. As an example it shows a structure prediction (calculation) of HPr from S. faecalis from a rather small number of restraints created from the X-ray structure (1PTF) of the protein. Only 427 torsion angle restraints together with 41 hydrogen bond restraints can be sufficient to determine the various secondary structure elements together with the global fold of the molecule. Even the loop regions for which no hydrogen bond restraints are present adopt native-like conformations. Only the third α-helix is rotated away from the core of the protein since its orientation is solely defined by the angle restraints of residues 67–69.
Discussion
In this contribution we have presented a new program for homology modeling of protein structures. Using restraint molecular dynamics simulations together with spatial restraints derived from template structures we calculated homology structures of HPr from E. coli and of human Ppar γ. An advantage of the proposed method is the use of spatial restraints with individual upper and lower limits depending on the local structural conservation in the template structures. This becomes especially evident for the obtained bundle of Ppar γ model structures where one can easily distinguish between the mostly well-defined secondary structure elements and less ordered regions e.g. some of the larger loop regions.
At first glance it appears to be a disadvantage of the proposed method that not a unique, seemingly perfect structure is the result of the calculations as in the case of threading methods. However, the structure bundle produced by our approach gives an idea of the conformational subspace determined by the available experimental basis and the physical model. This is a safeguard against typical over-interpretations of model structures where data in badly predictable regions are used for the detailed interpretation of functional data or are used during the drug design process.
An additional advantage of the simulated annealing approach is that restraint violations are not treated explicitly but contribute to the overall "energy" which is minimized. In contrast to other methods in the approach used in PERMOL the mean torsion angles and their errors provide the main information. A few distance restraints are used to define the long-range relations which cannot be described sufficiently well by the local data. Accordingly, details of the selection of these restraints are not critical. Thus, the selection of pairwise restraints between all conserved residues seems to be plausible. The same is true for conserved hydrogen bonds. However, the PERMOL software also allows to define a custom selection of restraints and thus an adaptation to specific needs. As an example all hydrophobic contacts between amino acid residues observed in the template structures could be selected to serve as restraints. The automated calculation of individual weighting factors during the calculation of the expectation values and standard errors of the individual restraints would permit to introduce information about the local and global sequence conservation and the precision of the used structures. Currently, we are undertaking efforts to address this question. The high quality of the structure models generated with PERMOL illustrates that the same MD programs used for the determination of NMR structures can also be utilized for homology modeling. The programs and strategies developed for NMR structure determination have evolved to efficient optimizers even when only limited information (i. e. small number of structural restraints) is available. This has been recognized for example by Dominguez et al. [47] who use restrained molecular dynamics together with the ARIA protocol [48] for solving the docking problem. While in the case of NMR structure determination the restraints that enter the molecular dynamics simulation are derived from experimental observables like NOE cross-peaks, J-couplings, and residual dipolar couplings, in the case of homology modeling synthetic restraints are generated from previously determined structures of homologous template proteins. The use of standard MD programs and protocols also has a disadvantage since it is not possible to directly introduce properties in the calculation which are not provided for by the programs. An example would be the use of specific potential forms with multiple minima which describe the homology-derived information in more detail as it is done e. g. by MODELLER [28].
We compared the HPr homology structure we obtained with PERMOL to a structural model of HPr from E. coli calculated using MODELLER (version 6v2). When the same alignment file and template structures were used, homology models of similar quality were obtained with the two programs.
A specific advantage of the approach presented here is that it can be well used in the context of standard structure determination by NMR. The restraint files generated by PERMOL are editable and can be easily combined with other data and be adapted for use with different programs. As the same MD programs are used both for modeling with PERMOL and for NMR structure determination, incomplete experimental data can be conveniently combined with spatial restraints derived from homologous template proteins. The validity of the resulting structure models can be checked by calculating NMR R-factors [46]. Different force fields and annealing protocols which are available for the NMR MD programs can also be utilized for homology modeling. In this way recent advances like the structure refinement in explicit solvent [49,50] can be readily exploited to derive more accurate homology structures.
Conclusion
In summary, we have presented a new method for homology modeling capable of producing high-quality structure models. Compared to many other homology structure prediction programs it is based on a different philosophy since its aim is not to predict a unique best structure but a bundle of structures representing the locally different degrees of reliability of the structure prediction. Since the homology-derived restraints are mainly used to reduce the conformational space to be searched by the MD calculation, their relative importance for obtaining a correct homology model is expected to decrease in future time as the physical model employed in these calculations is improved. Another advantage of the approach described here is its flexibility, conveniently allowing several template structures to be included as sources of structural restraints. Furthermore, the PERMOL software permits to determine which kinds of structural restraints enter the molecular dynamics calculation in a controlled fashion. We demonstrated that the standard MD programs used in the course of structure determination by NMR can also be well utilized for the purpose of homology modeling. Prediction on the basis of averaged torsion angles is a powerful tool which efficiently makes use of the structural information available in the protein data base and leads to well-defined structures.
Recently, a homology model determined with PERMOL was used in the resonance assignment [51] and structure determination process of a mutant form of HPr from S. carnosus [52] and to obtain an initial estimate for the molecular alignment tensor describing the partial orientation of the HPr molecule in anisotropic solution [53,54]. PERMOL has also been integrated in the NMR structure determination package AUREMOL [39]. In this molecule-centered top-down approach one starts with a trial structure e.g. a homology model obtained by PERMOL that is iteratively refined until it fits the experimental data sufficiently as verified by the calculation of NMR R-factors.
Methods
Calculation of the restraints for simulated annealing
Structural information obtained from a set of homologous structures j (j = 1,..,Ni) must be expressed in form of restraints. The restraint of a parameter αi is usually defined by its expectation value and the upper and lower limits and , respectively. PERMOL offers several ways to calculate these quantities from the expectation values observed in the template proteins <αi> and the corresponding standard deviations si. For non-cyclic parameters <αi> and si can be simply calculated according to eqs. (1) and (2).
and
with the weighting factor for a given event i and the total number of events Ni. For cyclic parameters like dihedral angles, which are mainly used within PERMOL such a definition does not directly apply but can be extended by the approach described by Döker et al. (1999) [34].
Here, the origin of the coordinate system is shifted to fulfill the condition
and the standard deviation is calculated according to eq. (2). The expectation value is obtained by
The parameters determine the statistical weight of a given homology structure used to calculate a restraint. In principle, their value will depend on factors such as the local and global sequence conservation and the quality of a structure, e. g. when comparing X-ray and NMR-structures.
Implementation overview
In order to facilitate the determination of structural restraints for homology modeling the software package PERMOL was developed. PERMOL was written in Perl/Tk and has been tested with the operating systems SGI IRIX, Linux and Windows. The software and a detailed manual explaining its use can be obtained free of charge from the authors . Sequence alignment is done by using the program CLUSTALX [18]. Structure calculations are performed with output data files generated by PERMOL which can be imported by the molecular dynamics programs DYANA [32] and CNS [33]. Dihedral angles from different structures are averaged following the algorithm described by Döker et al. [34]. The typical computing time for setting up the restraint and parameter files for the MD-calculation is negligible using a modern PC. The calculation of the structures strongly depends on the MD-program used, the number of structures calculated and the actual simulated annealing protocol. In the examples presented here structures were calculated on a standard Linux-PC using the MD program DYANA. The corresponding calculation times for a single structure model were around 30 and 160 seconds for HPr and Ppar γ, respectively. Figures 1, 2, 3, and 4 have been prepared with MOLMOL and rendered with PovRay .
Validation of homology models
Modeled structures can be quantitatively compared to their respective target structures by calculating NMR R-factors according to [46]. Analogous to crystallography R-factors, NMR R-factors are used to quantify how well a three-dimensional structure accounts for the spectral signals occurring in an experimental NMR spectrum. Using an implementation of the complete relaxation matrix analysis (RELAX, [56,57]) artificial NMR spectra are calculated for the given three-dimensional structure and compared to the experimental spectra. R-factors quantify the deviations between the two types of spectra and are therefore a measure for the quality of the trial structure. In the case of perfectly matching spectra the R-factor adopts a value of 0. Analogous, R-factor analysis can also be employed to quantify the agreement between two protein structures. In that case artificial NMR spectra are calculated for both structures and are compared to each other.
The agreement between two structures can be further assessed by determining the root mean square deviations (RMSD) between the atom positions of the structures. The program MOLMOL [55] is used to fit the structures atop of each other and to calculate RMSD values. The stereo-chemical quality of the obtained models was validated using the program PROCHECK-NMR [41].
Abbreviations
HPr: histidine-containing phosphocarrier protein, MD: molecular dynamics, NMR: nuclear magnetic resonance, NOE: nuclear Overhauser effect, PDB: Protein Data Bank Brookhaven, Ppar γ: human peroxisome proliferator activated receptor γ, PTS: phosphoenolpyruvate carbohydrate phosphotransferase system, RMSD: root mean square deviation
Authors' contributions
TM, WG, and HRK conceived the project. DW and TM performed initial feasibility studies and refined the overall modeling strategy. AM wrote the PERMOL software and a manual. AM, DW, and WG calculated the homology structures. AM drafted the manuscript. WG and HRK coordinated the study and wrote the manuscript. All authors read and approved the final manuscript.
Acknowledgements
The authors thank Dipl. Phys. K. Brunner, Dr. R. Döker, Dr. W. Kremer and Dipl. Phys. J. Trenner for helpful discussions. Financial support by the European Commission (SPINE-project) is gratefully acknowledged.
Figures and Tables
Figure 1 Homology structures of HPr from E. coli determined by PERMOL. Ensemble of the 10 homology structures with the lowest pseudo-energy out of 200 structures calculated with DYANA. (left) A superimposition of the Cα atom traces is shown. (right) A cartoon representation of the mean structure of the 10 models is displayed.
Figure 2 Comparison of the model structure of Ppar γ from human with the corresponding X-ray structure. Overall good agreement between the bundle of final model structures (helices in red and yellow, β-strands in blue and loops in grey) and the X-ray structure (orange) is obtained. Deviations are mainly seen in larger loop regions, the unstructured N-terminus and at the C-terminal end.
Figure 3 Comparison of the model structure of HPr from E. coli with the corresponding X-ray and NMR structures. A comparison of the modeled HPr homology structure with the structures experimentally determined by NMR spectroscopy (1HDN) and X-ray crystallography (1POH). The structures are shown in the same orientation as in Fig. 1 with the radius of the backbone splines indicating the RMSD of the Cα atom positions in the respective structures. (A) Overall good agreement between the model structure (yellow) and the X-ray structure (blue) is obtained. Deviations are mainly seen in loop regions and in the orientation of helices a and b. RMSD values for the Cα atom positions of the X-ray structure 1POH have been derived from the crystallographic B-factors, fB, using the Debye-Waller equation where isotropic displacement from the mean atom positions was assumed. (B) Comparison of the model (yellow) and the NMR structure (red). Deviations are seen in the same regions as before. (C) X-ray (blue) and NMR (red) structures superimpose well. Interestingly, deviations between them are mainly observed in regions where the two structures also diverge from the homology model.
Figure 4 Importance of torsion angle restraints exemplified on HPr from Streptococcus faecalis. On the left hand side the model structure calculated with PERMOL using 427 torsion angle restraints and 41 hydrogen bonds is displayed, while on the right hand side the target X-ray structure 1PTF is shown. The RMSD value for the heavy atoms of the two structures is 0.328 nm. Restraints for torsion angles and hydrogen bonds were directly generated from the X-ray structure 1PTF.
Table 1 Statistics of PDB structure files used for HPr
PDB code Organism Method Resolution [nm]a Reference
1HDN E. coli NMR 0.20 [36]
1POH E. coli X-ray 0.20 [37]
1PTF S. faecalis X-ray 0.16 [58]
1QFR E. faecalis NMR 0.27 [59]
1QR5 S. carnosus NMR 0.28 [60]
2HID B. subtilis NMR 0.19 [61]
aThe equivalent resolution of the NMR structures was calculated using PROCHECK-NMR [41].
Table 2 Statistics of PDB structure files used for Ppar γ
PDB code Organism Method Resolution [nm] Reference
3PRG human X-ray 0.29 [38]
1K7L human X-ray 0.25 [42]
1KKQ human X-ray 0.30 [43]
1I7G human X-ray 0.22 [44]
1GWX human X-ray 0.25 [45]
3GWX human X-ray 0.24 [45]
Table 3 Restraints for molecular dynamics calculation for HPr
Type of restraint Number
inter-atomic distances 186
hydrogen bonds 50
backbone dihedral angles 164
Table 4 Restraints for molecular dynamics calculation for Ppar γ
Type of restraint Number
inter-atomic distances 1391
hydrogen bonds 153
backbone dihedral angles 528
Table 5 Structural statistics for HPr
RMSD values for the ten lowest-energy structures RMSD [nm]
backbone atoms Cα, C', N 0.041
heavy atoms 0.111
Residues in the Ramachandran plot Incidencea
most favored regions 87.2 %
additional allowed regions 12.8 %
generously allowed regions 0.0 %
disallowed regions 0.0 %
aThe dihedral angles have been analyzed using the program PROCHECK-NMR.
Table 6 Structural statistics or Ppar γ
RMSD values for the sixteen lowest-energy structures RMSD [nm]
backbone atoms Cα, C', N 0.135
heavy atoms 0.191
Residues in the Ramachandran plot Incidencea
most favored regions 84.1 %
additional allowed regions 14.3 %
generously allowed regions 1.4 %
disallowed regions 0.2 %
aThe dihedral angles have been analyzed using the program PROCHECK-NMR.
Table 7 Comparison between model structures and experimental structures for HPr
Structures Quantitiesa NMR target structure X-ray target structure
X-ray structure backbone RMSD [nm] 0.106 0
heavy atom RMSD [nm] 0.273 0
R-factor 0.073 0
best NMR structure backbone RMSD [nm] 0 0.106
heavy atom RMSD [nm] 0 0.273
R-factor 0 0.072
best model structure backbone RMSD [nm] 0.169 0.147
heavy atom RMSD [nm] 0.273 0.253
R-factor 0.093 0.076
model structure bundle backbone RMSD [nm] 0.178 0.154
heavy atom RMSD [nm] 0.277 0.258
R-factor 0.097 0.081
aBackbone RMSDs include NH, Cα, and C' atoms. Heavy atoms include all atoms except protons. RMSDs are pairwise RMSDs. R-factors are calculated using the R-factor R3 according to [46] including only signals arising from backbone protons.
Table 8 Comparison between model structures and experimental structures for Ppar γ
Structures Quantitiesa X-ray target structure
best model structure backbone RMSD [nm] 0.262
heavy atom RMSD [nm] 0.317
R-factor 0.260
model structure bundle backbone RMSD [nm] 0.299
heavy atom RMSD [nm] 0.355
R-factor 0.231
aBackbone RMSDs include NH, Cα, and C' atoms. Heavy atoms include all atoms except protons. RMSDs are pairwise RMSDs. R-factors are calculated using the R-factor R3 according to [46] including only signals arising from backbone protons.
==== Refs
Baker D A surprising simplicity to protein folding Nature 2000 405 39 42 10811210 10.1038/35011000
Bonneau R Baker D Ab initio protein structure prediction: progress and prospects Annu Rev Biophys Biomol Struct 2001 30 173 189 11340057 10.1146/annurev.biophys.30.1.173
Hardin C Pogorelov TV Luthey-Schulten Z Ab initio protein structure prediction Curr Opin Struct Biol 2002 12 176 181 11959494 10.1016/S0959-440X(02)00306-8
Murzin AG Brenner SE Hubbard TJP Chothia C SCOP: a structural classification of proteins database for the investigation of sequences and structures J Mol Biol 1995 247 536 540 7723011 10.1006/jmbi.1995.0159
Lo Conte L Brenner SE Hubbard TJP Chothia C Murzin AG SCOP database in 2002: refinements accommodate structural genomics Nucleic Acids Res 2002 30 264 267 11752311 10.1093/nar/30.1.264
Chothia C Lesk AM The relation between the divergence of sequence and structure in proteins EMBO J 1986 5 823 826 3709526
Sander C Schneider R Database of homology-derived protein structures and the structural meaning of sequence alignment Proteins 1991 9 56 68 2017436
Martin ACR MacArthur MW Thornton JM Assessment of comparative modeling in CASP2 Proteins 1997 Suppl 1 14 28 9485491 10.1002/(SICI)1097-0134(1997)1+<14::AID-PROT4>3.0.CO;2-O
Vitkup D Melamund E Moult J Sander C Completeness in structural genomics Nat Struct Biol 2001 8 559 566 11373627 10.1038/88640
Martin-Renom M Stuart RC Fiser A Sanchez R Melo F Sali A Comparitive Protein Structure Modeling of Genes and Genomes Annu Rev Biophys Biomol Struct 2000 29 291 325 10940251 10.1146/annurev.biophys.29.1.291
Al-Lazikani B Jung J Xiang Z Honig B Protein structure prediction Curr Opin Chem Biol 2001 5 51 56 11166648 10.1016/S1367-5931(00)00164-2
Westbrook J Feng Z Chen L Yang H Berman HM The Protein Data Bank and structural genomics Nucleic Acids Res 2003 31 489 491 12520059 10.1093/nar/gkg068
Orengo CA Bray JE Buchan DW Harrison A Lee D Pearl FM The CATH protein family database: A resource for structural and functional annotation of genomes Proteomics 2002 2 11 21 11788987 10.1002/1615-9861(200201)2:1<11::AID-PROT11>3.3.CO;2-K
Pearson WR Comparison of methods for searching protein sequence databases Protein Sci 1995 4 1145 1160 7549879
Altschul SF Gish W Miller W Myers EW Lipman DJ Basic local alignmnet search tool J Mol Biol 1990 215 403 410 2231712 10.1006/jmbi.1990.9999
Altschul SF Madden TL Schaffer AA Zhang J Zhang Z Miller W 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
David R Korenberg MJ Hunter IW 3D-1D threading methods for protein fold recognition Pharmacogenomics 2000 1 445 455 11257928 10.1517/14622416.1.4.445
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
Browne WJ North ACT Phillips DC Brew K Vanaman TC Hill RC A possible three-dimensional structure of bovine lactalbumin based on that of hen's egg-white lysosyme J Mol Biol 1969 42 65 86 5817651 10.1016/0022-2836(69)90487-2
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
Greer J Comparative modeling methods: application to the family of the mammalian serine proteases Proteins 1990 7 317 334 2381905
Jones TA Thirup S Using known substructures in protein model building and crystallography EMBO J 1986 5 819 822 3709525
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
Claessens M Van Cutsem C Lasters I Wodak S Modelling the polypeptide backbone with 'spare parts' from known protein structures Protein Eng 1989 2 335 345 2928296
Levitt M Accurate modeling of protein conformation by automatic segment matching J Mol Biol 1992 226 507 533 1640463 10.1016/0022-2836(92)90964-L
Havel TF Snow ME A new method for building protein conformations from sequence alignments with homologues of known structure J Mol Biol 1991 217 7 10.1016/0022-2836(91)90603-4
Srinivasan S March CJ Sudarsanam S An automated method for modeling proteins on known templates using distance geometry Prot Sci 1993 2 277 289
Sali A Blundell TL Comparitive protein modelling by satisfaction of spatial restraints J Mol Biol 1993 234 779 815 8254673 10.1006/jmbi.1993.1626
Brocklehurst SM Perham RN Prediction of the three-dimensional structures of the biotinylated domain from yeast pyruvate carboxylase and the of the lipoylated H-protein from the pea leaf glycine cleavage system: a new automated method for the prediction of protein tertiary structure Prot Sci 1993 2 626 639
Aszodi A Taylor WR Homology modelling by distance geometry Fold Des 1996 1 325 334 9080179
Kolinski A Betancourt MR Kihara D Rotkiewicz P Skolnick J Generalized comparitive modeling (GENECOMP): A combination of sequence comparison, threading, and lattice modeling for protein structure prediction and refinement Proteins 2001 44 133 149 11391776 10.1002/prot.1080
Güntert P Mumenthaler C Wüthrich K Torsion Angle Dynamics for NMR Structure Calculation with the New Program DYANA J Mol Biol 1997 273 283 298 9367762 10.1006/jmbi.1997.1284
Brünger AT Adams PD Clore GM DeLano WL Gros P Grossekunstleve RW Crystallography & NMR System: A New Software Suite for Macromolecular Structure Determination Acta Cryst 1998 D54 905 921
Döker R Maurer T Kremer W Neidig K-P Kalbitzer HR Determination of Mean and Standard Deviation of Dihedral Angles BBRC 1999 257 348 350 10198215
Zhang C Hou J Kim S-H Fold prediction of helical proteins using torsion angle dynamics and predicted restraints Proc Natl Acad Sci U S A 2002 99 3581 3585 11904420 10.1073/pnas.052003799
van Nuland NAJ Hangyi IW van Schaik RC Berendsen HJ van Gunsteren WF Scheek RM The High-resolution Structure of the Histidine-containing Phosphocarrier Protein HPr from Escherichia coli Determined by Restrained Molecular Dynamics from Nuclear Magnetic Resonance Nuclear Overhauser Effect Data J Mol Biol 1994 237 544 559 8158637 10.1006/jmbi.1994.1254
Jia Z Quail JW Waygood EB Delbaere LT The 2.0-A resolution structure of Escherichia coli histidine-containing phosphocarrier protein HPr. A redetermination J Biol Chem 1993 268 22490 22501 8226757
Uppenberg J Svensson C Jaki M Bertilsson G Jendeberg L Berkenstam A Crystal Structure of the Ligand Binding Domain of the Human Nuclear Receptor Ppargamma J Biol Chem 1998 273 31108 31112 9813012 10.1074/jbc.273.47.31108
Gronwald W Kalbitzer HR Automated structure determination of proteins by NMR spectroscopy Prog NMR Spectrosc 2004 44 33 96 10.1016/j.pnmrs.2003.12.002
Postma PW Lengeler JW Jacobson GR Phosphoenolpyruvate:carbohydrate phosphotransferase systems of bacteria Microbiol Rev 1993 57 543 594 8246840
Laskowski RA Rullmann JAC MacArthur MW Kaptein R Thornton JM AQUA and PROCHECK-NMR Programs for checking the quality of protein structures solved by NMR J Biomol NMR 1996 8 477 486 9008363 10.1007/BF00228148
Xu HE Lambert MH Montana VG Plunket KD Moore LB Collins JL Structural determinants of ligand binding selectivity between the peroxisome proliferator-activated receptors Proc Natl Acad Sci U S A 2001 98 13919 13924 11698662 10.1073/pnas.241410198
Xu HE Stanley TB Montana VG Lambert MH Shearer BG Cobb JE Structural basis for antagonist-mediated recruitment of nuclear co-repressors by PPARalpha Nature 2002 415 813 817 11845213
Cronet P Petersen JFW Folmer R Blomberg N Sjoblom K Karlsson U Structure of the PPARalpha and -gamma ligand binding domain in complex with AZ 242; ligand selectivity and agonist activation in the PPAR family Structure 2001 9 699 706 11587644 10.1016/S0969-2126(01)00634-7
Xu HE Lambert MH Montana VG Parks DJ Blanchard SG Brown PJ Molecular recognition of fatty acids by peroxisome proliferator-activated receptors Mol Cell 1999 3 397 403 10198642 10.1016/S1097-2765(00)80467-0
Gronwald W Kirchhofer R Gorler A Kremer W Ganslmeier B Neidig KP RFAC, a program for automated NMR R-factor estimation J Biomol NMR 2000 17 137 151 10921778 10.1023/A:1008360715569
Dominguez C Boelens R Bonvin AMJJ HADDOCK: A Protein-Protein Docking Approach Based on Biochemical or Biophysical Information J Am Chem Soc 2003 125 1731 1737 12580598 10.1021/ja026939x
Linge JP Habeck M Rieping W Nilges M ARIA: automated NOE assignment and NMR structure calculation Bioinformatics 2003 19 315 316 12538267 10.1093/bioinformatics/19.2.315
Linge JP Williams MA Spronk CAEM Bonvin AMJJ Nilges M Refinement of protein structures in explicit solvent Proteins 2003 50 496 506 12557191 10.1002/prot.10299
Nabuurs SB Nederveen AJ Vranken W Doreleijers JF Bonvin AMJJ Vuister GW DRESS: a Database of REfined Solution NMR Structures Proteins 2004 55 483 486 15103611 10.1002/prot.20118
Gröger C Möglich A Pons M Koch B Hengstenberg W Kalbitzer HR NMR-Spectroscopic Mapping of an Engineered Cavity in the I14A Mutant of HPr from Staphylococcus carnosus Using Xenon J Am Chem Soc 2003 125 8726 8727 12862458 10.1021/ja030113t
Möglich A Koch B Gronwald W Hengstenberg W Brunner E Kalbitzer HR Solution structure of the active-centre mutant I14A of the histidine-containing phosphocarrier protein from Staphlococcus carnosus Eur J Biochem 2004 271 4815 4824 15606769 10.1111/j.1432-1033.2004.04447.x
Tjandra NL Bax A Direct Measurement of Distances and Angles in Biomolecules by NMR in a Dilute Liquid Crystalline Medium Science 1997 278 1111 1114 9353189 10.1126/science.278.5340.1111
Brunner E Residual dipolar couplings in protein NMR Concepts Magn Reson 2001 13 238 259 10.1002/cmr.1012
Koradi R Billeter M Wüthrich K MOLMOL: a program for display and analysis of macromolecular structures J Mol Graphics 1996 14 51 55 10.1016/0263-7855(96)00009-4
Görler A Kalbitzer HR Relax, a Flexible Program for the Back Calculation of NOESY Spectra Based on Complete-Relaxation-Matrix Formalism J Magn Reson 1997 124 177 188 9169213 10.1006/jmre.1996.1033
Görler A Gronwald W Neidig KP Kalbitzer HR Computer assisted assignment of 13C or 15N edited 3D-NOESY-HSQC spectra using back calculated and experimental spectra J Magn Reson 1999 137 39 45 10053131 10.1006/jmre.1998.1614
Jia Z Vandonselaar M Hengstenberg W Quail JW Delbaere LT The 1.6 A structure of histidine-containing phosphotransfer protein HPr from Streptococcus faecalis J Mol Biol 1994 236 1341 1355 8126724 10.1016/0022-2836(94)90062-0
Maurer T Döker R Görler A Hengstenberg W Kalbitzer HR Three-dimensional structure of the histidine containing phosphocarrier protein (HPr) from Enterococcus faecalis in solution Eur J Biochem 2001 268 635 644 11168402 10.1046/j.1432-1327.2001.01916.x
Görler A Hengstenberg W Kravanja M Beneicke W Maurer T Kalbitzer HR Solution Structure of the Histidine-Containing Phosphocarrier Protein from Staphylococcus carnosus Appl Magn Reson 1999 17 465 480
Jones BE Rajagopal P Klevit RE Phosphorylation on histidine is accompanied by localized structural changes in the phosphocarrier protein, HPr from Bacillus subtilis Protein Sci 1997 6 2107 2119 9336834
| 15850481 | PMC1127111 | CC BY | 2021-01-04 16:38:05 | no | BMC Microbiol. 2005 Apr 25; 5:19 | latin-1 | BMC Microbiol | 2,005 | 10.1186/1471-2180-5-19 | oa_comm |
==== Front
Cardiovasc DiabetolCardiovascular Diabetology1475-2840BioMed Central London 1475-2840-3-91557163710.1186/1475-2840-3-9Original InvestigationThe need for a large-scale trial of fibrate therapy in diabetes: the rationale and design of the Fenofibrate Intervention and Event Lowering in Diabetes (FIELD) study. [ISRCTN64783481] The FIELD Study Investigators [email protected] NHMRC Clinical Trials Centre, Mallett St Campus, University of Sydney NSW 2006, Australia2004 1 12 2004 3 9 9 28 9 2004 1 12 2004 Copyright © 2004 The FIELD Study Investigators; licensee BioMed Central Ltd.2004The FIELD Study Investigators; 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
Fibrates correct the typical lipid abnormalities of type 2 diabetes mellitus, yet no study, to date, has specifically set out to evaluate the role of fibrate therapy in preventing cardiovascular events in this setting.
Methods
Subjects with type 2 diabetes, aged 50–75 years, were screened for eligibility to participate in a long-term trial of comicronized fenofibrate 200 mg daily compared with matching placebo to assess benefits of treatment on the occurrence of coronary and other vascular events. People with total cholesterol levels 3.0–6.5 mmol/L plus either a total-to-HDLc ratio >4.0 or triglyceride level >1.0 mmol/L with no clear indication for lipid-modifying therapy were eligible.
Results
A total of 9795 people were randomized into the Fenofibrate Intervention and Event Lowering in Diabetes (FIELD) trial. All received dietary advice, followed by a 6-week single-blind placebo run-in, then a 6-week active run-in period before randomization. Participants are being followed up every 6 months for outcome events and safety assessments. The study is designed to yield at least 500 coronary events (primary endpoint: first nonfatal myocardial infarction or coronary death) over 5 years, to have 80% power to identify as statistically significant at 2P = 0.05 a 22% reduction in such events, using intention-to-treat methods.
Conclusions
Type 2 diabetes is the most common endocrine disorder worldwide, and its prevalence is increasing. The current evidence about use of fibrates in type 2 diabetes, from around 2000 people treated, will increase with FIELD to evidence from around 12000. FIELD will establish the role of fenofibrate treatment in reducing cardiovascular risk in people with type 2 diabetes. The main results are expected to be available in late 2005.
diabetes mellitus, type 2fibratecardiovascular diseaserandomized controlled trialcoronary heart disease
==== Body
Introduction
Type 2 diabetes mellitus is an increasingly common condition associated with a high cardiovascular risk. To date, very few trials of lipid-lowering therapy have focused on this condition, and in particular, no large trials of fibrate therapy in diabetes have been conducted. As fibrates are known to correct the typical dyslipidaemia of diabetes, their role in cardiovascular risk reduction in diabetes may be especially important. The Fenofibrate Intervention and Event Lowering in Diabetes (FIELD) study is a multicentre, double-blind, placebo-controlled trial evaluating the effects on coronary morbidity and mortality of long-term treatment with fenofibrate to elevate high-density lipoprotein (HDL) cholesterol levels and lower triglyceride (TG) levels in patients with type 2 diabetes and total blood cholesterol between 3 and 6.5 mmol/L (115 and 250 mg/dL) at study entry. In type 2 diabetes, rates of coronary heart disease (CHD) are 3 to 4 times higher than those of persons without diabetes at any given level of blood cholesterol, and at any given age [1,2]. Evidence also suggests that in women with diabetes the natural protection against CHD afforded by sex may be lost [3,4]. Further, people with type 2 diabetes have both higher in-hospital mortality after myocardial infarction (MI) and a poorer outcome in the subsequent years [5,6], losing on average between 5 and 10 years of life expectancy. It follows that type 2 diabetes contributes significantly to the overall burden of premature CHD morbidity and mortality, far in excess of its prevalence in the community.
Diabetes and blood lipids
Blood total cholesterol levels are not substantially different between patients with type 2 diabetes and those of nondiabetic populations of similar age and sex [7]. However, evaluation of other lipoprotein fractions shows that those with diabetes more often have a below-average HDL cholesterol level and elevation of TG levels in the blood [8,9], which together confer an independent additional risk of CHD [10,11]. Furthermore, although low-density lipoprotein (LDL) cholesterol levels are not substantially raised, the LDL particle is often smaller and denser than in similar nondiabetic populations, which is considered to be a more atherogenic state [12]. An increased number of LDL particles, as seen in diabetes, is reflected in an elevated level of plasma apolipoprotein B, a more powerful predictor of risk for cardiovascular events than either total cholesterol or LDL cholesterol [13].
The strength of the cholesterol-CHD relationship is very similar for those with type 2 diabetes as for nondiabetics, although at a higher background rate of CHD [2]. Evidence from the Helsinki Heart Study [14], which tested long-term fibrate (gemfibrozil) use in hypercholesterolaemic men and women without prior coronary disease, showed a significant reduction in coronary events, with the reduction among the small numbers of people with diabetes not being separately significant, but appearing somewhat greater [15]. The reductions in events observed were greater than would have been expected on the basis of lowering of LDL cholesterol alone. So, whether substantially increasing low HDL cholesterol levels and reducing elevated triglyceride levels independently reduces cardiovascular events and mortality and should be a specific target for therapy remains less well agreed.
Why a large trial of fibrates?
For patients with type 2 diabetes and its typical dyslipidaemia, many physicians believe that fibrates are the logical first choice of drug treatment. The fibrates have been in clinical use for a long time, being well tolerated and with few short-term side-effects. Fenofibrate has been widely used and marketed for more than 20 years and is an effective agent for reducing plasma triglyceride and raising HDL cholesterol [16]. Although the effects on lipid fractions may vary with the population under study, a fall of 15% or more in total cholesterol, mediated through a reduction in LDL cholesterol, is often seen with long-term use [16]. In parallel, HDL cholesterol elevation of 10–15% is common, together with large reductions in plasma triglycerides of 30–40%. In addition, a reduction in plasma fibrinogen of about 15% has been observed [16].
FIELD is designed to provide the first properly randomized evidence as to whether the substantial effects of fenofibrate confer a benefit on clinical cardiovascular events in persons with type 2 diabetes. A clearly favourable result might be expected to help physicians determine which type of lipid-modifying drug therapy is likely to be most cost-effective for such people.
The FIELD study design
FIELD is a randomized, double-blind, placebo-controlled parallel-group trial among middle-aged to elderly people with type 2 diabetes mellitus considered to be at increased risk of CHD. Those with and without pre-existing vascular disease or other lipid abnormalities, such as low HDL cholesterol and elevated TG, were eligible, provided the total blood cholesterol level at screening fell between 3.0 and 6.5 mmol/L (about 115–250 mg/dL) plus either a total-to-HDL cholesterol ratio of >4.0 or a blood TG level >1.0 mmol/L (88.6 mg/dL) (Table 1). The study is being conducted in 63 clinical centres in Australia (39), Finland (9) and New Zealand (15) (see Appendix).
Table 1 Inclusion and exclusion criteria for the FIELD study
Individuals were eligible for this study provided they had the following characteristics:
• male or female, aged 50–75 years inclusive
• non-insulin dependent diabetes mellitus (type 2) with age at diagnosis >35 years (currently using any of diet, tablets or insulin); for Maori, Pacific Islanders, Australian Aborigines and Torres Strait Islanders, the eligible age of diagnosis was >25 years, provided there had been at least 1 year of treatment without insulin
• on the basis of diabetes, considered to be at higher risk for coronary heart disease than the general population
• no clear indication for any cholesterol-lowering treatment: the patient was not already taking any cholesterol-lowering drug and neither the patient nor the patient's doctor considered there to be any definite need to do so
• total cholesterol level 3 to 6.5 mmol/L, plus either
a total cholesterol-to-HDL cholesterol ratio of ≥ 4.0
a blood triglyceride level >1.0 mmol/L
• no clear contraindication to study therapy in the view of the treating physician
• no other predominant medical problem that might limit compliance with 5 years of study treatment or compromise long-term participation and clinic attendance in the trial
Individuals were not eligible if they had any of the following characteristics:
• serum triglyceride >5 mmol/L in the baseline visit fasting blood sample
• concurrent treatment with any other lipid-lowering agent
• serum creatinine >130 μmol/L
• known chronic liver disease, transaminases >2 × upper limit of normal or symptomatic gall-bladder disease
• myocardial infarction or hospital admission for unstable angina within 3 months
• female, of child-bearing potential, unless sterilized or on reliable approved methods of contraception, including oral contraceptives
• concurrent cyclosporin treatment (or a condition likely to result in organ transplantation and the need for cyclosporin during the next 5 years)
• known allergy to any fibrate drug or known photosensitivity
• unwilling or unable to consent to enter the study, with the understanding that follow-up was planned to continue for more than 5 years
The underlying principle guiding recruitment of patients into the study was that of clinical uncertainty: that is, patients were only to be considered if the patients' treating physicians were substantially uncertain about the value of lipid-modifying therapy for that particular individual and felt that there was no indication for lipid-modifying therapy. Therefore, none of the participants was on lipid-lowering therapy at study entry.
Following clinical and laboratory screening for eligibility, informed consent, and completion of the run-in period, patients were randomized to receive either fenofibrate (200 mg comicronized formulation) or matching placebo as one capsule daily with breakfast. There was no formal restriction on randomization related to compliance during the run-in period. Randomization was carried out using a dynamic allocation method [17] with stratification for important prognostic factors, including age, sex, prior MI, lipid levels and urinary albumin excretion. All patients are being followed up through regular clinic visits to a clinic set in place for the purposes of the study as well as by routine health care provided by a regular diabetes clinic or specialist.
The run-in phase for the study consists of a 4-week diet-only period, followed by a 6-week single-blind placebo period, then a 6-week single-blind active run-in period on comicronized fenofibrate 200 mg once daily for all patients, before randomization (Figure 1). This was to allow patients time to discuss long-term participation with their families and their usual doctors and for evaluation of the benefits of fenofibrate treatment on a background of recommended dietary advice. Further, the active run-in period was to determine to what extent any long-term clinical benefits of treatment correlate with the short-term effects of the drug to modify different lipid fractions.
Figure 1 Study flow schema. CVD = cardiovascular disease.
Follow-up in the study will be for not less than 5 years of median duration and until a total of at least 500 first coronary events have accumulated in the trial, unless the study is terminated earlier by advice from the Safety and Data Monitoring Committee.
Study outcomes
The principal study outcome is the combined incidence of first nonfatal MI or CHD death among all randomized patients during the scheduled treatment period (Table 2). Secondary outcomes include the effects of comicronized fenofibrate on major cardiovascular events (CHD events, total stroke and other cardiovascular death combined), total cardiovascular events (major cardiovascular events plus coronary and carotid revascularization), CHD death, total cardiovascular deaths, haemorrhagic and nonhaemorrhagic stroke, coronary and peripheral revascularization procedures, cause-specific non-CHD mortality (including cancer, suicide), and total mortality. All deaths, possible MIs and possible strokes are adjudicated in blinded fashion by the Outcomes Assessment Committee.
Table 2 Definitions for primary outcome assessment in the FIELD study
Myocardial infarction
Definite myocardial infarction = criterion 1; or any two of criteria 2 to 4; or criterion 5
1. New Q waves: new pathological Q waves (or Q-S pattern) of at least 0.03 seconds in width in at least 2 leads in the same lead group (in the absence of left bundle branch block)
2. Evolutionary ST-T wave changes: evolution of an injury current lasting more than one day and present in at least 2 leads in the same lead group: for example, ST elevation of 2 mm or more in anterior leads, or 1 mm or more in inferior leads followed by T-wave inversion of 1 mm or more; this requires a minimum of two traces taken at least one day apart
3. Ischemic pain: history of typical ischemic pain lasting for at least 15 minutes and unresponsive to sublingual nitrates (if given)
4. Biochemical markers: elevation of CK or CKMB enzymes to over twice the upper limit of normal (for the laboratory) after the attack or elevation of troponin T to more than 0.1 μg/L or troponin I to levels above the upper limit of normal (for the laboratory)
5. Postmortem diagnosis: autopsy showing evidence of acute myocardial infarction.
Death
Coronary heart disease death = any of 1.1 to 1.7
1. Coronary
1.1 Definite fatal myocardial infarction: death following definite acute myocardial infarction in the preceding 28 days (and without an unrelated noncoronary cause of death), or autopsy-proven recent acute myocardial infarction
1.2 Sudden cardiac death: death occurring within one hour of onset of new cardiac symptoms or unwitnessed death after last having been seen without new symptoms; in each case, without any noncoronary disease that could have been rapidly fatal and without having been confined to hospital or other institution because of illness within 24 hours of death
1.3 Possible myocardial infarction: death in hospital with possible myocardial infarction (that is, typical ischaemic pain and ECG and enzyme results do not fulfil the criteria for definite myocardial and there is no good evidence for another event)
1.4 Resuscitated sudden death: documented cardiac arrest (in or out of hospital), after being resuscitated from what would have been sudden death; patient lives for more than one hour (hours to weeks).
1.5 Heart failure: death due to heart failure (prior grade 3–4 dyspnoea, NYHA) without any defined noncoronary cause
1.6 Death after coronary revascularisation: death (in the same admission) after any coronary revascularisation procedure (CABG or PTCA).
1.7 Other coronary: death where the underlying cause is certified as coronary (and there is no evidence for a noncoronary cause of death, clinically or at autopsy)
2. Noncoronary cardiac: death for which the underlying cause is certified as noncoronary cardiac disease
3. Vascular (noncardiac): death which is certified as vascular but not coronary disease: for example, cerebrovascular accident, pulmonary embolism, complications of peripheral vascular disease or uncontrolled hypertension
4. Cancer: death for which the underlying cause is certified as malignant neoplasm
5. Trauma: death where the underlying cause is certified as a wound or injury either accidental or inflicted
6. Suicide: death for which the underlying cause is certified as deliberate and voluntary taking of one's own life
7. Other: other cause of death not specified above.
Tertiary outcomes include the effects of treatment on development of vascular and neuropathic amputations, nonfatal cancers, the progression of renal disease, laser treatment for diabetic retinopathy, hospitalization for angina pectoris, and numbers and duration of all hospital admissions. The effects of treatment on the outcome of total cardiovascular events will be examined inThe rates of events various subgroups of particular interest, such as men and women, those <65 years and ≥ 65 years of age, by subgroup of each of baseline total cholesterol, HDL cholesterol, triglyceride and fibrinogen, baseline insulin use, or not, and the presence, or absence, at baseline of microalbuminuria.
The primary analysis will be of time to first study outcome, using standard log-rank methods [18,19], and where appropriate, proportional-hazards models with adjustment for covariates. Intention-to-treat methods, comparing all those allocated to comicronized fenofibrate with all those allocated to placebo, will be used.
Sample size
The rates of events used for the original study power calculations were based on information from a variety of sources. During recruitment, when the numbers of participants with prior MI was falling well short of the number originally planned (in about 2000), the sample size was extended from the original total of 8000 to a final number of 9795 reached in November 2000. In late 2002, the statistical power of the trial was reviewed again. These reviews were planned in the original protocol design and were undertaken by reviewers completely blinded to all treatment allocation. The reassessment included information on final sample size, overall rate of discontinuation of study medication and commencement of open-label cholesterol treatment, and overall event rates in relation to CHD death, MI, and stroke.
After the review it was clear that the trial would yield too few deaths from CHD to retain sufficient power, over its planned duration of around 5 years, to show a significant reduction in this endpoint. The FIELD Management Committee determined that the primary outcome of the trial should be amended from CHD death to CHD events (that is, CHD death plus nonfatal MI, a decision made in December 2002). It was also decided to change the principal outcome for subgroup analyses to look at the effects of fenofibrate in subjects with and without prior cardiovascular disease (CVD) (originally those with and without prior MI).
For a primary outcome of CHD events (CHD death plus nonfatal MI), it is projected that approximately 500 CHD events will have occurred when 5 years median follow-up has elapsed (during the first quarter of 2005); by this time the trial will have 80% power to detect an observed 22% reduction in CHD events (based on the intention-to-treat method of analysis). This will also provide 90% power to detect a 25% relative reduction in CHD events (based on intention-to-treat analysis). Both calculations allow for the effects of an average drop-out rate from active treatment over the course of the study of 10% and a similar drop-in rate of 10% from placebo to open cholesterol-lowering therapy (Table 3). These allowances for loss of compliance require an increase in sample size of approximately 60% when compared with a study with no drop-outs from, or drop-ins to, active treatment.
Table 3 Predicted numbers of events and corresponding power in the FIELD study among 9795 people with diabetes, based on a median follow-up of 5 years
Allocated treatment Power to detect effect at 2P < 0.05*
Risk category Fenofibrate Placebo
Primary prevention (7683 with no prior CVD)
Total CVD events† 298 385 93%
Secondary prevention (2112 with prior CVD)
Total CVD events 229 288 83%
Men (6139)
Total CVD events 392 503 98%
Women (3656)
Total CVD events 133 171 60%
All patients (9795)
Total CVD events 525 675 99%
Total CHD events 219 281 80%
* Calculations assume a reduction in risk for each endpoint of around 27% with full compliance, resulting in an observed risk reduction of approximately 22% on intention-to-treat analysis; the risk in the prior CVD group is 2.75 times that for patients with no prior CVD, the risk in men is 1.75 times that in women.
† Expanded endpoint 'total CVD events' comprises cardiovascular death, nonfatal myocardial infarction, coronary artery bypass grafting, percutaneous transluminal coronary angioplasty, stroke and carotid revascularization.
CVD = cardiovascular, CHD = coronary heart disease
If the uptake of cholesterol-lowering therapy proves to be greater in the placebo group than in the fenofibrate-allocated group, the observed treatment effect of fenofibrate will underestimate its true efficacy.
Safety and event monitoring
The trial has an independent Safety and Data Monitoring Committee to safeguard the patients' interests and to formally evaluate from time to time on a regular basis whether, for any reason, they would recommend that the study should be modified or stopped. Up to 5 formal interim analyses are planned, at time points to be determined by the Safety and Data Monitoring Committee, with a stringent nominal significance level (3 standard deviations; 2P = 0.003) to preserve an overall type 1 error probability of no more than 0.05. The events to be used for these analyses are counts of death from CHD. The Management Committee, the collaborators, the study sponsor and all the central administrative staff, with the exception of the unblinded statistician, will remain ignorant of the interim results for mortality and major morbidity. During the study, the group effects of treatment on biochemical parameters, such as lipid fractions, and other surrogate endpoints may be published, subject to prior approval of the Management Committee, provided that individual patient treatment assignments are not revealed. Patients are being monitored regularly by lipid profiles, liver function tests, creatine phosphokinase, fasting glucose, HbA1c, and urinary microalbumin. The study has been approved by local ethics committees at each participating institution, which also approved the information discussed and informed-consent procedures. The first patient in FIELD was registered in November 1997 and randomized in February 1998. The study has recruited 9795 patients; the final patient was randomized on 3 November 2000.
Study sponsorship and organisation
The main sponsor of the trial and supplier of the fenofibrate and matching placebo medication, is Laboratoires Fournier S.A., Dijon, France. The study is also supported by the National Health and Medical Research Council of Australia through Unit, Program and Fellowship grants to the NHMRC Clinical Trials Centre. The study is being coordinated independently of the sponsors by the NHMRC Clinical Trials Centre, University of Sydney, Sydney, Australia and overseen by the study Management Committee. The study has been endorsed by the National Heart Foundation of Australia, Diabetes Australia, the New Zealand Society for the Study of Diabetes, and the Finnish Diabetes Association.
Conclusion
In 1997, before the FIELD study commenced, the role of lipid modification in diabetes remained uncertain, except possibly for hypercholesterolaemic people with a prior MI. Two large-scale trials, the Scandinavian Simvastatin Survival Study (4S) [20] and the Cholesterol and Recurrent Events (CARE) [21] study, had showed that the use of 3-hydroxy-3-methylglutaryl-coenzyme A (HMG-CoA) reductase inhibitors, simvastatin and pravastatin, respectively, substantially reduced cardiovascular events hypercholesterolaemic and in general post-MI populations. But neither study included sufficient numbers of patients with diabetes (n = 202 and n = 586, respectively) to have the power to show reliably whether these benefits would translate into reductions in CHD mortality in the setting of diabetes, nor the effects in them of treatment on noncoronary events and mortality. Further, the West of Scotland (WOSCOPS) study of pravastatin in hypercholesterolaemic men with no prior CHD, which reported a marginally significant reduction in overall mortality, had fewer than 100 subjects with diabetes [22].
Since that time, numerous other trials of statin treatment have been reported, with randomized data now reported on over 18000 persons with diabetes. Those involving more than 1000 people with diabetes include the Long-Term Intervention with Pravastatin in Ischaemic Disease (LIPID) study [6,23], the Heart Protection Study [24,25], the Anglo-Scandinavian Cardiac Outcomes Trial (ASCOT) [26] and Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT-LLT) [27]. Another trial, the Collaborative Atorvastatin Diabetes Study (CARDS), has stopped early, after about 4 years of follow-up, with results showing clear benefits of reduced cardiac and stroke events of using atorvastatin among 2838 people with diabetes and high cardiovascular risk [28]. Important new results have been communicated to investigators and patients so that it can be considered whether, during the follow-up of FIELD, statin therapy is now indicated for any individual. The protocol allows for statin therapy to be added at any time after randomization and recommends continuing study medication; thus the study is evaluating the role of fenofibrate on a background of usual care. This feature of the study design will contribute to the evidence about the safety of combined statin and fibrate therapy.
Two large-scale trials of fibrate therapy have also been completed: the Veterans Low-HDL Cholesterol Intervention Trial (VA-HIT) [29,30] and the Bezafibrate Infarct Prevention (BIP) [31] trial. Both studies were limited to people with prior MI and have reported reductions in major cardiovascular events among participants with low HDL and high TG at baseline, which were greater than those seen with use of the same fibrate among those without dyslipidaemia. The VA-HIT trial also reported reduced CHD mortality in those with diabetes receiving gemfibrozil and a reduced rate of cardiovascular events, although rates of nonfatal MI did not change significantly [32]. A third trial, the Diabetes Atherosclerosis Intervention Study (DAIS), showed reduced progression of established coronary atherosclerosis among those randomized to fenofibrate compared with those receiving matching placebo, over 3 years [33].
At the same time, our understanding of the mechanism of action of fibrates has grown, with identification of the peroxisome proliferator-activated receptor alpha (PPAR-alpha) transcription factor as the primary pathway through which fibrate-mediated effects are triggered [34,35]. The abundance of desirable effects of PPAR-alpha activation by fibrates has generated extraordinary interest in their role in the prevention of atherosclerosis via regulation of lipid metabolism, vascular inflammation, and haemostatic factors. The importance of changes in apolipoprotein B and non-HDL-cholesterol levels appears greater with fibrate therapy than with statin use [36], particularly in patients with type 2 diabetes [37]. Increased interest in the FIELD study has resulted, as it will generate clinical data on similar numbers of persons with diabetes to that available for the statins (Table 4) and will enlarge the range of lipid profiles studied and the number of events in such populations (Table 5).
Table 4 Unconfounded randomized controlled trials of lipid-lowering therapy, showing numbers of subjects with diabetes
Study Population Year of primary publication Therapy Total no. No. with diabetes Reference
4S Prior CHD 1994 Simvastatin 20–40 mg 4444 202 20, 38
CARE Prior CHD 1996 Pravastatin 40 mg 4159 586 21
Post-CABG* Prior CHD 1997 Lovastatin 40–80 mg vs 2.5–5 mg 1351 122 39, 40
LIPID Prior CHD 1998 Pravastatin 40 mg 9014 1077 6, 23
GISSI-P* Prior CHD 2000 Pravastatin 20 mg 4271 582 41
GREACE* Prior CHD 2002 Atorvastatin 10–80 mg 1600 313 42
PROSPER Mixed 2002 Pravastatin 40 mg 5804 623 43
ALLHAT-LLT* Mixed 2002 Pravastatin 20–40 mg 10355 3638 27
HPS Mixed 2003 Simvastatin 40 mg 20536 5963 24, 25
ASCOT-LLA Mixed 2003 Atorvastatin 10 mg 10305 2532 26
WOSCOPS Primary 1995 Pravastatin 40 mg 6595 76 22
AFCAPS/TexCAPS Primary 1998 Lovastatin 20–40 mg 6605 1 55 44
CARDS Primary 2004 Atorvastatin 10 mg 2838 2838 28
Total – – Any statin 87877 18707
VA-HIT Prior CHD 1999 Gemfibrozil 1200 mg 2531 769 29, 30, 32
BIP Prior CHD 2000 Bezafibrate 400 mg 3090 309 31
DAIS Mixed 2001 Fenofibrate 200 mg 418 418 33
LEADER Mixed 2002 Bezafibrate 400 mg 1568 268 45
SENDCAP Primary 1998 Bezafibrate 400 mg 164 164 46
HHS Primary 1992 Gemfibrozil 1200 mg 4081 135 14, 15
Total – – Any fibrate 11852 2063
* No placebo used; lipid lowering compared with less treatment in Post-CABG study, with no treatment in GISSI-P study, and with usual care in ALLHAT-LLT and GREACE studies.
CHD = coronary heart disease
Table 5 Entry criteria and outcomes in trials of fibrates
Study Demographic entry criteria Lipid entry criteria Outcomes
SENDCAP men and women, 35–65 years, no cardiovascular disease • serum cholesterol ≥ 5.2 mmol/L
• triglyceride ≥ 1.8 mmol/L
• HDL ≤ 1.1 mmol/L
• total/HDL ≥ 4.7 change in carotid intima-media thickness, lipid changes, CHD events, at 3 years
VA-HIT men, <74 years, documented history of CHD • HDL ≤ 1.0 mmol/L
• LDL ≤ 3.6 mmol/L
• triglyceride ≤ 3.4 mmol/L nonfatal MI or CHD death over median 5.1 years
BIP men and women, 45–74 years, MI 6 months to 5 years before, no insulin-dependent diabetes • serum cholesterol 4.7-6.5 mmol/L
• LDL ≤ 4.7 mmol/L
• HDL ≤ 1.2 mmol/L
• triglyceride ≤ 3.4 mmol/L fatal or nonfatal MI or sudden death over mean 6.2 years
DAIS men and women, 40–65 years, type 2 diabetes • total/HDL ≥ 4
• LDL 3.5–4.5 mmol/L + triglyceride ≤ 5.2 mol/L or LDL ≤ 4.5 mmol/L + triglyceride 1.7–5.2 mol/L change in coronary artery lumen diameters, by angiography, and lipid changes after 3 years
LEADER men with lower-extremity arterial disease, no lipid-lowering drug • serum cholesterol 3.5–8.0 mmol/L CHD events and stroke over median 4.6 years
HHS men, 40–55 years, asymptomatic • non-HDL cholesterol >5.2 mmol/L lipid changes, MI, cardiac death at 5 years
CHD = coronary heart disease, MI = myocardial infarction, HDL = high-density lipoprotein, LDL = low-density lipoprotein
Approximately 140 million adults were estimated to be suffering from diabetes mellitus, the most common endocrine disorder worldwide, in 1997. By 2010, projections put diabetes prevalence about 60 percent higher, at 221 million. Just as many persons again have an elevated fasting glucose level, or impaired fasting glucose, which can progress rapidly to diabetes. Without the FIELD study, doctors would remain uncertain about the merits of using a fibrate when confronted with a patient with diabetes at risk of clinical CHD. It is expected that the main results of FIELD will be reported in late 2005.
Declaration of competing interests
Of the Management Committee of the FIELD Study:
PB, YAK and RS have received reimbursements, fees, funding, or salary in the past five years from an organization that may in any way gain or lose financially from the publication of this paper;
No authors hold or have held stocks or shares in such an organization;
No authors have other financial competing interests;
AK has the following nonfinancial competing interests: Advisory board membership.
Authors' contributions
The FIELD Management Committee conceived and developed the study protocol and are the responsible authors of this manuscript.
Appendix 1: Study organization
Management Committee
P Barter*, J Best*, P Colman, M d’Emden, T Davis, P Drury, C Ehnholm, P Glasziou, D Hunt, A Keech* (study chairman and principal investigator), YA Kesaniemi, M Laakso, R Scott*, RJ Simes*, D Sullivan, M-R Taskinen*, M Whiting; J-C Ansquer, B Fraitag (non-voting sponsor representatives). * Executive Committee members.
Outcomes Assessment Committee
N Anderson, G Hankey, D Hunt (chairman), S Lehto, S Mann, M Romo; LP Li (outcomes officer, in attendance).
Safety and Data Monitoring Committee
C Hennekens, S MacMahon (chairman), S Pocock, A Tonkin, L Wilhelmsen; P Forder (unblinded statistician, in attendance).
Site principal investigators
Australia: H Akauola, F Alford, P Barter, I Beinart, J Best, S Bohra, S Boyages, P Colman, H Connor, D Darnell, T Davis, P Davoren, F Lepre, F De Looze, M d'Emden, A Duffield, R Fassett, J Flack, G Fulcher, S Grant, S Hamwood, D Harmelin, R Jackson, W Jeffries, M Kamp, L Kritharides, L Mahar, V McCann, D McIntyre, R Moses, H Newnham, G Nicholson, R O'Brien, K Park, N Petrovsky, P Phillips, G Pinn, D Simmons, K Stanton, B Stuckey, D R Sullivan, M Suranyi, M Suthers, Y Tan, M Templer, D Topliss, J H Waites, G Watts, T Welborn, R Wyndham; Finland: H Haapamaki, A Kesaniemi, M Laakso, J Lahtela, H Levanen, J Saltevo, H Sodervik, M Taskinen, M Vanhala; New Zealand: J Baker, A Burton, P Dixon, J Doran, P Drury, P Dunn, N Graham, A Hamer, J Hedley, J Lloyd, P Manning, I McPherson, S Morris, C Renner, R Scott, R Smith, M Wackrow, S Young.
Co-investigators and site coordinators
Australia: F Alard, J Alcoe, F Alford, C Allan, J Amerena, R Anderson, N Arnold, T Arsov, D Ashby, C Atkinson, L Badhni, M Balme, D Barton, B Batrouney, C Beare, T Beattie, J Beggs, C Bendall, C Bendall, A Benz, A Bond, R Bradfield, J Bradshaw, S Brearley, D Bruce, J Burgess, J Butler, M Callary, J Campbell, K Chambers, J Chow, S Chow, K Ciszek, P Clifton, P Clifton-Bligh, V Clowes, P Coates, C Cocks, S Cole, D Colquhoun, M Correcha, B Costa, S Coverdale, M Croft, J Crowe, S Dal Sasso, W Davis, J Dunn, S Edwards, R Elder, S El-Kaissi, L Emery, M England, O Farouque, M Fernandez, B Fitzpatrick, N Francis, P Freeman, A Fuller, D Gale, V Gaylard, C Gillzan, C Glatthaar, J Goddard, V Grange, T Greenaway, J Griffin, A Grogan, S Guha, J Gustafson, P S Hamblin, T Hannay, C Hardie, A Harper, G Hartl, A Harvey, S Havlin, K Haworth, P Hay, L Hay, B Heenan, R Hesketh, A Heyworth, M Hines, G Hockings, A Hodge, L Hoffman, L Hoskin, M Howells, D Hunt, A Hunt, W Inder, W Inder, D Jackson, A Jovanovska, K Kearins, P Kee, J Keen, D Kilpatrick, J Kindellan, M Kingston-Ray, M Kotowicz, A Lassig, M Layton, S Lean, E Lim, F Long, L Lucas, D Ludeman, D Ludeman, C Ludeman-Robertson, M Lyall, L Lynch, C Maddison, B Malkus, A Marangou, F Margrie, K Matthiesson, J Matthiesson, S Maxwell, K McCarthy, A McElduff, H McKee, J McKenzie, K McLachan, P McNair, M Meischke, A Merkel, C Miller, B Morrison, A Morton, W Mossman, A Mowat, J Muecke, P Murie, S Murray, P Nadorp, S Nair, J Nairn, A Nankervis, K Narayan, N Nattrass, J Ngui, S Nicholls, V Nicholls, JA Nye, E Nye, D O'Neal, M O'Neill, S O'Rourke, J Pearse, C Pearson, J Phillips, L Pittis, D Playford, L Porter, L Porter, R Portley, M Powell, C Preston, S Pringle, W A Quinn, J Raffaele, G Ramnath, J Ramsden, D Richtsteiger, S Roffe, S Rosen, G Ross, Z Ross, J Rowe, D Rumble, S Ryan, J Sansom, C Seymour, E Shanahan, S Shelly, J Shepherd, G Sherman, R Siddall, D Silva, S Simmons, R Simpson, A Sinha, R Slobodniuk, M Smith, P Smith, S Smith, V Smith-Orr, J Snow, L Socha, T Stack, K Steed, K Steele, J Stephensen, P Stevens, G Stewart, R Stewart, C Strakosch, M Sullivan, S Sunder, J Sunderland, E Tapp, J Taylor, D Thorn, D Thorn, A Tolley, D Torpy, G Truran, F Turner, J Turner, J van de Velde, S Varley, J Wallace, J Walsh, J Walsh, J Walshe, G Ward, B Watson, J Watson, A Webb, F Werner, E White, A Whitehouse, N Whitehouse, S Wigg, J Wilkinson, E Wilmshurst, D Wilson, G Wittert, B Wong, M Wong, S Worboys, S Wright, S Wu, J Yarker, M Yeo, K Young, J Youssef, R Yuen, H Zeimer, R W Ziffer; Finland: A Aura, A Friman, J Hanninen, J Henell, N Hyvarinen, M Ikonen, A Itkonen, J Jappinen, A Jarva, T Jerkkola, V Jokinen, J Juutilainen, H Kahkonen, T Kangas, M Karttunen, P Kauranen, S Kortelainen, H Koukkunen, L Kumpulainen, T Laitinen, M Laitinen, S Lehto, R Lehto, E Leinonen, M Lindstron-Karjalainen, A Lumiaho, J Makela, K Makinen, L Mannermaa, T Mard, J Miettinen, V Naatti, S Paavola, N Parssinen, J Ripatti, S Ruotsalainen, A Salo, M Siiskonen, A Soppela, J Starck, I Suonranta, L Ukkola, K Valli, J Virolainen; New Zealand: P Allan, W Arnold, W Bagg, K Balfour, T Ball, B Ballantine, C Ballantyne, C Barker, C Barker, F Bartley, E Berry, G Braatvedt, A Campbell, T Clarke, R Clarke, A Claydon, S Clayton, P Cresswell, R Cutfield, J Daffurn, J Delahunt, A Dissnayake, C Eagleton, C Ferguson, C Florkowski, D Fry, P Giles, M Gluyas, C Grant, P Guile, M Guolo, P Hale, M Hammond, M Hammond, P Healy, M Hills, J Hinge, J Holland, B Hyne, A Ireland, A Johnstone, S Jones, G Kerr, K Kerr, M Khant, J Krebs, L Law, B Lydon, K MacAuley, R McEwan, P McGregor, B McLaren, L McLeod, J Medforth, R Miskimmin, J Moffat, M Pickup, C Prentice, M Rahman, E Reda, C Ross, A Ryalls, D Schmid, N Shergill, A Snaddon, H Snell, L Stevens, A Waterman, V Watts.
Coordinating centre teams
NHMRC Clinical Trials Centre, Sydney: K Jayne, E Keirnan, P Newman, G Ritchie, A Rosenfeld (project directors), E Beller, P Forder, V Gebski, A Pillai (study statisticians), C Anderson, S Blakesmith, S-Y Chan, S Czyniewski, A Dobbie, S Doshi, A Dupuy, S Eckermann, M Edwards, N Fields, K Flood, S Ford, C French, S Gillies, C Greig, M Groshens, J Gu, Y Guo, W Hague, S Healy, L Hones, Z Hossain, M Howlett, J Lee, L-P Li, T Matthews, J Micallef, A Martin, I Minns, A Nguyen, F Papuni, A Patel, J Pearse, R Pike, M Pena, K Pinto, D Schipp, J Schroeder, B Sim, C Sodhi, T Sourjina, C Sutton, R Taylor, P Vlagsma, S Walder, R Walker, W Wong, J Zhang, B Zhong, A Keech (deputy director), RJ Simes (director); Helsinki Project Office: A Kokkonen, P Narva, E-L Niemi, A Salo, A-M Syrjanen, M-R Taskinen (director); Christchurch Project Office: C Lintott, R Scott (director).
Central laboratories
Adelaide: R Tirimacco, M Whiting; Helsinki: C Ehnholm, M Ikonen, M Kajosaari, L Raman, J Sundvall, M Tukianen. .
Laboratoires Fournier SA liaison: Dijon: J-C Ansquer, B Fraitag, D Crimet, I SirugueSydney: P Aubonnet.
Acknowledgements
This study is supported by a grant from Laboratoires Fournier SA, Dijon, France and is being coordinated independently by the National Health and Medical Research Council Clinical Trials Centre, University of Sydney, Sydney, Australia, and overseen by the study Management Committee.
The study is also supported by the National Health and Medical Research Council (NHMRC), Australia (Unit grant, Project grant and Fellowships to A. Keech and J. Simes), without which it would not be possible. We thank the National Heart Foundation, Australia, Diabetes Australia, Diabetes New Zealand, and the Finnish Diabetes Association for endorsing the study. Investigators express their thanks to Rhana Pike and Christelle Foucher for their assistance with the preparation of this manuscript, and the staff of Kadima, Sydney, for efficient management and distribution of study materials. Finally, the many patients participating in the FIELD study are thanked for their untiring contributions.
==== Refs
Barrett-Connor E Orchard T Diabetes and heart disease Diabetes Data Compiled 1984. National Diabetes Data Group. NIH publication no. 85-1468. Washington DC: US Department of Health and Human Services, XVI-XVI-41 1985
Stamler J Vaccaro O Neaton JD Westworth D Diabetes and other risk factors and the 12-year cardiovascular mortality for men screened for the Multiple Risk Factor Intervention Trial Diabetes Care 1993 16 434 444 8432214
Kannel WB Lipids, diabetes and coronary disease from the Framingham Study Am Heart J 1985 110 1100 1107 4061265 10.1016/0002-8703(85)90224-8
Barrett-Connor EL Cohn BA Wingard DL Edelstein SL Why is diabetes mellitus a stronger risk factor for fatal ischaemic heart disease in women than in men? The Rancho Bernardo Study JAMA 1991 265 627 631 1987413 10.1001/jama.265.5.627
Zuanetti G Latini R Maggioni AP Santoro L Franzosi MG Influence of diabetes on mortality in acute myocardial infarction: data from the GISSI-2 study J Am Coll Cardiol 1993 22 1788 1794 8245329
Keech AC Colquhoun D Best J Kirby A Simes RJ Hunt D Hague W Beller E Arulchelvam M Baker J Tonkin A for the LIPID study group Secondary prevention of cardiovascular events with long-term pravastatin in patients with diabetes or impaired fasting glucose: results from the LIPID trial Diabetes Care 2003 26 2713 2721 14514569
Laakso M Voutilainen E Sarlund H Aro A Pyorala K Penttila I Serum lipids and lipoproteins in middle aged non-insulin dependent diabetics Atherosclerosis 1985 56 271 281 3902037
Wilson PWF Kannel WB Anderson KM Lipids, glucose intolerance and vascular disease: The Framingham Study Monogr Atheroscler 1985 13 1 11 4088270
Howard BV Lipoprotein metabolism in diabetes mellitus J Lipid Res 1987 28 613 628 3302085
Fontbonne A Eschwege E Cambien F Richard JL Ducimetiere P Thibult N Warnet JM Claude JR Rosselin GE Hypertriglyceridemia as a risk factor for coronary heart disease mortality in subjects with impaired glucose tolerance or diabetes Diabetologia 1989 32 300 304 2666216 10.1007/BF00265546
Assmann G Schulte H Triglycerides and atherosclerosis; results from the Prospective Cardiovascular Munster Study Arterioscler Rev 1991 22 51 57
Feingold KR Grunfeld C Pang M Doerrler W Krauss RM LDL subclass phenotypes and triglyceride metabolism in non-insulin dependent diabetes Arterioscler Thromb 1992 12 1496 1502 1450181
Sniderman AD Furberg CD Keech A Roeters van Lennep JE Frohlich J Jungner I Walldius G Apolipoproteins versus lipids as indices of coronary risk and as targets for statin treatment Lancet 2003 361 777 780 12620753 10.1016/S0140-6736(03)12663-3
Frick MH Elo O Haapa K Heinonen OP Heinsalmi P Helo P Huttunen JK Kaitaniemi P Koskinen P Manninen V and others Helsinki Heart Study: primary prevention trial with gemfibrozil in middle-aged men with dyslipidemia. Safety treatment, changes in risk factors, and incidence of coronary heart disease N Engl J Med 1987 317 1237 1245 3313041
Koskinen P Mänttäri M Manninen V Huttunen JK Heinonen OP Frick MH Coronary heart disease incidence in NIDDM patients in the Helsinki Heart Study Diabetes Care 1992 15 820 825 1516498
Keating GM Ormrod D Micronised fenofibrate. An updated review of its clinical efficacy in the management of dyslipidaemia Drugs 2002 62 1909 1944 12215067
Signorini DF Leung O Simes RJ Beller E Gebski VJ Callaghan T Dynamic balanced randomizatio for clinical trials Stat Med 1993 12 2343 2250 8134737
Peto R Pike MC Armitage P Breslow NE Cox DR Howard SV Mantel N McPherson K Peto J Smith PG Design and analysis of randomised clinical trials requiring prolonged observation of each patient. I. Introduction and design Br J Cancer 1976 34 585 612 795448
Peto R Pike MC Armitage P Breslow NE Cox DR Howard SV Mantel N McPherson K Peto J Smith PG Design and analysis of randomized clinical trials requiring prolonged observation of each patient. II. Analysis and examples Br J Cancer 1977 35 1 39 831755
Scandinavian Simvastatin Survival Study Group Randomised trial of cholesterol lowering in 4444 patients with coronary heart disease; the Scandinavian Simvastatin Survival Study (4S) Lancet 1994 344 1383 1389 7968073
Sacks FM Pfeffer MA Moye LA Rouleau JL Rutherford JD Cole TG Brown L Warnica JW Arnold JM Wun CC Davis BR Braunwald E The effect of pravastatin on coronary events after myocardial infarction in patients with average cholesterol levels N Engl J Med 1996 335 1001 1009 8801446 10.1056/NEJM199610033351401
Shepherd J Cobbe SM Ford I Isles CG Lorimer AR MacFarlane PW McKillop JH Packard CJ Prevention of coronary heart disease with pravastatin in men with hypercholesterolemia. West of Scotland Coronary Prevention Study Group N Engl J Med 1995 333 1301 1307 7566020 10.1056/NEJM199511163332001
LIPID Study Group Prevention of cardiovascular events and death with pravastatin in patients with coronary heart disease and a broad range of initial cholesterol levels N Engl J Med 1998 339 1349 1357 9841303 10.1056/NEJM199811053391902
Heart Protection Study Collaborative Group MRC/BHF Heart Protection Study of cholesterol lowering with simvastatin in 20 536 high-risk individuals: a randomised placebo-controlled trial Lancet 2002 360 7 22 12114036 10.1016/S0140-6736(02)09327-3
Heart Protection Study Collaborative Group MRC/BHF Heart Protection Study of cholesterol lowering with simvastatin in 5963 people with diabetes: a randomised placebo-controlled trial Lancet 2003 361 2005 2016 12814710 10.1016/S0140-6736(03)13636-7
Sever PS Dahlöf B Poulter NR Wedel H Beevers G Caulfield M Collins R Kjeldsen SE Kristinsson A McInnes GT Mehlsen J Nieminen M O'Brien E Ostergren J ASCOT investigators Prevention of coronary and stroke events with atorvastatin in hypertensive patients who have average or lower-than-average cholesterol concentrations, in the Anglo-Scandinavian Cardiac Outcomes Trial-Lipid Lowering Arm (ASCOT-LLA): a multicentre randomised controlled trial Lancet 2003 361 1149 1158 12686036 10.1016/S0140-6736(03)12948-0
ALLHAT Officers and Coordinators for the ALLHAT Collaborative Research Group: The Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial: Major outcomes in moderately hypercholesterolemic, hypertensive patients randomized to pravastatin vs usual care: The Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT-LLT) JAMA 2002 288 2998 3007 12479764 10.1001/jama.288.23.2998
Colhoun HM Betteridge DJ Durrington PN Hitman GA Neil HAW Livingstone SJ Thomason MJ Mackness MI Charlton-Menys V Fuller JH on behalf of the CARDS Investigators Primary prevention of cardiovascular disease with atorvastatin in type 2 diabetes in the Collaborative Atorvastatin Diabetes Study (CARDS): multicentre randomised placebo-controlled trial Lancet 2004 364 685 696 15325833 10.1016/S0140-6736(04)16895-5
Rubins HB Robins SJ Collins D Fye CL Anderson JW Elam MB Faas FH Linares E Schaefer EJ Schectman G Wilt TJ Wittes J Gemfibrozil for secondary prevention of coronary heart disease in men with low levels of high-density lipoprotein cholesterol. Veterans Affairs High-density lipoprotein cholesterol Intervention Trial study group N Engl J Med 1999 341 410 418 10438259 10.1056/NEJM199908053410604
Robins SJ Collins D Wittes JT Papademetriou V Deedwania PC Schaefer EJ McNamara JR Kashyap ML Hershman JM Wexler LF Rubins HB Relation of gemfibrozil treatment and lipid levels with major coronary events: VA-HIT: a randomized controlled trial JAMA 2001 285 1585 1591 11268266 10.1001/jama.285.12.1585
The BIP Study Group Secondary prevention by raising HDL cholesterol and reducing triglycerides in patients with coronary artery disease. The Bezafibrate Infarction Prevention (BIP) Study Circulation 2000 102 21 27 10880410
Rubins HB Robins SJ Collins D Nelson DB Elam MB Schaefer EJ Faas FH Anderson JW for the VA-HIT Study Group Diabetes, plasma insulin and cardiovascular disease. Subgroup analysis from the Department of Veterans Affairs high-density lipoprotein intervention trial (VA-HIT) Arch Intern Med 2002 162 2597 2604 12456232 10.1001/archinte.162.22.2597
Diabetes Atherosclerosis Intervention Study Investigators Effect of fenofibrate on progression of coronary artery disease in type 2 diabetes: the Diabetes Atherosclerosis Intervention Study, a randomised study Lancet 2001 357 905 910 11289345 10.1016/S0140-6736(00)04209-4
Schoonjans K Staels B Auwerx J Role of the peroxisome proliferator-activated receptor (PPAR) in mediating the effects of fibrates and fatty acids on gene expression J Lipid Res 1996 37 907 925 8725145
Staels B Dallongeville J Auwerx J Schoonjans K Leitersdorf E Fruchart JC Mechanisms of action of fibrates on lipid and lipoprotein metabolism Circulation 1998 98 2088 2093 9808609
Aguilar-Salinas CA Fanghanel-Salmon G Meza E Montes J Gulias-Herrero A Sanchez L Monterrubio-Flores EA Gonzalez-Valdez H Gomez Perez FJ Ciprofibrate versus gemfibrozil in the treatment of mixed hyperlipidemias: an open-label, multicenter study Metabolism 2001 50 729 733 11398153 10.1053/meta.2001.23308
Branchi A Rovellini A Torri A Sommariva D Accuracy of calculated serum low-density lipoprotein cholesterol for the assessment of coronary heart disease risk in NIDDM patients Diabetes Care 1998 21 1397 1402 9727883
Pyorala K Pedersen TR Kjekshus J Faergeman O Olsson AG Thorgeirsson G for the Scandinavian Simvastatin Survival Study (4S) Group Cholesterol lowering with simvastatin improves prognosis of diabetic patients with coronary heart disease: A subgroup analysis of the Scandinavian Simvastatin Survival Study (4S) Diabetes Care 1997 20 614 620 9096989
Post Coronary Artery Bypass Graft Trial Investigators The effect of aggressive lowering of low-density lipoprotein cholesterol levels and low-dose anticoagulation on obstructive changes in saphenous-vein coronary-artery bypass grafts N Engl J Med 1997 336 153 62 8992351 10.1056/NEJM199701163360301
Hoogwerf BJ Waness A Cressman M Canner J Campeau Domanski M Geller N Herd A Hickey A Hunninghake DB Knatterud GL White C Effects of aggressive cholesterol lowering and low-dose anticoagulation on clinical and angiographic outcomes in patients with diabetes. The Post Coronary Artery Bypass Graft Trial Diabetes 1999 48 1289 1294 10342818
GISSI Prevenzione Investigators (Gruppo Italiano per lo Studio della Sopravvivenza nell'Infarto Miocardico) Results of the low-dose (20 mg) pravastatin GISSI Prevenzione trial in 4271 patients with recent myocardial infarction: do stopped trials contribute to overall knowledge? Ital Heart J 2000 1 810 20 11302109
Athyros VG Mikhailidis DP Papageorgiou AA Mercouris BR Athyrou VV Symeonidis AN Basayannis EO Demitriadis DS Kontopoulos AG Attaining United Kingdom-European Atherosclerosis Society low-density lipoprotein cholesterol guideline target values in the Greek Atorvastatin and Coronary-Heart-Disease Evaluation (GREACE) Study Curr Med Res Opin 2002 18 499 502 12564661 10.1185/030079902125001317
Shepherd J Blauw GJ Murphy MB Bollen EL Buckley BM Cobbe SM Ford I Gaw A Hyland M Jukema JW Kamper AM Macfarlane PW Meinders AE Norrie J Packard CJ Perry IJ Stott DJ Sweeney BJ Twomey C Westendorp RG PROSPER study group Prospective Study of Pravastatin in the Elderly at Risk. Pravastatin in elderly individuals at risk of vascular disease (PROSPER): a randomised controlled trial Lancet 2002 360 1623 630 12457784 10.1016/S0140-6736(02)11600-X
Downs JR Clearfield M Weis S Whitney E Shapiro DR Beere PA Langendorfer A Stein EA Kruyer W Gotto AM Jr Primary prevention of acute coronary events with lovastatin in men and women with average cholesterol levels. Results of AFCAPS/TexCAPS JAMA 1998 279 1615 1622 9613910 10.1001/jama.279.20.1615
Meade T Zuhrie R Cook C Cooper J on behalf of MRC General Practice Research Framework Bezafibrate in men with lower extremity arterial disease: randomized controlled trial BMJ 2002 325 1139 12433762 10.1136/bmj.325.7373.1139
Elkeles RS Diamond JR Poulter C Dhanjil S Nicolaides A Mahmood S Richmond W Mather H Sharp P Feher MD Cardiovascular outcomes in type 2 diabetes: A double-blind placebo-controlled study of bezafibrate: The St. Mary's, Ealing, Northwick Park Diabetes Cardiovascular Disease Prevention (SENDCAP) Study Diabetes Care 1998 21 641 642 9571357
| 15571637 | PMC1129022 | CC BY | 2021-01-04 16:38:07 | no | Cardiovasc Diabetol. 2004 Dec 1; 3:9 | utf-8 | Cardiovasc Diabetol | 2,004 | 10.1186/1475-2840-3-9 | oa_comm |
==== Front
BMC GenomicsBMC Genomics1471-2164BioMed Central London 1471-2164-6-481580435710.1186/1471-2164-6-48Research ArticleIdentification and characterization of a novel mammalian Mg2+ transporter with channel-like properties Goytain Angela [email protected] Gary A [email protected] Department of Medicine University of British Columbia Vancouver, B.C. Canada2005 1 4 2005 6 48 48 23 11 2004 1 4 2005 Copyright © 2005 Goytain and Quamme; licensee BioMed Central Ltd.2005Goytain and Quamme; 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
Intracellular magnesium is abundant, highly regulated and plays an important role in biochemical functions. Despite the extensive evidence for unique mammalian Mg2+ transporters, few proteins have been biochemically identified to date that fulfill this role. We have shown that epithelial magnesium conservation is controlled, in part, by differential gene expression leading to regulation of Mg2+ transport. We used this knowledge to identify a novel gene that is regulated by magnesium.
Results
Oligonucleotide microarray analysis was used to identify a novel human gene that encodes a protein involved with Mg2+-evoked transport. We have designated this magnesium transporter (MagT1) protein. MagT1 is a novel protein with no amino acid sequence identity to other known transporters. The corresponding cDNA comprises an open reading frame of 1005 base pairs encoding a protein of 335 amino acids. It possesses five putative transmembrane (TM) regions with a cleavage site, a N-glycosylation site, and a number of phosphorylation sites. Based on Northern analysis of mouse tissues, a 2.4 kilobase transcript is present in many tissues. When expressed in Xenopus laevis oocytes, MagT1 mediates saturable Mg2+ uptake with a Michaelis constant of 0.23 mM. Transport of Mg2+ by MagT1 is rheogenic, voltage-dependent, does not display any time-dependent inactivation. Transport is very specific to Mg2+ as other divalent cations did not evoke currents. Large external concentrations of some cations inhibited Mg2+ transport (Ni2+, Zn2+, Mn2+) in MagT1-expressing oocytes. Ca2+and Fe2+ were without effect. Real-time reverse transcription polymerase chain reaction and Western blot analysis using a specific antibody demonstrated that MagT1 mRNA and protein is increased by about 2.1-fold and 32%, respectively, in kidney epithelial cells cultured in low magnesium media relative to normal media and in kidney cortex of mice maintained on low magnesium diets compared to those animals consuming normal diets. Accordingly, it is apparent that an increase in mRNA levels is translated into higher protein expression.
Conclusion
These studies suggest that MagT1 may provide a selective and regulated pathway for Mg2+ transport in epithelial cells.
==== Body
Background
Magnesium is the second most abundant cation within the cell and plays an important role in many intracellular biochemical functions [1]. Despite the abundance and importance of magnesium, little is known about how eukaryotic cells regulate their magnesium content.
Intracellular free Mg2+ concentration is in the order of 0.5 mM which is 1–2% of the total cellular magnesium [2]. Accordingly, intracellular Mg2+ is maintained below the concentration predicted from the transmembrane electrochemical potential. Intracellular Mg2+ concentration is finely regulated likely by precise controls of Mg2+ entry, Mg2+ efflux, and intracellular storage compartments [3]. The transporters comprising these pathways have only begun to be identified.
Few magnesium transporters have been identified at the molecular level. Schweyen and colleagues have demonstrated that the mitochondrial RNA splicing2 (Mrs2) gene encodes a protein that is present in yeast and mammalian inner mitochondrial membranes [4,5]. Mrs2 mediates high capacity Mg2+ influx in isolated yeast mitochondria driven by the inner membrane potential but also transports a range of divalent cations such as Ni2+, Co2+, and Cu2+ [6]. Overexpression of Mrs2 increases influx while deletion of the gene abolishes uptake suggesting that it is the major mitochondrial system. This data suggests that Mrs2 protein may mediate Mg2+ transport in mammalian mitochondria. Nadler et al first identified TRPM7, a widely expressed member of the transient receptor potential melastatin (TRPM) ion channel family, that produces a Mg2+ current in a wide variety of cells [7]. TRPM7 is regulated by intracellular Mg·ATP levels and is similarly permeable to both major divalent cations, Ca2+ and Mg2+, but also many of the trace elements, such as Zn2+, Mn2+, and Co2+ [8]. Using a positional cloning approach, Schlingmann et al [9] and Walder et al [10] found that hypomagnesemia with secondary hypocalcemia (HSH) was caused by mutations in TRPM6, a new member of the TRPM family. HSH is an inherited disease affecting both intestinal and renal Mg2+ absorption [3]. The functional characteristics of the TRPM6 transporter have not been fully elucidated [11,12]. Other magnesium transporters have been functionally described but they have not been characterized at the molecular level [13-18]. It is disparaging that, despite the significance of cellular Mg2+, only three specific magnesium transporters have been described in mammalian cells to date.
Mammalian magnesium homeostasis is a balance of epithelial intestinal magnesium absorption and renal magnesium excretion. The kidney plays a major role in control of vertebrate magnesium balance, in part, by active magnesium transport within the distal tubule of the nephron [2]. Using the Madin-Darby canine kidney (MDCK) cell line obtained from canine distal tubules and immortalized mouse distal convoluted tubule cells (MDCT), we have shown that Mg2+ entry is through specific and regulated magnesium pathways that are controlled by a variety of hormonal influences [19]. However these hormones do not provide selective control as they also affect calcium and in some cases sodium and potassium transport [19]. Selective and sensitive control of cellular Mg2+ transport is regulated by intrinsic mechanisms so that culture in media containing low magnesium results in upregulation of Mg2+ uptake in these cells. This adaptive increase in Mg2+ entry was shown to be dependent on de novo transcription since prior treatment of the epithelial cells with actinomycin D prevented the adaptation to low extracellular magnesium [20]. The data suggest that epithelial cells can somehow sense the environmental magnesium and through transcription- and translation-dependent processes alter Mg2+ transport and maintain magnesium balance. These conclusions using isolated epithelial cells are consonant with our views of magnesium conservation in the intact kidney [2].
In an attempt to identify genes underlying cellular changes resulting from adaptation to low extracellular magnesium, we used oligonucleotide microarray analysis to screen for magnesium-regulated transcripts in epithelial cells. This approach revealed one transcript whose relative level was dramatically altered by extracellular magnesium. Thus, this transcript potentially represented a species of mRNA whose synthesis was regulated by changes in cell magnesium. In this study, we describe the identification and characterization of this novel transcript referred to as MagT1. Our data indicate MagT1 may mediate Mg2+ transport in a wide variety of cells and may play a role in control of cellular magnesium homeostasis.
Results
Identification of MagT1
With the knowledge that differential gene expression is involved with selective control of epithelial cell magnesium conservation, our strategy was to use microarray analysis to identify candidates that were up-regulated with low magnesium. Using Affymetrix GeneChipR technology, we showed that 116 DNA fragments were significantly increased (p < 0.0002) from the 24,000 mouse ESTs represented on the chips. The RNA of one of these was significantly increased, greater than 2-fold, n = 3, determined by real-time RT-PCR. The full length human cDNA was identified from clone DKFZp564K142Q3 obtained from RZPD Resource Center, Berlin, in pAMP1 vector and bidirectionally sequenced at NAPS, University of British Columbia. Based on the cDNA sequence, electrophysiological properties and cation selectivity of the encoded protein, we designated it as MagT1 for Magnesium Transport protein, subtype 1. MgT was not used to avoid confusion with the bacterial MgtA/B and MgtE magnesium transporters [21,22].
Primary structure of MagT1
MagT1 cDNA comprises 2241-base pairs (bp) with an open reading frame of 1005 bp that predicts a protein of 335 amino acids with a relative molecular mass of 38,036 Da (Fig. 1). Hydropathy profile analysis suggested that MagT1 is an integral membrane protein containing five hydrophobic transmembrane-spanning (TM) α helical regions, the first of which is likely cleaved to form the final product with four TM domains (Fig. 1). MagT1 contains a N-linked glycosylation site at residue 215 located in the first extracellular loop. The N-terminal region of MagT1 contains four putative cAMP-dependent protein kinase phosphorylation sites at residues S73, S108, T153 and S162 and four possible protein kinase C phosphorylation sites at residues S38, T48, S103, T111. The short C-terminal cytoplasmic region does not possess any cAMP-dependent or protein kinase C phosphorylation sites. The presence of putative phosphorylation sites for protein kinase A and protein kinase C in the cytoplasmic domain suggests that transport might be regulated by phosphorylation.
Figure 1 Primary amino acid sequence of human hMagT1. Human MagT1 was aligned with human candidate tumor suppressor sequence, N33, and the human implantation associated protein, designated IAP. The six predicted transmembrane domains are overlined and numbered. The amino acid numbers corresponding to the MagT1 protein are shown on the left side.
MagT1 is a novel gene located at Xq13.1–13.2
The human origin, chromosomal location, and intron-exon organization of the MagT1 gene were deduced from the expressed sequence tag (EST) database and the human genome data. There may be an alternative splicing of MagT1 but only one transcript could be seen on the Northern blot (Fig. 2). Mouse mMagT1 gene is comprised of 10 exons spanning 41,680 bp located on the X chromosome (unplaced). The human hMagT1 gene is composed of 11 exons spanning 69,137 bp and is also on the X chromosome (Xq13.1–13.2).
Figure 2 Tissue distribution of mMagT1 mRNA. A, Northern blot analysis of mMagT1 mRNA in MDCT cells or mouse tissues. Tissues were harvested and poly(A)+ RNA prepared by standard techniques. Each lane was loaded with 8 µg of poly(A)+ RNA. The same blot was stripped and hybridized with 32P-labeled β-actin as a control for loading. B, real-time reverse transcription PCR analysis of mMagT1 RNA in tissues harvested from mice maintained on normal magnesium diet. mMagT1 and murine β-actin RNA was measured with Real-Time RT PCR (AB7000TM, Applied Biosystems) using SYBR GreenTM fluorescence. Standard curves for MagT1 and β-actin were generated by serial dilution of each plasmid DNA. The expression level of the mMagT1 transcript was normalized to that of the mouse β-actin transcript measured in the same 1.0 μg RNA sample. Results are normalized to the small intestine and expressed as fold-difference. Mean mRNA levels of kidney, colon, heart, brain, lung, and liver tissues were significantly greater, p>0.01, than small intestine ans spleen.
A BLAST search yielded a number of poorly characterized proteins with similar amino acid sequences to MagT1 (Fig. 1). Using the BESTFIT sequence alignment program, MagT1 shows 100% identity to a human unnamed protein (GenBank™ CAB66571.1, BAC11592.1), 88% to a mouse implantation associated protein (GenBank™ NP_080228.1, BAB28739.1, BAB31313.1, AAH03881.1), 87% to a rat implantation associated protein (GenBank™ IAG2_RAT, NP_446398.1, AAB63294.2), 66% (first 131 amino acids) to a human implantation associated protein (GenBank™ XP_497668) and to an unknown protein MGC:56218 from the zebra fish (AAH46002.1). MagT1 shares some similarity (65–67%) to the human (GenBank™ AAH10370.1, AAB18376.1, AAB18374.1, G02297, N33_HUMAN, NP_006756.1, AAB18375.1), mouse (GenBank™ BAC25795.1), and rat (GenBank™ XP_214356.1) putative prostate cancer tumor suppressor protein. There is also some similarity (23–54%) to a number of un-characterized proteins in Anopheles (GenBank™ EAA13927.1), Drosophila melanogaster (GenBank™ AAL68198.1, AAF52636.2, NP_609204.2), Ochlerotatus trisertiatus (GenBank™ AF275675.1), and Caenorhabditis elegans (GenBank™ NP_498691.1, AAA28222.1, S44911, Y013_CAEEL). None of these proteins, with similar amino acid sequences to MagT1, are sufficiently characterized to suggest a common functional purpose. MagT1 has a more distant relationship (P = 3 × 10-12) to the OST3 gene of Saccharomyces cervisiae that encodes a regulatory subunit of the endoplasmic reticulum oligosaccharyltransferase complex [23]. A gapped alignment of these sequences showed only 21% identical residues between the hMagT1 and OST3 sequences extending throughout most of both proteins.
Tissue distribution of MagT1 expression
Northern analysis of cultured mouse distal convoluted tubule cells and tissues harvested from mice revealed a single strong transcript of about 2.4 kb (Fig. 2). The kidney, colon, heart and liver possessed relatively high levels of MagT1 mRNA and smaller amounts were found in intestine, spleen, brain, and lung (Fig. 2). Accordingly, MagT1 mRNA appears to be widely expressed among tissues but the transcript is variably expressed among these tissues.
The MagT1 antibody recognized two protein bands, 35 and 38 kDa, in tissues expressing the MagT1 transcript (Fig. 3). Two bands were apparent in kidney and liver tissue whereas one was evident in heart, colon, and brain. The molecular size of MagT1 calculated from cDNA is 38 kDa. A significant difference in the calculated molecular size and that the smaller band found by immunoblot analysis raises the possibility that MagT1 may be cleaved to yield the 35 kDa carboxyl-terminal protein detected by MagT1 antibody. There was very little MagT1 protein in the small intestine (Fig. 3). Other than liver tissue, there appears to be a good correlation between the respective amounts of transcripts and the protein content. The discrepancy between the levels of MagT1 mRNA and protein expression in liver (abundant mRNA detected but little protein detected) suggests that a posttranslational mechanism may play a role in tissue-specific expression of the MagT1 protein. In summary, the 38 kDa MagT1 protein is expressed to a variable extent in all of the tissues sampled (Fig. 3) but the 35 kDa band appears to be present in only some of the tissues. Although this is a limited survey of tissues, the results suggest that MagT1 is expressed in many tissues with an apparent correlation of mRNA and protein but expression may be post-translationally modified in a tissue-specific fashion such as the liver.
The specificity of the affinity-purified polyclonal anti-MagT1 antibody was assessed by Western blots of the total protein extract from the MDCT cells probed with a preimmune serum. No protein of the predicted size (~35 kDa) was detected when the preimmune serum was used (Fig. 3). Taken together, the results indicate that the affinity-purified anti-MagT1 antibody specifically reacts with the endogenous MagT1 protein.
Human MagT1 elicits Mg2+-evoked currents in Xenopus oocytes
The functional properties of MagT1-evoked Mg2+ currents were characterized using two-microelectrode voltage clamp analysis in Xenopus oocytes injected with hMagT1 cRNA. The electrophysiological data gave evidence for a rheogenic process with inward currents in hMagT1 cRNA-injected oocytes whereas there were no appreciable currents in control H2O- or total poly(A)+RNA-injected cells from the same batch of oocytes (Fig. 4). Human MagT1-mediated Mg2+-evoked uptake was linear for at least 20 min and did not display any time-dependent decay during repetitive stimulation with voltage steps (data not shown). The reversal potential was significantly shifted to the right as would be expected of a magnesium transporter (Fig. 5). In consonant with the notion that MagT1 protein mediates the observed Mg2+ currents is the association of the magnitude of the Mg2+-evoked current with the quantity of MagT1 protein in oocytes injected with MagT1 cRNA (Fig. 6). In this study oocytes were selected according to the size of the Mg2+-evoked current and Western blotting performed on the same oocyte. Both 38 and 35 kDa molecular size bands were correlated with the measured currents. Steady-state Mg2+-evoked currents were saturable (Fig. 7). The Michaelis constant (Km) was 0.23 mM, n = 29, when measured at -125 mV holding potential (Fig. 7, insert). The Michaelis constant was independent of the Vm used to determine the saturation kinetics. The Michaelis constants (Km) were +25 mV, 0.22 mM; -50 mV, 0.19 mM; -75, 0.20 mM; -100 mV, 0.19 mM; -125 mV, 0.23; -150 mV, 0.23 mM (data not shown).
Figure 3 Tissue distribution of mMagT1 protein. A. Western blots of membrane proteins from tissue extracts. Extracts were prepared from tissues as described under “Experimental Procedures”. MagT1 bands were probed with anti-MagT1antibody. Molecular sizes are expressed in kDa of pre-stained standards are shown on the left of each of the representative blots. B, summary of 38 kDa MagT1 protein in 15 μg total protein from various mice tissues. Data were obtained from 3 different mice and are indicated as the mean ± SEM. C, specificity of anti-MagT1 antibody. The fractions isolated from normal and magnesium-depleted MDCT cells were blotted with MagT1 antibody and MagT1 antibody preadsorbed with excess antigen peptide. The signal was reduced to background levels when preadsorbed antibody was used indicating that the antibody was specific to MagT1.
The Mg2+-evoked currents were not altered with deletion of external sodium by substitution with choline (89 ± 9 %, n = 3, of control currents) or replacement of chloride with nitrate (100 ± 1 %, n = 3, of control) suggesting that transport does not depend on extracellular Na+ or Cl- (data not shown). Niflumic acid (0.5 mM), a Cl- channel antagonist, did not affect Mg2+ currents (data not shown). Next, we determined the effect of transmembrane H+ gradients on Mg2+-evoked currents in MagT1-injected oocytes (Fig. 8). Currents are maximal at physiological pH, 7.4, and diminished with acidic and alkaline pH values (Fig. 8). Moreover, amiloride (0.1 mM), a Na+/K+ exchange inhibitor, did not influence expressed Mg2+ currents in oocytes (data not shown). This data suggests that Mg2+-evoked currents are not coupled to H+ movement but are sensitive to external pH. On balance, these data indicate that Mg2+-evoked currents in MagT1-injected oocytes are not coupled to Na+, Cl-, or H+ but are influenced by external pH values.
Figure 4 Mg2+-evoked currents in Xenopus oocytes expressing hMagT1 RNA transcripts. Current was continuously monitored in a single oocyte expressing hMagT1 clamped at -100 mV and superfused for the period indicated, first with modified Barth’s solution containing 0 mM magnesium then with 2.0 mM magnesium and finally returning to magnesium-free solution.
Large concentrations (2 mM) of Ca2+, or its analogs, Sr2+ and Ba2+ or the other divalent cations tested, Fe2+, Cu2+, Co2+, Zn2+, Mn2+, and Ni2+, did not produce appreciable currents in the absence of Mg2+ in hMagt1-expressing oocytes (Fig. 9). In the experiments shown, the permeability ratios (Erev for tested cation relative to Erev for Mg2+) were corrected for changes in membrane resistance caused by the respective divalent cation using values from H2O-injected oocytes (Fig. 9).
Some divalent cations inhibited Mg2+-evoked currents at relatively large concentrations of the respective inhibitor, 0.2 mM, in the presence of 2.0 mM MgCl2 (Fig. 10). The cations Ni2+and Zn2+ markedly shifted the ΔErev to the left whereas Mn2+was less effective and Gd3+, Cd2+, Co2+ and Cu2+ were modest inhibitors (Fig. 10). The multivalent cation, Gd3+, is a nonselective channel blocker that inhibits most Ca2+-permeable channels and known TRP channels [24]. The presence of 0.2 mM (Fig. 10) or 10.0 mM Ca2+, 98 ± 8 % (data not shown), was without effect on the amplitude of Mg2+-evoked currents. Fe2+ had no influence on MagT1-mediated currents (Fig. 10). On balance, these data indicate that hMagT1 cRNA-induced transport in oocytes is highly selective for Mg2+. Other divalent cations may be blockers but our evidence is that they are at most very weak agonists.
We have shown that relatively high concentrations of 1,4-dihydropyridine analogues, organic blockers of L-type Ca2+ channels, inhibit Mg2+ entry into distal tubule epithelial cells [19,20]. In the present experiments, nifedipine (10 μM) did not inhibit Mg2+-evoked currents (0.61 ± 0.08 μA at -125 mV, n = 5) but its analogue nitrendipine (10 μM) was an effective inhibitor (0.15 ± 0.02 μA, n = 7) in MagT1 expressed oocytes (Fig. 11). Control Mg2+-induced currents were 0.59 ± 0.09 μA, n = 6, in this series of experiments (Fig. 11). These findings were similar to our experience with MDCK and MDCT epithelial cells [19,20]. Again in consonant with our previous studies, the channel agonist, BAY K8644 (10 μM) stimulated Mg2+-evoked currents in expressing oocytes (0.80 ± 0.18 μA, n = 5) supporting the above electrophysiological data that MagT1 is a channel-like protein (Fig. 11).
MagT1 expression is responsive to magnesium
The rationale for these studies is based on the observation that renal magnesium conservation is principally regulated by differential expression of genes encoding magnesium transport proteins. Accordingly, we determined the response of MagT1 to changes in magnesium at the messenger and protein levels. These studies were performed with distal tubule epithelial cells, MDCT, cultured in media containing normal (1.0 mM) or low (nominally magnesium-free) magnesium concentrations for 16 h and on kidney cortex tissue harvested from mice maintained on either normal or magnesium-restricted diets for 5 days. The mRNA and protein expression was relatively stronger in cells cultured in low magnesium media and in tissue of mice maintained on low magnesium diets (urine and plasma magnesium concentration, 1.1 ± 0.3 and 0.13 ± 0.01 mM, respectively) compared to normal cells and tissue of animals on normal diets (urine and plasma magnesium, 13.2 ± 1.2 and 0.75 ± 0.09 mM, respectively). MDCT and tissue mMagT1 mRNA, as measured by real-time RT-PCR was increased by 2.1-fold and 2.3-fold, respectively (Figure 12). In association with the increases in mRNA, MagT1 protein was increased by 31 ± 12% in the cultured epithelial cells and 33 ± 6 % in kidney cortex with low magnesium relative to the respective controls (Figure 13). Accordingly, it is apparent that an increase in mRNA levels is translated into higher protein expression and by inference leads to greater magnesium transport (the latter conclusion is based on the urinary magnesium excretion of animals maintained on low magnesium relative to normal diets).
Discussion
Despite the extensive evidence for unique mammalian Mg2+ transporters, few proteins have been biochemically identified to date that fulfill this role. Moreover, functional characterization has not been fully investigated for those that have been reported [11,12]. With the knowledge that the kidney, particularly the distal tubule, regulates magnesium conservation through transcriptional mechanisms, we used oligonucleotide microarray analysis to identify MagT1, a novel Mg2+ transporter [2,19]. The MagT1 transcript is a 2.4-kb mRNA that encodes a protein comprising a relatively long N-terminal segment, a putative region of four TM domains, and a short C-terminal sequence. The cytoplasmic segments possess a number of characteristic phosphorylation motifs. MagT1 shows no structural similarity to any known transporter. Functional expression of MagT1 in oocytes results in large Mg2+-evoked currents with little permeability to other divalent cations. However, some divalent cations, Ni2+, Zn2+, and Mn2+inhibit Mg2+-evoked currents at relatively large external concentrations. These cations are not found in the extracellular or intracellular fluid at the concentrations used here, 0.2 mM. The other major extracellular divalent cation, Ca2+, was neither transported nor were Mg2+-evoked currents inhibited by extracellular Ca2+. MagT1 is widely distributed among tissues particularly those of epithelial structure. Finally, MagT1 expression is regulated in these tissues by external magnesium as predicted by our starting premise. Accordingly, MagT1 fulfills the role of a dedicated mammalian magnesium transporter. The function of MagT1 in cellular Mg2+ balance remains to be determined.
The electrophysiological characteristics of MagT1 expressed in Xenopus oocytes are reminiscent of our observations of Mg2+ transport in intact renal epithelial cells measured by microfluorescence [19]. There is not a suitable isotope of Mg2+ for use in physiological experiments so that we have used fura-mag-2 fluorescence to investigate Mg2+ transport [25]. We have shown that Mg2+ uptake in a variety of epithelial cells is driven by the electrochemical gradient of Mg2+. Membrane hyperpolarization stimulates Mg2+ transport whereas depolarization abrogates uptake (19). There was no evidence in renal distal tubule cells for coupling of apical Mg2+ entry to other ions such as Na+, Cl-, or H+ [19]. Magnesium transport in immortalized mouse distal convoluted tubule (MDCT) cells is dependent on the transmembrane concentration gradient and uptake is saturable, as determined by fluorescence. The apparent affinity constant is in the order of 0.5 mM that is similar to that observed for MagT1 expressed in Xenopus oocytes (Fig. 5). This affinity is appropriate for a physiological role of the transporter in cellular Mg2+ conservation [19]. Mg2+-evoked currents in oocytes expressing MagT1 is highly specific for Mg2+, an observation that is again consonant with our views of Mg2+ transport in MDCT cells and in vivo kidney [19]. The microfluorescence experiments suggest that there may be some variability in cationic inhibition of Mg2+ uptake depending on the cell-type used so that other transporters may be present with differing selectivity that are tissue specific [26]. Relatively large concentrations of nitrendipine, a 1,4-dihydropyridine channel blocker, inhibited Mg2+-evoked currents in MagT1-expressing oocytes not unlike the inhibition of Mg2+ entry into distal epithelial cells [19]. Intriguingly, nifedipine did not influence Mg2+-induced currents in MagT1-expressing oocytes that is similar to our previous reports using MDCT cells [19]. Although both antagonists are dihydropyridines, they have differing efficacy based on their structural differences [27]. Again, reminiscent of our observations using MDCK and MDCT epithelial cells, the channel agonist, BAY 8644, increased Mg2+-evoked currents [19]. The 1,4-dihydropyridines analogues are not highly selective channel blockers/activators but these findings support the notion that Mg2+ entry into MagT1-expressed oocytes or distal epithelial cells is via channel-like proteins. Two other characteristics are noteworthy. First, Mg2+-evoked currents in MagT1-expressed oocytes are greater at physiological pH values relative to acidic pH. This is also true for Mg2+ uptake in distal tubule epithelial cells and magnesium conservation by the intact kidney in vivo [19]. Magnesium reabsorption is greater and urinary excretion is less in metabolic alkalosis than acidosis. Indeed, magnesium wasting may be sufficient in chronic metabolic acidosis to lead to hypomagnesemia [2]. Second, the presence of multiple putative protein kinase A and C phosphorylation sites in MagT1 may suggest phosphorylation-dependent regulation. We have shown that Mg2+ entry into epithelial cells is stimulated by peptide hormones, such as parathyroid hormone, glucagon and calcitonin, that act through protein kinases A and C [19]. Further studies are needed to elucidate the mechanisms underlying these phenomena. On balance, many of the functional characteristics of MagT1 expressed in oocytes are harmonious with our earlier physiological observations using kidney distal convoluted tubule cells.
MagT1 is a membrane protein that may comprise ER, early and late endosomes or apical and basolateral plasma membrane fractions. The role of each of these structures in cellular magnesium homeostasis is poorly understood. Using single cell spatial imaging, we have previously shown that intracellular ionized Mg2+ concentration is heterogenously distributed across the cell [28]. The ER or sarcoplasmic reticulum normally contains high concentrations of Mg2+, ranging from 0.4–2.0 mM, relative to the cytosolic concentration, 0.5 mM, and nucleus, 0.32 mM. It is clear that Mg2+ is transported into and out of a variety of intracellular compartments and there is likely dedicated magnesium transporters for each event. Further studies are required to establish the subcellular localization and intracellular trafficking of Mg2+ and the role of MagT1 protein.
Our evidence is that the expression of MagT1 mRNA and protein is responsive to cellular magnesium. The ability of epithelial cells to selectively respond to the availability of essential nutrients, such as Zn2+ and Fe2+, is not unique but the cellular mechanisms are unknown [29,30]. Presumably epithelial cells may sensitively sense intracellular nutrient concentration and through transcriptional and post-translational mechanisms adjust transport rates appropriately [19,29,30]. Our studies indicate that this response within the cell is the basis for sensitive and selective control of magnesium balance in the kidney [19].
Epithelial cells comprising the intestine and kidney are primarily involved with dietary magnesium absorption, urinary magnesium excretion, and total body magnesium homeostasis [2]. Accordingly, MagT1 may, in part, be responsible for intestinal and renal tubular Mg2+ conservation. In support of this is the observation that the MagT1 transcript is present in these tissues (Fig. 2). However, magnesium is necessary in all cells and the wide-spread distribution of the MagT1 transcript may suggest a housekeeping role for this transporter. It is also germane to note that MagT1 mRNA is regulated in all cells investigated. Further studies are needed to define the function of MagT1 in intestine and kidney and the role in overall cellular magnesium balance.
Conclusion
We have identified a novel magnesium transporter, probably a channel, that is regulated by extracellular magnesium. To our knowledge this is the first report of a highly selective Mg2+ transporter. Its role in cellular magnesium homeostasis and transepithelial magnesium absorption is unknown but our evidence from our differential gene expression studies indicate that it plays an important in cellular magnesium homeostasis.
Methods
Cell culture and oligonucleotide microarray analysis
Mouse distal convoluted tubule (MDCT) cells were isolated from kidneys and immortalized by Pizzonia et al (31). The MDCT cell line has been extensively used by us to study renal magnesium transport [21]. Cells were grown in Basal Dulbecco's minimal essential medium (DMEM)/Ham's F-12, 1:1, media (GIBCO) supplemented with 10% fetal calf serum (Flow Laboratories, McLean, VA), 1 mM glucose, 5 mM L-glutamine, 50 U/ml penicillin, and 50 μg/ml streptomycin in a humidified environment of 5% CO2- 95% air at 37°C. Where indicated, subconfluent MDCT cells were cultured in Mg2+-free media (Stem Cell Technologies Inc., Vancouver, BC) for 4 h. Other constituents of the Mg2+-free culture media were similar to the complete media.
Microarray analysis was performed according to the protocol recommended by Affymetrix . Poly(A)+ RNA was extracted with Poly(A)Pure (Ambion) from cells cultured in high and low magnesium media. Twenty Fg RNA was used for cDNA synthesis followed by in vitro transcription. The cRNA was biotin-labeled, fragmented, and the probes hybridized to Affymetrix MG U74 Bv2 and MG U74 Cv2 arrays (Affymetrix, Santa Clara) representing approximately 24,000 mouse transcripts. Detailed protocols for data analysis, documentation of sensitivity, reproducibility and other aspects of the quantitative microarray analysis are those given by Affymetrix. Gene categorization was based on the NetAffx Database.
Northern blot analysis
Cells were harvested by scraping and total RNA isolated using TRIzol (Life Technologies, Inc.). In some experiments poly(A)+ RNA was isolated using the Poly(A)Pure mRNA isolation system (Ambion) following the manufacturer's instructions. Samples of total RNA (20 μg) or poly(A)+ RNA (8 μg) were denatured in 2.2 M formaldehyde, 50% (v/v) formamide buffer and electrophoresed on 0.8% agarose 3 M formaldehyde, 0.02 M MOPS, 8 mM Na acetate, 1 mM EDTA, pH 7.0 gels. The size-fractionated RNA was transferred to GeneScreen nylon membranes (NEN) by downward alkali transfer and UV crosslinked (Stratagene Stratalinker 1800). Membranes were probed with 32P-labelled probes made from gene specific inserts represented in the microarray analytical results. The probe templates were prepared from PCR products representing inserts using specific primers on cDNA prepared from MDCT cells. The inserts were ligated into pGEM-t vector (Promega) following QiaexII gel (Qiagen) purification. Blots were prehybridized in 50% formamide, 5 X SSPE, 100 μg/ml denatured sonicated salmon sperm DNA, 5 X Denhardt's solution, 0.1% SDS for 1 h at 42EC in a rotating hybridization oven (Tyler HI-16000). Probe was heated to 95EC for 5 min, then added to the prehybridization solution. Membranes were hybridized for 16 h at 42EC then washed at high stringency sequentially: 2X [1X SSPE, 0.2% SDS, 28EC] 2X [1X SSPE, 0.4% SDS, 37EC] 1X [0.1X SSPE, 0.2% SDS, 55EC]. Membranes were exposed on Kodak X-AR-2 film. In most cases, after images were obtained, membranes were incubated at 95°C for 1 h in 0.1% SDS to remove the bound probe and hybridized with a 32P-labelled β-actin probe in order to normalize loading.
Quantitative analysis of MagT1 transcripts by real-time RT PCR
Total RNA of cells was extracted by TRIzol (Invitrogen). Genomic DNA contamination was removed by DNA-free™ kit (Ambion) prior to making first strand cDNA. Standard curves were constructed by serial dilution of a linear pGEM-T vector (Promega) containing the MagT gene. The primer set of mouse MagT1 was: forward, 5'-CCAAAGGGGCTGATACATA-3' and reverse, 5'-ATAGAAGAACGATGTGTG-3' and the human MagT1: forward, 5'-GCAAACTCCTGGCGATACTCC-3' and the human reverse 5'-ACTGGGCTTGACTGCTTCC-3'. PCR products were quantified continuously with AB7000™ (Applied Biosystems) using SYBR Green™ fluorescence according to the manufacturer's instructions. The relative amounts of MagT1 RNA were normalized to the respective human and mouse β-actin transcripts.
Genomic sequence analysis
The MagT1 cDNA sequence was determined by standard methods. Data base searching and alignments were performed using BLAST. The nonredundant and EST data bases were sourced. Protein homology searches were performed by comparing the amino acid query sequence against SWISSPROT data base. The full-length MagT1 cDNA sequence has been deposited in the GenBank™ data base (accession human DQ000004, mouse DQ000005).
Western blot analysis
A rabbit polyclonal antibody, anti-MagT1, was raised against the N-terminal domain of the final cleaved human MagT1 protein using a synthetic peptide, INFPAKGKPKRGDTYELQV (amino acid residues 140–158), coupled to keyhole limpet hemocyanin. Affinity-purified rabbit anti-human MagT1 antibody was diluted in TBS (Tris-buffered saline, 20 mM Tris, 200 mM NaCl, pH 7.6) containing 0.5 % BSA at a final concentration about 0.7 μl/ml. For subcellular fractionation, cells were suspended in lysis buffer (0.25 M sucrose, 10 mM triethanolamine-acetic acid pH 7.6, 1 mM EDTA) containing protease inhibitors (1 mM PMSF, 2 μg/ml leupeptin, 2 μg/ml aprotinin). Protein concentrations were determined using Bio-Rad protein assay reagent. SDS-PAGE was performed according to Laemmli. For immunobblotting, the proteins were electrophoretically transferred to polyvinylidene difluoride membranes (HybondR, Amersham Biosciences) by semidry electroblotting for 45 min. Western analysis was performed by incubating the blots with antiMagT1 antibody or anti-MagT1 antibody preabsorbed with 50 × antigen peptide (control for antibody specificity) overnight at 4EC followed by three washes with TBS/0.1% Tween-20, 10 min each. The blots were then incubated with 1/10,000 horseradish peroxidase-conjugated donkey anti-rabbit secondary (Sigma Aldrich) antibody for 1 h. After washing three times with TBS/Tween-20, 10 min each, the blots were visualized with ECL (Amersham Biosciences) according to the manufacturer's instructions.
Expression of MagT1 in Xenopus oocytes and current measurements
The cDNA comprising the open reading frame (ORF) of MagT1 was amplified from the pAMP1 vector using the cloning primers (sense: 5'-GATTGGTACCGTGAACATGGCCTC-3'; antisense: 5'-CTTGTCGACCCTCTTTAACTCATC-3') and was subcloned into the KpnI and ApaI restriction sites of the pEYFP-N1 expression vector. The constructs were linearized and then transcribed with SP6 polymerase in the presence of m7GpppG cap using the mMESSAGE MACHINE™ SP6 KIT (Ambion) transcription system. Oocytes were injected with MagT1 complementary RNA (cRNA) or for control observations, H2O or kidney total poly(A)+ RNA; no Mg2+-induced currents were detected in the latter.
Xenopus oocytes were prepared and injected with cRNA and electrophysiological recordings were preformed according to previously described techniques [32]. Briefly, defolliculated stage V-VI oocytes were typically injected with 25 ng cRNA in 50 nl H2O. Oocytes were incubated at 18°C for 3–6 days in multiwell tissue culture plates containing Barth's solution (88 mM NaCl, 1.0 mM KCl, 2.4 mM NaHCO3, 1.0 mM MgSO4, 1.0 mM CaCl2, 0.3 mM Ca(NO3)2, 10 mM Hepes-NaOH, pH 7.6, 2.5 mM Na-pyruvate, 0.1 % BSA, 10,000 U/l penicillin, 100 mg/l streptomycin). To record expressed membrane currents, the oocytes were placed in a recording chamber (0.3 ml) and perfused with modified Barth's (96 mM NaCl, 10 mM Hepes-NaOH) containing various concentrations of MgCl2, as indicated, in substitution for osmotically equivalent amounts of NaCl. All experiments were performed at room temperature (21°C).
Steady-state membrane currents were recorded with the two-microelectrode voltage-clamp technique using a Geneclamp 500 amplifier (Axon Instruments, Inc.). Electrophysiology consisted of a voltage clamp step profile consisting of a holding potential of -15 mV, followed by 8 episode series of +25 mV steps of 2 s duration, from -150 mV to +25 mV within an episode duration of 6.14 sec. Each episode recorded 1536 data points collected at 4 ms intervals. The data was filtered at the appropriate frequency before digitization. In order to assess the permeability of different divalent cations, we used the shift in the reversal potentials of the respective cation from the reversal potentials of Mg2+ currents, ΔErev, and calculated by the permeability ratio by:
Px/PMg = exp/(ΔErev X F/RT)
where R, T, and F have their standard meanings. Voltage clamp episodes in the presence of extracellular test cations were corrected against episodes in the absence of external test cations.
All experimental conditions were performed on oocytes harvested from a minimum of 3 different animals.
Authors' contributions
Authors contributed equally in all parts of this study. All authors read and approved the final manuscript.
Figure 5 Mg2+-evoked currents in Xenopus oocytes expressing hMagT1. Current-voltage relationships obtained from linear voltage steps from -150 mV to +25 mV in the presence of Mg2+-free solutions or those containing the indicated concentrations of MgCl2. Oocytes were clamped at a holding potential of -15 mV and stepped from -150 mV to +25 mV in 25 mV increments for 2 s at each of the concentrations indicated. Shown are average I-V curves obtained from control H2O-injected (n = 13) or MagT1-expressing (n =/>7) oocytes. Note, the positive shift in reversal potential, indicated by arrows, with increments in magnesium concentration. Values are mean ± SEM of observations measured at the end of each voltage sweep for the respective Mg2+ concentration.
Figure 6 Association of Mg2+ currents with the expression of 38 kDa MagT1 protein in Xenopus oocytes injected with MagT1 cRNA. Oocytes were selected from one frog according to the expressed Mg2+ currents as shown. Results illustrated is representative of four oocyte preparations from different animals. The relative amplitude of Mg2+ currents was associated with the amount of MagT1 protein determined by Western blot analysis.
Figure 7 Summary of concentration-dependent Mg2+-evoked currents in MagT1-expressing oocytes using a holding potential of -125 mV. Mean ± SEM values are those given in Fig. 1A. Inset illustrates an Eadie-Hofstee plot of concentration-dependent Mg2+-evoked currents demonstrating a Michaelis constant of 0.23 mM.
Figure 8 Characterization of Mg2+-evoked currents in Xenopus oocytes expressing hMagT1. A, effect of pH on Mg2+-evoked currents. Currents were measured in standard solutions containing 2.0 mM MgCl2 at the pH values indicated. B, summary of mean currents with external pH at a holding potential of -125 mV. Mg2+ did not evoke currents in H2O-injected oocytes at any of the pH values tested.
Figure 9 Substrate specificity of MagT1 following application of test cations, 2.0 mM, in the absence of external Mg2+. For clarity, only Mg2+,Cu2+, Mn2+, and Sr2+ are represented in panel A. Oocytes were clamped at a holding potential of -15 mV and stepped from -150 mV to +25 mV in 25 mV increments for 2 s for each of the cations. Values are mean ± SEM of currents measured at the end of each voltage sweep for the respective divalent cation. B, summary of permeabilities of the tested divalent cations. Figure illustrates average permeability ratios (Erev for tested cation relative to Erev for Mg2+) given in Fig. 9A.
Figure 10 Inhibition of MagT1-mediated currents. A,inhibition of Mg2+-evoked currents with 0.2 mM test cation in the presence of external 2.0 mM Mg2+. For clarity, only Cu2+, Mn3+, and Zn2+ relative to Mg2+ are represented. Values are mean ± SEM of currents measured at the end of each voltage sweep for the respective cation. B, summary of inhibition by multivalent cations of Mg2+ currents based on the change in Erev represented in Fig. 10A. The inhibitor was added with MgCl2 and voltage-clamp was performed about 5 min later.
Figure 11 Effect of voltage-dependent channel antagonists on MagT1-mediated currents. A, the antagonists nifedipine (10 µM) and nitrendipine (10 µM), or the agonist, Bay K8644 (10 µM), were added prior to determining Mg2+-evoked currents. B, summary of mean currents (I µA) with the respective inhibitors at a holding potential (Vm) of -125 mV (n=7). The analogues were added 5 min prior to voltage-clamping.
Figure 12 MagT1 mRNA expression is responsive to magnesium. Where indicated MDCT cells were cultured in normal (1.0 mM) or low (<0.01 mM) magnesium media for 16 h. Kidney cortical tissue was harvested from mice on normal (0.05% by weight) or low magnesium (<0.01%) diets for 5 days. MagT1 and murine β-actin RNA was measured with Real-Time RT PCR (AB7000TM, Applied Biosystems) using SYBR GreenTM fluorescence. Data is from 10-12 PCRs performed on five separate cultures or animals in each group maintained on low and normal magnesium.
Figure 13 MagT1 protein expression is responsive to magnesium. Western blots of membrane proteins from cells and tissues as described under “Experimental Procedures”. MagT1 bands were probed with anti-MagT1antibody. Data are from four Western blots performed on five separate cultures or animals in each group maintained on low and normal magnesium.
Acknowledgements
This work was supported by a research grants from the Canadian Institutes of Health Research, MOP-53288, and the Kidney Foundation of Canada. We acknowledge Genomic Sciences Center, Riken Yokohama Institute, Japan for EST clones A530029P05, A330056M18, and A530032I23 and RZPD Deutsches Ressourcenzentrum für Genomforschung GmbH, Berlin, Germany for clone DKFZp564K142Q3.
==== Refs
Flatman PW Magnesium transport across cell membranes J Membr Biol 1984 80 1 14 6384523
Quamme GA Renal magnesium handling: New insights in understanding old problems Kidney Int 1997 52 1180 1195 9350641
Cole DEC Quamme GA Inherited disorders of renal magnesium handling J Am Soc Nephrol 2000 11 1937 1947 11004227
Bui DM Gregan J Jarosch E Ragnini A Schweyen RJ The bacterial magnesium transporter CorA can functionally substitute for its putative homologue Mrs2p in the yeast inner mitochondrial membrane J Biol Chem 1999 274 20438 20443 10400670 10.1074/jbc.274.29.20438
Zsurka G Gregan J Schweyen RJ The human mitochondrial Mrs2 protein functionally substitutes for its yeast homologue, a candidate magnesium transporter Genomics 2001 72 158 168 11401429 10.1006/geno.2000.6407
Kolisek M Zsurka G Samaj J Weghuber J Schweyen RJ Schweigel M Mrs2p is an essential component of the major electrophoretic Mg2+ influx system in mitochondria EMBO J 2003 22 1235 1244 12628916 10.1093/emboj/cdg122
Nadler MJS Hermosura MC Inabe K Perraud A-L Zhu Q Stokes AJ Kurosaki T Kinet J-P Penner R Scharenberg AM Fleig A Hypomagnesemia with secondary hypocalcemia is caused by mutations in TRPM6, a new member of the TRPM gene family Nature 2001 411 590 595 11385574 10.1038/35079092
Monteilh-Zoller MK Hermosura MC Nadler MJS Scharenberg AM Penner R Fleig A TRPM7 provides an ion channel mechanism for cellular entry of trace metal ions J Gen Physiol 2003 121 49 60 12508053 10.1085/jgp.20028740
Schlingmann KP Weber S Peters M Nejsums LN Vitzthum H Klingel K Kratz M Haddad E Ristoff E Dinour D Syrrou M Nielsen S Sassen M Waldegger S Seyberth HW Konrad M Hypomagnesemia with secondary hypocalcemia is caused by mutations in TRPM6, a new member of the TRPM gene family Nat Genet 2002 31 166 171 12032568 10.1038/ng889
Walder YW Landau D Meyer P Shalev H Tsolia M Borochowitz Z Boettger MB Beck GE Englehardt RK Carmi R Sheffield VC Mutation of TRPM6 causes familial hypomagnesemia with secondary hypocalcemia Nat Genet 2002 31 171 174 12032570 10.1038/ng901
Voets T Nilius B Hoefs S van der Kemp AW Droogmans G Bindels RJ Hoenderop JG TRPM6 forms the Mg2+ influx channel involved in intestinal and renal Mg2+ absorption J Biol Chem 2004 279 19 25 14576148 10.1074/jbc.M311201200
Chubanov V Waldegger S Mederos y Schnitzler M Vitzthum H Sassen MC Seyberth HW Konrad M Gudermann T Disruption of TRPM6/TRPM7 complex formation by a mutation in the TRPM6 gene causes hypomagnesemia with secondary hypocalcemia Proc Nat Acad Sci U S A 2004 101 2894 2899 10.1073/pnas.0305252101
Cefaratti C Romani A Scarpa A Differential localization and operation of distinct Mg(2+) transporters in apical and basolateral sides of rat liver plasma membrane J Biol Chem 2000 275 3772 3780 10660526 10.1074/jbc.275.6.3772
Günther T Mechanisms and regulation of Mg2+ efflux and Mg2+ influx Miner Electrolyte Metab 1993 19 259 265 8264512
Schweigel M Vormann J Martens H Mechanisms of Mg2+ transport in cultured ruminal epithelial cells Am J Physiol 2000 278 G400 G408
Rasgado-Flores H Gonzales-Serratos H Plasmalemmal transport of magnesium in excitable cells Front Biosci 2000 5 D866 D879 10966876
Touyz RM Mercure C Reudelhuber TL Angiotensin II type I receptor modulates intracellular free Mg2+ in renally derived cells via Na+-dependent Ca2+-independent mechanisms J Biol Chem 2001 276 13657 13663 11278387
Tashiro M Konishi M Iwamoto T Shigekawa M Kurihara S Transport of magnesium by two isoforms of the Na+-Ca2+ exchanger expressed in CCL39 fibroblasts Pflügers Archiv Eur J Physiol 2000 440 819 827 10.1007/s004240000384
Dai L-j Ritchie G Kerstan D Kang HS Cole DEC Quamme GA Magnesium transport in the renal distal convoluted tubule Physiol Rev 2001 81 51 84 11152754
Dai L-j Quamme GA Intracellular Mg2+ and magnesium depletion in isolated renal thick ascending limb cells J Clin Invest 1991 88 1255 1264 1655827
Smith RL Thompson LJ Maguire ME Cloning and characterization of MgtE, a putative new class of Mg2+ transporter from Bacillus firmus OF4 J Bacteriol 1995 177 1233 1238 7868596
Moncrief MB Maguire ME Magnesium transport in prokaryotes J Biol Inorg Chem 1999 4 523 527 10550680 10.1007/s007750050374
Knauer R Lehle L The oligosaccharyltransferase complex from Saccharomyces cerevisiae. Isolation of the OST6 gene, its synthetic interaction with OST3, and analysis of the native complex J Biol Chem 1999 274 17249 17256 10358084 10.1074/jbc.274.24.17249
Lee N Chen J Sun L Wu S Gray KR Rich A Huang M Lin J-H Feder JN Janovitz EB Levesque PC Blanar MA Expression and characterization of human transient receptor potential melastatin 3 (hTRPM3) J Biol Chem 2003 278 20890 20897 12672827 10.1074/jbc.M211232200
Quamme GA Dai L-j Presence of a novel influx pathway for Mg2+ in MDCK cells Am J Physiol 1990 258 C521 C525 2399971
Quamme GA Rabkin SW Cytosolic free magnesium in cardiac myocytes: Identification of a Mg2+ influx pathway Biochim Biophys Res Comm 1990 167 1406 1412 10.1016/0006-291X(90)90679-H
Hockerman GH Peterson BZ Johnson BD Catterall WA Molecular determinants of drug binding and action on L-type calcium channels Ann Rev Pharmacol Toxicol 1997 37 361 396 9131258 10.1146/annurev.pharmtox.37.1.361
Quamme GA Dai L-j Rabkin SW Dynamics of intracellular free Mg2+ changes in vascular smooth muscle cells Am J Physiol 1993 265 H281 H288 8342644
Eide DJ The SLC39 family of metal ion transporters Pflügers Archiv Eur J Physiol 2004 447 796 800 10.1007/s00424-003-1074-3
Roy CN Andrews NC Recent advances in disorders of iron metabolism: mutations, mechanisms and modifiers Hum Mol Gen 2001 10 2181 2186 11673399 10.1093/hmg/10.20.2181
Pizzonia JH Gesek FA Kennedy SM Coutermarsh BA Bacskai BJ Friedman PA Immunomagnetic separation, primary culture and characterization of cortical thick ascending limb plus distal convoluted tubule cells from mouse kidney In Vitro Cell Dev Biol 1991 27A 409 416 1649164
Quamme GA Chlorpromazine activates a chloride current in Xenopus oocytes Biochem Biophys Acta 1997 1324 18 26 9059494
| 15804357 | PMC1129089 | CC BY | 2021-01-04 16:39:53 | no | BMC Genomics. 2005 Apr 1; 6:48 | utf-8 | BMC Genomics | 2,005 | 10.1186/1471-2164-6-48 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1589883210.1371/journal.pbio.0030181Research ArticleBioinformatics/Computational BiologyEvolutionGenetics/Genomics/Gene TherapyImmunologyMolecular Biology/Structural BiologyNoneRAG1 Core and V(D)J Recombination Signal Sequences Were Derived from Transib Transposons V(D)J Recombination and Transib TransposonsKapitonov Vladimir V [email protected]
1
Jurka Jerzy [email protected]
1
1Genetic Information Research Institute, Mountain ViewCaliforniaUnited States of AmericaNemazee David Academic EditorScripps Research InstituteUnited States of America6 2005 24 5 2005 24 5 2005 3 6 e18123 9 2004 17 3 2005 Copyright: © 2005 Kapitonov and Jurka.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.
Uncovering the Ancient Source of Immune System Variety
The V(D)J recombination reaction in jawed vertebrates is catalyzed by the RAG1 and RAG2 proteins, which are believed to have emerged approximately 500 million years ago from transposon-encoded proteins. Yet no transposase sequence similar to RAG1 or RAG2 has been found. Here we show that the approximately 600-amino acid “core” region of RAG1 required for its catalytic activity is significantly similar to the transposase encoded by DNA transposons that belong to the Transib superfamily. This superfamily was discovered recently based on computational analysis of the fruit fly and African malaria mosquito genomes. Transib transposons also are present in the genomes of sea urchin, yellow fever mosquito, silkworm, dog hookworm, hydra, and soybean rust. We demonstrate that recombination signal sequences (RSSs) were derived from terminal inverted repeats of an ancient Transib transposon. Furthermore, the critical DDE catalytic triad of RAG1 is shared with the Transib transposase as part of conserved motifs. We also studied several divergent proteins encoded by the sea urchin and lancelet genomes that are 25%−30% identical to the RAG1 N-terminal domain and the RAG1 core. Our results provide the first direct evidence linking RAG1 and RSSs to a specific superfamily of DNA transposons and indicate that the V(D)J machinery evolved from transposons. We propose that only the RAG1 core was derived from the Transib transposase, whereas the N-terminal domain was assembled from separate proteins of unknown function that may still be active in sea urchin, lancelet, hydra, and starlet sea anemone. We also suggest that the RAG2 protein was not encoded by ancient Transib transposons but emerged in jawed vertebrates as a counterpart of RAG1 necessary for the V(D)J recombination reaction.
RAG1 and RAG2 activity is central to adaptive immunity, but the precise evolutionary origins of these proteins were unknown; the authors argue that RAG1 evolved from the Transib family of transposons.
==== Body
Introduction
The immune system of jawed vertebrates detects and destroys foreign invaders, including bacteria and viruses, by a specific response to an unlimited number of antigens expressed by them. The antigens can be identified after they are specifically bound by surface receptors of vertebrate B and T immune cells (BCRs and TCRs, respectively). Because the vast repertoire of BCRs and TCRs cannot be encoded genetically, ancestors of jawed vertebrates adopted an elegant combinatorial solution [1]. The variable portions of the BCR and TCR genes are composed of separate V (variable), D (diversity), and J (joining) segments, which are represented by fewer than a few hundred copies each. In a B and T cell site-specific recombination reaction, commonly known as V(D)J recombination, one V, one D, and one J segment are joined together into a single exon encoding the variable antigen-binding region of the receptor. In addition to this combinatorial diversity, further diversity is generated by small insertions and deletions at junctions between the joined segments. In V(D)J recombination, DNA cleavage is catalyzed by two proteins encoded by the recombination-activating genes, approximately 1040-amino acid (aa) RAG1 and approximately 530-aa RAG2 [2,3]. The site specificity of the recombination is defined by the binding of RAG1/2 to RSSs flanking the V, D, and J segments [4]. All RSSs can be divided into two groups, referred to as RSS12 and RSS23, and consist of conserved heptamer and nonamer sequences separated by a variable spacer either 12 ± 1 (RSS12) or 23 ± 1 (RSS23) bp long [4–7].
During V(D)J recombination, RAG1/2 complex binds one RSS12 and one RSS23, bringing them into juxtaposition, and cuts the chromosome between the RSS heptamers and the corresponding V and D, D and J, or V and J coding segments [3,8]. A rule requiring that efficient V(D)J recombination occur between RSS12 and RSS23 is known as the “12/23” rule [1]. Even prior to the discovery of RAG1 and RAG2, it had been suggested that the first two RSSs were originally terminal inverted repeats (TIRs) of an ancient transposon whose accidental insertion into a gene ancestral to BCR and TCR, followed by gene duplications, triggered the emergence of the V(D)J machinery [4]. Later, this model was expanded by the suggestion that both RAG1 and RAG2 might have evolved from a transposase (TPase) that catalyzed transpositions of ancient transposons flanked by TIRs that were precursors of RSSs [9]. This model has received additional support through observations of similar biochemical reactions in transposition and V(D)J recombination [10,11]. Finally, it was demonstrated that RAG1/2 catalyzed transpositions of a DNA segment flanked by RSS12 and RSS23 in vitro [12,13] and in vivo in yeast [14]. In vertebrates, in vivo RAG-mediated transpositions are strongly suppressed, probably to minimize potential harm to genome function. So far, only one putative instance of such a transposition has been reported [15]. However, given the lack of significant structural similarities between RAGs and known TPases, the “RAG transposon” model [9,12,13,16] remained unproven. Here we demonstrate that the RAG1 core and RSSs were derived from a TPase and TIRs encoded by ancient DNA transposons from the Transib superfamily [17].
The Transib superfamily is one of ten superfamilies of DNA transposons detected so far in eukaryotes [17]. Like other DNA transposons, Transib transposons exist as autonomous and nonautonomous elements. The autonomous Transib transposons are 3–4 kb long and code for an approximately 700-aa TPase that is not similar to TPases from any other transposon superfamilies. Computational analysis of Transib elements, including their numerous insertions into copies of other transposons, demonstrated that Transib transposons are flanked by 5-bp target site duplications (TSDs), which also distinguishes this superfamily from all the others [17]. Transib transpositions are expected to be catalyzed by the binding of the TPase to TIRs of autonomous and nonautonomous transposons [17]. As discussed in this paper, in addition to the fruit fly (Drosophila melanogaster) and African malaria mosquito (Anopheles gambiae) genomes, in which Transib transposons were originally discovered, these genes are also present in diverse animals (Table S1), including other species of fruit fly (e.g., Drosophila pseudoobscura, Drosophila willistoni), yellow fever mosquito (Anopheles aegypti), silkworm (Bombyx mori), red flour beetle (Tribolium castaneum), dog hookworm (Ancylostoma caninum), freshwater flatworm (Schmidtea mediterranea), hydra (Hydra magnipapillata), sea urchin (Strongylocentrotus purpuratus), and soybean rust (Phakopsora pachyrhizi). Genomes of plants and vertebrates seem to be free of any recognizable Transib transposons (Figure 1).
Figure 1 Schematic Presentation of Transib transposons, RAG1, RAG2, and RAG1-Like Proteins in Eukaryotes
The basic timescale of the evolutionary tree is based on published literature [49–51]. Red circles mark species in which Transib TPases were found. Gray squares indicate RAG2; orange and blue ellipses show the RAG1 core and RAG1 N-terminal domain, respectively. Overall taxonomy, including common and Latin names, is reported on the right side of the figure. A question mark at the lamprey lineage indicates insufficient sequence data. A lack of any labels means that the Transib TPase and RAG1/2 are not present in the sequenced portions of the corresponding genomes. Among branches lacking Transib TPases, only lamprey and crocodile genomes are not extensively sequenced to date. In sea anemone, the RAG1 core–like protein is capped by the ring finger motif, which also forms the C-terminus in the RAG1 N-terminal domain. In fungi, the Transib TPase was detected in soybean rust only.
Results
Detection of Similarity between Transib TPases and RAG1
Using protein sequences of seven known Transib TPases (Transib1 through Transib4 and Transib1_AG through Transib3_AG from D. melanogaster and A. gambiae, respectively) [17] as queries in a standard BLASTP search against all GenBank proteins, we found that the approximately 60-aa C-terminal portion of the Transib2_AG TPase was 35%−38% identical to the C-terminal portion of the RAG1 core (Figure S1). However, this similarity was only marginally significant (E = 0.07 where the E-value is an expected number of sequences matching by chance; Table 1). In another search against GenBank, using PSI-BLAST [18] (see Materials and Methods) with the Transib2_AG TPase as a query, we found that two unclassified proteins (GenBank gi 30923617 and 30923765; annotated as hypothetical proteins) and RAG1s constituted the only group of any GenBank proteins similar to the Transib2_AG TPase (Table 1). The statistical significance of similarity between the TPase and RAG1s was measured by Ei = 0.025, where Ei is the E-value threshold for the first inclusion of RAG1 sequences into the PSI-BLAST iterations [18] (Materials and Methods). The observed improvement in significance of the Transib/RAG1 similarity (from E = 0.07 in BLASTP to Ei = 0.025 in PSI-BLAST; Table 1) was due to the fact that both 151-aa and 123-aa hypothetical GenBank proteins were apparent remnants of Transib TPases (approximately 40% identity to the Transib2_AG TPase, E < 10−10 in BLASTP). The RAG1 proteins appeared to be more similar to the position-specific scoring matrix (PSSM) created by PSI-BLAST based on multiple alignment of the Transib2_AG TPase and two Transib TPase-like proteins, than to the solo Transib2_AG TPase in the BLASTP search.
Table 1 Significance of Similarities between the Transib TPases and RAG1 Core
The first column lists all 18 Transib TPases used as queries in our analysis, and the shaded areas indicate those added to the original set of all GenBank proteins in subsequent PSI-BLAST searches. The original GenBank set included two incomplete Transib TPase-like proteins. Column 2 lists E-values of best matches between RAG1s and Transib TPases detected in BLASTP searches against the original GenBank set. Column 3 reports Ei-values of best matches between RAG1s and a PSSM derived from the chosen query sequence and the two GenBank TPase-like proteins in PSI-BLAST searches against the original set of all GenBank proteins (see Materials and Methods). Columns 4–6 report the Ei-values for best matches between RAG1s and a Transib-derived PSSM after adding 7, 13, and 18 Transib TPases to the GenBank set, respectively. The numbers of the PSI-BLAST iterations after which the entire RAG1 core significantly aligned with the TPases are indicated in parentheses. Ei-values greater than 1 are indicated by dashes. Each empty cell indicates that the corresponding TPase query was not used at the particular stage of PSI-BLAST analysis.
Given the latter observation, we decided to improve the quality of the PSSM constructed by PSI-BLAST for different Transib TPase sequences. To achieve that, we combined protein sequences of the seven known Transib TPases with the set of all GenBank proteins. As a result, Ei-values for matches of RAG1s to a new PSSM based on alignment of nine Transib TPases (the two GenBank TPase-like proteins plus seven added TPases) noticeably dropped in comparison with the Ei-values obtained for the PSSM constructed in the previous step based on alignment of the three TPases (Table 1).
To support the observation that Ei-values of matches between RAG1s and the Transib TPase PSSM decrease as the number of TPase sequences used for construction of the PSSM increases, we identified six new Transib TPases (Transib5, Transib3_DP, Transib4_DP, Transib1_AA, Transib2_AA, Transib3_AA; Figure S2). During the next step of the PSI-BLAST analysis, the original GenBank set was combined with 13 Transib TPases. Again, Ei-values of matches between RAG1s and the new PSSM derived from multiple alignment of 15 Transib TPases (the two GenBank proteins plus all our TPases) were much smaller (approximately 10−6–10−3; Table 1) than those obtained based on the PSSM constructed from the nine TPases at the preceding step (approximately 10−3–10−2). In the final step, we identified one more set of five new Transib TPases (Transib1_DP, Transib2_DP, Transib4_AA, Transib5_AA, and Transib1_SP). When all 18 TPases were combined with the original GenBank set, the Ei values of matches between RAG1s and the Transib PSSM dropped significantly further (10−9–10−4; Table 1). During the final revision of this manuscript, we identified an intermediate RAG1-like sequence in Hydra magnipapillata, called RAG1L_HM, which is significantly similar to both RAG1 and Transib TPase, as shown later. This direct result represents an independent validation of our analysis.
The PSI-BLAST PSSM of Transib TPases approximates conservation/variability of the Transib TPase consensus sequence. The more diverse the TPases used in determining the PSSM, the more accurate is the approximation; some of the insect Transib TPases are less than 30% identical to each other, as shown in Figure 2. The RAG1 Ei values decreased as the number of Transib TPases used for the PSSM construction increased due to the fact that RAG1 evolved from a Transib TPase. In all cases, the E values obtained after several rounds of iterations were less than 10−20 at the point of convergence. Nearly the entire sequences of several Transib TPases, excluding their 100–140-aa N-terminal domains, converged with an approximately 600-aa portion of RAG1 defined by positions approximately 360–1010 (Figure S3). This portion of RAG1 corresponds to the “RAG1 core,” hereafter numbered relative to human RAG1 (residues 387–1011), which along with RAG2 is known to be sufficient to perform V(D)J cleavage even after deletions of the 383-aa N-terminal and 32-aa C-terminal portions of RAG1 [19,20].
Figure 2 Diversity of the Transib TPases and RAG1 Core–Like Proteins in Animals
The phylogenetic tree was obtained by using the neighbor-joining algorithm implemented in MEGA [44]. Evolutionary distance for each pair of protein sequences was measured as the proportion of aa sites at which the two sequences were different. Its scale is shown by the horizontal bar. Bootstrap values higher than 60% are reported at the corresponding nodes. Species abbreviations are as follows: AA, yellow fever mosquito; AG, African malaria mosquito; BF, lancelet; CL, bull shark; DP, D. pseudoobscura fruit fly; FR, fugu fish; HM, hydra; HS, human; NV, starlet sea anemone; SP, sea urchin; XL, frog. (Transib1 through Transib5 are from D. melanogaster fruit fly).
During studies reported here, we identified 11 additional new families of Transib transposons and TPases (see Figure S2) that are well preserved in the genomes of fruit flies (Transib5 in D. melanogaster; and Transib1_DP, Transib2_DP, Transib3_DP, and Transib4_DP in D. pseudoobscura), mosquitoes (Transib1_AA, Transib2_AA, Transib3_AA, Transib4_AA, and Transib5_AA from A. aegypti) and sea urchin (Transib1_SP). Transib1_SP is the first Transib transposon identified outside of insect genomes. A well-preserved 4132-bp Transib1_SP element (contig 7839, positions 376–4506) is flanked by a 5-bp
CGGCG TSD, and it encodes a 676-aa TPase (two exons) that is most similar to the Transib2 TPase (34% identity). Based on the currently available sequence data, we also reconstructed portions of TPases that were missed in previous studies [17] (Materials and Methods; see Figure S2). Using the Transib1_SP TPase as a query in TBLASTN searches against all GenBank sections (NR, HTGs, WGS, dbGSS, dbEST, dbSTS, and Trace Archives) we also found diverse Transib TPases in silkworm, red flour beetle, dog hookworm, freshwater flatworm, soybean rust, and hydra (Table S1). At the same time, recently sequenced genomes of honeybee, roundworms, fish, frog, mammals, sea squirts, plants, and fungi (except soybean rust) do not contain any detectable Transib transposons (see Figure 1). The observed patchy distribution could be caused by horizontal transfers and extinctions of Transib transposons in eukaryotic species.
Common Structural Hallmarks of the Transib TPase and RAG1 Core
All three core residues from the catalytic DDE triad in the RAG1 proteins (residues 603, 711, and 965) that are necessary for V(D)J recombination [21,22] are conserved in the Transib TPases (Figures 3 and Figure S3). This includes the distances between the second D and E residues, which are much longer in Transib transposons (206–214 aa) and RAG1 (253 aa) than in DDE TPases from other studied superfamilies (e.g., approximately 35-aa in Mariner/Tc1 [23], 2-aa in P [23], approximately 35-aa in Harbinger [24], with hAT as an exception (325-aa, [25]). Moreover, each catalytic residue is a part of a motif that is conserved in the Transib TPases and RAG1 (motifs 4, 6, and 10 in Figures 3 and Figure S3). The RAG1 core is composed of the N-terminal region and the central and C-terminal domains ([26,27]. The N-terminal region includes the RSS nonamer-binding regions (residues 387–480), referred to as NBR [28,29]. The two terminal motifs of RAG1 NBR are conserved in the Transib TPases (Figure S3), which indicates that they may be important for their binding to the Transib TIRs during transposition (the RSS-like structure of TIRs is described below; Figure 4). The central domain of the RAG1 core (residues 531–763) includes two aspartic acid residues from the DDE triad and is also thought to be involved in binding to the RSS heptamer and RAG2 [30,31].
Figure 3 Multiple Alignment of Ten Conserved Motifs in the RAG1 Core Proteins and Transib TPases
The motifs are underlined and numbered from 1 to 10. Starting positions of the motifs immediately follow the corresponding protein names. Distances between the motifs are indicated in numbers of aa residues. Black circles denote conserved residues that form the RAG1/Transib catalytic DDE triad. The RAG1 proteins are as follows: RAG1_XL (GenBank GI no. 2501723, Xenopus laevis, frog), RAG1_HS (4557841, Homo sapiens, human), RAG1_GG (131826, Gallus gallus, chicken), RAG1_CL (1470117, Carcharhinus leucas, bull shark), RAG1_FR (4426834, Fugu rubripes, fugu fish). Coloring scheme [43] reflects physiochemical properties of amino acids: black shading marks hydrophobic residues, blue indicates charged (white font), positively charged (red font), and negatively charged (green font); red indicates proline (blue font) and glycine (green font); gray indicates aliphatic (red font) and aromatic (blue font); green indicates polar (black font) and amphoteric (red font); and yellow indicates tiny (blue font) and small (green font). The species abbreviations for the Transib transposons are as follows: AA, yellow fever mosquito; AG, African malaria mosquito; DP, D. pseudoobscura fruit fly. (Transib1 through Transib5 are from the fruitfly D. melanogaster).
Figure 4 Structural Similarities between the Transib TIRs and V(D)J RSS Signals
The species abbreviations are: AA, yellow fever mosquito; AG, African malaria mosquito; DM, D. melanogaster fruit fly DP, D. pseudoobscura fruit fly; SP, sea urchin. (Transib1 through Transib5 are from the fruit fly D. melanogaster).
(A) Frequencies of the most frequent nucleotides at each position of the consensus sequence of the 5′ TIRs of transposons that belong to 20 families of Transib transposons identified in fruit flies and mosquitoes. The RSS23 consensus sequence is shown immediately under the TIRs consensus sequence. The most conserved nucleotides in the RSS23 heptamer and nonamer, which are necessary for efficient V(D)J recombination, are highlighted. The 23 ± 1 bp variable spacer is marked by Ns.
(B) Non-gapped alignment of consensus sequences of 5′ TIRs from 21 families of Transib transposons.
(C) The 12/23 rule follows from the basic structure of TIRs of the consensus sequences of transposons that belong to the Transib5, Transib2_AG, TransibN1_AG, TransibN2_AG, and TransibN3_AG families. The 5′ TIRs of these transposons are aligned with the corresponding 3′ TIRs. Structures of the 5′ and 3′ TIRs resemble RSS12 and RSS23, respectively.
The C-terminal domain of RAG1 (residues 764–1011) is the portion of RAG1 that is most conserved between RAG1 and Transib TPases. In addition to the catalytic activity attributed to the last residue of the DDE triad, this domain has a strong nonspecific DNA-binding affinity because it binds to coding DNA upstream of the RSS heptamer, and is thought to be involved in RAG1 dimerization [26,27]. This domain is predicted to function analogously in Transib transposons. Several other motifs conserved in Transib TPases and RAG1 include aa residues that have been shown experimentally to be important for specific functions in V(D)J recombination (Figure S3). Based on this information, the function of these motifs in Transib TPases is expected to be similar to that in RAG1. Among the most conserved motifs, motif 5 (see Figures 3 and Figure S3) is of particular interest because its function is not known yet but is expected to play a role both V(D)J recombination and Transib transposition.
In conjunction with detailed studies of the Transib superfamily, we also analyzed the remaining nine known superfamilies of DNA transposons defined by diverse TPases (see Table 1 in [24]). Some of these TPases, including Mariner, Harbinger, P, and hAT, also contain the catalytic DDE triad [23]. However, based on PSI-BLAST searches, no significant similarities between these nine TPases and RAG1 protein were found (data not shown). Therefore, given that the only significant similarity of the RAG1 core was to the Transib TPase, the RAG1 core was re-confirmed as belonging to the Transib superfamily.
In addition to the statistically significant similarity between the approximately 600-aa RAG1 core and Transib TPases, there are two other lines of evidence suggesting evolution of the V(D)J machinery from Transib DNA transposons. They include the characteristic TSDs and structure of the TIRs discussed in the next two sections.
Similar Length of TSDs and Target Site Composition in Transib and RAG1/2-Mediated Transpositions
It has been known that RAG1-mediated transposition in vitro, both intermolecular and intramolecular, is most frequently accompanied by 5-bp TSDs [12,13]. In one study [12], 35 of 38 (92%) TSDs generated during RAG-mediated intermolecular transposition were 5 bp long, and the remaining 8% were either 4 or 3 bp long. Also, 69% of 36 TSDs recovered during RAG-mediated intramolecular transpositions were 5 bp in length; of the remaining ones, 28% were 4 bp and 3% were 3 bp long. In another study [13], six of six TSDs detected in the intermolecular transposition were 5 bp long. Intramolecular transposition mediated by murine RAG1/2 proteins was also studied recently in vivo in yeast [14]. Again, 60% of TSDs recovered in 26 events were 5 bp long [14]. Given the predominance of 5-bp TSDs, it is striking that Transib transposons belong to the only superfamily of eukaryotic DNA transposons with 5-bp TSDs generated upon insertions into the genome [17,24]. To illustrate the characteristic 5-bp TSDs, we show copies of Transib transposons with intact 5′ and 3′ TIRs from diverse families of Transib transposons present in the D. melanogaster, D. pseudoobscura, A. gambiae, and S. purpuratus genomes (Figure S4). Moreover, some families show high target site specificity, e.g., Transib-N1_AG and Transib-N2_AG integrate preferentially at cCASTGg and cCAWTGc, respectively (TSDs are capitalized). RAG1/2-mediated transpositions also show significant target specificity, presumably reflecting the original specificity of the Transib TPase [12]. Indigenous properties of the Transib TPase, that were not related directly to RAG1 functions, including those responsible for the precise 5-bp length of TSDs, might have been altered during evolution of RAG1, leading to occasional 4-bp and 3-bp TSDs that are atypical for Transib transposons. Both RAG1/2-mediated and Transib transpositions show strong preference for GC-rich target sites [12–14,32], even though genomes hosting Transib transposons are AT-rich (Figure S4; Table 2).
Structure of Transib TIRs
The structure and conservation patterns of the 38-bp termini of Transib transposons from 21 different families closely resemble those of RSSs, suggesting that the latter were derived from termini of ancient Transib transposons (Figures 4 and S4). The 38-bp consensus TIR of Transib transposons consists of a conserved 5′-
CACAATG heptamer separated by a variable 23-bp spacer from an
AAAAAAATC-3′ nonamer. This corresponds closely to the structure of RSSs, which are composed of the conserved heptamers 5′-
CACAGTG separated by a variable 22-bp spacers from
ACAAAAACC-like nonamers [1,5–7]. Only bases at positions 1 through 3 in the heptamer and at positions 5 and 6 in the nonamer are universally conserved in RSSs and absolutely essential for efficient V(D)J recombination [5–7]. The corresponding positions are perfectly conserved in all Transib transposons (Figure 4A and 4B; excluding the 85% conserved position 34 in the Transib consensus that corresponds to position 5 in the RSS nonamer). The probability of the observed match between the RSS and Transib termini to occur by chance is less than 10−3 (see Materials and Methods). Although most Transib families are represented by transposons flanked by TIRs similar to RSS23 (Figure 4A), several families include transposons with 5′ and 3′ termini similar to RSS12 and RSS23, respectively (Figure 4C). Therefore, even the 12/23 rule [1] can be derived directly from the sequence structure of known Transib transposons.
RAG1 Core–Like Sequences in the Sea Urchin, Lancelet, Starlet Sea Anemone, and Hydra Genomes
Using RAG1 proteins as query sequences in a WU BLAST search against sea urchin contigs sequenced at Baylor College (see Materials and Methods), we identified eight proteins approximately 30% identical to portions of the RAG1 core and approximately 50% identical to each other (see Figures 2, 5, and S5). Only one protein is present in two copies, which are 94% identical to each other at the DNA level (contigs 81987 and 6797). Both copies appear to be encoded by pseudogenes damaged by a stop codon at the same position of each protein. Interestingly, the 6,690-bp contig 6797 harbours two additional defective pseudogenes coding for different RAG1 core–like proteins (Figure 5). We also identified a 597-aa protein sequence encoded by a single open reading frame (contig 29068, positions 1157–2944), which is 28% identical to nearly the entire RAG1 core (positions 461–1002 in the human RAG1, Figure S5). Extensive analysis of the flanks failed to show any hallmarks of putative transposons that might be associated with this RAG1-like protein, and we did not find any evidence indicating that other RAG1 core–like proteins are encoded by transposable elements (Figure 5).
Figure 5 Schematic Structure of the Sea Urchin RAG1-Like Sequences
Contig accession numbers are shown in the left column. Inverted complement contigs are marked by “c” followed by the contig number. In each contig, RAG1-like proteins (white rectangle) are schematically aligned with the human RAG1 core (top rectangle). Nucleotide positions of the RAG1-like sequences are shown beneath the white rectangles. Three pairs of recently duplicated sequences (nucleotide identity is higher than 95%) are underlined by red, green, and black lines, respectively. Transposable and repetitive elements detected in the flanking regions are marked by painted rectangles. Names of these elements are shown above the rectangles. Asterisks denote stop codons in the corresponding RAG1-like sequences. BLASTP E-values characterizing similarities between the sea urchin and RAG1 proteins are shown above the white rectangles. Multiple alignment of these protein sequences is reported in Figure S5.
Table 2 Preferential Insertion of Transib transposons into GC-Rich Sites
Each of the 35-bp insertion sites corresponds to two 20-bp DNA fragments flanking a genomic Transib element at its 5′ and 3′ termini. One of the 5-bp TSDs flanking the 3′ terminus of a Transib was excluded in each case. Analogously, the 15-bp insertion sites were composed of two 10-bp flanking fragm
Using FGENESH [33], we detected that the RAG1 core–like open reading frame (ORF) in the contig 29068 forms a terminal exon (positions 1154–2947) of an incomplete hypothetical gene composed of two exons (internal and terminal; see Figure S6). The 3′ terminal portion of the internal exon encodes a protein sequence that appears to be marginally similar to an approximately 50-aa fragment of the RAG1 core (positions 394–454 in human RAG1; Figure S5). The RAG1 core–like protein in whole genome shotgun (WGS) contig 12509 (Figure 5) also seems to be encoded by the last exon starting at position 1650 of a hypothetical RAG1-like gene. Although the two proteins are only 38% identical to each other, they share common features: (1) their N-terminal portions are missing and the RAG1-like sequences start at positions 17 or 18; (2) in both proteins the first aa residue overlaps with the acceptor splice site; and (3) their similarity to RAG1 starts at positions corresponding to position 470 of the human RAG1. Remarkably, the acceptor splice site positions in the sea urchin RAG1 core–like proteins closely correspond to those in RAG1 from teleosts (i.e., most of the living ray-finned or bony fish), in which RAG1 is split by an intron at position homologous to Gly460 in human RAG1 [34].
Using the same RAG1 query sequences in a TBLASTN search against WGS trace sequences from the lancelet (Branchiostoma floridae) genome recently sequenced at the Joint Genome Institute (see Materials and Methods), we found that the lancelet genome encodes protein sequences approximately 35% identical to the RAG1 core (Figure S5; RAG1L_BF; BLASTP E-value is equal to 10−34). Again, as in the case of the sea urchin sequences, the lancelet RAG1 core–like elements show no hallmarks of transposons (data not shown). However, unlike highly conserved RAG1 proteins, the RAG1 core–like proteins are remarkably diverse (see Figure 2).
During the second review of the manuscript of this article, we were kindly informed by Dr. Hervé Philippe of a RAG1 core–like sequence present the starlet sea anemone (Nematostella vectensis). After that, we screened all available Trace Archives (Materials and Methods) and detected additional RAG1-like proteins. In starlet sea anemone, several approximately 1000-bp WGS trace sequences were found (e.g., GenBank Trace Archive IDs 668021618, 558173651, 568641192, and 599572062), which encode protein, called RAG1L_NV, that is approximately 30% identical to the human RAG1 core (positions 284–802, TBLASTN, 10−26 < E < 10−7). We also found several approximately 1000-bp WGS trace sequences of Hydra magnipapillata (Trace Archive IDs 688654311, 647073738, 666995387, 687186526, 688683890, and 688948453), coding for protein sequences 26%−30% identical to the RAG1 core (positions 753−995, E-value is approximately equal to 10−7 in a BLASTX search against GenBank). Using these trace sequences, we partially assembled a hydra gene, called RAG1L_NM, which encodes the RAG1 core–like protein.
Remarkably, the hydra RAG1L_NM protein turned out to be significantly similar to the Transib TPase (26% identity; E-value is approximately equal to 10−14 in a BLASTX search against GenBank proteins combined with the Transib TPase sequences). Therefore, the hydra RAG1 core–like protein provides the first direct link between the RAG1 core and Transib TPase.
N-Terminal–Like Domain of RAG1 in the Sea Urchin, Lancelet, Starlet Sea Anemone, and Hydra Genomes
A separate analysis of the assembled sea urchin sequences yielded seven sequences encoding three diverse proteins that were significantly similar to the 380-aa N-terminal domain of RAG1 (BLASTX, E < 10−4), excluding the 100-aa N-terminus (Figure 6). The first 305-aa protein is encoded by contig 1226, and its recently duplicated copies are on contigs 1219 and 1222 (approximately 95% identical to each other at the protein level.) The second, 195-aa protein (contig 83099) is the shortest. It is only approximately 26% identical to the first protein and more than 90% identical at the DNA level to its duplicate on contig 86231. We also found a third protein on contig 768 that contains unique motifs in its N-terminal regions that best match the homologous regions of RAG1. Furthermore, we found that unassembled WGS trace sequences encode two other proteins, P4_SP and P5_SP, similar to the N-terminal RAG1 domain (Figure 6).
Figure 6 Multiple Alignment of the RAG1 N-Terminal Domain and Sea Urchin Protein Sequences
RAG1_HS, RAG1_PD, RAG1_SS, RAG1_RM, and RAG1_LM mark the human (GenBank accession number NP_000439), lungfish (AAS75810), pig (BAC54968), stripe-sided rhabdornis or Rhabdornis mysticalis bird (AAQ76078), and latimeria (AAS75807) proteins, respectively. The sea urchin and lancelet proteins are marked by “_SP” and “_BF” following the identification numbers of the corresponding contigs. Protein sequences assembled from the sea urchin and lancelet WGS Trace Archives are denoted as P4-P5_SP and P1-P5_BF, respectively. Three conserved motifs are underlined and numbered. The third conserved motif is known as the ring finger. Distances from the protein N-termini are indicated by numbers.
By analyzing the lancelet WGS traces, we also found that the lancelet genome encodes five different proteins similar to the N-terminal domain of RAG1 (BLASTP E values in searches against all GenBank proteins were in a range of 10−14–10−7). DNA sequences coding for these proteins, P1_BF through P5_BF, were manually assembled from overlapping WGS sequences (data available upon request).
The proteins detected in the sea urchin and lancelet genome share a ring finger motif as well as two novel motifs matching the N-terminal RAG1 domain (Figure 6) and remotely resembling C-x2-C zinc finger motifs. The new conserved motifs are H-x3-L-x3-C-R-x-C-G and D-x3-I-h-P-x2-F-C-x2-C, and their function remains to be determined. It is thought that the ring finger motif of RAG1 functions as a zinc-binding domain, is involved in dimerization [30,35], and acts as an E3 ligase in the ubiquitylation [36]. It also likely that the N-terminal RAG1 and RAG1-like proteins share an additional conserved motif W-x-p-h-x(3–6)-C-x2-C that resides between conserved motif 2 and the ring finger (Figure 6).
None of the sea urchin and lancelet proteins align to the approximately 100-aa N-terminus of RAG1, which may indicate that this portion is missing from the genome or highly diverged and difficult to detect. It is also worth noting that this portion corresponds to a separate exon in some teleosts (see Discussion). The ring finger motif itself is also present in several sea urchin proteins unrelated to RAG1 but significantly similar to diverse proteins associated with immune and developmental systems as well as regulation of transcription. To test whether the reported sea urchin sequences represent a true RAG1-like match, we cut off the ring finger motif and repeated the BLASTP search against all GenBank proteins. Even without the finger, the remaining portions of the sea urchin sequences were significantly similar to the corresponding portions of RAG1. BLASTP E-values were 9×10−9, 7×10−5, and 10−3 for the P5_SP, P4_SP, and 768_SP sequences, respectively; because both the low-complexity filter and composition-based statistics were applied, the corresponding E-values were estimated very conservatively. BLASTP searches of the sea urchin sequences against all GenBank proteins, excluding RAG1, detected only the ring finger domain of the sea urchin sequences. E-values of these matches were much higher than the E-values of similarities to the RAG1 proteins (SP_768: 0.04 versus 7×10−7; SP_86231: 3·10−4 versus 7×10−7; SP_1226: 10−4 versus 2×10−7; P4_SP: 10 versus 2×10−7; P5_SP does not have ring finger and matches RAG1 only, E-value = 9×10−7).
Based on the same approach, our study found that the starlet sea anemone and hydra genomes also encode several families of the N-terminal RAG1 domain that appear to be separate from the RAG1 core–like proteins (data not shown). The only exception was the already mentioned sea anemone RAG1 core–like sequence. The approximately 90-aa N-terminus of the latter sequence is the ring finger (E < 10−7, multiple BLASTP matches against known ring fingers in GenBank).
Discussion
The significant similarity between the Transib TPases and RAG1 core, the common structure of the Transib TIRs and RSSs, as well as the similar size of TSDs characterizing transpositions of Transib transposons and transpositions catalyzed by RAG1 and RAG2, directly support the 25-year-old hypothesis of a transposon-related origin of the V(D)J machinery. Previously, the “RAG transposon” hypothesis was open to challenge by alternative models of convergent evolution. Because there were no known TPases similar to RAG1, it could be argued that RAG1 independently developed some TPase-like properties, rather than deriving them from a TE-encoded TPase [24]. These arguments can now be put to rest.
As shown in this paper, the RAG1 core was derived from a Transib TPase, but given the low identity between the Transib TPase and the RAG1 core (14%–17%) it is not clear whether the ancestral transposon was a member of the group of canonical Transib transposons preserved in modern genomes of insects, hydra, and sea urchin (see Figure 1), or a member of an unknown group of Transib transposons that encoded a TPase that was more similar to RAG1 core than to the canonical TPase from the currently known Transib transposons. Furthermore, after its recruitment, the RAG1 core most likely went through a period of intensive transformations due to diversifying/positive selection, which further decreased its similarity to Transib TPase. Afterwards, the RAG1 genes continued to evolve at a slow and steady pace under stabilizing selection, as indicated by the observed conservation of the RAG1 core (79% identity between sharks and mammals).
Some of the intermediate stages of RAG1 evolution can be inferred from analysis of the sea urchin in which RAG1-like proteins were recently observed [37], and from analysis of the lancelet, starlet sea anemone, and hydra genomes. Based on the presence of stop codons disrupting some of the RAG1-like sequences, it has been suggested [37] that the sea urchin sequences represent remnants of transposable elements. Typically, TPase-coding autonomous DNA transposons are present in only a few complete copies per genome. At the same time, sequences homologous to their terminal portions, including specific TIRs, are usually abundant due to the proliferation of nonautonomous DNA transposons fueled by the TPase expressed by the corresponding low-copy autonomous elements. Therefore, even if only 30% of the sea urchin genome has been sequenced to date, it is expected that the regions flanking the TPase portions of potential autonomous elements should be similar to numerous nonautonomous elements. So far, we have found no evidence of such similarities. Detailed analysis of regions flanking the sea urchin RAG1-like DNA coding sequences revealed a variety of different transposable elements inserted in the proximity of the coding sequences (see Figure 5). Nevertheless, based on the orientations and relative positions of these transposons, none of them appears to be associated with the RAG1-like sequences (see Figure 5). We also could not identify the 5-bp TSDs and TIRs characteristic of the Transib superfamily. Still, given that only one third of the sea urchin genome is currently assembled as a set of contigs longer than several thousand nucleotides (the remaining portion is represented by short WGS sequences), we cannot rule out the possibility that the sea urchin RAG1-like proteins are remnants of an unknown branch of Transib transposons. Given that the genomes of lancelet, hydra, and starlet sea anemone are currently available only as unassembled WGS traces, the question whether the corresponding RAG1-like sequences are remnants of transposons or genes/pseudogenes must be left open.
The alternative possibility is that the sea urchin RAG1 core–like sequences represent diverse genes and pseudogenes that belong to a rapidly evolving multigene family. This opens the tantalizing possibility that the RAG1 core was recruited from a Transib TPase in a common ancestor of Bilaterians and Cnidarians, and subsequently lost in nematodes, insects, and sea squirts (see Figure 1). Furthermore, given that the sea urchin, lancelet, hydra, and starlet sea anemone genomes harbor several highly divergent N-terminal–like domains, separate from the RAG1 core–like sequences and known transposable elements, it is very likely that the N-terminal–like domains of RAG1 also form a multigene family that can be traced back to a common ancestor of Deuterostomes (see Figure 1). If so, then both N-terminal and core domains of RAG1 might have been derived from different genes present in a common ancestor of Deuterostomes. Alternatively, the N-terminal domain of RAG1 might have been derived from a separate, unknown transposon. The N-terminal domain of RAG1 has long been viewed as distinct from the core domain due to its lack of direct involvement in the V(D)J recombination reaction. In the sea urchin, lancelet, hydra, and starlet sea anemone genomes, the RAG1 core–like sequences and the N-terminal domain–like sequences do not appear to be linked to each other or to any other proteins. The only notable exception is the anemone RAG1 core–like protein sequence, which is capped by the 90-aa ring finger motif. Taken together with the fact that only the RAG1 core is significantly similar to Transib TPase, the data suggest that the vertebrate RAG1 represents a fusion of once separate proteins. This is consistent with the observation that in teleosts, (bony fish) the RAG1 gene is divided into exons by either one or two introns. As a result, the RAG1 core is split into separate exons at the aa position that corresponds to position 460 in the human RAG1gene [29,34,38]. The core-like sequences encoded by the sea urchin WGS sequence contigs 29068 and 12509 correspond to either the second or third RAG1 exon in teleosts (depending on the number of introns), which is remarkably consistent with the fusion model. The same model predicts that the N-terminal domain of RAG1 could also be assembled from two separate domains based on the presence of the second intron in some teleosts, splitting the N-terminal domain into the 102-aa N-terminal subdomain and the rest [34]. As indicated above, this subdomain, corresponding to the first exon in the genes split by two introns, appears to be missing in the sea urchin, lancelet, hydra, and starlet sea anemone N-terminal–like proteins. It may be encoded by a separate exon that is difficult to detect given its short length and the high level of sequence divergence between these species and vertebrates, or it might have been added in vertebrates. Similarly, the RAG1 core–like protein in the sea urchin genome is shorter in its N-terminal part than the core domain in vertebrates and the corresponding Transib TPase. Again, it is unclear if this part is not present in sea urchins or simply undetectable due to its small size and the high sequence divergence.
It is currently believed that both RAG1 and RAG2 proteins were originally encoded by the same transposon recruited in a common ancestor of jawed vertebrates [3,12,13,16]. However, none of the Transib transposons identified so far encode any proteins other than the Transib/RAG TPase. Also, we could not find any RAG2-like sequences in the recently sequenced sea urchin, lancelet, hydra, and sea anemone genomes, which encode RAG1-like sequences. Autonomous DNA transposons from the MuDR, Harbinger, and En/Spm superfamilies are each known to encode a second regulatory protein [23,24], whereas some transposons from these superfamilies encode the TPase only. Therefore, it is in principle possible that an ancient vertebrate Transib that was a direct ancestor of the RAG1 core also encoded a second protein, the direct ancestor of RAG2. Nevertheless, the apparent lack of RAG2-like proteins in the sequenced portion of the sea urchin, lancelet, hydra, and sea anemone genomes, as well as in Transib transposons suggests that RAG2 was introduced in a separate event in jawless vertebrates. However, given the low 30% identity between the RAG1 and sea urchin/lancelet/sea squirt RAG1-like proteins, we cannot exclude the possibility that the ancestral RAG2 protein went through a period of strong diversification driven by positive selection, and it can no longer be identified by sequence comparisons but may still be present in invertebrates. In any case, the origin of the V(D)J recombination system in jawless vertebrates appears to be a culmination of earlier evolutionary processes rather than an isolated event associated with insertion of a single transposon. If so, detailed studies of individual components, including active Transib transposons and invertebrate proteins homologous to RAG1 elements can bring new breakthroughs in our understanding of evolutionary and mechanistic aspects of V(D)J recombination.
The observed sequence similarity between the RAG1 and Transib TPase protein can help to identify aa residues in the TPase that are crucial for transposition of Transib transposons. For instance, on the basis of the TPase comparison to RAG1 (see Figures S1 and S3), we were able to identify correct positions of the last two aa residues in the DDE catalytic triad (see Figure 2 in [17]), missed in our previous study due to insufficient data. Interestingly, only two cysteines of the zinc finger B (ZFB) C2H2 motif in RAG1 (residues 695–761) involved in its binding to RAG2 [30,31] are perfectly conserved in the Transib TPases (motif 7; see Figures 3 and Figure S3). The remaining portion of the ZFB motif was probably lost in TPases of insect Transib transposons, which do not encode RAG2-like proteins. Notably, two ZFB cysteines are part of the conserved SxxCxxC motif, and mutations of the serine from the same motif cause severe defects in RAG1 transpositions in vitro [32]. Therefore, the presence of serine in this motif is expected to be crucial to Transib transpositions.
After submission of our manuscript, additional biochemical evidence favoring evolution of V(D)J recombination from transposable elements was reported [25]. Analogously to V(D)J recombination, transposition of the fly Hermes transposon, which belongs to the hAT superfamily, is also characterized by a double-strand break via hairpin formation on flanking DNA and 3′ OH joining to the target DNA [25]. However, although the observed biochemical relationship between the hAT TPase and V(D)J recombination is a step forward in our understanding of transposition reaction, several arguments strongly suggest that V(D)J machinery evolved from a Transib rather than from hAT transposon. First, as we mentioned previously, there is no significant sequence identity between hAT TPases and RAG1, even if one employs a PSI-BLAST search with most relaxed parameters (i.e., E < 10, no filters, no composition-based statistics). Second, although RAG1/2-mediated transpositions are characterized by 5-bp (sometimes 4-bp) TSDs, all known hAT transposons are characterized by 8-bp TSDs. Third, unlike in the case of Transib transposons, TIRs of hAT transposons are different from RSS both in terms of DNA sequence similarities and their conservation patterns (Figure S7). Fourth, hAT- and RAG1/2-mediated transpositions differ dramatically in terms of the GC content of their target sites: Unlike Transib transposons and RAG1 transpositions occurring in GC-rich DNA, hAT transposons tend to be integrated into AT-rich regions (Table S2). All four arguments strongly favor evolution of V(D)J machinery from a Transib transposon. Most likely, the Transib transpositions are also characterized by hairpin intermediates formed by the ends of the donor DNA double-strand breaks, as observed during V(D)J recombination and hAT transposition.
