Datasets:

health360

/

Healix-Shot

like 5

Follow

Health 360 7

publicly maintained biological databases, which are typically funded by grants with limited resources to invest in 'infrastructure'. Findings To identify biocuration requirements, we carried out a detailed analysis of biocuration pipelines in preparation for the workshop. At the workshop, there were presentations from the organizers, but also from a number of groups experimenting with text mining and the curation workflow. The workshop also included a discussion session where biocurators and developers were able to discuss the challenges from the perspectives of text mining and biocuration. Following the workshop, we surveyed Pre-workshop analysis of biocuration workflows A major stumbling block for the application of text mining tools is the need to integrate any new tool into the curation workflow, and to tailor it to produce the kind of output needed by that database. To understand better the needs from the perspective of the biocurators, we undertook a detailed study of the biocuration workflow for eight biological databases listed in *bold in Table 2. Our series of interviews showed that detailed workflow differs from database to database, reflecting differences in requirements, volume of literature to be curated, length of time the databases have been operating and the scale and complexity of the curation activities. For example, some databases require the annotation of additional entities and relations relevant to the experimental setup, such as tissue types and cell lines, as well as patient-related information. Many databases only curate findings that have experimental evidence provided in the article. Access to and processing of textual materials may be a problem, particularly for tables and figures, and for information provided in the article's supplementary material. Workshop and follow-up At the workshop, the organizers summarized their findings on biocuration workflows and provided an overview of text mining terms and methods. This was followed by talks addressing practical experiences applying text mining to biocuration focused on two themes: people who had built tools that had the potential to make a contribution to biocuration work, by Lourenço (14) and Wiegers (7); also two groups reported on their successes and failures in applying text mining to biocuration work: Dowell (15) and Veuthey (16). In addition, Chatr-aryamontri reported on an experiment with author curation for the MINT database (17) and Cohen presented Bada and Hunter's talk on annotation (18). The final segment of the workshop was devoted to discussion, including an informal poll to get a biocurator 'wish list'. A number of biocurators expressed interest in having text mining tools capture other kinds of information, such as phenotype, chemicals or Gene Ontology (GO) terms. Survey of biocurators The initial pre-workshop interviews with curators and the lively discussion at the well-attended workshop at the Biocuration Conference motivated the workshop organizers to explore further the integration of text mining into the biocuration workflow. Following the workshop, the organizers put together a survey on current annotation processes and existing bottlenecks, in order to get more detailed insight into biocuration practices, experiences with text mining and priorities for new tools from the biocurator perspective. The survey covered four areas: (i) information about the curator and curation task; (ii) information about the curation workflow, including article selection, strategy for curating individual abstracts/articles and bio-entities to Table 2). The key findings from the responses to the survey were as follows: biocurators are adopters of text mining technology. Over 70% had tried text mining, and almost 50% were using it in some form. The most widely used system was Textpresso, with 7 out of 28 curators using it for some aspect of curation (survey question 8); the application of greatest interest to curators was document selection and prioritization: 19 out of 27 curators responded that they make or would make heavy (14) or moderate (5) use of text mining for this purpose (survey question 9); identification of underlying evidence was also of great interest: 19 out of 27 curators would make heavy (9) or moderate (10) use of this (survey question 9); and aids to link biological entities to underlying biological resources, including ontological resources were also of high interest: curators would make heavy (8) or moderate (10) use of aids to link to resources such as EntrezGene or GO (survey question 9). The survey also identified a number of interesting issues including the following. Ability to handle full text was a top priority; 27 out of 29 respondents curated from full text routinely (21) or as needed (6) (survey question 4). The need to handle full text imposed related requirements, including ability to handle multiple file formats (Microsoft Word .doc, Adobe Acrobat .pdf, Excel .xls), as well as access to and persistence of supplementary materials. Curation from figures and tables was a standard practice (23 and 24 out of 24 respondents, respectively, in response to survey question 5). Ontologies and standardized terminologies are in widespread use across diverse organisms and tasks. For example, 23 out of 29 respondents were using GO (question 7); other frequently mentioned resources included EntrezGene, ChEBI, PSI-MOD, UniProt and the Plant Ontology. Interestingly, a number of groups were doing phenotype or anatomy, but each group was using a species-specific vocabulary. There was strong interest for using text mining tools in batch processing mode (25 out of 28 respondents said that they would use this feature moderately/ frequently/all the time-question 10). However, 22 out of 25 respondents also said that they would use interactive tools moderately or more frequently. Discussion Adaptation Biocuration workflows have important commonalities and differences. Commonalities include document triage or prioritization, extraction and linkage of important biological entities, and extraction of relations and the underlying evidence for the relations. However, despite these commonalities, each biocuration workflow is different-in its inventory of biological entities, in its designation of what is 'curation-relevant', in the way that articles are prioritized for curation (by journal, by gene or protein, by novelty, etc.) and in how the workflow is divided among curators. Curators expressed a need for tools that could be easily adapted to the specific needs of their workflow and database, such as extensible lexicons that could be edited to include new relevant terms or to exclude terminological resources not relevant to the task. Another need was for tools that could tag the database-specific inventory of biological entities and relations, including numerical descriptions and parameters such as kinetic information. If adaptation is needed, then a key question is: who is responsible for doing the adaptation-the tool developer or the curation team? Adaptation is a complex process and requires well-engineered, well-documented software, as well as sophisticated users/developers on the curation team. As mentioned above, it may require the construction of new lexicons and synonym lists, the writing of new hand-crafted patterns (for rule-based systems) and-for machine learning based systems-the 'training' of the system on application specific training data. Acquiring such training data and doing the training requires familiarity with annotation tools as well as experience in machine learning-based systems for natural language applications. A developer supporting a specific curated database may not have the time or expertise required to adapt natural language processing tools to the specific needs of their database. Site-specific adaptation would also require each database to maintain its own curation pipeline and associated software. This could make it more difficult for curated databases to leverage 'general purpose' tools and could ultimately slow progress by making it more laborious to incorporate new tools or to upgrade the pipeline supporting the curation workflow. Due to these issues, adaptation requires close cooperation between the tool development team and the adopters of the tool. Literature access Literature access is still a stumbling block for both biocurators and text mining developers. As noted above, curators need access to the full articles, including figures, tables and supplementary materials. Many research groups developing text mining tools have focused on abstracts, because these are easily accessible and can be downloaded as ASCII or XML. In contrast, access to full journal articles is complicated by difficulties in handling pdf and obtaining xml versions of the articles, as well as intellectual property issues. Although there are an increasing number of open access publications, curation teams need access to all of the relevant literature, not just to those journals that are more easily accessible. What curators want from text mining tools Through interviews, presentations at the workshop and the follow-up post-workshop survey, we have identified some curator desiderata. Curators wanted tools that were easy to use, easy to install and easy to maintain by the intended end user (ideally, a developer associated with the curation team, who will not necessarily be an expert in text mining or natural language processing). The tools do not have to be perfect, but they need to complement (not replace) the biocurator's function. A number of curation groups indicated that they would use the tools to do an initial batch processing, followed by biocurator validation, where the biocurator makes a yes/no decision and avoids having to type or look names up in a large database. Another important use was linking mentions of biological entities in text with the correct identifiers in biological databases, as well as linkage to the appropriate ontology terms. A number of curators felt that they would like text mining tools to aid in identifying and prioritizing papers for curation, to avoid wasting time on papers that did not have 'relevant' (e.g. curatable or novel) results. They also wanted tools to identify the sections of full-text papers containing curatable information. Biocurators were also concerned about interoperability and data exchange, including formats that could communicate with other bioinformatics resources, either through the use of Web services, or via links to external resources and databases. Curators were interested in using text mining tools that could produce confidence scores, linkage to evidence passages in the text and ranking of automatically generated results, together with visualization aids, such as a customizable color-coding scheme for highlighting different levels annotations contained in a given article under curation. What text mining developers need from curators The biocurator community can assist by providing formalized descriptions of their workflows. Findings from our initial workflow studies indicated that each database may have a unique workflow-since databases typically differ in their criteria for what gets curated and in what order they do the various steps. Instrumentation of the curation interfaces would make it possible to gather data from curators on timing, throughput and patterns of use. This, in turn, would help to identify the major 'choke points' in the workflow. Based on such a workflow description and associated data on patterns of usage, the curators could work with the tool developers to identify appropriate insertion points for text mining in the workflow. From the text mining tool developer point of view, it would be useful to have curators provide a more detailed description (and examples) of data selected as relevant and data designated as nonrelevant during the curation process. If annotations were saved on textual data that had been manually reviewed but deemed not curation relevant, this could serve as negative training data, crucial for the development and evaluation of text mining applications. It would also support comparison of current database content and automatically extracted annotations. Page 8 of 10 Original article Database, Vol. 2012, Article ID bas020, doi:10.1093/database/bas020 through workshops and challenge evaluations, we expect to see significant progress in this critical area. There is now substantial momentum behind these interactions. Since the workshop in the spring of 2009, there have been two additional evaluations that have continued the exploration of these issues, with a third workshop planned for April 2012 and BioCreative IV planned for spring 2013. BioCreative II.5 (11) (October 2009) compared curation of FEBS Letters articles on protein-protein interaction by authors, expert biocurators and automated systems. This work was inspired in part by community discussions around structured digital abstracts and the feasibility of author curation (19,20). The evaluation was organized with active participation of FEBS Letters, including both the editor (Gianni Cesareni) and the publisher (Elsevier), as well as a number of authors who participated in the author curation experiment. Two findings of relevance were that (i) authors had particular difficulty with the protein normalization step (the assignment of an appropriate UniProt identifier to a protein described in the article) and (ii) a post hoc combination of author plus automated system outperformed either one individually-in part, because the authors and the

automated systems made very different kinds of mistakes. These results suggest that existing automated systems may be good enough now to help authors link genes or proteins mentioned in an article to the correct unique identifier; this might be a good candidate insertion point that could save time even for an experienced biocurator. BioCreative III was held in September 2010, introducing a new 'Interactive Annotation Task', inspired in part by the findings from the April 2009 workshop. This interactive task focused on identifying which genes were being studied in an article and linking those genes to standard database identifiers. The task was designated as a demonstration task, with the goal of laying the groundwork for a rigorous evaluation of an interactive system for BioCreative IV (planned for spring 2013). To provide input from the end user and biocurator perspective, a User Advisory Group was organized to assess the six participating interactive systems and provided detailed feedback to the developer teams (12). In addition to the above activities, the dialog is broadening to include the scientific publishing community, which is becoming an increasingly active partner. Both BioCreative II.5 and BioCreative III had active participation from the publisher community, and the Intelligent Systems for Molecular Biology (ISMB) conference has held successful sessions on Scientific Publishing for the last 3 years. However, the importance of this topic is not confined to the text mining, biocuration and scientific publishing communities. The maintenance of timely, high-quality computable resources provided by the growing number of curated databases derived from the scientific literature is critical to the entire scientific enterprise. A direct follow on to the April 2009 Biocuration workshop will be held in association with the fifth International Biocuration Conference (spring 2012). This is organized as a BioCreative Satellite Workshop, with organizers including biological curators (Wu, Arighi from PRO and UniProt; Mattingly and Wiegers from the Comparative Toxicogenomic Database). The workshop will consist of three Tracks: Triage (Track 1): a collaborative biocuration-text mining development task for document prioritization for curation; Biocuration Workflows (Track 2): a collection of detailed descriptions of biocuration workflows and identification of insertion points for text mining, from the perspective of biocurators; and Interactive Text Mining (Track 3): an interactive text mining and user evaluation task, with evaluation by biocurators. Each of these tracks will have 6-9 participating groups. The spring 2012 workshop described above will set the stage for BioCreative IV, to be held in the spring of 2013. We believe that these activities are greatly increasing communication among the diverse communities involved in biocuration. This, in turn, will lead to improved tools inserted into the biocuration workflow-driven by the needs and the insights of the biocurators. Supplementary Data Supplementary Data are available at Database Online. Periodontitis is associated with higher subclinical atherosclerosis in patients with systemic lupus erythematosus Abstract Aim To determine periodontitis prevalence in patients with systemic lupus erythematosus (SLE) and to assess whether periodontitis in SLE patients is associated with a greater subclinical atherosclerosis. Methods An observational case–control study was conducted in SLE (cases) and patients without any rheumatic diseases (controls), matched for sex. Sociodemographic and cardiometabolic variables were gathered, and SLE activity was assessed through several indexes. Periodontal examination registered probing pocket depth, clinical attachment level, bleeding on probing, plaque index, and tooth loss. Subclinical atherosclerosis was assessed by measuring the carotid–femoral pulse wave velocity (PWV) by Doppler velocimetry, homocysteine levels, C‐reactive protein (CRP), and erythrocyte sedimentation rate (ESR). Bivariate analyses and logistic regression were used to assess the association of any of the studied variables with SLE. Results Seventy‐one cases and 72 controls were included in the study. Thirty‐nine SLE patients (54.9%) were diagnosed with periodontitis, compared with 16 controls (22.2%). High levels of PWV (≥7.7 m/s, 75th percentile) were shown by 44.3% of the cases vs. 22.4% of the controls (p = .011). Among SLE patients, those with periodontitis showed higher PWV values (8.1 ± 1.52 vs. 7.16 ± 1.11 m/s, p = .006) and higher homeostasis model assessment index (indicative of insulin resistance) (1.7 ± 0.73 vs. 2.92 ± 3.05, p = .028) compared to those with periodontal health. Logistic regression showed that waist circumference (OR 1.06, 95% CI 1.01–1.12, p = .015); ESR (OR 1.09, 95% CI 1.03–1.16, p = .003); and bleeding on probing (OR 1.1, 95% CI 1.01–1.19, p = .018) were associated with the risk of SLE. Conclusion Systemic lupus erythematosus patients showed a higher periodontitis percentage than controls. Higher PWV values were found in SLE patients with periodontitis, indicating a higher prevalence of subclinical atherosclerosis. Patients with higher gingival bleeding showed a higher risk of SLE. | INTRODUC TI ON Systemic lupus erythematosus (SLE) is a prototype of a systemic autoimmune disorder characterized by inflammation, which mainly affects women of reproductive age. 1 Early diagnosis and the improvement of the available treatments have led to a greater survival of these patients, mainly through a good control of infections and SLE outbreaks. 2 Cardiovascular disease (CVD) is currently the leading cause of death, in the period beginning 5 years after initial diagnosis, due to the high prevalence of atherosclerosis and faster atherosclerosis progression among SLE patients. 3 A good cardiovascular risk assessment is necessary in order to achieve lower CVD mortality in SLE patients. 4 Periodontitis is an inflammatory disease, associated with and probably caused by a multifaceted dynamic interaction among dysbiotic biofilm, host immune responses, hazardous environmental exposure, and genetic propensity. 5 A controversial two-way relationship has been proposed between periodontitis and SLE, and the reported rates of periodontitis in SLE patients have been compared to the ones found between periodontitis and type 2 diabetes mellitus. 6 The use of systemic immunosuppressive drugs in patients with SLE may contribute to an earlier manifestation of periodontitis. Periodontitis, on the other side, would aggravate SLE-related cardiovascular complications such as endothelial dysfunction and atherosclerosis. Two metanalyses on the relationship between periodontitis and SLE reported contradictory results. One meta-analysis performed on four studies reported a risk of periodontitis in SLE cases compared with controls significantly greater, with a risk ratio of 1.76, but no differences were found in clinical periodontal parameters, like probing pocket depth (PPD) or clinical attachment loss. 7 The other metanalysis showed a significant association between periodontitis and SLE (OR 5.32), and SLE patients presented higher bleeding on probing (BOP) and higher mean clinical attachment loss compared with controls. 8 Scientific evidence has widely linked periodontitis with a higher incidence and prevalence of CVD and specifically with acute myocardial infarction. Marfil-Álvarez et al. 9 demonstrated that severe and extensive periodontitis led to more severe acute myocardial infarction. This relationship has been recently explained in two metanalyses by atherosclerosis. A higher atherosclerotic process has been found in patients with periodontitis, measured by pulse wave velocity (PWV). 10,11 PWV is the gold standard and the most validated noninvasive method for the assessment of arterial stiffness. 12 Endothelial damage and inflammation are one of the first steps in CVD development, and many novel biomarkers related to those initial steps have been investigated. Homocysteine has been proposed as a disruptor of the endothelial function, leading to vessel damage, atherosclerosis progression and, ultimately, to CVD. 13 Maintained high erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) levels are other biomarkers associated with atherosclerosis and CVD predictors. 14,15 Our study hypothesis is that patients with SLE have a higher risk of presenting periodontitis. Also, patients with SLE and periodontitis present more CVD affections due to an aggravated atherosclerotic process. The objectives of our study were to determine the prevalence of periodontitis in patients with SLE and assess whether the diagnosis of periodontitis in SLE patients is associated with a more severe atherosclerotic process, measured by PWV, ESR, and CRP, and also determine whether periodontitis or any periodontal clinical variable could be considered as risk factors for SLE. | Participants An observational case-control study was conducted on patients from the Autoimmune Diseases Unit at the University Hospital | Variables Sociodemographic and cardiometabolic variables were gathered from each participant in both groups: sex, age, higher education level, waist circumference (cm), smoking habit, menopausal status, and sedentary lifestyle according to the level of daily physical activity (no physical activity or less than 30 min/day). CVD diagnosis was considered positive when subjects presented at least one of the following parameters in their medical record: coronary heart disease (acute myocardial infarction or angina), valvular heart disease, chronic heart failure, or a cerebrovascular accident. A family history of premature CVD was defined as the presence of a first-degree relative who had experienced an acute myocardial infarction or cerebrovascular accident before the age of 55 in men or 65 in women. Blood pressure was measured twice using a calibrated automatic device (HEM-7051T, Omrom Health Care). Patients were considered hypertensive when systolic blood pressure (SBP) was higher than 140 mmHg or diastolic blood pressure (DBP) was higher than 90 mmHg. Carotid-femoral PWV was measured by Doppler velocimetry (Complior-Analyse, ALAM-MEDICAL). The cutoff values for PWV were established as the 75th percentile of the control group. Cardiovascular risk score was determined using the Framingham criteria. 15 Treatment with statins, antihypertensive drugs, immunosuppressants, and hydroxychloroquine was also registered in all participants. SLE Disease Activity Index (SLEDAI) 18 where "i" is the site, "d" is the PPD of the site in mm, "n" is the absolute frequency of the sites, and "t" is the number of remaining teeth. 9 BOP was determined as the percentage of bleeding teeth, following the criteria by Ainamo and Bay, 21 and the presence of plaque was registered scored using plaque disclosing tablets according to plaque index by O'Leary. 22 At the same appointment or in morning next to it, blood samples were gathered from each participant to determine biochemical and immunological parameters. Total cholesterol (mg/dl), HDL-cholesterol (mg/dl), LDL-cholesterol (mg/dl), and triglycerides (mg/dl) were assessed, as well as fasting blood glucose (mg/dl). Creactive protein (mg/dl), ESR (mm/h), uric acid (mg/dl), homocysteine (µmol/L), and insulin (µU/ml) levels were assessed. Homeostasis Model Assessment (HOMA) index for insulin resistance was calculated. 23 All samples were processed, and all laboratory tests were performed at the clinical analysis laboratory of the "Virgen de las Nieves" University Hospital in Granada (Spain). | Statistical analysis A conservative sample size calculation was made, taking into account that the prevalence of periodontitis according to recent evidence in general population is 35%, and in patients with SLE is 60%. The 25 Taking these values into account, and in order to achieve a statistical power of 85%, necessary to detect differences in the null hypothesis test H0: p1 = p2, using a bilateral Chi-squared test for two independent samples with a significance threshold of 5%, A minimum of 71 patients would be needed in each group, making a total of 142 subjects in the study. | RE SULTS Seventy-one patients with SLE and 72 control subjects were enrolled in the study from January to December of 2015, and there were no losses to follow-up during the study ( Figure S1). All participants were In comparison with SLE patients, control subjects presented lower waist circumference, higher levels of physical activity, and higher percentage of participants with higher education (university). | DISCUSS ION The carotid-femoral PWV is a measure of subclinical atherosclerosis and the gold standard for determining arterial stiffness. In our study, Their results showed a significant association between periodontitis and SLE, concluding that there is a causal relationship between periodontitis and SLE and that propose periodontitis as a prelude of these autoimmune diseases rather than a consequence. 34 We explored a different way that could be plausible in a bidirectional relationship between SLE and periodontitis, as proposed by other authors. Wang et al. found that Treponema denticola and Porphyromonas gingivalis and the combination of both in patients with SLE was associated with increased titers of anti-β2-glycoprotein I and anti-cardiolipin antibodies. 35 Porphyromonas gingivalis-specific genotypes fimA-Ib, fimA-II, and fimA-IV, the ones more associated with periodontitis, were found as the most prevalent ones in SLE patients compared with controls. 36 Aggregatibacter actinomycetemcomitans is also a bacterial species known to be a trigger of the autoimmune response in diseases such as SLE and rheumatoid arthritis. 7 Of the three independent variables that showed to be risk factors for SLE in our study (waist circumference, ESR, and BoP), according to the results of the adjusted logistic regression

model, BoP is the variable that showed the greater OR (1.102) for suffering SLE. In a recently published randomized clinical trial, the authors reported that periodontitis treatment improved the response to immunosuppressive therapy in SLE patients. This result, however, must be considered with caution, since the patients of the study were diagnosed with periodontitis using only the bleeding gingival index as criterion. The mean PPD was 1.7 mm in the group of treated patients, and the sample size and duration of the study were very short (32 patients during 3 months). 25 A limitation of our study is that SLE patients were, on average, older than controls (40 years vs. 34 years), which could have affected the prevalence of periodontitis; however, we think that the age group where it occurs the difference does not affect this frequency measurement. Another potential limitation is that the SLE group could not be representative of general SLE patients from a Caucasian population, due to a selection bias, but the demographic characteristics of this group are similar to the ones of other previously published studies with larger samples. 6,37,38 Our study, the second largest in the European population, provides the information that periodontitis is a treatable factor, so early intervention can prevent at least to a certain extent the development of an atherosclerotic process that constitutes the basis of future cardiovascular events in patients with SLE. | CON CLUS ION In this case-control study, periodontitis was more common in patients with SLE compared with controls. PWV is higher in patients with SLE with periodontitis, which indicates a higher percentage of subclinical atherosclerosis. Gingival BOP was the only modifiable periodontal variable that was associated 1.102 times with suffering SLE. ACK N OWLED G EM ENTS We would like to thank the patients from the Unit of Autoimmune Diseases of the "Virgen de las Nieves" University Hospital. This investigation has not received funds from any private entity. All procedures in this were performed from the regular care, with resources of the Spanish National Health System. All authors declare no conflicts of interest, and all authors have approved the final article. Open access funding is provided by Universidad de Granada/CBUA. DATA AVA I L A B I L I T Y S TAT E M E N T The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions. Concomitant boost IMRT-based neoadjuvant chemoradiotherapy for clinical stage II/III rectal adenocarcinoma: results of a phase II study Aim This study was designed to evaluate the efficacy and toxicities of concomitant boost intensity-modulated radiation therapy (IMRT) along with capecitabine and oxaliplatin, followed by a cycle of Xelox, in neoadjuvant course for locally advanced rectal cancer. Materials and methods Patients with histologically confirmed, newly diagnosed, locally advanced rectal adenocarcinoma (cT3-T4 and/or cN+) located within 12 cm of the anal verge were included in this study. Patients received IMRT to the pelvis of 50 Gy and a concomitant boost of 5 Gy to the primary tumor in 25 fractions, and concurrent with oxaliplatin (50 mg/m2 d1 weekly) and capecitabine (625 mg/m2 bid d1–5 weekly). One cycle of Xelox (oxaliplatin 130 mg/m2 on d1 and capecitabine 1000 mg/m2 twice daily d1–14) was given two weeks after the completion of chemoradiation, and radical surgery was scheduled eight weeks after chemoradiation. Tumor response was evaluated by tumor regression grade (TRG) system and acute toxicities were evaluated by NCI-CTC 3.0 criteria. Survival curves were estimated using the Kaplan-Meier method and compared with Log-rank test. Results A total of 78 patients were included between March 2009 and May 2011 (median age 54 years; 62 male). Seventy-six patients underwent surgical resection. Twenty-eight patients underwent sphincter-sparing lower anterior resection and 18 patients (23.7%) were evaluated as pathological complete response (pCR). The incidences of grade 3 hematologic toxicity, diarrhea, and radiation dermatitis were 3.8%, 10.3%, and 17.9%, respectively. The three-year LR (local recurrence), DFS (disease-free survival) and OS (overall survival) rates were 14.6%, 63.8% and 77.4%, respectively. Initial clinical T stage and tumor regression were independent prognostic factors to DFS. Conclusion An intensified regimen of concomitant boost radiotherapy plus concurrent capecitabine and oxaliplatin, followed by one cycle of Xelox, can be safely administered in patients with locally advanced rectal cancer, and produces a high rate of pCR. A prognostic score model is helpful to distinguish different long-term prognosis groups in early stage. Introduction Preoperative chemoradiotherapy (CRT) followed by total mesorectal excision (TME) is the standard treatment for patients with locally advanced rectal cancer (LARC). Some significant benefits of neoadjuvant CRT, such as better local control and sphincter preservation, have been shown in patients with stage II/III rectal cancer [1][2][3]. To obtain a better tumor response, elevating treatment dose has been considered a feasible method. EORTC22921 and FFCD9203 studies showed that preoperative radiotherapy combined with fluorouracil (5-FU) can significantly improve the treatment effect compared with radiotherapy alone [3,4]. In a retrospective analysis of 3,157 patients enrolled in seven randomized Phase III trials and 45 Phase II trials, the use of continuous infusion 5-FU, a second drug based on 5-FU and a higher radiation dose was associated with higher rates of pCR [5]. However, the next five randomized phase III trials, ACCORD 12/0405-Prodige 2 [6], STAR-01 [7], NSABP R-04 [8], CAO/ARO/AIO-04 [9] and PETACC-6 [10], demonstrated conflicting results as to whether oxaliplation increased the rate of pCR. In our center, most patients receiving neoadjuvant CRT were clinical T4 or N+, and might have more opportunities to benefit from a high intensity treatment, whether chemotherapy or radiotherapy [11][12][13]. To decrease the additional toxicities from a high-dose treatment, especially diarrhea, intensity modulated radiotherapy (IMRT) was used to lessen radiation-associated toxicities by decreasing the volume of high irradiation dose of surrounding normal tissues [14][15][16]. IMRT allows higher radiation doses to be focused on regions within the tumor while minimizing the dose to surrounding normal critical structures. The data from dosimetric studies of IMRT in rectal cancer are encouraging. Compared to conventional 2D or 3D radiation therapy, IMRT showed similar target coverage with reduced dose to the small bowel, bladder, pelvic bone and femoral heads [14,17,18]. By decreasing the dose delivered to normal structures, IMRT may provide a potential for increasing treatment dose to improve tumor response. Therefore, we designed this study to examine the use of IMRT, escalating the primary lesion's dose to 55 Gy together with the whole pelvis dose of 50 Gy in 25 fractions, along with weekly capecitabine and oxaliplatin. Two weeks after the end of chemoradiotherapy, a cycle of Xelox (capecitabine and oxaliplatin) was prescribed before surgery. The efficacy and toxicity of this modality were evaluated to explore the feasibility of high-dose intensity in preoperative treatment. This phase II study was approved by our institutional review board. Eligibility criteria Patients with histologically confirmed, newly diagnosed, locally advanced rectal adenocarcinoma (cT3-T4 and/or cN+) located within 12 cm from the anal verge were included in this study at the Fudan University Shanghai Cancer Center. All patients were ≥ 18 years of age and had a Karnofsky Performance Status score of ≥ 60, no evidence of distant metastases, adequate bone marrow function (leukocyte count > 4,000/mL and platelet count > 100,000/mL), and adequate renal and hepatic function (creatinine clearance > 50 mL/min and bilirubin ≤ 2 mg/mL). Patients were excluded if they were older than 75 years of age, had undergone previous pelvic radiotherapy or previous chemotherapy, or had previous or synchronous tumors other than nonmelanoma skin cancer. Patients suffering of the following medical conditions were also ineligible: ischemic heart disease, inflammatory bowel disease, malabsorption syndrome, peripheral neuropathy, or psychological disorders. Signed informed consent was obtained from all patients before inclusion on this study. The institutional review board of Fudan University approved the study. Baseline evaluation Pretreatment evaluation was performed within two weeks before initiation of chemoradiation. The evaluation included a complete history and physical examination, including digital rectal examination, complete blood count, hepatic and renal function tests, tumor marker measurement, colonoscopy and biopsy, computed tomography (CT) of the thorax and abdomen, magnetic resonance imaging (MRI) of the pelvis, and, in selected patients, endorectal ultrasound. All patients were clinically staged with the AJCC 7th version manual. Combined chemoradiotherapy Intensity modulated radiation therapy (IMRT) All patients were immobilized in the prone position using a belly board and underwent a non-contrast-enhanced, planning CT with 5-mm slices from the L3-L4 junction to 2 cm below the perineum. The image data were transferred to the PINNACLE planning system (Philips Radiation Oncology Systems, Milpitas, CA). The definitions of volumes were in accordance with the ICRU Report #83 [19]. The gross tumor volume (GTV) was defined as all known gross disease determined from CT and MRI. The clinical target volume 1 (CTV1) included the gross tumor volume and the corresponding mesorectum plus 2 cm cranio-caudally. The CTV2 included the CTV1 plus the entire mesorectum, entire pre-sacral space, internal iliac nodes and high-risk anatomical and nodal sub-sites, based on the distance of the tumor from the anal margin [20]. Based on our institution set-up data, the planning target volume (PTV) was defined as the CTV with 10-mm margins superiorly and inferiorly and 8-mm margins in all other directions. Organs at risk (OARs) were contoured as follows: 1) the small intestine was defined as all intestinal loops below the sacral promontory (rectosigmoid junction excluded); 2) femoral heads were contoured from the cranial extremity to the level of the lower margin of ischial tuberosities; and 3) the bladder was contoured entirely with no distinction between the wall and its content [16]. The IMRT plans were generated using the inverse planning module of PINNACLE for a 6-MV liner accelerator, with five to seven coplanar fields. The planned doses to the PTV1 and PTV2 were 55 Gy and 50 Gy, respectively, in 25 fractions, five times per week (Monday through Friday) for five weeks. The D2%, D50%, and D98% to PTV1 and PTV2 were set at 52.25 Gy and 55 Gy, 57.75 Gy and 47.5 Gy, and 50 Gy and 52.5 Gy, respectively. The doses of the OARs were set as low as possible and had to at least meet the following constraints: bladder, ≥ 45 Gy in 15% volume and ≥ 40 Gy in 40% volume; femoral heads, ≥ 45 Gy in 25% volume and ≥ 40 Gy in 40% volume; and small bowel, ≥ 45 Gy in 65 cc volume, ≥ 40 Gy in 100 cc volume, and ≥ 35 Gy in 180 cc volume. The positioning and isocenter of each patient were verified on electronic portal imaging device (EPID) films for the anterior and lateral gantry positions by visually comparing the digitally reconstructed radiographs. Concurrent and neoadjuvant chemotherapy Capecitabine combined with oxaliplatin was administered concurrently with pelvic radiation. Capecitabine was given at a dose of 625 mg/m 2 twice daily from Monday to Friday throughout the entire course of IMRT. Oxaliplatin at a dose of 50 mg/m 2 was administered weekly during the five-week course of radiotherapy. Two weeks after concurrent chemoradiation, one cycle of Xelox (oxaliplatin 130 mg/m 2 on d1 and capecitabine 1000 mg/m 2 twice daily d1-14) was administered ( Figure 1). Surgery and histopathology Surgery was scheduled eight weeks after the completion of CRT. Total mesorectal excision (TME) was mandatory, whereas the form of surgery (anterior resection or abdominal-perineal resection) and whether a temporary colostomy should be performed were decided by the surgeon. All lymph nodes were examined according to standard procedures. If the number of lymph nodes was less than 12, two pathologists were needed to sign to ensure the reliability of the detection result. The circumferential rectal margin (CRM) was assessed according to the method of Quirke et al. [2], and a margin of < 1 mm was considered CRM-positive. All sections of the surgical specimens were reviewed by two pathologists. The pathologic stage (ypTN) was recorded according to the Union for International Cancer Control (UICC) TNM system. Tumor regression grading (TRG) was evaluated according to the criteria by Dworak et al. as follows [21]: Grade 0, no regression; Grade 1, dominant tumor mass with obvious fibrosis and/or vasculopathy; Grade 2, dominant fibrotic changes with few tumor cells or groups (easy to find); Grade 3, very few (difficult to find microscopically) tumor cells in fibrotic tissue

with or without mucous substance; and Grade 4, no tumor cells, only a fibrotic mass (total regression or response). Adjuvant chemotherapy and follow-up All patients were recommended to receive postoperative chemotherapy regardless of pathological stages. Adjuvant chemotherapy was recommended consisting of five cycles of Xelox. Patient follow-up was scheduled every three months during the first two years, and then every six months over the next three years. After five years, the frequency of follow-up was extended to once each year. Figure 1 The workflow of preoperative chemoradiotherapy in patients with locally advanced rectal cancer. Toxicity and measurement Toxicities were evaluated and recorded weekly according to the CTC 3.0 criteria. If grade 3 toxicities occurred, the physicians determined causes and decided the response. In general, the sequence of dose reduction or suspension moved from oxaliplatin to capecitabine to radiotherapy, unless an adverse effect was strongly associated with a particular treatment. Endpoints and statistics The primary endpoint for this trial was pCR rate. This study was a phase II trial of 78 patients to evaluate the treatment feasibility and efficacy of this dosing regimen. Based on a literature review, the pCR rate is approximately 10-15% for patients treated with neoadjuvant CRT. We determined that an experimental arm with a pCR rate of at least 18% would merit further study. In this study, if more than 17 cases were evaluated as pCR, we had 85% power to reject the null hypothesis that our strategy could not reach the pCR of 18%, with a type I error level of 5%. Secondary endpoints included safety, sphincter preservation rate, TRG, LR (local recurrence), DFS (disease-free survival) and OS (overall survival). Sphincter preservation was defined as any procedure in which the rectal tumor was removed while leaving behind the anal sphincter. All characteristics were described by the frequency for classified variables, by mean and standard deviations for normal distributional continuous data, and by the median for non-normal distributional continuous data. Survival time was calculated from the beginning of CRT to the date of event or the last follow-up. Survival curves were estimated using the Kaplan-Meier method and compared with Log-rank test. Cox proportional hazards regression was used for multivariate modeling and for examining the prognostic significance of the variables identified in the models. P values of less than 0.05 were taken to indicate statistically significant differences. Clinical characteristics Between March 2009 and May 2011, a total of 78 patients were included in the study. All patients were diagnosed with locally advanced rectal cancer: 50 with cT3 and 28 with cT4 primary tumor. Lymph node involvements were detected in 75 patients. Of the total 78 patients, 62 were men and 16 were women; the median age was 54 years (range, 30-76 years). Fifty-six patients (71.8%) had tumors located ≤ 5 cm from the anal verge (Table 1). Treatment compliance and acute toxicities All patients completed the prescribed radiation treatment to a total dose of 55 Gy in 25 fractions. The median total radiation duration was 37 days (range, 33-41). All patients completed five weeks of capecitabine, and 48 cases received five cycles oxaliplatin and the rest received four cycles. In addition, all patients received a scheduled single cycle of Xelox two weeks after the completion of chemoradiotherapy without dose adjustment. Most of the adverse events during CRT were mild (grade 1 or 2). No grade 4-5 toxicities were observed. The most common grade 3 toxicity was radiation dermatitis (17.9%), while grade 3 diarrhea and hematological toxicities were evaluated in eight (10.3%) and three cases (3.8%) ( Table 2). Surgical procedures and pathological response Seventy-six patients underwent a surgical resection according to the schedule. One patient refused surgery because of good response, and another case did not receive an operation because of being evaluated as unresectable lesions. The median interval between the completion of CRT and primary tumor surgery was 52 days (range, 46-67 days). Twenty-eight patients (36.8%) underwent sphincter-sparing or poor responders (defined as TRG 1-2). More than or equal to 12 lymph nodes were found in half of the patients. Lymphatic/vascular invasion and neural invasion were confirmed in five and eight cases, respectively ( Table 3). The overall rate of postoperative complications was 17.1%. Delayed sacral-wound healing, postoperative bleeding and anastomotic leakage occurred in 9, 3 and 1 patients, respectively. Follow-up With a median follow-up of 30 months (range, 9-48 months), 10 patients were diagnosed with local recurrence and 19 patients were confirmed with distant metastases (5 in the liver, 13 in the lung, and 1 in bone). Fourteen patients died of rectal cancer. For the two patients that did not receive surgery, one patient was confirmed of tumor failure at 27 months and died 28 months after the beginning of CRT, and the other patient did not present any evidence of failure at the last visit of 9 months. The 3-year LR, DFS and OS rates were 14.6%, 63.8% and 77.4%, respectively ( Figure 2). Univariate and multivariate analysis for LR and DFS All potential prognostic factors, including age, gender, distance from anal verge, cT stage, cN stage, ypT stage, ypN stage and TRG score were evaluated using the Kaplan-Meier method (compared with Log-rank test). cT stage and pCR status demonstrated a correlation with LR ( Figure 3). No LR was observed in pCR cases. In the next multivariate Cox regression analysis, only cT stage was left in the model. cT stage, ypT stage and TRG score exhibit a correlation with DFS. YpT stage was excluded in the further multivariate Cox regression analysis. Hazard ratios of these two factors (cT stage and TRG score) were very close to one another (Table 4). A prognostic scoring system was produced, with each of the two unfavorable prognostic factors allocated one point, including cT4 stage and poor responder. Use of the scoring system led to the identification of three risk groups: low risk (score of 0), intermediate risk (score of 1), and high risk (score of 2). The three-year DFS rates were significantly different among the groups: the low-risk group, 81.1%; intermediate-risk group, 60.0%; and the high-risk group, 28.6% (P = 0.000, Figure 4). Discussion Our study findings demonstrated that the IMRT technique, which decreases radiation-induced toxicities by a lower high-dose irradiation volume, and an intensified preoperative CRT followed by a cycle of Xelox resulted in a pCR rate of 24.7%. These results are higher than those of studies on preoperative conventional CRT concurrent with oxaliplatin (15-20%) [6][7][8][9][10]. Regarding toxicities, however, the high rate of pCR did not translate into high local control compared to other trials. This may be attributed to two factors, First, our study had a small sample size, and therefore some certain remains. Second, the percentage of cT4 and cN + tumors in our study was significantly higher than some other trials (cT4: 35.9% vs. 5-15%, cN+: 96.2% vs. 40-70%, respectively) [1,3,5,[7][8][9][10][11]. Our study found that the incidence of grade 3 diarrhea, hematologic toxicity, and radiation dermatitis was 10.3%, 3.8%, and 17.9%, respectively. Consistent with our previous study [22], the incidences of diarrhea and hematologic toxicity were slightly lower compared with other reported stage III clinical trials. A significant increase in the incidence of radiation dermatitis was observed in our study, which might be attributed to a lower irradiation field due to a distal rectal tumor location in most cases. Finally, with a median follow-up of 30 months, our data showed that the baseline T stage and treatment response were associated with long-term prognosis. Patients with cT4 stage had a higher chance of LR and distance metastases, but no LR was observed in pCR cases, regardless of initial T stage. In addition, baseline T stage and tumor response had equal effects to predict DFS. Patients with cT3 and good response had a three-year DFS of 81.1%, but for those with cT4 and poor response, the three-year DFS declined to 28.6%. The IMRT technique has been used widely in some solid tumors. In our study, IMRT was used to concomitantly boost the total irradiation dose to 55 Gy for the gross tumor. The rationale was based on several dosimetric studies that showed that IMRT could significantly decrease the surrounding organs' high-dose irradiation volume, especially the small bowel, compared with conventional radiotherapy or 3DCRT [14,15]. Therefore, IMRT provided the potential of elevating the dose to improve the tumor response. This is consistent with results from Hartley's meta-analysis, which included a total of 3157 rectal cancer patients [5]. Our previous study of stage IV rectal cancer demonstrated the feasibility to deliver 45 Gy to the pelvis and a concomitant 10 Gy boost to the gross tumor using IMRT [23]. Oxaliplatin is an effective drug for colorectal cancer when combined with 5-FU. However, the results of five recent large sample phase III trials are unclear as to whether oxaliplatin is an appropriate radiation sensitizer. Three studies, including NSABP R-04 [8], STAR-01 [7] and PETACC-6 [10], all reported that additional oxaliplation based on conventional CRT failed to increase pCR rate and caused more toxicities, especially diarrhea. The ACCORD 12/0405-Prodige 2 trial also reported a higher toxicity rate in the oxaliplatin group, together with a significantly higher pCR rate [6]. The CAO/ARO/ AIO-04 trial was the only phase III trial supporting additional oxaliplatin in preoperative CRT, which showed a better tumor regression in the oxaliplatin group without any additional toxicities [9]. However, our study demonstrated an encouraging pCR and tumor shrinkage, without high incidence of toxicities. This also illustrated that IMRT was an effective method to offset toxicities induced by high-dose CRT. In several previous studies, early surrogate indicator was focused to help to decide the strategy of adjuvant therapy. Tumor regression after CRT, especially pCR, was regarded as an important prognostic factor. The EORTC 22921 trial showed that patients with good response had a significantly good prognosis compared with those with poor response [24]. Capirci's retrospective study of 566 patients with pCR also demonstrated an encouraging prognosis [25]. In a retrospective study by MD Anderson Cancer Center, tumor response was associated with the five-year recurrence free survival, distant-metastasis rate and LR [26]. Thus, after neoadjuvant therapy, conventional adjuvant chemotherapy (oxaliplatin plus fluorouracil) might be over-treatment to patients with good prognosis, but no use to poor responders. Based on our data, a prognostic score model including initial clinical T stage and tumor response might be helpful to determine the adjuvant chemotherapy regimen. Conclusion An intensified regimen of concomitant boost radiotherapy plus concurrent capecitabine and oxaliplatin, followed by one cycle of Xelox, can be safely administered in patients with locally advanced rectal cancer, and produces a high rate of pCR. A prognostic score model is helpful to distinguish different long-term prognosis groups in early stage. Figure 4 Disease-free survival rates by clinical T stage (panels a), ypT stage (panels b), tumor regression (panels c) and prognostic score (panels d). TRG: tumor regression grade. Building personal resilience in primary care paramedic students, and subsequent skill decay Introduction Paramedics are routinely exposed to traumatic incidents that include physical injuries; these events may manifest into psychosocial injury. Proactive and preventive measures have the potential to mitigate the negative impact of exposure to traumatic events. Enhancing an individual’s capacity to effectively manage stressful/adverse life events through an online resilience resource (ORR) offers a promising option for paramedics. The aim of this study is to investigate the initial impact of an ORR on resilience and to explore the potential skill decay following this self-guided online resource among pre-employment paramedic trainees. Methods Through a repeated measures design, 227 primary care paramedics from British Columbia, Canada completed a baseline resilience assessment and ORR. A subset of participants completed follow-up resilience assessments at 3 to 6 month or 9-month intervals. Results Between the baseline and 3-month follow-up tests, results indicate that self-report resilience scores showed a slight improvement. However, as time increased to 6 or 9 months, a statistically significant decrease in resilience scores in comparison to the baseline was observed. Conclusion This study presents evidence to suggest that an educational tool such as an online self-paced training program for building resilience may be an effective strategy for improving short-term personal resilience among primary care paramedic students. Given the gradual skill decay associated with an ORR, we can highlight the temporal limits of resilience training. Developing additional resilience training programs to be delivered throughout students’ pre-employment education may

help reduce skill decay. Introduction Due to the nature of their work, public safety personnel (PSP) such as police officers, firefighters, correctional workers, emergency medical technicians (EMTs) and paramedics are highly likely to be exposed to potentially traumatic and stressinducing events (1). As a result, they are at increased risk of developing post-traumatic stress injuries (PTSI), which may include mental health issues such as symptoms consistent with anxiety, depression, as well as the adoption of maladaptive coping styles (2)(3)(4)(5)(6). In general, PTSI could have deleterious effects on PSP, potentially leading to a reduced quality in occupational performance, increased absenteeism, attrition, sleep difficulties, interpersonal issues, burnout, substance use, and suicidal thoughts and behaviour (2)(3)(4)6,7). Among EMTs/paramedics in particular, mental health issues and suicidal thoughts and behaviour are prominent (2,(5)(6)(7)(8) which may be -at least partially -attributable to the fact that they are often exposed to varying forms of human suffering to which they feel some level of responsibility (2,6,9). Carleton and colleagues (2) reported that 49.1% of paramedics screened positive for symptoms consistent with one or more mental disorders, with symptoms of depression and PTSI being the most commonly occurring mental health problems among this group of PSP. Carleton and colleagues (7) have also shown that paramedics reported statistically significantly higher prevalence of past-year and lifetime suicidality compared to other PSP. Given the nature of their work, and the prevalence of mental health and suicide-related issues, EMTs/ paramedics would benefit from formal training that aims to build resilience and provide the necessary tools to maintain healthy psychological functioning. Research has shown that resilience -an individual's capacity to effectively manage stressful/adverse life events (10-15) -is associated with wellbeing and that this association may be enhanced through appropriate interventions (16). However, the effectiveness of preventive/proactive programs that aim to educate EMTs/paramedics such as those that focus on building resilience to mitigate the effects of working in potentially traumatic and stressful environments are limited. Accordingly, researchers at the Justice Institute of British Columbia (JIBC) designed one of the first self-paced online resilience-building courses in Canada (hereafter referred to as the online resilience resource; ORR) to help PSP build resilience to increase mental wellness and psychological readiness for working in trauma-exposed fields (17). The aim of this study is to investigate the initial impact of the ORR on resilience and to explore the potential skill decay following this self-guided online resource among preemployment paramedic trainees. Training and educational requirements are complex policies in PSP contexts and are often fixed to resource and staffing levels. Identifying the point at which certain skills, such as those adopted through resilience training, may begin to decay, provides stakeholders with knowledge of when it may be appropriate to refresh the skills of PSP. Study design Through a repeated-measures research design, this study investigated the effectiveness of the ORR (17) as an educational tool for building personal resilience among a sample of paramedic students enrolled in a primary care paramedic (PCP) program. Setting PCP students were recruited from JIBC, a post-secondary institution that provides training to professionals in the justice, public safety and social service fields. The study was based entirely online. Specifically, participants were contacted via email to ask for participation, and were prompted to complete a baseline assessment survey if they provided consent to participate. Once baseline data were established, participants were directed to complete the self-paced online resilience course. Following completion of these components, participants were later contacted via email to complete a follow-up assessment survey at one of the randomly assigned timepoints (3, 6 or 9 months following the ORR). Participants A total of 227 students enrolled in the Province of British Columbia's PCP certificate program delivered by the JIBC participated in this study. Of these 227 respondents, 34 paramedic students completed all three phases of the study: baseline assessment; self-paced online resilience course; and follow-up assessment. Accordingly, 193 respondents completed only the baseline assessment and did not respond to the follow-up assessment. All participation in this study was voluntary and thus had no impact on participants' performance in the PCP program. Instrumentation The baseline assessment incorporated measures on age, gender and education. The Resilience Scale for Adults (RSA) (18) is a 33-item self-report scale that measures inter-and intra-personal protective factors believed to play an important role in one's adaptation to adversity. This instrument consists of 17 positively worded items and 16 negatively worded items, measuring five dimensions of resilience: personal strength (10 items) -perception of oneself and one's future; structured style (four items) -perception of one's level of structure and organisation in life; social competence (six items) -perception of one's sociability; family cohesion (six items) -perception of one's level of connection to their family; social resources (seven items) -perception of one's access to support from friends/family members. All items on the RSA are scored according to a 5-point Likert scale (ranging from 1 to 5). The full RSA scale, along with its subscales, are created by reversecoding negatively worded items and summing across pertinent items. Possible range for the full RSA scale is from 33 to 165, with higher scores indicative of greater perceived resilience. With respect to psychometric properties, Windle et al (19) ranked the 33-item RSA (18) highly on reliability and validity. Procedures The researcher team randomly assigned classes to three separate study cohorts based on when they would receive the follow-up assessment: 3 months after completing the ORR, 6 months after completing the ORR, or 9 months after completing the ORR. All communication for baseline and follow-up survey data collection was conducted through Qualtrics (a secure platform for distributing online surveys and storing data). The baseline assessment survey included questions on sociodemographic characteristics (eg. age, gender, education), as well as the instrument used to measure resilience (the RSA). At the end of the baseline assessment survey, participants were provided a link to the online resilience course and were directed to complete the course at their own pace (the course is estimated to take approximately 6-8 hours to complete). A certificate of completion was awarded to participants who completed all learning modules in the ORR. Depending on their randomly assigned cohort, participants who completed the baseline assessment survey and the ORR were contacted via Qualtrics at either 3, 6 or 9-month intervals to participate in the follow-up assessment survey. The follow-up survey included only the measure of resilience (the RSA). Outcome This study investigated whether, after having completed a self-paced online resilience course (17), paramedic students' scores on a measure of personal resilience would increase or decrease between the baseline and the 3, 6 or 9-month followup assessments. Data analysis The Statistical Package for the Social Sciences (Version 25) was used to clean the data and conduct all statistical analyses. A missing values analysis was conducted on the RSA, showing that data were missing completely at random at the baseline (Little's MCAR, X 2 (2680) = 2644.27, p=0.685) and follow up (Little's MCAR, X 2 (277) = 257.31, p=0.796) assessments. Given our small analytic sample (n=34) and Little's MCAR test was not statistically significant, we opted to use data imputation methods to handle missing data on RSA items as opposed to deleting cases with missing values. Specifically, expectationmaximisation data imputation was used to handle missing cases. The RSA scale and subscales were created after data imputation. With respect to descriptive statistics, analyses were conducted on sociodemographic characteristics, stratified according to whether or not respondents completed all three phases of the study (responders; n=34) and those who completed only the first phase of the study (non-responders; n=193); chi-square test for independence and independent samples t-test were used to compare these two groups on sociodemographic characteristics (Table 1). Descriptive and reliability statistics are also provided for the RSA (Table 2). Paired samples t-tests were conducted to examine change, if any, in RSA score from baseline to follow-up assessment for the total subset of respondents who completed all three phases of the study (n=34), as well as according to individual study cohort (Table 3). Results Sociodemographic characteristics were stratified according to the subset of respondents who completed all three phases of the study. As shown in Table 1, responders were, on average, older than non-responders; nearly two-thirds of responders were female, whereas a greater proportion of non-responders were male; and responders had a higher level of education than non-responders. Descriptive and reliability statistics for the RSA at baseline and follow-up assessments are reported in Table 2. Among those who completed the baseline assessment (n=227), baseline scores on the RSA ranged from 79 to 162 with a mean of 127.54 (SD=15.95; 95% CI: 125-130) and internal consistency reliability was excellent (Cronbach's α=0.900). Among the subset who completed all three phases of the study (n=34), baseline scores on the RSA ranged from 94 to 156 with a mean of 126.38 (SD=14.40; 95% CI: 122-131) and internal consistency reliability was good (Cronbach's α=0.878). An independent samples t-test revealed that responders (n=34; M=126.38, SD=14.40; 95% CI: 122-131) and non-responders (n=193; M=127.75, SD=16.24; 95% CI: 125-130) did not significantly differ on baseline RSA score, t (225) = 0.460, p=0.646. Furthermore, among the subset who completed all three phases of the study (n=34), follow-up scores on the RSA ranged from 89 to 156 with a mean of 121.73 (SD=16.36; 95% CI: 116-127) and internal consistency reliability was excellent (Cronbach's α=0.912). Table 3 reports findings for paired samples t-tests, comparing RSA scores at baseline to follow-up assessment for the total subset of respondents who completed all three phases of the study (n=34), along with the individual 3-month (n=8), 6-month (n=20), and 9-month (n=6) study cohorts. For the total subset of paramedic students who completed all three phases of the 4.1 a Participants who provided data both before and after using the ORR; b participants who provided data only before using the ORR; c chi-square test for independence and independent samples t-test, comparing responders to non-responders on sociodemographic characteristics Figure 1a and 1b further illustrate the patterns reported in Table 3. Figure 1a shows that the 3-month and 6-month study cohort reported a similar mean RSA score at baseline, whereas the 9-month study cohort reported a slightly lower mean RSA score at baseline compared to these two cohorts. When comparing Figure 1a and 1b, the mean RSA score increases between baseline to follow-up assessment for the 3-month study cohort, whereas mean RSA score decreases between baseline and follow-up assessment for the 6-month and 9-month cohorts. Figure 1b further shows that, in general, the 3-month study cohort reported the highest mean RSA score at follow-up, followed by the 6-month cohort and 9-month cohort. Discussion The results from this study suggest that the ORR may be an effective strategy for building personal resilience among paramedic students in the short-term. However, much like previous research that highlights a significant skill decay after 90 days for training in healthcare contexts (20,21), the diminished impact of the ORR over time was also illustrated in this study. . In addition, it is noteworthy to highlight that the decrease in mean scores on the RSA was larger for the 9-month compared to the 6-month study cohort. Although these findings are only preliminary, stakeholders should acknowledge that it may be important to establish guidelines for refreshing resilience-based skills among PSP. Perhaps it would be most beneficial to ensure PSP are consistently exposed to resilience- 109.50 (13.99) RSA = Resilience Scale for Adults; Time 1 is RSA score before using the ORR; Time 2 is RSA score after using the ORR; a N=34; b n=8; c n=20; d n=6 building strategies (eg. over 3 or 6-month intervals) following initial involvement in a resilience training program. Such practices would allow for these professionals to draw on resilience-building techniques if they desire. In addition, more general recommendations for research and current practice may be extracted from the current study. First, EMT/paramedic programs should consider implementing mandatory resilience training services/programs to better help professionals manage exposure to potentially traumatic and stressful work-related events. In this case, a 6 to 8-hour selfpaced online resilience course may be enough for improving personal resilience in the short-term. Further research will need to identify an appropriate time interval for refreshing resiliencebased skills following initial involvement in a resilience training service/program. Researchers should also continue to investigate the effectiveness of online educational tools, such as the ORR (17), among varying