Materials and Methods
DNA and protein sequences.
Assembled D. pseudoobscura sequences were downloaded from the Human Genome Sequencing Center at Baylor College of Medicine through the Web site at http://hgsc.bcm.tmc.edu/projects/drosophila/ on 2 March 2004. Preliminary A. aegypti sequence data were obtained from The Institute for Genomic Research through the Web site at http://www.tigr.org on 4 March 2004. Assembled D. melanogaster sequences were downloaded from the Berkeley Drosophila Genome Project at http://www.fruitfly.org/sequence/download.html on 17 February 2004. Partially assembled S. purpuratus contig sequences were downloaded on 12 August 2004 from the Baylor College of Medicine through the Web site at ftp://ftp.hgsc.bcm.tmc.edu/pub/data/Spurpuratus/blast/Spur20030922-genome. In addition to the assembled contigs, Baylor College of Medicine, Human Genome Sequencing Center (http://www.hgsc.bcm.tmc.edu) produced an approximately 8-Gb set of short unassembled WGS sequences, called “traces”, which cover nearly the entire sea urchin genome. We downloaded these sequences from the GenBank Trace Archive at the National Center for Biotechnology Information (NCBI; ftp://ftp.ncbi.nih.gov/pub/TraceDB/strongylocentrotus_purpuratus/) on 17 November 2004. Also, we downloaded an approximately 5-Gb set of unassembled traces that cover almost completely the 600-Mb genome of Florida lancelet (ftp://ftp.ncbi.nih.gov/pub/TraceDB/branchiostoma_floridae
/; 3 December 2004). These sequences were produced and deposited in the GenBank Trace Archive by Department of Energy Joint Genomic Institute (http://www.jgi.doe.gov/). All other DNA and protein sequences were accessed from GenBank (NCBI) through the server at http://www.ncbi.nih.gov/Genbank/ and from Ensembl (EMBL-EBI and Sanger Institute) via the server at http://www.ensembl.org. Sequences of the Transib1 through Transib4 and Transib1_AG through Transib3_AG transposons [17] were obtained from the D. melanogaster (drorep.ref) and A. gambiae (angrep.ref) sections of Repbase Update [39] at Genetic Information Research Institute (http://www.girinst.org).
Sequence analysis.
Computer-assisted identification and reconstruction of the Transib transposons was done as described previously [17,40–42]. DNA sequence analysis including local sequence alignments, multiple alignments, and reconstruction of the Transib consensus sequences was done using software developed at Genetic Information Research Institute (available upon request) and WU-BLASTN 2.0 (http://blast.wustl.edu). To avoid background noise introduced by mutations, Transib relics, whose TPase-coding regions contained numerous stop codons and indels, were ignored unless several copies were available. (We included in the analysis incomplete relics of the Transib2–5_AA TPases represented by single DNA copies). Prediction of putative exons and introns encoded by the Transib consensus sequences was done with FGENESH [33] (at http://www.softberry.com). Multiple alignments of distantly related RAG1 and Transib TPase protein sequences were created by T-Coffee [40]. Shading and minor manual refinements of the aligned sequences were done using Genedoc [43]. Phylogenetic trees were produced by using MEGA3 [44]. Some of the sea urchin sequences encoding the RAG1 N-terminal domain were assembled from traces based on the Baylor BAC-Fisher server at http://www.hgsc.bcm.tmc.edu/BAC-Fisher
/ (the results of assembly were verified manually).
All GenBank proteins were downloaded from ftp://ftp.ncbi.nih.gov/blast/DB/fasta/nr (February 2004) and were combined into a single set with the identified Transib TPases. No Transib TPases had been deposited or annotated previously in GenBank, except for two short hypothetical proteins predicted automatically during annotation of the D. melanogaster genome: 151-aa gi:30923617 and 123-aa gi:30923765. These proteins are apparent fragments of Transib TPases encoded by relics of Transib transposons, including Transib5_DM.
A standalone 2001 version of PSI (Position-Specific Iterating)-BLAST [18,45] was used for detection of proteins that were significantly similar to TPases encoded by Transib and other superfamilies of DNA transposons. The PSI-BLAST program [18,45] is much more sensitive than a regular BLAST search due to the use of PSSM) PSI-BLAST first performs a standard BLASTP search of a protein query against a protein database and constructs a multiple alignment of matches exceeding a certain E-value threshold (called Ei value for the inclusion of sequences into PSI-BLAST iterations). From this alignment, a PSSM is constructed. The PSSM is a weight matrix indicating the relative occurrence of each of the 20 aa at each position in the alignment. This new PSSM is used as the score matrix for a new BLAST search in a second iteration. The process is repeated for a specific number of iterations or until convergence, when no additional proteins are added on successive iterations. The use of a PSSM in place of a fixed generic substitution matrix such as BLOSUM62 results in a much more sensitive BLAST search [18,45]. Important practical aspects of using PSI-BLAST were recently described [46].
To ensure that a conservation profile for the Transib TPases and RAG1 proteins was not produced by a systematic error, we employed a procedure of “step-wise” PSI-BLAST iterations. In this procedure we studied dependence of Ei values on the number of the Transib TPases combined with the GenBank proteins. The following protocol describes the procedure: (1) Use a GenBank set combined with N number of Transib TPases (in our studies, N was equal to 7, 13, and 18), (2) run PSI-BLAST against GenBank combined with TPases using each TPase as a query or seed, (3) select only Transib TPase sequences with E-values less than 10−5 to define the PSSM, (4) take the best E-value (Ei) obtained by PSI-BLAST for RAG1s when PSSM is constructed without RAG1, then (5) repeat these operations for different numbers (N) of TPases.
Significant convergence of RAG1 and Transib TPases was observed to be independent of the particular type of substitution matrix (the same result was observed for both BLOSUM62 and PAM70 matrixes). To avoid detection of false similarities caused by simple repeats and coiled coils, the PSI-BLAST search was performed using stringent conditions with the SEG [47] and COILS [48] filters masking all low-complexity regions and coiled coils, respectively; composition-based statistics [45] were also employed.
The probability P1 that the 5′ terminus of a transposon from a particular Transib family would match by chance an RSS at its most conserved positions (positions 1−3 in the RSS heptamer, and positions 5 and 6 in the RSS nonamer) was estimated based on the following formula: P1 = fC × fA × fC × fA × fA, where fC (0.2) and fA (0.3) are frequencies of C and A in a set of 38-bp 5′ termini of Transib transposons from 21 families (see Figure 4). The value of P1 is 0.001, indicating a significant similarity between Transib TIRs and RSS. Indeed, given that these five positions conserved in RSS are conserved in all TIRs from 21 families of Transib transposons, and the average identity between these 38-bp TIRs is only 49%, the chance of randomly matching these positions in TIRs from all 21 families is extremely small.
TBLASTN searches against the Trace Archive were performed by using the BLAST client (blastcl3 or netblast at ftp://ftp.ncbi.nlm.nih.gov/blast/executables/LATEST/, which accesses the NCBI BLAST search engine. Names of all available Trace Databases were taken from a list of databases at http://www.ncbi.nlm.nih.gov/blast/mmtrace.shtml.
Supporting Information
Figure S1 Similarity between C-Terminal Portions of the Transib2_AG TPase and RAG1
Two examples extracted from the NCBI BLASTP output illustrate similarity between the approximately 60-aa C-terminal portions of the Transib2_AG TPase (which we used as a query in a BLASTP search against all GenBank proteins) and the RAG1 core.
(751 KB EPS).
Click here for additional data file.
Figure S2 Multiple Alignment of Transib TPases
The catalytic DDE triad is marked by black rectangles. Amino acids are shaded on the basis of their physiochemical properties according to the color scheme implemented in Genedoc [43]: Black shading marks hydrophobic residues, blue indicates charged (white font), positively charged (red font), and negatively charged (green font); red indicates proline (blue font) and glycine (green font); gray indicates aliphatic (red font) and aromatic (blue font); green indicates polar (black font) and amphoteric (red font); yellow indicates tiny (blue font) and small (green font). The species abbreviations are as follows: SP, sea urchin; DP, D. pseudoobscura fruit fly; AG, African malaria mosquito; AA, yellow fever mosquito. Transib1 through Transib5 are from the D. melanogaster fruit fly genome.
(3 MB EPS).
Click here for additional data file.
Figure S3 Multiple Alignment of the RAG1 Core and Transib TPase Proteins
The shading scheme is the same as in Figure S2. The catalytic DDE triad is marked by black rectangles. RAG1 aa whose replacements resulted in previously detected defects of V(D)J recombination [31] are marked by color rectangles indicated below the alignment blocks; red indicates DNA binding defect; green indicates nicking defect; cyan indicates hairpin defect; blue indicates joining mutants; yellow indicates catalytic mutants; gray indicates joining/transposition. Presence and absence of corresponding residues in the Transib TPases are indicated by + and −, respectively. Conserved motifs are marked by lines numbered from 1 to 10. The species abbreviations are as follows: DP, D. pseudoobscura fruit fly; AG, African malaria mosquito; AA, yellow fever mosquito; GG, chicken; HS, human; XL, frog; CL, bull shark; FR, fugu fish.
(3 MB EPS).
Click here for additional data file.
Figure S4 TSDs in Transposons from Different Transib Families
For each family, DNA copies of transposons are aligned to the corresponding consensus sequence. The consensus sequence is shown in the top line. Dots indicate nucleotide identity with the consensus sequence; hyphens represent alignment gaps. Internal portions of transposons are not shown and are marked by xxx. TSDs are highlighted. Coordinates of the reported elements are shown in the first two columns (sequence name, beginning to end).
(A) TransibN1_AG family from mosquito.
(B) TransibN2_AG family from mosquito.
(C) TransibN3_AG family from mosquito.
(D) TransibN1_DP family from fruit fly.
(E) Hopper family from fruit fly.
(F) TransibN1_DM family from fruit fly.
(G) TransibN1_SP family from sea urchin.
(179 KB PDF).
Click here for additional data file.
Figure S5 Multiple Alignment of the RAG1 Core and RAG1 Core–Like Proteins Encoded by the Sea Urchin and Lancelet Genomes
The shading scheme is the same as in Figure S2 and S3. The species abbreviations are as follows: SP, sea urchin; BF, lancelet; HS, human; CL, bull shark; GG, chicken; XL, frog; FR, fugu fish. The lancelet RAG1L_BF protein is encoded by several overlapping WGS trace sequences (for example, GenBank Trace Archive identification numbers 543943730, 538583629).
(2.8 MB EPS).
Click here for additional data file.
Figure S6 RAG1-Like Protein SP_29068 in the Sea Urchin Contig 29068
(A) Exon/intron structure of the SP_29068 gene is reported based on the FGENESH prediction.
(B) Alignment of the predicted protein and human RAG1 (29% identity, E = 10−43. The intron in SP_29068 is inserted between residues shaded in green and red. Gly460 that harbors the intron in the teleost RAG1 is shaded in black.
(1.5 MB EPS).
Click here for additional data file.
Figure S7 Structure of hAT 5′ Termini
Non-gapped alignment of consensus sequences of 5′ termini of transposons from 22 different families is shown beneath the RSS23 consensus sequence, composed of the RSS heptamer and nonamer. The most conserved nucleotides in the heptamer and nonamer, which are necessary for efficient V(D)J recombination, are highlighted. Among the necessary RSS nucleotides, only one, marked by a + corresponds to a nucleotide that is 100% conserved in hAT transposons. The critical third nucleotide of the hAT 5′ termini is always G, as opposed to C in the RSS heptamer. It is also clear from the alignment that the hAT termini do not have any second conserved block, which is expected to be preserved if RSSs have evolved from hAT termini. Hobo (GenBank number X04705), Homer (AF110403), Hermes (L34807), Ac9 (K01904), Tam3_AM (X55078), TAG1 (L12220), Pegasus (U47019) are active hAT transposons from fruit fly, Queensland fruit fly, house fly, maize, snapdragon, thale-cress, and African malaria mosquito, respectively. HOPPER_BD is from oriental fruit fly (GenBank AF486809). The consensus sequences of hAT-1N_DP and hAT-1N_DP (nonautonomous transposons from fruit fly, D. pseudoobscura); HAT1N_DR, hAT-2n1_DR, and hAT-N19_DR (nonautonomous transposons from zebrafish); CHARLIE1A and CHESHIRE (human); hAT-N1_SP (sea urchin); ATHAT1, ATHAT7, and ATHAT10 (thale-cress); PegasusA, HATN4_AG, and hAT-2N_AG (African malaria mosquito) were reported in Repbase Update.
(775 KB EPS).
Click here for additional data file.
Table S1
Transib TPase in Eukaryotes
Columns 1 and 2 list common and Latin names of species whose genomes contain Transib TPase sequences. Column 3 shows GenBank sections collecting corresponding sequences: "NR", "WGS", "EST", and "HTGS" are names of GenBank sections; "tr" stands for “Trace Archives.” Column 4 shows a range of E-values of matches between the sea urchin Transib TPase (Transib1_SP) and TPases encoded by the listed species that were detected in TBLASTN searches against corresponding sections of GenBank. Matches to the Transib TPase observed for Oryza sativa indica (seven sequences from Trace Archives, 10−48 < E < 10−13) were discarded as a likely sequencing contamination, based on the fact that these sequences were over 80% identical to Hydra magnipapillata traces (the hydra Trace Archive dataset contains over 100 sequences matching the TPase, and hydra Transib TPase sequences are also present in the dbEST section of GenBank). Analogously, matches to the Transib TPase detected in the AC011430 HTGs and AADC01054609 WGS GenBank sequences, which were annotated as portions of the human genome, were discarded as products of contamination (these sequences contain 100% identical copies of the non-long terminal repeat (LTR) retrotransposon G2_DM [17] from D. melanogaster).
(27 KB DOC).
Click here for additional data file.
Table S2 GC Content of Target Sites for hAT Transposons
The table shows that hAT transposons are inserted preferentially into GC-rich sites. Each of the 35-bp insertion sites corresponds to two 14-bp and 13-bp DNA fragments flanking a genomic hAT element at its 5′ and 3′ termini; one of the 8-bp TSDs (flanking the 3′ terminus of a transposon) was excluded in each case. Analogously, the 15-bp insertion sites were composed of two 4-bp and 3-bp flanking fragments. (1) GenBank accession number U47019; (2) Repbase Update, the angrep.ref section; (3) GenBank X04705; (4) Repbase Update, the drorep.ref section; (5) Repbase Update, spurep.ref; (6)Repbase Updates, the zebrep.ref section. Copies of Pegasus, HATN4_AG, and HAT2N_AG were identified in the mosquito A. gambiae genome; Hobo and hAT-1N_DP in the D. melanogaster and D. pseudoobscura fruit fly genomes, respectively; HAT-1N_SP in the sea urchin genome; and HAT1N_DR, HAT-2N1_DR, and HAT-N19_DR in the zebrafish genome.
(27 KB DOC).
Click here for additional data file.
Accession Numbers
The sea urchin Transib1_SP transposon, RAG1L_HM, RAG1L_BF, RAG1L_NV, 81978_SP, 12509_SP, 6797–1_SP, 6797–2_SP, 6797–3_SP, 8813_SP, 71716_SP, and 29068_SP genes/pseudogenes have been deposited on our website (http://girinst.org/server/publ/PLOS.2005) and in the Third Party Annotation (TPA) database of GenBank (http://www.ncbi.nih.gov/Genbank/TPA.html; accession numbers are pending. The Transib1, Transib2, Transib3, Transib4, Transib1_AG, Transib2_AG, Transib3_AG, Transib1_DP, Transib2_DP, Transib3_DP, Transib4_DP, Transib1_AA, Transib2_AA, Transib3_AA, Transib4_AA, Transib5_AA, Transib1_SP, TransibN1_SP, TransibN1_AG, TransibN2_AG, TransibN3_AG, TransibN1_DM, TransibN1_DP, TransibN2_DP, TransibN3_DP, TransibN4_DP, and TransibN5_DP transposons are deposited in the drorep (D. melanogaster), angrep (A. gambiae), spurep (S. purpuratus), and invrep (invertebrates) sections of Repbase Update (http://www.girinst.org/Repbase_Update.html).
We gratefully acknowledge David Schatz for detailed suggestions and encouragements, Hervé Philippe for constructive criticism and indicating that a RAG1 core–like sequence is present in the sea anemone genome, Andrew Gentles for critical reading of the manuscript, Adam Pavlicek and Michael Ponomarenko for discussions, anonymous reviewers for constructive criticism, Oleksiy Kohany for assistance with computational analysis, and Jolanta Walichiewicz for technical assistance. We also thank the Institute of Cytology and Genetics (Novosibirsk, Russia) for hospitality. This work was supported by the National Institutes of Health grant 2 P41 LM06252–04A1
Competing interests. The authors have declared that no competing interests exist.
Author contributions. VVK conceived and designed the experiments, and performed the experiments. VVK and JJ analyzed the data, contributed reagents/materials/analysis tools, and wrote the paper.
Citation: Kapitonov VV, Jurka J (2005) RAG1 Core and V(D)J Recombination signal sequences were derived from Transib transposons. PLoS Biol 3(6): e181.
Abbreviations
aaamino acid
BCRB cell receptor
Ei-valueE-value threshold for the first inclusion of matching sequences into the PSI-BLAST iterations
NCBINational Center for Biotechnology Information
PSI-BLASTposition-specific iterated BLAST
PSSMposition-specific score matrix
RSSrecombination signal sequence
TCRT cell receptor
TIRterminal inverted repeat
TPasetransposase
TSDtarget site duplication
WGSwhole genome shotgun
ZFBzinc finger B
==== Refs
References
Tonegawa S Somatic generation of antibody diversity Nature 1983 302 575 581 6300689
Oettinger MA Schatz DG Gorka C Baltimore D RAG-1 and RAG-2, adjacent genes that synergistically activate V(D)J recombination Science 1990 248 1517 1523 2360047
Gellert M V(D)J recombination: RAG proteins, repair factors, and regulation Annu Rev Biochem 2002 71 101 132 12045092
Sakano H Huppi K Heinrich G Tonegawa S Sequences at the somatic recombination sites of immunoglobulin light-chain genes Nature 1979 280 288 294 111144
Akira S Okazaki K Sakano H Two pairs of recombination signals are sufficient to cause immunoglobulin V-(D)-J joining Science 1987 238 1134 1138 3120312
Ramsden DA Baetz K Wu GE Conservation of sequence in recombination signal sequence spacers Nucleic Acids Res 1994 22 1785 1796 8208601
Lee AI Fugmann SD Cowell LG Ptaszek LM Kelsoe G A functional analysis of the spacer of V(D)J recombination signal sequences 2003 PLoS Biol 1 E1
Akamatsu Y Oettinger MA Distinct roles of RAG1 and RAG2 in binding the V(D)J recombination signal sequences Mol Cell Biol 1998 18 4670 4678 9671477
Thompson CB New insights into V(D)J recombination and its role in the evolution of the immune system Immunity 1995 3 531 539 7584143
van Gent DC Mizuuchi K Gellert M Similarities between initiation of V(D)J recombination and retroviral integration Science 1996 271 1592 1594 8599117
Melek M Gellert M van Gent DC Rejoining of DNA by the RAG1 and RAG2 proteins Science 1998 280 301 303 9535663
Agrawal A Eastman QM Schatz DG Transposition mediated by RAG1 and RAG2 and its implications for the evolution of the immune system Nature 1998 394 744 751 9723614
Hiom K Melek M Gellert M DNA transposition by the RAG1 and RAG2 proteins: A possible source of oncogenic translocations Cell 1998 94 463 470 9727489
Clatworthy AE Valencia MA Haber JE Oettinger MA V(D)J recombination and RAG-mediated transposition in yeast Mol Cell 2003 12 489 499 14536087
Messier TL O'Neill JP Hou SM Nicklas JA Finette BA In vivo transposition mediated by V(D)J recombinase in human T lymphocytes EMBO J 2003 22 1381 1388 12628930
Lewis SM Wu GE The old and the restless J Exp Med 2000 191 1631 1636 10811857
Kapitonov VV Jurka J Molecular paleontology of transposable elements in the Drosophila melanogaster genome Proc Natl Acad Sci U S A 2003 100 6569 6574 12743378
Altschul SF Madden TL Schaffer AA Zhang J Zhang Z Gapped BLAST and PSI-BLAST: A new generation of protein database search programs Nucleic Acids Res 1997 25 3389 3402 9254694
Sadofsky MJ Hesse JE McBlane JF Gellert M Expression and V(D)J recombination activity of mutated RAG-1 proteins Nucleic Acids Res 1993 21 5644 5650 8284210
Silver DP Spanopoulou E Mulligan RC Baltimore D Dispensable sequence motifs in the RAG-1 and RAG-2 genes for plasmid V(D)J recombination Proc Natl Acad Sci U S A 1993 90 6100 6104 8327489
Kim DR Dai Y Mundy CL Yang W Oettinger MA Mutations of acidic residues in RAG1 define the active site of the V(D)J recombinase Genes Dev 1999 13 3070 3080 10601033
Landree MA Wibbenmeyer JA Roth DB Mutational analysis of RAG1 and RAG2 identifies three catalytic amino acids in RAG1 critical for both cleavage steps of V(D)J recombination Genes Dev 1999 13 3059 3069 10601032
Craig NL Craigie R Gellert M Lambowitz AM Mobile DNA II 2002 Washington, DC ASM Press 1204
Kapitonov VV Jurka J Harbinger transposons and an ancient HARBI1 gene derived from a transposase DNA Cell Biol 2004 23 311 324 15169610
Zhou L Mitra R Atkinson PW Hickman AB Dyda F Transposition of hAT elements links transposable elements and V(D)J recombination Nature 2004 432 995 1001 15616554
Arbuckle JL Fauss LA Simpson R Ptaszek LM Rodgers KK Identification of two topologically independent domains in RAG1 and their role in macromolecular interactions relevant to V(D)J recombination J Biol Chem 2001 276 37093 37101 11479318
Mo X Bailin T Sadofsky MJ A C-terminal region of RAG1 contacts the coding DNA during V(D)J recombination Mol Cell Biol 2001 21 2038 2047 11238939
Difilippantonio MJ McMahan CJ Eastman QM Spanopoulou E Schatz DG RAG1 mediates signal sequence recognition and recruitment of RAG2 in V(D)J recombination Cell 1996 87 253 262 8861909
Spanopoulou E Zaitseva F Wang FH Santagata S Baltimore D The homeodomain region of Rag-1 reveals the parallel mechanisms of bacterial and V(D)J recombination Cell 1996 87 263 276 8861910
Rodgers KK Bu Z Fleming KG Schatz DG Engelman DM A zinc-binding domain involved in the dimerization of RAG1 J Mol Biol 1996 260 70 84 8676393
Aidinis V Dias DC Gomez CA Bhattacharyya D Spanopoulou E Definition of minimal domains of interaction within the recombination-activating genes 1 and 2 recombinase complex J Immunol 2000 164 5826 5832 10820261
Tsai CL Drejer AH Schatz DG Evidence of a critical architectural function for the RAG proteins in end processing, protection, and joining in V(D)J recombination Genes Dev 2002 16 1934 1949 12154124
Solovyev VV Jiang T Smith T Xu Y Zhang M Finding Genes by Computer: Probabilistic and Discriminative Approaches Current Topics in Computational Biology: MIT Press 2002 361 401
Willett CE Cherry JJ Steiner LA Characterization and expression of the recombination activating genes (rag1 and rag2) of zebrafish Immunogenetics 1997 45 394 404 9089097
Bellon SF Rodgers KK Schatz DG Coleman JE Steitz TA Crystal structure of the RAG1 dimerization domain reveals multiple zinc-binding motifs including a novel zinc binuclear cluster Nat Struct Biol 1997 4 586 591 9228952
Yurchenko V Xue Z Sadofsky M The RAG1 N-terminal domain is an E3 ubiquitin ligase Genes Dev 2003 17 581 585 12629039
Cannon JP Haire RN Rast JP Litman GW The phylogenetic origins of the antigen-binding receptors and somatic diversification mechanisms Immunol Rev 2004 200 12 22 15242392
Venkatesh B Erdmann MV Brenner S Molecular synapomorphies resolve evolutionary relationships of extant jawed vertebrates Proc Natl Acad Sci U S A 2001 98 11382 11387 11553795
Jurka J Repbase update: A database and an electronic journal of repetitive elements Trends Genet 2000 16 418 420 10973072
Notredame C Higgins DG Heringa J T-Coffee: A novel method for fast and accurate multiple sequence alignment J Mol Biol 2000 302 205 217 10964570
Kapitonov VV Jurka J Rolling-circle transposons in eukaryotes Proc Natl Acad Sci U S A 2001 98 8714 8719 11447285
Kapitonov VV Jurka J The esterase and PHD domains in CR1-like non-LTR retrotransposons Mol Biol Evol 2003 20 38 46 12519904
Nicholas KB Nicholas HB Deerfield DW II GeneDoc: Analysis and Visualization of Genetic Variation EMBNEW News 1997 4 14
Kumar S Tamura K Nei M MEGA3: Integrated software for Molecular Evolutionary Genetics Analysis and sequence alignment Brief Bioinform 2004 5 150 163 15260895
Schaffer AA Aravind L Madden TL Shavirin S Spouge JL Improving the accuracy of PSI-BLAST protein database searches with composition-based statistics and other refinements Nucleic Acids Res 2001 29 2994 3005 11452024
Koonin EV Galperin MY Sequence–evolution–function. Computational approaches in comparative genomics 2003 Norwell (Massachusetts) Kluwer Academic Publishers
Wootton JC Federhen S Analysis of compositionally biased regions in sequence databases Methods Enzymol 1996 266 554 571 8743706
Lupas A Predicting coiled-coil regions in proteins Curr Opin Struct Biol 1997 7 388 393 9204281
Douzery EJ Snell EA Bapteste E Delsuc F Philippe H The timing of eukaryotic evolution: Does a relaxed molecular clock reconcile proteins and fossils? Proc Natl Acad Sci U S A 2004 101 15386 15391 15494441
Peterson KJ Lyons JB Nowak KS Takacs CM Wargo MJ Estimating metazoan divergence times with a molecular clock Proc Natl Acad Sci U S A 2004 101 6536 6541 15084738
Pires-daSilva A Sommer RJ Conservation of the global sex determination gene tra-1 in distantly related nematodes Genes Dev 2004 18 1198 1208 15155582
| 15898832 | PMC1131882 | CC BY | 2021-01-05 08:21:23 | no | PLoS Biol. 2005 Jun 24; 3(6):e181 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030181 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1589883310.1371/journal.pbio.0030193Research ArticleEcologyEvolutionGenetics/Genomics/Gene TherapyMedical HistoryStatisticsHomo (Human)On the Number of New World Founders: A Population Genetic Portrait of the Peopling of the Americas Peopling of the AmericasHey Jody
1
1Department of Genetics, Rutgersthe State University of New Jersey, Piscataway, New JerseyUnited States of AmericaClark Andy G. Academic EditorCornell UniversityUnited States of America6 2005 24 5 2005 24 5 2005 3 6 e19312 7 2004 25 3 2005 Copyright: © 2005 Jody Hey.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.
From Few to Many: New World Founded by Surprisingly Small Population
The founding of New World populations by Asian peoples is the focus of considerable archaeological and genetic research, and there persist important questions on when and how these events occurred. Genetic data offer great potential for the study of human population history, but there are significant challenges in discerning distinct demographic processes. A new method for the study of diverging populations was applied to questions on the founding and history of Amerind-speaking Native American populations. The model permits estimation of founding population sizes, changes in population size, time of population formation, and gene flow. Analyses of data from nine loci are consistent with the general portrait that has emerged from archaeological and other kinds of evidence. The estimated effective size of the founding population for the New World is fewer than 80 individuals, approximately 1% of the effective size of the estimated ancestral Asian population. By adding a splitting parameter to population divergence models it becomes possible to develop detailed portraits of human demographic history. Analyses of Asian and New World data support a model of a recent founding of the New World by a population of quite small effective size.
A new population-genetic method for assessing human demographic history reveals that the effective size of the founding population of the New World comprised less than 80 individuals.
==== Body
Introduction
Archeological evidence, as well as anatomical, linguistic, and genetic evidence, have shown that the original human inhabitants of the Western Hemisphere arrived from Asia during the Late Pleistocene [1–4]. However, there persists uncertainty on the source, within Asia, of peoples who migrated to the New World [5], on the timing of the earliest migration [6–10], and on whether there have been multiple migrations [3,11–13].
For complex historical subjects such as the colonization of the Americas, there are many ways that models can be constructed, examined, and compared. One approach is to develop a portrait based on a particular kind of data, such as linguistic [6], skeletal [14], or archaeological [15] data, or on DNA sequence data from a particular portion of the human genome such as the mitochondria [4,16–19] or the Y chromosome [9]. Yet each source of data has unique sources of variation. In the case of genetic data there occurs a large stochastic variance of the coalescent history among genes that causes different loci to vary widely in levels of genetic variation and in apparent patterns of relationships among populations [20–22]. This stochastic variance is sometimes overlooked, for example in discussions of the histories of the individual DNA sequence haplotypes [18], and it is easy to underestimate the many possible histories that are consistent with a finding that haplotypes are shared by different populations [23–25].
To accommodate the stochastic variance among loci, population geneticists have turned in recent years to Bayesian and likelihood methods that explicitly take into account the range of possible gene tree histories that are consistent with a given dataset [26–30]. For questions on population divergence, the focus has been on models of population splitting in which an ancestral population divides into two descendant populations, after which there may be gene flow between the descendant populations. These “isolation with migration” (IM) models can have a large number of parameters, and they offer the possibility of capturing many of the dynamics that occur in the early stages of population divergence or speciation [30–33].
Figure 1A shows the basic IM model, in which the ancestral and descendant populations each have a constant size. Each of the terms in the model is explained in Table 1. Basic limitations of this model are that it cannot provide details on how descendant populations were founded or whether population sizes have changed. Certainly for human populations there is considerable genetic evidence that population sizes have grown [34–37], and it would be helpful if it were possible to capture information on the sizes of descendant populations as they are formed. For example, if one descendant population formed as a small founder population that later grew to a large size, such dynamics would not be revealed in the fitting of the basic IM model. To allow the study of such histories, an additional parameter has been added to the IM model. Figure 1B shows a model in which an ancestral population splits in two, with the relative sizes of those two new populations reflected in the parameter s, where 0 < s < 1. At the time of the split, descendant population 1 has size sNA from which it moves to size N1 at the time of sampling. Similarly, population 2 begins with size (1 − s)NA from which it moves to size N2 at the time of sampling. Figure 1B depicts one population growing and the other shrinking, but in fact either population is free to either grow or shrink under this model.
Figure 1 Isolation with Migration Models
(A) The basic IM model. The demographic terms are effective population sizes (N1, N2, and NA), gene flow rates (m1 and m2), and population splitting time (t). Also shown are parameters scaled by the neutral mutation rate (u), as they are actually used in the model fitting. Terms for basic demographic parameters, including N, m, t, and u, are not italicized. Note that the migration parameters are identified by the source of migrants as time goes backward in the coalescent. In other words, the migration rate from population 1 to population 2 (i.e., m1) actually corresponds to the movement of genes from population 2 to population 1 as time moves forward.
(B) The IM model with changing population size. An additional parameter, s, is the fraction of NA that forms N1 (i.e., the fraction 1 − s gives rise to N2)
Table 1 Parameter Summary and Description
These models were applied to questions on the founding of New World populations from Asia. A total of nine DNA sequence datasets that included Asian and Native American (Amerind-speaking) samples were drawn from the literature (Figure 2 and Table 1) and analyzed jointly using a procedure that provides posterior probability distributions for each of the model parameters [30,33]. The stochastic variance among loci is clearly evident in the variation of F
ST values (between Asian and New World samples) observed among the loci. Of the nine loci included in the present study, three have fairly high F
ST values, while the remainder are either negative (undefined) or near zero (Table 1).
Figure 2 Approximate Geographic Locations, and Sample Sizes per location, for Each Locus Listed in Table 1
In some cases locations are based on actual geographic locations, in other cases the locations are the approximate center of the geographic region occupied by ethnic groups identified in the original references (Table 1).
Asian samples were arbitrarily designated as being from population 1 and the New World samples from population 2. In this case, 1 − s is the fraction of the ancestral population that founded the New World population. The analyses also require that prior distributions be specified for each model parameter. It was assumed that the New World was founded by a minority of the ancestral Asian population, corresponding to a specified uniform prior distribution for s between 0.5 and 1. For the other parameters, flat prior distributions were selected that would span the entire range of the posterior densities (i.e., uninformative priors) [30]. However, in some cases the posterior distributions were quite flat over the highest portions of parameter ranges. In these cases the choice of the upper bound on the prior distribution does affect the posterior distribution, and we are not able to use an uninformative prior distribution. However, parameters can still be estimated on the basis of the locations of peaks in the parameter regions that can be assessed, and the effect of altering the prior distribution on these estimates can be determined.
The overall picture that emerges is one in which the New World was very recently founded by a small number of individuals (effective size of about 70), and then grew by a factor of about 10. The data do suggest that there has been gene exchange between Asia and the New World since that time; however, the likelihood surfaces are quite flat, so confidence in gene flow estimates is low.
Results
The method assumes that the loci have not been subject to recombination or to directional or balancing selection. For recombination, we used only those loci that showed no evidence of recombination by the four-gamete test [38]. It is possible that this has missed some recombination since the time of common ancestry. Regarding natural selection, the study was limited to loci that had not individually been reported to show evidence of directional or balancing selection. However, it is possible that when considered together, and polymorphism and divergence from chimpanzees are considered under a common neutral model, that there is evidence of selection. An HKA test [39] of the eight loci with estimates of divergence from chimpanzees (Table 2) yielded a p value of 0.054, which is nearly statistically significant. This test assumes, as do the models analyzed in this study, that all loci are sampled from the same panmictic population [39], and it is possible that the differing geographic sources of the loci included in the study may have contributed some variation.
Table 2 Information on Loci Used in the Study
See Dataset S1 and Protocol S1 for more detail.
a The inheritance scalar was set to reflect the expected effective population size experienced by a locus relative to an autosome, assuming equal sex ratios and variance in reproductive success: autosomal loci, 1.0; X-linked loci, 0.75; maternally or paternally inherited loci, 0.25.
b The percentage of basepairs that differ between a human and a chimpanzee sequence.
c
F
ST is the proportion of variation that lies between samples pooled for Asia and the New World for each locus [70,71]. When divergence is low, calculation may yield a negative value (Undef).
d Data from [72]. The β-globin locus falls near a recombination hotspot [73]. Of the 3,011 bases of a large population genetic study of the β-globin region [72], the 5′ half shows ample evidence of historical recombination by the four-gamete criterion [38], whereas the 3′ half that was used for this study showed no evidence of historical recombination. Divergence from chimpanzees was measured over this region from the available chimpanzee sequence [74].
e Full-length mtDNA sequences were used [75,76]. Because of the need for an absence of homoplasy by the computer program fitting the model, control region sequences were removed and only transversion differences were used.
f Concatenated data from several noncoding regions of the nonrecombining portion of the Y chromosomes (NRY) [48]. Human-chimpanzee divergence for the NRY was estimated from 4,758 noncoding basepairs of the SMCY locus [77].
g Data from [78,79].
h Data from [80,81].
i Haplotypes were determined over multiple points across this locus [82]. A data summary was provided by Yvonne Thorstenson. The region used for this analysis included pieces scattered over 96 kilobasepairs that showed no evidence of recombination in Asian and New World samples. This locus was not included in the estimate of mutation rate per year because of length ambiguity of the sampled sequence and uncertainty over human-chimpanzee divergence.
j Data from [83].
The estimated posterior distributions are shown in Figure 3. For the initial analysis, allowing for exponential population size changes, the posterior distribution for t yielded both a major and a minor peak (the curve for t with a high t
upper, Figure 3D). Given the mutation rate estimates (see Table 1), the location of the major peak (t = 0.032) corresponds to 7,130 y, whereas the location of the minor peak (t = 0.27) corresponds to 44,400 y. Given the remote possibility of such an ancient time as the latter, analyses were also done with a smaller upper bound on t of 0.2 (identified as “low t
upper” in Figure 3), which corresponds to 33,000 y. Analyses were done with this reduced upper limit for t for both models in Figure 1, allowing for population size change and for the case of fixed population sizes. In the case of constant population sizes, the distribution for t shows a peak (t = 0.038) very near those for the analyses under population size change; however, the highest posterior density is found at the upper limit of t. When the constant population size model was run with a higher upper limit on t, the posterior distribution showed the same low value peak as well as a steadily rising curve for higher values of t (unpublished data).
Figure 3 Marginal Posterior Probability Densities
Probability densities for each of the parameters described in Figure 1 are shown, as follows: (A) θ1; (B) θ2; (C) θA; (D) t (i.e., t/u); (E) t shown on a scale of years over the range corresponding to a maximum t value of 0.2; (F) s; (G) m
1; and (H) m
2. The analysis in which a high upper limit on the prior distribution for t was used is identified as “high tupper,” while those analyses with a smaller upper limit on the prior distribution of t are identified as “low tupper.” Each curve is based upon the results of multiple simulations over millions of Markov chain updates (see Materials and Methods), and is plotted over the specified prior range of that parameter.
The archaeological portrait of early New World populations has largely centered around widespread Clovis sites that have an earliest estimated age of about 13,000 y before the present [15,40,41]. The oldest generally agreed upon New World archaeological date is from the non-Clovis Monte Verde site in Southern Chile, which has been dated to about 14,000 y before the present [10,42,43]. Clearly the time points associated with our estimates of t are more recent than expected, given the archaeological estimates. However, these distributions do span the time periods that have been most discussed. For example, a time of 14,000 y has a relatively high probability in each of the analyses (Figure 3E). Given that people have lived in the New World probably for only several hundred generations, it is noteworthy both that the posterior densities for t do show clear peaks in the expected time period and that the probability estimates drop to zero as t approaches zero. In other words, the data contain a clear signal of a nonzero, albeit recent, founding time of New World populations.
With regard to migration, each of the three analyses show nonzero peaks for both directions of gene flow. These may well reflect the occurrence of more than one episode of migration to the New World. For example, it has been suggested on the basis of mitochondrial DNA haplotypes and glaciation history that an initial migration along a coastal route may have been followed later by another migration, possibly through an ice-free noncoastal corridor [13]. However, the posterior distributions shown here have little resolution, as all of the curves for m
1 and m
2 are broad, and all have high probability at the lower limit of resolution, indicating that zero gene flow is nearly as well supported by the data as are nonzero gene flow levels.
The ancestral population parameter, θA, shows a relatively narrow distribution with a very consistent peak location across the three analyses. These attributes are partly to be expected, given that the very large majority of the variation in the samples is older than t. In effect, more information is available for θA than for the other parameters. The estimated effective size of the ancestral population is about 9,000 (Table 3), which is roughly consistent with previous estimates for Asian samples [44]. The current Asian population parameter (θ1) revealed broad distributions and estimates that are near those for the ancestral population. Although the estimates of current effective size in Asia vary among the analyses (Table 3), they are all fairly close to the ancestral size estimates, suggesting that there has not been much population growth in Asia since t. Also consistent with the apparent constancy of population size is the distribution of s, the splitting parameter, which shows a peak at 0.992, signifying that only a small portion (less than 1%) of the ancestral Asian population left to found the New World population.
Table 3 Model Parameter Estimates
Parameter estimates are shown for three models described in the text. For those parameters in which the complete posterior distribution appeared to be estimated, the 90% highest posterior density interval was also determined and given as a range (in parentheses). This range is the shortest interval that contains 90% of the probability.
a The location of the highest value of t is at the right margin of the distribution. The location of the secondary peak is also given in parentheses.
NA, not applicable
In contrast to the Asian population, the New World population parameter (θ2) is much smaller, and suggests a recent New World effective population size of less than 1,000 (Table 3). However, given the estimate of the effective size of the founding New World population (about 70; Table 4), the overall picture is of a nearly 10-fold growth in the New World effective size since t.
Table 4 Estimates of Demographic Quantities
The conversion of model parameters to demographic terms is described in “Analyses” in Materials and Methods.
a The estimated time associated with the highest value of t which is at the right margin of the distribution. The estimated time associated with the secondary peak is given in parentheses.
In order to gain a sense of how consistent the data actually are with the model and the parameter estimates, 500 simulated datasets were generated under the model in Figure 1B, with sample sizes and true parameter values (see Table 2, column 3) that were the same as for the actual data. From each simulated dataset, the average number of pairwise differences between sequences were calculated within each population (Asia and the New World) and between these populations. The average of these values from the 500 simulated datasets, and the observed values from the actual data, are shown in Table 5. In general, the observed and expected values are similar; however, one consistent pattern of departure is that the data from the New World, for most loci, show more variation by this measure than were found in the simulated data.
Table 5 Contrasting Observed and Expected Levels of Variation
Shown, both within and between populations, are the values of the average number of differences between pairs of sequences.
Exp, expected; Obs, observed
Discussion
The method described is one of several new approaches that can glean information about ancestral population sizes [30,45–47]. By including a new parameter for population splitting, it is possible to generate estimates not only of the size of the ancestral population, but also of the founding size of each founder population.
Taken together, the analyses in this study suggest a recent founding of the New World Amerind-speaking peoples by a small population of effective size near 70, followed by population growth in the New World. It is interesting that the analyses do not suggest much population size change in Asia since the time of the founding of the New World population. Given the very broad distributions for θ1, it is possible that the true value of this parameter is much higher than suggested by the peak location, and that there has been considerable population growth in Asia. The analyses reveal very broad distributions for migration parameters, and although the peak locations suggest that gene flow has been fairly high (2Nm values greater than 1; see Table 3), the estimated probabilities of migration rates having been zero are also high (Figure 3G and 3H). Also, because Eskimo-Aleut and Na Déne speakers were not included in this study, the question of separate migrations for these groups has not been addressed [3].
As parameter-rich as the method is, neither this nor any mathematical model can be expected to fully represent the complex history of two related populations. However, the same is essentially true of narrative models, as investigators are always constrained by limited data and the need to keep explanations as simple as possible given their data. In this light, the IM model provides a fairly complete framework for some oft-debated questions on human history. With the addition of a new parameter, the IM framework can now also be used to address questions about the founding size of populations and of population size change.
In the context of human demographic history, the most problematic assumption under the IM model is that each population is panmictic. Certainly this is not the case today, and it is likely to have even been less true in times past. This raises the general and important question of how local patterns of population structure affect regional or global estimates of diversity [44,48,49]. Although this question cannot be answered here, the analyses do suggest that some kinds of departures from panmixia have not occurred. For example, if the New World had been founded by a local population that had long been separated from other Asian populations, then the estimate of t would be expected to reflect this older population structure, rather than the founding of the New World. Our generally low estimates of t argue against this scenario. Similarly, if the sampled Asian populations had been highly structured, with many long-separated local populations, then this would have inflated the estimates of NA and N1, respectively. However, the generally low estimates of effective population size argue against this particular kind of population structure.
The analyses presented here share with some other genetic studies estimated dates for the peopling of the Americas that are more recent than archeologically based estimates [8,9,16]. However, the difficulty of estimating such recent events using genetic data alone should not be overestimated [18]. When considering human populations within the past few tens of thousands of years, two gene copies that share the same haplotype will often have had a common ancestor far longer ago than any of the dates in question. Similarly, genetic evidence on the peopling of the Americas has been interpreted both as consistent with multiple migrations [12] and as indicating just a single founder event [16,19,50]. Divergent interpretations are understandable, given that a finding of two populations that share sequence haplotypes at a locus can be taken as evidence of two quite different models: (1) a recent population separation; or (2) gene exchange between populations.
The available data do not yet allow precise estimates of founding time nor of whether there has been gene flow between the New World and Asia following the initial founding event. However, the new method implements a parameter-rich model of divergence and has the potential to recover the history of complex divergence processes. The method can also be applied to a large number of loci, with large sample sizes, and in the future can be expected to provide increasingly detailed portraits of human population divergence.
Materials and Methods
Selected loci and samples
Given the prevailing model of the founding of New World populations via a Bering land bridge, the descendant populations were defined as the Amerind speakers of the New World and the peoples of northeastern Asia. Greenberg et al. [3] proposed that New World populations include three linguistic groups (Eskimo-Aleut, Na Déne, and Amerind), each associated with a separate episode or period of migration. Because of the limited number of published comparative DNA sequence studies that include samples from Eskimo-Aleut and Na Déne group, the present study was limited to samples from Amerind-speaking populations. Asian samples were limited to those from China, Mongolia, Korea, and Siberia. These are partly arbitrary boundaries selected as a balance between the need to include as many loci as possible and uncertainty about the present locations of descendants of those Asian populations that gave rise to the founders of the New World.
The model fitting requires data from loci that do not show evidence of recombination and that do not show clear evidence of directional or balancing selection. All available datasets from the literature that met these criteria and that had multiple DNA sequences from both of the designated sample regions were selected. The selected loci are listed in Table 1. The input data file is provided in Dataset S1, and a list of sample locations is provided in Protocol S1.
Model development.
At the center of the method for estimating the parameters is an expression for the posterior probability distribution of model parameters Θ, given the data. For the case of multiple loci
where Θ refers to the vector of parameters of the model, Xi refers to the data for locus i, and G
i is the genealogy for locus i [33]. With n loci, the full set of parameters includes six or seven demographic parameters, depending on the inclusion of s, as well as n locus-specific mutation rate scalars [33]. A genealogy includes the topology of an ultrametric tree, the associated coalescence times, and the times of migrations on each branch of the tree [30]. For a given locus i, the probability f(X
i|G
i) is calculated using the mutation model for that locus and the branch lengths in the genealogy. The probability f(G
i|Θ) is calculated using expressions from basic coalescent theory [30,51–55]. By integrating over all possible genealogies that are consistent with the data, the results obtained are not conditioned on any particular estimate of the genealogy, and they necessarily incorporate all of the stochastic variance that arises among independent loci under the model.
The integration in Equation 1 cannot be solved directly for any but the simplest of models, but it can be approximated using a Markov chain simulation [56]. This approach was originally applied to the IM model by Nielsen and Wakeley [30], and then augmented to include multiple loci [33] and additional mutation models [32,57].
Over the course of a simulation the genealogy for a given locus varies for topology, branch lengths, and migration times. However, the probability of the data for a locus given a particular genealogy, f(X
i|G
i), depends only upon the branch lengths and the mutation model for that locus [30]. Although inclusion of s will affect the genealogies that arise in the course of the simulation, there will be no effect on the calculation of the probability of the data for a given genealogy (i.e., f(X
i|G
i) is not a function of s), and thus including s has no effect on the applicability of the method to diverse mutation models. In contrast, the probability of a genealogy given a set of parameter values, f(G
i|Θ), depends strongly on s because the probability of individual coalescent and migration times are functions of population size.
The calculation of f(G
i|Θ) is most directly done by taking the product of the probabilities of each of the coalescent and migration events that occur in the genealogy. Griffiths and Tavare [55] developed the general theory for the probability distribution of coalescence times when the population size is changing. Given a function v(τ)= Nτ/N0 of the population size at time τ, relative to that at time 0, they provide a general expression for the distribution of coalescent times. For population 1, the effective size goes from N1 at time zero, to sNA at time t. If it is assumed that the size change is exponential over this period, then for population 1,
and for population 2,
One additional complication that arises is that when the population is growing exponentially back into the past (decreasing in size as time moves forward), there is a finite probability that the time to coalescence will be infinity [58]. Thus, for population 1 when sNA is less than N1, it is necessary to calculate the probability of coalescence time conditioned on there being a coalescent event.
Migration under an exponentially changing population size can also be incorporated under this same framework with two changes. First, unlike coalescence, where the rate is inversely proportional to population size, the rate of migration is directly proportional to population size. Second, as time goes backward in the coalescent, the migration rate from population 1 to population 2 (i.e., m
1) actually corresponds to the movement of genes from population 2 to population 1 as time moves forward. This means that in the coalescent under changing population size, we expect that the migration rate from population 1 to 2 will vary with the size of population 2. Thus the corresponding relative rate function for migration from population 2 to population 1 is
and for migration in the reverse direction it is
These intensity functions for coalescence and migration were used to develop an expression for f(G
i|Θ) that includes s, and that could be directly incorporated into the update criteria for all of the demographic, mutation, and inheritance scalars described in Hey and Nielsen [33]. Also needed, in order to allow for changing population size, are the update criteria for s and the update criteria for the genealogies. For s, updates are drawn from a uniform distribution over the user-specified prior range (e.g., in the current study, an interval within the range of 0.5 to 1). An update from s to s* will affect the probability of all genealogies and thus has an acceptance probability, under the Metropolis Hastings criterion, of
where n is the number of loci and G
i is the current genealogy for locus i (see Equation 3 in Hey and Nielsen [33]). If we assume a uniform prior distribution for s, such that the prior probability of s, f(s), is constant for all s, and if we choose updates such that the q(s* → s) = q(s → s*) [30], then this simplifies to
For genealogy updates the same proposal distribution of genealogies that was used in the case without s was retained, and then this proposal distribution was incorporated into the update criteria [59]. If f(G
i|s) denotes the probability of the genealogy for locus i, given the other parameters including s, and f(G
i) is the Hastings term for the proposal probability of the genealogy for locus i, given the other parameters1 excluding s, then the update criteria for the genealogy for locus i is
Performance.
The IM computer program [33] was modified to include the additional parameter. The program is available from http://lifesci.rutgers.edu/~heylab/HeylabSoftware.htm#IM. For the Markov chain simulation that is implemented by the program, it is difficult to assess how well the method works, because of the need to generate large numbers of simulated datasets and because of the long run times required [33]. To conduct testing, a program was written to generate simulated datasets under the models in Figure 1. Datasets were simulated in groups of 10 or 20, each having 10–20 loci, for a given set of parameter values, and for a range of parameter values. Figure 4 shows the marginal posterior densities estimated from each of 20 independent simulations for a case of modest population growth with the following parameter values. θ1 = 10; θ2, = 10; θA = 10; t = 2.5; s = 0.2; m
1 = 0.04; and m
2 = 0.1. For each parameter, the mean of the 20 estimates is shown, and in general these are fairly close to the true value, though there is considerable variance for the peak locations in individual runs. To test whether the locations of these distributions are consistent with the true values of the parameters (i.e., the values used in the simulations), probabilities were combined by treating each simulation as an independent test of the same hypothesis [60]. For each posterior density pi, i = 1,…,20, is the chance that a parameter value is more extreme (i.e., departs more from the mean of the distribution) than is the actual true value. That is, if x is the area of the curve to the left of the true value then pi = 2x if x < 0.5 and pi = 2(1 − x) if x > 0.5. If the pi's are uniformly distributed, then the quantity
Figure 4 The Marginal Densities Obtained by Fitting the Model with Population Size Change to Simulated Data
The input parameters for the simulations were as follows: (A) θ1 = 10; (B) θ2 = 10; (C) θA = 10; (D) t =2.5, (E) s = 0.2, (F) m
1= 0.04; (G) m
2= 0.2 ; and t = 5 (t/2NA = 0.5). For each simulated dataset, coalescent simulations were done for each of 20 loci with identical mutation rates under an infinite sites mutation model, each with sample sizes of 10 for each of the two populations. Each simulated dataset was analyzed using wide uniform prior distributions for each parameter. Each analysis began with a burn-in period of 300,000 steps followed by a primary chain of 3 million to 10 million steps. The curves for parameters θ1 through m2 are shown in (A) through (G), respectively. For each figure, the true parameter value used in the simulations is shown as a black vertical bar, and the mean of the estimates for the 20 simulations (based on peak locations) is shown as a gray vertical bar.
is χ2 distributed with 40 degrees of freedom (i.e., two times the number of densities). The z values were as follows: θ1, 35.5; θ2, 26.4; θA, 41.7; t, 41.1; s, 26.4; m
1, 29.9; m
2, 44.1; and the mean of the seven values was 35.0. In the corresponding χ2 distribution, 90% of the probability mass falls above 29.05; 50% falls above 39.3; and 10% falls above 51.8 [61]. Clearly these values are not entirely independent of each other, but they all fall in the middle of the χ2 distribution with a mean (35.0) close to the 50% point of the χ2 distribution (39.3).
From these simulations, and many others (additional results provided in Protocol S1), it is clear that sample sizes do need to be large for the posterior distributions to be informative. With data from fewer than five loci or fewer than ten individuals per population per locus, it is often the case that distributions are very flat or that there are multiple peaks. There is a tradeoff in sampling effort required for different kinds of histories. When t is small, sampling effort should be shifted to larger sample sizes per locus, whereas when t is large, sampling effort should be shifted toward more loci. This tradeoff is a byproduct of the fact that the stochastic variance among loci, that is associated with coalescent and migration events in genealogies at times near t, goes up as t increases. Another tradeoff that arises is between s and the migration rate parameters. Just as the frequency of polymorphic sites can be used to estimate changes in population size [62], it can also be appreciated that the information for s must reside in the distribution of times of node intervals in the descendant populations. Migration can have dramatic effects on node interval times within populations. In practice, via simulation, the method does not resolve a sharp peak for s for populations that have had more than moderate amounts of migration (e.g., 2Nm values are greater than 0.5; see Protocol S1).
Analyses
Each of the three analyses were done using at least three independent runs, with ten or more independent chains under Metropolis coupling [33] as described by Geyer [63]. Each chain was initiated with a burn-in period of 100,000 updates, and the total run length of each analysis was between 10 million and 30 million updates. The mixing properties of individual runs were monitored by measuring the autocorrelation of individual parameters over the course of the run, and by estimating the effective sample size for each of the parameters as a function of the autocorrelation estimates (see p. 499 in [64]). Analyses were taken to have converged upon the stationary distribution if independent runs generated similar distributions, with each having a lowest effective sample size of 50 for the time parameter (the parameter to show the slowest rate of mixing).
To convert estimates of parameters that include the mutation rate to more easily interpreted units, a value of 6 million y since the splitting of human and chimpanzee lineages was used [65–69]. The geometric mean of the human-chimpanzee DNA sequence divergence of each locus, except ATM (see Table 2), was calculated and then used as a molecular clock calibration for converting the estimate of the time parameter, t, to divergence in years. The geometric mean mutation rate across these loci was estimated to be 4.66 × 10−6 mutations per year. The geometric mean is used rather than an arithmetic mean, because under the multilocus model, the mutation rate by which demographic parameters are scaled is the geometric mean of the individual locus-specific mutation rates [33].
To convert the estimates of the population mutation rate parameters (θ1, θ2, and θA) to estimates of effective population size (N1, N2, and NA, respectively) a measure of mutation rate on a scale of generations is needed. Thus, an assumption was made of 20 y per generation, and the geometric mean divergence between humans and chimpanzees for each species contrast was divided by 12 million y then multiplied by 20 y per generation. These calculations yielded a geometric mean value of 9.32 × 10−5 mutations per generation. These mutation rate values were then used to convert individual θ estimates to effective population size estimates (i.e., θ = 4Nu, and N = θ/4u).
Migration parameters in the model can be used to obtain population migration rate estimates (i.e., M = 2Nm, the product of the effective number of gene copies and the per gene copy migration rate) using an estimate of the population mutation rate (θ = 4Nu). Thus θ × m/2 = (4Nu × m/u)/2 = 2Nm [32].
Supporting Information
Dataset S1 Peopling of Americas Data File: Nine Loci
This is the input file that contains all of the data and that was analyzed using the IM computer program.
(582 KB TXT).
Click here for additional data file.
Protocol S1 Additional Simulations and List of Sample Locations
(92 KB DOC).
Click here for additional data file.
John Wakeley, Tad Schurr, and David Meltzer provided input on an early draft of the paper. Rasmus Nielsen provided some helpful suggestions on parameter updating. Thanks also to three reviewers for very helpful suggestions and critique.
Competing interests. The author has declared that no competing interests exist.
Author contributions. JH conceived and designed the model and analyses, selected the datasets, wrote the computer programs, performed the analyses, and wrote the paper.
Citation: Hey J (2005) On the number of New World founders: A population genetic portrait of the peopling of the Americas. PLoS Biol 3(6): e193.
Abbreviation
IMisolation with migration
==== Refs
References
Brace CL Nelson AR Seguchi N Oe H Sering L Old World sources of the first New World human inhabitants: A comparative craniofacial view Proc Natl Acad Sci U S A 2001 98 10017 10022 11481450
Goebel T Pleistocene human colonization of Siberia and peopling of the Americas: An ecological approach Evol Anthropol 1999 8 208 227
Greenberg JH Turner CG Zegura SL The settlement of the Americas: A comparison of the linguistic, dental and genetic evidence Curr Anthropol 1986 27 477 497
Wallace DC Garrison K Knowler WC Dramatic founder effects in Amerindian mitochondrial DNAs Am J Phys Anthropol 1985 68 149 155 2998196
Goebel T Waters MR Dikova M The archaeology of Ushki Lake, Kamchatka, and the Pleistocene peopling of the Americas Science 2003 301 501 505 12881567
Nichols J Linguistic diversity and the first settlement of the New World Language 1990 66 475 521
Nettle D Linguistic diversity of the Americas can be reconciled with a recent colonization Proc Natl Acad Sci U S A 1999 96 3325 3329 10077683
Starikovskaya YB Sukernik RI Schurr TG Kogelnik AM Wallace DC mtDNA diversity in Chukchi and Siberian Eskimos: Implications for the genetic history of Ancient Beringia and the peopling of the New World Am J Hum Genet 1998 63 1473 1491 9792876
Lell JT Brown MD Schurr TG Sukernik RI Starikovskaya YB Y chromosome polymorphisms in native American and Siberian populations: Identification of native American Y chromosome haplotypes Hum Genet 1997 100 536 543 9341867
Fiedel SJ The peopling of the New World: Present evidence, new theories, and future directions J Archaeol Res 2000 8 39 103
Tarazona-Santos E Santos FR The peopling of the americas: A second major migration? Am J Hum Genet 2002 70 1377 1380 11951181
Lell JT Sukernik RI Starikovskaya YB Su B Jin L The dual origin and Siberian affinities of Native American Y chromosomes Am J Hum Genet 2002 70 192 206 11731934
Schurr TG Sherry ST Mitochondrial DNA and Y chromosome diversity and the peopling of the Americas: Evolutionary and demographic evidence Am J Human Biol 2004 16 420 439 15214060
Gonzalez-Jose R Dahinten SL Luis MA Hernandez M Pucciarelli HM Craniometric variation and the settlement of the Americas: Testing hypotheses by means of R-matrix and matrix correlation analyses Am J Phys Anthropol 2001 116 154 165 11590587
Taylor RE Haynes CV Stuiver M Clovis and Folsom age estimates: Stratigraphic context and radiocarbon calibration Antiquity 1996 70 515 525
Bonatto SL Salzano FM Diversity and age of the four major mtDNA haplogroups, and their implications for the peopling of the New World Am J Hum Genet 1997 61 1413 1423 9399887
Merriwether DA Clark AG Ballinger SW Schurr TG Soodyall H The structure of human mitochondrial DNA variation J Mol Evol 1991 33 543 555 1685753
Malhi RS Eshleman JA Greenberg JA Weiss DA Schultz Shook BA The structure of diversity within New World mitochondrial DNA haplogroups: Implications for the prehistory of North America Am J Hum Genet 2002 70 905 919 11845406
Silva WA Bonatto SL Holanda AJ Ribeiro-Dos-Santos AK Paixao BM Mitochondrial genome diversity of Native Americans supports a single early entry of founder populations into America Am J Hum Genet 2002 71 187 192 12022039
Takahata N Nei M Gene genealogy and variance of interpopulational nucleotide differences Genetics 1985 110 325 344 4007484
Hudson RR Futuyma D Antonovics J Gene genealogies and the coalescent process Oxford Surveys in Evolutionary Biology 1990 New York Oxford University Press 1 44
Hudson RR Turelli M Stochasticity overrules the “three-times rule”: Genetic drift, genetic draft, and coalescence times for nuclear loci versus mitochondrial DNA Evolution 2003 57 182 190 12643581
Knowles LL Maddison WP Statistical phylogeography Mol Ecol 2002 11 2623 2635 12453245
Templeton AR Statistical phylogeography: Methods of evaluating and minimizing inference errors Mol Ecol 2004 13 789 809 15012756
Hey J Machado CA The study of structured populations—New hope for a difficult and divided science Nat Rev Genet 2003 4 535 543 12838345
Felsenstein J Estimating effective population size from samples of sequences: A bootstrap Monte Carlo integration method Genet Res 1992 60 209 220 1286805
Kuhner MK Yamato J Felsenstein J Estimating effective population size and mutation rate from sequence data using Metropolis-Hastings sampling Genetics 1995 140 1421 1430 7498781
Griffiths RC Tavare S Ancestral inference in population genetics Stat Sci 1994 9 307 319
Beerli P Felsenstein J Maximum-likelihood estimation of migration rates and effective population numbers in two populations using a coalescent approach Genetics 1999 152 763 773 10353916
Nielsen R Wakeley J Distinguishing migration from isolation. A Markov chain Monte Carlo approach Genetics 2001 158 885 896 11404349
Wakeley J Hey J DeSalle R Schierwater B Testing speciation models with DNA sequence data Molecular Approaches to Ecology and Evolution 1998 Basel Birkhäuser Verlag 157 175
Hey J Won Y-J Sivasundar A Nielsen R Markert JA Using nuclear haplotypes with microsatellites to study gene flow between recently separated Cichlid species Mol Ecol 2004 13 909 919 15012765
Hey J Nielsen R Multilocus methods for estimating population sizes, migration rates and divergence time, with applications to the divergence of Drosophila pseudoobscura and D. persimilis
Genetics 2004 167 747 760 15238526
Pritchard JK Seielstad MT Perez-Lezaun A Feldman MW Population growth of human Y chromosomes: A study of Y chromosome microsatellites Mol Biol Evol 1999 16 1791 1798 10605120
Harpending HC Batzer MA Gurven M Jorde LB Rogers AR Genetic traces of ancient demography Proc Natl Acad Sci U S A 1998 95 1961 1967 9465125
Sherry ST Rogers AR Harpending H Soodyall H Jenkins T Mismatch distributions of mtDNA reveal recent human population expansions Hum Biol 1994 66 761 775 8001908
Watkins WS Ricker CE Bamshad MJ Carroll ML Nguyen SV Patterns of ancestral human diversity: An analysis of alu-insertion and restriction-site polymorphisms Am J Hum Genet 2001 68 738 752 11179020
Hudson RR Kaplan NL Statistical properties of the number of recombination events in the history of a sample of DNA sequences Genetics 1985 111 147 164 4029609
Hudson RR Kreitman M Aguadé M A test of neutral molecular evolution based on nucleotide data Genetics 1987 116 153 159 3110004
Meltzer DJ Clocking the first Americans Annu Rev Anthropol 1995 24 21 45
Anderson DG Faught MK Palaeoindian artefact distributions: Evidence and implications Antiquity 2000 74 507 513
Dillehay TD editor Monte Verde: A late Pleistocene settlement in Chile. Vol 2: The archaeological context and interpretation 1996 Washington, D C Smithsonian Insititution Press 1071
Meltzer DJ Monte Verde and the Pleistocene peopling of the Americas Science 1997 276 754 755
Tishkoff SA Verrelli BC Patterns of human genetic diversity: Implications for human evolutionary history and disease Annu Rev Genomics Hum Genet 2003 4 293 340 14527305
Rannala B Yang Z Bayes estimation of species divergence times and ancestral population sizes using DNA sequences from multiple loci Genetics 2003 164 1645 1656 12930768
Rosenberg NA Feldman MW Slatkin M Veuille M The relationship between coalescence times and population divergence times Modern developments in theoretical population genetics 2002 Oxford Oxford University Press 130 164
Takahata N Lee SH Satta Y Testing multiregionality of modern human origins Mol Biol Evol 2001 18 172 183 11158376
Hammer MF Blackmer F Garrigan D Nachman MW Wilder JA Human population structure and its effects on sampling y chromosome sequence variation Genetics 2003 164 1495 1509 12930755
Zietkiewicz E Yotova V Gehl D Wambach T Arrieta I Haplotypes in the dystrophin DNA segment point to a mosaic origin of modern human diversity Am J Hum Genet 2003 73 994 1015 14513410
Merriwether DA Rothhammer F Ferrell RE Distribution of the four founding lineage haplotypes in Native Americans suggests a single wave of migration for the New World Am J Phys Anthropol 1995 98 411 430 8599378
Kingman JFC The coalescent Stochastic Processes Appl 1982 13 235 248
Kingman JFC On the genealogy of large populations J Appl Probab 19A 1982 27 43
Tavare S Line-of-Descent and genealogical processes, and their applications in population genetics models Theor Popul Biol 1984 26 119 164 6505980
Hudson RR Properties of a neutral allele model with intragenic recombination Theor Popul Biol 1983 23 183 201 6612631
Griffiths RC Tavare S Sampling theory for neutral alleles in a varying environment Philos Trans R Soc Lond B Biol Sci 1994 344 403 410 7800710
Gilks WR Richardson S Spiegelhalter DJ Markov chain Monte Carlo in practice 1996 Boca Raton, FL Chapman and Hall 486
Palsbøll PJ Berube M Aguilar A Notarbartolo-Di-Sciara G Nielsen R Discerning between recurrent gene flow and recent divergence under a finite-site mutation model applied to North Atlantic and Mediterranean Sea fin whale (Balaenoptera physalus) populations Evolution Int J Org Evolution 2004 58 670 675
Kuhner MK Yamato J Felsenstein J Maximum likelihood estimation of population growth rates based on the coalescent Genetics 1998 149 429 434 9584114
Hastings WK Monte Carlo sampling methods using Markov chains and their applications Biometrika 1970 57 97 109
Fisher RA Statistical methods for research workers 1954 Edinburgh Oliver and Boyd 360
Rohlf FJ Sokal RR Statistical tables 1981 San Francisco W. H. Freeman and Company 192
Tajima F The effect of change in population size on DNA polymorphism Genetics 1989 123 597 601 2599369
Geyer CJ Keramidas EM Markov chain Monte Carlo maximum likelihood Computing science and statistics. Proceedings of the 23rd Symposium on the Interface; April 21–24, 1991 1991 Seattle, Washington Interface Foundation of North America 156 163
Robert CP Casella G Monte Carlo statistical methods 2004 New York, New York Springer 645
Glazko GV Nei M Estimation of divergence times for major lineages of primate species Mol Biol Evol 2003 20 424 434 12644563
Vignaud P Duringer P Mackaye HT Likius A Blondel C Geology and palaeontology of the Upper Miocene Toros-Menalla hominid locality, Chad Nature 2002 418 152 155 12110881
Brunet M Guy F Pilbeam D Mackaye HT Likius A A new hominid from the Upper Miocene of Chad, Central Africa Nature 2002 418 145 151 12110880
Wildman DE Uddin M Liu G Grossman LI Goodman M Implications of natural selection in shaping 99.4% nonsynonymous DNA identity between humans and chimpanzees: Enlarging genus Homo
Proc Natl Acad Sci U S A 2003 100 7181 7188 12766228
Chen F-C Li W-H Genomic divergences between humans and other hominoids and the effective population size of the common ancestor of humans and chimpanzees Am J Hum Genet 2001 68 444 456 11170892
Wright S The genetical structure of populations Ann Eugenics 1951 15 323 354
Hudson RR Slatkin M Maddison WP Estimation of levels of gene flow from DNA sequence data Genetics 1992 132 583 589 1427045
Harding RM Fullerton SM Griffiths RC Bond J Cox MJ Archaic African and Asian lineages in the genetic ancestry of modern humans Am J Hum Genet 1997 60 772 789 9106523
Chakravarti A Buetow KH Antonarakis SE Waber PG Boehm CD Nonuniform recombination within the human β-globin gene cluster Am J Hum Genet 1984 36 1239 1258 6097112
Savatier P Trabuchet G Faure C Chebloune Y Gouy M Evolution of the primate beta-globin gene region. High rate of variation in CpG dinucleotides and in short repeated sequences between man and chimpanzee J Mol Biol 1985 182 21 29 3999143
Mishmar D Ruiz-Pesini E Golik P Macaulay V Clark AG Natural selection shaped regional mtDNA variation in humans Proc Natl Acad Sci U S A 2003 100 171 176 12509511
Ingman M Kaessmann H Paabo S Gyllensten U Mitochondrial genome variation and the origin of modern humans Nature 2000 408 708 713 11130070
Shen P Wang F Underhill PA Franco C Yang WH Population genetic implications from sequence variation in four Y chromosome genes Proc Natl Acad Sci U S A 2000 97 7354 7359 10861003
Kaessmann H Heissig F von Haeseler A Paabo S DNA sequence variation in a non-coding region of low recombination on the human X chromosome Nat Genet 1999 22 78 81 10319866
Kaessmann H Wiebe V Paabo S Extensive nuclear DNA sequence diversity among chimpanzees Science 1999 286 1159 1162 10550054
Jaruzelska J Zietkiewicz E Batzer M Cole DE Moisan JP Spatial and temporal distribution of the neutral polymorphisms in the last ZFX intron. Analysis of the haplotype structure and genealogy Genetics 1999 152 1091 1101 10388827
Jaruzelska J Zietkiewicz E Labuda D Is selection responsible for the low level of variation in the last intron of the ZFY locus? Mol Biol Evol 1999 16 1633 1640 10555294
Thorstenson YR Shen P Tusher VG Wayne TL Davis RW Global analysis of ATM polymorphism reveals significant functional constraint Am J Hum Genet 2001 69 396 412 11443540
Hammer MF Garrigan D Wood E Wilder JA Mobasher Z Heterogeneous patterns of variation among multiple human X-linked loci: The possible role of diversity-reducing selection in non-Africans Genetics 2004 167 1841 1853 15342522
| 15898833 | PMC1131883 | CC BY | 2021-01-05 08:21:23 | no | PLoS Biol. 2005 Jun 24; 3(6):e193 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030193 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030212SynopsisBioinformatics/Computational BiologyEvolutionGenetics/Genomics/Gene TherapyImmunologyMolecular Biology/Structural BiologyNoneUncovering the Ancient Source of Immune System Variety Synopsis6 2005 24 5 2005 24 5 2005 3 6 e212Copyright: © 2005 Public Library of Science.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.
RAG1 Core and V(D)J Recombination Signal Sequences Were Derived from Transib Transposons
==== Body
Animals with adaptive immunity have a secret for deaing with foreign invaders like viruses and bacteria—variety. Their immune systems generate a diverse array of receptors to detect the enormous number of components (antigens) that make up an invader. But with so many potential antigens, it would be difficult for the immune system to anticipate every one and thereby encode a receptor gene for each of them. Instead, the immune system employs a strategy of combinatorial diversity, recombining a few genes to give an unlimited supply of different receptors.
The portions of immune receptor genes that recombine are called V (variable), D (diversity), and J (joining) segments. The immune system randomly recombines these segments in a process called V(D)J recombination. This extraordinary reorganization is undertaken by two enzymes: RAG1 and RAG2. How this process evolved in animals is a mystery, although it has been theorized that RAG1 and RAG2 might have evolved from an ancient enzyme, called transposase, that could move or transpose gene segments. But proof of this theory for the origin of RAG's activity has remained elusive.
In a new study, Vladimir Kapitonov and Jerzy Jurka have found that RAG1 is similar to transposases encoded by transposons (jumping genes that encode transposases necessary for their mobility) found in both terrestrial and marine organisms: the fruit fly and malaria-carrying African mosquito and the sea urchin and hydra. These potentially ancient relatives of RAG1 are all called Transib transposons. The discovery of their relation to RAG1 supports the decades-old hypothesis that V(D)J recombination sprung from a transposase.
A number of different types (superfamilies) of transposons exist in nature, but no one has been able to show that RAG1 or RAG2 evolved from them. Kapitonov and Jurka took advantage of the recently discovered Transib superfamily of transposons to reexamine this problem. They used seven known Transib transposases from the fruit fly and malaria-carrying African mosquito to search the protein database GenBank, finding that part of one Transib transposase, Transib2_AG, was 35%–38% identical to part of RAG1.
This initial relationship only suggested that RAG1 might be related to Transib2_AG, since the similarity between the two was only “marginally” statistically significant, leaving the possibility that it occurred by chance. To find more statistical evidence of a relationship, Kapitonov and Jurka searched for more Transib proteins. They found a diverse family of Transib transposases in various animals, including silkworm, red flour beetle, dog hookworm, soybean rust, and hydra. The authors also found that plants and vertebrates appear not to contain Transib proteins.
With the new proteins in tow, Kapitonov and Jurka found that a 600-amino-acid region of RAG1 was statistically similar to Transib transposases. This 600-amino-acid region of RAG1 forms the core region that mediates V(D)J recombination. Three important amino acids, which underlie RAG1's ability to recombine gene segments, are also conserved in Transib transposases. Furthermore, RAG1 and RAG2 are known to recombine V, D, and J segments by binding to specific signals in these genes (called recombination signal sequences), which appear to have been derived from the ends of Transib transposons. It was previously thought that both RAG1 and RAG2 likely evolved from two proteins encoded by the same transposon. However, Kapitonov and Jurka could not find any RAG2-like proteins encoded by Transib transposons. The authors therefore suggest that RAG2 appeared later in jawed vertebrates as a necessary component for the evolution of V(D)J recombination.
A gene involved in V(D)J recombination—which allows immune cells to recognize an unlimited number of antigens by reshuffling immune receptor gene segments—evolved from an ancient gene-transposing enzyme
With the use of similarity searches (using computer programs to identify comparable parts of proteins and transposons), Kapitonov and Jurka have provided support for the transposon origin of V(D)J recombination. This theory was previously up for debate, as it was possible that RAG1 and RAG2 could have independently evolved to function like transposons. But the authors suggest that “these arguments can now be put to rest,” as it appears RAG1 evolved from a transposon currently found in flies and other organisms. Future experiments on how Transib transposons work may allow further understanding into how RAG1 and RAG2 evolved and how they function in vertebrates.
| 0 | PMC1131884 | CC BY | 2021-01-05 08:21:23 | no | PLoS Biol. 2005 Jun 24; 3(6):e212 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030212 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030227SynopsisEcologyEvolutionGenetics/Genomics/Gene TherapyMedical HistoryStatisticsHomo (Human)From Few to Many: New World Founded by Surprisingly Small Population Synopsis6 2005 24 5 2005 24 5 2005 3 6 e227Copyright: © 2005 Public Library of Science.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.
On the Number of New World Founders: A Population Genetic Portrait of the Peopling of the Americas
==== Body
About 14,000 years ago—a few hundred thousand years after our putative modern forebears spread out from Africa—descendants of archaic humans crossed the Bering land bridge from Siberia to North America. Several lines of evidence support this model, but that's where the consensus ends. The details remain hotly debated, focusing mostly on which Asian population migrated, when they did it, and whether they did it more than once.
Part of the challenge in reconstructing this history stems from the dynamic nature of human populations—which experience unpredictable changes in size, composition, density, and mating patterns—and the difficulty in interpreting genetic history. To get a better picture of the range of possible scenarios, scientists are using new statistical approaches that require computer simulations. Jody Hey now extends this approach in a novel method for the study of the origins of New World populations. Along with DNA analysis and computer simulations, Hey adds a new twist to an old model to reveal how the sizes of the first New World populations have changed since they were founded. His results fall in line with archeological, genetic, and linguistic evidence, pointing to a relatively recent colonization of the Americas. But they go a step further by showing that the New World was colonized by a small population with an effective size of only about 70 individuals. (The effective size refers to the number of individuals likely to contribute genes to the next generation.)
Hey's approach addresses shortcomings in population genetic studies that rely on just one gene and that assume that population sizes have been constant over time. Studying levels of DNA sequence variation at a single genomic region, or loci, can offer insight into the history of that gene, but the stochastic nature of gene evolution means that different genes have different histories. And as simplified versions of a very complex reality, population genetics models, such as the “isolation with migration” (IM) model, that aim to capture the population dynamics during the early stages of divergence or speciation have necessary limitations. The widely used IM model, for example, assumes that a founding population splits into two descendant populations that may interbreed, and incorporates a large number of parameters. But until now the IM model has required the assumption that all of the populations were constant in size, and therefore it has not been useful for assessing how descendant populations arose or changed in size.
Data from nine different regions in the human genome chart the journey of the first immigration to the New World
Hey analyzed DNA sequences from nine loci, so that the population genetic history could be found despite the variation among genes. He also added an additional parameter to the standard IM model to incorporate changes in the size of the ancestral population and of each founder population through time. The genetic data included DNA sequences from Asian and American Indian individuals. Hey varied the parameters in his model, which included founding population size, changes in population size, time of population formation (and splitting), and gene exchange between the populations, to work out the demographic scenario that best fit the available genetic data.
His analysis suggests that only about 70 individuals left their ancestral Asian population, estimated at about 9,000 individuals, to reach America 7,000 to 14,000 years ago. Archeological evidence places the earliest American inhabitants in the New World at around 14,000 years ago. Though Hey's estimates are more recent, they also indicate a high probability at this time period. Hey did not include genetic data from Eskimo-Aleut and Na Déne speakers, so the number of migrations was not addressed. But with this new approach, researchers will be able to explore this and many other questions to fill in the details of the first American immigration.
| 0 | PMC1131885 | CC BY | 2021-01-05 08:21:23 | no | PLoS Biol. 2005 Jun 24; 3(6):e227 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030227 | oa_comm |
==== Front
Ann Gen PsychiatryAnnals of General Psychiatry1744-859XBioMed Central London 1744-859X-4-101588244810.1186/1744-859X-4-10Primary ResearchN-terminal fragment of B-type natriuretic peptide (NT-proBNP), a marker of cardiac safety during antipsychotic treatment Kropp Stefan [email protected] Argyro [email protected] Udo [email protected] Ralf [email protected] Department of Clinical Psychiatry and Psychotherapy, Hannover Medical School, 30623 Hannover, Germany2 Department of Psychiatry and Psychotherapy, Lübbecke Medical Hospital, Virchowstr. 65, 32312 Lübbecke, Germany3 Department of Clinical Chemistry, Hannover Medical School, 30623 Hannover, Germany2005 9 5 2005 4 10 10 6 3 2004 9 5 2005 Copyright © 2005 Kropp et al; licensee BioMed Central Ltd.2005Kropp 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 potential cardiotoxicity of antipsychotic drugs is well known. The N-terminal fragment of B-type natriuretic peptide (NT-proBNP) is considered to be a possible biomarker in clinical practice for the diagnosis and prognosis in patients with suspected heart failure. This pilot evaluation tests the influence of antipsychotic drugs on NT-proBNP concentration in view of the hypothesis that NT-proBNP could be used as marker for the tolerability and safety of antipsychotic medications.
Methods
On a routine basis, patient's blood samples were examined for NT-proBNP on days 0, 7 and 21 after initiation of a new antipsychotic monotherapy. All plasma samples were analysed for NT-proBNP using an electrochemiluminiscence immunoassay "ECLIA" (proBNP kit, Roche Diagnostics, Mannheim, Germany) on an Elecsys 2010 analyser.
Results
A difference was found in NT-proBNP values at day 0 between patients younger versus older than 40 years. Also women had comparatively lower NTproBNP on days 7 and 21. Smokers' levels of NT-proBNP values decreased more from day 0 to day 7.
Conclusion
Our results suggest that antipsychotic medication influences the plasma concentration of NT-proBNP, suggesting a possible method to identify high-risk-patients for cardiovascular adverse effects due to antipsychotic medication. Larger studies should further test this hypothesis.
==== Body
Background
The potential cardiotoxicity of antipsychotic drugs has been recognized since the 1960s [1]. The most known of these cardiological side effects is the QT-prolongation in the electrocardiogram (ECG), which predisposes to a life-threatening ventricular arrhythmia known as Torsades de Pointes (TdP) and sudden death. Other cardiac adverse effects related to antipsychotic medication such as myocarditis and cardiomyopathy with sometimes fatal effect have been recently reported [2]. Brain or B-type natriuretic peptide (BNP) belongs to a family of vasoactive peptides and is primarily synthesized by the ventricular myocardium [3]. It acts as a key regulator in the homeostasis of water and salt excretion and in the maintenance of blood pressure [4] mainly by inhibiting the renin-angiotensin-aldosteron-axis and blocking the cardiac sympathetic nervous activity [4,5]. Its synthesis and secretion as proBNP is activated by myocyte stretch [6]. In this process it is split into physiologically active BNP and the N-terminal fragment NT-proBNP. Both are considered to be valuable biomarkers in clinical practice for the prediction of disease state and prognosis in patients with suspected heart failure [5]. Although adequate comparisons standing shoulder to shoulder have not been done [5], amino-terminal pro-brain natriuretic peptide (NT-pro-BNP) seems to provide very similar information to BNP. It is therefore a promising alternative marker for the detection of left-ventricular dysfunction [7]. According to other authors [8], the proportional and absolute rise of NT-proBNP values above normal plasma levels in cardiac impairment (including NYHA Class I) exceeds the rise of BNP levels. This suggests that NT-proBNP may be a more accurate marker of early cardiac dysfunction than BNP.
The aim of this clinical evaluation was to test the influence of antipsychotic drugs on NT-proBNP concentration with the hypothesis that NT-proBNP could be used as marker for tolerability and safety of antipsychotic medication.
Methods
Blood samples of 36 patients, who were treated with first (FGAs) or second-generation antipsychotics (SGAs), were selected on a routine basis.
Inclusion Criteria
Patients had the diagnosis of schizophrenia, schizoaffective or affective disorder according to ICD-10 with the need of an antipsychotic treatment. Their age ranged from 18 to 66 years. Patients with a previous history of major head injuries or neurological disorders, diabetes, current or previous substance misuse and patients receiving a combination of antipsychotics were excluded from this analysis. There was no washout period in patients who were treated with other antipsychotics before.
Blood samples examination
The blood samples had to be examined for NT-proBNP during the routine laboratory tests in a three-week-pattern. Day 0 was the day of the new treatment with an antipsychotic, whether a first or a second-generation antipsychotic drug. Blood samples were scheduled for day 0, 7 and 21 for each individual patient.
Analysation technique
Venous blood was drawn in the early morning after an overnight fast and centrifuged at 2000 g for 15 minutes to remove RBCs; the obtained clear plasma fraction was stored at -20° until the time of assay. All plasma samples were analysed for NT-proBNP using an electrochemiluminiscence immunoassay "ECLIA" (proBNP kit, Roche Diagnostics, Mannheim, Germany) on an Elecsys 2010 analyser. The assay had a measuring range from 0.6 to 4130 pg/ml and a functional sensitivity of <50 pg/ml. All assays were performed blind to clinical information on the patients.
Statistical Analysis
Data were analysed using nonparametric statistics, because data were only partly normally distributed. Group comparisons were examined using the Mann-Whitney test (two-tailed) for unpaired and Wilcoxon test for paired groups. A Bonferonni correction was taken into account in case of multiple tests. We performed 39 tests in total. After a Bonferonni correction all statistical tests were considered significant at the 0,0013 (0,05/39) probability level. The SPSS 10.0 package was used throughout.
Results
Patients and treatment
The mean age of the patients was 42,3+/-15,6 years (range:19–74). There were 16 men (44,4%) and 20 women (55,6%). Seventeen patients were smokers (47,2%) and 7 (19,4%) had cardiovascular disease (5: hypertension, 1: heart failure, 1: pacemaker). The administered medication was as follows: 8 patients received FGAs (haloperidol: 3, flupentixol: 4), 5 patients amisulpride, 11 patients risperidone, 4 patients clozapine, 6 patients olanzapine and 2 patients received quetiapine.
NT-proBNP measurement
The median measured NT-proBNP value for the group of FGAs was 28,00 at day 0 (mean = 55,63 +/-67,45, range: 5–165), 13,50 at day 7 (mean = 18,13 +/-15,70, range 5–54) and 24,00 at day 21 (mean = 59,00 +/-75,92, range: 5–188). The median measured NT-proBNP value for the group of amisulpiride and risperidone was 20,50 at day 0 (mean = 31,75 +/-35,64, range: 9–151), 15,00 at day 7 (mean = 26,94 +/-40,73, range: 5–176) and 23,50 at day 21 (mean = 24,31 +/-18,94, range: 5–75). The median measured NT-proBNP value for the group of olanzapine, clozapine and quetiapine was 25,00 at day 0 (mean = 27,33 +/-17,72, range: 6–59), 28,50 at day 7 (mean = 31,25 +/-20,24, range: 6–69) and 25,50 at day 21 (mean = 55,92 +/-92,13, range: 5–335). The performance of the Mann and Whitney test showed no statistical differences in NT-proBNP values of each day between the different groups of antipsychotics.
The impact of age
The median NT-proBNP value of patients younger than 40 years was 13,50 on day 0 (mean = 17,25 +/-10,27, range: 6–40), 14,50 at day 7 (mean = 19,83 +/-17,17, range: 6–69), and 23,50 on day 21 (mean = 35,33 +/-49,44, range: 6–188). For patients older than 40 years the median NT-proBNP value was 28,50 on day 0 (mean = 44,75 +/-47,28, range: 5–165), 21,50 on day 7 (mean = 29,71 +/-34,85, range: 5–176) and 26,50 on day 21 (mean = 46,17 +/-72,50, range: 5–335).
Comparing the median NT-proBNP values between the two age groups (Mann and Whitney test) on day 0 (13,50 vs. 28,50, p = 0,032), on day 7 (14,50 vs. 21,50, p = 0,311) and on day 21 (23,50 vs. 26,50, p = 0,987), no significant differences were found (after Bonferonni correction).
In the subgroup of patients younger than 40 years old, the NT-proBNP values showed a trend to increase but the performance of the Wilcoxon test demonstrated no significant differences of the NT-proBNP values between days 0 and 7 (13,50 vs. 14,50, p = 0.894), 7 and 21 (14,50 vs. 23,50, p = 0.119) and 0 and 21 (13,50 vs. 23,50, p = 0.197).
In the subgroup of older patients there were also no significant differences (after Bonferonni correction) of NT-proBNP values between day 0 and day 7 (28, 50 vs. 21,50, p = 0,042), from day 7 to 21 (21,50 vs. 26,50, p = 0.417) and between day 0 and 21 (28,50 vs. 26,50, p = 0.533).
The impact of sex
The median NT-proBNP value of men was 14,00 on day 0 (mean = 19,63 +/-14,79, range: 5–59), 18,00 on day 7 (mean = 23,38 +/-19,48, range: 5–69), and 18,50 on day 21 (mean = 52,56 +/-89,14, range: 5–335). The median NT-proBNP value of women was 28,00 on day 0 (mean = 48,35 +/-50,26, range: 9–165), 20,00 on day 7 (mean = 28,85+/- 36,99, range: 5–176), and 27,00 on day 21 (mean = 34,55+/-37,46, range: 5–172).
Comparing the median NT-proBNP values between men and women (Mann and Whitney test) on day 0 (14,00 vs. 28,00, p = 0,018), on day 7 (18,00 vs. 20,00, p = 0,789) and on day 21 (18,50 vs. 27,00, p = 0,459), no significant differences were found (after Bonferonni correction).
The NT-proBNP value in men showed a trend to increase over time, but the performance of the Wilcoxon test revealed no significant differences of NT-proBNP value from day 0 to 7 (14,00 vs. 18,00, p = 0,506), from day 7 to 21 (18,00 vs. 18,50, p = 0,348) and from day 0 to 21 (14,00 vs. 18,50, p = 0,300).
Women showed a decrease in NT-proBNP value between day 0 and 7 (28,00 vs. 20,00, p = 0,017) but the difference was not significant (after Bonferonni correction). In the same way no significant differences were revealed in NT-proBNP values from day 7 to 21 (20,00 vs. 27,00, p = 0.396) and from day 0 to 21 (28,00 vs. 27,00, p = 0.422).
Smoking and NT-proBNP
The median NT-proBNP value of non-smokers was 30,00 at day 0 (mean = 41,05+/-44,58, range: 9–165), 26,00 at day 7 (mean = 35,74+/- 39,05, range: 6–176), and 26,00 at day 21 (mean = 53,74+/-79,46, range: 5–335). The median NT-proBNP value of smokers was 21,00 at day 0 (mean = 29,47+/-36,79, range: 5–161), 15,00 at day 7 (mean = 16,00 +/- 7,97, range: 5–31), and 16,00 at day 21 (mean = 30,06+/-43,35, range: 5–188).
Comparing the median NT-proBNP values between smoking and non-smoking patients (Mann and Whitney test) at day 0 (21,00 vs. 30,00, p = 0,271), at day 7 (15,00 vs. 26,00, p = 0,045) and at day 21 (16,00 vs. 26,00, p = 0,285) no significant differences were found (after Bonferonni correction).
The decrease of the NT-proBNP value from day 0 to 7 in smoking patients was greater (from 21,00 to 15,00, p = 0.038) than in non-smoking patients (from 30,00 to 26,00, p = 0.647) but this decrease was in neither group significant (after Bonferonni correction).
History of cardiovascular disease and NT-proBNP
The median NT-proBNP value of patients with a positive cardiovascular history (hypertension, heart failure, arrhythmias) was 67,00 on day 0 (mean = 90,57+/-66,48, range: 9–165), 39,00 on day 7 (mean = 56,29+/-56,91, range: 9–176) and 27,00 on day 21 (mean = 37,71+/-36,13, range: 5–101). The median NT-proBNP value of patients with a negative cardiovascular history was 20,00 on day 0 (mean = 22,31+/-14,42, range: 5–59), 16,00 on day 7 (mean = 19,21+/-13,05, range: 5–69) and 25,00 on day 21 (mean = 43,72+/-70,89, range: 5–335).
Patients with a positive cardiovascular history had higher NT-proBNP values in comparison to patients without cardiovascular diseases history on day 0 (67,00 vs. 20, 00, p = 0,005) at day 7 (39,00 vs. 16, 00, p = 0,026) and at day 21 (27,00 vs. 25,00, p = 0,725). The mean age was 60,7+/-8,4 and 37,7+/-13,5 years for the subgroup with positive and negative cardiovascular history respectively, which was significant different (p = 0,000). There were neither differences between the sexes (p = 0,433) nor differences in therapy (p = 0.387), nor in smoking habits (p = 0.105) between the two groups. There were no statistical differences in NT-proBNP values over the 3 weeks in neither of the two groups. In patients with a negative cardiovascular history the median NT-proBNP value on day 21 was greater than the one in the respective group at baseline measurement.
Discussion
The aim of this pilot study testing the use of NT-proBNP in clinical routine was to investigate whether antipsychotics influence NT-proBNP concentrations. This might lead to the use of NT-proBNP as a marker for the detection of high-risk patients regarding cardiovascular adverse effects in patients receiving antipsychotic drugs. No statistical differences in the NT-proBNP values were found among the different groups of antipsychotics. Patients older than 40 years had higher values in comparison to younger patients (mean = 44,75+/-47,28 vs.17,25+/-10,27, p = 0,032 at day 0, mean = 29,71+/-34,85 vs. 19,83+/-17,17 p = 0,311, on day 7, mean = 46,17+/-72,50 vs. 35,33+/-49,44, p = 0,987 at day 21). In younger patients NT-proBNP values showed a trend to increase over time. Women had higher values in comparison to men (mean = 48,35+/-50,26 vs. 19,63+/-14,79, p = 0,018 at day 0, mean = 28,85+/-36,99 vs. 23,38+/-19,48, p = 0,789 at day 7, mean = 34,55+/-37,46 vs. 52,56+/-89,14, p = 0,459 on day 21). NT-proBNP values in men showed a trend to increase over time. Non-smoking patients had higher values in comparison to smoking ones (mean = 41,05+/-44,58 vs. 29,47+/-36,79, p = 0,271 at day 0, mean = 35,74+/-39,05 vs. 16,00+/-7,97, p = 0,045 at day 7, mean = 53,74+/-79,46 vs. 30,06+/-43,35, p = 0,285 at day 21). Smoking patients showed a greater decrease of the NT-proBNP values from day 0 to day 7(mean = 29,47+/-36,79 at day 0 to mean = 16,00+/-7,97 at day 7, p = 0,038) in comparison to non-smoking ones (mean = 41,05+/-44,58 at day 0 to mean = 35,74+/-39,05 at day 7, p = 0,647). Patients with a positive cardiovascular history had higher values in comparison to patients with a negative one (mean = 90,57+/-66,48 vs. 22,31+/-14,42, p = 0,005 on day 0, mean = 56,29+/-56,92 vs. 19,21+/-13,05, p = 0,026 at day 7, mean = 37,71+/-36,13 vs. 43,72+/-70,89, p = 0,725 at day 21). These differences were reduced over time.
BNP and NT-proBNP are new cardiac markers with a number of potential applications in both the clinical diagnosis and prognostic assessment of heart failure. In early pilot studies raised concentrations of BNP (with a sensitivity of 97% and a specificity of 84% (9)) distinguished heart failure from other causes of dyspnoea more accurately than left-ventricular ejection fraction, atrial natriuretic peptide (ANP) and N-terminal ANP did. In comparison with history, clinical signs and tests a high BNP concentration was the strongest predictor of underlying heart failure [10]. In patients with dyspnoea on exercise NT-proBNP measurement showed a sensitivity of 75% with a specificity of 79% and a negative predictive value of 99% for the detection of high-grade left-ventricular pump-dysfunction [11]. Because of the high negative predictive value of the marker a high-grade left-ventricular dysfunction could be safely ruled out in symptomatic patients with normal concentrations of NT-proBNP [10]. Elevated NT-proBNP concentration has been proven to be a good prognostic marker after acute coronary syndromes or myocardial infarction as well as a marker for patients with chronic heart failure and decreased left-ventricular dysfunction [10]. This fact could facilitate the identification of patients at risk and improved care during follow-up of these patients. Interestingly a recent study has shown that not only the initial concentrations but also the follow-up measurements compared with the initial ones are of prognostic importance [10]. Furthermore Throughton et al. [12] have shown that in patients with impaired left-ventricular systolic function and established symptomatic heart failure drug treatment guided by plasma NT-proBNP concentrations reduced the total number of cardiovascular events more than a treatment guided by clinical judgment did.
Minor cardiovascular adverse effects from antipsychotic drugs are common. They include postural hypotension and tachycardia due to anticholinergic or alpha1-adrenoreceptor blockade. They may occur in the majority of patients at therapeutic dosages [13]. Among several ECG abnormalities induced by antipsychotic drugs (AV-Blocks, widening of QRS-Complexes) the QT interval prolongation is the most vital. Most of antipsychotic drugs have been associated with QT prolongation, sometimes in a dose-dependent fashion, and some have been linked (with varying levels of confidence) to TdP and sudden death [14]. At the same time not only antipsychotics but also other psychotropic drugs such as tricyclic and tetracyclic antidepressants can cause prolongation of the QT interval [15]. The QT interval on the ECG is the time from the onset of ventricular depolarization to completion of reporalization. The prolongation of the QT interval is associated with an increased risk of dysrhythmias, especially to mention TdP, and of sudden cardiac death. The risk of electrical heart instability can be increased in pathological myocardial tissue, as for example in myocardial hypertrophy and ischaemia and in coronary atherosclerosis. This is because of the loss of membrane integrity, which disrupts both depolarization and repolarization [16]. Heart muscle disorders such as myocarditis and cardiomyopathy have been recently reported as adverse effects of clozapine, but also of other antipsychotic drugs (i.e. risperidone, haloperidol, olanzapine, quetiapine), although these associations were much weaker than for clozapine [2]. These adverse effects, which potentially lead to a heart failure, could add to the already increased cardiovascular risk of schizophrenic patients [16], resulting in lethal effects.
Our results suggest that antipsychotic medication influences the plasma concentrations of NT-proBNP. NT-proBNP concentrations are normally higher in women and in older people [17]. This impact of age and sex on NT-proBNP plasma levels, though not significant, can be seen in the baseline measurements of our patients. Older patients had higher NT-proBNP values on day 0 in comparison to younger patients(mean = 44,75+/-47,28 vs. 17,25+/-10,27, p = 0,032). Women had higher NT-proBNP values on day 0 in comparison to men (mean = 48,35+/-50,26 vs. 19,63+/-14,79, p = 0,018). The trend of an increase of the NT-proBNP values over time in male and younger patients diminished these differences in follow up measurements one and three weeks after administering the antipsychotic medication. That could not be expected, because antipsychotics differ importantly in pharmacology and widely in chemical structure. Because of that, it is unlikely that all of them have the same effects on heart function and accordingly on NT-proBNP concentrations in all patients. Among the different groups of antipsychotics patients of the group who received clozapine showed a remarkable increase of NT-proBNP plasma levels on day 7 in contrast to the decreasing values in the other groups of antipsychotics by comparable values on day 0. This is consistent with literature about an association of clozapine with cardiomyopathy and myocarditis to a severe [18] and a greater degree as other antipsychotics [2]. Smoking patients had higher values in comparison to non-smoking ones. The decrease of NT-proBNP plasma levels in smoking patients one week after receiving antipsychotic medication was greater than the respective one of non-smoking patients (from 29,47+/-36,79 to 16,00+/-7,97, p = 0,038 vs. from 41,05+/-44,58 to 35,74+/-39,05, p = 0,647). Nicotine induces the liver enzyme system (CYP1A2), which is used for the metabolisation of several antipsychotics, resulting in lower plasma levels of these drugs. The lower plasma levels of the quickly degraded antipsychotics could cause the lower NT-proBNP plasma levels in smokers.
Our paper has certain limitations. The number of patients was small. There was no washout period for patients taken other antipsychotic drugs before the start of the evaluation and patients with other than antipsychotic co-medication or other medical illnesses, which might influence the NT-proBNP levels, were not excluded. NT-proBNP is proved to be stable in EDTA plasma for a period between 6 and 24 hours [4] or even for 3 days at room temperature or longer at 4°C [19]; whether the stability of NT-proBNP decreases when stored at -20°C for a longer time is not known.
Conclusion
Despite the limitations of this study and the non-significant results in this small sample the measurement of the NT-proBNP concentration at baseline and after the beginning of antipsychotic medication seems to be a promising method to identify patients with an increased risk of dangerous cardiovascular adverse effects due to antipsychotic medication. Studies with larger number of patients, which would also examine the clinical impact of the NT-proBNP balances on the heart dysfunction in patients treated with antipsychotics, should test the hypothesis of this evaluation.
Competing interests
The authors have obtained the NT-proBNP reagent from Roche Diagnostics.
Authors' contributions
SK conceived and designed the study and helped to draft the manuscript. AT participated in designing the study, performed the statistical analysis and drafted the manuscript. US collected and interpreted the clinical data and revised the manuscript. RL carried out the NT-proBNP tests and corrected the manuscript. All authors read and approved the final manuscript.
Acknowledgements
The authors would like to thank F. Dsiosa and K. Burfeind for their expert technical assistance. We thank Dr. Spanuth (Roche Diagnostics) for supplying us with NT-proBNP reagent.
==== Refs
Menkes DB Knight JC Cardiotoxicity and prescription of thioridazine Austral New Zealand J Psychiatry 2002 36 492 494 10.1046/j.1440-1614.2002.01045.x
Coulter DM Bate A Meyboom RHB Lindquist M Edwards IR Antipsychotic drugs and heart muscle disorder in international pharmacovigilance: Data Mining Study BMJ 2001 322 1207 1209 11358771 10.1136/bmj.322.7296.1207
Yasue H Yoshimura M Sumida H Kikuta K Kugiyama K Jougasaki M Ogawa H Okumura K Mukoyama M Nakao K Congestive Heart Failure/Hypertension/Hypertrophy: Localization and Mechanism of Secretion of B-Type Natriuretic Peptide in Comparison With Those of A-Type Natriuretic Peptide in Normal Subjects and Patients With Heart Failure Circulation 1994 90 195 203 8025996
Hammerer-Lercher A Puschendorf B Mair J Cardiac Natriuretic Peptides: New Laboratory Parameters In Heart Failure Patients Clin Lab 2001 47 265 277 11405605
de Lemos JA McGuire DK Drazner MH B-type natriuretic peptide in cardiovascular disease Lancet 2003 362 316 322 12892964 10.1016/S0140-6736(03)13976-1
Bruneau BG Piazza LA de Bold AJ BNP gene expression is specifically modulated by stretch and ET-1 in a new model of isolated rat atria Am J Physiol 1997 273 2678 2686
Hammerer-Lercher A Neubauer E Muller S Pachinger O Puschendorf B Mair J Head-to-head comparison of N-terminal pro-brain natriuretic peptide, brain natriuretic peptide and N-terminal pro-atrial natriuretic peptide in diagnosing left ventricular dysfunction Clin Chim Acta 2001 310 193 197 11498085 10.1016/S0009-8981(01)00578-2
Hunt PJ Richards AM Nicholls MG Yandle TG Doughty RN Espiner EA Immunoreactive amino-terminal pro-brain natriuretic peptide (NT-PROBNP): a new marker of cardiac impairment Clin Endocrinol (Oxf) 1997 47 287 296 9373449 10.1046/j.1365-2265.1997.2361058.x
Cowie MR Struthers AD Wood DA Coats AJS Thompson SG Pool-Wilson PA Value of natriuretic peptides in assessment of patients with possible new heart failure in primary care Lancet 1997 350 1349 1353 9365448 10.1016/S0140-6736(97)06031-5
Luchner A Holmer S Schunkert H Riegger GA Bedeutung der Herzinsuffizienzmarker BNP und NT-proBNP für die Klinik Dtsch Arztebl 2003 50 A3314 3321
Mc Donagh TA Holmer S Raymond I Dargie H Hildebrant P Luchner A NT-proBNP and the diagnosis of heart failure: a pooled analysis of three European epidemiological studies J Am Coll Cardiol 2003 41 1013 1018 12651051
Throughton RW Frampton CM Yandle TG Espiner EA Nicholls MG Richards AM Treatment of heart failure guided by plasma aminoterminal brain natriuretic (N-BNP) concentrations Lancet 2000 355 1126 1130 10791374 10.1016/S0140-6736(00)02060-2
Buckley NA Sanders P Cardiovascular adverse effects of antipsychotic drugs Drug Saf 2000 23 215 228 11005704
Taylor DM Antipsychotics and QT prolongation Acta Psychiatr Scand 2003 107 85 95 12534433 10.1034/j.1600-0447.2003.02078.x
Tan HL Hou CJY Lauer MR Sung RJ Electrophysiologic Mechanisms of the Long QT Interval Syndromes and Torsade de Pointes Ann Intern Med 1995 122 701 714 7702233
Davidson M Risk of Cardiovascular Disease and Sudden Death in Schizophrenia J Clin Psychiatry 2002 63 5 11 12088174
Raymond I Groeming BA Hildebrandt PR Nillson JC Baumann M Trawinski J Pedersen F The influence of age, sex and other variables on the plasma level of N-terminal pro brain natriuretic peptide in a large sample of the general population Heart 2003 89 745 751 12807847 10.1136/heart.89.7.745
Kilian JG Kerr K Lawrence C Celermajer DS Myocarditis and cardiomyopathy associated with clozapine Lancet 1999 354 1841 1845 10584719 10.1016/S0140-6736(99)10385-4
Yeo KT Wu AH Apple FS Kroll MH Christenson RH Lewandrowski KB Sedor FA Butch AW Multicenter evaluation of the Roche NT-proBNP assay and comparison to the Biosite Triage BNP assay Clin Chim Acta 2003 338 107 115 14637274 10.1016/j.cccn.2003.08.016
| 15882448 | PMC1131886 | CC BY | 2021-01-04 16:39:07 | no | Ann Gen Psychiatry. 2005 May 9; 4:10 | utf-8 | Ann Gen Psychiatry | 2,005 | 10.1186/1744-859X-4-10 | oa_comm |
==== Front
AIDS Res TherAIDS Research and Therapy1742-6405BioMed Central London 1742-6405-2-41587781810.1186/1742-6405-2-4ResearchVitamin supplementation for prevention of mother-to-child transmission of HIV and pre-term delivery: a systematic review of randomized trial including more than 2800 women Mills Edward J [email protected] Ping [email protected] Dugald [email protected] Gordon H [email protected] Department of Clinical Epidemiology & Biostatistics, McMaster University, Hamilton, Canada2 London School of Hygiene & Tropical Medicine, London, UK3 Division of Clinical Epidemiology, Canadian College Of Naturopathic Medicine, Toronto, Canada4 Hospital for Sick Children, University of Toronto, Toronto, Canada2005 6 5 2005 2 4 4 20 1 2005 6 5 2005 Copyright © 2005 Mills et al; licensee BioMed Central Ltd.2005Mills 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
Observational studies have suggested that low serum vitamin levels are associated with increased mother-to-child transmission (MTCT) of HIV and increased preterm delivery. We aimed to determine the efficacy of vitamins on the prevention of MTCT and preterm delivery by systematically reviewing the available randomized controlled trials [RCTs]. We conducted systematic searches of 7 electronic databases. We extracted data from the RCTs independently, in duplicate.
Results
We included 4 trials in our review. Of the three trials on Vitamin A, two suggested no difference in MTCT, while the third and largest trial (n = 1078) suggested an increased risk of MTCT (Relative Risk 1.35, 95% Confidence Interval [CI], 1.11–1.66, P = 0.009). Two of the vitamin A trials addressed the impact of supplementation on pre-term delivery; one suggested a benefit (RR 0.65, 95% CI, 0.44–0.94) and the other no difference. All three vitamin A trials found no significant effect on infant mortality at 1 year. Of the two trials that looked at multivitamin use, only one addressed the prevention of MTCT, and found a non-significant RR of 1.04 (95% CI, 0.82–1.32). Two of the multivitamin trials found no significant effects on pre-term delivery. The single multivitamin trial examining children's mortality at 1 year yielded a non-significant RR of 0.91 (95% CI, 0.17–1.17).
Conclusion
Randomized trials of vitamins to prevent MTCT have yielded conflicting results without strong evidence of benefit and have failed to exclude the possibility of harm.
HIVVitaminsVitamin AMother-to-child transmissionPreterm delivery
==== Body
Introduction
In Africa, 55% of HIV-1-positive adults are women, most of childbearing age [1]. Data from antenatal clinics show that in several parts of southern Africa, more than 30% of pregnant women are infected with HIV-1. The fastest growth has been in South Africa, where the prevalence of infection in adults increased from 5% in 1990, to over 25% in 2002 [1]. Mother-to-child transmission (MTCT) of HIV-1 can occur during pregnancy, delivery, and post-partum through breastfeeding. In observational cohort studies, the cumulative rates of transmission are between 25% and 45% of all children born to HIV-1-infected mothers in Africa compared with 10–30% in wealthier countries [1]. This difference is greatly but not totally accounted for by the risk of postnatal transmission in populations in which breastfeeding is common.
MTCT is responsible for 5–10% of the total of new HIV infections in many developing countries, with more than 500,000 children being infected each year [1]. In many industrialized countries, the introduction of antiretroviral (ARV) drugs for the prevention of MTCT has dramatically reduced rates of transmission among non-breastfeeding mothers. Improvement is evident as more women enter pregnancy while on combination ARV therapy [2,3]. The limited access to ARV's throughout Africa has, however, led to a search for cheaper alternatives. Observational studies demonstrating an association between low biochemical and dietary levels of micronutrients and MTCT have fueled the hypothesis that micro-nutrient supplementation, particularly with Vitamin A and multivitamin combinations, may reduce vertical transmission [4-9].
Vitamin supplementation may reduce vertical transmission through either intrapartum or breastfeeding routes by reducing HIV viral load in lower genital tract secretions and in breast milk, respectively [10]. Other potential therapeutic mechanisms include improved placental and lower genital tract integrity [11], and improved fetal and newborn gastrointestinal immunity [12]. Investigators have undertaken several randomized trials addressing the impact of vitamin supplementation on MTCT. In order to determine the effectiveness of these treatments in preventing MTCT and pre-term delivery, we conducted a systematic review of these randomized trials. In addition, we addressed the effect of Vitamin A and multivitamins on childhood mortality.
Methods
With the aid of an information specialist, we (EM, PW) performed a systematic, all language search of the following electronic databases independently, in duplicate: MedLine (1966- January 2005), AMED (1985- January 2005), AltHealthWatch (1990- January 2005), CinAhl (1982- January 2005), Embase (1980- January 2005), and the Cochrane Library (2004, issue 2). We supplemented this search by reviewing reference sections of relevant articles, and by searching for unpublished trials on the National Research Register (UK) (October 1998- January 2005) and Clinicaltrials.gov (February 2000- January 2005).
Selection of abstracts
Two of us (EM, PW) independently evaluated the abstracts of retrieved articles. Eligible studies met the following criteria: (1) were original randomized controlled trials examining HIV+ patients using either Vitamin A or multivitamin treatment during pregnancy; (2) examined the outcomes of MTCT or pre-term delivery. We excluded any previous analyses of the same trial in our meta-analysis and used the most recent data available [13]. Kappa scores reflected chance-adjusted inter-observer agreement in the study identification process.
Quality assessment
Pre-specified quality criteria included: methods of randomization, allocation concealment, blinding status of patients and assessors, use of placebo, informed consent, a priori sample size estimations, use of intention-to-treat, and sources of funding. In addition, we contacted the study authors for clarification of study methods.
Assessing the quality of trials included in a systematic review is important in determining trial validity, potential for introducing bias and heterogeneity and exploring subgroup analysis. Quality assessment was performed independently, in duplicate (EM, PW). Quality assessment items were used as a priori explanations of heterogeneity.
Data abstraction
We extracted data independently, in duplicate (EM, PW) [14]. Data abstraction sheets were developed and piloted amongst the group (EM, PW, DS) to determine outcomes of interest and reproducibility.
Statistical analysis
We determined agreement between reviewers using the kappa statistic. We report on study sample size and dosing using descriptive data. Outcomes measured were the number of live births, not number of pregnancies. Our primary endpoint for meta-analysis of MTCT was children's infection status at the latest time point reported. In order to provide a best-estimate of treatment effects, we conducted a meta-analysis. Our primary endpoint for the meta-analysis of pre-term delivery was pre-term delivery defined as <37 weeks. We also determined childhood mortality at 1 year where reported. All outcomes were treated as dichotomous outcomes [15] and the appropriate relative risks (RR) and applicable 95% confidence intervals [CI] were determined. We calculated RR from raw data, when provided. Pooled analysis of relative risk was conducted using a random effects model. We tested for heterogeneity using the Zalen test and the I2 test [16]. A priori explanations of heterogeneity included quality assessment items, design, and length of follow-up. StatsDirect was used for all meta-analytic procedures (StatsDirect, Copyright 1993–2004, Manchester).
Results
Figure 1 displays the yield of our systematic searches. Of 27 clinical trial abstracts that appeared relevant, we examined 16 full text articles. Four trials met inclusion criteria and are included in this systematic review. κ for initial decisions on the inclusion of studies was 0.9, suggesting near-perfect agreement. Three [17-19] studies met our inclusion criteria of examining mother-to-child transmission of HIV and three studies [13,17,20] met our inclusion for pre-term delivery. Three studies examined the role of vitamin A for prevention of MTCT [17,18] and 1 study [19] examined the role of both vitamin A and a multivitamin using a 2 × 2 factorial design. We identified 1 unpublished and unreported study [21], from which however no results could be obtained.
Figure 1 Flowchart depicting study selection and inclusion/exclusion.
Study characteristics
Details for each of the trials can be found in table 1 (additional file 1) with regards to: the intervention; standards of care for all participants; the number of mothers randomized and the gestational period in which they were enrolled; the number of live births; compliance; outcomes measured; and results. The table is split into 2 sections to reflect details with reference to vitamin A supplementation alone or a combination of multivitamins. A brief overview of the studies found is provided here.
Vitamin A
Fawzi et al published, in several analyses of the same factorial trial assessing the impact of vitamin A, and of multivitamins excluding Vitamin A, on vertical transmission of HIV-1 and child mortality on pregnant women in Dar es Salaam, Tanzania (n = 1078) [19]. HIV+ women presenting at antenatal clinics between 12 and 27 weeks of pregnancy were randomized to receive (i) vitamin A alone or matching placebo and, (ii) multivitamins excluding vitamin A or matching placebo. 985 children were born alive from the total sample with 898 having at least one specimen for HIV testing. Of these, 268 tested positive for HIV-1 at 6 weeks of age. Details for the earlier analyses in 1998 are also provided in Table 1 in the section on vitamin A.
In a parallel group randomized trial in Durban, South Africa, Coutsoudis et al randomized 728 pregnant HIV infected women to either placebo (n = 360) or vitamin A retinyl palmitate + B-carotene, with additional vitamin A at delivery(n = 368) [17]. Data on HIV infection at 3 months were available for 502 children of the total 661 live births.
Another parallel group trial by Kumwenda et al randomized 697 HIV infected pregnant women in Blantyre, Malawi to an intervention where vitamin A was added to their supplements (n = 340), or placebo (n = 357) [18]. There were a total of 622 live births (including 14 pairs of twins), however, 32 infants died to prior to 6 weeks of age, making HIV status undeterminable.
Multivitamins
As described above, the factorial trial of Fawzi et al (2002) [19] performed examined the impact of both vitamin A and multivitamins of MTCT on infant mortality. The characteristics and results from the earlier analysis by Fawzi et al (1998) are also listed in table 1 [13]. In a subgroup analysis, not listed in table 1, multivitamin supplementation reduced death and prolonged HIV-free survival in women with low maternal immunological and nutritional status (RR of death 0.30, 95% CI, 0.10–0.92).
Friis et al (2004) [20] conducted a parallel randomized trial of micronutrients versus placebo. They examined a subgroup of pregnant women with HIV infection (n = 360) enrolled between the period of 22 and 36 weeks gestation (active group n = 189, control group n = 171). The study was hampered by not examining infant HIV infections or reporting specific number of births by HIV group.
Methodological reporting
Three studies described sequence generation [13,18,20] Two reported allocation concealment [18,20]. Only 1 study described who was blinded [20]. Four studies reported obtaining informed consent [13,17,18,20]. Five studies reported an a priori sample size estimation [13,18-20,22] and 4 reported analysis by intention-to-treat [13,17,19,20]. All studies disclosed the sources of funding.
Meta-analysis
The combined RR of vitamin A for prevention of MTCT yielded a RR of 1.05 (95% CI, 0.78–1.41, p = 0.2, I2 = 75%, heterogeneity P = 0.01) (figure 2). The impressive variability in results is reflected in the largely non-overlapping confidence intervals between the two studies that suggested no difference between treatment and control, and the Fawzi study that suggested harm. Two trials examined the protection of vitamin A for pre-term delivery and yielded a non-significant pooled RR of 0.85 (95% CI, 0.53–1.37, P = 0.5, I2 = 77%, heterogeneity P = 0.03) (figure 3). Three trials examined the role of maternal vitamin A supplementation on children's mortality at 1 year. The pooled RR was 1.05 (95%CI, 0.88–1.27, P = 0.5, I2 = 0%, heterogeneity P = 0.8).
This single trial by Fawzi et al. examining a multivitamin for prevention of MTCT yielded a non-significant RR of 1.04 (95% CI, 0.82–1.32) (figure 2). The single trial examining maternal multivitamin intake on children's mortality at 1 year yielded a non-significant RR of 0.91 (95% CI, 0.17–1.17). Two trials examined the role of multivitamins for prevention of pre-term delivery. The combined RR yielded a non-significant RR of 0.88 (95% CI, 0.73–1.06, P = 0.1, I2 = 0%, heterogeneity P = 0.8) (figure 3).
Figure 2 Meta-analysis of MTCT.
Figure 3 Meta-analysis of pre-term delivery.
Discussion
The results of this review should be of interest to clinicians and policy makers alike. We found that, despite early observational studies suggesting an association between vitamin A deficiency and decreased risk of MTCT [4-9], RCTs show no such effect and actually raise the possibility of increased risk. Similarly, the single trial examining supplementation with multivitamins did not decrease MTCT. Supplementation with vitamin A or multivitamins was not associated with a reduction in childhood mortality. While one trial suggested multivitamins might decrease pre-term delivery, the results are not consistent.
There are several limitations to consider in this review. Due to the small number of studies included in each separate analysis, available methods for exploring the likelihood of publication bias are uninformative. We attempted to reduce this potential impact by systematically searching the databases, contacting authors, and searching for unpublished studies through registries. A further limitation is the impact that multiple childbirths from the same mother had on the results of our analyses [23]. This information was not provided consistently across studies or through contact with authors and, although systematic evaluations of this have shown it does not significantly confound meta-analyses, could theoretically affect the estimates of effect [23]. Finally, all of the studies compared vitamin supplementation vs. placebo. It is possible that a trial examining ARVs plus micronutrients vs ARVs alone would yield results generalizable to the current desired situation.
A strength of our meta-analysis is that we used a random effects model as this assumes a different underlying effect for each study and takes between-study variability into consideration as an additional source of variation. These effects are assumed to be randomly distributed and the central point of the distribution is the focus of the combined effect estimate. Thus, the random effects model gives greater weight to smaller studies than does the fixed effects model, and results in wider confidence intervals and a more conservative estimate of effect than the fixed effects model. This is especially warranted in this study as we identified significant heterogeneity in our pooled analysis of vitamin A on MTCT (I2 = 75%, heterogeneity P = 0.01) and pre-term delivery (I2 = 77%, heterogeneity P = 0.03).
We found large and unexplained heterogeneity between studies in the meta-analysis of vitamin A for prevention of MTCT. We were unable to explain this heterogeneity using our a priori determined explanations of heterogeneity. However, biological evidence may best explain this occurrence. MTCT is most likely to occur during the process of vaginal birth. Thus, it is important that a further investigation of the trial by Fawzi et al. demonstrated that vaginal HIV-1 viral shedding actually increased in women who were given Vitamin A supplementation but not in the case of other micronutrient supplementation (74.8% vs. 65.1%, P = 0.04) [24]. This supports the plausibility that vitamin A may contribute to an increased risk of transmission. There does not appear to be evidence demonstrating the same risk with the use of other vitamins and multivitamins may still provide some level of protection for women living with HIV [25,26].
There is also another explanation for the difference between the trial by Fawzi et al (2002) [19] and the trial in South Africa and Malawi [17,18]. In the trials that found no effect, the supplements were given during the antenatal period only, whereas in the Tanzania trial supplementation continued during the antenatal and breastfeeding periods. It may be that a longer period of supplementation on a larger n resulted in greater power to detect effects. Indeed, earlier analysis of this sample by Fawzi (2000) did not reveal this effect [22]. It is additionally possible that geographical differences exist from between Tanzania and the other countries (South Africa and Malawi). It is possible that nutritional status regarding important nutritional supplementation associated with HIV progression, such as selenium [27-29], is different in Tanzania.
Conducting trials to assess the impact of interventions on MTCT is an ethically challenging, yet politically eye-opening area. Section 29 of the Helsinki Declaration ethical principles for conducting research on human subjects states that "the benefits, risks, burdens and effectiveness of a new method should be tested against those of the best current prophylactic, diagnostic, and therapeutic methods. This does not exclude the use of placebo, or no treatment, in studies where no proven prophylactic, diagnostic or therapeutic method exists [30]" However, in many impoverished nations, supplying antiretrovirals would also result in inducement to participate, a factor that is largely considered unethical to recruitment. Were antiretroviral treatment provided to these developing nation populations, vitamins would have to be tested in the presence of antiretroviral treatments, such as single-dose nevirapine or short-dose zidovudine [2]. However, access to antiretroviral treatments in developing nations is extremely limited and although the Global Fund for AIDS, Malaria and Tuberculosis is making great strides at providing access to antiretrovirals for impoverished nations, the likelihood of effective treatment even in pregnancy is not guaranteed. The investigators of the trials reviewed here have provided evidence in a pragmatic fashion as they provide results from the population with which we would aim to generalize.
More than 95% of HIV-1-infected children acquired their infection from their mother [1]. Mother-to-child transmission is largely preventable with interventions that are accessible to resource-poor countries: prevention of sexual transmission of HIV-1 through education for women of childbearing age, especially very young women; access to HIV-1 testing and reduction of unwanted pregnancies by HIV infected women informed of their serostatus; and ARV-based prevention of mother-to-child transmission. Prevention of mother-to-child transmission is the most cost-effective antiretroviral method and one of the most attractive interventions for prevention of HIV-1. A rapid scaling-up of implementation is crucial to allow programs to prevent mother-to-child transmission to affect the burden of paediatric HIV/AIDS. Such national initiatives should build a comprehensive continuum of care, including access to ARVs, for all members of affected families.
Using vitamins as a therapy to prevent MTCT seems inadvisable given the current state of evidence indicating a lack of consistent effect in prevention of vertical transmission [2,3,31]. However, in settings where poverty and social circumstances prevent adequate nutrition, the implementation of nutritional programs for pregnant women may play a role in preventing other harmful pregnancy outcomes. Future trials assessing the impact of effective nutrition on pregnant women living with HIV are not only an important effort in stemming the epidemic and improving the quality of life of patients, but also a human rights imperative [32]. Specific trials aimed at women with low nutritional status may provide an additional armament in the fight against HIV/AIDS.
In summary, the findings from our systematic review and meta-analysis do not support the use of vitamin A supplementation as an aid in reducing the risk of mother-to-child transmission of HIV-1, and may in fact increase the risk. With respect to protection against pre-term delivery, vitamin A supplementation demonstrated a non-statistically significant protective trend. No role was found for maternal vitamin A supplementation in reducing childhood mortality at 1 year. We also found that multivitamin supplementation showed no effect on mother-to-child transmission, childhood mortality at 1 year, or prevention of pre-term delivery.
Competing interests
The author(s) declare that they have no competing interest.
Authors' contributions
Concept, protocol: EM, PW, GG
Data searching and abstraction: EM, PW, DS
Data analysis: EM, PW, GG, DS
Manuscript drafts: EM, PW, GG, DS
Approval of final manuscript: EM, PW, GG, DS
Supplementary Material
Additional File 1
Table 1. Study characteristics.
Click here for file
Acknowledgements
The authors thank Dr. Peter Brocklehurst for critical revisions.
==== Refs
UNAIDS Report on the global AIDS epidemic 2004
Brocklehurst P Interventions for reducing the risk of mother-to-child transmission of HIV infection Cochrane Database Syst Rev 2002 CD000102 11869564
Brocklehurst P Volmink J Antiretrovirals for reducing the risk of mother-to-child transmission of HIV infection Cochrane Database Syst Rev 2002 CD003510
Sherr L Preventing HIV transmission during pregnancy and delivery: a review AIDS STD Health Promot Exch 1997 4 6 12348385
Semba RD Miotti PG Chiphangwi JD Dallabetta G Yang LP Saah A Hoover D Maternal vitamin A deficiency and infant mortality in Malawi J Trop Pediatr 1998 44 232 234 9718911 10.1093/tropej/44.4.232
Semba RD Miotti PG Chiphangwi JD Liomba G Yang LP Saah AJ Dallabetta GA Hoover DR Infant mortality and maternal vitamin A deficiency during human immunodeficiency virus infection Clin Infect Dis 1995 21 966 972 8645848
Semba RD Miotti PG Chiphangwi JD Saah AJ Canner JK Dallabetta GA Hoover DR Maternal vitamin A deficiency and mother-to-child transmission of HIV-1 Lancet 1994 343 1593 1597 7911919 10.1016/S0140-6736(94)93056-2
Greenberg BL Semba RD Vink PE Farley JJ Sivapalasingam M Steketee RW Thea DM Schoenbaum EE Vitamin A deficiency and maternal-infant transmissions of HIV in two metropolitan areas in the United States Aids 1997 11 325 332 9147424 10.1097/00002030-199703110-00010
Dreyfuss ML Fawzi WW Micronutrients and vertical transmission of HIV-1 Am J Clin Nutr 2002 75 959 970 12036800
Fawzi WW Hunter DJ Vitamins in HIV disease progression and vertical transmission Epidemiology 1998 9 457 466 9647913 10.1097/00001648-199807000-00015
John GC Nduati RW Mbori-Ngacha D Overbaugh J Welch M Richardson BA Ndinya-Achola J Bwayo J Krieger J Onyango F Kreiss JK Genital shedding of human immunodeficiency virus type 1 DNA during pregnancy: association with immunosuppression, abnormal cervical or vaginal discharge, and severe vitamin A deficiency J Infect Dis 1997 175 57 62 8985196
Filteau SM Rollins NC Coutsoudis A Sullivan KR Willumsen JF Tomkins AM The effect of antenatal vitamin A and beta-carotene supplementation on gut integrity of infants of HIV-infected South African women J Pediatr Gastroenterol Nutr 2001 32 464 470 11396815 10.1097/00005176-200104000-00014
Fawzi WW Msamanga GI Spiegelman D Urassa EJ McGrath N Mwakagile D Antelman G Mbise R Herrera G Kapiga S Willett W Hunter DJ Randomised trial of effects of vitamin supplements on pregnancy outcomes and T cell counts in HIV-1-infected women in Tanzania Lancet 1998 351 1477 1482 9605804 10.1016/S0140-6736(98)04197-X
Meade MO Richardson WS Selecting and appraising studies for a systematic review Ann Intern Med 1997 127 531 537 9313021
Altman DG Systematic reviews of evaluations of prognostic variables Systematic reviews in Healthcare, Meta-analysis in context 2001 BMJ London
Higgins JP Thompson SG Quantifying heterogeneity in a meta-analysis Stat Med 2002 21 1539 1558 12111919 10.1002/sim.1186
Coutsoudis A Pillay K Spooner E Kuhn L Coovadia HM Randomized trial testing the effect of vitamin A supplementation on pregnancy outcomes and early mother-to-child HIV-1 transmission in Durban, South Africa. South African Vitamin A Study Group Aids 1999 13 1517 1524 10465076 10.1097/00002030-199908200-00012
Kumwenda N Miotti PG Taha TE Broadhead R Biggar RJ Jackson JB Melikian G Semba RD Antenatal vitamin A supplementation increases birth weight and decreases anemia among infants born to human immunodeficiency virus-infected women in Malawi Clin Infect Dis 2002 35 618 624 12173139 10.1086/342297
Fawzi WW Msamanga GI Hunter D Renjifo B Antelman G Bang H Manji K Kapiga S Mwakagile D Essex M Spiegelman D Randomized trial of vitamin supplements in relation to transmission of HIV-1 through breastfeeding and early child mortality Aids 2002 16 1935 1944 12351954 10.1097/00002030-200209270-00011
Friis H Gomo E Nyazema N Ndhlovu P Krarup H Kaestel P Michaelsen KF Effect of multimicronutrient supplementation on gestational length and birth size: a randomized, placebo-controlled, double-blind effectiveness trial in Zimbabwe Am J Clin Nutr 2004 80 178 184 15213046
Joubert G Steinberg H van der Ryst E Chikobvu P Consent for participation in the Bloemfontein vitamin A trial: how informed and voluntary? Am J Public Health 2003 93 582 584 12660201
Fawzi WW Msamanga G Hunter D Urassa E Renjifo B Mwakagile D Hertzmark E Coley J Garland M Kapiga S Antelman G Essex M Spiegelman D Randomized trial of vitamin supplements in relation to vertical transmission of HIV-1 in Tanzania J Acquir Immune Defic Syndr 2000 23 246 254 10839660
Gates S Brocklehurst P How should randomised trials including multiple pregnancies be analysed? Bjog 2004 111 213 219 14961881
Fawzi W Msamanga G Antelman G Xu C Hertzmark E Spiegelman D Hunter D Anderson D Effect of prenatal vitamin supplementation on lower-genital levels of HIV type 1 and interleukin type 1 beta at 36 weeks of gestation Clin Infect Dis 2004 38 716 722 14986257 10.1086/381673
Fawzi WW Msamanga GI Spiegelman D Wei R Kapiga S Villamor E Mwakagile D Mugusi F Hertzmark E Essex M Hunter DJ A randomized trial of multivitamin supplements and HIV disease progression and mortality N Engl J Med 2004 351 23 32 15229304 10.1056/NEJMoa040541
Fawzi W Micronutrients and human immunodeficiency virus type 1 disease progression among adults and children Clin Infect Dis 2003 37 Suppl 2 S112 6 12942384 10.1086/375882
McClelland RS Baeten JM Overbaugh J Richardson BA Mandaliya K Emery S Lavreys L Ndinya-Achola JO Bankson DD Bwayo JJ Kreiss JK Micronutrient Supplementation Increases Genital Tract Shedding of HIV-1 in Women: Results of a Randomized Trial J Acquir Immune Defic Syndr 2004 37 1657 1663 15577425
van Lettow M Harries AD Kumwenda JJ Zijlstra EE Clark TD Taha TE Semba RD Micronutrient malnutrition and wasting in adults with pulmonary tuberculosis with and without HIV co-infection in Malawi BMC Infect Dis 2004 4 61 15613232 10.1186/1471-2334-4-61
Kupka R Msamanga GI Spiegelman D Morris S Mugusi F Hunter DJ Fawzi WW Selenium status is associated with accelerated HIV disease progression among HIV-1-infected pregnant women in Tanzania J Nutr 2004 134 2556 2560 15465747
WMA Ethical Principles for Medical Research Involving Human Subjects WORLD MEDICAL ASSOCIATION DECLARATION OF HELSINKI 2004 http://www.wma.net/e/policy/b3.htm (Accessed March 25, 2005)
Brocklehurst P French R The association between maternal HIV infection and perinatal outcome: a systematic review of the literature and meta-analysis Br J Obstet Gynaecol 1998 105 836 848 9746375
Nations U Universal Declaration of Human Rights United Nations 1998 Article 25
| 15877818 | PMC1131887 | CC BY | 2021-01-04 16:38:33 | no | AIDS Res Ther. 2005 May 6; 2:4 | utf-8 | AIDS Res Ther | 2,005 | 10.1186/1742-6405-2-4 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1051585047710.1186/1471-2105-6-105Research ArticleDetailed protein sequence alignment based on Spectral Similarity Score (SSS) Gupta Kshitiz [email protected] Dina [email protected] SV [email protected] KV [email protected] S [email protected] Department of Computer Science & Engineering, Indian Institute of Technology, Bombay, Mumbai, India2 Department of Chemical Engineering, Indian Institute of Technology, Bombay, Mumbai, India3 School of Biosciences & Bioengineering, Indian Institute of Technology, Bombay, Mumbai, India4 Department of Physics, Indian Institute of Science, Bangalore, India5 Bioinformatics Center, Indian Institute of Science, Bangalore, India2005 23 4 2005 6 105 105 1 2 2005 23 4 2005 Copyright © 2005 Gupta 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 chemical property and biological function of a protein is a direct consequence of its primary structure. Several algorithms have been developed which determine alignment and similarity of primary protein sequences. However, character based similarity cannot provide insight into the structural aspects of a protein. We present a method based on spectral similarity to compare subsequences of amino acids that behave similarly but are not aligned well by considering amino acids as mere characters. This approach finds a similarity score between sequences based on any given attribute, like hydrophobicity of amino acids, on the basis of spectral information after partial conversion to the frequency domain.
Results
Distance matrices of various branches of the human kinome, that is the full complement of human kinases, were developed that matched the phylogenetic tree of the human kinome establishing the efficacy of the global alignment of the algorithm. PKCd and PKCe kinases share close biological properties and structural similarities but do not give high scores with character based alignments. Detailed comparison established close similarities between subsequences that do not have any significant character identity. We compared their known 3D structures to establish that the algorithm is able to pick subsequences that are not considered similar by character based matching algorithms but share structural similarities. Similarly many subsequences with low character identity were picked between xyna-theau and xyna-clotm F/10 xylanases. Comparison of 3D structures of the subsequences confirmed the claim of similarity in structure.
Conclusion
An algorithm is developed which is inspired by successful application of spectral similarity applied to music sequences. The method captures subsequences that do not align by traditional character based alignment tools but give rise to similar secondary and tertiary structures. The Spectral Similarity Score (SSS) is an extension to the conventional similarity methods and results indicate that it holds a strong potential for analysis of various biological sequences and structural variations in proteins.
==== Body
Background
Comparison and alignment of primary structures has become the prime tool for protein sequence analysis [1]. Comparative analysis of primary structures of amino acids can reveal useful information regarding the structure and function of proteins. Many algorithms therefore have been developed and databases designed to search for similar proteins, but most of them are based on character-matching techniques. In this technique, the amino acids are considered to be distinct characters.
However, there are certain limitations of character based similarity measure approaches that cannot allow insights in the structural aspects of the protein. Though two sequences with a high character based similarity are expected to depict similar structures and show close biological functions, the reverse is not always true. Instances have been found where structurally closely related sequences do not provide good character based similarity measure [2]. Two protein sequences with low sequential identity may show similarities in their physiochemical properties, tertiary structure and biological activities. There could be many reasons for this observation. The one most widely hypothesized is that nature sometimes retains the biological functions but changes the amino acids as the protein evolves. Also, most of the times, researchers are interested in the active site of the protein, and not its overall backbone structure. The active site may occupy just a small part of the overall protein, therefore it is important to capture the structure and local variations in properties of amino acids at a certain location. Overall similarity score based on character matching may not be able to capture the local similarities, particularly if the amino acids differ in the location but provide similar overall structure.
Many algorithms have been developed based on character based similarity, though differing in their approaches. BLAST attempts to fragment protein sequences and establishes matches between them using substitution matrices for thresholds. PSI-BLAST [3], an extension to BLAST [4,5], uses similarity matrices (called profiles) based on specificity of position of an amino acid, and is probably the most widely used sequence similarity tool. All BLAST algorithms are based on consideration of sequences as long strings of alphabets. In addition, various heuristics are employed based on biological observations as extensions to purely character based approaches. Similarly, FASTA [6] algorithms using optimized gap penalties are used to find homologous sequences from protein databases. SSearch [7] engine implements Smith-Waterman [8] algorithm, an extension to the N-W algorithm [9] for establishment of protein similarity. PRIDE [10] establishes similarity score by considering Cα - Cα distances between residues separated within a threshold of amino acids. An interesting holistic approach to protein alignment developed by Taylor and Orengo [2] present an algorithm that considers structural aspects inducing hydrogen bonding like solvent exposure, torsion angle apart from the traditional character based methods, and does indeed presents appreciable alignments for proteins with low sequence similarity. Tonges et al. [11] presents a general method for sequence alignment based on conventional dynamic programming and building of secondary matrices by their results. However, it works best for highly homologous sequences and therefore is of not much use for less homologous sequences. Double dynamic programming approach, an interesting extension to the N-W [9] algorithm is used to increase the accuracy in multiple sequence alignment by Tailor et al. [12]. T-Coffee [13] also shows appreciable enhancement in accuracy over traditional alignment methods by prearchiving of alignment information. CHAIN [14] uses monte carlo optimization of a hidden markov model to establish gapped alignment of primary structures. A whole range of CLUSTAL [15,16] softwares are available for protein alignment customized for specific needs and available resources. Further, machine learning approaches [17] have been used to improve the similarity searches. Pearson [18] and Shpaer et al. [19] provide an extensive review and comparison of the existing tools for searching primary protein sequence databases. However, the algorithms fail to extract subsequences that are not identical in characters but share common secondary structure. In all of the above, similarity is very closely related to identity except while incorporating discrete properties like acidic, basic, aromatic to which an aligned amino acid may belong to.
Non character based approaches to establish similarity between polypeptides have also been tried with limited success like by capturing the repetitions of amino acids by considering sequences in the frequency domain using the acclaimed Fast Fourier Transformation [20-22]. Various repeats in the protein sequences can be adequately captured by using FFT and its various versions, but we lose the sequence information in such attempts.
Most of the algorithms for similarity detection are primarily alignment tools and are based on string managements of protein sequences that are considered as words of 20 characters. The algorithm presented here attempts to remove this limitation by considering the properties of the amino acids and also their variation directly during matching of sequences. Our approach is inspired by a few recent researches in the field of music retrieval and the commercial success of Music Database and Retrieval Systems [23] (MDR) based on the Spectral Analysis of audio signals. We have attempted to use the ideas in the field advantageously along with the traditional methods to adapt to protein sequence similarity estimation. Since the MDRs have been commercialized, new algorithms and heuristics may not be available in the public domain. The developed algorithm is capable of evaluating similarity based on any or a combination of the 256 attributes listed down in the AA index database [24,25] and is intended to detect local variations in the property in the sequence along with global alignment. We present this method as an extension to traditional character based matching algorithm.
Results
The algorithm was coded, with Sz and F kept as variable parameters. A single property, i.e. hydrophobicity [26] was taken as the property, F is kept more than twice the Sz so that no information is lost while the neighborhood around the highest peak is considered. βp, the penalty factor can be changed to accommodate the parameters and can be tuned to consider the 'not so similar' segments in the sequences. The threshold for selection of subsequences of size 8 amino acids with β = 2.5, was kept as a function of the actual character identities in the subsequences. The threshold t was taken as SSS < = 3.5 - n * 0.4, so that if there is no character identity, subsequence matches with SSS < = 3.5 were looked for. This non-fixed threshold function was evolved as matchings with high character identity did produce low matches, but the "interesting" matches are typically the ones with low identity of amino acids. A detailed analysis of the matching presents subsequences that are alphabetically dissimilar, and are therefore not detected by traditional algorithms, but share common 3D structures.
1. Various branches of the evolutionary tree of Human Kinome [27] were generated by tree-generating algorithm, after finding the distance measure for various kinases. As an illustration, when closely related kinases (with SwissProt accession no. in brackets), PAK4 (O96013), PAK5 (O95547), PAK6 (Q9NQU5) and a distant neighbor PLK1 (SwissProt acc.no: P53350) are run through the automation of the algorithm, expected results are obtained (see table 2). This establishes the global alignment capability which is due to the Dynamic Programming Algorithm. Similarly evolutionary relationships were found for the PKC series of human kinases (see table 3) with F doubled. The global alignment capability does not seem to be dependent on the F measure significantly.
2. PKCd (pdbid [28,29] (accession number in the Protein Data Bank [28]): 1bdy) and PKCe (pdbid: 1GMI) (BLAST identity 40%, similarity 57%) human kinases are considered as evolutionarily similar but do not produce close alignments (GAP 55.472%, SSS .7149). The algorithm was able to identify many subsequences that are not identical but share close secondary structure similarity. Results are tabulated in table 4 alongwith the alignment found in BLAST. Also, results are compared with those of Smith-Waterman algorithm [8], using the standard software called SSearch [7]. In both the cases, it was seen that SSS was able to identify subsequences that are alphabetically dissimilar but gives low SSS scores, but are structurally similar. The value of segment size Sz was kept 8 and F 16. The tertiary structures of the subsequences within the threshold were found to be closely similar using Swiss pdbviewer (SPBDV) [30,31]. The references of the figure showing alignments are given in each row in table 4. The alignments shown is between the subsequences by a simple "Magic Fit" in SPDBV using the actual pdb files of the proteins, and most of the fits obtained for SSS within the threshold validate our results. Therefore, it is possible that even when the subsequences have complete identity, they may theoretically not fit at all in the actual protein owing to the non-alignment of other regions.
PKCd and PKCe, though share a similar fold, do not superimpose well using SPDBV but our experiments suggest that the subsequences picked up by SSS within the threshold do produce good fits with low rms (root mean square) value apart from their similarity in the secondary structure (also shown in the table 4). Figure 4 shows the fits obtained using SPDBV for subsequences that were picked by the algorithm with the exception of Figure 6 which reported a high SSS value, and also has reported a high rms value during pdb fitting. Matches found with high character identity are not shown in the table, but in general their SSS value is lower which is taken care of by the threshold. This demonstrates that the algorithm's ability to pick non identical subsequences if they are similar in their tertiary structure. The accounting of subsequences through SSS that are found below threshold would increase the BLAST similarity score by more than 10% in this particular example and more than 5% in most other protein pairs. However, the potency of the algorithm essentially remains in capturing "interesting" subsequences and not perse at global alignment.
3. SSS consistently was found to capture subsequences with similar secondary structures, and most of the times with similar tertiary structures purely by the primary structure. In xyna-psefl (pdbid: 1clx) and xynz-clotm chain A (pdbid: 1xyz) we found interestingly subsequences that do not get aligned in BLAST but still show similar tertiary structures using the algorithm. Table 6 shows subsequences that are not aligned in BLAST and do not share sequential similarity but are similar in tertiary structures as seen through their pdb coordinates. Similar conclusions can be drawn by comparison with the results obtained by Smith-Waterman algorithm. SSearch engine was used for comparative analysis. This strongly suggests the potency of the algorithm to even find non aligned subsequences that are structurally similar and renders SSS as a useful test after traditional alignment algorithms. This seems to be a result of the inadequacy of the simplistic dynamic programming approach compared to BLAST which is a better alignment tool, but depicts that SSS with better alignment tools as an abstraction (like the way dynamic programming is used as a wrapper) can be used effectively for finding alignments between proteins where homology is not detected using traditional algorithms.
4. The algorithm was run on xyna-theau (pdbid: 1gor) and xynz-clotm chain A (pdbid: 1xyz) and compared with the results from BLAST. Subsequences that were found to be matching with large distance values (meaning that the similarity is not very high, but reported in the matching segments) were looked for their secondary structures. Appreciable similarity in secondary structures were reported though alignment was not perfect (see table 5). Figures 5 shows the fits obtained for the individual subsequences picked by the SSS using SPDBV. Xyna-theau and xynz-clotm are abound in H (Helix), but the algorithm is able to catch the subsequences where for short duration β strands were located within two bends and align them with a similar stretch in the other sequence. It must be considered, that interesting results may be expected by the algorithm (and those not expected from character based alignment) only when the distance value D is not very small, and a micro analysis of the matching segments may produce results that are unobtainable otherwise.
5. xyna-theau (pdbid: 1gor) and xyna-strli (pdbid: 1eov) when run over by the algorithm also produced subsequences that were dissimilar in characters but highly similar in their overall structure. In Table 7, subsequences 3 and 4 were completely dissimilar sequences but were obtained by the algorithm and were found to be very similar in their tertiary structure with very low rms values. Both the subsequences produce α helical structures.
This illustrates the chief advantage of the algorithm, wherein not only direct character alignment but similarity between subsequences is captured. Analysis in the spectral domain after conversion to an orthogonal plane of property using FFT allows SSS to establish similarity where traditional character based algorithm may not succeed. This holds true for BLAST and many other algorithms based on a similar approach. Though, essentially SSS is suited for detailed analysis of sequences in a locality and can be wrapped over by other global alignment tools (like N-W dynamic programming or BLAST), but within the locality it scores over other algorithm due to its emphasis on the local variation of the property besides the property itself. As has been demonstrated in the results, local variation of a group of properties can also have an effect on determining the structural and functional properties of the protein in a locality. Therefore, it scores over even Smith-Waterman [8] in the cases where alphabetical similarity is either low or does not exist. Further, any purely character based similarity approach cannot capture the local variation of multiple properties in a local region. If two subsequences register an appreciably low SSS score, and are sequentially different, it depicts the local variation of property (here hydrophobic effect for residue burial) to be similar in both the subsequences, which might be of interest to the analyst. Taking a greater frequency component (F being doubled) and subsequent analysis at such locations might give useful insight into the similarity pattern where character matching is not evident. The flexibility to use the algorithm with a healthy compromise between the frequency and position offers another advantage of the developed algorithm. Further, other properties like α helical propensity, β strand propensity may be used in conjunction with hydrophobicity as different property planes.
Conclusion
We present a novel method to establish similarity between two amino acid sequences that goes further than the conventional character based similarity approaches and purely frequency based similarity approaches based on repetitions of amino acids. The algorithm derives its inspiration from spectral similarity approaches employed successfully in music database retrieval systems and attempts to establish similarity based on the Spectral Similarity Score on any general attribute of amino acids. We have demonstrated that the approach is capable of picking subsequences of amino acids as similar though they may not be identical in nature. Further, tertiary structures of these picked subsequences have shown appreciable similarity and fit, though the overall structure of the protein may not fit well. This demonstrates that the algorithm is capable of establishing similarity in tertiary structure purely by processing primary structures even when the primary subsequences do not match well. Further, as SSS is able to find even subsequences that do not align through BLAST or SSsearch but are nevertheless similar, it can be used as a useful tool after operation by traditional alignment algorithms. Further, SSS without dynamic programming can be used to pick a subsequence of interest from a corpus of subsequences that alignment algorithms would fail to achieve.
A distinct advantage of the algorithm is its ability to detect subsequences that are not similar in characters but in the property under consideration, and even in the profile of the local variation of the property in a localized region. Therefore, it is able to establish similarity in those subsequences where character based similarity is not possible to establish. The algorithm is flexible and allows alteration of size of subsequences as powers of 2. If FFT is replaced by other fourier transformation algorithm (at the cost of time complexity) then this constraint on the size of the subsequence may also be eliminated. Another advantage of the algorithm is its ability to encode any property of the amino acids as given in the AAindex database. Therefore different indices may be used in different contexts to establish similarity in function, fold, structural, or evolutionary or superfamily relationships. These indices may be normalized to compare the results from different indices. Further, multiple properties may be handled at a time either by generating property profiles in different planes or by creating a new property as a linear combination of multiple properties. Effects of such extensions are currently being explored.
The Dynamic Programming approach can be replaced by other approaches used in character based similarity establishment with suitable modifications. Smith-Waterman algorithm performs an exhaustive search of all possible gapped alignments between a pair of sequences using a set of scoring parameters, and therefore can be used more effectively with SSS. It is noteworthy, that though there are frequency conversion mechanisms other than FFT, but the latter is a linear time algorithm and is therefore, faster. If Smith-Waterman algorithm is used as a wrapper for an exhaustive search of gapped alignments (here, SSS similarity alignments), usage of FFT would become critically important. Penalty, windowing and normalization parameters may be further tuned to get better depth in the results. Histograms can be generated for a better visualization of similarity and to avoid detailed analysis of the SSS results. Color coding of alignment, as done in BLAST, can be employed and algorithms used in MDR may be used in filtering and linearity enforcing. This approach, we believe, can be used in many fields in bioinformatics to establish similarity. This algorithm can be effectively used to find similarity in genomes after suitable estimation of the parameters, and can also be used to find similarities in the 3-dimensional structures of proteins by using variations in relatively accessible surface areas of proteins.
Methods
The focus of the SSS algorithm is to capture subsequences in amino acid sequences that are not similar on alphabetical scale, but are similar on some property(s) scale of which a choice can be made during the course of the algorithm. SSS involves preprocessing of the primary structure, and conversion to the frequency domain followed by matching and estimation of the similarity score.
Preprocessing of inputs
The algorithm intends to find the similarity measure based on any general attribute of the amino acid. Therefore, the amino acid in the input sequence is replaced with its attribute measure, such as the hydrophobicity [26] value. This generates a property profile of the protein in one dimension, which is a sequence of floating point numbers of length equal to the number of amino acids in the protein. If more than one property is to be considered simultaneously, then the property profile is a multi dimensional sequence.
Formally, for a protein Pr of size n (number of amino acids) let the p properties considered be {P1, P2,..., Pp}. Let function Pp(n) give the property value of type Pp of the amino acid at position n in protein Pr. Then the property profile of Pr is designed as
The sequences of floating point values thus generated is plotted with the position of amino acid as abscissa and its attribute measure as ordinate for each dimension p. The attribute is analogous to the amplitude of a time-varying non static signal, and the generated graph to the amplitude profile of the signal. Figure 1 describes the hydrophobicity profile of two closely related kinases PAK4 (Swiss-Prot [32] accession no: Q8N4E1) and PAK5 (Swiss-Prot accession no: O95547). Thereafter, the profile is segmented in equal segments of fixed length and the local maximum is found in each segment. The width of the segment would matter in the quality of results.
For each dimension p pertaining to property Pp, let the sequence be divided in N equal segments denoted by sp,i where i ∈ {1, 2 ..., N} and size of each segment be Sz. Also let the positions in each segment where local maximum was found be mp,i where i ∈ {1, 2, ..., N}. The maxima is found within the segment in the abcissa by simply comparing the peaks of the property values, as represented in figure 2.
The purport of identifying local maximum mp,i in each segment sp,i is to do away with bogus peaks in the neighborhood. It is assumed that an amino acid with the highest value of say, hydrophobicity would be able to influence the property of the protein the most in the vicinity. It should be noted that a local minima (instead of maxima) in each segment can also be considered for evaluation in the case where a lower value of the property determines the strength. For example, in the property considered here, the minima would mean the highest hydrophilicity. However, it is possible that the local maximum is not able to catch the property in a limited neighborhood, but that aspect is considered in the step that follows.
Conversion to frequency domain
Around each position mp,i a neighborhood of a size F is taken and converted to the frequency domain by using Fast Fourier Transformation (FFT) algorithm [33-35]. FFT is faster than other frequency conversion mechanisms and is a linear time algorithm rendering SSS faster [34]. This procedure constraints the value of F to a power of 2 (there are other ways with higher time order for fourier transformation that would not put this constraint on the value of F). The global alignment during matching is to be done for segments sp,i and not for individual amino acids. Positional information of the amino acids within a segment is not available after fourier transformation. Therefore, F can be used as a useful manoeuvering parameter while analysis of the alignment output.
The property profile PP on segmentation and fourier transformation generates a vector <vp,i >. We normalize each segment <vp,i > so that their mean is 0 and variance is 1. This procedure is conducted for each dimension p. For the two protein sequences to be compared, such two vectors are generated, of say size n and m.
Matching
We use Minimum distance matching method, a version of the Needleman-Wunsch Algorithm [9]. Let us surmise by considering two lists of vectors <xp,1, xp,2,..., xp,n > and <yp,1, yp,2, ..., yp,m > respectively. Let ep,i,j be the mean square distance between xp,i and yp,j. The mean square distance describes the extent of dissimilarity between the two complex frequency vectors.
Let Mp,k = {(xp,i, yp,j)} be defined as a matching of size k, pairing xp,i with yp,j. We need to get the largest matching with the lowest value of dissimilarity. Given the subsets = {xp,1, xp,2,..., xp,a}, = {yp,1, yp,2,..., yp,b} and a matching Mp,k s.t. (k ≤ a ≤ n, k ≤ b ≤ m), distance between the sets and wrt Mp,k is defined as:
and minimum distance between Xa and Yb can be calculated by finding the minimum over Mp,k. In effect, a penalty of βp is imposed on each non-matching vector, while the dissimilarity measure (msd) is imposed on those which are matching.
The distance measure between the two sequences can be found by using a dynamic programming approach [36] employing a recursive strategy as shown in figure 3.
We determine the optimal matching set Mp,k which gives the most optimal distance using dynamic programming approach. The optimal matching for all properties is a simple summation of optimal matching for all p dimensions. Therefore, after normalization gives us the Spectral Similarity Score (SSS). Note that the focus of the method is to capture the "interesting" subsequences with similarity in structure, but may not be similar in the alphabetical plane. Hence, this dynamic programming algorithm, which is not the chief concern of the method, can well be replaced suitably by any other matching algorithm for more accurate global alignment.
Time complexity analysis
The time order of an algorithm refered by O is defined as the number of operations required as an order of the input size of data. The preprocessing of inputs to replace with attribute amplitudes, and subsequently to identify local maximae in segments is O(n), while identifying the neighborhood of size F takes O(n) time for n residues. FFT takes Flog2(F) time for each vector in the list, and hanning, normalization take O(F) time for each vector. Since there are m = n/F vectors in all, it takes m * (O(F) + Flog2(F)) in all for a sequence. Dynamic Programming requires O(m2) time, if both sequences are assumed to be of equal length. Matching set can also be found in linear time over the number of segments m.
If the algorithm is implemented in a database, and queried for fixed values of F and segment size, then for a database of size n the time required is approximated to O(np), or linear in time for p properties considered at a time.
Authors' contributions
KG developed the idea into the algorithm, coded the software and tested on examples. Also he interpreted the results and jointly wrote the manuscript. DT fine tuned the parameters of the algorithm and did large scale testing on proteins besides assisting in writing the manuscript. SVV developed perl scripts for automation of testing. KVV supervised the testing of examples, fine tuning of the algorithm and jointly wrote the manuscript. SR assisted in developing the idea and guided the software development and testing.
Acknowledgements
We would like to extend our warm thanks and acknowledgements to Prof. Petety V. Balaji, School of Biosciences & Bioengineering, Indian Institute of Technology, Bombay for his assistance in the testing of the algorithm and criticisms on the manuscript.
Figures and Tables
Figure 1 Hydrophobicity Profiles generated before preprocessing for PAK4 & PAK5. Hydrophobicity profiles of the sequences of kinases PAK4 and PAK5 generated by substituting the amino acid characters with their respective property value (hydrophobicity values given in table 1). The two sequences are known to be closely similar. These profiles would subsequently be divided in equal segments and the neighborhood around the maximum peak in each segment would be converted to an orthogonal plane using Fast Fourier Transformation.
Figure 2 Preprocessing of Inputs in a single property plane. The property profile of one of the input sequences in a plane is subjected to segmentation of equal sizes. Maximum peak in each segmented is identified by simple comparison of the heights of the peaks and the a neighborhood of size F around the position containing the peak is taken. Each neighborhood is then collectively subjected to fourier transformation. This preprocessing is implemented in each plane of the property profile.
Figure 3 Matching of segments using dynamic programming. Matching of the Sequence vectors generated through Dynamic Programming. The method used is a version of the N-W Algorithm. A penalty of β is imposed on each non matching of segments while for an accepted match the distance score is increased by the dissimilarity measure between the segments. A matching is defined as an ordered map between the two ordered sets of segments.
Figure 4 3D matching for PAKd PAKe using SPDBV magic fit. 3D images of fit obtained by using SPDBV [30, 31] software's "magic fit" tools. The first value in the bracket is the SSS for the subsequence and second refers to rms value obtained by the tool in 0A. Color red is used for PKCd and yellow for PKCe. The subsequences in the figures are (a)MKEALSTE & DDSRIGQT (b) ANQPFCAV & QTFLLDPY (c) GKAEFWLD & ANCTIQFE (d) QAKVLMSV & RVYVIIDL (e) RVIQIVLM & RKIELAVF belonging to PKCd and PKCe respectively. All subsequences are completely dissimilar using character based approaches but are found to be similar using SSS. Appreciably low rms values confirms that the subsequences in fig 4a-4e are similar subsequences.
Figure 5 3D matching for xyna-theau xynz-clotm using SPDBV magic fit. 3D images of fit obtained by using SPDBV [30, 31] software's "magic fit" tools. The first value in the bracket is the SSS for the subsequence and second refers to rms value obtained by the tool in 0A Color red is used for xyna-theau and yellow for xynz-clotm. The two proteins are similar proteins with high BLAST score and overlapping 3D structures. SSS however is still able to catch subsequences that are left as dissimilar by BLAST, and low rms values for captured subsequences confirm the findings. The subsequences in the figures are (a) SCVGITVM & NCNTFVMW (b) GITVWGVA & TFVMWGFT (c) RVKQWRAA & MIKSMKER (d) EDGSLRQT & SGNGLRSS belonging to xyna-theau and xynz-clotm respectively.
Figure 6 3D matching for xyna-psefl xynz-clotm using SPDBV magic fit. 3D images of fit obtained by using SPDBV [30, 31] software's "magic fit" tools. The first value in the bracket is the SSS for the subsequence and second refers to rms value obtained by the tool in 0A. Color red is used for xyna-psefl and green for xynz-clotm. The subsequences for which structures are shown are (a) NCNTFVMW & RRGGITVW (b) RDSLLAVM & ENGAKTTA (c) YNSILQRE & RQSVFYRQ belonging to xynz-clotm and xyna-psefl respectively. All the subsequences found to be similar are left by traditional algorithms as dissimilar (or unidentical). Interestingly, the subsequences paired up in fig 6b and 6c are not aligned by BLAST but were still found to be similar by SSS and are captured by the same.
Figure 7 3D matching for xyna-theau xyna-strli using SPDBV magic fit. 3D images of fit obtained by using SPDBV [30, 31] software's "magic fit" tools. The first value in the bracket is the SSS for the subsequence and second refers to rms value obtained by the tool in 0A. Color red is used for xyna-theau and green for xyna-strli. Subsequences for which structures are shown are (a) TTPLLFDG & QTPLLFNN (b) SQTHLSAG & FQSHFNSG (c) VLQALPLL & YNSNFRTT belonging to xyna-theau and xyna-strli respectively. Fig 7a shows a structure refering to matching subsequences that shows that SSS is able to capture subsequences like traditional algorithms also, though it is also capable of picking subsequences like in fig 7c that are not similar on the basis of amino acid characters.
Table 1 Estimated Hydrophobic Effect for residual burial. Estimated Hydrophobic Effect for residual burial shown in the second column for each amino acid. These values are substituted for individual amino acid forming a property plane for further preprocessing of inputs in SSS. The values are in kilocalories/mol.
Amino Acid value [kcal/mol]
Gly 1.18
Ala 2.15
Val 3.38
Ile 3.88
Leu 4.10
Pro 3.10
Cys 1.20
Met 3.43
Phe 3.46
Trp 4.11
Tyr 2.81
His 2.45
Thr 2.25
Ser 1.40
Gln 1.65
Asn 1.05
Glu 1.73
Asp 1.13
Lys 3.05
Arg 2.23
Table 2 Distance Matrix for human kinases PAK series. Distance measures D between various human kinases. PAK series are closely similar kinases, while PLK1 is a distant relative in the kinome evolutionary tree [27]. Smaller SSS values correspond to strong similarity. GAP (Needleman-Wuntch [9] algorithm implemented in gcg package) scores are in percentage similarity. F = 8, Sz = 4, β = 0.502. It can be seen that the dynamic programming approach used in the SSS algorithm is a simple but effective approach to ascertain global similarity. A replica of the branch of the kinome tree can be generated using the matrix.
PLK1 PAK4 PAK5 PAK6
SSS GAP SSS GAP SSS GAP SSS GAP
PLK1 0.000 100 0.981 39.688 0.976 37.813 0.969 39.264
PAK4 0.981 39.688 0.000 100 0.681 69.898 0.845 63.776
PAK5 0.976 37.813 0.681 69.898 0.000 100 0.870 58.045
PAK6 0.969 39.264 0.845 63.776 0.870 58.045 0.000 100
Table 3 Distance Matrix for PKC series in human kinome. Distance matrix for PKC series in Human Kinome. These proteins occur as a distinct branch in the phylogenetic tree of the Human Kinome. GAP results are given as percentages while SSS scores are fractions. Lower SSS scores refer to higher similarity detection. It is seen that SSS with the dynamic programming approach is able to capture phylogenetic relationships between human kinases in the PKC subfamily of proteins. F = 16, Sz = 8, β = 2.5
PKCa PKCb PKCd PKCe
SSS GAP SSS GAP SSS GAP SSS GAP
PKCa 0.000 100 0.4678 85.949 0.7238 61.835 0.7391 63.851
PKCb 0.4678 85.949 0.000 100 0.7254 61.029 0.6904 62.944
PKCd 0.7238 61.835 0.7254 61.029 0.000 100 0.7149 55.472
PKCe 0.7391 63.851 0.6904 62.944 0.7149 55.472 0.000 100
PKCg 0.5137 81.081 0.5951 79.464 0.7348 60.589 0.7550 61.695
PKCh 0.7160 64.794 0.7371 - 0.7371 53.506 0.6146 76.035
PKCi 0.7498 52.072 0.7568 50.357 0.7338 45.098 0.7599 52.909
PKCt 0.7113 61.847 0.7474 59.370 0.6068 73.333 0.7451 55.043
Table 4 SSS results for PKCd and PKCe kinases. SSS results for the human kinases PKCd and PKCe (BLAST identity score 40%, similarity 57%). Similar subsequences are shown where BLAST is not able to find appreciable similarity with pure character matching strategies. None of the good alignment detected by BLAST were found to be with high SSS scores. Only the sequences with low SSS scores but low BLAST alignments are shown. Smith-Waterman algorithm application SSearch results are also shown. Figures in the last column are created by Magic Fit using the SPDBV software with real pdb files downloaded from the PDB Databank. F = 16, Sz = 8, β = 2.5. PDBids : PKCd = 1BDY, PKCe = 1GMI. The assignments for secondary structure are: h = helix; b = residue in isolated beta bridge; e = extended beta strand; g = 310 helix; i = pi helix; t = hydrogen bonded turn; s = bend [37].
Seq Segment Subseq msd Blast Result SSearch Results rms Image
1 PKCd
PKCe (4) [31–39]
(6) [46–54] 2.31 MKEALSTE
DDSRIGQT 1.67 fig 4a
2 PKCd
PKCe (3) [22–30]
(4) [33–41] 4.38 ANQPFCAV
QTFLLDPY ANQPFCA V
QTFLLDP Y 1.17 fig 4b
3 PKCd
PKCe (12) [99–107]
(12) [95–103] 3.96 GKAEFWLD
ANCTIQFE GKAEFWL D
ANCTIQF E 2.08 fig 4c
4 PKCd
PKCe (14) [110–118]
(15) [111–119] 3.85 0.83 fig 4d
5 PKCd
PKCe (8) [66–74]
(10) [76–84] 1.77 0.66 fig 4e
Table 5 SSS results for xyna-theau and xynz-clotm. SSS results for the human kinases xyna-theau and xynz-clotm (BLAST identity score 41%, similarity 59%). Similar subsequences are shown where BLAST is not able to find appreciable similarity with pure character matching strategies. None of the good alignment detected by BLAST were found to be with high SSS scores. Only the sequences with low SSS scores but low BLAST alignments are shown. Figures in the last column are created by Magic Fit using the SPDBV software with real pdb files downloaded from the PDB Databank. F = 16, Sz = 8, β = 2.5. PDBids : xyna-theau = 1GOR, xynz-clotm = 1XYZ. The assignments for secondary structure are: h = helix; b = residue in isolated beta bridge; e = extended beta strand; g = 310 helix; i = pi helix; t = hydrogen bonded turn; s = bend [37].
Seq Segment Subseq msd Blast Result rms Image
1 x-theau
x-clotm (32) [259–267]
(30) [242–250] 3.89 0.76 fig 5a
2 x-theau
x-clotm (36) [288–296]
(31) [245–253] 1.96 0.61 fig 5b
3 x-theau
x-clotm (27) [215–223]
(20) [158–166] 2.45 0.17 fig 5c
4 x-theau
x-clotm (20) [160–168]
(13) [103–111] 2.97 1.59 fig 5d
Table 6 SSS results for xyna-psefl and xynz-clotm. SSS results for the F/10 xylanases xyna-psefl and xynz-clotm (BLAST identity score 33%, similarity 52%). Similar subsequences are shown where BLAST is not able to find appreciable similarity with pure character matching strategies. Interestingly the second subsequence does not find alignment in BLAST and is not sequentially similar but produces good alignment. SSearch alignment results are based on Smith-Waterman algorithm. SSearch also did not align the presented subsequences, though it is more sensitive to local and detailed alignments of sequences. Smith-Waterman score was 510 while similarity score in SSearch was 32.984%. Referenced figures show the fit obtained using SPDBV Magic Fit. F = 16, Sz = 8, β = 2.5, SSS = 0.783. PDBids : xyna-psefl = 1CLX, xynz-clotm = 1XYZ. The assignments for secondary structure are: h = helix; b = residue in isolated beta bridge; e = extended beta strand; g = 310 helix; i = pi helix; t = hydrogen bonded turn; s = bend [37].
Seq Segment Subseq msd Blast Result SSearch Results rms Image
1 x-clotm
x-psefl (37) [297–305]
(37) [297–305] 2.10 1.29 fig 6a
2 x-clotm
x-psefl (16) [124–132]
(23) [184–192] 2.51 RDSLLAVMR
ENGAKTTAE DSLLAVM
NGAKTTA 1.96 fig 6b
3 x-clotm
x-psefl (07) [054–062]
(18) [145–153] 2.59 YNSILQREY
RQSVFYRQR NSILQRE
QSVFYRQ 2.58 fig 6c
Table 7 SSS results for xyna-theau and xyna-strli. SSS results for the F/10 xyna-theau and xyna-strli (BLAST identity score 47%, similarity 62%). Both the xylanases are exceedingly similar in their structure (RMS = 2.130A using SPDBV) and therefore close identity in the primary structure is expected as depicted by high BLAST identity score. Similarly, SSearch produces good alignment where character based identity is high. Highly identical subsequences do produce low SSS score (row 1) but non identical subsequences producing low scores are interesting. BLAST does not detect similarity in row 3 subsequences but aligns them. SSearch, however, does not align the two subsequences in row 3. Most of them show similar secondary and tertiary structures. Note the similarity in secondary structures shown below each subsequence. F = 16, Sz = 8, β = 2.5. PDBids: xyna-theau = 1GOR, xyna-strli = 1EOV. The assignments for secondary structure are: h = helix; b = residue in isolated beta bridge; e = extended beta strand; g = 310 helix; i = pi helix; t = hydrogen bonded turn; s = bend [37]. SSS = 0.731.
Seq Segment Subseq msd Blast Result SSearch Results rms Image
1 x-theau
x-strli (38) [304–312]
(41) [329–337] 0.60 0.21 fig 7a
2 x-theau
x-strli (29) [231–223]
(32) [255–263] 2.17 1.51 fig 7b
3 x-theau
x-strli (30) [243–239]
(33) [265–273] 3.92 VLQALPLL
YNSNFRTT VLQALPLL
YNSNFRTT 1.69 fig 7c
4 x-theau
x-strli (27) [215–223]
(30) [239–247] 2.93 0.19 fig 7d
==== Refs
Altschul SF Boguski MS Gish W Wootton JC Issues in searching molecular sequence databases Nature Genet 1994 6 119 129 8162065 10.1038/ng0294-119
Taylor WR Orengo CA A holistic approach to protein structure alignment Protein Eng 1989 2 505 519 2748567
Altschul SF Madden TL Schffer 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
Altschul SF Gish W Miller W Myers EW Lipman DJ Basic local alignment search tool J Mol Biol 1990 215 403 410 2231712 10.1006/jmbi.1990.9999
McGinnis S Madden TL BLAST: at the core of a powerful and diverse set of sequence analysis tools Nucleic Acids Res 2004 32 W20 W25 15215342 10.1093/nar/gnh003
Pearson W Lipman DJ Improved tools for biological sequence comparison Proc Natl Acad Sci U S A 1988 85 2444 2448 3162770
Pearson WR Searching protein sequence libraries: comparison of the sensitivity and selectivity of the Smith-Waterman and FASTA algorithms Genomics 1991 11 635 650 1774068 10.1016/0888-7543(91)90071-L
Smith TF Waterman MS Identification of Common Molecular Subsequences J Mol Bio 1981 147 195 197 7265238 10.1016/0022-2836(81)90087-5
Needleman SB Wunsch CD A general method applicable to the search for similarities in the amino acid sequence of two proteins J Mol Biol 1970 48 443 453 5420325 10.1016/0022-2836(70)90057-4
Carugo O Pongor S Protein fold similarity estimated by a probabilistic approach based on C([alpha])-C([alpha]) distance comparison J Mol Biol 2002 315 887 898 11812155 10.1006/jmbi.2001.5250
Tonges U Perrey SW Stoye J Dress AW A general method for fast multiple sequence alignment Gene 1996 172 GC33 41 8654965 10.1016/0378-1119(96)00123-0
Taylor WR Saelensminde G Eidhammer I Multiple protein sequence alignment using double-dynamic programming Comput Chem 2000 24 3 12 10642876 10.1016/S0097-8485(99)00043-1
Notredame C Higgins DG Heringa J T-Coffee: A novel method for fast and accurate multiple sequence alignment J Mol Biol 2000 302 205 217 10964570 10.1006/jmbi.2000.4042
Neuwald AF Liu JS Gapped alignment of protein sequence motifs through Monte Carlo optimization of a hidden Markov model BMC Bioinformatics 2004 5 157 15504234 10.1186/1471-2105-5-157
Higgins DG Thompson JD Gibson TJ Using CLUSTAL for multiple sequence alignments Methods Enzymol 1996 266 383 402 8743695
Thompson JD Higgins DG Gibson TJ Clustal W: improving the sensitivity of progressive multiple sequence alignments through sequence weighting, position-specific gap penalties and weight matrix choice Nucleic Acids Res 1994 22 4673 4680 7984417
Karwath A King RD Homology Induction: the use of machine learning to improve sequence similarity searches BMC Bioinformatics 2002 3 11 11972320 10.1186/1471-2105-3-11
Pearson WR Comparison of methods for searching protein sequence databases Protein Sci 1995 4 1145 1160 7549879
Shpaer EG Robinson M Yee D Candlin JD Mines RTH T H Sensitivity and selectivity in protein similarity searches: a comparison of Smith-Waterman in hardware to BLAST and FASTA Genomics 1996 38 179 191 8954800 10.1006/geno.1996.0614
Pasquier CM Promponas VI Varvayannis NJ J HS A Web server to locate periodicities in a sequence Bioinformatics 1998 14 749 750 9789101
de Trad CH Fang Q Cosic I Protein sequence comparison based on wavelet transform Protein Engineering 2002 15 193 202 11932490 10.1093/protein/15.3.193
Shepherd AJ Gorse D Thornton JM A Novel Approach to the Recognition of Protein Architecture from Sequence Using Fourier Analysis and Neural Networks PROTEINS: Structure, Function, and Genetics 2003 50 299 302 10.1002/prot.10290
Cheng Y Music Database Retrieval Based on Spectral Similarity Stanford University Database Group technical report 2000 2001 2014
AAindex: Amino Acid Index Database, Release 6.0, September 2002
Kawashima S Kanehisa M AAindex: amino acid index database Nucleic Acids Res 2000 28 374 10592278 10.1093/nar/28.1.374
Karplus PA Hydrophobicity Regained Protein Science 1997 6 1302 1307 9194190
Manning G Whyte DB Martinez R Hunter T Sudarsanam S The Protein Kinase Complement of the Human Genome Science 2002 298 1912 1934 12471243 10.1126/science.1075762
Berman H Henrick K Nakamura H Announcing the Worldwide Protein Data Bank Nature Struct Bio 2003 10 980 14634627 10.1038/nsb1203-980
Berman HM Westbrook J Feng Z Gilliland G Bhat TN Weissig H Shindyalov IN Bourne PE The Protein Data Bank Nucl Acids Res 2000 28 235 242 10592235 10.1093/nar/28.1.235
Guex N Peitsch MC SWISS-MODEL and the Swiss-PdbViewer: An environment for comparative protein modeling Electrophoresis 1997 18 2714 1723 9504803
Deep View Swiss-PdbViewer
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 Nucl Acids Res 2003 31 365 370 12520024 10.1093/nar/gkg095
Press WH Teukolsky SA Vetterlong WT Flannery BP Fast Fourier Transformation Numerical Recipes in C 2002 2 Cambridge University Press 496 524
Elliott DF Rao KR Fast Transforms: Algorithms, Analyses, Applications 1982 New York: Academic Press
Heideman MT Johnson DH Burris CS Gauss and the history of fast Fourier Transform IEEE ASSP Magazine 1984 1 14 21
Cormen TH Leiserson CE Rivest RL Stein C Dynamic Algorithms Introduction to Algorithm 2000 2 2 MIT Press
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
| 15850477 | PMC1131888 | CC BY | 2021-01-04 16:02:48 | no | BMC Bioinformatics. 2005 Apr 23; 6:105 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-105 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1081585751010.1186/1471-2105-6-108Research ArticleScoredist: A simple and robust protein sequence distance estimator Sonnhammer Erik LL [email protected] Volker [email protected] Center for Genomics and Bioinformatics, Karolinska Institutet, Berzelius väg 35, 171 77 Stockholm, Sweden2005 27 4 2005 6 108 108 30 12 2004 27 4 2005 Copyright © 2005 Sonnhammer and Hollich; 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
Distance-based methods are popular for reconstructing evolutionary trees thanks to their speed and generality. A number of methods exist for estimating distances from sequence alignments, which often involves some sort of correction for multiple substitutions. The problem is to accurately estimate the number of true substitutions given an observed alignment. So far, the most accurate protein distance estimators have looked for the optimal matrix in a series of transition probability matrices, e.g. the Dayhoff series. The evolutionary distance between two aligned sequences is here estimated as the evolutionary distance of the optimal matrix. The optimal matrix can be found either by an iterative search for the Maximum Likelihood matrix, or by integration to find the Expected Distance. As a consequence, these methods are more complex to implement and computationally heavier than correction-based methods. Another problem is that the result may vary substantially depending on the evolutionary model used for the matrices. An ideal distance estimator should produce consistent and accurate distances independent of the evolutionary model used.
Results
We propose a correction-based protein sequence estimator called Scoredist. It uses a logarithmic correction of observed divergence based on the alignment score according to the BLOSUM62 score matrix. We evaluated Scoredist and a number of optimal matrix methods using three evolutionary models for both training and testing Dayhoff, Jones-Taylor-Thornton, and Müller-Vingron, as well as Whelan and Goldman solely for testing. Test alignments with known distances between 0.01 and 2 substitutions per position (1–200 PAM) were simulated using ROSE. Scoredist proved as accurate as the optimal matrix methods, yet substantially more robust. When trained on one model but tested on another one, Scoredist was nearly always more accurate. The Jukes-Cantor and Kimura correction methods were also tested, but were substantially less accurate.
Conclusion
The Scoredist distance estimator is fast to implement and run, and combines robustness with accuracy. Scoredist has been incorporated into the Belvu alignment viewer, which is available at .
==== Body
Background
Estimating divergence time of protein sequences is one of the fundamental problems in bioinformatics. Evolutionary distance estimates are used by many of the most commonly used phylogenetic tree reconstruction algorithms [1-3]. In current research, phylogenetic trees are used for many types of subsequent analysis, e.g. orthology inference [4-6]. Early models for sequence evolution focussed on nucleotides. They commonly employ Markov chains and assume independent evolution at every site. Each of the four nucleotides is identified by one state and the substitution probability is modelled as a state transition probability from one state to another. In the most straightforward approach, the same state transition probability is assigned to every substitution [7]. Subsequent models take account of more nucleotide specific properties, e.g. transitional and transversional substitutions as well as GC content (see [8] for an introduction). These more advanced approaches are bound to nucleotide sequences and cannot be directly used with protein sequences.
Markov chain models for protein evolution differ from nucleotide approaches in their larger number of states and transitions for which parameters need to be estimated. The protein sequence Jukes-Cantor model assigns the same probability to each substitution and is hence a rather poor approximation. This method essentially takes the observed differences between two sequences and corrects this value to the estimated evolutionary distance using a logarithmic function. Other similar methods exist that also correct observed differences, e.g. Kimura's method [9]. Although they produce rather inaccurate distance estimates, correction-based distance estimators are popular because of their simplicity. More advanced protein evolution models estimate parameters from protein sequence alignments. Assuming the same substitutions for closely and distantly related sequences leads to the construction of the Dayhoff matrix series [10]. Following this approach, it suffices to collect data from alignments of closely related sequences to build an evolutionary model of amino acid substitution.
Dayhoff and co-workers introduced the term Percent Accepted (point) Mutation (PAM), which denotes a commonly used measure for evolutionary distance between two aligned sequences (insertions and deletions are ignored). In other words, two sequences at a distance of 150 PAM are related to each other by 1.5 substitutions per position on average. As substitution is a stochastic process, some positions will experience multiple substitutions while others will experience none. It is also possible that secondary substitutions at one site will result in the original residue, making the evolutionary steps invisible. This is in essence the reason why estimating evolutionary distance is so hard – multiple substitutions cannot be observed directly. An evolutionary distance of 250 PAM corresponds roughly to 80% observed differences. The term PAM is found in literature for both the matrix series given by Dayhoff et al. as well as for evolutionary distance unit. In this publication we refer to the matrices as Dayhoff matrices and reserve the term PAM for distance units.
There are two major shortcomings connected with the derivation of the Dayhoff matrices. First, potential errors inherent in the experimental data will be magnified by extrapolation. Additionally, it is questionable whether substitution probabilities observed on closely related sequences can accurately reflect the evolution of more distantly related sequences. The efforts of researchers since the publication of the Dayhoff matrices have led to several other matrix series, sharing the idea of an underlying Markov chain. They differ in terms of the data they are built upon and account for the above-mentioned shortcomings in various ways [11-13].
The approach behind the BLOSUM matrices [14] is different from Dayhoff's evolutionary model. Whereas the Markov model assumes that any transition probability matrix may be derived from another matrix in the same series, the BLOSUM matrices do not imply any evolutionary time. There is no direct mathematical relationship between matrices in the BLOSUM series. Sequences with identities above a given identity cutoff are clustered and used to derive score matrices. The BLOSUM matrices are known as a good general-purpose choice. Especially, BLOSUM62 is frequently chosen for the alignment of sequences.
Results
We here introduce Scoredist, a novel correction-based distance estimator for protein sequences. It applies a correction function to an observed reduction in normalised score, rather than to observed differences as other correction-based methods. This gives a better estimate of the divergence in the well-established PAM measure and allows the popular BLOSUM matrix series to be used. Other matrices could in principle be used, but the BLOSUM matrix has proved to be the most universal. Scoredist distance estimates are calculated directly by a simple equation and do not require cumbersome computational approximations, which is needed for e.g. Maximum Likelihood (ML) and Expected Distance (ED) estimates [15]. Additional calibration opens the possibility to make Scoredist tuned to other evolutionary models.
In order to evaluate our novel protein distance estimator Scoredist against other estimators, we generated a large testset of artificial sequence alignments. Simulation is the only way to exactly know an alignment's evolutionary distance. The substitutions were made by ROSE [16] according to an evolutionary model that can be chosen arbitrarily. It is to be expected that a distance estimator based on a particular evolutionary model will perform optimally on a testset generated with the same model. We therefore generated testsets using four different matrix series: Dayhoff [10], MV [12], JTT [11], and WAG [13]. For each model, 2000 alignments were created for evolutionary distances between 1 and 200 PAM units, i.e. 10 alignments for each distance. The Scoredist, Maximum Likelihood, and Expected Distance estimators can all be tuned towards a particular evolutionary model. We therefore used three evolutionary models which were also used to generate the testsets for these distance estimators, and use a shorthand to refer to these as "method-model". For instance, Maximum Likelihood using the MV model is denoted ML-MV. The Jukes-Cantor and Kimura estimators can not be tuned to a specific model but were tested on all four datasets.
Table 1 shows a compressed summary of the results. For each combination of distance estimator and dataset, the average root mean square deviation from the true distance was calculated for all 2000 alignments. The Expected Distance results were similar to ML, as the methods are akin in nature, but ML was generally more accurate and is much more widely used. As expected, low RMSD values as a sign of good distance estimates were generally obtained when using the same model for alignment creation and subsequent distance estimation. This is seen in the diagonal of low RMSD values from Dayhoff/Dayhoff to JTT/JTT. The only exception to this rule was observed when the testset was generated with Dayhoff. Here, ML-JTT was slightly better than ML-Dayhoff. This result was also verified for distances up to 250 and 300 PAM (data not shown). Comparing Scoredist and ML accuracies when training and testing using the same model resulted in a tie. Scoredist was better for MV, ML was better for JTT, and they were equally accurate for Dayhoff. When comparing accuracies for different training and testing models, however, Scoredist dominates. Here, Scoredist performed better than ML in five of the six cases. For the MV testset the difference was very big. The only case where ML was better than Scoredist was again when running ML-JTT on the Dayhoff testset, which for unclear reasons produced very accurate distance estimates.
The Jukes-Cantor and Kimura correction methods are generally less accurate than Scoredist and ML estimators. In some cases they reached higher accuracy than Scoredist and ML trained on the "wrong" model. For instance, on the Dayhoff testset Kimura was better than Scoredist-MV and ML-MV, and on the MV testset Jukes-Cantor was better than Scoredist and ML trained on Dayhoff or JTT. However, Jukes-Cantor and Kimura never came near the Scoredist and ML accuracy when trained on the "right" model. In a real situation, it is of course not known which evolutionary model is most appropriate. Therefore, taking the average RMSD values for each training model reveals the generality and robustness of the method on different testsets. The average accuracy of Scoredist is consistently better than for ML, and Jukes-Cantor and Kimura are even further behind.
Figure 1 two shows a more detailed picture of the different distance estimators. The average of 10 estimates from 10 independent simulations at each evolutionary distance is plotted for data generated with the Dayhoff matrices. The variance among the 10 estimates is not shown for clarity; they are however reflected by the RMSD values in Table 1 which may give a slightly different picture. For instance, it is possible that the average deviation is close to zero if the individual estimates have large positive and negative deviations that cancel each other out. Therefore, the RMSD values should be trusted more than the deviation plots when in doubt. Figure 1A shows the dependence on evolutionary model for Scoredist and ML. Testing on the Dayhoff testset, Scoredist-Dayhoff and ML-Dayhoff stayed reasonable accurate in the entire range (below 5% error). In contrast, Scoredist-MV and ML-MV deviated considerably from the true distance. It is however clear that ML is more affected by switching model than Scoredist is. In Figure 1B the testset was generated with the MV model. Again, the corresponding deviation was observed for "wrong model" estimators. Here it is even more pronounced that ML is more dependent on the model, and generalizes poorly. Scoredist was less affected by the change of model – Scoredist-Dayhoff was considerably more accurate on the MV testset than ML-Dayhoff. As expected, when Scoredist and ML had been trained on MV data, the accuracy is very good for both estimators. In conclusion, we observed that although the Scoredist method is very simple compared to the ML method, it is approximately equally accurate when testing and training using the same evolutionary model. However, when testing on a different model, Scoredist is considerably more accurate.
Implementation
The Scoredist estimator was implemented in Belvu, which is a general-purpose multiple alignment viewer that allows basic alignment editing. Belvu can calculate and display phylogenetic trees. The tree reconstruction can be based on Scoredist or other common correction-based distance estimators available within Belvu.
Multiple alignments can be coloured in Belvu according to conservation using average BLOSUM62 score in the column, or by residue-specific colours. User-specified cutoffs can be employed to fine-tune the display. Belvu has a range of functions for sorting, colouring, marking up, and printing alignments. In Figure 2, the alignment is coloured according to conservation, and sorted according to the tree. The effect of distance correction with Scoredist is illustrated.
Belvu can also be utilised for batch mode operations on the multiple alignment, or for producing distance matrices or phylogenetic trees without graphical output. It is available for the most common UNIX operating systems and can be obtained from [20]. A Windows version exists but is less frequently maintained. See [17] for instructions, and [18] for information on the Stockholm format, which is used by the Pfam project.
Discussion
Our analysis was based on four different evolutionary models – Dayhoff, MV, JTT and WAG. We chose these because they represent the spectrum of models well. The only tuning done in the Scoredist method is the estimation of the calibration factor c. This factor can be seen as a scaling factor for the logarithm base in equation (5) that needs to be set empirically.
The difference between Scoredist and ML becomes particularly apparent in the MV dataset. There are several hypotheses for this behaviour. The Dayhoff matrices were constructed with the limited data available at the time. Given the substantial increase of research output in this field particularly during the last decade, it is not surprising that the Müller-Vingron model (published in 2000) reports substantially other results than the Dayhoff (1978) and JTT (1992) matrices. Additionally, the calibration factor c can also be interpreted as measure for the similarity of the respective models. Following this argument, JTT and Dayhoff are more akin given a Δc ≈ 0.05. The MV model is more distant to both JTT (Δc ≈ 0.11) and Dayhoff (Δc ≈ 0.16).
The Expected Distance estimator generally overestimates distances. For instance, among Dayhoff-calibrated estimators on the MV testset, Expected Distance is more than 10 PAM RMSD units (over 50%) poorer than the best method Scoredist. Similar values are observed for JTT calibrated estimators. Generally, MV-trained estimators are prone to underestimate evolutionary distances (Figure 1A). In combination with the ED higher distance estimation, this rather fortuitously leads to good results for ED – MV. However, the scope of this research was to identify a robust method that performs well on various data sources. An estimator which is highly sensitive to the data source or possible incorrect calibration is of less value. The best single estimator was JTT-calibrated Scoredist. If the method per se is measured by averaging over all calibrations and testsets, Scoredist receives 15.39, ED 16.89, and ML 17.14 PAM RMSD units. This highlights Scoredist as the most robust estimator, with the distance between Scoredist and ED (ΔScoredist, ED ≈ 1.50) being 6 fold the difference between ED and ML (ΔED, ML ≈ 0.25).
We here only present Scoredist results using BLOSUM62 for calculating the score σ between two sequences. In principle one could use some other score matrix, but we found that this had little effect on the results. Since the goal was to make a general-purpose method, BLOSUM62 was an obvious choice. The key to Scoredist is the usage of scores rather than identities, and the choice of somewhat arbitrary parameters is not of primary concern. At present, gaps in the alignments are not included in the Scoredist calculation. Traditionally, gaps have been difficult to embody in evolutionary models. In the models used here, they are at best crudely modelled by treating every gap equally. An inherent problem is that the probabilities for insertions and deletions (indels) are not necessarily synchronized with the substitution probabilities. Some protein families are more prone to indels than others, hence it is hard to make a generalizable model that suits all protein types. We have experimented with affine gap penalties in the Scoredist method (this is an option in the implementation), but this resulted in decreased accuracy. We therefore do not recommend using gaps to estimate protein distances.
Conclusion
We have developed the score matrix based distance estimator Scoredist for aligned protein sequences. Its main advantages are computational simplicity and high robustness. Most other distance estimators produce good results for certain evolutionary models but perform poorly on others. The Maximum Likelihood and Expected Distance were found to overfit their estimates to the evolutionary model so much that the results on testsets generated with other models suffered heavily. The correction-based methods Jukes-Cantor and Kimura also favoured a particular evolutionary model, but were not competitively accurate on any testset. It seems that Scoredist achieved the best compromise between accuracy and generalization power.
Methods
For the estimation of divergence time, let s1 and s2 be two aligned protein sequences (gaps are ignored) of identical length l. A similarity score σ is defined as
where S is a log-odds score matrix. Log-odds score matrices are constructed such that substitutions by the same or a similar amino acid receive a positive score, whereas substitutions to dissimilar amino acids are attributed a negative score. The expected value for this kind of matrix is negative. This ensures that the comparison of unrelated sequences returns a negative score. For two random sequences of length l the expected score σr(l) = σ0 * l, where is the expected value of the score matrix. As we strive to measure scores above the scores for the null model of sequence independence, the score σ(s1, s2) is deducted by the expected score σr, giving the normalised score σN
σN = σ(s1, s2) - σr (l). (2)
For two random sequences of length l the expected score σr (l) = σ0 * l, where σ0 is the expected value of the score matrix.
The expected score σr for unrelated sequences can be regarded as lower limit. The upper limit of the score between s1 and any other sequences is given by σ(s1, s1). For two different sequences, the upper limit of the score σU is, for the sake of symmetry, assumed to be
and normalised
σUN = σU (s1, s2) - σr (l). (4)
Any sound score σN is situated within the interval [0, σUN]. The validity of the upper boundary follows from the score's definition. The lower boundary might, however, get violated if two sequences receive a score σ(s1, s2) <σr (l). As the model assumes independent evolution already for σr (l), a score below σr does not contain any additional information. A lower score is therefore set to σ(s1, s2) = σr (l). We model the raw distance as a modified Poisson process
As seen in Figure 3, dr is linearly related to the true distance, deviating only by a constant factor. The Scoredist evolutionary distance estimate of two sequences is given as the product of the raw distance and a calibration factor
ds = c * dr. (6)
Evolutionary distances of 250–300 PAM units are commonly considered as the maximum for reasonable distance estimation and, therefore, the Scoredist estimate ds is restricted to the interval [0, 300] PAM.
Calibration factors can be determined for various evolutionary models. We used the ROSE program [16] to simulate evolution with three different matrix series and generated 2000 sample sequence alignments for distances up to 200 PAM units. The calibration factor c was calculated by least squares fitting on this data, using the BLOSUM62 score matrix for calculating the score σ in the estimator (Table 2, Figure 3). The simulated evolution started with a random sequence of 200 residues. For each integer distance within the interval [1, 200] PAM, we produced 10 alignments, yielding 2000 alignments per dataset. The default gap parameters of ROSE V1.3 were applied. Each dataset was generated with the transition probability matrix and the stationary frequencies of the respective evolutionary model.
Calculation of Maximum Likelihood (ML) and Expected Distances (ED): ML distances were estimated by applying the Newton-Raphson method to the derivative of the likelihood of the evolutionary distance given an alignment. To calculate ED, the same likelihood function was numerically integrated, to get its "center of gravity" [15]. Both methods are implemented in the program lapd (L. Arvestad, unpublished), which uses Perl and Octave. The Jukes-Cantor and Kimura distance estimators were run as implemented in Belvu. The popular PROTDIST program from the PHYLIP package [19] calculates only ML-Dayhoff and Kimura distances. We therefore chose to use lapd in order to assess Scoredist by a broader range of distance estimators.
Authors' contributions
ES had the initial idea and implemented the method. VH carried out the evaluation and wrote the first manuscript draft. All authors read and approved the final manuscript.
Acknowledgements
We thank Lars Arvestad for the lapd program, for helpful discussions, and for reading the manuscript.
Figures and Tables
Figure 1 Stratified accuracy analysis of Scoredist and ML. To illustrate how estimated distance depends on the model, the average deviation is plotted as a function of true distance for two evolutionary models, Dayhoff and Mueller-Vingron. For each evolutionary distance between 1 and 200 PAM, 10 alignments were generated. For each alignment, the deviation was calculated as the difference between the estimated distance and the true distance used for data generation by ROSE [16]. The average of the 10 deviations was plotted using a running average with a window of 10 residues. Note that positive and negative deviations at the same true distance can cancel each other out – the curve only shows the average deviation and not the variability. The values in Table 1 measure the accuracy more correctly by using RMSD of every datapoint. The testset data was created with the matrices given by Dayhoff (A) or Müller-Vingron (B). In both cases, the estimators using the same evolutionary model as the testset data perform well. However, when switching the model in the estimator, Scoredist diverges less than ML, indicating that Scoredist is more robust. The curves show that ML-MV is more different from ML-Dayhoff than Scoredist-MV is from Scoredist-Dayhoff, particularly for the MV dataset in (B). The less difference between estimates using different models, the more robust is the method.
Figure 2 The Belvu multiple sequence alignment viewer. Belvu is a multiple sequence alignment viewer that implements the Scoredist distance estimator. The alignment window (A) shows a subset of the Pfam family DNA_pol_A (PF00476). Uniprot IDs are shown throughout. A sequence with known structure is included (DPO1_ECOLI) – the SA line showing surface accessibility and the SS line showing secondary structure. The neighbour-joining tree in (B) used uncorrected distances (observed differences), while the tree in (C) used Scoredist correction. Belvu assigns a colour to each species if provided with species markup information. The distance correction mainly affects the longer branches, and affects the tree topology in some cases, e.g. the placement of DPOQ_HUMAN. Structural markup and taxonomic information were embedded in the Stockholm format alignment provided by the Pfam database.
Figure 3 Estimation of the calibration factor c in Scoredist. This factor rescales the raw distance dr to optimally fit true evolutionary distances. The plot shows how c is estimated by least-squares fitting of raw distances dr to true distances for 2000 artificially produced sequence alignments, using the Dayhoff matrix series. The linear relationship between the raw distance dr and the true distance of the sequence samples justifies the introduction of the calibration factor c, which was here determined to cDayhoff = 1.3370 (See Table 2).
Table 1 Accuracy as average RMSD values for combinations of data modelsand estimators
testset
Dayhoff MV JTT WAG average
Scoredist – Dayhoff 12.68 20.85 13.67 12.81 15.00
ML – Dayhoff 12.70 28.40 14.75 15.15 17.75
ED – Dayhoff 13.57 31.36 16.10 16.63 19.41
Scoredist – MV 19.28 13.15 16.29 18.73 16.86
ML – MV 19.96 13.44 19.36 19.21 17.99
ED – MV 15.68 13.35 13.95 14.75 14.43
Scoredist – JTT 13.67 17.16 12.89 13.47 14.30
ML – JTT 12.15 25.07 12.10 13.44 15.69
ED – JTT 12.56 27.71 12.70 14.37 16.84
Jukes-Cantor 23.92 16.28 19.88 22.48 20.64
Kimura 16.24 29.81 22.36 19.16 21.89
For each testset and method, the average root mean square deviation from the true distance was calculated for 2,000 alignment samples in the interval 1–200 PAM units. Lower RMSD values indicate higher accuracy on a single testset. The column 'average' gives the mean of the four evaluated testsets. A low value in this column shows the estimator's robustness as it measures the accuracy over all four models (including "wrong" data models). Scoredist was more robust than ML, as it for each training set always had higher accuracy on average. The ED estimator gave good results when trained with MV, but was poor in all other cases (see Discussion for details). Scoredist, Jukes-Cantor, and Kimura distances were calculated with the Belvu alignment viewer. The Maximum Likelihood (ML) and Expected Distance (ED) estimates were produced by lapd (L. Arvestad, unpublished).
Table 2 Calibration factors for three evolutionary models
c
Dayhoff 1.3370
JTT 1.2873
MV 1.1775
The raw distance dr is scaled by the calibration factor c, which was obtained by least squares fitting of 2000 artificial protein sequence alignments generated for the matrices as given by Dayhoff, JTT (Jones-Taylor-Thornton), and MV (Müller-Vingron).
==== Refs
Bruno WJ Socci ND Halpern AL Weighted Neighbor Joining: A Likelihood-Based Approach to Distance-Based Phylogeny Reconstruction Mol Biol Evol 2000 17 189 197 10666718
Gascuel O BIONJ: An Improved Version on the NJ Algorithm Based on a Simple Model of Sequence Data Mol Biol Evol 1997 14 685 695 9254330
Saitou N Nei M The Neighbor-joining Method: A New Method for Reconstructing Phylogenetic Trees Mol Biol Evol 1987 4 406 425 3447015
Zmasek C Eddy S RIO: analyzing proteomes by automated phylogenenomics using resampled inference of orthologs BMC Bioinformatics 2002 3 14 12028595 10.1186/1471-2105-3-14
Hollich V Storm CE Sonnhammer ELL OrthoGUI: graphical presentation of Orthostrapper results Bioinformatics 2002 18 1272 1273 12217923 10.1093/bioinformatics/18.9.1272
Storm CE Sonnhammer ELL Automated ortholog inference from phylogenetic trees and calculation of orthology reliability Bioinformatics 2002 18 92 99 11836216 10.1093/bioinformatics/18.1.92
Jukes TH Cantor CR Munro HN Evolution of protein molecules Mammalian Protein Metabolism 1969 Academic Press 21 132
Nei M Kumar S Molecular Evolution and Phylogenetics 2000 New York: Oxford University Press
Kimura M The Neutral Theory of Molecular Evolution 1983 Cambridge: Cambridge University Press
Dayhoff MO Schwartz RM Orcutt BC Dayhoff MO A model of Evolutionary Change in Proteins Atlas of Protein Sequence and Structure vol 5 supplement 3 1978 National Biomedical Research Foundation, Washington 353 352
Jones DT Taylor WR Thornton JM The rapid generation of mutation data matrices from protein sequences Comput Appl Biosci 1992 8 275 282 1633570
Müller T Vingron M Modeling amino acid replacement J Comput Biol 2000 7 761 776 11382360 10.1089/10665270050514918
Whelan S Goldman N A general empirical model of protein evolution derived from multiple protein families using a maximum-likelihood approach Mol Biol Evol 2001 18 691 699 11319253
Henikoff S Henikoff JG Amino acid substitution matrices from protein blocks Proc Natl Acad Sci USA 1992 89 10915 10919 1438297
Agarwal P States JS A Bayesian Evolutionary Distance for Parametrically Aligned Sequences J Comput Biol 1996 3 1 17 8697232
Stoye J Evers D Meyer F Rose: generating sequence families Bioinformatics 1998 14 157 163 9545448 10.1093/bioinformatics/14.2.157
Belvu website
Stockholm data format
Felsenstein J PHYLIP – Phylogeny Inference Package (Version 3.2) Cladistics 1989 5 164 166
Belvu download site
| 15857510 | PMC1131889 | CC BY | 2021-01-04 16:02:50 | no | BMC Bioinformatics. 2005 Apr 27; 6:108 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-108 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1111586970810.1186/1471-2105-6-111Research ArticleRefined repetitive sequence searches utilizing a fast hash function and cross species information retrievals Reneker Jeff [email protected] Chi-Ren [email protected] Department of Computer Science, University of Missouri, Columbia, USA2 Department of Health Management & Informatics, University of Missouri, Columbia, USA2005 3 5 2005 6 111 111 28 12 2004 3 5 2005 Copyright © 2005 Reneker and Shyu; 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
Searching for small tandem/disperse repetitive DNA sequences streamlines many biomedical research processes. For instance, whole genomic array analysis in yeast has revealed 22 PHO-regulated genes. The promoter regions of all but one of them contain at least one of the two core Pho4p binding sites, CACGTG and CACGTT. In humans, microsatellites play a role in a number of rare neurodegenerative diseases such as spinocerebellar ataxia type 1 (SCA1). SCA1 is a hereditary neurodegenerative disease caused by an expanded CAG repeat in the coding sequence of the gene. In bacterial pathogens, microsatellites are proposed to regulate expression of some virulence factors. For example, bacteria commonly generate intra-strain diversity through phase variation which is strongly associated with virulence determinants. A recent analysis of the complete sequences of the Helicobacter pylori strains 26695 and J99 has identified 46 putative phase-variable genes among the two genomes through their association with homopolymeric tracts and dinucleotide repeats. Life scientists are increasingly interested in studying the function of small sequences of DNA. However, current search algorithms often generate thousands of matches – most of which are irrelevant to the researcher.
Results
We present our hash function as well as our search algorithm to locate small sequences of DNA within multiple genomes. Our system applies information retrieval algorithms to discover knowledge of cross-species conservation of repeat sequences. We discuss our incorporation of the Gene Ontology (GO) database into these algorithms. We conduct an exhaustive time analysis of our system for various repetitive sequence lengths. For instance, a search for eight bases of sequence within 3.224 GBases on 49 different chromosomes takes 1.147 seconds on average. To illustrate the relevance of the search results, we conduct a search with and without added annotation terms for the yeast Pho4p binding sites, CACGTG and CACGTT. Also, a cross-species search is presented to illustrate how potential hidden correlations in genomic data can be quickly discerned. The findings in one species are used as a catalyst to discover something new in another species. These experiments also demonstrate that our system performs well while searching multiple genomes – without the main memory constraints present in other systems.
Conclusion
We present a time-efficient algorithm to locate small segments of DNA and concurrently to search the annotation data accompanying the sequence. Genome-wide searches for short sequences often return hundreds of hits. Our experiments show that subsequently searching the annotation data can refine and focus the results for the user. Our algorithms are also space-efficient in terms of main memory requirements. Source code is available upon request.
==== Body
Background
Many algorithms have been developed recently that search DNA sequences looking for various types of subsequences [1-10]. All of these algorithms were developed to help life science researchers efficiently and accurately find short segments of DNA within entire genomes. Some search for short tandem nucleotide repeat (STNR) sequences that are commonly found throughout the genomes of higher organisms [1-6]. Others search for variable length tandem repeats (VLTR) and multi-period tandem repeats (MPTR) [8]. Most of them require that the sequence be repeated at least a few times in order to guarantee being detected and all of them require that the repeat be of a certain minimum length. Many of the algorithms can identify repeats without a priori knowledge of the repeat pattern. They do this by identifying short segments of DNA, termed words, and then manipulating the properties of these words as a process moves down the length of the sequence. A word's properties include, for instance: location, distance to the nearest identical word, and Hamming distance [11] to similar words. Once the process is finished, the results are displayed with accompanying statistics. Each new search must proceed from the beginning of the sequence. However, Ning, Cox, and Mullikin [10] have developed a different type of algorithm called 'Sequence Search and Alignment by Hashing Algorithm', SSAHA, that stores information about the locations of DNA words into a hash table. During a homology search, these locations are sorted and then examined for a series of numbers that are word-length apart. Since the hash map is in main memory, this search can be quite fast. For instance, a search through 2.7 GBases of DNA took only 2.20 seconds per query on average while searching for homology with 177 query sequences (104,755 total bases). However, in this experiment, the system required a minimum homology of 2k - 1 bases (where the word size k = 10) to guarantee a match. Furthermore, under no circumstances can the SSAHA algorithm find matches for sequence lengths less than k. Also, since the hash map is in main memory, scaling becomes a problem for multiple species searches.
In addition to SSAHA, other algorithms employ a static index of the database to speed retrievals. For instance, BLAT [12] is very effective for aligning mRNA and genomic DNA taken from the same species. It uses an index of non-overlapping length k DNA words (and their positions in the database) to find short matching sequences that can be extended into longer matches. As with the SSAHA algorithm, BLAT cannot locate hits smaller than length k. Also, BLAT excludes from the index k-mers that occur too often as well as k-mers containing ambiguity codes in order to reduce the number of initial matches to extend. This practice improves performance at the cost of missing some hits.
The FLASH [13] and TEIRESAIS [14] algorithms were designed to manage mismatches between a query sequence and sequences in the database. FLASH uses a strategy where non-contiguous DNA words are concatenated and then indexed. The algorithm generates a very large number of concatenated DNA words from each portion of an original string. TEIRESAIS was built to discover patterns in biological sequences by scanning input sequences to locate the set of all patterns with at least minimal support. Then, it builds larger patterns by matching prefixes and suffixes of patterns in the initial set. However, like the SSAHA algorithm, scaling becomes a problem for FLASH and TEIRESAIS for multiple species searches.
While accuracy and efficiency are important to life science researchers, they undoubtedly would like to do more than just locate their query sequences. For example, suppose a researcher knows the sequence of a transcription factor binding site and wishes to search through several species to see what genes this factor might be controlling. Current algorithms could easily generate tens of thousands of hits but leave the researcher with weeks of additional work in order to locate more specific information. If, however, the researcher also had an idea about possible annotation terms that would accompany specific genes of interest, then they could substantially narrow the search results by simultaneously searching for those terms. The expected results would be all genes that contain both the sequence and the annotation terms. This would add value to the search and help the researcher to understand a biological meaning in the results.
We have developed a search algorithm based on a unique and fast hash function that can search for a query sequence of any length in any number of genomic sequences. Like most other recent search algorithms, our algorithm uses the properties of DNA words, or more specifically their location. We have also incorporated an information retrieval function into our hash structure for fast retrieval of the annotation data that accompanies genomic sequences. This can possibly help facilitate knowledge discovery through cross-species conservation of sequences.
Results
Algorithm efficiency and utility
We report three experiments in this paper; the first experiment demonstrates the efficiency of our algorithm while the second and third demonstrate the potentials to the life science community. Our current implementation features a Dual Xeon 2.4 GHz processor with 512 KB cache, 2 GB RAM, and 120 GB EIDE 7200 rpm hard drive.
Experiment 1: Efficiency study
For our efficiency study, a random number generator was used to select locations from within Arabidopsis thaliana chromosome 5, which is over 26,000,000 base pairs. Then a sequence was retrieved from each location. For each sequence length, ten thousand different sequences were obtained and searched. Presently, our database consists of 49 different chromosome sequences from Arabidopsis thaliana, Haemophilus influenzae, Helicobacter pylori 26695, Helicobacter pylori J99, Homo sapien and Saccharomyces cerevisiae. Each chromosome was searched for each query. Since we obtained query sequences from a chromosome within our database, we can expect to find at least one hit per query. The average search times and the average number of hits over all chromosomes for each query length are listed in Table 1. For instance, length 4 queries require approximately 56 seconds to retrieve 15,428,878 hits on average. Length 8 queries require just over one second to retrieve an average of 98,141 hits while length 1024 queries require approximately 4.7 second to retrieve 1 hit on average. The discussion below details an explanation of these results.
Experiment 2: Refining repetitive sequence searches
Our abstract mentioned that a whole genomic array analysis in yeast had revealed 22 PHO-regulated genes. The promoter regions of all but one of them contain at least one of the two core Pho4p binding sites, CACGTG and CACGTT [15]. Using these sequences, we tried a simple test of the utility of our algorithm. First, the yeast genome was searched for both sequences without adding any annotation terms to find 4027 total hits. Assume that we are interested in learning more about these sites. We can try to focus the search by adding annotation terms. Pho-regulated genes are involved in phosphate metabolism in yeast so the term 'phosphatase' was added to the search which narrowed the results to 51 total hits. We found 10 hits for CACGTG, 20 hits for CACGTT, and 21 hits for AACGTG (the reverse complement of CACGTT). If the promoter region is defined as 1500 base pairs upstream of the start codon, then CACGTG was located in 4 different promoters, CACGTT was in 7 different promoters and AACGTG was in 8 promoters. In other words, 19 genes in the yeast genome contain at least one of the two core Pho4p binding sites in their promoter region and also contain the term 'phosphatase' in their annotation data. This figure is 90.5% (19 / 21) of the results reported in reference [15] and gives a high level of confidence for further investigation of these genes. If the promoter region is extended by 1500 base pairs, then there are 8 more hits – Protein IDs 6321642, 14318551, 6320067, 6324664, 6319971, 6321700, 6322880, and 6325061. These searches were repeated for the other genomes in our database and the results are shown in Table 2.
Experiment 3: Cross-species study
In this experiment, the findings from a published work involving phase-variable sites in Helicobacter pylori were used as an impetus to identify a potentially new phase-variable site in Haemophilus influenzae. Furthermore, there is supporting evidence for the new site in another published work involving Haemophilus influenzae. Although the experiment presented is specific to phase-variable sites in two species of microorganisms, the methodology is flexible enough to be applied to other similar biological problems.
A recent study [18] of Helicobacter pylori 26695 and J99 reported 46 candidate phase-variable genes that were identified by either having a homopolymeric tract greater than or equal to (G)7, (C)7, (A)9, and (T)9 or having a dinucleotide repeat greater than or equal to four copies. Seven of the 46 genes reported in the study were classified as lipopolysaccharide-biosynthesis related. A search of our database was conducted for the same set of repetitive sequences in the same species. We refined the search by concurrently searching for the term 'lipopolysaccharide' similarly to the previous experiment. The search returned eight distinct genes – two of them (Protein ID 15611217, and 15611264) were identified by having a (CT)4 repeat in their open reading frames. This finding was used as an impetus for a new search in a different species.
Assuming that (CT)4 also plays a role in the phase-variable properties of lipopolysaccharide genes in other microorganisms, a search for (CT)4 was conducted in Haemophilus influenzae. Of the 20 total hits, one gene (Protein ID 16273438) is described as a 'Lic-1 operon protein'. This gene is a lipopolysaccharide biosynthesis-related gene because 'Lic-1' is a homologue of lipopolysaccharide and 'operon' is a group of closely linked genes that produces a single messenger RNA molecule in transcription. The paper by Hood et.al. [19] reports three phase-variable lipopolysaccharide biosynthesis genes (Lic1, Lic2, and Lic3) in Haemophilus influenzae that were identified by multiple (>4) CAAT repeats. A search for (CAAT)4 in Haemophilus influenzae with a concurrent search for 'Lic-1' returned Protein ID 16273437 which abuts to Protein ID 16273438 and is part of the same operon. Thus, the 'Lic-1 operon' of Haemophilus influenzae contains both (CAAT)4 repeats as well as (CT)4 repeats – separated by approximately 1400 base pairs.
A molecular biological analysis would be required to verify whether the (CT)4 repeat in the Lic-1 operon plays a role in phase-variability, but since it is known in Helicobacter pylori [18] it might also in Haemophilus influenzae. The purpose of the experiment in this paper is to illustrate how cross-species searches using our database can quickly help discern hidden correlations in the data.
The search algorithms presented here are available on our web site [16] so that users can conduct experiments similar to the ones presented above. Users wanting a brief introduction to the web site can read the paper [17]. The current paper describes the hash function in detail and gives example searches. Here we also present detailed experiments which are not found in the previous paper that demonstrate the usefulness of our approach to the research community.
Discussion
Our hash function described below processes DNA words that are eight bases in length. Therefore, length 8 queries will retrieve one hash bin in an average of 1.15 seconds. Length 4 queries must retrieve 256 contiguous hash bins because there are 256 possible combinations that make a length 4 query into a length 8 query. For longer queries, although the query lengths double, the search times do not because the recursive search function has been optimized to reduce IO operations. When it has been determined that the query is not present, no further IO operations are performed. The search times presented grow at O(n) time, where n is the number of chromosomes searched. The chromosomes are searched sequentially so searching a very large number of chromosomes should be possible with our system. After the hits have been retrieved for the current chromosome, its pages are free to be swapped by the operating system.
Conclusion
We have presented a novel algorithm for homology searches in DNA sequences. Our hash function approach can quickly locate exact matches with a query sequence of any length. Also, our new search engine has several information retrieval features to assist researchers in finding functional homologies across species. Several more data mining features are in development. As more species are added to our database, we anticipate a richer data mining experience for the life science research community.
Methods
Definitions
Necessary concepts
We define a mapping function m(A) = 1, m(T) = 2, m(G) = 3, and m(C) = 4 to map DNA bases into digits. We can convert a DNA word into a number by applying the general positional number system conversion function f() to Q = {q0q1q2...q|Q|-1}:
where Q is a DNA word, m is the mapping function, q is one base of the word, r is the radix (four in our case), and |Q| is the length of the word. This conversion differs slightly from the usual number system conversion since zero is not used in order to achieve a one-to-one mapping between a DNA word and the radix 10 number system. If m(A) = 0, the function could not distinguish between one A and a string of A's. Table 3 lists twenty sequences and their corresponding radix 10 numbers. One caveat to this conversion in our implementation is that we proceed from left-to-right down the DNA word for coding convenience instead of the usual right-to-left. For instance, the codon ATG is (1 * 40) + (2 * 41) + (3 * 42) = 57.
Algorithms
Obtaining sequence and annotation data
We used sequences and annotation data from NCBI [20]. When available, genomic sequence from a FASTA nucleic acid file (*.fna) and annotation data from a protein table file (*.ptt) were used. Human genomic sequence data from build 34.2 was obtained from the BLAST database in FASTA format [21]. Human annotation data for this build was obtained from two different files. The file named gene.q contained the gene annotation which was listed by GeneID. The file named seq_gene.md listed the beginning and ending locations along a chromosome for each GeneID.
Constructing the hash table
Sequences in our database are preprocessed as follows. At each base along the sequence, the algorithm counts up k consecutive bases to make a k-word. Then, it converts the word into an integer key as discussed above. If the sequence has n bases, then there are n - k + 1 keys in the sequence. The location of each key is the number of bases from the beginning of the sequence to the first base of the word that made the key. Thus, the pair <key, location> describes each base along the length of the sequence. For our application, we add one more descriptive attribute to this pair which is the protein ID (PID) of the nearest gene to this location. For human sequences, this ID came from the seq_gene.md file while for other sequences it came from the *.ptt file. A hash table file is used to store locations and PIDs while a related indexing file stores information about the hash table and is accessed via the key.
Our database contains a set of sequences S = {s1, s2, ..., sNd} where si is the sequence for the i-th chromosome in the database. Each sequence in S has two associated files, a list of <location, PID> pairs and an array of offsets into which serves as an index for . There are 4k bins in , one for each of the 4k possible k-words. After sorting, the algorithm writes each <location, PID> pair into and counts the number of pairs written for each key. This information is written into according to the following format:
key|offset,occurrences,bytes,numbersize
where key is discussed above, offset is the starting position within of the first <location, PID> pair for this key, occurrences is the number of <location, PID> pairs for this key, bytes is the number of bytes written to for this key, i.e. the hash bin size, and numbersize is the size in bytes of each location and PID. Figure 1 shows how offset plus bytes equals offset for the next key. Each line of is extended with spaces on the right up to an appropriate number (we use 30) so that all lines have the same length. In this way, when using as an index into , we can easily determine which line to read based on the key. See lines 4–10 and 20–23 of the search algorithm presented in Figure 2. Using larger k-words during processing will increase the size of but will not change the size . It will, however, decrease the size of each hash bin within .
To distinguish locations from PIDs within , one can either use delimiters between them or add leading zeros to shorter numbers to make all numbers the same size. Using delimiters saves space for shorter sequences while adding leading zeros saves space for longer sequences. For instance, if all PIDs are 8 bytes and all locations are extended to 8 bytes with leading zeros, then this approach begins to save space for sequences longer than approximately 1,000,000 base pairs. The average chromosome length in our database is 65.79 million base pairs. Therefore, our implementation uses leading zeros.
Sequences in FASTA format allow for bases other than A, T, G, and C. For instance, R = (A or G), Y = (T or C), and K = (G or T), etc. However, a radix of 4 will only accommodate A, T, G, and C during the conversion. Therefore, for other bases, the algorithm converts the DNA word into each possible combination. For instance, the word YA is converted into two words, TA and CA, which are both stored into their respective hash bins. Query sequences are similarly converted into each possible combination before searching and these combinations are searched sequentially. Therefore, in sequence s1 of Figure 1, a user could search for YAA and recover two CAA hits (at positions 0 and 19) and no TAA hits. This approach differs from the SSAHA system which converts all non-standard bases into A's and would not recover any hits since there are no AAA's in s1. The next two sections show how to use and to search for a query sequence within .
We also have a table in a relational database to hold gene annotation information. The schema of the Protein table is as follows:
Protein(species: NUMBER(3), chr: NUMBER(2), begin: NUMBER(8),
end: NUMBER(8), strand: CHAR(1), length: NUMBER(5),
PID: NUMBER(10), product: VARCHAR2(3000),
PRIMARY KEY (PID));
Species is a number that can be de-referenced when needed. The chromosome number, the beginning of the gene, the end of the gene, and the length are self-explanatory. The strand attribute is either + or - depending on the direction of the gene. The PID attribute is the universally unique Protein ID number that allows us to access information on other databases. Finally, the gene product contains a brief description of the function of the gene. Some examples include: Proline /pyrroline-5-carboxylate dehydrogenase, carbonic anhydrase, and lipopolysaccharide biosynthesis protein. Both the PID and the product are indexed based on the frequency of the queries on these attributes.
Hash function
The function below takes as input a DNA word q and returns the key(s) necessary to locate the correct hash bin(s). It is used by the search function which is presented in the next section. Assume that k is still the length of the DNA words that were used to make the hash table in the previous section. If |q| <k, we would like to expand q into a larger sequence domain q', where |q'| = k, while still retaining all of the properties of q. One possible way is to add every combination of A, T, G, and C to the left of q such that the final length is k. Once again, our algorithm works from left to right for coding convenience. Another simpler way is to add A's to the left of q to get the word
qLow = A1A2...Ak-|q|q
and to add C's to the left of q to get the word
qHigh = C1C2...Ck-|q|q
so that |qLow| = |qHigh| = k. By applying the conversion function to qLow and qHigh, the low and high values of a range of numbers that represents q within q' are found. Furthermore, the numbers in this range are consecutive. The input to the function, q, is the entire query, Q, if |Q| < = k and is a k-length subsequence of Q if |Q| >k. The search function presented below determines which case is appropriate.
Hash(q, k)
offset = f((A)k)
if(|q| <k) //Case 1
qLow = A0A1...Ak-|q|q
qHigh = C0C1...Ck-|q|q
h(q) = {f(qLow) - offset, f(qHigh) - offset}
else if(|q| = k) //Case 2
h(q) = {f(q) - offset}
return h(q)
Case 1 returns two keys that bound 4k-|q| contiguous hash bins. Case 2 returns only one key for one hash bin. An offset is subtracted from each conversion so that the smallest key, (A)k, will equal zero, the next larger key will equal one, and so forth.
Searching the database
We now present an algorithm called Search to locate all occurrences of a query sequence Q within the database (see Figure 2). The database S contains a set of sequences {s1, s2, ..., sNd} where s is the sequence of a chromosome. All sequences in S are searched separately with their union giving the final result. Each sequence sx has two associated files, the array index and the list of offsets , which are used differently during a search based on the length of Q. Since our implementation uses k bases of DNA per k-word, the three possibilities for length are shorter than k, equal to k, and longer than k. In all three cases, the algorithm returns <location, PID> pairs of all occurrences of Q. By taking the reverse complement of Q and reapplying the search algorithm, we can search the opposite strand of each sequence.
Case 1 – Query length shorter than k bases
There is a set of keys derived by extending Q to k bases with all possible combinations of sequence and then applying the hash function to each sequence. A range for this set is found by adding A's (or C's) to the left hand side of Q until length equals k and converting to get the low key (or the high key). Since is ordered, this range is continuous from low to high. Next, is consulted to determine the offset, the numbersize in bytes, and the count of the numbers to read from . Then, the hard drive's read/write head seeks offset bytes into and loads into main memory only the pages necessary for reading. See lines 3–18 of Figure 2.
Case 2 – Query length exactly k bases
Q is converted into its key and is consulted to find the offset, the numbersize in bytes, and the count of the numbers to read. Then, the read/write head seeks to the correct position within and loads into main memory only the pages necessary for reading. See lines 19–29 of Figure 2.
Case 3 – Query length longer than k bases
The search algorithm divides Q into two parts, Q1 and Q2, where Q1 is the first k bases of Q and Q2 is the last length - k bases of Q. Recursive calls to Search with Q1 and Q2 retrieve two sets of results having the format <location, PID>. They are named R1 and R2. After retrieval, we find matches by comparing locations. If length is greater than k and less than 2k, then Q1 and Q2 will overlap and a match will be a location from R2 that is (length - k) greater than a location from R1. If length is greater than or equal to 2k, then Q1 and Q2 will abut and a match will be a location from R2 that is k greater than a location from R1. See lines 30–57 of Figure 2.
Retrieval IO is the most time-consuming step in our algorithm. Let n be the length of a sequence. If we assume an even distribution of each k-word in the sequence then there will be approximately n / 4k k-words per hash bin. If each word and each PID are 8 bytes, for instance, then there will be approximately 16n / 4k bytes per hash bin. As an example, with n = 200,000,000 bases and k = 8 there are 48,828 bytes per hash bin on average. Page sizes usually range from 4,096 bytes up to 4,194,304 bytes. If we assume a page size from the upper half of this range then in this example we can safely estimate for short queries (6 to 16 bases) that the algorithm will only retrieve one or two pages per search. This keeps retrieval times low and allows our system to perform well even with very large hash map files each with 4+ GB made from 200+ million base pair sequences.
Examples of searches
As an example of each category of search with k = 2, we will search for Q of lengths 1, 2, 3, and 4 from s1 of Figure 1. Line numbers in this section refer to the pseudo code in Figure 2 unless indicated otherwise. Lines 3 through 18 are used to search for "G" within s1. Hash("G", 2) in line 2 returns the smallest and biggest keys from the set of keys {13 ("AG"), 14 ("TG"), 15 ("GG"), 16 ("CG")}. The key for (A)2 from Table 3 is 5. Therefore, in line 4 and 5, index line low = 13 - 5 = 8 and index line high = 16 - 5 = 11. Line 8 from in Figure 1 is 13|120,1,8,4 (the lines begin at zero) and line 11 is 16|144,1,8,4. Applying lines 11 and 12 of the algorithm, we get count = ((144 - 120) / 4) + (1 * 2) = 8. Thus, we seek 120 bytes into and read 8 numbers which are all 4 bytes long. The result set is [<0008, 1234> <0013, 5678> <0022, 5678> <0006, 1234>] as <location, PID> pairs. Going back to s1, all occurrences of "G" have been found (see table 6):
Next, lines 19 through 29 will find exact matches to "TT". The key for "TT" = 10 so the index line from in Figure 1 is 10 - 5 = 5 which is 10|88,1,8,4. In line 25 through 27 of the algorithm, we seek 88 bytes into and read 2 numbers which are both 4 bytes long. The result set is [<0003, 1234>] as a <location, PID> pair. From the original sequence s1 the only occurrence of "TT" is at location 3.
Now we will illustrate finding "CAA" as a disperse repeat. Disperse repeats are repeats that are not adjacent. From lines 32 and 33 of the search algorithm, Q1 = "CA" and Q2 = "AA". Lines 34 and 35 are recursive calls to Search that return R1 = [<0000, 1234> <0019, 5678>] and R2 = [<0001, 1234> <0020, 5678>] as <location, PID> pairs. The locations in R1 are 0 and 19 and in R2 are 1 and 20. Applying lines 36–41, R1 [0] + 3 - 2 = R2 [0]. Similarly, R1[2] + 3 - 2 = R2[2]. So, the pairs [<0000, 1234> <0019, 5678>] are returned as the result set.
Finally, we will illustrate finding the tandem repeat (CT)2. Tandem repeats are adjacent repeats. So this example will find "CTCT" as a tandem repeat of "CT". By applying lines 46 and 47 of the algorithm, Q1 = "CT" and Q2 = "CT", too. In this example, R1 = R2 = [<0010, 1234> <0012, 1234> <0016, 5678>]. Applying lines 50–55, R1 [0] + 2 = R2[2]. This is the only pair of locations that meet the criterion, so the result set is [<0010, 1234>].
Information retrieval using query expansion
We conducted the following experiments to examine the best approach for query expansion [11] using the Gene Ontology (GO) database [22]. The first step was to determine the correlation between individual NCBI annotation terms and GO terms. There were 14,349 unique NCBI annotation terms in our database. Of these terms, we found 160 (1.12%) matching terms within the GO database. There were also found 160 inexact matching terms by searching for similarity (i.e. using the LIKE keyword in the WHERE clause instead of equality). Next, we broadened the matching criterion even further by testing unique NCBI terms similarity to GO phrases. The NCBI term was surrounded by wildcard characters (i.e. %term%) and again tested for similarity. This search gave 561,294 matching phrases. We conducted similar experiments with two and three term windows. The results are shown in Table 4. Next, we expanded the GO terms retrieved from the previous experiments by including their unique siblings. Siblings are defined as terms (or phrases) having the same parent in the term2term table in GO. For example, "apoptosis" and "hypersensitive response" are siblings of "autophagic cell death" which share the same parent, "programmed cell death". The results are presented in Table 5. Presently, based on these experiments, we expand annotation terms by querying the GO database with two term windows and including all siblings of GO phrases. This is in an attempt to include other relevant terms in the search.
Authors' contributions
The hash function algorithms were devised and implemented by JR who also prepared the manuscript. CRS provided essential guidance during development of the algorithms as well as during preparation of the manuscript.
Acknowledgements
We would like to thank Dr. Peiyu Zeng of the University of Rhode Island and Dr. Walter Gassman and Dr. Joe Polacco of the University of Missouri-Columbia for their insight into bacterial pathogens and Arabidopsis thaliana and their assistance with several other biological aspects of our system. We would also like to thank Tao Tao of NCBI user services for his help with the human genome data. JR was funded from National Library of Medicine (NLM) Bioinformatics and Health Informatics Training (BHIRT) Grant No. 5 T15 LM07089 13.
Figures and Tables
Figure 1 Index file A and data file L for sequence s1: CAATTACGAGCTCTGCCTACAATGAT. The format for and are discussed in the text. To demonstrate how different regions map to different genes, the first 13 bases map to the gene with PID = 1234 and the last 13 bases map to the gene with PID = 5678. We add leading zeroes to each location so that all numbers in are four bytes and we record this as numbersize in each line in . Keys in this example are made from two bases of sequence so there are 42 = 16 lines in ranging from m(AA) = 5 through m(CC) = 20. Key number m(GT) = 11 and number m(GG) = 15 are not present in the sequence. For clarity, each offset in is repeated in the correct position above the line in and each PID is underlined. Two arrows map two different lines from into by pointing to two bubbles that show the content of two hash bins.
Figure 2 Database search pseudo code. The length of the query sequence Q determines which block of code will execute. Lines 3 – 18 execute for |Q| <k (wordsize) while lines 19 – 29 execute for |Q| = k. If k < |Q| < 2k then Q is divided into two k length pieces for recursive calls to Search in lines 32 – 35 and the results from these calls are further tested in lines 38 – 41 to obtain the final answer in line 42. If |Q| > = 2k, a similar block of code is executed in lines 44 – 57. However, a different comparison is made in line 54 as compared to line 40.
Table 1 Time study over 3.224 GBases using randomly obtained sequences
Query Length (Bases) Average Search Time (Seconds) Average Number of Hits
4 56.40 15,428,878
8 1.147 98,141
12 2.254 2308
16 2.235 121
32 2.451 1.55
64 2.517 1.15
128 2.648 1.07
256 2.956 1.03
512 3.598 1.01
1024 4.738 1.01
Table 2 Cross-species search for PHO4P binding sites, CACGTG and CACGTT (with reverse complement)
Arabidopsis thaliana Haemophilus influenzae Helicobacter pylori 26695 Helicobacter pylori J99 Homo sapien Saccharomyces cerevisiae
Total Hits In Genome 30,056 1077 81 119 67,881 4027
Hits With 'Phosphatase' Present In Annotation 267 1 0 0 535 51
Percentage Of Hits With 'Phosphatase' Where Hit Is In Promoter (3000 Base Pairs Upstream) 53.9 100 0 0 28.8 52.9
Table 3 Conversions between DNA sequences and the radix 10 number system
SEQ NUMBER SEQ NUMBER SEQUENCE NUMBER SEQUENCE NUMBER
A 1 (C)3 84 AATGCT 3,301 GCTCACTG 62,003
C 4 (A)4 85 AGTGTCA 8,941 GGGAGGTGAA 388,991
(A)2 5 (C)4 340 ATGGGGT 12,281 GGGCGGAATT 679,999
(C)2 20 (A)5 341 GTAGATAA 23,003 GTTTCCTGCG 1,111,211
(A)3 21 (C)5 1,364 GCGGCTGA 32,003 GCTAAAAGGC 1,299,827
Table 4 NCBI terms correlation to gene ontology (GO) terms
14,349 Unique NCBI One Term Windows 18,934 Unique NCBI Two Term Windows 24,747 Unique NCBI Three Term Windows
GO Term Equality 160 164 73
GO Term Similarity 160 164 73
GO Phrase Similarity 561,294 6840 818
Table 5 Gene ontology (GO) term expansion with siblings
14,349 Unique NCBI One Term Windows 18,934 Unique NCBI Two Term Windows 24,747 Unique NCBI Three Term Windows
GO Term Equality 16,001 9096 4313
GO Term Similarity 16,001 9096 4313
GO Phrase Similarity 19,137,006 228,624 28,556
Table 6 Location 2-Word
08 AG
13 TG
22 TG
06 CG
==== Refs
Benson G Tandem repeat finder: a program to analyze DNA sequences Nucleic Acids Research 1999 27 573 580 9862982 10.1093/nar/27.2.573
Adebiyi E Jiang T Kaufmann M An efficient algorithm for finding short approximate non-tandem repeats Bioinformatics 2001 17 S5 S12 11472987
Landau G Schmidt J Sokol D An algorithm for approximate tandem repeats Journal of Computational Biology 2001 8 1 18 11339903 10.1089/106652701300099038
Castelo A Martins W Gao G TROLL – Tandem Repeat Occurrence Locator Bioinformatics 2002 18 634 636 12016062 10.1093/bioinformatics/18.4.634
Kolpakov R Bana G Kucherov G mreps: efficient and flexible detection of tandem repeats in DNA Nucleic Acids Research 2003 31 3672 3678 12824391 10.1093/nar/gkg617
Altschul S Gish W Miller W Myers E Lipman D Basic local alignment search tool Journal of Molecular Biology 1990 215 403 410 2231712 10.1006/jmbi.1990.9999
Gusfield D Algorithms on Strings, Trees and Sequences: Computer Science and Computational Biology 1997 Cambridge, UK: Cambridge University Press
Hauth A Joseph D Beyond tandem repeats: complex pattern structures and distant regions of similarity Bioinformatics 2002 18 S31 S37 12169528
Altschul S Madden T Schaffer A Zhang J Zhang Z Miller W Lipman D Gapped BLAST and PSI-BLAST: a new generation of protein database search programs Nucleic Acids Research 1997 25 3389 3402 9254694 10.1093/nar/25.17.3389
Ning Z Cox A Mullikin J SSAHA: A Fast Search Method for Large DNA Databases Genome Research 2001 11 1725 1729 11591649 10.1101/gr.194201
Baeza-Yates R Ribeiro-Neto B Modern Information Retrieval 1999 New York, NY: ACM Press
Kent WJ BLAT – the BLAST-like alignment tool Genome Research 2002 12 656 664 11932250 10.1101/gr.229202. Article published online before March 2002
Califano A Rigoutsos I FLASH: A fast look-up algorithm for string homology Proceedings of the 1st International Conference on Intelligent Systems for Molecular Biology, Bethesda, MD 1993 353 359
Rigoutsos I Floratos A Combinatorial pattern discovery in biological sequences: the TEIRESAIS algorithm Bioinformatics 1998 14 55 67 9520502 10.1093/bioinformatics/14.1.55
Ogawa N DeRisi J Brown P New Components of a System for Phosphate Accumulation and Polyphosphate Metabolism in Saccharomyces cerevisiae Revealed by Genomic Expression Analysis Molecular Biology Cell 2000 12 4309 4321
Advanced Content Match Engine for Sequences (ACMES)
Reneker J Shyu CR Zeng P Polacco JC Gassmann W ACMES: fast multiple-genome searches for short repeat sequences with concurrent cross-species information retrieval Nucleic Acids Research 2004 W649 53 15215469
Salaün L Linz B Suerbaum S Saunders N The diversity within an expanded and redefined repertoire of phase-variable genes in Helicobacter pylori Microbiology 2004 150 817 830 15073292 10.1099/mic.0.26993-0
Hood D Deadman M Jennings M Bisercic M Fleischmann R Venter C Moxon R DNA repeats identify novel virulence genes in Haemophilus influenzae Proceedings of the National Academy of Science 1996 93 11121 11125 10.1073/pnas.93.20.11121
National Center for Biotechnology Information (NCBI)
National Center for Biotechnology Information (NCBI) BLAST database
Gene Ontology (GO) Consortium
| 15869708 | PMC1131890 | CC BY | 2021-01-04 16:02:48 | no | BMC Bioinformatics. 2005 May 3; 6:111 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-111 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-791579978210.1186/1471-2105-6-79SoftwareMAPPER: a search engine for the computational identification of putative transcription factor binding sites in multiple genomes Marinescu Voichita D [email protected] Isaac S [email protected] Alberto [email protected] Children's Hospital Informatics Program, Children's Hospital Boston, Harvard Medical School,300 Longwood Avenue, Boston, MA 02115, USA2005 30 3 2005 6 79 79 6 8 2004 30 3 2005 Copyright © 2005 Marinescu 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
Cis-regulatory modules are combinations of regulatory elements occurring in close proximity to each other that control the spatial and temporal expression of genes. The ability to identify them in a genome-wide manner depends on the availability of accurate models and of search methods able to detect putative regulatory elements with enhanced sensitivity and specificity.
Results
We describe the implementation of a search method for putative transcription factor binding sites (TFBSs) based on hidden Markov models built from alignments of known sites. We built 1,079 models of TFBSs using experimentally determined sequence alignments of sites provided by the TRANSFAC and JASPAR databases and used them to scan sequences of the human, mouse, fly, worm and yeast genomes. In several cases tested the method identified correctly experimentally characterized sites, with better specificity and sensitivity than other similar computational methods. Moreover, a large-scale comparison using synthetic data showed that in the majority of cases our method performed significantly better than a nucleotide weight matrix-based method.
Conclusion
The search engine, available at , allows the identification, visualization and selection of putative TFBSs occurring in the promoter or other regions of a gene from the human, mouse, fly, worm and yeast genomes. In addition it allows the user to upload a sequence to query and to build a model by supplying a multiple sequence alignment of binding sites for a transcription factor of interest. Due to its extensive database of models, powerful search engine and flexible interface, MAPPER represents an effective resource for the large-scale computational analysis of transcriptional regulation.
==== Body
Background
Identifying the combinatorial logic of transcriptional regulation is key for understanding the mechanisms of development, cell commitment and differentiation and the way in which external and internal signals are converted into specific patterns of gene expression. Transcriptional regulation is accomplished by the coordinated activity of specific regulatory proteins that recognize and bind regulatory elements – short DNA motifs located in the untranscribed regions of the genes [1]. Regulatory elements such as TFBSs, enhancers, and silencers, are commonly located in the promoter region of genes, while others, such as splicing control elements, may be located within the introns or exons of a gene. As more sequence and expression data have become available, the task of understanding gene regulation has come to rely on a combination of experimental and computational approaches. Among the many bioinformatics approaches aimed at understanding the role of regulatory elements in transcriptional control, several research themes have emerged [2-5]. They include search algorithms that extract putative regulatory elements [6-15], search engines for their retrieval [16-19] and databases of experimentally characterized or computationally derived regulatory elements [20-24]. Moreover, combinations of regulatory elements that occur in close proximity to each other form cis-regulatory modules that control gene expression. Their presence suggests the existence of a combinatorial code for transcriptional regulation [25], with ample effort being devoted to developing algorithms for its elucidation [26-34].
In sequences of orthologous genes, regulatory elements that have a functional role are often conserved throughout evolution by selective pressure. This led to the development of many algorithms that use 'phylogenetic footprinting' – a method for inferring regulatory elements based upon sequence conservation of orthologous genes [35-45]. However, recent evidence [46] shows that functional elements are not necessarily located in conserved regions; this requires the development of computational methods to detect binding sites with high specificity without relying primarily on sequence conservation data.
One of the most common strategies for identifying putative TFBSs in DNA sequences relies on matching a general pattern abstracted from sequences of experimentally characterized binding sites and expressed in the form of a probability weight matrix that describes the probability distribution of the four possible nucleotides at each location [47]. Several programs, such as Patch, Match, MatInspector and TESS, rely on the nucleotide weight matrices (NWMs) of TRANSFAC – a large and frequently updated database that contains information on the transcription factors (TFs) and their binding sites in target genes [20]. The assumption underlying the construction and use of NWMs is that each nucleotide contributes independently to the binding site consensus and that the contribution of the nucleotides to the site is additive [48]. This assumption was tested experimentally in the case of binding sites for two transcription factors – the Mnt repressor protein and the mouse EGR1 protein. Nucleotides at positions 16 and 17 in the Mnt repressor protein binding site [49] and the central nucleotide triplet in the mouse EGR1 binding site [50] were systematically mutated to all possible combinations and the binding affinity of the respective TFs (or its mutants) for these sites was determined. The results pointed out that the assumption of independence of nucleotides within a site is not entirely accurate, but that although NWMs do not capture the dependencies between nucleotides within a site they represent a good enough approximation for modeling it[48,49]. Nevertheless, it is generally recognized that using NWMs to identify putative TFBSs often leads to the retrieval a very high number of false positives [47].
In this work, we rely instead on Hidden Markov Models (HMMs) as a more accurate probabilistic method to model the sequence of nucleotides within a binding site that in addition to abstracting the probability distribution of the nucleotides at each site can also model insertion or deletions and retrieve fragment matches to the model in the search procedure [51]. HMMs are statistical models able to represent stochastic sequences of symbols and can be used to generate sequences that conform to a given model, or to determine the likelihood that a given sequence was generated by that model. HMM techniques have become the basis of many bioinformatics applications for recognizing conserved domains in amino acid sequences or gene features in DNA sequences [51,52]. Several publicly available implementations such as HMMER [51], SAM [53] and Meta-MEME [54] build HMMs based on multiple sequence alignments and, among other functions, search input sequences for matching domains. The HMMER package was used to generate the large collection of annotated protein domains of the Pfam database [55] and consists of several modules for which the source code is available, well commented and easily modifiable. HMMs were previously considered for modeling and searching for TFBSs. The reports so far were either theoretical in nature [50,56], where not extended to genome-wide searches [57], or focused only on a small number of transcription factors [58-60]. A HMM for CREB binding sites was used to scan upstream sequences of 10 kb in length from the human and mouse genomes [58], while a Markov model for the hepatocyte nuclear factor 4 (HNF4) was used to scan sequences between positions -500 to +100 relative to the transcription start site of confirmed genes in the human genome [59]. Recently, a method for identifying nuclear hormone receptor bindings site was developed based on the use of classification HMMs [61]. However, to date no study used HMMs of multiple TFBSs in a large-scale search across three or more genomes.
The work described in this paper focuses on developing methods to generate accurate and complete information on putative TFBSs in genes across multiple genomes (H. sapiens, M. musculus, D. melanogaster, C. elegans and S. cerevisiae in the initial implementation). Our methodology relies on combining the information on experimentally determined binding sites contained in curated databases such as TRANSFAC and JASPAR with the pattern matching power of HMMs. As TRANSFAC and JASPAR provide manually curated representative nucleotide sequences of binding sites for most TFs included in the databases, we leveraged this information to recreate the alignments used to calculate the NWMs, and we used them as input to HMMER instead. We thus generated a library of 1,079 HMM profiles containing one model for each TRANSFAC matrix or factor entry (see below) or JASPAR matrix for which alignments of binding sites were available. The performance of selected models was evaluated by their ability to retrieve experimentally characterized binding sites, and the sensitivity and specificity of the method were assessed in a large-scale comparison with a NWM-based method using synthetic data. A flexible interface was implemented (and is publicly available at ) that allows the user to search a sequence in FastA format or a gene and its orthologs across five genomes against the library of 1,079 HMM models or against a model built by the user starting with a multiple sequence alignment of binding sites. Although the tools developed for this work were used to find putative TFBSs, they are directly applicable for identifying other types of regulatory elements for which sequence alignments are available.
Results
The MAPPER HMM library
HMMs were built using multiple sequence alignments of binding sites compiled from the TRANSFAC [20] and JASPAR [21] databases. TRANSFAC provides two sources of information regarding the binding sites for TFs: nucleotide sequences of binding sites referenced in the description of the TRANSFAC matrices that were optimally aligned and used to derive NWMs (designated below as matrix-derived alignments and catalogued with accession numbers starting with "M"), and nucleotide sequences of binding sites referenced as part of the description of the TFs – also referred to as "factors", used to extract alignments designated below as factor-derived alignments and catalogued with accession numbers starting with "T". By parsing the TRANSFAC flat files (see Methods for details) we obtained 402 alignments corresponding to matrices and 588 alignments corresponding to factor entries. In addition, 89 alignments were obtained from data downloaded from the JASPAR database. Thus, the total number of alignments used to build HMMs was equal to 1,079. Figure 1A shows the distribution of the length of the models and of the number of sequences and size of the nucleotide matrix used to train them. The models have an average length of 10 nucleotides and were trained on an average of 22 sequences. TRANSFAC assigns a quality value to the sites used to build the factor-derived models, based on the existing biological evidence of the binding (see Figure 1B legend for details). The distribution of the median and average quality of the sites used for the factor-derived models suggests that the large majority contains high quality sites (categorical values smaller than 4).
The 1,079 models retrieved correspond to 888 transcription factors entries with distinct names in TRANSFAC and JASPAR. Table 1 in Additional File 1 lists the names of all TF entries in the two databases for which HMMs were built and the models that describe them. It is important to note that different databases (or even the same database) often use different names for the same TF or for isoforms of the same TF (e.g. p65, RelA; HNF-1, HNF-1alpha); nevertheless, for the purpose of this paper, entries with different names were considered as distinct.
While the matrix-derived models are generated by combining binding sites for homologous factors from multiple organisms, every factor-derived model and JASPAR model is derived from sites from a single organism. Our search engine does not place restrictions on the use of a model associated with a TF from one organism when searching a sequence from a different organism. The rationale for this is that ortholog TFs from different organisms usually show very high structural and functional conservation that extends to their binding site specificities. This allows the user to use all available models for a given TF when searching, and to evaluate a posteriori whether the resulting hits are significant.
Evaluation of the method
Evaluating the sensitivity and specificity or our method compared to other commonly used ones is not straightforward, for a variety of reasons. First, in order to measure the false positive and false negative rates we would need to be able to reliably classify occurrences of the motif (also referred to as "hits") into "true" and "false" positive categories. This is obviously impossible by computational means, and too expensive and time consuming to pursue experimentally for a large number of transcription factors and binding sites. On the other hand, the experimental data sets that make available genome-wide positions of "true" hits pertain to a limited number of factors of interest and usually report a region for which binding was detected and not the precise locations and sequences of the binding sites [62-65]. For a limited number of well-characterized factors, collections of binding sites were compiled from the literature (see below) but the total number of such sites is still too small to enable a statistically significant comparison. As such sites usually come from promoter regions that are rich in regulatory sites, we were forced to use short flanking sequences to avoid including extraneous additional sites. The results of testing a method on such short sequences are not entirely predictive of its performance when long genomic regions are used as input, as in most experiments. Moreover, for all these datasets and even for other ones that contain binding data for multiple factors [66], no information is available about "false" hits (i.e. sites that match a consensus but are not functional). Although it is clear that only a large-scale biological validation of the predictions of this method can provide a definite estimation of its performance, such goal is beyond the scope of this paper.
Given these constraints, we performed three types of evaluations of our method. First, for three different factors we performed control runs to determine if our method is able to correctly identify a total of 17 experimentally characterized binding sites in 9 different genes with the exact sequence and at the positions reported in the literature. Secondly, for a collection of 89 experimentally characterized sites for six other transcription factors, we determined the number of sites retrieved and the percentage of false positives reported by our method and compared them with the results of four other methods: Match [19], Patser [9], LMM [67] and ScanACE [6]. Finally, we conducted a large-scale evaluation based on synthetic data for 491 models in our database in order to compare, using modified ROC curves, the sensitivity and specificity of our method with the ones of a NWM-based method (in this case Match).
Control runs
To obtain a preliminary evaluation of the performance of our method in detecting TFBSs we used control sequences consisting of promoter regions of genes in which binding sites for specific factors were determined experimentally and were shown to play a role in the regulation of the gene. We selected as controls HMM models for three transcription factors – p53, Su(H) and MEF-2 and used them to scan control sequences chosen so that they contain experimentally determined TFBSs for the factors whose nucleotide sequence were not included in the multiple sequence alignments used to train the models.
A HMM corresponding to p53 was built based on the alignment for matrix M00761 of TRANSFAC, and was used to scan promoter regions from the following genes: the human 14-3-3 sigma protein gene for which two p53 binding sites, BDS-1 and BDS-2, were characterized [68], the mouse cyclinG1 gene that contains one p53 binding site [69], and the mouse B99/Gtse1 gene, encoding the G two S phase expressed protein 1 for which a p53-responsive element containing three p53 half-sites was reported [70]. With the exception of one half-site in the latter sequence, HMMER retrieved all the described p53 binding sites at the positions indicated in the literature, with significant scores and E-values (for the definitions of these parameters see the Methods section). Model M00234 for the Drosophila Su(H) (Suppressor of Hairless) was used to scan the following sequences: the promoter of Drosophila him [71] containing four Su(H) binding sites that were identified computationally and confirmed experimentally, the sequence of an enhancer containing three Su(H) sites located 3.5 kb upstream of the first exon in Drosophila yan [72] and the promoter sequence of the human erbb-2 gene containing one Su(H) binding site [73]. While the sites in Drosophila him where identified with positive scores, the binding sites in the last two sequences were reported with negative scores, pointing out that, as mentioned in the HMMER documentation [74], real matches can in some cases have negative scores (one of the three Su(H) sites in the Drosophila yan enhancer was missed). Model T00505 corresponding to human MEF-2 (myocyte-specific enhancer factor 2A) was used to search the promoter sequences from the following genes, each containing one well characterized MEF-2 site: Drosophila Actin57B gene [75], mouse mef2c gene [76] and human c-jun [77]. The characterized MEF-2 binding sites in these sequences were also retrieved by HMMER (with a negative score for c-jun). In total our method identified correctly 15 out of the total 17 binding sites (the complete list of hits can be accessed following the appropriate link from Additional File 1).
Small-scale evaluation
The purpose of this evaluation was to measure the performance of our method compared to other widely used tools for the computational detection of TFBSs, using a high-quality dataset of experimentally validated binding sites. The criteria we used to compare the TFBS detection methods in our experiment are the following: first, we required a method to be able to detect all the experimentally validated sites (true positives) in the input sequences, except for one at most. Next, we counted the number of hits not corresponding to true sites (false positives) having a score greater then the lowest-scoring true positive, and we expressed it as a percentage of the sum of the number of true positives and false positives. According to this definition, the false positives represent those hits that cannot be separated from the true ones on the basis of their score, since raising the score threshold to exclude them would cause the method to miss some true positives.
The dataset used for this evaluation contained 110 experimentally characterized binding sites for six factors: E2F, the estrogen receptor (ER), the glucocorticoid receptor (GR) and the hepatocyte nuclear factors 1, 3 and 4 (HNF-1, HNF-3 and HNF-4) previously compiled from the literature [78-80]. For each factor the dataset contained a 50 bp long sequence extracted from a given gene in which the binding site was always located in the same positions throughout the dataset (e.g. for all sequences in the E2F dataset the actual site was found between positions 20–31). The length of the sites was 11 for HNF-3, 12 for E2F, 13 for ER and HNF-1, 14 for GR, and 15 for HNF-4. Adopting this consistent format for the datasets allowed us to define a true positive as a hit that overlaps at least 60% of the sequence of the site and a false positive as any other hit retrieved in the sequence. Hits retrieved on both strands that overlapped more than 75% were considered equivalent and only the best scoring one was reported.
We included in this evaluation four other methods for TFBS detection: Match [19] and Patser [9] that scan a sequence using a supplied NWM, LMM (Local Markov Method) [67] that uses a p-value-based scoring measuring the similarity of the hit to the known binding sites for the factor and its contrast to the local genomic context, and ScanACE [6] that scans a sequence for matches for a given motif using a scoring method based on a maximum a priori log likelihood score. In order to conduct a fair evaluation for each factor we used as input either the HMM (for our method) or the corresponding NWM (for Match, Patser and LMM – see exceptions below) that were built on the same alignment. This alignment was also used directly as input for ScanACE. For three factors (ER, GR and HNF-1) the matrix used in the analysis was not found in the LMM matrix library, therefore we used the closest matrix available as judged by examining its consensus sequence. To preserve this correspondence we were in some cases forced to use a sub-optimal model in the HMMER run; in these cases we also included an alternative HMM model in the analysis (the best one available to HMMER for the given factor). We also filtered the collection of binding sites in order to eliminate the ones that appear in the training set (i.e., that were used in the alignments on which the matrices and all HMMs used were built), resulting in a dataset that contained a total of 89 sites. Information regarding the binding sites in the alternative matrices used for LMM is not available in TRANSFAC, so it is possible that in these cases the alternative matrix used overfits the dataset.
The results of the small-scale evaluation are presented in Table 1. The complete listing of the hits found (including their position, sequence and score) as well as the consensus of the models and matrices used is provided in the Additional File 2. Table 2 summarizes the results: for each method tested we report the overall percentage of true positives identified, and the lowest and highest percentage of false positives in each of the six datasets. In several cases, two of the methods tested (LMM and ScanACE) did not reach a sufficient number of true positives, so their performance is not directly comparable to the one of the other methods. Their performance in terms of false-positive hits suggests that they are too specific, and therefore prone to missing true sites. The other two programs that we tested against, Match and Patser, show a minimum false-positive percentage of 12% and 3% respectively. In contrast, HMMER reaches a minimum value of 0 false-positive hits, while detecting all (or almost all) true sites in all cases. The highest percentage of false positive hits ranges from 16.7% for HMMER (selecting the optimal model) to 43.7% for Patser. While on individual models other methods might perform better in particular cases, these results indicate that HMMER is powerful enough to detect the target binding sites in all the datasets tested, and that its sensitivity-specificity trade-off is consistently better than those of the other methods.
Large-scale evaluation
To add to the rigor of this analysis and perform it in a context closer to a real-life biological investigation, we resorted to a computational evaluation based on synthetic data generated similarly to the technique described by Barash et al. [81]. This allowed us to compare the proportion of false positive hits returned by our method and by a TFBS prediction program based on NWMs, in this case Match [19]. We applied the following procedure to the 491 HMMs in our database (corresponding to all TRANSFAC matrix-derived and JASPAR-derived models) for which a corresponding NWM built using the same multiple sequence alignment was available: we generated a random nucleotide sequence of a fixed length (50,000 bp) and we inserted in it, at random locations, 100 "synthetic" binding sites. These binding sites were not generated by sampling from the matrix nor from the HMM as that could have conferred an advantage to one of the two methods. Instead the sites were generated by sampling from the alignment with an algorithm designed to make the test as fair as possible for both methods by preserving both the dependencies between nucleotides in the sequences and the core matrix if it exists (see Methods for details). Before each hit was planted the random sequence was scanned to eliminate any other occurrence of its sequence that might have been present by chance. We defined our planted hits as the "true positives", while any other reported instance of the pattern was considered a "false positive". We then scanned the resulting nucleotide sequence with the Match program and with HMMER, we repeated the experiment 20 times for each model independently and we averaged the results to eliminate fluctuations due to randomization.
We measured the performance of the two methods using modified ROC curves – ROC50 curves [82] and compared the areas obtained with both methods by using a Bonferroni-corrected Wilcoxon signed rank test. Out of the original 491 models tested 105 were eliminated due to the fact that the average ROC50 area for either method was smaller than 0.25, leaving 386 filtered models on which the comparison was conducted. At a significance level of 0.05 and with a Bonferroni correction of 491, our method performed better than Match for 96% of the models for which the result was significant (71% of the filtered models), suggesting that it can provide a better trade-off between sensitivity and specificity that translates into being able to retrieve more true positives with fewer false positives. The results of this test are presented in Table 2 of Additional File 1. It should be noted that in many cases the length of the matrix used by Match differs from the length of the HMM even though the same alignments were used to construct both. This situation confers an advantage to Match because the planted hits have the same length as the alignment.
We then analyzed the results of these runs to determine, for each method and model, the percentage of true positive hits within the first n reported hits (the greater this percentage, the more sensitive the method is), and the amount of false positives expressed as a percentage that needs to be accepted in order to retrieve the first m true positive hits (the smaller this percentage, the more specific the method is); these tests are referred below to as the first and second TPP (true positive proportion) tests. The values used for n and m were 30, 50, 70 and 90. Tables 3 and 4 of Additional File 1 show these results for all 386 filtered models tested. While the ROC tests give an indication of the overall performance of the method, the TPP tests assess its performance in the "early" and "late" stages of the search, respectively. By the stringent criteria used for comparison (see Methods) in the first TPP test our method performs better for 78% of the cases for which a difference was noted suggesting that these models are better suited for retrieving a relatively small number of hits that are true positives (as for example when analyzing the promoter of a given gene) while reporting a minimal number of false positives. In the second TPP test our method performs better for 85% of the cases in which a di fference in performance between the two methods was noticed, suggesting that these models can retrieve a large number of hits while limiting the number of false positives, as for example in the case of a genome-wide search for putative TFBSs for a given factor. It should be noted that since each of the three tests measures a different aspect of the performance of the method, the list of models that perform better in each case might not entirely overlap.
The MAPPER interface
We designed and implemented a web-based application, called MAPPER (Multi-genome Analysis of Positions and Patterns of Elements of Regulation), to facilitate the retrieval of putative TFBSs in a given sequence based on the library of 1,079 HMM models described above. The interface takes as input a gene identifier (e.g. NCBI Gene ID, RNA accession number) or a user supplied sequence in FastA format. The user then selects the models to be used (all, TRANSFAC or JASPAR only) and has the option to build his/her own model starting with a multiple sequence alignment of binding sites in FastA format. The search can be performed on the entire gene region flanked by a user-specified distance upstream and downstream, on a specified gene region (promoter, introns, exons, 3'-UTR) or within a certain distance upstream of the ATG or the start of the transcript (Figure 2). If a gene identifier is provided the program will also display the actual nucleotide sequence scanned (in FastA and Genbank format), a useful option in the case of those genes for which discrepancies exist between different annotations.
The user can choose to display all hits for a given sequence, or only the hits for factors that are common across the orthologs of that sequence (if present in the HomoloGene database). The output of the system is the list of putative hits found in the specified conditions – default score and E-value thresholds are 0 and 10 respectively (Figure 3). For each hit the system displays the model used to retrieve it, its location, score and E-value and (in a pop-up window) the alignment between the model and the sequence at that site. The hit set can be sorted by position (from the ATG or the start of the transcript for genes supplied via an identifier, or from the beginning of the sequence for FastA sequences), by model name, model accession, score or E-value. In addition the results page highlights adjacent sites (situated within 50 bp from each other) retrieved for TFs that are known to physically interact with each other (as annotated in the TRANSFAC database), and also the classes of TFs for which putative sites were found. For each TRANSFAC factor-derived or JASPAR-derived model the organism in which the factor was described is specified. The user can choose to highlight hits in evolutionarily conserved regions representing the most conserved elements between sets of organisms provided by the UCSC Genome Browser annotations (see Methods for details). Figure 4A shows a graphical representation of the position and orientation of the hits listed in Figure 3. Arrows are drawn to scale and in some cases represent the sum of overlapping sites for the different transcription factors that are listed above or beneath them. Hits occurring in evolutionarily conserved regions are displayed in red. While the results page in Figure 3 lists and sorts all putative TFBSs independently from each other, the graphical representation in Figure 4A makes it easy to identify regions in which more than one model (corresponding to the same factor or to closely related factors) detects a binding site making it, therefore, more likely to represent a true binding site. The set of hits can also be exported and displayed in the UCSC Genome Browser using its "custom tracks" feature (Figure 4B). This allows the user to view the TFBSs in the general context of the genomic region in which they appear and to take advantage of the powerful visualization tools of the UCSC Genome Browser in order to highlight important features of the genomic region.
Figures 3 and 4 present the output of MAPPER when the human MCM5 gene (Entrez Gene ID 4174) is used as an example. The promoter of the human MCM5 gene contains multiple experimentally characterized binding sites for the E2F transcription factor. These binding sites were retrieved by our search, and were found to be conserved across the human, mouse and Drosophila MCM5 orthologs. MCM5 genes code for proteins involved in the initiation of DNA replication [83], and are members of the MCM family of chromatin-binding proteins that participate in cell cycle regulation. The E2F family of transcription factors plays a critical role in the control of cell proliferation and consists of six factors, E2F-1 to E2F-6, that heterodimerize with two other subunits, DP-1 and DP-2; the activity of these complexes is modulated by the retinoblastoma tumor suppressor protein (pRB) that binds E2F [84]. TRANSFAC and our database contain multiple models describing the binding sites in target genes characterized for different combinations of E2F and DP proteins, complexed or not with pRB. Below, we refer to these models generically as "E2F" models given the fact that, while the transcriptional role of E2F family members is different given the identity of the E2F and DP moieties that forms the complex [85], no specificity has been detected in vivo for the association of particular complexes to known E2F-regulated promoters [86,87]. Experimental evidence showed that the upregulation of the human MCM5 gene in response to growth stimulation is mediated by the binding of E2F to four sites within the MCM5 promoter, and that mutations in these sites abolish this response [88]. The four E2F binding sites consist of two sets of overlapping sequences running on opposite strands and were mapped by RNase protection assays to positions -194 to -183 and +2 to +13 respectively, relative to the start of the transcript [88]. In our search, three models for the E2F family (MA0024 for E2F, T05206 for E2F-4:DP-1 and T05208 for Rb:E2F-1:DP-1) retrieved all four E2F sites at the location and in orientation described in the literature (Figures 3 and 4). To simplify the display in these figures and to highlight the four E2F binding sites retrieved by the three models, a more stringent set of parameters was used for the query (500 bp upstream of the ATG, score > 3, E-value < 6.8). Figure 3 shows the list of all TFBSs retrieved given these input parameters with the four E2F binding sites described by Ohtani et al. boxed, as well as the list of factors for which putative binding sites where found also in the other two MCM5 homologs selected (mouse and Drosophila MCM5). For each hit in the listing the model identifier is displayed as a double link, to a pop-up window showing the match between the sequence and the model (Figure 3) and to a separate page giving detailed information regarding the model including its length, the number of sequences in the training set, associated models, HMM logo [89], and the references used to build the alignment (Figure 5).
Discussion
Our method offers several advantages over other similar tools, with respect to the extent and quality of the models it uses, its sensitivity and specificity, and the overall functionality of the web-based interface.
MAPPER includes a large database of profile HMMs corresponding to 888 TF entries that was built using the data provided by the TRANSFAC and JASPAR databases, consisting of sets of experimentally validated binding sites for several hundred TFs. In addition to the models based on optimal alignments provided by TRANSFAC and JASPAR, our database includes a large number of additional models generated by extracting the representative motif from the "raw" binding site sequences contained in TRANSFAC with the program MEME (Multiple Expectation-maximization for Motif Elicitation) [90] that usually provide a tighter definition of the binding site specificity. As a result our method can make use of a larger number of models that provides an increased ability to detect putative binding sites. In many cases, several models are available for a single TF; in addition to increasing the probability of detecting a binding site for the factor, this redundancy also allows the user to evaluate whether a putative site is a "true" one (if it is detected by multiple models) or a potential false positive. Although the Plan7 architecture on which HMMER is built does not take into account the dependencies between the nucleotides within a site and, similarly to NWMs, weights each state independently [74], several features of the HMMER modeling and search procedure confer an added level of generality to HMMs as compared to NWMs built upon the same alignments. First, profile HMMs model insertions, deletions and allow fragment matches to the model [74]. This property becomes significant in the case of those TFs that bind to sites comprised of half sites separated by spacer regions of variable length (as for example nuclear receptors); while insertion and deletions are rare in the functional half sites they can occur with higher frequency in the spacer regions that are much more divergent [91]. Moreover, allowing fragment matches to the model ensures that binding sites that may contain a well defined half-site and an imperfect one or half-sites separated by long spacers can still be retrieved as fragment matches to the model. Secondly, all hits returned by HMMER are subject to a bias composition filtering based on a second null model that is computed for each alignment and leads to a rescoring of the hits penalizing the ones for which the nucleotide composition is biased [74]. Even in equal performance conditions, as could be the case for short alignments that do not allow insertions or deletions or fragment matches to the model, this filtering alone would still confer an advantage to using HMMER versus NWMs.
Using profile HMM for modeling bindings sites has also limitations. To build a model HMMER converts the observed counts in the training set into probabilities by combining the actual counts with pseudocounts from priors, in this case single-component Dirichlet priors [74]. The latter can have a more pronounced effect and can bias the model in the case in which the number of sequences in the training set is low. These cases would be difficult to model accurately by any statistical approach (including NWMs) and their suitability for the desired analysis will have to be evaluated case by case by the user. To facilitate this, we report for each model the length, the number of sequences in the training set, the HMM consensus (for matrix-derived models), the HMM logo [89], the other associated MAPPER models, and the references used to curate the binding sites used in the training set. The presence of models trained on small number of sequences usually does not represent a problem, as in the large majority of cases MAPPER makes available multiple models for any given TF.
We compared the predictive performance of our method with that of several other similar computational tools, by testing them on a dataset of over 100 experimentally determined binding sites as well as on synthetic data. The factors for which experimentally characterized sites were tested were selected so that they bind sites with different overall organization and belong to different categories of TFs such as fork head TFs (E2F and HNF-3), MADS box (MEF-2), helix-turn-helix/homeo domain (HNF-1), Cys4 Zn finger of nuclear receptor type (ER, GR and HNF-4), beta-scaffold factors with minor groove contacts (p53), and IPT/TIG domain (Su(H)). As presented in the results section our method correctly identified 15 out of 17 binding sites reported in the literature for p53, Su(H) and MEF-2. Moreover, from a collection of 89 binding sites for six other TFs (E2F, ER, GR, HNF-1, HNF-3 and HNF-4) our method identified 98% to 100% of the true positives with false positive ratios ranging from 0% to 16% (or 24% when a non-optimal model was used). The other methods tested (Match, Patser, LMM and ScanACE) either retrieved a comparable number of true positives at the expense of higher false positive ratios (as for example Match and Patser) or attained lower false positive ratios at the expense of missing a high number of true positives (as for example ScanACE and, in some cases, LMM).
Although encouraging, the results of this evaluation cannot be easily extrapolated to a scenario in which very long sequences (up to an entire genome) are scanned with hundreds of models. In order to make our analysis more general, we performed a large-scale evaluation using synthetic data for 491 models in our database (46% of the total number) for which a HMM and a NWM was available and compared the performance of our method with the one of Match in scanning sequences of 50 kb in length. Models that performed very poorly for one method or the other or both were filtered out and among the remaining 386 models 74% showed a statistically significant difference based on a Bonferroni corrected Wilcoxon signed rank test. Among the latter our method performed better in 96% of the cases.
However, we recognize that the entire enterprise of TF binding site annotation is burdened by the challenge of a robust definition of what constitutes a true positive, even those based on binding studies. For example, Tronche [92]and others note the evolutionary conservation of binding sites for genes transcribed in tissues that do not even express the transcription factor. In these instances as in others, computational or even biochemical binding assays are only a first step on the path to focused functional studies.
Finally, the MAPPER interface offers several advantages over other similar tools. It accepts as input a user-supplied FastaA sequence or a gene identifier for any annotated gene in the human, mouse, fly, worm or yeast genomes. It can use in the search all or each of the different categories of models in our database (based on TRANSFAC matrices, TRANSFAC factors or JASPAR matrices) or a model built on a multiple-sequence alignment supplied by the user. The results are presented in a simple yet comprehensive manner providing detailed information regarding the gene, the sequence scanned and the putative sites retrieved, and powerful graphical and export options facilitate the analysis and the interpretation of the results.
Conclusion
The purpose of our work was to establish a methodology for the detection of TFBSs in multiple genomes endowed with enough sensitivity and specificity to be effective in large-scale analysis (such as generating a whole-genome map of binding sites for a collection of TFs). Accomplishing this requires a large library of high-quality TFBS models and a computational method able to reliably detect instances of the models in a given DNA sequence.
The model library used by our program was created from the data contained in the TRANSFAC and JASPAR databases, with a procedure that generated over a thousand high-quality models. The computational method we implemented relies on HMM profiles built from nucleotide sequence alignments. Using HMM profiles instead of NWMs is a powerful way for capturing the characteristics of a binding site and several observations suggest that our method is reliable and performs well. First, HMM profiles for selected factors retrieved binding sites in the promoter regions of genes used as controls with high specificity, as described in the Results section. Secondly, on an extended collection of experimentally characterized TFBSs, our method identified 98% to 100% of the true positives with a false positive ratio that was consistently smaller than the ones reported by the other methods tested. Thirdly, ROC and True Positive Proportion tests performed on a large number of models for which both a NWM and a HMM was available showed that in the majority of the cases our method performs significantly better than a NWM-based program such as Match. This translates into an increased ability to detect true binding sites while reducing the number of false positive sites reported. Finally, our method takes advantage of a larger set of models for a given TF, and this results in an increased ability to detect true hits.
The web-based interface was design to maximize usability and to facilitate the analysis of the retrieved hits; it has simple and flexible input requirements, a clear and comprehensive display of the results and powerful graphical and export options.
The current work and its future extensions make available a novel and reliable method for the identification of TFBSs that, used in combination with existing molecular genetic information and biological validation, represents a powerful tool for understanding the logic of combinatorial regulation. The search engine can be seen as the foundation for more advanced applications, such as highlighting patterns of TFBSs involved in the regulation of particular genes, assessing the conservation of such patterns across multiple genomes, or identifying the TFBSs overrepresented in a set of coexpressed genes.
Methods
Genomic sequences, homology and conservation information
Genomic sequences and annotations were downloaded from the UCSC Genome Bioinformatics site [93,94] and correspond to the following releases: Homo sapiens – hg17, Mus musculus – mm5, Drosophila melanogaster – dm1, Caenorhabditis elegans – ce2 and Saccharomyces cerevisiae – sg1. Homology information was obtained from the HomoloGene database Build 38.1 [95], containing clusters of genes that share a consistent ortholog relationship across three or more organisms [96]. Evolutionary conservation information was obtained from the UCSC Genome Browser [97] and consisted in the location and scores of the most conserved elements identified using the phastCons program [98] between the following sets of organisms respectively: human, chimp, mouse, rat, dog, chick, fugu and zebrafish; Drosophila melanogaster, D. yakuba, D. pseudoobscura and A. gambiae; Caenorhabditis elegans and C. briggsae; Saccharomyces cerevisiae, S. paradoxus, S. mikatae, S. kudriavzevii, S. bayanus, S. castelli and S. kluyveri. The option "highlight hits in evolutionarily conserved regions" of the results page of the interface emphasizes hits that fall in or within a distance of 100 bp upstream to 100 bp downstream of these elements for each appropriate genome.
Generating the multiple sequence alignments of binding sites
The flat files of TRANSFAC Professional version 8.1 were parsed to extract two types of alignments: nucleotide sequences used to generate the TRANSFAC NWMs and the nucleotide sequences referenced in the description of the TRANSFAC factors (see below). The vast majority of matrix entries in TRANSFAC lists the accession numbers of the factor(s) associated with that matrix (multiple factors are usually orthologs from different organisms) and the accession numbers of the nucleotide sequences used to generate the matrix referred to below as "sites". Moreover, for each factor TRANSFAC lists which organism the factor belongs to and the accession numbers of the sites described for the factor in target genes. One factor can be linked with more than one matrix, and more than one matrix can describe the same factor. Not all matrices have associated site identifiers, and, more importantly, not all factors that have associated sites were used to build NWMs. Therefore, to extract the maximum amount of information, the TRANSFAC files were parsed following not only the links from "matrices" to "sites" but also the links from "matrices" to "factors" and from there to "sites". We called the alignments retrieved following the links from "matrices" to "sites" matrix-derived alignments. These were optimal multiple sequence alignments that were used as such to build HMMs called matrix-derived models and having accession numbers starting with "M". Nucleotide sequences retrieved following the links from "matrices" to "factors" and from there to "sites" were first processed in order to extract the underlying motif using the MEME program [90] downloaded from [99]. For each set of sequences, the MEME search was conducted separately on the forward and on the forward and reverse strands and the best motif was selected taking into account its length and E-value; this selection was also verified by manual curation. The resulting MEME alignments, called factor-derived alignments, were used to build HMMs called factor-derived models that have accession numbers starting with "T".
Motifs extracted using Gibbs sampling from nucleotide sequences of binding sites and used to build the JASPAR matrices were extracted by parsing the matrix site files downloaded from [100]. The resulting alignments, called JASPAR-derived alignments, were checked against the Jaspar matrices and used to build HMMs called JASPAR-derived models designated with accession numbers starting with "MA". The accession numbers of the HMM models are the same as the corresponding entries in the TRANSFAC and JASPAR databases. To estimate the number of TFs that have corresponding models in MAPPER, we counted them as distinct if they had different names, although in several cases in TRANSFAC and JASPAR entries with different names may refer to the same TF or TF family, or slightly different names may refer to isoforms of the same TF.
Generating HMMs from alignments using HMMER
Profile Hidden Markov models were generated using the HMMER package (version 2.2 August 2001) available at [74]. The null model used to generate the models employed equal probabilities for all four nucleotides and took into account the fact that TFBSs can occur frequently throughout the sequence scanned. Therefore we used in the null model a p1 value for the G→G transition controlling the expected length of the target sequences [74] equal to 0.98 instead of the default value of 0.999, thus assuming that two sites for the same TF may occur 50 bp and not 1000 bp apart as in the default model. This significantly decreased the likelihood of retrieving true positive hits with negative scores (S.R. Eddy, personal communication).
The HMMER function hmmpfam searches a sequence or a database of sequences against a library of HMM models, and characterizes each hit it returns by two parameters: the score and the E-value. The score is the logarithm in base 2 of the ratio P(seq|HMM)/P(seq|null), where P(seq|HMM) is the probability of the target sequence according to the HMM model and P(seq|null) is the probability of the sequence according to a null model distribution. The greater the score the better the match between the hit and the model is. The E-value, computed with respect to the number of the sequences in the database queried, is a measure of the expected number of false positives that will have scores equal to or larger than the score of the hit. The smaller the E-value, the more significant the hit is.
HMMER control runs
First, a qualitative evaluation of the performance of the method was carried out by searching promoter sequences of selected genes that contain well characterized binding sites for specific TFs against the HMMs built for these factors. The factors selected as controls were mouse p53, human MEF-2 and Drosophila Su(H) – Suppressor of Hairless for which the alignment files M00761, T00505 and M00234 respectively, were used to construct and calibrate HMMs. For the promoter sequences of the genes used as controls the nucleotide positions and sequences of the characterized binding sites for these TFs were available in the literature and were compared with the ones of the hits returned by HMMER.
Small-scale evaluation
As a starting point for this analysis, six datasets were prepared containing a total of 110 experimentally characterized binding sites for the following transcription factors: E2F, the estrogen receptor (ER), the glucocorticoid receptor (GR), the hepatocyte nuclear factors HNF-1, HNF-3 and HNF-4. The sequences for 27 E2F and 25 ER binding sites, located between well defined positions in the center of a 50 bp sequence containing flanking regions from the corresponding genes [78], were downloaded from [101]. For consistency and to facilitate the analysis of the results the remaining datasets were processed and written in the same format. The sequences of 16 GR binding sites were downloaded from [79,102]. 19 bindings sites for HNF-1, 11 for HNF-3 and 12 for HNF-4 [80] were obtained by parsing the datafiles at [103]. For these datasets the sequence of the sites listed in the Gibbs sampling log files were matched to the original fasta sequences and the flanking nucleotides extracted in a sequence of 50 bp total.
The methods used for comparison were accessed as follows: the Match code was supplied with the TRANSFAC professional 8.2 suite [19], Patser [9] was used at [104], LMM [67] was downloaded from [105], and ScanACE [6] was downloaded from [106]. For Match a value of 0.7 was used for both the matrix and the core similarity thresholds. Patser, ScanACE and the hmmpfam function of HMMER were used with default parameters. LMM was used with a window size of 15 and default values for the other parameters.
For each dataset we used the NWM provided by TRANSFAC as input for the NWM-methods and its corresponding HMM model as input for HMMER. An alternative, better performing HMM model for the factor (designated with accession numbers starting with "T") was always included for HMMER. In three cases, for LMM that contains only the publicly available TRANSFAC matrices we had to use the closest available matrix for the factor as an input. The following matrices and corresponding HMM models were used for this analysis: for E2F V$E2F_02 (equivalent to M00050) and T05206; for ER V$ER_Q6_02 (equivalent to M00775), T00258 and V$ER_Q6 (for LMM); for GR V$GR_Q6_01 (equivalent to M00921), T05076 and V$GR_Q6 (for LMM); for HNF-1 V$HNF1_Q6 (equivalent to M00790), T01211 and V$HNF1_01 (for LMM); for HNF-3 V$HNF3ALPHA_Q (equivalent to M00724) and T00371; for HNF-4 V$HNF4ALPHA_Q6 (equivalent to M00638) and T00372. For each factor the binding sites included in the test datasets were checked against the ones in the training set on which the corresponding matrix and HMMs were built. 21 sites that were in common were eliminated from the test set resulting in a filtered dataset of 89 binding sites that can be downloaded following the appropriate link from Additional File 1. No information regarding the exact sites used to build the alternative LMM matrices was available in TRANSFAC so these matrices may very well overfit the test set.
The output of each method was parsed to identify the true and the false positives among the hits retrieved. Hits retrieved on both strands and overlapping more than 75% were counted as one hit. True positives were defined as hits that are either contained in or overlap at least 60% of the sequence of the known binding site; all other hits were considered false positives. For each run the number of distinct sequences containing at least one true positive was reported. Hits were sorted by score and the percent of false positives was calculated as the ratio between the false positives and the sum of false positives and true positives that have to be retrieved until at least one true positive was found for each of n sequences from the dataset. The value of n was chosen for each dataset based on the following rule: if in four cases corresponding to three different methods (the two HMMER cases, Match and Patser) the maximum number of unique sequences was identified, we used this value as a cutoff. Otherwise we used the value immediately below the maximum number. This percentage was not computed for methods that missed more than 2 sequences from the dataset.
Large-scale evaluation
We compared the sensitivity and specificity of our method against those of a NWM-based method (Match) by carrying out a large-scale analysis of their performance on synthetic data using 491 models in our database corresponding to the TRANSFAC and JASPAR matrices for which both a NWM and a HMM, built from the same multiple sequence alignment, were available. For each of these models we generated a 50,000 bp random sequence and we planted 100 "synthetic" binding sites into it, at random locations [81]. Before a hit was planted the random sequence was scanned to eliminate any potential matches that could occur by chance. The algorithm used to generate the synthetic binding sites builds a simple Markov chain by reading the original multiple sequence alignment and represents each step in the chain as a 6-by-6 matrix of transition frequencies (the states include A, C, G, T, N and gap). Each sequence in the alignment is scanned sequentially, and the matrix element corresponding to each transition is incremented. In the end, the counts are converted into probabilities by normalizing them to 1. The Markov chain was used to generate new sequences by choosing a starting base at random according to the marginal probabilities of the bases in the first position, and by selecting at random a transition from each successive matrix at each step. This procedure prevents transitions that never appear in the alignment from being generated, as could instead happen if the synthetic sites had been generated by sampling from the probability distribution described by the NWM, while at the same time preserving the "core" sequence (the most conserved nucleotides) that Match relies on. Therefore, this method is well suited for generating sequences that can be recognized by both HMMER and Match, without giving an unfair disadvantage to any of the two methods. This method was favored over inserting at random binding sites from the alignments used to generate the NWM or HMM in order to keep the training and test set separate and to evaluate the two methods based on their ability to detect sequences that are similar but not identical with the one already reported as it would be expected for novel bona fide binding sites. We scanned the resulting sequence with both HMMER (with a threshold on the E-value equal to 20) and Match (with both core matrix and similarity matrix thresholds equal to 0.7), and we compared the results of both programs against the known locations of the synthetic binding sites. The whole process was repeated 20 times, and the results were analyzed in two different ways, by using modified ROC curves [82] and True Positive Proportion (TPP) tests described below.
We generated ROC plots to obtain an indication of the overall performance of both methods. However, while our methodology provides a definition of true and false positives, it does not explicitly define the set of true negatives. Since both programs scan the nucleotide sequence assigning a score to every position in the sequence and moving one base at a time, the effective number of true negative hits should be the length of the sequence minus the number of planted TFBSs, that is, the number of positions that were tested and were not found to match the TFBS pattern. This obviously results in a heavy imbalance between the number of true positives and the number of false positives, making it hard to evaluate the ROC curves in the commonly used way. For example, the areas under the curves that are normally used as a measure of predictive performance will always be very close to 1 and very similar to each other. Therefore, following the method of Gribskov and Robinson [82] we used ROC50 curves, plotted until 50 false positives are found, and we computed the areas under them (ROC50 areas). Models for which one or both method attained average areas below 0.25 were filtered out as their comparison was not meaningful. To determine if the values of the two sets are statistically different we performed a Wilcoxon signed rank test at significance value α equal to 0.05 with a Bonferroni correction of α divided by the number of independent tests (491). The result of this test is presented in Table 2 of Additional file 1.
In addition to the ROC curves, we also used two alternative tests, assessing the True Positive Proportion of hits retrieved by each method. For each sequence, we sorted the list of predicted hits by score, from highest to lowest, and we determined the percentage of true hits retrieved by each method within the first n reported hits and the amount of false positive hits (expressed as percentage) that one needs to accept in order to identify the first m true positive hits. The values used for n and m were 30, 50, 70 and 90. One method was considered to outperform the other in the TPP tests if it scored strictly better (showed a higher or lower percentage, depending on the test) for three out of the four n or m values and equally or better in the remaining one. The complete results of these tests are reproduced in Tables 3 and 4 of Additional File 1.
Database construction, website development and software environment
Genomic annotations from the UCSC Genome Browser, TRANSFAC, JASPAR and HomoloGene information were used to build a MySQL relational database storing data about genes, transcription factors, and their binding sites. We implemented a web-based system, accessible at , that allows users to search for the putative TFBSs in any region of a gene and of its orthologs, or in an arbitrary user-supplied sequence. This resource also offers access to a previously described database of pre-computed TFBSs found in the upstream sequences of all genes in the human, mouse and Drosophila genomes [107], that was generated using a methodology similar to the one described in this paper. The application is written in Common Lisp and relies on a development environment for web-based applications developed by the authors.
Abbreviations
HMM – hidden Markov model; NWM – nucleotide weight matrix; ROC curve – receiver operating characteristic curve; TF – transcription factor; TFBS – transcription factor binding site; TPP test – true positive proportion test.
Supplementary Material
Additional File 1
Word file containing links to the factors table and the results of the small-scale and large-scale evaluations.
Click here for file
Additional File 2
Excel file containing detailed results of the small-scale evaluation.
Click here for file
Acknowledgements
We thank Dr. Sean Eddy for extremely helpful suggestions and practical help in using HMMER more efficiently for this task, and Dr. Sven Rahmann and Benjamin Schuster-Böckler for making available the source code to generate the HMM logos. We also thank Drs. Paola Sebastiani, Marco Ramoni, Alvin Kho, John Brownstein and Peter Park for helpful discussions and suggestions, and the anonymous reviewers for their useful comments.
Figures and Tables
Figure 1 Quality measures for the alignments retrieved. A. Distribution of the parameters characterizing the model (length, number of sequences and the size of the nucleotide matrix used to train the model). B. Distribution of the median and average quality of the nucleotide sequences used to build the alignments for the TRANSFAC factor-derived models. The quality variable is categorical and represents "1 – functionally confirmed factor binding site; 2 – binding of pure protein purified or recombinant, 3 – immunologically characterized binding activity of a cellular extract, 4 – binding activity characterized via a known binding sequence, 5 – binding of uncharacterized extract protein to a bona fide element, 6 – no quality assigned" (cf. TRANSFAC documentation).
Figure 2 The selection page of the search engine. The selection page for the MCM5 gene displays detailed information on the gene and its homologs available in our database, and allows the user to select the gene region to be scanned. The same region will be scanned for all homologs included in the search.
Figure 3 The output of the query for the human MCM5 gene. The output was edited to highlight the E2F binding sites discussed in the text. The hit alignment window shows the match between the sequence at positions +2 to +12 from the transcript start and model T05206. The set of hits can be sorted by position, name or accession number of the factor. The position of the hits can be displayed with respect to the start of the transcript, the ATG or as absolute coordinates on the chromosome. The page can display the list of common factors that bind to the same selected region in the homologs included in the analysis, the factors on the list that are known to physically interact or the different classes to which they belong. In addition, the hits occurring in evolutionarily conserved regions can be highlighted.
Figure 4 Different representations of the set of putative TFBSs in the human MCM5 gene promoter. A. Graphical representation of the hit set presented in Figure 3. B. The hit set was exported to the UCSC Human Genome Browser as a custom track. The region displayed in this image extends to 500 bp upstream of the coding sequence start. Note that the clusters of predicted binding sites correspond to peaks in the human/mouse conservation track at the bottom, suggesting that those regions are functional. The positions of the most conserved elements displayed in the conservation track are the ones used in the previous page to highlight hits in evolutionary conserved regions (see Methods for details).
Figure 5 The page for model T05206 for E2F-4:DP-1. The model page displays detailed information regarding the model including the name and (if available) organism and classification of the factor, the model length, the number of sequences in the alignment used to train the model and the references used to select these sequences. The page also displays the HMM logo generated using the LogoMat-M software [89].
Table 1 Results of the small-scale evaluation.
Factor Sites Method Model Distinct sequences identified Target True positives False positives False positive ratio
E2F 27 HMMER M00050 27 27 36 8 18.20%
T05206 27 27 27 4 12.90%
Match V$E2F_02 27 27 36 7 16.30%
Patser V$E2F_02 27 27 36 8 18.20%
LMM V$E2F_02 18 27 18 3 n/a
ScanACE M00050 12 27 12 3 n/a
ER 17 HMMER M00959 16 16 22 7 24.10%
T00258 17 16 16 2 11.10%
Match V$ER_Q6_02 17 16 24 7 22.60%
Patser V$ER_Q6_02 17 16 24 8 25.00%
LMM V$ER_Q6 15 16 15 0 0.00%
ScanACE M00959 8 16 11 1 n/a
GR 7 HMMER M00921 7 7 10 2 16.70%
T05076 7 7 7 1 12.50%
Match V$GR_Q6_01 7 7 9 3 25.00%
Patser V$GR_Q6_01 7 7 9 7 43.70%
LMM V$GR_Q6 6 7 6 1 14.30%
ScanACE M00921 4 7 4 1 n/a
HNF-1 18 HMMER M00790 18 18 19 0 0.00%
T01211 18 18 18 0 0.00%
Match V$HNF1_Q6 18 18 22 3 12.00%
Patser V$HNF1_Q6 18 18 29 1 3.30%
LMM V$HNF1_01 16 18 16 0 0.00%
ScanACE M00790 11 18 11 0 n/a
HNF-3 10 HMMER M00724 10 10 10 1 9.10%
T01049 10 10 10 2 16.70%
Match V$HNF3ALPHA_Q6 10 10 12 4 25.00%
Patser V$HNF3ALPHA_Q6 10 10 10 1 9.10%
LMM V$HNF3ALPHA_Q6 9 10 9 4 30.80%
ScanACE M00724 8 10 8 0 0.00%
HNF-4 10 HMMER M00638 9 9 9 2 18.20%
T00372 10 9 9 0 0.00%
Match V$HNF4ALPHA_Q6 9 9 9 2 18.20%
Patser V$HNF4ALPHA_Q6 10 9 9 2 18.20%
LMM V$HNF4ALPHA_Q6 7 9 7 0 0.00%
ScanACE M00638 3 9 3 0 n/a
The "Sites" column contains the number of sequences containing experimentally validated binding sites provided as input. "Target" represents the number of binding sites to be retrieved by a method to be considered successful, and "Distinct sequences identified" is the number of distinct sequences in which at least one true positive was detected. Because of partially overlapping hits, the actual number of true positives reported may be higher than the true number of sites. Not all matrices tested were available in the LMM matrix library; for those cases, the results obtained using the closest available LMM matrix are displayed in italics.
Table 2 Comparative performance of the TFBS detection methods tested.
Performance value Method
HMMER M model HMMER T model Match Patser LMM ScanACE
% true positives identified 98% 100% 99% 100% 80% 52%
minimum false positive ratio 0.00% 0.00% 12.00% 3.30% 0.00% 0.00%
maximum false positive ratio 24.10% 16.70% 25.00% 43.70% 30.80% 0.00%
A summary of the results presented in Table 1. The first row displays the overall percentage of validated binding sites detected by the different methods. The next two rows display the minimum and maximum percentage of false positive hits found by each method across the six different datasets tested.
==== Refs
Ghazi A VijayRaghavan KV Developmental biology. Control by combinatorial codes Nature 2000 408 419 420 11100709 10.1038/35044174
Bulyk ML Computational prediction of transcription-factor binding site locations Genome Biol 2003 5 201 14709165 10.1186/gb-2003-5-1-201
Qiu P Recent advances in computational promoter analysis in understanding the transcriptional regulatory network Biochem Biophys Res Commun 2003 309 495 501 12963016 10.1016/j.bbrc.2003.08.052
Pennacchio LA Rubin EM Comparative genomic tools and databases: providing insights into the human genome J Clin Invest 2003 111 1099 1106 12697725 10.1172/JCI200317842
Pennacchio LA Rubin EM Genomic strategies to identify mammalian regulatory sequences Nat Rev Genet 2001 2 100 109 11253049 10.1038/35052548
Hughes JD Estep PW Tavazoie S Church GM Computational identification of cis-regulatory elements associated with groups of functionally related genes in Saccharomyces cerevisiae J Mol Biol 2000 296 1205 1214 10698627 10.1006/jmbi.2000.3519
Bailey TL Elkan C The value of prior knowledge in discovering motifs with MEME Proc Int Conf Intell Syst Mol Biol 1995 3 21 29 7584439
Lawrence CE Altschul SF Boguski MS Liu JS Neuwald AF Wootton JC Detecting subtle sequence signals: a Gibbs sampling strategy for multiple alignment Science 1993 262 208 214 8211139
Hertz GZ Stormo GD Identifying DNA and protein patterns with statistically significant alignments of multiple sequences Bioinformatics 1999 15 563 577 10487864 10.1093/bioinformatics/15.7.563
Buhler J Tompa M Finding motifs using random projections J Comput Biol 2002 9 225 242 12015879 10.1089/10665270252935430
Brazma A Jonassen I Vilo J Ukkonen E Predicting gene regulatory elements in silico on a genomic scale Genome Res 1998 8 1202 1215 9847082
Liu XS Brutlag DL Liu JS An algorithm for finding protein-DNA binding sites with applications to chromatin-immunoprecipitation microarray experiments Nat Biotechnol 2002 20 835 839 12101404
Workman CT Stormo GD ANN-Spec: a method for discovering transcription factor binding sites with improved specificity Pac Symp Biocomput 2000 467 478 10902194
Bussemaker HJ Li H Siggia ED Building a dictionary for genomes: identification of presumptive regulatory sites by statistical analysis Proc Natl Acad Sci U S A 2000 97 10096 10100 10944202 10.1073/pnas.180265397
Frith MC Fu Y Yu L Chen JF Hansen U Weng Z Detection of functional DNA motifs via statistical over-representation Nucleic Acids Res 2004 32 1372 1381 14988425 10.1093/nar/gkh299
Thompson W Rouchka EC Lawrence CE Gibbs Recursive Sampler: finding transcription factor binding sites Nucleic Acids Res 2003 31 3580 3585 12824370 10.1093/nar/gkg608
Liu Y Wei L Batzoglou S Brutlag DL Liu JS Liu XS A suite of web-based programs to search for transcriptional regulatory motifs Nucleic Acids Res 2004 32 W204 7 15215381
Hu Z Fu Y Halees AS Kielbasa SM Weng Z SeqVISTA: a new module of integrated computational tools for studying transcriptional regulation Nucleic Acids Res 2004 32 W235 41 15215387 10.1093/nar/gkh117
Kel AE Gossling E Reuter I Cheremushkin E Kel-Margoulis OV Wingender E MATCH: A tool for searching transcription factor binding sites in DNA sequences Nucleic Acids Res 2003 31 3576 3579 12824369 10.1093/nar/gkg585
Matys V Fricke E Geffers R Gossling E Haubrock M Hehl R Hornischer K Karas D Kel AE Kel-Margoulis OV Kloos DU Land S Lewicki-Potapov B Michael H Munch R Reuter I Rotert S Saxel H Scheer M Thiele S Wingender E TRANSFAC: transcriptional regulation, from patterns to profiles Nucleic Acids Res 2003 31 374 378 12520026 10.1093/nar/gkg108
Sandelin A Alkema W Engstrom P Wasserman WW Lenhard B JASPAR: an open-access database for eukaryotic transcription factor binding profiles Nucleic Acids Res 2004 32 Database issue D91 4 14681366 10.1093/nar/gkh012
Steffens NO Galuschka C Schindler M Bulow L Hehl R AthaMap: an online resource for in silico transcription factor binding sites in the Arabidopsis thaliana genome Nucleic Acids Res 2004 32 Database issue D368 72 14681436 10.1093/nar/gkh017
Kolchanov NA Ignatieva EV Ananko EA Podkolodnaya OA Stepanenko IL Merkulova TI Pozdnyakov MA Podkolodny NL Naumochkin AN Romashchenko AG Transcription Regulatory Regions Database (TRRD): its status in 2002 Nucleic Acids Res 2002 30 312 317 11752324 10.1093/nar/30.1.312
Kel-Margoulis OV Kel AE Reuter I Deineko IV Wingender E TRANSCompel: a database on composite regulatory elements in eukaryotic genes Nucleic Acids Res 2002 30 332 334 11752329 10.1093/nar/30.1.332
Levine M Tjian R Transcription regulation and animal diversity Nature 2003 424 147 151 12853946 10.1038/nature01763
Bailey TL Noble WS Searching for statistically significant regulatory modules Bioinformatics 2003 19 Suppl 2 II16 II25 14534166
Sinha S Van Nimwegen E Siggia ED A probabilistic method to detect regulatory modules Bioinformatics 2003 19 Suppl 1 I292 I301 12855472 10.1093/bioinformatics/btg1040
Frith MC Hansen U Weng Z Detection of cis-element clusters in higher eukaryotic DNA Bioinformatics 2001 17 878 889 11673232 10.1093/bioinformatics/17.10.878
Rajewsky N Vergassola M Gaul U Siggia ED Computational detection of genomic cis-regulatory modules applied to body patterning in the early Drosophila embryo BMC Bioinformatics 2002 3 30 12398796 10.1186/1471-2105-3-30
Berman BP Nibu Y Pfeiffer BD Tomancak P Celniker SE Levine M Rubin GM Eisen MB Exploiting transcription factor binding site clustering to identify cis-regulatory modules involved in pattern formation in the Drosophila genome Proc Natl Acad Sci U S A 2002 99 757 762 11805330 10.1073/pnas.231608898
Aerts S Thijs G Coessens B Staes M Moreau Y Moor BD Toucan: deciphering the cis-regulatory logic of coregulated genes Nucleic Acids Res 2003 31 1753 1764 12626717 10.1093/nar/gkg268
Frith MC Spouge JL Hansen U Weng Z Statistical significance of clusters of motifs represented by position specific scoring matrices in nucleotide sequences Nucleic Acids Res 2002 30 3214 3224 12136103 10.1093/nar/gkf438
Sharan R Ben-Hur A Loots GG Ovcharenko I CREME: Cis-Regulatory Module Explorer for the human genome Nucleic Acids Res 2004 32 W253 6 15215390
Alkema WB Johansson O Lagergren J Wasserman WW MSCAN: identification of functional clusters of transcription factor binding sites Nucleic Acids Res 2004 32 W195 8 15215379
Dieterich C Herwig R Vingron M Exploring potential target genes of signaling pathways by predicting conserved transcription factor binding sites Bioinformatics 2003 19 Suppl 2 II50 II56 14534171
Sandelin A Wasserman WW Lenhard B ConSite: web-based prediction of regulatory elements using cross-species comparison Nucleic Acids Res 2004 32 W249 52 15215389 10.1093/nar/gkh012
Liu Y Liu XS Wei L Altman RB Batzoglou S Eukaryotic regulatory element conservation analysis and identification using comparative genomics Genome Res 2004 14 451 458 14993210 10.1101/gr.1327604
Bigelow HR Wenick AS Wong A Hobert O CisOrtho: a program pipeline for genome-wide identification of transcription factor target genes using phylogenetic footprinting BMC Bioinformatics 2004 5 27 15113408 10.1186/1471-2105-5-27
Loots GG Ovcharenko I rVISTA 2.0: evolutionary analysis of transcription factor binding sites Nucleic Acids Res 2004 32 W217 21 15215384 10.1093/nar/gkh095
Grad YH Roth FP Halfon MS Church GM Prediction of similarly-acting cis-regulatory modules by subsequence profiling and comparative genomics in D. melanogaster and D. pseudoobscura Bioinformatics 2004
Blanchette M Tompa M Discovery of regulatory elements by a computational method for phylogenetic footprinting Genome Res 2002 12 739 748 11997340 10.1101/gr.6902
Prakash A Blanchette M Sinha S Tompa M Motif discovery in heterogeneous sequence data Pac Symp Biocomput 2004 348 359 14992516
Jegga AG Sherwood SP Carman JW Pinski AT Phillips JL Pestian JP Aronow BJ Detection and visualization of compositionally similar cis-regulatory element clusters in orthologous and coordinately controlled genes Genome Res 2002 12 1408 1417 12213778 10.1101/gr.255002
Lenhard B Sandelin A Mendoza L Engstrom P Jareborg N Wasserman WW Identification of conserved regulatory elements by comparative genome analysis J Biol 2003 2 13 12760745 10.1186/1475-4924-2-13
Wasserman WW Palumbo M Thompson W Fickett JW Lawrence CE Human-mouse genome comparisons to locate regulatory sites Nat Genet 2000 26 225 228 11017083 10.1038/79965
Emberly EG Rajewsky N Siggia ED Conservation of Regulatory Elements between two species of Drosophila BMC Bioinformatics 2003 4 57 14629780 10.1186/1471-2105-4-57
Stormo GD DNA binding sites: representation and discovery Bioinformatics 2000 16 16 23 10812473 10.1093/bioinformatics/16.1.16
Benos PV Bulyk ML Stormo GD Additivity in protein-DNA interactions: how good an approximation is it? Nucleic Acids Res 2002 30 4442 4451 12384591 10.1093/nar/gkf578
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 2478 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 1261 11861919 10.1093/nar/30.5.1255
Eddy SR Profile hidden Markov models Bioinformatics 1998 14 755 763 9918945 10.1093/bioinformatics/14.9.755
Durbin R Eddy S Krogh A G. M Biological sequence analysis: probabilistic models of proteins and nucleic acids 1998 Cambridge, UK, Cambridge University Press
Hughey R Krogh A Hidden Markov models for sequence analysis: extension and analysis of the basic method Comput Appl Biosci 1996 12 95 107 8744772
Grundy WN Bailey TL Elkan CP Baker ME Meta-MEME: motif-based hidden Markov models of protein families Comput Appl Biosci 1997 13 397 406 9283754
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
Raman R Overton GC Application of hidden Markov modeling to the characterization of transcription factor binding sites Proceedings of the 27th Hawaii International Conference on Systems Sciences 1994
Yada T Totoki Y Ishikawa M Asai K Nakai K Automatic extraction of motifs represented in the hidden Markov model from a number of DNA sequences Bioinformatics 1998 14 317 325 9632826 10.1093/bioinformatics/14.4.317
Conkright MD Guzman E Flechner L Su AI Hogenesch JB Montminy M Genome-wide analysis of CREB target genes reveals a core promoter requirement for cAMP responsiveness Mol Cell 2003 11 1101 1108 12718894 10.1016/S1097-2765(03)00134-5
Ellrott K Yang C Sladek FM Jiang T Identifying transcription factor binding sites through Markov chain optimization Bioinformatics 2002 18 Suppl 2 S100 S109 12385991
Price CW Fawcett P Ceremonie H Su N Murphy CK Youngman P Genome-wide analysis of the general stress response in Bacillus subtilis Mol Microbiol 2001 41 757 774 11532142 10.1046/j.1365-2958.2001.02534.x
Sandelin A Wasserman WW Prediction of nuclear hormone receptor response elements Mol Endocrinol 2005 19 595 606 15563547 10.1210/me.2004-0101
Fernandez PC Frank SR Wang L Schroeder M Liu S Greene J Cocito A Amati B Genomic targets of the human c-Myc protein Genes Dev 2003 17 1115 1129 12695333 10.1101/gad.1067003
Orian A van Steensel B Delrow J Bussemaker HJ Li L Sawado T Williams E Loo LW Cowley SM Yost C Pierce S Edgar BA Parkhurst SM Eisenman RN Genomic binding by the Drosophila Myc, Max, Mad/Mnt transcription factor network Genes Dev 2003 17 1101 1114 12695332 10.1101/gad.1066903
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 3654 12789273 10.1038/sj.onc.1206477
Cawley S Bekiranov S Ng HH Kapranov P Sekinger EA Kampa D Piccolboni A Sementchenko V Cheng J Williams AJ Wheeler R Wong B Drenkow J Yamanaka M Patel S Brubaker S Tammana H Helt G Struhl K Gingeras TR Unbiased Mapping of Transcription Factor Binding Sites along Human Chromosomes 21 and 22 Points to Widespread Regulation of Noncoding RNAs Cell 2004 116 499 509 14980218 10.1016/S0092-8674(04)00127-8
Wasserman WW Fickett JW Identification of regulatory regions which confer muscle-specific gene expression J Mol Biol 1998 278 167 181 9571041 10.1006/jmbi.1998.1700
Huang H Kao MC Zhou X Liu JS Wong WH Determination of local statistical significance of patterns in Markov sequences with application to promoter element identification J Comput Biol 2004 11 1 14 15072685 10.1089/106652704773416858
Hermeking H Lengauer C Polyak K He TC Zhang L Thiagalingam S Kinzler KW Vogelstein B 14-3-3 sigma is a p53-regulated inhibitor of G2/M progression Mol Cell 1997 1 3 11 9659898 10.1016/S1097-2765(00)80002-7
Jensen MR Factor VM Zimonjic DB Miller MJ Keck CL Thorgeirsson SS Chromosome localization and structure of the murine cyclin G1 gene promoter sequence Genomics 1997 45 297 303 9344652 10.1006/geno.1997.4947
Utrera R Collavin L Lazarevic D Delia D Schneider C A novel p53-inducible gene coding for a microtubule-localized protein with G2-phase-specific expression Embo J 1998 17 5015 5025 9724637 10.1093/emboj/17.17.5015
Rebeiz M Reeves NL Posakony JW SCORE: a computational approach to the identification of cis-regulatory modules and target genes in whole-genome sequence data. Site clustering over random expectation Proc Natl Acad Sci U S A 2002 99 9888 9893 12107285 10.1073/pnas.152320899
Rohrbaugh M Ramos E Nguyen D Price M Wen Y Lai ZC Notch activation of yan expression is antagonized by RTK/pointed signaling in the Drosophila eye Curr Biol 2002 12 576 581 11937027 10.1016/S0960-9822(02)00743-1
Chen Y Fischer WH Gill GN Regulation of the ERBB-2 promoter by RBPJkappa and NOTCH J Biol Chem 1997 272 14110 14114 9162037 10.1074/jbc.272.22.14110
Eddy SR HMMER User's Guide: Biological sequence analysis using profile hidden Markov models 2003
Kelly KK Meadows SM Cripps RM Drosophila MEF2 is a direct regulator of Actin57B transcription in cardiac, skeletal, and visceral muscle lineages Mech Dev 2002 110 39 50 11744367 10.1016/S0925-4773(01)00586-X
Wang DZ Valdez MR McAnally J Richardson J Olson EN The Mef2c gene is a direct transcriptional target of myogenic bHLH and MEF2 proteins during skeletal muscle development Development 2001 128 4623 4633 11714687
Shin HM Han TH CD28-mediated regulation of the c-jun promoter involves the MEF2 transcription factor in Jurkat T cells Mol Immunol 1999 36 197 203 10403485 10.1016/S0161-5890(99)00030-9
Frith MC Hansen U Spouge JL Weng Z Finding functional sequence elements by multiple local alignment Nucleic Acids Res 2004 32 189 200 14704356 10.1093/nar/gkh169
Frech K Quandt K Werner T Software for the analysis of DNA sequence elements of transcription Comput Appl Biosci 1997 13 89 97 9088714
Krivan W Wasserman WW A predictive model for regulatory sequences directing liver-specific transcription Genome Res 2001 11 1559 1566 11544200 10.1101/gr.180601
Barash Y Elidan G Friedman N Kaplan T Modeling dependencies in protein-DNA binding sites: Seventh annual international conference in computational molecular biology (RECOMB) 2003 Berlin, Germany 28 37
Gribskov M Robinson N Use of receiver operating characteristic (ROC) analysis to evaluate sequence matching Comput Chem 1996 20 25 33 10.1016/S0097-8485(96)80004-0
Hennessy KM Botstein D Regulation of DNA replication during the yeast cell cycle Cold Spring Harb Symp Quant Biol 1991 56 279 284 1819492
Dyson N The regulation of E2F by pRB-family proteins Genes Dev 1998 12 2245 2262 9694791
Trimarchi JM Lees JA Sibling rivalry in the E2F family Nat Rev Mol Cell Biol 2002 3 11 20 11823794 10.1038/nrm714
Takahashi Y Rayman JB Dynlacht BD Analysis of promoter binding by the E2F and pRB families in vivo: distinct E2F proteins mediate activation and repression Genes Dev 2000 14 804 816 10766737
Wells J Boyd KE Fry CJ Bartley SM Farnham PJ Target gene specificity of E2F and pocket protein family members in living cells Mol Cell Biol 2000 20 5797 5807 10913163 10.1128/MCB.20.16.5797-5807.2000
Ohtani K Iwanaga R Nakamura M Ikeda M Yabuta N Tsuruga H Nojima H Cell growth-regulated expression of mammalian MCM5 and MCM6 genes mediated by the transcription factor E2F Oncogene 1999 18 2299 2309 10327050 10.1038/sj.onc.1202544
Schuster-Bockler B Schultz J Rahmann S HMM Logos for visualization of protein families BMC Bioinformatics 2004 5 7 14736340 10.1186/1471-2105-5-7
Bailey TL Elkan C Fitting a mixture model by expectation maximization to discover motifs in biopolymers Proc Int Conf Intell Syst Mol Biol 1994 2 28 36 7584402
Sinha S Tompa M Discovery of novel transcription factor binding sites by statistical overrepresentation Nucleic Acids Res 2002 30 5549 5560 12490723 10.1093/nar/gkf669
Tronche F Ringeisen F Blumenfeld M Yaniv M Pontoglio M Analysis of the distribution of binding sites for a tissue-specific transcription factor in the vertebrate genome J Mol Biol 1997 266 231 245 9047360 10.1006/jmbi.1996.0760
Karolchik D Baertsch R Diekhans M Furey TS Hinrichs A Lu YT Roskin KM Schwartz M Sugnet CW Thomas DJ Weber RJ Haussler D Kent WJ The UCSC Genome Browser Database Nucleic Acids Res 2003 31 51 54 12519945 10.1093/nar/gkg129
UCSC Genome Bioinformatics Sequence and Annotation Downloads
HomoloGene
Wheeler DL Church DM Federhen S Lash AE Madden TL Pontius JU Schuler GD Schriml LM Sequeira E Tatusova TA Wagner L Database resources of the National Center for Biotechnology Nucleic Acids Res 2003 31 28 33 12519941 10.1093/nar/gkg033
UCSC Genome Bioinformatics Table Browser
Siepel A Haussler D Nielsen R Phylogenetic hidden Markov models Statistical Methods in Molecular Evolution 2004 , Springer
MEME/MAST Download
JASPAR Sites Download
GLAM Datasets
Software for the Analysis of DNA Sequence Elements of Transcription
A Predictive Model for Regulatory Sequences Directing Liver-specific Transcription, Supplementary Material
RSA-tools - Patser
Sequence Analysis: Local Markov Method
AlignACE Accessory Software
Marinescu VD Kohane IS Riva A The MAPPER database: a multi-genome catalog of putative transcription factor binding sites Nucleic Acids Res 2005 33 D91 7 15608292
| 15799782 | PMC1131891 | CC BY | 2021-01-04 16:02:48 | no | BMC Bioinformatics. 2005 Mar 30; 6:79 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-79 | oa_comm |
==== Front
BMC BiolBMC Biology1741-7007BioMed Central London 1741-7007-3-121584514810.1186/1741-7007-3-12Research ArticleThe Arabidopsis AtRaptor genes are essential for post-embryonic plant growth Anderson Garrett H [email protected] Bruce [email protected] Maureen R [email protected] Molecular Biology and Genetics, Cornell University, Ithaca, 14853, USA2 AgResearch, Private Bag 11008, Tennent Drive, Palmerston North, 5301, New Zealand2005 21 4 2005 3 12 12 23 1 2005 21 4 2005 Copyright © 2005 Anderson 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
Flowering plant development is wholly reliant on growth from meristems, which contain totipotent cells that give rise to all post-embryonic organs in the plant. Plants are uniquely able to alter their development throughout their lifespan through the generation of new organs in response to external signals. To identify genes that regulate meristem-based growth, we considered homologues of Raptor proteins, which regulate cell growth in response to nutrients in yeast and metazoans as part of a signaling complex with the target of rapamycin (TOR) kinase.
Results
We identified AtRaptor1A and AtRaptor1B, two loci predicted to encode Raptor proteins in Arabidopsis. Disruption of AtRaptor1B yields plants with a wide range of developmental defects: roots are thick and grow slowly, leaf initiation and bolting are delayed and the shoot inflorescence shows reduced apical dominance. AtRaptor1A AtRaptor1B double mutants show normal embryonic development but are unable to maintain post-embryonic meristem-driven growth. AtRaptor transcripts accumulate in dividing and expanding cells and tissues.
Conclusion
The data implicate the TOR signaling pathway, a major regulator of cell growth in yeast and metazoans, in the maintenance of growth from the shoot apical meristem in plants. These results provide insights into the ways in which TOR/Raptor signaling has been adapted to regulate plant growth and development, and indicate that in plants, as in other eukaryotes, there is some Raptor-independent TOR activity.
==== Body
Background
Plant development is remarkably plastic. Groups of totipotent cells termed meristems, which are maintained throughout the life of the plant, give rise to all post-embryonic organs from roots and leaves to petals and fruit. This allows plants, unlike metazoans, to change their final body plans dramatically in response to environmental, hormonal and nutritional cues. While much has been learned about the determination of cell fates in the embryo [1] and the apical meristems [2-4], less is known about the genes that control the growth that occurs in cells emerging from the plant meristem.
TOR proteins were originally identified in budding yeast as the targets of rapamycin, a potent antibiotic that disrupts cell growth [5,6]. In both yeast and metazoans, TOR proteins mediate translation in response to nutrients [7]. Yeast TOR2 and mammalian mTOR also regulate cytoskeletal organization [8-13].
In both yeast and mammalian cells, TOR proteins form a complex, TORC1, with GβL [14] and Raptor (regulatory associated protein of TOR) [15-17]; their yeast homologues are LST8 and KOG1, respectively [17]. Raptor contains protein-binding domains, and the Raptor N-terminus shows similarity to a caspase domain [18], though catalytic activity has yet to be shown. It has been reported that the strength of the TOR-Raptor interaction is regulated by nutrients, though there is not yet consensus on this point [15-17]. Raptor functions in TORC1 to recruit substrates for phosphorylation by TOR; Raptor binds TOR substrates S6kinase and eIF4E-BP, and is necessary for full TOR phosphorylation of these substrates in vitro [16,19]. The TOR-Raptor complex is thought to mediate the nutrient-sensitive regulation of cell growth; mutants in yeast lacking a functional complex cease cell growth in a manner that mimics rapamycin treatment [17]. A second TOR complex, TORC2, involves GβL, Rictor (rapamycin-insensitive companion to TOR) and perhaps other proteins [13,17,20]. This complex is unaffected by rapamycin, and is thought to mediate TOR cytoskeletal regulation.
In Arabidopsis the single TOR homologue, AtTOR, is critical for plant development [21,22]. AtTOR insertion homozygote embryos undergo cell division but are unable to gain cell volume or undergo apical-basal differentiation. Although AtTOR transcripts accumulate in all tissues assayed [22], an AtTOR-Gus fusion transcript is only translated in meristematic tissue and the immediately adjacent expanding cells [21]. A TOR-specific microRNA conserved between Arabidopsis and rice [23] implicates post-transcriptional regulation in AtTOR expression. Rapamycin perturbs the activity of the nutrient-sensitive TOR complex in yeast and mammals. Arabidopsis, however, is insensitive to rapamycin, [21] precluding this approach to the elucidation of Arabidopsis TOR signaling.
Here we report a characterization of AtRaptor1A and AtRaptor1B, which encode the two Arabidopsis Raptor homologues. We found defects in both root and shoot growth in AtRaptor1B disruption lines, resulting in slow leaf initiation, late flowering and increased branching. We also found that AtRaptor1A AtRaptor1B double mutant homozygotes undergo normal seedling development, but exhibit only minimal post-embryonic meristem activity. We contrast these results with those found previously for the AtTOR mutant, and discuss the implications of this work for TOR signaling in plants.
Results
Characterization of the Arabidopsis Raptor homologues
To search for plant homologues of mammalian Raptor, we surveyed the completed Arabidopsis genome and found two loci that might encode proteins highly similar to Raptor at the north ends of chromosomes 3 and 5. We did not find an expressed sequence tag (EST) for the first locus (AtRaptor1A; At5g01770). However, we used the 5'-rapid amplification of cDNA ends (RACE) to establish that the locus is transcribed to an mRNA with a 5' end consistent with the predicted open reading frame (ORF) of AtRaptor1A. The AtRaptor1A protein is predicted to be 1,346 amino acids in length. The second locus (AtRaptor1B; At3g08850) was represented by a single EST (accession no. AY769948). We used 5' RACE to confirm that the EST contained the full ORF of AtRaptor1B. The AtRaptor1B locus is transcribed to a 4.8 kb mRNA containing 23 exons, and the protein is predicted to be 1,344 amino acids in length. Predicted AtRaptor1A and AtRaptor1B proteins show 80% identity over their entirety.
To determine the time of divergence between the two AtRaptor loci, we searched for Raptor homologues in available plant genome data using AtRaptor1A and AtRaptor1B as query sequences. Full-length Raptor loci were discovered in the available rice (Oryza sativa subspecies japonica) and alfalfa (Medicago truncatula) genome sequences. Using AtRaptor1B and partial cDNA sequence (where available) as guides, we determined putative protein sequences from these loci. Alignment of these sequences with AtRaptor1A, AtRaptor1B, mammalian Raptor, the budding yeast Raptor homologue KOG1 and the fission yeast Raptor homologue Mip1p showed that there is a striking degree of conservation among all Raptor homologues (Fig. 1A; for the complete alignment [see additional file 1]). AtRaptor1B and the Saccharomyces cereviseae Raptor homologue KOG1, the most divergent member included in the analysis, show 28% identity throughout their length. All plant Raptor homologues encode the Raptor N-terminal Conserved / Caspase (RNC/C) motif, HEAT repeats and WD40 motifs first identified in fission yeast Mip1 and characteristic of all Raptor proteins [25,26](Fig. 1B). From this alignment we generated a phylogeny of the Raptor homologues (Fig. 1C). The predicted AtRaptor proteins resolve to a single clade with a high degree of confidence, indicating that the duplication of the AtRaptor loci post-dates the divergence of Arabidopsis from Medicago, and that the loci are likely to encode functionally redundant proteins.
Identification of knockout mutants in AtRaptor1A and AtRaptor1B
To gain insight into the function of the AtRaptor proteins, we searched for insertion alleles of each locus among the publicly available sequenced T-DNA insertion lines. Insertion mutant lines SALK_043920 and SALK_078159, with disruptions to the AtRaptor1A and AtRaptor1B loci, were obtained from the Arabidopsis Biological Resource Center (USA). Lines homozygous for each insertion (referred to as 1A -/- and 1B -/-, respectively) were identified via polymerase chain reaction (PCR), and RNA from floral buds of insertion homozygotes was used for reverse-transcriptase-PCR (RT-PCR) to assay for accumulation of wild-type transcripts from the disrupted locus. AtRaptor1A transcripts could not be detected in 1A -/- buds, but were detected in 1B -/- buds. AtRaptor1B transcripts were not detected in 1B -/- buds, but were detected in 1A -/- buds. Both transcripts were detected in wild-type Columbia (Col) buds (Fig. 2).
1B-/- seedling roots grow slowly
As a first step toward characterizing these mutant lines, Col, 1A-/- and 1B-/- seedlings were germinated on culture medium. Col and 1A-/- seedlings were phenotypically indistinguishable. However, B-/- seedlings showed distinct root growth defects: roots were thick, coiled and densely covered with hairs, and the roots had difficulty penetrating the medium (Fig. 3A). Plate-grown 1B -/- seedlings were repeatedly shorter than Col or 1A-/- seedlings. Although dark grown 1B-/- seedlings were also shorter than Col or 1A-/-, the difference in size was much less (Fig. 3B). While dark-grown etiolated seedlings grow primarily through the expansion and division of embryonic hypocotyl cells, light-grown seedlings grow primarily through the production of cells from the root apical meristem. Thus the difference in size between Col, 1A-/- and 1B-/- seedlings results primarily from a reduction in root apical meristem growth rather than a general defect in cell expansion or metabolism.
The roots of 1B-/- seedlings grown on 90° inclined plates were morphologically wild type but shorter than Col roots. When the plates were rotated to lay flat horizontally, new growth from the root apex of 1B-/- seedlings produced thick, coiled roots (though the roots were not densely covered in hairs), while the previously produced root tissue remained morphologically wild type (Fig. 3C). Only those sections of the root not in contact with the medium were hairy (Fig. 3A). Thus the root growth disturbance was a defect in growth into agar rather than in gravitropism, though in the absence of resistance, roots still grew more slowly than wild type.
Col and 1B-/- root apical meristems (RAM) were examined using bright-field microscopy (Fig. 3I). 1B-/- seedling RAMs contained all cell types present in Col RAMs but their root morphology was much more blunt; this difference appeared to be localized to the 1B-/- zone of elongation. 1B-/- root tips showed a tendency to shed their root caps, although shedding was sometimes observed in Col root tips as well.
1B-/- shoots show developmental and morphological defects
To measure developmental defects resulting from AtRaptor insertions, we grew Col, 1A-/- and 1B-/- lines on soil, and scored their rates of leaf initiation, time to floral bud initiation, rate of cauline leaf initiation, time to flowering and number of floral shoot apices. 1B-/- plants were smaller, had a slower rate of leaf initiation, and bolted and flowered later than did Col or 1A-/- plants (Fig. 4A–C).
Mature 1B-/- plants were conspicuously bushier than Col or 1A-/- homozygotes (Fig. 5). The primary shoot apex was shorter than that of Col plants and ended prematurely in a terminal inflorescence of infertile flowers. The growth of the plant was then taken over by axillary shoots and by secondary shoots from the basal rosette.
To quantify this phenotype, Col and 1B-/- lines were grown to maturity in 16-hr days, and shoot architecture for an individual plant was scored at the shattering of the first silique. 1B-/- plants showed reduced shoot height, reduced primary stem length, increased axillary branch number and an increased number of secondary shoots compared to Col and 1A-/- plants (Fig. 5B, E). 1B-/- branch numbers, though variable, were significantly larger than those of wild type or 1A-/-. Student's t-Tests of sample pairs assuming equal variance yielded P-values of 0.0027 and 0.0004 for 1B-/- vs. Col and 1B-/- vs. 1A-/- axillary branch number, respectively, and 0.0026 and 0.0013 for 1B-/- vs. Col and 1B-/- vs. 1A-/- secondary shoot number. Axillary branch length did not differ significantly from Col values (Fig. 5F). This phenotype became more pronounced later in plant development and was more conspicuous under short days.
Complementation of the AtRaptor1B-/- phenotype
To confirm that the collection of phenotypes observed in the AtRaptor1B-/- homozygotes was due to the mutant 1B allele, 1B-/- plants were transformed with a construct containing the AtRaptor1B ORF and 5' UTR and driven by the AtRaptor1B promoter. Some of these transformants showed wild-type transition to bolting, shoot branching and root growth (data not shown). We conclude that the 1B-/- phenotypes described above result from homozygosity for the lesion at the AtRaptor1B locus.
AtRaptor transcripts accumulate in dividing and expanding cell tissues
A combination of in situ RNA hybridization and in silico expression analysis was used to determine the RNA accumulation pattern of AtRaptor transcripts in Arabidopsis. The AtRaptor1B cDNA sequence was used to generate an AtRaptor probe to assay transcript accumulation in the shoot tips of wild-type plants. Since AtRaptor1A and AtRaptor1B show 80% identity through the length of their transcripts, this probe likely detected expression of both loci.AtRaptor transcripts were detected throughout the shoot tip, in all organs of the differentiating floral bud, and deep into the differentiated inflorescence stem (Fig. 6A). Signal intensity faded with the distance from the shoot apex. This accumulation pattern differed from that of actin (Fig. 6B), which was more prominent in dividing cells of the apex. Notably, AtRaptor accumulation is not restricted to the primary shoot apex.
To obtain an estimate of the relative levels of AtRaptor1A and AtRaptor1B in the signal seen in Fig. 6A, AtRaptor1A and AtRaptor1B locus IDs were used to query the Genevestigator Arabidopsis thaliana Microarray Database and Analysis Toolbox [27]. AtRaptor1A and AtRaptor1B accumulate in all developmental stages (Fig. 6C). However, AtRaptor1B accumulation levels are consistently fourfold higher than those of AtRaptor1A. A second analysis of accumulation by tissue rather than developmental stage produced similar results. We conclude that AtRaptor1A and AtRaptor1B show similar expression patterns.
1A-/- 1B-/- double homozygote mutant seedlings show minimal meristem growth
The high degree of similarity between AtRaptor1A and AtRaptor1B led us to suspect that they may be at least partially functionally redundant. To eliminate all AtRaptor activity in a single line, 1A-/- and 1B-/- lines were crossed. No 1A-/- 1B-/- plants could be identified among the F2 of this cross on soil. Phenotypically wild-type 1A-/- 1B+/- plants were identified via PCR using the primer pairs described in Materials and Methods and their progeny (205 seedlings) were examined on agar plates. Of these, 165 seedlings (80%) appeared wild type (Fig. 7A). It was determined by PCR that these seedlings carried a wild-type 1B allele (32/32 tested). The remaining seedlings (40/205; 20%) germinated but showed minimal postembryonic shoot meristem growth (Fig. 7A). It was determined by PCR that these seedlings were genetically 1A-/- 1B-/- (32/32 tested). This genotyping confirmed that the seedling arrest mutant phenotype co-segregated with homozygosity for the AtRaptor1B mutant allele in the AtRaptor1A mutant background. We concluded that this seedling phenotype corresponded to the 1A-/- 1B-/- double mutant.
We examined 1A-/- 1B-/- double mutant seedling roots via bright field microscopy to ascertain the extent of their post-embryonic growth defect. 1A-/- 1B-/- roots were conspicuously narrower than wild type. The columella, quiescent center, vasculature, pericycle, endodermis and cortex are present though all smaller than wild type; the outer layer of cells is difficult to identify. Further up the root, root hairs were clearly visible, indicating that the RAM had produced some mature root cell types (Fig. 7D, E).
To gain a better understanding of the double mutant shoot apical meristem (SAM) defect, Col and 1A-/- 1B-/- 7-day seedlings were fixed and coated, and their shoot apexes were observed with a scanning electron microscope. In Col seedlings, leaves 1 and 2, with trichomes, were clearly visible; primordia for leaves 3, 4 and 5 were visible under higher magnification. 1A-/- 1B-/- shoot apexes showed minimal SAM activity: leaves 1 and 2 were produced, but were smaller than wild type and did not separate to reveal any further leaf primordia (Fig. 7F, G, H, I).
In order to determine whether a plant hormone or signaling molecule could restore post-embryonic growth to these arrested seedlings, progeny of 1A-/- 1B+/- individuals were germinated on plates containing a variety of plant hormones and signaling molecules and scored for their total seedling lengths at 10 days post germination. Signaling molecules tested included the gibberellin GA3 (100 nM to 10 μM), abscisic acid (100 nM to 10 μM), the auxin 2,4-D (1 nM to 10 μM), the ethylene precursor ACC (1 μM and 10 μM), sucrose (1%) and glucose (6%). We also attempted dark treatment. None of these treatments restored wild-type SAM growth to the quarter of the progeny showing seedling arrest. 1A-/- 1B-/- seedlings did show a significant response to sucrose (Fig. 7B, C). Addition of 1% sucrose, which is a growth signaling factor as well as a carbon source [28], to the growth medium promoted growth in Col, 1A-/-, 1B-/- and 1A-/- 1B-/-. In sibling seedlings, the addition of sucrose elicited a twofold increase in seedling length; in the double mutants the increase was five-fold. Germination on 6% sucrose, which signals growth arrest in Col plants [28], yielded sibling plants 1–1.5× the size of seedlings grown on 1/2 Murashige & Skoog (MS) salts with no sucrose added. 1A-/- 1B-/- seedlings on 6% sucrose were 2.3× the length of those grown on 1/2 MS salts. A similar result was observed among 1A-/- 1B+/- progeny germinated in the dark on 1/2 MS salts plates, and the length of 1A-/- 1B-/- seedlings was increased more than 3× under these conditions. The absolute length of the double mutant seedlings was in all cases still substantially smaller that that of sibling seedlings for a given treatment, and the increase in length was largely due to hypocotyl elongation or root growth. Minimal SAM activity was observed.
Discussion
Axillary branch growth in Arabidopsis and other vascular plants is governed indirectly by auxin produced at the primary shoot apex, which acts through an undetermined secondary messenger to repress axillary meristem growth [29]. The shoot morphology of 1B-/- plants indicates a defect in primary SAM maintenance. The primary shoot is shorter than wild type and ends in a whorl of sterile flowers. The number and length of axillary shoots, both cauline and emerging from the basal rosette, is significantly increased. Notably, axillary meristems are quite viable, producing branches as long as or longer than wild type. The viability of axillary shoot apices points to a failure to maintain primary SAM activity, and a subsequent failure to repress axillary meristem activity. This defect is not shared by the axillary SAMs, which match or surpass the activity of their wild-type counterparts. A similar phenomenon is seen in erecta mutants [30] and may be quite common. The increased branching in AtRaptor1B mutant plants may result directly from reduced auxin production at the shoot apex or from a change in the auxin/cytokinin ratio. However, 1B-/- mutants did not show defects in their ability to sense exogenous auxin (data not shown).
The endogenous AtRaptor1A locus cannot fully complement the disruption of AtRaptor1B despite the fact that their predicted proteins are 80% identical, share all conserved Raptor motifs and strongly resolve to a single clade in phylogenetic analysis. In silico analysis suggests that the failure of AtRaptor1A to mask AtRaptor1B lesions may be due to lower levels of AtRaptor1A expression globally rather than limited tissue-specific accumulation of AtRaptor1A. We did not assay for the complementation of AtRaptor1B lesions by transforming AtRaptor1B-/- plants with an AtRaptor1A construct. In contrast, AtRaptor1B can fully complement the loss of AtRaptor1A expression; indeed, a single wild-type AtRaptor1B allele in 1A-/- 1B+/- plants is sufficient for wild-type growth.
Our ability to isolate AtRaptor1B mutant homozygotes conflicts with a recent report [24]. In that report, mutants homozygous for AtRaptor1B disruptions were lethal at or before the first asymmetric embryonic division, which is even earlier than the lethality of AtTOR mutant homozygotes. We are uncertain of the reason for the discrepancy in these results. We are able to recover 1B-/- plants, but Deprost et al. [24], working with the same AtRaptor1B allele (SALK_043920), described them as embryo-lethal. One possible explanation might be a variation in growth conditions, which could lead to a higher degree of stress on developing embryos in the experiments by Deprost et al. The fact that these authors observed some embryo arrest in wild type control lines supports this possibility. Disruption of AtRaptor1A yielded no significant phenotype in either our growth conditions or those of the previous report [24].
Simultaneous disruption of both AtRaptor loci resulted in growth-arrested seedlings, which undergo normal embryonic organogenesis but are unable to maintain post-embryonic growth from their SAMs. AtRaptor1A-/- AtRaptor1B-/- double mutants' SAMs cease activity after the production of a few leaf primordia. We conclude that AtRaptor activity is essential for the maintenance of SAM activity, but not for SAM assembly or for the initiation of SAM activity.
Our results indicate that AtRaptor activity is not essential for embryonic organogenesis. In contrast, AtTOR-/- mutants arrest development early in embryogenesis [21]. The disparity between the AtRaptor1A-/- 1B-/- and the AtTOR-/-phenotypes indicates that AtTOR activity in embryonic development does not require AtRaptor. Thus if AtTOR is acting in a complex in embryonic development, this complex does not require AtRaptor.
In both yeast and mammalian cells, TOR functions in two distinct complexes (Fig. 8A). The first of these, TORC1, involves Raptor and regulates cell growth and translation in response to nutrients [14-17]. The second of these complexes, TORC2, regulates cytoskeletal organization and does not involve Raptor [13,17,20].
AtRaptor-independent AtTOR activity in the plant embryo is consistent with the existence of two AtTOR complexes in Arabidopsis: 1) a Raptor-independent complex critical for early embryogenesis (and perhaps all stages of plant development), and 2) a Raptor-dependent complex that is dispensable for embryonic development but which is necessary for post-embryonic, meristem-driven plant growth (Fig. 8B).
Conclusion
By identifying the phenotypes resulting from partial and total disruption of AtRaptor activity, we have generated valuable tools for the study of plant TOR signaling. When viewed in the context of previous work on AtTOR [21,22], our work provides evidence for Raptor-independent TOR activity in land plant embryonic development – development that, notably, occurs in an environment made nutritionally and environmentally stable via maternal input to growth. We propose that in plants as in other eukaryotes there are two (or more) TOR complexes, only one of which involves Raptor. We further propose that the plant homologue of TORC1, involving TOR and Raptor, has been co-opted in plants from its ancestral role in nutrient sensing and cell growth to regulate the highly plastic post-embryonic growth driven by the plant shoot apical meristem.
Methods
Insertion lines
The mutant lines SALK_043920 and SALK_078159, in which Agrobacterium-mediated T-DNA insertions disrupted the AtRaptor1A and AtRaptor1B loci in the Columbia (Col) genetic background, were generated by Joseph R. Ecker and the Salk Institute Genomic Analysis Laboratory (USA) and distributed by the Arabidopsis Biological Resource Center (USA) [31]. Lines were genotyped using the following primers: 1Asm5 5'aaaaagtctcttagatgtagtttcagatg 3' and 1Asm3 5' attcagaatatacaatccaagcattagt 3' to identify the AtRaptor1A wild-type locus; 1Bsm5 5' ctgaccataacattctcttgtaggtaagg 3' and 1Bsm3 5' aggcctgaactctaatgaacaaactctcc 3' to identify the AtRaptor1B wild-type locus; and 1Bsm5 or 1Asm5 and pROK-737 (5' gggaattcactggccgtcgttttacaa 3') to identify the respective AtRaptor mutant alleles. Mutant lines were crossed to Col wild type plants; the AtRaptor1B mutant phenotype cosegregated with the 1B insertion allele in the F2 population.
Growth conditions
Plants were grown in a greenhouse with supplemental light to 16 hrs, with temperatures held at 22°C days and 17°C nights. Seeds for plate-grown seedlings were surface sterilized with 20% Chlorox, washed in H20, stratified for four days at 4°C and grown on 1/2× MS salts, 0.3% Phytagel under 12 hr. daylight cycles.
RT-PCR
Buds from Col and insertion line homozygotes were snap-frozen in liquid N2. Total RNA was extracted from 0.3 g of tissue using TRIzol Reagent (Invitrogen) according to manufacturer's instructions. Resuspended RNA was thrice treated with DNA-free™ DNase treatment (Ambion). RNA was quantified with a spectrophotometer. cDNA was generated from 2 μg of total RNA using Omniscript Reverse Transcriptase (Qiagen) and a poly-dT18 primer. PCR was performed using ExTaq polymerase (Takara) in a PTC-100 thermocycler. Primer sequences were as follows: 1A+1828, 5' gctgcgtttattttggctgttattgtc 3' and 1A-2800, 5' ctaggccagccagaggagtgtgagatg 3'; 1B+1379, 5' aggccggcaaaacgatcgtaagacatt 3' and 1B-2774, 5' catcagcccagaggagccaagagg 3'. PCR cycling parameters were 95°C for 5 min followed by 35 cycles of 94°C for 30 sec, 62°C for 30 sec, and 72°C for 1 min.
AtRaptor1B cDNA clone
EST clone RZL03b06 corresponding to the 5' end of AtRaptor1B was ordered from Kazusa DNA Research Institute and sequenced (Accession number AY769948). RNA ligase-mediated RACE was performed on total RNA extracted from bulk shoot tissues using a GeneRacer™ Kit (Invitrogen) according to manufacturer's instructions in order to confirm that RZL03b06 represented a full-length clone.
Assembly of the AtRaptor1B complementation vector
Primers were designed to amplify the region from the end of the transcript adjacent to the AtRaptor1B locus to a site within the ORF, spanning a region from 1145 bases upstream of the transcript initiation site through to the sixth exon of the AtRaptor1B transcript. Primers used and restriction sites added are as follows: 1B-8189BglII 5' agatctgaggaaccagaagaaccc 3'; 1B+5104HindIII 5' aagcttcggcggagtaggaaaac 3'. PCR was performed using ExTaq polymerase (Takara) in a PTC-100 thermocycler on Col-0 genomic DNA with the following parameters: 96°C for 5 min, then 35 cycles of 94°C for 30 sec, 60°C for 30 sec and 72°C for 2 min, followed by a final step of 72°C for 10 minutes. HindIII BglII digested fragments of the resulting PCR products were ligated into pCambia1301.
Next, the 1301B construct was digested with PmlI and PmeI. The EST containing the AtRaptor1B ORF was digested with PmeI and SmaI, and the resulting fragment was ligated into 1301B to create the complementation vector 1301B:Raptor. 1301B:Raptor was transformed into Agrobacterium line GV3101, and then into Col or 1B-/- plants via floral dip [32].
Microscopy
SEM fixation was performed using standard methods [33]. Bright field microscopy was performed on a Zeiss microscope, and images were collected on a BioRad Confocal System.
In situ hybridization
Digoxigenin-labelled probes were hybridized to paraffin-sectioned material using previously described methods [34].
In silico analysis
Locus identifiers were submitted to the Genevestigator Arabidopsis thaliana microarray database and analysis toolbox [27] at , where they were assayed against 1434 developmental and tissue-specific Arabidopsis microarray experiment results [35-38].
Authors' contributions
GHA and MRH formulated the experiments, analyzed the data, and wrote the manuscript, GHA obtained the data for the mutant phenotypes, and BV performed the in situ hybridizations.
Supplementary Material
Additional File 1
Alignment of Raptor homologues in plants, fungi and mammals Additional file 1 is an image of the Raptor homologue alignment. Shown are predicted protein sequences for the plant Raptor proteins AtRaptor1A and AtRaptor1B (Arabidopsis), MtRaptor (Medicago truncatula), OsRaptor (Oryza sativa), the fungal raptor homologue Mip1 (S. pombe), and mammalian Raptor. Sequences were aligned in Megalign; the image was created in Genedoc. Putative sequences for AtRaptor1A and MtRaptor are based on genomic predictions rather than EST or cDNA sequence. Residues that are 100% conserved (either identical or biochemically similar) are shown as white text on black. More than 80% conservation is shown as white on dark grey, and more than 60% is shown as black on light grey.
Click here for file
Acknowledgements
The authors gratefully acknowledge Barbara Eaglesham for assistance with scanning electron microscopy, Paulina Melechkina for the drawing in Fig. 8 and Nena Alvarez for assistance with in situ hybridization. This work was funded by a fellowship from the National Science Foundation/Department of Energy (DOE)/United States Department of Agriculture (USDA) Training Group in Molecular Mechanisms of Plant Processes to GHA, and USDA Hatch Program and DOE Energy Biosciences (DE-F602-89ER14030) grants to MRH. BV was supported by a Marsden award from the Royal Society of New Zealand and by the New Zealand Foundation for Research Science and Technology.
Figures and Tables
Figure 1 Raptor proteins in eukaryotes are highly conserved. (A) Similarity plot of Raptor homologues from the vascular plants Arabidopsis, Medicago truncatula and Oryza sativa, the fungus S. pombe (Mip1p), and mammals. The X-axis represents residue number; the Y-axis represents percent identity at that residue from 0% (0) to 100% (1). (B) Schematic diagram showing the position of the Raptor N-terminal Conserved / putative Caspase domain (RNC/C) region, HEAT repeats (H), and WD-40 repeats (WDx7) common to all Raptor proteins. (C) Phylogeny of plant, animal and fungal Raptor proteins. Bootstrap values, calculated using both parsimony (left) and maximum likelihood (right) are shown to the left of the clades they describe. The two Arabidopsis Raptor proteins, AtRaptor1A and AtRaptor1B, resolve as a single clade with 100% confidence. The alignment was generated using Megalign (DNAStar), the similarity plot was generated from this alignment using VectorNTI, and bootstrap values were calculated using PAUP*4.0b.
Figure 2 AtRaptor loci and insertion allele characterization. (A) AtRaptor1A and AtRaptor1B loci. Genomic sequence is depicted as a thin central line. Thick blocks indicate exons. Coding exons span the central line; exons encoding untranslated regions are fully below the central line. The positions of the T-DNA insertions are depicted with inverted triangles. (B) Reverse-Transcribed RNA-template Polymerase Chain Reactions (RT-PCR) on plants homozygous for both wild-type AtRaptor alleles (Col), the AtRaptor1A insertion allele (A-) or the AtRaptor1B insertion allele (B-), using primers spanning the AtRaptor1A insertion site, the AtRaptor1B insertion site, or control primers. Both AtRaptor insertion alleles abolish accumulation of the wild-type transcript from their locus.
Figure 3 Seedling root phenotype of AtRaptor1B-/- (B-) mutants. (A), (B). Col and B- seedlings on growth medium, four days after germination. The B- root has not penetrated the medium and is thick, hairy and coiled. (C) Col, A- and B- seedlings at 8 days after germination in light. Scale bar is in mm. (D) Same genotypes and age as C, grown in the dark. (E) Measurements of populations grown as in C, D. Root length is indicated in tan; shoot length is dark green. (F) B- and Col seedlings grown on vertical plates for 12 days, and then returned to horizontal growth for three days. B- roots are thin, straight and hairless on vertical plates (compare to 3B), and revert to coiled growth only in tissue generated after being placed horizontally. (G) Quantification of results in (F). B- seedlings grown on vertical plates are intermediate in length between flat-grown B- seedlings and Col seedlings. (H), (I) Col and B- root tips, viewed under bright field microscopy. Scale bar = 100 μm. B- root tips contain all the cell types seen in Col root tips, but the overall morphology is blunt and rounded compared to Col.
Figure 4 AtRaptor1B-/- plants grow slowly. (A), (B) Col and B- plants at 15 days after germination on soil. (C) B- plants bolt later than Col or A-. Shown are shoots from plants 1 month after germination. (D) Growth curve of Col, A- and B- plants. The X-axis represents time after production of the first leaf. The Y-axis represents the number of rosette leaves up to 11; presence of a floral bud is 12; number of cauline leaves plus 12 is 13–16, and values above 16 are the number of shoot apices harboring flowers plus 15. B- plants show slower leaf initiation, later bolting (though at a similar rosette leaf number as Col and A-) and later flowering.
Figure 5 AtRaptor1B-/- plants show altered shoot architecture. (A) B- plant at flowering. The primary shoot apex, center, has ceased growth and is surpassed by axillary branches. Compare to Col, A- in 3C. (B) Col, A- and B- primary shoot length. (C) Mature B- plant, showing a bushy phenotype due to decreased primary shoot growth and increased branching. (D) Shoot apex of the plant in C, magnified 4.5×. Axillary branches outgrow the primary shoot, which arrests in a whorl of sterile flowers. (E) Col, A- and B- cauline and rosette branch number. (F) Col, A- and B- cauline and rosette branch length. B- primary shoots are smaller than Col or A-, and secondary shoots initiate more frequently than Col or A- but are not significantly longer than Col or A-.
Figure 6 AtRaptor accumulation pattern. (A) AtRaptor transcripts accumulate throughout the floral shoot apex, stem and differentiating floral buds. Accumulation is not confined to dividing or meristematic cells, but fades in intensity away from the apex. (B) Adjacent tissue slice, probed with actin. AtRaptor and actin transcript accumulation patterns differ. (C) In silico analysis of AtRaptor1A (left) and AtRaptor1B (right) accumulation from 1434 developmental gene chip experiments. Results are given by developmental stage (X-axis) and in terms of gene chip-normalized expression levels (Y-axis). Expression levels are shown to scale. Developmental stages are as follows: 1, 1.0–5.9 days; 2, 6.0–13.9 days; 3, 14.0–17.9 days; 4, 18.0–20.9 days; 5, 21.0–24.9 days; 6, 25.0–28.9 days; 7, 29.0–35.9 days; 8, 36.0–44.9 days; 9, 45.0–50.0 days. Analyses performed via the genevestigator website .
Figure 7 AtRaptor1A-/- 1B-/- double mutants. (A) Col, A-, B- and A-B- seedlings at seven days on growth medium with no sucrose. (B) 1A-/- 1B+/- progeny germinated on growth medium supplemented with 0%, 1% or 6% sucrose. Shown for each treatment is an A-B- seedling and an A-/- B+ sibling. A-B- seedlings on 1% sucrose show significant root growth and minimal leaf buds. Scale bar for A, B = 5 mm. (C) Quantification of results in (B). (D) A-B- root tips grown on 1% sucrose lack an epidermal cell layer. (E) A-B- roots form root hairs on 1% sucrose. Scale bar = 100 μm. Compare to 3H, I. (F, G, H, I) Scanning electron microscopy on Col, A-B- shoot apices from growth medium plates. As in (B), A-B- seedlings show minimal (SAM) activity. Primordia for leaves 1 and 2 form but do not expand significantly. Scale bar = 50 μm (F, G) or 15 μm (H, I).
Figure 8 TOR functions in two complexes in eukaryotes. (A) TOR participates in two complexes in yeast and mammals. The first of these, TORC1, regulates cell growth in response to nutrient and hormonal signals. Raptor is integral for TORC1 activity. The second of these, TORC2, regulates cytoskeletal organization. Its activity is nutrient-independent, and Raptor is not a component of TORC2. (B) Model of TOR function in plants. Embryonic development is indicated by the single horizontal arrow from zygote to seedling; meristem-driven post-embryonic development is indicated by the arrows emanating from the seedling root and shoot apices. TOR, acting independent of Raptor in a putative complex homologous to yeast and mammalian TORC2, is essential for embryonic development. TOR via TORC2 may play a role in post-embryonic development as well; the embryonic lethal AtTOR knockout phenotype precludes a definitive answer on this point. Raptor activity, in a putative plant homologue of TORC1, is dispensable for embryonic development but is essential for meristem-driven post-embryonic growth.