PSP working in the field as well as students in educational/training programs for varying types of public health and safety professions. Continued focus on online educational tools designed for building personal resilience, enhancing mental wellness and psychological readiness among those working in trauma-exposed fields is imperative, especially given that this method of delivery increases accessibility to PSP who may experience challenges in accessing such resources (eg. those located in remote or rural communities and/or completing their education online). Finally, given that several studies have reported that a considerable proportion of EMTs/ paramedics experience mental health issues and struggle with suicidal thoughts and behaviour (2,(5)(6)(7), it is recommended that educational tools for building personal resilience among this group of PSP provide psychoeducation, especially with respect to trauma or stressor-related disorders (eg. PTSI), mood disorders (eg. anxiety and depression) and suicidal thoughts and behaviour. Limitations There are several limitations that must be considered when interpreting the findings from this study. First, attrition is a common issue with longitudinal study designs. Relatedly, it is plausible that findings are somewhat biased as participants who responded to the follow-up assessments may differ in some respects from those who did not. In this case, it is clear that responders and non-responders significantly differed on some sociodemographic characteristics; however, analyses also showed that responders and non-responders did not significantly differ on baseline assessment of resilience. Second, the analytic sample size (n=34) was quite small, especially when further disaggregated by individual study cohort; for instance, the 3-month and 9-month study cohorts comprised of eight and six participants, respectively. Such a small analytic sample size limits the power of inferential tests, as well as limits generalisability and validity of the findings. Collecting data on PCP participants at 3, 6 or 9 months after ORR completion was particularly challenging as the participants had graduated from the program at that point in time and were no longer students. There are a range of methods to improve retention for longitudinal studies, particularly for webdelivered training for student paramedics who are likely to be bombarded with new information in their studies. Financial compensation may have been effective at increasing response rates as perhaps the addition of a paper-based version of the questionnaire. Third, there are some limitations related to the instrument used to capture participants' level of personal resilience (ie. the RSA). Although it is understood that a higher score on the RSA is indicative of a greater level of resilience, the scale does not have 'cut-off' points which could aid in interpretation of the score (eg. 'cut-off' points for low, medium or high resilience). Fourth, the current study did not control for potential confounding effects that may also explain why scores changed between baseline and follow-up assessment. Specifically, it is unclear whether improvement or deterioration in personal resilience is due to a true change in skill over time or if this change is related to exposure to other factors between baseline and follow-up assessment. Conclusion This study presents evidence to suggest that an educational tool, such as a self-paced ORR, may be an effective strategy for improving short-term personal resilience among PCP students. This study also presents evidence to suggest that any improvement in personal resilience following initial involvement in a resilience training service/program may decay over time. Although the present findings support the use of proactive training programs to build resilience among paramedic students, as well as the need to consider refresher practices to prevent skill decay, further research is required to identify whether the skills acquired through resilience training are effective when put into practice (eg. identifying whether PSP with resilience training are less likely to develop PTSI related to exposure to potentially traumatic and stressful work-related events). Systematic Review and Meta-analysis on COVID-19 Vaccine Hesitancy Background: The presented meta-analysis was developed in response to the publication of several studies addressing COVID-19 vaccines hesitancy. We aimed to identify the proportion of vaccine acceptance and rejection, and factors affecting vaccine hesitancy worldwide especially with the fast emergency approval of vaccines. Methods: Online database search was performed, and relevant studies were included with no language restriction. A meta-analysis was conducted using R software to obtain the random effect model of the pooled prevalence of vaccine acceptance and rejection. Egger regression test was performed to assess publication bias. Quality assessment was assessed using Newcastle-Ottawa Scale quality assessment tool. Results: Thirty-nine out of 12246 articles met the predefined inclusion criteria. All studies were cross-sectional designs. The pooled proportion of COVID-19 vaccine hesitancy was 17% (95% CI: 14-20) while the pooled proportion of COVID-19 vaccine acceptance was 75% (95% CI: 71-79). The vaccine hesitancy and the vaccine acceptance showed high heterogeneity (I 2 =100%). Case fatality ratio and the number of reported cases had significant effect on the vaccine acceptance as the pooled proportion of vaccine acceptance increased by 39.95% (95% CI: 20.1-59.8) for each 1% increase in case fatality (P<0.0001) and decreased by 0.1% (95% CI: -0.2-0.01) for each 1000 reported case of COVID-19, P= 0.0183). Conclusion: Transparency in reporting the number of newly diagnosed COVID-19 cases and deaths is mandatory as these factors are the main determinants of COVID-19 vaccine acceptance. Introduction: The wide use of vaccines has led to decreased mortality and morbidity of different transmissible diseases, this was a crucial factor in elimination of poliomyelitis in the Americas and the worldwide eradication of smallpox (1). Vaccination programs depend on mass vaccination to be able to decrease incidence and prevalence of Vaccine Preventable Diseases (VPD). In addition to the proposed direct protection for vaccinated candidates, wide vaccination scope results in indirect shielding for the overall community by declined conveyance of VPD, thereby dampening the risk of infection for vulnerable individuals in the community (2). One of the main limiting factors for wide-spread of vaccination programs (especially for newly emerging vaccines) is vaccine hesitancy. The World Health Organization (WHO) named vaccine hesitancy as one of the top ten threats to global health in 2019, calling for research to identify the factors associated with this phenomenon (3). Vaccine hesitancy is defined as a behavior of a delayed vaccine approval or even declined vaccination despite accessible vaccination services (4,5). The pandemic COVID-19 caused by the recently discovered coronavirus-2019 (SARS-CoV-2) is strongly influencing the worldwide public health, culture, economy, and human social behavior. Despite all efforts since the beginning of the pandemic there is no approved medicine or treatment to cure COVID-19 till now, whereas vaccine development efforts are taking the highest priority as it can potentially save humanity by inducing immunity against COVID-19 (6). According to WHO, herd immunity against COVID-19, which is known as population immunity, can be achieved naturally by the exposed people who recovered from the virus by their own protective antibodies or by providing COVID-19 vaccination (7,8). Herd immunity for COVID-19 can be achieved on 70% of the single vaccinated dose individuals and 90% of the two vaccinated dose individuals (9). Vaccines typically require years of research and testing before reaching the clinic, but in 2020, scientists were racing against time to produce safe and effective coronavirus vaccines. Currently we have 14 approved vaccines for full use, 6 authorized in early or limited use, 27 vaccines in phase 3 trials, 36 vaccines in phase 2, 48 vaccines in phase 1 and 4 abandoned vaccines after trials. In addition, at least 77 preclinical vaccines are under active investigation in animals (10). Unfortunately, the newly emerging vaccines for COVID-19 are faced nowadays with hesitancy to use in different countries. People showed concerns about both efficacy and possible side effects of these recently approved vaccines. Such hesitancy can have a heavy influence on vaccine delivery and the aimed wide uptake to control the pandemic (11). After the announcement of several pharmaceutical manufactures the production of COVID-19 vaccines, social media started to discuss vaccine content widely across different platforms. The propagated information provides mostly non-factual data and from non-medical individuals (12). The presented systematic review & meta-analysis was developed in response to the publication of several studies addressing COVID-19 vaccines hesitancy. Identification of independent factors affecting vaccine hesitancy worldwide especially with the fast emergency approval of these vaccines. Data sources This meta-analysis was guided by the 2020 Cochrane Handbook of Systematic Review and Meta-Analysis (13), with respect to the preferred reporting items of the systematic review and meta-analysis (PRISMA) checklist (14). Search was conducted for the hesitancy or refusal of COVID-19 vaccination through the published and grey literature using multiple databases; PsycINFO, ScienceDirect, Embase, Scopus, EBSCO, MEDLINE central/PubMed, ProQuest, SciELO, SAGE, Web of science, and Google scholar. Search terms were determined and approved after the consultation of PubMed help desk. The used keywords were added to Annex 1. Study selection All studies reporting COVID-19 vaccine hesitancy, were included with no language restriction. Abstract-only papers as proposals, conference, editorials, author . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) Investigations of heterogeneity: Cochrane Q test and (I²) test was used to assess and measure heterogeneity between studies, considering I 2 ≥ 75% represents substantial heterogeneity and strength of evidence for heterogeneity is the P-value ≤ 0.05 from the Q test; according to Cochrane Handbook for Systematic Reviews of Interventions (13). Due to substantial heterogeneity, DerSimonian and Laird random-effects models were applied to pool the outcomes. Publication bias: . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 18, 2021. ; https://doi.org/10.1101/2021.05.15.21257261 doi: medRxiv preprint Publication biases were assessed by visual inspection of the funnel plot and statistically by Begg's modified funnel plot and Egger's regression test (13). Quality assessment Quality assessment (QA) was assessed using Newcastle-Ottawa Scale quality assessment tool customized for cross-sectional studies (15). The assessment was performed by two independent reviewers (DMH, EE) and further checked by two additional reviewers (SO EI-ganainy, AA). Statistical analysis and data synthesis: R software was used to perform the meta-analysis and to pool the effect size (proportion); fixed or random effect model were used according to the studies' consistency. Meta-regression analysis was performed to examine the impact of confounders on the effect of vaccine hesitancy such as age, sex, and country. Results were presented in the Forest plots to visualize the degree of variation between studies. Leave one-out sensitivity analysis was conducted to test the effect of each study on the pooled effect to determine the robustness of the obtained outcomes. Sub-group analysis was performed to categorize the vaccine hesitancy according to sample size studies. To investigate the sources of high heterogeneity in the pooled prevalence of vaccine acceptance and hesitancy, meta-regression analysis was performed with different models including the main predictors of vaccine acceptance and hesitancy reported in included studies such as age, sex, educational level and setting. Additionally, number of reported cases, number of reported deaths, case fatality ratio and number of vaccinated people within each country until the end of January 2021 (16, 17), were examined as potential modifiers of vaccine acceptance and hesitancy and included in the meta-regression model. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) Search results: The flow diagram of the selection process is shown in figure 1. From a total of 12246 potentially relevant articles, 1621 duplicate articles and 2944 citations published before 2019 were excluded. A total of 7681 citations were eligible for title screening. Only 51 articles were eligible for full-text screening after removing irrelevant (7627) and duplicate articles (3). In total 34 articles were excluded after full text screening (2 duplicates and 29 irrelevant), 3 were retracted. Another 22 articles were added manually For quantitative assessment, there were 39 eligible articles. The inter-rater agreement for inclusion was κ=0.87 and for the quality assessment was κ=0.91. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. . CC-BY 4.0 International license It is made available under a is

the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 18, 2021. ; Table 1 shows the main findings of included studies; all the studies were cross-sectional surveys. The total sample size was 173213 ranging from 351 in the study of Sharun et al, 2020 (18) reaching 32361 in the study of Paul et al, 2021 (19). The highest presentation of female sex was in the study of Kowk, 2021 (20) followed by Wang 2020 (21) while the lowest proportion of females was in the study of Malik et al 2020 (22). Age range was 15->85 in the study of Taylor 2020, the mean age of the study participants was the highest in the study of Taylor 2020 (23) Predictors of COVID-19 vaccine acceptance and hesitancy Multiple factors were associated with vaccine hesitancy Table(1). Previously receiving influenza vaccine is the main factor that determines the acceptance of COVID-19 vaccine. Individuals reporting intake of influenza vaccine were more likely to accept COVID-19 vaccine than those who did not receive it previously (21,28,33).Some socio-demographic characteristics were considered to influence the acceptance of the vaccine. Being young was associated with no or not sure response towards the intake of COVID-19 vaccine (28,38,41), while older individuals were more likely to accept the vaccine intake (24, 33). Regarding the gender, males were more likely to accept the vaccine rather than females (21,33,38,45).. Low education levels and income, being not employed in a full time job or retired were associated with refusal of the vaccine (19,28,38,41), while those with professional private work were more likely to accept the vaccine (21). The marital status also affects the response to vaccine acceptance, being single or widowed were associated with hesitancy (38), while married individuals were more likely to accept the vaccine (24). Racial and ethnic groups were noticed to affect the acceptance of vaccine. Black race and mixed ethnicity were associated with hesitancy towards the vaccine (28,38).Other factors that increase the acceptance towards the vaccine is the presence of trusted health systems (24), the fear from getting infected with the virus (33) and having chronic diseases (21). While factors that increase the refusal of the vaccine involve the suspicion from its efficacy and effectiveness (21), individuals may think . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 18, 2021. ; about the side effects and do not believe that the vaccine will work, or they trust their immune system and are not afraid of getting sick (45). On the other hand, the pooled proportion of COVID-19 vaccine acceptance ( Figure (4) shows the results of the meta-regression models between the case fatality (%) and the proportion of vaccine hesitancy and vaccine acceptance, respectively by type of setting and study sample size. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Discussion The vaccine for COVID-19 availability is a critical step to face the COVID-19 pandemic. But vaccine hesitancy represents a great threat to global health during this pandemic and limits the power of health systems to control the COVID-19 pandemic. Hence, estimating the COVID-19 vaccine hesitancy represents a tool to design an action plan to improve the vaccine acceptance. In this meta-analysis, there was large variability between the studies discussing COVID-19 hesitancy in terms of vaccine acceptance. We aimed to determine the reported that about 20% of the participants refused COVID-19 vaccine. They observed that differences across countries were very substantial and resulted in a heterogeneity above 90%. Furthermore, they declared that the trend of rejection increased with time. The main determinants of COVID19 vaccine rejection were being female, of low educational level, or belonging to minor ethnicity. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 18, 2021. ; https://doi.org/10.1101/2021.05.15.21257261 doi: medRxiv preprint were common factors. Impacts of regional infection rates, gender, and personal COVID-19 experience were inconclusive. Unique COVID-19 factors included political party orientation, doubts toward expedited development/approval process, and perceived political interference. Many receptive participants preferred to wait until others have taken the vaccine; mandates could increase resistance. We speculate that the difference in vaccine acceptance may be affected by vaccine efficacy and side effects. Vaccines' side effects range between local to systemic, and short to long term events. The reported common side effects are generally mild to moderate and last for a few days. These include injection site pain, fatigue, rigors, fever, muscle and joints pains. Less commonly, a vaccine recipient may develop allergic reaction or anaphylaxis, and neurological side effects; however they are rarely reported (63). There is a rising concern particularly related to reported thrombo-embolic events, particularly after administration of AstraZeneca vaccine in Europe, but the European Medicines Agency concluded that the benefits of the vaccine overweighs the potential risk of this rare side effect (64). In this context, Kaplan et al, (65) underlined that vaccine acceptance improved when vaccine efficacy exceeds 70%. Moreover, they addressed that minor side effects, such as a sore arm or fever lasting for a day did not affect vaccine acceptance, while major side effects in 1/100000 greatly affected vaccine acceptance. These side effects may vary according to the type of vaccine used in each country. Emerging evidence suggests that both exposure to misinformation about COVID-19 and public concerns over the safety of vaccines may be contributing to the observed decline in intentions to be vaccinated, and this highlights the need for measures to address public acceptability, trust and concern over the safety and benefit of approved vaccines (66,67). This finding highlights the power of social media. Some studies emerged in the last months discussing the vaccine confidence in several populations, especially in countries with high burden of diseases like Pakistan (68) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 18, 2021. ; https://doi.org/10.1101/2021.05.15.21257261 doi: medRxiv preprint education, whether medical or nonmedical students, on their decision (35).With the development of multiple effective vaccines, Immunization programs are only successful when there are high rates of acceptance and coverage (69). To accomplish this, it is critical to understand vaccine-acceptance messaging to effectively control the pandemic and prevent thousands of additional deaths (70).Individuals commonly considered COVID-19 to be a very severe disease, although they expected to experience less severe symptoms themselves. Individuals also worried more about transmitting the disease to others than about falling ill personally (71). The strongest predictor of intentions to accept a COVID-19 vaccine recommended by authorities was the degree to which respondents trusted the vaccine to be safe. Perceived vaccine safety explained 52% of the variance in intentions to vaccinate (72).The study of Malik et al. shows that COVID-19 vaccine acceptance can be predicted with relatively high accuracy by readily available demographic characteristics. Since the beginning of the COVID-19 pandemic in the United States, it has been clear that low-income and communities of color are at higher risk for infection and death from COVID-19 (22). Strengths and limitations One of the main strength points in this study is the search strategy, we searched 12 different databases. Each citation was screened by two reviewers and disagreement was solved by a senior author. The same was done for quality assessment to ensure robust evidence. A large proportion of the included studies used quota (as opposed to probability-based sampling) and were pre-prints yet to be peer reviewed (as opposed to published journal articles). However, the type of sampling method used (quota vs. probability) had minimal impact on intentions estimates and that studies reported in pre-prints produced similar effect estimates as peer-reviewed journals. One of the main limitations was different tools used to assess vaccine acceptance in addition, the data collected either through face-to-face interview or through online data collection tools. We think that this may affect the internal validity of the study. However, we segregated analysis based on the method of data collection and the difference was not significant. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 18, 2021. ; https://doi.org/10.1101/2021.05.15.21257261 doi: medRxiv preprint Conclusions COVID-19 vaccine rejection is low; however, continuous health education and social support is necessary to maintain the high acceptance rates. Time and residency have no significant effect on vaccine acceptance. However, the country-level case fatality and the officially reported number of cases were significant predictors of COVID-19 vaccine acceptance. That's why encouraging the health authorities to accurately follow & announce case fatalities could be a major contributing factor to increasing vaccine acceptance. We believe that this study will demonstrate public hesitancy and help further . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 18, 2021. ; Early palliative care: taking ownership and creating the conditions. The evidence for early integration of palliative care into standard oncology care is growing. All things considered, the benefits to patients, families, and the health care system of initiating palliative care earlier in the illness trajectory and not relegating it to the terminal phase in the last days or weeks of life far outweigh any potential burdens. Kain and Eisenhauer, in this edition of Current Oncology, add a noteworthy voice to the call for action. Notably, that call is neither new nor confined to cancer care; the need to provide palliative care and to initiate it earlier in the illness trajectory is echoing ever louder, to include many non-cancer conditions ranging from advanced heart, lung, liver, renal, and neurologic diseases to conditions such as dementia and frailty1–4. In 2002, the World Health Organization modified its definition of “palliative care” by replacing the term “terminal illnesses” with “life-threatening illnesses.” This intentional alteration emphasizes the need to implement palliative care earlier in the illness trajectory. That same year, the Canadian Hospice Palliative Care Association embraced the approach, embedding it in its model in which palliative care starts earlier, alongside treatments to control the disease and manage its complications. When the disease progresses and disease-modifying treatments become less effective, the palliative care approach starts to take more prominence. The two approaches are therefore complementary in a dynamic process rather than mutually exclusive, as in the “old” model in which palliative care was considered only in the terminal, end-of-life phase. Identifying and addressing physical, psychological, social, and spiritual needs should not be confined to the terminal phase of disease and neither should advance care planning. Patients whose diseases are potentially controllable or even curable also benefit from a supportive and palliative care approach that aims to assist quality of life. Likewise, recognizing disease progression and engaging in honest goals-of-care discussions accompanied by wise treatment choices in a way that promotes realistic hope and prepares for future possibilities should not be relegated only to end of life. Hawley 5 has adapted the Canadian Hospice Palliative Care Association model in what is called the “bow-tie” model (Figure 1). It, too, emphasizes the complementary nature of disease-modifying treatments and palliative care. Unfortunately, however, the model relegates palliative care units to end-of-life care. In many parts of Canada and internationally, admissions to palliative care units are not limited to end-of-life patients, but are open to patients with complex symptomatic and psychosocial needs across the illness trajectory; on the other hand, residential hospices generally provide end-of-life care. Of particular note in the “bow-tie” model is that it accommodates rehabilitation and survivorship. Palliative care

rehabilitation programs for patients with palliative care needs who are not at end of life have highlighted the “early” palliative care model and its accompanying benefits, particularly through maintenance of functional status, mobility, and quality of life 6. Whose responsibility is it, then, to provide palliative care? Kain and Eisenhauer appropriately highlight the need, across Canada, for more specialist palliative care resources to address current gaps and future needs. Ontario, for example, has 218 medical oncologists and 225 radiation oncologists 7. It has only 276 “palliative care” physicians—too few to provide all the palliative care The evidence for early integration of palliative care into standard oncology care is growing. All things considered, the benefits to patients, families, and the health care system of initiating palliative care earlier in the illness trajectory and not relegating it to the terminal phase in the last days or weeks of life far outweigh any potential burdens. Kain and Eisenhauer, in this edition of Current Oncology, add a noteworthy voice to the call for action. Notably, that call is neither new nor confined to cancer care; the need to provide palliative care and to initiate it earlier in the illness trajectory is echoing ever louder, to include many non-cancer conditions ranging from advanced heart, lung, liver, renal, and neurologic diseases to conditions such as dementia and frailty [1][2][3][4] . In 2002, the World Health Organization modified its definition of "palliative care" by replacing the term "terminal illnesses" with "life-threatening illnesses." This intentional alteration emphasizes the need to implement palliative care earlier in the illness trajectory. That same year, the Canadian Hospice Palliative Care Association embraced the approach, embedding it in its model in which palliative care starts earlier, alongside treatments to control the disease and manage its complications. When the disease progresses and disease-modifying treatments become less effective, the palliative care approach starts to take more prominence. The two approaches are therefore complementary in a dynamic process rather than mutually exclusive, as in the "old" model in which palliative care was considered only in the terminal, end-of-life phase. Identifying and addressing physical, psychological, social, and spiritual needs should not be confined to the terminal phase of disease and neither should advance care planning. Patients whose diseases are potentially controllable or even curable also benefit from a supportive and palliative care approach that aims to assist quality of life. Likewise, recognizing disease progression and engaging in honest goals-of-care discussions accompanied by wise treatment choices in a way that promotes realistic hope and prepares for future possibilities should not be relegated only to end of life. Hawley 5 has adapted the Canadian Hospice Palliative Care Association model in what is called the "bow-tie" model ( Figure 1). It, too, emphasizes the complementary nature of disease-modifying treatments and palliative care. Unfortunately, however, the model relegates palliative care units to end-of-life care. In many parts of Canada and internationally, admissions to palliative care units are not limited to end-of-life patients, but are open to patients with complex symptomatic and psychosocial needs across the illness trajectory; on the other hand, residential hospices generally provide end-of-life care. Of particular note in the "bow-tie" model is that it accommodates rehabilitation and survivorship. Palliative care rehabilitation programs for patients with palliative care needs who are not at end of life have highlighted the "early" palliative care model and its accompanying benefits, particularly through maintenance of functional status, mobility, and quality of life 6 . Whose responsibility is it, then, to provide palliative care? Kain and Eisenhauer appropriately highlight the need, across Canada, for more specialist palliative care resources to address current gaps and future needs. Ontario, for example, has 218 medical oncologists and 225 radiation oncologists 7 . It has only 276 "palliative care" physicians-too few to provide all the palliative care required for cancer and non-cancer patients 8 . Some hospitals in Ontario have inadequate specialist palliative care physician or nurse coverage (and in some cases none at all). Shortages of palliative care physicians or inadequate funding account for that lack. The Champlain region (1.3 million population and 30,000 km 2 ) in southeastern Ontario, which includes Ottawa, has a 1.0 full-time equivalent palliative care physician positions in its community palliative care consultation team. In contrast, the Edmonton region in Alberta, similar in population size and area, has a much more appropriate complement of 5 full-time physician equivalents and 5 full-time nurse equivalent positions. Other Ontario community teams have no physicians, because those teams are funded as nurse-only models by the Ontario Ministry of Health and Long-Term Care (mohltc). Advocacy efforts to the mohltc, Cancer Care Ontario, and the Ontario Medical Association to address the gaps have, with some exceptions, generated few concrete results in recent years. In 2013 for example, Cancer Care Ontario opted to exclude palliative care physicians from the provincial alternative funding plan for cancer care specialists negotiated as part of its Models of Care initiative. The mohltc provides hospital on-call funding for some hospital-based palliative care teams, but not for others. The praiseworthy decision by the College of Family Physicians of Canada and the Royal College of Physicians and Surgeons of Canada to recognize palliative medicine as a subspecialty will likely increase interest in the field, but the insufficient number of palliative care residency and physician positions in Ontario could create bottlenecks. Bold action by policymakers, health services administrators, and educators is needed to overcome those funding and resource barriers. However, even with much-needed increases in the palliative care specialist workforce (and the administrative support that is often lacking in many centres), this relatively small workforce will not be able to provide for all the palliative care needs (including early integration) for all patients with life-limiting cancer and non-cancer diagnoses. The responsibility for providing palliative care is not solely that of palliative care specialists or specialist teams, it is everyone's business. All health care professionals who provide any form of care to patients with life-threatening cancer and non-cancer illnesses-including oncologists, family physicians, internists, and cardiologists, to name only a few-should take ownership of the challenge by providing a palliative care approach to their patients. "Palliative care approach" refers to primary-level (that is, generalist-level), high-quality palliative care that could be provided by health care professionals who are not palliative care specialists. For several years now, the American Society of Clinical Oncology has endorsed that approach, and more recently, it further elucidated the competencies that the generalist-level palliative care for oncologists and cancer care providers should include 9 . A case has also been made for a palliative care approach to be provided by family physicians 10 . Most oncologists and oncology nurses acknowledge a role in providing the palliative care approach to their patients 11 . Cancer Care Ontario has developed a care pathway that integrates the palliative care approach and psychosocial care in everyday cancer care 12 . Gardiner et al. 13 outlined the important role for generalist providers (including oncologists and clinicians in many different specialty areas and in family medicine) in the earlier incorporation of palliative care and transitions to palliative care. Apart from increasing access to palliative care, other benefits accrue when attending non-palliative-care clinicians provide a palliative care approach to their own patients. They often have close rapport with their patients and are trusted by the patients and families. Continuity of care is valued by palliative care patients 14 . Leaving it up to a third party, particularly the palliative care team, to disclose bad news and to discuss goals of care in the presence of disease progression or to initiate advance care planning might leave the patient feeling abandoned and the palliative care clinician feeling vilified as the harbinger of doom. Many palliative care clinicians can share experiences of being fired by patients and families whose diseases have progressed, but who were not fully informed of the situation by their attending clinicians. Several conditions have to be met for non-palliativecare providers such as oncologists, oncology nurses, and family physicians to effectively provide a palliative care approach to their patients. Those conditions include education, support, tools, and incentives and supporting policies. Many health professionals have not received adequate basic palliative care training and feel uncomfortable and ill equipped to provide such care 15 . A survey of hospital-based staff reported that 19.3% of staff time was spent caring for end-of-life patients, but that only 19% of respondents had received formal palliative care training 16 . Almost three quarters wanted formal training in the area and perceived that confidence in palliative care delivery was significantly greater for the staff with formal palliative care training. The palliative care approach should be an integral component of undergraduate, postgraduate, and continuing professional development curricula. Some standardized programs are available in Canada for these purposes, including Pallium Canada's leap (Learning Essential Approaches to Palliative Care) courses (interprofessional), the Educating Physicians on End-of-Life Care course and other intensive interprofessional courses offered by McMaster University (Victoria Hospice), as well as Life and Death Matters (for support workers). Hospital and community settings should have adequately staffed specialist palliative care teams to provide support with complex patients, to verify care plans when needed, and to take over the care of the small proportion of cases that are extremely challenging. A recent systematic review of the impact of hospital-based palliative care consultation teams in which the partnership model was typically used showed improvements in patient care and cost savings 17 . Several community benefits have been noted as well. For example, Kain and Eisenhauer quote a study by Seow et al. 18 of the impact of community-based specialist palliative care teams in Ontario. In that work, 2 of the 11 teams studied used a consultation or shared-care model in which family physicians and community nurses provided the palliative care with the support of specialist palliative care teams. Those 2 teams achieved benefits similar to those achieved by the other teams, whose members largely used a specialist takeover model; significant reductions in emergency department visits, hospitalizations, and hospital deaths were observed. Other studies have reported similar results using the "partnership" model between non-palliative-care clinicians and palliative care specialists 19,20 . Funding models that undermine this consultation support role should be replaced with models and funding incentives that support it. In Ontario, for example, the absence of an adequate alternative salary-type funding plan for palliative care physicians by the mohltc is pushing more palliative care physicians to a fee-for-service model, which could ultimately undermine capacity-building for a primary-level palliative care approach as palliative care physicians intentionally or unintentionally take over all the palliative care, including primary-level palliative care. A large dose of self-awareness and acknowledgment of competency limitations is also required. Oncologists and other non-palliative-care specialist providers should acknowledge their skill gaps in this area and seek programs to address those gaps. The organizations they work for should provide opportunities to acquire the necessary skills, and other incentives should be put in place. Timely referrals to specialist palliative care teams should be provided when patients require specialist-level palliative care interventions. Guidelines to assist in identifying such situations have been published 21 . It requires commitment to participate in education programs to update one's palliative care approach skills. By the same token, palliative care physicians require a substantial dose of humility and respect for the work done by non-palliative-care specialists and should see referrals as opportunities for support and collaboration. We are all in this together, and the article by Kain and Eisenhauer brings to the fore the need for effective collaboration and integration to improve patient outcomes. Kain and Eisenhauer identify the "surprise question" as a useful tool. Its utility as an approach for the earlier identification of patients who could benefit from a palliative care approach has been demonstrated in various settings for cancer and non-cancer diagnoses. However, it is important to note that the "surprise question" neither limits palliative care to the last year of life nor acts as a prognostication tool. It is simply a useful alert that helps in moving palliative care upstream. Other tools to identify patients with palliative care needs have been reported and tested in various settings. Weissman et al. 20 highlighted the usefulness of tools to screen patients for unmet palliative care needs, combined with education initiatives and other system-change work. They go on to posit

that hospital staff engaged in day-to-day patient care can identify and address most such needs, reserving specialty palliative care services for more complex problems 20 . That focus on palliation in day-to-day care in turn raises a question: What percentage of patients should be referred to a specialist palliative care service? Clearly, 100% would not be realistic, but too-small numbers would also raise concerns, particularly in the absence of the conditions already described 22 . In a population-based survey in Belgium 23 , palliative care services were not used for 79% of patients with organ failure, 64% with dementia, and 44% with cancer. Among the most common reasons for not referring to a palliative care service was the perception by some clinicians that they had already provided sufficient palliative care (raising the question of whether the clinicians had been prepared to provide a palliative care approach) and that insufficient time to initiate palliative care was available (indicating that palliation was being considered too late). For patients who were referred, the timing of referral varied from a median of 6 days before death (organ failure) to 16 days (cancer). Further studies are needed to determine what an acceptable range would be if all the conditions were to be in place. Kain and Eisenhauer highlight the challenges related to the fear associated with the term "palliative care." Cherny 24 summarized three different approaches to the concerns about the name. They can be summarized as "getting over it," "living with it," and "getting around it." In the "getting over it" approach, the term "palliative" continues to be used, but patients, colleagues, and the public are educated about what it actually means. In the "living with it approach," a term such as "supportive care" is used to refer to palliative care earlier in the illness trajectory and "palliative" is reserved for end of life, when the word might be less threatening for some. In the "getting around it" approach, the term is replaced with another term such as "supportive care," "pain and symptom management," or "quality of life." We are concerned about the latter term, because it might simply introduce a euphemism that allows people to avoid talking about the issues and realities at hand. In any case, the measure is a temporary one, because, within a generation or less, the new word will come to acquire the same stigma as "palliative care" bears now so long as there is no education and open dialogue about what palliative care actually is. We are of the opinion that the second approach, calling a service "supportive and palliative care," is the optimal pragmatic approach. The language used is important. For example, confusion does arise because "palliative care" refers both to an approach to care and to a specialized care team. All patients with life-threatening illness should receive palliative care as an approach to care, but, as posited earlier in this article, that approach does not always have to be provided by specialized physicians or teams. The specific words used are also indications of the culture that exists. The phrases "the patient is not palliative yet" or "the patient is now palliative" are often used in everyday care, and yet they are inappropriate in a new paradigm in which a palliative care approach has to start earlier, because they suggest that palliative care is confined to the end of life. Phrases such as "this patient would also benefit from a palliative care approach" or "this patient now needs end-of-life palliative care" would be more appropriate. To summarize, a paradigm shift is needed. The call is not new, but the emerging evidence for initiating palliative care earlier and more broadly is ever more compelling. The required culture change requires ownership by and collaboration between many stakeholders, including health care professionals who are not palliative care specialists, palliative care specialists, administrators, policymakers, funders, and educators. Kain and Eisenhauer make several important recommendations, including implementation of quality improvement initiatives across many care settings to achieve the paradigm shift. Cancer Care Ontario's integrate project is one such quality improvement initiative, as is the Speak Up campaign (http://www.advance careplanning.ca) 25 . Future studies to gather more evidence about the impact of the early palliative care model provided by non-palliative-care clinicians will be needed. Depth relationships and measures of tissue thickness in dorsal midbrain Abstract Dorsal human midbrain contains two nuclei with clear laminar organization, the superior and inferior colliculi. These nuclei extend in depth between the superficial dorsal surface of midbrain and a deep midbrain nucleus, the periaqueductal gray matter (PAG). The PAG, in turn, surrounds the cerebral aqueduct (CA). This study examined the use of two depth metrics to characterize depth and thickness relationships within dorsal midbrain using the superficial surface of midbrain and CA as references. The first utilized nearest‐neighbor Euclidean distance from one reference surface, while the second used a level‐set approach that combines signed distances from both reference surfaces. Both depth methods provided similar functional depth profiles generated by saccadic eye movements in a functional MRI task, confirming their efficacy for delineating depth for superficial functional activity. Next, the boundaries of the PAG were estimated using Euclidean distance together with elliptical fitting, indicating that the PAG can be readily characterized by a smooth surface surrounding PAG. Finally, we used the level‐set approach to measure tissue depth between the superficial surface and the PAG, thus characterizing the variable thickness of the colliculi. Overall, this study demonstrates depth‐mapping schemes for human midbrain that enables accurate segmentation of the PAG and consistent depth and thickness estimates of the superior and inferior colliculi. requiring a depth-mapping scheme for probing the anatomy and function of the human midbrain. Both SC and IC have an intricate laminar cytoarchitecture and exist in a complex 3D topology that is folded into four small hillocks on the dorsal surface of the midbrain. The inner boundaries of the colliculi abut another nucleus, the periaqueductal gray (PAG), which surrounds the ventricular cerebral aqueduct (CA). The laminar functional organization of mammalian colliculi has been clearly established based on several electrophysiology and lesion studies in animal models. Experiments in nonhuman primates and cats have shown that the layers of SC can be divided into three groups with independent functional purposes. Superficial layers have been shown to receive direct visual input and have retinotopically organized receptive fields (Cynader & Berman, 1972;Feldon & Kruger, 1970). Intermediate layers mediate oculomotor control (Robinson, 1972), while deep layers are associated with multimodal inputs, primarily multisensory, and visuomotor neurons (Meredith & Stein, 1986;Sprague & Meikle, 1965). Likewise, electrophysiological studies in IC of rats and primates have shown that the central nucleus organizes auditory inputs of varying frequency in a laminar fashion (Baumann et al., 2010;Schreiner & Langner, 1997). Invasive studies have exhibited this tonotopic organization along the dorsal-to-ventral direction, which roughly corresponds to laminar depth (Baumann et al., 2011;Cheung et al., 2012;Loftus, Malmierca, Bishop, & Oliver, 2008;Malmierca et al., 2008;Schreiner & Langner, 1997). While functional MRI (fMRI) has been used extensively in human cerebral cortex, the human brainstem has been relatively neglected despite its critical role in brain function. Functional contrast-to-noise ratio (CNR) is often low due to its deep location (Singh, Pfeuffer, Zhao, & Ress, 2017) and functional data is susceptible to partial volume effects due to the small size of important brainstem structures compared with conventional fMRI voxel sizes of 3-4 mm. However, several recent high-resolution fMRI studies using millimeter-scale voxels have demonstrated detailed analysis of the functional laminar topography of the human colliculi with reasonable CNR. Indeed, these studies have confirmed the functional properties of the three major layer groups of SC (Katyal & Ress, 2014;Linzenbold & Himmelbach, 2012;Loureiro et al., 2017;Savjani, Katyal, Halfen, Kim, & Ress, 2018;Schneider & Kastner, 2010), as well as the auditory organization of IC (De Martino et al., 2013;Moerel, De Martino, U gurbil, Yacoub, & Formisano, 2015;Ress & Chandrasekaran, 2013), encouraging interest in resolving these laminar variations of human colliculi in more detailed experiments. A recent fMRI study showed depth-dependent variations of the BOLD response with 1-mm sampling in SC (Loureiro et al., 2017). This work utilized binary mask erosion techniques to generate depth-dependent regions-of-interest (ROI). The approach resolves depth at the 1-mm scale but does not permit associations between laminae at different depths. In previous work, we utilized a nearest-neighbor Euclidean pointto-surface definition of depth to define layers in SC by creating depth kernels that associate a particular locus of superficial tissue with deeper tissue. These kernels were small computational cylinders extending from the surface of the brainstem inwards, normal to the surface of SC. With this method, we successfully demonstrated depth-dependent BOLD responses from visual stimulation and attention (Katyal & Ress, 2014;Katyal, Zughni, Greene, & Ress, 2010) as well as saccadic activity in SC (Savjani et al., 2018) and frequencydependent auditory stimulation in IC (Ress & Chandrasekaran, 2013). However, defining a laminar neighborhood of tissue using nearest-neighbor associations is a drastic oversimplification. In the small convoluted colliculi, the cylinders are susceptible to oversampling or undersampling deep tissue, depending on the gradient of the curvature. In order to capture the physical structure of collicular layers as observed in histology and MR microscopy, a more topologically consistent method of defining depth within the colliculi is desirable. Several methods have been developed to compute depth in cortex, including solutions of Laplace's equation (Jones, Buchbinder, & Aharon, 2000) and equi-volume methods (Waehnert et al., 2014). However, they are computationally expensive to solve in a stable and accurate fashion, and have never been applied to midbrain. Here, we compare two methods for depth-analysis in human midbrain: our previous Euclidean approach, and an algebraic level-set algorithm that utilizes two surfaces. We adapted the level-set method originally implemented in cortex (Kim, Taylor, & Ress, 2017) to human midbrain using the superficial brainstem tissue-CSF boundary as the outer surface and the CA as the inner surface to create a depth coordinate normalized to the thickness of the tissue under investigation. First, we compared the ability of both methods to delineate visualand saccade-evoked function in superficial SC, concluding that both methods perform similarly. We also delineated the inner boundaries of the colliculi in order to set a precedent for future functional studies in the deepest layers of SC. This was accomplished by localizing the outer boundary of the PAG using a combination of Euclidean and level-set methods applied to gray-matter tissue-probability maps obtained at 9.4T. Lastly, we evaluated anatomical thickness of the colliculi, using both methods to measure the distance from the collicular surface to the outer boundary of the PAG. In this case, the level-set approach provided more precise measures of tissue depth in the superior and inferior colliculus. Altogether, we show that our level-set method for depth analysis meets or exceeds the performance of more typical Euclidean approaches and suggests new applications of our method for mapping the human brainstem. | Subjects We applied our level-set scheme to 20 subjects. Eight of these participated in subcortical fMRI experiments at 3T (Savjani et al., 2018), giving informed consent under procedures reviewed and authorized by the Baylor College of Medicine Institutional Review Board. The other 12 subjects participated in quantitative MRI experiments at 9.4T and underwent a physical and psychological check-up by a local physician and provided written informed consent following local research ethics policies and procedures. These investigations were conducted in agreement with the World Medical Association Declaration of Helsinki in its most recent version (2013). | Acquisition and segmentation of anatomical images For each of the eight subjects scanned at 3T, we acquired highresolution (0.7-mm voxels) T 1 -weighted anatomical volumes using an MP-RAGE sequence (TI = 900 ms, TR = 2,600 ms, 9 flip angle, TA = 22 min) on a 3T Siemens (Siemens Medical Solutions, Erlangen, Germany) Trio scanner with a product 32-channel head coil. The remaining 12 subjects were scanned at 9.4T using a Siemens Magnetom scanner equipped with a 16-channel, dual-row transmit array operating in the circularly polarized mode, and a 31-channel receive array (Shajan et al., 2014). We acquired B 1 field maps by the actual flip angle imaging method (Yarnykh, 2010;Yarnykh & Yuan, 2004) with nominal flip angle FA = 60 ; repetition time TR 1 /TR 2 = 20/100 ms; echo time TE = 7 ms, voxel size = 3 × 3 × 5 mm