==== Refs
Willemsen V Sheres B Mechanisms of pattern formation in plant embryogenesis Annu Rev Genet 2004 38 587 614 15568987 10.1146/annurev.genet.38.072902.092231
Zhao Y Medrano L Ohashi K Fletcher JC Yu H Sakai H Meyerowitz EM HANABA TARANU Is a GATA transcription factor that regulates shoot apical meristem and flower development in Arabidopsis Plant Cell 2004 16 2586 600 15367721 10.1105/tpc.104.024869
Hake S Smith HM Holtan H Magnini E Mele G Ramirez J The role of knox genes in plant development Annu Rev Cell Dev Biol 2004 20 125 51 15473837 10.1146/annurev.cellbio.20.031803.093824
Poethig RS Phase change and the regulation of developmental timing in plants Science 2003 301 334 6 12869752 10.1126/science.1085328
Thomas G Sabatini D Hall M TOR Target of Rapamycin Current Topics in Microbiology and Immunology 279 2004 New York: Springer-Verlag
Jacinto E Hall MN Tor signalling in bugs, brain and brawn Nat Rev Mol Cell Biol 2003 4 117 26 12563289 10.1038/nrm1018
Schmelzle T Hall MN TOR, a central controller of cell growth Cell 2000 103 253 62 11057898 10.1016/S0092-8674(00)00117-3
Heitman J Movva NR Hall MN Targets for cell cycle arrest by the immunosuppressant rapamycin in yeast Science 1991 253 905 9 1715094
Kunz J Henriquez R Schneider U M Deuter-Reinhard Movva NR Hall MN Target of rapamycin in yeast, TOR2, is an essential phosphatidylinositol kinase homolog required for G1 progression Cell 1993 73 585 96 8387896 10.1016/0092-8674(93)90144-F
Barbet NC Schneider U Helliwell SB Stansfield I Tuite MF Hall MN TOR controls translation initiation and early G1 progression in yeast Mol Biol Cell 1996 7 25 42 8741837
Helliwell SB Howald I Barbet N Hall MN TOR2 is part of two related signaling pathways coordinating cell growth in Saccharomyces cerevisiae Genetics 1998 148 99 112 9475724
Murakami M Ichisaka T Maeda M Oshiro N Hara K Edenhofer F H Kiyama Yonezawa K Yamanaka S mTOR is essential for growth and proliferation in early mouse embryos and embryonic stem cells Mol Cell Biol 2004 24 6710 8 15254238 10.1128/MCB.24.15.6710-6718.2004
Jacinto E Loewith R Schmidt A Lin S Ruegg MA Hall A Hall MN Mammalian TOR complex 2 controls the actin cytoskeleton and is rapamycin insensitive Nat Cell Biol 2004 6 1122 8 15467718 10.1038/ncb1183
Kim DH Sarbassov DD Ali SM Latek RR Guntur KV Erdjument-Bromage H Tempst P Sabatini DM GbetaL, a positive regulator of the rapamycin-sensitive pathway required for the nutrient-sensitive interaction between raptor and mTOR Mol Cell 2003 11 895 904 12718876 10.1016/S1097-2765(03)00114-X
Kim DH Sarbassov DD Ali SM King JE Latek RR Erdjument-Bromage H Tempst P Sabatini DM mTOR interacts with raptor to form a nutrient-sensitive complex that signals to the cell growth machinery Cell 2002 110 163 75 12150925 10.1016/S0092-8674(02)00808-5
Hara K Maruki Y Long X Yoshino K Oshiro N Hidayat S Tokunaga C Avruch J Yonezawa K Raptor, a binding partner of target of rapamycin (TOR), mediates TOR action Cell 2002 110 177 89 12150926 10.1016/S0092-8674(02)00833-4
Loewith R Jacinto E Wullschleger S Lorberg A Crespo JL Bonenfant D Oppliger W Jenoe P Hall MN Two TOR complexes, only one of which is rapamycin sensitive, have distinct roles in cell growth control Mol Cell 2002 10 457 68 12408816 10.1016/S1097-2765(02)00636-6
Ginalski K Zhang H Grishin NV Raptor protein contains a caspase-like domain Trends Biochem Sci 2004 29 522 4 15450605 10.1016/j.tibs.2004.08.006
Nojima H Tokunaga C Eguchi S Oshiro N Hidayat S Yoshino K Hara K Tanaka N Avruch J Yonezawa K The mammalian target of rapamycin (mTOR) partner, raptor, binds the mTOR substrates p70 S6 kinase and 4E-BP1 through their TOR signaling (TOS) motif J Biol Chem 2003 278 15461 4 12604610 10.1074/jbc.C200665200
Sarbassov DD Ali SM Kim DH Guertin DA Latek RR Erdjument-Bromage H Tempst P Sabatini DM Rictor, a novel binding partner of mTOR, defines a rapamycin-insensitive and raptor-independent pathway that regulates the cytoskeleton Curr Biol 2004 14 1296 302 15268862 10.1016/j.cub.2004.06.054
Menand B Desnos T Nussaume L Berger F Bouchez D Meyer C Robaglia C Expression and disruption of the Arabidopsis TOR (target of rapamycin) gene Proc Natl Acad Sci U S A 2002 99 6422 7 11983923 10.1073/pnas.092141899
Robaglia C Menand B Lei Y Sormani R Nicolai M Gery C Teoule E Deprost D Meyer C Plant growth: the translational connection Biochem Soc Trans 2004 32 581 4 15270681 10.1042/BST0320581
Bonnet E Wuyts J Rouze P Van de Peer Y Detection of 91 potential conserved plant microRNAs in Arabidopsis thaliana and Oryza sativa identifies important target genes Proc Natl Acad Sci U S A 2004 101 11511 6 15272084 10.1073/pnas.0404025101
Deprost D Truong HN Robaglia C Meyer C An Arabidopsis homolog of RAPTOR/KOG1 is essential for early embryo development Biochem Biophys Res Commun 2005 326 844 50 15607746 10.1016/j.bbrc.2004.11.117
Shinozaki-Yabana S Watanabe Y Yamamoto M Novel WD-repeat protein Mip1p facilitates function of the meiotic regulator Mei2p in fission yeast Mol Cell Biol 2000 20 1234 42 10648609 10.1128/MCB.20.4.1234-1242.2000
Yonezawa K Tokunaga C Oshiro N Yoshino K Raptor, a binding partner of target of rapamycin Biochem Biophys Res Commun 2004 313 437 41 14684181 10.1016/j.bbrc.2003.07.018
Zimmermann P Hirsch-Hoffmann M Hennig L Gruissem W GENEVESTIGATOR. Arabidopsis Microarray Database and Analysis Toolbox Plant Physiol 2004 136 2621 32 15375207 10.1104/pp.104.046367
Rolland F Moore B Sheen J Sugar sensing and signaling in plants Plant Cell 2002 14 S185 205 12045277
Ward SP Leyser O Shoot branching Curr Opin Plant Biol 2004 7 73 8 14732444 10.1016/j.pbi.2003.10.002
Shpak ED CT Berthiaume Hill EJ Torii KU Synergistic interaction of three ERECTA-family receptor-like kinases controls Arabidopsis organ growth and flower development by promoting cell proliferation Development 2004 131 1491 501 14985254 10.1242/dev.01028
Alonso JM Stepanova AN Leisse TJ Kim CJ Chen H Shinn P Stevenson DK Zimmerman J Barajas P Cheuk R Gadrinab C Heller C Jeske A Koesema E Meyers CC Parker H Prednis L Ansari Y Choy N Deen H Geralt M Hazari N Hom E Karnes M Mulholland C Ndubaku R Schmidt I Guzman P Aguilar-Henonin L Schmid M Weigel D Carter DE Marchand T Risseeuw E Brogden D Zeko A Crosby WL Berry CC Ecker JR Genome-wide insertional mutagenesis of Arabidopsis thaliana Science 2003 301 653 7 12893945 10.1126/science.1086391
Clough SJ Bent AF Floral dip: a simplified method for Agrobacterium-mediated transformation of Arabidopsis thaliana Plant J 1998 16 735 43 10069079 10.1046/j.1365-313x.1998.00343.x
Jackson D Veit B Hake S Expression of maize KNOTTED1 related homeobox genes in the shoot apical meristem predicts patterns of morphogenesis in the vegetative shoot Development 1994 120 405 413
Weigel D Glazebrook J Arabidopsis : A Laboratory Manual 2002 Cold Spring Harbor Laboratory Press
Edgar R Domrachev M Lash AE Gene Expression Omnibus: NCBI gene expression and hybridization array data repository Nucleic Acids Res 2002 30 207 10 11752295 10.1093/nar/30.1.207
Rocca-Serra P Brazma A Parkinson H Sarkans U Shojatalab M Contrino S Vilo J Abeygunawardena N Mukherjee G Holloway E Kapushesky M Kemmeren P Lara GG A Oezcimen Sansone SA ArrayExpress: a public database of gene expression data at EBI C R Biol 2003 326 1075 8 14744115
Parkinson H Sarkans U Shojatalab M Abeygunawardena N Contrino S Coulson R Farne A Garcia Lara G Holloway E Kapushesky M Lilja P Mukherjee G Oezcimen A Rayner T Rocca-Serra P Sharma A Sansone S Brazma A ArrayExpress – a public repository for microarray gene expression data at the EBI Nucleic Acids Res 2005 33 D553 5 15608260 10.1093/nar/gki056
Craigon DJ James N Okyere J Higgins J Jotham J May S NASCArrays: a repository for microarray data generated by NASC's transcriptomics service Nucleic Acids Res 2004 D575 7 14681484 10.1093/nar/gkh133
| 15845148 | PMC1131892 | CC BY | 2021-01-04 16:02:56 | no | BMC Biol. 2005 Apr 21; 3:12 | utf-8 | BMC Biol | 2,005 | 10.1186/1741-7007-3-12 | oa_comm |
==== Front
BMC BiotechnolBMC Biotechnology1472-6750BioMed Central London 1472-6750-5-101585047810.1186/1472-6750-5-10Methodology ArticleDetection of target DNA using fluorescent cationic polymer and peptide nucleic acid probes on solid support Raymond Frédéric R [email protected] Hoang-Anh [email protected] Régis [email protected] Luc [email protected] Maurice [email protected] François J [email protected] Mario [email protected] Michel G [email protected] Centre de recherche en infectiologie de l'Université Laval, Centre hospitalier universitaire de Québec, Pavillon CHUL, Sainte-Foy, Québec, G1V 4G2, Canada2 Canada Research Chair in Electroactive and Photoactive Polymers, Département de Chimie, Université Laval, Sainte-Foy, Québec, G1K 7P4, Canada2005 25 4 2005 5 10 10 17 8 2004 25 4 2005 Copyright © 2005 Raymond 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
Nucleic acids detection using microarrays requires labelling of target nucleic acids with fluorophores or other reporter molecules prior to hybridization.
Results
Using surface-bound peptide nucleic acids (PNA) probes and soluble fluorescent cationic polythiophenes, we show a simple and sensitive electrostatic approach to detect and identify unlabelled target nucleic acid on microarray.
Conclusion
This simple methodology opens exciting possibilities for applied genetic analysis for the diagnosis of infections, identification of genetic mutations, and forensic inquiries. This electrostatic strategy could also be used with other nucleic acid detection methods such as electrochemistry, silver staining, metallization, quantum dots, or electrochemical dyes.
==== Body
Background
Classical strategies for nucleic acid detection require labelling the analyte or the probes with fluorophores or other reporter molecules [1,2]. Labelling steps add cost, often complexify the reaction mixture, and are not useful when rapidity of the assay is essential, such as for some molecular diagnostic applications. Fluorescent conjugated polymers have already been shown to allow the sensitive detection of DNA and RNA in liquid phase by complexation of the cationic fluorescent polymer to negatively-charged nucleic acids [3-8]. The use of DNA/RNA-binding fluorescent polymers on microarray could allow unlabelled nucleic acid detection, and thus prevent the need for a nucleic acid labelling step. Using PNA capture probes, we show a simple and sensitive electrostatic approach which enables the direct detection and specific identification of unlabelled target nucleic acid analyte using a standard microarray scanner.
Results and discussion
Experiments with commercially available aldehyde-functionalized glass slides, while giving strong signal when fluorophore-labelled target DNA hybridized to DNA probes, gave no signal when detection was conducted using PNA probes (data not shown). We developed aminoalkylsilane slides activated with carbonyldiimidazole and compared them with commercial aldehyde slides. Our chemistry was compatible with detection on PNA capture probes using Cy3-labelled oligonucleotides or using the fluorescent cationic polymer, while aldehyde slides gave no signal. The aminated slides activated by carbodiimidazole were thus chosen to immobilize both DNA and PNA capture probes for all experiments described here.
Fluorescent conjugated polythiophene derivative used in this study binds nucleic acids via electrostatic interactions with negatively-charged phosphate groups of the DNA backbone [4,5]. This is illustrated in Figure 1 where single-stranded and double-stranded DNA oligonucleotides (Figure 1a and 1b) both produce fluorescent signals due to the formation of DNA-polythiophene complexes. Thus, discrimination between hybridized and non-hybridized DNA capture probes is not possible using a conventional microarray fluorescence scanner.
Capture probes of neutral charge, such as PNA [9-14], have been successfully used in microarray experiments [6,15]. Since PNA do not have a charged backbone, they could be used to allow the detection of nucleic acids with cationic polymeric biosensors such as polythiophenes. Using PNA as capture probes provide a simple and sensitive electrostatic approach on solid support which enables the direct detection and specific identification of unlabelled target nucleic acid analyte with a standard microarray scanner.
The cationic polymer does not bind to unhybridized neutral PNA capture probes (Figure 1c), but strongly interacts with the negatively-charged backbone of the complementary oligonucleotides bound to PNA probes, allowing transduction of hybridization into a fluorescence signal (Figure 1d). This clearly demonstrates the usefulness of PNA capture probes for the detection of hybridization events with positively-charged fluorescent conjugated polymers on solid support. In a recent study, Gaylord et al. have shown detection in solution of a complementary DNA hybridized to a PNA probe using Förster resonance energy transfer (FRET) between a water soluble conjugated polymer and a PNA probe labelled with a reporter chromophore [6]. The present study shows similar results without the need for labelled PNA. Moreover, we demonstrate that this detection can be performed using PNA capture probes tethered onto a solid support. Those results suggest that this electrostatic strategy could also be used with other nucleic acid detection methods such as electrochemistry [16], silver staining [17,18], metallization [19,20], quantum dots [21-23], or electrochemical dyes [24].
Specificity of detection was investigated by hybridizing mismatched oligonucleotides to PNA probes. After room temperature hybridization of oligonucleotides with PNA probes, the fluorescent polythiophene polymeric biosensor gave a strong signal over background when target oligonucleotide was fully complementary to the capture probe. Oligonucleotides with two mismatches and non complementary oligonucleotides produced near-background signals easily distinguishable from the much stronger signal observed with perfectly matched hybrids (21-fold stronger) (Figure 2). For a single mismatch, discrimination is related to the position of the mismatch in the capture PNA probe. When the mismatch is located at the probe extremity, signal intensity is reduced 2.5 fold as compared to the perfect match. By contrast, a ratio of 6 is observed when the mismatch is located close to the center of the capture probe (Figure 2).
The sensitivity of the detection scheme described here is approximately 2.5 × 10-13 mole of oligonucleotide in a volume of 20 μL. In a recent report, Nilsson and Inganäs [8] have described the use of a zwitterionic polythiophene derivative able to detect 2 × 10-8 mole of oligonucleotide within a hydrogel matrix. Our approach, based on standard glass slide microarray technologies, is approximately five orders of magnitude more sensitive. Moreover, it is expected that further progress in terms of sensitivity should be obtained by reducing the size of the microarray spots and of the hybridization reaction volume. Also, the detection of larger DNA molecules (e.g. amplicons) should increase sensitivity since the amount of complexed fluorescent cationic polymer is theoretically proportional to the amount of possible electrostatic interactions. The development of scanners specifically fabricated to excite at the wavelength of this fluorescent cationic polymer should also contribute to increase the analytical sensitivity. Indeed, recent optimizations of the fluorometric detection applied to our polymer technology has enabled the detection of only a few hundred molecules of genetic material in 3 mL of aqueous solution [3]. This clearly indicates the great potential of cationic conjugated polymers as highly sensitive fluorescent transducers.
Conclusion
In this study, we presented a simple and specific nucleic acid detection method onto solid support which does not require labelling of the analyte prior to hybridization. This methodology opens new possibilities for genetic analysis applied for the diagnosis of infections, identification of genetic mutations, and forensic inquiries. For instance, this technology would be useful for the identification of pathogens and related antimicrobial resistance genotypes using microarrays. Finally, the electroactivity of the present fluorescent cationic polymer could be exploited for a real-time electrical discrimination of single nucleotide polymorphisms (SNPs) onto solid support.
Methods
Materials
All chemical reagents were obtained from Sigma-Aldrich Co. (St. Louis, MO) and were used without further purification unless otherwise stated. Fluorescent polythiophene used in this study, poly(1H-imidazolium, 1-methyl-3-[2-[(4-methyl-3-thienyl)oxy]ethyl]-, chloride), was prepared following published procedures [3,4]. Oligodeoxyribonucleotide capture probes, which were 5'-modified by the addition of two nine carbon spacers and an amino-linker, were synthesized by Biosearch Technologies (Novato, CA). PNA capture probes having a N-terminal amine and two O linkers were synthesized by Applied Biosystems (Foster City, CA). The amino-linker modification allowed covalent attachment of probes onto a functionalized glass surface. The capture DNA or PNA probe of 15-mer (5'-CCGCTCGCCAGCTCC-3') targeted a polymorphic region of the blaSHV-1 gene associated with β-lactam antibiotic resistance. Target oligonucleotides (i) fully complementary to the capture DNA or PNA probe (5'-GGAGCTGGCGAGCGG-3'), (ii) having two mismatched bases (5'-GGCGCTGACGAGCGG-3'), (iii) having a central single mismatch (5'-GGAGCTGACGAGCGG-3'), (iv) having a single mismatch at one extremity (5'-GGCGCTGGCGAGCGG-3'), and (v) non complementary (5'-CGCTCTGCTTTGTTATTCGG-3') were synthesized by Biosearch Technologies.
Preparation of glass slide
The aldehyde-functionalized commercial glass slides were purchased from CEL Associates (Pearland, TX). The home made aminoalkylsilane-functionalized glass slides were prepared as follows. All chemical reactions were carried out in polypropylene jars. Surfaces used were 25 mm × 75 mm microscope glass slides (VWR International, West Chester, PA). After sonication (1 hour) in deionized water, the slides were sonicated in 40 mL of 10% sodium hydroxide for 1 hr, washed several times with deionized water, and dried under a stream of nitrogen. The slides were then sonicated in an aminopropyltrimethoxysilane solution (2 mL water, 38 mL methanol and 2 mL aminopropyltrimethoxysilane) for 1 hr, washed with methanol, dried and baked at 110°C for 15 min. The amine modified slides were activated by sonication overnight in 40 mL of 1.4-dioxane containing 0.32 g (2 mmoles) of carbonyldiimidazole as coupling agent, washed with dioxane and diethyl ether, and dried under a stream of nitrogen.
Microarray production
The probes were diluted two-fold by the addition of Array-it Microspotting Solution Plus (TeleChem International, Sunnyvale, CA), to a final concentration of 5 μM. Capture probes were spotted in triplicate, using a SDDC-2 arrayer (Bio-Rad Laboratories, Hercules, CA) with SMP3 pins (TeleChem International). Upon spotting, each spot had a volume of 0.6 nL, a diameter between 140 and 150 μm, and contained approximately 1.8 × 109 amino-modified probes. After spotting, slides were dried overnight, washed by immersion in boiling 0.1% Igepal CA-630 (Sigma-Aldrich, St. Louis, MO) for 5 min, rinsed in ultra-pure water for 2 min, and dried by centrifugation for 5 min under vacuum (SpeedVac plus; Thermo Savant, Milford, MA). Slides were stored at room-temperature in a dry and oxygen-free environment.
DNA microarray hybridization, polymeric detection and data acquisition
Prehybridization and hybridization were performed in 15 × 13 mm Hybri-well self-sticking hybridization chambers (Sigma-Aldrich). Microarrays were first prehybridized for 15 min at room temperature with 20 μL of 1X hybridization solution (6X SSPE [Omnipur, EM Science, Gilbstown, NJ], 0.03% polyvinylpyrrolidone [PVP], and 30% formamide). Subsequently, the prehybridization buffer was blown out of the chambers and replaced with the same buffer containing the target oligonucleotide (complementary, central mismatch or two mismatches) at a final concentration of 2.5 μM, except for sensitivity experiments for which concentrations ranged from 0.25 nM to 2.5 μM. Hybridization was carried out at 22°C for 15 min. After hybridization, the liquid was expelled from the chambers and replaced by an aqueous solution of cationic polymer (7.3 × 10-4 M based on monomeric units). After a 15 min incubation period, the slides were washed with deionized water containing 0.1 % Igepal CA-630. Microarrays were dried by centrifugation at 1350 × g for 3 min. Slides were scanned using the Cy3 configuration (excitation wavelength at 530 nm) of ScanArray 4000XL (Packard Bioscience Biochip Technologies, Billerica, MA) and the fluorescent signals were analyzed using QuantArray software (Packard Bioscience Biochip Technologies).
Authors' contributions
FRR participated in the design of the experiments, performed all microarray experiments and co-drafted the manuscript with HAH. HAH synthesized the cationic polythiophene, worked on surface chemistry, and co-drafted the manuscript with FRR. RP supervised some of the work and participated in the design of the experiments. LB conceived the study and participated in the design of the experiments. RP, MB, FJP, ML, and MGB provided guidance and suggestions for experimental design, analyzed data and edited the manuscript. ML holds the Canada Research Chair in Electroactive and Photoactive Polymers. MGB is head of the Centre de recherche en infectiologie de l'Université Laval. All authors read and approved the final version of the manuscript.
Acknowledgements
This work was supported by the Natural Sciences and Engineering Research Council of Canada, the Canadian Institutes of Health Research, Infectio Diagnostic Inc. (Sainte-Foy, Québec, Canada), Génome Québec and Genome Canada, and the Chemical, Biological, Radiological and Nuclear Research and Technology Initiative.
Figures and Tables
Figure 1 Cationic polythiophene transducer for the fluorometric detection of hybridization on microarrays. A) Schematic depiction of the interaction between cationic polymers and a) single-stranded DNA, b) double-stranded DNA, c) single-stranded PNA and d) PNA-DNA duplex. Fluorescent cationic polymer is shown in yellow, DNA probes are shown in green and PNA probes are shown in red. B) Experimental results for fluorometric detection on microarray when cationic polythiophene transducer is reacted with a) single-stranded DNA, b) double-stranded DNA, c) single-stranded PNA and d) PNA-DNA duplex. Results are shown in triplicate. C) Graphs showing the fluorescence intensity with standard deviation for each triplicate shown in B.
Figure 2 Specificity of oligodeoxyribonucleotide hybridization to PNA probes. Hybridizations were performed at room temperature with a concentration of 7.5 × 1010 targets per μL using the fluorescent cationic polymer for detection. Hybridization of PNA probes to perfectly complementary, or complementary oligonucleotides presenting a terminal mismatch, a central mismatch, or two mismatches were performed in triplicate. Fluorescence intensities from hybridized probes were corrected by substraction of background fluorescence intensity.
==== Refs
Pirrung MC How to make a DNA chip Angew Chem Int Ed 2002 41 1276 1289
Epstein JR Biran I Walt DR Fluorescence-based nucleic acid detection and microarrays Anal Chim Acta 2002 469 3 36
Doré K Dubus S Ho HA Lévesque I Brunette M Corbeil G Boissinot M Boivin G Bergeron MG Boudreau D Fluorescent polymeric transducer for the rapid, simple, and specific detection of nucleic acids at the zeptomole level J Am Chem Soc 2004 126 4240 4244 15053613
Ho HA Boissinot M Bergeron MG Corbeil G Doré K Boudreau D Leclerc M Colorimetric and fluorometric detection of nucleic acids using cationic polythiophene derivatives Angew Chem Int Ed 2002 41 1548 1551
Ho HA Leclerc M New colorimetric and fluorometric chemosensor based on a cationic polythiophene derivative for iodide-specific detection J Am Chem Soc 2003 125 4412 4413 12683798
Gaylord BS Heeger AJ Bazan GC DNA detection using water-soluble conjugated polymers and peptide nucleic acid probes Proc Natl Acad Sci U S A 2002 99 10954 10957 12167673
Gaylord BS Heeger AJ Bazan GC DNA hybridization detection with water-soluble conjugated polymers and chromophore-labeled single-stranded DNA J Am Chem Soc 2003 125 896 900 12537486
Nilsson KP Inganas O Chip and solution detection of DNA hybridization using a luminescent zwitterionic polythiophene derivative Nat Mater 2003 2 419 424 12754497
Livache T Roget A Dejean E Barthet C Bidan G Teoule R Preparation of a DNA matrix via an electrochemically directed copolymerization of polypyrrole and oligonucleotide bearing a pyrrole group Nucl Acids Res 1994 22 2915 2921 8065902
Nielsen PE Applications of peptide nucleic acids Curr Opin Biotechnol 1999 10 71 75 10047504
Nielsen PE Egholm M An introduction to peptide nucleic acid Curr Issues Mol Biol 1999 1 89 104 11475704
Ratilainen T Holmen A Tuite E Haaima G Christensen L Nielsen PE Norden B Hybridization of peptide nucleic acid Biochemistry 1998 37 12331 12342 9724547
Stender H Fiandaca M Hyldig-Nielsen JJ Coull J PNA for rapid microbiology J Microbiol Methods 2002 48 1 17 11733079
Wang J PNA biosensors for nucleic acid detection Curr Issues Mol Biol 1999 1 117 122 11475696
Brandt O Feldner J Stephan A Schroder M Schnolzer M Arlinghaus HF Hoheisel JD Jacob A PNA microarrays for hybridisation of unlabelled DNA samples Nucl Acids Res 2003 31 e119 14500847
Liu A Anzai J Use of polymeric indicator for electrochemical DNA sensors: poly(4-vinylpyridine) derivative bearing [Os(5,6-dimethyl-1,10-phenanthroline)2Cl]2+ Anal Chem 2004 76 2975 2980 15144212
Braun E Eichen Y Sivan U Ben-Yoseph G DNA-templated assembly and electrode attachment of a conducting silver wire Nature 1998 391 775 778 9486645
Brust M Kiely CJ Some recent advances in nanostructure preparation from gold and silver particles: a short topical review Colloids Surfaces 2002 202 175 186
Warner MG Hutchison JE Linear assemblies of nanoparticles electrostatically organized on DNA scaffolds Nat Mat 2003 2 272 277
Storhoff JJ Mirkin CA Programmed materials synthesis with DNA Chem Rev 1999 99 1849 1862 11849013
Alivisatos P Colloidal quantum dots. From scaling laws to biological applications Pure Appl Chem 2000 72 3 9
Chan WC Nie S Quantum dot bioconjugates for ultrasensitive nonisotopic detection Science 1998 281 2016 2018 9748158
Penner RM Hybrid electrochemical/chemical synthesis of quantum dots Acc Chem Res 2000 33 78 86 10673315
Kricka LJ Stains, labels and detection strategies for nucleic acids assays Ann Clin Biochem 2002 39 114 129 11928759
| 15850478 | PMC1131893 | CC BY | 2021-01-04 16:02:57 | no | BMC Biotechnol. 2005 Apr 25; 5:10 | utf-8 | BMC Biotechnol | 2,005 | 10.1186/1472-6750-5-10 | oa_comm |
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-441586212710.1186/1471-2407-5-44Research ArticleA phase I study of hydralazine to demethylate and reactivate the expression of tumor suppressor genes Zambrano Pilar [email protected] Blanca [email protected] Enrique [email protected] Lucely [email protected] Alma [email protected] Lucía [email protected] Alma [email protected] Enrique [email protected] Gustavo [email protected] Karina [email protected] Catalina [email protected] Jose [email protected]ález Alfonso [email protected] Unidad de Investigación Biomédica en Cáncer, Instituto Nacional de Cancerología/Instituto de Investigaciones Biomédicas, UNAM, Mexico2 Division of Clinical Research, Instituto Nacional de Cancerología, Mexico3 Laboratorio de Desarrollo de Métodos Analíticos, FES-Cuautitlán, UNAM, Mexico4 Laboratorio de Química Medicinal FES-Cuautitlán, UNAM, Mexico2005 29 4 2005 5 44 44 25 1 2005 29 4 2005 Copyright © 2005 Zambrano 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 antihypertensive compound hydralazine is a known demethylating agent. This phase I study evaluated the tolerability and its effects upon DNA methylation and gene reactivation in patients with untreated cervical cancer.
Methods
Hydralazine was administered to cohorts of 4 patients at the following dose levels: I) 50 mg/day, II) 75 mg/day, III) 100 mg/day and IV) 150 mg/day. Tumor biopsies and peripheral blood samples were taken the day before and after treatment. The genes APC, MGMT; ER, GSTP1, DAPK, RARβ, FHIT and p16 were evaluated pre and post-treatment for DNA promoter methylation and gene expression by MSP (Methylation-Specific PCR) and RT-PCR respectively in each of the tumor samples. Methylation of the imprinted H19 gene and the "normally methylated" sequence clone 1.2 was also analyzed. Global DNA methylation was analyzed by capillary electrophoresis and cytosine extension assay. Toxicity was evaluated using the NCI Common Toxicity Criteria.
Results
Hydralazine was well tolerated. Toxicities were mild being the most common nausea, dizziness, fatigue, headache and palpitations. Overall, 70% of the pretreatment samples and all the patients had at least one methylated gene. Rates of demethylation at the different dose levels were as follows: 50 mg/day, 40%; 75 mg/day, 52%, 100 mg/day, 43%, and 150 mg/day, 32%. Gene expression analysis showed only 12 informative cases, of these 9 (75%) re-expressed the gene. There was neither change in the methylation status of H19 and clone 1.2 nor changes in global DNA methylation.
Conclusion
Hydralazine at doses between 50 and 150 mg/day is well tolerated and effective to demethylate and reactivate the expression of tumor suppressor genes without affecting global DNA methylation
==== Body
Background
Cancer is considered to be a disease of the genome that results from a plethora of genetic and epigenetic lesions. Among the epigenetic alterations, DNA hypermethylation is thought to play an important role in tumor development and progression [1]. In this regard, at least three functional DNA methyltransferases (DNMTs) have been identified, the most abundant is DNMT1 which preferentially methylates hemi-methylated DNA [2], and plays a key role in imprinting and X-chromosome inactivation during embryogenesis [3,4]. DNTM1 localizes to replication foci [5], at least in part by interacting with proliferating cell nuclear antigen (PCNA), a protein closely involved in DNA replication. It is therefore responsible for maintaining proper methylation levels during replication and possibly repair [6]. Other known functional methyltransferases are DNMT3a and DNMT3b, which are responsible for de novo methylation during embryogenesis [7]. DNMT3a and DNMT3b have equal preferences for hemi-methylated and non-methylated DNA, and so have been classified as de novo methyltransferases [8]. DNA methylation can directly interfere with transcriptional factor binding and thus inhibit replication [9], with methyl-CpG binding proteins which bind methylated DNA and with regulatory proteins that inhibit transcription [10]. In addition, both DNMT1 and methyl-binding proteins (MBP), such as methyl-CpG-binding protein 2 (MeCP2) recruit histone deacetylases which deacetilate histone core tails leading to tighter chromatin packaging, reducing the access of transcriptional factors to DNA [11,12].
Cancer cells are considered to have global hypomethylation and regional hypermethylation. Hypermethylated regions are CpG islands, CpG and GpC rich sequences 1 kb long found proximal to gene promoters involved in transcriptional control [13]. These islands are associated with roughly half of all genes [15], their methylation can repress transcription in a manner analogous to a mutation or deletion (16). It is thought that tumor suppressor gene promoter hypermethylation contributes to their transcriptional silencing [14]. Furthermore, there is a growing list of tumor suppressor genes in both sporadic and familial cancers which are found to be transcriptionally silenced by hypermethylation [17].
In this regard, tumor suppressor gene transcriptional reactivation through promoter de-methylation represents an attractive strategy for anticancer treatment. Substantial preclinical studies characterizing DNA methylation inhibitors have shown cancer cell line growth arrest in vitro and antitumor effects in animal models, including survival prolongation [18-20]. These concepts are supported by the transforming effect of exogenous DNA methyltransferase gene expression observed in fibroblasts [21] as well as by the malignant phenotype reversion documented using antisense oligonucleotides against this gene [22]. These findings have paved the way for the clinical testing of demethylating agents in cancer.
Nucleoside deoxycytidine analogs formerly known as classic cytotoxic agents and later known as DNA methylation inhibitors show poor activity against solid tumors [23] however, 5-aza-2'-deoxycytidine has recently gained considerable attention and is presently being tested as a demethylating agent for the treatment of hematological neoplasms [24]. MG98, is an antisense oligodeoxynucleotide directed against the 3' untranslated region of the DNA methyltransferase-1 enzyme mRNA that has been tested in clinic [23]. A phase I study using biweekly administration of this agent, showed no consistent decrease of mRNA levels in the peripheral blood cells of patient [25]. Although this agent has shown activity in xenografts models of nude mice, demonstration of antitumor efficacy in humans is pending.
Our group has recently shown in vitro and in vivo promoter demethylation and tumor suppressor gene transcriptional reactivation mediated by the antihypertensive compound hydralazine [26]. Its DNA demethylating activity can be explained by the interaction between its Nitrogen atoms with residues Lys162 and Arg240 of the DNA methyltransferase active site as showed in a silico model [27]. Hydralazine is a well-tolerated drug devoid of the common side effects of cytotoxic chemotherapy agents, however, its hypotensive effects could limit its use in a clinical setting, we thus felt desirable to determine the dose at which its demethylating activities were observed in a set of genes known to be methylated in cervical cancer.
Methods
Patient selection
Previously untreated patients with histological diagnosis of carcinoma of the cervix were entered into this phase I study. The following inclusion criteria were applied: 1) age between 18 and 75 years; 2) Karnofsky status 70% or higher; 3) hematological, renal and hepatic functions as follows: hematological: Hemoglobin equal or higher than 10 g/L, leukocytes >4000/mm3, platelets >100 000/mm3, total bilirubin and transaminases <1.5× the normal upper limit, and normal serum creatinine; 4) a normal chest X-ray and 5) signed informed consent. Exclusion criteria were: 1) history of allergy to hydralazine; 2) any past or current cardiovascular condition (higher blood pressure, heart failure, etc., that required pharmacological treatment; 3) any past or current rheumatic or autoimmune disease; 4) uncontrolled infection or other systemic diseases; 5) concomitant treatment with any experimental drug; 6) pregnant or nursing women; 7) mental illness; and 8) previous or concomitant malignant diseases other than non-melanoma skin cancer. The Institutional Regulatory Board approved the study protocol.
Clinical samples and nucleic acids extraction
Biopsies were taken from areas with visible macroscopic cervical tumor using a sterile biopsy punch the day before and the day after the 10 days of hydralazine treatment. The specimen was immediately frozen at -20°C for further processing. In addition, a blood sample of 10 mL was drawn from the arm by venipuncture for subsequent mononuclear cell DNA extraction. In addition, the surgical specimen of an early stage cervical cancer patient undergoing radical hysterectomy was multi-sampled in the surgical room by taking 4 small fragments of the macroscopic tumor from separate areas. Genomic DNA was isolated from the tumor tissues and from the buffy coat layer of blood samples using the standard method of proteinase K digestion and phenol-chloroform extraction. RNA from tumor biopsies was obtained using the TriReagent (Gibco BRL Grand Island, New York) RNA extraction kit following the manufacturer instructions.
Hydralazine treatment
Twenty four hours after tumor and blood sampling patients where divided into the following groups and started on oral hydralazine for a ten day period: I) 25 mg every 12 hours, II) 25 mg every 8 hours; III) 50 mg every 12 hours; and IV) 50 mg every 8 hours. Toxicity was assessed at the end of study period (10 days), with special emphasis on the presence, throughout the treatment days, of known signs and symptoms associated with hydralazine treatment (hypotension, tachycardia, palpitations, syncope, sweating, headache, dizziness and fluid retention). The study period ended at day ten and patients went to receive definitive treatment.
Analysis of DNA methylation in the tumor
The methylation status of genes was determined in the biopsies pre and post-treatment. The gene set studied was: p16, RARβ, MGMT, ER, FHIT, APC, DAPK and GSTp1 which were analyzed by methylation-specific PCR as previously described [28]. In addition, the four fragments of the tumor of the untreated patient were analyzed for methylation of the DAPK gene. Briefly, 1 μg of DNA in a volume of 100 μl of each sample was denaturated with freshly prepared NaOH at a final concentration of 0.2 M, and modified according to the manufacturer instructions of the DNA Modification kit (Intergen, Purchase, New York). The PCR mixture contained 2 μl of 10X PCR buffer, 0.5 U of Taq Gold polymerase, dNTPs (each 1.25 mM), 300 ng of primers and bisulfite-modified DNA in a final volume 20 μl. Products were visualized in a 2% agarose gel under UV light. The primers and conditions for PCR are shown in Table 1. A gene was deemed methylated whenever a band was present after amplification with the methylated, or both methylated and unmethylated set of primers. On the contrary, a gene was deemed unmethylated when a band was present after amplification with the unmethylated set in the absense of band with the methylated set. An exception to these criteria was when in the pretreatment biopsy there was only a methylated band but both unmethylated and methylated bands after treatment.
Analysis of DNA methylation in mononuclear cells of peripheral blood
The methylation status of the imprinted H19 gene as well as the clone 1.2 which is known to be normally methylated and not subjected to imprinting [29] was also analyzed by methylation specific PCR. The positive control for unmethylation in the clone 1.2 sequence was DNA from cultured cells treated with ethionine. Primers and conditions are shown in Table 1.
Analysis of gene expression
Total RNA obtained from pre and post-treatment biopsies was reverse transcribed using a RT-PCR kit (Perkin Elmer; Branchburg, New Jersey) following the manufacturer instructions. Primers and conditions for amplifications are shown in Table 2.
Quantification genomic 5-methylcytosine DNA content by capillary electrophoresis
Adenine, thymine, guanine, uracyl, cytosine and 5-methyl-cytosine were purchased from Sigma (St. Louis, MO; USA). Bases were dissolved at 2.4 mM in 0.1 M HCL, 0.01 M HCL, or Milli-Q-grade water (Millipore, Bedford, MA) and filtered through 0.45 μm pore size filters (Millipore). Briefly, extracted DNA samples were resuspended in TE buffer at (1 μg/μl). Hydrolysis of DNA was carried out by incubating 40 μg of DNA in 2 mL 88% v/v formic acid at 140°C into a sealed ampoule for 90 min. After hydrolysis, samples were reduced to dryness by speedvac concentration and redissolved in 30 μL of Milli-Q-grade water. DNA samples of pre and post-treatment biopsies were analyzed for 5-methylcytosine content by capillary electrophoresis (CE) as reported by us [30]. For the CE procedure, an uncoated fused-silica capillary (Beckman-Coulter; 60 cm × 75 μm; effective length, 44.5 cm) was used in a CE system (P/ACE MDQ; Beckman-Coulter) connected to a data-processing station (32 Karat software). The running buffer was 20 mM NaCO3 (pH 9.6 ± 1) containing 80 mM SDS. Running conditions were 25°C with an operating voltage of 20 kV. On-column absorbance was monitored at 223 nm. Before each run, the capillary system was conditioned by washing with the running buffer for 2 min. Buffers and washing solutions were prepared with Milli-Q water and filtered throughout 0.45-μm filters. Hydrolyzed samples, previously filtered through 0.45 μm pore filters, were injected under pressure (0.5 p.s.i.) for 15s. The relative methylation of each DNA sample was taken as the percentage of mC in total cytosine: mC peak area × 100/(C peak area + mC peak area).
Methylation pattern in global DNA
Global DNA methylation was evaluated by the cytosine-extension assay [31] which is based on the use of methylation-sensitive restriction endonuclease that leave a 5' guanine overhang after DNA cleavage, with subsequent single nucleotide extension with radiolabeled [(3)H]dCTP. Briefly 1 μg of DNA from tumor tissue was digested overnight with BssHII according to the manufacturer. Single nucleotide extension reaction was performed in a 25 μl reaction mixture containing 0.25 μg of DNA, 1X buffer II, 1 mM MgCl, 0.25 U of DNA polymerase (Perkin Elmer; Branchburg, New Jersey [3H]dCTP (Ci/mmol) and incubated at 56°C for 1 hour, then placed on ice. The reaction mixture was then applied to sephadex G25 column. For the column chromatography, each sephadex G25 column was centrifuged for 10 sec., at 5000 rpm to remove the buffer and loaded with the reaction mixture. After loading the samples to the column, the radiolabeled DNA was collected by centrifugation of the column and mixed with liquid scintillation for the determination of radioactivity. Results were expressed as the percentage change from the average controls which were tumor samples of patients pre and post-treatment without BssHII digestion.
Results
Study group
A total of 16 patients were studied. All of them were chemotherapy or radiation naive and had a macroscopic tumor accessible for punch biopsy. Their mean age was 51.8 (35–75 years), all cases were squamous histology and staged as FIGO stage IIB and IIIB. The status performance was 0–1 in all patients (Table 3).
Treatment compliance and side effects
All patients completed the prescribed medication and reported that hydralazine treatment was well tolerated. Side effects as evaluated by the Common Toxicity Criteria are showed in Table 4. Grade 1 fatigue, headache and palpitations were observed in 50% of patients; grade 1 nausea was present in 37% of cases, being grade 2 in only one patient. Only 2 out of 16 patients presented dizziness grades 1 and 2 respectively. No patient stopped treatment due to adverse effects. We observed no side-effect dose relationship.
Gene promoter methylation
Methylation analysis of the biopsies taken before and after 10 days of hydralazine administration was performed in all 16 patients. Methylation results involving the genes analyzed were variable. Overall, 70% (89 out of 128) of the pretreatment samples analyzed (8 genes for each of the 16 pre-treatment biopsies) had at least one methylated gene, and all 16 patients had at least one methylated gene. Analysis by individual genes showed the following rates of methylated genes: APC (94%), ER (25%) FHIT (88%), GSTp1 (88%), MGMT(81%), p16(19%), RARβ (62%), and DAPK (100%), irrespective of the dose of hydralazine used, the post-treatment biopsies showed a variable demethylation rate according to the gene, varying from 15% (2/13 samples) for the MGMT gene, to 67% of demethylation for the p16 gene (2 out of 3 samples), (Figure 1A). Representative cases of the results are shown in Figure 1B. Correlation between demethylation analysis and dose level revealed the following results: at 50 mg a day, 40% of methylated genes suffered demethylation; at 75 mg it was 52%, at 100 mg dose the rate was 43%, and 32% for the 150 mg dose (Figure 1C).
Gene expression
All cases showed expression of β actin. Gene expression analysis showed that 90% (116 out of 128) of the tumor samples expressed the messenger in the pre-treatment and post-treatment biopsies regardless of methylation status, hence were not informative. On the other hand, there were only 12 informative cases. Of these 12 cases, three (25%) (having only the methylated band pretreatment) showed no re-expression after treatment (2/ DAPK, 1/GSTp1 cases). These three cases received the 50 mg dose. In the remaining 9 cases (75%) expression of the gene was re-induced after treatment. These nine cases behaved as follows: Five cases were RT-PCR negative pretreatment (only methylated band) and converted positive post-treatment, displaying methylated and unmethylated bands (FHIT two cases -100 and 150 mg doses- MGMT two cases -75 and 150 mg doses, GSTp1 one case -50 mg dose). Three cases RT-PCR negative pre-treatment with methylated band which converted to unmethylated and expression positive in the post-therapy biopsy (DAPK and ER genes at 75 mg, GSTp1 at 150 mg dose), and finally, a RT-PCR negative in the pre-treatment biopsy despite having methylated and unmethylated bands which converted to RT-PCR positive accompanied by only unmethylated band (GSTp1 case at 75 mg). Table 5, Figure 2D. Three representative cases are shown in Figure 2A, B, C.
Methylation heterogeneity in tumor samples
Because the sampling of tumor may lead to tissue specimens with different degree of "contamination" with non-malignant that may affect the result of methylation in the post-treatment, biopsies, all samples were analyzed by a pathologist. The results showed that in all biopsy samples the content of malignant cells varied from 30 to 70%. A representative set is shown in Figure 3A. In addition, we reasoned that if tumor heterogeneity were of a magnitude to yield different methylation patterns in the tumors regardless of the treatment, then multisampling a tumor with subsequent methylation analysis of each of the tumor fragments will render zones with different methylation pattern. The analysis of the surgical specimen of an untreated patient showed that the methylation of the DAPK gene promoter was the same in the four fragments of the tumor. Figure 3B.
Methylation of imprinted and "normally methylated" genes
The methylation status of the imprinted gene H19 was investigated in the DNA extracted from peripheral blood cells. As shown in Figure 4, no change in the expected pattern of bands was observed after treatment, all patients showed bands with the methylated and unmethylated set of primers. A consistent pattern of methylation also was observed in all patients analyzed for the "normally methylated" sequence clone 1.2. No case showed demethylation at this locus.
Global methylation
Global tumor DNA methylation was evaluated by two methods, capillary electrophoresis and cytosine extension. Enough DNA to perform capillary electrophoresis was only available in five cases (four patients taking 100 mg, and one taking 150 mg/day respectively). The relative methylation of each DNA sample was taken as the percentage of dmC in total cytosine. Figure 5A shows the relative methylation in theses cases. Remarkably, relative methylation in all cases was between 34.2 and 38.4% and there was no change in methylation levels after hydralazine treatment (37.3% ± 0.81 and 36.3% ± 1.1 pre and post-treatment respectively). An electropherogram showing the separation of the analytes of interest, C and mC peak are shown in Figure 5B. In the cytosine extension assay, the extent of [(3)H]dCTP incorporation after restriction enzyme treatment is directly proportional to the number of unmethylated (cleaved) CpG sites. The results of this assay showed no statistically significant changes in the tumor samples pre and post-treatment as the percentage increase in radiolabel incorporation were 15% ± 31.3% and 7% ±16.7% respectively, Figure 5C.
Discussion
Malignant tumors frequently silence genes that control cell growth, differentiation and apoptosis through DNA methylation at their promoter regions; consequently, their reactivation by DNA methylation inhibitors has raised considerable interest as anti-tumor therapy. The present focus on DNA methylation has opened a new way to clinically test the ability of diverse compounds as tumor suppressor gene transcriptional reactivators. We previously reported that hydralazine induces in vitro and in vivo, demethylation and transcriptional reactivation at the mRNA and protein levels of the ER, RARβ and p16 genes [26]. The present study confirms our previous findings in a larger number of patients receiving different hydralazine dosage levels for a 10 day period. Our results demonstrate that hydralazine can demethylate and re-express tumor suppressor genes in previously untreated cervical cancer patients in all the tested dosages. Theses effects are accompanied by no change in global tumor DNA methylation evaluated by two methods, and lack of demethylation in the imprinted gene H19 and the "normally methylated" 1.2 clone in peripheral mononuclear blood cells.
Hydralazine, a widely available peripheral vasodilator agent, has been extensively used for high blood pressure, heart failure and pregnancy-associated hypertensive disorders [32-34]. Hence, its evaluation as a demethylating agent in a clinical trial involving cancer patients could proceed with no major concerns regarding unexpected toxicity and long-term side effects aside of its known capacity to induce a lupus-like condition [35]. Evaluation of tumor DNA demethylating agents in patients bearing solid tumors is troublesome due to the need of repeated tumor sampling. Cervical cancer is easily accessible for repeated tumor sampling and thus we chose to study patients with newly diagnosed cervical cancer. In order to avoid treatment delay in the studied group, we scheduled the present study between the date of diagnosis and the beginning of chemoradiation. In addition, three recent publications have consistently shown a number of tumor suppressor genes to be methylated in primary cervical tumors [36-38]. Our results demonstrate that 70% of the pretreatment samples analyzed had at least one gene methylated, and all 16 patients had at least one methylated gene. The methylation frequency ranged from 3/16 (19%) to 16/16 (100%) for p16 and DAPK genes respectively. (Figure 1A). Despite the fact that our global gene methylation frequency is remarkably similar to the frequencies reported by other authors [36-38] who found 86%, 74% and 79% respectively, we encountered differences in individual gene methylation frequencies which most likely stem from the different patient populations analyzed such as invasive versus pre-invasive disease, tumor histologies, and clinical stages (Table 6). Nevertheless, our results and those obtained in other studies as shown in Table 6, suggest that cervical cancer is a good tumor model to evaluate the effect of DNA demethylating agents upon a small number of individual genes found consistently methylated in high proportion. As expected, there were genes "fully" unmethylated or "fully" methylated but we also observed cases showing both methylated and unmethylated bands. Such ambiguous methylation pattern as evaluated by methylation-specific PCR is commonly observed when patient tumor samples are studied which probably results from certain degree of normal tissue/cell biopsy contamination. This may rise concerns on whether the demethylating effect showed post-treatment is consequence of the treatment and not due tumor heterogeneity, however, in all pretreatment and post-treatment biopsies the major component of the tumor was malignant cells (Figure 3A); On the other hand, if demethylation were the result of chance alone we would expect to have cases in both directions, that is, methylated tumors that demethylated, and demethylated tumors that methylated. The latter never occurred, which strongly supports that the demethylating effect was caused by hydralazine. To gain further insight into the issue of tumor heterogeneity, we multi-sampled a tumor and submitted to methylation analysis each of the tumor fragments containing malignant cells. As shown in Figure 3B, there were no changes in the pattern of methylation in the four separate tumor areas of the tumor. This result strongly supports the observation that demethylation is an effect of hydralazine.
Hydralazine is considered a safe drug with hypertension and hearth failure dosing ranging from 50 mg day to 400 mg/day [33,34]. However, higher doses produce symptomatic side effects mainly derived from its cardiovascular effects. Based on this, we felt important to determine the lowest possible dose that induced DNA demethylation to be used in future clinical trials. In the present study we demonstrate that at a dose range between 50 mg and 150 mg a day for 10 days, hydralazine not only induces gene demethylation in roughly half the evaluated gene/tumors, but we also observed re-expression in two-thirds of the informative cases. However, due to the limited sample size of this study we could not establish whether the effectiveness of this demethylating agent is gene or patient-dependent.
At the present time there is limited information regarding cancer treatment efficacy, demethylating and gene re-expressing profiles of the various DNA demethylating agents in the clinical setting. There are three clinical trial reports evaluating 5-aza-2'-deoxycytidine. Two of these were performed in patients with myelodysplasic syndrome and relapsed leukemia patients. Daskalakis et al., using the quantitative assay Ms-SNuPE found basal hypermethylation of p15 gene in 15/23 patients (65%) which decreased in nine of 12 patients sequentially analyzed after at least a course of low-dose decitabine. As expected, reactivation of the p15 protein expression was found in bone marrow biopsies of four out of eight patients analyzed in this study. Interestingly, response (3 CR) was observed on all nine patients in whom p15 gene demethylation occurred [39]. More recently, a phase I study with escalating doses of decitabine (5, 10, 15 or 20 mg/m2 IV over 1 hour daily, 5 days a week for two consecutive weeks) found responses in 11/18 patients (61%), however, no significant decline in p15 methylation after treatment with decitabine regardless of the response was observed. Moreover, in the three patients that showed >50% demethylation, evaluated by the COBRA assay there was no response. Authors speculated on the poor sensitivity of this technique to explain their results [40]. Evaluation of these nucleoside analogs in solid tumors has also shown inconsistent results. Aparicio et al., reported a phase I study using 5-aza-2'-deoxycytidine in patients with advanced solid tumors using escalating doses of 20, 30, 40 mg/m2 using a 72-hour continuous infusion every 28 days. The quantitative Methyl-Light reaction was used to evaluate changes in promoter methylation in 19 genes but no consistent evidence of gene demethylation was documented despite grade 4 neutropenia was found in almost a third of the patients. This latter finding argues against its use as DNA demethylating agent in solid tumors because despite such toxicity the steady-state levels reached during the study (0.1–0.2 μM) are below the levels needed in vitro to demethylate gene promoters [41]. On the other hand, a number of genes showed increased methylation which could be derived from the cytotoxicity of this nucleoside analog. It is well-known that most of the cytotoxic agents lead to an increase in DNA methylation in vitro and in cancer patients [42].
Among the non-nucleoside demethylating agents, MG98, a DNMT1 antisense oligonucleotide, has been tested in a phase I study to treat solid tumors. In contrast to the studies discussed above no attempt was done to evaluate gene methylation in tumors although DNMT1 mRNA levels were investigated in peripheral blood cells; no consistent decrease was observed [25].
Our results demonstrate that hydralazine is quite promising in regards to its gene demethylating and tumor suppressor gene reactivating activities without significant side effects. It is noteworthy that a dose-effect level in the frequency of demethylation was not observed. A pharmacokinetic correlation analysis would have been useful for the interpretation of these findings, nonetheless, the pharmacokinetics of this drug is relatively well-known and some assumptions can be made. After a single 1 mg/kg oral dose and after the fifth 1 mg/kg dose given every 12 hours, the peak concentrations achieved are in the range of 0.12–1.31 μM and 0.10–1.39 μM respectively and AUC values range between: 4.0–30.4 and 3.2–38.5 μM-minute, respectively [43]. These concentrations were likely achieved in our patients at the doses used which are close to the concentrations needed to achieve gene demethylation and re-expression in vitro [26]. It is not clear why we observed such variable gene demethylation rates irrespective of the doses used, this may be the result of the small number of patients studied as well as the additional variability imposed by the acetylator phenotype which may vary from 30 to 70% according to some studies performed in Latin American Hispanic populations [44].
After years characterizing the role of DNA methylation in cancer, it has become clear that both hypermethylation and hypomethylation are by their own cancer causative or at least cancer promoting, in animal models. This has led to consider the use of demethylating agents for cancer treatment a two-edge sword [45]. Such consideration is debatable as the profound DNA demethylation levels achieved in animal models can be hardly pharmacologically reproduced in patients. The present study demonstrates that hydralazine at the doses used not only does not demethylate genes which are "normally methylated" such as H19 and clone 2.1 but also fails to produce global DNA demethylation as evaluated by capillary electrophoresis which is a powerful and accurate method to quantitate methylated and unmethylated cytosines [46], and by cytosine extension assay which is also a sensitive method to underscore abnormal methylation patterns [31]. We can only speculate on the nature of these findings. At the individual gene level, a plausible explanation could be that certain "normally methylated" genes could lie within a milieu prone to rapid re-methylation and/or that certain genes have higher "availability" of the elements that compose the methylation machinery, therefore the demethylating agent would fail to produce demethylation of these sequences at least at a detectable levels by the current methods of analysis. This view is supported by the reporting that in mice 5-azacytidine does not change the methylation status of H19 [47]. In regard to global DNA methylation changes most likely is just a matter of dose as we have in vitro evidence that hydralazine produces global DNA hypomethylation in MCF-7 cells as evaluated by the same methods here used. Alternatively, changes in global mC content induced by hydralazine are not sufficient to be detected by the methods we used.
Conclusion
In conclusion, hydralazine used at standard doses for the treatment of cardiovascular conditions is an effective demethylating and tumor suppressor gene transcriptional reactivator causing no decrease in DNA mC content for solid tumors. Our results are in agreement with those reported by Chan et al., who found substantial degrees of demethylation at all latent and lytic Epstein-Barr virus promoters examined by methylation-specific PCR in nasopharyngeal cancer patients treated with 5-azacytidine [48]. A phase II clinical study using hydralazine in combination with standard cytotoxic chemotherapy is being planned as proof of concept that the reactivation of tumor suppressor genes silenced by DNA methylation increases chemotherapy efficacy in solid tumors.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
P Z, B S-P, E P-C, C T-B, L TC, and A C-B performed the methylation analysis and gene expresión experiments; LC cared for the patients; AR-V and KS performed the capillary electrophoresis analysis; EA and GC critically read and contributed to the discussion; J C-V, performed the pathological analysis; and A D-G conceived the study and wrote the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This work was supported by CONACyT grants: SALUD-2002-C01-6579, AVANCE C01-294, and by Psicofarma S.A. de C.V. Mexico.
Figures and Tables
Figure 1 1A. Pre- (dark bars) and post-hydralazine treatment (light bars). The bars represent the number of patients that showed methylation for each studied gene from each of the 16 patients. 1B. Representative cases of genes (M methylated, U unmethylated; pre/post): M/M; M/U; U/U, M/U; MU/U; M/ U-M. 1C. Percentage of demethylation after treatment according to the dose. Percentage was calculated considering 100% methylation the total number of pre-treatment methylated genes in each cohort of 4 patients
Figure 2 Representative cases correlating methylation and re-expression before and after hydralazine treatment. 2A is a patient treated with 75 mg/day that demethylated and re-expressed the DAPK gene. 2B corresponds to a patient receiving 150 mg/day who showed only the methylated band pre-treatment, but both bands after treatment, which correlated with re-expression of MGMT. 2C is a 50 mg/day patient which failed to demethylate the DAPK gene and therefore lacked expression. 2D represents the distribution of informative cases. From the 128 genes/cases, 116 were RT-PCR positive regardless of the methylation status, hence were not informative. In the remaining 12 cases, nine demethylated and re-expressed the gene.
Figure 3 3A. Photomicrography of a representative set of pre and post-treatment tumor biopsies showing that the malignant component represents almost the half of the tumor. 3B. Methylation analysis of the DAPK gene in the four fragments of the tumor biopsy of an untreated cervical cancer patient. Despite all fragments contained different proportions of malignant cells and stroma the four samples show methylated and unmethylated bands.
Figure 4 DNA methylation analysis obtained from peripheral mononuclear cells of genes "imprinted or normally methylated". Clone 1.2 remained methylated in all cases whereas for the imprinted H19 gene, the pattern of U and M alleles did not change.
Figure 5 5A. Capillary electrophoretic analysis of global methylation. Relative methylation showed no variation in percentage of mC after treatment (37.3% versus 36.3%). 5B is a electropherogram showing the separation of C and mC. 5C is the percent increase in radiolabeled incorporation pre and post-treatment as compared to the control of undigested DNA.
Table 1 Primers and conditions used for MSP.
Primer set Sense 5'-- 3' Antisense 5'--- 3' Size Annealing
p16 M TTATTAGAGGGTGGGGCGGATCGC GACCCCGAACCGCGACCGTAA 150 62°C
p16 U TTATTAGAGGGTGGGGTGGATTGT CAACCCCAAACCACAACCATAA 151 62°C
RARβM TCGAGAACGCGAGCGATTCG GACCAATCCAACCGAAACGA 146 55.5°C
RARβU TTGAGAATGTGAGTGATTTGA AACCAATCCAACCAAAACAA 146 55.5°C
MGMT M TTTCGACGTTCGTAGGTTTTCGC AACCAATCCAACCAAAACAA 81 55.7°C
MGMT U TTTGTGTTTTGATGTTTGTAGGTTT AACTCCACACTCTTCCAAAAAC 93 55.7°C
FHIT M TTGGGGCGCGGGTTTGGGTTTTTA CGTAAACGACGCCGACCCCACTA 74 55.5°C
FHIT U TTGGGGTGTGGGTTTGGGTTTTTA CATAAACAACACCAACCCCACTA 74 55.5°C
DAPK M GGATAGTCGGATCGAGTTAACGT CCCTCCCAAACGCCGA 98 55.7°C
DAPK U GGAGGATAGTTGGATTGAGTTAAT CAAATCCCTCCCAAACACCAA 106 55.7°C
APC M TATTGCGGAGTGCGGGTC TCGACGAACTCCCGACGA 100 56°C
APC U GTGTTTTATTGTGGAGTGTGGGTT CCAATCAACAAACTCCCAACAA 110 56°C
GSTp1 M TTCGGGGTGTAGCGGTGGTC GCCCCAATACTAAATCACGACG 91 56°C
GSTp1 U GATGTTTGGGGTGTAGTGGTTGTT CCACCCCAATACTAAATCACAACA 97 56°C
H19 M TTATAAAATCGAAAATTACGCGCGA AGATGATTTTCGTGAATTTTGCG 136 55°C
H19 U TATAATTATAAAATCAAAAATTACA TTTTAGATGATTTTTGTGAATTTT 145 55°C
ER M CGGTTGGAGTTTTTGAATCGTTC CTAGCGTTAACGACGACCG 151 55°C
ER U ATGAGTTGGAGTTTTTGAATTGTTT ATAAACCTACACATTAACAACAACCA 158 55°C
Clone 1.2 M ATGAGTTGGAGTTTTTGAATTGTTT AATAATAAACGTAACGCCCGCGAAC 258 63°C
Clone 1.2 U GGTTGTTTGGTTTTTATTGGGATGTTTTT CCTAAATAATAAACATAACACCCACAAAC 258 63°C
Table 2 Primers and conditions used for RT-PCR
Primer set Sense 5'-- 3' Antisense 5'--- 3' Size Annealing
p16 AGCCTTCGGCTGACTGGCTGG CTGCCCATCATCATGACCTGG 150 60°C
RARβ GACTGTATGGATGTTCTGTCAG ATTTGTCCTGGCAGACGAAGCA 146 50°C
MGMT GCTGCAGACCAVCTCTGTGGCACG GCCGCCTCTTCACCATCCCG 81 50°C
FHIT ATGTCGTTCAGATTTGGCCAAC TCATAGATGCTGTCATTCCTGT 340 53°C
DAPK AACCCATCATCCATGCCATC TCTCTCCTTCTCGGTTCTTGA 200 51°C
APC GAGACAGAATGGAGGTGCTGC GTAAGATGATTGGAATTATCTTCT 180 56°C
GSTp1 TCCGCTGCAAATACATCTCC TGTTTCCCGTTGCCATTGAT 320 50°C
ER GGAGACATGAGAGCTGCCAAC CCAGCAGCATGTCGAAGATC 480 55°C
Table 3 Clinical characteristics of patients
Number 16
Mean Age 51.8 (35–75 years)
Histology
Squamous 16 (100%)
FIGO Stage*
IIB 8 (50%)
IIIB 8 (50%)
Performance Status**
0 10 (62%)
1 6 (38%)
*International Federation of Gynecology and Obstetrics,
** World Health Organization Criteria
Table 4 Toxicity to hydralazine expressed by number of patients suffering the event (16 patients)
Toxicity 4 Patients in each dose level (mg
50 75 100 150
Nausea 2/4 - 2/4* 2/2
Vomiting - - - -
Dizziness - 1/4* 1/4 -
Fatigue 2/4 2/4 3/4 1/4
Flushes - - - -
Headache 2/4 1/4 3/4 2/4
Edema - - - -
Palpitations 1/4 2/4 2/4 3/4
* Grade 2 toxicity. All other were grade 1.
Table 5 Methylation and gene re-expression of the 12 informative cases.
Pre-treat Post-treat Re-expression
Gene dose level M U M U Yes/No
DAPK 50 + - + - No
DAPK 50 + - + - No
GSTp1 50 + - + - No
FHIT 100 + - + + Yes
FHIT 150 + - + + Yes
MGMT 75 + - + + Yes
MGMT 150 + - + + Yes
GSTp1 50 + - + + Yes
DAPK 75 + - - + Yes
ER 75 + - - + Yes
GSTp1 150 + - - + Yes
GSTp1 75 + + - + Yes
Table 6 Frecuency of promoter methylation of individual genes (%).
APC ER FHIT GSTp1 MGMT P16 RARB DAPK ANY
This study 94 25 88 88 81 19 62 100 70
Narayan (ref 35) 11 - 11 - 5 8.5 29 45 86
Virmani (ref 34) - - 32 21 26 42 26 - 74
Dong (ref 36) 32 - - - 8 30 - 51 79
==== Refs
Robertson KD DNA methylation, methyltransferases and cancer Oncogene 2001 20 3139 55 11420731 10.1038/sj.onc.1204341
Pradhan S Bacolla A Wells RD Roberts RJ Recombinant human DNA (cytosine-5) methyltransferase.I. Expression, purification, and comparison of de novo and maintenance methylation J Biol Chem 1999 274 33002 10 10551868 10.1074/jbc.274.46.33002
Beard C Li E Jaenisch R methylation activates Xist in somatic but not in embryonic cells Genes Dev 1995 9 2325 34 7557385
Li E Beard C Jaenisch R Role for DNA methylation in genomic imprinting Nature 1993 366 362 5 8247133 10.1038/366362a0
Leonhardt H Page AW Weier HU Bestor TH A targeting sequence directs DNA methyltransferase to sites of DNA replication in mammalian nuclei Cell 1992 71 865 73 1423634 10.1016/0092-8674(92)90561-P
Chuang LS Ian HI Koh TW Ng HH Xu G Li BF Human DNA-(cytosine-5) methyltransferase-PCNA complex as a target for P21WAF1 Science 1997 277 1996 2000 9302295 10.1126/science.277.5334.1996
Okano M Bell DW Haber DA Li E DNA methyltransferases Dnmt3a and Dnmt3b are essential for de novo methylation and mammalian development Cell 1999 99 247 57 10555141 10.1016/S0092-8674(00)81656-6
Okano M Xie S Li E Dnmt2 is not required for de novo and maintenance methylation of viral DNA in embryonic stem cells Nucleic Acids Res 1998 26 2536 40 9592134 10.1093/nar/26.11.2536
Tate PH Bird AP Effects of DNA methylation on DNA-binding proteins and gene expression Curr Opin Genet Dev 1993 3 226 31 8504247 10.1016/0959-437X(93)90027-M
Nan X Campoy FJ Bird A MeCP2 is a transcriptional repressor with abundant binding sites in genomic chromatin Cell 1997 88 471 81 9038338 10.1016/S0092-8674(00)81887-5
Rountree MR Bachman KE Baylin SB DNMT1 binds HDAC2 and a new co-repressor, DMAP1, to form a complex at replication foci Nature Genet 2000 25 269 77 10888872 10.1038/77023
Nan X Ng HH Johnson CA Turner BM Eisenman RN Bird A Transcriptional repression by the methyl-CpG-binding protein MeCP2 involves a histone deacetylase complex Nature 1998 393 386 9 9620804 10.1038/30764
Bird A The essentials of DNA methylation Cell 1992 70 5 8 1377983 10.1016/0092-8674(92)90526-I
Clark SJ Melki J DNA methylation and gene silencing in cancer: which is the guilty party? Oncogene 2002 21 5380 7 12154400 10.1038/sj.onc.1205598
Antequera F Bird A Number of CpG islands and genes in human and mouse Proc Natl Acad Sci USA 1993 90 11995 9 7505451
Baylin SB Herman JG DNA hyper methylation in tumorigenesis: epigenetics joins genetics Trends Genet 2000 16 168 74 10729832 10.1016/S0168-9525(99)01971-X
Santini V Kantarjian HM Issa JP Changes in DNA methylation in neoplasia: pathophysiology and therapeutic implications Ann Intern Med 2001 134 573 86 11281740
Bigey P Knox JD Croteau S Bhattacharya SK Theberge J Szyf M Modified oligonucleotides as bona fide antagonists of proteins interacting with DNA. Hairpin antagonists of the human DNA methyltransferase J Biol Chem 1999 274 4594 606 9988694 10.1074/jbc.274.8.4594
Lantry LE Zhang Z Crist KA Wang Y Kelloff GJ Lubet RA You M 5-Aza-2'-deoxycytidine is chemopreventive in a 4-(methyl-nitrosamino)-1-(3-pyridyl)-1-butanone-induced primary mouse lung tumor model Carcinogenesis 1999 20 343 6 10069475 10.1093/carcin/20.2.343
Covey JM Zaharko DS Comparison of the in vitro cytotoxicity (L1210) of 5-aza-2'-deoxycytidine with its therapeutic and toxic effects in mice Eur J Cancer Clin Oncol 1985 21 109 117 2578963 10.1016/0277-5379(85)90207-X
Wu J Issa JP Herman J Bassett DE Nelking BD Baylin SB Expression of an exogenous eukaryotic DNA methyltransferase gene induces transformation of NIH3T3 cells Proc Natl Acad Sci USA 1993 90 8891 5 8415627
Ramchandani S MacLeod AR Pinard M Von Hofe E Szyf M Inhibition of tumorigenesis by a cytosine-DNA methyltransferase, antisense oligodeoxynucleotide Proc Natl Acad Sci USA 1997 94 684 9 9012845 10.1073/pnas.94.2.684
Goffin J Eisenhauer E DNA methyltransferase inhibitors-state of the art Ann Oncol 2002 13 1699 716 12419742 10.1093/annonc/mdf314
Lyons J Bayar E Fine G McCullar M Rolens R Rubinfeld J Rosenfeld C Decitabine: development of a DNA methyltransferase inhibitor for hematological malignancies Curr Opin Investig Drugs 2003 4 1442 50 14763130
Stewart DJ Donehower RC Eisenhauer EA Wainman N Shah AK Bonfils C MacLeod AR Besterman JM Reid GK A phase I pharmacokinetic and pharmacodynamic study of the DNA methyltransferase 1 inhibitor MG98 administered twice weekly Ann Oncol 2003 14 766 74 12702532 10.1093/annonc/mdg216
Segura-Pacheco B Trejo-Becerril C Perez-Cardenas E Taja-Chayeb L Mariscal I Chavez A Acuna C Salazar AM Lizano M Duenas-Gonzalez A Reactivation of tumor suppressor genes by the cardiovascular drugs hydralazine and procainamide and their potential use in cancer therapy Clin Cancer Res 2003 9 1596 603 12738711
Angeles E Vazquez-Valadez VH Vazquez-Valadez O Velazquez-Sanchez AM Ramirez A Martinez L Diaz-Barriga S Romero-Rojas A Cabrera G Lopez-Castañares R Duenas-Gonzalez A Computational studies of 1-Hydrazinophtalazine (Hydralazine) as antineoplastic agent. Docking studies on methyltransferase Lett Drug Design Discov 2005
Herman JG Graff JR Myohanen S Nelkin BD Baylin SB Methylation-specific PCR: a novel PCR assay for methylation status of CpG islands Proc Natl Acad Sci USA 1996 93 9821 6 8790415 10.1073/pnas.93.18.9821
Strichman-Almashanu LZ Lee RS Onyango PO Perlman E Flam F Frieman MB Feinberg AP A genome-wide screen for normally methylated human CpG islands that can identify novel imprinted genes Genome Res 2002 12 543 54 Article published online before print in March 2002 11932239 10.1101/gr.224102
Sandoval Guerrero K Revilla Vazquez A Segura-Pacheco B Duenas-Gonzalez A Determination of 5-methyl-cytosine and cytosine in tumor DNA samples of cancer patients Electrophoresis 2005 26 1057 62 15765481 10.1002/elps.200410040
Pogribny I Yi P James SJ A sensitive new method for rapid detection of abnormal methylation patterns in global DNA and within CpG islands Biochem Biophys Res Commun 1999 262 624 8 10471374 10.1006/bbrc.1999.1187
Kirsten R Nelson K Kirsten D Heintz B Clinical pharmacokinetics of vasodilators. Part II Clin Pharmacokinet 1998 35 9 36 9673832
Magee LA Cham C Waterman EJ Ohlsson A von Dadelszen P Hydralazine for treatment of severe hypertension in pregnancy: meta-analysis BMJ 2003 32 7955 60
Klein L O'Connor CM Gattis WA Gattis WA Zampino M de Luca L Vitarelli A Fedele F Gheorghiade M Pharmacologic therapy for patients with chronic heart failure and reduced systolic function: review of trials and practical considerations Am J Cardiol 2003 9 18F 40F 10.1016/S0002-9149(02)03336-2
Yung R Chang S Hemati N Johnson K Richardson B Mechanisms of drug-induced lupus. IV. Comparison of procainamide and hydralazine with analogs in vitro and in vivo Arthritis Rheum 1997 40 1436 43 9259423
Virmani AK Muller C Rathi A Zoechbauer-Mueller S Mathis M Gazdar AF Aberrant methylation during cervical carcinogenesis Clin Cancer Res 2001 7 584 9 11297252
Narayan G Arias-Pulido H Koul S Vargas H Zhang FF Villella J Schneider A Terry MB Mansukhani M Murty VV Frequent promoter methylation of CDH1, DAPK, RARB, and HIC1 genes in carcinoma of cervix uteri: Its relationship to clinical outcome Mol Cancer 2003 2 24 30 12773202 10.1186/1476-4598-2-24
Dong SM Kim HS Rha SH Sidransky D Promoter hypermethylation of multiple genes in carcinoma of the uterine cervix Clin Cancer Res 2001 7 1982 6 11448914
Daskalakis M Nguyen TT Nguyen C Guldberg P Kohler G Wijermans P Jones PA Lubbert M Demethylation of a hypermethylated P15/INK4B gene in patients with myelodysplastic syndrome by 5-Aza-2'-deoxycytidine (decitabine) treatment Blood 2002 100 2957 64 12351408 10.1182/blood.V100.8.2957
Issa JP Garcia-Manero G Giles FJ Mannari R Thomas D Faderl S Bayar E Lyons J Rosenfeld CS Cortes J Kantarjian HM Phase 1 study of low-dose prolonged exposure schedules of the hypomethylating agent 5-aza-2'-deoxycytidine (decitabine) in hematopoietic malignancies Blood 2004 103 1635 40 14604977 10.1182/blood-2003-03-0687
Aparicio A Eads CA Leong LA Laird PW Newman EM Synold TW Baker SD Zhao M Weber JS Phase I trial of continuous infusion 5-aza-2'-deoxycytidine Cancer Chemother Pharmacol 2003 51 231 9 12655442
Nyce JW Drug-induced DNA hypermethylation: a potential mediator of acquired drug resistance during cancer chemotherapy Mutat Res 1997 386 153 61 9113116
Ludden TM McNay JL JrShepherd AM Lin MS Variability of plasma hydralazine concentrations in male hypertensive patients Arthritis Rheum 1981 24 987 93 7284051
Dipierri JE Alfaro E Bejarano IF Etchart AA Acetylator phenotypes: allele frequency in northwestern Argentina and review of acetylator distribution in the Americas Hum Biol 1998 70 959 64 9780522
Feinberg AP Tycko B The history of cancer epigenetics Nat Rev Cancer 2004 4 143 53 14732866 10.1038/nrc1279
Stach D Schmitz OJ Stilgenbauer S Benner A Dohner H Wiessler M Lyko F Capillary electrophoretic analysis of genomic DNA methylation level Nucleic Acids Res 2003 31 E2 12527791 10.1093/nar/gng002
Hu JF Nguyen PH Pham NV Vu TH Hoffman AR Modulation of Igf2 genomic imprinting in mice induced by 5-azacytidine, an inhibitor of DNA methylation Mol Endocrinol 1997 11 1891 8 9415394 10.1210/me.11.13.1891
Chan AT Tao Q Robertson KD Flinn IW Mann RB Klencke B Kwan WH Leung TW Johnson PJ Ambinder RF Azacitidine induces demethylation of the Epstein-Barr virus genome in tumors J Clin Oncol 2004 22 1373 81 15007085 10.1200/JCO.2004.04.185
| 15862127 | PMC1131894 | CC BY | 2021-01-04 16:03:05 | no | BMC Cancer. 2005 Apr 29; 5:44 | utf-8 | BMC Cancer | 2,005 | 10.1186/1471-2407-5-44 | oa_comm |
==== Front
BMC Cell BiolBMC Cell Biology1471-2121BioMed Central London 1471-2121-6-221585422510.1186/1471-2121-6-22Research ArticleThe use of time-resolved fluorescence imaging in the study of protein kinase C localisation in cells Stubbs Christopher D [email protected] Stanley W [email protected] Simon J [email protected] Anthony W [email protected] Department of Pathology and Cell Biology, Thomas Jefferson University, Philadelphia PA 19107 USA2 The Central Laser Facility, Rutherford Appleton Laboratory, CCLRC, Chilton, OX11 OQX UK2005 26 4 2005 6 22 22 24 12 2004 26 4 2005 Copyright © 2005 Stubbs et al; licensee BioMed Central Ltd.2005Stubbs 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
Two-photon-excitation fluorescence lifetime imaging (2P-FLIM) was used to investigate the association of protein kinase C alpha (PKCα) with caveolin in CHO cells. PKCα is found widely in the cytoplasm and nucleus in most cells. Upon activation, as a result of increased intracellular Ca2+ and production of DAG, through G-protein coupled-phospholipase C signalling, PKC translocates to a variety of regions in the cell where it phosphorylates and interacts with many signalling pathways. Due to its wide distribution, discerning a particular interaction from others within the cell is extremely difficult
Results
Fluorescence energy transfer (FRET), between GFP-PKCα and DsRed-caveolin, was used to investigate the interaction between caveolin and PKC, an aspect of signalling that is poorly understood. Using 2P-FLIM measurements, the lifetime of GFP was found to decrease (quench) in certain regions of the cell from ~2.2 ns to ~1.5 ns when the GFP and DsRed were sufficiently close for FRET to occur. This only occurred when intracellular Ca2+ increased or in the presence of phorbol ester, and was an indication of PKC and caveolin co-localisation under these conditions. In the case of phorbol ester stimulated PKC translocation, as commonly used to model PKC activation, three PKC areas could be delineated. These included PKCα that was not associated with caveolin in the nucleus and cytoplasm, PKCα associated with caveolin in the cytoplasm/perinuclear regions and probably in endosomes, and PKC in the peripheral regions of the cell, possibly indirectly interacting with caveolin.
Conclusion
Based on the extent of lifetime quenching observed, the results are consistent with a direct interaction between PKCα and caveolin in the endosomes, and possibly an indirect interaction in the peripheral regions of the cell. The results show that 2P-FLIM-FRET imaging offers an approach that can provide information not only confirming the occurrence of specific protein-protein interactions but where they occur within the cell.
==== Body
Background
In common with other signalling proteins, PKC has been suggested to associate with caveolin-1 [1,2], a key component of caveoli and membrane rafts, that are proposed to exist as signalling platforms in the plasma membrane and elsewhere in the cell. However, where in the cell the PKC-caveolin interaction might occur and under what conditions remains unclear.
Caveolae are vesicular organelles are that are involved in a wide range of cellular functions, serving as platforms or rafts, wherein reside a wide variety of signalling molecules [3]. The caveolin proteins (caveolin-1, -2, and -3) act as the structural components of caveolae. They also function as scaffolding proteins and as such recruit numerous signalling molecules to caveolae where their activity is regulated. PKC is a signalling molecule of major importance in cells, which in the form of twelve isoforms, regulates numerous signalling cascades by virtue of its ability to phosphorylate target proteins that include receptors, G-proteins, ion channels as well as other kinases [4-6]. This leads to control of numerous cellular processes, such as secretion, proliferation, differentiation, apoptosis, permeability, migration, hypertrophy etc [4,5,7-11].
While it has been shown that isolated caveoli interact with purified PKCα [12], PKCγ [13] and PKCε [14] using immunoprecipitation, where in the cell this occurs is not known. Caveolin contains a sequence that is a consensus site for phosphorylation by PKC [15], while down-regulation of plasma membrane-translocated PKCα involvesinternalization of the active enzyme that involves ubiquitination, through a caveolae-dependent mechanism, followed by multisite dephosphorylation and down-regulation in a perinuclear compartment in a time dependent manner (~30 min after stimulated translocation to the plasma membrane) [16]. It has also been shown that endocytic trafficking via caveolae may be a PKCα-dependent process [17]. These observations lead to the question of whether PKCα interacts directly with caveolin, and where in the cell this occurs, a question we examined in this present study.
The classic biochemical or immunoprecipitation approaches for determining the location of signalling molecules in cells, while commonly used, is severely limited for several reasons. The main drawback is that it involves destruction of the cell, resulting in loss of spatial information. Staining the cells with fluorescent antibodies provides a useful advance enabling apparent "co-localisation" to be obtained when two different fluorophores are used. However, the two fluorophores may still be a considerable distance apart without any protein-protein interaction between the pair occurring. The method also suffers from problems of photobleaching and when the probes are in abundance in the cell false co-localisation data can result.
Recently, with the advent of green fluorescent protein (GFP) technology, considerable new information has become available using imaging approaches. This allows the protein of interest to be tagged by expression in the cell, allowing it to be functionally located within the cell without recourse to antibodies. Numerous papers have shown signalling molecules in cells can be tracked as they move between subcellular locations using expression of GFP-tagged proteins and fluorescence microscopy. PKC has been studied in a wide range of situations using GFP-tags (e.g. [18-20]. Lifetime imaging has been used as a localisation tool for GFP-tagged proteins [21-23] and using this approach both PKCα activation levels, along with localisation, has been detected through the binding of fluorescently tagged phosphorylation site-specific antibodies using fluorescence energy transfer (FRET), measured through a donor fluorophore on the PKC [24]. In most cases the localisation was either followed in real-time or after fixation, following initiation of intracellular signalling. There are now a number of different fluorescent proteins available in the "GFP" family, for example cyan fluorescent protein (CFP), yellow fluorescent protein (YFP), red fluorescent protein (DsRed) etc.
A considerable advance over the co-localisation approach is to use steady state FRET between suitably tagged proteins. A further improvement over steady state FRET is achieved by monitoring the lifetime of the donor, which is independent of changes in concentration, photobleaching and various limitations over intensity-based detection. Donor lifetime quenching is evidence for a direct physical interaction and in addition does not require corrections due to spectral over-lap that are required in steady state FRET. The use of this approach FRET-FLIM for localisation using different GFP-type constructs has been described recently in the literature [25-27]. This approach in combination with 2-photon (2P) excitation provides a powerful imaging technique. It does not require complex corrections that intensity-based FRET entails, allows excitation within a narrow focussed plane without contributions outside this region, as would be the case in one-photon excitation, as well as other advantages, such as the ability to image deeper in tissue etc.
In this work the interaction of PKCα with caveolin, was investigated using two-photon-fluorescence lifetime imaging (2P-FLIM). This was determined by investigating the FRET between GFP-tagged PKCα and DsRed-caveolin (DsRed-cav). Using the quenched lifetime of the GFP-tag, areas showing co-localisation in CHO cells was identified after activation of the PKC had occurred. When the GFP-PKC is induced to translocate to membranes, excitation of the GFP at 850 nm, for 2P excitation, should lead to FRET to the DsRed construct if the two are co-localised. This would be seen as a reduced GFP lifetime, in contrast to areas in which the GFP-PKC resides but not with caveolin. This information was determined using a FLIM time-correlated single photon counting set up (TCSPC) with the frequency doubled output of a Tsunami pulsed laser (110 fs), coupled to an inverted microscope. Based on the lifetime quenching data the results were consistent with three populations of activated GFP-PKC. One within the cytoplasmic area of the cells, which was probably undergoing a direct interaction with caveolin, likely in endosomes, a second population in the cytoplasm and in the nucleus that was not interacting with caveolin and a third at the plasma membrane possibly indirectly interacting with caveolin.
Results
The epifluorescence images in Figure 1 show GFP-PKC and DsRed-cav localisation as green and red colours respectively. The images were taken before and after exposure to the phorbol ester TPA for 3 min. It can be seen that PKC accumulated at the cell periphery and in organelles in the perinuclear region.
Figure 1 Fluorescence intensity images of GFP-PKC and DsRed-cav expressed in CHO cells. GFP-PKC and/or DsRed-cav were transiently expressed in CHO cells on glass coverslips by culturing for 48 hr. The cells were then treated with Ca2+-ionophore (1 nM) or TPA (100 nM) for 3 min before the cells were fixed on microscope slides as described under Methods. Images were acquired using a Olympus IX-70 inverted epifluorescence microscope with a 60× objective and GFP/DsRed filter sets and a Olympus C3030 camera. (a) a representative image before and (b) after Ca2+-induced translocation by ionophore of GFP-PKC to the peripheral membrane and to discrete regions in the perinuclear region. (c) image of DsRed-cav and (d) GFP-PKC, after treatment with 100 nM TPA.
A lifetime image of a CHO cell transiently transfected with GFP-PKCα for 48 hr is shown in Figure 2, with the corresponding epifluorescence image. It can be seen that PKCα is widely and abundantly distributed in the cell cytoplasm and nucleus. The data collection time for this and other lifetime images was optimally 2–3 minutes. The lifetime data for the various conditions examined is also summarised in Table 1.
Figure 2 Fluorescence lifetime imaging of GFP-PKC expressed in CHO cells. GFP-PKC was transiently expressed in CHO cells as described in the legend in Figure 1 and detailed under Methods. Images were acquired using a Nikon 2000 inverted epifluorescence microscope with a 60× objective and GFP/DsRed filter sets and an ICAM camera. The 2P-FLIM images were collected using the TCSPC fast scanning imaging mode as described under Methods. (a) Conventional epifluorescence image of GFP-PKC distribution in a resting CHO cell, (b) lifetime image of the same cell with the analysis area enclosed by the red line (cytosol) shown (with colour coding) in the inset and giving an average lifetime of ~2.2 ns and (c) with the analysis area enclosed by the red line (nucleus) shown in the inset giving an average lifetime of ~2.0 ns. Cells shown are representative images from replicate experiments.
Table 1 Summary of average lifetimes for GFP emission under various conditions
Transfection Fig Treatment Area analysed Avg Lifetime (ns)
GFP-PKC 2 None cytoplasm 2.2
None nucleus 2.2
3 TPA cytoplasm 2.2
TPA nucleus 2.3
GFP-PKC/DsRed-cav 4 None cytoplasm 2.4
None cytoplasm 2.5
5 Ca2+ cytoplasm 1.6*
Ca2+ nucleus 2.0
6 TPA cytoplasm excluding peripheral 1.5*
TPA nucleus 2.0
TPA peripheral 1.8*
7 Bradykinin cytoplasm 1.6*
*reduced lifetime indicating PKC-GFP-DsRed-cav interaction
The lifetime of the GFP averaged for the cell area was around 2.2 ns (Figure 2b). The software analysis program optionally allows for whole or part of the field to be subjected to a lifetime analysis. By this means contributions from the background can be effectively excluded, as well as different areas with the cell compared. It is also possible to analyse the lifetimes on a pixel by pixel basis in discrete areas. Both approaches were used in this study. If the analysis was not restricted to the cell area and the whole field is included the background contributions became significant and a background component with a lifetime of ~1.4 ns became apparent. A single pixel analysis, within the cell area, yielded 2.0 ns for a single exponential with a reduced χ2 of 1.22, a two exponential analysis did not give an improved fit. Based on this, and observations from the literature, we analysed the lifetime of GFP as a single exponential. Analysis of the GFP-PKC located within the nucleus showed the same lifetime as that in the cytoplasm (Figure 2c).
We next examined the effect of the phorbol ester TPA on the GFP-PKC lifetime. If a FRET analysis is to be usefully undertaken it is first essential that while the PKC may redistribute in the cell, as a result of the TPA treatment, it should not result any changes to the lifetime. This was in fact the case as shown in Figure 3 where the lifetime for GFP-PKC across the cell area remained unchanged in the ~2.2 ns region.
Figure 3 Fluorescence lifetime imaging of GFP-PKC expressed in CHO cells: effect of TPA. 2P-FLIM images were collected as described in the legend to Figure 2 except cells were treated with TPA (100 nM) for 3 min. Treatment with the phorbol ester did not affect the fluorescence lifetime of GFP attached to PKC. (a) Fluorescence lifetime image with the analysis area enclosed by the red line (nucleus) shown in inset giving an average lifetime of ~2.1 to 2.2 ns. (b) Fluorescence lifetime image with the analysis area enclosed by the red line (cytosol) shown in inset giving an average lifetime of ~2.2 ns. Insets: lifetime distributions with colour coding.
The images in Figure 4 are from cells with GFP-PKC co-expressed with DsRed-cav. The epifluorescence images show that both PKC and caveolin were widely distributed in most cell areas, except that caveolin was not found in the nucleus and was concentrated in the perinuclear region. The lifetime images of GFP-PKC were acquired under conditions that would produce no contributions from DsRed fluorescence. This was achieved since the 2-P excitation was nominally at 425 nm, where there is virtually no excitation of the DsRed, also a 500 nm band-pass filter was used to exclude DsRed emission. Therefore only the GFP emission was collected. If the DsRed were to be close enough to PKC then a shortened lifetime would result due to FRET occurring. Again the average lifetime of the GFP-PKC was >2 ns, showing that under conditions when the cell is not stimulated (as defined here by Ca2+-mobilisation or phorbol ester activation of PKC), then PKC and caveolin do not co-associate in the cell.
Figure 4 Fluorescence lifetime imaging of GFP-PKC co-expressed with DsRed-cav in CHO cells. 2P-FLIM images were collected as described in the legend to Figure 2. Co-expression of the GFP-PKC with DsRed-cav does not affect the lifetime of the GFP showing that in the unstimulated state PKC is not associated with caveolin. Epifluorescence images for excitation of DsRed (a) and GFP along with DsRed (b) showing that the PKC and caveolin co-distributed in the cytosol. Fluorescence lifetime images with the analysis area enclosed by the red line, (cytosol) (c) or nucleus both essentially showing a lifetime as for Figs 2–3 centred around ~2.2 ns. Cells shown are representative images from replicate experiments.
When Ca2+ ionophore is added to cells Ca2+ gains entry and this induces immediate PKC translocation to various locations within the cells, including the perinuclear regions and peripheral membranes, as has been extensively shown in the literature (e.g. see ref [28,16,29]). The localisation of caveolin did not appear to change upon addition of Ca2+ ionophore to the cells (results not shown). For lifetime images it is important to note that what we are observing is the lifetime of the GFP, not its intensity. The lifetime images shown in Figure 5 show that the GFP-PKC fluorescence lifetime in the cytoplasm is reduced to ~1.6 ns. This is due to the quenching by DsRed, since the lifetime is unaffected in the absence of DsRed-cav. By contrast, the GFP-PKC lifetime in the nucleus was unaffected by the Ca2+ treatment, since there was little or no caveolin within the nucleus (Figure 5).
Figure 5 Lifetime imaging of GFP-PKC co-expressed with DsRed-cav in CHO cells: effect of Ca2+-ionophore. 2P-FLIM images were collected as described in the legend to Figure 2. Cells were treated with ionophore for 3 min before mounting and fixation as described in Methods. The epifluorescence image in the inset shows the DsRed-cav distribution (cytoplasmic) which was not affected by the Ca2+ ionophore. When the cytoplasmic area was analysed, (a) as shown by the area within the red line, both orange and green/blue areas are seen indicating the presence of both GFP-PKC and quenched GFP-PKC – note that only GFP lifetime can be observed in the lifetime images. This indicates that DsRed-cav was sufficiently close to the PKC-GFP to induce a quenching of the GFP by the DsRed, i.e. the PKC is translocating to caveolin containing areas. By contrast, in the nucleus (b) only GFP-PKC was expressed and the lifetime was unquenched (~2.2 ns). This is the same as the lifetime for GFP-PKC when only the latter is expressed (see Figure 2). Cells shown are representative images from replicate experiments.
The treatment of the cells with the phorbol ester TPA not only induces translocation of PKC to different cell compartments but also produces catalytically active PKC. In general, PKC moves to the outer membrane and to perinuclear regions, and associates with various signalling and cytostructural components in the cell. The lifetime image data revealed that, similar to the effect of increasing intracellular Ca2+, phorbol ester induced PKC to associate with caveolin, the GFP lifetime being reduced accordingly, as shown in Figure 6.
Figure 6 Lifetime imaging of GFP-PKC co-expressed with DsRed-cav in CHO cells: effect of phorbol ester. Epifluorescence and 2P-FLIM images were collected as described in the legend to Figure 2. Cells were treated with TPA (100 nM) for 3 min before mounting and fixation as described in Methods. (a) left: GFP-PKC and DsRed-cav co-distribution revealed in a green and red epifluorescence image (red and green filters) and right: DsRed-cav visualised using a red filter, the caveolin was mainly restricted to the perinuclear region with PKC more widely distributed. Three distinct GFP lifetimes are discernable in the lifetime image and were separately analysed. For the lifetime images shown in (b, c and d), representative single point analyses within the regions enclosed by dashed white lines are analysed in (f, g and h) respectively, as follows: (b and f) peripheral regions (τ avg: 1.8 ns; single point 1.87 ns [χ2 1.00]); (c and g) the nuclear region (τ avg: 2.0 ns; single point 2.00 [χ2 1.00]); (d and h) the cytoplasm (τ avg: 1.5 ns; single point 1.48 ns [χ2 1.56]). (e) example of derivation of average lifetime for one of the three areas, shown for the cytoplasm, with the analysis for the area enclosed in red (lifetime colour coding shown in the inset) and other similar areas in the cytoplasm indicated by white arrows. Cells shown are representative images from replicate experiments.
Three discrete areas could be ascertained from visual inspection of the FLIM images. Unfortunately when such areas are relatively small or scattered it is not possible to use the masking option of the software as used here to restrict analysis to the nucleus or cytoplasm since the data becomes noisy. Therefore a single pixel analysis within such areas has to be performed. These were found at the cell peripheral regions, the nuclear region and the perinuclear region (Figure 6b–d), These areas are outlined with dotted white lines. Figure 6f–h shows representative single pixel analyses for each of the three regions. The lifetimes (single exponential) were 2.0 ns (nuclear), 1.87 ns (peripheral) and 1.48 ns (perinuclear). This indicates that FRET, due to co-localisation of the caveolin and PKC, is predominantly occurring in the perinuclear region (see also analysis of a representative perinuclear region (depicted with white arrows in Figure 6e), and to some extent in the peripheral regions of the cell, with little or no co-localisation within the nuclear region, due to a lack of significant caveolin in that region.
The effect of the hormone bradykinin was also examined, a hormone that interacts with a G-protein coupled receptor and initiates intracellular Ca2+-release and also the production of DAG. Again the lifetime images (Figure 7) showed a reduced lifetime for GFP indicative of FRET to the DsRed and a co-localisation.
Figure 7 Lifetime imaging of GFP-PKC co-expressed with DsRed-cav in CHO cells: effect of bradykinin. 2P-FLIM images were collected as described in the legend to Figure 2. Cells were treated with 5 nM bradykinin for 3 min before mounting and fixation as described in Methods. The area in the cytoplasm outlined in red was analysed as shown in the colour coded inset (τ avg: 1.6 ns), showing PKC interaction with caveolin.
Discussion
Caveolin is the main structural component of caveoli and in addition performs an important role as a scaffold protein interacting with many key cellular signalling components [1,2]. In addition caveoli constitute a special class of membrane rafts, discrete areas in the cell membrane, within which related signalling molecules co-localise to optimise signalling events [30]. In other studies we have addressed the question whether PKC interacts with cell membrane rafts and have found that as a result of phorbol ester treatment the different PKC isoforms distribute into these regions, as defined by recovery in a detergent resistant fraction. (Slater, S. J. et al., unpublished observations) In the present study, we show for the first time, using a FRET-FLIM approach, that PKCα and caveolin co-localise upon increased intracellular Ca2+ or phorbol ester-induced interaction with PKC, however, in the 'resting' state the two molecules do not interact.
The translocation of PKC to various regions in the cell from a basically cytosolic location upon activation is a basic feature of signalling involving this molecule, and is important to many cell processes. Earlier studies considered that the main effects acted through IP3-induced release of intracellular Ca2+, induced by hormone stimulated G-protein coupled receptor-linked G-protein-induced stimulated release of IP3 and diacylglycerol (DAG) induced by the action of phospholipase C on phosphatidylinositol 4,5-bisphosphate. The PKC would then induced by the Ca2+ to move to the membrane, following which it would encounter with DAG which induces PKC to unfold and expose its substrate binding site which is followed by its phosphorylation. When PKCα is exposed to Ca2+ released from internal Ca2+ stores it transiently translocates to the outer membrane but returns to the cytoplasm [20,31], unless the activator levels are sustained, as occurs under the conditions used in the present work. It is now clear that PKC interacts with other cellular structures and organelles, including actin and RhoA for example [32,33]. Thus it is possible for a single isoform to be simultaneously involved in several distinct processes in the cell in different locations. Therefore to understand, and eventually bring under some form of control in therapeutic situations, new methodologies are required to unravel such complications. The combined approaches of fluorescence imaging and GFP-technology is now revolutionising our understanding of complex signalling events in cells such as presented by elements such as PKC. Conventional imaging techniques have been widely used in studies of signalling events in cells. Due to the equipment being widely available and inexpensive conventional epifluorescence imaging is widely utilised. However, due to problems such as photobleaching, spectral overlap, and lack of information on concentrations etc the information that can be gained is limited and often over-interpreted. A major question in cell biology is currently whether one particular protein interacts with another. In the present study using a 2P-FLIM-FRET approach we were able to unequivocally demonstrate that PKC moves to caveolin containing structures as a result of interaction with Ca2+ or phorbol ester.
The principle caveolin containing structures in the cell are caveoli, which transport to and from the plasma membrane and endosomes [16]. In the present work we showed that PKC translocates in a Ca2+-dependent manner. To examine this we used a 3 min time point. From other studies on CHO cells (Kelly, M. et al., in preparation) we have found that PKC translocation takes place over a 0–3 min time frame after which the cell morphology begins to change. Therefore for the purposes of this study, which was to prove the technology and to find out if PKC and caveolin interact, the 3 min time frame was optimal. It is important to note that PKC can translocate to a region as a result of Ca2+ interaction but this does not mean that it is catalytically active. It then has to encounter DAG or other signalling elements in order to gain phosphorylation ability. The nature of this element and whether DAG is produced in these regions of the cell has yet to be determined. In addition we do not yet know the role of PKC in endosomes or what other signalling molecules may be involved. It is possible that in the normal of the cell PKC is directed to endosomes as a part of a degradation pathway but this also remains to be determined.
In order to perform FLIM-FRET measurements it is important to determine that there are no spurious effects on the GFP donor lifetime properties. Here we showed that the translocation of PKC to its various locations in the cell had no effect on the GFP excited state lifetime in the absence of acceptor. It is also important to note that if there was a distribution of close GFP-DsRed distances, then a range of GFP decays for varying quenching would have occurred and a multi-exponential fit to the data would give an improved fit to the data. On the contrary, a single exponential fit to the data was optimal under FRET conditions indicating that the GFP and DsRed were in close contact. The next requirement is that since, compared to bulk fluorescence measurements, the signal is necessarily relatively weak, contributions from background and other factors must be minimized. In this study we found that the background interfered if the entire cell field was analysed. However if the cell region was isolated using the masking option in the software the background contribution could be eliminated. This can also be achieved by increasing the pixel resolution. Further, specific regions within the cell must be discerned for the imaging approach to be useful. Here we were able to isolate three discrete regions in the cells. These were PKC-GFP in the plasma membrane/peripheral regions, the perinuclear region (endosomes) and the nuclear region (see Table 1). The lack of caveolin in the nucleus allowed an internal control for each cell analysed and no quenching of the GFP lifetime was found in that region. In the endosome region significant quenching could be seen with the lifetime dropping from ~2.2 to ~1.5 ns. This is indicative of a direct interaction between the GFP and DsRed, attached to the PKC and Caveolin proteins. The peripheral region of the cell also showed a reduced GFP lifetime but by a lesser degree than in the perinuclear region. This could be due to the interaction being indirect but still close enough to allow FRET. Since FRET falls off in intensity according to the inverse forth law it will still have to be close but perhaps with a small intermediate protein. These possibilities remain to be further explored. Finally in this study we chose a point in time rather than following live cells since the PKC locality stabilised over a 2–3 min time course and could be conveniently analysed when the cells were fixed and mounted on slides. There is no indication, however, that the signal would be insufficient to follow time-dependent changes as collection times as low as 15 sec. could be used still allowing a signal sufficient for lifetime analysis.
Conclusion
In conclusion we have shown that PKC translocates to caveolin as a result of Ca2+ or phorbol ester interactions. The 2P-FLIM-FRET approach combined with GFP-technology offers a method for unambiguously determining the location of specific protein-protein interactions within the cell. Here we show that PKC translocates to and interacts with caveolin.
Methods
Chinese hamster ovary (CHO) cells were obtained from the ATCC (Manassas, VA) or were generously provided by Dr. Emma Leatherbarrow (MRC, Harwell, UK). Cells were cultured on glass cover slips in F-12 HAM's media (Sigma-Aldrich) supplemented with 10% Foetal calf serum (ICN Biomedicals or Gibco), 1% L-glutamine and 1% penicillin/streptomycin (Sigma Aldrich or Gibco) and cultured until reaching confluence.
For transfection 1–2 μg of pPKCα-EGFP vector DNA (Clontech) and/or DsRed1-cav-1, (kindly provided by Dr. R. Pagano [34], Mayo Clinic and Foundation) was added to cells in Lipofectin (Invitrogen) according to the manufacturer's protocol. Cells were incubated for a further 48 h and were then washed with phosphate buffered saline (PBS). Cells were then treated with 4β-12-O-Tetradecanoylphorbol-13-acetate (TPA), calcium ionophore or A23187, bradykinin (all from Sigma-Aldrich), as indicated, or were untreated (controls). The cells were then fixed in 3.7% formaldehyde solution in PBS for 20 minutes at room temperature washed again with PBS before mounting on slides with Crystal Mount before FLIM analysis.
Epifluorescence images were collected using a Nikon TE2000 U or Olympus XI microscope and a GFP or DsRed filter cube set and a 60× objective. Fluorescence lifetime images were obtained using a 2P-microscopy apparatus, using the Nikon microscope, constructed in the Central Laser Facility of the Rutherford Appleton laboratory, which has a Bio-Rad MRC600 confocal scanning system or external x, y galvanometers, [35,36]. Laser light at a wavelength of 850 nm was obtained from a Titanium Sapphire, 82 MHz, mode-locked laser (Spectra-Physics), with a pulse width of 120 fs. The light was focused to a diffraction-limited spot through an air (x40, n.a. 0.9) and specimens illuminated at the microscope stage by passing the beam through the MRC600 scan head or through a dichroic at the epifluorescence port.
Fluorescence emission was passed through a BG39 filter (Comar) to remove the laser line. The scan was operated in the normal mode, and line, frame and pixel clock signals were generated and synchronized with an external fast micro channel plate – photomultiplier tube, which was used as the detector (Becker & Hickl, GmbH Germany). These were linked via a Time-Correlated-Single Photon Counting PC module SPC700 (Becker & Hickl). With this system we were able to distinguish distinct lifetime regions within a single cell provided the lifetimes differed by greater than 0.2 ns. The set up was used to excite GFP fluorescence, the lifetime of which was measured as a function of the quenching by DsRed when the GFP-PKC and DsRed-cav were close enough for fluorophore tags to participate in FRET.
Images (6 bit, 256 × 256 pixels) were exported from Becker & Hickl software as bitmaps and converted into TIFF files. Image analysis was performed using either Irfanviewer or Image Pro 5.1 software (steady state), FLIM analyses were performed using SPC Image 2.0 software (Becker & Hickl).
Authors' contributions
CDS designed and performed all experiments and analyses and drafted the manuscript. SWB supervised the data collection instrumentation helped draft the manuscript. SJS helped to design the experiments and supervised pilot fluorescence studies. AWP and SWB designed and constructed the apparatus used for fluorescence lifetime measurements. AWP helped draft the manuscript. All authors have read and approved the manuscript.
Acknowledgements
This work was supported in part by United States Public Health Service Grants AA10990 and R37-AA10978. We are extremely grateful to Dr. R. Pagano for gift of DsRed-cav, Dr. Emma Leatherbarrow for CHO cells and M. Kelly and V Gaur for help in some of the imaging experiments.
==== Refs
Isshiki M Anderson RG Calcium signal transduction from caveolae Cell Calcium 1999 26 201 8 10643558 10.1054/ceca.1999.0073
Liu P Rudick M Anderson RGW Multiple Functions of Caveolin-1 J Biol Chem 2002 277 41295 41298 12189159 10.1074/jbc.R200020200
Cohen AW Hnasko R Schubert W Lisanti MP Role of Caveolae and Caveolins in Health and Disease Physiol Rev 2004 84 1341 1379 15383654 10.1152/physrev.00046.2003
Nishizuka Y Protein kinase C and lipid signaling for sustained cellular responses FASEB J 1995 9 484 496 7737456
Mellor H Parker PJ The extended protein kinase C superfamily Biochem J 1998 332 281 92 9601053
Newton AC Protein kinase C: structure, function, and regulation J Biol Chem 1995 270 28495 8 7499357 10.1074/jbc.270.43.25526
Clarke H Marano CW Peralta Soler A Mullin JM Modification of tight junction function by protein kinase C isoforms Adv Drug Deliv Rev 2000 41 283 301 10854687 10.1016/S0169-409X(00)00047-8
Dempsey EC Newton AC Mochly-Rosen D Fields AP Reyland ME Insel PA Messing RO Protein kinase C isozymes and the regulation of diverse cell responses Am J Physiol Lung Cell Mol Physiol 2000 279 L429 38 10956616
Jaken S Protein kinase C isozymes and substrates Curr Opin Cell Biol 1996 8 168 73 8791416 10.1016/S0955-0674(96)80062-7
Musashi M Ota S Shiroshita N The role of protein kinase C isoforms in cell proliferation and apoptosis Int J Hematol 2000 72 12 9 10979203
Naruse K King GL Protein kinase C and myocardial biology and function Circ Res 2000 86 1104 6 10850959
Mineo C Ying Y-S Chapline C Jaken S Anderson RGW Targeting of Protein Kinase C alpha to Caveolae J Cell Biol 1998 141 601 610 9566962 10.1083/jcb.141.3.601
Lin D Zhou J Zelenka PS Takemoto DJ Protein kinase C gamma regulation of gap junction activity through caveolin-1-containing lipid rafts Invest Ophthalmol Vis Sci 2003 44 5259 68 14638725 10.1167/iovs.03-0296
Wu D Foreman TL Gregory CW McJilton MA Wescott GG Ford OH Alvey RF Mohler JL Terrian DM Protein Kinase C epsilon Has the Potential to Advance the Recurrence of Human Prostate Cancer Cancer Res 2002 62 2423 2429 11956106
Tang ZL Scherer PE Lisanti MP The primary sequence of murine caveolin reveals a conserved consensus site for phosphorylation by protein kinase C Gene 1994 147 299 300 7926819 10.1016/0378-1119(94)90087-6
Prevostel C Alice V Joubert D Parker P Protein kinase Calpha actively downregulates through caveolae-dependent traffic to an endosomal compartment J Cell Sci 2000 113 2575 2584 10862715
Upla P Marjomaki V Kankaanpaa P Ivaska J Hyypia T Van Der Goot FG Heino J Clustering induces a lateral redistribution of alpha 2 beta 1 integrin from membrane rafts to caveolae and subsequent protein kinase C-dependent internalization Mol Biol Cell 2004 15 625 36 14657242 10.1091/mbc.E03-08-0588
Sakai N Sasaki K Ikegaki N Shirai Y Ono Y Saito N Direct visualization of the translocation of the gamma-subspecies of protein kinase C in living cells using fusion proteins with green fluorescent protein J Cell Biol 1997 139 1465 76 9396752 10.1083/jcb.139.6.1465
Oancea E Teruel MN Quest AF Meyer T Green fluorescent protein (GFP)-tagged cysteine-rich domains from protein kinase C as fluorescent indicators for diacylglycerol signaling in living cells J Cell Biol 1998 140 485 98 9456311 10.1083/jcb.140.3.485
Shirai Y Sakai N Saito N Subspecies-specific targeting mechanism of protein kinase C Jpn J Pharmacol 1998 78 411 7 9920197 10.1254/jjp.78.411
Bastiaens PI Squire A Fluorescence lifetime imaging microscopy: spatial resolution of biochemical processes in the cell Trends Cell Biol 1999 9 48 52 10087617 10.1016/S0962-8924(98)01410-X
Hanley QS Subramaniam V Arndt-Jovin DJ Jovin TM Fluorescence lifetime imaging: multi-point calibration, minimum resolvable differences, and artifact suppression Cytometry 2001 43 248 60 11260592 10.1002/1097-0320(20010401)43:4<248::AID-CYTO1057>3.0.CO;2-Y
Harpur AG Wouters FS Bastiaens PI Imaging FRET between spectrally similar GFP molecules in single cells Nat Biotechnol 2001 19 167 9 11175733 10.1038/84443
Ng T Squire A Hansra G Bornancin F Prevostel C Hanby A Harris W Barnes D Schmidt S Mellor H others Imaging protein kinase Calpha activation in cells Science 1999 283 2085 9 10092232 10.1126/science.283.5410.2085
Chen Y Periasamy A Characterization of two-photon excitation fluorescence lifetime imaging microscopy for protein localization Micro Res Tech 2004 63 72 80 10.1002/jemt.10430
Becker W Bergmann A Hink MA Konig K Benndorf K Biskup C Fluorescence lifetime imaging by time-correlated single-photon counting Microsc Res Tech 2004 63 58 66 14677134 10.1002/jemt.10421
Duncan RR Bergmann A Cousin MA Apps DK Shipston MJ Multi-dimensional time-correlated single photon counting(TCSPC) fluorescence lifetime imaging microscopy (FLIM) to detect FRET in cells J Microsc 2004 215 1 12 15230870 10.1111/j.0022-2720.2004.01343.x
Maasch C Wagner S Lindschau C Alexander G Buchner K Gollasch M Luft FC Haller H Protein kinase calpha targeting is regulated by temporal and spatial changes in intracellular free calcium concentration [Ca(2+)](i) Faseb J 2000 14 1653 63 10929000 10.1096/fj.14.11.1653
Hu T Exton JH Protein Kinase Calpha Translocates to the Perinuclear Region to Activate Phospholipase D1 J Biol Chem 2004 279 35702 35708 15187091 10.1074/jbc.M402372200
Quest AF Leyton L Parraga M Caveolins, caveolae, and lipid rafts in cellular transport, signaling, and disease Biochem Cell Biol 2004 82 129 44 15052333 10.1139/o03-071
Almholt K Arkhammar PO Thastrup O Tullin S Simultaneous visualization of the translocation of protein kinase C alpha-green fluorescent protein hybrids and intracellular calcium concentrations Biochem J 1999 337 211 8 9882617 10.1042/0264-6021:3370211
Slater SJ Milano SK Stagliano BA Gergich KJ Curry JP Taddeo FJ Stubbs CD Interaction of protein kinase C with filamentous actin: isozyme specificity resulting from divergent phorbol ester and calcium dependencies Biochemistry 2000 39 271 80 10630986 10.1021/bi9916527
Slater SJ Seiz JL Stagliano BA Stubbs CD Interaction of protein kinase C isozymes with Rho GTPases Biochemistry 2001 40 4437 45 11284700 10.1021/bi001654n
Singh RD Puri V Valiyaveettil JT Marks DL Bittman R Pagano RE Selective caveolin-1-dependent endocytosis of glycosphingolipids Mol Biol Cell 2003 14 3254 65 12925761 10.1091/mbc.E02-12-0809
Connelly JP Botchway SW Kunz L Pattison D Parker AW MacRobert AJ Time-resolved fluorescence imaging of photosensitiser distributions in mammalian cells using a picosecond laser line- scanning microscope J Photochem Photobiol A 2001 142 169 175 10.1016/S1010-6030(01)00511-1
Botchway SW Parker AW Barba I Brindle K Development of a time-correlated single photon counting multiphoton laser scanning confocal microscope Central Laser Facility Annual Report 2000–2001 2001 RAL-TR-2001-030 170 171
| 15854225 | PMC1131895 | CC BY | 2021-01-04 16:39:11 | no | BMC Cell Biol. 2005 Apr 26; 6:22 | utf-8 | BMC Cell Biol | 2,005 | 10.1186/1471-2121-6-22 | oa_comm |
==== Front
BMC Cell BiolBMC Cell Biology1471-2121BioMed Central London 1471-2121-6-221585422510.1186/1471-2121-6-22Research ArticleThe use of time-resolved fluorescence imaging in the study of protein kinase C localisation in cells Stubbs Christopher D [email protected] Stanley W [email protected] Simon J [email protected] Anthony W [email protected] Department of Pathology and Cell Biology, Thomas Jefferson University, Philadelphia PA 19107 USA2 The Central Laser Facility, Rutherford Appleton Laboratory, CCLRC, Chilton, OX11 OQX UK2005 26 4 2005 6 22 22 24 12 2004 26 4 2005 Copyright © 2005 Stubbs et al; licensee BioMed Central Ltd.2005Stubbs 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
Two-photon-excitation fluorescence lifetime imaging (2P-FLIM) was used to investigate the association of protein kinase C alpha (PKCα) with caveolin in CHO cells. PKCα is found widely in the cytoplasm and nucleus in most cells. Upon activation, as a result of increased intracellular Ca2+ and production of DAG, through G-protein coupled-phospholipase C signalling, PKC translocates to a variety of regions in the cell where it phosphorylates and interacts with many signalling pathways. Due to its wide distribution, discerning a particular interaction from others within the cell is extremely difficult
Results
Fluorescence energy transfer (FRET), between GFP-PKCα and DsRed-caveolin, was used to investigate the interaction between caveolin and PKC, an aspect of signalling that is poorly understood. Using 2P-FLIM measurements, the lifetime of GFP was found to decrease (quench) in certain regions of the cell from ~2.2 ns to ~1.5 ns when the GFP and DsRed were sufficiently close for FRET to occur. This only occurred when intracellular Ca2+ increased or in the presence of phorbol ester, and was an indication of PKC and caveolin co-localisation under these conditions. In the case of phorbol ester stimulated PKC translocation, as commonly used to model PKC activation, three PKC areas could be delineated. These included PKCα that was not associated with caveolin in the nucleus and cytoplasm, PKCα associated with caveolin in the cytoplasm/perinuclear regions and probably in endosomes, and PKC in the peripheral regions of the cell, possibly indirectly interacting with caveolin.
Conclusion
Based on the extent of lifetime quenching observed, the results are consistent with a direct interaction between PKCα and caveolin in the endosomes, and possibly an indirect interaction in the peripheral regions of the cell. The results show that 2P-FLIM-FRET imaging offers an approach that can provide information not only confirming the occurrence of specific protein-protein interactions but where they occur within the cell.
==== Body
Background
In common with other signalling proteins, PKC has been suggested to associate with caveolin-1 [1,2], a key component of caveoli and membrane rafts, that are proposed to exist as signalling platforms in the plasma membrane and elsewhere in the cell. However, where in the cell the PKC-caveolin interaction might occur and under what conditions remains unclear.
Caveolae are vesicular organelles are that are involved in a wide range of cellular functions, serving as platforms or rafts, wherein reside a wide variety of signalling molecules [3]. The caveolin proteins (caveolin-1, -2, and -3) act as the structural components of caveolae. They also function as scaffolding proteins and as such recruit numerous signalling molecules to caveolae where their activity is regulated. PKC is a signalling molecule of major importance in cells, which in the form of twelve isoforms, regulates numerous signalling cascades by virtue of its ability to phosphorylate target proteins that include receptors, G-proteins, ion channels as well as other kinases [4-6]. This leads to control of numerous cellular processes, such as secretion, proliferation, differentiation, apoptosis, permeability, migration, hypertrophy etc [4,5,7-11].
While it has been shown that isolated caveoli interact with purified PKCα [12], PKCγ [13] and PKCε [14] using immunoprecipitation, where in the cell this occurs is not known. Caveolin contains a sequence that is a consensus site for phosphorylation by PKC [15], while down-regulation of plasma membrane-translocated PKCα involvesinternalization of the active enzyme that involves ubiquitination, through a caveolae-dependent mechanism, followed by multisite dephosphorylation and down-regulation in a perinuclear compartment in a time dependent manner (~30 min after stimulated translocation to the plasma membrane) [16]. It has also been shown that endocytic trafficking via caveolae may be a PKCα-dependent process [17]. These observations lead to the question of whether PKCα interacts directly with caveolin, and where in the cell this occurs, a question we examined in this present study.
The classic biochemical or immunoprecipitation approaches for determining the location of signalling molecules in cells, while commonly used, is severely limited for several reasons. The main drawback is that it involves destruction of the cell, resulting in loss of spatial information. Staining the cells with fluorescent antibodies provides a useful advance enabling apparent "co-localisation" to be obtained when two different fluorophores are used. However, the two fluorophores may still be a considerable distance apart without any protein-protein interaction between the pair occurring. The method also suffers from problems of photobleaching and when the probes are in abundance in the cell false co-localisation data can result.
Recently, with the advent of green fluorescent protein (GFP) technology, considerable new information has become available using imaging approaches. This allows the protein of interest to be tagged by expression in the cell, allowing it to be functionally located within the cell without recourse to antibodies. Numerous papers have shown signalling molecules in cells can be tracked as they move between subcellular locations using expression of GFP-tagged proteins and fluorescence microscopy. PKC has been studied in a wide range of situations using GFP-tags (e.g. [18-20]. Lifetime imaging has been used as a localisation tool for GFP-tagged proteins [21-23] and using this approach both PKCα activation levels, along with localisation, has been detected through the binding of fluorescently tagged phosphorylation site-specific antibodies using fluorescence energy transfer (FRET), measured through a donor fluorophore on the PKC [24]. In most cases the localisation was either followed in real-time or after fixation, following initiation of intracellular signalling. There are now a number of different fluorescent proteins available in the "GFP" family, for example cyan fluorescent protein (CFP), yellow fluorescent protein (YFP), red fluorescent protein (DsRed) etc.
A considerable advance over the co-localisation approach is to use steady state FRET between suitably tagged proteins. A further improvement over steady state FRET is achieved by monitoring the lifetime of the donor, which is independent of changes in concentration, photobleaching and various limitations over intensity-based detection. Donor lifetime quenching is evidence for a direct physical interaction and in addition does not require corrections due to spectral over-lap that are required in steady state FRET. The use of this approach FRET-FLIM for localisation using different GFP-type constructs has been described recently in the literature [25-27]. This approach in combination with 2-photon (2P) excitation provides a powerful imaging technique. It does not require complex corrections that intensity-based FRET entails, allows excitation within a narrow focussed plane without contributions outside this region, as would be the case in one-photon excitation, as well as other advantages, such as the ability to image deeper in tissue etc.
In this work the interaction of PKCα with caveolin, was investigated using two-photon-fluorescence lifetime imaging (2P-FLIM). This was determined by investigating the FRET between GFP-tagged PKCα and DsRed-caveolin (DsRed-cav). Using the quenched lifetime of the GFP-tag, areas showing co-localisation in CHO cells was identified after activation of the PKC had occurred. When the GFP-PKC is induced to translocate to membranes, excitation of the GFP at 850 nm, for 2P excitation, should lead to FRET to the DsRed construct if the two are co-localised. This would be seen as a reduced GFP lifetime, in contrast to areas in which the GFP-PKC resides but not with caveolin. This information was determined using a FLIM time-correlated single photon counting set up (TCSPC) with the frequency doubled output of a Tsunami pulsed laser (110 fs), coupled to an inverted microscope. Based on the lifetime quenching data the results were consistent with three populations of activated GFP-PKC. One within the cytoplasmic area of the cells, which was probably undergoing a direct interaction with caveolin, likely in endosomes, a second population in the cytoplasm and in the nucleus that was not interacting with caveolin and a third at the plasma membrane possibly indirectly interacting with caveolin.
Results
The epifluorescence images in Figure 1 show GFP-PKC and DsRed-cav localisation as green and red colours respectively. The images were taken before and after exposure to the phorbol ester TPA for 3 min. It can be seen that PKC accumulated at the cell periphery and in organelles in the perinuclear region.
Figure 1 Fluorescence intensity images of GFP-PKC and DsRed-cav expressed in CHO cells. GFP-PKC and/or DsRed-cav were transiently expressed in CHO cells on glass coverslips by culturing for 48 hr. The cells were then treated with Ca2+-ionophore (1 nM) or TPA (100 nM) for 3 min before the cells were fixed on microscope slides as described under Methods. Images were acquired using a Olympus IX-70 inverted epifluorescence microscope with a 60× objective and GFP/DsRed filter sets and a Olympus C3030 camera. (a) a representative image before and (b) after Ca2+-induced translocation by ionophore of GFP-PKC to the peripheral membrane and to discrete regions in the perinuclear region. (c) image of DsRed-cav and (d) GFP-PKC, after treatment with 100 nM TPA.
A lifetime image of a CHO cell transiently transfected with GFP-PKCα for 48 hr is shown in Figure 2, with the corresponding epifluorescence image. It can be seen that PKCα is widely and abundantly distributed in the cell cytoplasm and nucleus. The data collection time for this and other lifetime images was optimally 2–3 minutes. The lifetime data for the various conditions examined is also summarised in Table 1.
Figure 2 Fluorescence lifetime imaging of GFP-PKC expressed in CHO cells. GFP-PKC was transiently expressed in CHO cells as described in the legend in Figure 1 and detailed under Methods. Images were acquired using a Nikon 2000 inverted epifluorescence microscope with a 60× objective and GFP/DsRed filter sets and an ICAM camera. The 2P-FLIM images were collected using the TCSPC fast scanning imaging mode as described under Methods. (a) Conventional epifluorescence image of GFP-PKC distribution in a resting CHO cell, (b) lifetime image of the same cell with the analysis area enclosed by the red line (cytosol) shown (with colour coding) in the inset and giving an average lifetime of ~2.2 ns and (c) with the analysis area enclosed by the red line (nucleus) shown in the inset giving an average lifetime of ~2.0 ns. Cells shown are representative images from replicate experiments.
Table 1 Summary of average lifetimes for GFP emission under various conditions
Transfection Fig Treatment Area analysed Avg Lifetime (ns)
GFP-PKC 2 None cytoplasm 2.2
None nucleus 2.2
3 TPA cytoplasm 2.2
TPA nucleus 2.3
GFP-PKC/DsRed-cav 4 None cytoplasm 2.4
None cytoplasm 2.5
5 Ca2+ cytoplasm 1.6*
Ca2+ nucleus 2.0
6 TPA cytoplasm excluding peripheral 1.5*
TPA nucleus 2.0
TPA peripheral 1.8*
7 Bradykinin cytoplasm 1.6*
*reduced lifetime indicating PKC-GFP-DsRed-cav interaction
The lifetime of the GFP averaged for the cell area was around 2.2 ns (Figure 2b). The software analysis program optionally allows for whole or part of the field to be subjected to a lifetime analysis. By this means contributions from the background can be effectively excluded, as well as different areas with the cell compared. It is also possible to analyse the lifetimes on a pixel by pixel basis in discrete areas. Both approaches were used in this study. If the analysis was not restricted to the cell area and the whole field is included the background contributions became significant and a background component with a lifetime of ~1.4 ns became apparent. A single pixel analysis, within the cell area, yielded 2.0 ns for a single exponential with a reduced χ2 of 1.22, a two exponential analysis did not give an improved fit. Based on this, and observations from the literature, we analysed the lifetime of GFP as a single exponential. Analysis of the GFP-PKC located within the nucleus showed the same lifetime as that in the cytoplasm (Figure 2c).
We next examined the effect of the phorbol ester TPA on the GFP-PKC lifetime. If a FRET analysis is to be usefully undertaken it is first essential that while the PKC may redistribute in the cell, as a result of the TPA treatment, it should not result any changes to the lifetime. This was in fact the case as shown in Figure 3 where the lifetime for GFP-PKC across the cell area remained unchanged in the ~2.2 ns region.
Figure 3 Fluorescence lifetime imaging of GFP-PKC expressed in CHO cells: effect of TPA. 2P-FLIM images were collected as described in the legend to Figure 2 except cells were treated with TPA (100 nM) for 3 min. Treatment with the phorbol ester did not affect the fluorescence lifetime of GFP attached to PKC. (a) Fluorescence lifetime image with the analysis area enclosed by the red line (nucleus) shown in inset giving an average lifetime of ~2.1 to 2.2 ns. (b) Fluorescence lifetime image with the analysis area enclosed by the red line (cytosol) shown in inset giving an average lifetime of ~2.2 ns. Insets: lifetime distributions with colour coding.
The images in Figure 4 are from cells with GFP-PKC co-expressed with DsRed-cav. The epifluorescence images show that both PKC and caveolin were widely distributed in most cell areas, except that caveolin was not found in the nucleus and was concentrated in the perinuclear region. The lifetime images of GFP-PKC were acquired under conditions that would produce no contributions from DsRed fluorescence. This was achieved since the 2-P excitation was nominally at 425 nm, where there is virtually no excitation of the DsRed, also a 500 nm band-pass filter was used to exclude DsRed emission. Therefore only the GFP emission was collected. If the DsRed were to be close enough to PKC then a shortened lifetime would result due to FRET occurring. Again the average lifetime of the GFP-PKC was >2 ns, showing that under conditions when the cell is not stimulated (as defined here by Ca2+-mobilisation or phorbol ester activation of PKC), then PKC and caveolin do not co-associate in the cell.
Figure 4 Fluorescence lifetime imaging of GFP-PKC co-expressed with DsRed-cav in CHO cells. 2P-FLIM images were collected as described in the legend to Figure 2. Co-expression of the GFP-PKC with DsRed-cav does not affect the lifetime of the GFP showing that in the unstimulated state PKC is not associated with caveolin. Epifluorescence images for excitation of DsRed (a) and GFP along with DsRed (b) showing that the PKC and caveolin co-distributed in the cytosol. Fluorescence lifetime images with the analysis area enclosed by the red line, (cytosol) (c) or nucleus both essentially showing a lifetime as for Figs 2–3 centred around ~2.2 ns. Cells shown are representative images from replicate experiments.
When Ca2+ ionophore is added to cells Ca2+ gains entry and this induces immediate PKC translocation to various locations within the cells, including the perinuclear regions and peripheral membranes, as has been extensively shown in the literature (e.g. see ref [28,16,29]). The localisation of caveolin did not appear to change upon addition of Ca2+ ionophore to the cells (results not shown). For lifetime images it is important to note that what we are observing is the lifetime of the GFP, not its intensity. The lifetime images shown in Figure 5 show that the GFP-PKC fluorescence lifetime in the cytoplasm is reduced to ~1.6 ns. This is due to the quenching by DsRed, since the lifetime is unaffected in the absence of DsRed-cav. By contrast, the GFP-PKC lifetime in the nucleus was unaffected by the Ca2+ treatment, since there was little or no caveolin within the nucleus (Figure 5).
Figure 5 Lifetime imaging of GFP-PKC co-expressed with DsRed-cav in CHO cells: effect of Ca2+-ionophore. 2P-FLIM images were collected as described in the legend to Figure 2. Cells were treated with ionophore for 3 min before mounting and fixation as described in Methods. The epifluorescence image in the inset shows the DsRed-cav distribution (cytoplasmic) which was not affected by the Ca2+ ionophore. When the cytoplasmic area was analysed, (a) as shown by the area within the red line, both orange and green/blue areas are seen indicating the presence of both GFP-PKC and quenched GFP-PKC – note that only GFP lifetime can be observed in the lifetime images. This indicates that DsRed-cav was sufficiently close to the PKC-GFP to induce a quenching of the GFP by the DsRed, i.e. the PKC is translocating to caveolin containing areas. By contrast, in the nucleus (b) only GFP-PKC was expressed and the lifetime was unquenched (~2.2 ns). This is the same as the lifetime for GFP-PKC when only the latter is expressed (see Figure 2). Cells shown are representative images from replicate experiments.
The treatment of the cells with the phorbol ester TPA not only induces translocation of PKC to different cell compartments but also produces catalytically active PKC. In general, PKC moves to the outer membrane and to perinuclear regions, and associates with various signalling and cytostructural components in the cell. The lifetime image data revealed that, similar to the effect of increasing intracellular Ca2+, phorbol ester induced PKC to associate with caveolin, the GFP lifetime being reduced accordingly, as shown in Figure 6.
Figure 6 Lifetime imaging of GFP-PKC co-expressed with DsRed-cav in CHO cells: effect of phorbol ester. Epifluorescence and 2P-FLIM images were collected as described in the legend to Figure 2. Cells were treated with TPA (100 nM) for 3 min before mounting and fixation as described in Methods. (a) left: GFP-PKC and DsRed-cav co-distribution revealed in a green and red epifluorescence image (red and green filters) and right: DsRed-cav visualised using a red filter, the caveolin was mainly restricted to the perinuclear region with PKC more widely distributed. Three distinct GFP lifetimes are discernable in the lifetime image and were separately analysed. For the lifetime images shown in (b, c and d), representative single point analyses within the regions enclosed by dashed white lines are analysed in (f, g and h) respectively, as follows: (b and f) peripheral regions (τ avg: 1.8 ns; single point 1.87 ns [χ2 1.00]); (c and g) the nuclear region (τ avg: 2.0 ns; single point 2.00 [χ2 1.00]); (d and h) the cytoplasm (τ avg: 1.5 ns; single point 1.48 ns [χ2 1.56]). (e) example of derivation of average lifetime for one of the three areas, shown for the cytoplasm, with the analysis for the area enclosed in red (lifetime colour coding shown in the inset) and other similar areas in the cytoplasm indicated by white arrows. Cells shown are representative images from replicate experiments.
Three discrete areas could be ascertained from visual inspection of the FLIM images. Unfortunately when such areas are relatively small or scattered it is not possible to use the masking option of the software as used here to restrict analysis to the nucleus or cytoplasm since the data becomes noisy. Therefore a single pixel analysis within such areas has to be performed. These were found at the cell peripheral regions, the nuclear region and the perinuclear region (Figure 6b–d), These areas are outlined with dotted white lines. Figure 6f–h shows representative single pixel analyses for each of the three regions. The lifetimes (single exponential) were 2.0 ns (nuclear), 1.87 ns (peripheral) and 1.48 ns (perinuclear). This indicates that FRET, due to co-localisation of the caveolin and PKC, is predominantly occurring in the perinuclear region (see also analysis of a representative perinuclear region (depicted with white arrows in Figure 6e), and to some extent in the peripheral regions of the cell, with little or no co-localisation within the nuclear region, due to a lack of significant caveolin in that region.
The effect of the hormone bradykinin was also examined, a hormone that interacts with a G-protein coupled receptor and initiates intracellular Ca2+-release and also the production of DAG. Again the lifetime images (Figure 7) showed a reduced lifetime for GFP indicative of FRET to the DsRed and a co-localisation.
Figure 7 Lifetime imaging of GFP-PKC co-expressed with DsRed-cav in CHO cells: effect of bradykinin. 2P-FLIM images were collected as described in the legend to Figure 2. Cells were treated with 5 nM bradykinin for 3 min before mounting and fixation as described in Methods. The area in the cytoplasm outlined in red was analysed as shown in the colour coded inset (τ avg: 1.6 ns), showing PKC interaction with caveolin.
Discussion
Caveolin is the main structural component of caveoli and in addition performs an important role as a scaffold protein interacting with many key cellular signalling components [1,2]. In addition caveoli constitute a special class of membrane rafts, discrete areas in the cell membrane, within which related signalling molecules co-localise to optimise signalling events [30]. In other studies we have addressed the question whether PKC interacts with cell membrane rafts and have found that as a result of phorbol ester treatment the different PKC isoforms distribute into these regions, as defined by recovery in a detergent resistant fraction. (Slater, S. J. et al., unpublished observations) In the present study, we show for the first time, using a FRET-FLIM approach, that PKCα and caveolin co-localise upon increased intracellular Ca2+ or phorbol ester-induced interaction with PKC, however, in the 'resting' state the two molecules do not interact.
The translocation of PKC to various regions in the cell from a basically cytosolic location upon activation is a basic feature of signalling involving this molecule, and is important to many cell processes. Earlier studies considered that the main effects acted through IP3-induced release of intracellular Ca2+, induced by hormone stimulated G-protein coupled receptor-linked G-protein-induced stimulated release of IP3 and diacylglycerol (DAG) induced by the action of phospholipase C on phosphatidylinositol 4,5-bisphosphate. The PKC would then induced by the Ca2+ to move to the membrane, following which it would encounter with DAG which induces PKC to unfold and expose its substrate binding site which is followed by its phosphorylation. When PKCα is exposed to Ca2+ released from internal Ca2+ stores it transiently translocates to the outer membrane but returns to the cytoplasm [20,31], unless the activator levels are sustained, as occurs under the conditions used in the present work. It is now clear that PKC interacts with other cellular structures and organelles, including actin and RhoA for example [32,33]. Thus it is possible for a single isoform to be simultaneously involved in several distinct processes in the cell in different locations. Therefore to understand, and eventually bring under some form of control in therapeutic situations, new methodologies are required to unravel such complications. The combined approaches of fluorescence imaging and GFP-technology is now revolutionising our understanding of complex signalling events in cells such as presented by elements such as PKC. Conventional imaging techniques have been widely used in studies of signalling events in cells. Due to the equipment being widely available and inexpensive conventional epifluorescence imaging is widely utilised. However, due to problems such as photobleaching, spectral overlap, and lack of information on concentrations etc the information that can be gained is limited and often over-interpreted. A major question in cell biology is currently whether one particular protein interacts with another. In the present study using a 2P-FLIM-FRET approach we were able to unequivocally demonstrate that PKC moves to caveolin containing structures as a result of interaction with Ca2+ or phorbol ester.
The principle caveolin containing structures in the cell are caveoli, which transport to and from the plasma membrane and endosomes [16]. In the present work we showed that PKC translocates in a Ca2+-dependent manner. To examine this we used a 3 min time point. From other studies on CHO cells (Kelly, M. et al., in preparation) we have found that PKC translocation takes place over a 0–3 min time frame after which the cell morphology begins to change. Therefore for the purposes of this study, which was to prove the technology and to find out if PKC and caveolin interact, the 3 min time frame was optimal. It is important to note that PKC can translocate to a region as a result of Ca2+ interaction but this does not mean that it is catalytically active. It then has to encounter DAG or other signalling elements in order to gain phosphorylation ability. The nature of this element and whether DAG is produced in these regions of the cell has yet to be determined. In addition we do not yet know the role of PKC in endosomes or what other signalling molecules may be involved. It is possible that in the normal of the cell PKC is directed to endosomes as a part of a degradation pathway but this also remains to be determined.
In order to perform FLIM-FRET measurements it is important to determine that there are no spurious effects on the GFP donor lifetime properties. Here we showed that the translocation of PKC to its various locations in the cell had no effect on the GFP excited state lifetime in the absence of acceptor. It is also important to note that if there was a distribution of close GFP-DsRed distances, then a range of GFP decays for varying quenching would have occurred and a multi-exponential fit to the data would give an improved fit to the data. On the contrary, a single exponential fit to the data was optimal under FRET conditions indicating that the GFP and DsRed were in close contact. The next requirement is that since, compared to bulk fluorescence measurements, the signal is necessarily relatively weak, contributions from background and other factors must be minimized. In this study we found that the background interfered if the entire cell field was analysed. However if the cell region was isolated using the masking option in the software the background contribution could be eliminated. This can also be achieved by increasing the pixel resolution. Further, specific regions within the cell must be discerned for the imaging approach to be useful. Here we were able to isolate three discrete regions in the cells. These were PKC-GFP in the plasma membrane/peripheral regions, the perinuclear region (endosomes) and the nuclear region (see Table 1). The lack of caveolin in the nucleus allowed an internal control for each cell analysed and no quenching of the GFP lifetime was found in that region. In the endosome region significant quenching could be seen with the lifetime dropping from ~2.2 to ~1.5 ns. This is indicative of a direct interaction between the GFP and DsRed, attached to the PKC and Caveolin proteins. The peripheral region of the cell also showed a reduced GFP lifetime but by a lesser degree than in the perinuclear region. This could be due to the interaction being indirect but still close enough to allow FRET. Since FRET falls off in intensity according to the inverse forth law it will still have to be close but perhaps with a small intermediate protein. These possibilities remain to be further explored. Finally in this study we chose a point in time rather than following live cells since the PKC locality stabilised over a 2–3 min time course and could be conveniently analysed when the cells were fixed and mounted on slides. There is no indication, however, that the signal would be insufficient to follow time-dependent changes as collection times as low as 15 sec. could be used still allowing a signal sufficient for lifetime analysis.
Conclusion
In conclusion we have shown that PKC translocates to caveolin as a result of Ca2+ or phorbol ester interactions. The 2P-FLIM-FRET approach combined with GFP-technology offers a method for unambiguously determining the location of specific protein-protein interactions within the cell. Here we show that PKC translocates to and interacts with caveolin.
Methods
Chinese hamster ovary (CHO) cells were obtained from the ATCC (Manassas, VA) or were generously provided by Dr. Emma Leatherbarrow (MRC, Harwell, UK). Cells were cultured on glass cover slips in F-12 HAM's media (Sigma-Aldrich) supplemented with 10% Foetal calf serum (ICN Biomedicals or Gibco), 1% L-glutamine and 1% penicillin/streptomycin (Sigma Aldrich or Gibco) and cultured until reaching confluence.
For transfection 1–2 μg of pPKCα-EGFP vector DNA (Clontech) and/or DsRed1-cav-1, (kindly provided by Dr. R. Pagano [34], Mayo Clinic and Foundation) was added to cells in Lipofectin (Invitrogen) according to the manufacturer's protocol. Cells were incubated for a further 48 h and were then washed with phosphate buffered saline (PBS). Cells were then treated with 4β-12-O-Tetradecanoylphorbol-13-acetate (TPA), calcium ionophore or A23187, bradykinin (all from Sigma-Aldrich), as indicated, or were untreated (controls). The cells were then fixed in 3.7% formaldehyde solution in PBS for 20 minutes at room temperature washed again with PBS before mounting on slides with Crystal Mount before FLIM analysis.
Epifluorescence images were collected using a Nikon TE2000 U or Olympus XI microscope and a GFP or DsRed filter cube set and a 60× objective. Fluorescence lifetime images were obtained using a 2P-microscopy apparatus, using the Nikon microscope, constructed in the Central Laser Facility of the Rutherford Appleton laboratory, which has a Bio-Rad MRC600 confocal scanning system or external x, y galvanometers, [35,36]. Laser light at a wavelength of 850 nm was obtained from a Titanium Sapphire, 82 MHz, mode-locked laser (Spectra-Physics), with a pulse width of 120 fs. The light was focused to a diffraction-limited spot through an air (x40, n.a. 0.9) and specimens illuminated at the microscope stage by passing the beam through the MRC600 scan head or through a dichroic at the epifluorescence port.
Fluorescence emission was passed through a BG39 filter (Comar) to remove the laser line. The scan was operated in the normal mode, and line, frame and pixel clock signals were generated and synchronized with an external fast micro channel plate – photomultiplier tube, which was used as the detector (Becker & Hickl, GmbH Germany). These were linked via a Time-Correlated-Single Photon Counting PC module SPC700 (Becker & Hickl). With this system we were able to distinguish distinct lifetime regions within a single cell provided the lifetimes differed by greater than 0.2 ns. The set up was used to excite GFP fluorescence, the lifetime of which was measured as a function of the quenching by DsRed when the GFP-PKC and DsRed-cav were close enough for fluorophore tags to participate in FRET.
Images (6 bit, 256 × 256 pixels) were exported from Becker & Hickl software as bitmaps and converted into TIFF files. Image analysis was performed using either Irfanviewer or Image Pro 5.1 software (steady state), FLIM analyses were performed using SPC Image 2.0 software (Becker & Hickl).
Authors' contributions
CDS designed and performed all experiments and analyses and drafted the manuscript. SWB supervised the data collection instrumentation helped draft the manuscript. SJS helped to design the experiments and supervised pilot fluorescence studies. AWP and SWB designed and constructed the apparatus used for fluorescence lifetime measurements. AWP helped draft the manuscript. All authors have read and approved the manuscript.
Acknowledgements
This work was supported in part by United States Public Health Service Grants AA10990 and R37-AA10978. We are extremely grateful to Dr. R. Pagano for gift of DsRed-cav, Dr. Emma Leatherbarrow for CHO cells and M. Kelly and V Gaur for help in some of the imaging experiments.
==== Refs
Isshiki M Anderson RG Calcium signal transduction from caveolae Cell Calcium 1999 26 201 8 10643558 10.1054/ceca.1999.0073
Liu P Rudick M Anderson RGW Multiple Functions of Caveolin-1 J Biol Chem 2002 277 41295 41298 12189159 10.1074/jbc.R200020200
Cohen AW Hnasko R Schubert W Lisanti MP Role of Caveolae and Caveolins in Health and Disease Physiol Rev 2004 84 1341 1379 15383654 10.1152/physrev.00046.2003
Nishizuka Y Protein kinase C and lipid signaling for sustained cellular responses FASEB J 1995 9 484 496 7737456
Mellor H Parker PJ The extended protein kinase C superfamily Biochem J 1998 332 281 92 9601053
Newton AC Protein kinase C: structure, function, and regulation J Biol Chem 1995 270 28495 8 7499357 10.1074/jbc.270.43.25526
Clarke H Marano CW Peralta Soler A Mullin JM Modification of tight junction function by protein kinase C isoforms Adv Drug Deliv Rev 2000 41 283 301 10854687 10.1016/S0169-409X(00)00047-8
Dempsey EC Newton AC Mochly-Rosen D Fields AP Reyland ME Insel PA Messing RO Protein kinase C isozymes and the regulation of diverse cell responses Am J Physiol Lung Cell Mol Physiol 2000 279 L429 38 10956616
Jaken S Protein kinase C isozymes and substrates Curr Opin Cell Biol 1996 8 168 73 8791416 10.1016/S0955-0674(96)80062-7
Musashi M Ota S Shiroshita N The role of protein kinase C isoforms in cell proliferation and apoptosis Int J Hematol 2000 72 12 9 10979203
Naruse K King GL Protein kinase C and myocardial biology and function Circ Res 2000 86 1104 6 10850959
Mineo C Ying Y-S Chapline C Jaken S Anderson RGW Targeting of Protein Kinase C alpha to Caveolae J Cell Biol 1998 141 601 610 9566962 10.1083/jcb.141.3.601
Lin D Zhou J Zelenka PS Takemoto DJ Protein kinase C gamma regulation of gap junction activity through caveolin-1-containing lipid rafts Invest Ophthalmol Vis Sci 2003 44 5259 68 14638725 10.1167/iovs.03-0296
Wu D Foreman TL Gregory CW McJilton MA Wescott GG Ford OH Alvey RF Mohler JL Terrian DM Protein Kinase C epsilon Has the Potential to Advance the Recurrence of Human Prostate Cancer Cancer Res 2002 62 2423 2429 11956106
Tang ZL Scherer PE Lisanti MP The primary sequence of murine caveolin reveals a conserved consensus site for phosphorylation by protein kinase C Gene 1994 147 299 300 7926819 10.1016/0378-1119(94)90087-6
Prevostel C Alice V Joubert D Parker P Protein kinase Calpha actively downregulates through caveolae-dependent traffic to an endosomal compartment J Cell Sci 2000 113 2575 2584 10862715
Upla P Marjomaki V Kankaanpaa P Ivaska J Hyypia T Van Der Goot FG Heino J Clustering induces a lateral redistribution of alpha 2 beta 1 integrin from membrane rafts to caveolae and subsequent protein kinase C-dependent internalization Mol Biol Cell 2004 15 625 36 14657242 10.1091/mbc.E03-08-0588
Sakai N Sasaki K Ikegaki N Shirai Y Ono Y Saito N Direct visualization of the translocation of the gamma-subspecies of protein kinase C in living cells using fusion proteins with green fluorescent protein J Cell Biol 1997 139 1465 76 9396752 10.1083/jcb.139.6.1465
Oancea E Teruel MN Quest AF Meyer T Green fluorescent protein (GFP)-tagged cysteine-rich domains from protein kinase C as fluorescent indicators for diacylglycerol signaling in living cells J Cell Biol 1998 140 485 98 9456311 10.1083/jcb.140.3.485
Shirai Y Sakai N Saito N Subspecies-specific targeting mechanism of protein kinase C Jpn J Pharmacol 1998 78 411 7 9920197 10.1254/jjp.78.411
Bastiaens PI Squire A Fluorescence lifetime imaging microscopy: spatial resolution of biochemical processes in the cell Trends Cell Biol 1999 9 48 52 10087617 10.1016/S0962-8924(98)01410-X
Hanley QS Subramaniam V Arndt-Jovin DJ Jovin TM Fluorescence lifetime imaging: multi-point calibration, minimum resolvable differences, and artifact suppression Cytometry 2001 43 248 60 11260592 10.1002/1097-0320(20010401)43:4<248::AID-CYTO1057>3.0.CO;2-Y
Harpur AG Wouters FS Bastiaens PI Imaging FRET between spectrally similar GFP molecules in single cells Nat Biotechnol 2001 19 167 9 11175733 10.1038/84443
Ng T Squire A Hansra G Bornancin F Prevostel C Hanby A Harris W Barnes D Schmidt S Mellor H others Imaging protein kinase Calpha activation in cells Science 1999 283 2085 9 10092232 10.1126/science.283.5410.2085
Chen Y Periasamy A Characterization of two-photon excitation fluorescence lifetime imaging microscopy for protein localization Micro Res Tech 2004 63 72 80 10.1002/jemt.10430
Becker W Bergmann A Hink MA Konig K Benndorf K Biskup C Fluorescence lifetime imaging by time-correlated single-photon counting Microsc Res Tech 2004 63 58 66 14677134 10.1002/jemt.10421
Duncan RR Bergmann A Cousin MA Apps DK Shipston MJ Multi-dimensional time-correlated single photon counting(TCSPC) fluorescence lifetime imaging microscopy (FLIM) to detect FRET in cells J Microsc 2004 215 1 12 15230870 10.1111/j.0022-2720.2004.01343.x
Maasch C Wagner S Lindschau C Alexander G Buchner K Gollasch M Luft FC Haller H Protein kinase calpha targeting is regulated by temporal and spatial changes in intracellular free calcium concentration [Ca(2+)](i) Faseb J 2000 14 1653 63 10929000 10.1096/fj.14.11.1653
Hu T Exton JH Protein Kinase Calpha Translocates to the Perinuclear Region to Activate Phospholipase D1 J Biol Chem 2004 279 35702 35708 15187091 10.1074/jbc.M402372200
Quest AF Leyton L Parraga M Caveolins, caveolae, and lipid rafts in cellular transport, signaling, and disease Biochem Cell Biol 2004 82 129 44 15052333 10.1139/o03-071
Almholt K Arkhammar PO Thastrup O Tullin S Simultaneous visualization of the translocation of protein kinase C alpha-green fluorescent protein hybrids and intracellular calcium concentrations Biochem J 1999 337 211 8 9882617 10.1042/0264-6021:3370211
Slater SJ Milano SK Stagliano BA Gergich KJ Curry JP Taddeo FJ Stubbs CD Interaction of protein kinase C with filamentous actin: isozyme specificity resulting from divergent phorbol ester and calcium dependencies Biochemistry 2000 39 271 80 10630986 10.1021/bi9916527
Slater SJ Seiz JL Stagliano BA Stubbs CD Interaction of protein kinase C isozymes with Rho GTPases Biochemistry 2001 40 4437 45 11284700 10.1021/bi001654n
Singh RD Puri V Valiyaveettil JT Marks DL Bittman R Pagano RE Selective caveolin-1-dependent endocytosis of glycosphingolipids Mol Biol Cell 2003 14 3254 65 12925761 10.1091/mbc.E02-12-0809
Connelly JP Botchway SW Kunz L Pattison D Parker AW MacRobert AJ Time-resolved fluorescence imaging of photosensitiser distributions in mammalian cells using a picosecond laser line- scanning microscope J Photochem Photobiol A 2001 142 169 175 10.1016/S1010-6030(01)00511-1
Botchway SW Parker AW Barba I Brindle K Development of a time-correlated single photon counting multiphoton laser scanning confocal microscope Central Laser Facility Annual Report 2000–2001 2001 RAL-TR-2001-030 170 171
| 15882451 | PMC1131896 | CC BY | 2021-01-04 16:29:46 | no | BMC Dermatol. 2005 May 9; 5:5 | latin-1 | BMC Dermatol | 2,005 | 10.1186/1471-5945-5-5 | oa_comm |
==== Front
BMC Fam PractBMC Family Practice1471-2296BioMed Central London 1471-2296-6-181586562410.1186/1471-2296-6-18Research ArticleGeneral Practitioners' opinions on their practice in mental health and their collaboration with mental health professionals Younes Nadia [email protected] Isabelle [email protected] Pierre [email protected] Marie-Pierre [email protected] Viviane [email protected] Bruno [email protected] Bayle Marie-Christine [email protected] Academic Unit of Psychiatry, Centre Hospitalier de Versailles, 177 Rue de Versailles 78157 Le Chesnay Cedex. France2 National Institute of Health and Medical Research (INSERM-U669), Hôpital Cochin, Paris, France3 Direction of Medical Policy, Assistance Publique – Hôpitaux de Paris, paris, France4 MGEN, Mental Health Foundation, Paris, France2005 2 5 2005 6 18 18 28 11 2004 2 5 2005 Copyright © 2005 Younes et al; licensee BioMed Central Ltd.2005Younes 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
Common mental health problems are mainly treated in primary care settings and collaboration with mental health services is needed. Prior to re-organisation of the mental health care offer in a geographical area, a study was organized: 1) to evaluate GPs' opinions on their day-to-day practice with Patients with Mental Health Problems (PMHP) and on relationships with Mental Health Professionals (MHPro); 2) to identify factors associated with perceived need for collaboration with MHPro and with actual collaboration.
Methods
All GPs in the South Yvelines area in France (n = 492) were informed of the implementation of a local mental health program. GPs interested in taking part (n = 180) were invited to complete a satisfaction questionnaire on their practice in the field of Mental Health and to include prospectively all PMHP consultants over an 8-day period (n = 1519). For each PMHP, data was collected on demographic and clinical profile, and on needs (met v. unmet) for collaboration with MHPro.
Results
A majority of GPs rated PMHP as requiring more care (83.4%), more time (92.3%), more frequent consultations (64.0%) and as being more difficult to refer (87.7%) than other patients. A minority of GPs had a satisfactory relationship with private psychiatrists (49.5%), public psychiatrists (35%) and social workers (27.8%). 53.9% had a less satisfactory relationship with MHPro than with other physicians.
Needs for collaboration with a MHPro were more often felt in caring for PMHP who were young, not in employment, with mental health problems lasting for more than one year, with a history of psychiatric hospitalization, and showing reluctance to talk of psychological problems and to consult a MHPro.
Needs for collaboration were more often met among PMHP with past psychiatric consultation or hospitalization and when the patient was not reluctant to consult a MHPro. Where needs were not met, GP would opt for the classic procedure of mental health referral for only 31.3% of their PMHP.
Conclusion
GPs need targeted collaboration with MHPro to support their management of PMHP, whom they are willing to care for without systematic referral to specialists as the major therapeutic option.
==== Body
Background
In developed countries, mental health problems, especially anxious and depressive disorders, are frequent and a leading cause of disability [1-4]. Since they are potentially remediable when adequately treated, they represent a major public health challenge [5,6]. A major obstacle to the instatement of adequate care is that when people do seek help, generally from their General Practitioner (GP), most of these problems are not recognized or not appropriately treated [4,7-9]. GPs have thus received special attention to improve mental health care because of their unique position [10]. Educational interventions have been proposed but have shown some limitations: temporary effect, no improvement in recognition of depression nor in patient recovery. They seem effective only when accompanied by organizational interventions [10-13]. Organizational interventions, based on the interaction between primary and secondary care, have been developed in several countries through local initiatives or national mental health reforms for improving depression care: in US [13,14], in UK [15], in Australia [16,17], in Canada [18,19]. They focus on the key role of GPs and on different forms of collaboration with mental health professionals (education, communication, on-site collaboration, collaborative care, stepped collaborative care, quality improvement, case management...). In France, collaboration is also encouraged by national government policies ("plan santé mentale" 2001 and 2005–2008). However collaboration of this sort requires pragmatic definition in clinical practice: for which patients with mental health problems (PMHP) do GPs need assistance from Mental Health Professionals (MHPro)? What sort of assistance? Why has this assistance not been organized up till now, i.e. what are the barriers to collaboration? Defining these issues is important before the development of quality improvement programs, considering some disappointing instances of collaboration between GPs and psychiatrists, where GPs have made limited use of opportunities for collaborative care with psychiatrists in spite of GP-reported perceived needs [20].
To design effective quality improvement programs based on targeted strategies among professionals and adapted to professional needs in the pilot area of South Yvelines, a survey was organized to gather information on some of these questions, with two objectives.
First, to evaluate satisfaction with mental health practice, exploring GPs' opinions on their patients with mental health problems (PMHP) and on relationships with Mental Health Professionals (MHPro).
Second, to measure factors associated with GPs' needs for collaboration with MHPro, with collaboration actually occurring, and with instances where needs are not met.
Methods
Population
All the 492 GPs of the area of "South Yvelines" (600 000 inhabitants) were approached by post in spring 2000 to recruit for the survey with a postage-paid reply envelope if they agreed to take part in this local area mental health program.
GPs were asked to include prospectively over an 8-day period all consulting patients over 15 years old "for whom a Mental Health Problem was the main current problem".
Data collected
GPs completed two questionnaires requiring approximately 30 minutes to complete:
First a questionnaire on their overall practice in the field of mental health, including data on their opinions on their PMHP compared to other patients, and on relationships with MHPro compared to other physicians.
A second questionnaire was completed for each PMHP included. Data was collected on demographic profile, clinical status, care provision and needs for collaboration with MHPro (met or unmet). To be feasible in daily practice, diagnoses were established using a classification developed by a working group of GPs and psychiatrists, secondarily translated into CIM-10 main diagnostic groups by 2 physicians independent from the study (a psychiatrist and a GP).
Statistical analysis
Descriptive and comparative analyses were performed with SAS 8.2. Three groups of PMHP were considered: no need for collaboration with MHPro expressed by the GP ("No need"); need for collaboration with MHPro but no actual collaboration ("Need Unmet") and need for collaboration with MHPro and actual collaboration ("Need Met"). Factors leading to a need for collaboration ("Need") and to actual collaboration ("Need met") were determined using two multivariate logistic regressions. The patients' demographic and clinical variables were entered into the regressions where chi-square tests (for categorical variables) and ANOVA tests (for continuous variables) produced a 5% level of significance. The "need" multivariate logistic regression, obtaining a non-significant result on the Hosmer and Lemeshow Goodness-of-Fit Test (p = 0.95), concerned 1007 patients. The "need met" multivariate logistic regression also producing a non-significant result on the Hosmer and Lemeshow Goodness-of-Fit Test (p = 0.87), concerned 532 patients.
Results
Characteristics of participating GPs and patients enrolled
One hundred and eighty GPs volunteered to participate to the mental health program (36.6% of local area GPs). They were predominantly male (69.4%), experienced providers (66.1% had been working for more than 10 years) and most were exclusively in private practice (79.4%). Compared to local area GPs, they were younger (p = 0.05) but did not differ for gender (Table 1).
Table 1 Characteristics of GPs responding to the survey (N = 180)
Respondents N = 182 Yvelines GPs N = 492 P
Male (%) 69.4 78.4 ns (0.3)
Age in years (%)
25 to 34 9.4 7.7
35 to 44 40.6 45.1
45 to 54 43.9 37.0 0.05
55 to 64 5.6 8.7
65 and over 0.6 1.6
Time in current practice (%)
< 5 years 15.0 na
5 to 10 18.9 na
> 10 66.1 na
Type of practice (%)
Private practice exclusively 79.4 na
Private and public practice 20.6 na
Mean working hours per week (sd) 51.5(15) na
na Data not available
The GPs enrolled 1519 MHP patients, representing 15.0 % of the overall number of consultations. Each participating GP saw 8 MHP patients on average (range 0–35). A majority of MHP patients were female (68.2%), mean age was 46.9 years (sd = 15.9). 61.4% had a current professional activity, 25.5% were living alone and 13.7% had a national disability allowance. The most frequent diagnoses were anxious and depressive disorders (33.7% and 31.3%). The disorders had lasted on average for 6.7 years (sd = 8.1). 18.3% of patients had a history of psychiatric hospitalization, 51% a history of care by psychiatrists. 71.6% had been managed by GPs for more than 2 years. Consultations lasted on average 23.2 minutes (sd = 8.9). According to the GPs, for 70.8% it was easy to talk about "psychological problems" but it was less easy to talk about a psychiatric consultation (proving easy for only 43.4%).
GPs' opinions on Patients with Mental Health Problems and on relationships with Mental Health Professionals
Four GPs out of five considered that patients with MHP have more expectations regarding care (83.4%), require more consultation time (92.3%) and are more difficult to refer to a specialist (87.7%) than other patients. A majority of GPs (64.2%) regretted having so many patients with MHP. 46.6% considered that PMHP expectations in terms of medical results are greater than among other patients. Few GPs complained about non-punctuality or unreliability of PMHP with regard to appointments (14.4%) (Table 2).
Table 2 GPs' opinion on their Patients with Mental Health Problems compared to their other patients (N = 182)
% Fully agree Rather agree Rather disagree Completely disagree No opinion
Have more expectations for care 30.0 53.4 10.5 1.1 5,0
Have more expectations for results 9.4 37.2 45.6 2.8 5,0
Expect more frequent consultations 22.2 42.8 26.1 3.3 5.6
Require more time 58.9 33.4 4.4 1.1 2.2
Are more difficult to refer 53.3 34.4 6.7 3.9 1.7
Are less punctual/reliable on appointments 7.7 6.7 27.8 51.7 6.1
While 78.4% of GPs were 'very' or 'mostly' satisfied with their relationships with other GPs, only 49.5% rated the same level of satisfaction for relationships with private psychiatrists, 35.0% for public psychiatrists and 27.8% for social workers. None of the GPs was 'very' satisfied with the information given by mental health professionals, and only 23.9% were mostly satisfied (Table 3).
Table 3 GPs' satisfaction of quality of exchanges with Health Professionals (N = 182)
% Very satisfied Fairly Satisfied Fairly Unsatisfied Very unsatisfied No opinion
Relationship with
...private psychiatrists 6.7 42.8 36.1 11.7 2.8
...public psychiatrists 3.5 31.5 21.8 8.1 35.1
...with social workers 1.1 26.7 37.8 10.6 23.8
...other primary physicians 17.8 60.6 16.7 1.7 3.3
...health professionals in general 9.5 73.9 14.4 1.7 0.6
Information received from mental health professionals in case of collaboration for a patient 0.0 23.9 42.8 28.9 4.4
Much better Better Same Worse Much worse
Relationship with mental health professionals in comparison with other health colleagues 0.6 4.5 41.0 40.4 13.5
Factors associated with GPs' needs for collaboration with Mental Health Professionals and with these needs being met
GPs felt a need for collaboration with a MHPro for 43.3% of their MHP patients. Within this group only 35.3% felt that their need was met (15.3% of the overall PMHP group).
Where needs were not met, for 64.1% of their patients GPs do not know what type of collaboration to seek, and for 31.3% they considered there was a need for care by MHPro, and for occasional advice for 4.6%. They would like to be able to refer mainly because they lack confidence with this type of care (48.3 %) but also because it requires too much time (17.8%). 70.5% cited a psychiatrist as the desired collaborator, and 22.7% a psychologist.
The need for collaboration with a MHPro (whether met or unmet) was more often felt by GPs for PMHP who were young (p < .0001), not in employment (p = .002), with mental health problems lasting for more than 1 year (p = .003) and past psychiatric hospitalization (p < .0001). GPs' needs for collaboration were more frequent when the patient was reluctant to consult a MHPro. Further to this, GPs seem to be rather more comfortable with patients suffering from anxiety than with other diagnoses. Finally, where a need for collaboration was felt, consultations were shorter (Table 4, ' [see Additional file 1]').
The need for collaboration was more often met in case of past psychiatric consultation (p = .0002) or hospitalization (p = .0004) and when the patient showed no reluctance to consult a MHPro (Table 4, ' [see Additional file 1]').
The more emphasis GPs put on collaboration, the more positive they evaluated their relationships with mental health professionals to be. 57.5% considered relationships with MHPro as less satisfactory than those with other health professionals when no need for collaboration was felt, 55.9% in case of unmet need and 48.1% when need was met (p = 0.004).
Discussion
Limitations
More than one third of GPs contacted volunteered for the local area mental health program and participated in the study. Results may reflect a particular population of GPs, younger than the average and probably already more involved in mental health care in their ordinary practice than non-respondents (who were however not contacted). It is likely that mental health actions targeting GPs can reach only a certain proportion. It has indeed been shown that the willingness to collaborate is greater among physicians under the age of 50 [21].
Caution is also required in the interpretation of this study on account of a second limitation. This resides in the fact that the results are based on GPs' reports on patients that they identified as PMHP. This use of assessment by the GPs could involve a recruitment bias, with a selection of the most severe patients. Indeed, external audits among general practice attendees have shown high unmet needs of mental health treatments but also PMHP as having less severe, less chronic and more readily treatable disorders [22-25]. The study option was to approach GPs' day to day practice with such patients and their subjective perceptions. The focus is on their attitudes towards patients they identify as PMHP and their attitudes towards the relevant specialist services, the aim being to adapt the mental health program to these particular attitudes.
GPs' opinions on their Patients with Mental Health Problems
In the study, PMHP identified by primary care respondents presented mainly anxious and depressive disorders. GPs have rather negative attitudes towards them. Previous papers have noted that complicated depressive symptoms are frequently encountered in primary care [26-28] and PMHP are time-consuming and require particular skills [29]. But as shown in our study, managing PMHP is a key part of a GP's job, and a part they are willing to take on if sufficient support and expertise are available.
GPs' collaboration with Mental Health Professionals
In the survey, GPs' needs for collaboration with MHPro have been reported to apply to half of their MHP patients. No publication on this point was found in the literature. Other studies conducted in ordinary practice have been focused on actual referrals from GPs to MHPro, or on actual utilization of mental health specialists, without reporting on GPs' perceived needs as is the case here [26,28,30,31]. Referral percentages have been estimated to be between 4 to 23% of primary care patients, and utilization of mental health specialists at 38% of depressed patients [26,30,31]. GPs' perceived needs for collaboration with MHPro are greater than needs for referral, probably because most patients are reluctant to consult a mental health professional [32].
This study sheds new light on factors related to GPs' collaboration with MHPro. According to a previous study on primary care patients with depressive symptoms, the best predictors of referral and utilization of mental health specialists were: more severe depressive symptoms, more long-standing problems (more than 1 year), prior visits to a mental health specialist, more years of education, being in the younger age groups, and being female [31]. The influence of the "psychiatric label" has been shown [33]. The present results on perceived needs for collaboration with MHPro may well apply to all mental pathologies encountered in primary care. Some of the above variables already reported to be related to perceived need (young age, prior mental health care) have been confirmed in the present study, and in addition this work has pinpointed the variable of not being in employment (which could correlate with disease severity). As has already been shown, GPs view patient-centered barriers as the most influential barrier to collaboration, more so than physician-centered barriers or system barriers [26,28-30,34,35]. But these patient-centered barriers could be associated with physician centered barriers, given GPs' dissatisfaction with relationships with mental health professionals. The dissatisfaction is greater than with other health professionals, and dissatisfaction is known to be associated with less frequent use of mental health services [30]. It is noteworthy that when needs are not met, only a third of GPs would opt for a referral to MHPro, suggesting that it is not a major therapeutic option for GPs. The classic pattern of referral to specialists as the major therapeutic option is often not relevant since it does not readily occur in day-to-day practice. The solution could be to develop other forms of collaboration between GPs and mental health professionals. Many MHP patients could be managed entirely by their GPs or treated in primary care if sufficient expertise is available (prompt psychiatric consultation, collaborative care) without actual referral [10,13,25]. To reinforce this notion, our results have shown that when there is actual collaboration, GPs' negative opinions on relationships with mental health professionals are less marked.
Conclusion
GPs are a key factor in the care of the commoner mental health problems. They are willing to care for this type of patient if they have more support for this job than they do at present. There is a need for collaboration, not in the form of the classic referral to specialists as the major therapeutic option, but in the form of emphasis on collaborative relationships with mental health specialists, to improve quality of the care provided in commoner mental health disorders[36]. Results from this survey have been integrated into the "South Yvelines Mental Health Network" created in June 2001, by promoting this type of collaborative relationships in the area. Further evaluations are underway.
List of abbreviations
GPs (General Practitioners). MHP (Mental health problems). PMHP (Patients with Mental Health Problems). MHPro (Mental Health Professionals).
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
Study concept and design : Gasquet, Kovess, Hardy-Bayle. Acquisition of data, study supervision : Chaillet. Analysis and interpretation: Younès, Gasquet. Drafting of the manuscript : Younès. Statistical expertise : Younès, Gaudebout, Falissard. Critical revision : Gasquet, Younès, Falissard, Kovess, Hardy-Bayle
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
Table 4 : Primary care patients' factors associated with Needs, Needs met and Needs unmet for collaboration with Mental Health Professionals. Univariate analysis and logistic regressions for demographic profile, clinical profile, modality of primary care, past psychiatric care, patient's attitude towards psychological problems.
Click here for file
Acknowledgements
All the professionals of the Réseau Santé Mentale Yvelines Sud and this survey was made possible through funding from local hospitals and by an unrestricted grant by Eli Lilly and Company, France.
==== Refs
Murray CJ Lopez AD Global mortality, disability, and the contribution of risk factors: Global Burden of Disease Study Lancet 1997 349 1436 1442 9164317 10.1016/S0140-6736(96)07495-8
Ustun TB Ayuso-Mateos JL Chatterji S Mathers C Murray CJ Global burden of depressive disorders in the year 2000 Br J Psychiatry 2004 184 386 392 15123501 10.1192/bjp.184.5.386
Alonso J Angermeyer MC Bernert S Bruffaerts R Brugha TS Bryson H Girolamo G Graaf R Demyttenaere K Gasquet I Haro JM Katz SJ Kessler RC Kovess V Lepine JP Ormel J Polidori G Russo LJ Vilagut G Almansa J Arbabzadeh-Bouchez S Autonell J Bernal M Buist-Bouwman MA Codony M Domingo-Salvany A Ferrer M Joo SS Martinez-Alonso M Matschinger H Mazzi F Morgan Z Morosini P Palacin C Romera B Taub N Vollebergh WA Prevalence of mental disorders in Europe: results from the European Study of the Epidemiology of Mental Disorders (ESEMeD) project Acta Psychiatr Scand Suppl 2004 21 27 15128384
Regier DA Narrow WE Rae DS Manderscheid RW Locke BZ Goodwin FK The de facto US mental and addictive disorders service system. Epidemiologic catchment area prospective 1-year prevalence rates of disorders and services Arch Gen Psychiatry 1993 50 85 94 8427558
Coulehan JL Schulberg HC Block MR Madonia MJ Rodriguez E Treating depressed primary care patients improves their physical, mental, and social functioning Arch Intern Med 1997 157 1113 1120 9164377 10.1001/archinte.157.10.1113
Ballenger JC Clinical guidelines for establishing remission in patients with depression and anxiety J Clin Psychiatry 1999 60 Suppl 22 29 34 10634353
Lepine JP Gastpar M Mendlewicz J Tylee A Depression in the community: the first pan-European study DEPRES (Depression Research in European Society) Int Clin Psychopharmacol 1997 12 19 29 9179630
Andrews G Henderson S Hall W Prevalence, comorbidity, disability and service utilisation. Overview of the Australian National Mental Health Survey Br J Psychiatry 2001 178 145 153 11157427 10.1192/bjp.178.2.145
Alonso J Angermeyer MC Bernert S Bruffaerts R Brugha TS Bryson H Girolamo G Graaf R Demyttenaere K Gasquet I Haro JM Katz SJ Kessler RC Kovess V Lepine JP Ormel J Polidori G Russo LJ Vilagut G Almansa J Arbabzadeh-Bouchez S Autonell J Bernal M Buist-Bouwman MA Codony M Domingo-Salvany A Ferrer M Joo SS Martinez-Alonso M Matschinger H Mazzi F Morgan Z Morosini P Palacin C Romera B Taub N Vollebergh WA Use of mental health services in Europe: results from the European Study of the Epidemiology of Mental Disorders (ESEMeD) project Acta Psychiatr Scand Suppl 2004 47 54 15128387
Gilbody S Whitty P Grimshaw J Thomas R Educational and organizational interventions to improve the management of depression in primary care: a systematic review Jama 2003 289 3145 3151 12813120 10.1001/jama.289.23.3145
Thompson C Kinmonth AL Stevens L Peveler RC Stevens A Ostler KJ Pickering RM Baker NG Henson A Preece J Cooper D Campbell MJ Effects of a clinical-practice guideline and practice-based education on detection and outcome of depression in primary care: Hampshire Depression Project randomised controlled trial Lancet 2000 355 185 191 10675118 10.1016/S0140-6736(99)03171-2
Katon W Von Korff M Lin E Simon G Walker E Bush T Ludman E Collaborative management to achieve depression treatment guidelines J Clin Psychiatry 1997 58 Suppl 1 20 23 9054905
Von Korff M Katon W Unutzer J Wells K Wagner EH Improving depression care: barriers, solutions, and research needs J Fam Pract 2001 50 E1 11401751
Rubenstein LV Jackson-Triche M Unutzer J Miranda J Minnium K Pearson ML Wells KB Evidence-based care for depression in managed primary care practices Health Aff (Millwood) 1999 18 89 105 10495595 10.1377/hlthaff.18.5.89
Meadows GN Establishing a collaborative service model for primary mental health care Med J Aust 1998 168 162 165 9507711
Hickie IB Davenport TA Naismith SL Scott EM SPHERE: a national depression project. SPHERE National Secretariat Med J Aust 2001 175 Suppl S4 5 11556435
Ellis PM Smith DA Treating depression: the beyondblue guidelines for treating depression in primary care. "Not so much what you do but that you keep doing it" Med J Aust 2002 176 Suppl S77 83 12065002
Kates N Craven MA Crustolo AM Nikolaou L Allen C Farrar S Sharing care: the psychiatrist in the family physician's office Can J Psychiatry 1997 42 960 965 9429067
Kates N Sharing mental health care. Training psychiatry residents to work with primary care physicians Psychosomatics 2000 41 53 57 10665268
Ungar TE Jarmain S Shared mental healthcare: a collaborative consultation relationship. The North York General Hospital experience Hosp Q 1999 3 34 40 11494604
Lucena RJ Lesage A Elie R Lamontagne Y Corbiere M Strategies of collaboration between general practitioners and psychiatrists: a survey of practitioners' opinions and characteristics Can J Psychiatry 2002 47 750 758 12420653
Hickie IB Davenport TA Naismith SL Scott EM Conclusions about the assessment and management of common mental disorders in Australian general practice. SPHERE National Secretariat Med J Aust 2001 175 Suppl S52 5 11556438
Hickie IB Davenport TA Scott EM Hadzi-Pavlovic D Naismith SL Koschera A Unmet need for recognition of common mental disorders in Australian general practice Med J Aust 2001 175 Suppl S18 24 11556431
Hickie IB Davenport TA Naismith SL Scott EM Hadzi-Pavlovic D Koschera A Treatment of common mental disorders in Australian general practice Med J Aust 2001 175 Suppl S25 30 11556432
Boardman J Henshaw C Willmott S Needs for mental health treatment among general practice attenders Br J Psychiatry 2004 185 318 327 15458992 10.1192/bjp.185.4.318
Orleans CT George LK Houpt JL Brodie HK How primary care physicians treat psychiatric disorders: a national survey of family practitioners Am J Psychiatry 1985 142 52 57 3966586
Klinkman MS Schwenk TL Coyne JC Depression in primary care--more like asthma than appendicitis: the Michigan Depression Project Can J Psychiatry 1997 42 966 973 9429068
Sorgaard KW Sandanger I Sorensen T Ingebrigtsen G Dalgard OS Mental disorders and referrals to mental health specialists by general practitioners Soc Psychiatry Psychiatr Epidemiol 1999 34 128 135 10327837 10.1007/s001270050123
Phongsavan P Ward JE Oldenburg BF Gordon JJ Mental health care practices and educational needs of general practitioners Med J Aust 1995 162 139 142 7854226
Williams JWJ Rost K Dietrich AJ Ciotti MC Zyzanski SJ Cornell J Primary care physicians' approach to depressive disorders. Effects of physician specialty and practice structure Arch Fam Med 1999 8 58 67 9932074 10.1001/archfami.8.1.58
Grembowski DE Martin D Patrick DL Diehr P Katon W Williams B Engelberg R Novak L Dickstein D Deyo R Goldberg HI Managed care, access to mental health specialists, and outcomes among primary care patients with depressive symptoms J Gen Intern Med 2002 17 258 269 11972722 10.1046/j.1525-1497.2002.10321.x
Jorm AF Mental health literacy. Public knowledge and beliefs about mental disorders Br J Psychiatry 2000 177 396 401 11059991 10.1192/bjp.177.5.396
Farmer AE Griffiths H Labelling and illness in primary care: comparing factors influencing general practitioners' and psychiatrists' decisions regarding patient referral to mental illness services Psychol Med 1992 22 717 723 1410095
Telford R Hutchinson A Jones R Rix S Howe A Obstacles to effective treatment of depression: a general practice perspective Fam Pract 2002 19 45 52 11818349 10.1093/fampra/19.1.45
Nutting PA Rost K Dickinson M Werner JJ Dickinson P Smith JL Gallovic B Barriers to initiating depression treatment in primary care practice J Gen Intern Med 2002 17 103 111 11841525 10.1046/j.1525-1497.2002.10128.x
Von Korff M Tiemens B Individualized stepped care of chronic illness West J Med 2000 172 133 137 10693379 10.1136/ewjm.172.2.133
| 15865624 | PMC1131897 | CC BY | 2021-01-04 16:29:13 | no | BMC Fam Pract. 2005 May 2; 6:18 | utf-8 | BMC Fam Pract | 2,005 | 10.1186/1471-2296-6-18 | oa_comm |
==== Front
BMC Fam PractBMC Family Practice1471-2296BioMed Central London 1471-2296-6-191586970410.1186/1471-2296-6-19Research ArticlePerceived barriers for treatment of chronic heart failure in general practice; are they affecting performance? Kasje Willeke N [email protected] Petra [email protected] Graeff Pieter A [email protected] Flora M [email protected] Department of Clinical Pharmacology, University Medical Center Groningen, A. Deusinglaan 1, 9713 AV Groningen, The Netherlands2 Department of Internal Medicine, University Medical Center Groningen, The Netherlands2005 3 5 2005 6 19 19 9 12 2004 3 5 2005 Copyright © 2005 Kasje et al; licensee BioMed Central Ltd.2005Kasje 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 aim of this study is to determine to what extent barriers perceived by general practitioners (GPs) for prescribing angiotensin-converting enzyme inhibitors (ACE-I) in chronic heart failure (CHF) patients are related to underuse and underdosing of these drugs in actual practice.
Methods
Barriers were assessed with a semi-structured questionnaire. Prescribing data were extracted from GPs' computerised medical records for a random sample of their CHF patients. Relations between barriers and prescribing behaviour were assessed by means of Spearman rank correlation and multivariate regression modelling.