3 ; and acquisition time TA = 225 s. We further acquired high-resolution (0.8-mm isotropic voxels) quantitative whole brain T 1 -maps using an MP2-RAGE sequence (TI1/TI2 = 900/3500 ms, TR = 6 ms, volume TR = 9 s, TE = 2.3 ms, GRAPPA = 3, partial-Fourier factor 6/8, 256 RF pulses, and TA = 9 min 40 s). At 9.4T, this approach permitted high-quality anatomical imaging with the much shorter acquisition time than at 3T. The MP2RAGE images were reconstructed off-line while correcting for deviations from the nominal excitation flip angle and for T 2dependent deviations in inversion efficiency of the adiabatic inversion pulse . From the quantitative T 1 maps, two sets of synthetic, B 1 -artifact-free, T 1 -weighted, MP2RAGE contrast images were generated pixel-wise from the analytical model equations described in the Appendix of Marques et al. (2010). The first image set was used for brain tissue segmentation in volume space using SPM12. The white-matter gray matter tissue contrast could be increased with respect to the image acquisition by setting the TI 2 to 2,650 ms, and the inversion efficiency to 0.9 while the remaining parameters were the same as in the MR-acquisition. The second image set was generated to improve the quality of the FreeSurfer segmentation (performed using 100 iterations and the following recon_all options: "-highres -3T -schwartzya3t-atlas") and was based on the following parameters: TI 1 /TI 2 = 950/2100 ms; flip angle = 5/3 ; volume TR = 6 s; and an inversion efficiency of 0.8451 (corresponding to the experimentally determined median value found in brain tissue). For all subjects, to generate the outer segmentation, the brainstem tissue (BS) of each subject was initially identified using a probabilistic Bayesian approach in the FreeSurfer 6.0 software package (Iglesias et al., 2015). These segmentations were then edited manually to obtain precise delineation of the four colliculi and their vicinity. To create the inner segmentation, the CA was identified using a mixture of manual and automatic region-growing tools implemented in ITK-SNAP (Yushkevich et al., 2006). | Surface modeling A surface model S 1 was estimated at the interface between the brainstem tissue and adjacent CSF or thalamic tissues ( Figure 1a). An initial isosurface was created directly from the segmentation using MATLAB R2016a (Mathworks Corp., Natick, MA). Then, to reduce voxelation artifacts, we applied five iterations of refinement using a variational, volume-preserving deformable surface algorithm (Bajaj, Xu, & Zhang, 2008;Khan et al., 2011;Xu, Pan, & Bajaj, 2006). The same procedure was applied to the CA segmentation to construct a second surface, S 2 ( Figure 1e). | 3D level-set depth-mapping We calculated a normalized signed distance function w using two separate physical distances relative to the corresponding surfaces ( Figure 2a,b). First, Euclidean distances were calculated for voxels in the volume to the nearest triangle of the designated surface S 1 and S 2 , giving a 3D mapping of each voxel to its distance value (Eberly, 1999). Second, the sign for each voxel was determined based on its location; voxels enclosed within the surface were assigned positive distance values, while voxels outside the surface were given negative distances. Thus, distances from S 1 (d 1 ) became increasingly positive from the surface to BS into deeper tissue, while distances from S 2 (d 2 ) were only positive within CA and negative in BS and surrounding CSF. F I G U R E 1 Isosurfaces of tissue proximal to the cerebral aqueduct constructed at five values of w for one subject scanned at 3T, with w = 0 representing the interface between brainstem tissue and CSF (S 1 ) and w = 1 representing the cerebral aqueduct (S 2 ) Finally, we calculated a weighted sum of the signed distance functions, with distance parameter w, using a level-set scheme that solves an Eikonal equation (Bajaj et al., 2008;Khan et al., 2011;Kim et al., 2017): where the level-set parameter w provides a normalized depth coordinate between the two surfaces. The depth metric w (Figure 2c) is zero on the brainstem surface (d 1 = 0) and unity at the surface of CA (d 2 = 0); w < 0 is outside BS and w > 1 within CA. Because the normalized depth coordinate is independent of the variable tissue thickness, the above equation also enables generation of isosurfaces that evolve F I G U R E 3 Left to right: axial, coronal, and sagittal cross sections of normalized depth maps in four subjects scanned at 3T, and location of cross sections (pink: axial, green: coronal, and blue: sagittal) on brainstem surface models (S 1 ) of each subject. Axial cross-sections show both superior colliculus (Subjects 1 and 2) and inferior colliculus (Subjects 3 and 4). Sagittal cross sections ran through the cerebral aqueduct (Subject 1), through the crown of the colliculi (Subjects 2 and 3), and through tissue adjacent to the cerebral aqueduct (Subject 4) F I G U R E 2 Distance maps generated for one subject overlaid on an axial T 1 MRI image of superior colliculus (far left) acquired at 3T. Shown in panels to the right are Euclidean signed-distance functions (a) d 1 , (b) d 2 , (c) normalized depth, w and (d) level-set depth derived from w by ray tracing. Black dashed lines show the surface representations of the superior colliculus (S 1 ) in (a) and the cerebral aqueduct (S 2 ) in (b), at signeddistance function values of zero smoothly from S 1 to S 2 (Figure 1a-e) as well as outside of the domain bounded by the surfaces. Taking advantage of this computational feature, the changes in curvature that distinguish the colliculi became less obvious with increasing w (moving from brainstem surface to CA), and the level-set surfaces of w ultimately morph into the cerebral aqueduct, S 2 . Normalized depth maps were consistent between subjects ( Figure 3), showing a smooth progression between the brainstem surface at w = 0 to the CA at w = 1. As evident in the surfaces shown in Figure 1, the characteristic curvature of the colliculi diminished with increasing w depth. | Generation of streamlines and depthaveraging kernels To obtain unique correspondences between S 1 and S 2 , we treated w as a pseudopotential. We calculated r streamlines that failed to make sufficient spatial progress after each iteration beyond a minimum threshold (typically 0.05 voxels) before reaching CA were removed. Any streamlines that remained incomplete (w < 1) were also discarded after a maximum number of iterations (typically 64 in the direction of positive w and 32 in the negative direction). To create depth-averaging kernels that associated voxels on the surface of BS with deeper tissue, we gathered all streamlines within a chosen radius, typically 0.7 mm, of manifold distance around each vertex on S 1 (Figure 4a). The coordinates of each streamline were then rounded to the nearest voxel and associated with their corresponding vertex of origin on the brainstem tissue surface. For comparison, we utilized our earlier approach that used only the Euclidean distance metric d 1 to define depth. In this method, depth kernels were computed by extending a 0.7-mm manifold radius disk of BS surface voxels along their mean normal in both directions, yielding a cylinder of voxels ( Figure 4b). | Depth profiles of visual stimulation-saccadeevoked activity in SC To test the new level-set streamline approach, we re-analyzed existing data on the polar-angle representation of saccadic eye movements in SC (Savjani et al., 2018). For both stimulation and saccades, we measured the peak BOLD response at depths between 0.5 mm outside BS and 3.5 mm inside BS at increments of 0.1 mm, averaging the values within each 1.2-mm-wide bin. Given that the nature of our stimulus was expected to elicit narrow bands of saccade-evoked activity in F I G U R E 4 (a) Quasi-axial view of a surface representation of superior colliculus (S 1 ) and cerebral aqueduct (S 2 ) in one subject scanned at 3T with three level-set depth-averaging kernels (radius = 0.7 mm, manifold distance). Their component streamlines (dark green) are regridded (light green) to the spatial resolution of the anatomical volume (0.7 mm isotropic voxels). (b) In comparison, cylindrical depth-averaging kernels demonstrate undesirable overlap ("contention", red arrow) and inappropriately directed sampling in deeper collicular tissue. (c) Quantification of the deviation metric, Δd, between level-set and Euclidean sampling methods. (d) Relationship between Euclidean nearest-neighbor depth and level-set depth in collicular tissue. (e) Surface representations of brainstem (S 1 ) of one subject, with deviation Δd values at three level-set depths mapped back to the surface vertices from which each streamline originates intermediate SC, as detailed in (Savjani et al., 2018), we automatically generated elliptical ROIs at the surface of SC to capture the most robust activity. We expected the profiles for saccades to be shifted deeper into SC compared with those from visual stimulation. The reliability of our data was established using a bootstrapping method. We resampled across runs with replacement over 2000 iterations, calculating a new laminar profile for each resampled average. This data were used to obtain the mean depth profiles across subjects with their corresponding 68% confidence intervals, and depth values of peak activity with their 68% confidence intervals and p-values ( Figure 5). Depth profiles generated using streamlines and level-set physical tissue depth were compared to profiles using our previous cylinders and Euclidean depth. | PAG analysis in midbrain We estimated the outer boundary of the PAG using quantitative structural data obtained at 9.4T. For this purpose, the unified segmentation approach (Ashburner & Friston, 2005) and a gray-matter tissue-probability atlas (Lorio et al., 2016) in SPM12 were applied to the MP2RAGE-based image sets with optimized white-gray matter contrast (as described above). This map was qualitatively discriminative of the PAG boundary at very high probability thresholds (Figure 6a). To quantify an optimal probability threshold (P thr ), we minimized spatial uncertainty (Ress, Harlow, Marshall, & McMahan, 2004) of the PAG boundary across subjects. Since the outer boundary of the deep PAG roughly forms an annulus surrounding CA, we chose to define it in units of our Euclidean distance metric from CA, d 2 . We used our new levelset depth kernels to quantify P as a function of depth, binning increasing values of d 2 from S 1 to S 2 , at increments of 0.05 mm (Figure 6b). We initially calculated depth profiles for comparison using normalized w distance, level-set path distance, and Euclidean distance from S 1 , as well as using the cylinder depth kernels. However, they provided almost no contrast between the PAG and the surrounding tissue, and using d 2 as a reference depth metric clearly proved to be the most useful method. From the streamline depth profiles, we then computed the spatial uncertainty U: where σ P is the SD of P across subjects at each value of d 2 . Examining the mean values of U with respect to gray matter probability, we determined that the optimal P thr = 0.96. We then constructed and characterized a 3D estimate of the PAG boundary (Figure 6c,d). The segmentation for CA was upsampled by a factor of two and skeletonized to create an axis. Using this axis as a normal vector, 2D slices were taken through the brainstem probability map at each point along the axis. These contours were then fit in a leastsquared error sense with ellipses centered around the cerebral aqueduct. These ellipses were characterized by three parameters: left-right axis length a, dorsal-ventral axis length b, and aspect ratio ε = b/a. (Figure 6e). These 2D ellipses were then flood filled and transformed back to the locations of the slices they were calculated from, to create a 3D segmentation of the elliptical fit of the PAG. We termed the resulting surface generated from this segmentation S ellipse . A second 3D representation of the PAG was also obtained as a smooth iso-probability surface, S PAG , of the gray-matter probability map at P thr using the methods described above (Figure 6d). Upon grossly comparing S PAG and S ellipse , the former was, as expected, very similar to the elliptical fit described above. | Measurements of collicular thickness We used path distances along the streamlines to create a

new distance map estimating physical tissue depth, obtained by contour integration along each streamline. The path integrals, which oversample the volume, are then re-gridded using a Delaunay triangulation approach (Amidror, 2002) onto the anatomical reference volume (Figure 2d). This level-set depth stopped increasing at CA, whereas Euclidean depth ( Figure 2a) continued to increase into the center of BS. For the 12 subjects from which we collected gray-matter probability maps to delineate the PAG, we measured collicular tissue thickness using three different depth metrics: Euclidean distance, level-set path distance, and normalized w distance. Then, collicular thickness was calculated by integrating path distance along each streamline from the surface of the colliculi to S PAG (7A). Table 1 of the bin with the most negative mean curvature defined the axial plane of separation. We used these planes to divide S 1A into four subsets, each containing a single collicular surface. Next, we defined a base-plane for each colliculus by taking all vertices with positive curvature values and fit these with a plane using a least-squares method. We then defined the peak location as the point with maximum Euclidean distance from this base-plane. We also experimented with defining the collicular peaks as the vertices with greatest positive curvature but found that these points of maximum curvature varied significantly across individual colliculi as well as across subjects. In contrast, the peak distance definition was far more stable, and was therefore used as our standard approach. where σ is the standard deviation of the Gaussian function, and FWHM = 2σ ffiffiffiffiffiffiffiffiffi ffi 2ln2 p . The final thickness value was then calculated as the weighted sum, T = P n 1 A n t n = P n 1 A n . | Euclidean versus level-set methods Unlike the kernels generated using the cylinder method, the level-set streamlines avoided overlap through their nonlinearity and compression with increased depth (Figure 4a With respect to metrics of physical tissue depth, the relationship between Euclidean and level-set depth was sub-linear (Figure 4d), reflecting increasing curvature in deeper collicular tissue. Additionally, level-set depth, or path distance along the streamlines, stopped increasing at CA, whereas Euclidean depth continued to increase into the center of BS (Figure 2a,d). We established a metric Δd to quantify the deviation between level-set and cylinder depths (Figure 4c). In addition to measuring deviation between the two methods, this comparison also quantified the non-linearity of the level-set depth trajectories. At each point along each streamline, Δd was the orthogonal distance to the S 1 normal vector of the vertex from which the streamline originated. Greater values of Δd at increasing level-set depth were consistent with the warping observed above. Δd was mapped back to S 1 at various depths ( Figure 4e). Non-linearity was very small (<<1 mm) for depths <3 mm and became substantial (>1 mm), at ≥5 mm depth. Largest error (and non-linearity) was in intercollicular regions, demonstrating where the Euclidean cylinder method is least appropriate. | Depth profiles of visual stimulation-and saccade-evoked activity in SC In superficial SC, our level-set depth was very similar to the original Euclidean depth. Our measurements of the BOLD response as a function of depth confirmed similar depth profiles, including the previous finding that the mean depth profiles across subjects for saccadic activity was shifted deeper into SC compared with those obtained from visual stimulation. This is consistent with the functional organization of SC delineated in animal models (Fuchs, Kaneko, & Scudder, 1985;Sparks & Hartwich-Young, 1989;Wurtz & Albano, 1980). The profiles generated using our new method did not show significant differences from those utilizing the cylinder method ( Figure 5). The laminar depth profiles generated using our streamline kernels and normalized F I G U R E 6 Gray-matter probability map of one subject scanned at 9.4T with contour lines drawn at p = .96 (red). (b) Left: Individual depth profiles (gray) sampling gray-matter probability maps using streamlines, with respect to distance from cerebral aqueduct (CA). Mean depth profiles across subjects is shown in black. Right: Individual values of spatial uncertainty U with respect to gray-matter probability. Mean across subjects is shown in black. (c) Slices of the gray-matter probability map of one subject taken normal to CA. Three sections are shown running through superior colliculus (SC; 1), intercollicular sulcus between SC and inferior colliculus (IC; 2), and IC (3). Contour lines drawn at p = .96 are shown in light red, and the resulting elliptical fits to these 2D slices are in dark red. (d) Surface representations of the PAG (dark gray) enclosing S 2 (CA; light gray), shown with S 1 (brainstem surface; light gray. Three orthogonal slices of one subject's T 1 anatomy volumes are shown. Light red depicts the actual PAG contour at p = .96 (S PAG ). Dark red depicts the boundary of the estimated PAG segmentation (S ellipse ) reconstructed from transforming the 2D elliptical fits back to their 3D locations. (e) Left: Mean ± SD of elliptical parameters (a = left-right semi-axis; b = dorsal-ventral semi-axis; ε = aspect ratio) across subjects. Right: Mean ± SD elliptical parameters a (orange) and b (green) along the length of the cerebral aqueduct, normalized to the mean length of CA w depth also demonstrate the depth differences between stimulationand saccade-evoked activity to a similar level of significance as the other methods. | Characterization of the PAG Upon examination the gray matter probability maps of the PAG, we found that a large proportion of the dorsal midbrain is attributed a high probability value, even including superficial tissue that would be considered to be colliculus (Figure 6a). However, after using a combination of the streamline depth kernels and Euclidean distance from CA, we were able to localize the PAG within this ambiguous region with a relatively small degree of uncertainty (Figure 6b). The PAG for all subjects was generally able to be fitted with ellipses resulting in a root-mean-squared error of 0.65 mm (Figure 6c However, the fraction of these outlier vertices, with separations >2 mm was very small, mean across subjects was 4.3 ± 2.7%. Thus, the PAG is well characterized as a series of smooth ellipses that vary along the length of the CA. Our elliptical fitting procedure smoothed out many of the irregularities present in the actual PAG contours derived from the gray-matter probability maps (Figure 6d), though some inhomogeneities are still observable in orthogonal cross-sections of S ellipse . However, particularly visible in a sagittal view, these "dimples" occur in the plane of the 2D ellipses, or in a direction that is normal to CA, reflecting some natural variance in the elliptical parameters along the length of CA. Overall, our results confirmed the utility of Euclidean distance from CA as a reference depth metric for this region of the midbrain. | Measurements of collicular tissue thickness Thickness maps calculated using level-set physical distance from S 1 to S PAG showed that the tissue of the colliculi tends to be thickest at its crown, as expected, but also extending laterally, particularly the superior lateral edges for SC. The tissue inbetween left and right colliculi, the intercollicular sulcus, was thinnest ( Figure 7a). Mean data across subjects for tissue thickness at the four peaks of SC and IC as well as the two sulci between left and right SC and IC are shown in Table 1 | DISCUSSION We compared an algebraic level-set scheme utilizing a normalized depth coordinate, w, to a Euclidean depth-mapping scheme for midbrain tissue between the surfaces of the brainstem and the cerebral aqueduct in humans. Previously, we used computational cylinders and Euclidean nearest-neighbor definitions of depth to estimate depth profiles of functional activity evoked by visual stimulation and attention (Katyal et al., 2010;Katyal & Ress, 2014;Savjani et al., 2018). However, the cylinder method was limited by contention of depth kernels, rendering it susceptible to oversampling deep tissue within individual colliculi and undersampling tissue in the inter-collicular sulcus. In contrast, the resulting nonlinear coordinate system in native brain space using our level-set method is specifically adapted to the 3D shape of the tectum. Our novel streamline depth kernels never crossed paths by nature of the pseudopotential and compress or expand along their depth, which avoided mis-association between deep and superficial voxels using the cylinder method. The streamlines give a one-to-one association between the brainstem surface and the cerebral aqueduct (CA) at all spatial locations. We also defined a new metric for physical tissue depth using streamline path distance. Both this level-set physical depth and the normalized w depth provide an analytic depth coordinate for each voxel. Additionally, the streamline depth kernels enable surface-based smoothing of the functional activity, provide mappings between the superficial surface (S 1 ) and deeper tissue, and allow quantitative data, such as gray matter probability, at different depths to be projected onto the surface (Dale, Fischl, & Sereno, 1999;Glasser et al., 2016). The parcellation of tissue into kernels provided depth relationships between deep and superficial tissue at a single spatial location, or over any functionally defined superficial surface. For example, the retinotopic organization of superficial superior colliculus (SC) (Cynader & Berman, 1972) could be related to the putative somatotopic organization in the deep layers of human SC suggested by experiments on animal models (Clemo & Stein, 1991;Jay & Sparks, 1987;Nagy, Kruse, Rottmann, Dannenberg, & Hoffmann, 2006;Wallace, Wilkinson, & Stein, 1996). Overall, the streamline kernels are superior to the cylindrical kernels in their ability to create logical associations between tissue. However, we show that both the Euclidean and level-set depth metrics have their merits in appropriate regions of the midbrain. Usage of streamline depth kernels and level-set physical depth produced depth profiles of visual stimulation-and saccade-evoked activity in superficial SC similar to those previous obtained using cylindrical kernels and Euclidean distance from the surface of the brainstem, d 1 , as the reference depth metric ( Figure 5). This similarity was maintained throughout the depth range over which we calculated the depth profiles, 0-3 mm, and negligible differences between the methods in superficial tissue suggest that either method works well at these superficial depths. Meaningful functional activity did not extend deep enough into SC for us to determine any significant differences between Euclidean and level-set depth metrics beyond these depths. This is consistent with our findings in Figure 4d, which demonstrate that substantial differences (e.g., >0.5 mm) between Euclidean and level-set path distance metrics only emerge at depths >3 mm. Depth profiles generated using normalized w distance and streamline kernels also produced results with similar degrees of significance and variability compared to those generated with physical measures of depth. According to Figure 4d, the medial regions of the colliculi are most susceptible to significant deviations between cylindrical and streamline sampling methods. Notably, upon visual inspection, the elliptical ROIs obtained in our fMRI experiments typically did not cover medial SC, which may explain the lack of significant differences in the depth profiles obtained using the two methods. On the other hand, the anatomical structure of the PAG is best defined using depth values according to Euclidean distance from CA, or d 2 ( Figure 6). As d 2 is straightforwardly defined using the boundary of CA as a reference surface, this is in agreement with the qualitative description of the PAG as roughly a cylindrical tube surrounding CA. Thus, Euclidean distances from an appropriate reference surface are sufficient to delineate both superficial colliculus and the deep PAG. Because of the differences between the methods that emerge at greater depths within the midbrain, characterized by our metric Δd, we predict that our streamlines and level-set depth metrics will provide the greatest utility in regions of intermediate depth, namely deep SC and IC. This hypothesis is made on the basis that our level-set streamline kernels more accurately follow the natural curvature of F I G U R E 7 (a) Surface representations of brainstem (S 1 ) of four subjects scanned at 9.4T, overlaid with collicular tissue thickness maps. (b) Determination of collicular peaks and intercollicular sulci. The collicular surface S 1A was divided into four quadrants using a quasisagittal (blue) and axial (fuchsia) plane. A base plane (yellow) through the vertices with zero curvature was defined for each colliculus, and collicular peaks (red) were defined to be the vertices with greatest distance to each base

plane Intercollicular sulci (cyan) were vertices on the mid-sagittal plane closest to the line between left and right peaks midbrain tissue. Particularly, Figure 4d demonstrates that the greatest deviations between Euclidean and level-set methods are located in the medial regions of the colliculi, where the surface normal of the brainstem tissue point away from CA, indicating where use of cylindrical depth kernels may be least appropriate. In regard to the laminar structure of the colliculi, isosurfaces of our normalized w distance may provide a more accurate representation of the anatomical morphology of midbrain tissue compared with Euclidean definitions of tissue depth. This should be validated through comparison studies between highresolution in vivo brain imaging and histology on post-mortem brains, similar to (Loureiro et al., 2018). Further confirmation of the utility of our level-set method in deeper midbrain structures, particularly for fMRI experiments, should be achieved through studies such as those involving reach-related activation in SC (Linzenbold & Himmelbach, 2012;Liu, Duggan, Salt, & Cordeiro, 2011;Sparks & Hartwich-Young, 1989 may appear to imply that some of our subjects suffered from unilateral degeneration, our method is one of many possibilities. Developing a way to accurately measure collicular tissue thickness depends on our ability to localize the entire boundary around these nuclei. Again, comparison studies between brain imaging and histology on postmortem brains (Loureiro et al., 2018) will be necessary to provide a complete analysis of collicular topology. We strove for a method to relate observed function back to the anatomical structures from which they originate. Studying brain function within a structural context, as in a standardized atlas space (Evans et al., 1994;Mazziotta et al., 2001;Talairach & Tournoux, 1988), allows for localization of relevant structures and, consequently, generalizability across populations and studies. For example, the MNI atlas allows non-linear warping of 3D MRI brain scans from individuals onto a common template and has become a standard tool for reporting fMRI results in volume space. The PAG particularly lends itself for atlas comparisons, since this brain structure is relatively stable across subjects and with age (Keuken et al., 2017). Our experiments assumed that the tissue probability maps we used were well-suited to study the PAG. Usage of these maps has been validated using other deep brain nuclei including the thalamus, caudate, putamen, substantia nigra, subthalamic nucleus, red nucleus, and cerebellar dentate (Lorio et al., 2016). To our knowledge, this work is the first application of these methods to the PAG. Though these maps were not explicitly designed for localization of the PAG, small spatial uncertainty values ( Figure 6b) suggest that this application is appropriate. Our more rigorous quantitative method confirms the use of high gray-matter probability values for tissue classification. Furthermore, under the assumption that the PAG boundary can be fit by a continuous surface, our estimation scheme using ellipses worked well. We observed low root-mean-squared errors for individual elliptical fits, and low coefficients-of-variation for the ellipse parameters across subjects. Further extension of our techniques to provide a fully 3D coordinate system may utilize normalized depth and angle from a point of origin on the mid-sagittal plane through CA. This could potentially allow localization of midbrain nuclei such as the PAG using polar coordinate shapes such as dimpled or convex limaçons, which may provide a greater degree of accuracy than the simple ellipses we utilized in our work. Our efforts towards characterizing and subsequently reconstructing the PAG provide a first step in the direction towards creating a generalized atlas of midbrain nuclei ( Figure 6). Using CA as a central axis, as we have done, could form the basis of a quantitative atlas of the midbrain in a normalized coordinate space such as the MNI-305 (Mandal, Mahajan, & Dinov, 2012), but at higher spatial resolution. While we concentrated our evaluation of this depth-mapping scheme to the dorsal portion of midbrain, our method can be also effective for ventral midbrain, for example, the red nucleus. It may be possible to interpolate the neuraxis between the exit of CA into the fourth ventricle and the geometric axis of the spinal cord at the base of the brainstem to facilitate similar methods in pons and medulla. Our depth analysis assumed that the laminar structure of SC is well-delineated by a depth metric derived from a combination of SC's superficial surface and the CA. In order to minimize oversampling and overlapping depth projections, selecting an appropriate reference structure was key for defining depth in the midbrain. The CA runs rostro-caudally through the midbrain ventral to the colliculi. It emerges developmentally from the neural canal, the cavity in the center of the neural tube. The midbrain develops around the CA, thus positioning it as an important anatomical feature in the midbrain. Ideally, the resulting isosurfaces of w should be parallel to the laminae of the colliculi. However, our choice of surfaces to calculate depth coordinates was also a matter of convenience. Although our present results are consistent across individuals and demonstrate the utility of our depth mapping approach in vivo, it remains to be seen how well the depth-mapping scheme aligns with laminarly organized collicular cytoarchitecture. In future work we, will apply our depth mapping approach to publicly available datasets, such as BigBrain and CIT168 MNI, and high-resolution post mortem MRI and histology (Loureiro et al., 2018;Sitek et al., 2019), where individual collicular laminae are distinguishable. This will allow us to validate our method's sensitivity to ground truth depth-dependent cytoarchitecture. | CONCLUSIONS We established a series of morphologically reasonable methods for determining tissue depth in human SC and IC based solely on structural MRI. The level-set scheme provides coordinates for both normalized and physical depth that accommodate the complex anatomical structure of midbrain tissue, and the associated streamline method was able to distinguish function in the superficial layers of SC at the same precision as our previous Euclidean method. Euclidean distance from CA provides an accurate way to localize the PAG. We also believe that our work is the first analytic method for in vivo determination of anatomical tissue thickness in human midbrain. Because the streamline depth kernels logically follow the curvature of the tissue to establish unique relationships between all layers, our method should be useful to describe associations between tissue at different depths, such as between the deep multi-sensory layers of superior colliculus and the superficial retinotopically organized layers. Furthermore, depth-analysis studies using our normalized depth coordinate applied to a large population of subjects may enable a fully 3D atlas of human brainstem. ACKNOWLEDGMENTS We gratefully acknowledge the support of lab members Elizabeth Halfen and Amanda Taylor in performing this work. This international collaborative work was supported by United States and German agencies: NIH grants R01 EB027586, K25 HL131997, and R01NS095933, BMBF CRCNS US-German Research Proposal Number 1822655. DATA AVAILABILITY STATEMENT The data that support the findings of this study are available from the corresponding author upon reasonable request. Differential expression of Cosmc, T-synthase and mucins in Tn-positive colorectal cancers Background The Tn neoantigen (GalNAcα1-O-Ser/Thr) is an O-glycan expressed in various types of human cancers. Studies in several Tn-expressing cancer cell lines and pancreatic tumors have identified loss of Cosmc expression caused by either mutations or promoter hypermethylation. In this study, we explored the mechanism(s) for Tn expression in human colorectal cancers (CRC). Methods Tn-expressing cell populations were isolated from CRC cell lines by Fluorescence-associated cell sorting (FACS). The expression of the Tn and sialylated Tn (STn) antigens, Cosmc, T-synthase, and mucins was characterized in paired specimens with CRC and in CRC cell lines by immunostaining, western blot, and qPCR. Results Using well-defined monoclonal antibodies, we confirmed prevalent Tn/STn expression in CRC samples. However, a majority of these tumors had elevated T-synthase activity and expression of both Cosmc and T-synthase proteins. Meanwhile, Tn antigen expression was not caused by mucin overproduction. In addition, we found that Tn-expressing CRC cell lines had either loss-of-function mutations in Cosmc or reversible Tn antigen expression, which was not caused by the deficiency of T-synthase activity. Conclusions Our results demonstrate multiple mechanisms for Tn expression in CRCs. Electronic supplementary material The online version of this article (10.1186/s12885-018-4708-8) contains supplementary material, which is available to authorized users. Background The Tn neoantigen (GalNAcα1-O-Ser/Thr) and its sialylated form (sialyl Tn, STn) are tumor-associated carbohydrate antigens (TACAs) expressed in a broad range of human cancers, including those in the colorectum, breast, prostate, lung, ovary, cervix, and pancreas [1,2]. The Tn/STn neoantigens have promise as tissue or serum biomarkers in cancer detection and prognosis, and in providing a tumor-specific epitope for targeted therapy [2]. In addition, they are involved in promoting cancer progression or protecting malignant cells from the surveillance of the immune system, hence being valuable therapeutic targets in clinical treatment [3]. Although Tn has been recognized as a neoantigen, few analyses have used paired normal and tumor samples to define its expression [2]. Some studies compared the healthy individuals and the patient group, which may not reflect the progression of the Tn antigen. In addition, the Tn positivity rate varies within a particular cancer type. For example, 68 of 146 (47%) colorectal cancers (CRCs) were reported to be Tn positive, while another study concluded 72-81% [4,5]. The differences were probably influenced by the detection approaches used, since many studies have used GalNAc-binding lectins, such as Vicia villosa agglutinin (VVA) and Helix pomatia agglutinin (HPA), or the antibodies that were privately in-house generated and often not extensively characterized for specificity [6]. We have utilized an IgM-type monoclonal antibody BaGs6 (CA3638) to the Tn antigen [7]. BaGs6 specifically recognizes glycoconjugates containing GalNAcα1-O-Ser/Thr but not blood group A and related glycans terminating in GalNAc, and BaGs6 also stains tissue sections from mice engineered to express the Tn antigen but does not stain normal tissues [8,9]. Therefore, BaGs6 is a reliable and well-characterized reagent to explore the Tn positivity in human cancers. The mechanisms of generating the Tn neoantigen in human cancers are unclear. The Tn antigen is a precursor structure biosynthesized in the Golgi apparatus by a family of twenty different polypeptide α-N-acetylgalactosaminyltransferases (ppGalNAc-Ts), which transfer GalNAc from the donor UDP-GalNAc to a Ser or Thr residue in glycoproteins [10]. In normal tissues, the Tn antigen is usually undetectable due to its efficient conversion into more extended glycans, primarily to the core 1 structure (Galβ1-3GalNAcα1-O-Ser/Thr, the T or TF antigen) [6]. This modification is catalyzed by a single enzyme, T-synthase (C1GALT1, UDP-Gal:GalNA-cα1-O-Ser/Thr glycopeptide β3-galactosyltransferase) in the Golgi apparatus [11]. The core 1 structure is further elongated to extended core 1 O-glycans, or branched to core 2 structures, or sialylated [6]. In the gastrointestinal tract (GI tract), GalNAcα1-O-Ser/Thr may be converted into core 3 O-glycans [12]. In addition, the Tn antigen can be sialylated to form STn [6]. The T-synthase is a unique enzyme whose correct folding requires an X-linked molecular chaperone, Cosmc (Core 1 β3-Gal-T-Specific Molecular Chaperone, also named C1GalT1C1) [13]. In the ER, Cosmc interacts co-translationally with non-native T-synthase to generate active T-synthase, which is then transported to and functions in the Golgi apparatus [14]. Defective Cosmc function results in aggregation and proteosomal degradation of T-synthase associated with expression of the Tn antigen [13]. Studies on Tn-expressing cancer cell lines and patients with Tn syndrome revealed loss-of-function mutations in the Cosmc gene and loss of T-synthase (Additional file 1). Promoter hypermethylation of Cosmc was also identified in Tn-positive human pancreatic cancers and the Tn4 cells, suggesting that reduction of Cosmc and T-synthase contributes to Tn neoantigen expression in human cancers [15,16]. Here we defined the expression of the Tn and STn antigens and characterized Cosmc and T-synthase in matched CRC specimens and in several CRC cell lines. We conclude that expression of the Tn antigen arises from multiple pathways, including mutation of Cosmc, as observed in some CRC cell lines such as LS 180 and HCT8, and alternative mechanisms in CRC specimens and the SW480 line. Human specimens and cancer cell lines Paraffin-embedded tissue sections from 39 colorectal cancer patients were obtained from the Emory Tissue bank and Dr. N. Volkan Adsay (Departments of Anatomic Pathology, Emory University School of Medicine, Atlanta), and frozen tissues were requested for 27 cases randomly. For each case, both tumor and its matched normal tissue (normal) were analyzed. Transitional mucosa (TM) was visible in the tumor sections of 11 cases, which is immediately adjacent to the cancer and exhibits

microscopic abnormalities without atypia [4]. Usage of these specimens was reviewed and approved by the Emory University Institutional Review Board (IRB) with informed consent from patients, and the research team did not receive any identifying patient information. Additional ATCC cell line included: HEK293T human embryonic kidney-ATCC CRL-3216; this cell line does not require ethics statements. LOX and LSC cells were obtained and used as previously described [17]. Tn4 cells were obtained and used as previously described [16]. Immunofluorescence Human CRC cells were cultured in Lab-Tek™ II-chamber slides (Thermo Fisher Scientific, Waltham, MA) for 48 h before fixation in 4% formaldehyde for 15 min at room temperature. After washing in PBS, cells were blocked for 1 h in PBS containing 10% (v/v) normal goat serum. Cells were then incubated with the anti-Tn antibody BaGs6 at 4°C overnight followed by Alexa Fluor® 488-or 568-conjugated goat anti-mouse IgM antibody (Invitrogen, Carlsbad, CA) for 60 min at 4°C. After four washes in PBS, nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) for 5 min. The chamber was then removed, and slides were mounted and imaged with a Zeiss Axioplan 2 fluorescent microscope (Zeiss, Oberkochen, Germany). Flow cytometry and fluorescent activated cell sorting (FACS) Cultured CRC cells were trypsinized, washed, and suspended in cold PBS. One million (1 × 10 6 ) cells were incubated with BaGs6 or isotype control (mouse IgM) for 40 min on ice, followed by incubation with Alexa Fluor® 488-conjugated goat anti-mouse IgM secondary antibody. After three washes with cold PBS, cells were analyzed in a Becton Dickinson FACscan flow cytometer (BD Biosciences, San Jose, CA). In FACS, ten to twenty million (1~2 × 10 7 ) cells were immunostained and sorted into 15 ml tubes by a SORP FACSAria II or into 96-well plates by a MoFlosorter (BD Biosciences). Immunohistochemistry (IHC) Tissue sections were deparaffinized, rehydrated, and washed with water. Antigen retrieval was done by heating slides in a pressure cooker for 3 min in citrated buffer (pH 6.0, 10 mM trisodium citrate). After cooling down at room temperature, tissue sections were incubated with 3% hydrogen peroxide and then blocked with 5% normal goat serum in Tris-buffered saline with 0.1% Tween-20 (TBST). Then tissue sections were incubated with primary antibodies at 4°C overnight, followed by HRP-conjugated secondary antibodies (KPL Inc., Gaithersburg, MA) at room temperature for 1 h. Primary antibodies used in this study included those against Tn (BaGs6, mouse IgM), STn (TAG-72, mouse IgG), blood group A antigen, mucin 2 (H-300) (Santa Cruz Biotechnology, Dallas, TX), and mucin 1 (Thermo Fisher Scientific). Signals were visualized by incubating sections with Aminoethylcarbazole (AEC) substrate solution (Invitrogen), and cell nuclei were counterstained with hematoxylin (Invitrogen). Whole tissue sections were mounted in CLEAR-MOUNT solution (Electron Microscopy Sciences, Hatfield, PA) and reviewed by microscopy. The signal intensity was indicated by a numerical scale of 0 to 3 (0 = no staining, 1 = weak staining, 2 = moderate staining, and 3 = strong staining), and the percentage of positive cells was estimated. The IHC score (IS) was calculated by multiplying the staining intensity by the percentage of positive cells. A sample is considered to be positive when the immunohistochemistry score is 50 or greater. Statistical analyses were performed using Paired t test. The correlations between two antibodies' IHC were determined using Pearson correlation coefficient (Pearson's r). Representative slides were scanned with VS120 Whole Slide Scanner (Harvard Medical School Neurobiology Imaging Facility), and pictures were captured using the OlyVIA 2.9 software (Olympus, Tokyo, Japan). Western blot (WB) Frozen human CRC tissues and cultured cells were sonicated or lysed in a Hepes buffer containing 0.5% Triton X-100 and protease inhibitors (Roche Diagnostics Corporation, Indianapolis, IN). The protein concentration was determined using a BCA kit (Thermo Scientific, Waltham, MA) following the manufacturer's instructions. Equal amounts of total protein were separated in SDS-PAGE and transferred onto nitrocellulose membranes. Western blot antibodies included those against Cosmc, T-synthase, β-actin, α-tubulin (Santa Cruz Biotechnology), and the Tn antigen (BaGs6). For human CRCs, each WB band was quantified for its intensity and area with ImageJ, and signal was calculated by multiplying band intensity by the area. T-synthase activity assay T-synthase activity assay was performed following the protocol described previously [18]. Briefly, total cell lysate was incubated with 4-Mu-α-GalNAc (acceptor), UDP-Gal (donor), MnCl 2 , Triton X-100, and O-glycosidase in MES buffer (pH 6.8) at 37°C for 2 h. Reactions were terminated with a stop solution (1 M Glycine-NaOH, pH 9.6). Relative fluorescence units (RFUs) were measured in a Victor Multiple-Label Counter using umbelliferone mode, e.g., Ex 355 nm and Em 460 nm. The specific activity of T-synthase was calculated by normalizing the total activity by the protein concentration and the incubation time. Genomic DNA preparation Genomic DNA of CRC tissues or cultured cells was prepared from the remaining tissue pellet of the protein extraction. Briefly, the pellet was re-suspended and digested in 1.0 mg/ml of proteinase K at 56°C overnight. Then genomic DNA was extracted and purified using the DNeasy blood and tissue kit (Qiagen, Valencia, CA). DNA concentrations were determined with a Nanodrop spectrophotometer (Thermo Scientific). Some human CRC cell lines contain Tn-positive cells To explore molecular mechanisms of Tn neoantigen expression in human CRCs, we examined 8 commonly used CRC cell lines (LS 180, SW480, SW620, SW1116, HCT8, HCT15, T84, and Caco-2) for expression of the Tn and T-synthase. LS174T-Tn(−) and LS174T-Tn(+)-II were used as negative and positive controls for Tn, respectively, and HEK293T was included as an additional negative control [17]. By western blot, the 8 CRC cell lines and HEK293T expressed no Tn antigen, but detectable T-synthase protein (Additional file 3 a). The enzyme activities of T-synthase in these cells were comparable to that in LS174T-Tn(−) (Additional file 3 b). However, by immunofluorescence, we found that three cell lines (LS 180, SW480 and HCT8) had approximately 1~2% of cells that were Tn(+) on the cell surface, while SW1116 expressed the Tn antigen surrounding a whole cell colony (Additional file 3 c). Therefore, several human CRC cell lines contain a small percentage of Tn(+) cells, indicating a mixed population of cells. Tn-positive (Tn(+)) subpopulations of LS 180 and HCT8 cells harbor a mutant Cosmc gene By fluorescence-activated cell sorting (FACS), we isolated Tn(−) and Tn(+) subpopulations from LS 180 and HCT8 parental cells using the anti-Tn antibody BaGs6, where only the cells with the strongest fluorescent signal (top 1%) were considered as Tn(+) for collection. For both cell lines, the majority of parental cells (> 97%) were Tn(−). Immunofluorescence ( Fig. 1a and b) and flow cytometry (Fig. 1c) analyses confirmed Tn antigen expression in the Tn(+) subpopulations. Furthermore, LS 180-Tn(+) cells expressed the STn antigen at the cell surface, while HCT8-Tn(+) cells did not (Additional file 4). Genomic sequencing revealed that LS 180-and HCT8-Tn(−) cells contain a wild type (WT) Cosmc gene, whereas LS 180-Tn(+) cells harbor a T deletion at nucleotide 473 (delT473), and HCT8-Tn(+) cells have delA482 (Fig. 1d). Both delT473 and delA482 resulted in open reading frame (ORF) shifts and truncated Cosmc proteins (Additional file 1). Accordingly, LS 180-and HCT8-Tn(+) cells, as predicted, exhibited significantly lower T-synthase activities when compared to the parental and Tn(−) cells (Fig. 1e). Western blot demonstrated that LS 180-and HCT8-Tn(+) cells were deficient in Cosmc and T-synthase proteins and acquired Tn expression on several glycoproteins (Fig. 1f ). Although the α-tubulin levels varied in different cell populations, staining with Ponceau S indicated equivalent amount of total proteins loaded for WB (Fig. 1f ). These results revealed that Tn expression in cancer cell lines is associated with loss-of-function mutations of Cosmc. Reversible Tn expression in SW480 cells are not due to loss of T-synthase. We conducted several FACS experiments on SW480 cells. As shown in Fig. 2a for SW480, the Tn(+) cells were remarkably enriched (85%) after sorting and collecting the top 1% of positive cells. However, we were unable to maintain the Tn(+) subpopulations. The majority of cells were Tn(−) after 3-4 weeks of expansion. We then sorted single SW480-Tn(+) cells into 96-well plates and obtained single-cell-derived clones (Fig. 2a), A total of 52 individual clones were analyzed for cell surface Tn expression, Among them, 19 (37%) clones were Tn(−), and 33 (63%) clones became heterogeneous for Tn expression (Fig. 2b). Representative histograms of Tn expression of these clones are shown in Fig. 2a. Tn heterogeneity developed from the single-cell derived clones demonstrates that Tn expression in SW480 cells is reversible. We then investigated whether transient expression of the Tn antigen in SW480 cells was caused by temporary reduction of T-synthase or Cosmc. However, western blot showed that SW480-Tn(+) cells had comparable levels of T-synthase and Cosmc to the parental and Tn(−) cells (Fig. 2c). Enzymatic activity assay further indicated that these Tn(+) cells remained a similar level of T-synthase activity as to the Tn(−) cells (Fig. 2d). Hence, unlike LS 180-and HCT8-Tn(+) cells, the transient Tn expression in some SW480 cells were not due to absence or reduction of T-synthase activity. While changes in methylation status can create changes in expression levels, the comparable protein levels coupled with previous data [16] showing that mutation or hypermethylation of Cosmc impairs T-synthase activity, did not indicate a role for methylation in the reversibility that we observed. Prevalent Tn and STn neoantigen expression in human colorectal cancers Using BaGs6, we examined both tumor and matched adjacent normal tissues (assigned as normal) from the same individual of a cohort of 39 CRC cases to determine Tn neoantigen status. Thirty-seven out of 39 (95%) of the tumors examined had detectable Tn antigen on the epithelial cell surface (Fig. 3a). Meanwhile, 7 (18%) of the adjacent normal tissues were Tn(+); the remaining normal sections had either weak intracellular or absent Tn antigen (Additional file 5). The blood group A antigen (BGA), which contains a terminal α-GalNAc, is often a confounding antigen in studying Tn expression, since some reagents may cross-react with BGA. Therefore, we stained the CRC samples with an anti-BGA antibody and observed distinct staining patterns from those observed for BaGs6 (Additional file 6). All cases are either negative for BGA staining or for those that were positive, staining was observed in blood vessels and red blood cells, while BaGs6 stained epithelial cells. BaGs6 does not recognize the BGA antigen, as shown in prior studies on its specificity, which is consistent with the data here. The tumor sections from 11 patients contained transitional mucosa (TM) regions, which are uninvolved histological "normal" crypts adjacent to the atypical cells. Notably, 7 out of the 11 proximal TMs had gradually increased intracellular Tn antigen, compared to the distal TM in the same section and the matched normal sections (Fig. 3a). These CRC tissues were also analyzed for the STn antigen using a monoclonal antibody TAG72. We observed robust cell surface STn antigen in most tumors, but rarely was it expressed in matched normal sections (Fig. 3b). All STn-expressing tumors were also Tn positive (Additional file 5). Similar to the Tn neoantigen, increased (Fig. 3b). Statistical analyses of the IHC score (IS) showed significant higher levels of cell surface Tn and STn antigens in tumors (both p < 0.0001) (Table 1, Fig. 3c) Lack of somatic mutations of Cosmc and B3GNT6 in human CRCs We then analyzed the genomic sequence (single exon) of Cosmc in 27 CRC samples (Case #1 -27). Unlike what we observed for most Tn-expressing cell lines, no somatic mutation was detected in the single coding exon of Cosmc. One tumor harbored a 1-bp deletion in a poly(G) consecutive sequence located in the promoter region, and its effect on Cosmc expression remains unknown. In the GI tract, in addition to the core 1 glycan, the epithelial cells also modify the Tn antigen to form the core 3 structure [1]. Therefore, we also analyzed the coding exon of the core 3 enzyme, B3GNT6 (UDP-GlcNAc:βGal β-1,3-N-Acetylglucosaminyltransferase 6) in the 27 cases of CRCs. No mutation was identified in B3GNT6, suggesting the two key glycosyltransferases in the core 1 and core 3 pathways are rarely mutated in CRCs. In addition, no putative CpG islands were identified in the B3GNT6 promoter, exon1, or intron1 by publically available predictive tools, therefore we did not test the methylation status of B3GNT6. Loss of heterozygosity (LOH) of the Cosmc locus in human CRCs LOH is a common mechanism for gene loss of

function in tumorigenesis. We investigated the LOH status of Cosmc and T-synthase using the SNP-based PCR approach. Six and 13 SNPs of Cosmc and T-synthase, respectively, are located at different genomic regions and were manually selected (Additional file 2). Since Cosmc is located on the X chromosome (Xq24), LOH could be assessed only in females. In normal tissues two alleles generated equal amounts of PCR products, while in tumors the allele with LOH produced less product at that location (Fig. 4a). Among 15 female CRCs examined, 8 (53%) showed LOH of Cosmc within at least one SNP ( Fig. 4b and Table 2). None of the cancer specimens examined contained LOH within the T-synthase (Fig. 4c). Thus, LOH occurred in Cosmc, but not the T-synthase, in a majority of specimens examined. Increased T-synthase and Cosmc expression in human CRCs To explore whether there are changes in Cosmc or T-synthase expression in human CRC, we examined Cosmc and T-synthase in CRC samples at both mRNA and protein levels. Total RNA from paired frozen tissues of 15 patients was subjected to real-time RT-PCR analyses. With a fold-change cut off of 2-fold, Cosmc mRNA levels increased in 8 of 15 tumors, compared to that in the matched normal tissues (p < 0.05), and T-synthase mRNA levels increased in 12 of 15 (p < 0.01, Fig. 5a and Table 2). The transcript levels of B3GNT6 in most of these samples were undetectable, making it difficult to draw a clear conclusion. Moreover, of 24 cases (Case #1-24) examined for protein expression by WB, the majority had increased Cosmc and T-synthase protein levels in the tumor compared to the normal, and only two (Case #3 and #23) showed a decrease in both protein levels in the tumor; Case #8 had decreased Cosmc but not T-synthase levels (Fig. 5b). After quantification of the blots with ImageJ, 18 of 24 cases exhibited elevated Cosmc/β-Actin ratio in the tumor at > 2-fold change, and the Cosmc protein level in overall tumors are higher than that in the normal (p = 0.0007, Fig. 5c). T-synthase protein levels showed similar alterations (p = 0.0015, Fig. 5c). Interestingly, the Cosmc protein level correlated with that of T-synthase, as determined by Pearson correlation coefficient analysis (r = 0.8887) (Fig. 5d). Furthermore, correlated to the protein levels detected by WB, a majority of the tumor samples had increased enzyme activities of T-synthase, compared to the matched normal tissues (Fig. 5e and Table 2). The two tumors (Case #3 and #23) that had reduced Cosmc and T-synthase proteins also exhibited decreased T-synthase activities. Therefore, in Tn-expressing CRCs, it cannot be simply defined by a loss of T-synthase or loss of enzyme activity, although it should be noted, that our analyses cannot assess the proper localization of the enzyme in the Golgi apparatus, where it normally functions, as the anti-T-synthase and anti-Cosmc antibodies are not suitable for immunohistochemistry. Tn antigen expression was not correlated with expression of MUC1 or MUC2 Mucins are heavily O-glycosylated and often differentially expressed in tumors [19]. We considered the possibility that overproduction of mucins may result in insufficient modification of terminal α-GalNAc on mucins by T-synthase, thus leading to Tn expression. To test this possibility, we used IHC to define expression of two major mucins produced in the GI tract, MUC1 and MUC2 ( Fig. 6a and b). Polyclonal antibodies were used to exclude the variables that may affect recognition of the epitopes. While the gel-forming MUC2 was abundant in normal epithelium, it was remarkably reduced or absent in the tumor (p = 0.0019, Fig. 6c, Table 1). MUC1 was significantly elevated in tumor sections (p = 0.0074, Fig. 6c, Table 1), but its expression level did not correlate with the IS of the Tn antigen (r = 0.0208, Fig. 6d). Hence, we observed no positive association between the MUC1 or MUC2 levels and level of Tn antigen expression. Increased ST6GALNAC1 mRNA levels in human CRCs Like the Tn neoantigen, the STn antigen was observed to frequently elevate in the CRC specimens, and all STn-bearing tumors were Tn-positive (Additional file 5). ST6GALNAC1 is the enzyme that is required for generating the STn antigen [20]. To determine whether the STn antigen in CRCs was due to increased expression of sialyltransferases [20], we measured the transcripts of ST6GAL-NAC1 and ST6GALNAC2 by quantitative RT-PCR. Nine of 15 tumor samples tested had remarkably elevated ST6GALNAC1 mRNA levels, but not ST6GALNAC2 (p = 0.0336 and 0.5665 respectively, Fig. 7). Our results Discussion In this study, using specific monoclonal antibodies and multiple approaches, we analyzed matched CRC specimens and several cancer cell lines for expression of the Tn/STn antigens and the core 1 biosynthetic pathway. Our study is unique as it focuses on the progressive expression of Tn in individual colorectal cancer patients [2] and, to our knowledge, it is the first to investigate expression and enzyme activities of the Cosmc/T-synthase pathway in human CRC specimens. Our results showed an increase of intracellular and cell surface Tn antigen in TM regions and tumors, respectively. It not only supports previous reports that both the Tn and STn antigens are prevalent in CRC samples, but also suggests that appearance of intracellular Tn and STn could be an early event during colorectal tumorigenesis. A few other studies also demonstrated the existence of the Tn antigen in gastrointestinal nonmalignant lesions, such as polyps and aberrant crypt foci (ACFs) [21,22]. Therefore, the Tn and STn antigens could be potential biomarkers for prediction or early detection of human CRCs. Although the Tn antigen is prevalent in a variety of human cancers, only a few cancer cell lines robustly express the Tn antigen, including Jurkat, Tn4, LSC and LOX [13,16,17]. LS174T and HT-29 have a small population of Tn(+) cells [17,23]. In this study we identified four additional CRC cell lines having a subpopulation of Tn-expressing cells (Additional file 3). Although LS 180 and SW1116 were reported to be Tn positive, the mechanisms for the Tn expression are unclear [24,25]. Here we identified frame shift mutations of Cosmc in LS 180-Tn(+) and HCT8-Tn(+) cells. The mechanistic explanation for this may be that several Cosmc mutations in cancer cell lines occur in repeated DNA sequences (microsatellites), suggesting that mononucleotide repeat tracts in the Cosmc coding region may be susceptible to the microsatellite instability (MSI) phenotype, like Bax and TGFβRII [26]. The cells with MSI are deficient in DNA mismatch repair and therefore exhibit higher mutation rates in microsatellites [26]. Indeed, LS174T, LS 180 and HCT8 are MSI cells [27]. This may explain the observation that Cosmc mutations in LS 180-Tn(+) and HCT8-Tn(+) cells were identical to that in Jurkat and LS174T-Tn(+)-II respectively, although they were derived from different patients; LSC, LS174T and LS 180 originated from the same patient but harbored different mutations. In addition to LS 180 and HCT8, we identified that Tn expression in SW480 was unstable and did not result from loss or reduction of T-synthase activity. These observations suggest that, while loss of functional Cosmc by mutation, deletion, or hypermethylation contributes to Tn expression in some cancer cell lines and certain types of malignancies (such as pancreatic cancer) observed so far [13,17,23,28,29], other mechanisms do exist to cause revertible expression of the Tn antigen. This may underlie the paradox that while the Tn antigen is a common marker of many tumors in situ, few tumor-derived cell lines express the Tn antigen. Unlike most of Tn-expressing cell lines, we did not detect any Cosmc somatic mutations in 27 cases of CRCs. Similarly, lack of Cosmc mutations were also reported in colorectal cancer in other studies [30,31]. Moreover, although a number of CRC specimens examined harbored LOH at the Cosmc locus, few expressed reduced levels of Cosmc and T-synthase. Conversely, the majority of the tumors had increased Cosmc and T-synthase mRNA and protein levels (Fig. 5). Since Cosmc is located on the X chromosome, our results suggest that these LOH-occurring samples had likely kept the active allele of Cosmc, which was further transcribed at a higher level in cancer cells. Elevated enzyme activities of T-synthase in these samples indicate that the enzyme is correctly folded, which is known to require functional Cosmc. Interestingly, the Cosmc protein levels in the CRC samples were well correlated with T-synthase (Fig. 5d), suggesting co-expression of the two proteins may be regulated in a coordinate fashion, consistent with studies on the promoter elements in the Cosmc and T-synthase genes [32]. Up-regulation of the T-synthase mRNA in CRCs was observed in several gene profiling studies, although the Tn antigen status in these samples was not determined [33]. The molecular mechanism underlying elevated Cosmc and T-synthase expression in Tn-positive tumors is unclear, but there are several possibilities to consider in future work. As a molecular chaperone in the ER, Cosmc may be induced by ER stress in cancer cells. ER stress commonly occurs in malignancies and contributes to many aspects of tumorigenesis. A number of molecular chaperones are up-regulated in cancer in response to ER stress, including heat shock proteins (such as GPR78 and GPR94) and lectin-like chaperones (such as calnexin and calreticulin) [34,35]. It is also possible that the increased Tn antigen regulates T-synthase/Cosmc production via a feedback loop. Our findings in CRC cell lines and specimens suggest more complicated mechanisms for Tn expression. Protein glycosylation is spatiotemporally controlled by glycosyltransferases and other enzymes. It is possible that in Tn(+) CRC specimens, the initiating enzymes ppGalNAc-Ts may be abnormally expressed or mislocalized. Several ppGalNAc-Ts including T1, T3, T6, and T13 are reported to be elevated in human cancers [36][37][38][39]. Since these ppGalNAc-Ts have overlapping yet distinct substrate specificities, their up-regulation may increase GalNAc linkage to Ser/Thr residues or generate aberrantly glycosylated proteins at novel or cryptic sites, resulting in unusual conformations that may not be recognized or modified by T-synthase or core 3 synthase (B3GnT6) [40]. For example, it was reported that translocation of ppGalNAc-Ts from Golgi to endoplasmic reticulum (ER) was associated with Fig. 7 Transcriptional expression of ST6GALNAC1 and ST6GALNAC2 in human CRCs. a, ST6GALNA1C mRNA levels. b, ST6GALNAC2 mRNA levels. Relative mRNA levels were calculated using the 2 -ΔΔCt method, with human β-Actin mRNA as the internal control. Paired samples are connected by a solid line. The ST6GALNAC1 mRNA level was significantly higher in the tumors. The significance of the difference between normal and tumor samples (p value) was calculated by paired t test high Tn levels in breast cancer, although the studies did not mechanistically explain the production of cell surface Tn antigen [41]. It is also possible that the in vivo 'functional' activity of T-synthase may be comprised in CRC samples due to mislocalization or enzyme inhibition in situ. Cancer cells often have lost the intact structure of the Golgi apparatus and exhibit an altered pH [42]. T-synthase, even correctly folded, may be mislocalized to aberrant compartments of the Golgi where access to newly synthesized Tn antigen-containing glycoproteins is compromised. Our current enzymatic assay and antibodies cannot assess activity of the T-synthase protein in situ. Furthermore, the access of T-synthase to the Tn antigen could be blocked either by an endogenous inhibitor, as observed for GnT1IP-L towards MGAT1 [43], or by a Tn antigenbinding molecule presented in cancer cells. In addition, Tn antigen expression may be caused by dysregulation of other enzymes, such as core 3 synthase (B3GnT6). B3GnT6 converts the Tn antigen paralleled to the T-synthase/Cosmc complex in the GI tract [12]. It suppresses the metastatic potential and was reported to be down-regulated in colon carcinoma [44]. Mice lacking core 3-derived O-glycans had increased susceptibility to colitis and colorectal tumors [45]. Although we did not observe any mutation of B3GNT6 coding region and its transcript level was undetectable in most CRC samples examined, it is still unknown whether B3GnT6 loses its function in Tn-positive CRCs. It is possible that dysregulation of core 3 synthase could compromise activity of T-synthase, leading to enhanced Tn expression. Finally, the in vivo microenvironment may affect Tn antigen expression. In patients, cancer cells grow in a microenvironment influenced by the oxygen level, cytokines, cellular polarity, and stromal contact. Recent studies showed that cytokine-initiating signaling may regulate the Tn antigen, and hypoxia promotes the STn antigen in bladder cancer [46,47]. In addition, hypoxia also up-regulates the transcription of UGT-1 in CRC cells, which transports UDP-Gal and probably

UDP-GalNAc, suggesting that hypoxia may affect nucleotide sugar pools to modulate glycosylation [48,49]. The difference between in vivo and in vitro environment may explain why most established cancer cell lines do not express the Tn neoantigen. Thus, it may be that a loss of functional T-synthase, through one or more putative pathways discussed above, leads to Tn expression, which further serves as a substrate of ST6GALNAC1 that was elevated in many CRC samples. Some of these hypotheses are currently under investigation. Clinical Features Associated with Frozen Shoulder Syndrome in Parkinson's Disease Background. Frozen shoulder syndrome is a common musculoskeletal disease of idiopathic Parkinson's disease (PD) that causes long-term pain and physical disability. A better understanding of the associated factors can help identify PD patients who will require prevention to improve their quality of life. Methodology. This prospective study evaluated 60 shoulders of 30 PD patients. Correlation analysis was used to evaluate the relationships between clinical factors and shoulder sonography findings. Results. Frozen shoulder syndrome was found in 14 of 30 PD patients affecting 19 shoulders, including bilateral involvement in five and unilateral involvement in nine. There was a significant positive correlation between the parameters of sonography findings and frozen shoulder syndrome (i.e., thickness of bicipital effusion and tendon thickness of the subscapularis and supraspinatus) and mean ipsilateral Unified Parkinson's Disease Rating Scale (UPDRS) III and its subscores (tremor, rigidity, and bradykinesia scores). Conclusions. Higher ipsilateral UPDRS and subscores are associated with increased effusion around the biceps tendon, with increased tendon thickness of subscapularis and supraspinatus. Preventing frozen shoulder syndrome in the high-risk PD group is an important safety issue and highly relevant for their quality of life. Introduction After Alzheimer's disease, Parkinson's disease (PD) is the second most common neurodegenerative disorder [1][2][3][4]. Associated musculoskeletal conditions, seen in as high as 70% of PD patients, are the most common cause of long-term pain and physical disability [5]. Frozen shoulder syndrome, also known as adhesive capsulitis, is characterized by stiffness and pain and is associated with increased fluid in the tendon sheath of the long head of the biceps [6]. It is one of its most disabling musculoskeletal problems severely affecting the quality of life of patients with PD. To date, very few clinical researches have focused specifically on PD patients with frozen shoulder syndrome [7][8][9][10]. Most of these studies are case reports [11] or retrospective studies [7] with less strict selection criteria (e.g., enrolled in combination with PD and multiple system atrophy or frozen shoulder and other shoulder disturbances) [12,13] or depend solely on clinical diagnosis [14]. Sonography enables the visualization of the internal structure of the shoulder tendons and has become an important diagnostic adjunct in the evaluation of shoulder conditions [15,16]. Although sonography and magnetic resonance imaging have comparable high sensitivity and specificity for diagnosing rotator pathologies [15,16], ultrasonography is noninvasive, more widely accessible, portable, and less expensive for assessing soft tissue involvement in musculoskeletal diseases [17]. Because of its possible benefits, an appropriate rehabilitation program should be routinely recommended, with emphasis on regular physical activity to reduce the risk of frozen shoulder syndrome. Nonetheless, a better delineation of potential prognostic factors in PD patients who should receive preventive intervention is warranted. At the same time, it is imperative to obtain accurate information on the relative frequency and severity of frozen shoulder syndrome, the severity and duration of PD, the daily dose of antiparkinsonian agents, and functional outcome. Thus, this hospital-based study aimed to analyze the clinical features, imaging findings, scientific clinical scores, and measurements to determine potential risk factors associated with frozen shoulder syndrome in PD patients. Study Design. This single-center prospective, casecontrol study was conducted at Chang Gung Memorial Hospital, Kaohsiung, a tertiary medical center and the main referral hospital serving a population of 3 million in southern Taiwan. Inclusion Criteria. This study evaluated 60 shoulders of 30 patients with clinical diagnosis of idiopathic Parkinson's disease with a good levodopa response [18], who were followed up at the Neurology Out-Patient Clinic for more than six months after titration of their daily antiparkinsonian agents to a steady dose in accordance with their clinical symptoms. There were 34 patients initially enrolled in this study. Four patients including rotator cuff tear in 2 and shoulder joints dislocation in 2 were excluded. Finally, only 30 cases were enrolled in this study. Seventeen shoulders of nine healthy age-matched patients who did not have PD were studied as the control group. All of the subjects in the control group were asymptomatic for shoulder pathologies and were pain-free. The hospital's Institutional Review Committees on Human Research approved the study protocol and all of the patients or their relatives provided written informed consent. Exclusion Criteria. Patients with preexisting shoulder pathology or other medical history of stroke, rheumatoid arthritis, trauma, dislocation, fracture or surgery, or cervical spine disease were excluded. Clinical Assessment. An experienced neurology nurse specialist who was blinded to the patients' clinical and biochemical data was trained to measure the functional scores upon enrollment. The clinical features recorded were age at disease onset (or age at the time of the first reported symptom attributable to the disease) and disease duration (time from onset until follow-up). The severity of PD was graded according to the scores of the Unified Parkinson's Disease Rating Scale (UPDRS) III and the Hoehn and Yahr staging [19,20]. The daily dose of antiparkinsonian agents was converted into the equivalent dose of levodopa [21]. In patients with fluctuating PD, the UPDRS III and Hoehn and Yahr scales were administered during the "off " situation (8-10 hours after the patient stopped the usual antiparkinsonian treatment) to evaluate the possible influence of disease severity. The Mini-Mental State Examination (MMSE) was used to assess general intellectual function, while cognitive outcomes were assessed using the Clinical Dementia Rating (CDR) scale. All of the subjects were assigned a CDR rating score, as follows: 0 for no dementia and 0.5, 1, 2, and 3 for questionable, mild, moderate, and severe dementia, respectively. All of the patients underwent 99m Tc-TRODAT-1 SPECT/ CT study according to previously published methods [22]. The Hoehn and Yahr disability scale and UPDRS III were used to determine the severity of the disease. Both shoulders were used separately for data analysis. Diagnostic Criteria of Adhesive Capsulitis. In this study, the diagnostic criteria for adhesive capsulitis were the following [23]: (1) insidious onset of pain associated with passive glenohumeral motion; (2) restricted range of glenohumeral motion both actively and passively, with external rotation <50% of the normal side; and (3) normal radiography and a shoulder ultrasound demonstrating no significant rotator cuff tear. Patients were excluded if they had secondary causes of adhesive capsulitis such as a fracture, recent shoulder surgery, calcific tendonitis, or rotator cuff tears; declined to participate; or were younger than 18 years of age. Four patients were excluded for shoulder dislocation in one, rotator cuff tear in two, and another shoulder disorder in one. Only 30 cases were finally included in this study. 99 Tc-TRODAT-SPECT/CT and Region of Interest Analysis. The participants were intravenously injected with a single bolus dose of 99m Tc- Tc-technetium). Four hours after the injection, images were obtained using a dual-head SPECT/CT equipped with low-energy high-resolution collimators (Symbia T, Siemens Medical Solutions, Erlangen, Germany). With the help of anatomic coregistration CT images, the regions of interest (ROIs) of the bilateral striata (including their subregions of caudate and putamen) were defined on composite images of the six highest striatum activity slices. The occipital cortex was drawn in the same way and served as background area. The radioactivity of the regions of interest was then calculated. Striatal dopamine transporter uptake ratios (TRODAT ratio) were calculated as the quotient of the mean counts per pixel in each striatum divided by the mean counts per pixel in the occipital cortex. All images were reviewed by an experienced nuclear physician who was blinded to the patients' data. Sonographic Examination. An experienced neurologist who was skilled in shoulder sonography examinations and blinded to the patients' clinical and biochemical data was trained to measure shoulder sonography studies at the time of enrollment. Ultrasonography of the shoulder was performed using a scanner with a 12/5 MHz linear array transducer (Philips HDI 5000; Philips Medical Systems, Bothell, WA). During the examination, the patient sat on a stool in a comfortable position and the long head of the biceps was examined with the patient's hands placed on the ipsilateral knee. The subscapularis tendon and biceps tendon grooves were examined with the patient seated and the arm held in neutral position, the elbow flexed to 90 ∘ , and the forearm resting supine on the ipsilateral thigh, dynamic with internal and external rotation [24][25][26][27]. The supraspinatus tendon was evaluated in full internal rotation and hyperextension, with the forearm behind the back and the palm facing posteriorly [24][25][26][27]. The infraspinatus tendon was evaluated with the bent arm placed in front of the patient's own chest, while the ipsilateral hand was on the contralateral arm [24][25][26][27]. All of the tendons were examined in transverse and longitudinal scans, and, after adjusting the angle of measurement, the brightest echogenic tendon was obtained. Tendon thickness was measured in sonograms at the point of the largest diameter. An increased hypoechoic area (thickness > 2 mm) around the long head of the biceps tendon in the transverse view was interpreted as effusion in the biceps tendon sheath [27]. Nonvisualization of the tendon or complete fiber discontinuity of the subscapularis, supraspinatus, and infraspinatus tendons was interpreted as full thickness tear [25]. Partial fiber discontinuity and mixed hypo-and hyperechoic focus in the critical zone of the tendon suggested partial thickness tear [26]. Tendonitis was defined as hypoechoic and swelling changes, with a difference in tendon thickness (>2 mm) compared to the healthy side [25]. Statistical Analyses. Data were expressed as mean ± SD or median (interquartile range) for continuous variables. Those with skewed deviation were logarithmically transformed to improve normality before comparison. Two separate statistical analyses were performed. First, parameters of sonography findings among three groups were compared using oneway ANOVA, followed by post hoc multiple comparison. Comparisons between two groups were made using independent -test. Second, correlation analysis was used to explore the relationship between parameters of shoulder sonography findings and clinical severity in PD patients. All statistical analyses were conducted using the SPSS software package, version 12 (SPSS, Inc., Chicago, IL). Results The 30 patients with idiopathic PD included 14 males (age range, 47-78 years; mean age, 66.6 years) and 16 females Values are expressed in mean ± SD unless otherwise indicated. (age range, 47-81 years; mean age, 63.2 years) ( Table 1). Frozen shoulder syndrome was found in 19 shoulders of 14 patients, including bilateral involvement in five and unilateral involvement in nine. Furthermore, the mean age between patients with and without frozen shoulder was 64.9 ± 10.9 and 64.7 ± 10.2, respectively ( = 0.965), and the mean educational levels between patients with and without frozen shoulder were 6.4±3.1 and 4.6±3.2, respectively ( = 0.130). The underlying diseases of the 30 patients included hypertension in 12, diabetes mellitus (DM) in six, and knee osteoarthritis in four (Table 1). Their percentage of cognitive impairment was 46.6%, including CDR 0.5 in 13 and CDR 1 in one. By the Hoehn and Yahr staging, 11 were of stage I, nine were of stage II, eight were of stage III, and two were of stage IV. Their mean UPDRS III score was 27.73 ± 13.9 (range, 5-62). The mean thickness of the tendon in the 19 shoulders with adhesive capsulitis was as follows: subscapularis 5.10 ± 1.2 mm, supraspinatus 5.45±1.2 mm, and infraspinatus 5.52± 1.3 mm. The mean thickness of the bicipital effusion in the 19 frozen shoulders was 2.72±0.5 mm. The ipsilateral UPDRS of PD patients with and without frozen shoulder syndrome was 17.00 ± 7.1 and 14.07 ± 8.8 ( = 0.210), respectively, while their other ipsilateral motor severities were listed in Table 2. Patients with frozen shoulder had a mean MMSE of 25.43±3.3 points, whereas patients without frozen shoulder had a mean MMSE of 23.81 ± 7.7 points. Discussion Differences in the relative prevalence of frozen shoulder syndrome in patients with idiopathic PD vary with case determination and inclusion criteria, disease severity, length of follow-up, underlying diseases, and diagnostic methods. The frequency of frozen

shoulder syndrome in PD patients in other studies varies and is estimated to be as high as 22.6% in one study [7]. In the present study, the frequency is 46.7% (14 out of 30). The present study examines risk factors associated with frozen shoulder syndrome in 30 patients with idiopathic PD and has two major findings. First, both the mean thickness of the subscapularis tendon and the supraspinatus and infraspinatus muscles and the thickness of bicipital effusion in the shoulders are largest in the frozen shoulders of PD patients, followed by the nonfrozen shoulders of PD patients and the shoulders of healthy controls. Second, both the mean thickness of the subscapularis and supraspinatus tendons and the thickness of bicipital effusion in the shoulders positively correlated with the mean ipsilateral UPDRS III and its subscores (i.e., tremor, rigidity, and bradykinesia scores). Frozen shoulder syndrome, also known as adhesive capsulitis, is the spontaneous onset of painful and progressively severe restriction of shoulder joint mobility, in the absence of any demonstrable intrinsic joint abnormality [28]. Although there are no specific ultrasound findings, findings associated with frozen shoulder syndrome may be correlated with increased fluid in the tendon sheath of the long head of the biceps [6]. Two reports have mentioned an association between chronic tendinopathy and the increase in thickness of the rotator cuff through indirect or direct measurements [24,29]. Regarding the sonography findings, there is a significantly thicker bicipital effusion in PD patients with asymptomatic shoulders than in the shoulders of the controls. Furthermore, bicipital effusion is significantly thicker in PD patients with symptomatic shoulders than in those with asymptomatic shoulders. A thicker mean bicipital effusion is observed in patients with significantly higher mean ipsilateral UPDRS III and its subscores for tremor, rigidity, and bradykinesia (scores indicated in advanced PD). This study also demonstrates the effects of ipsilateral UPDRS III and its subscores on the thickness of the supraspinatus and subscapularis tendons and the significantly higher tendon thickness of both the supraspinatus and the subscapularis in asymptomatic PD shoulders compared to those of the controls. The supraspinatus tendon is considered to be the most commonly affected site by degeneration and among rotator cuff age-related increases in thickness [24]. The present study demonstrates that not only the supraspinatus tendon but also the subscapularis tendon has changes associated with the clinical severity of PD, which is represented by ipsilateral UPDRS III and its subscores. One study indicates that the supraspinatus and subscapularis tendons are especially commonly affected and cause an anterosuperior subluxation of the shoulder joint in patients with PD [30]. Another study shows positive correlations between supraspinatus tendon tear and ipsilateral tremor and rigidity scores [9]. Consistent with such findings, results of the present study also reveal a positive correlation between the thickness of the supraspinatus tendon and ipsilateral UPDRS III and its subscores for tremor, rigidity, and bradykinesia. The mean infraspinatus tendon thickness is the only parameter that does not show any significant correlation with clinical motor scores because the thickness of the infraspinatus tendon is relatively difficult to measure due to the lack of anatomic landmarks. Several articles report risk factors for frozen shoulder syndrome, including prolonged illness, rigidity, immobilization, and akinesia [7,8,31]. One study demonstrates that akinesia is common as a first manifestation and is correlated with the laterality of the frozen shoulder [7]. Another study demonstrates that rigidity and bradykinesia in PD patients are supposed to lead to limited joint range of motion and predispose the patient to developing frozen shoulder syndrome [7,31]. The present study demonstrated the relationship between clinical severity and thickness of bicipital effusion. Patients with lower mean MMSE score, which may imply cognitive decline, have thicker bicipital effusion. Several important risk factors for frozen shoulders (e.g., age, sex, occupation, disease duration, and PD phenotypes) should be further discussed. In the present study, lower educational level and disease duration were found in patients without frozen shoulder. However the difference is not significant within our patients. Our study is limited to small patients' number and patients with lower educational level, and thus the occupation and PD phenotypes difference were not studied in the present study. Although this study demonstrates that the mean ipsilateral tremor score in UPDRS III is independently associated with frozen shoulder syndrome in patients with idiopathic PD, it has several limitations. First, PD is a slowly progressive neurodegenerative disorder. Hence, it is not possible to assess the effects of disease progression on the duration and severity of frozen shoulder in a cross-sectional study. Second, as a prospective observation study, it may be subject to bias of unmeasured factors. The frozen shoulders of some patients may be related to accidents or immobilization. Furthermore, underlying cognitive decline may cause potential frozen shoulder during follow-up. Third, PD patients with frozen shoulder syndrome associated with rotator cuff tear have been excluded and this may be subject to bias in statistical analysis. Lastly, the case numbers are small. Large-scale prospective studies to clarify the relationship are warranted. Considering the morbidity and mortality associated with frozen shoulder syndrome, prevention and evaluation of shoulder through sonography should be considered in high-risk patients with PD, especially in those with higher UPDRS III. These are important issues for patient safety and are highly relevant for improving their quality of life. Ethical Approval The study was approved by Chang Gung Memorial Hospital's Institutional Review Committee on Human Research. Building a Named Entity Recognizer in Three Days: Application to Disease Name Recognition in Bulgarian Epicrises We describe experiments with building a rec, ognizer for disease names in Bulgarian clinical epicrises, where both the language and the domain are different from those in mainstream research, which has focused on PubMed arti, cles in English. We show that using a general framework such as GATE and an appropriate pragmatic methodology can yield significant speed up of the manual annotation: we achieve F1=0.81 in just three days. This is the first step towards our ultimate goal: named entity nor, malization with respect to ICD,10. Introduction The problems of named entity recognition and normalization are central to biomedical text processing: as part of the typical preprocessing pipeline, they are key for any deep text analysis. The goal is to identify all mentions of named entities of a particular type, e.g., genes, proteins, diseases, drugs, and to propose a canonical name, or a unique identifier, for each mention. Solving this problem is important for many applications, e.g., enriching databases such as the Protein and Interaction Knowledge Base 1 (PIKB), part of LinkedLifeData, compiling gene-disease-drug search indexes for large document collections in KIM 2 (e.g., using the BioMedicalTagger 3 ) and MEDIE 4 , or building a search engine that can retrieve the effects of a drug on various diseases in research papers and patents. Being so central to biomedical text processing, the problems of named entity recognition and normalization have received a lot of research attention, e.g., there have been several related competitions at BioNLP 5 and BioCreAtIvE 6 . Moreover, high-quality manually annotated biomedical text corpora such as GENIA 7 have been created, which have enabled the development of a number of biomedical text processing tools that need such kind of data for training. Unfortunately, mainstream research has so far focused almost exclusively on English and on biomedical abstracts and full-text articles in PubMed. Thus, biomedical named entity recognition (NER) for languages other than English or for other types of biomedical texts faces the problem of the lack of manual text annotations and biomedical resources in general, which are needed for machine learning. While manually annotating some data is always a good idea, e.g., for analysis, parameter tuning and evaluation, it is hardly practical for more than just a few documents. It is thus important to make smart use of any existing resources and to facilitate the process of manual annotation as much as possible so that good results can be achieved quickly and with very little efforts. The best approach depends on the particular task as well as on the kinds of texts and resources that are available, and there is hardly a universal solution. Still, there are probably lessons to be learned from particular examples of efforts focusing on achieving good performance for NER in new languages and domains in a short period of time. Below we describe our experience with building a recognizer for disease names in Bulgarian clinical epicrisis, where both the language and the domain are different from those in mainstream research. We demonstrate good performance with little efforts and in a very short period of time. We further show how we can save about 57% of the efforts needed for manual annotation of instances of named entities in text. Finally, we discuss how this work can be extended to the task of named entity normalization. The remainder of the paper is organized as follows: Section 2 offers an overview of related work, Section 3 summarizes our methodology, Section 4 describes our experiments and presents the results, Section 5 goes into deeper analysis, Section 6 describes some potential applications, and Section 7 concludes with possible directions for future work. Related Work There have been several research efforts focused on making manual annotations over a text corpus or building a system for named entity recognition and information extraction in a short period of time and with limited resources. proposed a semiautomated approach to named entity annotation where first an n-best MIRA-based named entity recognizer is trained on the initial training set and then tuned for high recall by manipulating the MIRA loss function. Then, its output is checked by a human annotator, who makes yes/no decisions for each proposed entity. Their experiments show that this can speed up manual annotation by about 58% without loss in quality of annotation. We achieve a similar reduction of 57% in the required annotation time using a structured perceptron for named entity tagging; however, we start with no annotated data at all. Settles (2011) described the DUALIST system for semi-supervised annotation based on active learning. The system solicits and learns from labels on both features (e.g., words) and instances (e.g., entities). It has been evaluated on a number of annotation and classification tasks; on named entity recognition, it achieved 0.80 precision (unknown recall and F1), which is a bit lower than our 0.86-0.87 precision. Moreover, being based on active learning, DUALIST needs access to a large number of unlabeled documents from which to choose examples for annotation; such documents are not available in our case. Freedman et al. (2011) presented a bootstrapping system for what they call extreme extraction, where they start with an ontology defining the target concepts and relations they will need to extract and a limited number of training data. They achieve human-level accuracy in a week; this includes five hours of manual rule writing. In fact, our case is arguably more extreme since we start with no annotated data at all. We should also mention the early work done as part of the Surprise Language Exercises held in 2003, where sixteen teams tried to develop language technologies for two previously unanticipated languages, Cebuano and Hindi, in just ten and twenty-nine days, respectively. This work is described in two special issues of the ACM Transactions on Asian Language Processing journal (Oard, 2003). Finally, we should mention the 2007 Computational Medicine Challenge 8 , which focused on analyzing clinical epicrises but for English. However, it asked for assigning ICD-9 codes at the document level, while we want to find instances of IDC-10 disease mentions in text. Moreover, the challenge provided a lot of manually curated data, and thus there was no need to annotate additional data (Crammer et al., 2007). Method We started with a small number of Bulgarian epicrises and a list of diseases from an ontology. First, we analyzed and manually annotated a small number of documents to acquaint ourself with the data and the task and to produce datasets for development and evaluation. Next, we automatically induced contextual rules for finding additional names of diseases. We applied these rules on the development set, we inspected their output, and we incrementally restricted them in several iterations. Once we were satisfied with the precision, we applied the rules to new texts, and we collected their predictions to build a gazetteer of likely disease names.

Next, we applied the gazetteer to some unannotated documents, and we inspected and corrected the matches in context, thus ending up with more annotated documents for training. We then trained a sequence-based named entity recognizer on all training documents we had so far. Finally, we augmented that sequence recognizer with the predictions of the gazetteer as features, thus achieving 0.81 F1 score, which we found sufficient, and we stopped there. Initial Datasets We started with a collection of 100 Bulgarian documents describing clinical epicrises, which we analyzed manually. Based on this analysis, we developed annotation guidelines, which we used to annotate 20 of these documents with the names of diseases from the ICD-10 9 (International Classification of Diseases) ontology. There were a total of 441 disease names mentioned in these 20 documents. Contextual Rule Induction We selected 10 of the annotated documents for development, and used the remaining 10 documents for testing. We automatically learned contextual rules from the development set, which we then used to find named entities in the test set. The rules memorize three tokens to the left and to the right around a potential disease name instance. Here is an example (??? marks the target instance): We do not allow the context words to cross sentence boundaries, and thus the inferred rules can have access to a context of less than three words on either or both sides of ???, as in the following example: the therapy of ??? . Here are some inferred rules extracted from the original text in Bulgarian: Rules with a balanced context of three words on either side proved to be restrictive and very reliable. We indexed the annotated documents in GATE (Cunningham, 2000) so that we could perform fast search for annotations and context words on either side of a disease mention. The rules were implemented in JAPE format 10 : The preceding rule states that if the word "diagnosis" is followed by ":", all following tokens up to and not including "." should be considered part of a disease name. Figure 1 shows the user interface for searching and visualization of the results in the context of three tokens to the left/right of the candidate disease names. Fast searching allows us to find that two tokens on the right hand side and only one on the left hand side could yield better results. Inducing a Disease Gazetteer We executed the rules on the 80 unannotated documents and we created a gazetteer based on the recognized disease names. The resulting gazetteer was evaluated based on its ability to find correct disease mentions in text, on the development set and on the test set. The results are shown in Table 1. R P F1 Dev set 0.61 0.30 0.40 Test set 0.61 0.49 0.54 Table 1: Evaluation of the gazetteer that was induced using the context rules. As Table 1 shows, the low precision is a more important problem for the induced gazetteer than low recall. We thus focused on improving precision by adding rules that could filter out some bad extracted candidates. We first experimented with length-based filters, removing all candidates whose length is less than n symbols; we tried n = 5, 6,7,8,9,10. The results are shown in Table 2. We can see that using length filters significantly increases precision without negatively affecting recall: precision jumps from 0.3 to 0.65, which is higher than recall. As a result, F1 raises from 0.4 to 0.61. Next, we looked closer at the development dataset, and we found that many candidate disease names that started with numbers were in fact false positives. Thus, we tried adding a regular expression like "^[0-9] .*" as an additional filter to the length filter of 8. This improved precision from 0.58 to 0.72, but recall dropped from 0.60 to 0.56, and thus, F1 increased only slightly: from 0.61 to 0.62. We further tried removing candidates containing certain words like menopause and unencumbered, which our manual analysis has found to give rise to many false positives. Doing so led to another increase in precision to 0.87 on the development set while keeping the recall intact at 0.56. The F1 score for the gazetteer after these rules were applied increased from 0.62 to 0.68 (R=0.56, P=0.87) on the development set. The same F1 of 0.68 was also achieved on the test set, but there recall and precision were more balanced: R=0.64, P=0.73. These results are arguably not strong enough for a system that recognizes disease mentions in clinical epicrisis in a fully automated fashion. Still, the resulting gazetteer could potentially be useful for a number of tasks, e.g., for making more manual annotations faster. It could also help robustness since it captures most of the rules that were defined in the annotation guidelines created during the first run of the manual annotation: the 20 documents that we used for development and testing. Thus, we decided to use the gazetteer to help the annotation of 10 more documents. The annotation required about 1.6 minutes per document for the one annotator and 3.6 minutes for another one, or 2.6 minutes on average. As a comparison, in the first annotation run, the average time for annotating a document was about 6 minutes for both annotators. This is a 57% reduction in the time needed to annotate a new document. There was also an improvement in the quality of the manual annotation process. The average agreement between the gazetteer and the manual annotators was F1=0.84 (R=0.88, P=0.80) for the first annotator and 0.77 (R=0.67, P=0.91) for the second one. The inter-annotator agreement was F1=0.78 (R=0.88, P=0.70). This is comparable to the inter-annotator agreement between the two annotator on the first 10 documents where F1 was 0.80 (R=0.96, P=0.69) and to the second 10 where F1 was 0.77 (R=0.91, P=0.67). Training a Structured Perceptron for Disease Name Recognition in Text The resulting 30 documents (20 for training and 10 for testing) were used to train a structured perceptron (Freund and Shapire, 1999). This learning algorithm was selected for simplicity and because of its fast online training. We used a standard set of features that has been initially proposed by McDonald & Pereira (1996), and then successfully adapted to Bulgarian by Georgiev et al. (2009); shown in Table 3. Using this feature set, the perceptron achieved an F1 of 0.69, which is only slightly better than the 0.68 F1 of the rules/gazetteer approach. In order to improve the performance, we annotated 10 more documents by first applying the gazetteer and then doing manual annotations to create System 1, which was trained on 30 documents and tested on 10. We further built System 2, which was trained as System 1, but also used matches with the gazetteer as features. The results for System 1 in Table 4 show that adding more training data (i.e., 30 instead of 20 documents) yields only minor improvement in F1: from 0.69 to 0.71. However, also using features from the gazetteer in System 2 causes F1 to jump to 0.81. As we can see, this is due to a huge improvement in recall, which goes from 0.59 to 0.76, while precision remains stable. We can conclude that the gazetteer turned out to be an important information source, probably because it had analyzed more text (90 documents; all but the testing 10 ones), and thus it could help recall a lot. Table 4: Evaluation of the structured perceptron. System 1 is trained on 30 documents and tested on 10. System 2 is trained like System 1, but it also uses features based on matches with the gazetteer. Discussion The experiments above have shown that manually annotating data and building a system for named entity recognition in clinical epicrises written in Bulgarian is hard for a number of reasons, including but not limited to the following: (i) limited and chaotic general purpose text analysis resources for Bulgarian, (ii) our lack of experience with such texts, (iii) specificity of the domain language, (iv) specificity of the terminology, (v) specificity of the document structure, (vi) issues with extracting the text from the Microsoft Word format the epicrises were stored in. We should note that extracting disease names from epicrises would hardly have been much simpler for English, despite the existence of many biomedical corpora and tools for that language. The main problem here is the domain shift: the existing tools and resources for English are targeting almost exclusively journal papers, whose format, structure and vocabulary differs substantially from those of clinical epicrises. On the positive side, we have shown that even though the task looks complicated, it could be solved with usable F1 in just three days. This speed of building our system would have hardly been possible without our extensive use of the GATE framework for natural language engineering, which has saved us a lot of time and efforts. Among its features that have helped us the most were (i) its ability to extract text from Microsoft Word documents, (ii) its default Unicode tokenizer and (iii) its sentence splitter based on simple regular expressions, which we were able to adopt very quickly, thus overcoming the lack of general purpose text analysis tools and resources for our kind of biomedical text. We were further able to speed up the process of manual annotation by focusing on rules based on words/tokens rather than on part of speech or lemmata (for which we did not have ready tools that could handle the domain well). This was possible because of the particular structure of the documents and the specific language use. For example, clinicians tend to express the diagnosis at the beginning of the epicrisis, typically, in a paragraph that starts with the pattern "Diagnosis:" (or "Диагноза:" in Bulgarian). Here is an example: The diagnosis is followed by few paragraphs explaining why and how the patient was examined, which is further followed by additional information about how the presence of the disease was tested. Of course, a diseases can be extracted from other parts of the document, e.g., such that provide information about the examination of the patient by another specialist. For example, the structured paragraph below contains a list of diseases that have been suggested after a consultation with a neurologist: Консултация с невролог: Начален полиневропатен синдром. Терапия: контрол на кръвната захар. We should note that disease names can be mentioned not only in the diagnosis-related section(s) of an epicrisis, but can occur pretty much anywhere in the document. While catching all instances is generally hard, we were able to do it with the high F1 of 0.81 to a great extent because of the gazetteer. This is because most of the disease names mentioned outside of the diagnoses are likely to repeat those that have been already listed in the diagnosis section. Thus, once the gazetteer has been populated with the somewhat easy-to-extract disease names from the diagnosis-related sections, it can help find further instances of those disease names in contexts that are generally much more ambiguous. We have seen this effect above when comparing System 1 and System 2 in Table 4. Potential Applications As we have seen above, System 2 recognizes disease mentions in text with an F1 of 0.81, which is quite high and is arguably already usable for a number of practical applications. Still, generally, named entity recognition is just the first step in biomedical text analysis; we might also need normalization, which would allow us to get to the canonical names of the diseases mentioned in a particular epicrisis, thus enabling more sophisticated practical applications. For example, if a disease recognizer is coupled with a recognizer of dates and symptoms, we would be able to monitor disease progression and manifestation over time. Normalizing disease names to an identifier or a canonical name in an ontology, would also allow linking a particular clinical epicrisis to a whole web of linked data. One such example would be LinkedLifeData, which is a platform that integrates biomedical information for diseases, symptoms, proteins, genes, drug action information and clinical trials. Linking between an epicrisis and LinkedLifeData might facilitate knowledge acquisition and enrichment and could enable sophisticated queries and rich semantic search over a collection of epicrises. Thus, in order to enable such semantic annotations, we need not only the offsets and type of

each disease mention in a given epicrisis but also a mapping of the mention to a unique identifier. An obvious candidate in our case is the Bulgarian version of the ICD-10 ontology, which provides both unique disease name identifiers and canonical forms that can be used for disease name normalization. Motivated by the practical importance of the task, we did some preliminary assessment of the feasibility of the idea of mapping disease name mentions to identifiers in the ICD-10 ontology. Unfortunately, we found that this was not as simple as we thought initially since the disease names used in ICD-10 strongly disagreed with the names used in our clinical epicrisеs. One reason for this is the tendency of Bulgarian clinicians to describe their diagnoses in Latin. Unfortunately, the Bulgarian ICD-10 does not include disease names in Latin. Moreover, there were many abbreviations, both for Latin and Bulgarian. Thus, for this task, it is important to collect Latin medical terminology from other sources as well as synonyms of diseases for Bulgarian, which can be used to enrich the ICD-10 disease classification. It is worth mentioning that we partially handle this problem by automatically generating a gazetteer of diseases from the source documents. We further found that we needed to remove a number of identifier references from the ICD-10 names. In particular, we filtered out any disease names that looked like codes and were in parentheses, e.g., (J99.8*), (L40.5ї), (М45-М46ї, М48.-ї, М53-М54ї), (Е10-Е14ї с общ четвърти знак .4) and the like. We further filtered out some abbreviations that did not refer to the target disease but to molecular markers or abnormal proteins and other participants that can cause the disease. In order to increase the number of actual disease names for which we could provide ICD-10 identifiers and to create additional synonyms for some of the diseases, we reordered and selectively extracted some names in parentheses from the existing disease names in ICD-10 (rather than the description) of the disease in such cases. For example, we rewrote the two examples above as follows: In order to automatically prepare the corpus for manual annotation of disease mentions, we created a GATE processing tool that implements a number of string distance metrics based on SimMetric, an open source library of similarity and distance metrics, including Levenshtein, L2, Cosine, Jaccard, Jaro-Winkler, etc. SimMetric has a visual interface, which facilitates the selection of the most appropriate similarity measure for a particular task. After some preliminary experiments, we found Jaro-Winkler to be most fit for our data. Based on the score of the distance match between a disease mention in the text and the names in the ICD-10 dictionary, we ordered and select the top-3 names from ICD-10. We then fed these top-3 candidates in a GATE user interface, specially created for the purpose, which is shown on Figure 2. In Figure 2, the text and the diseases are shown in the upper part of the screen, while the disease names from ICD-10 with the top-3 Jaro-Winkler scores are shown in the bottom. A human annotator can delete the candidates that are incorrect in the given context with a single mouse click, e.g., "УЗД", which is a procedure/examination. In some rare cases, the annotator might also need to add new candidates, which can be done with a right click: see the case of "тиреоид на Хашимото", where the correct candidate is "E06.3 Автоимунен тиреоидит" instead of "E06.9 Тиреоидит, неуточнен" that is present in the top 3 candidates. Conclusion and Future Work In this work we focus on simple approaches to named entity recognition having limited or no prior example data. In this framework, we have demonstrated that a seemingly complicated named entity recognition task can be handled with satisfying quality in a fast and robust manner. We have achieved this by examining the structure and language expressions, as well as words and orthographic features found in clinical epicrises. We have further demonstrated that using general purpose frameworks such as GATE and an appropriate pragmatic methodology can significantly speed up the process of annotation. Our disease mentions recognizer, annotation guidelines and annotated epicrises are potentially useful for applications such as document categorization and search. Moreover, extending the disease mention recognition to semantic annotations with identifiers from an ontology such as ICD-10 would enable a number of applications such as monitoring disease progression and manifestation over a period of time and linking epicrises to a web of linked data like LinkedLifeData. We believe these are promising research directions and we plan to pursue them in future work. For the purpose of facilitating and speeding up manual semantic annotations, we have developed a new GATE-based processing tool that can calculate string similarity scores between disease mentions found in text and disease names listed in ICD-10. We have further coupled this with a GATE visual resource that allows a human annotator to delete wrong mentions at a given offset, thus only leaving the correct option(s), while also allowing the addition of more options. In future work, we plan to use this interface in a similar pragmatic approach to the task of normalizing disease names in context with respect to ICD-10. Localization of stomatal lineage proteins reveals contrasting planar polarity patterns in Arabidopsis cotyledons Many plant cells exhibit polarity, revealed by asymmetric localization of specific proteins within each cell. 1–6 Polarity is typically coordinated between cells across a tissue, raising the question of how coordination is achieved. One hypothesis is that mechanical stresses provide cues. 7 This idea gains support from experiments in which cotyledons were mechanically stretched transversely to their midline. 8 These previously published results showed that without applied tension, the stomatal lineage cell polarity marker, BREVIS RADIX-LIKE 2 (BRXL2), exhibited no significant excess in the transverse orientation. By contrast, 7 h after stretching, BRXL2 polarity distribution exhibited transverse excess, aligned with the stretch direction. These stretching experiments involved statistical comparisons between snapshots of stretched and unstretched cotyledons, with differentspecimensbeing imagedineachcase. 8 Here, weimage the samecotyledon beforeandafter stretching and find no evidence for reorientation of polarity. Instead, statistical analysis shows that cotyledons contain a In brief Fozard, Yu, Bezodis et al. reveal that cellular localization of stomatal polarity proteins is biased in cotyledons toward medial positions in stomatal lineages and toward proximal positions in nonstomatal lineages. They find no evidence that mechanical stretching reorients polarity, in contrast to a previous report. RESULTS AND DISCUSSION Stomatal development provides a model system for understanding the control of asymmetric division and patterning in plants. 9 Several stomatal proteins are localized in a polarized and dynamic manner during stomatal lineage development. 2,4,10,11 Tissue-wide mechanical forces have been proposed to play a role in controlling this localization pattern. 8 Following cell ablation, BREVIS RADIX-LIKE 2 (BRXL2) polarity protein becomes oriented toward the ablation site, suggesting that gradients in wall stresses generated by ablation may provide an orienting cue. An alternative explanation is that ablation generates a molecular wound signal that orients BRLXL2 toward the ablation site. More direct evidence for the role of mechanical stresses comes from experiments in which BRXL2 polarity was quantified in cotyledons mechanically stretched transverse to their midvein. 8 Cotyledons were chosen for this experiment because they show little evidence of tissue-wide polarity for BRXL2, in contrast to rosette leaves which show significant proximal bias (localization at the cell end nearer the leaf base). Cotyledons of live seedlings can also be readily stuck down on stretchable membranes. The previously published results showed that in unstretched controls, BRXL2 polarity exhibited no significant excess in the transverse orientation, whereas in stretched cotyledons, BRXL2 polarity became significantly aligned transversely, parallel to the stretch orientation. 8 This observation raises the question of how BRXL2 polarity in individual cells changes following stretching. Sequential imaging reveals no significant stretchinduced change in BRXL2 localization To answer this question, we developed a rig that allowed the same cotyledons to be imaged before and after stretching (Figure 1A). In addition to the BRXL2pro:BRXL2-YFP marker, we (B) Quantified deformation caused by stretching a cotyledon. Deformation calculated from nonlinear registration of cotyledons before and immediately after stretching. Black lines indicate local directions of maximum deformation. Color scale indicates stretch anisotropy x = x 2 y 1 =x 1 y 2 , where x 1 and x 2 are the lengths of a small region along the direction of maximum deformation before and after stretching respectively, and y 1 and y 2 are the lengths perpendicular to this direction for the same region. Deformation field computed using a cubic B-spline with 8 intervals in each direction, which may not resolve cellular-scale details. (C) Cotyledon before stretching. Cell outline (red) and BRXL2 (green) signal from projected confocal stacks. Inset shows magnified image of boxed region. Circles labeled (CI), (CII), and (CIII) indicate the cells in (F). (D) Same cotyledon as (C) immediately after stretching. (E) Same cotyledon as (C) imaged following 7 h stretching and released from the membrane. (F) Measurement of the angle b between the BRXL2 polarity vector (from the center of the signal to the cell centroid) and the transverse axis (perpendicular to the cotyledon midline). Left images show the cotyledon before stretching (as in C), right images show the cell after a 7 h stretch followed by release from the membrane (as in E). Db was calculated by subtracting b 1 before stretching from b 2 after a 7 h stretch. (legend continued on next page) ll OPEN ACCESS crossed in a mCherry membrane-marker to identify cell outlines. Before stretching ( Figure 1C), polarized BRXL2 signal on the abaxial surface could be seen in about 12% of cotyledon cells (1,701/14,535, 13 cotyledons from 9 seedlings 3-4 days old), reflecting the subset of cells undergoing stomatal lineage divisions. We then glued the cotyledons to an elastomer membrane (m in Figure 1A), which was stretched by 60% in a direction perpendicular to the cotyledon midline, as previously described. 8 As in the previous study, a 60% membrane stretch led to a more elliptical shape in the stretch direction ( Figure 1D). We calculated the deformation for each region of the cotyledon by nonlinear registration of images before and after stretching 12 and found that the degree of deformation anisotropy varied from over the cotyledon, indicating variable adhesion of the glue (Figure 1B). Previous experiments also described a much lower level of cotyledon deformation compared to the membrane. 8 Cotyledons were kept in the stretched condition for 7 h, as previously described, and then imaged under a coverslip after release from the stretch ( Figure 1E). A decrease in width/length ratio of the cotyledon upon stretch release was checked to ensure that the cotyledon remained under tension for 7 h. We analyzed 13 cotyledons from 9 individuals, before and after 7 h stretching. As controls, we also analyzed 10 cotyledons from 7 individuals that had not been stretched during the 7 h period. BRXL2 signal exhibits a highly dynamic pattern in stomatal lineages, appearing 5 ± 2.5 h before an asymmetric division and persisting for 9 ± 2.5 h afterward. 11 For cells that could be readily tracked, we detected BRXL2 in 14% of cotyledon cells before stretching (1,353/9,784 tracked cells). In 21% (281/1,353) of these cells, signal was still detected in the cell after the 7 h stretch, allowing shifts in BRXL2 polarity to be measured directly (individual cell images provided in Data S1 and S2). To quantify orientation, we drew arrows from the center of BRXL2 signal to the cell centroid (corrected by cell segmentation) ( Figure 1F). We then measured the magnitude of the angle b between the BRXL2 polarity vector and the line transverse to the midline (parallel to the stretch orientation). This angle varied from 0 (transverse to the midline) to 90 (parallel to the midline). We measured the value of b before ðb 1 Þ and after the stretch ðb 2 Þ and calculated the difference Db = b 2 À b 1 . Polarity reorientation in the direction of stretching should produce a decrease in b (i.e., mean Db should be negative). We found Db had a unimodal distribution that varied from about À45 to +45 , indicating fluctuations in BRXL2 localization and/or arrow positioning ( Figure 1G). Mean Db was +0.5 (SD 25.3 , n = 281; Figure 1G) and not significantly smaller