Results
GPs prescribed ACE-I to 45% of their patients and had previously initiated such treatment in an additional 3.5%, in an average standardised dose of 13.5 mg. They perceived a median of four barriers in prescribing ACE-I or optimising ACE-I dose. Many GPs found it difficult to change treatment initiated by a cardiologist. Furthermore, initiating ACE-I in patients already using a diuretic or stable on their current medication was perceived as barrier. Titrating the ACE-I dose was seen as difficult by more than half of the GPs. No significant relationships could be found between the barriers perceived and actual ACE-I prescribing. Regarding ACE-I dosing, the few GPs who did not agree that the ACE-I should be as high as possible prescribed higher ACE-I doses.
Conclusion
Variation between GPs in prescribing ACE-I for CHF cannot be explained by differences in the barriers they perceive. Tailor-made interventions targeting only those doctors that perceive a specific barrier will therefore not be an efficient approach to improve quality of care.
==== Body
Background
Despite several landmark studies showing that appropriate treatment of chronic heart failure (CHF) can improve morbidity and mortality, management in general practice is still not optimal. Persisting major problems are underuse and underdosing of angiotensin-converting enzyme inhibitors (ACE-I) [1].
General practitioners (GPs) perceive problems that may explain why they do not treat their CHF patients optimally [2-5]. These problems can be classified as internal or external barriers. Internal barriers include lack of knowledge, e.g. not knowing the target dose of ACE-I or lack of awareness of new recommendations, as well as certain attitudes, such as lack of confidence, doubts about benefits for very old patients, fear for adverse effects or reluctance to change the treatment when a patient is stable. External barriers may be related to organisational factors, including difficulties at the primary-secondary care interface.
It is often suggested that intervention programs for improving performance need to be targeted at perceived barriers [6,7]. However, several tailored interventions addressing identified barriers did not change professional performance [8,9]. It might be that not all barriers identified are as relevant for not achieving optimal management. In the case of CHF management, lack of knowledge appears not to be very pertinent [10,11]. Furthermore, doctors may not be fully aware of the factors influencing their performance, because self-insight in treatment decisions is limited [12]. To our knowledge, no study has tried to assess the relationship between the GPs' self-reported problems with specific treatment recommendations and their actual prescribing for heart failure. Better understanding of this relation may help to indicate areas in which an intervention could be most beneficial.
The aim of our study is to determine to what extent barriers that GPs perceive for prescribing ACE-I in CHF patients are related to their actual prescribing.
Methods
Study population and setting
This study was part of the baseline of a larger study conducted from September 2001 to May 2002 in the north of the Netherlands, evaluating two audit programs for peer review groups focussing on the treatment of CHF and treatment of hypertension in diabetic patients. In the Netherlands, nearly all GPs participate in such peer review groups. Of the 27 peer review groups in our region, 21 participated in the larger study. A total of ten peer review groups consisting of 97 GPs were randomised to the chronic heart failure program, and therefore eligible for this study (Figure 1).
Figure 1 General practitioners (GPs) and patients in study
Prescribing data were extracted from the GPs' computerised medical records. In most practices, GPs have personal lists of patients. Data were collected for a random sample of 10 CHF patients per practice using computer generated random numbers. The estimated prevalence of CHF lies between 15–30 patients for an average patient list in the Netherlands of 2400 patients per GP [13]. All patients with a diagnostic code for CHF or the text 'heart failure', 'cardiac asthma', 'cardiac decompensation' or 'left ventricular dysfunction' in their medical records were selected from the medical records as possible CHF patients. GPs were asked to verify the diagnosis. Because of the larger study, CHF patients with a co-morbidity of diabetes type 2 were excluded.
Perceived barriers
GPs present at the audit meeting of their peer review group were asked to complete a semi-structured questionnaire on their perceived problems with the recommended treatment for CHF. This questionnaire was developed with statements of possible internal and external barriers towards prescribing ACE-I in patients with heart failure as identified from the literature [2-5]. The literature-based barriers included three general beliefs supporting the recommended treatment. Disagreement with these was considered an internal barrier. Six specific beliefs and attitudes opposing the general recommendations were presented, each also representing a possible internal barrier. Furthermore, two external barriers were included which were related to the sharing of responsibilities between primary and secondary care (Table 1). An open-ended question was added to identify any other barriers the GPs perceived with implementing evidence-based recommendations for CHF treatment. These self-reported barriers were categorised in nine themes by the first two authors. A content analysis based on an inductive approach was conducted. First, the two authors independently identified the main issues described by the GPs. This resulted in thirteen issues that after comparison and discussion were reduced to nine separate themes. Next, both researchers independently classified all reported barriers to one of the nine themes. Discrepancies were discussed until agreement was reached.
Table 1 Perceived internal and external barriers for prescribing ACE-I for CHF, divided in literature-based and self-reported barriers (N = number of GPs reporting barrier)
Literature-based barriers N Self-reported barriers N
Internal Do not agree with: I believe that the standard therapy for new CHF patients should be an ACE-I, irrespective of the severity of the disease 1
I believe that the standard therapy for known CHF patients should be an ACE-I, irrespective of the severity of the disease 2
I believe that ACE-I should be prescribed in as high a dose as possible for CHF patients 2
Agree with: I believe one should be reserved in prescribing ACE-I to CHF patients, because of the risk of renal insufficiency 11 Starting, checking, and titrating ACE-I dose is difficult 3
I believe one should be reserved in prescribing ACE-I to CHF patients, because of the risk of hypotension 12 Fears about adverse effects of ACE-I 8
I find initiating ACE-I difficult in CHF patients already using a diuretic 18
I find it difficult to frequently titrate the ACE-I dose in CHF patients 25
I believe that CHF patients who are stable on their current medication, should not be put on an ACE-I 18 Not wanting to change treatment when patients are stable 4
I believe it is not useful to prescribe ACE-I to very old CHF patients 10 Doubts about usefulness of ACE-I, especially in elderly patients 3
Difficulties with treating complex cases (comorbidity/polyfarmacy) 3
External Problems with patient compliance or motivation 5
I believe that a cardiologist should initiate ACE-I therapy in CHF patients 3 Problems in interacting with specialist care 9
I find it hard to change treatment initiated by a cardiologist 33
Time constraints 1
Difficulties with screening for undertreated heart failure patients 4
Actual treatment
Actual treatment data were extracted from computerised medical records by third and fourth year medical students trained to copy all relevant medical and prescription data on structured forms. When two GPs shared their list of patients, prescribing decisions were assigned to both GPs reflecting their joint prescribing policy. Data collected for the patients included age, gender, date of CHF onset, specialist referrals and current medication. For each GP, the percentage of patients currently or previously treated with an ACE-I was calculated as outcome variable, as well as the average dosing of the ACE-I currently prescribed. The ACE-I dosages were first converted to standardised dosages according to target daily doses for heart failure as recommended in the Dutch desk reference book [14]. This method of standardisation, which has been used before, uses 20 mg of enalapril as reference dose [15]. Based on the conversion, enalapril 20 mg equivalents are captopril 150 mg, ramipril 10 mg, quinapril 20 mg, lisinopril 20 mg, fosinopril 40 mg, and perindopril 4 mg. This conversion is an alternative for the more commonly used defined daily dosage (DDD) method, which can not be applied in this case since the DDDs for ACE-I are based on their use for hypertension.
Data analysis
Prescribing of ACE-I was aggregated at GP level to assess the relationship between the perceived barriers and overall prescribing behaviour. We checked at patient level with chi-square tests whether ACE-I prescribing differed significantly for different age-groups, gender and comorbidity of the patients. Differences between participating and non-participating GPs were tested with t-tests or chi-square tests.
The number of different barriers perceived was related to the percentage of patients currently or previously being prescribed an ACE-I and to the average standardised ACE-I dose prescribed with Spearman rank correlations (ρ). This non-parametric statistic was used since the number of barriers perceived did not have a normal distribution. Two sub-analyses were conducted to assess whether relations differed for internal versus external barriers, and for literature-based versus self-reported barriers. Next, we looked at the relationship between individual barriers and actual ACE-I prescribing. The data were first explored univariate by means of Mann-Whitney tests. A stepwise multivariate linear regression model was used to assess the relevance of all 11 literature-based and five self-reported barriers for explaining differences in ACE-I prescribing aggregated at GP level. The other self-reported barriers overlapped with literature based barriers and were excluded from this regression analysis. Finally, the complete data analysis was repeated for ACE-I prescribing for a subgroup of patients who had not been referred to a cardiologist in the year prior to data collection. This was decided when it became clear that a substantial number of patients had been referred in the last year, and one might expect that treatment initiated by the specialist confounds the analysis [16].
Results
Fifty-eight GPs completed the questionnaire, and prescribing behaviour was measured for 43 of them, resulting in an overall response rate of 44% (Figure 1). The GPs participating in this study were mainly male (88%), and on average 47 years old (SD 6.9), which was not significantly different from the non-responding GPs (83% male, 48 years). They had an average practice size of 2485 patients (SD 244) and 60% were single-handed. In comparison, the average Dutch practice size in 2001/2002 was 2430 patients, and 40% of the GPs was single-handed. In thirteen practices, less than ten verified CHF patients could be identified. In another three practices, two GPs shared the responsibility for the same patients. Prescribing behaviour was therefore assessed for 339 patients. In 11% of the cases, the diagnosis was confirmed by a recorded echocardiography. The GPs prescribed an ACE-I to an average of 44.9% (SD 15.9) of their CHF patients in an average standardised dose of 13.5 mg (SD 6.6). Another 3.5% of the patients had been using an ACE-I prior to the study period, including 7 patients that had stopped using ACE-I because of various side effects and 5 who had stopped without a documented reason. Including these patients, the GPs prescribed or had previously prescribed an ACE-I to 48.6% of their patients (SD 17.9). This was not significantly different from GPs who did not attend the meeting (47.3%). ACE-I prescribing was significantly lower for patients over 85 years of age (32.8%). ACE-I prescribing did not significantly differ according to the patients' gender or comorbidity. Angiotensin-II-antagonists were prescribed to 6% of the patients.
All but two GPs considered ACE-I as standard therapy for all CHF patients, and most GPs agreed that ACE-I should be dosed as high as possible (Table 1). The median number of barriers perceived in prescribing ACE-I or optimising ACE-I dosage was four. All 43 GPs perceived at least one barrier; 41 GPs perceived at least one internal barrier, and 37 GPs perceived at least one external barrier (Table 2).
Table 2 Number of perceived barriers and average ACE-I prescribing in CHF patients (N = number of GPs)
Number of barriers N % patients on ACE-I standardised ACE-I dose (mg)*
1 1 80.0 13.7
2 9 49.7 13.5
3 7 42.2 10.9
4 11 48.1 15.0
5 7 41.4 15.4
6 4 67.1 13.4
7 2 31.7 11.2
8 1 62.5 6.2
10 1 55.6 9.2
median = 4.0 (SD 1.86) 43 48.6 13.5
median internal barriers = 3.0 (SD 1.64) 41 48.5 13.5
median external barriers = 1.0 (SD 0.88) 37 47.1 12.9
* = based on a conversion using a standardised target dose of 20 mg
Relationship between number of barriers and ACE-I prescribing
No relationship appeared to exist between the number of barriers and the ACE-I prescribing at GP level (Table 2). No significant correlations were found between the total number of barriers perceived by the GPs and the percentage of patients receiving an ACE-I (ρ = .02) or the average ACE-I dose prescribed (ρ = -.08). Also, no significant correlations were found with the number of internal or external barriers, nor with the number of literature-based or self-reported barriers.
Literature-based barriers
With regard to initiating an ACE-I, a substantial number of GPs (42%) reported that they were afraid of endangering a stable situation and reluctant to start an ACE-I when a patient already received a diuretic (Table 1). The most important barrier regarding the ACE-I dosing was the difficulty perceived with titrating this dose. A majority of the GPs (77%) found it hard to change treatment initiated by a cardiologist. The univariate analysis showed no significant relationships between the individual barriers and ACE-I prescribing, and the scatter plots also revealed no patterns. Even GPs who believed it is not useful to prescribe ACE-I to very old CHF patients did not have less patients of 85 years or older on these drugs (univariate correlation ρ = .19, p = 0.3). In the stepwise linear regression model using forward selection, none of the 11 literature-based barriers was found to be related to the percentage of patients currently of previously receiving an ACE-I. The model as a whole did not significantly explain the prescribing differences at GP level (R-square = 0.17). Also, a model including only the five internal barriers directly related to ACE-I prescribing, i.e. fear of renal insufficience, fear of hypotension, difficulty in initiating ACE-I in patients on diuretics, not wanting to start ACE-I in stable patients, not wanting to prescribe ACE-I to very old patients, could not predict ACE-I prescribing (R-square = 0.10).
Surprisingly, the two GPs who did not agree that the ACE-I dose should be as high as possible, prescribed higher doses than GPs who did agree with this recommendation. This barrier was significantly associated in the multivariate linear regression model explaining differences in ACE-I dosages (beta 0.42, p = .03). The GPs who agreed that it is not useful to prescribe ACE-I to very old CHF patients were prescribing lower ACE-I doses (beta 0.34, p = .04).
Self-reported barriers
The most common self-reported barrier concerned the sharing of responsibilities with specialists (Table 1). According to several GPs co-management with a cardiologist made it difficult to change the therapy, and some GPs felt the cardiologist interfered too much. Furthermore, eight GPs mentioned possible adverse effects of ACE-I as a barrier towards prescribing. A few GPs mentioned problems with patient motivation as a barrier. Addition of the self-reported barriers to the multivariate model did not significantly alter any of the findings.
Analysis of patients not referred to a cardiologist
Cardiologist treatment could have confounded our analysis. Patients not referred to a cardiologist in the year prior to data collection were prescribed an ACE-I less often than the 36% of patients that had been referred (44% versus 61% on ACE-I, t-test = -2.2, p = .03). No significant difference was found regarding ACE-I dosage. Analysis including only the non-referred patients, however, hardly changed our findings. Again no relationship was found between the number of barriers and ACE-I prescribing. In the multivariate model, there were no barriers significantly explaining differences in the percentage of patients currently or previously receiving an ACE-I. One additional factor was found to be associated in the model explaining differences in ACE-I dosages. GPs who believed that CHF patients stable on their current medication should not be put on an ACE-I prescribed higher dosages of ACE-I (beta -0.48, p = .02).
Discussion
In this study we found remarkably few relationships between perceived barriers and actual prescribing for CHF. The problems that certain GPs acknowledged, such as their reluctance to initiate ACE-I in already treated CHF patients or the difficulties with gradually increasing the ACE-I dose, were not reflected in their prescribing of these drugs. No matter what barrier GPs report, it does not seem to affect their management of CHF patients in general practice.
For some patients, GPs tried to initiate an ACE-I but treatment had been stopped for various reasons. We included these cases in our analysis, thereby focussing on all attempts of a GP to start ACE-I treatment in CHF patients. A third of the patients in our study were seen by a cardiologist in the year prior to data collection, which was found to be related to receiving more ACE-I. However, subgroup analysis including only prescriptions for patients not recently referred to a cardiologist did not show any concealed relationships.
In our study we took the overall prescribing of ACE-I for CHF patients at GP level as primary outcome measure, expecting to find relationships between perceived barriers and the general prescription pattern. Since ACE-I should be started in all CHF patients, this aggregated measure is considered a relevant performance indicator for the CHF treatment [17]. At patient level, however, we did observe a lower prescription rate for patients over 85 years of age. Therefore, we decided to look at the specific association between the barrier for prescribing ACE-I to very old patients and actual ACE-I prescribing in this subgroup. Even on this specific level no significant relationship could be found. Our findings are in line with those from a recent explanatory study on effective management of type 2 diabetes, where no relationship was found between the presence of barriers perceived and the number of recommendations followed by physicians [18].
The representativity of our GP population should be considered. The 43 responding GPs were representative regarding age and gender for the total of 97 GPs enrolled in the larger study, and there were also no differences regarding their prescribing of ACE-I for CHF. In comparison to all Dutch GPs, our study population included a relatively large proportion of single-handed, male GPs that is typical for our region. A previous study showed, however, that such general physician and practice characteristics did not determine ACE-I prescribing for CHF [19]. Therefore, we do not expect that the regional selection limited the analysis of the relationship between perceived barriers and actual prescribing. Another matter of concern is the power of our study to detect relevant associations. We analysed all data on the GP level, since barriers were measured at this GP level and not linked to individual patients. Our sample size of 43 achieved an 80% power to detect moderate correlations of 0.41 in the univariate analyses. In the linear regression model, this sample size achieved an 80% power to detect an R-squared of 0.34 attributed to 11 independent variables or an R-squared of 0.26 attributed to 5 variables. This implies that there may have been weaker associations that we have missed both in the statistical analysis and by inspecting at the univariate scatter plots. We used medical records to measure the GPs' prescribing behaviour for CHF patients. As is the case with more than 90% of Dutch GPs, the GPs in this study prescribe electronically, and these prescriptions are automatically stored in the medical records. Patients were selected as having CHF according to their GP, without independently confirming the diagnosis. This was done since our study sought to relate perceived barriers with current management for those patients who the GP considered as having CHF. The low percentage of diagnoses confirmed by echocardiography reflects reality in Dutch primary care [11,20]. The average number of 33 CHF patients identified per practice was in accordance to prevalence rates found in other general practice registrations in The Netherlands [13]. In thirteen practices, less than 10 verified CHF patients could be identified. This was partly due to exclusion of patients with diabetes co morbidity, and partly due to a relatively young patient population in these practices.
Conclusion
Interventions to improve quality of care often focus on education and addressing perceived barriers for optimal performance. The findings from our study imply that targeting only those doctors that perceive a specific barrier with a tailor-made programme will not be an efficient approach. Variation in the quality of care between GPs can not be explained by differences in the barriers they perceive. It might even be true that being aware of a barrier stimulates some doctors to be more active in dealing with that barrier.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
WK carried out the data collection, analyzed the data and drafted the manuscript. PD participated in the design of the study, participated in the statistical analysis, assisted to draft and revise the manuscript. PG participated in drafting the manuscript. FH-R participated in the study design and helped to draft the manuscript. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Pont LG Sturkenboom MC Van Gilst WH Denig P Haaijer-Ruskamp FM Trends in prescribing for heart failure in Dutch primary care from 1996 to 2000 Pharmacoepidemiol Drug Saf 2003 12 327 334 12812013 10.1002/pds.809
Fuat A Hungin AP Murphy JJ Barriers to accurate diagnosis and effective management of heart failure in primary care: qualitative study BMJ 2003 326 196 12543836 10.1136/bmj.326.7382.196
Horne R Coombes I Davies G Hankins M Vincent R Barriers to optimum management of heart failure by general practitioners Br J Gen Pract 1999 49 353 357 10736884
Hickling JA Nazareth I Rogers S The barriers to effective management of heart failure in general practice Br J Gen Pract 2001 51 615 618 11510388
Khunti K Hearnshaw H Baker R Grimshaw G Heart failure in primary care: qualitative study of current management and perceived obstacles to evidence-based diagnosis and management by general practitioners Eur J Heart Fail 2002 4 771 777 12453549 10.1016/S1388-9842(02)00119-8
Grol R Implementing guidelines in general practice Quality Health Care 1992 1 184 191
Grimshaw JM Thomas RE MacLennan G Fraser C Ramsay CR Vale L Whitty P Eccles MP Matowe L Shirran L Wensing M Dijkstra R Donaldson C Effectiveness and efficiency of guideline dissemination and implementation strategies Health Technol Assess 2004 8 iii iv 1-72. 14960256
Baker R Reddish S Robertson N Hearnshaw H Jones B Randomised controlled trial of tailored strategies to implement guidelines for the management of patients with depression in general practice Br J Gen Pract 2001 51 737 741 11593835
Flottorp S Havelsrud K Oxman AD Process evaluation of a cluster randomized trial of tailored interventions to implement guidelines in primary care – why is it so hard to change practice? Fam Pract 2003 20 333 339 12738704 10.1093/fampra/cmg316
Hobbs FD Jones MI Allan TF Wilson S Tobias R European survey of primary care physician perceptions on heart failure diagnosis and management (Euro-HF) Eur Heart J 2000 21 1877 1887 11052861 10.1053/euhj.2000.2170
Cleland JGF Cohen-Solal A Aguilar JC Dietz R Eastaugh J Follath F Freemantle N Gavazzi A van Gilst WH Hobbs FD Korewicki J Madeira HC Preda I Swedberg K Widimsky J Improvement of Heart Failure Programme Committees and Investigators, Improvement programme in evaluation and management, Study Group on Diagnosis of the Working Group on Heart Failure of The European Society of Cardiology Management of heart failure in primary care (the IMPROVEMENT of Heart Failure Programme): an international survey Lancet 2002 360 1631 1639 12457785 10.1016/S0140-6736(02)11601-1
Harries C Evans JSBT Dennis I Measuring doctors' self-insight into their treatment decisions Appl Cognit Psychol 2000 14 455 477 10.1002/1099-0720(200009)14:5<455::AID-ACP667>3.0.CO;2-V
Gijsen R Poos MJJC Background and details of GP monitoring (Achtergronden en details bij cijfers uit huisartsenregistraties) Volksgezondheid Toekomst Verkenning, Nationaal Kompas Volksgezondheid 2003 Bilthoven: RIVM; mei Gezondheidstoestand\ Ziekten en aandoeningen\ Ziekten van het hartvaatstelsel\ Hartfalen
CVZ (National Health Insurance Board) Desk reference book 2003 (Farmocotherapeutisch Kompas) Utrecht 2003
Luzier AB Forrest A Adelman M Hawari FI Schentag JJ Izzo JL Jr Impact of angiotensin-converting enzyme inhibitor underdosing on rehospitalization rates in congestive heart failure Am J Cardiol 1998 82 465 469 9723634 10.1016/S0002-9149(98)00361-0
Rutten FH Grobbee DE Hoes AW Differences between general practitioners and cardiologists in diagnosis and management of heart failure: a survey in every-day practice Eur J Heart Failure 2003 5 337 344 10.1016/S1388-9842(03)00050-3
NHS Quality indicators 2004
Grant RW Hamrick HE Sullivan CM Dubey AK Chueh HC Cagliero E Meigs JB Impact of population management with direct physician feedback on care of patients with type 2 diabetes Diabetes Care 2003 26 2275 2280 12882848
Kasje WN Denig P Stewart RE Graeff PA Haaijer-Ruskamp FM Physician, organisational and patient characteristics explaining the use of angiotensin converting enzyme inhibitors in heart failure treatment: a multilevel study Eur J Clin Pharmacol 2005 Mar 11
Rutten FH Grobbee DE Hoes AW Differences between general practitioners and cardiologists in diagnosis and management of heart failure: a survey in every-day practice Eur J Heart Fail 2003 5 337 44 12798832 10.1016/S1388-9842(03)00050-3
| 15869704 | PMC1131898 | CC BY | 2021-01-04 16:29:13 | no | BMC Fam Pract. 2005 May 3; 6:19 | utf-8 | BMC Fam Pract | 2,005 | 10.1186/1471-2296-6-19 | oa_comm |
==== Front
BMC GenetBMC Genetics1471-2156BioMed Central London 1471-2156-6-181581999010.1186/1471-2156-6-18Methodology ArticlePower and sample size calculations in the presence of phenotype errors for case/control genetic association studies Edwards Brian J [email protected] Chad [email protected] Mark A [email protected] Stephen J [email protected] Derek [email protected] Laboratory of Statistical Genetics, Rockefeller University, New York, NY 10021, USA2 Department of Applied Math and Statistics, Stony Brook University, Stony Brook, NY 11794, USA2005 8 4 2005 6 18 18 12 10 2004 8 4 2005 Copyright © 2005 Edwards et al; licensee BioMed Central Ltd.2005Edwards 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
Phenotype error causes reduction in power to detect genetic association. We present a quantification of phenotype error, also known as diagnostic error, on power and sample size calculations for case-control genetic association studies between a marker locus and a disease phenotype. We consider the classic Pearson chi-square test for independence as our test of genetic association. To determine asymptotic power analytically, we compute the distribution's non-centrality parameter, which is a function of the case and control sample sizes, genotype frequencies, disease prevalence, and phenotype misclassification probabilities. We derive the non-centrality parameter in the presence of phenotype errors and equivalent formulas for misclassification cost (the percentage increase in minimum sample size needed to maintain constant asymptotic power at a fixed significance level for each percentage increase in a given misclassification parameter). We use a linear Taylor Series approximation for the cost of phenotype misclassification to determine lower bounds for the relative costs of misclassifying a true affected (respectively, unaffected) as a control (respectively, case). Power is verified by computer simulation.
Results
Our major findings are that: (i) the median absolute difference between analytic power with our method and simulation power was 0.001 and the absolute difference was no larger than 0.011; (ii) as the disease prevalence approaches 0, the cost of misclassifying a unaffected as a case becomes infinitely large while the cost of misclassifying an affected as a control approaches 0.
Conclusion
Our work enables researchers to specifically quantify power loss and minimum sample size requirements in the presence of phenotype errors, thereby allowing for more realistic study design. For most diseases of current interest, verifying that cases are correctly classified is of paramount importance.
==== Body
Background
One technique used in gene localization is the case-control genetic association study [1]. In this method, genotype and phenotype data are collected for case and control individuals [2]. Both genotype and phenotype data often contain misclassification errors [3,4], adversely affecting statistical tests used to locate disease genes [5-9]. Though phenotype misclassification has been widely studied in conjunction with disease (e.g. cancer, depression, heart disease), such studies have primarily focused on environmental association, not genetic association [10-13]. We are aware of only one recent publication considering phenotype misclassification for a test of genetic association [14].
Page et al. [3] emphasize the importance of studying phenotype errors in the context of genetic studies. They note that more than 1300 National Institutes of Health (NIH)-funded studies of complex genetic diseases have yielded fewer than 50 causative polymorphisms in humans [15,16]. More importantly, only 16%–30% of initially reported associations are confirmed without evidence of between-study heterogeneity or bias [15,17,18].
The problem of phenotype misclassification is particularly important, given the high error rates encountered in some studies. Lansbury [19] reports that postmortem pathological studies estimate that greater than 15% of Alzheimer's Disease and Parkinson's Disease cases are misdiagnosed in the clinic. Duffy et al. [12] report that in a breast cancer study conducted by Press et al. [20], nearly half (34 out of 69) of the individuals containing over expression of the immunohistochemical marker c-erbB-2 were misclassified. Burd et al. [21] found that 5%–12% of individuals previously diagnosed with Tourette syndrome were misdiagnosed. They further note that in their three-step model for linkage analysis, a 5% misclassification rate in the first step leads to a 20% error rate by the third step.
In the presence of random errors that are non-differential with respect to trait status (case or control), the type I error rate is constant [5]. That is, there is no change in significance of the classic chi-square test of independence on 2 × n contingency tables (the statistic of interest in this work). Here and elsewhere, n is the number of observed genotypes at the marker locus. However, there is a reduction in the power of the chi-square test and an increase in the minimum sample size needed to maintain constant asymptotic power at a fixed significance level [5,22,23]. A key issue that arises then is a quantification of power loss in the presence of phenotype errors.
Formulas allowing researchers to perform realistic power and sample size calculations in the presence of errors benefit researchers in the design of case-control studies by saving them the cost of excessive genotyping and phenotyping due to underpowered initial conditions. Mote and Anderson [22] computed power in the presence of what we call genotype error (although in a more general statistical setting) and proved that the power of the chi-square test of independence on r × c contingency tables (r = number of rows; c = number of columns) is always less than or equal to the power of the test when data are perfectly classified. Carroll et al. [24] developed methods for estimating the parameters of a prospective logistic model given a binary response variable and arbitrary covariates with case/control data when the covariates have measurement error. Gordon et al. [6,7] developed formulas for power and sample size calculations for the specific situation of genotype error. They used Mitra's equation for the non-centrality parameter [6,7,25] to compute the power and minimum sample size both for data with and without genotype errors. Gordon et al. [6,7] showed that a one percent increase in the sum of genotypic error rates typically results in a two to eight percent increase in the minimum sample size for the parameters and error models considered and that the increase in minimum sample size is larger when the allele frequencies are more extreme [7]. Kang et al. [8] extended this work by determining a linear approximation for the sample size increase needed to maintain constant asymptotic power at a fixed significance level. Kang et al. [8] found that (i) the cost of genotype misclassifications is a function of the true genotype frequencies and the ratio of controls to cases; (ii) in general, misclassifying a more common genotype as a less common genotype is more costly than the reverse error; and (iii) certain types of misclassification have costs that approach infinity as the minor SNP allele frequency approaches 0.
Our goal in this research is therefore two-fold: (i) to quantify power and sample size for the chi-square test of independence on 2 × n contingency tables in the presence of phenotype errors; and (ii) to quantify the cost of each type of phenotype error.
We present formulas to facilitate accurate power and sample size calculations in the presence of phenotype errors. We perform a genotypic test of association using the Pearson chi-square test statistic on 2 × n contingency tables. The degrees of freedom (in our case n-1) and the non-centrality parameter completely describe the power of the chi-square test. We express the non-centrality parameter in terms of the case and control sample sizes, genotype frequencies, and phenotype error model parameters. Rearranging the equation for the non-centrality parameter gives an equation for the minimum sample size. Additionally, this work extends Kang et al.'s [8] findings to the cost of phenotype errors.
Results
As noted in the Methods section (Distinguishing case from affected and control from unaffected), we use the term case to refer to an individual who has been diagnosed as being affected with a given disease, whether or not that individual is truly affected. Similarly, we use the term control to refer to an individual who has been diagnosed as being unaffected with a given disease, whether or not that individual is truly unaffected. We use the term affected (respectively, unaffected) to refer to an individual who is truly affected (respectively, unaffected) with the disease of interest.
All notation in the Results section is defined in the Methods section (Notation).
Design of simulation program – null and power calculations for a fixed sample size
We performed power simulations for di-allelic and tetra-allelic loci using the parameter specifications (Table 1) in the Methods section (Design of the simulation program). For the null situation, we computed the proportion of replicates for a given set of parameter specifications whose chi-square statistic exceeded the cutoff determined assuming the appropriate asymptotic null distribution (central chi-square distribution with either 2 or 9 df for di-allelic and tetra-allelic simulations, respectively). We call this proportion the empirical significance level for a given setting (either 5% or 1%). The median (respectively, maximum) absolute difference observed over all parameter specifications in table 1 (di-allelic and tetra-allelic) was 0.0005 (respectively, 0.002; full results not shown). That means, the empirical significance level was always within 0.002 of the significance level assuming the appropriate asymptotic null distribution. These results confirm Bross's findings [5] that non-differential phenotype misclassification does not affect the size of the chi-square test of independence.
Table 1 Parameter settings for null and power simulations with di-allelic and tetra-allelic loci
Low High
True case and control genotype frequencies p = 0.05 p = 0.15
Pr(affected misclassified as a control) (θ) 0.05 0.15
Pr(unaffected misclassified as a case) (φ) 0.05 0.15
Disease prevalence (K) 0.005 0.05
Number of cases () 500 1000
Number of controls () 500 1000
Significance level 5% 1%
Genotype frequency parameter for tetra-allelic loci (power simulations)
d 1 2
This table presents the low and high parameter settings we consider for null and power simulation calculations for di-allelic and tetra-allelic loci. As per the 27 factorial design, null and power simulations are performed on 128 distinct sets of parameter settings. Each simulation uses 100,000 iterations to determine empirical significance level (null) or simulation power. For di-allelic loci, case and control genotype frequencies are determined by the parameter p (see Methods – design of simulation program – power calculations for a fixed sample size). For tetra-allelic loci, genotype frequencies are determined by the parameter d (see Methods – Design of simulation program – power calculations for a fixed sample size).
For the power simulations, we compared the asymptotic power with the simulation power using absolute difference. That is, the absolute difference in power, defined as |simulation power - asymptotic power|, was calculated for each simulation. In table 2, we report the minimum, 10th percentile, 25th percentile, median, 75th percentile, 90th percentile, and maximum differences at the 5% and 1% significance levels. There were 27 = 128 data points for each simulation. For the majority of simulations, the absolute difference was very small. For both di-allelic loci and tetra-allelic loci at both significance levels, the median absolute difference was 0.001. For di-allelic loci, the maximum absolute difference observed was 0.012 (at the 1% significance level) while for the tetra-allelic loci, the maximum absolute difference was 0.011 (also at the 1% significance level).
Table 2 Percentiles for absolute difference between asymptotic power and simulation power
5% significance level 1% significance level
Di-allelic locus
Minimum 0.0000 0.0000
10th percentile 0.0002 0.0002
25th percentile 0.0005 0.0004
50th percentile 0.0010 0.0011
75th percentile 0.0028 0.0026
90th percentile 0.0065 0.0057
Maximum 0.0099 0.0119
Tetra-allelic locus
Minimum 0.0000 0.0000
10th percentile 0.0000 0.0000
25th percentile 0.0007 0.0008
50th percentile 0.0012 0.0014
75th percentile 0.0028 0.0032
90th percentile 0.0072 0.0081
Maximum 0.0102 0.0111
Power simulations are performed at 100,000 iterations for each set of parameter specifications in the Methods section. Here we report various percentiles of the absolute difference |simulation power - asymptotic power| for our simulations. For each locus type (di-allelic, tetra-allelic), percentiles are computed using 27 = 128 settings documented in table 1.
Although the asymptotic power is a good enough approximation to the simulation power so that it can be used for design purposes, this difference is somewhat larger than would be expected in the event that the simulated power followed a binomial variation with probability equal to the asymptotic power (based on computation of 95% confidence intervals – results not shown). We discuss this issue below (see Discussion).
Cost functions
Using the mathematics presented in the Methods section (Cost functions), we compute the following formulas:
In table 3, we present the values of these cost coefficients for the parameters considered in table 1. One finding becomes immediately clear. It is that the cost of misclassifying an unaffected as a case is much larger than the cost of misclassifying an affected as a control. For example, for a disease prevalence K = 0.05, the minimum cost coefficient Cφ regarding misclassification of an unaffected as a case is approximately 40, occurring when R* = 2 and p = 0.15. The maximum cost coefficient Cθ for the same prevalence is 0.10, occurring for the same values of R* and p.
Table 3 Cost coefficients for different types of misclassification
K R* p Cθ Cφ
0.005 0.5 0.05 0.01 540.29
0.15 0.01 458.99
1 0.05 0.01 478.32
0.15 0.01 432.67
2 0.05 0.01 440.18
0.15 0.01 415.60
0.05 0.5 0.05 0.09 51.59
0.15 0.10 43.82
1 0.05 0.08 45.67
0.15 0.10 41.31
2 0.05 0.08 42.03
0.15 0.10 39.68
The column heading for this table are as follows: K = prevalence; R* = ratio of controls to cases; p = SNP minor allele frequency in affected population; Cθ = Cost coefficient corresponding to misclassification parameter θ – this is a lower bound of the percent increase in sample size necessary to maintain constant asymptotic power for every 1% increase in θ Cφ = Cost coefficient corresponding to misclassification parameter φ – this is a lower bound of the percent increase in sample size necessary to maintain constant asymptotic power for every 1% increase in φ. The cost coefficients are computed using equation (1).
When the prevalence K = 0.005, the cost coefficient Cφ becomes larger by an order of magnitude. The minimum value of Cφ is 415, occurring as above when R* = 2 and p = 0.15. That means that a 1% increase in the value of φ requires at least a 415% increase in cases and controls to maintain the same power at any significance level.
A second finding that becomes clear from studying equation (1) is that the cost coefficient Cφ has an infinite limit as the prevalence K approaches 0 (for any set of fixed values of the other parameters), while the cost coefficient Cθ has a limit of 0. This results comes from the observation that the dominating terms for the cost coefficients Cφ and Cθ in equation (1) are (1 - K)/K and K/(1 - K), respectively.
It should be noted that the linear Taylor approximation is not very accurate for even small values of φ. The linear Taylor approximation is useful, though, in that it serves as a lower bound for the percentage sample size increase. That is, percent increase in sample size is at least Cφ for any value of φ. We illustrate this point in the next section.
Minimum sample size requirements in presence of phenotype misclassification – Alzheimer's disease ApoE example
Figure 1 presents a contour plot of the minimum sample size necessary to maintain a constant power of 95% at the 5% significance level using the parameter values taken from the methods section (see Methods – Minimum sample size requirements in presence of phenotype misclassification – Alzheimer's disease ApoE example). Each approximately horizontal line represents a constant minimum number of cases (as a function of the misclassification parameters φ and θ). For two consecutive horizontal lines, the values in between those lines (represented by different colors) have sample sizes that are between the sample sizes indicated by the two horizontal lines. For example, consider the consecutive, approximately horizontal lines labeled 3394.9 and 4365.9 (third and fourth lines up, respectively, in figure 1). All values of θ and φ whose Cartesian coordinate(θ, φ) lies between these two lines have a corresponding minimum sample size between 3395 and 4365. An example of such a pair is the coordinate (0.00,0.075). Note that the minimum sample size of 484 occurs when φ = θ = 0 and the maximum sample size of 10,187 occurs when φ = θ = 0.15.
Figure 1 Contour plot of minimum number of cases needed to maintain constant asymptotic power of 95% at a 5% significance level in the presence of phenotype misclassification for Alzheimer's disease ApoE example. We compute the increase in minimum cases () needed to maintain constant 95% asymptotic power at the 5% significance level (using a central χ2 distribution with 5 degrees of freedom) in the presence of errors. Sample sizes are computed using equation (3). The affected and unaffected genotype frequencies are taken from a previous publication [9, 14]. In that work, the marker locus considered was ApoE and the disease phenotype was Alzheimer's disease. We use the LRTae estimates from table 5 of that work [9]. Six genotypes are observed in most populations. The frequencies we use to perform the sample size calculations in figure 1 are presented in the Methods section (Minimum sample size requirements in presence of phenotype misclassification – Alzheimer's Disease ApoE example). We assume that equal numbers of cases and controls are collected. Also, we specify a prevalence K = 0.02, which is consistent with recent published reports for Alzheimer's Disease in the U. S. [32]. Sample sizes are calculated for each misclassification parameter θ, φ ranging from 0.0 to 0.15 in increments of 0.01. The number of cases ranges from 484 when θ = φ = 0 to 10,187 when θ = φ = 0.15. In this figure, each (approximately) horizontal line represents a constant sample size as a function of the misclassification parameters θ and φ. For two consecutive horizontal lines, the values in between those lines (represented by different colors) have sample sizes that are between the sample sizes indicated by the two horizontal lines.
Our results for the cost functions are consistent with the findings here. For values of φ less than 0.02, sample size increase appears to be constant in the parameterθ. That is, misclassification of an affected as a control does not affect the sample size estimates at all. However, even a 1% misclassification of an unaffected as a case requires a sample size increase from 486 to 921 (φ = 0.01, θ = 0.0 in figure 1; exact results not shown) to maintain constant power, an approximately 90% increase. As the probability of misclassifying an unaffected as a case φ increases, there appears to be an interaction between the two misclassification parameters, requiring even larger sample size increases than would be expected if the sample size increase were linear in each misclassification parameter (figure 1).
Comparison of power loss for fixed sample size when only one misclassification parameter is non-zero
Another way of interpreting cost is by considering the power loss for fixed sample size. We demonstrate this point in figure 2. In that figure, we present the power in the presence of phenotype misclassification when either the θ or φ parameter is set to 0 and the other parameter ranges from 0 to 0.15 in increments of 0.01. Power is calculated at the 1% significance level assuming 250 cases and 250 controls, a SNP locus with case minor allele frequency 0.05, control minor allele frequency 0.15 (Hardy Weinberg equilibrium in both populations), and two settings of disease prevalence (K = 0.05, 0.01). Power is determined through calculation of the non-centrality parameter (equation (2)).
Figure 2 Power to detect association for two different settings of prevalence when only one phenotype misclassification parameter is non-zero. In this figure, the horizontal axis refers to the misclassification probability for one parameter when the second parameter is 0. For example, the graphs labeled "φ = 0" provide power calculations at two settings of disease prevalence (K = 0.05, K = 0.01) as a function of θ values ranging from 0.0 to 0.15 on the horizontal axis. Similarly, the graphs labeled "θ = 0" provide power calculations at two settings of disease prevalence (K = 0.05, K = 0.01) as a function of φ ranging from 0.0 to 0.15 on the horizontal axis.
The results of figure 2 further illustrate the importance of distinguishing between the two types of misclassification. When the φ parameter is 0, the asymptotic power is virtually independent of the value of the φ parameter and the disease prevalence K. Power values for all settings of φ and K are approximately 99%. When the θ parameter is 0, the asymptotic power reduces to 91% when φ = 0.01, K = 0.05 and to 33% when φ = 0.01, K = 0.01. When φ = 0.02, power reduces to 76% when K = 0.05 and to 11% when K = 0.01. These examples further document the dominating effect that disease prevalence has on power and/or sample size requirements in the presence of phenotype misclassification error.
Discussion
As we noted above (Results – Design of simulation program – power calculations for a fixed sample size), the asymptotic power is a good enough approximation to the simulation power so that it can be used for design purposes. However, the difference is somewhat larger than would be expected in the event that the simulated power followed a binomial variation with probability equal to the asymptotic power. One possible explanation may be that our simulation studies were "under-powered" so that the asymptotic theory did not hold. Indeed, the median power value at the 5% significance level for our simulation studies (table 1) was 13% (full results not shown). Given such low overall power levels and also the fact that, for the SNP minor allele frequency of 0.05, Cochran's condition of a minimal expected cell count of 5 is not achieved [26], it is conceivable that effective sample sizes are not sufficient for power values based on asymptotic theory to hold. Other authors studying misclassification error have also observed this phenomenon [27].
While we have considered a genetic model-free framework here, we note that our work easily extends to a genetic model-based framework as well [6,7]. We will implement calculations using a genetic model-based framework in our web tool (next paragraph).
Given the accuracy of our method (absolute errors no larger than 0.012, based on simulations), we conclude that researchers may use our method to accurately determine power and sample size calculations for case/control genetic association studies in the presence of phenotype misclassification. We have developed a web tool that performs these calculations online. The URL for this tool is: .
Conclusion
In this work, we developed a method for performing realistic power and sample size calculations in the presence of phenotype errors. Simulation results suggest that our formulas (equations (2) and (3)) may be used to design case/control genetic association studies incorporating phenotype misclassification. We confirmed that phenotype misclassification always reduces the power of the chi-square test of association (as was first shown by Bross [5]), and consequently, increases the minimum sample size needed to maintain constant asymptotic power.
Our cost calculations reveal two significant findings. The first is that power and/or sample size is most significantly altered by a change in disease prevalence. Specifically, the cost coefficient for misclassifying an affected as a control is of the order of magnitude K/(1 - K) and the cost coefficient for misclassifying an unaffected as a case is of the order of magnitude (1 - K)/K, where K is the disease prevalence (equation (1)). This finding suggests that, for many diseases of current interest, where prevalence is usually less than or equal to 0.10, it is much more important to insure that cases are truly cases rather than controls being truly controls. Zheng and Tian [14] made this same observation (without the explicit computation of cost coefficients) for the linear test of trend applied to cases and controls genotyped at a SNP marker.
Methods
Distinguishing case from affected and control from unaffected
Throughout this work, we use the term case to refer to an individual who has been diagnosed as being affected with a given disease, whether or not that individual is truly affected. Similarly, we use the term control to refer to an individual who has been diagnosed as being unaffected with a given disease, whether or not that individual is truly unaffected. We use the term affected (respectively, unaffected) to refer to an individual who is truly affected (respectively, unaffected) with the disease of interest. A key assumption we make through the paper is that we collect only cases and controls for our test of genetic association.
Notation
We use the following notation:
Count parameters
a = Number of alleles at the marker locus. The number of genotypes at the marker locus is always a(a + 1)/2 = n.
= Number of cases; this quantity is a fixed parameter in our design.
= Number of controls; this quantity is a fixed parameter in our design.
= Ratio of controls to cases.
Probability parameters
K = Prevalence of disease.
p0j = Frequency of genotype j at the marker locus for the affected group, 1 ≤ j ≤ a(a+1)/2.
p1j = Frequency of genotype j at the marker locus for the unaffected group, 1 ≤ j ≤ a(a+1)/2.
= Frequency of genotype j at the marker locus for the case group, 1 ≤ j ≤ a(a+1)/2.
= Frequency of genotype j at the marker locus for the control group, 1 ≤ j ≤ a(a+1)/2.
Error model parameters
θ = Pr (affected individual classified as control) = 1 - Se, where Se is the sensitivity of the phenotype measurement instrument.
φ = Pr (unaffected individual classified as case) = 1 - Sp, where Sp is the specificity of the phenotype measurement instrument. This notation was used by Bross [5].
A key assumption we make here is that these errors are random and independent. Furthermore, they are non-differential with respect to a particular genotype [14].
Cost parameters
Cθ = Cost of misclassifying an affected individual as a control. This value is the percent increase in minimum sample size necessary to maintain constant power for every one percent increase in the value of θ.
Cφ = Cost of misclassifying an unaffected individual as a case. This value is the percent increase in minimum sample size necessary to maintain constant power for every one percent increase in the value of φ.
Expressing case and control genotype frequencies in terms of affected and unaffected genotype frequencies
We comment that the case and control genotype frequencies, ,, may be written in terms of the affected and unaffected genotype frequencies, p0j, p1j, the disease prevalence K, and the misclassification error probabilities, θ and φ. Using the law of total of probability, we have:
= [p0j (1 - θ) K + p1jφ(1 - K)]/[(1 - θ) K + φ(1 - K)], 1 ≤ j ≤ a(a + 1)/2
= [p0jθK + p1j(1 - φ)(1 - K)]/[θK + (1 - φ)(1 - K)]. 1 ≤ j ≤ a(a + 1)/2
For a derivation, see the Appendix.
It is interesting to note that determination of case and control genotype frequencies in the presence of only phenotype error differs from determination of the same frequencies in the presence of only genotype error in that one needs to specify disease prevalence for phenotype error (in addition to specifying the respective misclassification probabilities for phenotype and genotype) [7,14].
Test statistic for genotypic association
The test statistic considered in this work is Pearson's chi-square statistic on 2 × n contingency tables. Here, the two rows refer to the two possible classifications (case or control) and the n columns correspond to the n different genotypes, where n = a(a + 1)/2. Using this statistic on 2 × n contingency tables, we test for association between genotype and disease status. We selected the genotypic test of association because the null distribution of the allelic test of association cannot be determined when either the case or control group genotype frequencies deviate from Hardy-Weinberg Equilibrium (HWE) [28,29]. Let Grc equal the observed count of the cth genotype in the rth group, where 1 ≤ c ≤ n and r = 0 for the case population and r = 1 for the control population. Then, the chi-square statistic is given by the formula .
In this expression, the expected cell count of the cth genotype in the rth group, Erc, is determined by the equation Erc = SrDc/N, where is the row total for the rth group, is the column total for the cth genotype, and is the total sample size.
Under the null hypothesis of no association between the marker locus and the disease (p0j = p1j for all j), the statistic X2 is asymptotically distributed as a central χ2 with n - 1 degrees of freedom. We verify this statement in our simulations (see Results).
Asymptotic power calculations
In this section, we describe our method for computing asymptotic power in the presence of errors. The asymptotic power is summarized by a non-centrality parameter λ, which is a function of the case and control sample sizes and the respective genotype frequencies.
The asymptotic power is , where β is the probability of a type II error (accepting a false null hypothesis) and is the cumulative distribution function (CDF) for the non-central χ2 distribution with n-1 degrees of freedom evaluated at the α percentile of the null distribution, which is a central χ2 distribution with n - 1 degrees of freedom.
Asymptotic non-centrality parameter
Mitra [25] derived the asymptotic power function for the chi-square test for unmatched cases and controls. Under the alternative hypothesis, the distribution is a non-central χ2 with n -1 degrees of freedom and non-centrality parameter λ*. Mitra [25] showed that for perfectly classified data (i.e., θ = φ = 0)), the non-centrality parameter is given by
where the sample sizes and are fixed by design and the genotype frequencies and are equal to p0j and p1j respectively, for each j. In the presence of phenotype errors, the genotype frequencies and are biased away from their true values, as indicated by formula (1). We verify the accuracy of the non-centrality parameter formula (2) using simulations (see Methods – Design of simulation program – null and power calculations for a fixed sample size).
Increase in minimum sample size
We determine the minimum sample size needed to maintain constant power at a fixed significance level in the presence of phenotype errors. The minimum sample size for cases can be found by rearranging equation (2) and substituting . We obtain
Design of simulation program – null and power calculations for a fixed sample size
We perform simulations using 100,000 iterations to verify (i) the nominal significance levels under the null hypothesis; and (ii) the asymptotic power calculations provided by equation (2). We use a 27 factorial design [30] in which we set lower and upper bounds for each set of parameters. In the simulations, we consider both di-allelic and tetra-allelic loci. For each simulation, both the affected and unaffected genotype frequencies are in HWE. For the power simulations using di-allelic loci, the genotype frequencies are specified as follows using a parameter p: for the affected group, p01 = (1 - p)2, p02 = 2p(1 - p), p03 = p2, and for the unaffected group, p11 = (1 - p - 0.1)2, p12 = 2(p + 0.1)(1 - p - 0.1), p13 = (p + 0.1)2. That is, the SNP minor allele frequency in the unaffected population is equal to the sum of the SNP minor allele frequency in the affected population (p) and 0.1. For the null simulations, both the affected and unaffected groups have genotype frequencies as specified above for p0j, j ∈ {1,2,3}. Our parameter settings for the factorial design are shown in table 1.
For the tetra-allelic loci, the parameter settings are the same as for the di-allelic loci with the exception of the affected and unaffected genotype frequencies. For the tetra-allelic loci, we let p = 0.25 and specify the genotype frequencies for power simulations as follows using a parameter d. For the affected population, the probability of a homozygous genotype is p2+d(0.03) and the probability of a heterozygous genotype is 2p2 - d(0.02), where d = 1,2. For the control group, the probability of a homozygous genotype is 0.0625 and the probability of a heterozygous genotype is 0.125. For null simulations, we set d = 0.
Here, we briefly describe the algorithm used to simulate our phenotype and genotype data for each replicate of a particular simulation. Note that a simulation is completely described by the each of the 7 parameter settings provided in table 1. For each individual in each replicate, we first randomly assign the individual an affection status (affected or unaffected) using the disease prevalence K. We then randomly assign the individual a genotype conditional on the affection status using the conditional probabilities p0j and p1j. Once affection status and genotype are determined, we then randomly assign case or control status using the individual's affection status and the phenotype misclassification probabilities. Within each replicate, we repeat this procedure until we have the specified number of cases and controls. Because of the low prevalence, we invariably reach our required number of controls much more quickly than we reach our required number of cases. In such situations, we simply ignore all assigned control individuals after reaching our required number, and keep collecting cases until we achieve that required number.
Cost functions
We demonstrate how to compute the sample size cost coefficient of phenotype misclassification to gain insight into which type of misclassification requires the greater increase in sample size for fixed power. Let λ equal the non-centrality parameter when there is no phenotype misclassification and let λ* equal the non-centrality parameter in the presence of phenotype errors. To find the sample size adjustment needed to maintain constant power, we set λ = λ*. We considered this condition previously when studying the cost of genotype error [8]. Let and . Then the condition λ = λ* may be rewritten as or . Though the cost of misclassification for cases is mathematically defined as the ratio /NA, we instead consider the reciprocal ratio NA/ because the latter allows for more straightforward computation. We approximate NA/ using a first-order Taylor Series expansion centered at (θ, φ) = (0,0). We obtain . Here, (∂/∂θ)[f]|(0,0) is the partial differential operator (with respect to θ) acting on the function f and evaluated at the point (0,0). An identical definition holds for (∂/∂φ)[f]|(0,0).
Since , the previous equation can be rewritten as , where . We note that because , . We let .
Minimum sample size requirements in presence of phenotype misclassification – Alzheimer's disease ApoE example
We determine the minimum sample size necessary to maintain a constant power of 95% at the 5% significance level using formula (3) and considering estimated genotype frequencies from a recently published genetic association analysis of Alzheimer's Disease (AD) cases and controls genotyped at the ApoE marker locus [9]. In most populations there are three alleles at the ApoE locus. Conventionally, they are denoted ε2, ε3, and ε4 and we label them 2, 3, and 4 respectively in this work. In a well known and often replicated association finding, every copy of the 4 allele in a person's genotype increases that person's risk of getting late-onset AD by a factor of 2.5–3 [31]. Furthermore, recently published estimates of prevalence for Alzheimer's Disease in the US hover around the 2% range [32]. Thus, for our sample size calculations, we assume a prevalence K = 0.02.
If we index the six genotypes as 1 = 22, 2 = 23, 3 = 24, 4 = 33, 5 = 34, 6 = 44, then the genotype frequency values we use for our sample size calculations (taken from our previous work [9]) are:
p01 = 0.019, p11 = 0.000, p02 = 0.057, p12 = 0.118, p03 = 0.019, p13 = 0.024, p04 = 0.465, p14 = 0.699, p05 = 0.344, p15 = 0.159, p06 = 0.096, p16 = 0.000.
As it has been documented that phenotype misclassification in Alzheimer's Disease may run as high as 15% or more [19], we consider phenotype misclassification values 0 ≤ θ, φ ≤ 0.15, in increments of 0.01. It is assumed that there are equal numbers of cases and controls (R* = 1).
Authors' contributions
BJE performed all analyses and wrote the majority of the original manuscript. CH wrote all computer code for simulations. MAL wrote portions of the manuscript and contributed to the development of the results to be presented. SJF and DG formulated the original research question and supervised every stage of the research. They also re-wrote significant portions of the revised manuscripts.
Appendix
Here, we derive formulas for the case and control genotype frequencies, , , in terms of the affected genotype frequencies p0j, the unaffected genotype frequencies p1j, the disease prevalence K, and the misclassification error probabilities, θ and φ. Zheng and Tian derived similar results in a genetic-model based framework [14].
= Pr(genotype = j | case) = Pr(genotype = j, case)/Pr(case)
= [Pr(genotype = j, case, affected) + Pr(genotype = j, case, unaffected)]/Pr(case)
= [Pr(genotype = j | case, affected) Pr(case | affected) Pr(affected) + Pr(genotype = j | case, unaffected) Pr(case | unaffected) Pr(unaffected)]/[Pr(case | affected) Pr(affected) + Pr(case | unaffected) Pr(unaffected)]
= [p0j (1 - θ)K + p1jφ(1 - K)]/[(1 - θ)K + φ(1 - K)].
= Pr(genotype = j | control) = Pr(genotype = j, control)/Pr(control)
= [Pr(genotype = j, control, affected) + Pr(genotype = j, control, unaffected)]/Pr(control)
= [Pr(genotype = j | control, affected) Pr(control | affected) Pr(affected) + Pr(genotype = j | control, unaffected) Pr(control | unaffected) Pr(unaffected)]/[Pr(control | affected) Pr(affected) + Pr(control | unaffected) Pr(unaffected)]
= [p0jθK + p1j(1 - φ)(1 - K)]/[θK + (1 - φ)(1 - K)].
Acknowledgements
The authors gratefully acknowledge grants K01-HG00055 (DG) and HG00008 (to J. Ott) from the National Institutes of Health. BJE was supported by the Rockefeller University Science Outreach Program. The authors also gratefully acknowledge two anonymous reviewers whose comments led to significant improvements and simplifications in the research.
==== Refs
Breslow NE Day NE Statistical Methods in Cancer Research The Analysis of Case-Control Studies 1980 1 Eighth Lyon, International Agency for Research on Cancer 350
Ott J Analysis of Human Genetic Linkage 1999 Baltimore, The Johns Hopkins University Press
Page GP George V Go RC Page PZ Allison DB "Are we there yet?": Deciding when one has demonstrated specific genetic causation in complex diseases and quantitative traits Am J Hum Genet 2003 73 711 719 13680525 10.1086/378900
Rice JP Saccone NL Rasmussen E Definition of the phenotype Adv Genet 2001 42 69 76 11037314
Bross I Misclassification in 2 x 2 tables Biometrics 1954 10 478 486
Gordon D Levenstien MA Finch SJ Ott J Errors and linkage disequilibrium interact multiplicatively when computing sample sizes for genetic case-control association studies Pac Symp Biocomput 2003 490 501 12603052
Gordon D Finch SJ Nothnagel M Ott J Power and sample size calculations for case-control genetic association tests when errors are present: application to single nucleotide polymorphisms Hum Hered 2002 54 22 33 12446984 10.1159/000066696
Kang SJ Gordon D Finch SJ What SNP genotyping errors are most costly for genetic association studies? Genet Epidemiol 2004 26 132 141 14748013 10.1002/gepi.10301
Gordon D Yang Y Haynes C Finch SJ Mendell NR Brown AM Haroutunian V Increasing power for tests of genetic association in the presence of phenotype and/or genotype error by use of double-sampling Stat Appl Genet and Mol Biol 2004 3 Article 26 16646805
Brown RP Sweeney J Frances A Kocsis JH Loutsch E Age as a predictor of treatment response in endogenous depression J Clin Psychopharmacol 1983 3 176 178 6875025
Appels A Mulder P Imminent myocardial infarction: a psychological study J Human Stress 1984 10 129 134 6520393
Duffy SW Rohan TE Kandel R Prevost TC Rice K Myles JP Misclassification in a matched case-control study with variable matching ratio: application to a study of c-erbB-2 overexpression and breast cancer Stat Med 2003 22 2459 2468 12872302 10.1002/sim.1477
Jacobsen SJ Roberts RO Re: Effect of nonsteroidal anti-inflammatory agents and finasteride on prostate cancer risk J Urol 2003 169 1798 1799 12686844 10.1097/01.ju.0000057804.01025.13
Zheng G Tian X The impact of diagnostic error on testing genetic association in case-control studies Stat Med 2005 24 869 882 15568195 10.1002/sim.1976
Ioannidis JP Trikalinos TA Ntzani EE Contopoulos-Ioannidis DG Genetic associations in large versus small studies: an empirical assessment Lancet 2003 361 567 571 12598142 10.1016/S0140-6736(03)12516-0
Lohmueller KE Pearce CL Pike M Lander ES Hirschhorn JN Meta-analysis of genetic association studies supports a contribution of common variants to susceptibility to common disease. Nat Genet 2003 33 177 182 12524541 10.1038/ng1071
Hirschhorn JN Altshuler D Once and again-issues surrounding replication in genetic association studies J Clin Endocrinol Metab 2002 87 4438 4441 12364414 10.1210/jc.2002-021329
Ioannidis JP Genetic associations: false or true? Trends Mol Med 2003 9 135 138 12727138 10.1016/S1471-4914(03)00030-3
Lansbury PTJ Back to the future: the 'old-fashioned' way to new medications for neurodegeneration Nat Med 2004 10 Suppl S51 7 15298008 10.1038/nrn1435
Press MF Hung G Godolphin W Slamon DJ Sensitivity of HER-2/neu antibodies in archival tissue samples: potential source of error in immunohistochemical studies of oncogene expression Cancer Res 1994 54 2771 2777 7909495
Burd L Kerbeshian J Klug MG Neuropsychiatric genetics: misclassification in linkage studies of phenotype-genotype research J Child Neurol 2001 16 499 504 11453446
Mote VL Anderson RL An investigation of the effect of misclassification on the properties of chisquare-tests in the analysis of categorical data Biometrika 1965 52 95 109 14341284
Gordon D Ott J Assessment and management of single nucleotide polymorphism genotype errors in genetic association analysis Pac Symp Biocomput 2001 18 29 11262939
Carroll RJ Gail MH Lubin JH Case-control studies with errors in covariates J Am Stat Assoc 1993 88 185 199
Mitra SK On the limiting power function of the frequency chi-square test Ann Math Stat 1958 29 1221 1233
Cochran WG The chi-square test of goodness of fit Ann Math Stat 1952 23 315 345
Tosteson TD Buzas JS Demidenko E Karagas M Power and sample size calculations for generalized regression models with covariate measurement error Stat Med 2003 22 1069 1082 12652554 10.1002/sim.1388
Sasieni PD From genotypes to genes: doubling the sample size Biometrics 1997 53 1253 1261 9423247
Czika W Weir BS Properties of the multiallelic trend test Biometrics 2004 60 69 74 15032775 10.1111/j.0006-341X.2004.00166.x
Box GEP Hunter WG Hunter JS Statistics for Experimenters Wiley series in probability and mathematical statistics 1978 New York, John Wiley and Sons
Corder EH Saunders AM Strittmatter WJ Schmechel DE Gaskell PC Small GW Roses AD Haines JL Pericak-Vance MA Gene dose of apolipoprotein E type 4 allele and the risk of Alzheimer's disease in late onset families Science 1993 261 921 923 8346443
Sloane PD Zimmerman S Suchindran C Reed P Wang L Boustani M Sudha S The public health impact of Alzheimer's Disease, 2000-2050: potential implication of treatment advances Annu Rev Public Health 2002 23 213 231 11910061 10.1146/annurev.publhealth.23.100901.140525
| 15819990 | PMC1131899 | CC BY | 2021-01-04 16:38:18 | no | BMC Genet. 2005 Apr 8; 6:18 | utf-8 | BMC Genet | 2,005 | 10.1186/1471-2156-6-18 | oa_comm |
==== Front
BMC Health Serv ResBMC Health Services Research1472-6963BioMed Central London 1472-6963-5-311586012710.1186/1472-6963-5-31Research ArticleUnder-reporting of inpatient services utilisation in household surveys – a population-based study in Hong Kong Tsui Eva LH [email protected] Gabriel M [email protected] Pauline PS [email protected] Sarah [email protected] Su-Vui [email protected] Hospital Authority, 5/F, HA Building, 147B Argyle Street, Kowloon, Hong Kong, China2 Department of Community Medicine and School of Public Health, University of Hong Kong, 5/F, Academic & Administration Block, Faculty of Medicine Building, 21 Sassoon Road, Hong Kong, China3 Health Care Financing Study Group, Health and Welfare Bureau, Government of the Hong Kong Special Administrative Region, Murray Building, 3 Garden Road, Central, Hong Kong, China2005 28 4 2005 5 31 31 25 10 2004 28 4 2005 Copyright © 2005 Tsui et al; licensee BioMed Central Ltd.2005Tsui 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
Recognising that household interviews may produce biased estimates of health services utilisation, we examined for under- and over-reporting of hospitalisation episodes in three recent, consecutive population-based household surveys in Hong Kong.
Methods
Territory-wide inpatient service utilisation volumes as estimated from the 1999, 2001 and 2002 Thematic Household Surveys (THS) were benchmarked against corresponding statistics derived from routine administrative databases. Between-year differences on net under-reporting were quantified by Cohen's d effect size. To assess the potential for systematic biases in under-reporting, age- and sex-specific net under-reporting rates within each survey year were computed and the F-test was performed to evaluate differences between demographic subgroups. We modelled the effects of age and sex on the likelihood of ever hospitalisation through logistic regression to compare the odds ratios respectively derived from survey and administrative data.
Results
The extent of net under-reporting was moderately large in all three years amounting to about one-third of all inpatient episodes. However, there did not appear to be significant systematic biases in the degree of under-reporting by age or sex on stratified analyses and logistic regression modelling.
Conclusion
Under-reporting was substantial in Hong Kong's THS. Recall bias was likely most responsible for such reporting inaccuracies. A proper full-design record-check study should be carried out to confirm the present findings.
==== Body
Background
Population and health services research commonly relies on in-person household interviews as the main source of health and health care data, in terms of disease, disability and utilisation of services. These types of information are important for evidence-based health policy formulation, planning and evaluation. While medical chart review, insurance claims records and government macro statistics are potential alternative sources of such information, they cannot entirely replace the household interview given the often prohibitive expense of data abstraction exercises, lack of population coverage of single data sources especially in a mixed medical economy where there is a multiplicity of financial intermediaries and care providers, and the inability to study individual-level associations with ecologic data respectively.
However, self-reported data from household survey reports are subject to various types of error. Generally, random reporting error tends to increase the variance and thus uncertainty associated with the data, whereas systematic reporting error can bias survey estimates. Therefore, it is important to study the accuracy and validity of data obtained from household surveys.
There are two broad categories of reporting error in surveys on health services utilisation: under-reporting and over-reporting. Under-reporting refers to respondents forgetting or otherwise omitting (often due to the sensitive nature of the questions that may reflect socially undesirable or embarrassing behaviour) relevant episodes. Over-reporting occurs when interviewees attribute episodes outside the reporting period or survey definition to their response. They may mis-report an episode outside the reference period as if it had happened within that period in either the forward or backward direction – i.e. the "telescoping" phenomenon which could also lead to over-reporting.
In Hong Kong where a mixed medical economy of public and private providers operate in parallel, with the former delivering over 94% of total bed-days and the latter responsible for more than 70% of all ambulatory episodes, household surveys have been the only viable and sustainable option to follow utilisation trends over time, the findings from which form the basis for health policy decisions. We therefore examined the accuracy and validity of recall in such recent surveys by comparing survey responses for inpatient utilisation to aggregate statistics from the two main public sector service organisations, namely the Hospital Authority (HA) for public inpatient data and corresponding figures on private hospital admissions collated by the Department of Health (DH).
The specific aims of the study were to (i) benchmark survey results against territory-wide macro estimates, based on administrative records, of the number of hospitalisation episodes in both the public and private sectors; (ii) analyse the variability in agreement between survey and administrative estimates with respect to age and sex; and (iii) consider the effect of reporting errors on the application of survey estimates for planning and research purposes.
Methods
Data sources
The Thematic Household Surveys (THS) are a regular series of cross-sectional in-person surveys of the land-based population of Hong Kong, conducted by the Census and Statistics Department (C&SD) of the Hong Kong Special Administrative Region Government, People's Republic of China. Different topics are covered in each round of THS. Topics related to health services utilisation such as hospitalisation and/or doctor consultation episodes had been surveyed 13 times since 1982. The three most recent such surveys were conducted in 1999, 2001 and 2002 consecutively [1-3]. The 2002 THS covered the entire land-based population of Hong Kong including both institutional and non-institutional residents whereas the 1999 and 2001 rounds only included the non-institutional population. (Table 1)
Table 1 Sample size and age-sex demographics of the last three rounds of THS
Sept – Nov 1999 Jan – May 2001 May – July 2002
No. of households successfully enumerated 10,057 10,046 10,015 (and 2,111 institutional residents)
Response rate 77% 76% 78% (97%)
No. of individuals No. of individuals No. of individuals
Age (years)
Less than 5 1,506 1,298 1,108
5 – 14 4,682 4,385 3,849
15 – 24 4,875 4,808 3,974
25 – 34 5,471 5,418 4,360
35 – 44 6,684 6,472 5,774
45 – 54 4,420 4,856 4,551
55 – 64 2,514 2,493 2,524
65 or above 3,611 3,879 5,532
Sex
Male 16,601 16,484 15,321
Female 17,162 17,125 16,351
Total 33,763 33,609 31,672
Respondents were asked to recall the total number of hospital admissions and associated characteristics (e.g. provider type, reason for attendance, payment details and so on) of each episode. The recall periods were 12 months in the 2002 THS, but six months for 1999 and 2001. In addition to details of inpatient care episodes, the surveys also collected information on demographic and socio-economic characteristics, health status and medical need (e.g. presence of chronic conditions, regular medications taken), medical benefits and insurance coverage, among others. Proxy reporting by primary caretakers was allowed for respondents aged 12 and below and those who were mentally unfit to respond to the survey, except for the self-reported health status questions.
Survey responses were compared, on an ecologic or macro level, with the "gold-standard" administrative databases maintained by DH and HA. The Hospital Authority, the dominant inpatient service provider (with >94% market share in terms of total bed-days), has detailed data on all inpatient episodes in the public sector and DH collects aggregate utilisation statistics routinely from all 12 private hospitals. Together, there is total coverage of hospitalization episodes in terms of volume of inpatient utilisation in the territory.
All inpatient administration procedures in all 44 public hospitals under the management of HA are electronically processed through an Integrated Patient Administration System (IPAS). The first version of this standardized corporate-wide system has been fully implemented since early 1994. It consists of (1) the Hong Kong Patient Master Index which is a corporate database holding personal particular details of patients; (2) admission/discharge/transfer modules which provide timely and comprehensive information on patient movement; and (3) a medical record indexing module which provides an indexing tool to facilitate the tracing and administration of patient medical records. All patients are identified by their Hong Kong resident identity card number. For non-residents, a pseudo identity card number is generated by the system and intended for repeated use in subsequent admissions. A unique episode number is generated for each inpatient episode within all HA hospitals, and coupled with the patients' identification card number form the common identifiers for all inpatient episodes within the HA system. Moreover, the system is linked to the patient billing system. Hong Kong has a very straightforward inpatient financing mechanism whereby the cost is 98% subsidised through general taxation as a one-line vote to the HA annual operating budget and the all-inclusive per diem point-of-care user fee at a public hospital is HK$100 or $68 prior to 2003 (HK$7.8 = US$1 pegged exchange rate) which is payable by all except for civil servants, HA staff and the socially and medically indigent (all certified by special entitlement cards captured in IPAS) through out-of-pocket payments without financial intermediaries such as insurance. Non-residents are charged at a much higher full cost recovery rate. Therefore this relatively simple payment administrative mechanism acts as a confirmatory check on the clinical utilisation statistics in IPAS. With this infrastructure in place, we believe that HA has virtually complete and accurate data capture of all inpatient episodes in the public sector and therefore can be taken as a reliable "gold-standard" in this audit exercise, much more so than in the case of most other countries with more complicated systems involving financial intermediaries, free care or otherwise unrecorded episodes.
The additional file 1 contains further technical details on survey sampling design, weighting methodology, questionnaire extracts and utilisation volume estimation formulae used.
Statistical analysis
Territory-wide service utilisation volume as estimated from the 1999, 2001 and 2002 rounds of the THS was verified against corresponding statistics derived from the routine DH and HA administrative databases. The 2002 THS was undertaken between May and July, therefore the comparator period of July 2001 through June 2002 was adopted as the benchmark, to take into account possible seasonal effects of service utilisation. The comparator periods of observation were similarly determined for 1999 and 2001, albeit with six months as the duration of observation. (Table 2)
Table 2 Inpatient utilisation volumes derived from THS vs administrative data
1999 2001 2002 Cohen's d effect size index on under-reporting§
Admin data (a) Survey data (b) Net under-reporting 1- (b)/(a) Admin data (a) Survey data (b) Net under-reporting 1- (b)/(a) Admin data (a) Survey data (b) Net under-reporting 1- (b)/(a)
1999 vs 01 1999 vs 2002 2001 vs 2002
Comparator period May 1999 – Oct 1999 Sep – Nov 1999; last admission in past 6 months Nov 2000 – Apr 2001 Jan – May 2001; last admission in past 6 months Jul 2001 – Jun 2002 May – Jul 2002; all admissions in past 12 months
No. of persons ever discharged from public hospitals 322,400 (295,974)* 205,039 36.4% (30.7%)* 345,029 (317,128)* 208,952 39.4% (34.1%)* 602,673 (546,793)* 337,868 43.9% (38.2%)* 1.64 3.88 2.20
Total no. of discharges from private hospitals 94,366 50,533† 46.4% 98,693 51,285† 48.0% 197,738 105,850 46.5% 0.49 0.007 0.59
* After excluding non-residents and deceased patients as at 31 Oct 1999, 30 Apr 2001 and 30 June 2002 respectively for the 3 rounds of THS
†Assuming one hospital episode only for those who reported a last admission to a private hospital in the past 6 months
§Cohen's for group 1 and 2 under comparison; where a = ad min istrative data, b = survey and
We examined for differences in utilisation volumes between survey data and administrative statistics across corresponding years by calculating the Cohen's d effect size index, a standard statistical methodology, where a value of 0.2 indicates a small effect size, 0.5 a medium effect size and 0.8 or greater a large effect size [4].
To assess the potential for systematic biases in under-reporting, age- and sex-specific net under-reporting rates within each survey year were computed and the F-test was performed to evaluate differences between demographic subgroups at an overall significance level of 0.05. In addition, we modelled the effects of age and sex on the likelihood of ever hospitalisation in public hospitals by logistic regression to compare the odds ratios respectively derived from survey and administrative data. In the model based on survey data, individuals were dichotomised into ever and never hospitalised groups according to their survey responses. In the model using administrative data, the ever hospitalised group was defined based on HA's individual-level hospitalisation episode records whereas the never hospitalised group was derived from the difference between the territory-wide population figure and the former hospitalised headcounts. To evaluate the agreement between the two sources in the odds ratio of hospitalisation relative to a particular age subgroup as reference control, we assessed if there was overlap in the respective 95% confidence intervals of the odds ratios for each demographic subgroup. If there was no systematic difference in under-reporting between demographic subgroups, the odds ratios of hospitalisation for each age-sex subgroup derived from the survey data should be similar to the corresponding odds ratios derived from the administrative data. Lastly, an interaction term for age and sex was added to and tested for in the full models as health service utilisation might have varied in different gender and age groups. Although THS survey data contains other sociodemographic and patient characteristics which can be used for further comparison examining for systematic reporting bias by these variables, there is no corresponding information in the HA or DH routine databases that would allow for such.
All analyses were performed using SAS (SAS Institute Inc.) Version 8.0.
Results
Aggregate utilisation volumes in THS vs administrative data
Table 2 shows aggregate utilisation estimates derived from the THS in 1999, 2001 and 2002 and corresponding administrative databases for inpatient care episodes. The extent of net under-reporting was moderately large in all three years amounting to about one-third of all inpatient episodes, after adjustment by excluding deceased inpatients and non-residents during the survey periods from the denominator. Under-reporting appears to have been particularly acute in the 2002 round, perhaps due in part to the questionnaire design where only those who reported symptoms were asked about service utilisation, as opposed to documenting all care episodes regardless of the presence of symptoms. In addition, the recall period was longer (12 months vs six months) in the latest THS in 2002 as compared to the two previous rounds. Pairwise comparisons between years on the extent of net under-reporting, indicated significant differences between years, except for between 1999 and 2002 in terms of the total number of discharges from private hospitals. The magnitude of under-reporting was about one-third for inpatient episodes (30.7%, 34.1% and 38.2% respectively) after adjusting for non-Hong Kong residents (who were not covered by the THS) and those deceased as at the censor dates.
Net under-reporting was consistently higher for private compared to public sector inpatient admissions. This could have been an artefact where we had to assume only one hospitalisation episode for each person reporting a last admission in the previous six months whereas the DH administrative database contained information on the total number of discharges (i.e. not persons) from private hospitals. An individual with more than one hospitalisation episode would have generated only one count in the numerator but responsible for more than one in the denominator, thereby leading to artefactual under-reporting. In contrast, the HA database could accommodate person counts and therefore its data were directly comparable to the survey information. Moreover, due to the unavailability of detailed disaggregated data from the private hospitals, we could not adjust for deceased inpatients and non-residents in the estimation procedure.
Differences in reporting by age and sex
As Table 3 illustrates, there does not appear to be significant systematic biases in the degree of under-reporting by age or sex. Within each survey year, we did not detect statistically significant differences between males and females except for inpatient episodes in 1999 (p = 0.04). Under the hypothesis of equal degree of under-reporting across all age subgroups, if the under five age group was excluded, there were no significant differences at the 0.05 level. On the other hand, if the under five age group was included, significant differences were found in 2002 (p = 0.02). This difference can likely be explained by proxy reporting for the under five age group.
Table 3 Extent of under-reporting (%) by age and sex in each THS
1999 2001 2002
No. of persons discharged from public hospitals in past 6 months* No. of persons discharged from public hospitals in past 6 months* No. of persons discharged from public hospitals in past 12 months*
Age (years)
Less than 5 44.6% 51.8% 65.0%
5 – 24 36.2% 40.9% 38.2%
25 – 44 31.6% 31.9% 37.6%
45 – 64 27.3% 28.6% 38.0%
65 or above 25.7% 32.2% 31.7%
Sex
Female 34.7% 36.3% 39.4%
Male 26.1% 31.4% 36.8%
Overall 30.7% 34.1% 38.2%
F-test for the overall demographic subgroups differences within each year (at significance level 0.05)†
Age difference
Including the 'less than 5' age group (p-value) 0.47 0.18 0.02
Excluding the 'less than 5' age group (p-value) 0.53 0.28 0.60
Sex difference(p-value) 0.04 0.23 0.10
* Adjusted for non-residents and deceased patients as at 31 Oct 1999, 30 Apr 2001 and 30 June 2002 respectively for the 3 rounds of THS
† where ai = ad min istrative data, bi = survey data, for each demographic subgroup i up to k.
As an alternative approach, we modelled the effects of age and sex on the likelihood of ever hospitalisation in public hospitals. The full model with the interaction term of age-sex was first fitted, but was subsequently dropped due to insignificant age-sex interaction effects. Figures 1, 2 and 3 plot age- and sex-specific odds ratios of ever hospitalisation and 95% confidence intervals (CIs) using both survey and administrative data. Both sets of curves are very similar in both direction and magnitude and largely overlap in their 95% CIs, confirming that the two data sources show consistent relativity in ever hospitalisation rate by age and sex. It suggests that there are no substantial systematic biases in under-reporting among age and sex subgroups.
Figure 1 Comparison of odds ratios for age-sex effects on the likelihood of ever hospitalisation in public hospitals between administrative and survey data for year 1999.
Figure 2 Comparison of odds ratios for age-sex effects on the likelihood of ever hospitalisation in public hospitals between administrative and survey data for year 2001.
Figure 3 Comparison of odds ratios for age-sex effects on the likelihood of ever hospitalisation in public hospitals between administrative and survey data for year 2002.
Discussion
In this large population-based audit, our findings show that under-reporting was consistently substantial, amounting to about one-third of all inpatient episodes, in the last three rounds of THS benchmarked against administrative data on the aggregate level. Differences in age and sex did not influence the degree of under-reporting.
Of note, although we observed under-reporting overall, the possibility that this represented mixed under- and over-reporting on the individual level cannot be ruled out. Under this ecologic design, it was not possible to disentangle the relative contributions of each type of recall error. The usual ideal study design would be to compare questionnaire-derived data with information abstracted directly from medical records on a person or episodic basis, thus allowing for more detailed analysis of the different types of reporting inaccuracy. Such a design would enable the quantification of recall error by measuring proportions of agreement and kappa values as well as the generation of 2 × 2 contingency tables and associated statistics such as true positives and negatives. However, this is prohibitively resource-intensive and cannot be routinely carried out for audit and benchmarking purposes, at least in the Hong Kong and other rapidly developing economy settings where such a management culture has yet to take hold. Moreover, widespread public concern about data privacy and the perception of possible government intrusion into personal medical and payment records in this laissez-faire society (in the politico-economic sense) would result in a low participation rate thus rendering the whole exercise useless. Therefore, the next best pragmatic alternative is to use aggregate statistics benchmarked against routine statistics as we have done here, especially when our unique circumstances and simple administrative and clinical care infrastructure particularly lend themselves to adopt such data as a reliable "gold-standard".
By this same reasoning, we did not consider outpatient episodes in this exercise. Provision of outpatient services is shared by both private and public sectors in the ratio of 70:30. While HA is responsible for all public specialist and general outpatient clinics and has the requisite data for comparison, the majority of ambulatory episodes are provided by private, self-employed, solo practitioners who charge on a fee-for-service basis and patients mostly pay out-of-pocket [5]. There is no central information repository for the private sector and only about one-third of such solo or small-group clinics are computerised which could potentially support an audit exercise [6]. Managed care, in the various forms of contract medicine, prepaid plans and preferred provider networks, has grown in the last decade although their penetration is still very limited in scope and size. About 30% of the population have private insurance or benefits schemes coverage, mostly through employment-based programmes. The majority of such coverage comes in the form of riders to other types of insurance schemes, most commonly life policies. Taking all these factors into consideration, an ecologic benchmarking exercise is not feasible for the outpatient sector.
Another potential caveat concerns the fact that private hospital data did not adjust for deceased and non-resident patients and multiple care episodes. However, HA hospitals accounted for 87% of all 34,237 deaths in Hong Kong in 2002 whereas the remainder took place in private hospitals (8.2%) and other places outside hospitals (4.8%). For the same year, non-residents only accounted for 0.5% of total inpatient episodes in HA hospitals. Therefore, it is unlikely multiple care episodes by the same individual can account for most of the under-reporting observed, especially given the very small market share of private hospitalisation.
A variety of reasons could potentially explain the under-reporting observed, including recall bias, non-response bias, sampling and estimation biases and questionnaire length and content. First, recall bias is the systematic over- or under-reporting of recall behaviour in surveys, such as health services use in this context. Respondents may forget relevant episodes or they may report an episode from outside the period of interest as if it had happened within the period (forward telescoping) or vice versa (backward telescoping). They may report episodes that do not meet the survey definition or they may fail to report relevant episodes because they perceive that such episodes do not meet survey criteria. For example, hospital transfers within the public sector are counted as two separate episodes in administrative databases but respondents usually only report them as a single episode. This has been addressed in the data analysis as both the administrative and survey data for the public sector were counted on a person basis. Deliberate backward telescoping would be possible if the respondent wishes to shorten the interview. Motivation to report tends to increase with saliency and frequency of events, and may decrease with increasing number of events. A final possibility regarding recall bias is that people genuinely forget. It seems amazing that anyone would forget hospital treatment. But other surveys as discussed in this paper have reached the same conclusion. Also there are similar findings in other areas of health. For instance, a study of recall of a diagnosis of cancer concluded that 20% of people forget that they have had cancer [7]. Recall bias is likely most responsible for the under-reporting observed in the THS.
Second, for the non-institutional sample in the 2002 round of THS, the overall non-response rate was 21.6%. A total of 2263 households could not be contacted after repeated visits and 503 households refused to respond. Non-response is usually more likely in high-income and singleton households. Unless the incidence and level of health services utilisation of non-respondents were substantially different from those who responded, this potential effect on the survey estimates would be limited especially in view of the current low rate of non-response. Assuming that the sample mean of the non-respondents were +10%, +30% and +50% of the corresponding mean of the respondents in the 2002 round of THS, the relative bias of non-response would have been -2%, -6% and -10% of the true population total of health services utilisation. Conversely, the relative bias would have been +2%, +7% and +12% respectively if the non-respondents' mean were -10%, -30% and -50% of the respondents' mean.
Third, sampling bias would be an issue if a non-representative survey sample results. This is highly unlikely in the THS series of surveys given the whole population coverage, explicit and validated sampling methodologies and the application of weighting factors to the results to ensure representativeness for the general population of Hong Kong (see additional file for details). Moreover, we have further improved the validity of the estimation procedure by excluding deceased inpatients and non-residents from the denominator in calculating the net under-reporting rate, thereby optimising the comparability between survey and administrative data.
Finally, unlike most other similar health statistics or utilisation surveys overseas, the THS series usually combines two or even three sub-surveys of disparate topics into an "omnibus" type of questionnaire for economy of scale and efficiency. Therefore the resulting survey instrument is often very long and can take up to 45 minutes to an hour to complete for each household. Coupled with the anecdotal observation that most Hong Kong residents maintain a very busy and hectic daily schedule, it is perhaps not surprising that respondents might have under-reported in order to complete the survey in a shorter period of time, although we know that in other settings overseas, interviews of 45 minutes to an hour are an acceptable burden to respondents. It might also have been quite difficult to focus on recalling specific details accurately when there are multiple topics covered in the same interview.
The extent of under-reporting as documented is moderate to large compared to experiences elsewhere. For instance, Harlow and Linet [8] in their systematic review found high proportions (at least 90% in three studies) of positive matches between records and survey interviews for hospitalisation episodes in four studies. However, some have criticised the design of those studies in which either positive survey responses were verified against medical records or positive record values were checked against survey responses, producing estimates which were biased towards either over- or under-reporting respectively. A full-design record check study, as recommended by Marquis [9], which identifies a population and sample from it independently of records, obtains survey and record information for each sampled element and compares the two data sources should be the "gold-standard", where both interview over-report (false positives) and under-report (false negatives) could be detected. The Health Interview Evaluation Survey (HIES) conducted by the US National Center for Health Statistics in 1990 [10], employed such a full design. It aimed to evaluate the accuracy of two-week doctor visit reporting through record checks. The study universe consisted of members of a staff model health maintenance organisation in Washington, D.C. The 1000 self-responding adult samples were selected from the membership roll, with an over-sampling of persons with recent ambulatory visits. Significant findings from the HIES, which were consistent with other findings in the literature, included: (i) under-reporting ranged about 17–35% and over-reporting about 20–40% for the 2-week reference period, but there was no evidence of general net under- or over-reporting of visits at the person level; (ii) under-reports were about 13–15% more prevalent for visits in the earlier week of the reference period than for those in the later week; (iii) under-reporting was greater for persons with more visits in the reference period; (iv) statistically significant differences in the percentage of positive match between household members present for the interview (84.4%) and those not present (46.9%) suggested some under-reporting by proxy respondents; and (v) males tended to under-report consistently more than females [10].
In comparison, Cartwright [11] also found under-reporting and over-reporting from adult self-respondents to be both about 21% in respect of physician contacts in a 4-week reference period "bounded" by interviews both at the start and at the end. The corresponding rates recorded in the study by Sudman et al [12], using a combined interview and diary procedure with a 3-month reference period, were 24% and 17% respectively. Means and Loftus [13] found rates of under-reporting and over-reporting in excess of 50% in respect of medical visits and hospital stays using a 1-year reference period.
Importantly, the HIES and other previous studies cited above indicate few consistent patterns of under- or over-reporting by respondents' demographic characteristics. Findings about age and health status were not consistent, and other characteristics were typically not associated with significant differences in reporting. This non-systematic recall pattern is also true in the present audit exercise. Age and sex (except for the under five age group in both sexes) did not appear to have influenced the extent of recall error on the aggregate level to any substantive degree. We did not have other data such as income and education level to examine for possible differential reporting behaviour, although there is little reason to believe these would be present given the lack of such effects seen in other studies and for age and sex in this study.
Therefore, the application of results such as relative measures of association between various socioeconomic and demographic characteristics and health care utilisation from such survey data for health planning (e.g. in formulating target subsidies for certain groups to achieve equity in health financing) in this context is reasonably valid, where random or non-systematic error would produce a conservative under-estimation of the true effect size. If the key interest is however in the absolute rate or volume for service planning use, the total service volume derived from survey estimates would need to be grossed up pro rata according to the degree of net under-reporting, as a crude measure to correct for such recall error.
Conclusion
This audit has, for the first time in Hong Kong and elsewhere in Asia to the best of our knowledge, attempted to systematically ascertain the veracity and validity of health services utilisation estimates derived from household in-person interviews against routine administrative data. It is important that such an exercise be carried out on a regular basis as a continuous quality improvement initiative to ensure that data of the highest possible quality be used in the formulation of health care policy. Future research should explore the possibility of employing a full-design record check study to confirm the present findings and better understand other dimensions of recall and reporting behaviour. In addition, the current findings could be extended by analysing what sort of admissions people forget, e.g. are short stays less likely to be recalled compared to long stays? Are admissions for certain diagnoses not regarded as "proper" admissions and therefore not reported in surveys? Lastly, we should opt for psychological studies probing individuals for the mechanisms which suppress recall of an important health event.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
ELHT and GML conceived the study question and design. ELHT and PPSW carried out the statistical analysis, in consultation with GML. ELHT wrote the first draft and GML revised the manuscript. SC and SVL were responsible for the conduct and fieldwork of the Thematic Household Survey. All authors contributed to the critical evaluation of the methods, analysis and writing. GML is the guarantor of the study.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
Technical details of Thematic Household Surveys in 1999, 2001 and 2002 – descriptions concerning the sampling design, weighting method, estimation formulae, questionnaire wording and survey respondent inclusion criteria are provided.
Click here for file
Acknowledgements
The authors report the findings on behalf of the Health Care Financing Study Group. We thank the Research Office of the Health, Welfare and Food Bureau and the Census and Statistics Department of the Government of the Hong Kong Special Administrative Region for their support in commissioning and carrying out this study by providing us with the data of the three consecutive rounds of THS and advising on the technical details in relation to the survey; the Statistics Unit of the Department of Health and the Statistics and Research Unit of Hospital Authority Head Office for the provision of administrative data on health services utilisation for benchmarking against the survey data; and Marie Chi for expert secretarial support in the preparation of the manuscript.
==== Refs
Census and Statistics Department Thematic Household Survey Report No 3 – Health status of Hong Kong residents, doctor consultation, hospitalisation, dental consultation and the usage of Chinese medical products and food 2000 Hong Kong, Government Printing Department
Census and Statistics Department Thematic Household Survey Report No 8 – Health status of Hong Kong residents, doctor consultation, hospitalisation, dental consultation and provision of medical benefits by employers/companies and purchase of medical insurance by individuals 2002 Hong Kong, Government Printing Department
Census and Statistics Department Thematic Household Survey Report No 12 – Health status of Hong Kong residents, doctor consultation, hospitalisation, dental consultation, provision of medical benefits by employers/companies and coverage of medical insurance purchased by individuals and health status of institutional residents and their utilization of medical services 2003 Hong Kong, Government Printing Department
Cohen J Statistical power analysis for the behavioral sciences 1988 2 Hillsdale, NJ: Erlbaum
Leung GM Castan-Cameo S McGhee SM Wong IOL Johnston JM Waiting time, doctor-shopping and non-attendance at specialist outpatient clinics: case-control study of 6,495 individuals in Hong Kong Med Care 2003 41 1293 300 14583692 10.1097/01.MLR.0000093481.93107.C2
Leung GM Johnston J Ho LM Wong FK Castan-Cameo S Computerization of clinical practice in Hong Kong Int J Med Inform 2001 62 143 54 11470617 10.1016/S1386-5056(01)00158-7
Nord C Mykletun A Fossa SD Cancer patients awareness about their diagnosis J Public Health Med 2003 25 313 7 14747590 10.1093/pubmed/fdg076
Harlow SD Linet MS Agreement between questionnaire data and medical records, the Evidence for accuracy of recall Am J Epidemiol 1989 129 233 48 2643301
Marquis K Jabine T, Loftus E, Straf M Record checks for sample surveys Cognitive aspects of survey methodology: Building a bridge between disciplines 1984 Washington, DC, National Academy Press
National Centre for Health Statistics Evaluation of 2-week doctor visit reporting in the National Health Interview Survey Vital and health statistics 2(122) (DHHS publication no (PHS) 96-1396 1996 Atlanta, GA
Cartwright A Memory errors in morbidity surveys Milbank Memorial Fund Quarterly 1988 41 5 24 14018984
Sudman S Wallace W Ferber R The cost-effectiveness of using the diary as an instrument for collecting health data in household surveys 1974 Chicago, IL, University of Illinois
Means B Loftus EF When personal history repeats itself: Decomposing memories for recurring events Applied Cognitive Psychology 1991 5 297 318
| 15860127 | PMC1131900 | CC BY | 2021-01-04 16:31:49 | no | BMC Health Serv Res. 2005 Apr 28; 5:31 | utf-8 | BMC Health Serv Res | 2,005 | 10.1186/1472-6963-5-31 | oa_comm |
==== Front
BMC Health Serv ResBMC Health Services Research1472-6963BioMed Central London 1472-6963-5-331588246310.1186/1472-6963-5-33Research ArticleHospital service areas – a new tool for health care planning in Switzerland Klauss Gunnar [email protected] Lukas [email protected] Marcel [email protected] André [email protected] Institute for Evaluative Research in Orthopaedic Surgery (IEFO), University of Bern, Stauffacherstrasse 78, CH-3014 Bern, Switzerland2005 9 5 2005 5 33 33 24 12 2004 9 5 2005 Copyright © 2005 Klauss et al; licensee BioMed Central Ltd.2005Klauss 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 description of patient travel patterns and variations in health care utilization may guide a sound health care planning process. In order to accurately describe these differences across regions with homogeneous populations, small area analysis (SAA) has proved as a valuable tool to create appropriate area models. This paper presents the methodology to create and characterize population-based hospital service areas (HSAs) for Switzerland.
Methods
We employed federal hospital discharge data to perform a patient origin study using small area analysis. Each of 605 residential regions was assigned to one of 215 hospital provider regions where the most frequent number of discharges took place. HSAs were characterized geographically, demographically, and through health utilization indices and rates that describe hospital use. We introduced novel planning variables extracted from the patient origin study and investigated relationships among health utilization indices and rates to understand patient travel patterns for hospital use. Results were visualized as maps in a geographic information system (GIS).
Results
We obtained 100 HSAs using a patient origin matrix containing over four million discharges. HSAs had diverse demographic and geographic characteristics. Urban HSAs had above average population sizes, while mountainous HSAs were scarcely populated but larger in size. We found higher localization of care in urban HSAs and in mountainous HSAs. Half of the Swiss population lives in service areas where 65% of hospital care is provided by local hospitals.
Conclusion
Health utilization indices and rates demonstrated patient travel patterns that merit more detailed analyses in light of political, infrastructural and developmental determinants. HSAs and health utilization indices provide valuable information for health care planning. They will be used to study variation phenomena in Swiss health care.
==== Body
Background
Switzerland hosts the second most expensive health care system worldwide with health care expenditures accounting for approximately 6.2% of the GDP in 2001 [1]. In the same year overall cost of in-patient care reached 13.9 billion € with an average of 5800 € per case [2]. In order to curb costs, novel approaches to health care planning need to be assessed. We therefore investigated utilization patterns of hospitals and created population-based hospital service areas (HSAs). Such service areas take into account patient travel patterns and demonstrate how hospitals are utilized. They are a suitable area model for the analysis of variations in hospital use and health care expenditures. The topic is of national importance and should be of interest to other European countries, because the underlying methodology has proved to be an influential tool in health service research, especially in the United States [3-7].
Hospital and health service areas have been widely used to demonstrate variations in health care utilization rates and health care expenditures [8-26]. The large variation often cannot be attributed to underlying differences in the prevalence or incidence of diseases. Practice style differences, diagnostic and therapeutic uncertainties in clinical medicine and supplier-induced demand are the most widely employed hypotheses to explain these variations [27]. Especially the supplier-induced demand hypothesis makes it apparent that resource allocations should not ignore variation patterns: Two hypothetical regions might both maximally utilize local resources for a given diagnostic procedure. This taken by itself could wrongfully indicate the same need for granting further resources when in fact one region represents a very high per-capita procedure rate and the other region a very low per-capita rate. The procedure rate in the first region might be driven by supplier-induced demand. If this has led to an oversupply the granting of further resources is questionable. The example demonstrates the importance to incorporate knowledge of variation profiles into resource allocation decisions.
Measuring variation profiles requires suitable area models as units of analysis. Although it is tempting to rely on administrative areas, factors that influence hospital use – such as natural geographic boundaries, travel time, infrastructure and attractiveness of health services – may be independent of administrative areas (e.g. borough, county, or canton)[6]. Boarder-crossing residents who seek health care outside their own residential area make interpretation of low or high per-capita rates problematic due to a numerator-denominator mismatch [28]. Therefore the concept of health service areas was introduced. Several methods to form such areas are described in the literature [6,22,29] but only Wennberg's approach specifically accounts for patient utilization patterns [10]. His method creates areas according to the use of health services by a given population and is considered most suitable to study variation phenomena [22,28].
This paper discusses the methodological underpinning for creating population-based hospital service areas (HSAs), describes the steps to create them and characterizes each HSA geographically, demographically and by various health utilization indices and rates.
Methods
Data
We utilized federal discharge data from the Swiss hospital discharge master file from 1998 through 2001. This file was initiated in 1997. We chose a time period of four years to achieve stable estimates of discharge patterns and to avoid possible fluctuation over one-year periods. At the time of analysis no data later than 2001 were available for research purposes. The file contained over 4 million hospital discharges of persons residing in Switzerland. The unit of observation were individual hospital discharges. We excluded discharges of patients not living in Switzerland at the time of treatment.
Patient residences and hospital locations were recorded as census region which are in essence aggregated zip code areas. Switzerland is divided into 605 census regions; their precise name in the federal discharge dataset is 'MedStat-region'. These regions constituted the geographic building blocks of subsequent HSAs. Their creation accounted for demographic, socio-economic and geographic criteria to ensure uniformity and comparability. Two-hundred-and-fifteen census regions contained at least one hospital; we termed them hospital regions. Unfortunately, the Swiss hospital discharge master file does not contain zip codes for patient residence but only the respective MedStat-region. This limitation is attributable to strict confidentiality laws and is unlikely to change in the near future.
We obtained commercial GIS-compatible vector files for Swiss census regions [MicroGIS Ltd., Baar, Switzerland]. Age-, and sex-stratified population counts for each census region were derived from the 1990 census. Necessary age-, and sex-stratified population counts for each census region were not yet available from the 2000 census. The difficulty in obtaining these counts owes to the fact that census data are collected on a different area scale than the specific census regions used in the federal hospital discharge data file. Subsequent approximations, performed by the Swiss Statistical Office, require complex algorithms and make use of so-called hectare data sets. At the time of our analysis, these hectare data were not available from all cantons.
Definition of health service areas
We defined hospital service areas through a patient origin study by cross-tabulating the sum of discharges of every residential region with all possible hospital regions. Each matrix cell gave the sum of discharges of residents from a given residential region in a hospital region. This step identified the main hospital provider region – the one with the highest number of discharges – for each residential region. A three-step algorithm first assigned residential regions to their main provider hospital region. We then adjusted several assignments manually to achieve contiguous hospital service areas. Contiguity, a geographical convention, ensures the readability and interpretability of maps [30]. Preliminary HSAs were finally checked for the plurality rule: whether the sum of all discharges of HSA-residents within their HSA was higher than the sum of discharges to any other HSA. Failed plurality necessitated merging HSAs. Assigning a unique colour to every residential region with the same hospital region graphically displayed newly defined HSAs.
HSA characteristics
We characterized HSAs demographically, geographically, by health utilization indices, and rates. Our patient origin study provided four core variables for each HSA: (1) the population count; (2) the sum of discharges of HSA-residents irrespective of hospital location, i.e. including discharges in hospitals outside their residential HSA; (3) the sum of discharges of HSA-residents only in hospitals within their HSA; and (4) the sum of discharges in HSA-hospitals irrespective of residential HSA of patients, i.e., including patients living outside the HSA. These variables allowed the calculation of a series of discharge counts, health utilization indices, and health utilization rates. Depending on the denominator, health utilization indices can be sub-classified into population-based and hospital-based.
The localization index (LI) is the fraction of all discharges of HSA-residents that happened within their HSA. It is calculated by dividing discharges of HSA-residents within their HSA by the total discharges of HSA-residents. It is a population-based index and it indicates the degree of localization of hospital care provided for the population in a given HSA.
The inflow index (II) is the fraction of non-HSA-resident discharges of all discharges within HSA-hospitals. It is calculated by subtracting from 1 the quotient of non-HSA-residential discharges over the total number of discharges in HSA-hospitals. It is a hospital-based index that can be viewed as a crude measure of attractiveness. It also depends on the type of hospital services provided within a given HSA.
The market share index (MSI) is the fraction of HSA-resident discharges of all discharges within HSA-hospitals. It is calculated by dividing the number of residential HSA discharges by the total number of discharges in HSA hospitals. This too is a hospital-based index; it indicates the degree of localization of hospital care provision from a hospital perspective. MSI and II are complementary to each other; thus adding to one in each HSA. One is therefore sufficient and we favoured the II. Above indices can be multiplied by 100 to give percentages; we will present all indices as percentages.
The net patient flow (NPF) is the fraction of the overall resulting patient movement. It is calculated by dividing the sum of discharges in HSA-hospitals by the sum of discharges of HSA-residents, minus 1. A negative NPF indicates that more HSA-residents leave the HSA than non-resident patients move into the HSA for hospital services. A positive NPF indicates that the number of patients coming from outside for hospital services is greater than the number of HSA-residents leaving their HSA to receive hospital care. This index has no unit. To our knowledge the NPF is a novel measure which we calculated from a patient origin study.
We also used the four core variables to calculate the number of discharges of HSA residents that received hospital care outside their HSA, the number of discharges of non-HSA residents that received hospital care within a given HSA and the effective number of discharges that crossed HSA borders by subtracting the two discharge counts. Dividing the three counts by the population in each HSA and multiplying by 1000 gave three population based rates: the local-out rate quantifies the number of discharges per 1000 HSA-residents that receive hospital care outside their HSA, the nonlocal-in rate quantifies the number of nonresidents that are treated per 1000 HSA residents, and the net-rate quantifies the effective exchange of patients (either in or out) per 1000 HSA residents. Dividing each rate by the country-wide average provided a rate ratio to compare HSAs in terms of their patient movements.
Data management/statistics
We performed data handling and statistical analyses in Stata 8® [StataCorp., Texas, USA]; we used a GIS [ArcView 8.2®, ESRI, Redlands CA, USA] and vector files to map spatial data. This paper presents primarily descriptive statistics of characteristics of newly defined HSAs. We investigated the relationships among health utilization indices and among health utilization rates. Correlations of continuous variables with skewed distribution were assessed with Spearman's rho; two means were compared with the Student-t-test for normally distributed variables and the Wilcoxon rank sum test for skewed variables. We employed simple linear regression to analyse the extent of linear relationships between continuous variables.
Results
Swiss hospital service areas
Figure 1 gives the reader an overview of Swiss topography. The massive Alps run along the entire south and cover roughly half of the country's area. The Jura mountain range runs along the northwest. In between stretch the Swiss midlands, from southwest to northeast. They are densely populated and contain most urban centres. Figure 2 shows the 605 census regions, which are the building blocks for our HSA. We wish to point out the large census regions in the Alps and the Jura range. In comparison, census regions are smaller and more numerous in the Swiss midlands. Urban centres like Basel, Bern, Geneva, Lausanne, and Zurich are clearly recognizable.
Figure 1 Swiss Topography.
Figure 2 Census Regions; building blocks of Swiss HSAs.
The patient origin study yielded 100 HSAs as shown in Figure 3. Twenty two residential regions were reassigned for contiguity; one preliminary HSA was merged to achieve plurality. Fifteen HSAs incorporated a census region of a neighbouring canton which we defined as geographic extension over canton borders.
Figure 3 Swiss Hospital Service Areas.
Table 1 gives an overview of core variables we retrieved from the patient origin study. HSAs were ranked according to the number of discharges of HSA-residents (population discharges). We then chose five HSAs with the highest ranks, five with the lowest ranks, and five with ranks around the median for demonstration. Table 2 gives; also for these 15 HSAs; the calculated discharge counts, health utilization indices, and health utilization rates to enables the reader replication of calculations.
Table 1 Core variables of patient origin study per HAS
Lft HSA Pop Pop_d Local_d Hosp_d
1 GE20 Genève 375900 70836 67887 72637
2 VD13 Lausanne 266627 53320 42360 62779
3 BE09 Bern 365502 46928 41383 69197
4 ZH85 Zürich-Grünau 210799 32178 16803 37081
5 LU01 Luzern 224858 30984 25025 34019
51 SG11 Uznach 52679 7418 2659 2942
52 ZH12 Horgen 49791 7383 3729 5843
53 VS31 Sierre 38293 6801 4061 7074
54 SZ10 Schwyz 49187 6415 3512 4174
55 ZH04 Affoltern 35403 6174 3231 3635
96 GR03 Engiadina 6744 1101 683 852
97 BE67 Simmental 8806 1037 367 656
98 BE68 Oberhasli 8053 1021 642 2474
99 GR10 Poschiavo 4398 727 403 417
100 GR09 Val Müstair 1623 250 163 339
Legend of variable* abbreviations in Table 1
Pop Population
Pop_d Population discharges
Local_d Local discharges
Hosp_d Hospital discharges
* each variable is described in detail in the Methods section of the text
Table 2 Discharge counts, health utilization indices and health utilization rates retrieved per HSA
Discharge Counts Indices Rates
Lft HSA L_out NL_in Diff LI II NPF Pop Local L_out NL_in Net
1 GE20 2949 4750 1801 0.96 0.93 0.03 188 181 8 13 5
2 VD13 10960 20419 9459 0.79 0.67 0.18 200 159 41 77 35
3 BE09 5545 27814 22269 0.88 0.6 0.47 128 113 15 76 61
4 ZH85 15375 20278 4903 0.52 0.45 0.15 153 80 73 96 23
5 LU01 5959 8994 3035 0.81 0.74 0.1 138 111 27 40 13
51 SG11 4759 283 -4476 0.36 0.9 -0.6 141 50 90 5 -85
52 ZH12 3654 2114 -1540 0.51 0.64 -0.2 148 75 73 42 -31
53 VS31 2740 3013 273 0.6 0.57 0.04 178 106 72 79 7
54 SZ10 2903 662 -2241 0.55 0.84 -0.4 130 71 59 13 -46
55 ZH04 2943 404 -2539 0.52 0.89 -0.4 174 91 83 11 -72
96 GR03 418 169 -249 0.62 0.8 -0.2 163 101 62 25 -37
97 BE67 670 289 -381 0.35 0.56 -0.4 118 42 76 33 -43
98 BE68 379 1832 1453 0.63 0.26 1.42 127 80 47 227 180
99 GR10 324 14 -310 0.55 0.97 -0.4 165 92 74 3 -70
100 GR09 87 176 89 0.65 0.48 0.36 154 100 54 108 55
Legend of variable* abbreviations in Table 2
L_out Local-out discharges
NL_in Nonlocal-in discharges
LI Localization index
II Inflow index
NPF Net patient flow
Net Net rate
* each variable is described in detail in the Methods section of the text
Demographic/geographic characteristics
HSAs showed marked differences in demographic and geographic characteristics. The mean population size was 68,867 (median 47,273), ranging from 1,623 to 375,900. Area size was on average 399 km2 (median 311 km2), ranging from 35 to 2125 km2. Larger HSAs were more often seen in mountainous, decentralized regions. Thirteen HSAs consisted of a single census region; forty-one consisted of two, three, or four census regions. Combined they constituted 54% of HSAs. The number of hospitals per HSA ranged from one to sixteen, with 32 HSAs having only one serving hospital and 27 HSAs having five hospitals or more. HSAs with more than five hospitals were urban or contained agglomerations around cities. The ten HSAs with the highest population had on average larger area sizes and incorporated urban centres. The ten highest population densities were also seen in HSAs of large urban centres. The ten lowest population counts were seen in mountainous HSAs (irrespective of area size) and in smaller HSAs.
Health utilization indices
Figure 4 depicts a map of the localization indices of HSAs. The HSA with the highest LI was Geneva (LI = 96%), the HSA with the lowest LI (of 28%) was situated in the Swiss Midlands. Three geographic patterns emerged: (1) HSAs of mountainous regions had above-average LIs; (2) HSAs with below-average LIs were predominantly highly developed region; (3) HSAs which incorporated any of the 14 large urban centres or agglomerations showed above-average LIs (mean = 75.7%) compared to the remaining HSAs (mean = 56.1%; p < 0.0001).
Figure 4 Localization Indices (in %) of Swiss HSAs.
Figure 5 shows the inflow indices of HSAs as percentages. These ranged from 3% (mountainous HSA) to 81% (Swiss Midlands). A distinct geographic pattern is less easily discernible because high or low IIs could be found irrespective of demographic and geographic characteristics. We detected one weak geographic pattern: higher IIs were seen in a number of HSAs of urban centres, a distinct exception being the HSA of Geneva, which exhibits a very low II, combined with the highest LI. Other variables not discussed or measured in this study (e.g. hospital contracts, hospital specialties, rehabilitation centres) were likely more influential on the II of a given HSA.
Figure 5 Inflow Indices (in %) of Swiss HSAs.
Figure 6 shows the net patient flow of HSAs. Red colour indicates higher inflow of nonresident patients (NFP positive) than outflow of residents (NPF negative) to hospitals outside the HSA. Green colour indicates a higher overall outflow of residents. Yellow HSAs had a balanced NPF around zero (+/- 0.05). Most urban HSAs displayed high positive NPFs. HSAs surrounding the urban centres predominantly showed varying degrees of negative NPFs. Four mountainous HSAs with famous rehabilitation clinics (Davos, Meiringen, Crans-Montana, Walensee) also displayed high positive NPFs. Please be reminded that inflow of foreign patients is not considered in our study.
Figure 6 Patient Net Flow (Ratio) of Swiss HSAs.
Relationship of HSA utilization indices and utilization rates
We investigated the relationship among health utilization indices (Figure 7): LI, a population-based index, and II, a hospital-based index, demonstrated weak negative correlation (Spearman's rho = - 0.31; p = 0.018) which we decided to ignore (the corresponding graph appears rather like a cloud). Interestingly, NPF positively correlated with LI (Spearman's rho = 0.541; p < 0.0001), indicating that HSAs with higher degrees of localized hospital care also had a positive NPF or NPF around zero. HSAs with low localization of care, in comparison, had negative NPF with the exception of one outlier. II was, expectedly, positively correlated with NPF (Spearman's rho = 0.545; p < 0.0001).
Figure 7 Correlation of Health Utilization Indices for Swiss HSAs.
We also investigated the relationship among health utilization rates (Figure 8): nonlocal in-rate and local out-rate were, not surprisingly, uncorrelated. A positive correlation of nonlocal in-rate and net rate (Spearman's rho = 0.76; p < 0.0001) as well as a negative correlation of local out-rate and net rate (Spearman's rho = -0.56; p < 0.0001) were also expected findings. Because both graphs indicated acceptable linear association, we regressed local out-rate on net rate (regression coefficient = -0.89, t = -3.94; p < 0.0001; R2 = 0.13) and nonlocal in-rate on net rate (regression coefficient = 0.98, t = 21.65; p < 0.0001; R2 = 0.82). Interestingly, net rate is driven slightly stronger by the inflow of nonresidents into an HSA than by the outflow of HSA residents to hospitals outside as indicated by the absolute values of the regression coefficients. Also, net rate is explained more consistently by the inflow of nonresidents into an HSA than by the outflow of HSA residents to hospitals outside as indicated by R2.
Figure 8 Correlation of Health Utilization Rates (per 1000 Residents) for Swiss HSAs.
Localization of hospital care
The LI, a population-based index, gauges the tendency of patients within an HSA to use local hospitals. Plotting the cumulative population counts of HSAs against ranked LIs of HSAs visually demonstrated that 50% of the Swiss population lives in HSAs where at least 65% of hospitalizations occur locally (Figure 9). Likewise, about one fifth of the Swiss population lives in HSAs where less than 50% of hospitalizations occur locally and 25% live in HSAs with a localization of hospital care exceeding 80%.
Figure 9 Cumulative Percentages of Swiss Population according to ascending HSA LI-ranks
Discussion
Data
In 1996 federal health statistical reporting was introduced in Switzerland. Public hospitals have a duty to disclose of medical, administrative, and economic data to the Swiss Federal Statistical Office. Five years after introduction hospital participation reached 99% with approximately 85% of public hospital admissions being documented [2]. All submitted individual-level data (e.g. hospital discharges) are subjected to an algorithm to check internal validity. Reliability of data is also enhanced by standardized documentation and data collection. The high data quality and completeness of documentation ensure reliability of our results.
Methodology
We employed an established method to define hospital service areas derived from hospital discharge data [10,31,32]. An algorithm assigns residential areas to a hospital provider area in order to create populations congruent with respect to place of residence and use of hospital services.
Ideally, all hospital care for residents is provided within those residents' service area. As there are always patients seeking health care outside their region, the LI is a useful indicator of the validity of HSA definitions: High LIs are desirable and indicate congruent populations with respect to place of residence and hospital use. Likewise, patient net flow (PNF) and inflow from outside (II) should remain low. We observed moderate to high LIs and moderate to low IIs for the majority of HSAs. Nevertheless, 22 HSAs offered localized hospital care for less than 50% of residents and 15 HSAs had an inflow of patients living outside above 50%. Numerator-denominator mismatch might become a problem. It remains unclear whether the finding dispels this methodology because no hard criteria exist upon which to judge suitability. Health utilization indices may be used to judge reliability of utilization measures in the context of possible rate distortion, and to describe and understand patterns of health care use within and between HSAs.
HSAs must be sufficiently small to detect small area variation patterns. Yet, they need to be large enough to contain population sizes that will give utilization rates with acceptable reliability (i.e. sufficient number of observations). Balancing those competing interests is a challenge because area size is an important determinant of LI. Incorporating HSAs with low LIs into neighbouring HSAs to obtain a larger LI of the combined regions may fail to detect heterogeneous utilization patterns. Differences between two areas can be hidden in the overall rate. However, HSAs with low LIs may decrease the validity of per capita rates and make interpretation problematic. It needs to be shown whether a median population count of 47'273 and a median area size of 311 km2 yield reliable per-capita rates for both common and less common medical conditions and surgical procedures.
We currently have no comparable data on health utilization indices for service areas obtained through different methods or for administrative areas. The LIs of 3436 hospital service areas used in the Dartmouth Atlas Project have a slightly larger range (17.9% to 94%) and the lowest LI was 10% lower than our lowest. It should be noted, however, that utilization rates in the Dartmouth Atlas are calculated for hospital referral regions (HRR) which are aggregated HSAs. HRRs are considerably larger than HSAs, thus increasing LI. The concept of HRR is not feasible for a small country like Switzerland.
Apart from the underlying area model, sizes and shapes of service areas are determined by the availability of hospital services and the actual utilization patterns of patients. Assuming that the area model and availability of hospital services were fixed variables for our study period, HSA size, shape and distribution were ultimately a function of hospital use by the population over four years. HSAs are thus not stable geographic constructs in time as hospitals close, merge, or open. Infrastructural changes, political regulations, and altered insurance policies may also necessitate patient origin studies with more recent discharge data. The speed of these changes will indicate appropriate intervals after which HSA definitions need to be updated.
Literature
Various studies have defined health service areas using different methodological approaches [29,33,34]. Such service areas can be on a country [35], state, province [15,33] or even city [36-38] level, depending on the research question. In Europe a large body of literature has evolved around small area analysis [11,18,36-46], but to our knowledge no definition of population-based health service areas on a country level has been pursued to date. In The Netherlands the National Institute of Public Health and Environment has maintained an extensive web-based small area analysis project since 1999 [35]. Their area definitions are based on various spatial models, depending on subject matter. These areas are partly historical, not always geographically congruent, and to our knowledge not population-based.
Limitations
The study has several limitations. These can be attributed to (1) the underlying area model, (2) the intrinsic diversity of hospital utilization, and (3) the combination of discharges irrespective of medical specialty.
Census regions, our underlying area model, were specifically created for the federal health statistics by aggregating zip codes areas. Being a less aggregated area model, zip codes might have yielded more precise estimates of the influence of patient utilization upon the size and shape of HSAs. Nevertheless, their use was precluded due to data confidentiality laws. The formation of census regions aimed to achieve comparability of socio-demographic and geographic factors. Despite their aggregated nature we consider census regions a reasonably valid area model to create HSAs.
We created HSAs using hospital discharge data irrespective of diagnosis or medical specialty. We wanted to give an overall view of hospital utilization. Determining HSAs by the totality of discharges may not reflect the utilization pattern of a population with a specific diagnosis or a specific age group [7]. There is a solution to this problem: calculated LIs of discharges with a specific diagnosis can be compared to LIs that were calculated from the overall discharge data. If the two LIs are similar for a given HSA, travel patterns can be assumed to be alike and the service area model is valid. If they are meaningfully different, service areas for the specific diagnoses or patient groups may have to be created. This will be necessary for highly specialized services like cardiac, spinal, and neurosurgery. Guagliardo et al. evaluated the appropriateness of the US Medicare-based HSAs from the Dartmouth project for paediatric discharges in California using this approach. They spoke of the fit of discharge data to the area model and calculated an LI "index divergence" [7]. It still needs to be established whether there is a scientific or health planning need for such sub-speciality HSAs in Switzerland.
Implications
In the last decade the political structure of the Swiss health care system has come under scrutiny. The existence of 26 micro health systems, one for each canton, makes planning very complex. Patient movements over canton borders were estimated to exceed 13% in 2001. Nevertheless, such movements are not accounted for in the majority of planning processes. A canton-independent perspective for planning is therefore on the political agenda. In early 2004, the Cantonal Health Planning Board launched a study group of health care representatives from each canton plus federal representatives to assess novel approaches to hospital planning in Switzerland ["Arbeitsgruppe Leistungsorientierte Spitalplanung" der Gesundheitsdirektorenkonferenz]. Our institute collaborates with the above-mentioned board. HSAs are currently being reviewed for inclusion in a pending guideline on hospital planning strategies.
Conclusion
Switzerland possesses ideal data to perform patient origin studies and define population based hospital service areas. The federal health statistics was formed specifically to monitor health care performance and provide data for epidemiologic studies. Our newly defined HSAs should function as units of analysis to assess regional distribution of health care resources and measure variation in utilization rates. They offer a finer discrimination than the traditional area model – canton – thus giving planners better inside into patient movements.
We are convinced that HSAs and the variables we derived from the patient origin study will lead to a better understanding of hospital use. Health utilization indices and rates provide new information on travel patterns and hospital use. They may be used to assess the current situation for a given region and for projections of future need of resources. Thereby HSAs will help to establish more awareness of differences in the use of hospital services. National benchmarks based on variation studies may unveil possible under- or over-use of resources.
Abbreviations
GDP Gross domestic product
GIS Geographic information system
HRR Hospital referral region
HSA Hospital service area
II Inflow index
LI Localization index
MSI Market share index
NPF Net patient flow
SAA Small area analysis
Competing interests
All authors declare that they have no competing interests. The study was partially funded by the Swiss national science foundation (SNSF grant 405340-104607/1).
Authors' contributions
GK is responsible for drafting the manuscript. He obtained the hospital discharge data, carried out the statistics, and performed GIS operations. LS and MW substantially contributed to developing the final version of the manuscript. AB participated in the coordination of the study. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We would like to thank the faculty and in particular Dr. Goodman of CECS/Dartmouth college for teaching us small area analysis; the staff of the section santé and GEOSTAT of the Swiss federal statistical office for providing data and support; MicroGIS™ for providing vector data and information on spatial area models; our new colleague K. Matter-Walstra for invaluable advice; and the institute of geography of the university of Bern for support with GIS software.
==== Refs
OECD Health 2004, 3rd edition 3rd
Swiss-Federal-Statistical-Office StatSanté - Informationen über das Projekt "Statistik der stationären Betriebe des Gesundheitswesens" [German/French] 2001 Neuchâtel 3 5
Wennberg JE Dealing with medical practice variations: a proposal for action Health Affairs 1984 3 6 32 6432667 10.1377/hlthaff.3.2.6
Wennberg JE On the appropriateness of small-area analysis for cost containment.[comment] Health Affairs 1996 15 164 167 8991271 10.1377/hlthaff.15.4.164
Wennberg JE Unwarranted variations in healthcare delivery: implications for academic medical centres Bmj 2002 325 961 964 12399352 10.1136/bmj.325.7370.961
Goody B Defining rural hospital markets Health Serv Res 1993 28 183 200 8514499
Guagliardo MF Jablonski KA Joseph JG Goodman DC Do pediatric hospitalizations have a unique geography? BMC Health Serv Res 2004 4 2 14736335 10.1186/1472-6963-4-2
Glover AJ The incidence of Tonsillectomy in School Children. Proceedings of the Royal Society of Medicine 1938 31 1219 1236
Lewis CE Variations in the incidence of surgery. New England Journal of Medicine 1969 281 880 884 5812257
Wennberg J Gittelsohn Small area variations in health care delivery Science 1973 182 1102 1108 4750608
McPherson K Wennberg JE Hovind OB Clifford P Small-area variations in the use of common surgical procedures: an international comparison of New England, England, and Norway New England Journal of Medicine 1982 307 1310 1314 7133068
Chassin MR Brook RH Park RE Keesey J Fink A Kosecoff J Kahn K Merrick N Solomon DH Variations in the use of medical and surgical services by the Medicare population N Engl J Med 1986 314 285 290 3510394
Keller RB Soule DN Wennberg JE Hanley DF Dealing with geographic variations in the use of hospitals. The experience of the Maine Medical Assessment Foundation Orthopaedic Study Group Journal of Bone & Joint Surgery - American Volume 1990 72 1286 1293
Welch WP Miller ME Welch HG Fisher ES Wennberg JE Geographic variation in expenditures for physicians' services in the United States.[comment] New England Journal of Medicine 1993 328 621 627 8429854 10.1056/NEJM199303043280906
Gittelsohn A Powe NR Small area variations in health care delivery in Maryland Health Services Research 1995 30 295 317 7782218
Coyte PC Hawker G Wright JG Variations in knee replacement utilization rates and the supply of health professionals in Ontario, Canada J Rheumatol 1996 23 1214 1220 8823695
Birkmeyer JD Sharp SM Finlayson SR Fisher ES Wennberg JE Variation profiles of common surgical procedures Surgery 1998 124 917 923 9823407 10.1067/msy.1998.90268
Swart E Wolff C Deh S Häufigkeit und kleinräumige Variabilität von Operationen Chirurg 2000 71 109 114 10663014 10.1007/s001040051024
Wennberg JE Factors governing utilization of hospital services Hospital Practice 1979 14 115 21, 126-7 478525
Knickman JR Foltz AM Regional differences in hospital utilization. How much can be traced to population differences? Med Care 1984 22 971 986 6503399
Wennberg J Freeman JL Culp WJ Are hospital services rationed in New Haven and over-utilized in Boston? Lancet 1987 1 1185 1189 2883497 10.1016/S0140-6736(87)92152-0
Paul-Shaheen P Clark JD Williams D Small area analysis: a review and analysis of the North American literature J Health Polit Policy Law 1987 12 741 809 3323295
Stano M Evaluating the policy role of the small area variations and physician practice style hypotheses Health Policy 1993 24 9 17 10125815 10.1016/0168-8510(93)90084-3
McMahon LFJ Wolfe RA Griffith JR Cuthbertson D Socioeconomic influence on small area hospital utilization Med Care 1993 31 YS29 36 8492582
Veugelers PJ Hornibrook S Small area comparisons of health: applications for policy makers and challenges for researchers Chronic Dis Can 2002 23 100 110 12443566
Pekoz EA Shwartz M Iezzoni LI Ash AS Posner MA Restuccia JD Comparing the importance of disease rate versus practice style variations in explaining differences in small area hospitalization rates for two respiratory conditions Stat Med 2003 22 1775 1786 12720310 10.1002/sim.1398
Detsky AS Regional variation in medical care N Engl J Med 1995 333 589 590 7623911 10.1056/NEJM199508313330911
Makuc DM Haglund B Ingram DD Kleinman JC Feldman JJ The use of health service areas for measuring provider availability J Rural Health 1991 7 347 356 10116027
Garnick DW Luft HS Robinson JC Tetreault J Appropriate measures of hospital market areas Health Serv Res 1987 22 69 89 3570813
Goodman DC Wennberg JE Maps and health: the challenges of interpretation Journal of Public Health Management & Practice 1999 5 xiii xvii 10538409
Wennberg JE Gittelsohn A A small area approach to the analysis of health system performance Health Planning Methods and Technology Performance 1980 266
Goodman DC Mick SS Bott D Stukel T Chang CH Marth N Poage J Carretta HJ Primary care service areas: a new tool for the evaluation of primary care services Health Serv Res 2003 38 287 309 12650392 10.1111/1475-6773.00116
Ricketts TC Randolph R Howard HA Pathman D Carey T Hospitalization rates as indicators of access to primary care Health Place 2001 7 27 38 11165153 10.1016/S1353-8292(00)00035-6
Zwarenstein M Krige D Wolff B The use of a geographical information system for hospital catchment area research in Natal/KwaZulu S Afr Med J 1991 80 497 500 1948466
Zwakhals SL Giesbers H Mac Gillavry E van Boven PF van der Veen AA The Dutch National Atlas of Public Health Bundesgesundheitsblatt Gesundheitsforschung Gesundheitsschutz 2004 47 882 889 15378176
Rytkonen M Rusanen J Nayha S Small-area variation in mortality in the city of Oulu, Finland, during the period 1978--1995 Health Place 2001 7 75 79 11470221 10.1016/S1353-8292(00)00040-X
Benach J Yasui Y Borrell C Rosa E Pasarin MI Benach N Espanol E Martinez JM Daponte A Examining geographic patterns of mortality: the atlas of mortality in small areas in Spain (1987-1995) Eur J Public Health 2003 13 115 123 12803409 10.1093/eurpub/13.2.115
Garcia-Gil C Cruz-Rojo C Alvarez-Giron M Solano-Pares A Health inequalities in Seville, Spain: use of indicators of social deprivation and mortality in small areas Public Health 2004 118 11 20 14643623 10.1016/S0033-3506(03)00141-0
Lichtner S Pflanz M Appendectomy in the Federal Republic of Germany: epidemiology and medical care patterns. Medical Care 1971 9 311 330 5562429
Mackenbach JP Kunst AE Looman CW Habbema JD van der Maas PJ Regional differences in mortality from conditions amenable to medical intervention in The Netherlands: a comparison of four time periods Journal of Epidemiology & Community Health 1988 42 325 332 3256573
Mackenbach JP Kunst AE Looman CW Cultural and economic determinants of geographical mortality patterns in The Netherlands Journal of Epidemiology & Community Health 1991 45 231 237 1757767
Kunst AE Looman CW Mackenbach JP Determinants of regional differences in lung cancer mortality in The Netherlands Social Science & Medicine 1993 37 623 631 8211276 10.1016/0277-9536(93)90101-9
Ubido J Ashton J Small area analysis: abortion statistics Journal of Public Health Medicine 1993 15 137 143 8353002
Katalinic A Bartel C Uhlenkamp T Raspe H [Developing a small-area cancer atlas: process, validity and possible applications] Gesundheitswesen 1999 61 436 438 10535228
Gatrell A Lancaster G Chapple A Horsley S Smith M Variations in use of tertiary cardiac services in part of North-West England Health Place 2002 8 147 153 12135638 10.1016/S1353-8292(01)00044-2
Schober E Rami B Waldhoer T Small area variation in childhood diabetes mellitus in Austria: links to population density, 1989 to 1999 J Clin Epidemiol 2003 56 269 273 12725882 10.1016/S0895-4356(02)00607-8
| 15882463 | PMC1131901 | CC BY | 2021-01-04 16:31:49 | no | BMC Health Serv Res. 2005 May 9; 5:33 | utf-8 | BMC Health Serv Res | 2,005 | 10.1186/1472-6963-5-33 | oa_comm |
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-5-321588245310.1186/1471-2334-5-32Research ArticlePrevalence of oropharyngeal beta-lactamase-producing Capnocytophaga spp. in pediatric oncology patients over a ten-year period Jolivet-Gougeon Anne [email protected] Zohreh [email protected] Laurent [email protected] Virginie [email protected] Jean-Louis [email protected] Nolwenn [email protected] Michel [email protected] Martine [email protected] Equipe de Microbiologie, UPRES-EA 1254, Faculté des Sciences Pharmaceutiques et Biologiques, Université de Rennes 1, 2 avenue du Professeur Léon Bernard, 35043 Rennes, France2 Pediatric Oncology Department, CHU Hôpital Sud, 16 boulevard de Bulgarie, 35000 Rennes, France2005 9 5 2005 5 32 32 5 11 2004 9 5 2005 Copyright © 2005 Jolivet-Gougeon et al; licensee BioMed Central Ltd.2005Jolivet-Gougeon 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 aim of this study was to evaluate the prevalence of beta-lactamase-producing Capnocytophaga isolates in young children hospitalized in the Pediatric Oncology Department of Hôpital Sud (Rennes, France) over a ten-year period (1993–2002).
Methods
In neutropenic children, a periodic survey of the oral cavity allows a predictive evaluation of the risk of systemic infections by Capnocytophaga spp. In 449 children with cancer, 3,053 samples were collected by oral swabbing and plated on TBBP agar. The susceptibility of Capnocytophaga isolates to five beta-lactams was determined.
Results
A total of 440 strains of Capnocytophaga spp. were isolated, 309 (70%) of which were beta-lactamase producers. The beta-lactamase-producing strains were all resistant to cefazolin, 86% to amoxicillin, and 63% to ceftazidime. The proportion of strains resistant to third-generation cephalosporins remained high throughout the ten-year study, while susceptibility to imipenem and amoxicillin combined with clavulanic acid was always conserved.
Conclusion
These results highlight the risk of antibiotic failure in Capnocytophaga infections and the importance of monitoring immunosuppressed patients and testing for antibiotic susceptibility and beta-lactamase production.
==== Body
Background
Capnocytophaga spp. are capnophilic, gram-negative fusiform rods with gliding motility, common inhabitants of the oral cavity, but their role as an etiologic agent in juvenile periodontitis remains controversial [1]. In immunocompromised granulocytopenic patients, a number of complications, including septicemia, endocarditis, and peripheric lesions, have been reported [2-4]. Several authors have described cases of infections by Capnocytophaga strains resistant to antimicrobial agents [5-8]. An episode of bacteremia can be the consequence of bacterial translocation from oral flora. In the absence of bacterial strain isolation from blood cultures, empiric treatment might be adjusted according to the susceptibility of strains isolated during the survey. A sequential survey of the oral cavity during hospitalization was performed to evaluate the prevalence of beta-lactamase-producing Capnocytophaga isolates in children hospitalized in the Pediatric Oncology Department of Hôpital Sud (Rennes, France) over a ten-year period (1993–2002).
Methods
Samples were collected by swabbing the bucco-pharyngeal area of children hospitalized in the Pediatric Oncology Department of Hôpital Sud (Rennes, France) [9]. Samples were taken periodically with a minimum of 15 days between each collection. All cancer patients from one to 17 years were included whatever the type of cancer, chemotherapy, and clinical situation were. The number of samples taken per child depended on the oncological disease (Acute Lymphoblastic Leukemia, Acute Myeloblastic Leukemia, others), the number of cures, and the number and length of hospitalizations or treatments. Each sample was inoculated on TBBP agar plates [4% trypticase soy agar supplemented with 5% sheep blood, 0.1% yeast extract (AES Laboratory, France), 100 μg/ml polymyxin and 50 μg/ml bacitracin (Sigma)] [10], which was then incubated for two to five days in a 10% CO2 atmosphere. Isolates were identified on the basis of colony morphology, Gram staining, negative catalase and oxidase reactions, API ZYM (BioMérieux, France) [11], and fatty acid profiles (gas chromatography, SHERLOCK Microbial Identification System™ MIDI Inc., Newark, DE, USA). Beta-lactamase production was tested using the qualitative chromogenic cephalosporin disk test (Cefinase®, BBL Microbiology Systems, Cockeysville, MD, USA). The results were read after 30 minutes.
Susceptibility testing was determined by standard methods and break points using the criteria of Bremmelgaard et al. [12] for screening determinations, and NCCLS [13] for intermediate/resistant strains. Minimal Inhibitory Concentrations (MICs) were confirmed by the E-test method (AES Laboratory, Combourg, France) using the same incubation conditions. Researchers tested the following antibiotics: amoxicillin, amoxicillin combined to clavulanic acid, cefazolin, ceftazidime, and imipenem.
Results
Over the ten-year period of this study, researchers analyzed 3,053 samples from 449 hospitalized children (266 males and 183 females). Capnocytophaga spp (440 strains) were isolated in 232 children, on TBBP agar and identified with conventional methods. The annual percentage of children, who carried a Capnocytophaga strain at least once varied, with a minimum from 1995 to 1997 (17% and 22%, respectively), and maxima in 1993 (52%), 2001 (58%), and 2002 (61%). These results were also observed studying the number of Capnocytophaga isolates in the same periods: the number of Capnocytophaga isolates changed, with a minimum (11, 6%) in 1996 and a maximum (96, 22%) in 2001, without modification to the isolation and culture techniques (Table 1).
Table 1 Prevalence of Capnocytophaga spp. strains in periodic oral samples from children hospitalized in the Pediatric Oncology Department of Hôpital Sud (Rennes, France) from 1993 to 2002
Year Number of children included Number of Capno1 carriers2 (%) Number of samples collected Number of Capno isolates (%)
1993 60 31 (52) 305 56 (18)
1994 50 24 (48) 228 49 (21)
1995 45 10 (22) 195 14 (7)
1996 49 8 (16) 198 11 (6)
1997 58 9 (17) 251 13 (5)
1998 62 24 (39) 296 41 (14)
1999 68 24 (35) 267 38 (14)
2000 84 17 (20) 347 31 (9)
2001 85 49 (58) 429 96 (22)
2002 80 49 (61) 537 91 (17)
Total 449 232 (52) 3,053 440 (14)
1: Capnocytophaga
2: Number of children who were Capnocytophaga spp. carriers at least once during the studied period.
Whatever the number of isolates, there was a high prevalence of beta-lactamase-producing strains (70%), with a varying incidence from 54% in 1996 (based on three out of 49 children) to 78% in 1995 (similarly, 11/14 isolates describes 8/45 children) (Figures 1 and 2). The susceptibility of Capnocytophaga strains was always conserved with imipenem (MIC < 4 μg/ml) and amoxicillin combined with clavulanic acid (MIC < 4/2 μg/ml). All beta-lactamase producing strains were uniformly resistant to cefazolin (MIC > 8 μg/ml), but the authors noted different levels of resistance to amoxicillin and ceftazidime (Table 2).
Figure 1 Prevalence of beta-lactamase producing Capnocytophaga spp. strains in periodic oral samples from children hospitalized in the Pediatric Oncology Department of Hôpital Sud (Rennes, France) from 1993 to 2002.
Figure 2 Percentage of positive beta-lactamase Capnocytophaga spp. carriers, from children hospitalized in the Pediatric Oncology Department of Hôpital Sud (Rennes, France) from 1993 to 2002.
Table 2 Susceptibility of oral Capnocytophaga spp. isolated from 1993 to 2002 (Hôpital Sud, Rennes, France). Susceptibility testing was determined by standard methods and break points according to the criteria of NCCLS [13] and Bremmelgaard et al. [12]
Year Number of Resistant or Intermediate Capnocytophaga strains (%) (Total number of Capno strains = 440)
Amoxicillin MIC1> 4 μg/ml Ceftazidime MIC > 4 μg/ml
1993 31 (88) 28 (80)
1994 31 (84) 28 (75)
1995 11 (100) 11 (100)
1996 6 (100) 5 (83)
1997 10 (100) 9 (90)
1998 27 (96) 20 (71)
1999 23 (85) 10 (37)
2000 16 (84) 12 (63)
2001 54 (80) 38 (56)
2002 56 (81) 34 (49)
Total 265 (86) 195 (63)
Note: No strains were resistant to an imipenem and amoxicillin/clavulanic acid combination, and all beta-lactamase-producing strains were resistant to cefazolin.
1: Minimal Inhibitory Concentration (μg/ml).
Discussion
In immunocompromised children, a periodic survey of the oral cavity during hospitalization allows a predictive evaluation of the risk of systemic infections by Capnocytophaga spp.
In this study, the authors correlated the number of Capnocytophaga isolates with the number of child carriers of Capnocytophaga, indicating that protocol was correctly conducted. A decrease in incidence was observed during the 1995–1997 period, which could be due to use, in first-line antibiotic protocols, of beta-lactamase inhibitor combinations. During the other time periods, extended-spectrum antibiotic were given. Aminoglycosides were always associated to beta-lactam antibiotics in all protocols from 1993 to 2002. Incidence rates of total extended-spectrum beta-lactamase producing bacteria (ESBL) in gram-negative rods responsible for infection in Hôpital Sud (Rennes, France), calculated for 1,000 days of hospitalization, varied in 2002 from 0.04 to 0.7 depending on the department (higher rate in reanimation units) . Interestingly, all Capnocytophaga spp. strains collected in this study appeared as colonizing strains, because neither bacteremia, nor other systemic infections, were observed during this study. Results of MIC determinations agree with previous works reporting that beta-lactamases confer a high degree of resistance to a wide range of beta-lactam antibiotics [14] while having no effect on imipenem and beta-lactamase-inhibitor combinations [15,16]. In a Canadian study, Roscoe et al. [17] reported that 36% of the strains collected mainly from clinical sources were beta-lactamase producers, while in a study in Taiwan, Lin et al. [3] reported that 18% of Capnocytophaga strains isolated from patients with bacteremia were beta-lactam resistant. In our study, the prevalence of beta-lactamase producers did not increase linearly during the study period.
A first hypothesis to explain these changes in incidence of beta-lactamase production could be the impact of current and previous therapy especially antimicrobial treatments or pathology, on the oral carriage of Capnocytophaga (study in progress). In a previous French study, Maury et al. [18] reported a high prevalence of beta-lactamase-producing Capnocytophaga species (75%), which they associated with previous beta-lactam treatments. Even if Capnocytophaga spp. belongs to the so-called late colonizers in the normal flora, meaning that it is more frequently found in children at 12 months and later, in a follow-up study of healthy infants from the age of 2 to 12 months, [19] Nyfors et al. reported a positive correlation between antimicrobial exposure and beta-lactamase production in oral anaerobic gram-negative species, while reporting only two beta-lactamase-producing C. ochracea isolates. In a recent contradictory study, [20]Capnocytophaga spp. has been detected in 1.9% of all episodes of fever and neutropenia before antibiotic therapy, versus 0.3% during antimicrobial treatment.
A second hypothesis to explain these changes in incidence of beta-lactamase production could be linked to the immediate environment and close contact. These conditions have a great influence on the composition of the flora, and coming and going between home and hospital could modify the children's oral ecosystem. A turnover of the bacterial population from beta-lactamase-positive to beta-lactamase-negative strains may occur in young children with a developing oral ecosystem.
Another explanation of this high prevalence of beta-lactamase production could be conferred on transfer of encoding-resistance genes. Anaerobic bacteria are known to be able to exchange genetic material with aerobic bacteria, even though antibiotic-resistance genes are expressed differently in aerobic and anaerobic bacteria [21]. The spread of antibiotic resistance can also play a great part in nosocomial infections in neutropenic patients [22]. The high rate of resistance to third-generation cephalosporins observed in this study could be due to the dissemination of an epidemic clone. Some ceftazidime-resistant strains have the same susceptibility and plasmid profiles (data not shown). Other authors have described resistant clinical isolates [5,6,14,16], but Rosenau et al. [8] were the first to characterize a plasmid-encoded TEM extended-spectrum beta-lactamase in Capnocytophaga spp. Clavulanate-sensitive cephalosporinases belonging to class A group 2e from the classification of Bush et al. [23] have recently been described [7,24]. They differ by only one or two substitutions in their sequences (cfxA, Genbank accession No. U75371; cfxA2, Genbank accession No. AF118110; cfxA3, Genbank accession No. AF472622). Large transposons encode all the functions needed for their own conjugation and for resistance to antimicrobial agents, including third-generation cephalosporins.
Conclusion
These results show the importance of testing for the antibiotic susceptibility of and beta-lactamase production by clinical Capnocytophaga strains, in which beta-lactamase production has become very common. In neutropenic patients, resistance to third-generation cephalosporins should be taken into consideration upon hospitalization to adapt the empiric antimicrobial treatment previously dispensed.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
AJG carried out the microbiological studies and wrote the manuscript. ZTS, LD and NMG participated in the microbiological studies. VG was in charge of the clinical studies. Provision of advice was given by JLS, MC and MBM. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We would like to thank Gene Bourgeau and Céline Allaire for editorial assistance.
This work was supported by the Fondation Langlois and Conseil Régional de Bretagne.
==== Refs
Nonnenmacher C Mutters R de Jacoby LF Microbiological characteristics of subgingival microbiota in adult periodontitis, localized juvenile periodontitis and rapidly progressive periodontitis subjects Clin Microbiol Infect 2001 7 213 217 11422244 10.1046/j.1469-0691.2001.00210.x
Buu-Hoi AY Joundy S Acar JF Endocarditis caused by Capnocytophaga ochracea J Clin Microbiol 1988 26 1061 1062 3384900
Lin RD Hsueh PR Chang SC Luh KT Capnocytophaga bacteremia: clinical features of patients and antimicrobial susceptibility of isolates J Formos Med Assoc 1998 97 44 48 9481064
Parenti DM Snydman DR Capnocytophaga species: infections in nonimmunocompromised and immunocompromised hosts J Infect Dis 1985 151 140 147 3965585
Arlet G Sanson-Le Pors MJ Casin IM Ortenberg M Perol Y In vitro susceptibility of 96 Capnocytophaga strains, including a beta-lactamase producer, to new beta-lactam antibiotics and six quinolones Antimicrob Agents Chemother 1987 31 1283 1284 3498438
Gomez-Garces JL Alos JI Sanchez J Cogollos R Bacteremia by multidrug-resistant Capnocytophaga sputigena J Clin Microbiol 1994 32 1067 1069 8027314
Jolivet-Gougeon A Tamanai-Shacoori Z Desbordes L Burggraeve N Cormier M Bonnaure-Mallet M Genetic analysis of an ambler class A extended-spectrum beta-lactamase from Capnocytophaga ochracea J Clin Microbiol 2004 42 888 890 14766881 10.1128/JCM.42.2.888-890.2004
Rosenau A Cattier B Gousset N Harriau P Philippon A Quentin R Capnocytophaga ochracea: characterization of a plasmid-encoded extended-spectrum TEM-17 beta-lactamase in the phylum Flavobacter-bacteroides Antimicrob Agents Chemother 2000 44 760 762 10681352 10.1128/AAC.44.3.760-762.2000
Bonnaure-Mallet M Bunetel L Tricot-Doleux S Guerin J Bergeron C LeGall E Oral complications during treatment of malignant diseases in childhood: effects of tooth brushing Eur J Cancer 1998 34 1588 1591 9893633 10.1016/S0959-8049(98)00169-5
Mashimo PA Yamamoto Y Nakamura M Slots J Selective recovery of oral Capnocytophaga spp. with sheep blood agar containing bacitracin and polymixin B J Clin Microbiol 1983 17 187 191 6833474
Slots J Enzymatic characterization of some oral and nonoral gram-negative bacteria with the API ZYM system J Clin Microbiol 1981 14 288 294 7026598
Bremmelgaard A Pers C Kristiansen JE Korner B Heltberg O Frederiksen W Susceptibility testing of Danish isolates of Capnocytophaga and CDC group DF-2 bacteria APMIS 1989 97 43 48 2914105
NCCLS Methods for antimicrobial susceptibility testing of anaerobic bacteria: approved standard 1997 4th edition Wayne , NCCLS M11 A14
Jolivet-Gougeon A Buffet A Dupuy C Sixou JL Bonnaure-Mallet M David S Cormier M In vitro susceptibilities of Capnocytophaga isolates to beta-lactam antibiotics and beta-lactamase inhibitors Antimicrob Agents Chemother 2000 44 3186 3188 11036049 10.1128/AAC.44.11.3186-3188.2000
Rummens JL Gordts B Van Landuyt HW In vitro susceptibility of Capnocytophaga species to 29 antimicrobial agents Antimicrob Agents Chemother 1986 30 739 742 3800350
Foweraker JE Hawkey PM Heritage J Van Landuyt HW Novel beta-lactamase from Capnocytophaga sp Antimicrob Agents Chemother 1990 34 1501 1504 2221858
Roscoe DL Zemcov SJ Thornber D Wise R Clarke AM Antimicrobial susceptibilities and beta-lactamase characterization of Capnocytophaga species Antimicrob Agents Chemother 1992 36 2197 2200 1444299
Maury S Leblanc T Rousselot P Legrand P Arlet G Cordonnier C Bacteremia due to Capnocytophaga species in patients with neutropenia: high freqency of b-lactamase-producing strains Clin Infect Dis 1999 28 1172 1174 10452663
Nyfors S Kononen E Takala A Jousimies-Somer H Beta-lactamase production by oral anaerobic gram-negative species in infants in relation to previous antimicrobial therapy Antimicrob Agents Chemother 1999 43 1591 1594 10390208
Ammann RA Hirt A Luthy AR Aebi C Predicting bacteremia in children with fever and chemotherapy-induced neutropenia Pediatr Infect Dis J 2004 23 61 67 14743049
Guiney DG Hasegawa P Davis CE Plasmid transfer from Escherichia coli to Bacteroides fragilis: differential expression of antibiotic resistance phenotypes Proc Natl Acad Sci USA 1984 81 7203 7206 6095273
Girlich D Poirel L Leelaporn A Karim A Tribuddharat C Fennewald M Nordmann P Molecular epidemiology of the integron-located VEB-1 extended-spectrum beta-lactamase in nosocomial enterobacterial isolates in Bangkok, Thailand J Clin Microbiol 2001 39 175 182 11136767 10.1128/JCM.39.1.175-182.2001
Bush K Jacoby GA Medeiros AA A functional classification scheme for beta-lactamases and its correlation with molecular structure Antimicrob Agents Chemother 1995 39 1211 1233 7574506
Parker AC Smith CJ Genetic and biochemical analysis of a novel Ambler class A beta-lactamase responsible for cefoxitin resistance in Bacteroides species Antimicrob Agents Chemother 1993 37 1028 1036 8517690
| 15882453 | PMC1131902 | CC BY | 2021-01-04 16:28:14 | no | BMC Infect Dis. 2005 May 9; 5:32 | utf-8 | BMC Infect Dis | 2,005 | 10.1186/1471-2334-5-32 | oa_comm |
==== Front
BMC Med EducBMC Medical Education1472-6920BioMed Central London 1472-6920-5-141588245810.1186/1472-6920-5-14Research ArticleCardiac auscultation training of medical students: a comparison of electronic sensor-based and acoustic stethoscopes Høyte Henning [email protected] Torstein [email protected] Knut [email protected] University of Oslo, Faculty of Medicine, Oslo, Norway2 Department of Cardiology, Ullevål University Hospital, Oslo, Norway2005 9 5 2005 5 14 14 1 3 2005 9 5 2005 Copyright © 2005 Høyte et al; licensee BioMed Central Ltd.2005Høyte 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 whether the use of an electronic, sensor based stethoscope affects the cardiac auscultation skills of undergraduate medical students.
Methods
Forty eight third year medical students were randomized to use either an electronic stethoscope, or a conventional acoustic stethoscope during clinical auscultation training. After a training period of four months, cardiac auscultation skills were evaluated using four patients with different cardiac murmurs. Two experienced cardiologists determined correct answers. The students completed a questionnaire for each patient. The thirteen questions were weighted according to their relative importance, and a correct answer was credited from one to six points.
Results
No difference in mean score was found between the two groups (p = 0.65). Grading and characterisation of murmurs and, if present, report of non existing murmurs were also rated. None of these yielded any significant differences between the groups.
Conclusion
Whether an electronic or a conventional stethoscope was used during training and testing did not affect the students' performance on a cardiac auscultation test.
==== Body
Background
The French physician Rene Laennec invented the first stethoscope in 1816[1]. The use of a modified version was widespread among physicians in the 1830's[2], and the currently used binaural models were designed in the 1870's. Since then a number of attempts to improve the stethoscope have been made, the most recent being the advent of electronic sound transmission. "TheStethoscope®" is a sensor based electronic stethoscope introduced in 1999 by Meditron (Asker, Norway) in cooperation with Welch Allyn (Skaneateles Falls, USA). The stethoscope is equipped with a pressure sensitive sensor, and the signals are converted into sound waves. It is also equipped with a volume regulator and a possibility to apply frequency filtering. The filter has three modes, enhancing low (20–800 Hz), high (200–20000 Hz) or all frequencies. It can be connected to external devices (PC/co-listening unit) allowing recording or sharing of auscultatory findings.
Electronic stethoscopes offer potential advantages compared to conventional pneumatic stethoscopes[3], and several of the features unique to electronic stethoscopes could influence the performance in cardiac auscultation. The high sound quality, the possibility of applying personal adjustments to frequency[4] and volume, and education by simultaneous auscultation could improve the performance on a cardiac auscultation test. The volume regulator could also prove beneficial to students and doctors with organic hearing problems. Electronic stethoscopes are, however, sensitive to manipulation artefacts as well as electronic and ambient noise. The sound picture from an electronic stethoscope is also quite different from a conventional stethoscope, requiring training. Thus, some of the features could possibly influence the performance negatively. The volume adjuster is step-less, which could give rise to problems when grading the intensity of murmurs. The increased sensitivity to ambient noise and noise from handling of the stethoscope could increase the report of false murmurs, and lead to inaccurate characterisation of murmurs.
The aim of this study was to compare the auscultation skills of medical students using an electronic, sensor based stethoscope with a similar group using conventional stethoscopes.
Methods
Study population
The trial was conducted at Ullevål University Hospital (UUH) during the autumn 2001 and spring 2002, using third year medical students at the University of Oslo. Teaching groups, each comprising 6–8 students, were randomised to use either the electronic stethoscope (intervention group) or their own acoustic stethoscopes (control group) during a four month training period. A total of 48 students were enrolled, 24 in each group.
The teaching of cardiac auscultation
The students at the University of Oslo are introduced to cardiac auscultation during propedeutic clinical courses in the second year, and more extensively during rotations in cardiology in the third year. In addition to the regular course program the students in our trial received a two hour lecture and four hours of clinical bedside teaching. The intervention group using used simultaneous auscultation during bedside teaching.
The auscultation test
The students' auscultation skills were tested on patients at the university hospital. Each student completed a questionnaire (mainly multiple choice questions) on auscultation findings for each patient (table 1). Next to each patient was a brief survey of the patient's presenting complaints, and the patients were instructed not to reveal their diagnoses. The students were allotted ten minutes to examine each patient. They were alone with the patients during examination, and were instructed not to discuss their findings with any other student. A total of ten patients participated, one of them twice. They were recruited either from the ward or from the clinic's outpatient population. (Table 2)
Table 1 The questionnaire. shows the questionnaire the students had to complete.
Question No: Alternatives Points
1 Do you hear any murmur? Yes/no 6
2 If so, is the murmur: Systolic/ Diastolic /both 5
3 If you have heard a systolic murmur, how would you characterise it? Holosystolic/Crescendo-decrescendo 2
4 Describe the quality of the systolic murmur: 3
5 Where is the murmur loudest? Anatomical alternatives 3
6 Grade 1 – 6 4
7 Radiation? Anatomical alternatives 4
8 If you have heard a diastolic murmur, how would you characterise it? Rumbling, whistling etc. 3
9 Is the 2nd heart tone preserved? preserved/ diminished/ not audible 4
10 Is the 2nd heart tone constantly split? Yes/ no 1
11 Is a third heart tone present? Yes/ no 1
12 What is the most likely cause of the murmur? Options 2
13 Any comments? Max 2
Table 2 The patients' diagnoses and findings on auscultation. shows the patients relevant cardiac diagnoses, and the findings on auscultation as reported by two cardiologists.
Patient nr 1 Diagnoses Findings*
Day 1: Patient no 1 (1) Mitral valve insufficiency, possible low grade aortic sclerosis Holosystolic, rough high frequency, grade 3/6, max intensity in the fourth left intercostal space in the medioclavicular line with radiation to the left axilla. 2nd heart sound (S2) preserved, no constant splitting of S2. S3 not present.
Patient no 2 (2) Aortic stenosis Holosystolic rough murmur grade 4/6, loudest at the apex. Radiation to the left axilla and the carotid arteries. S2 weakened.
Patient no 3 (3) Control patient without cardiac murmurs -
Patient no 4 (4) Aortic insufficiency High frequency diastolic murmur distinct at the apex.
Day 2: Patient no 1 (1) Same as day 1 Same as day 1.
Patient no 2 (5) Control without murmur -
Patient no 3 (6) VSD Holosystolic ejection sound grade 3/6, loudest at the apex. No radiation. S3 present.
Patient no 4 (7) Mitral valve insufficiency and aortic sclerosis Holosystolic ejection sound grade 4/6. loudest at the apex, radiation to the left axilla and the carotid artery. S3 absent.
Day 3: Patient nr 1 (8) St. Jude prosthetic valve in mitral position, paravalvular leakage Holosystolic rough murmur grade 2, loudest in the left axilla.
Patient nr 2 (9) Control without murmurs -
Patient no 3 (10) Aortic stenosis Crescendo/decrescendo quincking systolic murmur grade 3/6, loudest parasternally at the right second intercostal space. Radiation to the left axilla and to the carotid arteries.
*The findings on auscultation as reported by the two cardiologists.
Scoring
The correct answers on the questionnaire were defined by consensus of two cardiology consultants who examined the patients with acoustic stethoscopes on the same day as the students were tested. Each questionnaire was interpreted and scored blindly by one person. When there was doubt about scoring, the questionnaire was in addition evaluated by a second person, and consensus was reached. A correct response to each of the questions was rewarded by a predefined number of points, ranging from one to six (table 1). The points obtained on each question were added, a total score for the questionnaire calculated, and total and average scores were obtained for each group.
Based on our own experiences with this electronic stethoscope, we also wanted to test whether there were differences between the two groups' regarding grading and characterising murmurs, and report of non-existing murmurs.
Different scores were allotted for each patient and the mean max score that could be obtained differed. For all days combined the mean maximum score was 25.3 points.
Statistics
Data are reported as means with confidential intervals (CI) or range. Differences between the study groups were evaluated using Student's t-test. When comparing categorical data Chi-square tests were used. Calculations regarding group size and statistical power were done in retrospect. The reason for this was the difficulty of estimating the standard deviation (SD) prior to the trial. Each group comprised 24 students, providing 80% power of detecting a difference of seven points between the study groups (SD = 8.6). P-values are two-sided, and values <0.05 are regarded significant.
Results
Each student contributed three or four questionnaires (depending on the day of participation). Forty-one of the students (85%) completed the trial. (Two had exrolled from the faculty, one was abroad, and four did not meet for other reasons. Four of these students were from the group using the electronic stethoscope.) Three questionnaires were incomplete and excluded from final analysis.
The number of questionnaires scored was 78 and equal in the two groups. The total score in the control group was 1341.5 points versus 1388.5 in the intervention group. Mean scores in the control group and intervention group were 17.2 (SD = 8.7, range 1–30.5) and 17.8 (SD = 8.8, range 0–35) points respectively. The difference is 0.6 points with a 95% CI of (-0.33 – 1.53) points (p = 0.65). (Figure 1)
Figure 1 The number of students distributed on intervals of points.
When grading the murmurs, the students using conventional stethoscopes had 27 correct and 27 incorrect responses, whereas the students using the electronic stethoscope had 29 correct and 22 incorrect responses (p = 0.47). On characterising murmurs the group using conventional stethoscopes had 45 correct and 66 incorrect responses, while the students using the electronic stethoscope had 42 correct and 66 incorrect responses (p = 0.76). When tested for report of false murmurs, the group using conventional stethoscopes had 12 correct and 8 incorrect responses. The group using the electronic stethoscope had 11 correct and 9 incorrect responses (p = 0.75). (Table 3)
Table 3 Overview of the results in the two groups. Per cent correct answers. The results in the intervention and control group reported as % correct answers.
Electronic stethoscope Control
Total score (mean ± SD) 17.8 ± 8.8 17.2 ± 8.7
Grading of murmurs, % correct 29 correct of 51 answers; 43% 27 correct of 54 answers; 50%
Characterization of murmurs, % correct. 42 correct of 109 answers; 39% 45 correct of 111 answers; 41%
Report of false murmurs, % correct. 11 correct of 20 answers; 55% 12 correct of 20 answers; 60%
Discussion
The aim of our study was to determine if the use of an electronic stethoscope would influence cardiac auscultation skills of undergraduate medical students. To investigate this we compared the performance on a cardiac auscultation test of a group of medical students using conventional stethoscopes to a group using electronic stethoscopes. No differences between the study groups were found, neither for general nor for more specific skills in cardiac auscultation, such as grading and characterising murmurs, and report of non existing murmurs. We are not aware of any similar studies comparing electronic and acoustic stethoscopes.
Several factors strengthen our results. We used a prospective, randomized study design with test conditions nearly similar to a regular clinical setting. The opportunity to gather anamnestic information was limited, as we did not want any of the patients to reveal their diagnoses. A note, however, next to each patient summarised the relevant clinical data. Thus the students received exactly the same information. 85% of the included students completed the test, which is a high percentage considering the logistic difficulties inherent in use of patient and student volunteers.
The results in the two groups are strikingly similar, both for the main (figure 1) and for the additional hypotheses tested (table 3). It is therefore unlikely that the limited size of the groups has introduced a type II error. The size of the groups gives an 80% chance of discovering a difference between the groups of seven points. A difference of seven points is quite much, 40% of the average total score. The question can be raised if detection of a smaller difference between the groups would be clinically relevant.
It can be objected that our diagnoses were based on auscultation and not verified by echocardiography. However, we were primarily testing auscultatory findings and not diagnostic interpretation. We justify the use of the cardiologists' auscultatory findings as a gold standard for the students since one should not expect that the students would have greater auscultatory proficiency than the cardiologists [5-7]. Some of the patients used in the auscultation test were, however, known to the cardiologists, and there is a possibility that their findings on auscultation could be biased by background information about these patients.
Simultaneous auscultation was used as an additional teaching tool in our study. The instructors did not, however, receive any special training in the use of this new technique. This might have limited the opportunity to get full effect of this method. It is also possible that the intensity of the teaching intervention was insufficient, and that additional hours of teaching using simultaneous auscultation would have improved the cardiac auscultation skills selectively.
We chose to test our main hypothesis using a system of points for each question. Wrong answers on questions testing the most important auscultatory skills lead to a loss of most points, whereas incorrect answers on questions testing more advanced skills were punished less[8]. As an example a student who was unable to separate a systolic from a diastolic murmur lost five points, whereas a student missing a third heart tone lost only one point[9]. A correct diagnosis gave only two points, as this also tests other skills not related to the stethoscope used. The weighting of the questions was done prior to the test. We attempted to accommodate what we could expect from students at this level and that we tested the students' ability to report auscultatory findings, not their skills in general clinical examination.
When using points to grade the question it is of importance that the groups are evenly distributed on the patients. Not all the questions are applicable to all the patients, and the maximum number of points achievable varied between the patients (table 4). As seen in table 5, the two groups are evenly distributed on the test days and thus on our test patients.
Table 4 The mean scores in the groups for the different patients.
Patient no. 1 2 3 4 5 6 7 8 9 10
Mean scores:
"The Stethoscope®": 20 26 7.6 18.9 5.8 19 27 14.5 7 21
Control group: 21.1 26 4.8 18.4 6.8 17 24 10.7 7.3 27
Table 5 The distribution of the groups on the test days. Shows the number of students present from each group the different test days.
Group Day 1 Day 2 Day 3
"The stethoscope®" n = 7 n = 12 n = 1
Control n = 5 n = 12 n = 4
Each student was represented by three or four questionnaires (depending on the day of participation), and each questionnaire was treated as an independent variable in the statistical analysis. This is likely to underestimate the spread in the groups, but the averages, and thus the comparison of the two groups, are not affected.
The students received the electronic stethoscopes four months prior to the auscultation test. This should be sufficient time to get accustomed to the electronic stethoscope, although it is possible that a longer period is needed to take full advantage of the additional features. It is also possible that the students' skills in cardiac auscultation are insufficient to reveal an existing significant difference between the stethoscopes. There is, however, no available documentation that cardiologists perform better with electronic compared to conventional stethoscopes, but it could be of interest to investigate if this could be the case.
Conclusion
The cardiac auscultation skills of undergraduate medical students were not influenced by the use of an electronic sensor-based stethoscope.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
HH participated in the design of the study, the acquisition of data, performed the statistical analysis, drafted the manuscript. KG and TJ concieved the study, contributed to the design, acquisition of data and the preparation 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
Meditron AS provided the Stethoscope® and the co-listening unit for the test group students.
==== Refs
Bloch H The inventor of the stethoscope: Rene Laennec J Fam Pract 1993 37 191 8336101
Reiser SJ The medical influences of the stethoscope. Sci Am 1979 240 148 150 368980
Grenier MC Gagnon K Genest JJ Durand J Durand LG Clinical comparison of acoustic and electronic stethoscopes and design of a new electronic stethoscope Am J Cardiol 1998 81 653 656 9514471 10.1016/S0002-9149(97)00977-6
Tavel ME Cardiac auscultation. A glorious past--but does it have a future? Circulation 1996 93 1250 1253 8653848
Mangione S Cardiac auscultatory skills of physicians-in-training: a comparison of three English-speaking countries Am J Med 2001 110 210 216 11182108 10.1016/S0002-9343(00)00673-2
Gaskin PR Owens SE Talner NS Sanders SP Li JS Clinical auscultation skills in pediatric residents Pediatrics 2000 105 1184 1187 10835055 10.1542/peds.105.6.1184
Mangione S Nieman LZ Gracely E Kaye D The teaching and practice of cardiac auscultation during internal medicine and cardiology training. A nationwide survey Ann Intern Med 1993 119 47 54 8498764
Jordan MD Taylor CR Nyhuis AW Tavel ME Audibility of the fourth heart sound. Relationship to presence of disease and examiner experience Arch Intern Med 1987 147 721 726 3827460 10.1001/archinte.147.4.721
Ishmail AA Wing S Ferguson J Hutchinson TA Magder S Flegel KM Interobserver agreement by auscultation in the presence of a third heart sound in patients with congestive heart failure Chest 1987 91 870 873 3581934
| 15882458 | PMC1131903 | CC BY | 2021-01-04 16:30:56 | no | BMC Med Educ. 2005 May 9; 5:14 | utf-8 | BMC Med Educ | 2,005 | 10.1186/1472-6920-5-14 | oa_comm |
==== Front
BMC MicrobiolBMC Microbiology1471-2180BioMed Central London 1471-2180-5-241588245610.1186/1471-2180-5-24Research ArticleThe antimicrobial susceptibility of Stenotrophomonas maltophilia isolates using three different methods and their genetic relatedness Tatman-Otkun Müşerref [email protected]ürcan Şaban [email protected]Özer Burçin [email protected] Bayram [email protected] Şebnem [email protected] Medical Faculty, Department of Microbiology and Clinical Microbiology, Trakya University, Turkey2 Institute of Health Science, Trakya University, Turkey3 Medical Faculty, Department of Clinical Bacteriology and Infectious Diseases, 22030- Edirne, TURKEY Trakya University, Turkey2005 9 5 2005 5 24 24 19 12 2004 9 5 2005 Copyright © 2005 Tatman-Otkun 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
Stenotrophomonas maltophilia is inherently resistant to many antimicrobials. So far, antimicrobial susceptibility tests for S. maltophilia have not been fully standardized. The purpose of the study was to compare the susceptibility of S. maltophilia isolates against seven different antimicrobials using three different methods and to investigate their genetic relatedness.
Results
Although trimethoprim/sulfamethoxazole (SXT) and ciprofloxacin have the lowest MIC values, SXT (98.1%) and ticarcillin/clavulanate (TLc) (73.1%) were found to be the most effective antimicrobials by agar dilution method, which was in accordance with the breakpoints established by NCCLS. Disc diffusion and E-test was in agreement with agar dilution method for SXT. When the isolation dates, clinics, antibiotyping, and AP-PCR data were investigated, two small outbreaks consisting of five and three cases were determined.
Conclusion
By using the NCCLS criteria, disc diffusion and E-test were unreliable alternative methods for S. maltophilia, except for SXT. However, the significance of these data should be confirmed by further experimental and clinical studies.
==== Body
Background
Infections with Stenotrophomonas maltophilia are increasingly encountered especially in debilitated or immune suppressed patients [1,2]. Nosocomial infections of the respiratory, urinary, central nervous, musculoskeletal, skin-soft tissue systems and of the gastrointestinal tract, and bacteremia, endocarditis and eye infections occur in sensitive individuals [2-4].
S. maltophilia is inherently resistant to many antimicrobials. Additional resistance develops against cephalosporins and aminoglycosides because of decreased outer membrane permeability and via at least two types of beta-lactamases [5]. Recently, resistance to floroquinolones via efflux system has been reported [6]. Also, antimicrobial susceptibility tests for S. maltophilia have not been so far fully standardized [2]. These problems raise difficulties for the choice of antimicrobials in S. maltophilia infections.
In this study, disc diffusion and E-test methods were compared with the agar dilution method in S. maltophilia strains against seven different antimicrobials. We also investigated the genetic relatedness of S. maltophilia isolates by the arbitrarily primed PCR (AP-PCR).
Results
S. maltophilia strains were mostly isolated from the lower respiratory tract (n = 20), and from urine (n = 11) and blood (n = 7) samples, and the majority were from the clinics of Paediatrics (n = 14), Neurology (n = 9), Nephrology (n = 8), Chest Diseases (n = 4) and Intensive Care Units (n = 4). Only six strains were isolated in 1998; however the isolation rate steadily increased throughout the years and reached 17 strains in 2002. In particular, the annual isolation rate in the clinics of Paediatrics and Neurology was only one or two until 2001, however it increased to six in 2001 and five in 2002 in Paediatrics. Also, six strains were isolated in Neurology in 2002.
Although trimethoprim/sulfamethoxazole (SXT) and ciprofloxacin (CIP) have the lowest minimal inhibitory concentration (MIC) values, the most effective antimicrobials were SXT and ticarcillin/clavulanate (TLc), which was in accordance with the breakpoints established by The National Committee for Clinical Laboratory Standards (NCCLS). The rate of susceptibilities obtained by the disc diffusion and E-test methods were similar to the agar dilution test; however, false susceptibility for CIP by both tests (p < 0.001) and for TLc by the E-test were obtained (p < 0.04) (Table 1). The disc diffusion and E-test methods showed a good agreement with agar dilution method for SXT (Table 2). Rates of correlations were poor for the other antimicrobials.
A total of 44 different patterns of 52 strains were obtained by The arbitrarily primed PCR (AP-PCR). Three small clusters were observed. All of five strains in pattern l were isolated in the Neurology clinic between March and September 2002. All three strains in pattern ll were isolated in the Paediatrics clinic between January and March 2001, and three strains in pattern lll were isolated in the Paediatrics, Nephrology and Chest Diseases clinics between January and November 2001. The antimicrobial susceptibility results supported the AP-PCR method for patterns l and ll, but not for pattern III. The strains in pattern I were resistant to ceftazidime (CAZ), cefepime (CPM), piperacillin (PIP) and piperacillin/tazobactam (PTZ), and susceptible to TLc, CIP, and SXT. The strains in pattern II were susceptible to all antimicrobial except for CIP. The dendrogram showed a Dice similarity coefficient ranging from14.8 to 100% (Figure 1).
Discussion
S. maltophilia has low pathogenicity, but it has emerged as an important nosocomial pathogen. Patients infected with S. maltophilia usually have underlying immunodeficiency or history of long-term or multiple hospitalizations, exposure to invasive devices and/or broad spectrum antimicrobials [2]. This organism, most frequently, causes lower respiratory and urinary tract infections and may result in secondary bacteremia [2]. Our results supported these observations.
According to the recommendations of the NCCLS [7,8], agar dilution method should be used in order to detect antimicrobial susceptibility of S. maltophilia strains. Since the dilution methods are more cumbersome or expensive than the disc diffusion or E-test methods in routine clinical microbiology laboratories, the aim of this study was to compare the performance of these latter methods with agar dilution method.
The NCCLS had not defined the criteria for disc diffusion method for S. maltophilia by the year 2004. So, breakpoints for other bacteria from the NCCLS comments have been tried to be adapted in various studies [6,14]. The best correlated results have been obtained with those recommended for Pseudomonas aeruginosa [6]. In 2004, the NCCLS recommended disc diffusion breakpoints for minocycline, levofloxacin, and SXT. Nevertheless, we tested other antimicrobials frequently used in nosocomial infections and interpreted these antimicrobial susceptibilities using the NCCLS criteria established for P. aeruginosa.
Our results showed that the most effective antimicrobials against S. maltophilia were SXT and TLc, as observed by several authors [6,14]. Nevertheless, the resistance rates for other antimicrobials in our study were extremely high. Resistance to beta-lactams in S. maltophilia is primarily intrinsic and mediated by inducible beta-lactamases (L1 and L2) that hydrolyses virtually all classes of beta-lactams [2]. Although many authors have tested piperacillin (with or without tazobactam) and cefepime against S. maltophilia, they are not suitable for the treatment of S. maltophilia infections. L2 beta-lactamase is susceptible to clavulanic acid, so TLc is preferred to PTZ [2,15]. However, Carrol et al. [1] determined that there was an obvious increase in the level of resistance to TLc when they prolonged the susceptibility tests up to 48 hours. This finding makes the in vitro efficiency of TLc disputable. The NCCLS has recently recommended 20–24 hours incubation for S. maltophilia, and so we did not evaluate the tests after 48 hours incubation [9].
The correlations between in vitro susceptibility methods for S. maltophilia show variations [10,14]. While Nicodemo et al. [10] stated that the disc diffusion tests have an excellent correlation with agar dilution for several antimicrobials; Pankuch et al. [14] used the breakpoint values for Enterobacteriaceae recommended by the NCCLS and found a high level of discordance for PTZ, TLc, CIP. Also, in our study, poor agreement was observed in the alternative test methods except for SXT. In a study where the correlation of E-test with agar dilution method for 16 antimicrobials in 176 clinical isolates were investigated, the authors found an excellent correlation and recommended E-test as an alternative susceptibility test [16]. Major and very major errors were very low in this latter study and were similar to our results for all the antimicrobials except for CIP and CAZ by both the disc diffusion and by the E-test. However most of the discordant susceptibility rates among the three methods evaluated were due to high occurrence of minor errors in our study. The MIC values of our strains cumulated close to breakpoints. For instance, the susceptibility breakpoint of CIP established by the NCCLS is 1 mg/L [7] and equals to MIC50 value of our strains. Moreover MIC values for CIP were 1 mg/L and 2 mg/L of 18 and 17 strains, respectively. The same applies for CAZ and CPM also. Therefore, minor variations caused to change in the susceptibility categories from intermediate susceptibility to susceptibility or resistance or vice versa. If variations in ± 1 doubling dilutions between different methods were to be accepted as essential agreement suggested by Pankuch et al [14] then our error rates would have been much more smaller.
Tracking of S. maltophilia isolates has a great importance to reveal their outbreaks, to determine the distribution routes and to take preventive measures. However, biotyping and antibiotyping methods are not reliable due to the relative metabolic inactivity and multiresistance of these isolates. More recently, genotypic methods have been developed and used with success to discriminate for phenotypically indistinguishable bacteria. AP-PCR is one of the most preferred molecular typing methods for this aim, because the results can be obtained rapidly even in a clinical laboratory. Also, it can be applied to a wide range of bacterial species by using almost the same materials and equipment [2,13]. We determined two small outbreaks consisting of five cases in the Neurology and three cases in the Paediatrics clinics by using AP-PCR method. Isolation dates, clinics, and antibiotyping data have also supported these results. On the other hand, the third cluster (strains from pattern III) suggests that the same strain can persist for a long time in hospital. The first strain from pattern III were isolated from the Pediatrics clinic in January 2001. The second and third strains were isolated from the Nephrology clinic in August 2001 and from the Chest Diseases clinic in November 2001, respectively. Reported nosocomial outbreaks due to S. maltophilia are generally short termed. [2]. There is no report of prolonged transmission extending up to 11 months for S. maltophilia. Valdezate et al [17] concluded that the epidemiological relationship among different S. maltophilia isolates needed to be analysed because unexpected results could be obtained. Antibiotic susceptibility profiles also supported that isolates in pattern III were independent isolates having same genotypes.
Conclusion
In conclusion, the disc diffusion and the E-test methods were unreliable alternative methods for S. maltophilia, except for SXT. However, significance of these data should be confirmed by further experimental and clinical studies.
Methods
Bacteria
The 52 S. maltophilia strains (one per patient) that were isolated from nosocomial infections between 1998 and 2002 in the Hospital of Trakya University were included in the study. The strains were identified by conventional bacteriological methods and were stored at -70°C in skim-milk media (Becton Dickinson, USA). Before the study, they were twice passaged onto 5% sheep blood agar and the identification was confirmed by Crystal ID Enteric-nonfermenter (Becton Dickson, USA).
Antimicrobial susceptibility tests
The drug powders for the agar dilution test were obtained from the following suppliers: Ceftazidime pentahydrate (Glaxo-Welcome, UK), CPM (Bristol-Myers Squibb, USA), PIP and tazobactam (Lederle, USA), ticarcillin disodium and clavulanate lithium (GlaxoSmithKline, UK), CIP (Bayer, Turkey), trimethoprim and sulfamethoxazole (Roche, Turkey). Standard antimicrobial discs (Oxoid, UK) were used for the disc diffusion tests and E-test strips were supplied by AB Biodisk, Sweden.
Antimicrobial susceptibility tests were carried out using the disc diffusion and agar dilution techniques as described by NCCLS [7,8]. The Agar dilution and E-test results were interpreted using the NCCLS criteria established for non-enterobacteriaceae, and the disc diffusion test was interpreted using the criteria established for P. aeruginosa [7-9]. The E-test technique was carried out according to the manifacturer's instructions. The tests were evaluated after 20–24 hours incubation at 35°C and were repeated if they were found to be discordant.
Escherichia coli the American Type Culture Collection (ATCC) 25922 and P. aeruginosa ATCC 27853 were used as quality control strains.
Definitions
The agar dilution method was accepted as the reference method. Categorical agreement was defined if the tests results were within the same susceptibility category, and errors of disc diffusion and E-test methods were determined as follows: Very major error; (resistant by reference method, susceptible by test method); major error; (susceptible by reference method, resistant by test method); and minor error; (intermediate result was obtained by one method but not the other) [10]. Percentage errors were calculated based on the total number of isolates which were tested. A good agreement was defined as complete category agreement over 90% and the total of very major and major errors below 5% [11].
Arbitrarily primed PCR
The method of vanCouwenberghe et al. [12] was used for the preparation of the DNA and AP-PCR, with minor modifications. Briefly, after an overnight culture at 37°C in 5% sheep blood agar, the bacteria were suspended in 1 ml TE buffer (10 mM Tris, 1 mM EDTA, pH: 8.0) to regulate the density to a 4 McFarland standard. Then, they were heated at 100°C for 10 min. The suspension was centrifuged at 2500 rpm for 10 minutes and the supernatant was used for AP-PCR. DNA in the supernatant was quantitated by spectrophotometry at an optical density of 260 nm. PCR mixtures were prepared in 100 μl of 1X buffer (10 mM Tris-HCl, 50 mM KCl, 2.5 mM MgCl2) and contained 1 μg DNA, 0.1 mM each dNTP, 2.5 U Taq polymerase and 30 pmol of pBR322 SalI primer (AGTCATGCCCCGCGC). PCR was initiated with five cycles of low stringency, which included a denaturing step at 95°C for 1 min, annealing of the primer at 28°C for 1 min, and 2 min of extension at 72°C. After the initial 5 cycles, 55 additional cycles were conducted with annealing of the primer at 50°C. The reaction was terminated with a final extension cycle at 72°C for 10 min.
Samples were electrophoresed in a 1.5% agarose gel (Sigma, Germany) in 1X Tris-borate-EDTA buffer for 90 min at 100 V and visualized under UV light after staining with ethidium bromide. To ensure reproducibility, all amplifications were done in duplicate and were also repeated using DNA extracted on different days. Dice coefficients of similarity were calculated for every pair of isolates by visual comparison of restriction patterns. If DNA profiles of isolates were indistinguishable or differing by only three or fewer DNA band shifts, then, the isolates were deemed same or related and included in the same pattern [13].
Statistical analysis
Chi-square test (Fisher's exact test when necessary) was used.
Abreviations
SXT: Trimethoprim/sulfamethoxazole
CIP: Ciprofloxacin
MIC: Minimal inhibitory concentration
TLc: Ticarcillin/clavulanate
NCCLS: The National Committee for Clinical Laboratory Standards
AP-PCR: The arbitrarily primed PCR
CAZ: Ceftazidime
CPM: Cefepime
PIP: Piperacillin
PTZ: Piperacillin/tazobactam
ATCC: The American Type Culture Collection
Authors' contributions
All the authors have read and approved the final manuscript and contributed equally to the manuscript. MTO drafted the manuscript, participated in the susceptibility test studies, searched the literature and reviews. SG, BO and BA collected bacteria and participated in the susceptibility test studies. SB carried out the AP-PCR studies. MTO and SG performed the statistical analysis and gave the final approval of the version to be published.
Acknowledgements
This study was granted by Trakya University Scientific Research Projects (TUBAP-449).
Figures and Tables
Figure 1 AP-PCR profiles of S. maltophilia strins including statistical analysis and dendrogram showing the genetic relationship between strains.
Table 1 Susceptibility of S. maltophilia obtained by the three methods studied (N = 52)
Percentage susceptibility* Agar MIC (mg/L)
Agent Disc diffusion E Test Agar dilution (MIC breakpoint) MIC50 MIC90 MIC Range
Ceftazidime 67.3 63.5 50.0 (≤ 8 mg/L) 8 256 0.25->256
Cefepime 30.8 42.3 34.6 (≤ 8 mg/L) 16 64 1–64
Piperacillin 34.6 15.4 26.9 (≤ 16 mg/L) 64 256 1->256
PTZ 61.5 46.2 42.3 (≤ 16/4 mg/L) 32/4 128/4 4/4->256/4
TLc 84.6 90.4 73.1 (≤ 16/2 mg/L) 8/2 64/2 0.5/2->256/2
Ciprofloxacin 92.3 92.3 53.8 (≤ 1 mg/L) 1 4 0.5–16
SXT 98.1 98.1 98.1 (≤ 2/38 mg/L) 0.5/9.5 1/19 0.125/2.375- 128/2432
PTZ: Piperacillin/tazobactam, TLc: Ticarcillin/clavulanate, SXT: Trimethoprim/sulfamethoxazol
Table 2 Correlation of susceptibility test methods for 52 S. maltophilia strains
% Discrepance
Very Major Major Error Minor Error % Correlation
N = 52 Agar-disc Agar-E-test Agar-disc Agar-E-test Agar-disc Agar-E-test Agar-disc Agar-E-test
Ceftazidime 5.8 5.8 0 0 17.3 13.4 76.9 80.8
Cefepime 0 1.9 0 0 11.5 23.1 88.5 75.0
Piperacillin 0 0 1.9 1.9 38.5 38.5 59.6 59.6
PTZ 1.9 0 0 1.9 34.6 23.1 63.5 75.0
TLc 0 1.9 1.9 0 17.3 19.2 80.8 78.9
Ciprofloxacin 5.8 5.8 0 0 32.7 32.7 61.5 61.5
SXT 0 0 0 0 0 0 100 100
PTZ: Piperacillin/tazobactam, TLc: Ticarcillin/clavulanate, SXT: Trimethoprim/sulfamethoxazole
==== Refs
Carroll KC Cohen S Nelson R Campbell DM Claridge JD Garrison MW Kramp J Malone C Hoffmann M Anderson DE Comparison of various in vitro susceptibility methods for testing Stenotrophomonas maltophilia Diagn Microbiol Infect Dis 1998 32 229 235 9884841 10.1016/S0732-8893(98)00089-3
Denton M Kerr KG Microbiological and clinical aspects of infection associated with Stenotrophomonas maltophilia Clin Microbiol Rev 1998 11 57 80 9457429
Benian Ö Alimgil L Erda N Two cases of Stenotrophomonas maltophilia endophthalmitis Ophthalmic Surg Lasers 2002 33 253 256 12027111
Munter RG Yinnon AM Schlesinger Y Hershko C Infective endocarditis due to Stenotrophomonas (Xanthomonas) maltophilia Eur J Clin Microbiol Infect Dis 1998 17 353 356 9721966 10.1007/s100960050081
In vitro activity against Stenotrophomonas maltophilia: Single versus combination agents
Krueger TS Clark EA Nix DE In vitro susceptibility of Stenotrophomonas maltophilia to various antimicrobial combinations Diagn Microbiol Infect Dis 2001 41 71 78 11687317 10.1016/S0732-8893(01)00281-4
National Committee for Clinical Laboratory Standards Methods for dilution antimicrobial susceptibility tests for bacteria that grow aerobically: Approved Standard NCCLS Document M7-A6 2003 6 Wayne, Pa, USA
National Committee for Clinical Laboratory Standards Performance standards for antimicrobial disk susceptibility tests: Approved Standard NCCLS Document M2-A8 2003 8 Wayne, Pa, USA
National Committee for Clinical Laboratory Standards Performance standards for antimicrobial disc susceptibility testing-fourteen informational supplement M100-S14 2004 Wayne, Pa, USA
Nicodemo AC Araujo MRE Ruiz AS Gales AC In vitro susceptibility of Stenotrophomonas maltophilia isolates: Comparison of disc diffusion, Etest and agar dilution methods J Antimicrob Chemother 2004 53 604 608 14973153 10.1093/jac/dkh128
Jorgensen JH Selection criteria for an antimicrobial susceptibility testing system J Clin Microbiol 1993 31 2841 2844 8263164
vanCouwenberghe CY Cohen SH Tang YJ Gumerlock PH Silva JJr Genomic fingerprinting of epidemic and endemic strains of Stenotrophomonas maltophilia (formerly Xanthomonas maltophilia) by arbitrarily primed PCR J Clin Microbiol 1995 33 1289 1291 7615743
Yao JDC Conly JM Krajden M Molecular typing of Stenotrophomonas (Xanthomonas) maltophilia by DNA macrorestriction analysis and random amplified polymorphic DNA analysis J Clin Microbiol 1995 33 2195 2198 7559978
Pankuch GA Jacobs MR Rittenhouse SF Appelbaum PC Susceptibilities of 123 strains of Xanthomonas maltophilia to eight β-lactams (including β-lactam- β-lactamase inhibitor combinations) and ciprofloxacin tested by five methods Antimicrob Agents Chemother 1994 38 2317 2322 7840563
Laing FPY Ramotar K Read RR Alfieri N Kureishi A Henderson EA Louie TJ Molecular epidemiology of Xanthomonas maltophilia colonization and infection in the hospital environment J Clin Microbiol 1995 33 513 518 7751349
Yao JDC Louýe M Louýe L Goodfellow J Simor AE Comparison of E test and agar dilution for antimicrobial susceptibility testing of Stenothrophomonas (Xanthomonas) maltophilia J Clin Microbiol 1995 33 1428 1430 7615774
Valdezate S Vindel A Martín-Dávila P Sánchez del Saz B Baquero F Cantón R High genetic diversity among Stenothrophomonas maltophilia strains despite their originating at a single hospital J Clin Microbiol 2004 42 693 699 14766838 10.1128/JCM.42.2.693-699.2003
| 15882456 | PMC1131904 | CC BY | 2021-01-04 16:03:39 | no | BMC Microbiol. 2005 May 9; 5:24 | utf-8 | BMC Microbiol | 2,005 | 10.1186/1471-2180-5-24 | oa_comm |
==== Front
BMC Musculoskelet DisordBMC Musculoskeletal Disorders1471-2474BioMed Central London 1471-2474-6-211585422110.1186/1471-2474-6-21Research ArticleSensori-motor adaptation to knee osteoarthritis during stepping-down before and after total knee replacement Mouchnino L [email protected] N [email protected] C [email protected] C [email protected] G [email protected] J-M [email protected] J-P [email protected] A [email protected] Laboratory of Movement and Perception, Faculty of Sport Sciences, 163 av. de Luminy 13288 Marseille cedex 9, France2 Department of Physical Medicine and Rehabilitation, Université de la Méditerranée, 92 rue A. Blanqui 13005 Marseille, France3 Department of Orthopedic Surgery, CHU Conception, bd. Baille, 13005 Marseille, France2005 26 4 2005 6 21 21 19 7 2004 26 4 2005 Copyright © 2005 Mouchnino et al; licensee BioMed Central Ltd.2005Mouchnino 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
Stepping-down is preceded by a shift of the center of mass towards the supporting side and forward. The ability to control both balance and lower limb movement was investigated in knee osteoarthritis patients before and after surgery. It was hypothesized that pain rather than knee joint mobility affects the coordination between balance and movement control.
Methods
The experiment was performed with 25 adult individuals. Eleven were osteoarthritic patients with damage restricted to one lower limb (8 right leg and 3 left leg). Subjects were recruited within two weeks before total knee replacement by the same orthopedic surgeon using the same prosthesis and technics of surgery. Osteoarthritic patients were tested before total knee replacement (pre-surgery session) and then, 9 of the 11 patients were tested one year after the surgery when re-educative training was completed (post-surgery session). 14 adult individuals (men: n = 7 and women: n = 7) were tested as the control group.
Results
The way in which the center of mass shift forward and toward the supporting side is initiated (timing and amplitude) did not vary within patients before and after surgery. In addition knee joint range of motion of the leading leg remained close to normal before and after surgery. However, the relative timing between both postural and movement phases was modified for the osteoarthritis supporting leg (unusual strategy for stepping-down) before surgery. The "coordinated" control of balance and movement turned to be a "sequential" mode of control; once the body weight transfer has been completed, the movement onset is triggered. This strategy could be aimed at shortening the duration-time supporting on the painful limb. However no such compensatory response was observed.
Conclusion
The change in the strategy used when supporting on the arthritis and painful limb could result from the action of nociceptors that lead to increased proprioceptor thresholds, thus gating the proprioceptive inputs that may be the critical afferents in controlling the timing of the coordination between balance and movement initiation control.
==== Body
Background
Anticipatory postural adjustments (APAs) precede voluntary lower limb movements, as shown by experiments in which the limb to be moved initially supported the body weight (leg flexion [1,2]; lateral leg raising [3]; gait initiation [4]). In these cases, movement is preceded by a shift of the center of mass (CM) towards the supporting side and forward as in gait initiation. This anticipatory CM shift, aimed at unloading the leg to be moved and creating the condition for progression, is initiated by the generation of forces (thrust exerted on the ground) that shift the CM. Although no specific receptors exist that detect CM position and shift, it can be indirectly estimated through measurements of the center of pressure (CP). Morasso and Schieppati [5] showed the acceleration of the CM correlates highly with the CM-CP difference, as a consequence of physical laws. It has been hypothesized that pressure afferent inputs play a major role in determining the actual position of body mass to be balanced over the feet (i.e. the CM).
Load-detecting and position-sense afferents might be candidates for monitoring balance regulation, as shown by Dietz et al. [6], Eklund [7] and Roll and Roll [8] for proprioceptive information and by Magnusson et al. [9] and more recently by Kavounoudias et al. [10] for plantar cues. Afferent inputs also have phase-dependent effects during gait (see [11] for review). Other afferents such as noxious sensory afferents can influence balance control and could have deteriorating effects on postural control mechanisms [12]. Nociceptive primary afferent fibers have a peripheral action by way of dorsal root reflexes. Rossi et al. [13] showed in the case of foot pain that proprioceptive activity is profoundly influenced by nociceptive reflex action, indicating how closely the two functions of the two systems may be associated during natural movements.
We investigated the ability to control both balance and lower limb movement initiation in knee osteoarthritis patients, in a stepping down task. Stepping down requires a transition from a bipedal to a monopedal stance (as in the leg raising task) in addition with a forward propulsion prior to heel-off (as in gait initiation). These APAs precede any movement of the leading lower leg. In knee osteoarthritis patients, knee joint mobility is impaired, as are the torques exerted by this joint when supporting the body weight. Knee joint excursion and muscle strength are intrinsic elements of stiffness and are both affected by knee osteoarthritis. Among these physical decay mechanisms, the main factor is pain of both inflammatory and mechanical origins. The inflammatory pain results from the effects of a variety of endogenous chemical agents released from damaged cells at the knee joint level [13] and entering the damaged region (i.e. at the thigh level).
It is hypothesized that pain rather than knee joint mobility and muscle strength could change the coordination between APAs and movement initiation leading to compensatory mechanisms in osteoarthritis patients.
Methods
Subjects
The experiment was performed with 25 adult individuals. Eleven were osteoarthritic patients (mean age: 69 years from 45 to 82, men: n = 5 and women: n = 6; mean weight: 75 kg from 63 to 109; mean height: 1.64 m from 1.50 to 1.87) with damage restricted to one lower limb (unilateral symptomatic knee arthritis; 8 right leg and 3 left leg). Subjects were recruited within two weeks before total knee replacement by the same orthopedic surgeon using the same type of posterior cruciate-sparing prosthesis and technics of surgery were used in all patients.
A control group of 14 healthy adults were tested in this experiment (men: n = 7 and women: n = 7) (mean age: 72 years old from 66 to 81; mean weight: 65 kg from 47 to 85; mean height: 1.66 m from 1.50 to 1.82).
Protocol
Subjects were instructed to step down from a platform with a standard step height of 170 mm. Twenty randomized trials (ten with each leg) per subject were performed. Light Emitting Diodes located to the left and to the right signaled to the subject when to start their movement and which leg they had to move first.
A wooden platform was positioned on a force platform and in relation to the edge of a second force platform (Fig. 1A) so that each subject could comfortably land in the middle of this second force plate. Subjects kept the landing platform in view with peripheral vision and aimed central vision forward to a go signal given by the experimenter.
Figure 1 Phases of the motor act in stepping down movement (A). Reference times were measured from 2 curves (B). Top, The reference times plotted on the vertical velocity curve of the malleolus marker of the leading leg (T2) and of the supporting leg (T3) correspond respectively to the onset and offset of the movement phase. Bottom: lateral CP curve (T1) corresponds to the onset of CP change and Tbal to the end of the ballistic CP shift.
A trial consisted of a subject stepping off the first platform with either the right or the left foot and landing with the "leading" foot in the middle of the force-plate. Touch down with the leading foot was followed by stepping off the platform and over the landing force plate with the "supporting" foot.
Osteoarthritic patients were tested before total knee replacement (pre-surgery session) and then, 9 of the 11 patients were tested one year after the surgery when re-educative training was completed (post-surgery session). The control group was tested during one session. For safety, a researcher and medical doctor stood behind and laterally to each patient while stepping, but all subjects were able to execute the task without assistance.
Data collection
The kinematic analysis was performed by an automatic TV-image processor (EL.I.TE. system). TV cameras worked at a sampling frequency of 100 Hz, while system accuracy was 1 part in 3000. Under the present experimental condition, the field of view explored was 2*2*3.5 m and the accuracy was less than 1 mm.
Sixteen light-reflecting markers were placed on anatomical landmarks: bilaterally, on the external edge of the orbits, the acromions, the anterior-superior iliac spines, the greater trochanters, the external part of the lateral femoral condyles, the anterior tibial tuberosities, the lateral malleoli, and the 5th metatarsal heads. In this study we did not use the markers placed on the orbits, on the acromions, and on the 5th metatarsal heads. (Figure 1A). Two cameras were placed 3.5 m in front of the subject.
Electromyographic recordings (EMG) were made from 2 muscles Vastus lateralis (VL, knee extensor) on both sides of the subject, by means of bipolar surface electrodes spaced 2 cm apart. Preamplifiers were placed next to the recording electrodes. The EMG signals were amplified (gain of 1000), band-pass filtered (10 Hz to 200 Hz), digitized at 500 Hz, and rectified.
The ground reaction forces were recorded at a frequency of 500 Hz with the subjects standing on an AMTI force platform and landing on a Kistler force platform.
Data analyses
The onset of the lateral shift of the center of pressure (T1) was taken as the onset of the APAs (Fig 1B). The postural phase started with the first CP change (T1) measured with the force platform and ended with the end of the ballistic CP shift towards the supporting side (Tbal), which corresponds to a breakdown in the M/L CP curve (Fig. 1B).
The vertical velocity profile at the ankle (malleolus marker) was approximately bell-shaped, with a single maximum. The onset of leg flexion (T2) was taken as the end of the initial period of zero vertical velocity (+/-5% of the maximal velocity) of the leading leg (malleolus marker). The end point (T3) was defined as the position at which the leg that will be the supporting one returned back to zero velocity after the movement (Fig. 1B). T2–T3 defined the movement phase.
EMG analysis was performed by calculating latencies and areas of integrated EMG activities or bursts. The resting activities were measured in each trial in the 300 ms of recording preceding the signal onset in order to determine the background EMG activity. The mean and standard deviation of this background activity were then calculated for each subject. Timing and intensity measurements were performed. For the timing measurement, the onset and the end of EMG bursts were defined as the times when the EMG activity increased above or decreased below a threshold level set at two times the standard deviation of the background activity. The duration of the burst was also calculated. The intensity of muscular activity was calculated by subtracting the baseline from the EMG activity level reached 300 ms after the activity had increased above the threshold level. In the analysis of the changes in muscle activity profiles, the first step involved two phases: the pre-activation phase lasting 300 ms prior to the landing (first signal recorded by the landing platform), and the activation phase computed during 300 ms following the ground contact.
Next, the EMG activity was windowed before ground contact (pre-activation: from -300 ms to -150 ms and -150 ms to ground contact) and after the ground contact (activation: from the ground contact to 150 ms and 150 to 300 ms).
Statistical analysis
To determine modifications caused by the surgery, dependent variables were tested in a first step, using a 2*2 (pre-, post-surgery * sound, arthritis leg) repeated measures ANOVA.
In a second step, comparison with a control group was done to document that, post-surgery, behavior of patients was not different from that of a control group. Dependent variables were assessed using a 2*2 (post-surgery patients, control group * 2 sides) between subjects ANOVA. The Newman-Keuls post-hoc test was used to assess the difference between factors. The level of significance was set at 5%.
Results
Pain intensity and passive joint mobility assessments
The average pre-surgery Hospital for Special Surgery (HSS) score was 59.1 (+/- 10.15) with a maximum of 100. The mean pain Visual Analog Scale (VAS) value was 49 mm (+/- 9) (VAS; worst pain ever 100 mm, no pain 0 mm). The mean post-surgery HSS score was 80.8 (+/- 8.4) and the VAS value was 7 (+/- 9). The VAS value vas obtained just after stepping-down task. The passive mobility of the knee joint was tested for all patients. The average pre- and post-surgery mobilities were 113 degrees (+/- 21) and 105° (+/- 18) respectively, whereas the mobility of the sound knee was 128° (+/- 11).
Anticipatory Postural Adjustments
The osteoarthritis patients exhibited an increase of the duration of the postural phase (T1-Tbal; Fig. 1B) after surgery. This effect was however not statistically significant [F(1,4) = 5,67; p = 0.075]. On average, the total duration of the postural phase was longer (835 ms +/- 207) in the post-surgery session, than in the pre-surgery session (652 ms +/-143) and no side-effect was observed within patients. Post-surgery patients did not recover a duration similar (p < 0.001) to that observed in the control group (543 ms +/- 107).
By contrast, the onset of this phase in terms of "thrust" exerted onto the ground (Fig. 2) was not different in patients before and after surgery [F(1,4) = 0.038; P = 0.85]. The A/P and M/L peaks remained synchronized before (22 ms+/-60) and after surgery (16 ms +/- 17). These events were tightly coupled in patients after surgery as in the control group (1 ms +/- 26). In addition, the M/L peak amplitude was not different in patients between pre- (270 mm +/-45) and post-surgery sessions (257 mm +/-37). After surgery, the M/L thrust was close to that observed in the control group (258 mm +/-35; p = 0.88).
Figure 2 Schema of the horizontal shift of the center of mass (CM) and associated center of pressure (CP) (left part) and description of the M/L and A/P CP curves (right part). The dotted lines show the time-relationships between each component. Note that the M/L thrust (T1-Peak) coincides with the first backward CP shift, and that during the unloading component of the M/L CP shift, the second backward shift occurs, which corresponds to heel off (T2).
Movement performance
The total duration of the movement phase (T2–T3) was not different in patients [F(1,4) = 1.80; p = 0.24] (Pre-surgery: 1694 ms +/- 355 ; Post-surgery 1502 ms +/- 230). The movement duration in post-surgery was similar (p = 0.066) to that observed in the control group (1402 ms +/- 193).
The maximal flexion reached by the leading leg did not differ statistically in patients before and after surgery [F(1,6) = 5.44; p = 0.058] (see table. 1) and no side-effect was observed within patients. After surgery, the maximal flexion was close to that observed in the control group (p = 0.37). By contrast, considering the flexion of the leg that was previously the supporting leg, after surgery, patients decreased the leg flexion [F(1,5) = 19,8; p = 0.006] (see table. 1) and no side-effect was observed. However, in post-surgery session, the maximal flexion of the supporting leg remained reduced compared to the control group (p = 0.0017).
Table 1 Maximal knee joint angle reached during the stepping-down performance of the leading leg and of the supporting leg during the swing phase.
Maximal knee joint angle
Leading leg Controls Patients before surgery Patients after surgery
Right / Sound 45.4° +/-4.7 49.2° +/-10.5 58.3° +/-23.4
Left / arthritis / operated 46.2° +/-7.3 33.7° +/-14 40.9° +/-12
Supporting leg Controls Patients before surgery Patients after surgery
Right / Sound 79.9° +/-10 81.2° +/-6.3 67.3° +/-29.1
Left / arthritis / operated 82.7° +/-4.4 55.3° +/-14 52° +/-21.2
Time-relationships between APAs and movement initiation
The stepping down movement of the leading leg was triggered while the unloading phase (peak-Tbal, Fig. 2) was being performed, before the M/L CP shift was completed. The time-relationships between unloading (Tbal) and stepping down initiation (T2) differed in patients before and after surgery [F(1,4) = 15.53; p = 0.016]. Before surgery, in patients who used the arthritis limb as the supporting limb (unusual strategy), the movement initiation was delayed and coincided (-64 ms +/-452) with the end of the lateral unloading. This result, however, varied widely, as shown by the high standard deviation. Post-surgery, stepping down is triggered largely before the unloading is completed (sound supporting leg: -514 ms +/-60; operated supporting leg: -492 ms +/-176). Post-surgery patients did not recover an anticipation similar to that observed in the control group (-214 ms +/-40) (p < 0.001).
The delayed movement initiation (T2) when supporting on the arthritis leg before surgery, might be aimed at shortening the duration of the supporting phase for the painful leg. This was not the case, however, because there was no clear side-effect [F(1,4) = 6.33; p = 0.065] on the duration of the monopodal stance. In addition, this duration was even longer [F(1,4) = 19.8; p = 0.011] before than after surgery (797 ms+/-197 and 681 ms+/-156, respectively). The post-surgery duration decreased to a value close (p = 0.38) to that observed in the control group (644 ms +/-49).
Weight acceptance
The adaptation of the weight acceptance is illustrated in Fig. 3. The ground impact, defined as the maximal value of the vertical ground reaction force and normalized to the body weight, did not differ in patients before and after surgery [F(1,4) = 3.37; p = 0.14]. However, in patients landing on the sound leg (i.e. using the arthritis leg as the supporting leg) before surgery, the ground impact increased (142 % +/-36) [side-effect F(1,4) = 7.59; p = 0.05] compared to those landing on the arthritis leg (118 %+/-37) (Fig. 3). This result indicated a reduced breaking capacity of the supporting knee joint during the monopodal stance, which enhanced the forthcoming ground impact. After surgery, the ground impact decreased to a value close to that observed in the control group (p = 0.78) (Fig. 3).
Figure 3 Schema of the vertical ground reaction force recorded on the landing force platform. Weight acceptance was from the ground contact to the peak and was calculated in percentage relative to the body weight to normalize the data for all the subjects.
There was no significant difference in "time to peak" of the vertical force (weight acceptance velocity) for both sides in patients before and after surgery [F(1,4) = 3.37; p = 0.14] (Fig 3).
EMG activities associated with ground contact
The comparison between kinetic events and associated EMG pattern points out some differences. First, during the swing phase, the moving limb exhibited a pre-activation of the VL before the ground contact. The leading VL pre-activation was correlated with increasing activity of the VL on the supporting side (Fig. 4). The onset of the pre-activation of the VL muscle of the leading limb did not differ in patients before and after surgery [F(1,5) = 1.63; p = 0.25]. However, in pre-surgery session, the pre-activation occurred earlier [side-effect F(1,5) = 11.84; p = 0.018] when landing on the arthritis leg (-414 ms+/-90) than when landing on the sound leg (-335 ms +/-90). This was also observed for the post-surgery sessions (345 ms +/-67 and -298 ms +/-74, respectively).
Figure 4 Kinetic and rectified EMG patterns recording with one control subject. The EMGs were recorded at a proximal level (VL, Vastus lateralis) for both sides. Note the supporting and leading VL activity prior to the ground contact.
VL muscle activation was not statistically different in patients before and after surgery [F(1,5) = 0.5; p = 0.50] (Fig. 5). The pre-activation increased [window-effect F(3,15) = 14.36; p < 0.001] from the first window (-300 ms to -150 ms) to the second (-150 ms to ground contact) and to the third (ground contact to 150 ms). However, when landing on the sound leg, the activity of the leading VL strongly increased before the ground contact (-150 ms to ground contact) [interaction side*window [F(3,15) = 5.13; p = 0.012]. Note that in this latter case, the VL of the leg to be stepped down supported 140% of the body weight. No such increase was observed in patients landing on the arthritis leg. Post-surgery (Fig. 5), this enhanced activity no longer exhibited differences compared to the control group (P = 0.39).
Figure 5 Dynamic profiles of VL activation recorded on the forthcoming landing leg before and after surgery. EMG data are windowed each 150 ms from 300 ms before ground contact to 300 ms after ground contact (Arbitrary Units, AU).
Discussion
The present study examined whether the coordination between balance and movement remained unchanged in patients as in controls. It was hypothesized that pain rather than knee joint mobility could change the coordination between balance and movement leading to compensatory mechanisms.
Lack of changes in the APAs
It was found that the way the Center of Mass shifts towards the supporting side and forward is initiated during the postural phase (T1-Tbal) did not changed in patients. The same observation was made for controls. These results are in agreement with that characterizing patients before surgery reported during lateral stepping by Viton et al. [15]. The proactive mechanism controlling the APAs may be triggered in absence of peripheral feedback as proposed by Forget and Lamarre [16,17] for a forearm unloading (waiter task) performed by a patient without proprioception. Bent et al. [18] reported before voluntary forward step performed without vision, that the initiation phase is run in a feedforward manner without vestibular influence. In the present experiment the presence of pain, mobility and muscle strength possible changes were not sufficient to influence the initial control of the APAs.
The effects of arthritis supporting leg on triggering the leg movement
The coordination between balance and movement initiation (T2) is changed before surgery : movement onset was delayed until the end of the postural phase when patients used an unusual strategy to support themselves on the arthritis limb. The previous coordinated control of balance and movement (i.e. movement onset is initiated while the unloading of the leading leg is being performed) changed to a sequential mode of control; once the CM lateral shift was completed, the movement onset then was triggered. This change in the coordination leads to a forward fall as shown by the ground impact enhancement associated with an increased amount of muscle activity just prior to ground contact. There are several interpretations to explain the sequential strategy.
One possible explanation would have been that the stepping down movement is delayed to decrease the duration of the monopodal stance phase on the painful limb. However, no such compensatory mechanism was observed as shown by the longer duration of the monopodal stance on the arthritis leg as compared to the sound supporting leg.
The second interpretation of the time lag between postural adjustments and movement initiation observed before surgery might result from the lack of knee joint mobility. This, however, is not shown by the knee joint maximal flexion during stepping down which is even greater in pre-surgery session than in post-surgery. The breaking ability of the supporting arthritis leg (absorption by the knee extensor) could not intervene because at that time both feet are on the ground.
A third possibility could be related to an adapted motor command due to fear of pain. During the pre-surgery session, the patients were asked whether they have pain in quiet standing (i.e. before stepping down). None of the subjects reported having perceived any pain at that time. However, they reported to suffer from the arthritis limb when exerting pressure. In that case, a less painful strategy might be to avoid pressure on the arthritis leg in decreasing the amplitude of the M/L thrust. No such compensatory response was observed. In addition, twenty trials per subjects were collected in the pre- as in the post-surgery sessions. If fear of pain is the main contributing reason to observed changes in the coordination, the thrust amplitude would decrease from the first to the following trials. No such decrease was observed.
The sequential organization between postural phase and movement performance together with the high variability of the timing accounted to a reduced accuracy in the integration of sensory information. Pain modulation of movement might happen in many ways. One of this mechanism may involved presynaptic inhibition [19] produced by nociceptive action. Presynaptic control of Ia afferents from extensor muscles may shape the amplitude, duration, and timing of the stance phase of locomotion [20]. Acute arthritis and associated nociceptive stimulation might lead to increased proprioceptor thresholds thus gating the proprioceptive inputs, as it was previously reported by Rossi et al. [13] during locomotion in the case of foot pain. After surgery, pain is removed and this coordination becomes normal in addition with a slowing strategy of the body weight transfer. These observations emphasize the deteriorating effects of a nociceptive stimulation in controlling the timing of the coordination between balance and movement initiation control. This timing is normally controlled by the afferents of proprioceptive origin (see for review [21]). In that case, the load feedback mechanisms play a crucial role in phase-switching during the leg movement task, as reported for the 1b activity of ankle extensors in the switch-phase of locomotion [11]. This could certainly lead to the hypothesis of a combined effect of nociceptive and proprioceptive afferents in the posture-movement coordination as it was reported by Blouin et al. [12] for balance control.
Conclusion
To conclude, pain more than fear of pain or knee joint mobility and muscle strength appears to be the relevant factor that disturbed the coordinated control between balance and movement. After surgery no more pain is noticeable and the motor patterns were restored to as close to "normal" as possible.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
LM, NG, JMV and AD participated in the conception and design of the study. LM, NG, CBl Cbo and GG performed data analyses. JPF was the orthopedic surgeon. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This work was supported by a grant from PHRC (Clinical Research Project)
==== Refs
Rogers MW Pai YC Dynamic transitions in stance support accompanying leg flexion movements in man Exp Brain Res 1990 81 398 402 2397765 10.1007/BF00228132
Rogers MW Pai YC Organization of preparatory postural responses for the initiation of lateral body motion during goal directed leg movements Neurosci Lett 1995 187 99 102 7783967 10.1016/0304-3940(95)11351-7
Mouchnino L Aurenty R Massion J Pedotti A Coordination between equilibrium and head-trunk orientation during leg movement: a new strategy built up by training J Neurophysiol 1992 67 1587 1598 1629766
Brenière Y Do MC Control of Gait Initiation J Mot Behav 1991 23 235 240 14766505
Morasso P Schieppatti M Can muscle stiffness alone stabilize upright standing? J Neurophysiol 1999 82 1622 1626 10482776
Dietz V Gollhofer A Kleiber M Trippel M Regulation of bipedal stance: dependency on "load" receptors Exp Brain Res 1992 89 229 231 1601101 10.1007/BF00229020
Eklund G Position sense and state of contraction: the effect of vibration J neurol Neurosurg Psychiat 1972 41 433 444
Roll JP Roll R Amblard B, Berthoz A, Clarac F From eye to foot: a proprioceptive chain involved in postural control Posture and Gait 1988 Elsevier Science Publisher B.V 155 164
Magnusson M Enbom H Johansson R Wiklund J Significance of Pressor Input from the Human Feet in Lateral Postural Control Acta Otolaryngol 1990 110 321 327 2284906
Kavounoudias A Roll R Roll JP The plantar sole is a "dynamometric map" for human balance control NeuroReport 1998 9 3247 3252 9831459
Duysens J Clarac F Cruse H Load-regulating mechanisms in gait and posture: comparative aspects Physiol Rev 2000 80 83 133 10617766
Blouin JS Corbeil P Teasdale N Postural stability is altered by the stimulation of pain but not warm receptors in humans BMC Musculoskelet Disord 2003 17 23 14565854 10.1186/1471-2474-4-23
Rossi A Decchi F Ginanneschi F Presynaptic excitability changes of group Ia fibers to muscle nociceptive stimulation in humans Brain Res 1999 818 12 22 9914433 10.1016/S0006-8993(98)01253-0
Barrett DS Cobb AG Bentley G Joint proprioception in normal, osteoarthritic and replaced knees J Bone Joint Surg 1991 53 56
Viton JM Timsit M Mesure S Massion J Franceschi JP Delarque A Asymetry of gait initiation in patients with unilateral knee arthritis Arch Phys Med Rehabil 2000 81 194 200 10668774
Forget R Lamarre Y Anticipatory postural adjustment in the absence of normal peripheral feedback Brain Res 1990 508 176 179 2337787 10.1016/0006-8993(90)91135-4
Forget R Lamarre Y Postural Adjustments Associated with Different Unloadings of the Forearm – Effects of Proprioceptive and Cutaneous Afferent Deprivation Can J Physiol Pharmacol 1995 73 285 294 7621367
Bent LR Inglis JT McFadyen BJ Vestibular contributions across the execution of a voluntary forward step Exp Brain Res 2002 143 100 105 11907695 10.1007/s00221-001-0967-7
Schmidt RF Schaible HG Rudomin P, Romo R, Mendell L Modulation of nociceptive information at the presynaptic terminals of primary afferent fibers Presynaptic inhibition and neural control 1998 New york 424 439
Gossard JP Control of transmission in muscle group Ia afferents during fictive locomotion in the cat J Neurophysiol 1996 76 4104 4112 8985904
Rossignol S Lund J Drew T Cohen A, Rossignol S, Grillner S The role of sensory inputs in regulating patterns of rhythmical movements in higher vertebrates Neural control of rhythmic movements in vertebrates 1988 New York: Wiley 201 283
| 15854221 | PMC1131905 | CC BY | 2021-01-04 16:32:04 | no | BMC Musculoskelet Disord. 2005 Apr 26; 6:21 | utf-8 | BMC Musculoskelet Disord | 2,005 | 10.1186/1471-2474-6-21 | oa_comm |
==== Front
BMC PharmacolBMC Pharmacology1471-2210BioMed Central London 1471-2210-5-101581397310.1186/1471-2210-5-10Research ArticleSildenafil citrate increases myocardial cGMP content in rat heart, decreases its hypertrophic response to isoproterenol and decreases myocardial leak of creatine kinase and troponin T Hassan Madiha AH [email protected] Amal F [email protected] Department of Pharmacology, Faculty of Medicine, University of Alexandria, Egypt2 Department of Medical Biochemistry, Faculty of Medicine, University of Alexandria, Egypt2005 6 4 2005 5 10 10 10 10 2004 6 4 2005 Copyright © 2005 Hassan and Ketat; licensee BioMed Central Ltd.2005Hassan and Ketat; 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
Cardiac hypertrophy is a major risk factor for morbidity and mortality in a number of cardiovascular diseases. Consequently, the signaling pathways that inhibit cardiac hypertrophy are currently receiving much interest. Among them, nitric oxide (NO), signaling via cGMP and cGMP-dependent protein kinase I, has been recognized as a negative regulator of cardiac hypertrophy. The present study investigated the in-vivo effect of sildenafil as a phosphodiestrase-5A (PDE-5A) inhibitor on the hypertrophic response of rat heart to isoproterenol and the relation of this effect to the level of myocardial cGMP and integrity of the constitutive nitric oxide synthase (cNOS) activity.
Results
The results showed that daily intraperitoneal administration of sildenafil per se for 10 days was without noticeable adverse effects on survival or myocardium. Conversely, daily subcutaneous administration of isoproterenol for 10 days caused significant myocardial hypertrophy, cell injury and decline in survival. When sildenafil was injected daily, one hour before isoproterenol, survival was significantly improved and the myocardium didn't show significant hypertrophy or cell injury. Interestingly, sildenafil was accompanied by significant rise in myocardial cGMP level, a parameter which was found in the present study to possess a significant negative correlation with cardiac hypertrophy and leak of cardiac troponin T into serum. At the same time, cGMP was found to possess a positive correlation with myocardial creatine kinase activity that reflects the efficiency of the energy utilization processes in the myocardium. However, in rats given Nω-nitro-L-arginine (L-NNA) as a competitive inhibitor of cNOS, sildenafil failed to show any favorable effect on survival or the myocardial injury parameters used to assess isoproterenol-induced injury.
Conclusion
The present study suggests that increased cardiac cGMP level by sildenafil have a cardioprotective effect probably through acting as a post-receptor negative regulator of cardiac sympathetic responsiveness. Integrity of NOS function was an essential prerequisite for sildenafil's mediated cardioprotection encountered in the present study.
==== Body
Background
Sildenafil is a selective inhibitor of phosphodiesterase-5A (PDE-5A), the enzyme that hydrolyzes cGMP. Sildenafil is orally effective in the treatment of erectile dysfunction. Its pharmacological action is due to prolonging the signaling actions of nitric oxide (NO) in penile smooth muscle [1]. Interestingly, a recent publication reported a pronounced infarct size-reducing effect of sildenafil in an in-vivo rabbit model of coronary occlusion [2]. Reduced infarct size by sildenafil has also been reported in mice and rat heart subjected to global ischemic/reperfusion (I/R) injury [3,4]. In those studies, opening of mitochondrial ATP-sensitive K channels, induction of NO synthase (NOS) isoforms and increased cGMP level by sildenafil were suggested to mediate a preconditioning-like cardioprotective effect [2-4].
As regards myocardial hypertrophy, a great number of in-vivo and in-vitro studies tested the role of many signaling pathways involved in its induction [5]. At the same time, the role of the negative regulators such as NO, atrial natriuretic peptide and cGMP have so far received much less attention until recently recognized to be of particular significance [6,7]. Alternatively, drugs that modulate the NO-cGMP signaling pathway like sildenafil and probably its congeners may be useful especially that sildenafil proved to be cardioprotective in models of cardiac I/R injury.
To address this issue, the present study examined the cardioprotective effect of sildenafil in a rat model of ventricular hypertrophy and myocardial cell injury induced by the β-adrenergic agonist isoproterenol. The model is characterized by technical simplicity, excellent reproducibility and an acceptable level of mortality [8]. Previous studies in rats revealed that isoproterenol induced a dose-dependent increase in ventricular end diastolic pressure and global wall stress that were associated with compensatory hypertrophy and repair fibrosis [5,9]. However, in the present study, the following parameters were assessed: rate of survival among experimental groups, heart coefficient, myocardial cGMP level, myocardial creatine kinase activity and serum level of cardiac troponin T (cTnT). In addition, the role of constitutive NOS was examined by adding the NOS inhibitor Nω-nitro-L-arginine (L-NNA) to drinking water of a group of sildenafil and isoproternol treated rats.
Results
Cumulative survival curves (Fig. 1)
Figure 1 Kaplan-Meier analysis of 10 days survival among 70 rats classified into vehicle (group I, n= 10; solid line), isoproterenol (group II, n = 15; thick dotted intermediate curve), isoproterenol/ sildenafil (group III, n = 15; thick dashed curve), sildenafil/vehicle(group IV, n = 15; solid line) and sildenafil/isoproternol/ Nω-nitro-L-arginine (group V, n = 15; thin dash-dot line to the left). Significant improvement in survival rate was found by log-rank test in group III compared to groups II and V (P < 0.05).
In group I (vehicle control) and group IV (sildenafil/vehicle), survival was 100 % till termination of study (10 days). In group II (isoproterenol) and group V (sildenafil/isoproternol/L-NNA), survival was 53.33% (8/15) in each group. In group III (sildenafil/isoproternol) survival was 86.67% (13/15). By log-rank test, significant improvement in survival rate was found in group III compared to groups II and V (P < 0.05).
Heart coefficient, myocardial creatine kinase (CK) activity, myocardial cGMP level and serum cardiac troponin T (cTnT)
Table I shows significant increase in heart coefficient in group II (isoproterenol) and group V (sildenafil/isoproterenol/L-NNA) compared to group I (vehicle control), group III (sildenafil/isoproternol) and group IV (sildenafil/vehicle) denoting occurrence of cardiac hypertrophy in the former groups. Also, significant myocardial cell injury in group II and group V was indicated by the decline of myocardial CK activity and leak of cTnT into serum in these groups compared to group I. At the same time, myocardial CK activity and serum level of cTnT in group III and group IV were normal and comparable to group I (homogenous subset). Elevated myocardial cGMP was found in two sildenafil groups (III and IV) compared to other groups (I, II and V).
Table 1 The heart coefficient (mg/g body wt.), myocardial cGMP level (pmoles/g wet tissue), CK activity (U/mg tissue protein) and serum level of cTnT (ng/ml) in rats after 10 days of treatment with: vehicle (group I), isoproterenol (group II), sildenafil/ isoproterenol (group III), sildenafil (group IV) and sildenafil/isoproternol/ Nω-nitro-L-arginine (group V).
Heart coefficient Myocardial cGMP Myocardial CK Serum cTnT
Group I n = 10 3.33 ± 0.18* 4.76 ± 1.03** 6.03 ± 0.35** 0.38 ± 0.13*
Group II n = 8 5.27 ± 0.17** 4.12 ± 1.36** 3.10 ± 0.35* 6.50 ± 1.29***
Group III n = 13 3.75 ± 0.48* 10.51 ± 3.05*** 5.62 ± 0.44** 3.71 ± 1.16**
Group IV n = 15 3.49 ± 0.67* 10.14 ± 3.24*** 5.80 ± 1.02** 0.75 ± 0.28*
Group V n = 8 5.31 ± 0.47** 1.80 ± 0.54* 3.47 ± 0.37* 6.34 ± 1.14***
NB: Isoproterenol was given at the dose of 5 mg/kg daily s.c. and sildenafil was given at the dose of 0.7 mg/kg daily i.p. and Nω-nitro-L-arginine was added to drinking water at the concentration of 10 mg/L.
Values are mean ± SEM. n: number of rats survived for 10 days
One-way ANOVA test followed by Student-Newman-Keuls post-hoc test. Values forming homogeneous subsets are as follow: *homogeneous subset 1, ** homogeneous subset 2 and *** homogeneous subset 3. Significant difference existed between different subsets (P < 0.05)
Correlation of myocardial cGMP level to other parameters
The linear regression curves shown in the figures 2, 3, 4, demonstrate significant correlations between myocardial cGMP level and heart coefficient (r = - 0.65, P < 0.001), myocardial CK activity (r = 0.59, P < 0.001) and serum cTnT level (r = - 0.43, P < 0.01). Individual data from all studied rats were included in the correlation analysis.
Figure 2 Linear regression curve for myocardial cGMP level and heart coefficient.
Figure 3 Linear regression curve for myocardial cGMP and myocardial creatine kinase (CK) activity.
Figure 4 Linear regression curve for myocardial cGMP and serum level of cardiac troponin T (cTnT).
Discussion
Dealing with cardiac hypertrophy is currently a major goal of cardiovascular research and clinical trials[5]. The cardiomyocytes undergo hypertrophy in some pathological conditions that impose overwork on the heart e.g. hypertension, heart valve diseases, myocardial infarction, and cardiomyopathy [16]. Such cardiac hypertrophy is initially compensatory for an increased work load, however, prolongation of this process eventually leads to congestive heart failure, arrhythmia, and sudden death [16]. Todate, drugs targeting the renin-angiotensin system are the chief class of cardiovascular agents of special clinical utility in settings predisposing to cardiac hypertrophy based on the important role of angiotensin II in growth of cardiomyocytes [17].
In an attempt to examine other signaling pathways and other drugs, the present study probed the NO-cGMP pathway and the effect of its modulation by the selective PDE-5A inhibitor sildenafil in cardiac hypertrophy. The β-adrenergic agonist isoproterenol was subcutaneously injected in rats for 10 days to induce cardiac hypertrophy. Obtained results showed that sildenafil administration one hour before daily s.c. injection of isoproterenol was associated with significant improvement in survival and significant inhibition of cardiac hypertrophy. Also, the myocardium of the sildenafil-isoproterenol treated rats showed higher activity of myocardial CK activity and less cTnT leaked into the blood compared to isoproterenol-treated rats. These findings suggest that sildenafil conferred a significant anti-hypertrophic and cytoprotective effect on cardiomyocytes. Conversely, the decline in myocardial CK activity and increased leak of cTnT into serum observed in the isoproterenol treated rats and in rats subjected to NOS inhibition denote significant impairment in tissue energy metabolism and loss of cell integrity, respectively. In the myocardium, ATP is synthesized mainly in the mitochondria through oxidative phosphorylation and transported to the contractile apparatus, where it is consumed by myosin ATPase to generate force. The creatine kinase system plays an important role in myocardial energy metabolism by maintaining ADP levels high at the mitochondria, where ATP is generated, and low at sites of ATP utilization [18]. This is postulated to contribute to the maintenance of a high free energy of ATP hydrolysis, thereby enhancing the efficiency of the energy utilization processes [18]. In addition, a CK shuttle has been proposed, in which high-energy phosphate transport within the cell is facilitated by the higher diffusibility of creatine and phosphocreatine relative to ADP [18].
As regards cardiac troponins T, I and C, they regulate muscle contraction by modulating calcium-dependent interaction of actin and myosin. These intracellular structural proteins are released into circulation following loss of cell integrity [19].
An interesting observation in the present study was that, in the presence of the cNOS inhibitor L-NNA, sildenafil was deprived from its antihypertrophic and cytoprotective effects in isoproterenol-treated rats. Also, the significant correlations found between tissue cGMP level on one hand and heart coefficient, myocardial CK activity and serum level of cTnT on the other hand, suggest that the integrity of the NO-cGMP signaling pathway plays a pivotal role in the cardioprotective effect of sildenafil.
The second messenger cGMP that was identified almost 40 years ago, is generated from GTP either by soluble guanyl cyclase or particulate guanyl cyclase [20]. The former is activated by nitric oxide (NO) or carbon monoxide, where as the latter binds a family of natriuretic peptides consisting of atrial, brain, and C-type natriuretic peptides [20]. On the other hand, various phosphodiesterases regulate cGMP catabolism, including PDE-5A, PDE-6, PDE-9A, PDE-10A, and PDE-11A according to their tissue specificity [21-23]. Among these, PDE-5A is the most widely studied, and its inhibition is a primary mechanism for efficacy of sildenafil in erectile dysfunction [1]. Previous studies have shown that inhibition of PDE-5A in the myocardium enhanced coronary blood flow during exercise-induced ischemia, blunted cardiac stimulation by dobutamine and reduced contractility of adrenergically stimulated papillary muscle [22-24]. In accordance with this, inhibition of PDE-5A in rat myocardium probably underlies the cardioprotective effect of sildenafil against isoproterenol-induced cardiac hypertrophy observed in the present study. At cellular level, the common denominator in all these studies is probably the increased cellular level of cGMP.
In non stimulated hearts, cGMP has been suggested to augment contractile function at low concentrations, likely via cross-talk with cAMP-dependent signaling, inhibiting PDE-3 and degradation of cAMP [25]. At higher concentrations, cGMP has a negative inotropic effect by antagonizing cAMP via protein kinase G (in mammals) or PDE-2 stimulation (in amphibians) [25]. With β-adrenergic activation, both cAMP and cGMP synthesis increase, with the net effect of cGMP being negative on the inotropic response (a brake) [26]. Conversely, reducing cGMP level e.g. by NOS inhibition, enhances β-adrenergic responsiveness [27]. In addition to the negative inotropic action of cGMP, it reduces oxygen consumption and offsets the development of cardiac hypertrophy [28,29].
Consequently, PDE-5A inhibition by sildenafil and cellular accumulation of cGMP would be the braking force against isoproterenol-induced cardiac hypertrophy found in the present study. Consistent with this assumption, over-expression of the catalytic fragment of the constitutively active guanylate cyclase domain of the atrial natriuretic peptide receptor in mouse heart increased the intracellular concentration of cGMP within cardiomyocytes and attenuated the effects of isoproterenol on cardiac wall thickness and prevented fetal gene expression program normally associated with cardiac hypertrophy [30]. Conversely, disruption of cardiac guanylate cyclase-A (GC-A) gene resulted in mice that displayed elevation of blood pressure, cardiac fibrosis and hypertrophy [31].
Conclusion
The present study suggests that sildenafil possesses a cardioprotective and antihypertrophic effect against isoproterenol-induced myocardial injury. Inhibition of cGMP degradation by sildenafil with a consequent accumulation of this signaling molecule may act as a negative regulator against cardiac hypertrophy in-vivo. Understanding the downstream molecular mechanism(s) of such effect of sildenafil, may expand the utility of this drug beyond the current use for treatment of erectile dysfunction in men. The study also revealed that integrity of function of cNOS is an essential prerequisite for the cardioprotective effect of sildenafil in the adrenergically stimulated heart.
Methods
Experimental animals
The present study was conducted on 70 male albino rats of 5 months age, weighing 200–230 g/rat, from those bred in the animal house of the Pharmacology Department, Faculty of Medicine, Alexandria University, Egypt. Rats were kept in galvanized iron cages in groups of 2–3 rats/cage, at room temperature of 22–25°C and allowed free access to standard chow diet and tap water. The study was carried out in accordance with the local guidelines for animal experimentation.
Drugs, chemicals and kits
Isoproterenol and chemicals were purchased from Sigma-Aldrich, Inc. (USA). Sildenafil tablets of Pfizer Inc. (Viagra 100 mg/tablet) were crushed and dissolved in saline for intraperitoneal injection. Direct cGMP enzyme immunoassay kit (CG-200, Sigma-Aldrich, Inc. USA) was used to measure myocardial cGMP level. Cardiac troponin T enzyme immunoassay kit (ES300, Boehringer Mannheim Immunodiagnostics, Germany) was used to measure serum cTnT.
Nω-nitro-L-arginine (L-NNA) was purchased from Cayman chemical, USA.
Experimental groups and treatment
Rats were randomly assigned into 5 groups:
Group I (n = 10): A control group that received vehicle (saline) instead of drug injection.
Group II (n = 15): Isoproterenol was injected s.c. at a dose of 5 mg/kg/day [10].
Group III (n = 15): Sildenafil was injected i.p. at a dose of 0.7 mg/kg/day, one hour prior to isoproterenol. The drug was given in an experimental dose approximating, on a mg/kg basis, the clinical dose of 50 mg administered to a 70-kg patient as described by Okcaili et al [2].
Group IV (n = 15): Sildenafil was injected i.p. at mentioned dose, one hour prior to s.c. injection of saline instead of isoproterenol. This group served as a control for Group III.
Group V (n = 15): Drugs given to this group were essentially similar to group III, however, the NOS inhibitor Nω-nitro-L-arginine (L-NNA) was added to drinking water of this group at the concentration of 10 mg/L. This concentration was selected on the previously reported insignificant effect on blood pressure in rats [11].
Determination of survival, serum level of cTnT and heart coefficient
The study continued for 10 days [10]. The number of survivors in each group was recorded daily and on day 11th, 3 ml-trunkal blood sample/rat was collected under light ether anesthesia from the inferior vena cava and centrifuged for 15 min at 3000 r.p.m. Sera were kept frozen at -20°C until assayed for cardiac troponin T according to general principles of ELISA technique [12] and the manufacturer's instructions (Boehringer- Mannheim Immunodiagnostics, Germany), with a lower limit of detection of 0.01 ng/ml. The means ± SEM were determined for triplicate samples.
From each rat, the heart was excised immediately after rat sacrifice and weighed to calculate the ratio of heart weight to body weight (heart coefficient) [10].
Determination of myocardial creatine kinase (CK) activity
Tissue extract of apical myocardium was prepared by homogenization in cold 20 mM Tris/HCl buffer at the ratio of 1:3 (w/v), pH 7.5, containing 1 mM EDTA and 1 mM β-mercaptoethanol. Cellular debris were removed by centrifugation at 4°C for 60 minutes at 15.000 r.p.m. and the supernatant was used for the assay of CK activity spectrophotometrically at 340 nm, as recommended by the International Federation of Clinical Chemistry [13]. Creatine kinase is primarily concerned with ATP regeneration through catalyzing the following reaction:
Adenosine diphosphate (ADP) + phosphocreatine ATP + Creatine.
The present assay of CK activity was based on the formation of ATP linked to the production of NADHP via hexokinase and glucose-6-phosphate dehydrogenase. The reaction mixture contained 100 mM imidazole acetate, 2 mM EDTA, 10 mM magnesium acetate, 2 mM ADP, 5 mM AMP, 20 mM D-glucose, 2 mM NADP, 30 mM phosphocreatine, hexokinase (3 U/ml), and glucose-6-phosphate dehydrogenase (2 U/ml), pH 6.7. One unit CK activity is that converts one micromole of creatine phosphate to creatine per minute at the given pH.
The protein concentration in supernatant tissue extract was determined by Lowry method [14].
Determination of myocardial cGMP level
Frozen myocardial tissue samples in liquid nitrogen were ground to a fine powder in a stainless steel mortar. After the liquid nitrogen was evaporated, the frozen tissue was weighed and homogenize in 10 volumes of 0.1 M HCl to stop the action of phosphodiesterases. Centrifugation was done at 30.000 r.p.m. at room temperature and the supernatant was then collected for quantitative immunoassay of cGMP level according to general principle of ELISA technique [12] and the manufacturer's instructions (CG-200, Sigma-Aldrich, Inc. USA). The assay used a polyclonal antibody to cGMP to bind, in a competitive manner, cGMP in samples and standards or cGMP covalently attached to alkaline phosphatase. After incubation with the p-nitrophenyl phosphate substrate, a microplate autoreader (Bio-Tek Instruments EL311) was used to measure the intensity of the bound yellow colour at 405 nm. The measured optical density was used to calculate the concentration of cGMP in samples by interpolation from Logit-Log paper plot of the percent bound (B/Bo) versus concentration of cGMP for the standards.
Statistical analysis
Data were analyzed by SPSS/PC, version 9 Chicago/L software. The survival curves were generated using the method of Kaplan and Meier, and the log-rank test was used to detect significant difference between survival curves. Values of other parameters were expressed as mean ± SEM (standard error of mean). Other parameters were compared by one-way ANOVA, with post-hoc comparisons by Student Newman-Keuls test. Correlation studies were carried out by linear regression with curve estimation. For all of the determinations, P < 0.05 was used to indicate statistical significance [15].
Abbreviations
ADP, adenosine diphosphate; ATP, adenosine triphosphate; cAMP, cyclic adenosine monophosphate; cGMP: cyclic guanosine monophosphate; CK, creatine kinase; cTnT: cardiac troponin T; GC, guanyl cyclase; s.c., subcuteanous; i.p., intraperitoneal; NO, nitric oxide; cNOS, constitutive nitric oxide synthase; PDE, phosphodiesterase; w/v, weight/volume.
Authors' contributions
MAHH designed the study, supervised drug administration and collection of samples, estimated the heart coefficient and participated in the biochemical procedures for cGMP, CK and cTnT measurement. MAHH also performed the statistical analysis and prepared the manuscript. AFK participated in the biochemical procedures and in the preparation of the manuscript.
==== Refs
Andersson KE Pharmacology of penile erection Pharmacol Rev 2001 53 417 450 11546836
Ockaili R Salloum F Hawkins J Kukreja RC Sildenafil (Viagra) induces powerful cardioprotective effect via opening of mitochondrial KATP channels in rabbits Am J Physiol 2002 283 H1263 H1269
Salloum F Yin C Xi L Kukreja RC Sildenafil induces delayed preconditioning through inducible nitric oxide synthase-dependent pathway in mouse heart Circ Res 2003 92 595 597 12637371 10.1161/01.RES.0000066853.09821.98
Das S Maulik N Das DK Kadowitz PJ Bivalacqua TJ Cardioprotection with sildenafil, a selective inhibitor of cyclic 3',5'-monophosphate-specific phosphodiesterase 5 Drugs Exp Clin Res 2002 28 213 219 12776574
Swynghedauw B Molecular mechanisms of myocardial remodeling Physiological Rev 1999 79 216 248
Calderone A Thaik CM Takahashi N Chang DL Colucci WS Nitric oxide, atrial natriuretic peptide and cyclic GMP inhibit the growth-promoting effects of norepinephrine in cardiac myocytes and fibroblasts J Clin Invest 1998 10 812 818 9466976
Holtwick R van Eickels M Skryabin BV Baba HA Bubikat A Begrow F Schneider MD Garbers DL Kuhn M Pressure-independent cardiac hypertrophy in mice with cardiomyocyte-restricted inactivation of the atrial natriuretic peptide receptor guanylyl cyclase-A J Clin Invest 2003 111 1399 1407 12727932 10.1172/JCI200317061
Mudersu F Elsner D Animal models of chronic heart failure: Pharmacol Res 2000 41 605 612 10816329 10.1006/phrs.1999.0652
Teerlink JR Pfeffer JM Pfeffer MA Progressive ventricular remodeling in response to diffuse isoproterenol-induced myocardial necrosis in rats Circ Res 1994 75 105 113 8013068
Shi YR Bu DF Qi YF Gao L Jiang HF Pang YZ Tang CS Du JB Dysfunction of myocardial taurine transport and effect of taurine supplement in rats with Isoproterenol-induced myocardial injury Acta Pharmacol Sin 2002 23 910 918 12542050
Verhagen AMG Attia DMA Koomans HA Joles JA Male gender increases sensitivity to proteinuria induced by mild NOS inhibition in rats: role of sex hormones Am J Physiol Renal Physiol 2000 279 F664 F670 10997916
Porstmann T Kiessing ST Enzyme immunoassay techniques. An overview J Immunol Methods 1992 150 5 21 1613258 10.1016/0022-1759(92)90061-W
Horder M Elser RC Gerhardt W Mathieu M Sampson EJ Approved recommendation on IFCC methods for the measurement of catalytic concentration of enzymes: part 7: IFCC method for creatine kinase Eur J Clin Chem Clin Biochem 1991 29 435 456 1932364
Lowry OH Rosenbrough NT Farr AL Randall RJ Protein measurements with the Folin phenol reagent J Biol Chem 1951 193 265 275 14907713
Field A Discovering statistics: using SPSS for Windows 2000 Sage Publications London
Lorell BH Carabello BA Left ventricular hypertrophy: pathogenesis, detection, and prognosis Circulation 2000 102 470 479 10908222
Carson P Giles T Higginbotham M Hollenberg N Kannel W Helmy M Siragy HM Angiotensin receptor blockers: Evidence for preserving target organs Clin Cardiol 2001 24 183 190 11288962
Wallimann T Wyss M Brdiczka D Nicolay K Eppenberger HM Intracellular compartmentation, structure and function of creatine kinase isoenzymes in tissues with high and fluctuating energy demands: the "phosphocreatine circuit" for cellular energy homeostasis Biochem J 1992 281 21 40 1731757
O'Brien PJ Dameron GW Beck ML Kang YJ Erickson BK Di Battista TH Miller KE Jackson KN Mittelstadt S Cardiac troponin T is a sensitive, specific biomarker of cardiac injury in laboratory animals Lab Anim Sci 1997 47 486 495 9355091
Feil R Lohmann SM de Jonge H Walter U Hofmann F Cyclic GMP-dependent protein kinases and the cardiovascular system: insights from genetically modified mice Circ Res 2003 93 907 916 14615494 10.1161/01.RES.0000100390.68771.CC
Conti M Jin SL The molecular biology of cyclic nucleotide phosphodiesterases Prog Nucleic Acid Res Mol Biol 1999 63 1 17 10506827
Traverse JH Chen YJ Du R Bache RJ Cyclic nucleotide phosphodiesterase type 5 activity limits blood flow to hypoperfused myocardium during exercise Circulation 2000 102 2997 3002 11113052
Senzaki H Smith CJ Juang GJ Isoda T Mayer SP Ohler A Paolocci N Tomaselli GF Hare JM Kass KA Cardiac phosphodiesterase 5 (cGMP-specific) modulates β-adrenergic signaling in vivo and is down-regulated in heart failure FASEB J 2001 15 1718 1726 11481219 10.1096/fj.00-0538com
Mohan P Brutsaert DL Paulus WJ Sys SU Myocardial contractile response to nitric oxide and cGMP Circulation 1996 93 1223 1229 8653845
Vila-Petroff MG Younes A Egan J Lakatta EG Sollott SJ Activation of distinct cAMP-dependent and cGMP-dependent pathways by nitric oxide in cardiac myocytes Circ Res 1999 84 1020 1031 10325239
Balligand JL Regulation of cardiac beta-adrenergic response by nitric oxide Cardiovasc Res 1999 43 607 620 10690332 10.1016/S0008-6363(99)00163-7
Hare JM Givertz MM Creager MA Colucci WS Increased sensitivity to nitric oxide synthase inhibition in patients with heart failure: potentiation of beta-adrenergic inotropic responsiveness Circulation 1998 97 161 166 9445168
Straznicka M Gong G Yan L Scholz PM Weiss HR Cyclic GMP protein kinase mediates negative metabolic and functional effects of cyclic GMP in control and hypertrophied rabbit cardiac myocytes J Cardiovasc Pharmacol 1999 34 229 236 10445674 10.1097/00005344-199908000-00008
Calderone A Thaik CM Takahashi N Chang DF Colucci WS Nitric oxide, atrial natriuretic peptide, and cyclic GMP inhibit the growth-promoting effects of norepinephrine in cardiac myocytes and fibroblasts J Clin Invest 1998 101 812 818 9466976
Zahabi A Picard S Fortin N Reudelhuber TL Deschepper CF Expression of constitutively active guanylate cyclase in cardiomyocytes inhibits the hypertrophic effects of isoproterenol and aortic constriction on mouse hearts JBC 2003 278 47694 47699 10.1074/jbc.M309661200
Kishimoto I Rossi K Garbers DL A genetic model provides evidence that the receptor for atrial natriuretic peptide (guanylyl cyclase-A) inhibits cardiac ventricular myocyte hypertrophy PNAS 2001 98 2703 2706 11226303 10.1073/pnas.051625598
| 15813973 | PMC1131906 | CC BY | 2021-01-04 16:32:59 | no | BMC Pharmacol. 2005 Apr 6; 5:10 | utf-8 | BMC Pharmacol | 2,005 | 10.1186/1471-2210-5-10 | oa_comm |
==== Front
BMC PhysiolBMC Physiology1472-6793BioMed Central London 1472-6793-5-71587781110.1186/1472-6793-5-7Research ArticlePhrenic nerve afferents elicited cord dorsum potential in the cat cervical spinal cord Chou Yang-Ling [email protected] Paul W [email protected] Department of Physiological Sciences, Box 100144, HSC, University of Florida, Gainesville FL 32610, USA2005 6 5 2005 5 7 7 20 10 2004 6 5 2005 Copyright © 2005 Chou and Davenport; 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 diaphragm has sensory innervation from mechanoreceptors with myelinated axons entering the spinal cord via the phrenic nerve that project to the thalamus and somatosensory cortex. It was hypothesized that phrenic nerve afferent (PnA) projection to the central nervous system is via the spinal dorsal column pathway.
Results
A single N1 peak of the CDP was found in the C4 and C7 spinal segments. Three peaks (N1, N2, and N3) were found in the C5 and C6 segments. No CDP was recorded at C8 dorsal spinal cord surface in cats.
Conclusion
These results demonstrate PnA activation of neurons in the cervical spinal cord. Three populations of myelinated PnA (Group I, Group II, and Group III) enter the cat's cervical spinal segments that supply the phrenic nerve
==== Body
Background
The diaphragm is innervated by the phrenic nerve. The well-studied motor innervation of the diaphragm by the phrenic nerve arises from motor neurons in the ventral horn of the cervical spinal cord (Jammes et al., 1995). The phrenic nerve motor innervation originates from C4 to C8 spinal segments in the cat. The diaphragm also has afferent innervation carried to the central nervous system by the phrenic nerve. There are both myelinated and non-myelinated afferents in the diaphragm. The myelinated afferents have conduction velocities consistent with Group Ia, Ib and II afferents. The diaphragm has relatively few Group Ia muscle spindles but a relatively large percentage of Group Ib golgi tendon organs (Duron et al., 1978; Corda et al., 1965a; Goshgarian et al., 1986). Group II mechanoreceptors have also been reported (Corda et al., 1965a). Thus, the diaphragm has innervation with afferents that provide muscle mechanical feedback to the CNS via the phrenic nerve. However, physiological evidence of phrenic afferent activation of the spinal cord dorsal horn is lacking.
The role of phrenic afferents in the regulation of diaphragm function has been studied with early observations suggesting that phrenic afferents in the diaphragm are not involved in controlling the respiratory muscle activity (Sant'Ambrogio et al., 1962; Corda et al., 1965b; Kohrman et al., 1947; Landau et al., 1962). However, electrical and mechanical stimulation of phrenic afferents are reported to activate thalamic neurons (Zhang et al., 2003) and elicit neural activity in the cat somatosensory cortex (Davenport et al., 1985). In addition, Zechman et al (1985) reported a correlation of transdiaphragmatic pressure (Pdi) and the perception of inspiratory loads. Knafelc and Davenport reported that increased Pdi correlated with the amplitude of the respiratory related evoked potential recorded from the human somatosensory cortex. Thus, there appears to be a projection of diaphragmatic afferents to the CNS and mechanical changes in the diaphragm correlate with somatosensory activation of the cerebral cortex. However, PnA activation of spinal sensory pathway(s) remains unknown.
Anatomical studies have found that phrenic nerve afferents (of unknown type) terminate in the dorsal horn lamina I-IV of C4 and C5 spinal cord in rat (Goshgarian et al., 1986; Malakhova et al., 2001). In a brief report, Larnicol et al (1984) reported immunoflourescent evidence of dorsal root entry of phrenic nerve afferents in the dorsal horn of the cervical spinal cord. Gill et al (1963) also demonstrated phrenic motorneuron activities were elicited by segmental phrenic nerve afferents. Corda et al (1965) demonstrated that the cervical spinal dorsal rootlets contain diaphragmatic mechanoreceptors, including muscle spindles and Golgi tendon organs. Electrophysiological studies have confirmed that Group I and Group II phrenic nerve afferents project to lateral reticular nucleus (Macron et al., 1985), external cuneate nucleus (Marlot et al., 1985), and both ventral and dorsal respiratory-related areas of brainstem (Macron et al., 1986; Speck et al., 1987). Moreover, in anesthetized cats, short-latency responses have been recorded in the thalamus (Zhang et al., 2003) and somatosensory cortex (Davenport et al., 1985) after electrical stimulation of the phrenic nerve afferents and mechanical probing of the diaphragm. If diaphragmatic proprioceptors project to higher somatosensory brain centers via pathways similar to Group Ia and Group Ib receptors (Landford and Schmidt, 1983), then phrenic afferents should enter the spinal cord ipsilaterally through the dorsal roots of the cervical spinal segments, terminate on dorsal horn neurons, dorsal horn neurons should then project centrally via the dorsal columns to the brainstem and then project to the somatosensory cortex and other supraspinal structures. If this is the phrenic afferent pathway to the somatosensory cortex, then electrical stimulation of the phrenic nerve will stimulate phrenic afferents and activate dorsal horn neurons. However, the activation of the cervical dorsal horn by stimulation of phrenic afferents has not been reported.
One method to investigate dorsal horn neuronal activation by afferent stimulation is recording the cord dorsum potential (CDP) (Yates et al., 1985). Simultaneous stimulation of peripheral afferents activates groups of dorsal horn neurons. The activation of a group of neurons produces a dipole referenced to the surface of the spinal cord. This dipole is the result of a change in polarity of the activated neurons creating a current flow with the cord surface. Thus, neurons in the dorsal horn of the spinal cord generate a field potential when an afferent volley arrives. These negative voltage evoked potentials generated from the dorsal horn neurons of the spinal cord by stimulation of peripheral nerves have been extensively studied for limb afferents (Yates et al., 1982; Manjarrez et al., 2002). The CDP was first discovered and described by Grasser and Graham (1933) as they recorded complex evoked potentials from the surface of the spinal cord. This potential appeared to be largest in the dorsal horn grey matter and is generated by a synchronous activation of a population of dorsal horn neurons that respond to stimulation of low-threshold cutaneous afferents.
Stimulation of nerve afferents at an intensity that activates only Group I muscle afferents has been shown to evoke a dorsal cord field potential consisting of a triphasic spike, a short duration negative wave, and a positive wave. Activation of Group II muscle afferent fibers resulted in a second short duration negative component of the CDP (Bernhard, 1953; Coombs et al., 1956). When a nerve is stimulated sufficiently distal to the spinal cord, the depolarization of the dorsal horn neurons occurs sequentially as a function of the arrival of afferents with different conduction velocities. Thus, it has been shown that the triphasic spike of the CDP occurs because of a separation of activation due to the arrival of Group Ia, Group Ib and Group II afferents. The different peak latencies allow for the determination of different populations of activated afferents.
We reasoned that, if the phrenic nerve contains Group Ia, Group Ib and Group II afferents, then stimulating the phrenic nerve as far distal from the spinal cord as possible (near the diaphragm) would elicit multiple CDP peaks. In addition, we hypothesized that if phrenic nerve afferents enter a cervical segment of the spinal cord, then stimulating PnA will elicit a CDP in that segment. However, the CDP for phrenic nerve afferents has never been reported. Therefore, recording the phrenic afferent CDP was hypothesized to provide evidence of segmental dorsal horn activation by phrenic afferents. This study recorded the CDP from C4 to C7 elicited by stimulating PnA to determine the cervical spinal segmental distribution of PnA elicited CPD, to characterize CDP latencies and infer the populations of PnA eliciting the CDP in cats.
Results
In all cats, electrical stimulation of the phrenic nerve elicited CDP's recorded at the dorsal surface of C4 to C7 cervical spinal segments, and at rostral, middle and caudal locations within each spinal segment. No CDP was observed in the C8 spinal segment. A primary CDP elicited by stimulation of PnA was observed and recorded in dorsal surface of C4 to C7 cervical spinal segments (Fig. 1). The N1 CDP was recorded in the C4 to C7 spinal segments; whereas three CDP peaks (N1, N2 and N3) were identified only in the C5 and C6 spinal segments (Fig. 2). The distributions of the onset latencies and peak amplitudes of individual cervical spinal cord segments are summarized in (Fig. 3) and (Fig. 4), respectively. The averaged N1 peak latency was 1.7 ± 0.1 ms for all cervical spinal segments. The averaged N2 peak latency was 2.3 ± 0.1 ms and the averaged N3 peak latency was 4.6 ± 0.3 ms in the C5 and C6 segments. The averaged conduction velocity was 94.1 ± 8.6 m/sec for the N1 peak, 70.6 ± 7.5 m/sec for the N2 peak and 35.0 ± 3.8 m/sec for the N3 peak. There was a significant difference between the latencies for N1, N2 and N3 peaks (P < 0.05). The average N1 peak amplitude was 22.8 ± 6.0 μV in the C4 segment and significantly less than the N1 peak amplitude for C5, C6 and C7 spinal segments (p < 0.05). The average N1 peak amplitude was 141.3 ± 12.1 μV in the C5 segment and 146.0 ± 27.2 μV in the C6 segment and not significantly different. The average N1 amplitude was 54.6 ± 30.1 μV in the C7 segment and significantly less than C5 and C6. N2 and N3 peaks were only observed in the C5 and C6 spinal segments and the averaged N2 peak amplitude was 79.2 ± 10.3 μV in the C5 segment and 82.1 ± 6.5 μV in the C6 segment. The averaged N3 peak amplitude was 72.0 ± 4.4 μV in the C5 segment and 69.0 ± 11.3 μV in the C6 segment. The amplitudes of the CDP peaks were significantly different between the N1, N2 and N3 peaks (P < 0.05).
Discussion
Neural activity was elicited by stimulation of PnA in C4 to C7 dorsal cervical spinal cord segments in the present study. The electrical stimulation of PnA served the important purpose of demonstrating the existence and locations of PnA- elicited CDP in C4 to C7 segments of the cat cervical spinal cord. This stimulation of PnA was initially observed by one or three negative peaks depending on different segments recorded. The presence of the three negative peaks in C5 and C6 spinal segments appears to be due primarily to the spinal input from different groups of PnA. The presence of only the N1 peak in C4 and C7 demonstrates a different number of afferent populations between the cervical spinal segments that contribute to the cat phrenic nerve.
Central projections of PnA
CDP recordings are evidence of the activation of neurons in the dorsal horn of the cervical spinal cord by PnA. Neural activity in the dorsal surface of the cervical spinal cord elicited by PnA found in this study is consistent with the previous studies of limb muscle afferents (Yates et al., 1985). These results are consistent with PnA entering the cervical spinal cord, projecting to the dorsal horn, which then relays PnA to the brainstem, and projects phrenic afferent information to the somatosensory cortex via a thalamocortical pathway (Zifko et al., 1995; Davenport et al., 1985; Zhang et al., 2003). The function of this putative pathway could be related to the proprioceptive control of respiratory muscles and respiratory movement control originating in the motor cortex (Frazier et al., 1991). Davenport et al (1985) proposed that the PnA projection to the postcruciate region of the cerebral cortex may play a role in higher brain center control of the respiratory pump. The sensory projection sites of the PnA were localized in area 3a and 3b of the sensorimotor cortex in cats. However, the PnA sensory sites in the cortex are not co-localized with motor sites. This means that the cortical regions receiving the sensory information from PnA are separated from the cortical regions of motor output to the phrenic spinal motor neurons. Recordings of evoked potentials using phrenic nerve stimulation provide a unique method for studying the potential pathways for cortical integration of respiratory afferent information and the projections of PnA to the central nervous system (Straus et al., 1997). Although the PnA projection pathways to the cerebral cortex remain unknown, the present study supports a dorsal column mediated pathway that likely involves a multisynaptic thalamic relayed projection (Zhang et al., 2003). It has been shown in the previous studies (Frazier et al., 1991) that C-fiber afferents have much higher threshold and longer latency for PnA stimulation than group Ia, Ib and group II afferents. Conduction velocities for C-fibers are less than 1 m/s. With the length of the phrenic nerve for these cats, the latency for a C-fiber elicited peak of the CDP would be approximately 30 ms, longer than the 25 ms sampling time used for recording the post-stimulus epochs in this study. In addition, the use of 0.1 ms stimulus pulse width does not elicit c-fiber action potentials. Therefore, it is unlikely that C fiber afferents contributed to the CDP in this experiment.
PnA elicited CDP
Results of this study provide evidence that a spinal CDP was elicited by phrenic nerve afferents input to the dorsal horn of the cat cervical spinal cord. This result is consistent with the report from Cuddon et al (1999) that the CDP is a measurement of spinal segmental interneurons and dorsal horn cell function. The rationale behind the CDP measurement as an assessment of sensory nerve afferent and dorsal horn neuronal functions is the phases of the CDP reflect the different population of afferent fibers activating dorsal horn neurons (Yates et al., 1982; Yates et al., 1985). The large negative peak (N1) represents the interneuronal depolarization of spinal dorsal horn neurons elicited by large myelinated sensory afferents (Group Ia & Ib). The activation of a N1 peak in C4 to C7 spinal segments suggests a broadly distributed input to the spinal cord from phrenic group I afferents (Corda, et al., 1965). A N1 peak was not recorded in the C8 spinal segment. There is a second negative peak (N2) that represents the depolarization of spinal dorsal horn neurons by slower conducting myelinated sensory afferents (Group II). A third negative peak (N3) most likely represents the depolarization of spinal dorsal horn neurons elicited by small myelinated sensory afferents (Group III). The presence of N2 and N3 in the C5 and C6 spinal segments suggests a preferential input of group II and group III afferents into these specific spinal segments.
Limb group Ia, Ib and II have been shown to project to somatosensory cortex of cat via a dorsal column pathway (Jones et al., 1980). Corda et al (1965) reported group Ia, Ib and II afferents in the phrenic nerve. It is very likely that PnA have a projection similar to limb proprioceptors. This conjecture is supported in the present study by the CDP elicited in the dorsal surface of cervical spinal cord by stimulation of PnA which is consistent with a dorsal column central neural projection pathway. It is therefore concluded that group Ia, group Ib, group II and possibly group III afferents elicit a CDP that is consistent with an ascending phrenic sensory pathway via the dorsal column and the dorsolateral funiculus (spinocervical tract) of the cervical spinal cord.
Somatosensory PnA pathway
Cortical projection of PnA has been shown in both cortical evoked potentials (Zhang et al., 2003; Davenport et al., 1985) and retrograde fluorescent (Yates et al., 1987) studies. Activation of the somatosensory cortex in cats after electrical stimulation of the contralateral phrenic nerve (Davenport et al., 1986) and intercostals muscles (Davenport et al., 1993) has been reported by this laboratory. One role of the somatosensory projections from phrenic nerve afferents may be to provide the sensory feedback to the cerebral cortex of respiratory pump function. The diaphragm, the intercostal muscles and accessory muscles of respiration provide the inspiratory pumping force for ventilation. Stimulation of PnA has been shown in humans to elicit somatosensory cortical evoked potentials (Zifko et al., 1996). Inspiratory occlusion produces a maximal load on the pumping action of the respiratory muscle and has been reported to elicit somatosensory respiratory related evoked potential (RREP) in humans (Davenport et al., 1986; Davenport et al., 2000) and lambs (Davenport et al., 2001). Knafelc and Davenport reported a correlation between RREP amplitude and the magnitude of the increase in Pdi when graded inspiratory resistive loads were applied in humans. These reports suggest that mechanical loading of the respiratory muscles, including the diaphragm, can elicit somatosensory cortical neural activity. Although the afferents mediating these evoked potentials are unknown, it is likely that respiratory muscle afferents are one population of receptors that mediate these responses. The results presented in the present study are therefore consistent with the hypothesized role of PnA in the somatosensation of inspiratory loads. Thus, neurons in dorsal spinal cord activated by stimulation of PnA may be related to respiratory muscle proprioception, similar to what has been found in other muscle systems.
Conclusion
In summary, the present study recorded a spinal CDP elicited by the activation of phrenic nerve afferents. The PnA project to dorsal horn neurons in the cervical spinal cord from C4 to C7 indicating these segments can function as a relay for the conduction of proprioceptive information from the diaphragm to the higher brain centers in cats. The first peak, N1, conduction velocity is consistent with large myelinated afferent activation, group Ia and group Ib. These PnA enter all the spinal segments that contribute to the phrenic nerve in cats. The second peak conduction velocity is consistent with myelinated afferents, group Ib and large group II. The third peak conduction velocity is consistent with myelinated afferents, group III. The second and third peaks of the CDP were observed only in the primary spinal origin (C5 and C6 segments) of the cat phrenic nerve. The CDP potentials described in this study reflect the first relay by the dorsal spinal cord of the projection of myelinated PnA to the higher brain centers in cats. Therefore, the results of this study support the hypothesis that PnA activation of neurons in the dorsal cervical spinal cord may be involved in the central projection of respiratory muscle afferent information.
Methods
General preparation
Experiments were carried out in adult cats (2.5–3.0 Kg). The University of Florida, Institutional Animal Use and Care Committee reviewed and approved this study. CDP recordings were made from cervical spinal segments C4, C5, C6, C7, and C8 in anesthetized, paralyzed, and artificially ventilated animals. Adult cats (n = 7) of either sex were anesthetized with inhalation of halothane-oxygen. The femoral artery and vein were catheterized. Gas anesthesia was then replaced with α-chloralose by slow i.v. infusion (25 mg/ml). The animals were tracheotomized, vagotomized and placed prone with the head and spine fixed into a stereotaxic apparatus. The body temperature was monitored with a rectal probe and maintained at 38° ± 1°C with the periodic use of a heating pad. Arterial blood pressure, expired CO2 and tracheal pressure were continuously monitored on a polygraph. The animals were connected to a mechanical ventilator and paralyzed (gallamine triethiodide). If fluctuation in blood pressure and heart rate were observed, all experimental procedures were suspended and supplemental anesthesia was administered (0.1 mg/kg iv per dose) until a surgical plane of anesthesia was reestablished. The lungs were periodically inflated to prevent atelectasis. Arterial blood gases and pH were measured and maintained within the normal range. The animals received a continuous infusion of lactated Ringers.
Protocol
The skin and muscles overlying the right lower ribs were incised. The 7th intercostal space was identified and opened by cutting the intercostal muscles. The intercostal space was opened with retractors. The right phrenic nerve caudal to the heart was identified, isolated and dissected free of the surrounding tissue. The phrenic nerve, about 1 cm cranial to its entry into the diaphragm, was placed across bipolar platinum stimulating electrodes. The cathode electrode was about 5–7 mm proximal to the anode electrode. Supramaximal single pulse stimuli (250–500 μA) were delivered at a rate of 0.6 Hz with stimulus duration at 0.1 milliseconds. The pulse simultaneously triggered the signal averager for collecting the 25 msec. post-stimulus spinal activity sample. The entire surgical field was covered with saline soaked gauze.
Spinal laminectomies were performed to expose the dorsal spinal cord from C2-T2. The dura was reflected. The exposed spinal surface was flooded with warm mineral oil. A silver-silver chloride ball electrode was lowered to the pial surface. This electrode was used to record the CDP. The electrode was connected to a high impedance probe, which was connected to an amplifier. The electrical signal was band pass filtered at 3 Hz – 3.0 kHz and amplified. The amplifier output was led into a signal averager (Model 1401, Cambridge Electronic Design, Ltd) At least 64 post-stimulus epochs were sampled at 10 kHz and recorded to provide a minimum of 32 evoked spinal epochs averaged by the computer system to obtain the CDP (Signal2, Model 1401 Cambridge Electronics Ltd.). The recording electrode was placed on the surface of the ipsilateral dorsal spinal cord medial to the dorsal root entry zone of the spinal segments. Each spinal segment was subdivided into 3 recording regions: rostral, middle and caudal based on counting the dorsal roots. The electrode was systematically moved over the spinal surface at each recording site between C4 and C8. The ground electrode was placed over a bony prominence. Phrenic nerve stimulation was performed at each point and the CDP recorded.
Data analysis
A minimum of 32 evoked epochs were averaged by a computer system (Signal2 Cambridge Electronics Ltd.) to obtain the CDP. The averaged CDP's were analyzed for peak presence and polarity (Signal-2, Cambridge Electronics Ltd). The peaks of the CDP were negative voltage changes that were initially identified in C5 (Fig. 1) and C6 recordings. The peaks were identified and then labeled based on latency ranges from the known conduction velocities of phrenic nerve afferents (Corda et al., 1965) and CDP peak analysis reported for limb afferents (Yates et al., 1982). The corresponding onset and peak latencies and amplitudes were determined. The 0-peak amplitudes were measured from the voltage difference between the peak voltage and the averaged voltage of the 5 msec period before the stimulus onset. The initial post-stimulus baseline was corrected for any DC offset. The means and standard errors for onset and peak latencies and amplitudes were then calculated. The onset latency of each CDP was measured from the onset of the stimulus to the start of first peak of the CDP. The latency of the first negative peak (N1) of each CDP was measured as the time from the stimulus to the peak. When present, the latency of the second negative peak (N2) and the third negative peak (N3) were measured in the same manner. The length of phrenic nerve was measured from the stimulating electrodes to the dorsal root entry zone to calculate the conduction velocity of the phrenic nerve afferents. Each component of CDP was averaged and statistical analysis was applied. The mean ± standard deviation (SD) was calculated for all CDP peak latencies. One-way repeated measure ANOVA was used to compare between the N1 peak, N2 peak, and N3 peak latencies and amplitudes from each recording site. The criterion for significance was p < 0.05.
Abbreviations
PnA – Phrenic nerve afferents
CDP – cord dorsum potential
C4, C5, C6, C7, C8 – Cervical spinal segments
CNS – Central nervous system
Pdi – Transdiaphragmatic pressure
CO2 – Carbon dioxide
Authors' contributions
YL performed the data analysis and drafted the manuscript. PD carried out the study and coordinated its design. Both authors read and approved the final manuscript.
Figures and Tables
Figure 1 The cord dorsum potential in response to phrenic nerve afferent stimulation in the C5 segment. The trace was the average of 32 stimulus epochs for the middle C5 segment from one animal. The enlarged area expands the large initial peak of the CDP to better illustrate the N1 and N2 peaks. The first negative wave indicates the electrical stimulus used as the zero time point for peak latency analysis. The stimulus artifact is followed by three negative peaks N1, N2, and N3. The amplitudes of the negative peaks were measured from the voltage difference between the peak voltage and the averaged voltage of the 5 ms period before the stimulus onset.
Figure 2 Representative illustration of phrenic nerve afferents stimulation related cord dorsum potential recorded at C4, C5, C6, and C7 spinal segments in one animal. The traces were the average of 32 stimulus epochs for each spinal segment. The N1 peak was recorded in all cervical spinal segments (C4 to C7); whereas N2 and N3 CDP peaks were identified only in C5 and C6 spinal segments. The amplitudes of the negative peaks were measured from the voltage difference between the peak voltage and the averaged voltage of the 5 ms period before the stimulus onset.
Figure 3 Histogram of the CDP peak latencies in different regions of the cervical spinal segments (C4 to C7) after phrenic nerve afferents stimulation in cats. The intensity of the stimulation was 500 μA. N1 is the first peak latency. N2 and N3 are the second and third peak latencies found in C5 and C6 only. * indicates a significant differences between N1 and N2 peaks. † indicates a significant difference between N2 and N3 peaks in the C5 and C6 segmental locations. # indicates a significant difference between N1 and N3 peaks in the C5 and C6 segmental locations.
Figure 4 Histogram of the CDP peak amplitudes in the cervical spinal segments (C4 to C7) after phrenic nerve afferent stimulation in cats. N1 is the first peak amplitude observed in the C4 to C7 spinal segments; N2 and N3 are the second and third peak amplitudes only observed in the C5 and C6 spinal segments. * indicates a significant differences between N1 and N2 peaks in the C5 and C6 segmental locations. # indicates a significant difference between N1 and N3 peaks in the C5 and C6 segmental locations. ∞ indicates significant differences between the spinal segments for N1 peak.
==== Refs
Bernhard CG The spinal cord potentials in leads from the cord dorsum in relation to peripheral source of afferent stimulation Acta Physiol Scand 1953 29 1 29
Coombs JS Curtis DR Landgren S Spinal cord potentials generated by impulses in muscle and cutaneous afferent fibers J Neurophysiol 1956 19 452 467 13367876
Corda M Ecklund G Von Euler C External intercostal and phrenic α-motor responses to changes in respiratory load Acta Physiol Scand 1965 63 391 400 14324074
Corda M Von Euler C Lennerstrand G Proprioceptive innervation of the diaphragm J Appl Physiol 1965 78 161 177
Cuddon PA Delauche AJ Hutchison JM Assessment of dorsal nerve root and spinal cord dorsal horn function in clinically normal dogs by determination of cord dorsum potentials Am J Vet Res 1999 60 222 226 10048556
Davenport PW Shannon R Mercak A Reep RL Lindsey BG Cerebral cortical evoked potentials elicited by cat intercostals muscle mechanoreceptors J Appl Physiol 1993 74 799 804 8458798
Davenport PW Thompson FJ Reep RL Freed AN Projection of phrenic nerve afferents to the cat sensorimotor cortex Brain Res 1985 328 150 153 3971173 10.1016/0006-8993(85)91334-4
Davenport PW Hutchison AA Cerebral cortical respiratory-related evoked potentials elicited by inspiratory occlusion in lambs J Appl Physiol 2001 93 31 362
Davenport PW Cruz M Stecenko A Kifle Y Respiratory related evoked potentials in children with life-threatening asthma Am J Respir Crit Care Med 2001 161 1830 1835
Davenport PW Friedman WA Thompson FJ Franzen O Respiratory-related cortical potentials evoked by inspiratory occlusion in humans J Appl Physiol 1986 60 1843 1848 3722053
Duron B Jung-Carillol MC Marlot D Myelinated nerve fiber supply and muscle spindles in the respiratory muscles of the cat: a quantitative study Anat Embryol (Berl) 1978 152 171 192 147637 10.1007/BF00315923
Frazier DT Revelette WR Role of phrenic nerve afferents in the control of breathing J Appl Physiol 1991 70 491 496 2022537
Gasser HS Graham HT Potentials produced in the spinal cord by stimulation of dorsal roots Am J Physiol 1933 103 303 320
Gill PK Kuno M Excitatory and inhibitory actions on phrenic motoneurons J Physiol (Lond) 1963 168 274 289 14062677
Goshgarian HG Roubal WR Origin and distribution of phrenic primary afferent nerve fibers in the spinal cord of the adult rat Exp Neurol 1986 92 624 638 3709737 10.1016/0014-4886(86)90304-3
Jammes Y Speck DF Dempsey JA, Pack AI Respiratory control by diaphragmatic and respiratory muscle afferents Regulation of Breathing 1995 2 New York: Dekker 543 582
Jones EG Porter P What is area 3a? Brain Res Rev 1980 2 1 43 10.1016/0165-0173(80)90002-8
Knafelc M Davenport PW Relationship between magnitude estimation of resistive loads, inspiratory pressures, and the RREP P(1) peak J Appl Physiol 1999 87 516 522 10444607
Kohuamn RM Nosasco JB Wiggers CJ Types of afferent fibers in the phrenic nerve Am J Physiol 1947 151 547 553
Landau BE Akert K Robert KS Studies on the innervation of the diaphragm J Comp Neurol 1963 119 1 10 10.1002/cne.901190102
Landford S Schmit RF An electron microscopic analysis of the left phrenic nerve in the rat Anat Rec 1983 205 207 213 6846871 10.1002/ar.1092050211
Larnicol N Dominique R Duron B Identification of phrenic afferents in the dorsal columns: A fluorescent double-labeling study in the cat Neurosci Lett 1984 52 49 52 6527837 10.1016/0304-3940(84)90349-5
Macron JM Marlot D Effects of stimulation of phrenic afferent fiber on medullary respiratory neurons in cat Neurosci Lett 1986 63 231 236 3951749 10.1016/0304-3940(86)90361-7
Macron JM Marlot D Duron B Phrenic afferent input to the lateralmedullary reticular formation of the cat Respir Physiol 1985 59 155 167 2984753 10.1016/0034-5687(85)90004-0
Malakhova OE Davenport PW c-Fos expression in the central nervous system elicited by phrenic nerve stimulation J Appl Physiol 2001 90 1291 1298 11247926
Manjarrez E Perez H Rojas-Piloni JG Velez D Martinez L Flores A Absence of coherence between cervical and lumbar spinal cord dorsal surface potential in the anaesthetized cat Neurosci Lett 2002 328 37 40 12123854 10.1016/S0304-3940(02)00446-9
Marlot D Macron JM Duron B Projections of phrenic nerve to the external cuneate nucleus in the cat Brain Res 1985 327 328 330 2985178 10.1016/0006-8993(85)91529-X
Rudomin P Presynaptic inhibition of muscle spindle and tendon organ afferents in the mammalian spinal cord Trends Neurosci 1990 13 499 505 1703681 10.1016/0166-2236(90)90084-N
Rudomin P Solodkin M Jimenez I Synaptic potentials of primary afferent fibers and motoneurons evoked by single intermediate nucleus interneurons in the cat spinal cord J Neurophysiol 1987 57 288 313
Sant'Ambrogio G Wilson MF Frazier DT Somatic afferent activity in reflex regulation of diaphragmatic function in the cat J Appl Physiol 1962 17 829 832
Speck DF Revelette WR Excitation of dorsal and ventral respiratory group neurons by phrenic nerve afferents J Appl Physiol 1987 62 300 307 3558189
Straus C Zelter M Derenne JP Pidoux B Willer JC Similowski T Putative projection of phrenic afferents to the limbic cortex in humans studied with cerebral-evoked potential J Appl Physiol 1997 82 280 290
Yates BJ Thompson FJ Properties of spinal cord processing of femoral venous afferent input revealed by analysis of evoked potentials J Auton Nerv Syst 1985 14 201 207 4067182 10.1016/0165-1838(85)90076-1
Yates BJ Mickle JP Hedden WJ Thompson FJ Tracing of afferent pathways from the femoral-saphenous vein to the dorsal root ganglia using transport of horseradish peroxidase J Auton Nerv Syst 1987 20 1 11 3655181 10.1016/0165-1838(87)90076-2
Yates BJ Thompson FJ Mickle JP Origin and properties of spinal cord field potentials Neurosurg 1982 11 439 350
Yates JS Davenport PW Reep RL Thalamocortical projections activated by phrenic nerve afferents in the cat Neurosci Lett 1994 180 114 118 7535404 10.1016/0304-3940(94)90500-2
Zechman FW Muza SR Davenport PW Wiley RL Shelton R Relationship of transdiaphragmatic pressure and latencies for detecting added inspiratory loads J Appl Physiol 1985 58 236 243 3968013
Zhang WR Davenport PW Activation of thalamic ventroposteriolateral neurons by phrenic nerve afferents in cats and rats J Appl Physiol 2003 94 220 226 12391131
Zifko UA Young BG Remtulla H Bolton CF Somatosensory evoked potentials of the phrenic nerve Muscle Nerve 1995 18 1487 1489 7477077 10.1002/mus.880181224
Zifko UA Slomka PJ Reid RH Young GB Remtulla H Bolton CF The cortical representation of somatosensory evoked potentials of the phrenic nerve J Neuro Sci 1996 139 197 202 10.1016/S0022-510X(96)00055-X
| 15877811 | PMC1131907 | CC BY | 2021-01-04 16:03:51 | no | BMC Physiol. 2005 May 6; 5:7 | utf-8 | BMC Physiol | 2,005 | 10.1186/1472-6793-5-7 | oa_comm |
==== Front
BMC SurgBMC Surgery1471-2482BioMed Central London 1471-2482-5-101586562310.1186/1471-2482-5-10Research ArticleEarly efficacy of CABG care delivery in a low procedure-volume community hospital: operative and midterm results Papadimos Thomas J [email protected] Robert H [email protected] Anoar [email protected] Thomas A [email protected] Christopher J [email protected] Samuel J [email protected] Aamir [email protected] Department of Anesthesiology, Medical College of Ohio, 3000 Arlington Avenue, Toledo, OH 43614, USA2005 2 5 2005 5 10 10 27 1 2005 2 5 2005 Copyright © 2005 Papadimos et al; licensee BioMed Central Ltd.2005Papadimos 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 Leapfrog Group recommended that coronary artery bypass grafting (CABG) surgery should be done at high volume hospitals (>450 per year) without corresponding surgeon-volume criteria. The latter confounds procedure-volume effects substantially, and it is suggested that high surgeon-volume (>125 per year) rather than hospital-volume may be a more appropriate indicator of CABG quality.
Methods
We assessed 3-year isolated CABG morbidity and mortality outcomes at a low-volume hospital (LVH: 504 cases) and compared them to the corresponding Society of Thoracic Surgeons (STS) national data over the same period (2001–2003). All CABGs were performed by 5 high-volume surgeons (161–285 per year). "Best practice" care at LVH – including effective practice guidelines, protocols, data acquisition capabilities, case review process, dedicated facilities and support personnel – were closely modeled after a high-volume hospital served by the same surgeon-team.
Results
Operative mortality was similar for LVH and STS (OM: 2.38% vs. 2.53%), and the corresponding LVH observed-to-expected mortality (O/E = 0.81) indicated good quality relative to the STS risk model (O/E<1). Also, these results were consistent irrespective of risk category: O/E was 0, 0.9 and 1.03 for very-low risk (<1%), low risk (1–3%) and moderate-to-high risk category (>3%), respectively. Postoperative leg wound infections, ventilator hours, renal dysfunction (no dialysis), and atrial fibrillation were higher for LVH, but hospital stay was not. The unadjusted Kaplan-Meier survival for the LVH cohort was 96%, 94%, and 92% at one, two, and three years, respectively.
Conclusion
Our results demonstrated that high quality CABG care can be achieved at LVH programs if 1) served by high volume surgeons and 2) patient care procedures similar to those of large programs are implemented. This approach may prove a useful paradigm to ensure high quality CABG care and early efficacy at low volume institutions that wish to comply with the Leapfrog standards.
==== Body
Background
For many years investigators have studied the relationship between the outcomes of high procedure-volume institutions with those of low procedure-volume institutions. This has been especially true for coronary artery bypass grafting (CABG) surgery with varied results [1-11]. Indeed, recent studies have indicated that low volume-procedure institutions performing CABG surgery may have good outcomes particularly if associated with high volume surgeons (> 125 cases per year) [12-15].
In this report we review our experience with a recently initiated small community hospital cardiac surgery program modeled after a similar high volume practice in the region. In this paradigm a group of high volume cardiac surgeons expanded their practice, protocols, and procedures to include the smaller institution. We reasoned that this paradigm can be used successfully by low volume hospitals. We tested this contention by comparing the operative and midterm results of an LVH to the national data as reported for the Society of Thoracic Surgeons (STS) National Cardiac Database over the same period.
Methods
Patients
We retrospectively reviewed the initial 504 patients of a new cardiac surgery program undergoing isolated CABG between February 1, 2001 – December 31, 2003 at Saint Luke's Hospital (Maumee, Ohio), a 189-bed community hospital in Northwest Ohio. The information was extracted from a local database. Patients were excluded if they had concomitant valve, other cardiac, or carotid surgery. Institutional Review Board approval was obtained for this ongoing clinical cardiac surgery database research. The requirement for informed consent was waived by the Institutional Review Board.
Clinical data / end points
Clinical data on patients undergoing revascularization have been systematically abstracted and recorded in a dedicated cardiac surgery database, and are regularly reported to the STS national cardiac surgery database. The primary end points were operative mortality and morbidity (complications and length of hospital stay), and those were compared to the STS 2001–2003 data. In addition, LVH 0- to 3-year all-cause mortality for survival data was collected and combined via the Social Security Death Index (conducted in August 2004). Allowing for a 3-month lag, this corresponds to a follow-up between 5 and 41 months.
Statistical analysis
Continuous data were expressed as mean ± SD. Baseline variables were compared by use of the Wilcoxon rank-sum test, t-test, or the χ2 test as appropriate. Actual, risk-adjusted and observer-to-expected operative mortality (OM, Adj. OM, and O/E, respectively) data are reported as per the latest STS CABG risk model [16]. Effects of explanatory variables on OM were derived by logistic regression. Survival was compared with Kaplan-Meier analysis (log rank test) and multivariable Cox proportional-hazards regression. The latter was done to assess the effects of the varying death hazard on long-term mortality predictors and their associated risk ratios (RR). Regression model selection was done with backward elimination (Wald statistic – confirmed using forward and stepwise selection). A P < 0.05 cutoff was used to indicate significance (SPSS version 10.0, SPSS Inc., Chicago, IL).
Results
A total of 504 CABG procedures were performed in the initial 35 months of this new cardiac surgery program. This LVH population demonstrated a similar distribution of age and gender to that of the STS, whereas race and body mass index differed significantly (Table 1). Our LVH patients had a higher incidence of three vessel disease, obesity, preoperative myocardial infarction and family history of coronary artery disease. The STS cohort exhibited more peripheral vascular disease, chronic obstructive pulmonary disease, angina, preoperative renal failure, hypercholesterolemia, congestive heart failure, and trended toward more cerebral vascular accidents (p = .06). These differences in preoperative morbidity led to a lower STS predicted mortality risk in our population, 2.94% vs. 3.13%. The LVH cohort had a relatively greater incidence of emergencies, use of internal mammary arteries, use of intra-aortic balloon pumps, and blood transfusions. The incidence of elective cases, redo surgery, and off-pump procedures were less than the STS. Total CPB (cardiopulmonary bypass) and cross-clamp times were shorter.
Table 1 Patient Demographics, Risk Factors and Operative Data
Variable Study Site n mean ± SD 2001–2003 % median (25%–75%) STS (2001–03) % median (25%–75%) P-Value
No. of Patients 504 448841
Demographics/Risk Factors
Age (yrs) 64 ± 11 65 (56–72) 66 (57–74)
Male 368 73.0 71.3 .433
Caucasian 474 94.0 87.1 <.001
Black 5 1 5.17 <.001
Hispanic/Other 25 4.96 6.84
BSA (m2) 2.08 ± 0.25 2.08 (1.89–2.26) 1.95 (1.82 – 2.13) <.001
Obese (BMI>35 kg/m2) 103 20.4 13.4 <.001
Current Smoker 101 20 22.1 .295
Family History of CAD 403 80 43.2 <.001
Diabetes 168 33.30 35.0 .469
Insulin-dependent 41 8.13 10.47 .102
Hypercholesterolemia 323 64 68.49 .038
Renal Failure 10 1.98 5.20 .002
Hypertension 358 71 74.78 .060
Peripheral Vascular Disease 35 6.94 15.80 <.001
Cerebrovascular Disease 60 11.90 13.21 .425
COPD 68 13.50 18.61 .004
Myocardial Infarction 266 52.80 45.53 <.001
Congestive Heart Failure 44 8.73 13.80 <.001
Unstable 222 44.10 47.05 .193
Arrhythmia (any) 42 8.30 9.43 .443
Triple Vessel Disease 422 83.70 74.63 <.001
Left Main Disease >50% 118 23.40 24.54 .594
Ejection Fraction (%) 48 ± 11 50 (40–55) 50 (40–60)
Previous CV intervention 116 23.00 20.36 .155
Operative Data
Elective 129 25.60 51.88 <.001
Emergent 63 12.50 4.10 <.001
Redo Surgery 15 2.98 8.93 <.001
No. of Grafts 3.58 ± 1.01
Arterial 1.63 ± 0.98
Vein 1.94 ± 0.98
ITA Used 473 93.80 89.84 .004
Left ITA used 471 93.50 89.34 .004
Aortic Cross-Clamp (min) 58 ± 22 55 (45–68) 63 (47 – 83) .004
Perfusion time (min) 90 ± 31 85 (70–106) 94 (73 – 119)
Off-pump 21 4.17 20.48 <.001
STS Predicted Mortality (%) 2.94 ± 5.0 3.13
Ejection fraction was available in 464 patients. CV intervention = any cardiovascular intervention (surgery, angioplasty, or stenting). The urgent cases (197/504) are not compared in the STS database.
Our LVH operative mortality (OM) did not statistically differ from the STS (2.38% vs. 2.53%). The corresponding O/E morality ratio was 0.81 and adjusted OM was 1.9%. Multivariate OM predictors were age, emergency status, and time on cardiopulmonary bypass (Table 2). The patients in the very low risk (<1%) and low risk (1–3%) categories fared better than the STS with actual mortality rates and O/E mortality ratios of 0%, 0.0 and 0.9%, 0.52, respectively. The moderate to high-risk category (>3%) had an actual mortality rate (7.3%) and an O/E mortality ratio (1.03) that were comparable to the STS.
Table 2 Multivariate predictors of operative mortality by logistic regression applied to 504 patients
95%C.I.
Variables B S.E. Wald Sig. OR Lower Upper
Age (yr) .092 .035 6.942 .008 1.096 1.024 1.173
Emergency 1.312 .674 3.788 .052 3.712 .991 13.908
Time on CPB (min) .026 .007 13.421 .000 1.027 1.012 1.041
Constant -12.977 2.788 21.673 .000 .000
Compared to the STS data, post-operative complications of leg wound infection, prolonged ventilation, renal dysfunction (no dialysis), and atrial fibrillation occurred more frequently at our LVH (Table 3). Also, LVH had a greater rate of 30-day readmissions. However, total length of stay, post-operative length of stay, post-operative ventilator hours, strokes, sepsis, pneumonia, urinary tract infection, and sternal wound infection were similar.
Table 3 Operative Outcomes
Variable Study Site n mean (SD) 2001–03 % median (25%–75%) STS (2001–03) % median (25%–75%) P – value
Intra-aortic Balloon pump (any) 73 14.5 9.2 <.001
Blood Transfusion (Any) 256 50.8 45.3 .014
Complication
ReOp Bleeding 14 2.78 2.50 .798
Perioperative myocardial infarction 3 0.60 1.07 .414
Sternal wound infection 1 0.20 0.47 .578
Leg wound infection 14 2.78 0.70 <.001
Septicemia 7 1.39 1.03 .570
Urinary Infection 9 1.79 1.67 .973
Permanent Stroke 4 0.79 1.53 .242
Transient Stroke 7 1.39 0.93 .406
Post-Op Ventilator (hours) 23.2 80.3 6.3 (4.0–15.8) 21.73
Ventilator Prolonged (>24 hrs) 53 10.5 7.5 .012
Pneumonia 15 2.98 2.87 .989
Renal Failure 34 6.75 3.47 <.001
Atrial Fibrillation 144 28.6 19.9 <.001
Operative Mortality 12 2.38 2.53 .940
Total LOS (days) 8.59 6.11 7 (5–10) 8.97
Post-Op LOS (days) 6.45 5.27 5 (4–7) 6.87
30-day Readmission 59 11.7 8.4 .009
Unadjusted Kaplan-Meier survival was 96%, 94 %, and 92% at 1, 2, and 3 years, respectively (Figure 1), which was comparable to the STS. Effects of six patient variables on survival are shown in Figure 2. Gender and diabetes did not affect midterm survival, while survival for older (> 65 years) patients, cerebral vascular disease, longer time on CPB, and higher STS predicted operative mortality risk all exhibited significantly worse midterm survival.
Figure 1 Kaplin-Meyer survival curve for 504 LVH CABG patients. Bars = standard error.
Figure 2 Effects of gender, age, diabetes, cerebrovascular disease, time on CPB, and operative mortality (OM) on midterm survival. P value reflects log-rank test results.
Predictors of 0–3 year mortality were derived by multivariate (proportional hazard) Cox regression analysis and included longer time on CPB (RR = 1.15), all vein grafts (RR = 5.74), cerebral vascular disease (RR = 3.79), age (RR = 1.63 per 10 years), redo surgery (RR = 3.46), and congestive heart failure (RR = 2.48). Preoperative renal failure approached significance (RR = 3.81, p = .078) (Table 4).
Table 4 Predictors of 0 – 3 year mortality derived by Multivariate (proportional hazard) Cox regression analysis.
Variables B SE Wald Risk Ratio P-value 95% C.I.
Time on CPB (per 10 minutes) 0.02 0.00 16.26 0.0001 1.15 1.08 1.23
All vein grafts 1.75 0.55 10.15 0.0014 5.74 1.96 16.82
Cerebrovascular Disease 1.33 0.43 9.82 0.0017 3.79 1.65 8.72
Age (per 10 years) 0.06 0.02 8.28 0.0040 1.63 1.20 1.91
Redo Surgery 1.24 0.55 5.00 0.0254 3.46 1.17 10.26
Congestive Heart Failure 0.91 0.45 4.07 0.0435 2.48 1.03 6.00
Pre-operative Renal Failure 1.34 0.76 3.11 0.0779 3.81 0.86 16.86
95% C.I = 95% confidence interval; CPB = cardiopulmonary bypass;
Predictors are arranged by decreasing Wald statistic
Discussion
While many researchers have found hospitals with higher CABG volumes to be associated with better outcomes, there has been significant interest in the potentially confounding influence of cardiothoracic surgeon procedure-volumes on this association [1,2,5,11,13-17]. Four states (California, New York, New Jersey, and Pennsylvania), which represent a quarter of the US population, have rigorous reporting systems for CABG procedures based on risk-adjusted mortality alone. In these states the Leapfrog standard includes only those hospitals in the top quartile. These "top" hospitals had overall mortality rates of 1.7% compared with 4.1% in the lower quartiles. In the rest of the nation, volume (>450 cases per annum) and mortality rates below 2.7% comprise the Leapfrog standard [18]. The use of standard risk-adjusted mortality rates may become the benchmark if rigorous outcomes measurement systems are implemented throughout the United States. Risk adjustment makes surgeon procedure volume very important. In this way, certain cases may be restricted to particular procedure volume surgeons, thus patient redistribution to regional centers may not be necessary.
Three recent studies of note have highlighted the critical importance of surgeon procedure volume [13-15]. Peterson et al found that hospital procedure-volume, as a quality marker in CABG surgery, was to be only modestly associated with risk-adjusted CABG mortality rates [14]. In fact, these researchers clearly identified many low-volume hospitals with low mortality rates and several high volume centers with higher than expected rates. In addition, Wu et al indicated that high volume cardiothoracic surgeons were associated with a lower risk of death for both low-risk and moderate-to-high risk CABG patients [15] and Birkmeyer et al demonstrated that the observed associations between hospital volume and operative mortality were modulated by surgeon volume [13].
The establishment of the STS National Cardiac Database has facilitated quality of care comparisons of cardiac surgery programs relative to national statistics [19-21]. The Center for Medicare and Medicaid Services (CMS) and the Leapfrog group have suggested that hospital volume be used as the indicator of the quality of CABG outcomes [22-24]. However, an insightful review of the evidence by Shahian and Normand lends merit to the argument that CABG surgery is so pervasive that its well-recognized techniques and well-understood pathophysiology make provision for its transportability to low volume institutions in the hands of skilled, high volume surgeons [25]. The recent Peterson et al investigation of the STS database accounted for clinical factors, differences in site variability and clustering within sites. They documented that (1) low volume institutions tended to operate more often on patients under emergent conditions, (2) the association between hospital volume and mortality was different for younger (<65 years) vs. older (>65 years) patients, (3) the hospital volume and outcome associations were confounded by the concomitant effect of surgeon volume, and (4) hospital volume per se was a poor predictor of CABG outcome [14].
In view of the above studies the issue of volume as a "proxy" yields itself to discussion. Some have criticized hospital volume as a crude indicator of surgical quality in that hospital volume is only a proxy for low mortality (high quality) [26-29]. Volume must be examined from the perspective of, not only the hospital, but also of the surgeon. The procedure volume of surgeons is an important proxy for quality CABG care. In this light, our example of a high quality, low volume center should not be unexpected, especially when our LVH used high volume surgeons.
Our study site was a small, low procedure volume hospital located in an increasingly affluent suburban community with a large rural catchment area. This accounted for the large number of Caucasians and the under-representation of African-Americans. The increased body surface area and excessive obesity that occurred in our population was to be expected according to findings that indicate Ohio is near the leading edge of the obesity experience in the United States [30,31]. The study site patients had more significant family histories of coronary artery disease (CAD), but we cannot comment as to the extent of the effect of genetic predisposition versus social impact (over-eating, smoking, lack of exercise) had upon this finding. However, the extent of obesity, and family histories of CAD could account for the significant deviation from the STS in regard to triple vessel disease and history of myocardial infarction.
The relatively recent initiation of this cardiac surgery program and its rural catchment area may account for the distribution of cases (fewer elective and redo surgeries, but more emergent cases) because: (1) the cardiac surgery program was actually put into place to support interventional cardiology, (2) the cadre of cardiologists was not as well established at this location and had fewer elective cases posted, (3) the proximity to several rural hospitals allowed rural emergent cases (whether initially presenting to the study site emergently or becoming emergent during their rural hospital stay) to receive expeditious care, (4) the surgeons did not have an office in the immediate area and this affected their contribution of elective cases to the institution, and (5) a newly established program would have less opportunity to perform redo surgeries.
The surgeons at the study location were very experienced and were strong advocates of mammary and radial artery use [32], and this is reflected in their shorter aortic cross-clamp (perfusion) times and their use of internal mammary grafts as compared to the STS. However, the surgical group was less enthusiastic for the use of off-pump coronary artery bypass procedures. The increased use of intra-aortic balloon pumps and blood transfusions in our population may be attributed to the increased severity of illness as represented by an increased percentage of emergent cases, preoperative myocardial infarctions, and triple vessel disease.
The postoperative complications of renal dysfunction (no dialysis), prolonged ventilation greater than 24 hours, leg wound infection, atrial fibrillation, and 30-day readmission deviated from the STS. Postoperative renal dysfunction was not associated with an increased need for dialysis. The higher rate of renal dysfunction may be linked to a higher incidence of emergent cases done soon after cardiac catheterization. Such prompt surgeon response may not allow sufficient time for elimination of the contrast dye that may exacerbate the associated nephropathy [33].
Postoperative atrial fibrillation in our population is greater than the STS, but is within the range reported by others [34,35]. This may be related to the greater incidence of emergent cases and shorter catheterization-to-surgery times. The latter may have compromised the efficacy our implementation of the amiodarone prophylaxis protocol in high-risk atrial fibrillation patients.
The prolonged rate of postoperative ventilation may be a function of anesthesiology practice. In the evening the cardiothoracic anesthesiologist covers the intensive care unit by beeper. Although protocols are in place regarding extubation and the anesthesiologist can be consulted at any hour, we have found that there is reluctance to extubate patients between midnight and 0600 on the part of the staff. However, this did not influence the length of stay or the rate of pneumonia.
The cardiovascular surgical assistants hired at the study location initially had limited experience with the surgical care of leg wounds. Their ability to care intra-operatively and post-operatively for leg wounds has improved and is under a continuous quality improvement process. The extent of obesity and emergent cases in our population may be a contributing factors to poor wound healing. Also, the emergent cases arrive in the operating room on many occasions after administration of antiplatelet drugs and this may contribute to postoperative hematoma formation and infection at the leg wound sites.
An increased 30-day readmission rate of the study site was also noted. The surgeons initiating this cardiac surgery program had heightened awareness as to potential patient complications in a "new" program and did not hesitate to readmit a patient of questionable physiologic status.
In our paradigm a group of high volume surgeons from a high volume hospital (HVH) established a new LVH cardiothoracic surgery program providing practice guidelines, protocols and data acquisition capabilities. In addition, the physical layout of the cardiac surgery wing and the organization of the cardiovascular intensive care unit (CVU) within a dedicated heart center were of paramount importance. It included two cardiac catheterization suites with a 10 bed holding area, two operating suites were immediately adjacent to an eight-bed CVU used as intensive care, step-down and floor beds if capacity permitted ("one-stop" for all patients). An experienced critical care nursing staff was recruited from the HVH. In addition, two anesthesiologists/intensivists with transesophageal echocardiography skills were dedicated exclusively to the heart center. They were responsible for the pre-, intra-, and postoperative critical care of all the patients. The CVU model was one that involved the cardiac surgeon as the primary physician; however the anesthesiologists were consulted on every case to support the global care given to the patient. This was particularly beneficial to the surgeons because their practice involved three institutions in addition to their office practice. Also, daily group rounds included a surgeon, anesthesiologist, nurse, respiratory therapist, and pharmacist. Each month a meeting was held, not only for case review, but to assess the program's performance against the STS database.
Clearly, surgeon procedure-volume examination will cause controversy [36-38], but studies of risk-adjusted data indicate that surgeon volume is of significant importance [13,14,16,17]. In the setting of a LVH where there are more emergent and high risk patients [14] there should be a continuous quality improvement effort in place to ensure "best practice", evidence-based care is offered to patients [12].
Conclusion
We demonstrated that high quality CABG care with good outcomes can be achieved at a LVH program provided that that it is served by high volume cardiac surgeons and backed up by a highly trained, dedicated support team, and a sophisticated data acquisition capability and review process. This approach may prove a useful paradigm to ensure high quality CABG care and early efficacy at low volume institutions that wish to be compliant with Leapfrog recommendations.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
Drs. Papadimos and Habib wrote the abstract, methods, background, discussion, results, discussion and conclusions. Drs. Zacharias, Schwann, Riordan, Durham, and Shah assisted in writing the results section.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
none.
==== Refs
Showstack JA Rosenfeld KE Garnick DW Luft HS Schaffarzick RW Fowles J Association of volume with outcome of coronary artery bypass graft surgery JAMA 1987 257 785 789 3492614 10.1001/jama.257.6.785
Hannan EL O'Donell JF Kilburn H JrBernard HR Yazici A Investigation of the relationship between volume and mortality for surgical procedures performed in New York state hospitals JAMA 1989 262 503 510 2491412 10.1001/jama.262.4.503
Zelen J Bilfinger TV Anagnostopoulos CE Coronary artery bypass grafting. The relationship of surgical volume, hospital location, and outcome N Y State J Med 1991 91 290 292 1876313
Grumbach K Anderson GM Luft HS Roos LL Brook R Regionalization of cardiac surgery in Canada. Geographic access, choice, and outcomes JAMA 1995 274 1282 1288 7563533 10.1001/jama.274.16.1282
Hannan EL Siu AL Kumar D Kilburn H Chassin MR The decline in coronary artery bypass graft surgery mortality in New York State. The role of surgeon volume JAMA 1995 273 209 213 7807659 10.1001/jama.273.3.209
Clark RE Crawford FA Anderson RP Grover FL Kouchoukos NT Waldhaussen JA Outcome as a function of annual coronary artery bypass graft volume Ann Thorac Surg 1996 61 21 26 8561556 10.1016/0003-4975(95)00734-2
Tu JV Naylor CD Coronary artery bypass mortality rates in Ontario: a Canadian approach to quality assurance in cardiac surgery: Steering Committee of the Provincial Adult Cardiac Care Network of Ontario Circulation 1996 94 2429 2433 8921784
Shroyer ALW Marshall G Warner BA Johnson RR Guo W Grover FL Hammermeister KE No continuous relationship between Veterans Affairs hospital coronary artery bypass grafting surgical volume and operative mortality Ann Thorac Surg 1996 61 17 20 8561546 10.1016/0003-4975(95)00830-6
Sollano JA Gelijns AC Moskowitz AJ Heitjan DF Culliname S Saha T Chen JM Roohan PJ Reemtsma K Shields EP Volume-outcome relationships in cardiovascular operations: New York state, 1990–1995 J Thorac Cardiovasc Surg 1999 117 419 430 10047643
Nallamothu BK Saint S Ramsey SD Hofer TP Vijan S Eagle KA The role of hospital volume n coronary artery bypass grafting: Is more always better? J Am Coll Cardiol 2001 38 1923 1930 11738295 10.1016/S0735-1097(01)01647-3
Birkmeyer JD Siewers AE Finlayson VA Stukel TA Lucas FL Batista I Welch HG Wennberg DE Hospital volume and surgical mortality in the United States N Engl J Med 2002 346 1128 1137 11948273 10.1056/NEJMsa012337
Fergusen BT Peterson ED Coombs LP Eiken MC Carey ML Grover FL DeLong ER Use of continuous quality improvement to increase use of process measures in patients undergoing coronary artery bypass graft surgery. A randomized controlled trial JAMA 2003 290 49 56 12837711 10.1001/jama.290.1.49
Birkmeyer JD Stukel TA Siewers AE Goodney PP Wennberg DE Lucas FL Surgeon volume and operative mortality in the United States N Engl J Med 2003 349 2117 2127 14645640 10.1056/NEJMsa035205
Peterson ED Coombs LP De Long ER Haan CK Ferguson BT Procedural volume as a marker of quality for CABG surgery JAMA 2004 291 195 201 14722145 10.1001/jama.291.2.195
Wu C Hannan EL Ryan TJ Bennett E Culliford AT Gold JP Isom WO Jones PH McNeil B Rose EA Subramanian VA Is the impact of hospital and surgeon volumes on the in-hospital mortality rate for coronary artery bypass graft surgery limited to patients at high risk? Circulation 2004 110 784 789 15302792 10.1161/01.CIR.0000138744.13516.B5
Hannan EL Wu C Ryan TJ Bennett E Culliford AT Gold JP Hartman A Isom WO Jones PH McNeil B Rose EA Subramanian VA Do hospitals and surgeons with coronary artery bypass graft surgery volumes still have lower risk-adjusted mortality rates? Circulation 2003 108 795 801 12885743 10.1161/01.CIR.0000084551.52010.3B
Hannan EL Kilburn H Bernard H O'Donnell JF Lukacik G Shields EP Coronary artery bypass surgery: the relationship between in-hospital mortality rate and surgical volume after controlling for risk factors Med Care 1991 29 1094 1097 1943270
Birkmeyer JD Dimick JB Potential benefits of the new Leapfrog standards: effect of process and outcome measures Surgery 2004 135 569 575 15179361 10.1016/j.surg.2004.03.004
Ferguson BT Dziuban SW Edwards FH Eiken MC Shroyer AL Pairolero PC Anderson RP Grover FL The STS national database: Current changes and challenges for the new millennium Ann Thorac Surg 2000 69 680 691 10750744 10.1016/S0003-4975(99)01538-6
Edwards FH Clark RE Schwartz M Practical considerations in the management of large multi-institutional databases Ann Thorac Surg 1994 58 1841 1844 7979779
Edwards FH Carey JS Grover FL Bero JW Hartz RS Impact of gender on coronary bypass operative mortality Ann Thorac Surg 1998 66 125 131 9692451 10.1016/S0003-4975(98)00358-0
Birkmeyer JD Finlayson EVA Birkmeyer CM Volume standards for high-risk surgical procedures: Potential benefits of the Leapfrog program Surgery 2001 130 415 422 11562662 10.1067/msy.2001.117139
Daley J Invited commentary: Quality of care and the volume-outcome relationship – What's next for surgery? Surgery 2002 131 16 18 11812958 10.1067/msy.2002.120237
Epstein AM Volume and outcome – it is time to move ahead N Engl J Med 2002 346 1161 1164 11948278 10.1056/NEJM200204113461512
Shahian DM Normand ST The volume-outcome relationship: From Luft to Leapfrog Ann Thorac Surg 2003 25 1048 1058 12645752 10.1016/S0003-4975(02)04308-4
Khuri SF Invited commentary: surgeons, not General Motors, should set standards for surgical care Surgery 2001 130 429 431 11562665 10.1067/msy.2001.117138
Russel TR Invited commentary: volume standards for high-risk operations: an American College of Surgeons' view Surgery 2001 130 423 424 11562663 10.1067/msy.2001.117137
Dudley RA Johansen KL Invited commentary: physician responses to purchaser quality initiatives for surgical procedures Surgery 2001 130 425 428 11562664 10.1067/msy.2001.117136
Christian CK Gustafson ML Betensky RA Daley J Sinner MJ The Leapfrog volume criteria may fall short in identifying high-quality surgical centers Ann Surg 2003 238 447 455 14530717
Mokdad AH Ford ES Bowman BA Dietz WH Vinicor F Bales VS Marks JS Prevalence of obesity, diabetes, and obesity-related health risk factors, 2001 JAMA 2003 289 76 79 12503980 10.1001/jama.289.1.76
Whitaker RC Predicting preschooler obesity at birth: the role of materal obesity in early pregnancy Pediatrics 2004 114 e29 36 15231970 10.1542/peds.114.1.e29
Zacharias A Habib RH Schwann TA Riordan CJ Durham SJ Shah A Improved survival with radial artery versus vein conduits in coronary artery bypass surgery with left internal thoracic artery to left anterior descending artery grafting Circulation 2004 109 1489 1496 15023868 10.1161/01.CIR.0000121743.10146.78
Provenchère S Plantefève G Hufnagel G Vicaut E de Vaumas C Lecharny JB Depoix JP Vrtovsnik F Desmonts JM Philip I Renal dysfunction after cardiac surgery with normothermic cardiopulmonary bypass: Incidence, risk factors, and effect on clinical outcome Anesth Analg 2003 96 1258 1264 12707117 10.1213/01.ANE.0000055803.92191.69
Leitch JW Thomson D Baird DK Harris PJ The importance of age as a predictor of atrial fibrillation and flutter after coronary artery bypass grafting J Thorac Cardiovasc Surg 1990 100 338 342 2391970
Creswell LL Schuessler RB Rosenbloom M Cox JL Hazards of postoperative atrial arrhythmias Ann Thorac Surg 1993 56 539 549 8379728
Shahian DM Improving cardiac surgery quality – volume, outcome, process? JAMA 2004 291 246 249 14722153 10.1001/jama.291.2.246
Berwick DM Public performance reports and the will for change JAMA 2002 288 1523 1524 12243641 10.1001/jama.288.12.1523
Landon BE Normand SL Blumenthal D Daley J Physician clinical performance barriers: Prospects and barriers JAMA 2003 290 1183 1189 12953001 10.1001/jama.290.9.1183
| 15865623 | PMC1131908 | CC BY | 2021-01-04 16:28:04 | no | BMC Surg. 2005 May 2; 5:10 | utf-8 | BMC Surg | 2,005 | 10.1186/1471-2482-5-10 | oa_comm |
==== Front
BMC SurgBMC Surgery1471-2482BioMed Central London 1471-2482-5-91585750310.1186/1471-2482-5-9Technical AdvanceThe minimally invasive open video-assisted approach in surgical thyroid diseases Ruggieri Massimo [email protected] Andrea [email protected] Alessandra [email protected] Mariapia [email protected]'Armiento Massimino [email protected] Patrizia [email protected] Angela [email protected] Pierpaolo [email protected] Department of Surgical Sciences and Applied Medical Technologies "Francesco Durante", University of Rome "La Sapienza", Rome, Italy2 Department of Experimental Medicine and Pathology, Chair of Endocrinology, University of Rome "La Sapienza", Rome, Italy3 Department of Medicine, University of Rome "La Sapienza", Rome, Italy2005 27 4 2005 5 9 9 20 10 2004 27 4 2005 Copyright © 2005 Ruggieri et al; licensee BioMed Central Ltd.2005Ruggieri 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 targets of minimally invasive surgery (MIVA) could be summarised by: achievement of the same results as those obtained with traditional surgery, less trauma, better post-operative course, early discharge from hospital and improved cosmetic results. The minimally invasive techniques in thyroid surgery can be described as either endoscopic "pure" approach (completely closed approach with or without CO2 insufflation), or "open approach" with central neck mini-incision or "open video-assisted approach". Traditionally, open thyroidectomy requires a 6 to 8 cm, or bigger, transverse wound on the lower neck. The minimally invasive approach wound is much shorter (1.5 cm for small nodules, up to 2–3 cm for the largest ones, in respect of the exclusion criteria) upon the suprasternal notch.
Patients also experience much less pain after MIVA surgery than after conventional thyroidectomy. This is due to less dissection and destruction of tissues.
Pathologies treated are mainly nodular goiter; the only kind of thyroid cancer which may be approached with endoscopic surgery is a small differentiated carcinoma without lymph node involvement.
The patients were considered eligible for MIVA hemithyroidectomy and thyroidectomy on the basis of some criteria, such as gland volume and the kind of disease. In our experience we have chosen the minimally invasive open video-assisted approach of Miccoli et al. (2002). The aim of this work was to verify the suitability of the technique and the applicability in clinical practice.
Methods
A completely gasless procedure was carried out through a 15–30 mm central incision about 20 mm above the sternal notch. Dissection was mainly performed under endoscopic vision using conventional endoscopic instruments. The video aided group included 11 patients. All patients were women with a average age of 54.
Results
We performed thyroidectomy in 8 cases and hemithyroidectomy in 3 cases. The operative average time has been 170 minutes.
Conclusion
Nowadays this minimally invasive surgery, in selected patients, clearly demonstrates excellent results regarding patient cure rate and comfort, with shorter hospital stay, reduced postoperative pain and most attractive cosmetic results.
minimally invasive thyroidectomyMIVAvideo-assisted surgery
==== Body
Background
The targets of minimally invasive surgery (MIVA) could be summarised by: achievement of the same results as those obtained with traditional surgery, less trauma, better post-operative course, early discharge from hospital and improved cosmetic results.
After the first endoscopic parathyroidectomy, performed and described by Gagner in 1996 [4], several surgeons reported their experiences with minimally invasive and video-assisted surgery of the neck [1-30]. Shimizu successfully treated more than 120 patients using the anterior neck lifting method (VANS) using the abdominal wall-lift method [6,7].
The minimally invasive techniques in thyroid surgery can be described as either endoscopic "pure approach" (completely closed approach with or without CO2 insufflation) [9,10,12,21,22,27,30], or "open approach" with central neck mini-incision [8]or "open video-assisted approach" [1-3,5,11,13-20,23-26,28,29]. Traditionally, open thyroidectomy requires a 6 to 8 cm, or bigger, transverse wound on the lower neck. The minimally invasive approach wound is much shorter (1.5 cm for small nodules, maximum 2–3 cm for the largest, in respect of the exclusion criteria) upon the suprasternal notch.
In many papers, Miccoli et al. confirm the suitability of the minimally invasive video-assisted (MIVA) approach in performing hemithyroidectomy and thyroidectomy [12-18]. In their procedure a 15–20 mm transversal skin incision was made 2 cm above the sternal notch.
In our experience we have chosen the minimally invasive open video-assisted approach of Miccoli et al. (2002). The aim of this work was to verify the suitability of the technique and the applicability in clinical practice.
Methods
Patients
In our first year of this study 11 patients were selected for video-assisted surgery. The patients selected were 11 female with an average age of 54 (range 32 to 78). Preoperative evalutation (biochemical assessment, ultrasonography, and fine needle aspiration biopsy, in some cases) was obtained in all cases. Preoperative diagnosis was multinodular goiter in 8 cases, toxic adenoma in 1 case and papillary carcinoma in 2 cases. Approximate thyroid volume was 12.6 ml.
Inclusion criteria are: (table 1)
Table 1 Inclusion criteria
single nodule or small goiter (toxic or not) of surgical competence
cranio-caudal axis of the lobes must not exceed 7 centimetres
largest transversal diameter of the nodule must not exceed 3,5 centimetres
total thyroid volume <15–25 ml
small (max 2 cm) differentiated carcinoma without lymph node involvement
Exclusion criteria are: (table 2)
Table 2 Exclusion criteria
Absolute Relative
previous neck surgery;
big goiter;
local advanced cancer;
lymph node metastasis;
medullary or undifferentiated carcinoma Previous neck radiation therapy;
Basedow disease;
cronic thyroiditis.
Surgical instruments
The instruments necessary for this kind of surgery are in part the same in use for the traditional one; however, this technique used, in particular, proper tools characterized by small diameter (max 2 mm) that could be also used in endoscopic vision: atraumatic spatulas, spatula-shaped aspirator, forceps and scissors. Nevertheless, for minimally invasive thyroidectomy, the primary instruments are the 30-degree 5-mm endoscope and the 14 cm-long Harmonic Scalpel Scissors (Ethicon ENDO-SURGERY, Inc.).
Surgical procedures
The neck is quite hyperextended. The surgical team consists of the surgeon and two assistants, one of whom must hold the camera. A 25–30 mm skin incision is performed about 2 cm above the sternal notch, in the middle line. The cervical linea alba is then opened as much as possible, making sure to avoid any minimal bleeding. The thyroid lobe on the affected side is then carefully dissected from the muscles. Two small retractors are used to medially retract and lift the thyroid and to laterally retract the muscles to maintain the operative space. A 30-degree 5-mm endoscope is inserted through the skin incision (Fig. 1).
Figure 1 Minimally invasive video-assisted thyroidectomy. Two small retractors are used to maintain the operative space. The endoscope and the instruments are all inserted through the single single skin incision (an intraoperative view).
Dissection of the thyrotracheal groove is completed under endoscopic vision by using small instruments.
The area must be completely bloodless, because even minimal bleeding makes the operation more difficult or impossible. To achieve haemostasis, we use small (3 mm) clips or the 5 mm, 14 cm-long Harmonic Scalpel scissors. The first vessel to be cut is the middle vein, if present, or the small veins between the jugular vein and the thyroid capsule. The spatula is used to separate the larynx from the vessels and to retract them laterally. The external branch of the superior laryngeal nerve can be easily identified during most procedures, once the different components of the upper pedicle have been prepared. The upper pedicle is then exposed and selectively cut by Harmonic Scalpel Scissors (Fig. 2).
Figure 2 Minimally invasive video-assisted thyroidectomy. Upper pedicle sectioned by Harmonic Scalpel. Scissor (an endoscopic view).
The inferior vessels are also clipped and cut off, exposing the antero-lateral side of the trachea. The recurrent laryngeal nerve generally lies in the thyrotracheal groove, behind the Zuckerkandl tubercle. The recurrent nerve and the parathyroid glands are dissected and freed from the thyroid – these structures are well visualized by virtue of endoscope magnification [3]. Now the operation is conducted as in open surgery: the lobe is freed from the trachea, the isthmus dissected from the trachea and divided by harmonic scalpel. Finally the lobe is removed by conventional open technique. For total thyroidectomy, the same technique is repeated in the controlateral side. The muscles incision is sutured with reabsorbable suture and the wound is closed by intradermic adsorbable suture. We use drainage tubes (3.3 mm) that are introduced laterally.
Results
The video assisted group included a total of 11 patients.
We performed in 8 cases a total thyroidectomy and in 3 an hemithyroidectomy. Operative average time was 170 minutes. In one case conversion to the traditional approach has been necessary for difficulties in finding the recurrent laryngeal nerve. No complications have happened, except from 2 cases of transitory hoarseness at the beginning of our experience. We have established, in our survey, that the length of the wound must not exceed 3 cm. This is because, untill this value, dissection of subplatysmal plane is not necessary, so avoiding postoperative pain or anterior neck discomfort. However the limit of 3 cm is largely conditioned by the experience of the operator and the duration of the intervention as well.
We obtained excellent results about patient cure rate and comfort, with short hospital stay, few postoperative pain and attractive cosmetic results.
Discussion
Thyroid diseases primarily occurs from young to middle-age women who usually pay much attention to cosmetic results after thyroid surgery.
Postoperative pain and recovery, following MIVA surgery, are shorter than those with conventional thyroidectomy, because there are fewer dissection and destruction of tissues and the dividing platysma doesn't complitely perform. Another aspect is the smaller number of cases of neck paresthesia (in the wound region) in the days following the operation. Of utmost importance, the minimally invasive approach wound is very short (1–2 cm for small nodules, up to 2–3 cm for the biggest, in respect of the exclusion criteria) upon the suprasternal notch, and is easily covered by a shirt (Fig. 3).
Figure 3 Minimally invasive video-assisted thyroidectomy. Using video-assisted endoscopic technique, the neck scar is only 1,5-maximum 3 cm in length on suprasternal notch, easily covered by a shirt.
However, the development of a new surgical technique that minimizes the wound size is already possible, but the learning period is very long and the surgical tecnique is very hard. The operation time for minimally invasive video-assisted thyroidectomy is becoming comparable with that of conventional open technique [17,20,30].
It remains unclear whether the MIVA-thyroidectomy is suitable for the management of thyroid carcinoma or not. It is not yet recommended to use minimally invasive video-assisted thyroidectomy to manage thyroid malignancy until this surgical technique is mature enough to confidently dissect lymph nodes along the carotid sheath. The only kind of thyroid cancer that may be treated with endoscopic surgery is a small differentiated carcinoma without lymph node involvement [13].
Of course, minimally invasive video-assisted thyroidectomy has its limitations. MIVA-thyroidectomy is not suitable for repeated thyroid surgery because adhesions might interfere with the access of endoscope into the plane of the thyroid. Thyroid size is an important factor determining how difficult MIVA-approach would be, because the working space provided by the technique is limited. At this time, the maximum vertical axis of the lobes must not exceed 7 cm and the largest transversal diameter of the nodule must not exceed 3.5 cm. The total thyroid volume must not exceed 15–20 ml, echographically determined. It's not recommended to perform MIVA technique for goiter larger this size. Though in literature minimally invasive video-assisted approaches on thyroids well over 25 ml of volume and nodules greater than 3,5 cm are reported, at the moment it is not considered wise to operate lesions of such dimensions. The last limitation is the presence of a thyroidits diagnosed by biochemical or echographic signs [13,19]. As for the anterior neck lifting method (VANS) [6,7], we agree that it avoids the possibility of complications from carbon dioxide insufflation, but we would rather create the working space using two conventional retractors.
Complications of traditional thyroidectomy and MIVA are not different.
Transient recurrent nerve palsy and transient hypocalcemia are the more frequent, but the rate of these complications follows the learning curve. The operative time was about 80 minutes for a lobectomy and 130 minutes for a total thyroidectomy [12-18].
These operative times are slightly longer than those registered for conventional surgery in our Department, especially for the initial cases, because the development of a new technique always implies a learning period
Conversion to the traditional approach may be required in some cases for problems related to the bleeding from the vessels or the thyroid dimension.
A further issue of video-assisted techniques is a greater cost of this type of intervention, mainly due to the instruments that are required.
Conclusion
This study demonstrates that MIVA thyroidectomy is a possible and safe procedure, when selection criteria are strictly followed. It can be considered a valid option because of its cosmetic advantages, which are particularly appreciated by young patients.
Follicular nodules, instead, are optimal candidates for this approach just because in Italy they are generally small. Besides, in patients affected by primary hyperparathyroidism due to a single adenoma and also presenting a thyroid nodule, this access allows treating both diseases with a single operation.
In our experience, the video-assisted surgery represents a remarkable improvment in the techniques of the surgery of the neck. Besides, a very important cosmetic improvment, especially for the patients turned to this surgery (young female patients in particular), this technique even reduces the invasivity of the surgical manoeuvres with a precocious reestablishment of the preoperatory well-being and with precocious de-hospitalization of the patients. Moreover, this method can be used nearly like routine surgery to facilitate the surgeon in preparation of the upper vascular pedicle, reducing the entity of the skin opening and in preparation of the superior miocutaneal edge, reducing the frequent paresthetic consequences ensuing.
This kind of technique can also be used in case of parathyroid glands disease, with less difficulties than thyroid surgery. In order to precisely define and clarify the role of minimally invasive video-assisted thyroidectomy or parathyroidectomy in the management of patients with thyroid and parathyroidectomy disease, larger studies and longer follow-up are requested. At present this kind of surgery clearly demonstrates excellent results regarding patient cure rate and comfort, with shorter hospital stay, fewer postoperative pain and attractive cosmetic results. In the future a way to optimise the benefits, might be the combination of the minimally invasive video-assisted surgery with minimal-aggressive anaesthesia, such as locoregional anaesthesia jointly with intravenous sedation [19].
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Bellantone R Lombardi C Raffaeli M Alesina P De Crea C Traini E Salvatori M Video-assisted thyroidectomy for papillary thyroid carcinoma Surg Endosc 2003 17 1604 1608 12874681 10.1007/s00464-002-9220-0
Bellantone R Lombardi C Raffaeli M Boscherini M De Crea C Traini E Video-assisted thyroidectomy J Am Coll Surg 2002 194 610 614 12022601 10.1016/S1072-7515(02)01138-9
Berti P Materazzi G Conte M Galleri D Miccoli P Visualization of the external branch of the superior laryngeal nerve durino video-assisted thyroidectomy J Am Coll Surg 2002 195 573 574 12375766 10.1016/S1072-7515(02)01338-8
Gagner M Endoscopic subtotal parathyroidectomy in patients with primary hyperparathyroidism Br J Surg 1996 83 875 8696772
Ikeda Y Takami H Sasaki Y Takayama J Niimi M Kan S Comparative study of thyroidectomies. Endoscopic surgery vs conventional open surgery Surg Endosc 2002 16 1741 1745 12140635 10.1007/s00464-002-8830-x
Shimizu K Minimally invasive thyroid surgery Best Pract Res Clin Endocrind Metab 2001 15 123 137 10.1053/beem.2001.0130
Shimizu K Akira S Tanaka S Video assisted neck surgery: endoscopic resection of benign thyroid tumor aiming at scarless surgery on the neck J Surg Oncol 1998 69 178 180 9846506 10.1002/(SICI)1096-9098(199811)69:3<178::AID-JSO11>3.0.CO;2-9
Ikeda Y Takami H Tajima G Sasaki Y Takayama J Kurihara H Niimi M Direct mini-incision thyroidectomy Biomed Pharmacother 2002 56 Suppl 1 60s 63s 12487254 10.1016/S0753-3322(02)00257-3
Ikeda Y Takami H Tajima G Sasaki Y Takayama J Kurihara H Niimi M Total endoscopic thyroidectomy: axillary or anterior chest approach Biomed Pharmacother 2002 56 Suppl 1 72s 78s 12487257 10.1016/S0753-3322(02)00274-3
Kataoka H Kitano H Takeuchi E Fujimura M Total video endoscopic thyroidectomy via the anterior chest approach using the cervical region-lifting method Biomed Pharmacother 2002 56 Suppl 1 68s 71s 12487256 10.1016/S0753-3322(02)00227-5
Lombardi C Raffaeli M Modesti C Boscherini M Bellantone R Video-assisted thyroidectomy under local anesthesia Am J Surg 2004 187 515 518 15041502 10.1016/j.amjsurg.2003.12.030
Miccoli P Minimally invasive surgery for thyroid and parathyroid diseases Surg Endosc 2002 16 3 6 11961594 10.1007/s00464-001-8140-8
Miccoli P Bellantone R Mourad M Walz M Raffaeli M Berti P Minimally invasive video-assisted thyroidectomy: multiistitutional experience World J Surg 2002 26 972 975 12016476 10.1007/s00268-002-6627-7
Miccoli P Berti P Conte M Bendinelli C Marcocci C Minimally invasive surgery for thyroid small nodules: preliminary report J Endocrinol Invest 1999 22 849 851 10710272
Miccoli P Berti P Raffaeli M Conte M Materazzi G Galleri D Minimally invasive video-assisted thyroidectomy Am J Surg 2001 181 567 570 11513788 10.1016/S0002-9610(01)00625-0
Miccoli P Berti P Raffaeli M Materazzi G Conte M Galleri D Impact of Harmonic Scalpel on operative time durino video-assisted thyroidectomy Surg Endosc 2002 16 663 666 11972210 10.1007/s00464-001-9117-3
Miccoli P Berti P Raffaeli M Materazzi G Baldacci S Comparison between minimally invasive video-assisted thyroidectomy: a prospective randomised study Surgery 2001 130 1039 1043 11742335 10.1067/msy.2001.118264
Miccoli P Elisei R Materazzi G Capezzone M Galleri D Minimally invasive video-assisted thyroidectomy for papillary carcinoma: a prospective study of its completeness Surgery 2002 132 1070 1074 12490857 10.1067/msy.2002.128694
Mourad M Pugin F Elias B Coche E Squifflet JP Contributions of the video-assisted approach to thyroid and parathyroid surgery Acta Chir Belg 2002 102 323 327 12471764
Musella M Lombardi S Caiazzo P Milone F Di Palma R De Franciscis S Jovino R La chirurgia video-assistita della tiroide: note di tecnica e analisi dei risultati Ann Ital Chir 2003 1 3 5 12870275
Nakano S Kijima Y Owaki T Shirao K Baba M Aikou T Anterior chest wall approach for video-assisted thyroidectomy using a modified neck skin lifting method Biomed Pharmacother 2002 56 96 99 10.1016/S0753-3322(02)00233-0
Park Y Han W Bae W 100 cases of endoscopic thyroidectomy: breast approach Surg Laparosc Endosc Percutan Tech 2003 13 20 25 12598753 10.1097/00129689-200302000-00005
Ruggieri M Straniero A Pacini F Luongo B Del Grammastro A Mascaro A Paolini A Extent of lymph node dissection in thyroid cancer Policlinico J Surg 2002 109 65 71
Ruggieri M Straniero A Pacini F Luongo B Del Grammastro A Mascaro A Genderini M Gargiulo P Paolini A Role of surgery in tratment of hyperthyroidism Policlinico J Surg 2001 108 111 118
Ruggieri M Straniero A Pacini F Mariuolo A Mascaro A Genderini M Video-assisted surgery of the thyroid diseases Eur Rev Med Pharmacol Sci 2003 7 91 96 15068231
Ruggieri M Straniero A Pacini F Mariuolo A Mascaro A Genderini M Comparison between video-assisted thyroidectomy and conventional thyroidectomy. Preliminary aspects Policlinico J Surg 2003 110 11 17
Shiamazu K Shiba E Tamaki Y Takiguchi S Takiguchi E Ohashi S Endoscopic thyroid surgery through the axillo-bilateral-breast approach Surg Laparosc Endosc Percutan Tech 2003 13 196 201 12819505 10.1097/00129689-200306000-00011
Yamashita H Watanabe S Koike E Ohshima A Uchino S Video assisted thyroidectomy through a small wound in the submandibular area Am J Surg 2002 183 286 289 11943128 10.1016/S0002-9610(02)00801-2
Yeh T Jan Y Hsu B Chen K Chen M Video-assisted endoscopic thyroidectomy Am J Surg 2000 180 82 85 11044518 10.1016/S0002-9610(00)00429-3
Masahide Y Akira S Hiroshi A Yutaka S Nobuhiro S Jun N Rie M Kazuyoshi S Endoscopic subtotal thyroidectomy for patients with Graves' disease Surg Today 2001 31 1 4 11213035 10.1007/s005950170211
| 15857503 | PMC1131909 | CC BY | 2021-01-04 16:28:04 | no | BMC Surg. 2005 Apr 27; 5:9 | utf-8 | BMC Surg | 2,005 | 10.1186/1471-2482-5-9 | oa_comm |
==== Front
BMC UrolBMC Urology1471-2490BioMed Central London 1471-2490-5-71586213210.1186/1471-2490-5-7Research ArticleAPF, HB-EGF, and EGF biomarkers in patients with ulcerative vs. non-ulcerative interstitial cystitis Zhang Chen-Ou [email protected] Ze-Liang [email protected] Chui-Ze [email protected] Division of Infectious Diseases, Department of Medicine, University of Maryland School of Medicine, Baltimore, MD, 21201, USA2 Department of Urology, First Hospital of China Medical University, Shenyang City, Liaoning, 110001, China2005 29 4 2005 5 7 7 7 3 2005 29 4 2005 Copyright © 2005 Zhang et al; licensee BioMed Central Ltd.2005Zhang 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
Interstitial cystitis (IC) is a chronic bladder disorder, with symptoms including pelvic and or perineal pain, urinary frequency, and urgency. The etiology of IC is unknown, but sensitive and specific biomarkers have been described, including antiproliferative factor (APF), heparin-binding epidermal growth factor-like growth factor (HB-EGF), and epidermal growth factor (EGF). However, the relative sensitivity of these biomarkers in ulcerative vs. nonulcerative IC is unknown, and these markers have yet to be validated in another laboratory. We therefore measured these markers in urine from patients with or without Hunner's ulcer, as well as normal controls, patients with bladder cancer, and patients with bacterial cystitis, at the First Hospital of China Medical University.
Methods
Urine specimens were collected from two groups of Chinese IC patients (38 IC patients with Hunner's ulcers, 26 IC patients without Hunner's ulcers), 30 normal controls, 10 bacterial cystitis patients and 10 bladder cancer patients. APF activity was determined by measuring 3H-thymidine incorporation in vitro, and HB-EGF and EGF levels were determined by ELISA.
Results
APF activity (inhibition of thymidine incorporation) was significantly greater in all IC patient urine specimens than in normal control specimens or in specimens from patients with bacterial cystitis or bladder cancer (p < 0.0001 for each comparison). Urine HB-EGF levels were also significantly lower and EGF levels significantly higher in both groups of IC patients than in the three control groups (p < 0.0001 for each comparison). Although APF and HB-EGF levels were similar in ulcerative and nonulcerative IC patients, EGF levels were significantly higher in IC patients with vs. without ulcers (p < 0.004).
Conclusion
These findings indicate that APF, HB-EGF and EGF are good biomarkers for both ulcerative and nonulcerative IC and validate their measurement as biomarkers for IC in Chinese patients.
==== Body
Background
Interstitial cystitis (IC) is a chronic bladder disorder, with symptoms including pelvic and or perineal pain, urinary frequency, and urgency [1]. The cause of IC is unknown, and IC is therefore generally diagnosed by the presence of certain clinical features in the absence of other identifiable causes for the symptoms (such as urinary tract infection) [2]. Patients may undergo a history, physical examination, urinalysis, urine culture, urodynamics, and cystoscopy under anesthesia with bladder distention during the work-up for IC. All the results of these tests are combined with the clinical judgment of the practitioner to make the diagnosis.
Three biomarkers have been described for IC in various patient groups including Chinese IC patients [3-5]. The first one is antiproliferative factor (APF), a low molecular weight frizzled-8 related peptide [6] that inhibits the growth of bladder epithelial cells. APF activity is found in the urine of over 90% of patients clinically diagnosed with IC [4,5]. The second biomarker is heparin-binding epidermal growth factor-like growth factor (HB-EGF), which is important for epithelial cell proliferation and wound healing. This factor is significantly decreased in urine from IC patients when compared with specimens from normal controls or patients with other urological conditions [4,5]. A third biomarker is epidermal growth factor (EGF), which is significantly increased in urine from IC patients when compared with the same control groups [4,5].
In this article we describe studies to determine the relative sensitivity of these markers in patients with ulcerative vs. nonulcerative IC. We also sought to confirm their utility in Chinese patients by measuring these factors in our laboratory using specimens from a group of Chinese IC patients who were not previously studied, as well as normal controls, and patients with bacterial cystitis or bladder cancer.
Methods
Patients
Two groups of Chinese IC patients who had previously undergone diagnostic cystoscopy and fulfilled the NIDDK diagnostic criteria for IC were studied; the first group included 38 IC patients had Hunner's ulcers, and the second group included 26 IC patients who did not have Hunner's ulcers [2]. Normal controls were hospital personnel with no known urinary tract disease, who were matched as a group to the IC patients by gender and age. Urine was also obtained from bacterial cystitis and bladder cancer patients for comparison. IC patients were asked questions about symptoms and related problems from the O'Leary-Sant questionnaire [7], and total symptom index and problem index scores determined. All urine samples were collected at the First Hospital of China Medical University in the Shenyang City of the Peoples Republic of China. All participants were at least 18 years old and were enrolled in accordance with guidelines of the Institutional Review Board of China Medical University.
Urine specimens
Clean-catch urine specimens were obtained in which each IC patient or control wiped the labial area with 10% povidone iodine solution and then collected a midstream urine into a sterile container. Specimens were initially kept at 4°C, then transported to the laboratory where cellular debris was removed by low speed centrifugation at 4°C, aliquoted under sterile conditions, and stored at -80°C until used.
Cell culture
Adult human bladder epithelial (HBE) cells were grown from cadaveric bladder tissue of a young (30 year old) female accident victim who had no history of bladder disorder. These cells were grown in DMEM-F12 medium containing 10% fetal bovine serum (FBS), 1% antibiotic/antimycotic solution, 1% glutamine, and 1.0 u/ml insulin (all from Sigma) at 37°C in a 5% CO2 atmosphere.
3H-thymidine incorporation
3H-thymidine incorporation was assayed as previously described by Keay, et al [4]. HBE cells were plated at a density of 1 × 104 cell per well onto 96 well tissue culture plates and incubated at 37°C overnight. The medium was then changed to MEM containing only 1% glutamine and 1% antibiotic/antimycotic solution, and the cells were incubated at 37°C overnight. On the third day, urine specimens from IC patients or controls were corrected to pH 7.2 and 300 mOSM, filtered through a 0.2 uM pore filter (Gelman Science, Ann Arbor, MI), diluted 1:2 in MEM (Serum-free MEM containing only glutamine and antibiotics/antimycotics) and applied to the cells; cell controls received serum free MEM medium only. After 48 hours of incubation at 37°C the cells were labeled with 1 uCi per well 3H-thymidine (NEN DuPont, Wilmington, DE) and incubated for another 4 hours at 37°C. Cells were then trypsinized and insoluble cell contents harvested and methanol-fixed onto glass filber filter paper, as previously described. The amount of radioactivity incorporated was determined as counts per minute using a Tri-Carb 2900 TR Scintillation counter (Packard Bioscience). A significant inhibition of 3H-thymidine incorporation was defined as a mean decrease in counts per minute of greater than 2 standard deviations from the mean of control cells for each plate.
Enzyme-linked immunosorbent assays
HB-EGF
Each well of a 96 well Immulon II plate (Dynatech Laboratories, Chantilly, VA) was coated with 200 ul urine at 4°C overnight as previously described by Keay, et al [4]. The next day the plate was washed 5 times with 1× PBS buffer, the plates were blocked with 5% FBS/1 mM EDTA/0.05% Tween 20 in 1× PBS buffer. Anti-HB-EGF antibody (1 ug/ml; R & D systems, Minneapolis, MN) was added and the plates were incubated for 2 hours at 37°C. After an additional 5 washes, biotinylated anti-goat IgG/avidin D horseradish peroxidase (HRP) was added and plates were incubated for 1.5 hours at room temperature, washed, and developed with ABTS (2.2'-Azino-bis-(3-ethylbenzothiazoline-6-sulfonic)) substrate; absorbance was read at 405 nm.
EGF
Urine from IC patients and controls was diluted 1:200 in RD5E diluent and pipetted into wells precoated with monoclonal anti-EGF antibody, according to the manufacturer's instructions (R&D Systems). After incubation at room temperature for 2 hours, plates were rinsed with wash buffer and incubated further with HRP-linked polyclonal anti-goat antibody, rinsed again, and developed using tetramethylbenzidine (TMB) substrate. Development was stopped with 2 N sulfuric acid, and absorbance read at 450 nm.
Linear absorbance versus concentration curves were prepared from results with known standard concentrations of EGF or HB-EGF (R & D Systems), and sample concentrations were determined by plotting absorbance values.
Statistical analysis
The comparisons of mean change in 3H-thymidine incorporation and growth factor levels in urine specimens from patients with IC vs. controls were made using a two-tailed student t test.
Results
Antiproliferative factor activity in urine specimens
To compare the effect of urine specimens from IC patients and controls on bladder epithelial cell proliferation, we measured 3H-thymidine incorporation in normal human bladder epithelial cells. Specimens were collected from two groups of Chinese IC patients, 38 IC patients with Hunner's ulcers, 26 IC patients without Hunner's ulcers, 30 normal controls, 10 bacterial cystitis patients and 10 bladder cancer patients, by the clean catch method. As shown in the table, IC patients and normal controls did not differ significantly in age or gender; most of the IC patients and normal controls were women, with five men in both IC groups, and two men in the normal control group.
HBE cells exposed to urine from the 38 Chinese IC patients with Hunner's ulcers had significantly less 3H-thymidine incorporation than cells incubated with urine from normal controls (-82.1 ± 2.2% vs. 1.6 ± 3.5%, p < 0.000001) (Figure 1), and cells exposed to urine from the 26 Chinese IC patients without Hunner's ulcers also had significantly less 3H-thymidine incorporation than cells incubated with urine from normal controls (-78.5 ± 2.8% vs. 1.6 ± 3.5%, p < 0.000001). The two groups of IC patients did not differ significantly from each other (p = 0.13). In comparison, HBE cells exposed to urine from bacterial cystitis or bladder cancer patients did not differ significantly from cells incubated with urine from normal controls (-8.4 ± 6.5%, and -8.5 ± 5.3% vs. 1.6 ± 3.5%), but each differed significantly from cells cultured with IC patients urine (p < 0.0001). When inhibition of thymidine incorporation greater than 2 standard deviations from the mean of cell controls was used as the definition for APF activity, 59/64 (92%) of IC patients had evidence for APF activity vs. only 1/30 (3%) of normal controls. None of the samples from bacterial cystitis or bladder cancer patients had any evidence for APF activity.
Figure 1 APF activity in urine from Chinese interstitial cystitis patients with Hunner's ulcers (IC-U), Chinese interstitial cystitis patients without Hunner's ulcers (IC-N), normal controls (Ctr), and patients with bacterial cystitis (BC), or bladder cancer (BCa). APF activity was measured as inhibition of 3H-thymidine incorporation in normal bladder epithelial cells. Each data point indicates the mean change in incorporation, with each specimen tested in triplicate. Horizontal line indicates value of mean; vertical line indicates standard error of the mean for each group.
Levels of HB-EGF and EGF in urine from IC patients and normal controls
We next measured HB-EGF and EGF levels in urine specimens from the same Chinese IC patients, normal controls, bacterial cystitis patients and bladder cancer patients by ELISA. As shown in Figure 2, the concentration of urine HB-EGF was significantly lower in the 38 IC patients with Hunner's ulcers (1.19 ± 0.20 ng/ml) and in the 26 IC patients without Hunner's ulcers (1.42 ± 0.23 ng/ml) as compared to normal controls (9.28 ± 1.04 ng/ml) or patients with bacterial cystitis or bladder cancer (5.34 ± 1.19 ng/ml, and 5.72 ± 0.87 ng/ml, p < 0.0001 for comparison of each IC group to each control group). Mean HB-EGF levels in the two groups of IC patients did not differ significantly from each other (p = 0.43). The mean concentration of urine EGF (Figure 3), however, was significantly higher in 38 IC patients with Hunner's ulcers (21.90 ± 1.19 ng/ml) than in 26 IC patients without Hunner's ulcers (16.32 ± 1.44 ng/ml) (p <0.004), and was markedly higher for both IC groups as compared to normal controls (6.49 ± 0.57 ng/ml) or patients with bacterial cystitis or bladder cancer (6.32 ± 1.26 ng/ml, and 8.03 ± 1.95 ng/ml, p < 0.0001 for comparison of each IC group to each control group).
Figure 2 HB-EGF levels in urine from IC patients, normal controls, bacterial cystitis patients and bladder cancer patients. HB-EGF levels were measured by ELISA in urine specimens from interstitial cystitis patients with Hunner's ulcers (IC-U), interstitial cystitis patients without Hunner's ulcers (IC-N), asymptomatic controls (Ctr), and patients with bacterial cystitis (BC), or bladder cancer (BCa). Each data point is the mean value for duplicate specimens. Horizontal line indicates value of mean; vertical line indicates standard error of mean for each group.
Figure 3 EGF levels in urine from IC patients, normal controls, bacterial cystitis patients and bladder cancer patients. EGF levels were measured by ELISA in urine specimens from interstitial cystitis patients with Hunner's ulcers (IC-U), interstitial cystitis patients without Hunner's ulcers (IC-N), asymptomatic controls (Ctr), and patients with bacterial cystitis (BC), or bladder cancer (BCa). Each data point is the mean value for duplicate specimens. Horizontal line indicates value of mean; vertical line indicates standard error of mean for each group.
Discussion
This report presents evidence that three urine biomarkers for IC (APF, HB-EGF and EGF) were confirmed by measurement in our laboratory at First Hospital of China Medical University using bladder epithelial cells from Chinese normal controls and urine specimens from a new group of Chinese IC patients and controls that had not previously been studied. Our data were comparable to those published previously [3-5], with 92% of IC patients having APF activity as compared to only 3% of controls.
APF is a low molecular weight peptide made by bladder epithelial cells from IC patients that inhibits the proliferation of human bladder epithelial cells [6,8], suggesting that it may cause the bladder epithelial thinning or ulceration common in IC [9]. Bladder cell proliferation is also known to be influenced by growth factors and their regulatory proteins. HB-EGF has been shown to be important for replication of a variety of epithelial cells including hepatocytes, keratinocytes, gastric epithelial cells, and uterine epithelial cells and is known to stimulate bladder epithelial replication in vitro [10]. It is therefore possible that decreased synthesis of HB-EGF by epithelial or other bladder cells contributes to the pathogenesis of IC by impairing normal bladder epithelial regeneration. APF can inhibit HB-EGF production by bladder epithelial cells [8], indicating a possible mechanism for APF's antiproliferative activity.
The reproducibility of APF, HB-EGF and EGF as biomarkers for IC in our patients indicates that one or more of these factors may be useful for the diagnosis of IC in patients from various racial backgrounds, and measured in different laboratories. As previously reported by Keay, et al [11], these biomarkers can clearly distinguish between IC and control groups whether they are normalized to urine creatinine or urine volume [11]. We therefore normalized the levels to urine volume for our study. We also compared APF, HB-EGF and EGF levels in IC patients with Hunner's ulcers vs. IC patients without Hunner's ulcers, and determined that mean values for all three markers were more abnormal in patients with ulcers, although the difference between IC patients with Hunner's ulcers and IC patients without Hunner's ulcers was only statistically significant for EGF in this study. Whether this finding indicates that ulcerative IC is a more severe form of the disorder than nonulcerative IC requires further investigation.
Several biomarkers have been described for IC as recently reviewed [12]. GP51, a glycoprotein urinary marker, reportedly is also specific for IC [13], but additional studies have not been done on this marker. Nitric oxide synthase (NOS) stimulates the production of nitric oxide (NO) which then increases cyclic GMP levels by activating guanylyl cyclase. In some studies, female IC patients have been shown to have significantly decreased NOS activity in their urine cell pellet than female controls, and urinary cyclic GMP levels were significantly lower in female IC patients than in female controls or females with urinary tract infections [5,14]. However, bladder luminal NO levels have also been shown to be markedly increased in patients with IC, and elevated bladder luminal NO levels are not specific for IC [15,16] but also occur in patients with various forms of cystitis. CD45RO positive lymphocytes are another biomarker for ulcerative IC patients. They are not found in the urine of healthy subjects, but are also found in bladder cancer patients treated with BCG [17,18]. Although they also are not specific for IC, these lymphocytes may be biomarkers that reflect the severity of interstitial bladder inflammation.
Urine APF activity and levels of HB-EGF, EGF, IL-6 and IGFBP3 were shown to be significantly different between IC patients and normal controls in one large comparison study [5]. However, APF, HB-EGF, and EGF were the most sensitive and specific for IC, with anti-proliferative factor activity most clearly separating the interstitial cystitis and control groups (5).
Conclusion
APF, HB-EGF and EGF have now been confirmed as good biomarkers for IC by another laboratory. Furthermore, we provide the first evidence that these IC biomarkers are present in both the urine of patients with ulcerative as well as nonulcerative IC, and the first comparison of these markers between these two groups of IC patients. Although EGF levels were significantly more abnormal in ulcerative than nonulcerative patients, APF and HB-EGF were not significantly different between the two groups, and the clinical significance of the differences in EGF levels is therefore uncertain at this time.
List of abbreviations used
IC – interstitial cystitis
APF – antiproliferative factor
HB-EGF – heparin-binding epidermal growth factor-like growth factor
EGF – epidermal growth factor
PBS – phosphate buffered saline
FBS – fetal bovine serum
ELISA – enzyme-linked immunosorbent assay
HBE – human bladder epithelial
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
CZ provided input into the design of these studies, assistance with development of the assays used for these experiments in the laboratory at China Medical University, data analysis, and writing of the manuscript. ZL provided assistance with patient diagnoses and collection of specimens and clinical data, and supervised all aspects of the work performed for this paper in his laboratory. CK provided assistance with patient diagnoses, as well as collection of specimens and clinical data, and assisted with data analysis.
Table 1 Patient Characteristics
IC-U* IC-N+ Control BC‡ BCa†
Age 41 ± 11.1 46 ± 10.6 36.6 ± 8.8 36.4 ± 9.6 55.4 ± 7.1
Gender (F/M) 36/2 23/3 28/2 8/2 3/7
Symptom Index Score points 19.6 ± 0.12 18.9 ± 0.29 - - -
Problem Index Score points 15.7 ± 0.15 15.5 ± 0.21 - - -
Average voiding volume (ml) 95 ± 15 98 ± 18 - - -
Average bladder capacity (ml)# 205 ± 16 216+17 - - -
* IC-U = IC patients with Hunner's ulcers
+ IC-N = IC patients without Hunner's ulcers
‡ BC = Bacterial Cystitis
† BCa = Bladder Cancer
# Bladder capacity determined while patients were awake, prior to distension
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This research was supported by China Health Ministry, Wu Jie-ping Funding.
==== Refs
Hanno PM Staskin DR Krane RJ Wein AJ eds Interstitial cystitis 1990 Springer-Verlag, London
Division of Kidney, Urologic, and Hematologic Diseases (DKUHD) of the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) Diagnostic criteria for research studies (interstitial cystitis) Am J Kidney Dis 1989 13 353 354
Zhang C-O Li Z-L Shoenfelt JL Kong C-Z Chai TC Erickson DE Peters KM Rovner ES Keay S Comparison of APF activity and epithelial growth factor levels in urine from Chinese, African American and Caucasian American IC patients Urology 2003 61 897 901 12735999 10.1016/S0090-4295(02)02597-9
Keay S Zhang C-O Shoenfelt JL Erickson DR Whitmore K Warren JW Marvel R Chai T Sensitivity and specificity of antiproliferative factor, heparin-binding epidermal growth factor-like growth factor, and epidermal growth factor as urine markers for interstitial cystitis Urology 2001 57 9 14 11378043 10.1016/S0090-4295(01)01127-X
Erickson DR Xie SX Bhavanandan VP Wheeler MA Hurst RE Demers LM Kushner L Keay SK A comparison of multiple urine markers for interstitial cystitis J Urol 2002 167 2461 2469 11992058 10.1097/00005392-200206000-00026
Keay S Szekely Z Conrads TP Veenstra TD Barchi JJ JrZhang C-O Koch KR Michejda CJ An antiproliferative factor from interstitial cystitis patients is a frizzled 8 protein-related sialoglycopeptide PNAS 2004 101 11803 11808 15282374 10.1073/pnas.0404509101
O'Leary MP Sant GR Fowler FJ Whitmore KE Spolarich-Kroll J The interstitial cystitis symptom index and problem index Urology 1997 49 58 63 9146003 10.1016/S0090-4295(99)80333-1
Keay S Kleinberg M Zhang C-O Hise MK Warren JW Bladder epithelial cells from interstitial cystitis patients produce an inhibitor of HB-EGF production J Urol 2000 64 2112 2118 11061938 10.1097/00005392-200012000-00074
Tomaszewski JE Landis JR Russack V Williams TM Wang LP Hardy C Brensinger C Matthews YL Abele ST Kusek JW Nyberg LM Biopsy features are associated with primary symptoms in interstitial cystitis: results from the interstitial cystitis database study Urology 2001 57 67 81 11378053 10.1016/S0090-4295(01)01166-9
Freeman MR Yoo JJ Raab G Soker S Adam RM Schneck FX Renshaw AA Klagsbrun M Atala A Heparin-binding EGF-like growth factor is an autocrine growth factor for human urothelial cells and is synthesized by epithelial and smooth muscle cells in the human bladder J Clin Invest 1997 99 1028 1036 9062361
Keay S Zhang C-O Kagen DI Hise MK Jacobs SC Hebel JR Gordon D Whitmore K Bodison S Warren JW Concentrations of specific epithelial growth factor in the urine of interstitial cystitis patients and controls J Urol 1997 158 1983 1988 9334654
Keay S Takeda M Tamaki M Hanno P Current and future directions in diagnostic markers in interstitial cystitis International J Urol 2003 10 S27 S30 10.1046/j.1442-2042.10.s1.8.x
Byrne DS Sedor JF Estojak J Fitzpatrick KJ Chiura AN Mulholland SG The urinary glycoprotein GP51 as a clinical marker for interstitial cystitis J Urol 1999 161 1786 1790 10332435
Smith SD Wheeler MA Foster HE JrWeiss RM Urinary nitric oxide synthase activity and cyclic GMP levels are decreased with interstitial cystitis and increased with urinary tract infections J Urol 1996 155 1432 1435 8632605 10.1097/00005392-199604000-00098
Ehren I Hosseini A Lundberg JO Wiklund NP Nitric oxide: A useful gas in the detection of lower urinary tract inflammation J Urol 1999 162 327 329 10411031 10.1097/00005392-199908000-00011
Ehren I Hosseini A Herulf M Lundberg JO Wiklund NP Measurement of luminal nitric oxide in bladder inflammation using a silicon balloon catheter: A novel minimally invasive method Urology 1999 54 264 267 10443722 10.1016/S0090-4295(99)00120-X
Ueda T Tamaki M Ogawa O Yamauchi T Yoshimura N Improvement of interstitial cystitis and problems that developed during treatment with oral IPD-1151T J Urol 2000 164 1917 1920 11061880 10.1097/00005392-200012000-00011
Chang SG Lee SJ Huh JS Lee JH Changes in mucosal immune cells of bladder tumor patient after BCG intravesical immunotherapy Oncol Rep 2001 8 257 261 11182036
| 15862132 | PMC1131910 | CC BY | 2021-01-04 16:30:02 | no | BMC Urol. 2005 Apr 29; 5:7 | utf-8 | BMC Urol | 2,005 | 10.1186/1471-2490-5-7 | oa_comm |
==== Front
BMC UrolBMC Urology1471-2490BioMed Central London 1471-2490-5-71586213210.1186/1471-2490-5-7Research ArticleAPF, HB-EGF, and EGF biomarkers in patients with ulcerative vs. non-ulcerative interstitial cystitis Zhang Chen-Ou [email protected] Ze-Liang [email protected] Chui-Ze [email protected] Division of Infectious Diseases, Department of Medicine, University of Maryland School of Medicine, Baltimore, MD, 21201, USA2 Department of Urology, First Hospital of China Medical University, Shenyang City, Liaoning, 110001, China2005 29 4 2005 5 7 7 7 3 2005 29 4 2005 Copyright © 2005 Zhang et al; licensee BioMed Central Ltd.2005Zhang 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
Interstitial cystitis (IC) is a chronic bladder disorder, with symptoms including pelvic and or perineal pain, urinary frequency, and urgency. The etiology of IC is unknown, but sensitive and specific biomarkers have been described, including antiproliferative factor (APF), heparin-binding epidermal growth factor-like growth factor (HB-EGF), and epidermal growth factor (EGF). However, the relative sensitivity of these biomarkers in ulcerative vs. nonulcerative IC is unknown, and these markers have yet to be validated in another laboratory. We therefore measured these markers in urine from patients with or without Hunner's ulcer, as well as normal controls, patients with bladder cancer, and patients with bacterial cystitis, at the First Hospital of China Medical University.