than zero (p = 0.62, one-sample one-sided t test). Controls that had not been stretched gave a similar distribution of Db ( Figure 1H). Thus, there was no evidence from direct tracking that BRXL2 polarity shifted to align with the orientation of stretching. Unstretched cotyledons exhibit transverse excess in BRXL2 localization Given our inability to detect changes in BRXL2 polarity through direct tracking, we made statistical comparisons between unstretched and stretched cotyledons to check whether our results replicated the previously described stretch-induced changes in polarity. 8 This previous study analyzed a total of 3,571 cells covering four treatments (i.e., about 900 cells per treatment, if all treatments had a similar sample size). The treatments comprised one unstretched control and three different stretch intensities (20%, 40%, and 60%), with 10 or more individuals per treatment. Measurements were made of the angle a between BRXL2 cell polarity and the midline vector of the cotyledon. Angles were classified as transverse (|a| = 80 -100 , parallel to the stretch orientation) or non-transverse (all other angles). For a uniform distribution, 20/180 = 11.1% of cells should exhibit transverse angles. The previous study showed that unstretched controls and cotyledons stretched by 20% had a mean of about 10%-11% transverse angles, consistent with a random distribution; whereas cotyledons stretched by 40% or 60% had means of 17% or 18%-19% transverse angles, significantly higher than expected from a random distribution. We first analyzed our unstretched controls. To reduce sampling effects, we analyzed 6,834 cells taken from 68 cotyledons from 61 individuals. Combining all data revealed a peak for |a| of around 80 -130 with a mode at 100 -110 ( Figure 2A). In contrast to the previous study, there was a significant excess of transverse orientations in these unstretched controls (13.7% transverse angles, one-sided binomial test p = 2.4 3 10 À11 ). The distributions of |a| varied considerably between cotyledons ( Figure 2B), demonstrating the importance of having a large sample to establish overall trends. We next analyzed BRXL2 polarity in stretched cotyledons (13 cotyledons from 9 individuals; Figures 2C and 2D). The distribution was not significantly different from the unstretched cotyledons (Kolmogorov-Smirnov test, D = 0.0378, p = 0.11, m = 6,834, n = 1,175). However, unlike the unstretched cotyledons, the stretched cotyledons showed no significant transverse excess (11.8% transverse angles, p = 0.23). To check whether this difference might be caused by the lower number of individuals analyzed in the stretched group (13 cotyledons) compared to the unstretched control (68 cotyledons), we randomly sampled 20 groups of 13 cotyledons from the unstretched controls. This sampling produced sample sizes of 1,023-1,411 cells, similar to the number of cells sampled in the previous study for each treatment. The resulting distributions varied, with 16/20 showing significant transverse excess ( Figures 2E, 2F, and S1). Thus, the previously reported differences between stretched and unstretched cotyledons 8 could have been caused by sampling too few cotyledons. (G) Histogram of Db for cells expressing BRXL2 both at the start and the end of the experiment (n = 281 cells from 13 cotyledons from 9 different seedlings). p = 0.62 from a one-sided t test for the distribution mean being less than zero. Blue line shows mean of the distribution (m = 0.5 ), dotted blue lines 95% confidence intervals of the mean from Student's t distribution (À2.5 < m < 3.4 ). Red line at zero. (H) Histogram of Db for cells expressing BRXL2 both at the start and the end of the control (no stretch) experiment (n = 294 cells from 10 cotyledons from 7 different seedlings). Blue line shows mean of the distribution (m = 2.3 ), dotted blue lines 95% confidence intervals of the mean from Student's t distribution (À0.6 < m < 5.3 ). p = 0.94 from a one-sided t test for the distribution mean being less than zero. (B)-(E) and insets: scale bars, 100 mm. (F) Scale bars, 25 mm: See also Figure S3 and Data S1 and S2. Report To check for variation caused by sampling cotyledons at different stages, control cotyledons were divided into three groups according to width. All groups had a significant transverse excess ( Figure S2). Medium-width cotyledons (550-650 mm) had a significantly different distribution of |a| compared to other groups, though this variation may reflect the small sample sizes per group (19-25 cotyledons per group). Transverse excess in BRXL2 localization reflects a weak medial bias in stomatal lineages To investigate the basis of the transverse excess in control cotyledons, we plotted the distribution of a separately for the left and right halves of the cotyledons. Positive values of a indicated polarity vectors pointing leftward and negative value polarity vectors pointing rightward. Left cotyledon halves had an a peak at about +110 whereas right halves had a peak at about À100 and differences between left and right were highly significant ( Figure 2G; p = 2.2 3 10 À39 , chi-squared for cells with positive versus negative a). Thus, transversely oriented BRXL2 polarities tend to point divergently from the midline, reflecting medial bias of BRXL2 localization. To quantify the spatial distribution of polarity, we translated, rotated, and scaled control cotyledons to a common template. We then superimposed a rectangular grid of 15 regions over the registered cotyledons and plotted the distributions of a within each region using rose histograms ( Figure 3A). For grid regions that showed significant difference between negative and positive a (p < 0.05), we plotted the modal orientation of polarity (within the half of the distribution with excess polarity vectors) with a white arrow, and we also indicated those grid regions that showed significant transverse excess with a yellow asterisk. This analysis showed that polarity diverged from the midline, particularly in marginal and lateral positions ( Figure 3A). Most of the white arrows pointed slightly downward, as expected from the peak of the distributions in Figure 2G, which were at about 100 rather than 90 . The mediolateral polarity pattern was not evident from inspection of single cotyledons ( Figure 3B), indicating medial bias is weak and can only be detected through analysis of many samples. Thus, unstretched cotyledons exhibit a coarse-grain mediolateral polarity field. so is the midline vector. Data from 68 unstretched cotyledons from 61 different seedlings. Red bar along the x axis indicates the range 80 -100 of transverse angles used for the significance tests. p = 2.40 3 10 À11 from a one-sided binomial test indicates an excess of the proportion of angles within the transverse range (r) over that expected by chance (11%). (B) Stacked heatmaps of jaj for each cotyledon, in order of increasing cotyledon width (number on left-hand side, in microns). Red crosses indicate the modal histogram bin for each cotyledon. (C) Histogram of jaj after stretching for 13 stretched cotyledons from 9 plants (data from these cotyledons before stretching is included in A and B). p = 2.29 3 10 À1 from a one-sided binomial test as in (A). Polarity bias in cotyledons varies between stomatal and non-stomatal lineages BRXL2 colocalizes with another stomatal polarity protein, BREAKING OF ASYMMETRY IN THE STOMATAL LINEAGE (BASL). 11 To determine whether BASL polarity also exhibits mediolateral bias in stomatal lineages of cotyledons, we imaged the distribution of BASLpro:BASL-GFP in 17 unstretched cotyledons from 17 different plants. The distribution of|a|showed a transverse excess but was significantly different to the BRXL2 distribution (compare Figure 2H to Figure 2A). Relative to BRXL2, there was a deficit of low values of|a|, indicative of a slight distal bias, and the peak was shifted slightly from 100 -110 to about 120 -140 . The difference between BRXL2 and BASL distributions may reflect the observation that BRXL2 polarizes earlier than BASL and stays polarized for longer. 11 As with BRXL2, there was a significant difference in a between left and right cotyledon halves for BASL, indicative of medial bias ( Figure 2K; p = 8.7 3 10 À5 ). The left halves had a peak at about +130 whereas right halves had a peak at about À110 . Grid rose plots indicated that this distribution reflected a coarsegrain mediolateral polarity, though the sample size within each grid was smaller than that used for BRXL2 ( Figure 4A). As with BRXL2, mediolateral polarity was not evident from inspection of a single cotyledon ( Figure 4B). Thus, in cotyledons both BRXL2 and BASL localization exhibit a weak but significant bias toward medial locations, giving a polarity that points away from the midline. The medial bias for BRXL2 and BASL in cotyledons contrasts with the situation in rosette leaves where both BRXL2 and BASL exhibit a proximal bias. 8,13 The proximal bias in rosette leaves is stronger when BASL is expressed outside stomatal lineages. 13 To determine the polarity pattern for ectopic BASL in cotyledons, we quantified a in 35S:BASL-GFP. Polarity was mainly derived from non-stomatal cells, both because these were in excess Figure 2J). The distribution was almost complementary to that of stomatal BASL (compare Figure 2J with Figure 2H), indicative of orthogonal polarity patterns. The proximodistal polarity field for ectopic BASL was evident both at the coarse-grain level ( Figure 4C) and from inspection of cells from individual cotyledons (Figure 4D), indicating a stronger tissue-wide bias compared to stomatal lineages. Polarity pointed slightly divergently from the midline at the base of the cotyledon, accounting for the slight but significant difference in the a peak between left and right halves of the cotyledon ( Figure 2L). Ectopic BASL persisted in a greater fraction of cells than endogenous BRXL2, allowing the change in orientation, Db, to be measured for most (400/504) tracked cells initially expressing BASL, after stretching for 7 h. For ectopic BASL, Db was not significantly different from 0 ( Figure 4E; p = 0.09 for one-sample two-sided t test, 95% confidence interval for mean change À0.15 < Db < 1.94), showing that stretching had no significant effect on ectopic BASL polarity orientation. Origin of polarity patterns Cotyledons exhibit two types of bias in BASL/BRXL2 polarity markers: strong proximal bias in non-stomatal lineages and weak medial bias in stomatal lineages. In rosette leaves, both markers exhibit proximal bias, though this is weak in stomatal lineages. 8,13 One explanation for the weaker bias within stomatal lineages is that two polarizing mechanisms compete in these cell types. Division patterns in primary stomatal lineages depend on a polarity switching mechanism, involving BASL. 10 Polarity switching generates diverse polarity orientations that would tend to disrupt tissue-wide polarity, accounting for the weak biases observed in stomatal lineages. However, it is unclear why the weak bias in stomatal lineages is proximal in leaves and medial in cotyledons. Although the weak biases observed in stomatal lineages may confer little or no selective advantage for stomatal spacing, the underlying polarity systems that generate such biases may play fundamental roles in patterning and orienting other processes, such as growth. In the case of cotyledons, the coarse-grain mediolateral polarity in stomatal lineages is approximately orthogonal to the polarity in non-stomatal lineages. Orthogonal tissue cell polarity fields have been described for markers such as PIN and SOSEKI proteins, 3,5 although these are apical-basal versus radial rather than operating within the same epidermal plane. Orthogonal planar polarity patterns have also been described during development of the Drosophila wing, where core PCP polarity proteins reorient to align with the proximodistal axis, while Fat system polarity proteins remain aligned with the anteroposterior (mediolateral) axis. 14 How might the orthogonal polarities observed in cotyledons be coordinated? One hypothesis is that chemical signals provide cues, [15][16][17][18] with one set of molecules propagating between cotyledon base and tip to coordinate proximodistal polarity, and another set between midline and cotyledon margin to coordinate mediolateral polarity. A variant of this hypothesis is that chemical signals establish one polarity field (e.g., proximodistal) and the second is defined orthogonal to this. However, this mechanism would require a system to ensure that orthogonal polarity is oriented oppositely for the left and right halves of the cotyledon. It is also possible that stress gradients provide cues. 7 If so, they must be robust to externally applied tissue-wide stresses, as both the mediolateral and proximodistal patterns are unperturbed by stretching. Further studies are needed to determine the underlying cues guiding orthogonal polarities and how they influence polarity coordination. STAR+METHODS Detailed

methods are provided in the online version of this paper and include the following: ACKNOWLEDGMENTS We would like to thank Martyn Hewett for help design and modify the stretching rig, Dominique Bergmann for helpful discussions and providing plant lines, the JIC Bioimaging facility for help with confocal microscopy, and Horticultural services for looking after Arabidopsis plants. RESOURCE AVAILABILITY Lead contact Further information and requests for plant lines, constructs and raw data should be directed to and will be fulfilled by the lead contact, Enrico Coen ([email protected]). Materials availability This study did not generate new unique reagents. Data and code availability d Microscopy data reported in this paper have been deposited in osf.io and are publicly available as of the date of publication. DOIs are listed in the key resources table. d All original code has been deposited at Zenodo and is publicly available as of the date of publication. DOIs are listed in the key resources table. d Any additional information required to reanalyze the data reported in this paper is available from the lead contact on request. EXPERIMENTAL MODEL AND SUBJECT DETAILS Experiments were performed using Arabidopsis thaliana (Col-0 background) plants expressing BRXL2pro:BRXL2-YFP 8 or 35Spro:GFP-BASL 2 crossed with plants expressing PM-mCherry (PM-rb 19 ). F2 And F3 seedlings were used for experiments. The Arabidopsis plants containing BASLpro:GFP-BASL together with mCherry marking the plasma membrane, and those containing 35Spro:GFP-BASL, were from Dong et al. 2 Growth conditions Plants were grown on plates containing 1 % (w/v) Agar, 0.43 % (w/v) Murashige & Skoog powdered medium including vitamins, 3 mM MES, PH 5.7. Seeds were stratified at 4 C for 2 days and then germinated in a growth room set at 20 C. The growth room was REAGENT METHOD DETAILS Preparation of elastomer membrane Elastomer membrane was prepared by mixing 45 mL of SYLGARD 182 silicone solutions and 5 ml of SYLGARD 182 Silicone Elastomer Curing agent (Ellsworth Adhesives LTD; 4019601). 10 ml of elastomer solution was poured into square petri dishes and allowed to set for 3 days. The membrane was then cut into 3 cm x 1.8 cm strips using a razor blade. Stretching cotyledons Stretch experiments were carried out using a modified calliper. The calliper contained clasps on the jaws for fixing the elastomer strip, and a screw for opening/closing and locking the calliper. The calliper was mounted onto a base that fitted onto the stage of the microscope. A hole in the base allowed light to be transmitted through the specimen. Seedlings expressing either BRXL2pro:BRXL2-YFP and PM-mCherry or 35Spro:GFP-BASL and PM-mCherry were glued on to the elastomer membrane using medical adhesive (Adapt 7730; 7730). This was done by dabbing a spot of glue onto the centre of an elastomer strip. A seedling was then placed upside-down with the adaxial surfaces of cotyledons touching the glue. The proximal-distal axes of the cotyledons were oriented at 90 degrees relative to the long axis of the elastomer strip. The cotyledons were then gently pushed against the glue using a cotton swab. The glue set after at least 3 min. Damp strips of filter paper were then placed over the hypocotyl/roots to prevent the seedlings from dehydrating. Individuals that were damaged during this process were discarded. The elastomer membrane with glued seedling was then attached to the modified calliper (see Figure 1A). The calliper was opened until the membrane became straight. The distance between the tips of the calliper was then measured and the stretch applied by opening the calliper by an appropriate amount to stretch the membrane by 60 % stretch. The jaws were then locked tight. A piece of damp filter paper was placed on the base beneath the calliper and a block of agar (cut from the plate used to grow the seedlings) was placed on top the roots. The calliper was wrapped in cling film or sealed in a large square petri dish with damp filter paper and placed in the growth room over the period of stretch. For each experiment, 5 images of the cotyledons were used: (1) before stretch (x20 lens), (2) immediately upon stretch (x10 lens), (3) after 7 hours of stretch prior to release (x10 lens), (4) after release of stretch before removal from calliper (x10 lens), (5) after release of 7 hour stretch (x20 lens). First images were taken before seedlings were glued on to membrane strips. Second, third and fourth images were taken by placing the calliper in the microscope. For the final image, seedlings were incubated in a few drops of distilled water for 5 min and, when possible, gently removed from the stretched membrane using forceps. If the cotyledons remained firmly attached, the whole membrane was removed from the calliper for imaging. First and final images were taken using a x20 lens and after mounting seedlings under a cover slip with water. These were used for segmentation and analysis. Other images were taken without coverslips using the x10 lens and were just used to ensure successful stretch and that the stretch held for 7 hours. Note that, for a subset of the data, the second and fourth images were not acquired. Additional images of cotyledons that had not undergone attachment to membrane strips and stretching, imaged under a cover slip using x20 lens, were used to supply the data used in Figures 2B-2F, 2H, and 3. Microscopy Cotyledons were imaged with a Zeiss 780 exciter confocal laser scanning microscope equipped with EC PLAN-NEOFLUAR x10 (NA of 0.3) and Plan-APOCHROMAT x20 (NA of 0.8) objective lenses or a Leica SP8X confocal laser scanning microscope equipped with a PL APO x20 (NA of 0.75) objective lens. GFP and YFP were excited at 488 nm using an Argon ion laser (Zeiss 780) or a Pulsed White Light Laser (Leica SP8) and the emission collected at 490-530 nm. mCherry was excited at 561 nm and the emission collected between 590-640 nm. Confocal sections were collected at a z-spacing of 1-4 mm for x20 images and 5-15 mm for x10 images, with sufficient z depth to capture the width of the cotyledons for subsequent measurement. To remove autofluorescence, images collected using the Pulsed White light laser were gated so that the first 0.2 ns of emission were not collected. GFP and mCherry were collected sequentially using the Zeiss 780 and simultaneously using the Leica SP8X confocal microscope. Image processing Initial image processing was performed using the Fiji distribution 21 of the ImageJ software (https://imagej.nih.gov/ij/). 22 Details of the custom image analysis pipeline used are detailed below. Figure panels were combined using Inkscape. QUANTIFICATION AND STATISTICAL ANALYSIS Confocal images of cotyledons were processed using a custom pipeline, described below. Two different procedures were used -one where cells were not tracked before and after stretching, used to generate the data for the histograms in Figure 2, and the spatial data (iii) Each cell in the T1 segmentation was linked to the cell which it maximally overlapped in the deformed T3 segmentation, provided this is more than 50% of the area of the cell. This captured events where two cells merge (through under-segmentation at T1). (iv) The links from (ii) and (iii) formed a bipartite graph (a graph in which each edge connects members of two different, disjoint sets). The connected components (which we refer to as "tracks") of this bipartite graph were extracted (by breadth-first search on the graph), and classified according to how many cells they contained at T1 and T3. (v) Components containing one cell at T1 and one or two cells at T3 were retained, and identified as successfully tracked cells. This was the majority of the dataset. Components not containing one cell at T1 and one or two cells at T3 were discarded. (vi) At this point, each track contained one cell at T1, and one or two cells at T3. Each cell was classified according to whether it had a marker arrow (i.e. if one of the marker arrows ends within the segmented region associated with this cell). For those tracks that divide, and where the cell at T1 has a marker arrow, the two daughter cells were ordered such that the cell whose centroid (in the deformed T3 segmentation) was nearest the marker arrow base (before stretching) appeared first. If the cell at T1 had no marker arrow, the two daughter cells were sorted such that the one with the marker arrow appeared first. (vii) This allowed each track to be classified as belonging to one of 11 distinct categories, shown in detail in Figure S3J. [1,1]), where the cell underwent division, the arrow end was adjusted to be at the centroid of the region occupied by both of the two daughter cells after stretching. (h) To test the hypothesis that b was decreased by stretching, a one-sample one-sided t-test using the R function "t.test 25 " was performed, with the alternative hypothesis that the true mean is less than zero (alternative=''less''), accessed via the RPy Python bridge. Confidence intervals on the mean value of Db were provided by a one-sample two-sided t-test from the same R function. For ectopic BASL, a two-sided t-test was performed, with the alternative hypothesis that the true mean is non-zero, again using ''t.test'' Analysis of spatial distribution of polarity (Figures 3 and 4) (a) As for the histograms of angles for untracked cells, the stack was projected to a two-dimensional image, the cells were segmented, arrows were manually added to indicate marker polarity, arrow ends were adjusted to be at the centroids of the segmented cells, arrows where the distance between the original and the adjusted end was more than 1.5 times the original length of the arrow, or where the start point was more than 5 pixels from the cell boundary were discarded, and the angles a between the polarity arrows and the positive x-axis were measured. (b) Angles a for each arrow, were calculated by subtracting the midline angle q from a, and remapping the result to the range À 180 + < a À q % 180 + . (c) Each cotyledon was manually aligned with a template through rotation, translation and uniform rescaling ( Figure S3H). (d) A 3x5 rectangular grid was placed over the region occupied by the cotyledon template. Each arrow was assigned to the rectangular grid element containing the arrow base. (e) Histograms of the angle a were generated for each grid element. (Note that this used the manually determined midline angle q, rather than the rotation angle of the alignment, for consistency with the other histograms in the manuscript.) (f) Statistical significance tests were applied to the distribution of arrows within each grid element. For BRXL2, two tests were performed for each grid element. Firstly, whether there was an excess of leftwards ða > 0 + Þ or rightwards ða % 0 + Þ pointing arrows was examined using a two-sided binomial test. Secondly, whether there were more medio-laterally oriented arrows Transplantation of induced pluripotent stem cell-derived renal stem cells improved acute kidney injury Background Acute kidney injury (AKI) is a severe disease with high morbidity and mortality. Methods that promote repair of the injured kidney have been extensively investigated. Cell-based therapy with mesenchymal stem cells or renal progenitor cells (RPCs) resident in the kidney has appeared to be an effective strategy for the treatment of AKI. Embryonic stem cells or induced pluripotent stem cells (iPSCs) are also utilized for AKI recovery. However, the therapeutic effect of iPSC-derived RPCs for AKI has yet to be determined. Methods In this study, we induced iPSCs differentiation into RPCs using a nephrogenic cocktail of factors combined with the renal epithelial cell growth medium. We then established the rat ischemia–reperfusion injury (IR) model and transplanted the iPSC-derived RPCs into the injured rats in combination with the hydrogel. Next, we examined the renal function-related markers and renal histology to assess the therapeutic effect of the injected cells. Moreover, we investigated the mechanism by which iPSC-derived RPCs affect AKI caused by IR. Results We showed that the differentiation efficiency of iPSCs to RPCs increased when cultured with renal epithelial cell growth medium after stimulation with a nephrogenic cocktail of factors. The transplantation of iPSC-derived RPCs decreased the levels of biomarkers indicative of renal injury and attenuated the necrosis and apoptosis of renal tissues, but resulted in the up-regulation of

renal tubules formation, cell proliferation, and the expression of pro-renal factors. Conclusion Our results revealed that iPSC-derived RPCs can protect AKI rat from renal function impairment and severe tubular injury by up-regulating the renal tubules formation, promoting cell proliferation, reducing apoptosis, and regulating the microenvironment in the injured kidney. Electronic supplementary material The online version of this article (doi:10.1186/s13578-015-0040-z) contains supplementary material, which is available to authorized users. Background Acute kidney injury (AKI) is characterized by acute tubular injury and a rapid decline in renal function [1]. Renal ischemia is one of the most common causes of AKI. AKI has recently received increasing attention due to the high mortality and morbidity rates of the syndrome [2]. Although pharmacological therapy, modern dialysis techniques, and intermittent or continuous renal replacement therapies exist, the therapeutic approaches for AKI remain very limited. Thus, the establishment of novel therapeutic strategies for ischemic AKI is urgently needed. In response to acute injury, the adult kidney shows some level of regeneration characterized by the proliferation of the surviving cells and the replacement of the necrotic tubular cells with functional tubular epithelium [3]. The process is hypothesized to involve epithelial cell dedifferentiation, interstitial cell transdifferentiation and/or activation of quiescent stem cells. Over the last few years, renal progenitor cells (RPCs) have been isolated from the kidney [4][5][6][7]. Several studies have indicated that RPCs appear to play important roles in kidney repair under various pathological conditions [4,[8][9][10]. However, progenitor cells in the adult kidney are rare, which limits the application of these cells. Therefore, we resorted to other types of stem cells. Stem cells, including embryonic stem cells (ESCs) and adult stem cells, have the potential of self-renewal and differentiation into multiple cell types and therefore hold great promise in regenerative medicine [11][12][13]. It was recently reported that stem cells exhibit therapeutic potential for AKI [14][15][16][17]. Several studies have demonstrated that the exposure of ESCs to factors required for renal specification, such as retinoic acid, activin A and bone morphogenic proteins (BMPs), does induce their differentiation into renal lineage cells in vitro [18][19][20]. Vigneau et al. [10] showed that the transplantation of mouse ESCs-derived RPCs can result in the stable integration into proximal tubules with normal morphology and normal polarization injection into developing live newborn mouse kidneys, suggesting the potential of ESCs for application in regenerative therapies. However, the use of ESCs for human therapy raises several safety and ethical concerns. Induced pluripotent stem cells (iPSCs) are a type of newly defined stem cell with properties similar to those of ESCs in terms of self-renewal and differentiation. These cells were first generated by Yamanaka and colleagues [21]. Since their discovery, iPSCs have shown their advantages in regenerative medicine. iPSCs have also been reported to be effective for AKI. Lee and colleagues [16] showed that the injection of iPSCs into the kidney improves renal function and attenuates renal tubular injury in AKI caused by ischemia-reperfusion injury (IR) and eventually improves the survival rate of rats with AKI. Although iPSCs are an alternative potential cell source for stem cell therapy to protect against AKI, there is also a risk of tumor formation. Therefore, we investigated whether iPSC-derived RPCs have a therapeutic effect on AKI. In the present study, we examined whether iPSCderived RPCs can protect against AKI in a rate model of IR. We showed that the differentiation efficiency of iPSCs to RPCs increased when cultured with Renal Epithelial Cell Growth Medium after stimulation with a nephrogenic cocktail of factors. The injection of iPSC-derived RPCs combined with hydrogel, which provides a 3-D structure to the cells, into the kidney of rats with IR could improve renal function and attenuate tubular injury by forming renal tubules in the recipient rat kidney directly, promoting cell proliferation, reducing apoptosis, and regulating the microenvironment in the injured kidney. Efficient differentiation of RPCs from mouse iPSCs To investigate whether iPSC-derived RPCs can improve AKI, we first used a nephrogenic cocktail of factors (including retinoic acid, activin A, and BMP7) to induce iPSCs differentiation into RPCs (Additional file 1: Figure S1). The iPS cell line used in this study is tagged with a ubiquitously expressed gfp gene. The real-time PCR analysis showed that the expression of the RPC markers PAX2, WT1, Six2 and CD24 increased after RAB treatment (Fig. 1a), indicating that RPCs were successfully induced. As shown by flow cytometry analysis, the PAX2, WT1 and CD24 positive cells were 33. 16,19.56 and 27.82 % respectively (Fig. 1b). Additionally, the mature renal markers AQP1 Fig. 1 Efficient differentiation of renal progenitor cells from mouse induced pluripotent stem cells (iPSCs). The renal differentiation of mouse iPSCs was initiated by the formation of embryoid bodies (EBs). After 2 days, retinoic acid, activin A, and BMP7 were added to the differentiation medium, and the cells were exposed to this supplemented media for 5 days (RAB group). After incubation with growth factors for 5 days, the EBs in the RAB + REGM group were cultured in renal epithelial cell growth medium for another 5 days. a Real-time PCR analysis of the expression of mesoderm (Bry) and renal lineage genes (Pax2, Osr1, WT1, Six2, CD24, Sal1, AQP1, E-cad and PDGFR). b Flow cytometry analysis of the expression of mesoderm renal progenitor-related genes. Black line, cells stained with isotype control; Red line, cells stained with indicated antibodies. *P < 0.05 and E-cadherin were up-regulated upon RAB stimulation (Fig. 1a). To improve the differentiation efficiency of iPSCs into RPCs, the cells were cultured in renal epithelial cell growth medium for another 5 days (Additional file 1: Figure S1). As expected, the expression of the RPCs markers was further up-regulated compared with that of the RAB group, as shown by real-time PCR analysis after 5 days of differentiation in RAB medium (Fig. 1a). Moreover, the PAX2, WT1, and CD24 single-positive cells increased to 52.56, 27.8 and 36.24 % respectively (Fig. 1b), indicating that more RPCs were generated in the RAB + REGM group. Therapeutic effect of iPSC-derived RPCs in rats with IR-induced AKI Next, the rat IR model was generated by occlusion of the left renal artery followed by reperfusion and right kidney removal. iPSC-derived RPCs from the RAB + REGM group were injected directly into the renal parenchyma of the injured rats in combination with the hydrogel. The control group (IR-control) was injected with PBS combined with the hydrogel. The levels of BUN and Scr were detected at the indicated times. The plasma levels of BUN and Scr reached their peak levels at day 1 of renal IR in all groups and then decreased gradually (Fig. 2a, b). After RPC transplantation, the plasma levels of BUN and Scr were lower than those of the IR-control group (Fig. 2a, b). To test whether the RPC transplantation has beneficial effects on regeneration after IR, the rats were sacrificed on days 3, 7, 14, and 28, and the renal histology was examined. H&E staining showed that the proximal tubule in the renal cortex in the control group exhibited acute injury and necrosis 3 days after surgery ( Fig. 3a): c The immunohistochemistry analysis indicated that the injected GFP-positive cells formed a tubular structure (black arrow) in the recipient organ at the indicated times after transplantation. *P < 0.05 vs. IR-Control. Scale bar 50 μm the brush border of some epithelial cells disappeared; the lumen expanded; some of the tubular epithelial cells detached from the basement membrane; the tubule was obstructed with cellular debris; and the interstitial tubule was infiltrated with inflammatory cells. The histological features of necrotic injury were still severe 7 days after ischemia and began to improve on days 14 and 28. The injury phenotype was decreased in the rat kidney as a result of RPC transplantation. The quantitative analysis of the renal tubular necrosis using the grading scores of Paller et al. [22] is shown in Fig. 3b. These results suggest that the transplantation of iPSC-derived RPCs could improve renal regeneration and function after AKI. Mechanistic study of the improvement in the renal tubular damage caused by IR as a result of iPSC-derived RPC transplantation We further investigated the mechanism through which RPCs relieve acute renal injury caused by IR. Previous studies have reported that mouse kidney progenitor cells accelerate renal regeneration after ischemic injury by differentiation into epithelial cells and incorporating into the renal tubule [9,[23][24][25]. Thus, we first examined the localization of the transplanted iPSC-derived RPCs. Because the iPS cell line used in our study carries a ubiquitously expressed gfp gene, it is convenient to trace the injected cells. Using immunohistochemistry, we showed that GFP-positive cells began to form a tubular structure in the recipient kidney 7 days after injection (Fig. 3c). Importantly, no tumor was detected even 3 months after transplantation. Next, the effects of the RPCs on the proliferation and apoptosis of cells in the injured kidney were examined by PCNA staining and TUNEL analyses, respectively. The results showed that the number of PCNA-positive cells increased in the transplanted group compared with that found in the control group on days 7 after injection (Fig. 4a, b), indicating the presence of a greater number of proliferating renal tubular epithelial cells in the kidney after RPC transplantation. In contrast, the cell apoptosis on days 7 was reduced in the RPC-transplanted group, as shown by TUNEL analysis (Fig. 4c, d). Furthermore, we detected the expression levels of the anti-inflammatory factors interleukin-10 (IL-10), basic fibroblast growth factor (bFGF), and transforming growth factor β1 (TGF-β1) and growth factors promoting renal tubular cell repair, namely epidermal growth factor (EGF), hepatocyte growth factor (HGF), and platelet-derived growth factor (PDGF). The real-time PCR analysis showed that the expression of these genes was upregulated to different extents after RPC transplantation (Fig. 5). Altogether, these data indicate that iPSCderived RPC transplantation improves renal function after IR by directly incorporating into the renal tubule, promoting cell proliferation, reducing apoptosis, and upregulating the expression of pro-renal factors. Discussion AKI is an emerging worldwide public health problem with high mortality and morbidity rates and has thus attracted increasing attention. Although traditional therapies show benefits in the treatment of AKI, the therapeutic approaches are still very limited. Recently, stem cell-based therapy appeared to be a potential strategy for the prevention of AKI [26]. In the present study, we provide the first demonstration that the transplantation of iPSC-derived RPCs improves renal regeneration and function in a rat model of IR-induced AKI, and the protective effect of the RPCs was mediated by their direct differentiation into renal lineage cells and incorporation into the renal tubule, which resulted in the promotion of cell proliferation and the reduction of apoptosis in the damaged zone. Furthermore, the transplantation of iPSC-derived RPCs promoted the expression of antiinflammatory factors and pro-renal tubular repair factors, suggesting that RPCs can regulate the microenvironment and facilitate cell survival in the injured kidney. After ischemic injury, RPCs and mesenchymal stem cells (MSCs) resident in the kidney are activated and show some level of regenerative ability. Several studies have shown the therapeutic potential of isolated primary RPCs or MSCs in experimental AKI models. However, these cells are not available in large volumes. Recent advances demonstrated that iPSCs may be more readily accessible for kidney regeneration [16]. Therefore, in our study, iPSCs were differentiated into renal lineage cells to obtain a sufficient number of RPCs. To the traditional induction medium, we added renal specific growth medium for further induction. Using this modified protocol, the differentiation efficiency of iPSCs into RPCs and mature renal epithelial cells increased. Because kidney regeneration requires many functionally distinct cell types of the nephron, the transplantation of all renal cell types derived from iPSCs may offer the advantage of generating more than one cell type of the adult kidney. In previous studies, stem cells for the treatment of AKI were mainly injected through the tail vein or the intrarenal artery [9,16], and the transplanted stem cells in the circulation are mobilized to the injury site in the kidney mediated by chemokines/cytokines and receptor interaction and thereafter enhance renal regeneration. However, the homing efficiency cannot be determined. In our study, RPCs were injected directly into the renal parenchyma in combination with a hydrogel. A hydrogel consists of a network of polymer chains, and the function and structure of a hydrogel are similar to those of a natural extracellular matrix. Therefore, hydrogels are

currently widely used as scaffolds in tissue engineering [27][28][29][30]. The hydrogel used in our work is composed of polypeptides and can be degraded into natural amino acids in vivo; thus, the hydrogel has no hazardous effect on the injured tissue. After being injected into the renal parenchyma with the RPCs, the liquid hydrogel will quickly transform into a solid structure, which provides a three-dimensional microenvironment for RPC adhesion, migration, proliferation and differentiation. Our results showed that the exogenous cells formed a tubular structure in the injured kidney, indicating that the RPCs could migrate out of the hydrogel and participate in renal regeneration. The therapeutic mechanisms of primary RPCs for AKI were recently investigated in several studies [9,23,31]. It has been reported that the transplanted RPCs can differentiate into renal tubule cells and vessel endothelial cells with the expression of E-cadherin and CD34 [23] and therefore accelerate renal regeneration and prolong survival after AKI. Our result that iPSC-derived RPCs can differentiate into renal tubule cells and form renal tubules in the injured kidney further supports this hypothesis. Moreover, we showed that the transplantation of iPSCderived RPCs promoted cell proliferation and decreased apoptotic events. The cells also exhibited an anti-inflammatory effect by increasing the expression of anti-inflammatory factors, including TGF-β1. A recent study [32] reported that the presence of TGF-β1 in the homogenate of acute IR-injured renal tissue markedly increases the CXCR4 surface expression of MSCs and promotes the migration of MSCs to SDF-1. Therefore, we hypothesized that the transplantation of RPCs increased the level of TGF-β1 in the injured zone, which in turn regulated the CXCR4 expression of the RPCs, contributing to the migration of the RPCs out of the hydrogel to the injured sites. However, this proposed mechanism requires further investigation. Collectively, the present study demonstrated that the transplantation of iPSC-derived RPCs is able to protect against kidney injury in an experimental IR rat model. These cells function in both differentiation-dependent and -independent manners by directly forming renal tubules and exerting pro-proliferative, anti-apoptotic, and anti-inflammatory effects. Our data suggest that iPSC-derived RPCs are a potential cellular resource for stem cell-based therapy for IR-induced AKI. Conclusions This study reports a modified method for the differentiation of iPSCs into RPCs and provides experimental evidence that iPSC-derived RPCs combined with a hydrogel can improve renal function in AKI and reduce tubular injury in an animal IR model. We believe that iPSCderived RPCs hold great potential to be used as seed cells for personalized treatment in regenerative medicine. IPSCs culture Undifferentiated iPSCs were routinely cultured and expanded on mitotically inactivated MEFs (50,000 cells/ cm 2 ) in six-well culture plates with mES medium: DMEM supplemented with 15 % FBS, l-glutamine, nonessential amino acids, β-mercaptoethanol (Gibco), and 1000 U/ml LIF (Millipore). The iPS cell line used in the present study was purchased from SiDanSai Biotechnology (Shanghai, China). The sequence of gfp gene, for tracing the iPSCs and iPSC-derived cells, was inserted into the chromosome of iPSCs. iPSCs transduced with this sequence showed robust expression of GFP. Renal differentiation of murine iPSCs iPSCs were differentiated through embryoid body (EB) formation in EB medium (DMEM supplemented with 15 % FBS, l-glutamine, nonessential amino acids, and β-mercaptoethanol). On day 3 of differentiation, the EBs were transferred to gelatin-coated culture dishes at a density of 10 EBs per 1 ml of medium. The cells in the control group were maintained in EB medium for 5 days, and the medium was changed every 2 days. The cells in the induction group were exposed to a nephrogenic cocktail of factors, including 0.1 μM RA (Sigma), 10 ng/ml activin-A (PeproTech), and 50 ng/ml BMP7 (PeproTech), in the EB medium for 5 days. The induced cells were then cultured in the EB medium alone (RAB group) or in Renal Epithelial Cell Growth Medium (RAB + REGM group) for another 5 days and harvested for further analyses. Quantitative reverse-transcriptase polymerase chain reaction (RT-PCR) The total RNA from cells or tissues was extracted using the TRIzol reagent according to the protocol delineated by the manufacturer (Invitrogen-Life Technologies). One microgram of total RNA was reverse transcribed into cDNA with oligo-dT primers and reverse transcriptase (Invitrogen-Life Technologies). Real-time quantitative PCR was performed using SYBR Green (Toyobo Co, Japan). The primers used in the amplification reaction are shown in Table 1. Preparation of hydrogel and cell mixture for injection Hydrogels belong to a class of biological materials and produce a nanoscale environment similar that of the native extracellular matrix [27][28][29][30]. A hydrogel consists of 1 % self-assembling peptides (w/v) and 99 % water (w/v). The detailed constituents and properties and the detailed preparation of the hydrogel used in this study are detailed by the manufacturer (Beaver Nano Technologies, China). Briefly, the hydrogel (1 %) was sonicated for 30 min before use. The cells were suspended with 50 µl of 10 % sucrose solution, and 50 µl of the cell suspension was rapidly and carefully mixed with 50 µl of 1 % hydrogel. Then, 100 μl of PBS (Corning) was carefully added to the top of the cell and hydrogel mixture for instantaneous gelation. The 1 % hydrogel (50 µl) diluted with 150 μl of PBS was used as a control. Rat renal IR model and injection of RPCs The renal IR model was generated in 8-week-old male Sprague-Dawley as described previously [16]. Briefly, thirty rats (3-3.5 kg) were anesthetized with chloral hydrate (0.2 mg/g, Sigma), and this step was followed by exposure of the abdomen. For induction of total ischemia in the kidney, the left renal artery was clamped with a small vascular clamp for 45 min. Reperfusion was initiated by removal of the clamp. The right kidney was removed simultaneously. Thirty-four rats survived the IR surgery and no rats died in the following experiments. Then, the hydrogel (200 μl; denoted "IR-control" group) or the suspension of iPSC-derived RPCs (1 × 10 5 cells in 200 μl of 0.25 % hydrogel per rat; denoted "IR-transplantation" group) was respectively injected into the renal parenchyma of rats (seventeen rats per group). The rats were injected with the immunosuppressant cyclosporine A (Sigma) at a dose of 5 mg/kg per day. Six sham-operated animals underwent similar operative procedures but without clamping of the left renal artery and right kidney nephrectomy. At the indicated time, the blood was collected to measure the levels of blood urea nitrogen (BUN) and serum creatinine (Scr). The rats were killed at the indicated days for the collection of kidney tissue samples. All care and handling of animals was performed with the approval of the Ethical Review Board of Shanghai Six People's Hospital Affiliated to Shanghai Jiaotong University. The animals were raised in a specific pathogen-free (SPF) environment. Histological analysis and tubular injury score For the histological analysis, the kidney tissues were fixed with 4 % paraformaldehyde overnight and embedded in paraffin. The paraffin sections (4 μm) were stained with hematoxylin and eosin (H&E, Sigma). The quantitative analysis of the renal tubular injury used the grading scores proposed in a previous study [22]. The sections were evaluated by assessing 10 randomly selected high-power fields (40 × objective), and higher scores represented more severe damage: tubular epithelial cell flattening (1 point), brush border loss (1 point), cell membrane bleb formation (1 or 2 points), interstitial edema (1 point), cytoplasmic vacuolization (1 point), cell necrosis (1 or 2 points), and tubular lumen obstruction (1 or 2 points). Immunohistochemistry Immunohistochemical staining was performed with mouse anti-GFP antibody (1:100; Cell Signaling Technology) and mouse anti-proliferating cell nuclear antigen (PCNA, 1:200; Abcam). Briefly, paraffin sections of the kidneys were deparaffinized with xylene and rehydrated in an alcohol series and water. The kidney sections were subjected to antigen retrieval and blocked with a peroxidase-blocking reagent. The sections were incubated with the primary antibody overnight at 4 °C. After washing, the kidney sections were incubated with horseradish peroxidase-conjugated secondary antibody (ZSJQ-BIO, China) for 1 h at room temperature. The sections were visualized with 3,3′-diaminobenzidine (DAB, ZSGQ-BIO) and counterstained with hematoxylin. The number of positive cells was evaluated by counting the stained cells per high-power field in at least 20 randomly selected fields. The apoptotic cells in the kidney were detected by terminal deoxynucleotidyl transferase (TdT)-mediated digoxigenin deoxyuridine nick-end labeling (TUNEL, Promega) staining according to the protocol delineated by the manufacturer. The TUNEL-negative controls were stained without TdT enzyme. Statistical analysis All of the experiments were performed at least three times. The data are shown as the mean ± SEM. Differences were analyzed using one-way analysis of variance (ANOVA). Values of P < 0.05 were considered statistically significant. Treating an intramuscular abscess following toothpick injury in a diabetic patient Abstract Rationale: Toothpick puncture (TPP) is a penetrating injury that can result in bringing pathogens to the deep space. Such penetrating wounds are typically of pinpoint size with initial symptoms appearing subtle. Consequently, the injury itself is often neglected by patients, or is not detected during physical examinations by medical doctors. Reported complications from such injuries include osteomyelitis and septic arthritis, mostly due to delayed treatment. Patient concerns: A diabetic patient aged 83-year-old presented a 2-day history of skin redness, swelling, and tenderness over his forearm following a TPP a week earlier. Laboratory investigations showed leukocytosis with neutrophilic predominance and a high level of C-reactive protein. Before his operation, cultures of aspirated fluid from the injured site revealed the presence of Streptococcus anginosus, Streptococci viridans, Prevotella intermedia, and Pavimonas (Peptostreptococcus) micra. Diagnosis: Intramuscular abscess associated with toothpick injury. Interventions: Surgical irrigation with debridement and adjunctive antibiotics of ceftriaxone and clindamycin were given with a satisfactory response. Cultures of debrided tissue showed the presence of P intermedia and P (Peptostreptococcus) micra. Outcomes: A split-thickness skin graft was done. Patient was discharged on the 30th postoperative day. Lessons: Toothpick injury, initial symptoms of which are subtle, can in some cases, lead to serious complications especially when managements are delayed. In such situations (including the present case), surgical irrigation and debridement are administrated for the eradication of infections, removal of potentially retained toothpick, and tissue cultures analyzed. Adjunctive antibiotics is recommended to combat both the aerobic and anaerobic microorganisms of the gastrointestinal tract, skin surface, and oral cavity. Introduction A 1-year survey in the United States reported 8176 toothpickrelated injuries (TRIs) in which 76% involved extremities and trunk. [1] Toothpick punctures (TPPs) can lead to infections from polymicrobial oral flora, in mixed infections by both aerobic and anaerobic bacteria. [2] TRIs are usually neglected by the affected subjects, because the wounds are seemingly small and often innocuous with subtle initial symptoms. Such injuries, especially Editor: N/A. Datasets used and/or analyzed in this study are available from the corresponding author on request. All procedures in this study were performed in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. This study was approved by the Institutional Review Board of Taichung Veterans General Hospital (No. CE18102A). A written informed consent was obtained from the patient to the use of his history and all the related images and information for scientific purposes. The authors have no conflicts of interest to disclose. with delayed management, can lead to serious complications. We reported here a case of a diabetic patient presented with a deep space infection (intramuscular abscess, IMA) after inflicting injury from a TPP. The patient recovered well after our treatments with surgical irrigation and debridement with adjunctive antibiotics, and finally with split-thickness skin grafts. Case report A male patient of 83-year-old had a type 2 diabetes mellitus (DM) with previous diagnosed (>20 years ago) complications of nephropathy and retinopathy, primary hypertension, and dyslipidemia. He came to our emergency department (ED) recently with a 2-day history of erythema, swelling, and tenderness felt over the left forearm. A week earlier, his left distal forearm was reportedly punctured by a used toothpick. At the ED, he complained of subjective fever and chills as well as severe forearm pain and swelling, which prompted his visit to our hospital. Vital signs at presentation were the following: body temperature 38.0°C, blood pressure 144/74 mm Hg, pulse rate 77/minute, and respiratory rate 18/minute. Physical examination revealed a tender lesion over the left distal forearm with erythematous changes, swelling, and local heat. However, we found no signs of an open wound, bullae