Methods
Urine specimens were collected from two groups of Chinese IC patients (38 IC patients with Hunner's ulcers, 26 IC patients without Hunner's ulcers), 30 normal controls, 10 bacterial cystitis patients and 10 bladder cancer patients. APF activity was determined by measuring 3H-thymidine incorporation in vitro, and HB-EGF and EGF levels were determined by ELISA.
Results
APF activity (inhibition of thymidine incorporation) was significantly greater in all IC patient urine specimens than in normal control specimens or in specimens from patients with bacterial cystitis or bladder cancer (p < 0.0001 for each comparison). Urine HB-EGF levels were also significantly lower and EGF levels significantly higher in both groups of IC patients than in the three control groups (p < 0.0001 for each comparison). Although APF and HB-EGF levels were similar in ulcerative and nonulcerative IC patients, EGF levels were significantly higher in IC patients with vs. without ulcers (p < 0.004).
Conclusion
These findings indicate that APF, HB-EGF and EGF are good biomarkers for both ulcerative and nonulcerative IC and validate their measurement as biomarkers for IC in Chinese patients.
==== Body
Background
Interstitial cystitis (IC) is a chronic bladder disorder, with symptoms including pelvic and or perineal pain, urinary frequency, and urgency [1]. The cause of IC is unknown, and IC is therefore generally diagnosed by the presence of certain clinical features in the absence of other identifiable causes for the symptoms (such as urinary tract infection) [2]. Patients may undergo a history, physical examination, urinalysis, urine culture, urodynamics, and cystoscopy under anesthesia with bladder distention during the work-up for IC. All the results of these tests are combined with the clinical judgment of the practitioner to make the diagnosis.
Three biomarkers have been described for IC in various patient groups including Chinese IC patients [3-5]. The first one is antiproliferative factor (APF), a low molecular weight frizzled-8 related peptide [6] that inhibits the growth of bladder epithelial cells. APF activity is found in the urine of over 90% of patients clinically diagnosed with IC [4,5]. The second biomarker is heparin-binding epidermal growth factor-like growth factor (HB-EGF), which is important for epithelial cell proliferation and wound healing. This factor is significantly decreased in urine from IC patients when compared with specimens from normal controls or patients with other urological conditions [4,5]. A third biomarker is epidermal growth factor (EGF), which is significantly increased in urine from IC patients when compared with the same control groups [4,5].
In this article we describe studies to determine the relative sensitivity of these markers in patients with ulcerative vs. nonulcerative IC. We also sought to confirm their utility in Chinese patients by measuring these factors in our laboratory using specimens from a group of Chinese IC patients who were not previously studied, as well as normal controls, and patients with bacterial cystitis or bladder cancer.
Methods
Patients
Two groups of Chinese IC patients who had previously undergone diagnostic cystoscopy and fulfilled the NIDDK diagnostic criteria for IC were studied; the first group included 38 IC patients had Hunner's ulcers, and the second group included 26 IC patients who did not have Hunner's ulcers [2]. Normal controls were hospital personnel with no known urinary tract disease, who were matched as a group to the IC patients by gender and age. Urine was also obtained from bacterial cystitis and bladder cancer patients for comparison. IC patients were asked questions about symptoms and related problems from the O'Leary-Sant questionnaire [7], and total symptom index and problem index scores determined. All urine samples were collected at the First Hospital of China Medical University in the Shenyang City of the Peoples Republic of China. All participants were at least 18 years old and were enrolled in accordance with guidelines of the Institutional Review Board of China Medical University.
Urine specimens
Clean-catch urine specimens were obtained in which each IC patient or control wiped the labial area with 10% povidone iodine solution and then collected a midstream urine into a sterile container. Specimens were initially kept at 4°C, then transported to the laboratory where cellular debris was removed by low speed centrifugation at 4°C, aliquoted under sterile conditions, and stored at -80°C until used.
Cell culture
Adult human bladder epithelial (HBE) cells were grown from cadaveric bladder tissue of a young (30 year old) female accident victim who had no history of bladder disorder. These cells were grown in DMEM-F12 medium containing 10% fetal bovine serum (FBS), 1% antibiotic/antimycotic solution, 1% glutamine, and 1.0 u/ml insulin (all from Sigma) at 37°C in a 5% CO2 atmosphere.
3H-thymidine incorporation
3H-thymidine incorporation was assayed as previously described by Keay, et al [4]. HBE cells were plated at a density of 1 × 104 cell per well onto 96 well tissue culture plates and incubated at 37°C overnight. The medium was then changed to MEM containing only 1% glutamine and 1% antibiotic/antimycotic solution, and the cells were incubated at 37°C overnight. On the third day, urine specimens from IC patients or controls were corrected to pH 7.2 and 300 mOSM, filtered through a 0.2 uM pore filter (Gelman Science, Ann Arbor, MI), diluted 1:2 in MEM (Serum-free MEM containing only glutamine and antibiotics/antimycotics) and applied to the cells; cell controls received serum free MEM medium only. After 48 hours of incubation at 37°C the cells were labeled with 1 uCi per well 3H-thymidine (NEN DuPont, Wilmington, DE) and incubated for another 4 hours at 37°C. Cells were then trypsinized and insoluble cell contents harvested and methanol-fixed onto glass filber filter paper, as previously described. The amount of radioactivity incorporated was determined as counts per minute using a Tri-Carb 2900 TR Scintillation counter (Packard Bioscience). A significant inhibition of 3H-thymidine incorporation was defined as a mean decrease in counts per minute of greater than 2 standard deviations from the mean of control cells for each plate.
Enzyme-linked immunosorbent assays
HB-EGF
Each well of a 96 well Immulon II plate (Dynatech Laboratories, Chantilly, VA) was coated with 200 ul urine at 4°C overnight as previously described by Keay, et al [4]. The next day the plate was washed 5 times with 1× PBS buffer, the plates were blocked with 5% FBS/1 mM EDTA/0.05% Tween 20 in 1× PBS buffer. Anti-HB-EGF antibody (1 ug/ml; R & D systems, Minneapolis, MN) was added and the plates were incubated for 2 hours at 37°C. After an additional 5 washes, biotinylated anti-goat IgG/avidin D horseradish peroxidase (HRP) was added and plates were incubated for 1.5 hours at room temperature, washed, and developed with ABTS (2.2'-Azino-bis-(3-ethylbenzothiazoline-6-sulfonic)) substrate; absorbance was read at 405 nm.
EGF
Urine from IC patients and controls was diluted 1:200 in RD5E diluent and pipetted into wells precoated with monoclonal anti-EGF antibody, according to the manufacturer's instructions (R&D Systems). After incubation at room temperature for 2 hours, plates were rinsed with wash buffer and incubated further with HRP-linked polyclonal anti-goat antibody, rinsed again, and developed using tetramethylbenzidine (TMB) substrate. Development was stopped with 2 N sulfuric acid, and absorbance read at 450 nm.
Linear absorbance versus concentration curves were prepared from results with known standard concentrations of EGF or HB-EGF (R & D Systems), and sample concentrations were determined by plotting absorbance values.
Statistical analysis
The comparisons of mean change in 3H-thymidine incorporation and growth factor levels in urine specimens from patients with IC vs. controls were made using a two-tailed student t test.
Results
Antiproliferative factor activity in urine specimens
To compare the effect of urine specimens from IC patients and controls on bladder epithelial cell proliferation, we measured 3H-thymidine incorporation in normal human bladder epithelial cells. Specimens were collected from two groups of Chinese IC patients, 38 IC patients with Hunner's ulcers, 26 IC patients without Hunner's ulcers, 30 normal controls, 10 bacterial cystitis patients and 10 bladder cancer patients, by the clean catch method. As shown in the table, IC patients and normal controls did not differ significantly in age or gender; most of the IC patients and normal controls were women, with five men in both IC groups, and two men in the normal control group.
HBE cells exposed to urine from the 38 Chinese IC patients with Hunner's ulcers had significantly less 3H-thymidine incorporation than cells incubated with urine from normal controls (-82.1 ± 2.2% vs. 1.6 ± 3.5%, p < 0.000001) (Figure 1), and cells exposed to urine from the 26 Chinese IC patients without Hunner's ulcers also had significantly less 3H-thymidine incorporation than cells incubated with urine from normal controls (-78.5 ± 2.8% vs. 1.6 ± 3.5%, p < 0.000001). The two groups of IC patients did not differ significantly from each other (p = 0.13). In comparison, HBE cells exposed to urine from bacterial cystitis or bladder cancer patients did not differ significantly from cells incubated with urine from normal controls (-8.4 ± 6.5%, and -8.5 ± 5.3% vs. 1.6 ± 3.5%), but each differed significantly from cells cultured with IC patients urine (p < 0.0001). When inhibition of thymidine incorporation greater than 2 standard deviations from the mean of cell controls was used as the definition for APF activity, 59/64 (92%) of IC patients had evidence for APF activity vs. only 1/30 (3%) of normal controls. None of the samples from bacterial cystitis or bladder cancer patients had any evidence for APF activity.
Figure 1 APF activity in urine from Chinese interstitial cystitis patients with Hunner's ulcers (IC-U), Chinese interstitial cystitis patients without Hunner's ulcers (IC-N), normal controls (Ctr), and patients with bacterial cystitis (BC), or bladder cancer (BCa). APF activity was measured as inhibition of 3H-thymidine incorporation in normal bladder epithelial cells. Each data point indicates the mean change in incorporation, with each specimen tested in triplicate. Horizontal line indicates value of mean; vertical line indicates standard error of the mean for each group.
Levels of HB-EGF and EGF in urine from IC patients and normal controls
We next measured HB-EGF and EGF levels in urine specimens from the same Chinese IC patients, normal controls, bacterial cystitis patients and bladder cancer patients by ELISA. As shown in Figure 2, the concentration of urine HB-EGF was significantly lower in the 38 IC patients with Hunner's ulcers (1.19 ± 0.20 ng/ml) and in the 26 IC patients without Hunner's ulcers (1.42 ± 0.23 ng/ml) as compared to normal controls (9.28 ± 1.04 ng/ml) or patients with bacterial cystitis or bladder cancer (5.34 ± 1.19 ng/ml, and 5.72 ± 0.87 ng/ml, p < 0.0001 for comparison of each IC group to each control group). Mean HB-EGF levels in the two groups of IC patients did not differ significantly from each other (p = 0.43). The mean concentration of urine EGF (Figure 3), however, was significantly higher in 38 IC patients with Hunner's ulcers (21.90 ± 1.19 ng/ml) than in 26 IC patients without Hunner's ulcers (16.32 ± 1.44 ng/ml) (p <0.004), and was markedly higher for both IC groups as compared to normal controls (6.49 ± 0.57 ng/ml) or patients with bacterial cystitis or bladder cancer (6.32 ± 1.26 ng/ml, and 8.03 ± 1.95 ng/ml, p < 0.0001 for comparison of each IC group to each control group).
Figure 2 HB-EGF levels in urine from IC patients, normal controls, bacterial cystitis patients and bladder cancer patients. HB-EGF levels were measured by ELISA in urine specimens from interstitial cystitis patients with Hunner's ulcers (IC-U), interstitial cystitis patients without Hunner's ulcers (IC-N), asymptomatic controls (Ctr), and patients with bacterial cystitis (BC), or bladder cancer (BCa). Each data point is the mean value for duplicate specimens. Horizontal line indicates value of mean; vertical line indicates standard error of mean for each group.
Figure 3 EGF levels in urine from IC patients, normal controls, bacterial cystitis patients and bladder cancer patients. EGF levels were measured by ELISA in urine specimens from interstitial cystitis patients with Hunner's ulcers (IC-U), interstitial cystitis patients without Hunner's ulcers (IC-N), asymptomatic controls (Ctr), and patients with bacterial cystitis (BC), or bladder cancer (BCa). Each data point is the mean value for duplicate specimens. Horizontal line indicates value of mean; vertical line indicates standard error of mean for each group.
Discussion
This report presents evidence that three urine biomarkers for IC (APF, HB-EGF and EGF) were confirmed by measurement in our laboratory at First Hospital of China Medical University using bladder epithelial cells from Chinese normal controls and urine specimens from a new group of Chinese IC patients and controls that had not previously been studied. Our data were comparable to those published previously [3-5], with 92% of IC patients having APF activity as compared to only 3% of controls.
APF is a low molecular weight peptide made by bladder epithelial cells from IC patients that inhibits the proliferation of human bladder epithelial cells [6,8], suggesting that it may cause the bladder epithelial thinning or ulceration common in IC [9]. Bladder cell proliferation is also known to be influenced by growth factors and their regulatory proteins. HB-EGF has been shown to be important for replication of a variety of epithelial cells including hepatocytes, keratinocytes, gastric epithelial cells, and uterine epithelial cells and is known to stimulate bladder epithelial replication in vitro [10]. It is therefore possible that decreased synthesis of HB-EGF by epithelial or other bladder cells contributes to the pathogenesis of IC by impairing normal bladder epithelial regeneration. APF can inhibit HB-EGF production by bladder epithelial cells [8], indicating a possible mechanism for APF's antiproliferative activity.
The reproducibility of APF, HB-EGF and EGF as biomarkers for IC in our patients indicates that one or more of these factors may be useful for the diagnosis of IC in patients from various racial backgrounds, and measured in different laboratories. As previously reported by Keay, et al [11], these biomarkers can clearly distinguish between IC and control groups whether they are normalized to urine creatinine or urine volume [11]. We therefore normalized the levels to urine volume for our study. We also compared APF, HB-EGF and EGF levels in IC patients with Hunner's ulcers vs. IC patients without Hunner's ulcers, and determined that mean values for all three markers were more abnormal in patients with ulcers, although the difference between IC patients with Hunner's ulcers and IC patients without Hunner's ulcers was only statistically significant for EGF in this study. Whether this finding indicates that ulcerative IC is a more severe form of the disorder than nonulcerative IC requires further investigation.
Several biomarkers have been described for IC as recently reviewed [12]. GP51, a glycoprotein urinary marker, reportedly is also specific for IC [13], but additional studies have not been done on this marker. Nitric oxide synthase (NOS) stimulates the production of nitric oxide (NO) which then increases cyclic GMP levels by activating guanylyl cyclase. In some studies, female IC patients have been shown to have significantly decreased NOS activity in their urine cell pellet than female controls, and urinary cyclic GMP levels were significantly lower in female IC patients than in female controls or females with urinary tract infections [5,14]. However, bladder luminal NO levels have also been shown to be markedly increased in patients with IC, and elevated bladder luminal NO levels are not specific for IC [15,16] but also occur in patients with various forms of cystitis. CD45RO positive lymphocytes are another biomarker for ulcerative IC patients. They are not found in the urine of healthy subjects, but are also found in bladder cancer patients treated with BCG [17,18]. Although they also are not specific for IC, these lymphocytes may be biomarkers that reflect the severity of interstitial bladder inflammation.
Urine APF activity and levels of HB-EGF, EGF, IL-6 and IGFBP3 were shown to be significantly different between IC patients and normal controls in one large comparison study [5]. However, APF, HB-EGF, and EGF were the most sensitive and specific for IC, with anti-proliferative factor activity most clearly separating the interstitial cystitis and control groups (5).
Conclusion
APF, HB-EGF and EGF have now been confirmed as good biomarkers for IC by another laboratory. Furthermore, we provide the first evidence that these IC biomarkers are present in both the urine of patients with ulcerative as well as nonulcerative IC, and the first comparison of these markers between these two groups of IC patients. Although EGF levels were significantly more abnormal in ulcerative than nonulcerative patients, APF and HB-EGF were not significantly different between the two groups, and the clinical significance of the differences in EGF levels is therefore uncertain at this time.
List of abbreviations used
IC – interstitial cystitis
APF – antiproliferative factor
HB-EGF – heparin-binding epidermal growth factor-like growth factor
EGF – epidermal growth factor
PBS – phosphate buffered saline
FBS – fetal bovine serum
ELISA – enzyme-linked immunosorbent assay
HBE – human bladder epithelial
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
CZ provided input into the design of these studies, assistance with development of the assays used for these experiments in the laboratory at China Medical University, data analysis, and writing of the manuscript. ZL provided assistance with patient diagnoses and collection of specimens and clinical data, and supervised all aspects of the work performed for this paper in his laboratory. CK provided assistance with patient diagnoses, as well as collection of specimens and clinical data, and assisted with data analysis.
Table 1 Patient Characteristics
IC-U* IC-N+ Control BC‡ BCa†
Age 41 ± 11.1 46 ± 10.6 36.6 ± 8.8 36.4 ± 9.6 55.4 ± 7.1
Gender (F/M) 36/2 23/3 28/2 8/2 3/7
Symptom Index Score points 19.6 ± 0.12 18.9 ± 0.29 - - -
Problem Index Score points 15.7 ± 0.15 15.5 ± 0.21 - - -
Average voiding volume (ml) 95 ± 15 98 ± 18 - - -
Average bladder capacity (ml)# 205 ± 16 216+17 - - -
* IC-U = IC patients with Hunner's ulcers
+ IC-N = IC patients without Hunner's ulcers
‡ BC = Bacterial Cystitis
† BCa = Bladder Cancer
# Bladder capacity determined while patients were awake, prior to distension
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This research was supported by China Health Ministry, Wu Jie-ping Funding.
==== Refs
Hanno PM Staskin DR Krane RJ Wein AJ eds Interstitial cystitis 1990 Springer-Verlag, London
Division of Kidney, Urologic, and Hematologic Diseases (DKUHD) of the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) Diagnostic criteria for research studies (interstitial cystitis) Am J Kidney Dis 1989 13 353 354
Zhang C-O Li Z-L Shoenfelt JL Kong C-Z Chai TC Erickson DE Peters KM Rovner ES Keay S Comparison of APF activity and epithelial growth factor levels in urine from Chinese, African American and Caucasian American IC patients Urology 2003 61 897 901 12735999 10.1016/S0090-4295(02)02597-9
Keay S Zhang C-O Shoenfelt JL Erickson DR Whitmore K Warren JW Marvel R Chai T Sensitivity and specificity of antiproliferative factor, heparin-binding epidermal growth factor-like growth factor, and epidermal growth factor as urine markers for interstitial cystitis Urology 2001 57 9 14 11378043 10.1016/S0090-4295(01)01127-X
Erickson DR Xie SX Bhavanandan VP Wheeler MA Hurst RE Demers LM Kushner L Keay SK A comparison of multiple urine markers for interstitial cystitis J Urol 2002 167 2461 2469 11992058 10.1097/00005392-200206000-00026
Keay S Szekely Z Conrads TP Veenstra TD Barchi JJ JrZhang C-O Koch KR Michejda CJ An antiproliferative factor from interstitial cystitis patients is a frizzled 8 protein-related sialoglycopeptide PNAS 2004 101 11803 11808 15282374 10.1073/pnas.0404509101
O'Leary MP Sant GR Fowler FJ Whitmore KE Spolarich-Kroll J The interstitial cystitis symptom index and problem index Urology 1997 49 58 63 9146003 10.1016/S0090-4295(99)80333-1
Keay S Kleinberg M Zhang C-O Hise MK Warren JW Bladder epithelial cells from interstitial cystitis patients produce an inhibitor of HB-EGF production J Urol 2000 64 2112 2118 11061938 10.1097/00005392-200012000-00074
Tomaszewski JE Landis JR Russack V Williams TM Wang LP Hardy C Brensinger C Matthews YL Abele ST Kusek JW Nyberg LM Biopsy features are associated with primary symptoms in interstitial cystitis: results from the interstitial cystitis database study Urology 2001 57 67 81 11378053 10.1016/S0090-4295(01)01166-9
Freeman MR Yoo JJ Raab G Soker S Adam RM Schneck FX Renshaw AA Klagsbrun M Atala A Heparin-binding EGF-like growth factor is an autocrine growth factor for human urothelial cells and is synthesized by epithelial and smooth muscle cells in the human bladder J Clin Invest 1997 99 1028 1036 9062361
Keay S Zhang C-O Kagen DI Hise MK Jacobs SC Hebel JR Gordon D Whitmore K Bodison S Warren JW Concentrations of specific epithelial growth factor in the urine of interstitial cystitis patients and controls J Urol 1997 158 1983 1988 9334654
Keay S Takeda M Tamaki M Hanno P Current and future directions in diagnostic markers in interstitial cystitis International J Urol 2003 10 S27 S30 10.1046/j.1442-2042.10.s1.8.x
Byrne DS Sedor JF Estojak J Fitzpatrick KJ Chiura AN Mulholland SG The urinary glycoprotein GP51 as a clinical marker for interstitial cystitis J Urol 1999 161 1786 1790 10332435
Smith SD Wheeler MA Foster HE JrWeiss RM Urinary nitric oxide synthase activity and cyclic GMP levels are decreased with interstitial cystitis and increased with urinary tract infections J Urol 1996 155 1432 1435 8632605 10.1097/00005392-199604000-00098
Ehren I Hosseini A Lundberg JO Wiklund NP Nitric oxide: A useful gas in the detection of lower urinary tract inflammation J Urol 1999 162 327 329 10411031 10.1097/00005392-199908000-00011
Ehren I Hosseini A Herulf M Lundberg JO Wiklund NP Measurement of luminal nitric oxide in bladder inflammation using a silicon balloon catheter: A novel minimally invasive method Urology 1999 54 264 267 10443722 10.1016/S0090-4295(99)00120-X
Ueda T Tamaki M Ogawa O Yamauchi T Yoshimura N Improvement of interstitial cystitis and problems that developed during treatment with oral IPD-1151T J Urol 2000 164 1917 1920 11061880 10.1097/00005392-200012000-00011
Chang SG Lee SJ Huh JS Lee JH Changes in mucosal immune cells of bladder tumor patient after BCG intravesical immunotherapy Oncol Rep 2001 8 257 261 11182036
| 0 | PMC1131911 | CC BY | 2021-01-04 16:37:36 | no | Biomed Eng Online. 2005 Apr 20; 4:28 | latin-1 | Biomed Eng Online | 2,005 | 10.1186/1475-925X-4-28 | oa_comm |
==== Front
Cardiovasc DiabetolCardiovascular Diabetology1475-2840BioMed Central London 1475-2840-4-51586213310.1186/1475-2840-4-5ReviewOxidative stress and the use of antioxidants in diabetes: Linking basic science to clinical practice Johansen Jeanette Schultz [email protected] Alex K [email protected] David J [email protected] Adviye [email protected] University of Tromso, Tromso, Norway2 University of Georgia College of Pharmacy, Athens, Georgia, USA3 Medical College of Georgia Vascular Biology Center, Augusta, Georgia, USA2005 29 4 2005 4 5 5 9 3 2005 29 4 2005 Copyright © 2005 Johansen 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.
Cardiovascular complications, characterized by endothelial dysfunction and accelerated atherosclerosis, are the leading cause of morbidity and mortality associated with diabetes. There is growing evidence that excess generation of highly reactive free radicals, largely due to hyperglycemia, causes oxidative stress, which further exacerbates the development and progression of diabetes and its complications. Overproduction and/or insufficient removal of these free radicals result in vascular dysfunction, damage to cellular proteins, membrane lipids and nucleic acids. Despite overwhelming evidence on the damaging consequences of oxidative stress and its role in experimental diabetes, large scale clinical trials with classic antioxidants failed to demonstrate any benefit for diabetic patients. As our understanding of the mechanisms of free radical generation evolves, it is becoming clear that rather than merely scavenging reactive radicals, a more comprehensive approach aimed at preventing the generation of these reactive species as well as scavenging may prove more beneficial. Therefore, new strategies with classic as well as new antioxidants should be implemented in the treatment of diabetes.
AntioxidantsDiabetesOxidative Stress
==== Body
Introduction
It is a well-established fact that diabetes is a risk factor for cardiovascular disease [1,2]. While microvascular complications of diabetes include nephropathy and retinopathy, macrovascular complications resulting in atherosclerotic cardiovascular disease such as coronary artery disease, cerebrovascular disease and peripheral vascular disease are the leading cause of death in the diabetic population [3,4]. The Diabetes Control and Complications trial (DCCT) demonstrated that tight control of blood glucose is effective in reducing clinical complications significantly, but even optimal control of blood glucose could not prevent complications suggesting that alternative treatment strategies are needed [4]. Since numerous studies demonstrated that oxidative stress, mediated mainly by hyperglycemia-induced generation of free radicals, contributes to the development and progression of diabetes and related contributions, it became clear that ameliorating oxidative stress through treatment with antioxidants might be an effective strategy for reducing diabetic complications. To this end, several clinical trials investigated the effect of the antioxidant vitamin E on the prevention of diabetic complications. However, these trials failed to demonstrate relevant clinical benefits of this antioxidant on cardiovascular disease [5-7]. The negative results of the clinical trials with antioxidants prompted new studies focusing on the mechanisms of oxidative stress in diabetes in order to develop causal antioxidant therapy. In this article, sources of free radicals contributing to oxidative stress and the natural defense mechanisms in diabetes are briefly reviewed. Experimental and clinical evidence with respect to the use of conventional antioxidants in diabetes is summarized and causal therapy approaches with novel antioxidants are discussed.
What is oxidative stress?
Oxidative stress is defined in general as excess formation and/or insufficient removal of highly reactive molecules such as reactive oxygen species (ROS) and reactive nitrogen species (RNS) [8,9]. ROS include free radicals such as superoxide (•O2-), hydroxyl (•OH), peroxyl (•RO2), hydroperoxyl (•HRO2-) as well as nonradical species such as hydrogen peroxide (H2O2) and hydrochlorous acid (HOCl) [8,10]. RNS include free radicals like nitric oxide (•NO) and nitrogen dioxide (•NO2-), as well as nonradicals such as peroxynitrite (ONOO-), nitrous oxide (HNO2) and alkyl peroxynitrates (RONOO) [8,10]. Of these reactive molecules, •O2-, •NO and ONOO- are the most widely studied species and play important roles in the diabetic cardiovascular complications. Thus, these species will be discussed in more detail.
•NO is normally produced from L-arginine by endothelial nitric oxide synthase (eNOS) in the vasculature [8]. •NO mediates endothelium-dependent vasorelaxation by its action on guanylate cyclase in vascular smooth muscle cells (VSMC), initiating a cascade that leads to vasorelaxation. •NO also displays antiproliferative properties and inhibits platelet and leukocyte adhesion to vascular endothelium [8]. Therefore, •NO is considered a vasculoprotective molecule. However, •NO easily reacts with superoxide, generating the highly reactive molecule ONOO-, and triggering a cascade of harmful events as discussed below [8,11]. Therefore its chemical environment, i.e. presence of •O2-, determines whether •NO exerts protective or harmful effects.
Production of one ROS or RNS may lead to the production of others through radical chain reactions. As summarized in Fig. 1. •O2- is produced by one electron reduction of oxygen by several different oxidases including NAD(P)H oxidase, xanthine oxidase, cyclooxygenase and even eNOS under certain conditions as well as by the mitochondrial electron transport chain during the course of normal oxidative phosphorylation, which is essential for generating ATP [12-15]. Under normal conditions, •O2- is quickly eliminated by antioxidant defense mechanisms. •O2- is dismutated to H2O2 by manganese superoxide dismutase (Mn-SOD) in the mitochondria and by copper (Cu)-SOD in the cytosol [12]. H2O2 is converted to H20 and O2 by glutathione peroxidase (GSH-Px) or catalase in the mitochondria and lysosomes, respectively. H2O2 can also be converted to the highly reactive •OH radical in the presence of transition elements like iron and copper.
Why are reactive species bad?
While ROS are generated under physiological conditions and are involved to some extent as signaling molecules and defense mechanisms as seen in phagocytosis, neutrophil function, and shear-stress induced vasorelaxation, excess generation in oxidative stress has pathological consequences including damage to proteins, lipids and DNA. These detrimental effects are briefly summarized in this section.
ROS can stimulate oxidation of low-density lipoprotein (LDL), and ox-LDL, which is not recognized by the LDL receptor, can be taken up by scavenger receptors in macrophages leading to foam cell formation and atherosclerotic plaques [16]. As will be discussed in greater detail in the next section, •O2- can activate several damaging pathways in diabetes including accelerated formation of advanced glycation end products (AGE), polyol pathway, hexosamine pathway and PKC, all of which have been proven to be involved in micro- and macrovascular complications. •O2- and H2O2 stimulate stress-related signaling mechanisms such as NF-κB, p38-MAPK and STAT-JAK resulting in VSMC migration and proliferation. In endothelial cells, H2O2 mediates apoptosis and pathological angiogenesis [15]. Furthermore, •O2- immediately reacts with •NO generating cytotoxic ONOO- and this reaction itself has several consequences. First, ONOO- alters function of biomolecules by protein nitration as well as causing lipid peroxidation [8]. For example, potassium channels, which regulate the vasorelaxation response, are inhibited by nitration [17,18]. As recently reviewed by Turko et al, increased levels of nitrotyrosine are associated with apoptosis of myocytes, endothelial cells and fibroblasts in diabetes [8]. Second, ONOO- causes single-strand DNA breakage which in turn activates nuclear enzyme poly(ADP-ribose) polymerase (PARP) [19]. Third, it decreases •NO bioavailability causing impaired relaxation and inhibition of the antiproliferative effects of •NO [9]. Furthermore, ONOO- oxidizes tetrahydrobiopterin (BH4), an important cofactor for NOS, and causes uncoupling of NOS, which produces •O2- instead of •NO [9]. ROS-induced peroxidation of membrane lipids alters the structure and the fluidity of biological membranes, which ultimately affects function [9,13-15]. All these pathological modifications contribute to the pathogenesis of vascular dysfunction.
Sources of oxidative stress in diabetes
Direct evidence of oxidative stress in diabetes is based on studies that focused on the measurement of oxidative stress markers such as plasma and urinary F2-isoprostane as well as plasma and tissue levels of nitrotyrosine and •O2- [11,20-23]. There are multiple sources of oxidative stress in diabetes including nonenzymatic, enzymatic and mitochondrial pathways. Thus, we will first discuss these mechanisms and conclude with the recently proposed working plan for the initiation of oxidative stress and related vascular complications in diabetes.
Nonenzymatic sources of oxidative stress originate from the oxidative biochemistry of glucose. Hyperglycemia can directly cause increased ROS generation. Glucose can undergo autoxidation and generate •OH radicals [8]. In addition, glucose reacts with proteins in a nonenzymatic manner leading to the development of Amadori products followed by formation of AGEs. ROS is generated at multiple steps during this process. In hyperglycemia, there is enhanced metabolism of glucose through the polyol (sorbitol) pathway, which also results in enhanced production of •O2-.
Enzymatic sources of augmented generation of reactive species in diabetes include NOS, NAD(P)H oxidase and xanthine oxidase [21,22,24]. All isoforms of NOS require five cofactors/prosthetic groups such as flavin adenine dinucleotide (FAD), flavin mononucleotide (FMN), heme, BH4 and Ca2+-calmodulin. If NOS lacks its substrate L-arginine or one of its cofactors, NOS may produce •O2- instead of •NO and this is referred to as the uncoupled state of NOS [9,21,22,24]. NAD(P)H oxidase is a membrane associated enzyme that consists of five subunits and is a major source of •O2- production [21,22,25,26]. Guzik et al. investigated •O2- levels in vascular specimens from diabetic patients and probed sources of •O2- using inhibitors of NOS, NAD(P)H oxidase, xanthine oxidase and mitochondrial electron transport chain. This study demonstrated that there is enhanced production of •O2- in diabetes and this is predominantly mediated by NAD(P)H oxidase. Furthermore, the NOS-mediated component is greater in patients with diabetes than in patients who do not have diabetes [22]. We have also observed that NAD(P)H oxidase activity is significantly higher in vascular tissue (saphenous vein and internal mammary artery) obtained from diabetic patients [27]. There is plausible evidence that PKC, which is stimulated in diabetes via multiple mechanisms, i.e. polyol pathway and Ang II, activates NAD(P)H oxidase [28].
The mitochondrial respiratory chain is another source of nonenzymatic generation of reactive species. During the oxidative phosphorylation process, electrons are transferred from electron carriers NADH and FADH2, through four complexes in the inner mitochondrial membrane, to oxygen, generating ATP in the process [29]. Under normal conditions, •O2- is immediately eliminated by natural defense mechanisms. A recent study demonstrated that hyperglycemia-induced generation of •O2- at the mitochondrial level is the initial trigger of vicious cycle of oxidative stress in diabetes [30,31]. When endothelial cells are exposed to hyperglycemia at the levels relevant to clinical diabetes, there is increased generation of ROS and especially •O2-, which precedes the activation of four major pathways involved in the development of diabetic complications. Nishikawa and colleagues elegantly demonstrated that generation of excess pyruvate via accelerated glycolysis under hyperglycemic conditions floods the mitochondria and causes •O2- generation at the level of Complex II in the respiratory chain. What is more important is that blockade of •O2- radicals by three different approaches using either a small molecule uncoupler of mitochondrial oxidative phosphorylation (CCCP), overexpression of uncoupling protein-1 (UCP1) or overexpression of Mn-SOD, prevented changes in NF-κB as well as polyol pathway, AGE formation and PKC activity. Based on this information, it has been postulated by several groups that mitochondrial •O2- is the initiating snowball that turns oxidative stress into an avalanche in diabetes by stimulating more ROS and RNS production via downstream activation of NF-κB-mediated cytokine production, PKC and NAD(P)H oxidase (Fig. 2). Thus, inhibition of intracellular free radical formation would provide a causal therapy approach in the prevention of oxidative stress and related vascular complications in diabetes.
Natural defense against oxidative stress and antioxidants
Reactive species can be eliminated by a number of enzymatic and nonenzymatic antioxidant mechanisms. As discussed above, SOD immediately converts •O2- to H2O2, which is then detoxified to water either by catalase in the lysosomes or by glutathione peroxidase in the mitochondria (Fig. 1). Another enzyme that is important is glutathione reductase, which regenerates glutathione that is used as a hydrogen donor by glutathione peroxidase during the elimination of H2O2. Maritim and colleagues recently reviewed in detail that diabetes has multiple effects on the protein levels and activity of these enzymes, which further augment oxidative stress by causing a suppressed defense response [9]. For example, in the heart, which is an important target in diabetes and prone to diabetic cardiomyopathy leading to chronic heart failure, SOD and glutathione peroxidase expression as well as activity are decreased whereas catalase is increased in experimental models of diabetes [9,32,33]. In patients with chronic heart failure, all three enzymes are decreased in the smooth muscle [34] and exercise training can upregulate the expression and activity of antioxidant enzymes. Increased isoprostane levels in diabetic patients with chronic heart failure are correlated with antioxidant status and disease severity [35]. Thus, modulation of these enzymes in target organs prone to diabetic complications such as heart and kidney may prove beneficial in the prevention and management of heart failure and kidney failure.
Nonenzymatic antioxidants include vitamins A, C and E; glutathione; α-lipoic acid; carotenoids; trace elements like copper, zinc and selenium; coenzyme Q10 (CoQ10); and cofactors like folic acid, uric acid, albumin, and vitamins B1, B2, B6 and B12. Alterations in the antioxidant defense system in diabetes have recently been reviewed [11]. Glutathione (GSH) acts as a direct scavenger as well as a co-substrate for GSH peroxidase. It is a major intracellular redox tampon system. Vitamin E is a fat-soluble vitamin that prevents lipid peroxidation. It exists in 8 different forms, of which α-tocopherol is the most active form in humans. Hydroxyl radical reacts with tocopherol forming a stabilized phenolic radical which is reduced back to the phenol by ascorbate and NAD(P)H dependent reductase enzymes [36,37]. CoQ10 is an endogenously synthesized compound that acts as an electron carrier in the Complex II of the mitochondrial electron transport chain. Brownlee et al reported that this is the site of •O2- generation under hyperglycemic conditions [30,31]. CoQ10 is a lipid soluble antioxidant, and in higher concentrations, it scavenges •O2- and improves endothelial dysfunction in diabetes [38-40]. Vitamin C (ascorbic acid) increases NO production in endothelial cells by stabilizing NOS cofactor BH4 [41]. α-Lipoic acid is a hydrophilic antioxidant and can therefore exert beneficial effects in both aqueous and lipid environments. α-lipoic acid is reduced to another active compound dihydrolipoate. Dihydrolipoate is able to regenerate other antioxidants such as vitamin C, vitamin E and reduced glutathione through redox cycling [41]. Thus, both experimental and clinical studies summarized in the next sections utilized these naturally occurring antioxidants, especially vitamins C, E and α-lipoic acid, in order to delineate the role of oxidative stress in the development of vascular complications of diabetes.
Evidence from experimental models
A multitude of in vivo studies have been performed utilizing antioxidants in experimental diabetic models. The effects of antioxidants on oxidative stress are measured through certain observable biomarkers. These markers include the enzymatic activities of catalase, SOD, GSH-Px, and GSH-reductase, as well as thiobarbituric acid reactants (TBARS) levels, an indirect measurement of free-radical production that has been shown to be consistently elevated in diabetes. Normalization of the activity levels of any of these markers, and ultimately, the balance of free-radical production/removal, would be an effective method to reduce ROS-induced damage. Many animal studies have been completed with this aim in mind and indeed have shown that diabetes-induced alterations of oxidative stress indicators can be reversed when the animals are treated with various antioxidants. It should be noted that a plethora of studies have been done with numerous antioxidant compounds. We will, however, only cover a select few within the scope of this review, specifically the compounds for which a corresponding human clinical trial has been conducted
Mekinova et al. demonstrated that supplementation of streptozotocin (STZ) diabetic rats with vitamins C, E, and beta-carotene for 8 weeks produced a significant reduction of TBARS levels, GHS, and GHS-Px, an increase in Cu-SOD, and no change in catalase activity in kidneys [42]. Treatment with vitamins C and E was also shown to decrease urinary albumin excretion, glomerular basement membrane thickness, and kidney weight in STZ diabetic rats [43]. In the same study, vitamins C and E significantly lowered malondialdehyde (TBARS) levels and GSH-Px activity while increasing catalase and SOD activities when compared to unsupplemented diabetic animals [43]. A study by Cinar et el. demonstrated that supplementation with vitamin E significantly lowered liver and lung TBARS levels and improved impaired endothelium-dependent vasorelaxation in STZ diabetic rat aorta [44].
α-lipoic acid, which is involved in mitochondrial dehydrogenase reactions, has gained a considerable amount of attention as an antioxidant. Studies have demonstrated that intraperitoneal administration of α-lipoic acid to STZ diabetic Wistar rats normalizes TBARS level in plasma, retina, liver, and pancreas [45]. In the same study, Obrosova et.al observed a reduction of GSH activity in diabetic retina and that supplementation with α-lipoic acid produced no change [45]. However, another study demonstrated an increase in aortic GSH-Px in STZ diabetic rats that was normalized by treatment with α-lipoic acid [46]. Additionally, increased maximum contractile responses in diabetic aortic rings were ameliorated with α-lipoic acid treatment [46].
SOD activity is undoubtedly important to the regulation of oxidative status in diabetes. However, there is variation as to the status of this enzyme in the diabetic state. Some studies have reported decreased SOD activity [43,45] while others have shown increases [47] or no change in the enzyme [42,48]. α-lipoic acid has been observed to normalize diabetes-induced decreases of SOD in rat heart [48] and retina [45]. One study demonstrated that treatment of STZ diabetic rats with α-lipoic acid reverses SOD-induced vasorelaxation, potentially due to the elimination of excess superoxide/hydrogen peroxide and the recovery of basal NO [46]. A recent study by Brands et al. investigated the effect of oxidative stress in the development of hypertension in diabetes using the SOD mimetic tempol in a Type 1 model of diabetes where NOS is pharmacologically inhibited with a NOS inhibitor, L-NAME [49]. In this model, hyperglycemia causes hypertension implicating an important role for NO. Results of this study showed that if •O2- is eliminated by tempol early in the disease process, the hypertension and decrease in glomerular filtration precipitated by diabetes are prevented.
In summary, there are differences in response to antioxidants in experimental diabetes in the prevention of cardiovascular complications. Studies in experimental models provide a foundation for the clinical studies but results should be interpreted cautiously since the experimental models of diabetes, duration and type of antioxidant treatment and markers of oxidative stress investigated in these studies exhibit a wide range.
Evidence from clinical trials
Although studies with antioxidants in experimental models as well as observational studies strongly suggest that antioxidants should confer beneficial effects in reducing cardiovascular complications in diabetes, clinical evidence for the use of antioxidants is not solid. It should be emphasized that clinical trials with antioxidants in diabetes are limited and majority of these trials focused on the use of vitamin E and C and lately α-lipoic acid. Thus, we will attempt to group the clinical trials by the antioxidants used.
Small trials with vitamin E demonstrated beneficial cardiovascular effects. In a double-blind, placebo-controlled, randomized study, vitamin E supplementation (1000 IU/day) for three months in patients with Type 1 diabetes (n = 41) significantly improved endothelium-dependent vasorelaxation [50]. In another study, Beckman et al. reported that administration of vitamin E (800 IU/day) and C (1000 mg/day) combination for six months had a positive effect on endothelium-dependent vasorelaxation in Type 1 diabetic patients (n = 26) but had no effect in Type 2 diabetes (n = 23) [51]. Gaede et al reported that vitamin E (680 mg/day) and C (1250 mg/day) combination significantly improved renal function in Type 2 diabetes [52].
Other clinical trials on a larger scale include the Heart Outcomes Prevention Evaluation (HOPE) trial [53], Secondary Prevention with Antioxidants of Cardiovascular Disease in End Stage Renal Disease (SPACE) trial [54], the Steno trial [55], the Primary Prevention Project (PPP) trial [56] and the Study to Evaluate Carotid Ultrasound Changes in Patients Treated With Ramipril and Vitamin E (SECURE) trial [57].
The HOPE trial enrolled patients 55 years of age or older who were at high risk for cardiovascular disease and recruited significant number of patients with diabetes. This study had a 2 × 2 factorial design where in one arm patients were randomized to vitamin E (400 IU/day) or placebo and in the other arm of the study patients were randomized to ramipril (10 mg/day) or placebo [53]. Results with vitamin E and ramipril were evaluated separately as compared to respective placebo groups. In the vitamin E arm, 4761 patients received vitamin E and 4780 patients were given placebo. In the treatment and placebo groups, the number of patients with diabetes was 1838 and 1816, respectively. The primary endpoint was a composite of myocardial infarction, stroke and death from cardiovascular causes. The trial was stopped for ethical reasons after 4.5 years follow-up by the recommendations of an independent data and safety monitoring board based on the beneficial effects of ramipril on cardiovascular events in the concurrent treatment group and lack of effect in the vitamin E treatment group. Results of the study were published in 2000 and demonstrated that there was no significant difference in the primary outcome between vitamin E and placebo groups [53]. Analyses of the secondary endpoints of the study, which included total mortality, hospitalizations for heart failure and unstable angina, revascularization and nephropathy, were recently published [58], and again vitamin E supplementation for 4.5 years failed to provide any benefit in cardiovascular outcomes or nephropathy. It was also reported that there were no significant adverse events associated with vitamin E. The HOPE trial was the largest trial conducted thus far for the use of antioxidants in diabetes. The SECURE trial was designed as a substudy of the HOPE trial to evaluate the effects of long-term treatment with ramipril and vitamin E on atherosclerosis progression in high-risk patients. In this trial, 732 patients who had vascular disease or diabetes were randomized to two doses of (2.5 or 10 mg/d) ramipril and vitamin E (400 IU/day) or placebo and progression of atherosclerosis was monitored by B-mode carotid ultrasound. While ramipril slowed down atherosclerotic changes, vitamin E had no effect as compared to placebo group.
The SPACE trial recruited 196 hemodialysis patients with preexisting cardiovascular disease who were assigned to either placebo (n = 99) or 800 IU/day vitamin E (n = 97) for 2 years. 43% of the patients in each group had diabetes. The primary endpoint was a composite of myocardial infarction, stroke, peripheral arterial disease or unstable angina. There was a 46% decrease in the primary end point events in the vitamin E group and this was mainly due to a 70% reduction in total myocardial infarction [54]. The PPP trial was a randomized trial again with a 2 × 2 design to evaluate the effect of low dose aspirin (100 mg/day) and vitamin E (300 mg/day) on the prevention of cardiovascular complications in high-risk patients. Similar to the studies discussed above, the primary endpoint was a composite of cardiovascular death, stroke or myocardial infarction. Out of the 4784 patients recruited, 1031 had diabetes. The PPP trial was stopped prematurely by the recommendations of an independent data and safety monitoring board based on the consistent beneficial effects of aspirin as compared to placebo group. However, there was no significant effect of vitamin E treatment either in diabetic or nondiabetic subjects. Lastly, the Steno-2 trial compared the effect of a multifactorial intensive therapy (n = 80) with that of conventional treatment (n = 80) on modifiable risk factors for cardiovascular disease in patients with Type 2 diabetes [55]. In the intensive treatment group, patients received pharmacotherapy that targeted hyperglycemia, dyslipidemia, hypertension and microalbuminuria including daily supplementation of vitamin C (250 mg), E (100 mg), folic acid (400 mg) and chromium picolinate (100 mg) as well as behavior modification including low-fat diet, exercise and smoking cessation. The control group received conventional therapy as recommended in national guidelines. The intensive therapy resulted in almost a 50% decrease in the risk of cardiovascular events providing evidence that a multifactorial approach is superior to conventional therapy for the prevention of oxidative stress-induced vascular complications in diabetes.
Studies with α-lipoic acid are approved for the treatment of diabetic neuropathy and results are more promising than those obtained with vitamin E. In the Alpha Lipoic Acid in Diabetic Neuropathy (ALADIN) study, infusion of α-lipoic acid (>600 mg) significantly improved patient symptoms [59]. The ALADIN II Study demonstrated that long-term (24 months) use of α-lipoic acid (600 or 1200 mg) improved nerve function [60]. ALADIN III, a randomized multicenter double-blind placebo controlled study, showed that in a cohort of 509 patients, 600 mg α-lipoic acid administration for 6 months improved neuropathy impairment score as early as 19 days, which was maintained up to 7 months [61]. The DEKAN (Deutsche kardiale autonome neuropathie) study evaluated the effect of 800 mg α-lipoic acid or placebo in diabetic patients with cardiac autonomic neuropathy for 4 months and showed that heart rate variability, an indicator of cardiac autonomic neuropathy, significantly improved with α-lipoic acid treatment [62]. The SYDNEY trial investigated the effect of α-lipoic acid treatment on sensory symptoms of diabetic polyneuropathy as assessed by the Total Symptom Score. Administration of this antioxidant over a 3-week period improved sensory symptoms such as pain, prickling and numbness [63]. A recent meta-analysis of trials with α-lipoic acid concluded that treatment with intravenous α-lipoic acid (600 mg/day) over a 3-week period is safe and effective in improving positive neuropathic symptoms as well as neuropathic deficits [64].
In summary, clinical trials with conventional antioxidants in diabetic patients are limited. For major cardiovascular outcomes, vitamin E failed to provide any benefit. However, when study population was limited to diabetic patients alone as done in diabetic neuropathy trials, α-lipoic acid has proven to be effective. As further discussed under Perspectives, this antioxidant may be a viable option in trials focusing on cardiovascular outcomes in diabetes.
In addition to the many antioxidants examined above, a number of commonly used drugs have shown promising antioxidant activity in addition to their primary pharmacological activity. These drugs include thiazolidinediones (TZDs), HMG-CoA reductase inhibitors (statins), and inhibitors of the renin-angiotensin system.
Thiazolidinediones (TZDs) have been shown in many animal studies to have antioxidant effect. In one study, pioglitazone-treated rats had reduced urinary excretion of isoprostane, a marker of oxidative stress [65]. In a trial with type-2 diabetic rats, Bagi et al demonstrated that treatment with rosiglitazone reduced NAD(P)H-derived ROS and increased the activity of catalase [66]. Another study using type-2 diabetic rats found that treatment with troglitazone lowered hydroperoxides and decreased SOD activity [67]. A study using troglitazone and pioglitazone in type-2 diabetic rats found that both agents reduced TBARS levels and increased the aortic vasorelaxation response [68].
There is substantial evidence from in vitro studies that statins exert an antioxidant effect. Studies have demonstrated that statin therapy markedly reduces oxidative stress markers (such as nitrated tyrosine) in animals [69]. Although the mechanisms for these actions are still being elucidated, Takayama et al have demonstrated in canine models that the antioxidant effect of statins is at least partially due to inhibition of NAD(P)H oxidase [70]. Statins have also been shown to stimulate the activity of the antioxidant enzyme thioredoxin [71]. Additionally, statin therapy has been shown to stimulate the activity of paraoxonase (PON), which has a putative role in protecting LDL from oxidation [72]. Oxidation of LDL ex vivo has been shown to be inhibited by long-term statin therapy, an effect thought to be partly due to the binding of the statins to the LDL itself. It seems likely from the above studies that the antioxidant actions of statins are manifested via a variety of mechanisms.
Inhibitors of Angiotensin II (Ang II) activity, such as Angiotensin Converting Enzyme Inhibitors (ACEIs) and Angiotensin II receptor blockers (ARBs) have shown some beneficial effects that may stem from their antioxidant properties. Angiotensin II has been shown to increase ROS levels in animal studies, through stimulation of NAD(P)H oxidase activity [15,73]. Studies have suggested that this effect also occurs in humans [73,74]. Ang II has also been implicated in upregulating the expression of the LOX-1 receptor, which is specific for oxidized LDL cholesterol. Inhibition of the generation of Ang II, whether by ACEI or ARB, should therefore attenuate these deleterious processes. Indeed, Berry et al have shown that treatment with ACEI or ARB decreases •O2- levels in the human vasculature [75].
In summary, many of the agents which are a mainstay of pharmacotherapy in diabetes have been shown to have antioxidant properties in addition to their primary pharmacological actions. These antioxidant properties may be a contributing factor to the therapeutic efficacy of these agents. Their antioxidant properties make the case for use of these drugs even more compelling. Particularly in light of the lackluster results seen in clinical trials with antioxidant supplementation, health care providers should redouble their efforts to ensure adequate usage of the demonstrably effective agents summarized above.
Perspectives- Is there a role for antioxidant treatment in diabetes?
Although the clinical trials conducted to date failed to provide adequate support for the use of antioxidants in diabetes, it is still to early to reach a definitive conclusion on this issue. As discussed above, with the exception of alpha-lipoic acid studies in diabetic neuropathy, data from clinical trials are limited. The majority of studies were not designed to assess the effect of antioxidant use specifically in diabetic patients. This is an important point because diabetic patients represent a population in whom oxidative stress is much higher than in the general population. As was seen in the SPACE trial of patients on hemodialysis, patients exposed to very high oxidative stress responded favorably to vitamin E supplementation [54]. It is possible that antioxidants would be more demonstrably effective in a patient population chosen on the basis of elevated levels of oxidative stress. Unfortunately, none of the studies to date effectively assessed the baseline oxidative stress of the enrolled patients using any of the commonly accepted markers of inflammation.
The human trials to date used endpoints that were not directly related to oxidative stress, but rather gross markers of overall cardiovascular health, such as effect on mortality. The studies failed to assess the duration of the diabetic disease states, arguably a large confounding variable. In assessing oxidative stress and the effects of antioxidants thereon, specific markers of oxidative stress should be measured.
With respect to the specific antioxidants studied, their selection was based on epidemiological and observational data, and in the absence of any solid grasp of the underlying mechanisms of action. Whereas observational studies are based on whole populations and reflect the lifelong influence of dietary habits, most of the studies were five years duration or less and included older patients (average age 65.4 years). It is possible that the study populations represented patients in whom the disease states had progressed too far to be amenable to antioxidant intervention.
In all likelihood, the choice and dose of antioxidant might be very important. The clinical trials focused mainly on the use of vitamin E. Negative results with vitamins cannot be generalized to all antioxidants. As has been eloquently argued elsewhere, treating the antioxidant vitamins as a single class of compounds with expected similar effects inappropriately disregards their wide range of chemical properties and pharmacodynamics [76]. Clinical trials to date have been conducted without any real understanding of the mechanisms of action or the concentrations of the various agents seen at different physiological sites. Indeed, there is not sufficient evidence to demonstrate that vitamin E reaches target cells.
Recently, it has been postulated that antioxidant potency of vitamins such as C and E is limited because these antioxidants work as scavengers of existing excess reactive species in a stoichiometric manner and this approach represents a symptomatic approach to oxidative stress-associated clinical problems [77]. Based on the new developments in our understanding of the pathophysiology of oxidative stress, it is clear that strategies to block the formation of reactive radicals will provide a targeted and causal approach to provide conclusive evidence whether antioxidants should be part of the cardiovascular treatment plan in diabetes. Candidate agents include low molecular weight mitochondrial and cytosolic SOD and catalase mimetics, L-propionyl carnitine, PKC-β inhibitor LY-333531, peroxynitrite catalyst FP15 and mitochondrial uncoupler DNP [9,77,78].
Given the number of shortcomings in the clinical trials, it seems clear that more research on the use of antioxidants in the prevention of cardiovascular complications in diabetes is necessary and strongly encouraged. From a clinical viewpoint, however, efforts for the prevention of diabetic complications should seek to maximize the benefits of proven therapeutic strategies including appropriate life style changes and controlling blood pressure, blood glucose and lipids.
In conclusion, the amount of evidence on the harmful effects of oxidative stress on vascular function and the link to pathophysiological mechanisms underlying diabetic complications is compelling. While the lack of clinical evidence on the beneficial effects of antioxidant vitamins in diabetes management should not deter us from more basic and clinical research on this issue, practice guidelines that are based on the results of numerous clinical trials should be our guide to evidence-based medicine in the prevention of cardiovascular disease in diabetes. The recent American Heart Association science advisory on the subject of antioxidant vitamins and cardiovascular disease asserted that there is insufficient evidence to justify the use of antioxidant vitamins for cardiovascular disease risk reduction [79]. Hopefully, further research into the pathophysiology of oxidative stress and the role of antioxidant therapy will lead to appropriately-designed clinical trials in which the promise of antioxidant therapy will be realized.
Competing Interests
The author(s) declare that they have no competing interests.
Authors' Contributions
AKH and JSJ contribute equally to writing the evidence-based sections and drafting of this review. DR was responsible for critical revision and formatting. AE participated in all aspects and areas of this review.
Acknowledgements
This work was supported by grants from NIH (HL076236-01), American Heart Association Scientist Development Grant and American Diabetes Association to Adviye Ergul and an AHA Southeast Affiliate Predoctoral Fellowship Award to Alex K. Harris.
Figures and Tables
Figure 1 Generation of reactive species in diabetes. Highlighted in gray are some of the most important ROS and RNS in vascular cells. Oxygen is converted to •O2- via the activation of enzymatic and nonenzymatic pathways, which is then dismutated to H2O2 by SOD. H2O2 can be converted to H2O by catalase or glutathione peroxidase (GSH-Px) or to •OH after reaction with Cu or Fe. Glutathione reductase regenerates glutathione (GSH). In addition, •O2- reacts rapidly with •NO to form ONOO-.
Figure 2 Current working model for the generation of reactive species and downstream targets in diabetes. Excess generation of mitochondrial ROS due to hyperglycemia initiates a vicious circle by activating stress-sensitive pathways such as NF-κB, p38 MAPK and Jak/STAT, polyol (sorbitol) and hexosamine pathways, PKC and AGEs. Enhanced production of AGEs, sorbitol and proinflammatory cytokines exerts a positive feedback on ROS and RNS synthesis and potentiates PKC-mediated vascular dysfunction by altering gene expression as well as vascular function and structure.
==== Refs
Pyorala K Laakso M Uusitupa M Diabetes and atherosclerosis: an epidemiologic view Diabetes Metab Rev 1987 3 463 524 3552530
Laakso M Hyperglycemia and cardiovascular disease in type 2 diabetes Diabetes 1999 48 937 942 10331395
American Diabetes Association National diabetes fact sheet 2002
The Diabetes Control and Complications Trial Research Group The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus N Engl J Med 1993 329 977 986 8366922 10.1056/NEJM199309303291401
Giugliano D Ceriello A Paolisso G Oxidative stress and diabetic vascular complications Diabetes Care 1996 19 257 267 8742574
Ceriello A New insights on oxidative stress and diabetic complications may lead to a "causal" antioxidant therapy Diabetes Care 2003 26 1589 1596 12716823
Ceriello A Motz E Is oxidative stress the pathogenic mechanism underlying insulin resistance, diabetes, and cardiovascular disease? The common soil hypothesis revisited Arterioscler Thromb Vasc Biol 2004 24 816 823 14976002 10.1161/01.ATV.0000122852.22604.78
Turko IV Marcondes S Murad F Diabetes-associated nitration of tyrosine and inactivation of succinyl-CoA:3-oxoacid CoA-transferase Am J Physiol Heart Circ Physiol 2001 281 H2289 2294 11709394
Maritim AC Sanders RA Watkins JB 3rd Diabetes, oxidative stress, and antioxidants: A review J Biochem Mol Toxicol 2003 17 24 38 12616644 10.1002/jbt.10058
Evans JL Goldfine ID Maddux BA Grodsky GM Oxidative stress and stress-activated signaling pathways: a unifying hypothesis of type 2 diabetes Endocr Rev 2002 23 599 622 12372842 10.1210/er.2001-0039
Vega-Lopez S Devaraj S Jialal I Oxidative stress and antioxidant supplementation in the management of diabetic cardiovascular disease J Investig Med 2004 52 24 32 14989367
Evans JL Goldfine ID Maddux BA Grodsky GM Are oxidative stress-activated signaling pathways mediators of insulin resistance and {beta}-cell dysfunction? Diabetes 2003 52 1 8 12502486
Griendling KK FitzGerald GA Oxidative stress and cardiovascular injury: Part I: Basic mechanisms and in vivo monitoring of ROS Circulation 2003 108 1912 1916 14568884 10.1161/01.CIR.0000093660.86242.BB
Griendling KK FitzGerald GA Oxidative stress and cardiovascular injury: Part II: animal and human studies Circulation 2003 108 2034 2040 14581381 10.1161/01.CIR.0000093661.90582.c4
Taniyama Y Griendling KK Reactive oxygen species in the vasculature: Molecular and cellular mechanisms Hypertension 2003 42 1075 1081 14581295 10.1161/01.HYP.0000100443.09293.4F
Boullier A Bird DA Chang MK Dennis EA Friedman P Gillotre-Taylor K Horkko S Palinski W Quehenberger O Shaw P Scavenger receptors, oxidized LDL, and atherosclerosis Ann N Y Acad Sci 2001 947 214 222 11795269
Liu Y Gutterman DD The coronary circulation in diabetes: influence of reactive oxygen species on K+ channel-mediated vasodilation Vascul Pharmacol 2002 38 43 49 12378822 10.1016/S1537-1891(02)00125-8
Liu Y Terata K Chai Q Li H Kleinman LH Gutterman DD Peroxynitrite inhibits Ca2+-activated K+ channel activity in smooth muscle of human coronary arterioles Circ Res 2002 91 1070 1076 12456494 10.1161/01.RES.0000046003.14031.98
Soriano FG Virag L Szabo C Diabetic endothelial dysfunction: role of reactive oxygen and nitrogen species production and poly(ADP-ribose) polymerase activation J Mol Med 2001 79 437 448 11511974 10.1007/s001090100236
Oberg BP McMenamin E Lucas FL McMonagle E Morrow J Ikizler TA Himmelfarb J Increased prevalence of oxidant stress and inflammation in patients with moderate to severe chronic kidney disease Kidney Int 2004 65 1009 1016 14871421 10.1111/j.1523-1755.2004.00465.x
Guzik TJ West NE Black E McDonald D Ratnatunga C Pillai R Channon KM Vascular superoxide production by NAD(P)H oxidase: association with endothelial dysfunction and clinical risk factors Circ Res 2000 86 E85 90 10807876
Guzik TJ Mussa S Gastaldi D Sadowski J Ratnatunga C Pillai R Channon KM Mechanisms of increased vascular superoxide production in human diabetes mellitus: role of NAD(P)H oxidase and endothelial nitric oxide synthase Circulation 2002 105 1656 1662 11940543 10.1161/01.CIR.0000012748.58444.08
Ceriello A Mercuri F Quagliaro L Assaloni R Motz E Tonutti L Taboga C Detection of nitrotyrosine in the diabetic plasma: evidence of oxidative stress Diabetologia 2001 44 834 838 11508267 10.1007/s001250100529
Aliciguzel Y Ozen I Aslan M Karayalcin U Activities of xanthine oxidoreductase and antioxidant enzymes in different tissues of diabetic rats J Lab Clin Med 2003 142 172 177 14532905 10.1016/S0022-2143(03)00110-0
Kitada M Koya D Sugimoto T Isono M Araki S Kashiwagi A Haneda M Translocation of glomerular p47phox and p67phox by protein kinase C-beta activation is required for oxidative stress in diabetic nephropathy Diabetes 2003 52 2603 2614 14514646
Etoh T Inoguchi T Kakimoto M Sonoda N Kobayashi K Kuroda J Sumimoto H Nawata H Increased expression of NAD(P)H oxidase subunits, NOX4 and p22phox, in the kidney of streptozotocin-induced diabetic rats and its reversibity by interventive insulin treatment Diabetologia 2003 46 1428 1437 13680125 10.1007/s00125-003-1205-6
Ergul A Schultz Johansen J Stromhaug C Harris AK Hutchinson J Tawfik A Rahimi A Rhim E Wells B Caldwell RW Vascular dysfunction of venous bypass conduits is mediated by reactive oxygen species in diabetes: Role of endothelin-1 J Pharmacol Exp Ther 2004 313 70 77 15608082 10.1124/jpet.104.078105
Amiri F Shaw S Wang X Tang J Waller JL Eaton DC Marrero MB Angiotensin II activation of the JAK/STAT pathway in mesangial cells is altered by high glucose Kidney Int 2002 61 1605 1616 11967010 10.1046/j.1523-1755.2002.00311.x
Green K Brand MD Murphy MP Prevention of mitochondrial oxidative damage as a therapeutic strategy in diabetes Diabetes 2004 53 S110 118 14749275
Nishikawa T Edelstein D Du XL Yamagishi S Matsumura T Kaneda Y Yorek MA Beebe D Oates PJ Hammes HP Normalizing mitochondrial superoxide production blocks three pathways of hyperglycaemic damage Nature 2000 404 787 790 10783895 10.1038/35008121
Brownlee M Biochemistry and molecular cell biology of diabetic complications Nature 2001 414 813 820 11742414 10.1038/414813a
Kaul N Siveski-Iliskovic N Hill M Khaper N Seneviratne C Singal PK Probucol treatment reverses antioxidant and functional deficit in diabetic cardiomyopathy Mol Cell Biochem 1996 160–161 283 288 8813704 10.1007/BF00240060
Hayden MR Tyagi SC Myocardial redox stress and remodeling in metabolic syndrome, type 2 diabetes mellitus, and congestive heart failure Med Sci Monit 2003 9 SR35 52 12883468
Linke A Adams V Schulze PC Erbs S Gielen S Fiehn E Mobius-Winkler S Schubert A Schuler G Hambrecht R Antioxidative effects of exercise training in patients with chronic heart failure: increase in radical scavenger enzyme activity in skeletal muscle Circulation 2005 111 1763 1770 15809365 10.1161/01.CIR.0000165503.08661.E5
Cristina Polidori M Pratico D Savino K Rokach J Stahl W Mecocci P Increased F2 isoprostane plasma levels in patients with congestive heart failure are correlated with antioxidant status and disease severity J Card Fail 2004 10 334 338 15309701 10.1016/j.cardfail.2003.11.004
Hensley K Robinson KA Gabbita SP Salsman S Floyd RA Reactive oxygen species, cell signaling, and cell injury Free Radic Biol Med 2000 28 1456 1462 10927169 10.1016/S0891-5849(00)00252-5
Hensley K Benaksas EJ Bolli R Comp P Grammas P Hamdheydari L Mou S Pye QN Stoddard MF Wallis G New perspectives on vitamin E: gamma-tocopherol and carboxyelthylhydroxychroman metabolites in biology and medicine Free Radic Biol Med 2004 36 1 15 14732286 10.1016/j.freeradbiomed.2003.10.009
Hodgson JM Watts GF Playford DA Burke V Croft KD Coenzyme Q10 improves blood pressure and glycaemic control: a controlled trial in subjects with type 2 diabetes Eur J Clin Nutr 2002 56 1137 1142 12428181 10.1038/sj.ejcn.1601464
Hodgson JM Watts GF Can coenzyme Q10 improve vascular function and blood pressure? Potential for effective therapeutic reduction in vascular oxidative stress Biofactors 2003 18 129 136 14695928
Watts GF Playford DA Croft KD Ward NC Mori TA Burke V Coenzyme Q(10) improves endothelial dysfunction of the brachial artery in Type II diabetes mellitus Diabetologia 2002 45 420 426 11914748 10.1007/s00125-001-0760-y
Heller R Unbehaun A Schellenberg B Mayer B Werner-Felmayer G Werner ER L-ascorbic acid potentiates endothelial nitric oxide synthesis via a chemical stabilization of tetrahydrobiopterin J Biol Chem 2001 276 40 47 11022034 10.1074/jbc.M004392200
Mekinova D Chorvathova V Volkovova K Staruchova M Grancicova E Klvanoca J Ondreicka R Effect of intake of exogenous vitamins C, E and beta-carotene on the antioxidative status in kidneys of rats with streptozotocin-induced diabetes Nahrung 1995 39 257 261 7477241
Kedziora-kornatowska K Szram S Kornatowski T Szadujkis-Szadurski L Kedziora J Bartosz G Effect of vitamin E and vitamin C supplementation of antioxidative state and renal glomerular basement membrane thickness in diabetic kidney Exp Nephrology 2003 95 134 143 10.1159/000074840
Cinar M Ulker S Alper G Evinc A Effect of dietary vitamin E supplementation on vascular reactivity of thoracic aorta in streptozotocin-diabetic rats Pharmacology 2001 62 56 64 11150923 10.1159/000056072
Obrosova I Fathallah L Greene D Early changes in lipid peroxidation and antioxidative defense in rat retina Eur J Pharm 2000 398 139 146 10.1016/S0014-2999(00)00286-7
Kocak G Aktan F Canbolat O Ozogul C Elberg S Yildizoglu-Ari N Karasu C Alpha-lipoic acid treatment ameliorates metabolic parameters, blood pressure, vascular reavtivity and morphology of vessels already damaged by streptozotocin-diabetes Diab Nutr Metab 2000 13 308 318
Rauscher F Sanders R Watkins JI Effects of coenzyme Q10 treatment on antioxidant pathways in normal and streptozotocin-induced diabetic rats J Biochem Mol Toxicol 2001 15 41 46 11170314 10.1002/1099-0461(2001)15:1<41::AID-JBT5>3.0.CO;2-Z
Maritim A Sanders R Watkins JI Effects of alpha-lipoic acid on biomarkers of oxidative stress in streptozotic-induced diabetic rats J Nutr Biochem 2003 14 288 294 12832033 10.1016/S0955-2863(03)00036-6
Brands MW Bell TD Gibson B Nitric oxide may prevent hypertension early in diabetes by counteracting renal actions of superoxide Hypertension 2004 43 57 63 14656952 10.1161/01.HYP.0000104524.25807.EE
Skyrme-Jones RA O'Brien RC Berry KL Meredith IT Vitamin E supplementation improves endothelial function in type I diabetes mellitus: a randomized, placebo-controlled study J Am Coll Cardiol 2000 36 94 102 10898419 10.1016/S0735-1097(00)00720-8
Beckman JA Goldfine AB Gordon MB Garrett LA Keaney JF JrCreager MA Oral antioxidant therapy improves endothelial function in Type 1 but not Type 2 diabetes mellitus Am J Physiol 2003 285 H2392 2398
Gaede P Poulsen HE Parving HH Pedersen O Double-blind, randomised study of the effect of combined treatment with vitamin C and E on albuminuria in Type 2 diabetic patients Diabet Med 2001 18 756 760 11606175 10.1046/j.0742-3071.2001.00574.x
Yusuf S Dagenais G Pogue J Bosch J Sleight P Vitamin E supplementation and cardiovascular events in high-risk patients. The Heart Outcomes Prevention Evaluation Study Investigators N Engl J Med 2000 342 154 160 10639540 10.1056/NEJM200001203420302
Boaz M Smetana S Weinstein T Matas Z Gafter U Iaina A Knecht A Weissgarten Y Brunner D Fainaru M Secondary prevention with antioxidants of cardiovascular disease in endstage renal disease (SPACE): randomised placebo-controlled trial Lancet 2000 356 1213 1218 11072938 10.1016/S0140-6736(00)02783-5
Gaede P Vedel P Larsen N Jensen GVH Parving H-H Pedersen O Multifactorial intervention and cardiovascular disease in patients with Type 2 diabetes N Engl J Med 2003 348 383 393 12556541 10.1056/NEJMoa021778
Sacco M Pellegrini F Roncaglioni MC Avanzini F Tognoni G Nicolucci A Primary prevention of cardiovascular events with low-dose aspirin and Vitamin E in Type 2 diabetic patients: Results of the Primary Prevention Project (PPP) trial Diabetes Care 2003 26 3264 3272 14633812
Lonn EM Yusuf S Dzavik V Doris CI Yi Q Smith S Moore-Cox A Bosch J Riley WA Teo KK Effects of ramipril and vitamin E on atherosclerosis : The study to evaluate carotid ultrasound changes in patients treated with ramipril and vitamin E (SECURE) Circulation 2001 103 919 925 11181464
Lonn E Yusuf S Hoogwerf B Pogue J Yi Q Zinman B Bosch J Dagenais G Mann JFE Gerstein HC Effects of vitamin E on cardiovascular and microvascular outcomes in high-risk patients with diabetes: Results of the HOPE study and MICRO-HOPE substudy Diabetes Care 2002 25 1919 1927 12401733
Ziegler D Hanefeld M Ruhnau KJ Meissner HP Lobisch M Schutte K Gries FA Treatment of symptomatic diabetic peripheral neuropathy with the anti-oxidant alpha-lipoic acid. A 3-week multicentre randomized controlled trial (ALADIN Study) Diabetologia 1995 38 1425 1433 8786016 10.1007/BF00400603
Reljanovic M Reichel G Rett K Lobisch M Schuette K Moller W Tritschler HJ Mehnert H Treatment of diabetic polyneuropathy with the antioxidant thioctic acid (alpha-lipoic acid): a two year multicenter randomized double-blind placebo-controlled trial (ALADIN II). Alpha Lipoic Acid in Diabetic Neuropathy Free Radic Res 1999 31 171 179 10499773
Ziegler D Hanefeld M Ruhnau K Hasche H Lobisch M Schutte K Kerum G Malessa R Treatment of symptomatic diabetic polyneuropathy with the antioxidant alpha-lipoic acid: a 7-month multicenter randomized controlled trial (ALADIN III Study). ALADIN III Study Group. Alpha-Lipoic Acid in Diabetic Neuropathy Diabetes Care 1999 22 1296 1301 10480774
Ziegler D Schatz H Conrad F Gries FA Ulrich H Reichel G Effects of treatment with the antioxidant alpha-lipoic acid on cardiac autonomic neuropathy in NIDDM patients. A 4-month randomized controlled multicenter trial (DEKAN Study). Deutsche Kardiale Autonome Neuropathie Diabetes Care 1997 20 369 373 9051389
Ametov AS Barinov A Dyck PJ Hermann R Kozlova N Litchy WJ Low PA Nehrdich D Novosadova M O'Brien PC The sensory symptoms of diabetic polyneuropathy are improved with {alpha}-Lipoic Acid: The SYDNEY Trial Diabetes Care 2003 26 770 776 12610036
Ziegler D Nowak H Kempler P Vargha P Low PA Treatment of symptomatic diabetic polyneuropathy with the antioxidant alpha-lipoic acid: a meta-analysis Diabet Med 2004 21 114 121 14984445 10.1111/j.1464-5491.2004.01109.x
Dobrian AD Davies MJ Schriver SD Lauterio TJ Prewitt RL Oxidative stress in a rat model of obesity-induced hypertension Hypertension 2001 37 554 560 11230334
Bagi Z Koller A Kaley G PPARgamma activation, by reducing oxidative stress, increases NO bioavailability in coronary arterioles of mice with Type 2 diabetes Am J Physiol 2004 286 H742 748
Fukui T Noma T Mizushige K Aki Y Kimura S Abe Y Dietary troglitazone decreases oxidative stress in early stage type II diabetic rats Life Sci 2000 66 2043 2049 10823344 10.1016/S0024-3205(00)00531-2
lida KT Kawakami Y Suzuki M Shimano H Toyoshima H Sone H Shimada K Iwama Y Watanabe Y Mokuno H Effect of thiazolidinediones and metformin on LDL oxidation and aortic endothelium relaxation in diabetic GK rats Am J Physiol 2003 284 E1125 1130
Wilson SH Simari RD Best PJ Peterson TE Lerman LO Aviram M Nath KA Holmes DR JrLerman A Simvastatin preserves coronary endothelial function in hypercholesterolemia in the absence of lipid lowering Arterioscler Thromb Vasc Biol 2001 21 122 128 11145943
Takayama T Wada A Tsutamoto T Ohnishi M Fujii M Isono T Horie M Contribution of vascular NAD(P)H oxidase to endothelial dysfunction in heart failure and the therapeutic effects of HMG-CoA reductase inhibitor Circ J 2004 68 1067 1075 15502390 10.1253/circj.68.1067
Haendeler J Hoffmann J Zeiher AM Dimmeler S Antioxidant effects of statins via S-nitrosylation and activation of thioredoxin in endothelial cells: a novel vasculoprotective function of statins Circulation 2004 110 856 861 15289372 10.1161/01.CIR.0000138743.09012.93
Deakin S Leviev I Guernier S James RW Simvastatin modulates expression of the PON1 gene and increases serum paraoxonase: a role for sterol regulatory element-binding protein-2 Arterioscler Thromb Vasc Biol 2003 23 2083 2089 14500290 10.1161/01.ATV.0000096207.01487.36
Touyz RM Reactive oxygen species and angiotensin II signaling in vascular cells – implications in cardiovascular disease Braz J Med Biol Res 2004 37 1263 1273 15273829 10.1590/S0100-879X2004000800018
Touyz RM Schiffrin EL Ang II-stimulated superoxide production is mediated via phospholipase D in human vascular smooth muscle cells Hypertension 1999 34 976 982 10523394
Berry C Anderson N Kirk AJ Dominiczak AF McMurray JJ Renin angiotensin system inhibition is associated with reduced free radical concentrations in arteries of patients with coronary heart disease Heart 2001 86 217 220 11454847 10.1136/heart.86.2.217
Steinberg D Witztum JL Is the oxidative modification hypothesis relevant to human atherosclerosis? Do the antioxidant trials conducted to date refute the hypothesis? Circulation 2002 105 2107 2111 11980692 10.1161/01.CIR.0000014762.06201.06
Cuzzocrea S Riley DP Caputi AP Salvemini D Antioxidant therapy: A new pharmacological approach in shock, inflammation, and ischemia/reperfusion injury Pharmacol Rev 2001 53 135 159 11171943
Szabo C Mabley JG Moeller SM Shimanovich R Pacher P Virag L Soriano FG Van Duzer JH Williams W Salzman AL Part I: pathogenetic role of peroxynitrite in the development of diabetes and diabetic vascular complications: studies with FP15, a novel potent peroxynitrite decomposition catalyst Mol Med 2002 8 571 580 12477967
Kris-Etherton PM Lichtenstein AH Howard BV Steinberg D Witztum JL Antioxidant vitamin supplements and cardiovascular disease Circulation 2004 110 637 641 15289389 10.1161/01.CIR.0000137822.39831.F1
| 15862133 | PMC1131912 | CC BY | 2021-01-04 16:04:34 | no | Cardiovasc Diabetol. 2005 Apr 29; 4:5 | utf-8 | Cardiovasc Diabetol | 2,005 | 10.1186/1475-2840-4-5 | oa_comm |
==== Front
Clin Mol AllergyClinical and molecular allergy : CMA1476-7961BioMed Central London 1476-7961-3-41583311010.1186/1476-7961-3-4ResearchSkin testing versus radioallergosorbent testing for indoor allergens Chinoy Birjis [email protected] Edgar [email protected] Sami L [email protected] Allergy and Immunology Section, Louisiana State University Health Sciences Center; Shreveport, Louisiana, USA2005 15 4 2005 3 4 4 27 1 2005 15 4 2005 Copyright © 2005 Chinoy et al; licensee BioMed Central Ltd.2005Chinoy 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
Skin testing (ST) is the most common screening method for allergy evaluation. Measurement of serum specific IgE is also commonly used, but less so by allergists than by other practitioners. The sensitivity and specificity of these testing methods may vary by type of causative allergen and type of allergic manifestation. We compared ST reactivity with serum specific IgE antibodies to common indoor allergens in patients with respiratory allergies.
Methods
118 patients (3 mo-58 yr, mean 12 yr) with allergic rhinitis and/or bronchial asthma had percutaneous skin testing (PST) supplemented by intradermal testing (ID) with those allergens suspected by history but showed negative PST. The sera were tested blindly for specific IgE antibodies by the radioallergosorbent test (Phadebas RAST). The allergens were D. farinae (118), cockroach (60), cat epithelium (90), and dog epidermal (90). Test results were scored 0–4; ST ≥ 2 + and RAST ≥ 1 + were considered positive.
Results
The two tests were in agreement (i.e., either both positive or both negative) in 52.2% (dog epidermal) to 62.2% (cat epithelium). When RAST was positive, ST was positive in 80% (dog epidermal) to 100% (cockroach mix). When ST was positive, RAST was positive in 16.3% (dog epidermal) to 50.0% (D. farinae). When RAST was negative, ST was positive in 48.5% (cat epithelium) to 69.6% (D. farinae). When ST was negative, RAST was positive in 0% (cockroach) to 5.6% (cat epithelium). The scores of ST and RAST showed weak to moderate correlation (r = 0.24 to 0.54). Regardless of history of symptoms on exposure, ST was superior to RAST in detecting sensitization to cat epithelium and dog epidermal.
Conclusion
For all four indoor allergens tested, ST was more sensitive than RAST. When both tests were positive, their scores showed poor correlation. Sensitizations to cat epithelium and dog epidermal are common, even in subjects who claimed no direct exposure.
AllergySkin testingRASTSpecific IgEMiteCockroachCatDog
==== Body
Background
Skin testing (ST) and specific serum IgE antibody measurement are commonly used in allergy evaluation. Percutaneous skin testing (PST) is the most common screening method. Intradermal testing (ID) is usually used for aeroallergens that show negative PST, yet are suspected by the patient or by the environmental history. ST requires the discontinuation of antihistamines and other drugs that have antihistaminic effect for intervals ranging from days to weeks before testing. Serum specific IgE measurement by the radioallergosorbent test (RAST) or its analogues is also frequently used, albeit more commonly so by non-allergists. In some situations, RAST may be preferred over ST [1]. In clinical practice, it is of importance to know the reliability of RAST compared to ST. Inhalation provocation testing would be the most reliable for respiratory allergies, but its clinical use in practice is limited to occupational cases. The objective of the present study was to compare ST with RAST for indoor aeroallergens in patients with respiratory allergies.
Methods
Patients
118 patients (ages 3 - 58 yr, mean 12 yr) with a history of respiratory allergies (allergic rhinitis and/or asthma) were routinely evaluated in the allergy clinic.
Skin Testing
ST was done with extracts of the common aeroallergens. Commercial crude extracts (1:10 in 50% glycerin; Hollister-Stier, Spokane, WA) were used for PST (scratch method). Aeroallergens that showed negative PST in spite of a suggestive history were tested intradermally (ID) with 1:1000 crude aqueous extracts. Positive and negative controls were included using histamine (1 mg/ml for PST and 0.01 mg/ml for ID) and normal saline solution, respectively. The test result was read at 20 minutes for PST and at 15 minutes for ID testing. ST (PST and ID) was scored 0–4 as compared to the negative and positive controls [2], ST reactions ≥ 2 + were considered positive.
Specific IgE
Sera from most patients were tested in a blind fashion for specific IgE antibodies by Phadebas RAST (Pharmacia Diagnostics, Kalamazoo, MI) and the result was scored 0–4 according to the manufacturer's criteria; scores ≥ 1 + (≥ 0.35 PRU/ml) were considered positive.
Allergens
Four common indoor allergens were studied, namely: Dermatophagoides farinae, cockroach mix, cat epithelium, and dog epidermal.
Statistics
Chi-square test was used for comparing frequencies (or percentages). Student's t-test was used for comparison of two means. Correlation coefficient was calculated for quantitative relationships.
Results
The concordances and discordances of ST (PST ± ID) and RAST are presented in Table 1. The two tests were in agreement (i.e., both positive or both negative) in 52.2% (dog epidermal) to 62.2% (cat epithelium). When RAST was positive, ST was also positive in 80% (dog epidermal) to 100% (cockroach mix). When ST was positive, RAST was also positive in 16.3% (dog epidermal) to 50.0% (D. farinae). When RAST was negative, ST was positive in 48.5% (cat epithelium) to 69.6% (D. farinae). When ST was negative, RAST was positive in 0% (cockroach) to 5.6% (cat epithelium). Comparisons of the RAST results with the results of PST and ID tests, separately or in combination, are presented in Figures 1, 2, 3, 4.
Table 1 Concordance and discordance between skin testing (PST ± ID) and RAST in all patients tested for D. farinae, cockroach mix, cat epithelium and dog epidermal.
ST & RAST comparison D. farinae
n = 118 Cockroach
n = 60 Cat epithelium
n = 90 Dog epidermal
n = 90
Concordance Both + or - 69/118 (58.5%) 32/60 (53.3%) 56/90 (62.2%) 47/90 (52.2%)
ST+ of RAST+ 48/49 (98.0%) 8/8 (100%) 22/24 (91.7%) 8/10 (80.0%)
ST- of RAST+ 1/49 (2.0%) 0/8 (0%) 2/24 (8.3%) 2/10 (20.0%)
ST+ of RAST- 48/69 (69.6%) 28/52 (53.8%) 32/66 (48.5%) 41/80 (51.3%)
ST- of RAST- 21/69 (30.4%) 24/52 (46.2%) 34/66 (51.5%) 39/41 (48.7%)
RAST+ of ST+ 48/96 (50.0%) 8/36 (22.2%) 22/54 (40.7%) 8/49 (16.3%)
RAST- of ST+ 48/96 (50.0%) 28/36 (77.8%) 32/54 (59.3%) 41/49 (83.7%)
RAST+ of ST- 1/22 (4.5%) 0/24 (0%) 2/36 (5.6%) 2/41 (4.9%)
RAST- of ST- 21/22 (95.5%) 24/24 (100%) 34/36 (94.4%) 39/41 (95.1%)
Figure 1 Comparison between skin testing & RAST for D. farinae in patients with respiratory allergy.
Figure 2 Comparison between skin testing & RAST for cockroach mix in patients with respiratory allergy.
Figure 3 Comparison between skin testing & RAST for cat epithelium in patients with respiratory allergy.
Figure 4 Comparison between skin testing & RAST for dog epidermal in patients with respiratory allergy.
For D. farinae (Fig. 1), when ST was positive by either PST or ID, RAST was positive in only 50.0%, whereas when PST and ID were both negative, RAST was negative in 95.5%. When only PST was positive, RAST was positive in 72%, whereas when PST was negative, RAST was negative in 86.0%.
For cockroach mix (Fig. 2), when ST was positive by either PST or ID, RAST was positive in only 22%, whereas when PST and ID were both negative, RAST was negative in 100%. When only PST was positive, RAST was positive in 15%, whereas when PST was negative, RAST was negative in 100%.
For cat epithelium (Fig. 3), when ST was positive by either PST or ID, RAST was positive in only 41%, whereas when PST and ID were both negative, RAST was negative in 94%. When only PST was positive, RAST was positive in 43%, whereas when PST was negative, RAST was negative in 0%.
For dog epidermal (Fig. 4), when ST was positive by either PST or ID, RAST was positive in only 16%, whereas when PST and ID were both negative, RAST was negative in 95%. When only PST was positive, RAST was positive in 7.0%, whereas when PST was negative, RAST was negative in 83%.
Regardless of history of symptoms on exposure, ST was superior to RAST in detecting sensitization to cat epithelium and dog epidermal (Table 2). In subjects who gave no history of significant exposure to cat or dog, sensitization was detected to cat epithelium in 45% by ST vs. 12% by RAST, and to dog epidermal in 36% by ST vs. 5% by RAST. In patients who had exposure to cat or dog, both ST and RAST tended to be more frequently positive when the patient was aware of symptoms on exposure. The positivity of ST or RAST to cat epithelium and dog epidermal did not differ much relevant to the patient's awareness of a cause-and-effect relationship.
Table 2 Skin test (PST+ID) and RAST positivity to cat epithelium and dog epidermal according to history of exposure & symptoms
History of exposure & symptoms Cat epithelium Dog epidermal
ST+ RAST+ ST+ RAST+
Symptoms on exposure 84% 47% 80% 27%
No symptoms on exposure 75% 45% 64% 11%
No history of exposure 45% 12% 36% 5%
The scores of ST (PST ± ID) and RAST (Table 3) for all patients generally showed weak to moderate correlations (r = 0.24 to 0.54). However, when the analysis was limited to patients in whom both tests were positive, there was a weak, non-significant correlation between the scores of the two tests (r = 0.04 to 0.37).
Table 3 Correlation coefficient (r) between Skin Test (PST ± ID) and RAST scores in patients with respiratory allergies
Allergen Patients tested Patients positive by both ST & RAST
r p r p
D. farinae 0.54 <0.001 0.20 NS
Cockroach mix 0.42 <0.001 0.05 NS
Cat epithelium 0.48 <0.0001 0.04 NS
Dog epidermal 0.24 <0.05 0.37 NS
r: strong 0.8+, moderate 0.4 to <0.8, weak < 0.4
Discussion
In the present study of patients with respiratory allergies, the ST and RAST results showed moderate concordance to the common indoor allergens studied (D. farinae, cockroach mix, cat epithelium and dog epidermal). The two tests were in agreement (either both positive or both negative) in 52.2% for dog epidermal to 62.2% for cat epithelium. Compared to RAST, ST was more commonly positive for all four allergens. When PST was positive, RAST was negative in 93% for dog epidermal, 85% for cockroach mix, 57% for cat epithelium and 28% for D. farinae. When ID was performed with the allergens that were negative by PST, the positivity of ST increased for all four allergens. When both the PST and ID tests were negative, RAST positivity did not exceed 6%. When both ST and RAST were positive, their scores showed weak non-significant correlations (r = 0.04 to 0.37).
Haahtela and Jaakonmäki [3] reported that in patients with positive ST to various allergens, RAST was positive in only 53%. Pascual et al [4] reported a positive ST and RAST in 55.6% for D. farinae and noted that RAST was negative in all patients who had a negative ST. Eriksson et al [5] reported a positive ST and RAST in 40% for dog dander and 73% for cat dander. They did not provide data on RAST positivity when ST was negative. In a study by Collins-Williams and Bremner [6], D. farinae RAST was negative in 6 who had positive ST, whereas RAST was positive in only 1 out of 41 patients with a negative ST. For cat hair, RAST was negative in 7 who had positive ST and was positive in none of 31 negative ST. For dog hair, RAST was negative in 12 who had positive ST and was positive in only 1 out of 31 whose ST was negative.
Tang and Wu [7] noted a strong concordance of 97% between ID testing and RAST for D. farinae, and ST was negative in 1 out of 30 patients with positive RAST. On the other hand, the concordance of ID testing and RAST for dog epidermal was 57%, and RAST was negative in 6 out of 23 positive ID tests. van der Zee et al [8] reported that D. farinae RAST was negative in 33 out of 281 (12%) patients with ID positive tests, and was positive in only 11 out of 379 (3%) with negative ID tests. For cat dander, RAST was negative in 45 out of 212 (21%) patients with positive ID tests, and was positive in only 2 out of 448 (0.4%) with negative ID tests. The poor correlation noted in our study between the scores of ST and RAST, even when both tests were positive, was also reported by Paggiaro et al [9].
The discrepancies between ST and RAST can be due to multiple factors. First, differences in the underlying immunologic basis of the two tests. ST is an in vivo biologic test that mimics the natural immediate-type hypersensitivity reaction, i.e., contact between the allergen and its specific IgE antibody on the mast cell, resulting in the local release of mediators and the formation of wheal-and-flare. On the other hand, RAST is an in vitro measurement of the level of circulating IgE antibodies in the serum, which may not reflect the tissue-fixed IgE antibodies. Second, differences in the allergenic quantity between the extracts used in ST and those used for in vitro testing [10]. When a purified and standardized D. farinae preparation was used for both ST and RAST, a high concordance of 84% was noted [11]. Nevertheless, RAST was negative in 8 out of 16 positive ST and was positive in only 1 out of 17 negative ST. Vanto et al [12] noted that the efficiency of RAST was increased by using dog dandruff instead of dog epithelium. Third, several studies reported marked variations in the efficiency of various in vitro assays for specific serum IgE antibodies [8,12-15], and of various ST techniques [16,17].
Both ST and RAST positivities to cat epithelium and dog epidermal were highest in patients who reported symptoms on exposure, followed by those who did not report such a relationship. The higher sensitivity of ST over RAST to cat epithelium and dog epidermal was noted regardless of the patient's awareness of causal relationship between symptoms and exposure. Interestingly, in patients who claimed no history of exposure to cat or dog, the ST was positive to cat epithelium in 45% and to dog epidermal in 36%. Such allergens are ubiquitous and have been noted in places where such animals do not exist, such as furniture stores [18] and schools [19].
It is of particular interest that RAST to cockroach was negative in 100% of cases that were negative to ST. To the best of our knowledge, there have been no relevant studies in the literature.
Finally, it is worth noting that our findings on specific IgE were based on using Phadebas RAST and should not be extrapolated to the more sensitive ImmunoCAP method (Pharmacia Diagnostics; Kalamazoo, MI) [1,20].
Conclusion
Skin testing, particularly when PST was supplemented with ID, was more sensitive than Phadebas RAST in identification of the four indoor allergens we studied. However, RAST (or its analogues) would be indicated as a substitute for ST in certain cases [1,15] such as patients with dermographism, dermatitis, or who cannot discontinue antihistamines. It may also be preferred in patients with phobia to ST or in infants who have a few suspected allergens. It would be also safer than ST in patients with severe reactions to trivial exposures through inhalation or skin contact [21]. The high sensitization rate to cat and dog allergens in spite of the lack of direct exposure to such pets, underscores the high prevalence of such unsuspected, ubiquitous allergens.
List of abbreviations
ST: skin testing
PST: percutaneous skin testing
ID: intradermal testing
RAST: radioallergosorbent test
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
Birjis Chinoy, MD: data analysis, literature search, abstract presentation, manuscript preparation.
Edgar Yee, MD: study design, laboratory work, data gathering, data analysis.
Sami Bahna, MD, DrPH: planning, supervision and participation throughout the study and manuscript preparation
==== Refs
Dolen WK Skin testing and immunoassays for allergen-specific IgE Clin Rev Allergy Immunol 2001 21 229 239 11725606 10.1385/CRIAI:21:2-3:229
Bahna SL Diagnostic tests for food allergy Clin Rev Allergy 1988 6 259 284 3052774
Haahtela T Jaakonmaki I Relationship of allergen-specific IgE antibodies, skin prick tests and allergic disorders in unselected adolescents Allergy 1981 36 251 256 6172046
Pascual HC Reddy PM Nagaya H Lee SK Lauridsen J Gupta S Jerome Agreement between radioallergosorbent test and skin test Ann Allergy 1977 39 325 331 920997
Eriksson NE Ahlstedt S Belin L Diagnosis of reaginic allergy with house dust, animal dander and pollen allergens in adult patients. I. A comparison between RAST, skin tests and provocation tests Int Arch Allergy Appl Immunol 1976 52 335 346 1017893
Collins-Williams C Bremner K Comparison of skin tests and RAST in the diagnosis of atopic hypersensitivity Ann Allergy 1976 36 161 164 1259202
Tang RB Wu KK Total serum IgE, allergy skin testing, and the radioallergosorbent test for the diagnosis of allergy in asthmatic children Ann Allergy 1989 62 432 435 2719352
van der Zee JS de Groot H van Swieten P Jansen HM Aalberse RC Discrepancies between the skin test and IgE antibody assays: study of histamine release, complement activation in vitro, and occurrence of allergen-specific IgG J Allergy Clin Immunol 1988 82 270 281 3261308 10.1016/0091-6749(88)91011-1
Paggiaro PL Bacci E Amram DL Rossi O Talini D Skin reactivity and specific IgE levels in the evaluation of allergic sensitivity to common allergens for epidemiological purposes Clin Allergy 1986 16 49 55 3955800
Ahlstedt S Eriksson N Lindgren S Roth A Specific IgE determination by RAST compared with skin and provocation tests in allergy diagnosis with birch pollen, timothy pollen and dog epithelium allergens Clin Allergy 1974 4 131 140 4858363
Moxnes A Dale S Andrew E Halvorsen R A new, purified Dermatophagoides farinae allergen preparation. Evaluation with SPT and RAST techniques Allergy 1984 39 339 349 6465481
Vanto T Viander M Koivikko A Schwartz B Lowenstein H RAST in the diagnostic of dog dander allergy. A comparison between three allergen preparations using two variants of RAST Allergy 1982 37 75 85 7137523
Weltman JK Laboratory tests for total and allergen-specific immunoglobulin E N Engl Reg Allergy Proc 1988 9 129 133 3292887
Williams PB Barnes JH Szeinbach SL Sullivan TJ Analytic precision and accuracy of commercial immunoassays for specific IgE: establishing a standard J Allergy Clin Immunol 2000 105 1221 230 10856158 10.1067/mai.2000.105219
Hamilton RG Adkinson NF Jr 23. Clinical laboratory assessment of IgE-dependent hypersensitivity J Allergy Clin Immunol 2003 111 S687 701 12592314 10.1067/mai.2003.123
Menardo JL Bousquet J Michel FB Comparison of three prick test methods with the intradermal test and with the rast in the diagnosis of mite allergy Ann Allergy 1982 48 235 239 7073028
Wood RA Phipatanakul W Hamilton RG Eggleston PA A comparison of skin prick tests, intradermal skin tests, and RASTs in the diagnosis of cat allergy J Allergy Clin Immunol 1999 103 773 779 10329809
Egmar AC Almqvist C Emenius G Lilja G Wickman M Deposition of cat (Fel d 1), dog can f 1), and horse allergen over time in public environments – a model of dispersion Allergy 1998 53 957 961 9821475
Almqvist C Larsson PH Egmar AC Hedren M Malmberg P Wickman M School as a risk environment for children allergic to cats and a site for transfer of cat allergen to homes J Allergy Clin Immunol 1999 103 1012 1017 10359879
Ahlsted S Understanding the usefulness of specific IgE blood tests in allergy Clin Exp Allergy 2001 32 1 7
Tan BM Sher MR Good RA Bahna SL Severe food allergies by skin contact Ann Allergy Asthma Immunol 2001 86 583 586 11379811
| 15833110 | PMC1131913 | CC BY | 2021-01-04 16:36:25 | no | Clin Mol Allergy. 2005 Apr 15; 3:4 | utf-8 | Clin Mol Allergy | 2,005 | 10.1186/1476-7961-3-4 | oa_comm |
==== Front
Clin Mol AllergyClinical and molecular allergy : CMA1476-7961BioMed Central London 1476-7961-3-51585751110.1186/1476-7961-3-5ResearchEffects of dexamethasone on TNF-alpha-induced release of cytokines from purified human blood eosinophils Uings Iain [email protected] Ilaria [email protected] Vladislav [email protected] Susan J [email protected] Dilniya [email protected] Keith P [email protected] Francesca [email protected] Cell Biology Unit, Glaxo Wellcome SKB, Gunnels Wood Stevenage, Herts, SG1 2NY, UK2 Department of Pharmacology, School of Pharmacy, The Hebrew University of Jerusalem, Jerusalem, Israel2005 27 4 2005 3 5 5 11 1 2005 27 4 2005 Copyright © 2005 Uings et al; licensee BioMed Central Ltd.2005Uings 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
TNF-alpha is an important mediator in allergy also for its effects on eosinophils.
Methods
The effect of dexamethasone on TNF-alpha induced eosinophils survival, degranulation (ECP), cytokines release (IL-8, GM-CSF) and adhesion to VCAM-1, ICAM-1 and IgG coated wells (EPO release) were evaluated.
Results
The drug inhibited IL-8 and GM-CSF production, but not viability, degranulation or adhesion in human peripheral blood eosinophils.
Conclusion
These results indicate that part of the activity of glucocorticosteroids on eosinophils may be mediated by their ability to inhibit cytokine secretion that in turn is important for the perpetuation of the allergic inflammation.
TNF-αeosinophilsdexamethasoneallergic inflammationcytokine
==== Body
Background
Eosinophils are bone marrow-derived granulocytes that play a crucial role in allergic inflammation. TNF-α is a pro-inflammatory cytokine synthesized by many inflammatory and structural cells. We previously demonstrated that mast cell-derived TNF-α induced eosinophil survival by autocrine production of GM-CSF [1]. TNF-α is also involved in eosinophil adhesion to endothelial cells and induces eosinophil activation, degranulation, and cytokines production. Glucocorticosteroids (GCS), the main anti-inflammatory drugs in allergic diseases, have been demonstrated to decrease circulating and tissue eosinophils. In vitro dexamethasone can inhibit eosinophil survival [2], expression of adhesion molecules [3], and cytokines production [4]. However, the effect of GCS on TNF-α induced eosinophil activation has only been partially investigated. The present study evaluated the effect of dexamethasone on TNF-α induced eosinophil degranulation, cytokines release and adhesion to VCAM-1, ICAM-1 and IgG.
Materials and Methods
Cells culture
Eosinophils were purified (>95%) from the peripheral blood of healthy non-atopic volunteers as previously described [5]. Freshly isolated eosinophils (viability >98%) were cultured in 96 well flat bottom tissue culture plates (Costar, High Wycombe, UK) (1.5 × 105/200 μl/well) in RPMI-1640 supplemented with 10% heat inactivated foetal calf serum (FCS) containing 2 mM L-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin in the presence or absence of rhTNF-α (0.01–100 ng/ml, R&D Systems, Abingdon, UK), or GM-CSF (10 ng/ml, R&D Systems, Abingdon, UK) as a positive control (37°C, 5% CO2). Dexamethasone (Glaxo-Wellcome-SKB, Stevenage, UK) (1 μM) was added to the eosinophil cultures together with TNF-α(50 ng/ml for degranulation and 20 ng/ml for survival and cytokines release) or with GM-CSF (10 ng/ml). After 22 hrs of culture in absence or presence of dexamethasone (1 μM) eosinophil survival was evaluated by Trypan blue exclusion test.
Eosinophils degranulation and cytokines release
ECP level was measured in the culture supernatants by a RIA kit (ECP, Pharmacia Upjohn, Milton Keynes, UK). GM-CSF and IL-8 content was detected in the culture supernatants by ELISA kit (R&D Systems, Abingdon, UK).
Adhesion assay
Eosinophils (104/100 μl/well) were cultured in medium alone or with TNF-α (20 ng/ml) or GM-CSF (10 ng/ml) in the presence or absence of dexamethasone (1 μM) (30 min, 37°C) in 96 well plates pre-coated with recombinant VCAM-1, ICAM-1 or IgG. As a marker of adhesion EPO was detected as previously described [5]. Results are expressed as mean ± SEM.
Statistical analysis
Statistical analysis was performed by Student's t paired test. A p value of <0.05 was considered statistically significant.
Results
TNF-α significantly increased eosinophil viability in a concentration-dependent fashion, compared to culture in medium alone, with a maximal effect at 20 ng/ml (Figure 1). This effect was not influenced by the addition of dexamethasone in the culture medium (13.3% vs 11.5%).
Figure 1 Effect of TNF-α on eosinophil viability in vitro. Eosinophils were incubated with different concentrations of TNF-α for 22 hrs. Eosinophil viability was evaluated by Trypan blue exclusion test. Values are expressed as percentage of survival increase in the presence of TNF-α vs medium alone. Values are mean ± SEM (n = 14).
Incubation of eosinophils with TNF-α induced a significant release of ECP compared to eosinophils cultured in medium alone (24.5 ± 8.9 vs 6.9 ± 1.2 pg/106; p < 0.05). However, addition of dexamethasone in the cultures did not affect TNF-α-induced ECP release (Figure 2). The release of IL-8 by TNF-α treated eosinophils was dose-dependently proportional to the concentrations of TNF-α. A maximal release was achieved at 100 ng/ml (1770,49 ± 129 pg/ml; p < 0.05) (Figure 3). TNF-α also induced GM-CSF release by eosinophils although to a lesser extent than that of IL-8 (data not shown). Treatment of the cultures with dexamethasone completely blocked the TNF-α-induced release of both IL-8 and GM-CSF (p < 0.05) (Figures 4A–B). TNF-α enhanced significantly the percentage of eosinophil adhesion to VCAM-1, ICAM-1 and IgG in comparison to medium alone, by 177%, 205% and 169%, respectively. However, this effect was not inhibited by dexamethasone (Figure 5).
Figure 2 Effect of dexamethsone on TNF-α-induced ECP release from eosinophils. Eosinophils were cultured in medium alone (control) or with dexamethsone (medium+Dexa) or in the presence of TNF-α alone (50 ng/ml) (TNF-α) or TNF-α with dexamethsone (1 μM) (TNF-α+Dexa). ECP release was evaluated by RIA. Values are mean ± SEM (n = 3).
Figure 3 Effect of TNF-α on IL-8 release from eosinophils. Eosinophils were cultured with different concentrations of TNF-α (0 – 100 ng/ml). IL-8 release was evaluated by ELISA. Values are mean ± SEM (n = 5).
Figure 4 Effect of dexamethasone on TNF-α-induced IL-8 and GM-CSF release from eosinophils. Eosinophils were cultured with medium (control), or with TNF-α (20 ng/ml) (TNF-α) in the presence (Dexa) or absence (0) of dexamethasone (1 μM). IL-8 (A) and GM-CSF (B) release was evaluated by ELISA. Values are mean ± SEM (n = 5). *P < 0.05.
Figure 5 Effect of dexamethsone on TNF-α-induced eosinophil adhesion to VCAM-1 and ICAM-1. Eosinophils were cultured in medium alone or with dexamethsone, or with TNF-α (20 ng/ml) or TNF-α +dexamethsone in 96-wells plate coated with IgG, VCAM-1 and ICAM-1. EPO release was detected by colorimetric assay. Value are mean ± SEM (n = 3).
Discussion
We have shown that dexamethasone inhibits the release of IL-8 and GM-CSF in TNF-α activated human peripheral blood eosinophils from non-atopic volunteers.
The roles of GM-CSF and IL-8 in allergic inflammation are well established. For example, GM-CSF is a potent survival factor for eosinophils and IL-8 is an important chemoattractant for neutrophils. The inhibitory effect of GCS on the production of GM-CSF, IL-8 and MCP-1 by eosinophils after different stimuli has been demonstrated [4]. However, its effect on eosinophils activated by TNF-α has not been fully investigated as yet. In our system we used dexamethasone to study the effect of GCS on TNF-α-induced eosinophil activation. It is important to note that different GCS have similar effects on inflammatory cells. Several reports have demonstrated that the inhibition of dexamethasone on eosinophil survival and activation parallels the one of inhaled GCS in vitro. For example, budesonide reduced the number of peripheral blood eosinophils by suppressing both their progenitors in the blood and colony-forming unit production in the bone marrow [6,7]. It is also known that inhalation of high doses of fluticasone reduced the number of blood eosinophils by increasing their apoptosis in vivo [8].
Although dexamethasone has been shown to induce eosinophil apoptosis, we have found that it did not decrease eosinophil survival after 18 h of treatment. Therefore, its effects on cytokine release observed in our system can not be attributed to the eosinophil death. Dexamethasone, as other GCS, inhibits cytokines release by eosinophils by interference with transcription factors such as NF-kB and AP-1 [9]. Since TNF-α is a potent inducer of NF-kB in eosinophils [10] we can hypothesize that dexamethasone inhibits GM-CSF and IL-8 release in TNF-α activated eosinophils by blocking NF-kB (genomic mechanism). In our study dexamethasone was unable to inhibit TNF-α-induced ECP release and their adhesion to immobilised VCAM-1, ICAM-1 and IgG. These data are in accordance with previous works in which dexamethasone did not affect the C5a- and IL-5 enhanced immunoglobulin-induced eosinophil release of EDN [11].
Recent evidence supports a direct and extremely rapid inhibitory effect of GCS on some activated inflammatory cells (i.e. basophils) via a non-genomic effect that results from the interaction of the GCS with biological membranes [12]. However, we have not observed any effect of dexamethasone on eosinophils degranulation and adhesion (rapid events).
In conclusion, from our data we can speculate that GCS exert beneficial effects in allergic inflammation also by selectively inhibiting TNF-α-induced eosinophils release of GM-CSF and IL-8, but not their survival, degranulation and adhesion.
List of abbreviations used
TNF-α: tumor necrosis factor-α
GM-CSF: granulocyte-macrophage colony-stimulating factor
GCS: glucocorticosteroids
FCS: foetal calf serum
ECP: eosinophil cationic protein
VCAM-1: vascular cell adhesion molecule-1
ICAM-1: intercellular adhesion molecule-1
EDN: eosinophil-derived neurotoxin
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
IU performed 80% of the experiments and organized the graphs. IP drafted the manuscript and organized the figures. VT performed 10% of the experiments. SJS contributed to the experiments and to draft the manuscript. DF performed the experiments of adherence. KPR participated in the design and coordination of the study. FLS performed 10% of the experiments, contributed to design and coordinate the study and to draft the manuscript.
Acknowledgements
F. Levi-Schaffer is affiliated with the David R. Bloom Center for Pharmacy at the HUJI. S.J. Smith was a recipient of the Isaac and Myrna Kaye Travelling Fellowship (UK-IL).
==== Refs
Levi-Schaffer F Temkin V Malamud V Feld S Zilberman Y Mast cells enhance eosinophil survival in vitro: role of TNF-α and granulocyte-macrophage colony-stimulating factor J Immunol 1998 160 5554 5562 9605160
Lamas AM Leon OG Scleimer RP Glucocorticosteroids inhibit eosinophil responses to granulocyte-macrophage colony-stimulating factor J Immunol 1991 147 254 259 2051022
Lim LH Flower RJ Pereti M Das AM Glucocorticoid receptor activation reduces CD11b and CD49d levels on murine eosinophils: characterization and functional relevance Am J Respir Cell Mol Biol 2000 22 693 701 10837366
Miyamasu M Misaki Y Izumi S Takaishi T Morita Y Nakamura Y Glucocorticoids inhibit chemokines generation by human eosinophils J Allergy Clin Immunol 1998 101 75 83 9449504
Fattah D Page KR Bezbararuah S Priest RC Horgan CM Solari R A rapid activation assay for human eosinophils based on adhesion to immobilised ICAM-1, VCAM-1 and IgG Cytokine 1996 8 248 259 8833040 10.1006/cyto.1996.0034
Woolley MJ Denburg JA Ellis R Dahlback M O'Byrne PM Allergen-induced changes in bone marrow progenitors and airway responsiveness in dogs and the effect of inhaled budesonide on these parameters Am J Respir Cell Mol Biol 1994 11 600 606 7946389
Inman MD Denburg JA Ellis R Dahlback M O'Byrne PM The effect of treatment with budesonide or PGE2 in vitro on allergen-induced increases in canine bone marrow progenitors Am J Respir Cell Mol Biol 1997 17 634 641 9374115
Meagher LC Cousin JM Seckl JR Haslett C Opposing effects of glucocorticosteroids on the rate of apoptosis in neutrophilic and eosinophilic granulocytes J Immunol 1996 156 4422 4428 8666816
Adcock IM Glucocorticosteroid-regulated transcription factors Pulm Pharmacol Ther 2001 14 211 219 11448148 10.1006/pupt.2001.0283
Fujihara S Ward C Dransfield I Hay RT Uings IJ Hayes B Farrow SN Haslett C Rossi AG Inhibition of nuclear factor-kappaB activation un-masks the ability of TNF-alpha to induce human eosinophil apoptosis Eur J Immunol 2002 32 457 466 11813164 10.1002/1521-4141(200202)32:2<457::AID-IMMU457>3.0.CO;2-1
Kita H Abu-Ghazaleh R Sanderson CJ Gleich GJ Effect of steroids on immunoglobulin-induced eosiniophil degranulation J Allergy Clin Immunol 1991 87 70 77 1991924
Falkenstein E Tillmann HC Christ M Feuring M Wehling M Multiple actions of steroid hormones: a focus on rapid, non-genomic effects Pharmacol Rev 2000 52 513 556 11121509
| 15857511 | PMC1131914 | CC BY | 2021-01-04 16:36:25 | no | Clin Mol Allergy. 2005 Apr 27; 3:5 | utf-8 | Clin Mol Allergy | 2,005 | 10.1186/1476-7961-3-5 | oa_comm |
==== Front
Health Qual Life OutcomesHealth and Quality of Life Outcomes1477-7525BioMed Central London 1477-7525-3-281585751310.1186/1477-7525-3-28ResearchThe relations between symptoms, somatic and psychiatric conditions, life satisfaction and perceived health. A primary care based study Al-Windi Ahmad [email protected] Family Medicine Stockholm, Karolinska Institute, Alfred Nobels allé 12, SE-141 83 Huddinge, Sweden2005 27 4 2005 3 28 28 13 2 2005 27 4 2005 Copyright © 2005 Al-Windi; licensee BioMed Central Ltd.2005Al-Windi; 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 spite of the fact that self-rated health is such an important factor, little is known about the aetiological background to poor perceived health and also less is known about the impact of life satisfaction on health in a primary care practice population. The aim of this study was to evaluate the effect of socio-demographic characteristics, lifestyle factors, symptoms, somatic and psychiatric conditions as well as health status measures and life satisfaction on perceived health in a multi-ethnic Swedish health practice population.
Methods
Four-hundred and seventy adult patients, who visited the Jordbro Health Care Centre District (JHC), Haninge Municipality, participated in this study. A general questionnaire with questions about socio-demographic characteristics, lifestyle, health status and chronic disease were used. In addition to that, the Primary Care Evaluation of Mental Disorders (PRIME-MD) was used. Furthermore, physical examinations were conducted. Unconditional logistic regression in successive models was used, adjusted for socio-demographic variables and other confounders.
Results
Life satisfaction is the strongest predictor of poor perceived health in addition to country of birth, number of symptoms and depression. Being born in Sweden or other Nordic countries were related to lower OR as compared to those born outside Europe. The OR for non-depressed vs. depressed was 0.29 (0.17–0.48) and for non-symptomatic vs. symptomatic (1–3 symptoms) 0.25 (0.46–0.48). The OR and 95% CI for low satisfaction with life was 15.40 (5.28–44.97) in comparison to those who are satisfied with life.
Conclusion
Country of birth, depression, number of symptoms and life satisfaction are factors related significantly and independently to perceived health. Life satisfaction is the strongest predictor of perceived poor health.
depressionperceived healthPrimary Care Evaluation of Mental Disorders (PRIME-MD)country of birthhealth care practicequestionnaire.
==== Body
Background
During the past two decades, interest in subjects' perceived health has become one of the important research fields in epidemiology and research concerning health services [1-4]. At the beginning of the 1980s, Kaplan and others reported an association between subjects' perceived health and mortality [5-7]. Several other international studies later confirmed these findings. Follow-up studies have shown similar results – i.e., poor self-rated health as a strong predictor of mortality [1,2,8]. Self-assessment of health status appears to be a good measure of current physical health, morbidity, disability and a predictor of health care utilisation [6,9-17].
Numerous reports have shown a strong association between various indicators of individual socio-economic positions [18-24]. A strong correlation has also been found between self-reported health status and both socio-economic status and increased risk of mortality for all ethnic groups [25]. In the United Kingdom a number of studies have found higher morbidity and mortality rates for most British ethnic minorities than in the white population [23,26-28]. A study from the United States observed that socio-economic status was a principal determinant of racial/ethnic disparities in health, but several other factors, for example, medical care, migration, stress and resources, also play a role [27]. However, the effect of ethnicity has been compared to social class [29].
It is evident that subjective health assessment is a valid indicator of health status in middle-aged populations, and that it can be used in cohort studies and monitoring of a population's health [3,25]. The use of socio-economic status has been recommended for screening of majority groups, but not for minority ones [30]. In spite of the fact that self-rated health is such an important factor, little is known about the aetiological background to poor perceived health and also less is known about the impact of life satisfaction on health in a primary care practice population. Palmore and Kivett reported in 1977 that life satisfaction at the end of a 4-year period was significantly related to initial levels of self-rated health among subjects aged 46–70 years [31]. However, some other studies on the elderly show relations between perceived health and life satisfaction [32-34]. For example, Wang et al. found that life satisfaction was related to mental health both in females and males [32]. Benyamini et al. reported from a study of an old population (mean age 73) that both self-rated oral health and self-rated health independently explained a significant amount in concurrent rates of self-esteem and life satisfaction [33]. Ho et al. found that satisfaction was related to higher social class, education, good perceived health and other factors [34].
The present investigation is part of a comprehensive programme entitled "Improving Health Care in Jordbro (IHCJ)" to assess the influence of socio-demographic characteristics, including country of birth, as well as morbidity on health care and drug utilisation in patients resident in Jordbro, a small multi-ethnic sub-community in Stockholm, Sweden. In this study we explore the relations between socio-demographic characteristics, lifestyle factors, somatic and psychiatric symptoms as well as health status measures and perceived health.
The aim of this study was to evaluate the effect of socio-demographic characteristics, lifestyle factors, symptoms, somatic and psychiatric conditions as well as health status measures, and life satisfaction on perceived health in a multi-ethnic Swedish health practice population. The Committee on Research Ethics at the School of Medicine, Karolinska Institute, approved the study.
Methods
Subjects and setting
A full description of the methodology is provided elsewhere [35]. The patient sample was recruited between October 2002 and April 2003 from adult patients (≥ 16 years old) presenting for routine visits at the Jordbro Health Care Centre (JHC). In total 470 adult patients who visited the JHC during the period participated in this study.
The JHC has a catchment area of 9 500 patients and approximately 9 000 patients are registered at the health centre. At the time of the study the proportion of patients >15 years old was 77%, i.e. 7 296 patients. The study population consists of patients who were registered consequently during the first four weeks of the study period and only six patients provided insufficient data, i. e. only six patients did not fulfil the survey completely. All adult patients (≥ 16 years) who visited the health care centre were included but not the emergency cases.
The study consisted of three parts: (a) a general questionnaire [35,36], (b) patient questionnaire (PQ) followed by a clinical interview [37] and (c) a medical examination.
a) The general questionnaire dealing with questions on socio-demographic characteristics, lifestyle, health status and medicine use was handed out to the patients. The socio-demographic variables included, in addition to age and gender, family situation (e.g. whether the patient lived alone, with another adult or with children) and country of birth. In addition to that, the questionnaire included questions on 16 common somatic diseases and one on whether the respondents perceived themselves to be healthy or not. They were also asked to indicate the degree of life satisfaction, perceived health and health status. Patients were asked to state whether they had any chronic disease.
b) The Primary Care Evaluation of Mental Disorders (PRIME-MD) questionnaire was used [37]. It is a two-stage process for diagnosing mental disorders by primary care physicians consisting of a patient questionnaire (PQ) as an initial screening followed by a structured clinical interview with a psychologist/physician, consisting of five diagnosing modules to further evaluate those groups of disorders whose presence had been indicated by positive answers to the initial questionnaire screening [37]. Psychiatric diagnoses were based on those listed in the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV) [38]. A more detailed description of the PRIME-MD can be found in Loerch et al. [37].
c) Medical examination: this consisted of weight, height, and laboratory analyses including fasting blood glucose, serum cholesterol and serum triglycerides, flow volume spirometry and electrocardiography.
Outcome variables
Subjects were asked to indicate their present degree of perceived health on a seven-point scale, ranging from score 1 "very bad" to score 7 "excellent, could not be better". The variables were dichotomised to "Good scores" score 5–7 and "Bad or poor scores (PPH)" score 1–4.
Explanatory variables
The general questionnaire
The dichotomised or ordinal forms for these variables were used in nominal logistic regression – i.e., age was grouped into: 16–44, 45–64 and 65+. Gender: male and female. Working: yes and no. Country of birth: Sweden, other Nordic countries, Other European countries and rest of the world. Living conditions: the variables were: Living alone, Living with another adult and Living with children < 18 years. The presence of diagnosis was regarded as "yes", 1 and the "absence" as no, 0. Reported health status: Healthy: yes, 1; no, 0 and in between or unsure, 2. Life satisfaction: was defined on a seven-point scale, ranging from score 1 "very bad" to score 7 "excellent, could not be better". The variables were dichotomised to "Good scores" score 5–7, "Fair 'scores 3–4 and "Bad or poor scores" score 1–2.
Psychiatric evaluation
The patient questionnaire (PQ) consisted of 28 yes/no questions about the presence of symptoms during the previous month, of which 15 were about somatic symptoms and the other about psychiatric symptoms and alcohol. Psychiatric diagnoses: mood disorders, anxiety, compulsive disorder, social phobia, probable alcohol abuse/dependence and eating disorders. The presence of diagnosis was regarded as "yes", 1, and the "absent" as no, 0.
Medical Examination
BMI: Weight in kilograms/(height in metres)2; blood pressure: systolic and diastolic; heart rate (pulse) during 1 minute; fasting blood glucose, serum cholesterol and triglycerides. Spirometry: mean values were calculated for vital capacity (VC), forced expiratory volume in 1 second (FEV1), Forced vital capacity (FVC) and peak expiratory flow (PEF). Electrocardiography: The results were judged and grouped in four categories: normal, not surely pathological, suspect pathological and pathological. Common somatic disease consisted of 16 common conditions and one general question on chronic disease. The presence of condition was regarded as "yes", 1 and the absent as "no", 0.
Statistical methods
The data were analysed with the SAS and JMP software packages [39,40]. Standard methods were used to obtain summary statistics, such as means, prevalence and other measures. Chi-square test or Fisher's exact test were used to calculate the p-values. Associations between perceived health and the continuous independent variables age, life satisfaction, BMI, blood pressure, heart rate, blood glucose, serum triglycerides and cholesterol were estimated as product moment correlations according to Spearman. All tests were two-tailed. Probability (p-) values less than 0.05 was regarded as statistically significant.
We analysed the relationships between perceived health and the explanatory variables, using the unconditional logistic regression in a successive model, adjusted for socio-demographic variables and other confounders. The interaction terms for chronic disease*depression, depression*symptoms and chronic disease*depression*symptoms were not significant.
The results are shown as odds ratios (OR) with 95% confidence intervals (CIs). The fit of the models was judged by the Hosmer-Lemeshow goodness-of-fit test. The models were considered acceptable if p > 0.05 and all models met this criteria. The results are shown as odds ratios (OR) with 95% confidence intervals (CIs).
Results
Socio-demographic characteristics, self-health report, health status measures, life-satisfaction and perceived health
Table 1 shows that 46.4% of respondents reported perceived poor health. All variables except living with or without children were significantly related to perceived health. A higher proportion of subjects aged 45–64 years reported poor health (PPH) as compared to younger subjects aged 16–44 years. The proportion of subjects reporting PPH is higher among females, subjects not working, subjects born outside Sweden, living alone and not living with another adult compared to males, working subjects, subjects born in Sweden, not living alone and living with another adult. A significantly higher proportion of subjects who report that they are not healthy or less satisfied with life have PPH in comparison with healthy and satisfied subjects. Life satisfaction was strongly correlated to perceived health (r = 0.66; p <.0001).
Table 1 Population distribution in per cent (%) by perceived health and sociodemographic characteristics, reported health status and life satisfaction.
Perceived poor health (scores 1–4)
Variable N % P-value
Age <0.01
1–44 120 40.8
45–64 212 55.2
65- 135 38.5
Gender <0.001
Men 169 36.1
Women 298 52.8
Working <0.005
Yes 205 41.2
No 252 52.8
Country of birth <0.01
Sweden 278 40.3
Nordic 59 49.1
Europe 58 67.2
Other 72 52.8
Living conditions <0.05
Alone Yes 128 57.8
No 336 42.6
With other adult Yes 318 41.5 <0.005
No 146 58.2
With children <18 Yes 123 42.3 NS
No 341 48.1
Healthy <0.0000
Yes 247 25.1
Fair 26 69.2
No 190 72.1
Life satisfaction <0.0001
Low (score1–2) 45 88.9
Fair (score 3–4) 128 78.9
High (score 5–7) 292 25.7
Population 470 46.4
* N not equal to 470 due to missing some values.
Apart from BMI and smoking status, no other variable was significantly related to perceived health, table 2. Subjects who report PPH have higher BMI and smoke more than those with perceived good health. BMI is negatively related to PPH. The correlation for BMI and PPH is (r = -0.10; p < .05).
Table 2 Population distribution in per cent (%) by perceived health and health status measures. Values are means unless otherwise indicated. Poor perceived health (N = 218), good (N = 250).
Perceived health
Variables Number Poor (scores 1–4) Good (scores 5–7) P-value
Body Mass Index (BMI)* 28.3 27.3 <0.05
Smoking Yes 133 57.1 42.9 <0.005
No 335 42.3 57.7
Blood pressure
Systolic BP 134.4 135.4 NS
Diastolic BP 80.5 80.0 NS
Heart rate/minute 70.7 68.8 NS
Blood glucose 5.99 5.78 NS
Serum triglycyrids 1.54 1.50 NS
Serum cholesterol 5.46 5.28 NS
Spirometry
Vital capacity (VC) 96.5 93.8 NS
Forced expiratory volume (FEV1) 92.6 93.9 NS
Forced vital capacity (FVC) 91.8 91.3 NS
PEF 89.3 91.8 NS
Electrocardiography** NS
Normal 72 52.8 46.2
Not sure pathological 49 36.7 63.3
Suspect pathologic 226 43.8 56.2
Pathological 111 52.3 47.8
* BMI = Weigh/length (m2)
**Per cent
Symptoms and perceived health
Having any of the 13 of 15 symptoms listed in the PQ was significantly related to PPH, i.e. those who report any of these symptoms more often have PPH than do non-symptomatic subjects. For some of these symptoms the risk is almost doubled or tripled, Table 3. For example, 87% of those reporting fainting spells have PPH compared to 44.4% among non-symptomatic subjects. The figures for feeling tired or having low energy and trouble sleeping were 62.6 and 62.9% as compared to 26.2 and 33.6% respectively.
Table 3 The number of subjects with/without symptoms and the percentage (%) of these reporting perceived poor health (PPH).
Number* PPH % P-value
Symptoms Yes No Yes No
Stomach pain 130 339 60.3 41.3 <0.0005
Back pain 218 251 59.0 35.5 <0.0001
Pain in arms, legs & joints 288 181 52.4 37.0 <0.005
Menstrual cramps 46 210 50.0 46.2 NS
Headaches 176 291 57.0 40.2 <0.001
Chest pain 100 368 66.7 41.0 <0.0001
Dizziness 133 333 64.9 39.3 <0.0001
Fainting spells 23 446 87.0 44.4 <0.0001
Palpitations 88 381 63.6 42.5 <0.001
Shortness of breath 86 383 65.1 42.3 <0.0005
Pain or problems during sexual intercourse 27 238 57.7 45.4 NS
Constipation, loose bowels, or diarrhoea 117 352 62.4 41.2 <0.0001
Nausea, gas, or indigestion 145 324 58.6 41.1 <0.001
Feeling tired or having low energy 263 206 62.6 26.2 <0.0001
Trouble sleeping 202 265 62.9 33.6 <0.0001
* N not equal to 470 due to missing some values.
Chronic disease or conditions and perceived health
Subjects who report having any chronic disease or condition report to a higher extent having PPH, table 4. In general about 50% of subjects with chronic disease report PPH as compared to 26.3% among non-diseased subjects. Having any of nine of these diseases or conditions was related to PPH, heart failure, asthma, neurological disease, musculoskeletal and joint disorders, pain syndrome, psychiatric disorders, gastrointestinal and urinary tract troubles. For example, about 80% of subjects with a psychiatric diagnosis report PPH and for heart failure the figure is 73.3%.
Table 4 The number of subjects with/without a disease or a condition and the percentage (%) of these reporting perceived poor health (PPH).
Number* PPH %
Disease or condition Yes No Yes No P-value
Chronic disease 394 76 50.3 28.5 <0.005
Blood pressure 84 354 49.6 45.8 NS
Angina 38 429 52.6 46.1 NS
Heart failure 15 452 73.3 45.8 <0.01
Diabetes 44 422 54.5 45.7 NS
Asthma 59 408 67.8 43.6 <0.005
Chronic obstructive disease 13 454 38.5 46.9 NS
Neurological disease 25 441 68.0 45.6 <0.0005
Musculoskeletal disease 89 378 68.5 41.5 <0.0001
Joint disease 255 212 55.3 36.3 <0.0005
Pain syndrome 112 355 65.1 39.4 <0.0001
Cancer 18 448 44.4 46.9 NS
Psychiatric disorder 40 427 80.0 43.6 <0.0001
Eye disease 65 402 53.8 45.5 NS
Ear disease 39 427 46.1 46.8 NS
Gastrointestinal disorders 120 347 59.2 42.4 <0.0005
Urinary tract disease 59 406 55.9 45.1 <0.05
* N not equal to 470 due to missing some values.
From table 5 is obvious that depression, anxiety and compulsive disorders are related to PPH. About 75% of subjects with one of these three disorders report PPH.
Table 5 The number of subjects with/without a psychiatric diagnoses and the percentage (%) of these reporting perceived poor health (PPH). Perceived poor health (N = 218), good (N = 250).
Number % PPH
Diagnose Yes No Yes No P-value
Depressive disorders 216 249 73.8 34.6 <0.0001
Anxiety 216 380 76.5 39.7 <0.0001
Compulsive disorders 23 441 76.5 44.4 0.0001
Social phobia 10 455 60.0 46.2 NS
Probable alcohol abuse/dependent 32 433 46.9 46.4 NS
Eating disorders 11 454 54.5 46.3 NS
Logistic regression analyses
Table 6 shows odds ratios (OR) with 95% confidence intervals (95% CI) for having poor health by age, gender, living and working status, country of birth, smoking, chronic disease, depression, number of symptoms and life satisfaction. The ORs are adjusted for all confounders. We analysed the relationships between perceived health and the explanatory variables, using the unconditional logistic regression in successive models, adjusted for age and gender in the first model. This shows that subjects aged 45–64 years have higher OR than older subjects, i.e. aged 65 or older. The OR was 1.62. The figure for males was 0.52 compared to females.
Table 6 Odds rations (OR) with 95% confidence intervals (95% CI) for having poor health by age, gender, living and working status, country of birth, smoking, chronic disease, depression, number of symptoms and life satisfaction. The ORs are adjusted for all confounders.
Perceived poor health (scores 1–4)
Model 1 Model 2 Model 3 Model 4
OR 95% CI OR 95% CI OR 95% CI OR 95% CI
Age (65+ years = reference) 16–44 0.96 0.59–1.57 1.51 0.83–2.76 0.65 0.31–1.33 0.65 0.29–1.45
45–64 1.62 1.05–2.50 2.47 1.45–4.21 1.73 0.95–3.18 1.44 0.74–2.81
Male (female= ref) 0.52 0.35–0.76 0.56 0,37–0.84 0.86 0.54–1.37 0.87 0.52–1.46
Living alone (no = ref) 0.52 0.33–0.83 0.67 0.40–1.11 0.74 0.42–1.30
Working (yes =ref) 2.10 1.35–3.26 1.42 0.86–2.34 1.21 0.69–2.11
Country of birth (other = ref) Sweden 0.50 0.28–0.88 0.52 0.26–1.02 0.41 0.20–0.84
Nordic 0.57 0.27–1.21 0.44 0.19–1.02 0.26 0.10–0.66
Europe 1.19 0.55–2.57 0.96 0.40–2.31 0.68 0.26–1.80
Smoking (yes = ref) 0.81 0.50–1.32 0.97 0.56–1.66
Chronic disease (yes = ref) 0.68 0.36–1.27 0.97 0.56–1.27
Depression (yes = ref) 0.29 0.17–0.48 0.55 0.31–0.98
Number of symptoms (>6 ref) 1–3 symptoms 0.25 0.13–0.48 0.30 0.15–0.60
4–6 symptoms 0.85 0.46–1.58 0.98 0.50–1.92
Life satisfaction* (high = ref) Low 15.40 5.28–44.97
Fair 7.02 3.98–12.38
*Low (scores 1–2), fair (scores 3–4) and high (scores 5–7)
** Interaction tests for chronic disease*depression, depression*symptoms and chronic disease*depression*symptoms were not significant.
In the second model, when living conditions and country of birth were added to the analysis, the OR for subjects aged 45–64 years and for males remained significant. The ORs for subjects not living alone, working subjects and those born in Sweden were lower than for those living alone, not working and born outside Europe.
In the third model, with the addition of smoking status, chronic disease, depression and number of symptoms, the pattern changed. Neither age nor gender was significant any longer. However, depression and number of symptom were significant. Non-depressed had OR = 0.29 as compared to depressed. This means that depressed people have 71% higher risk than non-depressed of having PPH. The OR for having 1–3 symptoms and PPH was 0.25 as compared to those with more than 6 symptoms. The OR for those with 4–6 symptoms was 0.85, as compared to those with more than 6 symptoms, but this is not significant. The linear trend is interesting. However, the OR for subjects born Sweden was still lower as than for those born outside Europe but the 95% CI overlapped 1 (1.02), which was not significant.
In the final model (model 4) when all confounders were included, i.e. life satisfaction in addition to the previous variables, being born in Sweden or other Nordic countries was related to lower OR as compared to those born outside Europe. The OR for those born in Sweden was 0.41 and for those born in other Nordic countries 0.26. The OR for non-depressed and those with 1–3 symptoms was somewhat higher but still significant.
The OR for low satisfaction with life was as high as 15.40 in comparison to those who are satisfied with life. However, the 95% CI was very wide, 5.28–44.97. The OR for those who report fair life satisfaction is 7.02 and the 95%CI is 3.98–13.38.
Discussion
This study has shown that various socio-demographic characteristics and country of birth affect the subject's perceived health. A substantially high percentage reported that they had poor perceived health, with the lowest percentage in those born in Sweden. These figures were higher than the national Swedish average (5–6%) and in other international reports [41,42]. However, Hjern et al. also found a higher prevalence of poor perceived health in subjects born outside Sweden [43]. Williams, in a study of women from the United States, found that socio-economic status was an important determinant of racial/ethnic disparities in health, but several other factors including, for example, medical care, migration, acculturation, stress and resources, also play a role [27]. Not surprisingly, perceived health was influenced by socio-demographic characteristics in this study, but it is noteworthy that another individual factor, "life satisfaction", showed a profile stronger than that of socio-demographic characteristics even after adjustment was made for age, gender, education, occupation, country of birth, chronic disease or condition and symptoms. This factor probably plays an important role in a subject's health and indicates the overall life satisfaction including health. The question about "life satisfaction" can also be used to assess health. Prospective studies are needed to confirm this. Indeed, this variable can probably be used together with perceived health, but needs to be investigated further. Many studies have shown the effect of age, gender, education, occupation and country of birth on health, which accord with our results [44,45,5].
An interesting finding in this study is that patients aged 45–64 years have poorer health than older subjects. We believe that in this multi-ethnic population this age group is more vulnerable and sicker than the older subjects. It is possible that specific factors related to the neighbourhood in this study have an impact on health. Future studies should elucidate this issue. This could have important health policy implications. However, in the logistic model, the impact of age was not any longer significant when the symptom variable was taken into account.
As is the case with all questionnaire surveys, there was the possibility that patients exaggerated or underreported a condition, partly due to difficulties in remembering (recall bias). In addition to the potential problem of recall bias, there is the possibility of selection bias. The participants in this investigation were voluntary and time-limited, and we could only include those who showed interest first. It is possible that some selection bias occurred as a result of this consecutive procedure. It is possible, for instance, that our study included only the healthier patients in the Jordbro Health District because the sickest patients were not able to come to the health centre.
The questionnaire used in the present study (PRIME-MD) has been validated previously and shown to have good accuracy for psychiatric disorders [51-53]. Also, different parts of the questionnaire used here were previously validated in other investigation [35,36,49,51]. We also tested the questionnaire as a whole in advance in a pilot study and judged it to be satisfactory.
Although this study is not prospective, causal relations cannot be drawn from this investigation. The most of the literature on this issue are cross-sectional studies and not suited for statement or implications. Further prospective research is needed to clarify the direction of association. The new finding in this study is the strong association between life satisfaction and perceived health apart from country of birth, symptoms and depression, and also the fact that perceived health has a stronger correlation with psychiatric than somatic conditions. The strength of this study is that this represents a primary care patient population which makes it unique. Furthermore, the fact that we have controlled for a large number of confounders makes it more unique.
In conclusion, country of birth, depression, number of symptoms and life satisfaction are factors related significantly and independently to perceived health. Life satisfaction was the strongest predictor of poor health.
Abbreviations
OR = Odds ratio
95% CI = 95% Confidence Interval
Prime-MD = Primary Care Evaluation of Mental Disorders
PQ = Patient questionnaire
JHC = Jordbro Health Care Centre
VC = Vital capacity
FEV1 = Forced vital capacity during one second
PEF = Peak expiratory flow
PPH = poor perceived health
r = Spearman's correlation coefficient
BMI = body mass index
Acknowledgements
This study was supported by grants from Stockholm County Council (Dagmar & ALF Fund) and Haninge Community Council (Economic Target to Large Cities).
==== Refs
Kaplan GA Camacho T Perceived health and mortality: a nine-year follow-up of the Human Population Laboratory cohort Am J Epidemiol 1983 117 292 304 6829557
Sundquist J Johansson SE Self-reported poor health and low educational level predictors for mortality: a population-based follow-up study of 39,156 people in Sweden J Epidemiol Community Health 1997 51 35 40 9135786
Miilunpalo S Vuori I Oja P Pasanen M Urponen H Self-rated health status as a health measure: the predictive value of self-reported health status on the use of physician services and on mortality in the working-age population J Clin Epidemiol 1997 50 517 528 9180644 10.1016/S0895-4356(97)00045-0
Idler EL Angel RJ Self-rated health and mortality in the NHANES-1 epidemiologic follow-up study Am J Public Health 1990 80 446 452 2316767
Kaplan G Barell V Lusky A Subjective state of health and survival in elderly adults J Gerontol 1988 43 S114 120 3385152
Idler EL Benyamini Y Self-rated health and mortality: a review of twenty-seven community studies J Health Soc Behav 1997 38 21 37 9097506
Benyamini Y Idler EL Community studies reporting associations between self-rated health and mortality Research in Aging 1999 21 393 401
Heistaro S Jousilahti P Lahelma E Vartiainen E Puska P Self-rated health and mortality: a long-term prospective study in eastern Finland J Epidemiol Community Health 2001 55 227 232 11238576 10.1136/jech.55.4.227
Angel R Gronfein W The use of subjective information in statistical models ASR 1988 53 464 473
Wannamethee G Shaper AG Self-assessment of health status and mortality in middle-aged British men Int J Epidemiol 1991 20 239 245 2066228
Ferraro KF Farmer MM Wybraniec JA Health trajectories: Long-term dynamics among black and white adults J Health and Soc Behav 1997 38 38 54 9097507
Al-Windi A Elmfeldt D Svärdsudd K The influence of sociodemographic characteristics on health care utilisation in a Swedish municipality Ups J Med Sci 2004 109 33 42 15124951
Idler EL Kasl S Self-ratings of health: Do they also predict change in functional ability? J Gerontol B Psychol Sci Soc Sci 1995 50B S344 353 7583813
Al-Windi A Elmfeldt D Svärdsudd K The influence of perceived well-being and reported symptoms on health care utilization: A population-based study J Clinical Epidemiol 2002 55 60 66 11781123 10.1016/S0895-4356(01)00423-1
Connelly JE Smith GR Philbrick JT Kaiser DL Healthy patients who perceive poor health and their use of primary care services J Gen Intern Med 1991 6 47 51 1999746
Kjoller M Rasmussen NK Keiding LM Self-reported health and morbidity among adult Danes 1987–1994 Ugeskr Laeger 1999 161 2948 2954 10354781
Borgquist L Hansson L Nettelbladt P Nordstrom G Lindelow G Perceived health and high consumers of care. a study of mental health problems in a Swedish primary health care district Psychol Med 1993 23 763 770 8234582
Sihvonen AP Kunst AE Lahelma E Valkonen T Mackenbach JP Socioeconomic inequalities in health expectancy in Finland and Norway in the late 1980s Soc Sci Med 1998 47 303 315 9681900 10.1016/S0277-9536(98)00020-3
Hallqvist J Lundberg M Diderichsen F Ahlbom A Socioeconomic differences in risk of myocardial infarction 1971–1994 in Sweden: Time trends, relative risks and population-attributable risk Int J Epidemiol 1998 27 410 415 9698128 10.1093/ije/27.3.410
Cavelaars AE Kunst AE Geurts JJ Helmert U Lundberg O Mielck A Matheson J Mizrahi A Mizrahi A Rasmussen N Spuhler T Mackenbach JP Morbidity differences by occupational class among men in seven European countries: an application of the Erikson-Goldthorpe social class scheme Int J Epidemiol 1998 27 222 230 9602402 10.1093/ije/27.2.222
Devlin N Hansen P Herbinson P Variations in self-reported health status: results from a New Zealand survey N Z Med J 2000 113 517 520 11198514
Morris JK Cook DG Walker M Shaper AG Non-consulters and high consulters in general practice: cardio-respiratory health and risk factors J Public Health Med 1992 14 131 137 1515196
Marmot MG Adelstein AM Bulusu L Immigrant mortality in England and Wales 1970–78. Causes of death by country of birth. Studies on Medical and Population Subjects No. 47 1984 London: OPCS; Her Majesty's Stationery Office
Nazroo JY Bartley MB, Blane D, Davey Smith G Genetic, cultural or socio-economic vulnerability? Explaining ethnic inequalities in health The sociology of health inequalities 1998 Oxford: Blackwell Publishers
McGee DL Liao Y Cao G Cooper RS Self-reported health and mortality in a multiethnic US cohort Am J Epidemiol 1999 149 41 46 9883792
Harding S Maxwell R Differences in the mortality of migrants. In F. Drever, M Whitehead, Health inequalities 1997 London Stationery Office
Williams DR Racial/ethnic variations in women's health: the social embeddedness of health Am J Public Health 2002 92 588 597 11919058
Nazroo JY The health of Britain's ethnic minorities 1997 London: Policy Studies Institute
Sundquist J Ethnicity, social class and health. A population-based study on the influence of social factors on self-reported illness in 223 Latin American refugees, 333 Finnish and 126 southern European labour migrants and 841 Swedish controls Soc Sci Med 1995 40 777 787 7747212 10.1016/0277-9536(94)00146-K
Kwok RK Yankaskas BC The use of census data for determining race and education as SES Indicators: A validation study AEP 2001 11 171 177 11293403
Palmore E Kivett V Change in life satisfaction: a longitudinal study of persons aged 46–70 J Gerontol 1977 32 311 316 850059
Wang CW Iwaya T Kumano H Suzukamo Y Tobimatsu Y Fukudo S Relationship of health status and social support to the life satisfaction of older adults Tohoku J Exp Med 2002 198 141 149 12597241 10.1620/tjem.198.141
Benyamini Y Leventhal H Leventhal EA Self-rated oral health as an independent predictor of self-rated general health, self-esteem and life satisfaction Soc Sci Med 2004 59 1109 1116 15186909 10.1016/j.socscimed.2003.12.021
Ho SC Woo J Lau J Chan SG Yuen YK Chan YK Chi I Life satisfaction and associated factors in older Hong Kong Chinese J Am Geriatr Soc 1995 43 252 255 7884112
Al-Windi A Depression in general practice Nord J Psychiatry
Tibblin G Peciva S Kullman S Svardsudd K "The Gothenburg quality of life instrument": An assessment of well-being and symptoms among men born 1913 and 1923. Methods and validity Scand J Prim Health Care 1990 1 33 38
Loerch B Szegedi A Kohnen R Benkert O The Primary Care Evaluation of Mental Disorders (PRIME-DM) German version: a comparison with CIDI J Psychiat Res 2000 34 211 220 10867116 10.1016/S0022-3956(00)00005-4
American Psychiatric Association. Diagnostic and statistical manual of mental disorders 1994 4 Author, Washington, DC
SAS: Introductory Guide; ed 3.1 Campus Drive, SAS Institute 31 ed Cary 1997
JMP-program: Statistical Discovery Software Introductory Guide, Version 4 SAS Institute Inc, Cary, NC, USA 2000
Statistics Livslängd, hälsa och sysselsättning (Length of life, health and occupation) Stockholm, Sweden, Statististka centralbyrån (Statistics Sweden) 2002
Markovic M Manderson L Kelaher M The health of immigrant women: Queensland women from the former Yugoslavia J Immigrant Health 2002 4 5 15 10.1023/A:1013003126561
Hjern A Haglund B Persson G Rosen M Is there equity in access to health services for ethnic minorities in Sweden? Eur J Public Health 2001 11 147 152 11420800 10.1093/eurpub/11.2.147
Eriksson I Undén A-L Elofsson S Self-rated health. Comparisons between three different measures. Results from a population study Int J Epidemiol 2001 30 326 33 11369738 10.1093/ije/30.2.326
Verbrugge LM Sex differences in health Prevention 1982 97 417 437
Gijsbers van Wijk CM Kolk AM Sex differences in perceived health Ned Tijdschr Geneeskd 1997 141 283 287 9148163
Joung IM van der Meer JB Mackenbach JP Marital status and health care utilization Int J Epidemiol 1995 24 569 575 7672898
Fylkenes K Determinants of health care utilisation: Visits and referrals Scand J Soc Med 1993 21 40 50 8469943
Al-Windi A Determinants of Health Care and Drug Utilisation: The Causes of Health Care Utilisation Study Uppsala, Uppsala University, PhD Thesis 2000
Bosworth HB Butterfield MI Stechuchak KM Bastian LA The relationship between self-rated health care service use among women veterans in a primary care clinic Women's Health Issues 2000 10 278 285 10980445 10.1016/S1049-3867(00)00056-6
Parker T May PA Maviglia MA Petrakis S Sunde S Gloyd SV PRIME-MD: its utility in detecting mental disorders in American Indians Intl J Psychiat Med 1997 27 107 128
Kobak KA Taylor LH Dottl SL Greist JH Jefferson JW Burroughs D Mantle JM Katzelnick DJ Norton R Henk HJ Serlin RC A computer-administered telephone interview to identify mental disorders JAMA 1997 278 905 910 9302242 10.1001/jama.278.11.905
| 15857513 | PMC1131915 | CC BY | 2021-01-04 16:38:13 | no | Health Qual Life Outcomes. 2005 Apr 27; 3:28 | utf-8 | Health Qual Life Outcomes | 2,005 | 10.1186/1477-7525-3-28 | oa_comm |
==== Front
Health Qual Life OutcomesHealth and Quality of Life Outcomes1477-7525BioMed Central London 1477-7525-3-281585751310.1186/1477-7525-3-28ResearchThe relations between symptoms, somatic and psychiatric conditions, life satisfaction and perceived health. A primary care based study Al-Windi Ahmad [email protected] Family Medicine Stockholm, Karolinska Institute, Alfred Nobels allé 12, SE-141 83 Huddinge, Sweden2005 27 4 2005 3 28 28 13 2 2005 27 4 2005 Copyright © 2005 Al-Windi; licensee BioMed Central Ltd.2005Al-Windi; 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 spite of the fact that self-rated health is such an important factor, little is known about the aetiological background to poor perceived health and also less is known about the impact of life satisfaction on health in a primary care practice population. The aim of this study was to evaluate the effect of socio-demographic characteristics, lifestyle factors, symptoms, somatic and psychiatric conditions as well as health status measures and life satisfaction on perceived health in a multi-ethnic Swedish health practice population.
Methods
Four-hundred and seventy adult patients, who visited the Jordbro Health Care Centre District (JHC), Haninge Municipality, participated in this study. A general questionnaire with questions about socio-demographic characteristics, lifestyle, health status and chronic disease were used. In addition to that, the Primary Care Evaluation of Mental Disorders (PRIME-MD) was used. Furthermore, physical examinations were conducted. Unconditional logistic regression in successive models was used, adjusted for socio-demographic variables and other confounders.
Results
Life satisfaction is the strongest predictor of poor perceived health in addition to country of birth, number of symptoms and depression. Being born in Sweden or other Nordic countries were related to lower OR as compared to those born outside Europe. The OR for non-depressed vs. depressed was 0.29 (0.17–0.48) and for non-symptomatic vs. symptomatic (1–3 symptoms) 0.25 (0.46–0.48). The OR and 95% CI for low satisfaction with life was 15.40 (5.28–44.97) in comparison to those who are satisfied with life.
Conclusion
Country of birth, depression, number of symptoms and life satisfaction are factors related significantly and independently to perceived health. Life satisfaction is the strongest predictor of perceived poor health.
depressionperceived healthPrimary Care Evaluation of Mental Disorders (PRIME-MD)country of birthhealth care practicequestionnaire.
==== Body
Background
During the past two decades, interest in subjects' perceived health has become one of the important research fields in epidemiology and research concerning health services [1-4]. At the beginning of the 1980s, Kaplan and others reported an association between subjects' perceived health and mortality [5-7]. Several other international studies later confirmed these findings. Follow-up studies have shown similar results – i.e., poor self-rated health as a strong predictor of mortality [1,2,8]. Self-assessment of health status appears to be a good measure of current physical health, morbidity, disability and a predictor of health care utilisation [6,9-17].
Numerous reports have shown a strong association between various indicators of individual socio-economic positions [18-24]. A strong correlation has also been found between self-reported health status and both socio-economic status and increased risk of mortality for all ethnic groups [25]. In the United Kingdom a number of studies have found higher morbidity and mortality rates for most British ethnic minorities than in the white population [23,26-28]. A study from the United States observed that socio-economic status was a principal determinant of racial/ethnic disparities in health, but several other factors, for example, medical care, migration, stress and resources, also play a role [27]. However, the effect of ethnicity has been compared to social class [29].
It is evident that subjective health assessment is a valid indicator of health status in middle-aged populations, and that it can be used in cohort studies and monitoring of a population's health [3,25]. The use of socio-economic status has been recommended for screening of majority groups, but not for minority ones [30]. In spite of the fact that self-rated health is such an important factor, little is known about the aetiological background to poor perceived health and also less is known about the impact of life satisfaction on health in a primary care practice population. Palmore and Kivett reported in 1977 that life satisfaction at the end of a 4-year period was significantly related to initial levels of self-rated health among subjects aged 46–70 years [31]. However, some other studies on the elderly show relations between perceived health and life satisfaction [32-34]. For example, Wang et al. found that life satisfaction was related to mental health both in females and males [32]. Benyamini et al. reported from a study of an old population (mean age 73) that both self-rated oral health and self-rated health independently explained a significant amount in concurrent rates of self-esteem and life satisfaction [33]. Ho et al. found that satisfaction was related to higher social class, education, good perceived health and other factors [34].
The present investigation is part of a comprehensive programme entitled "Improving Health Care in Jordbro (IHCJ)" to assess the influence of socio-demographic characteristics, including country of birth, as well as morbidity on health care and drug utilisation in patients resident in Jordbro, a small multi-ethnic sub-community in Stockholm, Sweden. In this study we explore the relations between socio-demographic characteristics, lifestyle factors, somatic and psychiatric symptoms as well as health status measures and perceived health.
The aim of this study was to evaluate the effect of socio-demographic characteristics, lifestyle factors, symptoms, somatic and psychiatric conditions as well as health status measures, and life satisfaction on perceived health in a multi-ethnic Swedish health practice population. The Committee on Research Ethics at the School of Medicine, Karolinska Institute, approved the study.
Methods
Subjects and setting
A full description of the methodology is provided elsewhere [35]. The patient sample was recruited between October 2002 and April 2003 from adult patients (≥ 16 years old) presenting for routine visits at the Jordbro Health Care Centre (JHC). In total 470 adult patients who visited the JHC during the period participated in this study.
The JHC has a catchment area of 9 500 patients and approximately 9 000 patients are registered at the health centre. At the time of the study the proportion of patients >15 years old was 77%, i.e. 7 296 patients. The study population consists of patients who were registered consequently during the first four weeks of the study period and only six patients provided insufficient data, i. e. only six patients did not fulfil the survey completely. All adult patients (≥ 16 years) who visited the health care centre were included but not the emergency cases.
The study consisted of three parts: (a) a general questionnaire [35,36], (b) patient questionnaire (PQ) followed by a clinical interview [37] and (c) a medical examination.
a) The general questionnaire dealing with questions on socio-demographic characteristics, lifestyle, health status and medicine use was handed out to the patients. The socio-demographic variables included, in addition to age and gender, family situation (e.g. whether the patient lived alone, with another adult or with children) and country of birth. In addition to that, the questionnaire included questions on 16 common somatic diseases and one on whether the respondents perceived themselves to be healthy or not. They were also asked to indicate the degree of life satisfaction, perceived health and health status. Patients were asked to state whether they had any chronic disease.
b) The Primary Care Evaluation of Mental Disorders (PRIME-MD) questionnaire was used [37]. It is a two-stage process for diagnosing mental disorders by primary care physicians consisting of a patient questionnaire (PQ) as an initial screening followed by a structured clinical interview with a psychologist/physician, consisting of five diagnosing modules to further evaluate those groups of disorders whose presence had been indicated by positive answers to the initial questionnaire screening [37]. Psychiatric diagnoses were based on those listed in the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV) [38]. A more detailed description of the PRIME-MD can be found in Loerch et al. [37].
c) Medical examination: this consisted of weight, height, and laboratory analyses including fasting blood glucose, serum cholesterol and serum triglycerides, flow volume spirometry and electrocardiography.
Outcome variables
Subjects were asked to indicate their present degree of perceived health on a seven-point scale, ranging from score 1 "very bad" to score 7 "excellent, could not be better". The variables were dichotomised to "Good scores" score 5–7 and "Bad or poor scores (PPH)" score 1–4.
Explanatory variables
The general questionnaire
The dichotomised or ordinal forms for these variables were used in nominal logistic regression – i.e., age was grouped into: 16–44, 45–64 and 65+. Gender: male and female. Working: yes and no. Country of birth: Sweden, other Nordic countries, Other European countries and rest of the world. Living conditions: the variables were: Living alone, Living with another adult and Living with children < 18 years. The presence of diagnosis was regarded as "yes", 1 and the "absence" as no, 0. Reported health status: Healthy: yes, 1; no, 0 and in between or unsure, 2. Life satisfaction: was defined on a seven-point scale, ranging from score 1 "very bad" to score 7 "excellent, could not be better". The variables were dichotomised to "Good scores" score 5–7, "Fair 'scores 3–4 and "Bad or poor scores" score 1–2.
Psychiatric evaluation
The patient questionnaire (PQ) consisted of 28 yes/no questions about the presence of symptoms during the previous month, of which 15 were about somatic symptoms and the other about psychiatric symptoms and alcohol. Psychiatric diagnoses: mood disorders, anxiety, compulsive disorder, social phobia, probable alcohol abuse/dependence and eating disorders. The presence of diagnosis was regarded as "yes", 1, and the "absent" as no, 0.
Medical Examination
BMI: Weight in kilograms/(height in metres)2; blood pressure: systolic and diastolic; heart rate (pulse) during 1 minute; fasting blood glucose, serum cholesterol and triglycerides. Spirometry: mean values were calculated for vital capacity (VC), forced expiratory volume in 1 second (FEV1), Forced vital capacity (FVC) and peak expiratory flow (PEF). Electrocardiography: The results were judged and grouped in four categories: normal, not surely pathological, suspect pathological and pathological. Common somatic disease consisted of 16 common conditions and one general question on chronic disease. The presence of condition was regarded as "yes", 1 and the absent as "no", 0.
Statistical methods
The data were analysed with the SAS and JMP software packages [39,40]. Standard methods were used to obtain summary statistics, such as means, prevalence and other measures. Chi-square test or Fisher's exact test were used to calculate the p-values. Associations between perceived health and the continuous independent variables age, life satisfaction, BMI, blood pressure, heart rate, blood glucose, serum triglycerides and cholesterol were estimated as product moment correlations according to Spearman. All tests were two-tailed. Probability (p-) values less than 0.05 was regarded as statistically significant.
We analysed the relationships between perceived health and the explanatory variables, using the unconditional logistic regression in a successive model, adjusted for socio-demographic variables and other confounders. The interaction terms for chronic disease*depression, depression*symptoms and chronic disease*depression*symptoms were not significant.
The results are shown as odds ratios (OR) with 95% confidence intervals (CIs). The fit of the models was judged by the Hosmer-Lemeshow goodness-of-fit test. The models were considered acceptable if p > 0.05 and all models met this criteria. The results are shown as odds ratios (OR) with 95% confidence intervals (CIs).
Results
Socio-demographic characteristics, self-health report, health status measures, life-satisfaction and perceived health
Table 1 shows that 46.4% of respondents reported perceived poor health. All variables except living with or without children were significantly related to perceived health. A higher proportion of subjects aged 45–64 years reported poor health (PPH) as compared to younger subjects aged 16–44 years. The proportion of subjects reporting PPH is higher among females, subjects not working, subjects born outside Sweden, living alone and not living with another adult compared to males, working subjects, subjects born in Sweden, not living alone and living with another adult. A significantly higher proportion of subjects who report that they are not healthy or less satisfied with life have PPH in comparison with healthy and satisfied subjects. Life satisfaction was strongly correlated to perceived health (r = 0.66; p <.0001).
Table 1 Population distribution in per cent (%) by perceived health and sociodemographic characteristics, reported health status and life satisfaction.
Perceived poor health (scores 1–4)
Variable N % P-value
Age <0.01
1–44 120 40.8
45–64 212 55.2
65- 135 38.5
Gender <0.001
Men 169 36.1
Women 298 52.8
Working <0.005
Yes 205 41.2
No 252 52.8
Country of birth <0.01
Sweden 278 40.3
Nordic 59 49.1
Europe 58 67.2
Other 72 52.8
Living conditions <0.05
Alone Yes 128 57.8
No 336 42.6
With other adult Yes 318 41.5 <0.005
No 146 58.2
With children <18 Yes 123 42.3 NS
No 341 48.1
Healthy <0.0000
Yes 247 25.1
Fair 26 69.2
No 190 72.1
Life satisfaction <0.0001
Low (score1–2) 45 88.9
Fair (score 3–4) 128 78.9
High (score 5–7) 292 25.7
Population 470 46.4
* N not equal to 470 due to missing some values.
Apart from BMI and smoking status, no other variable was significantly related to perceived health, table 2. Subjects who report PPH have higher BMI and smoke more than those with perceived good health. BMI is negatively related to PPH. The correlation for BMI and PPH is (r = -0.10; p < .05).
Table 2 Population distribution in per cent (%) by perceived health and health status measures. Values are means unless otherwise indicated. Poor perceived health (N = 218), good (N = 250).
Perceived health
Variables Number Poor (scores 1–4) Good (scores 5–7) P-value
Body Mass Index (BMI)* 28.3 27.3 <0.05
Smoking Yes 133 57.1 42.9 <0.005
No 335 42.3 57.7
Blood pressure
Systolic BP 134.4 135.4 NS
Diastolic BP 80.5 80.0 NS
Heart rate/minute 70.7 68.8 NS
Blood glucose 5.99 5.78 NS
Serum triglycyrids 1.54 1.50 NS
Serum cholesterol 5.46 5.28 NS
Spirometry
Vital capacity (VC) 96.5 93.8 NS
Forced expiratory volume (FEV1) 92.6 93.9 NS
Forced vital capacity (FVC) 91.8 91.3 NS
PEF 89.3 91.8 NS
Electrocardiography** NS
Normal 72 52.8 46.2
Not sure pathological 49 36.7 63.3
Suspect pathologic 226 43.8 56.2
Pathological 111 52.3 47.8
* BMI = Weigh/length (m2)
**Per cent
Symptoms and perceived health
Having any of the 13 of 15 symptoms listed in the PQ was significantly related to PPH, i.e. those who report any of these symptoms more often have PPH than do non-symptomatic subjects. For some of these symptoms the risk is almost doubled or tripled, Table 3. For example, 87% of those reporting fainting spells have PPH compared to 44.4% among non-symptomatic subjects. The figures for feeling tired or having low energy and trouble sleeping were 62.6 and 62.9% as compared to 26.2 and 33.6% respectively.
Table 3 The number of subjects with/without symptoms and the percentage (%) of these reporting perceived poor health (PPH).
Number* PPH % P-value
Symptoms Yes No Yes No
Stomach pain 130 339 60.3 41.3 <0.0005
Back pain 218 251 59.0 35.5 <0.0001
Pain in arms, legs & joints 288 181 52.4 37.0 <0.005
Menstrual cramps 46 210 50.0 46.2 NS
Headaches 176 291 57.0 40.2 <0.001
Chest pain 100 368 66.7 41.0 <0.0001
Dizziness 133 333 64.9 39.3 <0.0001
Fainting spells 23 446 87.0 44.4 <0.0001
Palpitations 88 381 63.6 42.5 <0.001
Shortness of breath 86 383 65.1 42.3 <0.0005
Pain or problems during sexual intercourse 27 238 57.7 45.4 NS
Constipation, loose bowels, or diarrhoea 117 352 62.4 41.2 <0.0001
Nausea, gas, or indigestion 145 324 58.6 41.1 <0.001
Feeling tired or having low energy 263 206 62.6 26.2 <0.0001
Trouble sleeping 202 265 62.9 33.6 <0.0001
* N not equal to 470 due to missing some values.
Chronic disease or conditions and perceived health
Subjects who report having any chronic disease or condition report to a higher extent having PPH, table 4. In general about 50% of subjects with chronic disease report PPH as compared to 26.3% among non-diseased subjects. Having any of nine of these diseases or conditions was related to PPH, heart failure, asthma, neurological disease, musculoskeletal and joint disorders, pain syndrome, psychiatric disorders, gastrointestinal and urinary tract troubles. For example, about 80% of subjects with a psychiatric diagnosis report PPH and for heart failure the figure is 73.3%.
Table 4 The number of subjects with/without a disease or a condition and the percentage (%) of these reporting perceived poor health (PPH).
Number* PPH %
Disease or condition Yes No Yes No P-value
Chronic disease 394 76 50.3 28.5 <0.005
Blood pressure 84 354 49.6 45.8 NS
Angina 38 429 52.6 46.1 NS
Heart failure 15 452 73.3 45.8 <0.01
Diabetes 44 422 54.5 45.7 NS
Asthma 59 408 67.8 43.6 <0.005
Chronic obstructive disease 13 454 38.5 46.9 NS
Neurological disease 25 441 68.0 45.6 <0.0005
Musculoskeletal disease 89 378 68.5 41.5 <0.0001
Joint disease 255 212 55.3 36.3 <0.0005
Pain syndrome 112 355 65.1 39.4 <0.0001
Cancer 18 448 44.4 46.9 NS
Psychiatric disorder 40 427 80.0 43.6 <0.0001
Eye disease 65 402 53.8 45.5 NS
Ear disease 39 427 46.1 46.8 NS
Gastrointestinal disorders 120 347 59.2 42.4 <0.0005
Urinary tract disease 59 406 55.9 45.1 <0.05
* N not equal to 470 due to missing some values.
From table 5 is obvious that depression, anxiety and compulsive disorders are related to PPH. About 75% of subjects with one of these three disorders report PPH.
Table 5 The number of subjects with/without a psychiatric diagnoses and the percentage (%) of these reporting perceived poor health (PPH). Perceived poor health (N = 218), good (N = 250).
Number % PPH
Diagnose Yes No Yes No P-value
Depressive disorders 216 249 73.8 34.6 <0.0001
Anxiety 216 380 76.5 39.7 <0.0001
Compulsive disorders 23 441 76.5 44.4 0.0001
Social phobia 10 455 60.0 46.2 NS
Probable alcohol abuse/dependent 32 433 46.9 46.4 NS
Eating disorders 11 454 54.5 46.3 NS
Logistic regression analyses
Table 6 shows odds ratios (OR) with 95% confidence intervals (95% CI) for having poor health by age, gender, living and working status, country of birth, smoking, chronic disease, depression, number of symptoms and life satisfaction. The ORs are adjusted for all confounders. We analysed the relationships between perceived health and the explanatory variables, using the unconditional logistic regression in successive models, adjusted for age and gender in the first model. This shows that subjects aged 45–64 years have higher OR than older subjects, i.e. aged 65 or older. The OR was 1.62. The figure for males was 0.52 compared to females.
Table 6 Odds rations (OR) with 95% confidence intervals (95% CI) for having poor health by age, gender, living and working status, country of birth, smoking, chronic disease, depression, number of symptoms and life satisfaction. The ORs are adjusted for all confounders.
Perceived poor health (scores 1–4)
Model 1 Model 2 Model 3 Model 4
OR 95% CI OR 95% CI OR 95% CI OR 95% CI
Age (65+ years = reference) 16–44 0.96 0.59–1.57 1.51 0.83–2.76 0.65 0.31–1.33 0.65 0.29–1.45
45–64 1.62 1.05–2.50 2.47 1.45–4.21 1.73 0.95–3.18 1.44 0.74–2.81
Male (female= ref) 0.52 0.35–0.76 0.56 0,37–0.84 0.86 0.54–1.37 0.87 0.52–1.46
Living alone (no = ref) 0.52 0.33–0.83 0.67 0.40–1.11 0.74 0.42–1.30
Working (yes =ref) 2.10 1.35–3.26 1.42 0.86–2.34 1.21 0.69–2.11
Country of birth (other = ref) Sweden 0.50 0.28–0.88 0.52 0.26–1.02 0.41 0.20–0.84
Nordic 0.57 0.27–1.21 0.44 0.19–1.02 0.26 0.10–0.66
Europe 1.19 0.55–2.57 0.96 0.40–2.31 0.68 0.26–1.80
Smoking (yes = ref) 0.81 0.50–1.32 0.97 0.56–1.66
Chronic disease (yes = ref) 0.68 0.36–1.27 0.97 0.56–1.27
Depression (yes = ref) 0.29 0.17–0.48 0.55 0.31–0.98
Number of symptoms (>6 ref) 1–3 symptoms 0.25 0.13–0.48 0.30 0.15–0.60
4–6 symptoms 0.85 0.46–1.58 0.98 0.50–1.92
Life satisfaction* (high = ref) Low 15.40 5.28–44.97
Fair 7.02 3.98–12.38
*Low (scores 1–2), fair (scores 3–4) and high (scores 5–7)
** Interaction tests for chronic disease*depression, depression*symptoms and chronic disease*depression*symptoms were not significant.
In the second model, when living conditions and country of birth were added to the analysis, the OR for subjects aged 45–64 years and for males remained significant. The ORs for subjects not living alone, working subjects and those born in Sweden were lower than for those living alone, not working and born outside Europe.
In the third model, with the addition of smoking status, chronic disease, depression and number of symptoms, the pattern changed. Neither age nor gender was significant any longer. However, depression and number of symptom were significant. Non-depressed had OR = 0.29 as compared to depressed. This means that depressed people have 71% higher risk than non-depressed of having PPH. The OR for having 1–3 symptoms and PPH was 0.25 as compared to those with more than 6 symptoms. The OR for those with 4–6 symptoms was 0.85, as compared to those with more than 6 symptoms, but this is not significant. The linear trend is interesting. However, the OR for subjects born Sweden was still lower as than for those born outside Europe but the 95% CI overlapped 1 (1.02), which was not significant.
In the final model (model 4) when all confounders were included, i.e. life satisfaction in addition to the previous variables, being born in Sweden or other Nordic countries was related to lower OR as compared to those born outside Europe. The OR for those born in Sweden was 0.41 and for those born in other Nordic countries 0.26. The OR for non-depressed and those with 1–3 symptoms was somewhat higher but still significant.
The OR for low satisfaction with life was as high as 15.40 in comparison to those who are satisfied with life. However, the 95% CI was very wide, 5.28–44.97. The OR for those who report fair life satisfaction is 7.02 and the 95%CI is 3.98–13.38.
Discussion
This study has shown that various socio-demographic characteristics and country of birth affect the subject's perceived health. A substantially high percentage reported that they had poor perceived health, with the lowest percentage in those born in Sweden. These figures were higher than the national Swedish average (5–6%) and in other international reports [41,42]. However, Hjern et al. also found a higher prevalence of poor perceived health in subjects born outside Sweden [43]. Williams, in a study of women from the United States, found that socio-economic status was an important determinant of racial/ethnic disparities in health, but several other factors including, for example, medical care, migration, acculturation, stress and resources, also play a role [27]. Not surprisingly, perceived health was influenced by socio-demographic characteristics in this study, but it is noteworthy that another individual factor, "life satisfaction", showed a profile stronger than that of socio-demographic characteristics even after adjustment was made for age, gender, education, occupation, country of birth, chronic disease or condition and symptoms. This factor probably plays an important role in a subject's health and indicates the overall life satisfaction including health. The question about "life satisfaction" can also be used to assess health. Prospective studies are needed to confirm this. Indeed, this variable can probably be used together with perceived health, but needs to be investigated further. Many studies have shown the effect of age, gender, education, occupation and country of birth on health, which accord with our results [44,45,5].
An interesting finding in this study is that patients aged 45–64 years have poorer health than older subjects. We believe that in this multi-ethnic population this age group is more vulnerable and sicker than the older subjects. It is possible that specific factors related to the neighbourhood in this study have an impact on health. Future studies should elucidate this issue. This could have important health policy implications. However, in the logistic model, the impact of age was not any longer significant when the symptom variable was taken into account.
As is the case with all questionnaire surveys, there was the possibility that patients exaggerated or underreported a condition, partly due to difficulties in remembering (recall bias). In addition to the potential problem of recall bias, there is the possibility of selection bias. The participants in this investigation were voluntary and time-limited, and we could only include those who showed interest first. It is possible that some selection bias occurred as a result of this consecutive procedure. It is possible, for instance, that our study included only the healthier patients in the Jordbro Health District because the sickest patients were not able to come to the health centre.
The questionnaire used in the present study (PRIME-MD) has been validated previously and shown to have good accuracy for psychiatric disorders [51-53]. Also, different parts of the questionnaire used here were previously validated in other investigation [35,36,49,51]. We also tested the questionnaire as a whole in advance in a pilot study and judged it to be satisfactory.
Although this study is not prospective, causal relations cannot be drawn from this investigation. The most of the literature on this issue are cross-sectional studies and not suited for statement or implications. Further prospective research is needed to clarify the direction of association. The new finding in this study is the strong association between life satisfaction and perceived health apart from country of birth, symptoms and depression, and also the fact that perceived health has a stronger correlation with psychiatric than somatic conditions. The strength of this study is that this represents a primary care patient population which makes it unique. Furthermore, the fact that we have controlled for a large number of confounders makes it more unique.
In conclusion, country of birth, depression, number of symptoms and life satisfaction are factors related significantly and independently to perceived health. Life satisfaction was the strongest predictor of poor health.
Abbreviations
OR = Odds ratio
95% CI = 95% Confidence Interval
Prime-MD = Primary Care Evaluation of Mental Disorders
PQ = Patient questionnaire
JHC = Jordbro Health Care Centre
VC = Vital capacity
FEV1 = Forced vital capacity during one second
PEF = Peak expiratory flow
PPH = poor perceived health
r = Spearman's correlation coefficient
BMI = body mass index
Acknowledgements
This study was supported by grants from Stockholm County Council (Dagmar & ALF Fund) and Haninge Community Council (Economic Target to Large Cities).