formation, or hemorrhagic changes. On palpitation, subcutaneous liquid accumulation was noted over the area of his left distal forearm. Aspirating the accumulated subcutaneous fluid revealed a straw-color liquid, a sample of which was sent for culture study. Empirical antibiotics with ampicillin/sulbactam were first administered intravenously. A complete blood count revealed 12,710 leukocytes/mm 3 with neutrophilic predominance. The level of C-reactive protein was 23.38 mg/dL. Serum electrolytes were normal, but his blood level of glucose was 321 mg/dL and hemoglobin A1c 8.7 mmol/mol. Serum creatinine level was 3.49 mg/dL. Urinalysis results were normal. Culture of the aspirated fluid sample showed the presence of polymicrobial organisms, including Streptococcus anginosus, Streptococci viridans, Prevotella intermedia, and Pavimonas (Peptostreptococcus) micra. On the 3rd day of admission, the erythematous changes progressed over his left forearm with increasing tenderness. Necrotizing fasciitis was highly suspected. We shifted the choice of antibiotics to ceftriaxone and clindamycin. Due to his condition of chronic kidney disease, computed tomographic scan with contrast media was not performed. He was taken to the operating room for surgical debridement. Wound cultures confirmed Prevotella intermedia and P (Peptostreptococcus) micra. IMA of the left distal forearm was diagnosed during the operation (Fig. 1). After operation, wound dressing with vacuumassisted closure was given with antibiotics maintained throughout his hospital stay. The formation of granulated tissues over his wound bed was observed on postoperative day 21. In order to cover the affected area of his left forearm, we performed a splitthickness skin graft, which was harvested from his left anterior thigh. His condition subsequently improved, and the patient was finally discharged on postoperative day 30. Discussion Searching with the key word "toothpick" on PubMed, we found a total of 325 citations, within which 8 (1 case series, 7 case reports) are on soft tissue infections of the extremities following injuries from TPPs. [2][3][4][5][6][7][8][9] Table 1 shows the results of the 15 relevant cases. Of these cases, the commonest infection is deep space infection, which developed into conditions like soft tissue abscess (n = 6), [2,[4][5][6] pyogenic tenosynovitis (n = 5) (2 cases complicated with abscess formation), [2] osteomyelitis (n = 3), [7][8][9] and septic arthritis (complicated with abscess formation) (n = 1). [2] Only 2 cases were diagnosed of superficial infection and cellulitis (n = 2). [2,3] Generally speaking, TPP is a penetrating injury and may have the potential to bring pathogens to deep space (i.e., bone, tendon, and joint space). Injuries extended beyond the subcutaneous tissues are also of high risks of infection. [10] In addition, the small penetrating wound from the toothpick is typically of pinpoint size and the initial symptoms are subtle, so the wound is often neglected by the affected subjects themselves, or undetected during physical examinations. The published cases often are characterized by complications including osteomyelitis and septic arthritis arising from delayed treatments. Intervals between injury and hospital admission lie between 7 and 40 days (average 17.6 days). [2] Of the 15 cases shown in Table 1 (including a case series by Chang et al), [2] the interval between injury and admission is between 2 days and 4 months, with a majority of them (n = 12) exceeding 1 week (Table 1). [2][3][4][5][6][7][8][9] All 13 patients published with deep space infections caused by TPPs received surgical interventions. [2,[4][5][6][7][8][9] Surgical irrigation and Table 1 Literature reports of infection caused by toothpick penetration into limbs (7 case reports and 1 case series). debridement in conjunction with antibiotics are typically mandatory for deep space infections (e.g., abscess, necrotic soft tissue, tenosynovitis, septic arthritis, and osteomyelitis). Surgical debridement is generally warranted in the case of toothpick retention, because the retained foreign object of puncture is also a risk for infection, in line with reports on lacerating injuries. [10] Surgical debridement allows the collection of tissue and fluid specimens for microbiological evaluation to identify the causative pathogens. As a general rule, since the penetrating wound from toothpick is minute, very limited amount of tissue can be collected for pathogen cultures. It is worth to point out that findings from superficial cultures of the wounds should be interpreted with caution, since cultures of superficial tissues do not always correlate in results with those of the deeper tissues. [11][12][13] Pathogens are identified in tissue cultures collected during surgical irrigation and debridement in 12 of the 13 published cases on deep space infections. [2,[4][5][6][7][8][9] In conclusion, surgical intervention should be considered for infections caused by TPPs. As to the best time to apply the surgical intervention, studies published are too limited to provide a good recommendation. [14] TPPs can be associated with infections by polymicrobial oral flora, with mixed aerobic and anaerobic infections reaching up to 71.4%. [2] Of the 15 cases found in our literature search (including 8 cases by Chang et al, [2] most of them (n = 8) have polymicrobial flora infections, [2,[4][5][6]8] and the anaerobic microorganisms, which are mostly oral flora, are identified in 11 cases. [2][3][4][5][6][7][8][9] Eikenella corrodens is the most common anaerobic microorganism (n = 6). E corrodens is a commensal of the human mouth and upper respiratory tract. With invasive infections of E corrodens, polymicrobial infections occur in 65% of the patients. Head and neck are the commonest infectious sites (56%) and nearly 2/3 of the patients (63%) have preexisting diseases, especially malignancy of head and neck. [15] The disease condition also leads to infections in patients with insulin-dependent DM and intravenous drug users who might lick their needles ("needlelicker's osteomyelitis"). [16] In some cultures (n = 6), anaerobic pathogens coexist with aerobic microorganisms such as Escherichia coli, Streptococcus group D, and Enterobacter agglomerans (which are common flora in gastrointestinal tract), Streptococcus viridians and Streptococcus constellatus (which are common oral flora), Staphylococcus epidermidis and Staphylococcus aureus (which are common flora on the skin surface). [2,5,8] Generally speaking, toothpicks can have serious consequences similar to those following human bites. Infections of human bites are associated with alpha-hemolytic Streptococci, S aureus, E corrodens, Haemophilus species, and anaerobic bacteria (as in over half of the cases). [17] Drugs, not able to suppress E corrodens, should be avoided. Such agents include the first-generation cephalosporins, macrolides, clindamycin, and aminoglycosides. Therefore, the treatment with amoxicillin/clavulanate, ampicillin/ sulbactam, or ertapenem is recommended. If patients also have a history of hypersensitivity to b-lactams, the prescription of a fluoroquinolone (such as ciprofloxacin or levofloxacin) plus metronidazole, or moxifloxacin as a single agent is recommended. [18] In patients with DM, the known mechanisms of defective immune defense include impaired tissue perfusion, chemotaxis, phagocytosis, and decreased bactericidal activity of neutrophils. Such weakness immunity results in sepsis and severe infections, like IMA. [19,20] To treat infections caused by TRIs, coverage of both aerobic and anaerobic pathogens (including E corrodens) is recommended since we found in our patient, infectious pathogens similar to those found with human bites. Conclusion TRIs may have subtle initial symptoms which are easily neglected due to the seemingly small innocuous wound. Nonetheless it can lead to serious complications in delayed managements. Deep space infections should be considered. Therefore, surgical irrigation and debridement are the choice of treatment, in addition to collecting specimens for reliable cultures, and the removal of potentially retained toothpick. Adjunctive antibiotics should be considered to cover both aerobic and anaerobic microorganisms, including those floras at the gastrointestinal tract, skin surface, and oral cavity. Characterization of Sialic Acid Affinity of the Binding Domain of Mistletoe Lectin Isoform One Sialic acid (Sia) is considered as one of the most important biomolecules of life since its derivatives and terminal orientations on cell membranes and macromolecules play a major role in many biological and pathological processes. To date, there is only a limited number of active molecules that can selectively bind to Sia and this limitation has made the study of this glycan challenging. The lectin superfamily is a well-known family of glycan binding proteins, which encompasses many strong glycan binding peptides with diverse glycan affinities. Mistletoe lectin (ML) is considered one of the most active members of lectin family which was initially classified in early studies as a galactose binding lectin; more recent studies have suggested that the peptide can also actively bind to Sia. However, the details with respect to Sia binding of ML and the domain responsible for this binding are left unanswered because no comprehensive studies have been instigated. In this study, we sought to identify the binding domain responsible for the sialic acid affinity of mistletoe lectin isoform I (MLI) in comparison to the binding activity of elderberry lectin isoform I (SNA), which has long been identified as a potent Sia binding lectin. In order to execute this, we performed computational carbohydrate-protein docking for MLB and SNA with Neu5Ac and β-Galactose. We further analyzed the coding sequence of both lectins and identified their glycan binding domains, which were later cloned upstream and downstream to green fluorescent protein (GFP) and expressed in Escherichia coli (E. coli). Finally, the glycan affinity of the expressed fusion proteins was assessed by using different biochemical and cell-based assays and the Sia binding domains were identified. Introduction The plant kingdom is a rich of source of glycan binding peptides with vast diversity spread throughout the kingdom and the occurrence of a specific peptide is not often restricted to a specific plant family. Lectins are one of the most abundant family of glycan binding peptides observed in almost all plant families which can actively bind to carbohydrates through different types of non-covalent interactions, such as hydrophobic interactions, hydrogen bonds and metal coordination. Furthermore, the existence of a large number of hydroxyl groups on the surface of glycans makes them an appropriate binding partner for lectins, especially in the complex networks of hydrogen bonds where the hydroxyl groups function as both donor and acceptor [1,2]. Whilst the multivalent nature of lectins probably provides them with ultimate selectivity, their specificity to a primary monosaccharide is a major factor in their carbohydrate recognition and it is usually an indicator for a particular branched carbohydrate to be recognized by a specific lectin. Thus, complex carbohydrates bearing glucose (Glc) or mannose (Man) are bound by the Glc/Man-specific lectins (e.g., Lathyrus ockrus and Con A) but galactose (Gal)-specific lectins (e.g., Erythrina corallodendron, EcorL) require the presence of Gal to perform this binding [3,4]. Lectins can also agglutinate cells because of their Protein Sequence Analysis The alignment of full sequences of MLI and SNA against the sequences of previously known glycan binding lectins helped to identify the exact position of the main toxic A and glycan binding B domains of MLI and SNA. Numbers of glycan binding subdomains were also identified among the main glycan binding domains as shown in Figure 1. Virtual Identification of the Glycan Binding Pockets The virtual Glycan-GBP docking was used as a first step to identify the potential Neu5Ac binding pocket on MLB. In order to conduct this, the structure of both Neu5Ac and β-Galactose (Gal) was docked against the binding domain of both MLI (MLB) and SNA (SNA-B). The detected binding poses with the highest affinity of 5.0 and 5.4 kcal/mol in the docking of MLB with Neu5Ac and Gal, respectively, showed that both glycans bind to the same pocket on MLB but with a short distance between their binding sites ( Figure 2a). Several binding interactions were detected between MLB and Neu5Ac as follows: the oxygen atoms on the carboxyl group of Neu5AC were observed to interact with the backbone NH of Ile 93 and side chain indole NH of Trp 94 through an ionic bond; on the other side of the glycan, the O7 and O8 atoms H-bonded to the side chain COO of Asp 2 and Asp 211, respectively; the O9 atoms H-bonded to the side chain carbonyl of Asn 35. However, the binding of Gal to MLB appeared only through H-bonding of O3 and O6 atoms to the side chain carbonyl of Asn 36. The binding of SNA to Neu5Ac and β-Galactose was observed to be different than the binding of MLB as both glycans attached to the exact site on the exact pocket of SNA with the affinity of 6.2 and 5.8 kcal/mol for SNA + Neu5Ac and SNA + Gal, respectively ( Figure 2b). The binding of both glycans was found to be solely though H-bonding to Asp 44, Ser 45 and Asp 47. The carboxylic group of Asp 44

was found H-bonded to O5 and O9 atoms of Neu5Ac and O3 and O4 atoms of Gal. The side chain OH of Ser 45 was found Hbonded to O4 and O3 atoms of Neu5Ac and Gal, respectively, but its backbone CO was found H-bonded to O2 and O4 atoms of Neu5Ac and Gal, respectively. Finally, the backbone NH of Asp 47 was found H-bonded to O9 atom of Neu5Ac but its backbone CO was found H-bonded to O6 atom of Gal. These results suggest that the sialic acid and galactose binding sites on MLB are located on two different areas of the same pocket, which allows the protein to bind to both glycans simultaneously. However, the binding of both glycans to the same site of the same pocket on SNA-B suggests that the protein has ability to bind to both glycans but the higher affinity of SNA toward Neu5Ac may increase the chance of binding with Nue5Ac more than galactose. Virtual Identification of the Glycan Binding Pockets The virtual Glycan-GBP docking was used as a first step to identify the potential Neu5Ac binding pocket on MLB. In order to conduct this, the structure of both Neu5Ac and β-Galactose (Gal) was docked against the binding domain of both MLI (MLB) and SNA (SNA-B). The detected binding poses with the highest affinity of 5.0 and 5.4 kcal/mol in the docking of MLB with Neu5Ac and Gal, respectively, showed that both glycans bind to the same pocket on MLB but with a short distance between their binding sites ( Figure 2a). Several binding interactions were detected between MLB and Neu5Ac as follows: the oxygen atoms on the carboxyl group of Neu5AC were observed to interact with the backbone NH of Ile 93 and side chain indole NH of Trp 94 through an ionic bond; on the other side of the glycan, the O7 and O8 atoms H-bonded to the side chain COO of Asp 2 and Asp 211, respectively; the O9 atoms H-bonded to the side chain carbonyl of Asn 35. However, the binding of Gal to MLB appeared only through H-bonding of O3 and O6 atoms to the side chain carbonyl of Asn 36. PCR Amplification of MLA and MLB The sequences of the toxic chain (MLA) and the binding chain (MLB) of mistletoe lectin, 762 bp and 789 bp, respectively, were successfully amplified using cDNA templates reverse transcribed from the total RNA of European mistletoe. The extracted RNA was visualized on a formaldehyde electrophoreses gel, as shown in Figure 3A, and the size of the amplified chains is confirmed by electrophoresis agarose gel as shown in Figure 3B. The binding of SNA to Neu5Ac and β-Galactose was observed to be different than the binding of MLB as both glycans attached to the exact site on the exact pocket of SNA with the affinity of 6.2 and 5.8 kcal/mol for SNA + Neu5Ac and SNA + Gal, respectively ( Figure 2b). The binding of both glycans was found to be solely though H-bonding to Asp 44, Ser 45 and Asp 47. The carboxylic group of Asp 44 was found H-bonded to O5 and O9 atoms of Neu5Ac and O3 and O4 atoms of Gal. The side chain OH of Ser 45 was found H-bonded to O4 and O3 atoms of Neu5Ac and Gal, respectively, but its backbone CO was found H-bonded to O2 and O4 atoms of Neu5Ac and Gal, respectively. Finally, the backbone NH of Asp 47 was found H-bonded to O9 atom of Neu5Ac but its backbone CO was found H-bonded to O6 atom of Gal. These results suggest that the sialic acid and galactose binding sites on MLB are located on two different areas of the same pocket, which allows the protein to bind to both glycans simultaneously. However, the binding of both glycans to the same site of the same pocket on SNA-B suggests that the protein has ability to bind to both glycans but the higher affinity of SNA toward Neu5Ac may increase the chance of binding with Nue5Ac more than galactose. PCR Amplification of MLA and MLB The sequences of the toxic chain (MLA) and the binding chain (MLB) of mistletoe lectin, 762 bp and 789 bp, respectively, were successfully amplified using cDNA templates reverse transcribed from the total RNA of European mistletoe. The extracted RNA was visualized on a formaldehyde electrophoreses gel, as shown in Figure 3A, and the size of the amplified chains is confirmed by electrophoresis agarose gel as shown in Figure 3B. PCR Amplification of MLA and MLB The sequences of the toxic chain (MLA) and the binding chain (MLB) of mistletoe lectin, 762 bp and 789 bp, respectively, were successfully amplified using cDNA templates reverse transcribed from the total RNA of European mistletoe. The extracted RNA was visualized on a formaldehyde electrophoreses gel, as shown in Figure 3A, and the size of the amplified chains is confirmed by electrophoresis agarose gel as shown in Figure 3B. Several sequencing reactions were also performed on three biological replicates of the amplified products to eliminate the possible experimental false mismatch. We found that the coding sequences of both chains were slightly different than compared to the query sequence on NCBI GeneBank (AY377890.1) and some base mismatches were detected in both MLA and MLB chains. The result of the alignment of amplified MLA and MLB sequences with the query sequences showed that MLA had seven amino acids mutated as follows: ATA (Isoleucine), GTT (Valine), ATG (Methionine) and CAG (Glutamine) were replaced with TTA (Leucine), ATT (Isoleucine), GTG (Valine) and CAT (Histidine), respectively ( Figure S1 in Supplementary Data). Moreover, MLB had three amino acids mutated as follow: GAA (Glutamine acid), ATC (Isoleucine) and GTG (Valine) were Several sequencing reactions were also performed on three biological replicates of the amplified products to eliminate the possible experimental false mismatch. We found that the coding sequences of both chains were slightly different than compared to the query sequence on NCBI GeneBank (AY377890.1) and some base mismatches were detected in both MLA and MLB chains. The result of the alignment of amplified MLA and MLB sequences with the query sequences showed that MLA had seven amino acids mutated as follows: ATA (Isoleucine), GTT (Valine), ATG (Methionine) and CAG (Glutamine) were replaced with TTA (Leucine), ATT (Isoleucine), GTG (Valine) and CAT (Histidine), respectively ( Figure S1 in Supplementary Data ). Moreover, MLB had three amino acids mutated as follow: GAA (Glutamine acid), ATC (Isoleucine) and GTG (Valine) were replaced with GGA (Glycine), GTC (Valine) and GCG (Alanine), respectively ( Figure S2 in Supplementary Data). We believe that the main reason behind those mutations was the origin of the mistletoe plants, which were from the United Kingdom and Ukraine for our sequence and the query sequence, respectively. Lectin-GFP Fusion Proteins Since the sialic acid binding activity of the binding domain of mistletoe lectin isoform I (MLB) alone and without the toxic domain (MLA) had never been examined separately, we N/C-terminally fused the MLB sequence to green florescent protein (GFP) in order to generate two recombinant fusion proteins for subsequent sialic acid binding characterization. The same fusion proteins were also generated for the binding domain (SNA-B) of elderberry lectin (Sambucus nigra), which is a well-known sialic acid binding lectin and used as a positive control ( Figure 4). All the fusion proteins were successfully expressed in E. coli. We believe that the main reason behind those mutations was the origin of the mistletoe plants, which were from the United Kingdom and Ukraine for our sequence and the query sequence, respectively. Lectin-GFP Fusion Proteins Since the sialic acid binding activity of the binding domain of mistletoe lectin isoform I (MLB) alone and without the toxic domain (MLA) had never been examined separately, we N/C-terminally fused the MLB sequence to green florescent protein (GFP) in order to generate two recombinant fusion proteins for subsequent sialic acid binding characterization. The same fusion proteins were also generated for the binding domain (SNA-B) of elderberry lectin (Sambucus nigra), which is a well-known sialic acid binding lectin and used as a positive control ( Figure 4). All the fusion proteins were successfully expressed in E. coli. On-plate screening of the fusion proteins was performed to assess the bacterial expression level of the fusion proteins. The results showed that the fusion proteins made from C-terminally fused MLB and SNA-B to GFP were expressed better as they showed higher GFP emission than the N-terminally fused MLB and SNA-B to GFP ( Figure 5). This might be because of the better folding of the proteins or less protein-protein interaction between GFP and the lectin domains in the C-terminal cloning orientation. Therefore, the C-terminally fused recombinant proteins were selected for the large-scale protein expression. Affinity Protein Purification The purpose of using affinity purification chromatography was to purify the recombinant fusion proteins from the total cell lysate as well as the instant assessment of the sialic acid binding of the fusion proteins by assessing their binding to the free sialic acid residues on the fetuin glycoprotein of the column. The fractions of the purified proteins were analyzed with SDS-PAGE and immunoblotting and showed that both N-erminus and C-terminus fusion proteins successfully bound to the sialic acid residues. However, a higher amount of purified protein was recovered from the elution steps of C-terminally On-plate screening of the fusion proteins was performed to assess the bacterial expression level of the fusion proteins. The results showed that the fusion proteins made from C-terminally fused MLB and SNA-B to GFP were expressed better as they showed higher GFP emission than the N-terminally fused MLB and SNA-B to GFP ( Figure 5). This might be because of the better folding of the proteins or less protein-protein interaction between GFP and the lectin domains in the C-terminal cloning orientation. Therefore, the C-terminally fused recombinant proteins were selected for the large-scale protein expression. We believe that the main reason behind those mutations was the origin of the mistletoe plants, which were from the United Kingdom and Ukraine for our sequence and the query sequence, respectively. Lectin-GFP Fusion Proteins Since the sialic acid binding activity of the binding domain of mistletoe lectin isoform I (MLB) alone and without the toxic domain (MLA) had never been examined separately, we N/C-terminally fused the MLB sequence to green florescent protein (GFP) in order to generate two recombinant fusion proteins for subsequent sialic acid binding characterization. The same fusion proteins were also generated for the binding domain (SNA-B) of elderberry lectin (Sambucus nigra), which is a well-known sialic acid binding lectin and used as a positive control ( Figure 4). All the fusion proteins were successfully expressed in E. coli. On-plate screening of the fusion proteins was performed to assess the bacterial expression level of the fusion proteins. The results showed that the fusion proteins made from C-terminally fused MLB and SNA-B to GFP were expressed better as they showed higher GFP emission than the N-terminally fused MLB and SNA-B to GFP ( Figure 5). This might be because of the better folding of the proteins or less protein-protein interaction between GFP and the lectin domains in the C-terminal cloning orientation. Therefore, the C-terminally fused recombinant proteins were selected for the large-scale protein expression. Affinity Protein Purification The purpose of using affinity purification chromatography was to purify the recombinant fusion proteins from the total cell lysate as well as the instant assessment of the sialic acid binding of the fusion proteins by assessing their binding to the free sialic acid residues on the fetuin glycoprotein of the column. The fractions of the purified proteins were analyzed with SDS-PAGE and immunoblotting and showed that both N-erminus and C-terminus fusion proteins successfully bound to the sialic acid residues. However, a higher amount of purified protein was recovered from the elution steps of C-terminally Affinity Protein Purification The purpose of using affinity purification chromatography was to purify the recombinant fusion proteins from the total cell lysate as well as the instant assessment of the sialic acid binding of the fusion proteins by assessing their binding to the free sialic acid residues on the fetuin glycoprotein of the column. The fractions of the purified proteins were analyzed with SDS-PAGE and immunoblotting and showed that both N-erminus and C-terminus fusion proteins successfully bound to the sialic acid residues.

However, a higher amount of purified protein was recovered from the elution steps of C-terminally fused recombinant proteins of both MLB and SNA. This indicates that the sialic acid binding of C-terminally fused proteins was tighter than the N-terminally fused protein as both proteins were expressed and purified in identical conditions. This was also compatible with our previous finding of the on-plate screening of the fusion proteins ( Figure 6). fused recombinant proteins of both MLB and SNA. This indicates that the sialic acid binding of C-terminally fused proteins was tighter than the N-terminally fused protein as both proteins were expressed and purified in identical conditions. This was also compatible with our previous finding of the on-plate screening of the fusion proteins ( Figure 6). Erythrocytes Hemagglutination by MLB-GFP and SNA-GFP The minimum agglutination concentrations (MAC) of MLB-GFP and SNA-GFP fusion proteins to agglutinate erythrocytes were identified by a hemagglutination assay using 1% sheep erythrocytes. Both fusion proteins agglutinated the erythrocytes at a MAC of 0.3 μM and 0.25 μM, respectively. This shows that MLB alone without the toxic domain retained the native hemagglutination activity similar to the full-length counterpart lectins. Both identified MAC of MLB-GFP and SNA-GFP were then used in the hemagglutination inhibition assays using a serial dilution of 1% erythrocytes and different hemagglutination inhibition agents such as D-Gal, Neu5Ac, D-Glc and sialidase enzyme. The fusion proteins were observed to actively agglutinate the erythrocytes when they were used alone; however, their hemagglutination activity was partially compromised when they were pre-incubated with D-Gal, especially MLB-GFP, which showed slightly less activity than SNI-B-GFP (Figure 7). Moreover, both MLB-GFP and SNA-GFP were observed to lose their hemagglutination activity when they were pre-incubated with Neu5Ac and the main reason could be the pre-occupation of their Sia binding sites by Neu5Ac, which prevented them from binding to the terminal Sia on the membrane of erythrocytes and subsequently caused reduced agglutination. A similar result was also detected in the case of sialidase treated erythrocytes because the cells did not possess terminal Sia on their membrane glycoproteins anymore and this stopped them from being bound and agglutinated by MLB-GFP and SNA-GFP. These results suggest that MLB can bind to sialic acid similar to SNA. On the other hand, the incubation of both fusion proteins with D-Glc had no impact on their agglutination activity; this agrees with the previous findings that both fusion proteins have zero affinity for glucose (Figure 7). Erythrocytes Hemagglutination by MLB-GFP and SNA-GFP The minimum agglutination concentrations (MAC) of MLB-GFP and SNA-GFP fusion proteins to agglutinate erythrocytes were identified by a hemagglutination assay using 1% sheep erythrocytes. Both fusion proteins agglutinated the erythrocytes at a MAC of 0.3 µM and 0.25 µM, respectively. This shows that MLB alone without the toxic domain retained the native hemagglutination activity similar to the full-length counterpart lectins. Both identified MAC of MLB-GFP and SNA-GFP were then used in the hemagglutination inhibition assays using a serial dilution of 1% erythrocytes and different hemagglutination inhibition agents such as D-Gal, Neu5Ac, D-Glc and sialidase enzyme. The fusion proteins were observed to actively agglutinate the erythrocytes when they were used alone; however, their hemagglutination activity was partially compromised when they were pre-incubated with D-Gal, especially MLB-GFP, which showed slightly less activity than SNI-B-GFP (Figure 7). Moreover, both MLB-GFP and SNA-GFP were observed to lose their hemagglutination activity when they were pre-incubated with Neu5Ac and the main reason could be the pre-occupation of their Sia binding sites by Neu5Ac, which prevented them from binding to the terminal Sia on the membrane of erythrocytes and subsequently caused reduced agglutination. A similar result was also detected in the case of sialidase treated erythrocytes because the cells did not possess terminal Sia on their membrane glycoproteins anymore and this stopped them from being bound and agglutinated by MLB-GFP and SNA-GFP. These results suggest that MLB can bind to sialic acid similar to SNA. On the other hand, the incubation of both fusion proteins with D-Glc had no impact on their agglutination activity; this agrees with the previous findings that both fusion proteins have zero affinity for glucose (Figure 7). MLB-GFP and SNA-GFP Binding to Mammalian Cells The overexpression of α2, 6 sialic acid on the membrane of WM-266-4 metastatic melanoma cells was previously confirmed in a study conducted by Hoja-Łukowicz and Link-Lenczowsk, 2012. Therefore, we believed that WM-266-4 cell line is a suitable candidate for mammalian cell-based activity assessment of the fusion proteins. The epifluorescence microscopy screening of WM-266-4 cells treated with MLB-GFP and SNA-GFP showed that both fusion proteins successfully attached to the membrane of the cells and were successfully internalized across their membranes as the GFP florescence was detected on the outer surface of the cell membranes as well as inside the cells (Figure 8). The cells were also observed to be agglutinated by both fusion proteins, which confirms their agglutination activity and this is compatible with the outcome of our hemagglutination inhibition assay. MLB-GFP and SNA-GFP Binding to Mammalian Cells The overexpression of α2, 6 sialic acid on the membrane of WM-266-4 metastatic melanoma cells was previously confirmed in a study conducted by Hoja-Łukowicz and Link-Lenczowsk, 2012. Therefore, we believed that WM-266-4 cell line is a suitable candidate for mammalian cell-based activity assessment of the fusion proteins. The epifluorescence microscopy screening of WM-266-4 cells treated with MLB-GFP and SNA-GFP showed that both fusion proteins successfully attached to the membrane of the cells and were successfully internalized across their membranes as the GFP florescence was detected on the outer surface of the cell membranes as well as inside the cells (Figure 8). The cells were also observed to be agglutinated by both fusion proteins, which confirms their agglutination activity and this is compatible with the outcome of our hemagglutination inhibition assay. Discussion The vast diversity of sialic acid distribution among different organisms and its crucial biological functions render this glycan an important biomolecule to investigate more extensively, while understanding the biofunctions of this glycan and its derivatives is still relatively limited. In order to further study the biological functions of Sia, selective and efficient sialic acid binding molecules need to be identified in order to be able to comprehensively study the undiscovered functions of the glycan. MLB-GFP and SNA-GFP Binding to Mammalian Cells The overexpression of α2, 6 sialic acid on the membrane of WM-266-4 metastatic melanoma cells was previously confirmed in a study conducted by Hoja-Łukowicz and Link-Lenczowsk, 2012. Therefore, we believed that WM-266-4 cell line is a suitable candidate for mammalian cell-based activity assessment of the fusion proteins. The epifluorescence microscopy screening of WM-266-4 cells treated with MLB-GFP and SNA-GFP showed that both fusion proteins successfully attached to the membrane of the cells and were successfully internalized across their membranes as the GFP florescence was detected on the outer surface of the cell membranes as well as inside the cells (Figure 8). The cells were also observed to be agglutinated by both fusion proteins, which confirms their agglutination activity and this is compatible with the outcome of our hemagglutination inhibition assay. Discussion The vast diversity of sialic acid distribution among different organisms and its crucial biological functions render this glycan an important biomolecule to investigate more extensively, while understanding the biofunctions of this glycan and its derivatives is still relatively limited. In order to further study the biological functions of Sia, selective and efficient sialic acid binding molecules need to be identified in order to be able to comprehensively study the undiscovered functions of the glycan. The contradictory outcomes of previous studies regarding the glycan affinity of mistletoe lectin left the affinity of this lectin as an answered question, particularly since older Discussion The vast diversity of sialic acid distribution among different organisms and its crucial biological functions render this glycan an important biomolecule to investigate more extensively, while understanding the biofunctions of this glycan and its derivatives is still relatively limited. In order to further study the biological functions of Sia, selective and efficient sialic acid binding molecules need to be identified in order to be able to comprehensively study the undiscovered functions of the glycan. The contradictory outcomes of previous studies regarding the glycan affinity of mistletoe lectin left the affinity of this lectin as an answered question, particularly since older studies classified the lectin as a galactose binding while more recent studies detected sialic acid binding activity from an isoform of the lectin. Furthermore, the concept of Sia binding of mistletoe lectin remained vague for a long time because only limited experiments were conducted and they only focused on sialic acid binding activity of the total MLI isolated from extract of mistletoe [22]. Therefore, we used a genetic fusion approach to precisely examine the Sia affinity of the binding domain (MLB) of mistletoe lectin isoform I (MLI) without the toxic MLA domain in this study. The first binding assay conducted in the affinity purification step utilized the terminal sialic acids on the surface of fetuin glycoproteins and we discovered that MLB can actively bind to Sia as observed for SNA, which is a well-known sialic acid binding lectin. This strongly suggests that the classification of mistletoe lectin as a galactose binding lectin was purely due to a lack of experimentation. Subsequently, hemagglutination and inhibition assays were necessary for comparing the Sia binding of MLB with galactose binding. The slight inhibition of the erythrocyte agglutination by galactose treated MLB-GFP confirmed that MLB has the affinity to galactose as well. This is compatible with the previously reported galactose affinity of mistletoe lectin [23]. However, the agglutination inhibition caused by galactose was not as strong as the inhibition caused by the Sia treated MLB-GFP and this provides a clear indication that MLB binds to the terminal Sia on erythrocytes stronger than binding to galactose. Moreover, the binding of both MLB-GFP and SNA-GFP to the terminal Sia on the membrane glycoproteins of WM-266-4 human melanoma cells confirmed the fact that both lectin domains can similarly bind to the Sia rich glycoproteins on the membrane of mammalian cells. Furthermore, both binding domains preserved their agglutination capacity, which is a well-known biological function possessed by their native full-length counterpart lectins and a crucial function for facilitating their cell membrane binding and cell internalization. The observed activities of mistletoe lectin treatment on human melanoma cells are in line with results of a study conducted by Hoja et al. 1995, which investigated the impact of mistletoe lectin treatment on the interleukin-12 (IL-12) secretion in human peripheral blood mononuclear cells and the treatment was observed to enhance the secretion of an active form of IL-12. The results of our computational screening and the assays conducted in this study confirm that the binding chain (MLB) of mistletoe lectin isoform I (MLI) can actively bind to sialic acid and that the glycan binding activity of both MLB and SNA alone and without their toxic domains remains intact. Therefore, both domains could be used as potent binding agents or monobody to specifically target sialic acid in different biological systems. This could eventually permit the use of the domains in biomedical studies as diagnostic tools or potentially as therapeutic agents in order to activate certain biological signals via binding to sialic acid on specific counter-receptors. Chemicals and Reagents Pre-activated CNBr Sepharose 4B and HR 16/10 FPLC columns were purchased from GE Healthcare. Monoclonal Anti-GFPuv antibody and pGFPuv plasmid were purchased from Clontech. Fetuin glycoprotein from fetal bovine serum, β-Galactosidase enzyme, sialidase enzyme, goat IgG Anti-Mouse IgG-HRP conjugated secondary antibody, sintered glass filter (porosity G3) and all the chemicals (unless it is mentioned) were purchased from Sigma-Aldrich (Sigma-Aldrich, Dorset, UK). Virtual Carbohydrate Docking The crystal structures of MLI (2RG9) and SNA (3C9Z) were derived from RCSB Protein Data Bank as bdp files and they were converted to the final docking structures on Glycoprotein Builder tool on the online platform of Glycam project ( www.glycam.org accessed on 3 February 2021). The structures of Neu5Ac and β-Galactose were drawn on ChemDraw software and converted to 3D structures on Carbohydrate Builder tool on Glycam project. The virtual carbohydrate docking was executed by using AutoDock tools. Total RNA Extraction and cDNA Synthesis Fresh leaves of European mistletoe plant were ground to a fine powder in liquid nitrogen by using mortar and pestle. The RNA extraction was performed by using ISOLATE II RNA plant

kit (Meridian Bioscience, London, UK) following the manufacturer's instructions. The quality of the extracted RNA was later assessed on agarose gel electrophoresis containing formaldehyde and NanoDrop2000 (Thermo Fisher scientific, Waltham, MA, USA). The cDNA was reverse transcribed using SuperScriptTM-II reverse transcriptase kit (Invitrogen, Waltham, MA, USA) following the manufacturer's instructions. Polymerase Chain Reactions (PCR) The coding sequences of MLA and MLB chains of MLI were amplified by using reverse transcribed cDNA as a DNA template and Q5 High-Fidelity polymerase (NEB) with two sets of primer shown in Table 1. Sequencing Reactions The sequencing of the cloned MLA and MLB was carried out using BigDye Terminator cycle sequencing chemistries (Thermo Fisher scientific, USA) and standard forward and reverse M13 primers. The Prism 3100 Genetic Analyzer was used to run the capillary gels and the analysis of the sequencing data was carried out using SnapGene (SnapGene, San Diego, CA, USA), Chroma Lite and CLC Sequence Viewer (Qiagen, Hilden, Germany) software. The full-length sequences of SNA and MLI (MLA and MLB) from the sequencing reactions were aligned against the sequences of previously known glycan binding lectins. Lectin-GFP Fusion Proteins Fusion proteins were generated from the binding chain (MLB) of both mistletoe lectin isoform one (MLI) and elderberry lectin isoform I (SNA-B). A reference gene of SNI-B from NCBI (U27122.1) was used as a template sequence for solid synthesis of the gene carried out by Eurofins Genomics (Eurofins, Luxembourg). The MLB and SNI-B sequences were then cloned in upstream and downstream to the coding sequence of GFP in pGFPuv plasmid ( Figure 9). On Plate Induction The transformed E. coli cells with the plasmids containing the fusion protein constructs were plated on LB-Agar plates, which were surface-covered with 500 uM IPTG. The visualization and imaging of the grown colonies were performed under standard UV light (360-400 nm). Expression of Recombinant MLB and SNA-B The MLB + GFP and SNA-B + GFP fusion constructs were transformed into BL21(DE3) strain and plated on antibiotic selection plates (Ampicillin 100 μg/mL). The transformed cells were grown overnight in 10 mL of liquid LB with 100 μg/mL ampicillin at 37 °C. The overnight culture was added to a litter of fresh LB medium with antibiotic and incubated at 30 °C until OD600 0.7 was reached, followed by induction using 500 μM On Plate Induction The transformed E. coli cells with the plasmids containing the fusion protein constructs were plated on LB-Agar plates, which were surface-covered with 500 uM IPTG. The visualization and imaging of the grown colonies were performed under standard UV light (360-400 nm). Expression of Recombinant MLB and SNA-B The MLB + GFP and SNA-B + GFP fusion constructs were transformed into BL21(DE3) strain and plated on antibiotic selection plates (Ampicillin 100 µg/mL). The transformed cells were grown overnight in 10 mL of liquid LB with 100 µg/mL ampicillin at 37 • C. The overnight culture was added to a litter of fresh LB medium with antibiotic and incubated at 30 • C until OD600 0.7 was reached, followed by induction using 500 µM IPTG for 18 h. BugBuster protein extraction reagent was used to extract the total soluble proteins from the bacterial cells as per manufacture's instruction. The extracted protein was later analyzed by SDS PAGE electrophoresis. Affinity Column Preparation Affinity column bed was prepared from the pre-activated CNBr Sepharose 4B which is an appropriate medium for immobilization of ligands containing primary amines. The medium was coupled with degalactosilated bovine fetuin as a heavily glycosylated protein containing bi-antennary, tri-antennary and tetra-antennary oligosaccharides with variable sialyation. Briefly, the required amount of lyophilized CNBr powder was weighed and suspended in 1 mM HCL. The medium swelled immediately and was washed on a sintered glass filter (porosity G3) with 1 mM HCl for 15 min. The required amount of fetuin (10 mg/mL) was also dissolved in dissolving buffer (0.1 M CH 3 COONa, pH 8.3; containing 0.5 M NaCl) and incubated at 37 • C overnight with an appropriate amount of β-galactosidase to remove the free galactose residues. The fetuin was then extensively dialyzed against a coupling buffer (0.1 M NaHCO3, pH 8.3; containing 0.5 M NaCl) and added to the suspended CNBr medium in a stoppered vessel. The mixture was subject to end-overend rotation overnight at 4 • C and washed later with 5 medium (gel) volumes of coupling buffer in order to wash away the excess ligands. The mixture was incubated in blocking buffer (0.1 M Tris-HCl) pH 8.0 for 2 h to block remaining active groups on Sepharose 4B and washed in three cycles of alternating pH with 5 medium volumes of 0.1 M acetic acid/sodium acetate, pH 4.0 containing 0.5 M NaCl, and 0.1M Tris-HCl, pH 8 containing 0.5 M NaCl. Finally, the coupled CNBr-fetuin bead was packed in HR 16/10 column. Affinity Purification Affinity column chromatography was performed in order to purify the fusion proteins from the total extracted proteins by using the pre-packed CNBr-fetuin column and AKTA prime FPLC platform. The column was equilibrated with running buffer (PBS, pH 6.0) and the effect of pH on the binding of MLB + GFP and SNA + GFP to the column was first analyzed through loading a small scale of MLB, SNA on the column and performing elution steps via three different pH (3.0, 6.0 and 9.0) of the elution buffer (PBS alone). The UV signal of the eluted fractions was monitored and compared to the elution step of a column loaded with only running buffer. At low pH points (3.0 and 6.0), none of the fusion proteins were observed to be eluted by the elution buffer and a negligible amount of the fusion proteins was eluted at pH 9.0. On the other hand, both fusion proteins were observed to be completely eluted from the column when sialic acid was added to the elution buffer at 6.0. Therefore, we decided to use pH 6.0 for the elution buffer, which was PBS containing 0.5 M sialic acid. Collected fractions containing the target fusion proteins were later analyzed on SDS-PAGE and Western blotting was used with the anti-GFPuv primary antibody. Hemagglutination Assay Erythrocytes were prepared by washing fresh sheep blood four times with 0.15 M NaCl and stored in Alsever's medium (0.8% sodium citrate, 0.42% sodium chloride, 2.05% glucose and 0.055% citric acid) at 4 • C. The blood was centrifuged at 1500 g for 5 min to precipitate the erythrocytes, which was later washed with PBS four times in the ratio of 1:5 (v/v). A batch of the washed erythrocytes was treated with 10 unites/mL of sialidase Draper/CED-1 Mediates an Ancient Damage Response to Control Inflammatory Blood Cell Migration In Vivo Summary Tissue damage leads to a robust and rapid inflammatory response whereby leukocytes are actively drawn toward the wound. Hydrogen peroxide (H2O2) has been shown to be an immediate damage signal essential for the recruitment of these inflammatory blood cells to wound sites in both Drosophila and vertebrates [1, 2]. Recent studies in zebrafish have shown that wound-induced H2O2 is detected by the redox-sensitive Src family kinase (SFK) Lyn within the responding blood cells [3]. Here, we show the same signaling occurs in Drosophila inflammatory cells in response to wound-induced H2O2 with mutants for the Lyn homolog Src42A displaying impaired inflammatory migration to wounds. We go on to show that activation of Src42A is necessary to trigger a signaling cascade within the inflammatory cells involving the ITAM domain-containing protein Draper-I (a member of the CED-1 family of apoptotic cell clearance receptors) and a downstream kinase, Shark, that is required for migration to wounds. The Src42A-Draper-Shark-mediated signaling axis is homologous to the well-established SFK-ITAM-Syk-signaling pathway used in vertebrate adaptive immune responses. Consequently, our results suggest that adaptive immunoreceptor-signaling pathways important in distinguishing self from non-self appear to have evolved from a more-ancient damage response. Furthermore, this changes the role of H2O2 from an inflammatory chemoattractant to an activator signal that primes immune cells to respond to damage cues via the activation of damage receptors such as Draper. In Brief Evans et al. reveal that Draper functions in concert with a Src family kinase (Src42A) and a Syk homolog (Shark) to regulate migration of macrophages to sites of damage in Drosophila. This function critically depends upon Draper's immunoreceptor tyrosine-based activation motif and is separable from Draper's canonical role in clearance of dying cells. SUMMARY Tissue damage leads to a robust and rapid inflammatory response whereby leukocytes are actively drawn toward the wound. Hydrogen peroxide (H 2 O 2 ) has been shown to be an immediate damage signal essential for the recruitment of these inflammatory blood cells to wound sites in both Drosophila and vertebrates [1,2]. Recent studies in zebrafish have shown that wound-induced H 2 O 2 is detected by the redoxsensitive Src family kinase (SFK) Lyn within the responding blood cells [3]. Here, we show the same signaling occurs in Drosophila inflammatory cells in response to wound-induced H 2 O 2 with mutants for the Lyn homolog Src42A displaying impaired inflammatory migration to wounds. We go on to show that activation of Src42A is necessary to trigger a signaling cascade within the inflammatory cells involving the ITAM domain-containing protein Draper-I (a member of the CED-1 family of apoptotic cell clearance receptors) and a downstream kinase, Shark, that is required for migration to wounds. The Src42A-Draper-Sharkmediated signaling axis is homologous to the wellestablished SFK-ITAM-Syk-signaling pathway used in vertebrate adaptive immune responses. Consequently, our results suggest that adaptive immunoreceptor-signaling pathways important in distinguishing self from non-self appear to have evolved from a more-ancient damage response. Furthermore, this changes the role of H 2 O 2 from an inflammatory chemoattractant to an activator signal that primes immune cells to respond to damage cues via the activation of damage receptors such as Draper. RESULTS AND DISCUSSION Because H 2 O 2 is an evolutionarily conserved, wound-induced damage signal and has been shown to be detected by the redox-sensitive Src family kinase (SFK) Lyn in zebrafish neutrophils [3], we began by investigating whether Lyn's closest relative played a similar role in the recruitment of inflammatory cells to wounds in Drosophila embryos. The SFK most closely related to zebrafish Lyn in Drosophila is Src42A, and the critical redoxsensitive cysteine residue necessary for H 2 O 2 detection in the fish (C466) is conserved in Src42A but is absent from the remaining SFKs and related non-receptor tyrosine kinases (src62B, abl, and btk29A) [3]. Live imaging of macrophage (hemocyte) responses to laser-induced epithelial wounds in src42A E1 mutant embryos revealed that these inflammatory cells essentially ignored such wounds, exhibiting directionalities close to zero (Figures 1A-1B 0 ; Movie S1). Immunostaining revealed a strong loss of Src activity in these mutant embryos ( Figure S1A), and the src42A E1 wound recruitment defect was phenocopied when placed in a heteroallelic combination with a src42A myri loss-offunction allele (Figures S1B 0 and S1B 00 ), revealing this inflammatory deficit as specific to src42A. Developmental dispersal of macrophages (Figures S1C and S1D) and migration speeds of src42A mutant macrophages following injury (Figure 1B 00 ) were indistinguishable from controls, suggesting that their migratory machinery remains intact; therefore defective motility is unlikely to underlie the failure of these cells to respond to wounds. Furthermore, src42A does not appear necessary for specification or proliferation of Drosophila macrophages, demonstrating a specific role in these cells for src42A in wound recruitment (Figures S1C and S1D). In order to test whether the wound recruitment defect observed in src42A mutants is due to a macrophage-specific requirement for Src42A, we expressed a dominant-negative version of Src42A (Src42A DN ) [4] specifically in macrophages and assessed their ability to respond to wounds. Disrupting Src42A function in this way was sufficient to impair inflammatory recruitment following wounding, demonstrating a cell-autonomous function for Src42A ( Figures 1C and 1D). Expression of Src42A DN within macrophages did not alter overall cellular morphology, though Src42A DN macrophages exhibited slightly larger spread areas in vivo (Figures S1E 0 and S1E 00 ) and migrated marginally faster than controls (data not shown). In contrast, macrophages in src42A zygotic mutants appeared slightly smaller when visualized live ( Figure S1F), potentially reflecting non-macrophage autonomous effects and an altered in vivo environment, which presumably explains the stronger inflammatory defects in these embryos compared to those observed in Src42A DN experiments; for example

src42A has a role in epidermal responses to injury [5]. In summary, src42A appears to have a specific role in governing macrophage responses to injury in Drosophila embryos and this function is consistent with the related role of zebrafish Lyn in neutrophils [3], where it operates as a redox sensor to alert blood cells to the presence of wound-induced H 2 O 2 . What signaling is occurring downstream of Src42A following the detection of H 2 O 2 ? In Drosophila glia responding to degenerating axons, src42A interacts genetically with the Drosophila CED-1 homolog draper, where a tyrosine in Draper's ITAM (immunoreceptor tyrosine-based activation motif) domain is critical for these responses [6]. Src42A can phosphorylate Draper at this residue in vitro, and this results in the recruitment of a Syk-related kinase called Shark [6]. In glia, shark also genetically interacts with src42A and draper [6], revealing a tri-partite signaling axis evocative of the well-established SFK-ITAM-Syksignaling paradigm employed in adaptive immune responses in vertebrates [7]. We therefore wondered whether detection of wound-induced H 2 O 2 by Src42A in macrophages might trigger the same signaling pathway and be important for their migration to wounds. Live imaging macrophage responses to wounds in draper mutants revealed that these inflammatory migrations are severely impaired with fewer macrophages present at the Scale bars represent 20 mm. Central lines and error bars on scatterplots represent mean and SD, respectively; ns, not significant; *p < 0.05 and ***p < 0.001 via Mann-Whitney test (B) or one-way ANOVA followed by Sidak's multiple comparisons test (D); Mf, macrophages; white ovals depict wound edges. See also Figures S1 and S3 and Movie S1. wound 1 hr after wounding (Figures 2A and 2B). To test whether draper was required cell autonomously in macrophages, we used RNAi-mediated knockdown: this approach efficiently depleted draper-I levels in macrophages as demonstrated by qPCR for draper-I transcripts in FACS-sorted macrophages ( Figure S2A), and overexpression of this construct also led to loss of Draper protein in stage 15/16 embryos (Figures S2B 0 and S2B 00 ). RNAi-mediated knockdown of draper specifically in macrophages led to the same reduction in inflammatory cells at wounds, demonstrating a macrophage-specific requirement for Draper in mediating efficient inflammatory responses to damage (Figures 2A and 2B). We next wanted to determine whether the third participant of the wellestablished SFK-ITAM-Syk immune signaling pathway was involved in this innate response to damage in vivo: live imaging of macrophage responses to laser wounds in embryos mutant for the Syk homolog shark showed that these also have an impaired ability to raise an inflammatory response, with less macrophages arriving at wounds 1 hr post-wounding ( Figures 2C, 2D, and S2C). Importantly, we were able to demonstrate specificity and a cell-autonomous role for Shark by re-expression of Shark in macrophages within a shark 1 mutant background, an approach that rescues wound responses to control levels ( Figures 2E and 2F). A reduction in macrophage numbers does not appear to explain the wound-recruitment phenotypes, because both local and total numbers of macrophages appear unaffected in draper, src42A, and shark mutants (Figures S1C, S1D, and S3). To confirm src42A and draper operate in the (legend on next page) same genetic pathway, we wounded transheterozygous src42A/draper mutant embryos and compared them to controls. The resulting wounds showed a reduction in inflammatory cells present at the wound sites when src42A E1 /+ heterozygotes were compared with src42A E1 /draper D5 transheterozygotes, suggestive of a genetic interaction ( Figure 2G). Taken together, these results suggest that a Src42A-Draper-Shark-signaling axis is critical for the efficient recruitment of inflammatory macrophages to wounds in vivo. During the late stages of glial responses to axonal injuries, an alternative splice variant, Draper-II, becomes highly upregulated. Rather than an ITAM, the cytoplasmic domain of Draper-II contains an ITIM (immunoreceptor tyrosine inhibitory motif) [8]. Draper-II uses its ITIM to attenuate Draper-I signaling via the recruitment of a phosphatase, Corkscrew, which dephosphorylates Shark [8]. We found that macrophage-specific expression of Draper-II also impaired inflammatory migration to laser-induced epithelial wounds in Drosophila embryos ( Figures 3A and 3B), further demonstrating a role for Draper signaling in wound responses and highlighting the importance of the ITAM-containing intracellular domain of the Draper-I isoform in this process. ITAMs are found in many mammalian immune receptors involved in adaptive immune responses, such as B cell and T cell receptors, which, as per Draper, can be directly phosphorylated by SFKs [6,7]. Taken together, these results demonstrate a requirement for SFK-ITAM-Syk signaling in innate immune cell inflammatory responses to damage-induced H 2 O 2 and places the ITAM-containing Draper-I variant at the center of this damage-induced signaling cascade. However, draper encodes a homolog of the C. elegans apoptotic cell clearance receptor CED-1 [9] and has been shown to play a role in the detection and/or the processing of apoptotic debris in both Drosophila macrophages and glia [10,11], the two predominant phagocytes within developing Drosophila embryos [12][13][14]. Indeed, Drosophila embryonic macrophages actively prioritize apoptotic cells above the growth factor signals that guide their developmental migrations [2]. A role for the CED-1 family in apoptotic cell clearance appears conserved through to higher vertebrates, because Jedi-1 and MEGF10, the clearest homologs of Draper in mammals, are also involved in the removal of apoptotic cells [15,16]. Consistent with a role in the efficient processing of engulfed apoptotic debris [11], we found that draper mutant and draper RNAi-expressing macrophages appeared vacuolated, containing increased numbers of apoptotic corpses per cell compared to controls (Figures 3C, 3D, and 4D). We have previously shown that efficient processing of engulfed apoptotic corpses is critical for normal macrophage migration to occur [17], and similarly, we were able to observe a reduction in basal migration speeds of macrophages in draper mutant embryos or upon macrophage-specific expression of either Draper RNAi or Draper-II (Figures 3E and 3F). Draper's role in the engulfment and degradation of apoptotic and axonal debris requires the adaptor Ced-6 [18][19][20], the recruitment of which depends on the presence of an NPXY motif within the cytoplasmic domain of Draper [20,21]. Our findings, however, suggest that Draper's role in wound detection may be more reliant on its ITAM domain, because despite the presence of an NPXY motif in the Draper-II isoform, expression of Draper-II antagonized wound recruitment. To attempt to separate Draper's role in clearance and migration, we asked whether expression of a form of Draper-I that lacked the Src phosphorylation site on its ITAM domain (Drpr-I Y949F ) [6] was able to rescue the ability of macrophages either to migrate to a wound or process engulfed apoptotic corpses in a draper-null mutant background. Crucially, both Draper-I constructs can be expressed at comparable levels, including when expressed in macrophages in a draper mutant background ( Figures S4A and S4B), while localization and expression levels were very similar in overexpressing macrophages cultured in vitro ( Figures S4C 0 and S4C 00 ). This suggests there are no intrinsic differences in expression levels as a result of transgene insertion sites, nor are there differences in protein localization between constructs that might undermine this approach. The Y949F mutation strongly perturbs Draper-Shark interactions in vitro [6], and we found that, whereas expression of full-length Draper-I WT in draper mutant macrophages was sufficient to rescue their ability to launch an inflammatory response to wounds, expression of the version lacking the Src phosphorylation site was not ( Figures 4A-4C). However, the same mutated version of Draper rescued vacuolation defects ( Figure 4D) and basal migration rates (Figure 4E) as robustly as the expression of full-length Draper-I WT , demonstrating that, whereas its ITAM domain is critical for Draper's function in macrophage recruitment to wounds, the ITAM domain is dispensable for its role in apoptotic corpse processing, which instead may rely more heavily on the NPXY motif. Our findings demonstrate a novel role for Draper in the innate immune inflammatory response to wounds and place it downstream of the early damage cue H 2 O 2 . Our results suggest that H 2 O 2 production at wounds is detected by Src42A acting as a redox sensor within macrophages and that this then triggers the phosphorylation of Draper-I on its ITAM domain and the downstream recruitment and activation of the kinase Shark (Figure 4F). SFK-ITAM-Syk signaling is a well-established immunesignaling pathway used in the mammalian adaptive immune response during B cell and T cell signaling. Our results suggest that this adaptive immune-signaling pathway important in distinguishing self from non-self appears to have evolved from a more-ancient damage response and changes the role of H 2 O 2 from an inflammatory chemoattractant to an activator signal that potentially primes immune cells to respond to damage cues via the activation of damage receptors such as Draper and possibly other ITAM domain-containing proteins. Live Imaging of Drosophila Macrophages For all live imaging experiments, stage 15 embryos were collected from overnight apple juice agar plates and mounted on slides in a minimal volume of 10S Voltalef oil (VWR), following dechorionation in bleach for 1 min and extensive washing in water. All imaging was carried out at room temperature. For dynamic imaging of wound responses, epithelial wounds were induced using a nitrogen-pumped Micropoint ablation laser tuned to 435 nm (Andor Technologies), as per Razzell et al. [36]. EGFP or nuclear-red-stingerexpressing macrophages were followed at 30-s intervals for 20 min postwounding using a 403 oil immersion objective lens on a PerkinElmer UltraView spinning disc system. Basal migration speeds were analyzed by making time-lapse movies of macrophages specifically expressing either nuclear red stinger or EFGP in embryos on the ventral side of the embryo at stage 15 of development using a Leica LSM510 confocal and a 403 oil immersion objective. Movies were as per Evans et al. [17], with z stacks collected every 30 s for 20 min. Wound responses were also quantified at 20 and 60 min post-wounding (numbers of macrophages per mm 2 of wound, normalized to the average wound response of the control). For a detailed description of image processing and analysis, see the Supplemental Experimental Procedures. Advanced materials for enamel remineralization Dental caries, a chronic and irreversible disease caused by caries-causing bacteria, has been listed as one of the three major human diseases to be prevented and treated. Therefore, it is critical to effectively stop the development of enamel caries. Remineralization treatment can control the progression of caries by inhibiting and reversing enamel demineralization at an early stage. In this process, functional materials guide the deposition of minerals on the damaged enamel, and the structure and hardness of the enamel are then restored. These remineralization materials have great potential for clinical application. In this review, advanced materials for enamel remineralization were briefly summarized, furthermore, an outlook on the perspective of remineralization materials were addressed. Introduction The enamel, consisting of 96-97 wt% inorganic hydroxyapatite (HA, Ca 10 (PO 4 ) 6 (OH) 2 ), 3wt% water and 1wt% organic material, is the hardest tissue in the human body (Bowen et al., 2018;Harper et al., 2021). However, enamel is susceptible to acid, causing enamel demineralization and even developing cavities (Pitts et al., 2017). Currently, hundreds of millions of people in the world is under the enamel damage (Peres et al., 2019). It is difficult to repair enamel on its own due to the lack of sufficient calcium and phosphate ions in saliva (Lawn et al., 2010;Lacruz et al., 2017). Therefore, artificial materials such as resin, metal or bioglass are commonly used for clinical repair of cavities (Dorri et al., 2017). In terms of composition, mechanical properties, and appearance, these composites differ significantly from enamel. By comparison, enamel remineralization can be an effective clinical method for restoring the natural properties and structure of enamel while avoiding the problems associated with filling materials. Remineralization requires replacing minerals lost during the early stages of demineralization to restore enamel hardness or structure. Remineralized materials are essential to enamel repair. Functional materials can promote and arrange the deposition of calcium and phosphate ions or alter the solubility of the HA. They can be divided into inorganic materials, organic materials, and polymeric materials ( Figure 1). These functional materials are designed to rebuild remineralized OPEN ACCESS EDITED BY tissue on damaged enamel surfaces, thereby preventing disease progression while also improving aesthetics and mechanical strength. Therefore, materials for enamel remineralization have a bright

future in clinic. Although several reviews of remineralized materials have been published (Cochrane et al., 2010;Ding et al., 2017;Pandya and Diekwisch, 2019), enamel remineralized materials have been innovated and developed. As a result, it is critical to review the relevant research progress in time for the construction and upgrading of the enamel remineralization system. In this review, the characteristics and working mechanism of remineralized materials are briefly summarized. The specific functions of various functional materials will be clarified by category, with reference opinions provided for future material design and synthesis. Functional inorganic materials Functional inorganic materials can induce the formation of apatite layers or release ions, which can promote the remineralization of enamel. When the remineralized layer forms, calcium phosphates (CaPs) provide exogenous ions to compensate minerals lost by enamel, while fluoride and magnesium ions can exchange with calcium ions in HA, changing the solubility and mechanical properties of iondoped HA. Therefore, the ability of inorganic materials to release ions and the change in enamel properties caused by their participation in HA are the primary focal point of researches. Calcium phosphates CaPs can provide ions to reconstruct damaged enamel. Remineralization solutions containing calcium and phosphorus ions are usually used in remineralization experiments, which must be replaced or replenished on a regular basis. Some stable CaPs materials can provide ions required for an extended time. Amorphous calcium phosphate (ACP), tricalcium phosphate (TCP), and nano-hydroxyapatite (nHA) are common CaPs materials used for remineralization. The type and size of the CaPs crystals can influence the ion supply capacity and the depth of ion entry into the lesion. Therefore, the mineralization effects of these materials are differrent. ACP, the precursor phase of biogenic HA of bone and tooth, is the basic mineralization unit in the biological mineralization process (Gelli et al., 2019). Aqueous ACP solutions contain abundant Ca 2+ and PO 4 3− ions, which form highly hydrated clusters. The structure and composition of the crystalline phase change after further aggregation of clusters until the thermodynamically stable crystalline HA (alkaline conditions) or carbon brushes (acidic conditions) formed . Usually, such reaction time is fast in the absence of external interference. Only ACP solutions failed to restore enamel (Shao et al., 2019). Therefore, enamel remineralization requires ensuring the stability of ACP in solution and prolonging its phase transition time. Acidic groups, such as carboxyl and phosphoric groups, can bind calcium ions in solution, preventing Ca 2+ and PO 4 3− from aggregating. Organic compounds with carboxyl or phosphate groups are the most common ACP stabilizers. It is a good method to use amino acids such as aspartate (Asp), glutamate (Glu), citrate (Delgado-López et al., 2014;Iafisco et al., 2015), and the phosphate stabilizer triethylamine (Shao et al., 2019) to maintain the size of ACP particles, ensuring ion supply in the subsequent mineralization process. In addition, the casein phosphopeptide (CPP) that containing four to seven phosphate groups can attach to ACP nanoclusters, forming CPP-ACP. CPP-ACP complexes have been used as common additives for caries prevention. Furthermore, CPP-ACP in combination with fluoride show advantages in remineralization of existing lesions (Bijle et al., 2018;Tao et al., 2018). However, CPP-containing products should be used with caution in individuals with lactose intolerance issues. TCP can be divided into α-TCP and β-TCP according to the crystal form. β-TCP is often used in dental materials because of its great biodegradability and biocompatibility. When exposed to acid, β-TCP degrades to release ions for enamel restoration. After surface functionalization by carboxylic acid and surfactants, functionalized TCP (fTCP) can prevent fluoride from binding with calcium ions on the enamel surface prematurely to build a low-dose fluoride release system (Karlinsey and Pfarrer, 2012;Shen et al., 2018;Viana et al., 2020). After combining with fumaric acid, fTCP can show significantly higher calcium bioavailability than β-TCP and better remineralization of subsurface enamel damage (Karlinsey et al., 2010). nHA is a bioactive and biocompatible material with a small particle size of 10-20 nm in diameter and 60-80 nm in length (Huang et al., 2011). The nanometer size enables nHA to penetrate deeper lesion layers through large lesion pores and repair enamel damage (Juntavee et al., 2018;Bossu et al., 2019;Memarpour et al., 2019). However, high-concentrating nHA tend to self-aggregate into large-sized nHA, which can affect the amount and depth of nHA entering the lesion (Huang et al., 2009). As a carrier, the gel effectively extends the contact time between the active ingredient and the enamel, allowing nHA to fill the small holes and depressions. Both silica-based glycerol hydrogel (Khonina et al., 2020) and carbomer-based gel (Sari et al., 2021) containing nHA can repair damaged enamel. Fluorinated compounds Fluorinated compounds have been commonly utilized since the previous century to reverse or prevent enamel defects from spreading. Consequently, the global incidence rate of dental caries has decreased dramatically (Jokstad, 2016;Clark et al., 2020). Fluoride reduces demineralization by altering enamel solubility (Lynch et al., 2004). Fluoride and calcium ions are more strongly bound than hydroxyl groups. Therefore fluoride can replace hydroxyl to form fluorapatite (FAP), which has high acid resistance and poor solubility (Clark et al., 2020). Fluorides, on the other hand, can promote remineralization by encouraging Ca 2+ in saliva to attach to the tooth surface. In addition, fluoride can reduce the adhesion and growth of germs by blocking the activities of numerous enzymes. Fluoride is primarily ingested through drinking water (75%). Fluoridation of home water is a typical measure to prevent dental caries in many countries, and it can successfully reduce the incidence of dental caries. Fluoride can also be found in a variety of oral care treatments and dental materials, including sodium fluoride (NaF), stannous fluoride, silver diamine fluoride, acidulated phosphate fluoride, ammonium fluoride, and others (Barrera-Ortega et al., 2020). Fluoride sustained-release ability could be altered by combining fluoride ions with different cations and complexing it with different organic molecules. In toothpaste and rinses, polyvalent fluorides with tin and titanium as cations exhibit excellent corrosion resistance (Zanatta et al., 2020). This is due to the fact that they can not only produce CaF 2 on the enamel surface, but can also generate metallic precipitates on the enamel surface, which contributes to the reduction of calcium ion loss when subjected to external erosion. However, fluorides have caused certain issues when they are used. Fluoride tends to develop a disordered layered structure of remineralization layer, which is considerably different from natural enamel. The mechanical characteristics of the remineralized layer can be weakened by these disordered formations. Organic compounds like amelogenin can help minimize the occurrence of disordered structures in the reaction system (Yu et al., 2019) (Figure 2). Moreover, excessive fluoride use can result in hazardous effects like dental fluorosis and skeletal fluorosis (Philip, 2019). It also has the possibility to make cariogenic bacteria resistant, diminishing the effectiveness of follow-up prophylaxis . Fluorinated hydroxyapatite is rapidly formed in the superficial enamel layer of very concentrated F − solutions, preventing Ca 2+ and PO 4 3− from penetrating deeper into the lesion. As a result, subsurface enamel lesions can fail to mineralize adequately. Therefore, fluoride slow-release systems made of copolymer acrylic reservoirs and glass ionomer cement can be great alternatives for promoting enamel mineralization by extending the trailing effect (Chong et al., 2018). Fluoride compounds, as traditional enamel remineralization materials, have a relatively well-studied mineralization mechanism, which facilitates the development of novel fluoride-mediated remineralization systems. However, the functions of fluoride compounds are still need to be improved, and thus, blends or composites of fluoride compounds, which can combine multifunction together to achieve satisfactory clinical results, are greatly needed in the future. Magnesium related materials Magnesium presents in the hard tissues of the body. In enamel, the content of Mg 2+ is ranging from 0.2 to 0.5 wt%. Mg 2+ is present near the grain boundaries as an intergranular phase of Mg substituted amorphous calcium phosphate (Mg-ACP) (La Fontaine et al., 2016). Such amorphous phases have been proved to make a significant impact on the mechanical characteristics and wear resistance of enamel (Gordon et al., 2015). Mg 2+ slows crystal growth by competing with calcium ions at the growth point during mineralization, affecting the production of apatite (Ren et al., 2010;Abdallah et al., 2016). As a result, Mg 2+ can act as a competitive inhibitor to guide narrower crystal columns, which promotes a highly ordered arrangement and increases mineralized tissue hardness. As the concentration of Mg 2+ on the enamel surface increases, the nanohardness of the enamel rises dramatically (Kis et al., 2021). Layer by layer mineralization process is used to create multilayer arrays of enamel-like FAP/polymer nanocomposites controlled by Mg 2+ (FPN-M) at room temperature (Figure 3). In the presence of Mg 2+ , the single nanorods are refined in size and a highly compact array is formed, eventually, (FPN-M) n exhibits excellent mechanical strength and transparency. The present researches have demonstrated Mg 2+ have great importance during the process of enamel remineralization, therefore, more and more attention should be paid to Magnesium related materials. Besides, further understanding of the relationship between Mg 2+ and biomineralization can help develop strategies to improve the mechanical properties of mineralized tissues and improve the functions of repaired tooth enamel. Functionalized organic materials Inorganic matter production and growth require a relatively constant environment, which organic materials can offer. Organic molecules are rich in acidic functional groups such as carboxyl, phosphate, and sulfonic acid. These functional groups can induce inorganic compound nucleation, inhibit overgrowth, or interact with hydroxyapatite on the surface of the enamel to increase adsorption capacity. Understanding the specific role of these organic compounds can help to clarify the mechanism of enamel mineralization and provide ideas for future remineralized material design. Amino acids Amino acid molecules contain different amounts of amino and carboxyl groups. Depending on the isoelectric point, amino acids can be classified as acidic, neutral, and basic amino acids. Among them, acidic amino acids are negatively charged in weakly acidic solutions, which can influence the nucleation, crystallization, growth, and crystal transformation of HA. Glu and Asp can operate as soft templates, connecting two calcium ions diagonally to generate ordered HA crystals that parallel to the enamel column while stabilizing calcium ions in the solution. The crystals on the enamel surface can grow more ordered with the amino acid concentration rises. Asp and Glu is used to deposit the CaCO 3 layer as a sacrificial template on the enamel's surface (Wu et al., 2015). The acidic amino acids then absorbed phosphate and carbonate ions, depositing HA into the CaCO 3 layer to form the rod crystal. Glycine (Gly), a highly hydrophilic amino acid, can also be used as a biological additive to create enamel-like structures . According to the molecular dynamics experiment, Gly exhibits the same adsorption abilities and coverage in all directions of the crystal surface, which maintains HA's c-axis propensity (Pan , 2007). In the carboxymethyl chitosan-stabilized ACP remineralization system, a rod-shaped crystal layer is successfully produced in artificial caries when Gly is introduced to the system, whereas the system without Gly fail . Arginine (Arg), a basic amino acid, positively affects pH homeostasis, bacterial ecology, and pathogenicity. Arg is metabolized in oral biofilms to produce ammonia mainly through the internal arginine deiminase system (ADS) of bacteria (Streptococcus sanguis and Streptococcus). Ammonia produced under this pathway has a significant pH-raising effect, while inhibiting tooth demineralization by neutralizing acids in the peripheral environment (Bijle et al., 2021b). It also facilitates the formation of arginine-friendly microorganisms while disrupting the internal homeostasis of caries-causing bacteria (Nascimento et al., 2019). The combination of Arg and fluoride can create a pH-responsive fluoride pool that inhibit acid production and has potential synergistic effects in maintaining a healthy oral microbial balance (Agnello et al., 2017;Bijle et al., 2021a). The pool can also significantly improve the fluoride uptake and surface hardness of damaged enamel compared to fluoride alone (Zheng et al., 2015;Bijle et al., 2020). Enamel matrix proteins and proteases Enamel matrix proteins (EMPs) and proteases control the formation of enamel (Jia et al., 2020;Shin et al., 2020). EMPs govern the parallelism between the glazing columns and organize them in a dense and slender hexagonal prism structure at the micro-level by regulating the creation and structure of HA crystals (Bartlett, 2013;Uskokovic, 2015;Bai et al., 2020). These highly co-oriented glaze columns give enamel its remarkable shear strength and make it resistant to

everyday abrasion (Yeom et al., 2017). Over 90% of the EMPs consists of amelogenin. Amelogenin can be enzymatically processed into different peptides. These peptides undergo a change in spatial Frontiers in Bioengineering and Biotechnology frontiersin.org conformation, manifested by α-helix unraveling and β-sheet and β-turns formation, at which point amyloid-like aggregation occur in the proteins (Carneiro et al., 2016;Bai et al., 2020). Then, they self-assemble into oligomers and nanospheres (Fang et al., 2011;Engelberth et al., 2018;Bai et al., 2020). These oligomeric nanospheres further form nanochains that concentrate Ca 2+ and PO 4 3− in the peripheral matrix, generating mineralized precursors during enamel development, which then serve as templates to guide the crystal phase transition, eventually generating HA (Gil-Bona and Bidlack, 2020) ( Figure 4). The enamel columns then elongate in one direction to form a hexagonal prismatic structure (Jokisaari et al., 2019). Enzymes are critical requirements for enamel biomineralization. Enzymes activate the biological function of amelogenin and degrade organic matter in the matrix until a sufficiently hard tissue formed (Prajapati et al., 2016). Matrix metalloproteinase 20 (MMP-20) cleaves amelogenin, and the product peptide controls the lengthening and growth of crystal nucleus and induces HA mineralization (Fukae et al., 1998;Nagano et al., 2009;Gil-Bona and Bidlack, 2020). Addition of MMP-20 to full-length porcine amelogenin can promote neatly aligned bundles of enamel-like HA, whereas in the absence of MMP-20, only ACP particles seen (Kwak et al., 2016). Another important enzyme is Kallikrein-related peptidase 4 (KLK4). During enamel maturation, KLK4 degrades the organic matrix in the mineral (Smith et al., 2017;Sari et al., 2021). The width and thickness of the microcrystals can increase when proteins are removed from mature enamel. If the enzyme is deficient, the enamel will undergo hypoplasia (Simmer et al., 2009;Smith et al., 2017). In-depth studies of the enzymatic cleavage products revealed three major functional domains of the amelogenin (Mukherjee et al., 2019;Dissanayake et al., 2020). N-terminal: a hydrophobic tyrosine-rich N-terminal region, known as tyrosine-rich amelogenin peptide (TRAP), is critical in the directed assembly of amelogenin (Buchko et al., 2018). The central region: the central hydrophobic proline-rich region is mainly composed of X-Y-proline (X and Y are usually glutamine) repeat motifs, which is rich in β-sheets and βturns. The C-terminal: a highly hydrophilic domain contains a large number of acidic amino acid residues. These residues could combine with Ca 2+ to provide nucleation sites and bind to the (100) face of octacalcium phosphate (OCP), the intermediate sub-stable phase in early enamel, thereby govern the direction in which the enamel column extends (Wu et al., 2017) (Figure 5). Leucine-rich amelogenin peptide (LRAP), which contains two self-assembled domains of fulllength amelogenin, is the most common alternative splicing product of amelogenin (Xia et al., 2016;Green et al., 2019). It is discovered that depending on the phosphorylated version of the peptide on serine 16, LRAP can perform distinct activities. Phosphorylated LRAP (+P) inhibits calcium phosphate crystallization and stabilizes ACP, whereas LRAP (-P) directs the production of aligned enamel crystals (Yamazaki et al., 2017;Le Norcy et al., 2018). In vitro, LARP and PP i are used to remineralize the eroded enamel, and acicular HA crystals are successfully regenerated on the surface (Kwak , 2017). Inspired by these structural domains, amelogenin analogs are designed and synthesized to induce in vitro bionic remineralization. After grafting different fragments, these synthetic functional peptides can be easier to obtain and show certain functional enhancements, such as adsorption ability. Non-amelogenins work in early enamel formation, including enamelin and tuftelin. Enamelin acts as a transport and nucleation protein that affects amelogenin to regulate early enamel development (Bartlett et al., 2006;Lacruz et al., 2017;Yan et al., 2017). Tuftelin is an acidic protein produced by ameloblasts during the early stages of enamel formation. It is concentrated near the dentin-enamel intersection, in which enamel mineralization begins. The tuftelin-derived peptide (TDP) is created based on the structure of tuftelin. The group repaired by TDP demonstrate comparable enamel hardness and lesion depth healing results after pH cycling to NaF groups (Ding et al., 2020). Functional peptides Functional peptides inspired by bioproteins can in some ways replicate the unique functions of these bioproteins, as well as easier access. Assembly of these peptides with different functions can produce multifunctional peptides, such as peptides with high enamel binding and remineralization capacity or peptides with antibacterial and remineralization activities (Table 1). Amelogenin analogs Amelogenin analogs are created by mimicking the functional domain of amelogenin. These synthetic peptides outperform fulllength amelogenin in synthesis, purification, and retention (Gungormus et al., 2012;Dogan et al., 2018). The focus of recreating enamel structure and function in vitro is inducing columns growth and elongation directly, which is predominantly regulated by the C-tail. Therefore, peptides with C-terminal can stimulate remineralization in vitro. Amelogenin inspired peptides of 26 and 32 amino acid residues (P26 and P32) with hydrophilic inner N-and C-terminal are produced to mimic the "nanosphere" structure in the enamel matrix . A firm mineralized layer is successfully produced on the enamel surface after 7 days of in-situ culturing with polypeptide solution. C-axis oriented nanorods are generated on the enamel surface by repeating the peptide application process. P32 can restore the hardness of etched enamel better because the crystals created by P26 are smaller than those produced by P32. A chimeric peptide is created by grafting the C-terminal onto HA6-1, which can be selectively attached to the enamel surface . The C-terminals of the chimeric peptide increases the peptide adsorption and facilitates the formation of a mechanically strong remineralized layer. QP5 is consisting of five highly conserved Gln-Pro-X repeat sequence in the center region and a hydrophilic C-tail (Lv et al., 2015;Chu et al., 2018;Ren et al., 2018;Li et al., 2020). When compared to amino acids, QP5 has a better remineralization impact, which effectively restored enamel surface hardness and reduced surface roughness value . Moreover, QP5 can enhance remineralization in a complicated oral environment, as demonstrated by the rat caries model (Han et al., 2017). Shortened amelogenin derived peptide 5 (shADP5) is employed as an active ingredient to generate a mineralized layer in solution. The enamel surface is healed after 1 h of mineralization, and the average hardness and elastic modulus are higher than control samples, with the hardness of 2.23 ± 0.23 GPa vs. 2.10 ± 0.26 GPa and elastic modulus of 58.6 ± 4.7 GPa vs. 55.1 ± 4.3 GPa (Dogan et al., 2018). A phase transfer lysozyme (PTL) membrane can be used to simulate the N-terminal of amelogenin . After the occurrence of amyloid aggregation, the internal structure of lysozyme is changed: the α-helixes unravel and the β-sheets is formed through hydrophobic interactions, which is similar to the spatial phase shift of amelogenin selfassembling. At the liquid/solid interface, those β-sheet-rich proteins are quickly organized into nanoparticles, forming a nanofilm that could be adsorbed on the enamel surface and serve as a scaffold for subsequent remineralization. The hydrophilic C-tail is then grafted onto PTL to produce PTL/ C-AMG, which can guide HA growing in a direction. A 2.0-2.8 μm thick remineralization layer is produced after applying 1 mg/ml PTL/C-AMG to demineralized tooth slices for 7 days. These remineralized layers have similar properties to natural enamel with a "fish-scale" structure. Statherin derived peptide Statherin, a tyrosine-rich peptide with 43 amino acid residues, is a salivary protein that is found in the oral acquired membrane. Because of a unique combination of high negative charged domains on the N-terminal and enamel surface, statherin can securely cling to the enamel surface (Raj et al., 1992;Gururaja and Levine, 1996). To replicate the property of high HA binding, several peptides derived from statherin are created (Shuturminska et al., 2017;Luo et al., 2019;Carvalho et al., 2020). Separating the N-terminal of statherin can yield the peptide SN15 (Dodds et al., 2005;Shimotoyodome et al., 2006;Luo et al., 2019). Grafting SN15 onto PAMAM can improve its absorption on enamel surface . The statherin stimulated peptide and tannic acid are used to (2014) Abbreviations: %SMH R , surface microhardness recovery ratio; ADP5, amelogenin derived peptide 5; LCPS-CP: LCPS-CP, low-complexity protein segments containing phosphonate group. Frontiers in Bioengineering and Biotechnology frontiersin.org make SAP-TA . Polyphenol groups in TA can grab Ca 2+ and trigger HA crystal renewal. Iron ions work in tandem with SAP-TA to generate a thick layer that boosts adsorption capacity. Therefore, SAP-TA/Fe (III) can improve the adhesion and mechanical properties of the remineralization interface (surface microhardness recovery >80%, binding force 64.85 N). Peptide-7 is designed and synthesized with a significant number of carboxyl groups on its side chain to help in firmly interacting with HA and directional elongation of HA crystals (Liu et al., 2018). Under the guidance of Peptide 7, a dense mineralized crystal layer with tight adhesion was formed. Antibacterial peptide inspired bioactive peptides Bioactive peptides have antibacterial and remineralization properties, which can be obtained by grafting units capable of promoting remineralization onto the active sequences of antimicrobial peptides. This bioactive peptide can protect enamel against demineralization while also promoting selfhealing regeneration in a remineralized environment. P-113 is the smallest antibacterial unit of histatin 5, which is a type of natural antimicrobial peptide . A study coupled dopamine (DA), SpSp (DPS) domains, and binding peptide binding peptide SKHKGGKHKGGKHKG on P-113 to find the most cost-effective peptide . The experiment has discovered that P-113-DPS show similar antibacterial effect to Sp-H5 and can kill the majority of Streptococcus mutans (S. mutans) at low concentrations. After a 24-hour remineralization experiment, an 8.5 μm thick needlelike remineralization layer is formed on the enamel surface in the P -113-DPS group, twice as thick as the control group (4 μm) ( Figure 6). The low-complexity protein segments (LCPSs) 37 SYSGYS 42 in the fused in sarcoma protein is capable of forming nucleation structures that form reversible amyloid fibrils (Hughes et al., 2018). LCPSs are highly hydrophilic and structurally flexible. Due to weak multivalent interactions, proteins are entangled and subsequently form web-like structures. LCPSs containing a phosphate or phosphonate group is named LCPS-OP and LCPS-CP. These acidified polypeptides can bind calcium ions and acts as soft templates to induce HA formation. At the same time, the hydrophilic negatively charged peptide coating can reduce the bacterial adhesion of caries-causing bacteria by virtue of the negative electric mutual repulsion (Chang et al., 2022). Given that this bioactive peptide may more effectively repair damaged enamel while inhibiting further erosion of dental cariogenic bacteria, it may be an ideal material for the prevention of dental caries. In addition, some antimicrobial materials have been added to the Frontiers in Bioengineering and Biotechnology frontiersin.org remineralization system to promote enamel remineralization. The first type of materials can cover the enamel surface with an antifouling layer, and the second type can be used to destroy the bacterial biomass through the positive charges. Table 2 summarizes these materials that combine antimicrobial and remineralization functions and Figure 7 shows their modes. Dentin phosphoprotein derived peptide Dentin phosphoprotein, made up of a significant amount of aspartate serine repeat sequences that have a strong affinity for HA, can act as a nucleation template in the process of dentin mineralization (He and George, 2004). Therefore, inspired by dentin phosphoprotein, the peptide 8DSS containing eight DSS repetitions is synthesized. 8DSS can capture Ca 2+ and act as a diffusion barrier that prevent CaP from dissolving. In the repair of both surface cavities and deep lesions, 8DSS demonstrated equal remineralization ability to NaF (Yang et al., 2014;Liang et al., 2015;Zheng et al., 2019). In addition, 8DSS is able to resist hydrolase assault and sustain its action in the mouth due to the short peptide chain length, which is conductive to clinical application. 3.3.5 Self-assembly peptide 11-4 (P 11 -4) Self-assembly peptide P 11 -4 is a well-studied small molecule peptide. When activated by external stimuli, P 11 -4 can selfassemble through intermolecular hydrogen bonds between peptide backbones and form three-dimensional scaffolds in lesions (Alkilzy et al., 2018a). At this time, the negative group formed by 4 Glu-residues on P 11 -4 can attract calcium ions and induce mineralization. According to µCT imaging, the mineralization of the samples treated by P 11 -4 increase by 68% in 14 days (Kind et al., 2017). It is worth noting that P 11 -4 guided remineralization occurs in the subsurface of lesions, which can compensate for the shortcomings of fluoride. As a result, the combination of P 11 -4

and fluoride varnish can produce good results in clinical applications. In an experiment on children over 5 years old with obvious active early caries, P 11 -4 + fluoride varnish are superior to fluoride in terms of vision and safety (Alkilzy et al., 2018b). In some other LCPS-CP Hydrophilic LCPSs eliminate adsorbed biomolecules by forming an anti-sewage ensemble; negatively charged phosphate coatings cause electrostatic repulsion between the bacterial film and the enamel, ultimately reducing adhesion Alendronate Alendronate (ALN), a powerful bone resorption inhibitor with a high affinity for HA, is used to treat and prevent osteoporosis . The phosphate of ALN exchanges with the phosphate of HA in enamel, forming coordination chains that tightly bind it to the enamel surface (Palazzo et al., 2007). Therefore, ALN can act as a "glue" in the mineralization system to increase the adsorption of materials. ALN modified poly (amino amine) dendrimers (Wu et al., 2013) and carboxymethyl chitosan can significantly increase their absorption on the enamel surface. In addition, after forming a HA layer around the ALN-modified polyacrylic acid (PAA), the outer layer of HA is zinc-modified to synthesize ZHA@ALN . Once ZHA@ALN dissolved by acid, a substantial number of calcium, phosphorus, and zinc ions can be released for remineralization and sterilizing. The inner layer of ALN-PAA adheres quickly to the enamel surface due to the ALN. Then, PAA serves as an antifouling layer with 75.05% bacteriostatic efficiency. Polymer materials Polymer materials have complex side groups and spatial structures, which enable them to mimic the enamel matrix and induce mineralization. Some are used to keep ions stable, and some can be made into gel carriers to transport functional peptides while forming a protective layer on the enamel surface. Due to their high biocompatibility and adaptability, polymer materials can be an ideal choice for promoting enamel remineralization. Non-collagenous protein analogs Non-collagenous proteins (NCPs) stabilize crystal precursors during dentin and bone collagen mineralization. NCP analogs, such as polyaspartic acid (pAsp), polyglutamic acid (PGA), and biocompatible polymers polyacrylic acid (PAA), contain a large number of carboxyl groups. These polymers can be used to create induced liquid precursors by stabilizing calcium ions. PGA and pAsp can also bind to calcium ions on the enamel surface, strengthening adhesion and providing nucleation sites (Ustriyana et al., 2020a;Ustriyana et al., 2020b). It has been discovered that the αhelical of pAsp or PGA can promote HA crystal nucleation by templating Ca 2+ distribution. The HPO and -COO-can work Frontiers in Bioengineering and Biotechnology frontiersin.org together to attract Ca 2+ and form stable Ca 2+ triangles, which can develop into the nucleation core of ACP (Zeng et al., 2021). PAA can also chelate with Ca 2+ while maintaining liquid phase stability and transporting ions continuously for subsequent biomineralization Xu et al., 2020;Li N. et al., 2021). Furthermore, PAA can direct the transformation of ACP to form acicular microcrystals . Poly (amino amine) (PAMAM) Poly (amino amine) (PAMAM) is a kind of synthetic protein with a dendritic structure. PAMAM can selfassemble into nanospheres, nanochains, and nanoribbons . Grafting different groups such as -NH 2 , -COOH and -OH onto PAMAM can improve its ability to bind Ca 2+ or promote its adsorption on the enamel surface. The mineralization effects decrease in the order of -NH 2 > -COOH > -OH (Fan et al., 2020). This is because positive charged PAMAM-NH 2 can be more adsorbed on the negatively charged enamel surface. In addition, the adherence of S. mutans is also evaluated. Both PAMAM-COOH and PAMAM-NH 2 are shown to be effective in forming a smooth remineralized layer and minimizing S. mutans adherence (Jia et al., 2020). Grafted with SN15, SN15-PAMAM can increase adsorption on the enamel surface and achieve 90% higher mineralization effect than the control group. . Polydopamine Polydopamine (PDA), the polymer of dopamine that rich in amino and phenolic groups, shows great hydrophilicity. It has been used as a functional agent to increase the wettability and biocompatibility of substrate (Barclay et al., 2017;Ghorbani et al., 2019). After being immersed in PDA solution, a dense film can be formed on the surface of the material in a short time (Kaushik et al., 2020). The film contains a large number of charged groups, to which calcium and phosphorus ions will be attracted and form a stable bond (Ryu et al., 2010;Murari et al., 2020). It is also observed that the HA crystals on the PDA modified enamel surface accumulated more closely, suggesting that PDA might help in inducing uniform crystal nucleation (Zhou et al., 2012). This may be because PDA can increase surface hydrophilicity, decrease the interfacial energy, and accelerate crystallization speed of HA (Qu et al., 2020). Based on the super adhesion of PDA, HA layer can be synthesized on the subsurface of different materials after being modified (Chen et al., 2019;Wong et al., 2022). Biopolymers Biopolymers, including proteins and polysaccharides, have been used for the bionic formation of HA. Most of these polymers are mostly used in mineralized systems in the form of gels. All of these biopolymers show excellent non-immune and biocompatible properties, meanwhile with the advantages of easy storage and clinic application. Chitosan (CS), a cationic polysaccharide, can rarely produce allergic or inflammatory reactions in humans (Younes and Rinaudo, 2015). Therefore, CS has been used to construct organic templates and scaffolds, which can ensure the bioactivity of peptides for mineralization guidance Ren et al., 2018). Furthermore, CS is also able to prevent bacteria from adhering and reproducing. The adherence of S. mutans may be decreased by 94.91% by employing CS alone (Ren et al., 2019). This is because chitosan is positively charged. When CS comes into touch with the negatively charged bacterial wall, the structure of the bacterial wall would be disrupted. Simultaneously, positive charged CS can adhere to the negatively charged enamel surface, preventing acid erosion (He et al., 2019;Boda et al., 2020). In addition, the antibacterial function can be enhanced when CS paired with fluoride (Wassel and Khattab, 2017;Ren et al., 2019). Carboxymethyl chitosan (CMC), formed by CS carboxylation, also has excellent ACP stabilization and can promote the formation of enamel remineralization layers. (Chen et al., 2015;Xiao et al., 2017). The mineralization system using chitosan as a gel carrier is summarized in Table 3. Agarose is a natural polysaccharide with -OH groups that can form a thermally reversible gel. Agarose aqueous gel is widely used in medical systems such as mineral regeneration and drug delivery (Zarrintaj et al., 2018). The abundant hydroxyl groups in agarose molecules have a strong mutual attraction with Ca 2+ , allowing agarose to control the formation of nano ACP precursors and act as an ion reservoir to transport mineral precursors to the enamel surface for mineral mesomorphic transformation. The average elastic modulus and nano hardness of enamel increased significantly to 89.46 ± 11.82 and 3.04 ± 0.75 GPA after 6 days of the interaction of agarose gel with 500 ppm F (Cao et al., 2014a). When chitosan is added to the agarose aqueous gel, the groups between the two gels are cross-linked with each other, forming a fiber structure together with calcium ions, which can further simulate the protein matrix for enamel repair. The regeneration layer is 7.5-8.5 μm thick and regained 77.4% of the natural enamel's microhardness (Musat et al., 2021). Agarose can also be combined with amelogenin to form oriented hexagonal prism enamel columns on the enamel surface (Cao et al., 2014b). Gelatin is a peptide molecular polymer. Gel peptides in gelatin can form salt bonds with phosphate groups on the surface of apatite, causing enamel-like minerals to regenerate. The limited directional diffusion of ions in the gel environment Frontiers in Bioengineering and Biotechnology frontiersin.org promotes heterogeneous nucleation, resulting in a crystal with a clear structure (Zhang et al., 2010). Using the bionic double-layer gel system assisted by anodic aluminum oxide, it is possible to successfully prepare HA crystals with good orientation (Chen et al., 2019). Silk fibroin (SF) governs the synthesis of calcium carbonate in mollusks and the creation of animal shells. SF contains a large number of β-sheets, which are rich in acid aspartic acid and have a high affinity for calcium ions. In the rotary thermal evaporation approach, SF serves as a template to guide heterogeneous nucleation of HA and mineral layers with natural enamel-like shape, organization, and mechanical characteristics swiftly formed . Abalone water-soluble matrix (AWSM) plays an important role in the formation abalone shells. The proportion of organic matter and inorganic minerals in abalone shells is very similar to enamel (95% calcium carbonate and less than 5% organic matter). Therefore, AWSM can promote the formation of crystals. High AWSM concentration can increase the content of calcium and phosphate in the mineralized layer and promote to form a parallel, dense and highly ordered structure (Wen et al., 2016). Other polymers Carboxy betaine (CB) polymers are amphoteric polymers with functional carboxyl and quaternary ammonium groups. The carboxyl groups can serve as Ca 2+ and PO 4 3− deposition sites, whereas the positive quaternary amino group has bactericidal properties . Furthermore, CB can resist bacterial adhesion via electrostatically induced hydration. ACP can be stabilized by amphoteric ionic poly (carboxy betaine acrylamide) (PCBAA). PCBAA/ACP nanocomposites contribute to the growth of HAP in the damaged sublayer of enamel by blocking spontaneous ion conversion on the enamel surface while releasing sufficient ions (He et al., 2022). Thus, PCBAA/ ACP nanocomposites performed admirably in both remineralization (10.08 μm thick remineralized layer in mice intraoral for 14 days) and antimicrobial experiments (almost no bacterial adherence). ACP and poly (vinylpyrrolidone) nanofibers are mixed, and making electrospun mats. The mats can be hydrated to form a gel in the savila containing fluoride. Then, calcium and phosphorus ions crystallize under the guidance of fluorine ions to form HA (Fletcher et al., 2011). Conclusion and perspective Enamel caries have been common problem in our daily life, great efforts have been paid to design new materials and realize the remineralization of enamel. However, it is still a great challenge to repair the defect enamel and restore its functions, as for the emerging materials for enamel remineralization, there is still a long way to go to satisfy the clinic applications. Firstly, most of the current materials used for the remineralization still need a long time, from several days to more than 10 days, secondly, the stability and mechanical properties are not satisfying enough, in addition, most of the remineralization systems depends a lot on the solution or concentration of mineralization medium. Therefore, it is critical important to design advanced materials that can be used in enamel remineralization and solve the clinic problems. Frontiers in Bioengineering and Biotechnology frontiersin.org native mineral structure or composition, should be a charming field, besides, on considering the complicated oral environment, bacterial infections also threaten the treatment of dental health, therefore, materials with multifunction should also be designed and may pave the way of enamel remineralization. In addition, further researches in the remineralization mechanisms are also much important, which may be helpful to direct the design of new materials and their final applications. Author contributions Conceptualization, JW and ZL; writing-original draft preparation, JX, HS, and JL; writing-review and editing, JW, HS, HY, PW; supervision, ZL; project administration, ZL and JW; funding acquisition, ZL and JW. All authors are involved in the revision and approved the submitted manuscript. Conflict of interest The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Publisher's note All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher. LARC methods: entering a new age of contraception and reproductive health Although IUDs and implants have been around for a long time, their use has been severely hampered and almost extinguished for periods of time due to early design flaws, difficult insertion and removal demands, or by the unacceptable side effect profiles. It is all very different now. The currently available LARC methods are easy to use, safe, long lasting, quickly reversible and 20 times more effective