==== Refs
Kaplan GA Camacho T Perceived health and mortality: a nine-year follow-up of the Human Population Laboratory cohort Am J Epidemiol 1983 117 292 304 6829557
Sundquist J Johansson SE Self-reported poor health and low educational level predictors for mortality: a population-based follow-up study of 39,156 people in Sweden J Epidemiol Community Health 1997 51 35 40 9135786
Miilunpalo S Vuori I Oja P Pasanen M Urponen H Self-rated health status as a health measure: the predictive value of self-reported health status on the use of physician services and on mortality in the working-age population J Clin Epidemiol 1997 50 517 528 9180644 10.1016/S0895-4356(97)00045-0
Idler EL Angel RJ Self-rated health and mortality in the NHANES-1 epidemiologic follow-up study Am J Public Health 1990 80 446 452 2316767
Kaplan G Barell V Lusky A Subjective state of health and survival in elderly adults J Gerontol 1988 43 S114 120 3385152
Idler EL Benyamini Y Self-rated health and mortality: a review of twenty-seven community studies J Health Soc Behav 1997 38 21 37 9097506
Benyamini Y Idler EL Community studies reporting associations between self-rated health and mortality Research in Aging 1999 21 393 401
Heistaro S Jousilahti P Lahelma E Vartiainen E Puska P Self-rated health and mortality: a long-term prospective study in eastern Finland J Epidemiol Community Health 2001 55 227 232 11238576 10.1136/jech.55.4.227
Angel R Gronfein W The use of subjective information in statistical models ASR 1988 53 464 473
Wannamethee G Shaper AG Self-assessment of health status and mortality in middle-aged British men Int J Epidemiol 1991 20 239 245 2066228
Ferraro KF Farmer MM Wybraniec JA Health trajectories: Long-term dynamics among black and white adults J Health and Soc Behav 1997 38 38 54 9097507
Al-Windi A Elmfeldt D Svärdsudd K The influence of sociodemographic characteristics on health care utilisation in a Swedish municipality Ups J Med Sci 2004 109 33 42 15124951
Idler EL Kasl S Self-ratings of health: Do they also predict change in functional ability? J Gerontol B Psychol Sci Soc Sci 1995 50B S344 353 7583813
Al-Windi A Elmfeldt D Svärdsudd K The influence of perceived well-being and reported symptoms on health care utilization: A population-based study J Clinical Epidemiol 2002 55 60 66 11781123 10.1016/S0895-4356(01)00423-1
Connelly JE Smith GR Philbrick JT Kaiser DL Healthy patients who perceive poor health and their use of primary care services J Gen Intern Med 1991 6 47 51 1999746
Kjoller M Rasmussen NK Keiding LM Self-reported health and morbidity among adult Danes 1987–1994 Ugeskr Laeger 1999 161 2948 2954 10354781
Borgquist L Hansson L Nettelbladt P Nordstrom G Lindelow G Perceived health and high consumers of care. a study of mental health problems in a Swedish primary health care district Psychol Med 1993 23 763 770 8234582
Sihvonen AP Kunst AE Lahelma E Valkonen T Mackenbach JP Socioeconomic inequalities in health expectancy in Finland and Norway in the late 1980s Soc Sci Med 1998 47 303 315 9681900 10.1016/S0277-9536(98)00020-3
Hallqvist J Lundberg M Diderichsen F Ahlbom A Socioeconomic differences in risk of myocardial infarction 1971–1994 in Sweden: Time trends, relative risks and population-attributable risk Int J Epidemiol 1998 27 410 415 9698128 10.1093/ije/27.3.410
Cavelaars AE Kunst AE Geurts JJ Helmert U Lundberg O Mielck A Matheson J Mizrahi A Mizrahi A Rasmussen N Spuhler T Mackenbach JP Morbidity differences by occupational class among men in seven European countries: an application of the Erikson-Goldthorpe social class scheme Int J Epidemiol 1998 27 222 230 9602402 10.1093/ije/27.2.222
Devlin N Hansen P Herbinson P Variations in self-reported health status: results from a New Zealand survey N Z Med J 2000 113 517 520 11198514
Morris JK Cook DG Walker M Shaper AG Non-consulters and high consulters in general practice: cardio-respiratory health and risk factors J Public Health Med 1992 14 131 137 1515196
Marmot MG Adelstein AM Bulusu L Immigrant mortality in England and Wales 1970–78. Causes of death by country of birth. Studies on Medical and Population Subjects No. 47 1984 London: OPCS; Her Majesty's Stationery Office
Nazroo JY Bartley MB, Blane D, Davey Smith G Genetic, cultural or socio-economic vulnerability? Explaining ethnic inequalities in health The sociology of health inequalities 1998 Oxford: Blackwell Publishers
McGee DL Liao Y Cao G Cooper RS Self-reported health and mortality in a multiethnic US cohort Am J Epidemiol 1999 149 41 46 9883792
Harding S Maxwell R Differences in the mortality of migrants. In F. Drever, M Whitehead, Health inequalities 1997 London Stationery Office
Williams DR Racial/ethnic variations in women's health: the social embeddedness of health Am J Public Health 2002 92 588 597 11919058
Nazroo JY The health of Britain's ethnic minorities 1997 London: Policy Studies Institute
Sundquist J Ethnicity, social class and health. A population-based study on the influence of social factors on self-reported illness in 223 Latin American refugees, 333 Finnish and 126 southern European labour migrants and 841 Swedish controls Soc Sci Med 1995 40 777 787 7747212 10.1016/0277-9536(94)00146-K
Kwok RK Yankaskas BC The use of census data for determining race and education as SES Indicators: A validation study AEP 2001 11 171 177 11293403
Palmore E Kivett V Change in life satisfaction: a longitudinal study of persons aged 46–70 J Gerontol 1977 32 311 316 850059
Wang CW Iwaya T Kumano H Suzukamo Y Tobimatsu Y Fukudo S Relationship of health status and social support to the life satisfaction of older adults Tohoku J Exp Med 2002 198 141 149 12597241 10.1620/tjem.198.141
Benyamini Y Leventhal H Leventhal EA Self-rated oral health as an independent predictor of self-rated general health, self-esteem and life satisfaction Soc Sci Med 2004 59 1109 1116 15186909 10.1016/j.socscimed.2003.12.021
Ho SC Woo J Lau J Chan SG Yuen YK Chan YK Chi I Life satisfaction and associated factors in older Hong Kong Chinese J Am Geriatr Soc 1995 43 252 255 7884112
Al-Windi A Depression in general practice Nord J Psychiatry
Tibblin G Peciva S Kullman S Svardsudd K "The Gothenburg quality of life instrument": An assessment of well-being and symptoms among men born 1913 and 1923. Methods and validity Scand J Prim Health Care 1990 1 33 38
Loerch B Szegedi A Kohnen R Benkert O The Primary Care Evaluation of Mental Disorders (PRIME-DM) German version: a comparison with CIDI J Psychiat Res 2000 34 211 220 10867116 10.1016/S0022-3956(00)00005-4
American Psychiatric Association. Diagnostic and statistical manual of mental disorders 1994 4 Author, Washington, DC
SAS: Introductory Guide; ed 3.1 Campus Drive, SAS Institute 31 ed Cary 1997
JMP-program: Statistical Discovery Software Introductory Guide, Version 4 SAS Institute Inc, Cary, NC, USA 2000
Statistics Livslängd, hälsa och sysselsättning (Length of life, health and occupation) Stockholm, Sweden, Statististka centralbyrån (Statistics Sweden) 2002
Markovic M Manderson L Kelaher M The health of immigrant women: Queensland women from the former Yugoslavia J Immigrant Health 2002 4 5 15 10.1023/A:1013003126561
Hjern A Haglund B Persson G Rosen M Is there equity in access to health services for ethnic minorities in Sweden? Eur J Public Health 2001 11 147 152 11420800 10.1093/eurpub/11.2.147
Eriksson I Undén A-L Elofsson S Self-rated health. Comparisons between three different measures. Results from a population study Int J Epidemiol 2001 30 326 33 11369738 10.1093/ije/30.2.326
Verbrugge LM Sex differences in health Prevention 1982 97 417 437
Gijsbers van Wijk CM Kolk AM Sex differences in perceived health Ned Tijdschr Geneeskd 1997 141 283 287 9148163
Joung IM van der Meer JB Mackenbach JP Marital status and health care utilization Int J Epidemiol 1995 24 569 575 7672898
Fylkenes K Determinants of health care utilisation: Visits and referrals Scand J Soc Med 1993 21 40 50 8469943
Al-Windi A Determinants of Health Care and Drug Utilisation: The Causes of Health Care Utilisation Study Uppsala, Uppsala University, PhD Thesis 2000
Bosworth HB Butterfield MI Stechuchak KM Bastian LA The relationship between self-rated health care service use among women veterans in a primary care clinic Women's Health Issues 2000 10 278 285 10980445 10.1016/S1049-3867(00)00056-6
Parker T May PA Maviglia MA Petrakis S Sunde S Gloyd SV PRIME-MD: its utility in detecting mental disorders in American Indians Intl J Psychiat Med 1997 27 107 128
Kobak KA Taylor LH Dottl SL Greist JH Jefferson JW Burroughs D Mantle JM Katzelnick DJ Norton R Henk HJ Serlin RC A computer-administered telephone interview to identify mental disorders JAMA 1997 278 905 910 9302242 10.1001/jama.278.11.905
| 15850487 | PMC1131916 | CC BY | 2021-01-04 16:36:32 | no | Immun Ageing. 2005 Apr 25; 2:7 | latin-1 | Immun Ageing | 2,005 | 10.1186/1742-4933-2-7 | oa_comm |
==== Front
Int Semin Surg OncolInternational seminars in surgical oncology : ISSO1477-7800BioMed Central London 1477-7800-2-101587782110.1186/1477-7800-2-10Case ReportTuberculosis masquerading as malignancy: a multimodality approach to the correct diagnosis – a case report Amukotuwa Shalini [email protected] Peter FM [email protected] Peter J [email protected] Gerard J [email protected] John [email protected] Stephen M [email protected] Department of Medical Imaging, St. Vincent's Hospital, Fitzroy 3065, Melbourne, Australia2 Department of Orthopaedics, St. Vincent's Hospital, Fitzroy 3065, Melbourne, Australia3 Department of Pathology, St. Vincent's Hospital, Fitzroy 3065, Melbourne, Australia2005 7 5 2005 2 10 10 8 4 2005 7 5 2005 Copyright © 2005 Amukotuwa et al; licensee BioMed Central Ltd.2005Amukotuwa 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
Extrapulmonary tuberculosis is one of the great mimickers of medicine, and often masquerades as malignancy. As a result, patients may be referred to oncologists and surgeons for further evaluation and management, delaying the institution of appropriate anti-tuberculous drug therapy.
Case presentation
We present the case of a 21 year old man with tuberculous osteomyelitis, who was referred to the Bone and Soft Tissue Sarcoma Service at our institution with a provisional diagnosis of malignancy. Further investigation revealed extensive retroperitoneal abdominal and pelvic lymphadenopathy. The recognition of certain patterns on imaging, and finally the isolation of Mycobacterium tuberculosis from tissue samples obtained under image guidance, enabled the correct diagnosis to be made.
Conclusion
This case highlights the importance of remaining cognisant of the protean manifestations of extrapulmonary tuberculosis, and illustrates the advantage of a clinically directed multi-modality imaging approach to diagnosis.
==== Body
Background
Over the past decade, there has been a significant rise in the prevalence of extrapulmonary tuberculosis in the developed world, owing to the emergence of multi-drug resistant strains of Mycobacterium tuberculosis and an increasing number of immune compromised individuals and immigrants from the developing world [1-3]. Despite increasing awareness and the availability of better imaging and other diagnostic tests, extrapulmonary tuberculosis remains a difficult diagnosis to make, due to its often non-specific and protean manifestations. It is not uncommon for this disease to mimic malignancy. For instance, skeletal tuberculosis can clinically simulate sarcoma, leading to the referral of patients suffering from this condition to oncologists and surgeons, delaying correct diagnosis and the institution of appropriate therapy. Therefore, the prompt recognition by these clinicians of distinguishing features is vital for correct diagnosis, to facilitate timely anti-tuberculous therapy. In particular, imaging is a key tool in helping to make the diagnosis of extrapulmonary tuberculosis, through the recognition of certain key radiological patterns. However, as there are no pathognomonic imaging findings, the diagnosis ultimately rests on histopathological and microbiological confirmation. Whilst in the past this necessitated open surgical biopsy, tissue samples can now be obtained with minimal invasion, under image-guidance.
To highlight these issues we present the case of a young male with pain and swelling of the right elbow, referred to the Bone and Soft Tissue Sarcoma Service at our institution with a presumptive diagnosis of malignancy. Further investigation revealed extensive retroperitoneal abdominal and pelvic lymphadenopathy. The case illustrates the importance of a clinically directed multi-modality imaging approach for the exclusion of malignancy and the accurate diagnosis of tuberculous infection, by identifying sites of pathology and obtaining adequate tissue samples for histological and microbiological confirmation.
Case presentation
Case report
A previously well 21 year old East Timorese male presented to his local medical officer with a 2 month history of right elbow pain, associated with swelling and progressive limitation of movement. He also reported a 6 kg loss of weight during this time, but denied any other constitutional symptoms.
He appeared well, was afebrile, and had no abnormality on chest or abdominal examination. Examination of the right elbow revealed a firm, tender swelling, approximately 2 cm in diameter, over the lateral humeral epicondyle, with associated limitation of elbow flexion to 60 degrees.
Initial investigation with plain XR of the right elbow joint revealed a joint effusion (Figure 1). As these findings were non-specific, further evaluation with different imaging modalities was undertaken.
Figure 1 Plain radiographs of the right elbow. A. AP view, demonstrating no changes. B. Lateral view demonstrating a joint effusion (arrowed).
Ultrasound examination of the elbow showed a large joint effusion (Figure 2). Scintigraphic imaging with Tc99m three phase bone scanning demonstrated increased blood pool activity, and diffusely increased osteoblastic activity, involving the right elbow joint, particularly the distal humerus and ulna, as well as the distal left radius (Figure 3).
Figure 2 Ultrasound examination of the right elbow. This confirms the presence of a hypoechoic joint effusion (arrowed).
Figure 3 Limited bone scan using Tc99m. This demonstrates diffusely increased radioactive tracer uptake by the right elbow, particularly at the lateral epicondyle (dashed arrow) and the proximal ulna, as well as by the distal left radius (solid arrow).
CT scan confirmed the presence of cortical erosions involving the right lateral epicondyle of the humerus and the ulna. Further, there was increased attenuation within the proximal medullary shaft of the ulna, and diffuse soft tissue thickening around the proximal ulna and radial head, with an associated joint effusion (Figure 4).
Figure 4 Multiplanar CT of the right elbow joint. A. Axial view showing a complex joint effusion (arrowed) and soft tissue swelling adjacent to the elbow. B. Axial slice through the proximal forearm, demonstrating the joint effusion surrounding the radial head (solid arrow) and increased soft tissue density involving the medullary shaft of the ulna (dashed arrow). C. Sagittal bone targeted reconstruction, showing subtle cortical erosions (arrowed) of the proximal ulna. D. Sagittal soft tissue reconstruction, demonstrating increased soft tissue density (arrowed) of the medullary shaft of the proximal ulna.
Further functional imaging with Thallium 201 chloride revealed intense early uptake in the right lateral epicondyle of the humerus and proximal ulna (Figure 5A). The delayed 4 hour images demonstrated tracer retention (Figure 5B). The delayed whole body views showed further abnormal thallium retention involving the mediastinum, para-aortic regions of the abdomen, the left inguinal region and the distal left radius (Figure 6).
Figure 5 Functional metabolic imaging, using Thallium 201 chloride. A. Early planar images of the right elbow, showing intense radioactive tracer accumulation in the lateral humeral epicondyle and the proximal ulna. B. Delayed images at 4 hours, showing retention of tracer.
Figure 6 Functional metabolic imaging, using Thallium 201 chloride. Delayed whole body images, demonstrating abnormal tracer retention involving the mediastinum. (solid arrow), para-aortic region and left inguinal area (dashed arrow).
MR demonstrated erosive changes involving the lateral epicondyle, a loculated joint effusion and thickening and enhancement of the joint capsule and the annular ligament of the superior radio-ulnar joint. These changes suggested either a neoplasm or possibly infection (Figure 7).
Figure 7 Multiplanar MR imaging of the right elbow. A. T1 weighted axial image at the level of the elbow joint, demonstrating a complex effusion (arrow), particularly adjacent to the proximal ulna. B. Contrast enhanced axial image showing enhancement around the joint (arrow). C. T2 weighted axial image again showing extensive joint effusion D. Coronal contrast enhanced image showing the complex effusion and abnormal signal intensity in the medullary canal of the proximal ulna (arrow).
The patient was therefore referred to the Bone and Soft Tissue Sarcoma service for further investigation.
Biopsy of the right elbow joint demonstrated chronic synovial inflammation and numerous polymorphonuclear cells in the synovial fluid, but no malignant cells (Figure 8). Microbiological examination, including staining for acid-fast bacilli and fungi, revealed no organisms.
Figure 8 CT guided core biopsy of the elbow. A. Tissue adjacent to proximal ulna, at the level of the elbow joint, was obtained under CT guidance. B. Histological examination, after haematoxylin and eosin staining, demonstrated non-specific chronic inflammation and numerous polymorphonuclear cells.
CT scanning of the chest, abdomen and pelvis, performed in light of the thallium findings, revealed extensive retroperitoneal lymphadenopathy. The lymphadenopathy was characterised by peripheral enhancement with prominent central areas of low attenuation (Figures 9 and 10).
Figure 9 Multiplanar chest CT. This demonstrates mediastinal lymphadenopathy and subtle pulmonary involvement. A. Contrast enhanced axial slice, showing enlarged lymph nodes (arrowed) in the para-aortic window. B. Axial slice showing subcarinal lymphadenopathy (arrowed). C. Coronal slice, showing small nodules in the left upper lobe (arrowed).
Figure 10 Multiplanar abdominal and pelvic CT, demonstrating extensive lymphadenopathy. A. Large low attenuation portal lymph node (solid arrow) showing peripheral enhancement, and associated para-aortic nodes (dashed arrows). B. Extensive heterogenous attenuation side wall pelvic adenopathy. C. Coronal image, demonstrating extensive low attenuation para-aortic and pelvic lymphadenopathy (arrowed)
Subsequent CT guided biopsy of the retroperitoneal lymphoid tissue revealed necrotising granulomatous lymphadenitis indicative of Mycobacterium tuberculosis infection (Figure 11). While no acid fast bacilli were visible on Ziehl-Nielsen staining, polymerase chain reaction was positive for M. tuberculosis DNA. The lymph node biopsy specimens cultured M. tuberculosis after 6 weeks.
Figure 11 CT guided biopsy of the para-aortic lymph nodes. A. 14 gauge core needle biopsy. B. The core of tissue obtained. C. Histological examination after haematoxylin and eosin staining, demonstrating classic caseating granulomata of tuberculous infection (arrowed).
A final diagnosis of disseminated extrapulmonary tuberculosis was therefore made. Anti tuberculous therapy was commenced with good clinical response.
Discussion
Extrapulmonary tuberculosis, a condition whose resurgence can be ascribed to the emergence of multi-drug resistance and an increasing number of immune compromised individuals and immigrants from the developing world, remains a difficult diagnosis to make [1-3]. It is one of the great mimickers in medicine, with non-specific clinical and radiological manifestations that can suggest numerous other disease entities in particular malignancy.
Although a positive tuberculin test and chest imaging findings are supportive of the diagnosis, absence of these does not exclude it [4,5]. Therefore other diagnostic tests need to be used, combined with a high index of clinical suspicion. As highlighted by the above case, a multimodality imaging approach assists in the early diagnosis of extrapulmonary tuberculosis, enabling the timely initiation of appropriate therapy.
Musculoskeletal involvement occurs in 1 to 3 percent of patients with tuberculosis, usually due to haematogenous seeding [4,6]. In more than 50 % of cases there is no evidence of concurrent active intrathoracic tuberculosis [4].
While the most common site of osseous involvement is the spine, followed by the femur, tibia and the small bones of the hands and feet, any bone can potentially be affected [1,4,6,9,10]. The most common presenting symptoms of tuberculous osteomyelitis are non-specific pain and swelling. Consequently, as in the above case, skeletal tuberculosis frequently mimics osteosarcoma, leading to incorrect initial diagnosis and delay in the institution of treatment.
Plain XR findings in bony tuberculosis include osteopenia, osteolytic foci with poorly defined edges, and varying amounts of sclerosis and periostitis [1,5,9]. The metaphysis is typically involved, although epiphyseal involvement also occurs [1]. A particular type of tuberculous osteomyelitis, cystic tuberculosis, produces round or oval radiolucencies with variable amounts of sclerosis [1,6]. These findings are, however, non-specific, and can be found in a range of pathological processes, including neoplasia. For instance, osteolytic and osteosclerotic foci can be found in metastases from various primary tumours, for instance prostate, breast and renal cell carcinoma, whilst sclerosis and a periosteal reaction may suggest a primary bone tumour. Even the radiographic characteristics of cystic tuberculosis can also be found in metastatic carcinoma or germ cell tumours, and plasma cell myeloma [1,4].
There are a few radiographic features, however, that favour tuberculous infection over neoplasia. These include the presence of small juxtacortical abscesses or rings of inflammatory tissue, due to cortical destruction and spread of infection to the extraosseous tissues [11]. Regardless, the features of tuberculous osteomyelitis are so variable and inconstant that further investigation is usually required.
Nuclear Medicine techniques of value in the evaluation of tuberculous bony involvement include Tc99m methylene diphosphonate bone scanning, particularly in multifocal bone involvement. Sensitivity is reduced significantly however in pure septic arthritis and discitis without concomitant osteomyelitis. In these clinical situations, MRI is the modality of choice.
Another useful radiotracer technique is functional metabolic imaging, which assesses the basal metabolic activity of inflammatory and neoplastic lesions. Thallium-201 chloride (TI-201) scanning is a good example of this modality. TI-201 is a potassium analogue that is actively concentrated in cells by the sodium-potassium ATPase pump and by a co-transport system mediated by a calcium-dependent ion channel [12]. As enhanced metabolic activity often increases the activity of these pumps, tumours frequently concentrate this tracer more avidly than normal soft tissue or bone, particularly on delayed images. Conversely, inflammatory processes usually show an early increase in tracer uptake, but with reduced metabolic activity, or a washout pattern, on the delayed images. Interestingly, tuberculous lesions often show significant delayed activity or retention as in this case.
Joint involvement in skeletal tuberculosis occurs commonly due to direct invasion from an adjacent focus of osteomyelitis, as in the case presented in this report [4,6]. Characteristically a monoarticular process, tuberculous arthritis usually involves the large weight bearing joints, namely the knee and the hip [1,5]. Phemister's triad of juxta-articular osteoporosis, marginal joint erosions and joint space narrowing is classically described as characteristic of this condition, however these features are again non-specific [1,13,14]. The multiplanar abilities of multislice of CT and MR are important in differentiating tuberculous arthritis from other disease processes. Bone targeted CT gives excellent bony detail, periosteal and cortical definition, while MR is the modality of choice for assessment of the presence and extent of bone marrow changes, effusions and synovial involvement.
Abdominal tuberculosis, like musculoskeletal involvement, is a difficult diagnosis to make, due to its varied clinical presentations, and is again often mistaken for malignancy. Lymphadenopathy is the most common manifestation, with up to two thirds of patients with abdominal tuberculosis demonstrating lymph node involvement [15,16]. The mesenteric, omental and peripancreatic groups most frequently involved, reflecting the lymphatic drainage of the most commonly affected sites in the small bowel and liver [1,16]. Lymph node involvement is usually detected on CT imaging, and this is also the imaging modality of choice for the evaluation of intraabdominal and pelvic lymphadenopathy [16,17]. The main differential diagnosis for diffuse lymph node enlargement is lymphoma, whilst less widespread intraabdominal lymphadenopathy may suggest metastatic malignancy [16].
Certain features on CT imaging help distinguish tuberculous lymphadenopathy from these neoplastic causes. In tuberculous infection, the nodes are usually multiple and large, averaging 2 to 3 cm in diameter [1]. Peripheral enhancement with central areas of low attenuation or loculation are seen on contrast enhanced CT in the majority of cases [6,16-18]. Conglomerate mixed density nodal masses may also occur, likely representing multiple confluent nodes with peri-nodal spread of inflammation [16]. In contrast, lymphomatous adenopathy is characteristically associated with homogenous attenuation [19]. However, whilst heterogeneity is characteristic of the caseous necrosis seen in tuberculous lymphadenopathy, it is by no means pathognomonic, also occurring in metastatic testicular carcinoma [16]. Likewise, although nodal calcification is highly suggestive of tuberculous disease, especially in endemic areas, it can also occur in teratomatous testicular metastases and in non-Hodgkin's lymphoma after treatment with radiotherapy [16].
Because of this overlap in imaging appearances between extrapulmonary tuberculosis and malignancy, even in cases where the imaging and clinical features strongly suggest tuberculosis, the diagnosis requires histopathological and bacteriological confirmation. In the past, this often necessitated open surgical biopsy. Now core needle biopsy has been shown to be quick, safe and effective alternative for obtaining tissue specimens under image guidance [20].
Conclusion
The clinical and radiologic features of extrapulmonary tuberculosis often mimic those of many other pathological processes, in particular malignancy. Oncologists and surgeons must therefore remain cognisant of the possibility of tuberculous infection, and of radiological features that help distinguish this condition from neoplastic processes. However, even with the use of multiple imaging modalities and laboratory diagnostic tools, a definitive diagnosis still requires a positive culture or histologic analysis of biopsy specimens in many cases. Image-guidance facilitates optimal tissue samples to be obtained for analysis.
Abbreviations
Tc99: Technicium 99
TI-201: Thallium 201
TB: tuberculosis
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
SAA drafted and revised the manuscript, and was involved in the editing of the images.
PFC, PJS, GJP and JS critically evaluated and revised the manuscript. They also provided information from their respective areas of expertise, which was critical to the discussion section of the manuscript.
SMS was responsible for the conception of this case report, and was also involved in the drafting and revision of the manuscript.
All authors have read and given final approval of the submitted version of the manuscript.
Acknowledgements
Written consent was obtained from the patient for publication of this case report, including images.
==== Refs
Engin G Acuna B Acuna G Tunaci M Imaging of extrapulmonary tuberculosis RadioGraphics 2000 20 471 488 10715344
Goodman PC Tuberculosis and AIDS Radiol Clin North Am 1995 33 707 717 7610240
Bloom BR Murray CJL Tuberculosis: commentary on a re-emergent killer Science 1992 257 1055 1064 1509256
Davidson P Horowitz I Skeletal tuberculosis Am J Med 1970 48 77 84 4906108 10.1016/0002-9343(70)90101-4
Yao DC Sartoris DJ Musculoskeletal tuberculosis Radiol Clin North Am 1995 33 679 689 7610238
Harisinghani MG McLoud TC Shepard JO Tuberculosis from head to toe RadioGraphics 2000 20 449 470 10715343
Barrington NA The radiology of bone and joint infection Br J Hosp Med 1988 40 464 472 3228664
Murudali D Gold WC Vellend H Multifocal osteoarticular tuberculosis: report of four cases and review of management Clin Infect Dis 1993 17 204 209 8399868
Martini M Quahes M Bone and joint tuberculosis: a review of 652 cases Orthopedics 1988 2 861 866 3387332
Moon MS Tuberculosis of the spine Spine 1997 22 1791 1797 9259793 10.1097/00007632-199708010-00022
Griffith JF Kumta SM Leung PC Imaging of musculoskeletal tuberculosis: a new look at an old disease Clin Orth Related Res 2002 398 32 39 10.1097/00003086-200205000-00006
Sehweil AM McKillop JH Milroy R Wilson R Abdel-Dayem HM Omar YT Mechanism of 201 Tl uptake in tumours Eur J Nucl Med 1989 15 376 379 2776798 10.1007/BF00449228
Campbell JA Hoffman EB Tuberculosis of the hip in children J Bone Joint Surg Br 1995 77 319 326 7706357
Dhillon MS Sharma S Gill SS Tuberculosis of bones and joints of the foot: an analysis of 22 cases Foot Ankle 1993 14 505 513 8314185
Hulnick D Megibow A Naidich D Abdominal tuberculosis: CT evaluation Radiology 1985 157 199 204 4034967
Suri S Gupta S Suri R Computed tomography in abdominal tuberculosis Br J Radiol 1992 72 92 98 10341698
Denton T Hossain J A radiological study of abdominal tuberculosis in a Saudi population, with special reference to ultrasound and computed tomography Clin Radiol 1993 17 409 414 8519148
Leder RA Low VHS Tuberculosis of the abdomen Radiol Clin N Am 1995 33 691 705 7610239
Yang ZG Min PQ Sone S Tuberculosis versus lymphoma in the abdominal lymph nodes: evaluation with contrast enhanced CT AJR Am J Roentgenol 1999 172 619 623 10063847
Radhika S Rajwanshi A Kochar R Abdominal tuberculosis: diagnosis by fine needle aspiration cytology Acta Cytol 1993 37 673 678 8362577
| 15877821 | PMC1131917 | CC BY | 2021-01-04 16:38:36 | no | Int Semin Surg Oncol. 2005 May 7; 2:10 | utf-8 | Int Semin Surg Oncol | 2,005 | 10.1186/1477-7800-2-10 | oa_comm |
==== Front
Malar JMalaria Journal1475-2875BioMed Central London 1475-2875-4-201585423310.1186/1475-2875-4-20MethodologyPractical PCR genotyping protocols for Plasmodium vivax using Pvcs and Pvmsp1 Imwong Mallika [email protected] Sasithon [email protected]üner Anne Charlotte [email protected]énia Laurent [email protected] Frank [email protected] Sornchai [email protected] Nicholas J [email protected] Georges [email protected] Department of Clinical Tropical medicine, Faculty of Tropical Medicine, Mahidol University, Bangkok, Thailand2 Département d'Immunologie, INSERM U567, CNRS UMR8104, Institut Cochin, Université René Descartes, Paris 75014, France3 Laboratoire Commun de Séquençage, Institut Cochin, Université René Descartes, Paris 75014, France4 Wellcome Unit, Faculty of Tropical Medicine, Mahidol University, Bangkok, Thailand5 Centre for Vaccinology and Tropical Medicine, Churchill Hospital, Oxford, UK6 Unité de Parasitologie Bio-Médicale, CNRS URA2851, Institut Pasteur, Paris, France7 Parasitologie Comparée et Modèles Expérimentaux USM307, CNRS IFR101, Muséum National d'Histoire Naturelle, CP52, 61 Rue Buffon, 75231 Paris Cedex 05, Paris, France2005 27 4 2005 4 20 20 12 1 2005 27 4 2005 Copyright © 2005 Imwong et al; licensee BioMed Central Ltd.2005Imwong 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
Plasmodium vivax is the second most prevalent malaria parasite affecting more than 75 million people each year, mostly in South America and Asia. In addition to major morbidity this parasite is associated with relapses and a reduction in birthweight. The emergence and spread of drug resistance in Plasmodium falciparum is a major factor in the resurgence of this parasite. P. vivax resistance to drugs has more recently emerged and monitoring the situation would be helped, as for P. falciparum, by molecular methods that can be used to characterize parasites in field studies and drug efficacy trials.
Methods
Practical PCR genotyping protocols based on polymorphic loci present in two P. vivax genetic markers, Pvcs and Pvmsp1, were developed. The methodology was evaluated using 100 P. vivax isolates collected in Thailand.
Results and Discussion
Analysis revealed that P. vivax populations in Thailand are highly diverse genetically, with mixed genotype infections found in 26 % of the samples (average multiplicity of infection = 1.29). A large number of distinguishable alleles were found for the two markers, 23 for Pvcs and 36 for Pvmsp1. These were generally randomly distributed amongst the isolates. A total of 68 distinct genotypes could be enumerated in the 74 isolates with a multiplicity of infection of 1.
Conclusion
These results indicate that the genotyping protocols presented can be useful in the assessment of in vivo drug efficacy clinical trials conducted in endemic areas and for epidemiological studies of P. vivax infections.
==== Body
Background
Plasmodium vivax is globally distributed and is the dominant species in many countries. Infection by this species is generally regarded as benign and mortality is a rare outcome. 1.2 billion inhabitants of the South East Asian and Pacific countries are at risk from malaria transmission, representing about a third of world population exposed to Plasmodium parasites [1,2]. Most of the populations at risk in these countries (corresponding to the WHO regional groupings SEARO and WPRO) reside in hypoendemic and mesoendemic areas, where half of the recorded infections are due to P. vivax, suggesting that the global burden of vivax malaria morbidity must be close to that of falciparum malaria. It has recently been established that P. vivax infections during pregnancy are associated with reduced birthweight [3], and thereby, increase the risk of neonatal deaths. Thus, the urgency to develop and implement measures to control P. vivax should equal that attending P. falciparum.
The emergence and global spread of P. falciparum parasites resistant to chloroquine and pyrimethamine – sulfadoxine has led to an increase in morbidity and mortality [4], forcing many countries to abandon these relatively cheap drugs. Recent reports of P. vivax resistance to these drugs [5-13] are of concern and steps to restrict the emergence of drug resistance need to be taken. A critical component of malaria control is surveillance for resistance. Despite some disadvantages, in vitro assessment of the susceptibility of P. falciparum to antimalarial drugs provides a rapid means to monitor general levels of drug resistance. Although protocols to culture P. vivax parasites ex-vivo have been developed [14,15], they are still unsuited for routine assays. Thus, well conducted clinical trials, an essential component of resistance monitoring, remain the only alternative for P. vivax. However, in vivo drug efficacy studies conducted in endemic areas are to some extent compromised by their inability to distinguish true recrudescences, i.e. treatment failures, from re-infections which become clinically or parasitologically manifest during the follow-up period, i.e. treatment successes. This has been largely overcome for P. falciparum by the application of genotyping based on well characterized polymorphic regions within the gene encoding msp1, msp2 and glurp [16].
A similar approach has not been adopted for P. vivax, a parasite species less well-studied at the molecular level than P. falciparum [17]. Four polymorphic P. vivax single copy genes have been independently used for molecular epidemiological studies: Pvgam1 coding for a gametocyte antigen, Pvcs coding for the circumsporozoite protein, Pvmsp1 and Pvmsp3α coding for the merozoite surface proteins 1 and 3 alpha, respectively. Amplification of Pvgam1 was found to be associated with artefacts [18] and should thus be excluded as a genetic marker. The suitability of Pvmsp3α as a marker has been recently validated [19]. The extent of diversity found for the remaining two genes was mainly established through sequencing of amplified fragments.
The objective of this work was to develop field applicable methods of genotyping which could also complement clinical trials in vivax malaria. The highly polymorphic genes Pvcs and Pvmsp1 were therefore selected to develop protocols for polymerase chain reaction (PCR) amplification and subsequent analysis of the polymorphic regions. The methodology was then assessed and validated with samples obtained from patients who acquired a P. vivax infection in Thailand.
Materials and Methods
Blood samples
Blood samples were collected from adult patients with symptomatic P. vivax malaria admitted to the Bangkok Hospital for Tropical Diseases, Thailand between 1995 to 1998 (n = 100). All patients gave fully informed consent to enrolment in these studies which were approved by the Ethics committee of the Faculty of Tropical Medicine, Mahidol university. Diagnosis was established by microscopy at the malaria laboratory through examination of thin and thick blood smears, stained with Field's stain. The blood samples were kept at -30°C until DNA extraction.
DNA template preparation
DNA was purified from blood sample using the commercially available DNA Blood Kit (QIAGEN, Germany). The final volume of the DNA solution used as a template for the amplification reactions was such that 1 μL corresponded to 5 μL of whole blood. Confirmation of the microscopic diagnosis as P. vivax and testing for the presence of other Plasmodium species were achieved by a previously described PCR-based protocol [20].
Amplification protocols
In order to increase the sensitivity of amplification, a nested or semi-nested PCR approach was adopted for all the fragments amplified except for one of the three Pvmsp1 regions. Oligonucleotide primers were designed using published sequences of Pvcs and of Pvmsp1 (their sequences are presented in Table 1). The optimal Mg2+ concentrations, annealing temperatures and numbers of cycles were individually determined for the different primer pairs and are presented in Table 1.
Table 1 Primers used for genotyping P. vivax parasites.
Gene Primer Sequencea mM Mg2+ b Annealing °C b Cyclesb
1st 2nd 1st 2nd 1st 2nd
Pvcs VCS-OF ATGTAGATCTGTCCAAGGCCATAAA 1 58 25
VCS-OR TAATTGAATAATGCTAGGACTAACAATATG 1 58 25
VCS-NF GCAGAACCAAAAAATCCACGTGAAAATAAG 1 62 30
VCS-NR CCAACGGTAGCTCTAACTTTATCTAGGTAT 1 62 30
Pvmsp1 F1 VM1-O1R CCACTCCATGAAACTGAAGTGTTTA 2 68 25
VM1-N1F CGATATTGGAAAATTGGAGACCTTCATCAC 2 2 68 68 25 30
VM1-N1R CTTTTGCGCCTCCTCCAGCTGGCTCGTGT 2 68 30
Pvmsp1 F2 VM1-O2F GATGGAAAGCAACCGAAGAAGGGAAT 1 50 25
VM1-O2R AGCTTGTACTTTCCATAGTGGTCCAG 1 1 50 64 25 35
VM1-N2F AAAATCGAGAGCATGATCGCCACTGAGAAG 1 64 35
Pvmsp1 F3 VM1-3F CAAGCCTACCAAGAATTGATCCCCAA 1 66 42
VM1-3R ATTACTTTGTCGTAGTCCTCGGCGTAGTCC 1 66 42
a Sequences are provided 5'- to 3'-end.p
b The two columns indicate conditions for the primary and secondary amplification reactions.
All amplification reactions were carried out in a total volume of 20 μL and in the presence of 10 mM Tris-HCl, pH 8.3, 50 mM KCl, 250 nM of each oligonucleotide primers, 125 μM of each of the four dNTPs, and 0.4 units of AmpliTaq polymerase (Perkin Elmer Cetus, USA). Primary amplification reactions were initiated with one μL of the template genomic DNA prepared from the blood samples, and the one μL of the product of these reactions was used to initiate the secondary amplification reactions. The cycling parameters for PCR were as follows: an initial denaturation step at 95°C for five minutes preceded the cycles of annealing at a temperature defined for each primer pair (Table 1) for two minutes, extension at 72°C for two minutes, and denaturation at 94°C for one minute. After a final annealing step followed by five minutes of extension, the reaction was stopped. PCR products were stored at 4°C until analysis.
The lack of cross-contamination was monitored by the inclusion of multiple, randomly distributed, negative control samples (human DNA or no template) in each amplification run. A subset of the samples were analysed in triplicate in order to confirm the consistency of the results obtained.
Analysis of the amplification product
The DNA fragments obtained following PCR were analysed by electrophoresis in agarose gels. For direct analysis of the fragments, 10 μL of the amplified PCR product were mixed with two μL of loading buffer and applied to 2 % agarose gels. For restriction fragment length polymorphism analysis of the PCR products (PCR-RFLP), 10 μL of the amplified PCR product were first digested with a restriction enzyme (New England Biolabs Inc., UK) according to the suppliers specifications, for three hours in a total volume of 20 μL, before adding five μL of loading buffer and applying to 1.5 % or 1.8 % agarose gels. Electrophoresis was performed in TBE buffer, and the DNA was visualised on an ultraviolet transilluminator following ethidium bromide staining. The size of the amplified fragments was estimated by comparison to a 100 bp ladder marker set.
Selected amplified fragments were purified using QIAquick gel extraction kits (QIAGEN, Germany) and cloned using the TOPO -TA Cloning kit (Invitrogen, U.S.A.). Plasmid DNA containing the fragment was purified from positive bacterial colonies using the QIAquick Miniprep spin kit (QIAGEN, Germany) and sequenced using an ABI automated sequencer. Sequence alignments were performed using the Gene Jockey II program (Biosoft, United Kingdom).
Results
Specificity and sensitivity of the amplification reactions
Primers specific to Pvcs and Pvmsp1 were designed to hybridize to sequences conserved in all variants known at that time. The specificity of all the primer pairs was confirmed since amplification products were only observed when P. vivax DNA was included in the reaction, but not when genomic DNA from P. falciparum, Plasmodium malariae, Plasmodium ovale or humans were used as a template. A set of a 10-fold serial dilution of a template prepared from a blood sample containing P. vivax only, calibrated using a previously described protocol [20], was used to assess the sensitivity of the reactions. Consistent detection of 10 parasite genomes was achieved for Pvcs-specific primers, and of 50 parasite genomes for Pvmsp1-specific primers. These levels of sensitivity were in agreement with subsequent analyses of samples harbouring a known number of parasites.
Genotyping using Pvcs
The circumsporozoite protein gene of P. vivax, as for other Plasmodium species, comprises a central repetitive domain flanked by two conserved domains. The majority of variations observed in Pvcs occur in the repeat region and the immediate pre- and post-repeat sequences. Thus the genotyping strategy was focused on an amplified fragment spanning these regions (Fig. 1A). In a given parasite line, the repetitive domain is composed of a 27 bp element repeated a variable number of times. Variations in the number of repeated elements result in size polymorphisms amenable to detection by electrophoresis. Seven allelic types distinguishable by size were observed in the Thai P. vivax isolates analysed (Fig. 1B), a number consistent with the variations in the number of repeat elements (15 to 21) previously observed for Pvcs [21,22]. In P. vivax, two types of repeat elements are found, VK210 and VK247 [23], and a given Pvcs gene will exclusively bear either the VK210 type (type I repeats based on GDRADGQPA) or the VK247 type (type II repeats based on ANGAGNQPG). In order to differentiate between amplified fragments carrying the VK210 from those carrying the VK247 repeat element, the exclusive presence of Alu I recognition sites in the VK210-type and of Bst NI recognition sites in the VK247-type repeated sequences were exploited. Thus, following digestion with the appropriate enzyme, the amplified fragments are degraded into fragments of < 150 bp, easily distinguishable from the original fragments of > 700 bp (Fig. 1C). In this manner the PCR fragments obtained from an isolate can be grouped by size and by repeat type.
Figure 1 A. Schematic representation of the Pvcs gene, which consists of a conserved region (blank box) flanking a repeated region (grey box) which is itself flanked by pre- and post-repeat specific sequences (cross-hatched boxes). The bold horizontal line represents the PCR-amplified fragment used for further analysis. B. Example of the fragments (denoted A to G) of distinct size observed in different isolates, as single or as mixed infections (some lanes were not labelled because of lack of space). C. Fragments digested (D) or undigested (U) with Alu I or BstN I, two restriction enzymes with multiples sites in VK 210 type or VK247 type repeat sequences, respectively. Digestion cuts the fragment in small fragments of less than 150 bp in size. D. PCR-RFLP using specific restriction enzymes analysis for the presence/absence (yes/no) of specific pre- and post-repeat sequences in fragments carrying the VK210 or VK247 type repeats. Undigested fragments are denoted by (U) and Pr and Po denote digests to determine the presence of pre- and post-repeat sequence types, respectively. A 100 bp ladder, where the 600 bp band stains most intensely, was used a molecular weight marker (M) for all the gels.
In order to increase the genotyping resolution of Pvcs as a marker, so as to subdivide the P. vivax population into a larger number of Pvcs allelic types, sequence variations previously observed in the pre- and post-repeat regions [24] were used to develop an additional PCR-RFLP protocol (Fig. 1D). For some Pvcs genes, the pre-repeat consists of the Region I sequence (KLKQP) that directly abuts the repeated region, while for others amino acid residues are inserted between the two regions, namely a T, V or A residue for VK210-type Pvcs genes, or ED residues for VK247-type Pvcs genes. The Scr FI restriction endonuclease cuts VK210-type Pvcs fragments that do not have the T/A/V insertion following Region I, thus shortening them by 41 bp. VK247-type Pvcs bearing the ED insertion after Region I are cut by the Mbo II restriction endonuclease and the amplified fragment is thereby shortened by 55 bp. For the post-repeat region, a 36 bp insertion has been found for some VK210-type Pvcs. This insertion harbours a recognition site for the Bbs I enzyme, and digestion truncates the corresponding PCR fragments by 115 bp.
Genotyping of the 100 P. vivax isolates collected from Thailand was achieved successfully using the protocols described above. In total, parasites with a Pvcs bearing the VK210 type repeats only were found in 90 isolates (representing seven size polymorphisms). Parasites with a Pvcs bearing the VK247-type repeats only were found in nine isolates (representing only three size polymorphisms), and in one isolate parasites of both types were found. Thus, 10 different allelic forms of Pvcs were detected by simple analyses of the fragment size and repeat type. When associated with RFLP analysis of the pre- and post-repeat types, these could be divided into 23 different allelic types (Table 2), 18 for the VK210 type and five for the VK247 type. Mixed genotype infections were found in 20 of the isolates, for each of which the multiplicity of infection (MOI), i.e. the total number of different allelic types observed, was two. Thus, for the samples analysed, a total of 120 bands were observed, with an mean MOI of 1.2. The allelic variants were generally randomly distributed between the 120 bands; the highest frequencies, 0.2 and 0.18, were found for VK210k and VK210n, respectively (Fig 2).
Table 2 Frequency of Pvcs allelic variants classed by size, repeat type and presence of pre- and post-repeat insertions.
Allele Sizea Pre-repeatb Post-repeatb n Frequencyc
VK210a A Yes No 1 0.008
VK210b B Yes No 8 0.067
VK210c C Yes No 7 0.058
VK210d C No No 1 0.008
VK210e C Yes Yes 1 0.008
VK210f D Yes No 24 0.200
VK210g D No No 10 0.083
VK210h D No Yes 4 0.033
VK210i E Yes No 22 0.183
VK210j E No No 7 0.058
VK210k E No Yes 3 0.025
VK210l F No No 2 0.017
VK210m F No Yes 1 0.008
VK210n F Yes No 2 0.017
VK210o B No No 2 0.017
VK210p B No Yes 2 0.017
VK210q G No No 1 0.008
VK210r C No Yes 1 0.008
VK247a B In ND 3 0.025
VK247b C In ND 2 0.017
VK247c D In ND 4 0.033
VK247d B No ND 1 0.008
VK247e E No ND 1 0.008
a The fragments were assigned to different size bins (labelled A to G in decreasing order) by visual inspection.
b "Yes" and "No" respectively denote the presence or absence of insertions in the pre- and post-repeat regions as ascertained by PCR-RFLP analysis.
c Calculated for the 120 bands observed in the 100 Thai isolates analysed.
Figure 2 Allele frequency of the distinct allelic variants of Pvcs observed in the 100 isolates from Thailand. Allelic types were defined according to repeat type (VK210 or VK247), fragment size and the presence or absence of defined pre- and post-repeat sequences.
Genotyping using Pvmsp1
The Pvmsp1 gene encodes a polypeptide of about 1,720 amino acids [25,26], and sequence comparison revealed 13 regions of interallele conserved blocks and variable blocks [27]. Three main regions of sequence divergence were found through comparison of the full length Pvmsp1 sequences from two distinct P. vivax lines (Sal-1 and Belem). Three segments (labelled F1 to F3) corresponding to these regions were thus amplified for further analysis (Fig. 3A).
Figure 3 A. Schematic representation of the Pvmsp1 gene for the localization of interallele conserved blocks (blank boxes) and variable blocks (black boxes). The position of the three amplified segments (F1, F2 and F3) is indicated by the horizontal bold lines. B. Gel electrophoresis of PCR products of a selection of fragments, corresponding to the three segments, amplified from different isolates. The molecular size of the largest and smallest bands are indicated. C. Analysis of the diversity of the F2 segment by PCR-RFLP using Alu I or Mnl I. A selection of lanes is labelled to indicate the type of variant observed (corresponding to the nomenclature in Table 3). A 100 bp ladder was used a molecular weight marker (M) for all the gels.
For the 100 Thai P. vivax isolates, five distinguishable size variants were observed for F1 (ca. 350 bp – 450 bp), and four for F3 (ca. 250 bp – 350 bp), but only two were observed for the larger (ca. 1087 bp -1150 kb) F2 segment (Fig. 3B). One allelic variant dominated in frequency for each of the three amplified segments (Table 3): band C for F1 (42 % of the bands observed), band A for F2 (78 %) and band D for F3 (76 %). Mixed genotype infections were observed infrequently and only for the F1 (n = 3, 2 MOI of 2 and 1 MOI of 3) and for the F3 (n = 6, 4 MOI of 2 and 2 MOI of 3) segments.
Table 3 Frequency of Pvmsp1 allelic variants found in segments F1, F2 and F3.
F1 F2 F3
Variant n Frequency Varianta n Frequency Variant n Frequency
A 4 0.038 Aa1 50 0.5 A 1 0.009
B 19 0.183 Aa3 6 0.06 B 12 0.112
C 44 0.423 Aa6 15 0.15 C 12 0.112
D 22 0.212 Aa7 4 0.04 D 82 0.766
E 15 0.144 Aa8 1 0.01
Aa9 1 0.01
Aa10 1 0.01
Ba1 6 0.06
Ba2 1 0.01
Ba3 5 0.05
Ba5 2 0.02
Ba6 4 0.04
Ba7 3 0.03
Ba10 1 0.01
Am1 38 0.38
Am2 6 0.06
Am3 2 0.02
Am4 4 0.04
Am5 11 0.11
Am6 2 0.02
Am8 14 0.14
Am9 1 0.01
Bm1 6 0.06
Bm2 3 0.03
Bm3 7 0.07
Bm4 2 0.02
Bm5 3 0.03
Bm8 1 0.01
a RFLP patterns obtained by digestion with Alu I or Mnl I were numbered in two series a1, a2, etc.. and m1, m2 etc..., respectively. A given pattern could be obtained irrespective of the size of the fragment amplified (A or B).
A selection of amplified product from F1 (n = 18) and F3 (n = 8) representative of all the size variants were sequenced and compared to previously published sequences (Fig. 4). The number of distinct F1 variants was found to be higher than that revealed by electrophoretic separation, since 11 different allelic variants were observed for the 18 fragments derived from the Thai isolates. Sequence differences were observed within the variants assigned to the B, D or E size classes. Some of the bands within each class might be distinguished by the use of higher resolution agarose gels, though others had the same size and only exhibited subtle sequence differences. A similar pattern was observed for the F3 fragments, where sequencing distinguished six different variants.
Figure 4 Alignment of amino acid sequences of amplified F1 and F3 fragments of Pvmsp1. Sequences were aligned against the largest fragment obtained; dots represent identical residues and dashes represent gaps. Sequences obtained during this study are labelled by their GenBank accession numbers, while the Belem and Sal-1 sequences had been previously published (AF435594 and AF435593, respectively).
The F2 segment, which was comparatively poorly polymorphic in size in the Thai isolates, encompassed regions established as polymorphic in previous studies [27]. An RFLP strategy was thus adopted to distinguish between the different allelic variants. The restriction endonucleases Alu I and Mnl I which recognize multiple sites in F2, were used to reveal extensive polymorphism at the nucleotide level (Fig. 3C). Thus, nine different Alu I and eight different Mnl I RFLP patterns were observed in the 100 Thai isolates (Table 3). The occurrence of patterns indicative of partial digestion, or the presence of mixed F2 genotypes in individual samples were excluded since the sum of the RFLP fragments' sizes was not found to be greater than that of the uncut product for any isolate and the patterns observed were not unique to isolates where mixed genotypes were detected through size polymorphism of the F1 or F3 segments. When the data from both analyses (size and RFLP) were combined, 36 Pvmsp1 F2 allelic variants could be differentiated (Table 4). When the frequency of the allelic variants was considered individually for the two RFLP patterns, 50 % and 38 % of the isolates were of a single Alu I- and Mnl I-classified allelic variant, respectively. The remaining isolates were randomly distributed between the different variants (Fig. 5A). However, by combining the results for both restriction enzymes, the most dominant allele was found in only 27 % of the isolates (Fig. 5B), while the remainder were found at lower frequency (5 % or less).
Table 4 Frequency of the F2 Pvmsp1 allelic variants classed following RFLP analysis.
Variant Size Alu I Pattern number Mnl Pattern number na Frequency
Aa A 1 1 27 0.27
Ab A 1 2 5 0.05
Ac A 1 4 3 0.03
Ad A 1 5 1 0.01
Ae A 1 6 1 0.01
Af A 1 8 12 0.12
Ag A 1 9 1 0.01
Ah A 3 1 4 0.04
Ai A 3 2 1 0.01
Aj A 3 4 1 0.01
Ak A 6 1 3 0.03
Al A 6 3 2 0.02
Am A 6 5 8 0.08
An A 6 6 1 0.01
Ao A 6 8 1 0.01
Ap A 7 1 2 0.02
Aq A 7 5 2 0.02
Ar A 8 1 1 0.01
As A 9 1 1 0.01
At A 10 8 1 0.01
Ba B 1 1 2 0.02
Bb B 1 2 1 0.01
Bc B 1 3 1 0.01
Bd B 1 4 1 0.01
Be B 1 8 1 0.01
Bf B 2 1 1 0.01
Bg B 3 1 2 0.02
Bh B 3 2 1 0.01
Bi B 3 3 1 0.01
Bj B 3 4 1 0.01
Bk B 5 1 1 0.01
Bl B 5 2 1 0.01
Bm B 6 3 4 0.04
Bn B 7 3 1 0.01
Bo B 7 5 2 0.02
Bp B 10 5 1 0.01
a Number of isolates where a particular pattern was observed amongst the 100 Thai isolates analysed
Figure 5 Allele frequency of the distinct allelic variants of the F2 segment of Pvmsp1 observed in the 100 isolates from Thailand. Allelic variants were defined according to the RFLP patterns observed. A. Variants were divided according to the digestions patterns obtained individually with each of the two restriction enzymes Alu I (a1, a2, etc...) or Mnl I (m1, m2, etc...) for A or B, the two different sized fragments amplified, as in Table 3. B) Variants were classed according to size and the combined RFLP patterns obtained for both restriction enzymes (Alu I and Mnl I), as in Table 4.
Two locus genotyping
When the genotype analyses from the four polymorphic regions of the two genes were combined, the 100 isolates collected from Thai patients proved to be highly genetically diverse. Mixed genotype infections were detected in 26 of the isolates, mainly through the Pvcs gene (n = 20). Only one of the isolates identified as mixed genotype infections by Pvmsp1 (n = 7) was classed as such by Pvcs analysis. The mean MOI for isolates with a mixed genotype was 2.11, and that for the complete set of isolates was 1.29.
There were 74 isolates where a single genotype was detected after full analysis of the Pvcs and Pvmsp1 markers. Comparison of the genotyping patterns obtained for these isolates allows to ascertain the relative contribution of the polymorphic loci to distinguish between P. vivax populations. When the complete genotyping analyses are taken into account, a total of 68 distinct genotypes are enumerated. Sixty-three genotypes were observed in only one isolate, with a maximum of three isolates sharing the same genotype (highest genotype frequency = 0.040). No evidence for linkage disequilibrium could be detected when Pvcs and Pvmsp1 were considered. If the results from the Pvmsp1 F1 segment are omitted from the analysis, 58 distinct genotypes would be found, with a maximum of four isolates sharing the same genotype (highest genotype frequency = 0.054). Incremental omission of the Pvmsp1 F3 segment followed by that of the Pvcs pre- and post-repeat PCR-RFLP results would reduce the total number of distinct genotypes to 55 with four isolates sharing the same genotype (highest genotype frequency = 0.054), and to 49 with seven isolates sharing the same genotype (highest genotype frequency = 0.095), respectively.
Discussion
Populations of P. vivax, like those of other Plasmodium species, comprise genetically distinct lines that exhibit diversity in a number of factors of epidemiological, biological and clinical relevance. For example, the proportion of hypnozoites and the timing and frequency of their activation, differ between temperate and tropical strains of this parasite [28]. Resistance to drugs is linked to genetic mutations of defined genes [29] and recently the Pvcs repeat type was found to be associated with transmissibility by defined species of anophelines [30]. The ability to differentiate genetically different parasite populations would enhance a wide spectrum of investigations of the biology and epidemiology of P. vivax.
In this article, the potential of two P. vivax genes, Pvcs and Pvmsp1, as genetic markers, were assessed and practical protocols for their use in genotyping parasites collected from the field were described. PCR-based protocols targeted four polymorphic regions from two genetic markers, one from Pvcs and three from Pvmsp1, and were further associated to RFLP analyses of two of these regions, the repeat region of Pvcs and the F2 fragment of Pvmsp1, that allow the division of parasite populations into an incremental number of genetically distinct sub-groups.
For Pvmsp1, three polymorphic regions were independently assessed. The fragments F1, located at variable block 2 and F3 located at variable block 10, displayed moderate allelic size variation, five and four types respectively. The frequency of the size variants, in particular those of F3, displayed a biased frequency distribution among the 100 Thai isolates considered in this study, thus undermining their usefulness as genetic markers. Sequencing was carried out for a subset of F1 and F3 fragments amplified from the Thai isolates. This revealed that 13 F1 and 10 F2 allelic variants could actually be distinguished. These results confirm the high degree of complex polymorphisms observed for Pvmsp1 block 2 and block 10 observed in 31 isolates collected in Thailand, Brazil, Oceania and India, where 19 and 13 distinct allelic types were observed for block 2 and block 10 respectively [27]. Thus it is likely that the F1 and F3 segments could be exploited as useful genetic markers through the development of type-specific oligonucleotides, that would then be used in a series of nested PCR reactions. The amplified large central F2 fragment, located between variable block 6 and 8, showed two size variants only. However, when combined to RFLP analyses, using two restriction endonucleases, the variants were subdivided into a total of 36 different allelic types. For Pvcs, populations were divided into 10 subgroups through size and repeat type determination, and into 23 subgroups when sequence variations in the pre- and post-repeat sequences were assessed by RFLP.
The conclusion of this work is that the polymorphic repeat region of Pvcs and that of the Pvmsp1 F2 region can be considered suitable genetic markers of P. vivax populations, alone or in combination. This is supported by the fact that practical PCR-RFLP genotyping protocols revealed the presence of numerous distinct allelic variants in a sample of 100 isolates, among which they were randomly distributed without evidence of linkage disequilibrium between the two loci.
Interesting epidemiological indications could be derived from genotyping of the P. vivax isolates. Malaria is considered to be hypoendemic in Thailand where transmission, confined to regions bordering Cambodia, Lao PDR and Myanmar, rarely exceeds a few infective bites per person per year. A study of P. falciparum diversity, based on three genetic markers, in a refugee camp located in Thailand close to the Myanmar border, revealed levels of diversity and MOI (1.6, with mixed genotype infections observed in 60 % of the isolates) that were higher than expected from the low effective inoculation rates [31]. A recent study of P. vivax parasites collected in the same region and based on two genetic markers, showed that populations of this parasite species were also highly diverse and mixed genotype infections were observed in 35 % of the isolates [32]. The results of the present study indicated that 26 % of the isolates were of mixed genotype, with an overall MOI for all 100 isolates of 1.29. It should be stated that the three studies are not strictly comparable since the number and type of genetic markers, the genotyping protocols and the collection strategies that were employed differed. Furthermore, in the current study the origin of the P. vivax isolates was not confined to a single area, since the patients from whom blood was collected acquired their infection in diverse endemic areas of Thailand. Nonetheless, these studies indicated that the two biologically distinct parasite species have similar population characteristics. The production of infectious gametocytes early during the primary infection in P. vivax versus their late appearance, following the acute phase, in P. falciparum, and the existence of hypnozoites solely in P. vivax, might not be sufficient to explain the maintenance of diversity in areas of low transmission. Substantially prolonged duration of the infections, probably as a result of low level drug resistance leading to parasitological (though not clinical) failures, as suggested for P. falciparum [31], and/or an underestimation of the transmission intensity, might underlie the high genetic diversity and MOI observed for the two parasite species. These two non-mutually exclusive scenarios are consistent with the detection of a high level of mixed species infections observed in Thailand [33-36].
It is hoped that the methodologies presented here can be adopted as standard protocols for the genotyping of P. vivax parasites, not only for in vivo drug efficacy trials, but also in field-based investigations aimed at elucidating the biology, pathology and epidemiology of this important parasite species.
Conclusion
These results indicate that the genotyping protocols presented can be useful in the assessment of in vivo drug efficacy clinical trials conducted in endemic areas and for epidemiological studies of P. vivax infections.
List of Abbreviations used
Pvcs-Circumsporozoite surface protein gene of Plasmodium vivax
Pvmsp1-Merozoite surface protein 1 gene of Plasmodium vivax
Pvmsp3α-Merozoite surface protein 3 alpha gene of Plasmodium vivaxPCR-Polymerase Chain Reaction
RFLP-Restriction Fragment Length Polymorphism
Authors' contributions
GS, NJW and SP designed the study. GS and SP were responsible for the day-to-day supervision of the work. ACG, FL and LR and MI collaborated to obtain and analyze DNA sequences. SL and SP were responsible for patient recruitment and clinical management. MI developed the protocols with the help of GS, and carried out the vast majority of the laboratory work. GS and MI analyzed the data and composed the manuscript. All authors read and approved the final manuscript.
Acknowledgements
This study was part of the Wellcome Trust-Mahidol University-Oxford Tropical Medicine Research Programme supported by the Wellcome Trust of Great Britain.
==== Refs
Mendis K Sina BJ Marchesini P Carter R The neglected burden of Plasmodium vivax malaria Am J Trop Med Hyg 2001 64 97 106 11425182
Hay SI Guerra CA Tatem AJ Noor AM Snow RW The global distribution and population at risk of malaria: past, present, and future Lancet Infect Dis 2004 4 327 336 15172341 10.1016/S1473-3099(04)01043-6
Nosten F McGready R Simpson JA Thwai KL Balkan S Cho T Hkirijaroen L Looareesuwan S White NJ Effects of Plasmodium vivax malaria in pregnancy Lancet 1999 354 546 549 10470698 10.1016/S0140-6736(98)09247-2
Trape JF The public health impact of chloroquine resistance in Africa Am J Trop Med Hyg 2001 64 12 17 11425173
Rieckmann KH Davis DR Hutton DC Plasmodium vivax resistance to chloroquine? Lancet 1989 2 1183 1184 2572903 10.1016/S0140-6736(89)91792-3
Baird JK Basri H Purnomo Bangs MJ Subianto B Patchen LC Hoffman SL Resistance to chloroquine by Plasmodium vivax in Irian Jaya, Indonesia Am J Trop Med Hyg 1991 44 547 552 1676566
Garavelli PL Corti E Chloroquine resistance in Plasmodium vivax: the first case in Brazil Trans R Soc Trop Med Hyg 1992 86 128 1440766 10.1016/0035-9203(92)90535-K
Thaithong S Chan SW Songsomboon S Wilairat P Seesod N Sueblinwong T Goman M Ridley R Beale G Pyrimethamine resistant mutations in Plasmodium falciparum Mol Biochem Parasitol 1992 52 149 157 1620155 10.1016/0166-6851(92)90047-N
Myat Phone K Myint O Myint L Thaw Z Kyin Hla A Nwe Nwe Y Emergence of chloroquine-resistant Plasmodium vivax in Myanmar (Burma) Trans R Soc Trop Med Hyg 1993 87 687 8296378 10.1016/0035-9203(93)90294-Z
Murphy GS Basri H Purnomo Andersen EM Bangs MJ Mount DL Gorden J Lal AA Purwokusumo AR Harjosuwarno S Vivax malaria resistant to treatment and prophylaxis with chloroquine Lancet 1993 341 96 100 8093414 10.1016/0140-6736(93)92568-E
Garg M Gopinathan N Bodhe P Kshirsagar NA Vivax malaria resistant to chloroquine: case reports from Bombay Trans R Soc Trop Med Hyg 1995 89 656 657 8594687 10.1016/0035-9203(95)90432-8
Baird JK Leksana B Masbar S Fryauff DJ Sutanihardja MA Suradi Wignall FS Hoffman SL Diagnosis of resistance to chloroquine by Plasmodium vivax: timing of recurrence and whole blood chloroquine levels Am J Trop Med Hyg 1997 56 621 626 9230792
Pukrittayakamee S Chantra A Simpson JA Vanijanonta S Clemens R Looareesuwan S White NJ Therapeutic responses to different antimalarial drugs in vivax malaria Antimicrob Agents Chemother 2000 44 1680 1685 10817728 10.1128/AAC.44.6.1680-1685.2000
Golenda CF Li J Rosenberg R Continuous in vitro propagation of the malaria parasite Plasmodium vivax Proc Natl Acad Sci U S A 1997 94 6786 6791 9192643 10.1073/pnas.94.13.6786
Chotivanich K Silamut K Udomsangpetch R Stepniewska KA Pukrittayakamee S Looareesuwan S White NJ Ex-vivo short-term culture and developmental assessment of Plasmodium vivax Trans R Soc Trop Med Hyg 2001 95 677 680 11816444 10.1016/S0035-9203(01)90113-0
Snounou G Beck HP The use of PCR-genotyping in the assessment of recrudescence or reinfection after antimalarial treatment. Parasitology today 1998 14 462 467 17040849 10.1016/S0169-4758(98)01340-4
Cui L Escalante AA Imwong M Snounou G The genetic diversity of Plasmodium vivax populations Trends Parasitol 2003 19 220 226 12763428 10.1016/S1471-4922(03)00085-0
Imwong M Pukrittakayamee S Looareesuwan S Poirriez J Pasvol G White NJ Snounou G Plasmodium vivax: polymerase chain reaction amplification artifacts limit the suitability of pvgam1 as a genetic marker Exp Parasitol 2001 99 175 179 11846528 10.1006/expr.2001.4646
Bruce MC Galinski MR Barnwell JW Snounou G Day KP Polymorphism at the merozoite surface protein-3alpha locus of Plasmodium vivax: global and local diversity Am J Trop Med Hyg 1999 61 518 525 10548283
Snounou G Viriyakosol S Zhu XP Jarra W Pinheiro L do Rosario VE Thaithong S Brown KN High sensitivity of detection of human malaria parasites by the use of nested polymerase chain reaction Mol Biochem Parasitol 1993 61 315 320 8264734 10.1016/0166-6851(93)90077-B
Mann VH Huang T Cheng Q Saul A Sequence variation in the circumsporozoite protein gene of Plasmodium vivax appears to be regionally biased Mol Biochem Parasitol 1994 68 45 52 7891747 10.1016/0166-6851(94)00148-0
Rongnoparut P Supsamran N Sattabongkot J Suwanabun N Rosenberg R Phenotype and genotype diversity in the circumsporozoite proteins of Plasmodium vivax in Thailand Mol Biochem Parasitol 1995 74 201 210 8719161 10.1016/0166-6851(95)02504-9
Rosenberg R Wirtz RA Lanar DE Sattabongkot J Hall T Waters AP Prasittisuk C Circumsporozoite protein heterogeneity in the human malaria parasite Plasmodium vivax Science 1989 245 973 976 2672336
Mann VH Good MF Saul AJ Diversity in the circumsporozoite protein of Plasmodium vivax: Does it matter? Parasitology today 1995 11 33 36 10.1016/0169-4758(95)80107-3
del Portillo HA Longacre S Khouri E David PH Primary structure of the merozoite surface antigen 1 of Plasmodium vivax reveals sequences conserved between different Plasmodium species Proc Natl Acad Sci U S A 1991 88 4030 4034 2023952
Gibson HL Tucker JE Kaslow DC Krettli AU Collins WE Kiefer MC Bathurst IC Barr PJ Structure and expression of the gene for Pv200, a major blood-stage surface antigen of Plasmodium vivax Mol Biochem Parasitol 1992 50 325 333 1371329 10.1016/0166-6851(92)90230-H
Putaporntip C Jongwutiwes S Sakihama N Ferreira MU Kho WG Kaneko A Kanbara H Hattori T Tanabe K Mosaic organization and heterogeneity in frequency of allelic recombination of the Plasmodium vivax merozoite surface protein-1 locus Proc Natl Acad Sci U S A 2002 99 16348 16353 12466500 10.1073/pnas.252348999
Garnham PC Bray RS Bruce-Chwatt LJ Draper CC Killick-Kendrick R Sergiev PG Tiburskaja NA Shute PG Maryon M A strain of Plasmodium vivax characterized by prolonged incubation: morphological and biological characteristics Bull World Health Organ 1975 52 21 32 764993
Imwong M Pukrittayakamee S Renia L Letourneur F Charlieu JP Leartsakulpanich U Looareesuwan S White NJ Snounou G Novel point mutations in the dihydrofolate reductase gene of Plasmodium vivax: evidence for sequential selection by drug pressure Antimicrob Agents Chemother 2003 47 1514 1521 12709316 10.1128/AAC.47.5.1514-1521.2003
Gonzalez-Ceron L Rodriguez MH Santillan F Chavez B Nettel JA Hernandez-Avila JE Kain KC Plasmodium vivax: ookinete destruction and oocyst development arrest are responsible for Anopheles albimanus resistance to circumsporozoite phenotype VK247 parasites Exp Parasitol 2001 98 152 161 11527438 10.1006/expr.2001.4626
Paul RE Hackford I Brockman A Muller-Graf C Price R Luxemburger C White NJ Nosten F Day KP Transmission intensity and Plasmodium falciparum diversity on the northwestern border of Thailand Am J Trop Med Hyg 1998 58 195 203 9502604
Cui L Mascorro CN Fan Q Rzomp KA Khuntirat B Zhou G Chen H Yan G Sattabongkot J Genetic diversity and multiple infections of Plasmodium vivax malaria in Western Thailand Am J Trop Med Hyg 2003 68 613 619 12812356
Mayxay M Pukritrayakamee S Chotivanich K Imwong M Looareesuwan S White NJ Identification of cryptic coinfection with Plasmodium falciparum in patients presenting with vivax malaria Am J Trop Med Hyg 2001 65 588 592 11716119
Siripoon N Snounou G Yamogkul P Na-Bangchang K Thaithong S Cryptic Plasmodium falciparum parasites in clinical P. vivax blood samples from Thailand Trans R Soc Trop Med Hyg 2002 96 70 71 11925999 10.1016/S0035-9203(02)90246-4
Mayxay M Pukrittayakamee S Newton PN White NJ Mixed-species malaria infections in humans Trends Parasitol 2004 20 233 240 15105024 10.1016/j.pt.2004.03.006
Snounou G White NJ The co-existence of Plasmodium: sidelights from falciparum and vivax malaria in Thailand Trends Parasitol 2004 20 333 339 15193565 10.1016/j.pt.2004.05.004
| 15854233 | PMC1131918 | CC BY | 2021-01-04 16:37:31 | no | Malar J. 2005 Apr 27; 4:20 | utf-8 | Malar J | 2,005 | 10.1186/1475-2875-4-20 | oa_comm |
==== Front
Med ImmunolMedical Immunology1476-9433BioMed Central London 1476-9433-4-61587781610.1186/1476-9433-4-6CommentaryThe continuing HIV vaccine saga: naked emperors alongside fairy godmothers Smith Kendall A [email protected] The Division of Immunology, Department of Medicine, Weill Medical College, Cornell University, New York, NY USA2005 6 5 2005 4 6 6 4 5 2005 6 5 2005 Copyright © 2005 Smith; licensee BioMed Central Ltd.2005Smith; 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 latest developments in the HIV vaccine field were aired at a Keystone Symposium recently. This Commentary summarizes some of the highlights from this meeting, and focuses on some of the developments that appeared particularly promising, as well as those that do not. Unfortunately, the "saga" continues.
==== Body
Introduction
Investigators working in the HIV vaccine field congregated at a Keystone Symposia meeting that was held in Banff, Canada April 9–15, 2005. David Montefiori, and Kent Weinhold, both from Duke University, and Carolyn Williamson, from the University of Cape Town chaired the meeting, which was held concomitantly with a meeting on HIV Pathogenesis, chaired by Michael Lederman from Case Western University, Richard Koup from NIAID/NIH, and Michael Malim, from King's College London.
The meetings were noteworthy for new insights into the structure-activity relationships of the viral envelope, additional characterizations of the functional aspects of both CD4+ and CD8+ T cells in successful immune responses to HIV and other viruses and vaccines, as well as functional defects of these same cells from subjects with persistent viremia and progressive disease.
Against this scientific backdrop, was a thrilling presentation by Judy Lieberman of Harvard Medical School, who appears to have solved the problem of delivery of small-interfering RNA (siRNA) to specific target cells in vivo.
However, the various academic groups and companies who are competing to develop the first effective prophylactic HIV vaccines created the drama of the meeting, which was played out in several presentations and posters over the course of the 6-day meeting. Accordingly, this commentary will focus on these reports and developments in the hopes of providing a flavor of the proceedings to those who were unable to attend.
The envelope
Bing Chen from Children's Hospital Laboratory of Molecular Medicine of Harvard Medical School presented data from a collaborative effort by his group and by Stephen Harrison's group at Harvard Medical School and by John Skehel from the National Institute of Medical Research The Ridgeway, Mill Hill, London. They have also recently published their findings [1,2]. As these investigators pointed out, structures of fragments of gp120 and gp41 from the envelope glycoprotein have been known for some time, in conformations corresponding to their states after attachment to CD4 molecules and after membrane fusion. By comparison, these investigators determined the crystal structure, at 4.0 Å resolution, of a fully glycosylated and unliganded SIV gp120 core protein, in a conformation representing its prefusion state, prior to interaction with CD4. Comparison of the new structure and the HIV gp120 core in the CD4-bound conformation revealed a striking structural rearrangement in parts of the protein, resulting in distinct antigenic surfaces. Their model predicts that upon binding to CD4, parts of gp120 will shift around the CD4-binding cavity with very large excursions, e.g. the tip of the V1-V2 stem moves over 40 Å. Their model predicts that until co-receptor binds, the V1-V2 stem is not docked against the β20-β21 ribbon from the outer domain, and thus the bridging sheet, which binds to the co-receptor is not properly formed.
All of these structure-activity aspects of the envelope were graphically depicted in molecular movies, which allowed the uninitiated to appreciate the huge distances that parts of the molecules travel after binding CD4, to form the co-receptor binding site. Hopefully, these new insights into the dynamic structure and activity of the envelope will provide for new approaches so as to craft molecular immunogens that will promote the generation of neutralizing antibodies.
T Cell Exhaustion
There is a consensus among almost all investigators now, that persistent viral infection leads to a qualitative defect in both CD4+ and CD8+ T cells, which is manifest by an incapacity to produce cytokines, especially IL2, when activated in vitro by viral peptides. By comparison, cells from the same individual can respond fully and appropriately to other antigens to which the individual is immune, e.g. antigens from cytomegalovirus (CMV) and Epstein-Barr virus (EBV). The consequence of the inability to produce IL2 is a poor proliferative response and an inability to differentiate into an "effector" capacity, whether monitored by cytokine/chemokine production or by cytolytic capacity.
Michael Betts from the Vaccine Research Center at NIH monitored cytokine production by 11-parameter flow cytometry to examine five antigen-specific CD8+ T cell functions simultaneously (degranulation-CD107a; cytokine/chemokine expression-INF-γ, TNF-α, IL2, MIP1β) in 9 HIV-infected Long-Term Nonprogressors (LTNP) and 79 Progressors. The LTNP maintained a polyfunctional CD8+ T cell response, expressing most of the gene products assayed, while the progressors were markedly deficient. As well, the polyfunctional response was found to be a normal component of effector CD8+ T cell responses to CMV, EBV, and influenza virus in normal individuals. Very noteworthy was the finding that the polyfunctional responses were not confined to a particular surface phenotype, whether thought to reflect a "central memory" phenotype or not. These findings may well augur the end of phenotypic analysis as a surrogate for lymphocyte function.
Giuseppe Pantaleo seconded these findings, and emphasized that the lack of control of viral replication is not due to the lack of HIV-specific CD4+ or CD8+ T cells, as originally proposed. Instead, Pantaleo's data mirrored Betts' data, and he stressed that the capacity to produce IL2 correlates best with a good immune response, whether to HIV or to CMV or EBV, by either HIV-infected or uninfected individuals.
Accordingly, the $64 question still unanswered is why there seems to be such a specific defect in HIV-specific T cells and why it is manifest by the incapacity to express the IL2 genes. However, all of these data point clearly to where research must focus in the future, for not until this defect is understood will it be approachable therapeutically. It remains to be determined whether IL2 administration can circumvent the defect in IL2 production by HIV-specific CD4+ and CD8+ T cells.
In this regard, Mark Connors from NIAID/NIH reported on further data from his group indicating that in addition to a defect in IL2 production, CD8+ T cells from Progressors cannot respond to IL2 by proliferating in vitro in response to stimulation by HIV antigens. He emphasized that 10% to 40% of circulating CD8+ T cells are HIV-reactive when individuals are viremic, which is a huge proportion of the total CD8+ T cell mass. Using HIV-peptide tetramers to identify HIV-specific cells, he could show that the CD8+ T cells from both LTNP and Progressors are CD25-, and gene expression in these cells as monitored by DNA arrays are identical when studied de novo, before activation. However, when antigen activated, even the addition of exogenous IL2 does not cause CD8+ T cells from Progressors to divide. He suggested that his data point to a defect in IL2-responsiveness, which could be attributable to changes in the IL2 receptor chains, the IL2R signaling pathways, or in the expression of IL2-regulated genes, in particular the IL2-reguated genes that promote cell cycle progression. He further indicated that such qualitative defects in the virus-specific CD8+ T cell response remains the most likely determinant of the loss of immunological restriction of virus replication. As such, his data indicate that a further understanding of these qualitative features may provide important information on how these changes might be avoided in vaccinees or reversed in infected individuals.
RNAi
Judy Lieberman presented breathtaking data, which indicated that it may be possible to use the siRNA technology to create effective microbicides, and as well to target only HIV-infected cells systemically in vivo, by formulating magic bullets to seek out and suppress infected cells. If her data hold up and can be repeated and substantiated, it is a quantum leap forward, not only for the prevention and therapy of HIV infection, but for almost all aspects of biomedical research and medicine.
As explained by Lieberman, RNAi is an ancient, evolutionarily conserved host defense against viruses and transposable elements, which uses small double-stranded RNAs, called small interfering RNAs (siRNA), to silence gene expression with high specificity by targeted degradation of homologous mRNAs [3]. Lieberman's group used a mouse model of Herpes Simplex Virus-2 (HSV-2) vaginal infection to develop an siRNA-based microbicide. When mixed with lipids, siRNAs are efficiently and uniformly taken up by mucosal and submucosal cells and effectively silence gene expression in the mouse vagina. siRNAs that individually silence a variety of HSV-2 genes and inhibit HSV-2 replication in vitro by 4-8-fold were identified. Vaginal installation of siRNAs targeting HSV-2 was well tolerated and protected mice from lethal HSV-2 infection.
In other experiments, Lieberman targeted HIV-infected cells with RNAi by using a chimeric molecule engineered to contain the antigen-binding region of an antibody reactive with gp120, and protamine, a positively charged molecule. When mixed with siRNA targeting HIV genes in aqueous solution, the negatively charged siRNA forms a stable complex with the antibody-protamine chimera. Subsequently she showed that this molecular complex would selectively bind to target cells that expressed gp120. Accordingly, the antibody is used to identify the correct cells and the protamine delivers the siRNA. The data presented by Lieberman suggest that it may now be possible to silence your gene of choice, thereby changing gene therapy from a discipline focused on expressing missing genes, to silencing unwanted gene expression! A paper is in press that describes these remarkable findings [4].
The HIV vaccine pipeline (The Emperors and the Fairy Godmothers)
Standing in stark contrast to these promising developments on the basic science scene, were disappointing, perplexing, but then promising developments on the more practical aspects of developing effective vaccines.
Andrew McMichael from Oxford supported by the International AIDS Vaccine Initiative (IAVI) was perhaps one of the first entrants into to the race to develop an effective vaccine. Most everyone now knows that Oxford/IAVI discontinued their large phase I/II trial in Kenya last year. McMichael reviewed the reasons for this decision for the participants of the Keystone conference. The Oxford group chose a combination of a DNA prime immunization, followed by a booster immunization with Modified Vaccinia Ankara (MVA). At the time that this approach was planned, now several years ago, the prevailing opinion was that it is important to select "immunodominant" epitopes to include in the vaccines. Thus, both the DNA priming vaccine and the MVA boost vaccine only contained gag p24, p17, and a "string" of selected CTL epitopes from other regions of the HIV genome.
McMichael reported that the responses in the Kenyon trial were disappointing, with only ~20% of vaccine recipients giving a positive immune response to the immunization in a highly stringent validated IFN-γ ELISPOT assay. He went on to say that perhaps a better assay to detect immune reactivity to the vaccine antigens could be obtained by culturing the PBMCs from vaccinated people in vitro with HIV peptides and IL2 for 7–14 days before testing for IFN-γ producing cells by ELISPOT. He suggested that perhaps this "reactivated response" would be more representative of long-term memory. This may be true, but this type of "reactivated response" precludes the capacity to quantify the frequencies of HIV antigen-reactive T cells from a subject, because after 1–2 weeks of selective expansion of antigen-reactive cells in IL2-dependent growth, and the demise of antigen nonreactive cells, any attempt at quantification is fruitless.
Searching for possible reasons as to why this immunization strategy failed, the most probable are the lack of breadth of antigens included in the vectors, and the fact that both vectors are replication incompetent. For a discussion of replication competent vs. incompetent vaccines, see [5].
Perhaps the next entrant into the HIV vaccine race was the Merck Corporation, which also adopted a DNA priming concept, followed by a boost with replication incompetent Adenovirus serotype 5 (Ad5). Merck elected to include only genes encoding the gag polyprotein in their vectors. John Shiver summarized their findings when immunizing normal volunteers. Whether assayed by ELISPOT or by flow cytometry monitoring the accumulation of intracellular IFN-γ, the immune responses were almost undetectable, with frequencies of only 0.1%–0.2% by flow cytometry (1000–2000 spot-forming cells/million PBMCs by ELISPOT assay). Because these normal volunteers cannot be challenged with HIV to test their antiviral immunity, it remains unknown whether these immune responses will reflect protective immunity or not. However, at best these immune responses are just at the Lower Limit of Detection (LLD) by the flow cytometry assay. As well, it is known the majority of the people in the US are already immune to the Ad5 serotype, and to detect even marginal immune responses to the immunogen, a huge dose had to be given, 100 billion virus particles. It is not clear why Merck is pursuing this line of investigation, given these marginal results. Evidently there is no one at Merck willing to stand up and say that the Emperor has no clothes [6].
The last entrant into the HIV vaccine race that reported on data in humans was the Vaccine Research Center (VRC) of the NIAID/NIH, which is directed by Gary Nabel. The VRC has adopted an approach very similar to Merck, with a DNA prime followed by an Ad5 boost. However, in contrast to Merck, the VRC has placed more HIV genes into their vaccines. They are using a separate DNA plasmid for the env genes from clades A, B, and C. Then a separate plasmid contains the genes encoding the clade B gag, pol, and nef genes. This immunization is followed by a combination of 4 Ad5 vectors encoding clade B gag and pol, and env from clades A, B, and C.
Barney Graham reported on the findings available to date. Remarkably, despite the addition of more HIV genes in the vectors, the immune responses were still barely detectable by flow cytometry, on the order of 0.1% to 0.2% positive cells by intracellular cytokine staining. Again, these data were presented as if one were dealing with strong immune responses, not the marginal immune responses that they actually represent. Accordingly, there was no mention of discontinuing this approach.
Given these marginal immune responses elicited in normal individuals, one wonders whether a "Gatekeeper" should be used, such as prophylactic trials in monkeys as was suggested by some at the meeting, or by obtaining an antiviral endpoint in HIV-infected humans, prior to undertaking large scale prophylactic trials in developing countries [5,7].
Fortunately there are some Fairy Godmothers waiting in the wings, anxious to wave their magic wands to take over after the Emperors retreat to hide their nakedness. Harriet Robinson's group from Emory University submitted 5 abstracts that resulted in as many posters. Of these reports, one of the most interesting was a study where they mapped and characterized the T cell responses associated with long-term control of a virulent SHIV-9.6P challenge in 22 macaques vaccinated with a Gag-Pol-Env DNA/MVA vaccine. Thus, they have employed an approach similar to the Oxford group, but they have included the 3 most important/prevalent genes in their vaccine. In addition, the MVA that they use is different from the Oxford MVA, in that it releases virus-like particles (VLP) [8]. Therefore, it expresses more of a viral replicative cycle, so that it should be more immunogenic. They tested for the frequency, breadth, and stability of anti-viral CD4+ and CD8+ T cells and for the protection of T cell function as evidenced by IFN-γ and IL2 co-production during a 200-week period of viral control (3.9 years). Even 2.5 years post challenge, IFN-γ-producing T cells were still detectable in frequencies up to 1.0% of total CD8+ or CD4+ T cells, which is 10-fold greater than the frequencies that the Oxford, Merck and VRC groups are reporting in humans. Moreover, these T cell responses were remarkably stable when tested a year later and maintained good function, as evidenced by the co-production of IFN-γ and IL2.
This group also presented data in humans on a phase I multi-center, randomized, placebo-controlled, double blind trial conducted under the auspices of the HVTN. A clade B DNA HIV vaccine developed for use as a priming vector for a DNA/MVA combination vaccine in humans was tested in 30 HIV-uninfected normal volunteers. This 9.5 kb DNA uses a CMV promoter and a bovine growth hormone polyadenylation sequence to express a single transcript that undergoes Rev-dependent subgenomic splicing and frame-shifting to express Gag, Pr, RT, Env, Tat, Rev and Vpu. In this phase I safety trial, doses of 0.3 mg and 3.0 mg were well tolerated. ELISPOT assays performed on frozen cells were negative after 2 immunizations. However, flow cytometry assays performed on fresh cells did detect some responses, although at low frequencies, 0.01–0.1%. Additional data from this group will be awaited with interest.
Perhaps the maverick of the group is Marjorie Robert-Guroff from the NCI. She has been developing replication-competent Adenoviruses as vectors for HIV vaccines [9]. In contrast to the Merck and VRC/NIH groups, she has focused her attention on two different serotypes, #s 4 and 7. The U.S. Army had used these two serotypes as vaccines for new recruits between 1972 and 1996, and as many as 10 million young men received these vaccines, apparently safely. The immunizations were discontinued in 1996 because the company manufacturing them ceased production. However, a large majority of young adults in the U.S have not been exposed to these serotypes, so that they are at risk of developing severe upper respiratory infections. Because of this, and because of the Iraqi war, the government is now planning to reintroduce these vaccines. Since most individuals are not already immune to these serotypes, they make for prime candidates for use as vectors for vaccines against other microbes, such as HIV. Robert-Guroff is planning initial clinical trials with Ad4-based HIV recombinants.
A comprehensive analysis of the replication-competent Ad vaccine approach in the HIV-chimpanzee and SIV-macaque models demonstrated that it elicits potent humoral, cellular, and mucosal immunity yielding strong, durable protective efficacy against pathogenic SIV. Comparing replication-competent with replication-incompetent Ad vectors, she demonstrated that 10-100-fold lower doses of replication-competent Ad vectors elicited significantly more, and longer-lasting IFN-γ-secreting cells in response to env peptides compared with replication-incompetent Ad vectors, and primed stronger T cell proliferative responses. As well, in a combined Ad-HIV recombinant priming/protein boosting approach using Chiron's oligomeric HIV gp140ΔV2 vaccine, enhanced humoral immunity was elicited with the replicating Ad recombinants, with higher anti-envelope binding and neutralizing antibody titers, and greater capacity to neutralize a broad spectrum of heterologous HIV R5-tropic primary isolates. Replicating Ad-recombinants also elicited antibodies that mediated significantly greater levels of antibody-dependent cellular cytotoxicity (ADCC), and she showed data that the ADCC activity correlated with antiviral protection after SIV challenge [10].
Robert-Guroff concluded by stating that their results should support continued development of the replicating Ad-HIV recombinant vaccine approach, and suggested that greater attention should be focused on replicating vectors in general. As well, she pointed out that replicating vectors have a strong adjuvant effect and elicit persistent immunity, as evidenced by almost all of the live attenuated replication competent viral vaccines that have already been licensed. Finally, she stressed that "the use of replication-competent vectors, alone or in combination with other replicating, or non-replicating vectors, coupled with well-designed envelope boosts, may provide the broad, potent immunity and rapid responsiveness necessary for an effective prophylactic HIV vaccine."
Perhaps the Emperors should take note when designing their next suit of clothes!
==== Refs
Chen B Vogan EM Gong H Skehel JJ Wiley DC Harrison SC Determining the structure of an unliganded and fully glycosylated SIV gp120 envelope glycoprotein Structure 2005 13 197 211 15698564 10.1016/j.str.2004.12.004
Chen B Vogan EM Gong H Skehel JJ Wiley DC Harrison SC Structure of an unliganded simian immunodeficiency virus gp120 core. Nature 2005 433 834 841 15729334 10.1038/nature03327
Dykxhoorn M Lieberman J The silent revolution: RNA interference as basic biology, research tool, and therapeuic. Ann Rev Med 2005 56 401 423 15660519 10.1146/annurev.med.56.082103.104606
Song E Zhu P Lee SK Chowdhury D Kussman S Dykxhoorn M Palliser D Weiner DB Shankar P Marasco WA Lieberman J Antibody-mediated in vivo delivery of small interfering RNAs via cell surface receptors Nature Biotechnology 2005 In press.
Smith KA The HIV vaccine saga Med Immunol 2003 2 1 12628020 10.1186/1476-9433-2-1
Andersen HC The Emperor's New Clothes
Smith KA Optimal clinical trial designs for immune-based therapies in persistent viral infections. Med Immunol 2002 1 4 12459051 10.1186/1476-9433-1-4
Wyatt LS Earl PL Liu JY Smith JM Montefiori DC Robinson HL Moss B Multiprotein HIV type 1 clade B DNA and MVA vaccines: connstruction, expression, and immunogenicity in rodents of the MVA component. AIDS Res Hum Retroviruses 2004 20 645 653 15242542 10.1089/0889222041217428
Malkevitch NV Robert-Guroff M A call for replicating vector prime-protein boost strategies in HIV vaccine design. Expert Rev Vaccines 2004 3(4 Suppl) S105 S117 15285710 10.1586/14760584.3.4.S105
Gomez-Roman VR Patterson LJ Venzon D Liewehr D Aldrich K Florese R Robert-Guroff M Vaccine-Elicited Antibodies Mediate Antibody-Dependent Cellular Cytotoxicity Correlated with Significantly Reduced Acute Viremia in Rhesus Macaques Challenged with SIVmac251 J Immunol 2005 174 2185 2189 15699150
| 15877816 | PMC1131919 | CC BY | 2021-01-04 16:39:09 | no | Med Immunol. 2005 May 6; 4:6 | utf-8 | Med Immunol | 2,005 | 10.1186/1476-9433-4-6 | oa_comm |
==== Front
Mol CancerMolecular Cancer1476-4598BioMed Central London 1476-4598-4-151586970610.1186/1476-4598-4-15ResearchClinical and biological characteristics of cervical neoplasias with FGFR3 mutation Rosty Christophe [email protected] Marie-Hélène [email protected] David [email protected] Jérôme [email protected] Isabelle [email protected] Jean Paul [email protected] Xavier [email protected] François [email protected] Département de Pathologie, Institut Curie, Section Médicale, 26 rue d'Ulm, 75248 Paris Cedex 05, France2 UMR 144, CNRS – Institut Curie, Section de Recherche, 26 rue d'Ulm, 75248 Paris Cedex 05, France3 Laboratoire du Dr René Cartier, 20 rue des Cordelières, 75013 Paris, France2005 3 5 2005 4 15 15 5 11 2004 3 5 2005 Copyright © 2005 Rosty et al; licensee BioMed Central Ltd.2005Rosty 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 have previously reported activating mutations of the gene coding for the fibroblast growth factor receptor 3 (FGFR3) in invasive cervical carcinoma. To further analyze the role of FGFR3 in cervical tumor progression, we extended our study to screen a total of 75 invasive tumors and 80 cervical intraepithelial neoplasias (40 low-grade and 40 high-grade lesions).
Results
Using single strand conformation polymorphism (SSCP) followed by DNA sequencing, we found FGFR3 mutation (S249C in all cases) in 5% of invasive cervical carcinomas and no mutation in intraepithelial lesions. These results suggest that, unlike in bladder carcinoma, FGFR3 mutation does not or rarely occur in non invasive lesions. Compared to patients with wildtype FGFR3 tumor, patients with S249C FGFR3 mutated tumors were older (mean age 64 vs. 49.4 years, P = 0.02), and were more likely to be associated with a non-16/18 HPV type in their tumor. Gene expression analysis demonstrated that FGFR3 mutated tumors were associated with higher FGFR3b mRNA expression levels compared to wildtype FGFR3 tumors. Supervised analysis of Affymetrix expression data identified a significant number of genes specifically differentially expressed in tumors with respect to FGFR3 mutation status.
Conclusion
This study suggest that tumors with FGFR3 mutation appear to have distinctive clinical and biological characteristics that may help in defining a population of patients for FGFR3 mutation screening.
==== Body
Background
Cervical cancer is the second leading cancer in women worldwide and a common cause of death among women in developing countries where 80% of cases occur [1]. Invasive cervical carcinoma develops through a well-defined progression model. Cervical intraepithelial neoplasia (CIN) is the premalignant lesion that always precedes invasive squamous cell carcinoma [2]. These precursor lesions are subdivided into 3 grades (CIN I-III) or 2 grades (low-grade squamous intraepithelial lesions, LSIL and high-grade squamous intra-epithelial lesions, HSIL). Half of low-grade lesions spontaneously regress within 6 months although 10–20% of high-grade lesions may progress to invasive carcinomas [3]. Molecular epidemiologic studies clearly demonstrated that sexually transmitted infection by HPV (human papillomavirus) is the principal cause of cervical carcinoma. Fifteen HPV types are considered as high-risk HPV and are associated with a higher risk of developing invasive cervical carcinoma from squamous intraepithelial lesions [4]. However, HPV infection is not sufficient to transform the normal cervix epithelial cells to invasive carcinomas and several additional events are necessary. Only a few genetic alterations have been reported in cervical carcinoma so far.
We previously reported specific FGFR3 missense mutations in 3 out of 12 invasive cervical carcinomas [5]. FGFR3 belongs to a family of structurally related tyrosine kinase receptors encoded by four different genes (FGFR1-4). FGFRs are glycoproteins composed of two or three extracellular immunoglobulin (Ig)-like domains, an hydrophobic transmembrane region and a cytoplasmic part that contains the tyrosine catalytic site. FGFRs are present as inactive monomers on the cell surface, upon ligand binding FGFRs dimerize, autophosphorylate and are able to transmit a series of intracellular signals [6]. An alternative splicing event in the second half of the juxtamembrane Ig-like domain of FGFR3 generates two mutually exclusive isoforms : FGFR3b the main form expressed in epithelial cells and FGFR3c the main form expressed in chondrocytes. Germinal activating FGFR3 mutations result in craniosynostoses and dwarfing chondrodysplasias of varying severity (hypochondroplasia, achondroplasia, SADDAN and thanatophoric dysplasia). Strikingly the same activating mutations have been reported at the somatic level in several types of cancer: multiple myeloma, bladder and cervical carcinomas. Very frequent in bladder carcinoma, particularly in non invasive papillary tumours (pTa tumors) (70% of cases harbor mutations), FGFR3 mutations are more rare in multiple myeloma and cervical carcinomas [5,7,8].
The goal of this work is to extend our previous study to screen a total of 75 patients for FGFR3 mutations and to identify clinical and/or pathological features associated with FGFR3 mutation. We also asked whether FGFR3 mutation could occur at earlier stages of cervical tumor progression, like in bladder tumors for which the highest rate of mutation is for low-stage non invasive pTa tumors. We thus analyzed 80 squamous intra-epithelial lesions (40 LSILs and 40 HSILs).
Results
FGFR3 mutation in invasive cervical carcinoma and squamous intraepithelial lesions
To extend our previous report of FGFR3 mutation in 3 of 12 cervical carcinomas [5], we selected 63 additional cases for a total of 75 screened DNAs. Those patients had the same characteristics than the initial cohort of 12 patients. SSCP analysis was performed on exons 7, 10, 15, and 20 of FGFR3, followed by direct DNA sequencing for cases with abnormal SSCP profiles. We found one additional case with FGFR3 mutation. Taken together, our study showed FGFR3 mutation in 4 of 75 invasive cervical cancer (5%) (Table 1). All mutations were C to G transversion at nucleotide 746, which changed serine 249 to cysteine. This S249C mutation creates a cysteine residue in the extracellular domain of the FGFR3 receptor.
Table 1 FGFR3 mutations in 75 invasive cervical carcinoma samples
Sample Age Stage Histology HPV Codon Mutation Predicted effect
6.96.1 65 IB SCC + * 249 TCC → TGC Ser → Cys
4.139 69 IIB SCC 33 249 TCC → TGC Ser → Cys
4.13 64 IIB SCC 16 249 TCC → TGC Ser → Cys
7.79.1 58 IIB SCC + * 249 TCC → TGC Ser → Cys
* Tumors with HPV DNA sequences of undetermined type (other than HPV 6, 11, 16, 18, 31, 33, 35, 39, 42, 45, 52 and 58). SCC : Squamous Cell Carcinoma
To further analyze the role of FGFR3 in cervical tumor progression, we investigated whether FGFR3 mutations are restricted to invasive carcinoma or may occur in squamous intraepithelial lesions. Hence we analyzed DNA from 40 LSILs and 40 HSILs. SSCP revealed no FGFR3 mutation in any of these cases. For all SSCP experiments, positive controls carrying all somatic FGFR3 mutations previously identified in tumors have been included. The rate of FGFR3 mutation in invasive cervical carcinoma is significantly different from the one in squamous intraepithelial lesions (P = 0.036, chi-square test).
Overall FGFR3 mutation rate in invasive cervical carcinoma
Studies about FGFR3 mutation in cervical cancer reported various rates of mutation, ranging from 0% to 25%. In order to evaluate an overall mutation rate, we added all results. Hence, we searched for all published papers on FGFR3 mutations in cervical cancer, by interrogating Medline with the terms "cervical cancer" and "FGFR3" (Table 2). Including this work, 6 studies analyzed a total of 349 cervical cancer samples and found 6 FGFR3 mutations (1.7%) [5,9-12]. All mutations were S249C missense mutations.
Table 2 Total number of cervical carcinoma cases screened for FGFR3 mutation, as of august 2004.
Study Method Total number of cases Mutated cases (percentage)
Cappellen et al. [5] SSCP (entire coding region) 12 3 (25%)
Yee et al. [9] Direct sequencing (exon 7) 104 0
Wu et al. [10] Direct sequencing (exons 7, 10, 13, 15, and 19) 51 1 (2%)
Dai et al. [11] Amplification created restriction site methodology for S249C mutation 91 0
Sibley et al. [12] Direct sequencing (exons 7, 10, and 15) 28 1 (3.5%)
Present study SSCP (exons 7, 10, and 15) 63 1 (1.5%)
Total 349 6 (1.7%)
Patients with FGFR3 mutation are older that patients with wild-type FGFR3
Clinical and pathological features were retrieved from patient's charts and correlated with FGFR3 status (Table 1). The 4 patients with S249C FGFR3 mutation tumor were significantly older at time of diagnosis than patients without the S249C FGFR3 mutation in the tumor (mean age 64 vs. 49.4 years, P = 0.02, Mann-Whitney U test). Clinical stage was IIB for 3 patients and IB for 1 patient. One patient died from the disease after 19 months. Follow-up information for the other 3 patients was available only for 3 months, 47 months, and 81 months, with no sign of relapse.
All 4 mutated tumors were well-differentiated invasive squamous cell carcinomas. HPV DNA of various types was detected in all cases : one HPV16, one HPV33 and two HPV of undetermined type (other than HPV 6, 11, 16, 18, 31, 33, 35, 39, 42, 45, 52 and 58). Distribution of HPV types associated with FGFR3 mutated tumors was rather different than the expected distribution in squamous cell cervical carcinoma. The prevalence of HPV16, HPV33 and HPV of another type of those determined by our PCR method in squamous cell carcinoma is approximately 55%, 2%, and less than 10%, respectively [4,13]. Thus, the probability to have 2 tumors with HPV of undetermined type in our series would be less than 1%.
FGFR3 mRNA expression is higher in tumors with S249C FGFR3 mutation
S249C FGFR3 mutation has been initially described in constitutional DNA from individuals with thanatophoric dysplasia [14]. It is suggested that activation of FGFR3 results from formation of intermolecular disulfide bonds between 2 mutant FGFR3 monomers. In order to get insight about association between S249C FGFR3 mutation and expression, we analyzed FGFR3b expression by semi-quantitative PCR in normal cervical samples and in invasive carcinomas (Figures 1, 2 and data not shown). In 17 normal cervix samples, FGFR3b/TBP mRNA ratios was 4.16 ± 0.72 in exocervical specimens and 0.35 ± 0.17 in endocervical specimens (Figure 1). Difference in of FGFR3b/TBP mRNA ratio means in exocervix and endocervix was statistically significant (P < 0.001, Mann-Whitney U test). FGFR3b mRNA levels was determined in 62 cervical carcinoma cases, including the 4 S249C mutated cases. Tumors with S249C mutation showed a significantly higher FGFR3b/TBP mRNA ratio than tumors without the S249C mutation (30.92 ± 21.74 vs. 2.27 ± 0.36, P = 0.002, Mann-Whitney U test) (Figure 2).
Figure 1 Schematic results of FGFR3b expression level in normal exocervix (n = 10) and normal endocervix (n = 7). FGFR3b expression is significantly higher in exocervix compared to endocervix.
Figure 2 FGFR3b mRNA expression in normal cervix and invasive cervical carcinoma, using semi-quantitative RT-PCR with TBP as reference gene. The level of FGFR3b mRNA expression was measured in 3 normal exocervix epithelia, 2 underlying connective tissues (C.T.), 4 invasive cervical carcinomas with S249C FGFR3 mutation (*) and 14 invasive cervical carcinomas with wildtype FGFR3.
We then sought for a correlation between survival and FGFR3b expression level. When patients were subdivided into 2 or 3 groups of equal size according to FGFR3b/TBP levels in cervical carcinoma, there was no statistically significant difference in survival between these groups. However, when patients were subdivided into 3 groups, there was a trend for patients with low levels of FGFR3b/TBP to have a shorter survival compared to the 2 other groups of patients with higher levels of FGFR3b/TBP (P = 0.12, Log-rank test) (Figure 3).
Figure 3 Survival stratified for the FGFR3b/TBP levels of invasive cervical carcinoma. Kaplan-Meier survival curves compare the cumulative probability of survival after diagnosis among patients with low FGFR3b/TBP levels to those with high FGFR3b/TBP levels.
Differentially expressed genes in FGFR3 mutated tumors compared to FGFR3 wildtype tumors
Finally, we used gene expression data from Affymetrix oligonucleotide microarray hybridization of invasive cervical carcinomas [15]. As all tumors with FGFR3 mutations were of the squamous type, we only used gene expression profiles from squamous cell carcinomas to compare tumors of same histological type and to minimize gene expression differences caused by different histology. We compared expression profiles of 3 FGFR3 mutated cervical carcinomas with the expression profiles of 17 wildtype FGFR3 cervical carcinomas, using SAM analysis as described in the methods section. With a threefold differential cutoff, we found 262 probe sets expressed at higher level and 552 probe sets at lower levels in FGFR3 mutated tumors compared to wildtype FGFR3 tumors (see additional files 1 and 2). To restrict the number of genes, we arbitrarily set a cutoff q value at 0.20 for higher expressed probe sets to end up with 61 probe sets, corresponding to 51 different known genes (Table 3). Genes with a role in regulation of transcription were the most frequently represented, including POU6F2, NHLH2, NR1D1, and ST18. FGFR3 was among the differentially expressed genes and was found upregulated, in agreement with RT-PCR results.
Table 3 Genes with higher expression in FGFR3 mutated cervical carcinomas compared to wildtype FGFR3 cervical carcinomas.
Gene Symbol Gene Title
ABCB9 ATP-binding cassette, sub-family B (MDR/TAP), member 9
ACACA acetyl-Coenzyme A carboxylase alpha
ANKRD6 ankyrin repeat domain 6
APIN APin protein
AVP arginine vasopressin
C8A complement component 8, alpha polypeptide
CACNA2D3 calcium channel, voltage-dependent, alpha 2/delta 3 subunit
CACNB2 calcium channel, voltage-dependent, beta 2 subunit
CALML3 calmodulin-like 3
CAMK1G calcium/calmodulin-dependent protein kinase IG
CAMK2B calcium/calmodulin-dependent protein kinase II beta
CCL13 chemokine (C-C motif) ligand 13
CPEB1 cytoplasmic polyadenylation element binding protein 1
DBP D site of albumin promoter (albumin D-box) binding protein
DDO D-aspartate oxidase
ED1 ectodermal dysplasia 1, anhidrotic
FBXO26 F-box only protein 26
FGFR3 fibroblast growth factor receptor 3
FTHFD formyltetrahydrofolate dehydrogenase
GNAO1 guanine nucleotide binding protein (G protein), alpha activating activity polypeptide O
GRIK4 glutamate receptor, ionotropic, kainate 4
GUCA2B guanylate cyclase activator 2B
HIG2 hypoxia-inducible protein 2
KCNE1L potassium voltage-gated channel, Isk-related family, member 1-like
LPHN3 latrophilin 3
MDM2 Mdm2, transformed 3T3 cell double minute 2
MEP1A meprin A, alpha (PABA peptide hydrolase)
MYF6 myogenic factor 6 (herculin)
NEFH neurofilament, heavy polypeptide 200kDa
NHLH2 nescient helix loop helix 2
NR1D1 nuclear receptor subfamily 1, group D, member 1
NRN1 neuritin 1
OGDHL oxoglutarate dehydrogenase-like
POU6F2 POU domain, class 6, transcription factor 2
PPFIA4 protein tyrosine phosphatase, receptor type, f polypeptide
PVRL1 poliovirus receptor-related 1
RAPGEF3 Rap guanine nucleotide exchange factor (GEF) 3
RHAG Rhesus blood group-associated glycoprotein
RNF6 ring finger protein (C3H2C3 type) 6
SLC4A3 solute carrier family 4, anion exchanger, member 3
SLC6A15 solute carrier family 6 (neurotransmitter transporter), member 15
SLC6A8 solute carrier family 6 (neurotransmitter transporter, creatine), member 8
SLC7A8 solute carrier family 7 (cationic amino acid transporter, y+ system), member 8
SMARCA4 SWI/SNF related, matrix associated, actin dependent regulator of chromatin, subfamily a, member 4
ST18 suppression of tumorigenicity 18 (breast carcinoma) (zinc finger protein)
TFAP2B transcription factor AP-2 beta
TRIM2 tripartite motif-containing 2
UPB1 ureidopropionase, beta
ZNF257 zinc finger protein 257
ZNF287 zinc finger protein 287
ZXDA zinc finger, X-linked, duplicated A
Discussion
Since its first description, the role of FGFR3 mutation in cervical carcinoma has been debated when some studies reported an absence of mutation in a total of nearly 200 cases analyzed [9,11]. Here we showed that FGFR3 is mutated in 5% (4 of 75 cases) of invasive cervical carcinomas from a French cohort of patients (including 3 cases previously reported [5]). When all study results are grouped, the overall FGFR3 mutation rate in cervical carcinoma worldwide is 1.7%.
In bladder neoplasias, FGFR3 mutation characterizes low-grade low-stage pTa bladder tumors and is rarely found in advanced stage bladder carcinoma [16,17]. The low rate of FGFR3 mutation in invasive cervical carcinoma could result from the same distribution pattern, FGFR3 mutation being found at higher frequency in intraepithelial non invasive lesions. Here, we show that no mutation was identified in 80 cervical squamous intraepithelial lesions.
To better characterize FGFR3 mutated cervical tumors and to find explanations for the apparent discrepancy in mutation rates, we looked for clinical and pathological features that may be specifically associated with FGFR3 mutation. We showed that FGFR3 mutated tumors are more likely to be associated with a non-HPV16/18 type and that patients with mutated FGFR3 cervical carcinoma are older than patients with wildtype FGFR3 tumor. Epidemiologic studies have found an increase in prevalence with age for minor HPV types, such as types 39, 52, and 58 [4]. Age and association with minor HPV types may thus not be independent characteristics of patients with FGFR3 mutated cervical tumor. We can hypothesize that variations in reported FGFR3 mutation rate are, at least in part, related to differences in HPV types and in age distribution of studied patients. We cannot exclude that variations in FGFR3 mutation rate could also be related to other parameters, such as ethnic differences.
To get more insight into the biological significance of difference in FGFR3 mutated tumors, we sought for gene expression of FGFR3 in cervical tissues. FGFR3b expression is significantly higher in normal exocervix compared to normal endocervix and in FGFR3 mutated tumors compared to FGFR3 wildtype tumors. It has been suggested that the substitution of serine to cysteine in codon 249 may result in formation of intermolecular disulfide bond in the extracellular domain of two mutants FGFR3, causing a constitutive activation of the receptor [14]. It has been reported that activated FGFR3 results in activation of Jak/STAT pathway with STAT1 and STAT5 phsphorylation [18,19]. Here, we show for the first time that activation of FGFR3 is associated with an increase in FGFR3b expression in S249C mutated tumors. In bladder carcinoma, we also found a significantly higher FGFR3b expression in mutated tumors compared to wildtype tumors (unpublished data). Supervised analysis of Affymetrix expression data using SAM algorithm identified a significant number of genes that are up- or down-regulated in FGFR3 mutated tumors. These genes are potential targets of activated FGFR3. The small number of tumors in one group (3 vs. 17) may account for a high 0.20 false discovery rate for genes at higher expression in FGFR3 mutated tumors. However, the presence of FGFR3 in this list provides some evidence for the significance of these genes.
Finally, we found that patients with low FGFR3b expression in tumors are more likely to have a shorter survival than patients with high FGFR3b levels. Dai et al reported similar findings in an immunohistochemical study of 73 cervical carcinomas [11]. They found that patients with intense FGFR3 immunolabeling in tumor cells had a better prognosis. These results, not statistically significant in both studies, require further investigations.
Conclusion
In summary, we demonstrated that S249C FGFR3 mutation defines a small subset of invasive cervical carcinoma and is not found in precursor intraepithelial lesions. Albeit small in number, patients with FGFR3 mutation appear to have distinctive clinical and biological characteristics that may help in defining a population for FGFR3 mutation screening. Gene expression analysis suggests that FGFR3 mutated tumors have a different biology with higher FGFR3b expression and differentially expressed genes which are potential targets of activated FGFR3.
Methods
Tissue specimens
Seventeen normal cervical mucosa and 75 invasive cervical carcinoma samples were obtained from the Institut Curie Hospital. The normal cervical mucosa samples were collected by scraping the cervix surface with a razor blade from 13 surgical resection specimens (hysterectomy or conization). Localization of normal samples was the exocervix (n = 10) and the endocervix (n = 7). Cervical carcinoma samples were collected from 75 patients with a median age of 48 years (range 23–82 years). Twelve of these samples had been previously studied for FGFR3 mutation [5]. Clinical stages according to the International Federation of Gynecology Obstetrics (FIGO) staging system were IB for 38 patients, IIA for 6 patients, IIB for 21 patients, IIIA for 3 patients, and IIIB for 6 patients. No staging information was available for 1 patient. Median follow-up was 32 months. In 73 cases, carcinoma samples originated from the initial cervical biopsy before any treatment, except in one case in which the patient previously received radiotherapy treatment. In 2 cases, the tumor sample was a pelvic recurrence after treatment. Final histological diagnoses were invasive squamous cell carcinoma (n = 66) and invasive adenocarcinoma (n = 9). HPV status was determined as previously described [20]. Briefly, Southern blot hybridization was the first-step procedure, using specific probes for HPV 6, 11, 16, 18, 31, 33, 35, 39, 42, 45, 52 and 58. A PCR with consensus primers in the L1 open reading frame was performed on cases negative by Southern blot hybridization. HPV DNA sequences were detected in 66/75 specimens (88%): 43 HPV16, 12 HPV18, 2 HPV33, 1 HPV31, 1 HPV58, and 7 undetermined HPV types.
Eighty cervical intra-epithelial neoplasia (CIN) biopsy or conization samples have also been selected: 40 high-grade squamous intraepithelial lesions (HSIL) from the Institut Curie Hospital (20 cases associated with HPV and 20 cases not associated with HPV); 40 low-grade squamous intraepithelial lesions (LSIL) from Dr. Isabelle Cartier Pathology Laboratory (20 cases associated with HPV and 20 cases not associated with HPV). Median age was 28 years (range 18–62 years) for patients with LSIL and 36 years (range 19–59) for patients with HSIL.
All normal samples, invasive carcinomas and squamous intraepithelial lesions were immediately frozen in liquid nitrogen and stored at -80°C until used.
SSCP analysis and direct DNA sequencing
PCR-SSCP analysis was carried out on exons 7, 10, 15, 20 of the FGFR3 gene because these exons harbor all the activating mutations found previously in bladder and cervix carcinomas and in most human skeletal disorders due to FGFR3 mutations [5,21,22]. Samples that showed mobility shifts in SSCP analysis were further analysed by direct bidirectional sequencing. SSCP and sequence analysis were performed as previously described [16].
Semi-quantitative RT-PCR
Total RNA was extracted from each sample by caesium chloride ultracentrifugation. Messenger RNA levels were determined by multiplex semi-quantitative RT-PCR using TBP (TATA binding protein) as internal control. Complementary DNA synthesis, PCR and analysis were performed as described [23]. Twenty two cycles were performed for the coamplification of FGFR3b and TBP. Primer sequences for TBP (TFIID), and FGFR3b (XF2 and TMR1) were as described [23,24].
Oligonucleotide microarray analysis
We used results from a global gene expression analysis of 30 primary invasive cervical carcinomas [15]. Histological type was squamous cell carcinoma (SCC) for 20 samples and adenocarcinoma (AC) for 10 samples. S249C FGFR3 mutation was associated with 3 of the analyzed SCCs. Complementary RNA target was prepared and labeled as described in the Affymetrix GeneChip Expression Analysis Technical Manual, and hybridized to U133A Affymetrix oligonucleotide array, representing 22,215 probe sets.
Gene expression data were normalized across all samples and all probe sets, and log2-transformed. Significance analysis of microarrays (SAM) [25] was used to perform the two-class comparison for differentially expressed genes between the 3 cervical SCCs with FGFR3 mutation and the 17 cervical SCCs with wildtype FGFR3. The K-nearest neighbor imputation was used to account for missing data within the dataset. Output criteria selected for SAM included at least threefold greater expression in the FGFR3 mutated samples as compared to the FGFR3 wildtype samples. The false discovery rate (q value) was arbitrarily set to 0.20.
Statistical Analysis
All results are given as mean ± standard error of the mean unless otherwise specified. The Mann-Whitney U test was used to compare continuous variables. For survival analysis, patients who died because of the cervical cancer were categorized as deaths. All other clinical evolution were censored. Patients were subdivided into 2 or 3 groups of equal size, according to the level of FGFR3b/TBP expression in cervical carcinoma. Survival analysis was carried out using the Kaplan-Meier method. The difference between survival curves was analyzed by the log-rank test. Probability values of 0.05 or less were considered significant.
Authors' contributions
CR found the relationship between the occurrence of FGFR3 mutations and an older age for the patients. He also analysed the microarray experiments. MHA performed the SSCP analysis. DC performed the RT-PCR analysis together with JB. JB prepared the normal and tumor cervical samples and performed the RT-PCR experiments together with DC. IC provided the cervical intra-epithelial neoplasia specimens. XSG set up the cervix tumor bank and initiated the work on cervical cancer. FR coordinated the work. CR, MHA and FR participated in the writing of the manuscript. All authors participated in the different discussions, the interpretation of the data, participated in the correction of the manuscript and approved the final version.
Supplementary Material
Additional File 1
Raw data of gene expression analysis of 20 cervical carcinomas. Gene expression results of 20 cervical squamous cell carcinomas (SCC), using Affymetrix U133A array.
Click here for file
Additional File 2
SAM analysis from gene expression data of 20 cervical carcinomas. genes differentially expressed in FGFR3 mutated tumors (n = 3) compared to FGFR3 wildtype tumors (n = 17), identified by SAM analysis.
Click here for file
Acknowledgements
This work was supported by the CNRS, the Institut Curie, the Ligue Contre le Cancer (laboratoire associé) and the "Programme Génomique du Ministère chargé de la Recherche" for the Affymetrix experiments. We thank Jean-Baptiste Lahaye for his help in the SSCP analysis.
==== Refs
Ferlay F Bray F Pisani P Parkin DM GLOBOCAN 2000: Cancer Incidence, Mortality and Prevalence Worldwide, Version 1.0. IARC CancerBase No.5. 2001 Lyon, IARCPress
zur Hausen H Papillomaviruses and cancer: from basic studies to clinical application Nat Rev Cancer 2002 2 342 350 12044010 10.1038/nrc798
Schlecht NF Platt RW Duarte-Franco E Costa MC Sobrinho JP Prado JC Ferenczy A Rohan TE Villa LL Franco EL Human papillomavirus infection and time to progression and regression of cervical intraepithelial neoplasia J Natl Cancer Inst 2003 95 1336 1343 12953088
Munoz N Bosch FX de Sanjose S Herrero R Castellsague X Shah KV Snijders PJ Meijer CJ Epidemiologic classification of human papillomavirus types associated with cervical cancer N Engl J Med 2003 348 518 527 12571259 10.1056/NEJMoa021641
Cappellen D De Oliveira C Ricol D de Medina S Bourdin J Sastre-Garau X Chopin D Thiery JP Radvanyi F Frequent activating mutations of FGFR3 in human bladder and cervix carcinomas Nat Genet 1999 23 18 20 10471491
Robertson SC Tynan J Donoghue DJ RTK mutations and human syndromes: when good receptors turn bad Trends Genet 2000 16 368 10904266 10.1016/S0168-9525(00)02077-1
Chesi M Nardini E Brents LA Schrock E Ried T Kuehl WM Bergsagel PL Frequent translocation t(4;14)(p16.3;q32.3) in multiple myeloma is associated with increased expression and activating mutations of fibroblast growth factor receptor 3 Nat Genet 1997 16 260 264 9207791 10.1038/ng0797-260
Van Rhijn BW Van Tilborg AA Lurkin I Bonaventure J De Vries A Thiery JP Van Der Kwast TH Zwarthoff EC Radvanyi F Novel fibroblast growth factor receptor 3 (FGFR3) mutations in bladder cancer previously identified in non-lethal skeletal disorders Eur J Hum Genet 2002 10 819 824 12461689 10.1038/sj.ejhg.5200883
Yee CJ Lin O Boyd J Analysis of fibroblast growth factor receptor 3 S249C mutation in cervical carcinoma J Natl Cancer Inst 2000 92 1848 1849 11078763 10.1093/jnci/92.22.1848
Wu R Connolly D Ngelangel C Bosch FX Munoz N Cho KR Somatic mutations of fibroblast growth factor receptor 3 (FGFR3) are uncommon in carcinomas of the uterine cervix Oncogene 2000 19 5543 5546 11114733 10.1038/sj.onc.1203934
Dai H Holm R Kristensen GB Abeler VM Borresen-Dale AL Helland A Fibroblast growth factor receptor 3 (FGFR3) - analyses of the S249C mutation and protein expression in primary cervical carcinomas Anal Cell Pathol 2001 23 45 49 11904459
Sibley K Stern P Knowles MA Frequency of fibroblast growth factor receptor 3 mutations in sporadic tumours Oncogene 2001 20 4416 4418 11466624 10.1038/sj.onc.1204543
Clifford GM Smith JS Plummer M Munoz N Franceschi S Human papillomavirus types in invasive cervical cancer worldwide: a meta-analysis Br J Cancer 2003 88 63 73 12556961 10.1038/sj.bjc.6600688
Tavormina PL Rimoin DL Cohn DH Zhu YZ Shiang R Wasmuth JJ Another mutation that results in the substitution of an unpaired cysteine residue in the extracellular domain of FGFR3 in thanatophoric dysplasia type I Hum Mol Genet 1995 4 2175 2177 8589699
Rosty C Sheffer M Tsafrir D Stransky N Tsafrir I Peter M de Cremoux P de La Rochefordiere A Salmon R Dorval T Thiery JP Couturier J Radvanyi F Domany E Sastre-Garau X Identification of a proliferation gene cluster associated with HPV E6/E7 expression level and viral DNA load in invasive cervical carcinoma Submitted
Billerey C Chopin D Aubriot-Lorton MH Ricol D Gil Diez de Medina S Van Rhijn B Bralet MP Lefrere-Belda MA Lahaye JB Abbou CC Bonaventure J Zafrani ES van der Kwast T Thiery JP Radvanyi F Frequent FGFR3 mutations in papillary non-invasive bladder (pTa) tumors Am J Pathol 2001 158 1955 1959 11395371
Kimura T Suzuki H Ohashi T Asano K Kiyota H Eto Y The incidence of thanatophoric dysplasia mutations in FGFR3 gene is higher in low-grade or superficial bladder carcinomas Cancer 2001 92 2555 2561 11745189 10.1002/1097-0142(20011115)92:10<2555::AID-CNCR1607>3.0.CO;2-M
Kong M Wang CS Donoghue DJ Interaction of fibroblast growth factor receptor 3 and the adapter protein SH2-B. A role in STAT5 activation J Biol Chem 2002 277 15962 15970 11827956 10.1074/jbc.M102777200
Lievens PM Liboi E The thanatophoric dysplasia type II mutation hampers complete maturation of fibroblast growth factor receptor 3 (FGFR3), which activates signal transducer and activator of transcription 1 (STAT1) from the endoplasmic reticulum J Biol Chem 2003 278 17344 17349 12624096 10.1074/jbc.M212710200
Lombard I Vincent-Salomon A Validire P Zafrani B de la Rochefordiere A Clough K Favre M Pouillart P Sastre-Garau X Human papillomavirus genotype as a major determinant of the course of cervical cancer J Clin Oncol 1998 16 2613 2619 9704710
Webster MK Donoghue DJ FGFR activation in skeletal disorders: too much of a good thing Trends Genet 1997 13 178 182 9154000 10.1016/S0168-9525(97)01131-1
Tavormina PL Bellus GA Webster MK Bamshad MJ Fraley AE McIntosh I Szabo J Jiang W Jabs EW Wilcox WR Wasmuth JJ Donoghue DJ Thompson LM Francomano CA A novel skeletal dysplasia with developmental delay and acanthosis nigricans is caused by a Lys650Met mutation in the fibroblast growth factor receptor 3 gene Am J Hum Genet 1999 64 722 731 10053006 10.1086/302275
Diez de Medina SG Chopin D El Marjou A Delouvee A LaRochelle WJ Hoznek A Abbou C Aaronson SA Thiery JP Radvanyi F Decreased expression of keratinocyte growth factor receptor in a subset of human transitional cell bladder carcinomas Oncogene 1997 14 323 330 9018118 10.1038/sj.onc.1200830
Murgue B Tsunekawa S Rosenberg I deBeaumont M Podolsky DK Identification of a novel variant form of fibroblast growth factor receptor 3 (FGFR3 IIIb) in human colonic epithelium Cancer Res 1994 54 5206 5211 7923141
Tusher VG Tibshirani R Chu G Significance analysis of microarrays applied to the ionizing radiation response Proc Natl Acad Sci U S A 2001 98 5116 5121 11309499 10.1073/pnas.091062498
| 15869706 | PMC1131920 | CC BY | 2021-01-04 16:36:35 | no | Mol Cancer. 2005 May 3; 4:15 | utf-8 | Mol Cancer | 2,005 | 10.1186/1476-4598-4-15 | oa_comm |
==== Front
Mol CancerMolecular Cancer1476-4598BioMed Central London 1476-4598-4-161587635010.1186/1476-4598-4-16ResearchAmplification and overexpression of the ID4 gene at 6p22.3 in bladder cancer Wu Qiong [email protected] Michèle J [email protected] Florian H [email protected] Wolfgang A [email protected] Dept. of Urology, Heinrich Heine University, Düsseldorf, Germany2005 5 5 2005 4 16 16 9 2 2005 5 5 2005 Copyright © 2005 Wu et al; licensee BioMed Central Ltd.2005Wu 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
Amplifications at 6p22.3 are prevalent in advanced stage bladder cancer (TCC). Previous studies have identified SOX4, CDKAL, and E2F3 as targets of this amplification and therefore potential oncogenes, but the more telomeric DEK gene too has been reported as overexpressed and amplified. We have therefore investigated whether the intermediate region harboring the oncogene candidate ID4 is also part of the amplicon.
Results
Expression of E2F3, DEK, and ID4 was investigated by real-time RT-PCR in 28 TCC compared to 6 normal bladder tissues and in 15 TCC cell lines compared to cultured normal urothelial cells. Expression of E2F3 as well as DEK increased on average in tumor vs. normal tissues (3-fold and 2.5-fold, resp.), but only the increase for E2F3 was statistically significant (p = 0.039). ID4 overexpression was observed in selected specimens. Each of the three genes was overexpressed in several cell lines, up to 150-fold (ID4), 30-fold (E2F3), and 9-fold (DEK), but these increases were not correlated to each other. Instead, moderate (DEK) to excellent (ID4) correlations were observed with copy number increases of microsatellites near each gene. Microsatellite copy number increases were highly heterogeneous across the investigated several Mb region revealing at least three subregions of amplification.
Conclusion
Extending previous reports, our data indicate that the 6p22.3 amplicon in TCC is highly heterogeneous and targets several genes in a variable fashion. Among these, expression of E2F3 and DEK appear to be generally increased in TCC, with additional increases caused by amplifications. In contrast, over-expression of ID4, which is normally predominantly expressed in testes and brain, appears to depend more strictly on gene amplification. Accordingly, the effect of amplifications at 6p22.3 in bladder cancer is expected to be non-uniform, thereby contributing to the highly variable biological and clinical behavior of advanced stage tumors. ID4 is a potential oncogene in a small subset of bladder cancers.
==== Body
Background
Urothelial carcinoma, which is commonly called bladder cancer, occurs in two forms, a more prevalent papillary subtype and a rarer, but much more invasive subtype [1,2]. Invasive bladder cancers usually develop from highly dysplastic carcinoma in situ, but some papillary tumors also progress to an invasive form. While papillary cancers often contain a limited number of chromosomal alterations, invasive cancers are characterized by a high degree of chromosomal instability [3,4]. Even T1 stage cancers, which have only invaded the lamina propria underlying the urothelium, often exhibit multiple chromosomal changes. Cancers at more advanced stages accumulate further chromosomal alterations. In particular, they harbor amplifications, e.g. of regions from chromosomes 5p, 6p, 8q, 11q, and 20q [4-7]. It is generally assumed that chromosomal segments consistently amplified in a cancer contain oncogenes [8]. Accordingly, genes amplified in advanced bladder cancers would be expected to contribute to the progression of this cancer.
One of the most consistently amplified region in advanced bladder cancers is located at 6p22.3 [5-7,9-12]. This amplification is detected in up to 25% of advanced stage bladder cancers and is present in many bladder cancer cell lines. The cell lines harboring this amplification provide a convenient experimental access to map the amplified region precisely and identify potential urothelial carcinoma oncogenes. Mapping of the 6p22.3 amplicon has been performed by several groups who have identified different genes as potential targets of the amplification (Figure 1). In a first study [5], SOX4 was identified as a frequent, but not entirely consistent amplification target. Further studies revealed that many amplifications also included E2F3 and the encoded protein was over-expressed, particularly in high stage and high grade urothelial cancers [6,11,12]. A high resolution analysis by microarray-based comparative genomic hybridization identified CDKAL1 located between SOX4 and E2F3 as the most commonly amplified gene [7]. Another study indicated that DEK located further telomerically (Figure 1) may be amplified in a substantial proportion of bladder cancer tissues [10]. DEK was also found to be over-expressed in a cDNA microarray study, albeit predominantly in early stage tumors [12].
Figure 1 The chromosome 6p22.3 region. Verified genes are drawn to size as grey boxes and the location of microsatellites used (the prefix D6S is omitted) is indicated.
This somewhat confusing state may owe partly to the fact that many studies were performed prior to the publication of the finished sequence of chromosome 6 in October 2003 [14] and partly to the use of different techniques. Alternatively, the differences between the studies could also mean that the region of amplification is not uniform and that multiple genes might be targets.
Most previous studies have focussed on a more centromeric region within 6p22.3 containing SOX4, CDKAL1, and E2F3 (Figure 1). DEK is located about 2 Mb more telomeric of these genes. The interval framed by E2F3 and DEK contains another plausible oncogene candidate, i.e. ID4. The ID proteins ('inhibitor of differentiation') are named for their ability to bind and inhibit cell-type specific helix-loop-helix transcriptional activators inducing cell differentiation. Accordingly, they tend to stimulate cell proliferation, and have been implicated in various cancers [15-18]. Compared to ID1 and ID2, ID4 is a less well characterized member of the family. It is expressed in a tissue-specific manner, with the highest levels in testes and brain [19]. In the present study, we have therefore investigated to which extent ID4 gene copy numbers and expression are affected by 6p22 amplifications in bladder cancer, in comparison to E2F3 and DEK.
Results
Expression of 6p22 genes in bladder cancer cell lines
First, expression of ID4 mRNA in comparison to E2F3 and DEK mRNAs was investigated by real-time PCR in 16 TCC cell lines (Figure 2). Normal urothelial cells (UP) proliferating in culture and testicular tissue samples served as controls. Expression at least twice as strong as in normal urothelial cells was considered as over-expression. According to this criterion, twelve cell lines over-expressed ID4, with a maximum >150fold increase in HT1376. In 6 cell lines, E2F3 was over-expressed. The E2F3 over-expressing cell lines included HT1376 and 5637 in line with previous reports [7,11]. 5637 displayed a >30fold increased level of E2F3 mRNA. Six cell lines over-expressed DEK, although the relative increases were in general more moderate, with a maximum 9fold increase in RT112. Inspecting Figure 2 may suggest that cell lines over-expressing one gene also tended to over-express one or both others. However, this tendency was not reflected in a statistically significant correlation. Specifically, expression did not significantly correlate for any pair of genes, the best correlation coefficient reaching 0.39 between ID4 and DEK. The divergence is strikingly illustrated by the cell lines HT1376 and 5637, which presented increased expression levels for all three genes, but with either ID4 or E2F3 displaying particularly pronounced increases (Figure 2).
Figure 2 Expression of 6p22 genes in bladder cancer cell lines. Levels of mRNAs for ID4 (grey bars), E2F3 (black bars), and DEK (white bars) were determined by real-time RT-PCR relative to β-actin mRNA. As controls, shown to the left and right of the cancer cell lines, two independent primary cultures of normal urothelial cells (UP93 and UP94) and two different normal testicular tissues were used. All values were determined by at least triplicate measurements, with further repeats, if deviations exceeded 20% of the mean. Arbitrary units are given; values for E2F3 and DEK were divided by 10.
Gene amplification analysis
To determine whether the increases in mRNA expression were due to gene amplification, copy numbers were investigated for eight microsatellite loci located in 6p22.3. Five were located around ID4, from D6S422 close to E2F3 to DS1946, and three were located around DEK, including the intragenic marker D6S2051 (Figure 1). The results (Figure 3) demonstrate a considerable variation in the microsatellite copy numbers across the region, even within the same cell line. For instance, in HT1376 the copy numbers of the eight microsatellites ranged from approximately 0.5 to 13 normal genome equivalents, which would correspond to 1 – 26 copies in a diploid cell. As all bladder cancer cell lines are aneuploid, typically hypo- or hypertriploid with a modal distribution, between 2 – 40 copies would be present in a single cell. In HT1376 specifically, two different segments of amplification are discernible, one telomeric to DEK and one around ID4. In fact, amplification of the region including E2F3 and of SOX4 has been shown previously in this line [7,9,11]. Thus, there are at least three distinct regions of amplification in this cell line. In contrast, no amplification of microsatellites telomeric of D6S422 was evident in 5637, which contains increased copy numbers of E2F3 and SOX4 [7,9,11]. A more homogeneous increase of all more telomeric markers including those close to ID4 and DEK was seen in J82. RT112 contained a selective amplification of the D6S2051 marker located in the DEK gene in accord with the maximum expression of this gene in this cell line. None of the microsatellites showed increased copy numbers in T24 or SD, although E2F3 as well as DEK, but not ID4 were overexpressed in T24.
Figure 3 Copy numbers of 6p22 microsatellites in bladder cancer cell lines. Microsatellite copy numbers (see Fig. 1 for their location) were determined as described in the methods section in the cell lines (A-F) HT1376, 5637, J82, RT112, SD, and T24. Normal genome equivalents determined from leukocyte DNA were set as 1.
Expression of ID4 correlated excellently with the copy numbers of each of the microsatellites around the gene, yielding coefficients between 0.89 and 0.95 (Figure 4A). A closer inspection shows however that this strong correlation is primarily caused by the cell lines with clear-cut amplifications of the region, i.e. J82 and HT1376 while at lower copy numbers the relationship is essentially random. DEK expression correlated moderately well with the copy number of the intragenic D6S2051 marker (Figure 4B). E2F3 expression showed the best correlation with the microsatellite marker D6S422 located most closely to the gene (Figure 4C). However, this apparent correlation was mostly due to the strongly increased copy number of D6S422 in the 5637 cell line; if this data point is removed, the relationship is essentially random.
Figure 4 Relationship of 6p22 gene expression changes to copy number changes of adjacent microsatellites in bladder cancer cell lines. A: ID4 mRNA vs. D6S1128E (r = 0.95), B: DEK mRNA vs. D6S2051 (r = 0.57), C: E2F3 mRNA vs. D6S422 (r = 0.79),
6p22 gene expression in bladder cancer tissues
To determine whether the results from the cell lines can be extended to bladder cancer tissues, the expression of the three genes was determined in 28 tumor tissue samples and 6 morphological normal samples from cystectomy specimens by real-time RT-PCR (Figure 5, Table 1). In accord with previous reports, expression of E2F3 mRNA and DEK mRNA were often increased in tumor compared to normal tissues. Median E2F3 mRNA expression was 2.24 arbitrary units in cancers compared to 0.72 in normal tissues, i.e. about threefold higher. This difference was statistically significant (p = 0.039). Median DEK mRNA expression was 1.26 compared to 0.52, i.e. 2.5fold higher, but the difference did not reach statistical significance. In contrast, ID4 expression essentially did not differ between normal and cancer tissues. Instead, individual cancer specimens showed strongly increased expression beyond the range of normal tissues.
Figure 5 Expression of genes at 6p22.3 in bladder cancer tissues. Box plot representation of ID4 (left), E2F3 (center), and DEK (right) mRNA expression relative to β-actin mRNA as determined by real-time RT-PCR in 28 bladder cancer tissues (see Table 1) and 6 morphologically normal bladder tissues. Note the different scales in the three figure parts.
Table 1 Tumor tissues investigated
Patient Tumor mRNA expression (AU)
No Sex Age Stage Lymph node status Metastasis status Grade E2F3 ID4 DEK
3 m 75 pT3b N0 M0 G2 5.59 0.48 4.88
6 m 76 pT3b pN2 M0 G3 10.37 0.36 1.37
12 m 70 pT3a N0 M0 G3 1.73 0.04 1.99
28 m 72 pT3b N0 M0 G3 3.73 0.24 2.27
41 f 54 pTa Nx Mx G2 11.69 0.49 3.29
47 m 76 pT3b N0 M0 G3 3.75 0.19 1.84
52 f 87 pT3b N0 M0 G3 2.86 0.25 1.43
55 m 83 pT4a N1 M0 G3 1.25 0.18 0.52
61 m 75 pT3b N0 M0 G3 1.98 0.12 1.48
62 f 78 pT3b N0 M0 G2 1.60 0.54 0.72
64 m 74 pT3b N2 M0 G3 3.34 0.11 1.26
67 f 94 >pT2 Nx Mx G3 4.38 0.33 1.50
69 m 77 pT3b N0 M0 G3 2.17 0.14 1.19
104 m 61 pT1 N0 M0 G2 1.49 0.24 0.72
105 m 83 pTa Nx M0 G2 3.80 0.27 1.88
109 m 68 pT4a N0 M0 G3 2.14 0.23 0.30
111 m 81 pT3b N0 M0 G2-3 2.24 0.52 0.61
115 m 74 pT3a N2 M0 G3 1.74 0.12 0.84
120 m 66 pT2 N0 Mx G2 3.12 0.15 0.66
150 f 78 pT3a N2 M0 G3 0.54 0.20 0.86
168 m 64 pT3a N2 M0 G3 1.68 0.12 1.42
170 f 63 pT2 N0 M0 G2 0.58 0.08 0.27
172 m 72 pT3a N0 M0 G3 1.73 0.13 0.85
205 f 65 pT2 N0 M0 G2 0.36 0.10 0.17
212 m 73 pT2a N1 M0 G2 0.62 0.16 0.79
224 m 65 pT2a N0 M0 G2 4.29 3.39 1.87
231 m 95 >pT2 Nx M0 G2 13.04 1.11 1.59
246 m 67 pT4a N2 M0 G3 3.07 0.27 1.29
Discussion
Taken together with previous analyses of 6p22.3 amplifications in bladder cancer, the present study has implications concerning the structure of 6p22.3 amplicons, the effect of the amplification on gene expression, and more generally, concepts of the significance of gene amplification in human cancers. Specifically, the findings raise interesting aspects with regard to ID4.
Our findings indicate that 6p22.3 amplifications in bladder cancer are even more heterogeneous than hitherto assumed. Previous studies have identified the centromeric region around CDKAL1 as part of an amplicon that often, but not consistently included SOX4 and E2F3 [7,9,11,12]. The telomeric region around DEK had not been investigated as well yet [10]. Our study confirms that this region is also subject to copy number gains and amplifications. Specifically, our findings are in good accord with ref. [10] describing frequent copy numbers gains of microsatellites around DEK, but at highly variable frequencies. The three markers investigated in that study were located within 0.5 Mb and were amplified contiguously in 24% of the specimens, whereas each of these three markers individually was amplified in 45%, 56%, and 64% of the cases. Thus, upon closer analysis, this region of amplification appears also heterogeneous in itself. The intermediate region harboring ID4 had been more or less disregarded in previous studies, but our data indicate that it is clearly gained or even amplified in a certain number of cases, sometimes concomitantly with, and sometimes independent of the other two regions. In summary, therefore, one might discern three main segments of amplification, which split up into further subregions in individual cancers. The 6p22.3 amplicon therefore seems to belong to a class which is characterized by pronounced heterogeneity and great structural complexity (see below).
Previous studies have variously identified SOX4, E2F3, CDKAL1, and DEK as potential targets of the 6p22.3 amplification; the present study adds ID4 to this list. In this regard, it is interesting to compare the cell line data, where ID4 emerged as the most frequently over-expressed gene with the tissue data which showed generalized increases in expression of E2F3 and of DEK (Figure 1 vs. Figure 5). This apparent discrepancy can be relatively simply resolved by two plausible assumptions. Normal bladder tissue is largely quiescent, albeit proliferation increases strongly in response to tissue damage [20]. Thus, urothelial cancers are distinguished from normal tissue not only by expression of cancer-specific genes, but also by generalized over-expression of genes associated with cell proliferation. The generally increased expression of E2F3 and DEK in the cancer tissues may reflect the latter effect, with further increases in individual cases due to deregulation and copy number gains of these genes. In contrast, early passage cultured urothelial cells proliferate as rapidly as cancer cell lines [21-23]. Therefore, the increases in E2F3 and DEK in cultured cancer vs. normal cells may turn out as comparatively moderate. In contrast, ID4 expression has probably to be considered as ectopic in bladder cancer, since it is normally restricted to other tissues including testes and brain [19]. Thus, overexpression may more strictly depend on amplification of the gene, particularly in tumor tissues.
It is commonly assumed that regions in the genome that are amplified in cancers harbor proto-oncogenes that are activated by overexpression as a consequence of increased gene dosage. Indeed, several bona fide oncogenes have been found in amplified regions and some have even been identified by cloning from amplicons. In such cases, amplicons consistently contain one particular gene, alone or together with a limited number of others, e.g. ERBB2 only or together with TOP2A. Interestingly, the structure of such amplicons can be quite simple [24-26]. The mechanism underlying such amplifications is not understood in detail, but appears to involve the re-replication of a single chromosome fragment, most likely via a circular double-minute intermediate [26].
Clearly, the 6p22 amplicon in bladder cancer belongs to a different class of amplicons which are characterized by great heterogeneity and instability. Such amplicons often contain different segments and accordingly different genes from one region and even sequences from different chromosomes [27-30]. The mechanism causing these amplifications is considered to be most likely breakage-fusion-bridge cycles initiated e.g. by hypoxia [31] or breakage at fragile sites [32].
Considering this background, the question which is the oncogene targeted by 6p22 amplifications in bladder cancer and the specific issue of the role of ID4 have to be approached with due caution, since in each individual case the amplicon may be influenced by random factors such as the location of an initiating double-strand break and structural factors such as preferred sites of breakage of dicentric chromosomes arising during breakage-fusion-bridge cycles. Nevertheless, the relatively high prevalence of 6p22 amplifications in bladder cancer and the relative specificity of this amplification for this cancer type argue for a functional selection as well.
For several 6p22 genes subject to amplification, it is plausible to assume that their overexpression may confer a more aggressive phenotype to bladder cancer cells. SOX factors determine cell fate and cell differentiation [33], so SOX4 overexpression might lead to further dedifferentiation. E2F transcription factors activate the transcription of genes required for DNA synthesis and E2F3 appears to repress some promoters that are activated by E2F1, including the ARF promoter in CDKN2A [34]. DEK is part of a oncogenic fusion protein resulting from t(6;9)(p23;q34) translocations in acute myeloid leukemia [35] and is implicated in regulation of chromatin structure, which is evidently aberrant in invasive bladder cancers. However, specific functional studies on the role of these proteins in urothelial cells are lacking. The function of CDKAL1 is entirely unknown.
ID4 belongs to a protein family whose members have been shown to interfere with cell differentiation by blocking the effects of HLH transcription factors and pocket proteins, including RB1. Several members have been reported to be over-expressed in human cancers [15-18]. ID4 in particular has been shown to be overexpressed in rat mammary carcinomas. Accordingly, overexpression of ID4 blocked the differentiation of HC11 mammary epithelial cells and stimulated their proliferation [36]. It is also the target of a specific chromosomal translocation in some cases of B-cell acute lymphoblastic leukemia [37]. Conversely, ID4 expression has been reported to be down-regulated by promoter hypermethylation in colon carcinomas [38]. Evidently, as for the other oncogene candidates on 6p22, more detailed studies are required on the biochemical and functional properties of ID4 in normal and cancerous urothelial tissue.
Conclusion
In conclusion, our study indicates that the 6p22.3 amplification prevalent in advanced bladder cancers is highly heterogeneous and contributes to the altered expression of several genes, including ID4, in a highly variable manner. Thus, this genetic change may contribute to the highly variable biological and clinical behaviour of invasive bladder cancers.
Methods
Tissues
Cancerous and normal bladder tissues were used from a previous study [39]. Normal tissues were identified by gross morphology, with microscopic verification in case of extended tumors. Important clinical parameters of the cancer tissues are summarized in Table 1. RNA from testicular normal and cancer tissues used as a control for ID4 expression was also prepared in the course of a former study [40].
Cell lines and primary cultures
The bladder cancer cell lines 253J, 639v, 5637, BFTC905, BFTC909, EJ, HT1376, J82, MGHU4, RT112, SD, SW1710, T24, UMUC3, VMCub1, and VMCub2 and primary urothelial cells were cultured as described previously [23].
DNA and RNA extraction
High-quality DNA and RNA were extracted by standard methods using commercial kits from Qiagen (Hilden, Germany) and Peqlab (Erlangen, Germany).
RT-PCR
Following photometric quantification, 2 μg mRNA were transcribed into first strand cDNA using SuperscriptII (Invitrogen, Karlsruhe, Germany) according to the manufacturer's protocol with oligo-dT primers. Real-time RT-PCR was carried out using a LightCycler instrument (Roche Diagnostics, Mannheim, Germany) and the primers indicated in Table 2. The amplification mixture consisted of 1x reaction mix (LightCycler-FastStart DNA Master PLUS SYBR Green I; Roche Diagnostics), 10 pmoles (DEK, E2F3, ID4) or 5 pmoles (β-actin) of each primer and 20 ng cDNA in a final volume of 10 μl. The reaction was monitored between the annealing and elongation steps at 640 nm. After the final cycle, melting-point analysis of the samples was performed over the range of 69°C – 99°C. Turning-point values for the specific genes were related to those for β-actin.
Table 2 RT-PCR primers and conditions
Designation Sequence Tm(°C) Product size (bp)
GAPDH 350s TCCCATCACCATCTTCCA 62.3 379
GAPDH 350as CATCACGCCACAGTTTCC 61.9
Aktin661S TGACGGGGTCAC 72.2 661
Aktin661AS CTAGAAGCATTT 70.9
ID4 fw CCGCCCAACAAGAAAGTCAG 59.4 188
ID4 rv GGTGTTGAGCGCAGTGAG 58.2
E2F3 fw ACGTCTCTTGGTCTGCTCAC 59.4 155
E2F3 rv TCTTAATGAGGTGGATGCCT 55.3
DEK fw GTGGGTCAGTTCAGTGGC 58.2 291
DEK rv AGGACATTTGGTTCGCTTAG 55.3
Microsatellite analysis
Microsatellites located at 6p22 (see Figure 1) were amplified using published primer sets (see the ensembl database for sequences and Tms) in duplex reactions with one microsatellite from chromosome 12 (D12S1650) or chromosome 15 (D15S127) as control. These chromosome are rarely affected by allelic imbalances in bladder cancer [3]. One primer from each pair was labeled with IRD-800 fluorescence and the products were resolved and detected on a LiCOR 4200S automated sequencer. Reactions were carried out in the linear phase of PCR with 25–30 cycles, the precise number being determined for each primer pair. Band intensities were quantitatively determined by ONE-D-SCAN (MWG-Biotech, Ebersberg, Germany). Leukocyte DNA standards were included in each set of reaction. The ratio of intensities of chromosome 6 and chromosome 15 microsatellites in these was set as 1 for each pair.
Authors' contributions
QW performed most experiments and most of the data evaluation, aided and supported by MJH; FHH contributed and evaluated the clinical data; WAS conceived and supervised the study and drafted the manuscript.
Acknowledgements
Excellent technical assistance by Ms. Andrea Prior is gratefully acknowledged.
==== Refs
Knowles MA What we could do now: molecular pathology of bladder cancer Mol Pathol 2001 54 215 221 11477133 10.1136/mp.54.4.215
Dinney CP McConkey DJ Millikan RE Wu X Bar-Eli M Adam L Kamat AM Siefker-Radtke AO Tuziak T Sabichi AL Grossman HB Benedict WF Czerniak B Focus on bladder cancer Cancer Cell 2004 6 111 116 15324694 10.1016/j.ccr.2004.08.002
Knowles MA Elder PA Williamson M Cairns JP Shaw ME Law MG Allelotype of human bladder cancer Cancer Res 1994 54 531 538 8275491
Hovey RM Chu L Balazs M DeVries S Moore D Sauter G Carroll PR Waldman FM Genetic alterations in primary bladder cancers and their metastases Cancer Res 1998 58 3555 3560 9721860
Bruch J Wöhr G Hautmann R Mattfeldt T Brüderlein S Möller P Sauter S Hameister H Vogel W Paiss T Chromosomal changes during progression of transitional cell carcinoma of the bladder and delineation of the amplified interval on chromosome arm 8q Genes Chromosomes Cancer 1998 23 167 174 9739020 10.1002/(SICI)1098-2264(199810)23:2<167::AID-GCC10>3.0.CO;2-L
Veltman JA Fridlyand J Pejavar S Olshen AB Korkola JE DeVries S Carroll P Kuo WL Pinkel D Albertson D Cordon-Cardo C Jain AN Waldman FM Array-based comparative genomic hybridization for genome-wide screening of DNA copy number in bladder tumors Cancer Res 2003 63 2872 2880 12782593
Hurst CD Fiegler H Carr P Williams S Carter NP Knowles MA High-resolution analysis of genomic copy number alterations in bladder cancer by microarray-based comparative genomic hybridization Oncogene 2004 23 2250 2263 14968109 10.1038/sj.onc.1207260
Schwab M Oncogene amplification in solid tumors Semin Cancer Biol 1999 9 319 325 10448118 10.1006/scbi.1999.0126
Bruch J Schulz WA Häussler J Melzner I Brüderlein S Möller P Kemmerling R Vogel W Hameister H Delineation of the 6p22 amplification unit in urinary bladder carcinoma cell lines Cancer Res 2000 60 4526 4530 10969802
Evans AJ Gallie BL Jewett MAS Pond GR Vandezande K Underwood J Fradet Y Lim G Marrano P Zielenska M Squire JA Defining a 0.5-Mb region of genomic gain on chromosome 6p22 in bladder cancer by quantitative-multiplex polymerase chain reaction Am J Pathol 2004 164 285 293 14695341
Feber A Clark J Goodwin G Dodson AR Smith PH Fletcher A Edwards S Flohr P Falconer A Roe T Kovacs G Dennis N Fisher C Wooster R Huddart R Foster CS Cooper CS Amplification and overexpression of E2F3 in human bladder cancer Oncogene 2004 23 1627 1630 14716298 10.1038/sj.onc.1207274
Oeggerli M Tomovska S Schraml P Calvano-Forte D Schafroth S Simon R Gasser T Mihatsch MJ Sauter G E2F3 amplification and overexpression is associated with invasive tumor growth and rapid tumor cell proliferation in urinary bladder cancer Oncogene 2004 23 5616 5623 15122326 10.1038/sj.onc.1207749
Sanchez-Carbayo M Socci ND Lozano JJ Li W Charytonowicz E Belbin TJ Prystowsky MB Ortiz AR Childs G Cordon-Cardo C Gene discovery in bladder cancer progression using cDNA microarrays Am J Pathol 2003 163 505 516 12875971
Mungall AJ Palmer SA Sims SK Edwards CA Ashurst JL Wilming L Jones MC Horton R Hunt SE Scott CE Gilbert JG Clamp ME Bethel G Milne S Ainscough R Almeida JP Ambrose KD Andrews TD Ashwell RI Babbage AK Bagguley CL Bailey J Banerjee R Barker DJ Barlow KF Bates K Beare DM Beasley H Beasley O Bird CP Blakey S Bray-Allen S Brook J Brown AJ Brown JY Burford DC Burrill W Burton J Carder C Carter NP Chapman JC Clark SY Clark G Clee CM Clegg S Cobley V Collier RE Collins JE Colman LK Corby NR Coville GJ Culley KM Dhami P Davies J Dunn M Earthrowl ME Ellington AE Evans KA Faulkner L Francis MD Frankish A Frankland J French L Garner P Garnett J Ghori MJ Gilby LM Gillson CJ Glithero RJ Grafham DV Grant M Gribble S Griffiths C Griffiths M Hall R Halls KS Hammond S Harley JL Hart EA Heath PD Heathcott R Holmes SJ Howden PJ Howe KL Howell GR Huckle E Humphray SJ Humphries MD Hunt AR Johnson CM Joy AA Kay M Keenan SJ Kimberley AM King A Laird GK Langford C Lawlor S Leongamornlert DA Leversha M Lloyd CR Lloyd DM Loveland JE Lovell J Martin S Mashreghi-Mohammadi M Maslen GL Matthews L McCann OT McLaren SJ McLay K McMurray A Moore MJ Mullikin JC Niblett D Nickerson T Novik KL Oliver K Overton-Larty EK Parker A Patel R Pearce AV Peck AI Phillimore B Phillips S Plumb RW Porter KM Ramsey Y Ranby SA Rice CM Ross MT Searle SM Sehra HK Sheridan E Skuce CD Smith S Smith M Spraggon L Squares SL Steward CA Sycamore N Tamlyn-Hall G Tester J Theaker AJ Thomas DW Thorpe A Tracey A Tromans A Tubby B Wall M Wallis JM West AP White SS Whitehead SL Whittaker H Wild A Willey DJ Wilmer TE Wood JM Wray PW Wyatt JC Young L Younger RM Bentley DR Coulson A Durbin R Hubbard T Sulston JE Dunham I Rogers J Beck S The DNA sequence and analysis of human chromosome 6 Nature 2003 425 805 811 14574404 10.1038/nature02055
Israel MA Hernandez MC Florio M Andres-Barquin PJ Mantani A Carter JH Julin CM Id gene expression as a key mediator of tumor cell biology Cancer Res 1999 1726s 1730s 10197587
Norton JD ID helix-loop-helix proteins in cell growth, differentiation and tumorigenesis J Cell Sci 2000 113 3897 3905 11058077
Lasorella A Uo T Iavarone A Id proteins at the cross-road of development and cancer Oncogene 2001 20 8326 8333 11840325 10.1038/sj.onc.1205093
Sikder HA Devlin MK Dunlap S Ryu B Alani RM Id proteins in cell growth and tumorigenesis Cancer Cell 2003 3 525 12842081 10.1016/S1535-6108(03)00141-7
Riechmann V van Crüchten I Sablitzky F The expression pattern of Id4, a novel dominant negative helix-loop-helix protein, is distinct from Id1, Id2 and Id3 Nucleic Acids Res 1994 22 749 755 8139914
Mysorekar IU Mulvey MA Hultgren SJ Gordon JI Molecular regulation of urothelial renewal and host defenses during infection with uropathogenic Escherichia coli J Biol Chem 2002 277 7412 7419 11744708 10.1074/jbc.M110560200
Southgate J Hutton KA Thomas DF Trejdosiewicz LK Normal human urothelial cells in vitro: proliferation and induction of stratification Lab Invest 1994 71 583 594 7967513
Freeman MR Yoo JJ Raab G Soker S Adam RM Schneck FX Renshaw AA Klagsbrun M Atala A Heparin-binding EGF-like growth factor is an autocrine growth factor for human urothelial cells and is synthesized by epithelial and smooth muscle cells in the human bladder J Clin Invest 1997 99 1028 1036 9062361
Swiatkowski S Seifert HH Steinhoff C Prior A Thievessen I Schliess F Schulz WA Activities of MAP-kinase pathways in normal uroepithelial cells and urothelial carcinoma cell lines Exp Cell Res 2003 282 48 57 12490193 10.1006/excr.2002.5647
Simon R Atefy R Wagner U Forster T Fijan A Bruderer J Wilber K Mihatsch MJ Gasser T Sauter G HER-2 and TOP2A coamplification in urinary bladder cancer Int J Cancer 2003 107 764 772 14566826 10.1002/ijc.11477
Kuwahara Y Tanabe C Ikeuchi T Aoyagi K Nishigaki M Sakamoto H Hoshinaga K Yoshida T Sasaki H Terada M Alternative mechanisms of gene amplification in human cancers Genes Chromosomes Cancer 2004 41 125 132 15287025 10.1002/gcc.20075
Vogt N Lefebre SH Apiou F Dutrillaux AM Cor A Leuraud P Poupon MF Dutrillaux B Debatisse M Malfoy B Molecular structure of double-minute chromosomes bearing amplified copies of the epidermal growth factor receptor gene in gliomas Proc Natl Acad Sci USA 2004 101 11368 11373 15269346 10.1073/pnas.0402979101
Janssen JW Imoto I Inoue J Shimada Y Ueda M Imamura M Bartram CR Inazawa J MYEOV, a gene at 11q13, is coamplified with CCND1, but epigenetically inactivated in a subset of esophageal squamous cell carcinomas J Hum Genet 2002 47 460 464 12202983 10.1007/s100380200065
Volik S Zhao S Chin K Brebner JH Herndon DR Tao Q Kowbel D Huang G Lapuk A Kuo WL Magrane G De Jong P Gray JW Collins C End-sequence profiling: sequence-based analysis of aberrant genomes Proc Natl Acad Sci USA 2003 100 7696 7701 12788976 10.1073/pnas.1232418100
Yang ZQ Albertson D Ethier SP Genomic organization of the 8p11-p12 amplicon in three breast cancer cell lines Cancer Genet Cytogenet 2004 155 57 62 15527903 10.1016/j.cancergencyto.2004.03.013
Zatkova A Ullmann R Rouillard JM Lamb BJ Kuick R Hanash SM Schnittger S Schoch C Fonatsch C Wimmer K Distinct sequences on 11q13.5 and 11q23-24 are frequently coamplified with MLL in complexly organized 11q amplicons in AML/MDS patients Genes Chromosomes Cancer 2004 39 263 276 14978788 10.1002/gcc.20002
Coquelle A Toledo F Stern S Bieth A Debatisse M A new role for hypoxia in tumor progression: induction of fragile site triggering genomic rearrangements and formation of complex DMs and HSRs Mol Cell 1998 2 259 265 9734364 10.1016/S1097-2765(00)80137-9
Hellman A Zlotorynski E Scherer SW Cheung J Vincent JB Smith DI Trakhtenbrot L Kerem B A role for common fragile site induction in amplification of human oncogenes Cancer Cell 2002 1 89 97 12086891 10.1016/S1535-6108(02)00017-X
Dong C Wilhelm D Koopman P Sox genes and cancer Cytogenet Genome Res 2004 105 442 447 15237232 10.1159/000078217
Aslanian A Iaquinta PJ Verona R Lees JA Repression of the Arf tumor suppressor by E2F3 is required for normal cell cycle kinetics Genes Dev 2004 18 1413 1422 15175242 10.1101/gad.1196704
von Lindern M Fornerod M van Baal S Jaegle M de Wit T Buijs A Grosveld G The translocation (6;9), associated with a specific subtype of acute myeloid leukemia, results in the fusion of two genes, dek and can, and the expression of a chimeric, leukemia-specific dek-can mRNA Mol Cell Biol 1992 12 1687 1697 1549122
Shan L Yu M Qiu C Snyderwine EG Id4 regulates mammary epithelial cell growth and differentiation and is overexpressed in rat mammary gland carcinomas Am J Pathol 2003 163 2495 2502 14633621
Bellido M Aventin A Lasa A Estivill C Carnicer MJ Pons C Matias-Guiu Bordes R Baiget M Sierra J Nomdedéu JF Id4 is deregulated by a t(6;14)(p22;q32) chromosomal translocation in a B-cell lineage acute lymphoblastic leukemia J Hematol 2003 88 994 1001
Umetani N Takeuchi H Fujimoto A Shinozaki M Bilchik AJ Hoon DS Epigenetic inactivation of ID4 in colorectal carcinomas correlates with poor differentiation and unfavorable prognosis Clin Cancer Res 2004 10 7475 7483 15569977
Florl AR Franke KH Niederacher D Gerharz CD Seifert HH Schulz WA DNA methylation and the mechanisms of CDKN2A inactivation in transitional cell carcinoma of the urinary bladder Lab Invest 2000 80 1513 1522 11045568
Schmidt BA Rose A Steinhoff C Strohmeyer T Hartmann M Ackermann R Up-regulation of cyclin-dependent kinase 4/cyclin D2 expression but down-regulation of cyclin-dependent kinase 2/cyclin E in testicular germ cell tumors Cancer Res 2001 61 4214 4221 11358847
| 15876350 | PMC1131921 | CC BY | 2021-01-04 16:36:36 | no | Mol Cancer. 2005 May 5; 4:16 | utf-8 | Mol Cancer | 2,005 | 10.1186/1476-4598-4-16 | oa_comm |
==== Front
Reprod Biol EndocrinolReproductive biology and endocrinology : RB&E1477-7827BioMed Central London 1477-7827-3-141584514910.1186/1477-7827-3-14ResearchEndometriosis in adolescence: A long-term follow-up fecundability assessment Ventolini Gary [email protected] Gary M [email protected] Ronald [email protected] Department of Obstetrics and Gynecology, Wright State University, ayton, Ohio 45409-2793, USA2005 21 4 2005 3 14 14 6 10 2004 21 4 2005 Copyright © 2005 Ventolini et al; licensee BioMed Central Ltd.2005Ventolini 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.
Objective
A long-term, follow-up study comparing mild and severe forms of endometriosis and their fecundability, on 28 women diagnosed with endometriosis in adolescence.
Methodology
Twenty-eight patients were identified from a prospective cohort of 52 adolescents (ages 12 to 18 years) with operative diagnosis of endometriosis between July 1993 and December 1995. All patients presented with chronic pelvic pain unresponsive to conservative medical management. Diagnosis of pregnancy was made by sonographic identification of intrauterine pregnancy, positive serum human chorionic gonadotropin or pathological confirmation of products of conception. Patients were categorized as fertile or sub-fertile by having > 12 months of unprotected intercourse without conception. Follow-up was done for 8.6 years.
Results
Staging of endometriosis was performed according to the American Society for Reproductive Medicine standards. Stage I = 14.3%; Stage II = 39.3%; Stage III = 42.8%; Stage IV = 3.6%. Fecundability rates in each stage were statistically significant: Stage I (75%), Stage II (55%), Stage III (25%), Stage IV (0%) (p < .05). Rates of spontaneous abortion were not statistically significant.
Conclusion
In our cohort, even at the earliest point in the natural life cycle of endometriosis there is an inverse relationship between stage of disease at diagnosis and fecundability.
==== Body
Background
Recent studies [1] have given us more insight into the pathophysiology of endometriosis. Fragments of functional endometrium reflux through the fallopian tubes and reach the essentially hostile environment of the peritoneal cavity. Proteolytic activity, activated macrophages and natural killer cells all combine to degrade and digest the regurgitated tissue fragments. Occasionally whole fragments of endometrial tissue succeed in evading the peritoneal defense lines, perhaps by their sheer number, perhaps by an intrinsic defect in the defense system. Microtraumas to the peritoneum will expose the extracellular matrix; successful endometrial tissue implantation may occur, and the cyclic angiogenesis that occurs in the female reproductive tract, will allow survival and growth of the ectopic endometrial tissue. Then subsequent cyclic tissue remodeling and new tissue development will take place. The intricate interaction between the implant and the host tissue will eventually result in the condition called endometriosis. If the local immune system fails to repair its surface, then the development of the disease endometriosis characterized by pain, menstrual cycle disturbance and impaired fertility, will eventually prevail [2]. Endometriosis in early adolescence could be explained by the theory of metaplasia of embryonic mullerian remnants in an extrauterine location [3].
The role of adolescent endometriosis in the impairment of fertility in adult life has not been extensively evaluated.
The purpose of our study was to compare mild and severe forms of endometriosis and their fecundability in a cohort of 28 patients diagnosed with endometriosis in adolescence and followed for 8.6 years.
Methods
The 28 patients of the present report were identified from a prospective cohort of 52 adolescents, assembled between July 1993 and December 1995. They were operatively diagnosed with endometriosis during their evaluation for chronic pelvic pain unresponsive to conservative medical management. The diagnosis of endometriosis was made through diagnostic laparoscopy and biopsies. No other interventions beside diagnosis were performed. The original staging of endometriosis was done at the time of the surgery, using the American Fertility Society revised standard classification system [4]. Confirmation of staging was done by review of the photographic material obtained during the diagnostic procedures by experienced gynecologic laparoscopists.
All patients received the same therapy consistent of continuous uninterrupted combined oral contraceptives for 6 months (Low Ovral®): 1 tablet orally per day, and Naproxyn Sodium 500 milligrams orally to be used occasionally for pain exacerbation.
In June 2002 after Institutional Review Board approval and informed consent was obtained, a telephone interview with each one of the patients was conducted to follow-up fertility assessment. All the patients were questioned regarding their general health status, their history of sexually transmitted diseases (STDs), use of contraception, other therapies utilized for the endometriosis, their desire for conception, fertility evaluation of their male partner, and their own fertility status.
Patients were considered to be fertile if they had received obstetric sonographic identification of an intrauterine pregnancy, had a positive serum beta human chorionic gonadotropin ≥ 25 International Units per milliliter or had pathologic confirmation of fetal or placental tissue after a spontaneous abortion or instrumental dilation and curettage. The patients were categorized as subfertile if they had more than twelve months of unprotected sexual intercourse during their fertile periods without conception.
Fecundability was defined as the probability of achieving a pregnancy with in one year period, with the presence of desire for conception.
Twenty-seven out of twenty-eight patients were successfully interviewed, and their clinical data was completely collected and processed. One patient transferred to Mexico and was not possible to be located. She was initially diagnosed with stage III disease.
The patient data was then tabulated and analyzed. Statistical analysis was performed using the Fisher exact chi square trend to compare fecundability and to evaluate differences between the groups. Statistical significance was taken as a P value of less than .05. Confidence intervals (95%) were calculated using the Modified Wald Method (The American Statistician) [5].
Results
Their endometriosis was classified as: Stage I, four patients (14.3%); Stage II, eleven patients (39.2%), Stage III, twelve patients (42.8%) and Stage IV, one patient (3.6%).
Population characteristics regarding age at cohort entrance, age at follow-up, number of years trying to conceive, and age at diagnosis for endometriosis are displayed. (Table 1)
Table 1 Population Characteristics
Characteristics Stage I Stage II Stage III Stage IV P value
Age at Cohort Entrance 14.3 ± 2.3 15.4 ± 2.9 15.2 ± 2.6 16 .06
Caucasian 3 9 10 1 .73
African American 1 2 2 0 .35
STD's 0 2 2 0 .50
Contraception Use 1 4 4 1 .08
Desire for Conception 3 9 10 - .73
Male Partner's Evaluation 0 1 1 - .12
Other Therapies Used 1 2 3 - .25
Age at Stage of Endometriosis 15.2 ± 2.1 16.8 ± 2.7 16.7 ± 2.3 16.9 .535
Time Trying to Conceive (Month) 15.1 ± 4.5 12.8 ± 3.8 13.1 ± 4.2 - .310
Age at Follow-up 22.3 ± 5.4 24.7 ± 4.6 25.3 ± 4.9 - .125
Patient in stages I, II, III and IV were not statistically different regarding age, race, general health status, history of STDs, use of contraception, desire for conception, evaluation for infertility, male partner's fecundity evaluation, and therapies utilized for the treatment of their endometriosis. (Table 1)
Fecundability rates in each stage were: Stage I = 3 patients (75%), Stage II = 6 patients (55%), Stage III = 3 patients (25%), Stage IV = 0 patients (0%) (P < .05) (Figure 1).
Figure 1 Relationship between stage of endometriosis and fecundability. (CI = Confidence Interval)
The rate of spontaneous abortions was not statistically significant among the stages:
Stage I = 1 patient (16.7%), Stage II = 1 patient (33.4%), Stage III = 1 patient (33.4%) (P < 0.05).
Discussion
Endometriosis in adolescence is a cause of chronic pelvic pain and dysmenorrhea. The goal of therapy is to minimize pelvic pain and dysmenorrhea, primarily through long term medical therapy. Although endometriosis is associated with infertility, a clear causal relationship has yet to be established, except when adhesive disease is found [2].
The epidemiologic study of endometriosis presents researchers with unique Challenges. As a result, few well-designed studies have been published since the pathogenesis of endometriosis is still not yet fully understood [6].
Endometriosis is a progressive disease without a definitive cure. Therefore, adolescents with endometriosis require long-term medical management until they have completed their own fertility goal at childbearing age [6].
It is now well established that surgical management of endometriosis in the early stage of disease increases pregnancy rates (level I evidence based) [2]. Treatment of moderate to severe endometriosis also confers benefit although the evidence to support this treatment is using the United States Preventive Services Task Force classification of level II – III [2].
The above statements were not known nor applied to this cohort of adolescents since they were diagnosed with endometriosis between 1993 and 1995. The state of the art therapy for adolescence endometriosis during the referenced years was the use of oral contraceptive pills in a pseudo pregnancy fashion (continuous medication without withdrawal period) and the use of non-steroidal anti-inflammatory medications [7].
Our study, evaluating the fecundability of adolescents with endometriosis is unique because of the long follow-up 8.6 years. Also, the time elapsed between symptoms and diagnosis was relatively short. A recent study by Arruda et al. [8] showed a delay in diagnosis of endometriosis of 7.4 years for patients with pelvic pain.
Our findings, although the sample population is relatively small, confirm the well demonstrated fact observed in the adult population that there is an inverse relationship between endometriosis disease progression and fecundability [9]. Evidence supports surgical intervention, especially in stage 1 and 2 endometriosis to improve fecundity rates. In infertile women, laparoscopic resection or ablation of minimal and mild endometriosis practically doubles fecundity when compared to diagnostic laparoscopy alone [10]. Although a recent Italian study [11] was unable to show statistical improvement in birth rate after laparoscopic resection or ablation of endometriosis.
Our study suggests that the inverse relationship between stage of the endometriosis at diagnosis and fecundability seems to be present even at the earliest point in the natural life cycle of endometriosis.
Conclusion
In our cohort study, even at the earliest point in the natural life cycle of endometriosis there is an inverse relationship between stage of disease at diagnosis and fecundability.
==== Refs
Evers JL The defense against endometriosis Fertil Steril 1996 66 351 353 8751728
Pritts EA Taylor RN An evidence-based evaluation of endometriosis-associated infertility Endocrinol Metab Clin North Am 2003 32 653 667 14560892 10.1016/S0889-8529(03)00045-8
Batt RE Mitwally MF Endometriosis from the menarche to mid-teens: pathogenesis and prognosis, prevention and pedagogy J Pediatr Adolesc Gynecol 2003 16 337 347 14642954 10.1016/j.jpag.2003.09.008
Schenken RS Modern concepts of endometriosis. Classification and its consequences for therapy J Reprod Med 1998 43 269 275 9564660
The American Statistician 1998 52 119 126
Missmer SA Cramer DW The epidemiology of endometriosis Obstet Gynecol Clin North Am 2003 30 1 19 12699255 10.1016/S0889-8545(02)00050-5
Laufer MR Sanfilippo J Rose G Adolescent endometriosis: diagnosis and treatment approaches J Pediatr Adolesc Gynecol 2003 16 S3 S11 12742180 10.1016/S1083-3188(03)00066-4
Arruda MS Petta CA Abrao MS Benetti-Pinto Time elapsed from onset of symptoms to diagnosis of endometriosis in a cohort study of Brazilian women Hum Reprod 2003 18 756 759 12660267 10.1093/humrep/deg136
Tokushige M Suginami H Taniguchi F Kitaoka Y Laparoscopic surgery for endometriosis: a long-term follow-up J Obstet Gynaecol Res 2000 26 409 416 11152325
Marcoux S Maheux R Bérubé S Laparoscopic surgery in infertile women with minimal or mild endometriosis. Canadian Collaborative Group on Endometriosis N Engl J Med 1997 337 217 222 9227926 10.1056/NEJM199707243370401
Parazzini F Ablation of lesions or no treatment in minimal-mild endometriosis in infertile woman: a randomized trial. Gruppo Italiano per lo Studio dell' Endometriosis Hum Reprod 1999 14 1332 1334 10325288 10.1093/humrep/14.5.1332
| 15845149 | PMC1131922 | CC BY | 2021-01-04 16:37:13 | no | Reprod Biol Endocrinol. 2005 Apr 21; 3:14 | utf-8 | Reprod Biol Endocrinol | 2,005 | 10.1186/1477-7827-3-14 | oa_comm |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.