than oral contraceptive pills [1–4] (Fig. ​(Fig.1).1). The LARC methods have high patient acceptability [5], have limited contraindications for use, and are often recommended, in some cases, due to their dramatically improved bleeding control. All of the LARC methods can be inserted right after delivery or abortion and following removal, fertility is rapidly restored. Although the LARC methods have a high up-front cost, most or all of these costs are often covered by a 3rd party. In any case, compared to other options, they are highly cost-effective in the long-term [6–10]. Open in a separate window Fig. 1 Perfect vs Typical Use LARCs Introduction Although IUDs and implants have been around for a long time, their use has been severely hampered and almost extinguished for periods of time due to early design flaws, difficult insertion and removal demands, or by the unacceptable side effect profiles. It is all very different now. The currently available LARC methods are easy to use, safe, long lasting, quickly reversible and 20 times more effective than oral contraceptive pills [1][2][3][4] (Fig. 1). The LARC methods have high patient acceptability [5], have limited contraindications for use, and are often recommended, in some cases, due to their dramatically improved bleeding control. All of the LARC methods can be inserted right after delivery or abortion and following removal, fertility is rapidly restored. Although the LARC methods have a high up-front cost, most or all of these costs are often covered by a 3 rd party. In any case, compared to other options, they are highly cost-effective in the long-term [6][7][8][9][10]. Use of the LARC methods in the U.S. has traditionally trailed dramatically behind Europe [11], Asia and developing countries. Worldwide, the IUD is the currently the most popular means of reversible birth control in the world with 160 million users (2/3 of the users are in China) [12]. The good news is that as education and understanding of these new generations of LARC methods is getting to U.S. consumers, their popularity has had remarkable continual growth [13]. Troubled history of long-acting reversible contraceptive methods IUD history The first IUD was reported from Poland in 1909. This intrauterine ring was modified in 1920-30 by a Germany scientist and separately by a Japanese physician in1934. In 1949 another German scientist who had immigrated to the US developed a stainless steel ring IUD. The U.S. FDA first approved IUDs in 1968 eventually bringing the Lippes Loop, the Tatum-T, The Saf-T-Coil, Gynekoil and Copper 7 to the U.S. market. These IUDs were made with plastic, allowing the IUD to bend for insertion and then regain its shape. These IUDs also had a monofilament nylon string added to facilitate removal. The Lippes Loop, one of the most popular first generation IUDs, was introduced in the mid 1960's. From 1960-1970, 12 million women worldwide had IUDs inserted including 3 million living in the US. By 1970 there were over 17 IUDs in development worldwide. With the discovery that the addition of a copper band to the IUD increased effectiveness, a new generation of IUDs was introduced. Researchers also addressed the bleeding and cramping problems associated with copper containing IUDs with a new wave of technology directed at developing IUDs that released micro-dose progestin. By adding the hormone, these IUDs were noted to decrease cramping and bleeding by an average of 90 %. But the popularity and use of all IUDs was soon to be in serious jeopardy, particularly in the U.S., as a new IUD with a design flaw was already gaining popularity in the U.S. market. The Dalkon Shield, designed with fins to resist expulsion, was introduced in 1968. These fins had the disadvantage of making removal difficult and a super-strong multifilament string was added. By 1974, after over 2.2 million had been sold in the US, A.H. Robins suspended their sales. Reportedly six women had died due to complications of the IUD and thousands more had suffered serious infections. Scientific work eventually discovered a serious design flaw in the multifilament string. Studies using scanning electron microscopy reported that all strings that had been removed from users showed deterioration of the outer sheath and bacterial contamination of the underlying multifilaments [14,15]. This finding is the likely cause of ascending infection seen by so many Dalkon Shield users, resulted from contamination from either an ascending infection up the string or from direct contamination when an IUD was pulled into the endometrium as with an ongoing pregnancy [16,17]. Even though the Dalkon Shield was the only IUD implicated, the ensuing debacle affected the reputation of all IUDs. By 1986, all but 1 IUD (Progestasertmanufactured until 2001) was pulled from the U.S. market. For 2 years the U.S. market for IUDs almost disappeared. In 1988 GynoPharma introduced a new copper IUD, the T-380 A (ParaGard), to the U.S. market. This introduction started the rebound of IUD use in the U.S. A new hormonal IUD (Mirena) was approved by the FDA in 2000 and the smaller 3-year version, Skyla, was approved in 2013. 17 The most recent approval from the FDA was on March 2, 2015 for a new IUD that was designed with affordability in mind. Liletta is very similar to the Mirena although it is now only approved for 3 years of use. Ongoing studies should result in a longer approved use. Due to efforts by public health professionals including the American College of Obstetrics and Gynecology and the Centers for Disease Control and Prevention (CDC), public opinion and use of IUDs has been steadily improving [18][19][20]. Implant history Implants have also had their controversies and challenges. In 1990 the six rods implant (Norplant) was approved by the U.S. FDA. By 2002 Norplant was taken off the market due to limitations of product supplies, issues of coercion in women convicted of drug or child abuse, as well as issues related to difficult removals [21,22]. For the next 4 years there were no implants available in the US. The return of contraceptive implants began in 2006 when the FDA approved the one rod implant system (Implanon). Clinical trials had shown that the average insertion time was only one minute and the average removal time was only 3 minutes. In 2010 Implanon was replaced with the next generation implant, Nexplanon. The newer implant featured an improved, easier to use inserter and the addition of a barium marker on the implant making it more detectable by imaging techniques. The ongoing efforts of large numbers of public health professionals, national organizations, and healthcare providers to educate the public on the safety and efficacy of the LARC methods has benefited both IUDs and the implant [18][19][20][23][24][25][26][27][28]. While the popularity of the implant is still lagging behind the rapid increases seen in IUD use, implant demand is steadily climbing. Updated guidelines now encourage the use of LARCs in all age groups In 2009, LARC methods became first-line options when the American College of Obstetricians and Gynecologists recommended LARC methods for the majority of women [29]. Since then the growing support had been clear and wide-spread. The CDC Eligibility Criteria for Contraceptive Use recommends LARC methods for the majority of women who have their first menses, regardless of whether they have had any pregnancies [11]. The American Academy of Pediatrics recommends LARC methods for adolescents as "prevention is the cornerstone of pediatric practice" [20]. A recently published article in Pediatrics: the Official Journal of the American Academy of Pediatrics stated "We suggest that in response to the improvement in the effectiveness and i.e. IUD and implants ), pediatricians have a special opportunity to prevent unintended pregnancy" [30]. ACOG revised one of their practice guidelines on LARCs in 2012. The new guidelines recommended that sexually active adolescents at high risk for unintended pregnancy should be encouraged to consider LARCs [27]. In its Family Planning Handbook for Providers, WHO recommends the implants and IUDs for women with or without children of any age, including adolescents and women over 40 [28]. In an editorial in the Association for Reproductive Health Professional's Contraceptive Journal, the authors decry the "outdated perceptions about appropriate patient candidates for LARC among health care providers continue to negatively impact their use." An emerging body of research has disproved a number of contraindications to IUC use. Specifically women of any age or parity and those who are postpartum or post first or second trimester abortion are eligible for IUC [31]. Efforts to increase LARC use in young women Public health professionals have recently been developing programs to increase knowledge and use of LARC methods among young women. The Contraceptive CHOICE project was conducted by researchers at Washington University in Saint Louis. These researchers noted the high rates of unintended pregnancies and abortions in the U.S. including 273,105 babies born to teens 15-19 in 2013 [32]. Their goal was to eliminate cost barriers and increase patient access to LARC methods particularly in young women in their region. The CHOICE researchers developed a standardized script that included tiered counseling (most effective methods first) of all reversible methods. The "menu of options" listed hormonal IUD first, followed by the copper IUD and implant, then injections, pills, patch, vaginal ring, condoms and lastly emergency contraception. In their overall cohort 75 % of patients chose LARC methods while 72 % of teens opted for LARC methods. The overall 12 month continuation rate of the LARC methods was 86 % (with the LNG-IUS with the highest continuation rate 87.5 %). All of the other methods had 56-49 % twelve month continuation rate. The satisfaction rate in the overall cohort was 78-85 % in the LARC methods and between 54-44 % for the other method. In the 14-19 year old group, satisfaction was between 74-77 % for the LARC methods and between 31-46 % for the other methods [33,34]. These researchers projected that if the CHOICE model was adopted nationally among all sexually active teens in the U.S., the pregnancy rate of 67.8 per 1,000 teens (in 2008) [33] could be reduced to 29.6. [33,34] Confirmation of this assertion was a recent paper showing reductions in teenage pregnancy and abortion rates in England as LARC usage increases [35]. Overall use According to the Guttmacher Institute, the use of LARC methods has in the U.S. has jumped to 12 %, "the highest ever recorded" [36]. While the overall use of contraceptive use in reproductive-aged women has not changed, the newer data shows an ongoing shift towards the LARC methods. For comparison, in 2002 only 2.4 % of contraceptive users relied on LARC methods and in 2007, this number was 8.5 %. The top methods of choice are the OCPS (26 %) followed by the condom (15 %) and now LARC methods (12 %) [36]. The good news behind this upward trend of LARC use is the accompanying downward slope of unwanted pregnancies in the U.S. and 13 % decline in abortions between 2008 and 2011 [36]. CDC data reporting increased long-term use of LARC methods is coupled with data showing declines in overall births and abortions. The Affordable Care Act (ACA) that guarantees coverage of contraceptives for most women, including LARC methods, had been instrumental in the dramatic increase in LARC use. Protecting this funding source is a priority if this downward trend in unintended pregnancy rates in the US is to continue. Other areas of priority include Title X, Medicaid, and extension of the ACA into all states [36,37]. Worldwide use Worldwide use of contraceptives shows uneven progress. Most developing countries have generally seen consistent growth of contraceptive use, particularly the more effective methods, while in others, such as Nepal and Jordan, the use has leveled off or even fallen slightly. In many of the less and least developed countries there are large unmet needs for family planning [38,39]. It is interesting, however, to compare the use of modern methods, particularly LARC use in countries around the world (compiled in 2013). There is a great deal of variation of contraceptive use worldwide with high unmet needs among young unmarried women especially in less developed poorer countries. Effectiveness LARC methods are around 20 times more effective than any other type of reversible birth control excluding the DMPA injection [1][2][3]. In a 2012 study of over 7400 participants, the failure rates in participants using oral contraceptive pills, birth control

patch, or the vaginal ring was 17-20 times higher than the risk of those using LARC methods. For those under 21 using the pills, patch or ring, the risk of failure was almost twice as high as the older participants. But rates of unintended pregnancy regardless of age remained low in the LARC (and depo-provera group) [1]. Contraindications For common conditions, there are very few contraindications to LARC methods. ACOG endorses the US Medical Eligibility Criteria from the CDC which report the following contraindications. Contraindications to insertion of any IUD is the presence of cervicitis, current chlamydial infection or gonorrhea, distorted uterine cavity, current PID, cervical cancer awaiting treatment, suspicious (for serious disease) unexplained vaginal bleeding, puerperal sepsis, pregnancy, gestational trophoblastic disease, AIDS (category 3 -risks may outweigh benefits), complicated organ transplant (category 3), hepatocellular adenoma or malignant liver tumor (3) [40]. Contraindications for the hormonal IUD are history or current breast cancer, severe decompensated cirrhosis, SLE with positive or unknown antiphospholipid antibodies (category 3). Contraindication for copper IUD insertion is severe thrombocytopenia (category 3), [40] Implants also have few contraindications, including history or current breast cancer, severe decompensated cirrhosis, suspicious for serious cause unexplained vaginal bleeding (category 3 , SLE with positive or unknown antiphospholipid antibodies (category 3), hepatocellular adenoma or malignant liver tumor (category 3) The bleeding pattern after implant insertion is most commonly less bleeding or similar bleeding but in a small percentage of women, heavier, unpredictable bleeding may occur [40]. Importantly, IUDs and Implant can be used in the following conditions [40]: Insertion in presence of inflammatory bowel disease, current history of ischemic heart disease multiple risk factors for arterial CV disease ovarian cancer past history of PID immediate postpartum severe dysmenorrhea vaginitis stroke valvular heart disease fibroids anticoagulation therapy most antiretroviral therapy antimicrobial therapy sickle cell disease DVT/PE established on anticoagulant therapy for 3 months family history DVT/PE major surgery depressive disorders diabetes with or without vascular disease migraines with or without aura HIV infected or high risk for getting HIV The copper IUD can be used in women after breast cancer Counseling: risks and benefits The currently available LARC methods are safe, easy to use, long lasting, quickly reversible and 20 times more effective than oral contraceptive pills [1][2][3][4]. The LARC methods have high patient acceptability, 5 and generally have few contraindications for use [40]. The hormonal IUD is associated with dramatic reductions in bleeding, often amenorrhea, reduced cramping and reduced pain associated with endometriosis. All of the LARC methods can be inserted right after delivery or abortion and following removal, fertility is rapidly restored. Although the LARC methods have a high up-front cost, most or all of these costs are often covered by a 3 rd party. In any case, compared to other options, they are highly cost-effective in the long-term [6][7][8][9][10]. Importantly, modern IUDs do not carry a risk of pelvic infection after the first 20 days after insertion . There is a very small risk of uterine perforation (much less common than in the past due to advanced IUD designs), or expulsion. Heavier bleeding is a side effect of the copper-IUD while a dramatic reduction in menses is a benefit of the hormonal IUD. A small percentage of women with the hormonal IUD may experience progestin side effects such a mood changes or increased acne [42]. Contraceptive implants generally reduce bleeding and a complete cessation of menstrual flow can occur. However, some women may experience irregular and/or increased bleeding. Progestin side effects are not common with the implant but may include increased acne or mood changes [43,44]. Future directions: identifying barriers and increasing LARC Use The preceding discussion has provided evidence that LARC methods are safe, reversible, have high patient acceptability, and very few contraindications. Newer generation LARCs, such as the levonorgestrel containing IUDs, also have non contraceptive benefits. Perhaps most importantly, these methods are 20 times more effective than oral contraceptive pills [1][2][3][4]. Despite the numerous demonstrated advantages of LARCs, uptake remains low in the U.S. and many countries throughout the world (Table 1). In the following section, Future Directions, we will examine the barriers to LARC use and identify current and potential strategies for increasing LARC uptake and adherence. There are multiple barriers to LARC use [45,46]. Some of these barriers are global in nature and some of them are unique to specific populations such as post-partum women, minority and low-income women or geographic locations which are resource limited [47,48]. A knowledge deficit about LARCs exists among many healthcare professionals. This lack of education and training may result in significant barriers to LARC access that are linked to the provider or pharmacist [49]. In a recent American College of Obstetricians and Gynecologists Practice Bulletin some of the primary barriers to LARC use cited included cost, provider experience, and interest, as well as patient interest [50] (Table 2). Cost Women living in poverty have higher rates of unintended pregnancy as well as abortion [51,52]. The CHOICE study clearly demonstrated that unintended pregnancy rates can be dramatically reduced by eliminating financial barriers. When cost is not a factor this study found that 75 % of eligible women chose a LARC method [53]. Although multiple studies have shown that investment in contraception leads to significant costsavings, access to LARC methods by poor women remains challenging. In a frequently cited publication by Trussell et al, a decision model was used to evaluate cost of contraceptive methods from a payer perspective. The study results demonstrated that the use of any contraceptive method resulted in cost-savings compared to use of no method. This included LARCs, which despite a higher upfront cost, are more cost effective as a result of their high efficacy [54]. Foster et al provided additional support for the findings of Trussell and colleagues in their study of a California Medicaid amendment. The California Family Planning, Access, Care, and Treatment (Family PACT), is a Medicaid State Plan Amendment that serves more than 1.8 million clients per year at or below 200 % of the federal poverty level. The authors found that public cost-savings for each dollar spent on contraception ranged from $1.58 for barrier methods to approximately $5 for LARC methods. Short-term hormonal methods and DMPA demonstrated intermediate costsavings resulting in cost savings from $2.12 for the patch to $4.00 with DMPA, Most significantly this study demonstrated use of LARC methods is costeffective even if methods are not used for their full durations of efficacy [55]. Provider interest, knowledge and training Multiple authors have shown that provider beliefs and practices pose a significant barrier to the widespread use of LARCs. An ACOG supported survey of fellows demonstrated that even when providers consider it appropriate to provide LARCs, a much smaller percentage of Table 2 Barriers to LARC Use providers actually offer them in practice. The vast majority of obstetrician gynecologists offer IUDs (95.8 %), but a majority require two or more visits, which is a potential barrier to more wide spread use. Although 67 % of respondents felt it was appropriate to offer IUDs after spontaneous or therapeutic abortion, only 10.9 % offered them to patients. Nearly half (43.5 %) of fellows surveyed thought it was appropriate to offer IUDs in the immediate postpartum period, only 7.2 % offered this method to patients (Table 3) [56]. Education and training is a significant predictor of the use of LARCs. A lack of training, particularly training during residency, is an important barrier to the use of these contraceptives. Nearly a third of obstetrician gynecologists who were surveyed reported a lack of insertion training as a barrier to the use of contraceptive implants. Although a large majority (92 %), reported IUD insertion training during residency, only 50 % reported training on implants. Continuing education within the past two years was, in fact, the best predictor of implant provision. This is critically important because as the authors point out, "clinicians are gatekeepers of LARC services" [56]. Patient interest Patient interest is a critical factor in increasing LARC uptake. The CHOICE study provided clear evidence that patient interest in LARCs can be dramatically increased with education and reduction of economic barriers. 1 When obstetrician gynecologists were surveyed about the use of contraceptive implants, 46 % cited lack of patient interest as the reason for not offering this method [56]. In a study of predictors of LARC use among unmarried young adults, women with high IUD knowledge were six times more likely to be current LARC users (OR 6.3). High IUD knowledge was the strongest predictor of LARC use in the adjusted model [57]. Age is also an important predictor of interest in LARCs. Women ages 18-19 are less likely to report current LARC use compared with 25-29 year olds (OR 0.1 confidence interval 0.02-0.4) [57]. This statistic highlights the important role of adolescent health care providers in encouraging the use of LARCs in younger women. In a study of adolescent health care providers residency training in obstetrics and gynecology or family medicine was the strongest predictor of LARC provision, especially IUDs [58]. Education and training in LARC procedures appears to be an important factor in the provider's role in encouraging interest in LARCs. Provider knowledge appears to be a prerequisite for patient education about LARCs [58]. Strategies for increasing LARC use The increased uptake of LARCs has the potential to be an important strategy to decrease the unintended pregnancy rate. Multiple clinical studies have indicated that enhanced access to these methods has a dramatic impact [1,2,7]. In the following section strategies to increase LARC access and use will be discussed. Reducing cost An important consideration in evaluating the cost of LARCs is documentation of the cost savings associated with these methods, and not just the upfront cost of the contraceptive. In a recent article in the N.Y. Times the author examined the remarkable success of the Colorado state effort to reduce teen pregnancies. The article reported that the state health department estimated that every dollar spent on the long-acting birth control initiative saved $5.85 for the state's Medicaid program, which covers more than three-quarters of teenage pregnancies and births [7]. Multiple scientific manuscripts have also documented the cost-effectiveness of these methods [54,55]. Cost analysis studies that have assessed the cost benefit of contraception have nearly universally found that contraception, particularly highly effective contraception, is cost effective. Studies by the Guttmacher Institute found that for every dollar spent on family planning services, $1.30 would be saved on maternal and newborn health care [59]. It is critical that part of the advocacy process for LARCs includes education of those involved in healthcare related legislation and healthcare services. Another important strategy is the reduction of cost through large scale purchases. The Department of Defense healthcare services and large HMOs such as Kaiser Permante, as well as non-governmental organizations (NGOs) such as the Bill and Melinda Gates Foundation illustrate the potential to reduce LARC costs when these drugs and devices are bought in bulk. More than 200 million patients in the U.S. receive their health care through managed care organizations. This statistics illustrates the important role of managed care pharmacies in securing lower drug and device prices. HMOs have the ability to negotiate volume discounts, which results in lower drug expenditures and greater profits for the HMOs. These cost savings can be achieved in a Table 3 Provider Beliefs and Practices number of ways including, discount off invoice or bulk discounts, and rebate agreements [60]. Prescription drug coverage is one of the ten essential benefits required by the Affordable Health Care Act. Because the ACA makes drug coverage a core part of health insurance, it eliminates the insurers' ability to tack on a prescription drug benefit plan to a health care plan at an additional cost. As a part of preventative care, prescription birth control is now free if generic, and available through a co-pay if brand name [61]. Increasing provider interest, knowledge and training Although the majority of gynecologists have training in IUD insertion, a smaller percentage have training in implant insertion, and many who have training do not have adequate knowledge of patient eligibility [62,63]. The American College of Obstetricians and Gynecologists (ACOG) has created a number of programs to enhance provider education. ACOG provides a list of LARC Clinical Training Opportunities as a LARC Slide Set and

a LARC for Adolescents Slide Set. The organization also provides a list of Family Planning Speakers with special expertise in LARCs [62]. Family physicians also provide a great deal of the contraceptive counseling and provision in the U.S. Although the vast majority of family physicians believe patients are receptive to learning about IUDs, one study found that less than half offer counseling or the method. Both gynecologists and family physicians were found to have inadequate knowledge of IUD eligibility as gauged by the CDC and Prevention Medical Eligibility Criteria for contraception. Family physicians did report an interest in updating contraceptive skills. There is clearly an opportunity to increase LARC uptake through training and education of physician providers [63]. Nurse practitioners are often the primary providers of contraception for a number of women. It is important that women's health nurse practitioners are trained not only in contraceptive counseling, but also in IUD insertion. In a study of LARC counseling and provision by women's health nurse practitioners, two thirds (66 %), were trained in IUD insertion. This compared to only 12 % of primary care nurse practitioners. Contraceptive counseling, however, included IUDs in 43 % of cases. Nurse practitioners were found to use overly restrictive patient eligibility criteria which was inconsistent with the CDC guidelines. Both insertion training and knowledge of patient eligibility were significantly associated with IUD provision. Contraceptive implant provision was also low, with only 42 % of NPs in women's health and 10 % of primary care NPs providing implants to their patients [64]. Increasing training and education for nurse practitioners who provide contraception will play a critical role in increasing LARC usage as healthcare reforms focused on affordable primary care are put into practice. Pharmacists are becoming important providers of contraceptive services, but they are often not considered as advocates for increasing LARC uptake. For a number of years pharmacists have provided emergency contraception. Now that emergency contraception is available over the counter, patients are more likely to look to pharmacists for medical advice. Women who seek emergency contraception provide an opportunity for pharmacists to provide LARC education and referrals. Pharmacists, as providers of emergency contraception, are well positioned to intervene with patients at high risk for unintended pregnancy. Both the American College of Clinical Pharmacy (ACCP) and the Women's Health Practice and Research Network (PRN) advocate an expanded role for pharmacists in advocating for and facilitating the use of LARC methods [65]. Increasing patient interest The CHOICE study is a model of increasing patient interest and LARC use through patient education. Through the development of a standardized script that included tiered counseling [most effective methods first] of all reversible methods, LARC usage in this study was markedly increased. The investigators found 75 % of patients chose LARC methods while 72 % of teens opted for LARC methods. The overall 12 month continuation rate of the LARC methods was very high at 86 %. The satisfaction rate was also high at 78-85 % in the LARC methods compared to 54-44 % for the other methods. Among adolescents the satisfaction rate was between 74-77 % for the LARC methods and between 31-46 % for the other methods [33,34]. To successfully increase LARC uptake outside the setting of a clinical trial, other strategies for increasing patient education and interest must be identified. In a study from the United Kingdom examining the etiology of low uptake of LARCs, the authors identified dissemination of information in multiple venues such as health centers, schools, peer education, as well as the use of multiple media forms. Increasing primary healthcare nurses' role in contraceptive counseling and provision was also considered an important strategy [65]. Conclusions The safety, convenience and ability of LARC use to impact high rates of unintended pregnancy have been documented in numerous clinical trials as well as in health care settings that provide increased access to these methods [1,2,7]. The challenges to increase LARC uptake are many and include decreasing costs, insuring easy access to training, providing increased patient knowledge and encouraging patient interest. Overcoming these barriers requires a multifaceted strategy which involves patients, providers, healthcare administrators, legislators and the community. GluN2A/B ratio elevation induced by cortical spreading depression: electrophysiological and quantitative studies of the hippocampus Cortical spreading depression (CSD), an underlying mechanism of migraine aura, propagates to the hippocampus, and might explain hippocampusassociated symptoms during migraine attack. We hypothesised that this process is, some parts, mediated by NMDA receptors. By using a rat model, CSD was elicited by solid KCl for 45 minutes prior to electrophysiological and quantitative analyses. The result from electrophysiological study was the ratio of glutamate NMDA receptor 2A and 2B subunits (GluN2A/B). Total NMDA receptor response was isolated using an AMPA antagonist, prior to a GluN2B receptor antagonist. The GluN2A/B ratio was calculated by dividing the remaining NMDA-mediated field-excitatory synaptic potentials (fEPSP) with the subtracted difference of NMDAmediated fEPSP. Western blot analysis of the hippocampus was performed to confirm the quantitative change of GluN2A/B ratio. In hippocampal slice study (n = 12), the GluN2A/B ratio of hippocampal fEPSP was significantly increased in CSD group. Western blot analysis (n = 30) revealed an increase in GluN2A subunits and a decrease in GluN2B subunits in the hippocampus ipsilateral to the CSD induction. Our current study revealed that GluN2A/B ratio was shown to be elevated following CSD stimulation by increasing the total number of GluN2A while reducing the total number of GluN2B subunits. This ratio was demonstrated to be associated with synaptic plasticity of the hippocampus in numerous studies. In conclusion, we showed that CSD increased GluN2A/B ratio, in turn, would result in altered synaptic plasticity. Our findings provide a probable implication on the correlation of migraine aura and hippocampusassociated symptoms. Introduction Various cerebral insults (e.g. epileptic crises, trauma, ischemia, haemorrhage, and migraine) were shown to produce a transient disturbance in cortical activity, socalled cortical spreading depression (CSD). Cortical spreading depression is caused by massive redistribution of ions, particularly potassium and hydrogen ions between intracellular and extracellular compartments [1] resulting in cortical depolarisation that can spread to the adjacent areas. By adopting a model of migraine with aura, spreading depression (SD) in each cortical area is accountable for different clinical manifestations seen in the patients. For instance, CSD in the occipital cortex can cause visual metamorphopsia [2,3], whereas those occur in the somatosensory cortex result in paraesthesia or hemi-anaesthesia [4]. Spreading depression propagated to the hippocampus is believed to cause amnesia, emotional and behavioural changes (e.g. hyperactivity, yawning) during migraine attack [5]. In case of acute amnesia, it seems to be transient (4-8 hours in duration) and may impair both anterograde and retrograde memory. Long-term association of migraine and amnesia was also demonstrated in a retrospective cohort study suggesting that migraine is a risk factor of developing transient global amnesia (TGA) [6]. For decades, Olesan and Jorgensen proposed that the association between migraine and TGA may be explained by presence of spreading depression in the hippocampus [7]. The hypothesis was later proven by groups of ex vivo and in vitro experiments showing that CSD propagated to the hippocampus. Limited studies were published regarding the presence of SD in the in vivo hippocampus following CSD except in familial hemiplegic migraine type 1 (FHM1) mutant mice [8]. Although the molecular mechanisms underlying the correlation of migraine and TGA are still unclear, existing evidence suggested that the process may involve actions of glutamatergic receptors [9]. Glutamatergic transmission is known to play an important role in inducing plastic change in the hippocampal synapse. Repetitive activation of the hippocampal synapse can result in a long-lasting change in synaptic activity known as long-term potentiation (LTP). This process is an important step in the registration and consolidation of new memories. Various classes of glutamatergic receptors are involved in LTP GHYHORSPHQW VSHFLILFDOO\ Į-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and N-methyl-Daspartic acid (NMDA) receptors. Previously, our group showed that CSD significantly reduced LTP magnitude by decreasing post-synaptic AMPA receptor response [9]. However, the effect of CSD on hippocampal NMDA receptor activity were not demonstrated. Differential subunits of NMDA receptors mainly detected in the hippocampus are GluN2A and GluN2B subunits [10]. Thus, these receptors are of our particular interest. Additional studies have demonstrated that GluN2A/B response ratio is strongly associated with synaptic plasticity by modifying LTP induction threshold [11][12][13]. In this study, we aimed to demonstrate the changes in synaptic transmission of NMDA receptors. We identified existence of SD observed in the rat hippocampus and compared their differences in electrical properties with the original CSD. Sequential changes in NMDA receptor activity were reported in terms of GluN2A/B response ratio. Quantitative assays of GluN2A and GluN2B subunits were also performed using Western blot analysis. The findings of this study may imply a clinical correlation between migraine and hippocampusassociated symptoms. The animals were acclimatised to the housing facility for at least seven days prior to the experiments. The study was conducted according to the guideline for experimental animals suggested by the National Research Council of Thailand. The study protocol was approved by the Ethics Committee of the Faculty of Medicine, Chulalongkorn University (No.012/2553). Animal preparations Each rat was anaesthetised with 60 mg/kg of sodium pentobarbital (Ceva Sante Animale, Libourne, France) intraperitoneally. We avoided using inhaled isoflurane or IV dexmedetomidine as surgical anaesthetics due to their effects on suppressing CSD frequency [14]. Physiological parameters were monitored and only animals that were in stable condition throughout the preparation were included in the experiment. All anatomical landmarks guided by Paxinos & Watson's brain atlas [15]. CSD induction The rat's head was fitted to a stereotaxic apparatus (Narishige, Tokyo, Japan). After the right parietal bone had been exposed, a 2-mm craniotomy was performed at 6 mm posterior to the bregma and 2 mm lateral to the sagittal suture using an electric dental driller (NSK, Tokyo, Japan). Since propagation of CSD into the hippocampus usually occurred under hyper-excitable conditions, induced either pharmacologically or genetically, increased dose of KCl was employed in our study. For CSD induction, 3 mg of solid KCl (Sigma-Aldrich, St. Louis, MO, USA) was topically applied onto the dura mater for 45 minutes. In control rats, 3 mg of solid NaCl (Merck, Darmstadt, Germany) was used instead of solid KCl. In vivo cortical DC recording For the DC recording (n = 4), a 2-mm diameter craniotomy was performed in the right frontal bone (from bregma: anterior-posterior, +3 mm; lateral, 2 mm; and dorsal-ventral, 0.5 mm). A recording glass microelectrode for detecting the DC potential was inserted into the frontal neocortex to a depth of 500 μm. Analogue data were converted into digital format using. The data were then analysed using MP100 (Biopac Systems Inc., Goleta, CA, USA) and AcqKnowledge acquisition software (Biopac Systems Inc., Goleta, CA, USA). The measured variables included the area under the curve (AUC) and the amplitude of each CSD wave as well as the number of CSD waves that occurred within a 45-minute period. In vivo hippocampal DC recording In a separated set of experiments (n = 4), the DC potential was recorded in the CA1 region of the hippocampus instead of the neocortex (from bregma: anterior-posterior, -4 mm; lateral, 2 mm). A recording glass microelectrode was inserted with a depth of 2.2 mm with the aid of rat brain atlas [15] and previous electrophysiological study [16]. The histological position of the electrode was confirmed microscopically. Hippocampal slice preparation After 45 minutes of CSD stimulation using solid KCl, the animals (n = 6, each group) were decapitated and their ipsilateral hippocampal tissues were entirely dissected. These tissues were then quickly loaded and sectioned using a Vibratome tissue slicer (Vibratome, Richmond, IL, USA). The tissues were processed in cooled artificial CSF solution (119 mM NaCl, 26.2 mM NaHCO 3 , 11 mM glucose, 2.5 mM KCl, 2.5 mM CaCl 2 , 1.3 mM MgSO 4 , 1.0 mM NaH 2 PO 4 , 0.1 mM picrotoxin; a GABA A receptor antagonist), bubbled with carbogen (95 % O 2 , 5 % CO 2 ). Fresh slices were moved to a humidified interface-type holding chamber and recovered for at least 1.5 hours prior to the performance of the electrophysiological study. Electrophysiological recording A continuation between the CA1 and CA3 region was terminated in order to prevent epileptiform activity originating from the CA3 region. A bipolar tungstenstimulating electrode was placed in the Schaffer collaterals

to evoke a postsynaptic response by delivering a square-pulse stimulus at 0.1 Hz for 0.2 msec. Discharges from presynaptic fibres followed by fEPSPs were recorded using a glass microelectrode. Only slices that produced fEPSP amplitudes of more than 1 mV and were stable for at least 15 minutes were included in this study. Response ratio of GluN2A/B After stable baseline fEPSPs were recorded for at least 15 minutes, NMDA receptor-mediated fEPSPs were isolated by bath application of a potent AMPA receptor antagonist, 6-cyano-7-nitroquinoxaline-2,3-dione; CNQX (10 μM in 0.1 % DMSO; Tocris Bioscience, Bristol, UK) to exclude signals from AMPA components. Ten minutes after application of the drug, input stimulation was delivered at 0.033 Hz and its intensity was adjusted to evoke stable NMDA receptor-mediated fEPSPs. Ten minutes later, the GluN2A component of NMDA receptor-mediated fEPSPs was isolated by bath application of CNQX (10 μM) and GluN2B subunitselective NMDA receptor antagonist, ifenprodil (3 μM in 0.1 % DMSO; Tocris Bioscience, Bristol, UK) for 1 hour. Total component of NMDA receptor responses was the averaged AUC of the NMDA receptor-mediated fEPSPs during 10 minutes prior to ifenprodil application. The amplitude of NMDA receptor-mediated fEPSPs was normalised in the range of 0.5-1.5 mV. GluN2A component of NMDA receptor responses was the averaged AUC of the NMDA receptor-mediated fEPSPs during 50-60 minutes after ifenprodil application (i.e., ifenprodil-insensitive component). Western blot analysis Another set of adult male Wistar rats was divided into control and CSD group (n = 15, each group). Protocols for CSD stimulation using 3 mg of solid KCl for 45 minutes, combined with solid NaCl in control groups, were repeated. After 45 minutes of SD stimulation by solid KCl, the isolated hippocampi were then homogenised in a solution containing RIPA buffer (lysis buffer; 150 mM NaCl, 20 mM Tris-HCl, 2 mM EDTA, 1 % Triton X-100, 0.05 % SDS, 1 mM PMSF, pH 8; Cell Signaling Technology, Table 1 Comparison between CSD and Hippocampal SD Total number of spreading depression and amplitude were significantly reduced in the hippocampus compared to those in the cortex, while wave interval was significantly elevated in hippocampal SD. However, there was no significant changes in duration and AUC. Parameters CSD Hippocampal SD P-(n = 4) (n = 4) value Membranes that were intended to determine the quantity of GluN2A were blocked with 5 % bovine serum albumin (BSA) prior to incubation with the primary antibody in 5 % BSA at 4ºC overnight and incubated with horseradish peroxidase (HRP)-conjugated secondary antibody (1:10,000 dilution; anti-rabbit IgG antibody; Sigma-Aldrich, St. Louis, MO, USA) for 1 hour. For GluN2B detection, the membranes were blocked with 5 % TBST-MLK at room temperature for 1 hour and incubated with a primary antibody in 5 % TBST-MLK at 4 ºC overnight, and incubated with a HRP-conjugated secondary antibody (1:10,000 dilution; anti-mouse IgG antibody; Sigma-Aldrich, St. Louis, MO, USA) for 1 hour. The dilutions of the primary antibodies were 1:500 for both GluN2A (rabbit monoclonal antibody; Millipore, Billerica, MA, USA) and GluN2B (mouse monoclonal antibody; Millipore, Billerica, MA, USA) and 1:2,000 for ȕ-actin (mouse monoclonal antibody; Sigma-Aldrich, St. Louis, MO, USA). Protein bands were sequentially detected using enhanced chemiluminescent (ECL) reagents (GE Healthcare Life Science, Little Chalfont, UK) exposed onto a hyperfilm (GE Healthcare Life Science, Little Chalfont, UK). The quantity of GluN2A and GluN2B were eventually measured using Image J software (NIH, Bethesda, MD, USA). These signals were QRUPDOLVHG DJDLQVW ȕ-actin bands. Statistical analysis All data were reported in the format of mean ± standard error of the mean (SEM). Statistical analysis was performed using IBM SPSS software version 20. Independent sample t-test was adopted in the analyses to establish a statistical correlation. Only probability values less than 0.05 (P < 0.05) were considered to be statistically significant. In vivo cortical and hippocampal DC recordings Our data indicated that CSD propagated and reached the CA1 area of the hippocampus in an in vivo model with alteration of electrical properties. Several parameters characterising electrical properties of CSD and SD measured at the hippocampus (e.g. total number of SD waves, amplitude, duration, AUC, and wave interval) are displayed in Table 1. In the cortical DC recording, the results showed that multiple shifts of negative DC characterised as SD were detected in the frontal neocortex in all rats (n = 4) which indicated that solid KCl application consistently induced multiple waves of CSD (Fig. 1A). In hippocampal DC recording, we illustrated that a series of negative DC potentials characterised as SD was detected in the hippocampal CA1 in all rats (n = 4; Fig. 1B). Ratio of GluN2A/B response Analysis of our current study revealed that GluN2A/B ratio of the CSD group significantly increased in comparison with the control group. After application of ifenprodil, NMDA receptor-mediated fEPSPs were reduced in both control and CSD slices (Fig. 2A). The reduced magnitudes of NMDA receptor-mediated fEPSPs in CSD and control slices were 23.5 ± 2.1 and 34.4 ± 5.5 %, respectively (P = 0.092; independent samples two-tailed t-test). However, the extent of this reduction was not significantly different between the CSD and control groups. In control slices, ifenprodil decreased NMDA receptor-mediated fEPSPs on the hippocampal CA1 by a degree comparable to other studies [17]. The GluN2A/B ratios in CSD and control groups were 3.376 ± 0.361 and 1.968 ± 0.346, respectively (P = 0.018; n = 6 each group; independent . These tracings illustrated that SDs originated from the neocortex were shown to be able to spread to the hippocampus. The waves also appeared to be morphologically different between CSD and hippocampal SD. Recorded parameters were previously described in Table 1. samples two-tailed t-test; Fig. 2B). This result showed that CSD altered the responses of NMDA receptors in the hippocampal CA1 towards greater GluN2A/B response ratio. After treatment with ifenprodil, we also demonstrated that isolated fEPSPs were mediated by NMDA receptors, because the remaining fEPSPs were abolished by the NMDA receptor antagonist, APV (25 μM). Importantly, this GluN2A/B ratio represents the 'response' of GluN2A over GluN2B subunits on the neuronal plasma membrane. Quantitative assay of GluN2A and GluN2B receptors Our results obtained from Western blotting analysis showed that the total number of both ipsilateral GluN2A and GluN2B subunits of the NMDA receptor were significantly altered in CSD groups (n = 15; right KClplaced hippocampus) compared to control groups (n = 15). We observed a significant increase in total number of GluN2A subunits and a reduction of those GluN2B subunits. The averaged intensity of GluN2A protein band UHODWLYH WR ȕ-actin in ipsilateral CSD and control groups was 0.777 ± 0.040 and 0.655 ± 0.044, respectively (P = 0.048; independent samples two-tailed t-test). The DYHUDJHG LQWHQVLW\ RI *OX1% SURWHLQ EDQG UHODWLYH WR ȕ-actin in ipsilateral CSD and control groups was 0.589 ± 0.027 and 0.713 ± 0.024, respectively (P = 0.002; independent samples two-tailed t-test; Fig. 3). We also compared the total number of GluN2A and GluN2B subunits of the NMDA receptor between ipsilateral and contralateral sides of the hippocampus in CSD group. Total number of GluN2A subunits was significantly increased in the ipsilateral side of the hippocampus (P = 0.042; independent samples two-tailed t-test). The averaged values of protein intensity in both ipsilateral and contralateral side of the hippocampus were 0.777 ± 0.040 and 0.650 ± 0.047, respectively. In contrast, a significant reduction in the total number of GluN2B subunits was demonstrated with an averaged band intensity in both ipsilateral and contralateral side of 0.589 ± 0.027 and 0.699 ± 0.031, respectively (P = 0.012; independent samples two-tailed t-test). In addition, we compared a number of GluN2A and GluN2B subunits in contralateral CSD-induced hippocampi with those control hippocampi. These results were undoubtedly insignificant for both GluN2A and GluN2B subunits (P = 0.826 for GluN2A component; P = 0.805 for GluN2B component; independent samples two-tailed t-test). Discussion Our study demonstrated that induction of CSD resulted in trains of DC shifting, compatible with hippocampal spreading depression. The results are consistent with previous ex vivo studies that CSD was induced by 2 M KCl microinjection [18,19], or 3 mg solid KCl [9]. However, microinjection of 0.5 M KCl was shown not to produce DC shifting in the hippocampus [20]. Another study revealed that single CSD induced by 300 mM KCl topical application resulted in waves of SD in in vivo hippocampus only in FHM1 mutant mice, but not the wild-type mice [8]. These evidence support that propagation of SD from the neocortex into the hippocampus is increased in dose-dependent fashion. According to the DC recordings, wave frequency and amplitude of hippocampal DC waves were diminished, while there were no significant changes in both duration and AUC. The duration of SD refers to how long ion channels remain open to enable prolongation of depolarisation. The AUC is the sum of the amplitude and duration. Wave frequency, which refers to the induction threshold, and amplitude were significantly diminished in hippocampal SD. Possible explanations may lie in anatomical difference [18,21] and conduction sensitivity of the two structures. Based on our electrophysiological studies, we observed an enhanced GluN2A/B ratio secondarily to CSD stimulation. Combining this information with our previous research [9], we pointed that LTP magnitude was significantly reduced in CSD group compared to the control group. These findings suggest that CSD may be able to alter hippocampal synaptic transmission by interfering GluN2A/B response ratio. Some evidence strongly suggest that ratio of GluN2A/B response governs bidirectional modification of LTP induction threshold in the CA1 of hippocampus [23,24]. An increase in GluN2A/B ratio was shown to impair LTP, in which an increase of GluN2A/B ratio by one unit is associated with approximately 9 % reduction of LTP. [12,25] Findings from Western blotting analysis support our electrophysiological result of enhanced GluN2A/B ratio. We demonstrated that the total number of GluN2A subunits of the NMDA receptor was elevated whilst those of GluN2B were significantly diminished in the CSD group. Because CSD was elicited for only 45 minutes, we hypothesised that CSD causes posttranslational modifications to GluN2A and GluN2B proteins rather than interfering with transcriptional processes. Although little is known regarding the precise mechanism by which GluN2A/B ratio alters the plasticity threshold, we propose that it involves the individual properties of both GluN2A and GluN2B subunits. Furthermore, we also showed that CSD may not travel to the contralateral hippocampus, because we observed no The protein band intensities were exposed and normalised in associa-WLRQ WR WKH ȕ-actin bands at 43 kDa. For both GluN2A and GluN2B, the signals derived from contralateral and ipsilateral hippocampi are depicted as well as those for CSD and control group. (B) ProporWLRQV RI SURWHLQ VLJQDO LQWHQVLW\ RYHU WKH ȕ-actin were calculated and demographically presented in comparison with those of CSD and control groups. In CSD group, total number of GluN2A subunits of the NMDA receptor are significantly elevated while those of GluN2B are diminished. These findings are consistent with the result from our electrophysiological study towards greater GluN2A/B response ratio. significant changes of ipsilateral (right) and contralateral (left) in either the total number of individual GluN2A or GluN2B subunits. Evidence from molecular experiments suggests that GluN2B receptors have longer activation duration than GluN2A, which results in a greater Ca 2+ influx [26]; thus, overexpression of GluN2B led to the enhanced LTP in the hippocampus [27]. In addition, activation of GluN2B subunits of the NMDA receptor could generate LTP in GluN2A-knockout mice [23,28] and hippocampal LTP was not observed in GluN2B-knockout mice [29]. Some limitations should be considered. First, since we used CSD as a model, the interpretation of our study may not only be constrained to migraine with aura, since various cerebral insults has been shown to produce CSD. Second, although the knowledge obtained from this study may explain hippocampus-associated symptoms during migraine aura, the behaviour or memory in animals were not evaluated. This, however, are being studied in our further research. Taken together, our study revealed that CSD increased GluN2A/B ratio by modifying the numbers of GluN2A and GluN2B subtypes. Our previous studies [9] showed that repetitive CSD resulted in a reduction of LTP which, in turn, is correlated to impaired memory processes. Thus, it is suggested that increased GluN2A/B ratio is associated with reduced LTP. This physiological finding may be used to imply a temporal correlation between migraine with aura and hippocampus-associated symptoms. However, our study was conducted in animals, whether or not

the possibility of our findings hold true in human remains unanswered. Dominant negative retinoic acid receptor initiates tumor formation in mice Background Retinoic acid suppresses cell growth and promotes cell differentiation, and pharmacological retinoic acid receptor (RAR) activation is anti-tumorigenic. This begs the question of whether chronic physiological RAR activation by endogenous retinoids is likewise anti-tumorigenic. Results To address this question, we generated transgenic mice in which expression of a ligand binding defective dominant negative RARα (RARαG303E) was under the control of the mouse mammary tumor virus (MMTV) promoter. The transgene was expressed in the lymphoid compartment and in the mammary epithelium. Observation of aging mice revealed that transgenic mice, unlike their wild type littermates, developed B cell lymphomas at high penetrance, with a median latency of 40 weeks. MMTV-RARαG303E lymphomas were high grade Pax-5+, surface H+L Ig negative, CD69+ and BCL6- and cytologically and phenotypically resembled human adult high grade (Burkitt's or lymphoblastic) lymphomas. We postulated that mammary tumors might arise after a long latency period as seen in other transgenic models of breast cancer. We tested this idea by transplanting transgenic epithelium into the cleared fat pads of wild type hosts, thus bypassing lymphomagenesis. At 17 months post-transplantation, a metastatic mammary adenocarcinoma developed in one of four transplanted glands whereas no tumors developed in sixteen of sixteen endogenous glands with wild type epithelium. Conclusion These findings suggest that physiological RAR activity may normally suppress B lymphocyte and mammary epithelial cell growth and that global RAR inactivation is sufficient to initiate a stochastic process of tumor development requiring multiple transforming events. Our work makes available to the research community a new animal resource that should prove useful as an experimental model of aggressive sporadic lymphoma in immunologically uncompromised hosts. We anticipate that it may also prove useful as a model of breast cancer. Background Most of the biologic effects of vitamin A are mediated by all-trans retinoic acid (RA) receptors (RARs) encoded by three genes, RARα, β and γ, each of which give rise to multiple receptor isoforms. RARs regulate gene transcription upon heterodimerization with retinoid X (9-cis RA) receptors (RXRs). RARs and RXRs are members of the superfamily of ligand-dependent transcription factors. In the absence of ligand, RXR-RAR represses gene transcription by recruiting corepressors; upon ligand binding, corepressors are displaced and coactivators recruited, resulting in activation of gene transcription [1,2]. RXR-RAR activation regulates cell proliferation, apoptosis, and differentiation, events that figure prominently in cancer [3][4][5][6][7][8]. Pharmacological RXR-RAR activation prevents or slows down tumor formation in rodent models of breast cancer [9,10]. We wondered whether chronic physiological RAR activation by endogenous retinoids might exert a qualitatively similar effect. In other words, we wondered whether the anti-cancer effect of pharmacological retinoid doses might represent an enhancement of chronically ongoing, physiological RAR effects. We hypothesized that this might be the case and thus that inhibition of physiological RAR activation might prompt tumor formation. Studies showing that the cellular retinol-binding protein I and RARβ2 are frequently epigenetically silenced in cancer lent indirect support to our hypothesis [11][12][13][14][15][16][17][18]. More pertinent support came from studies showing that RARα antisense [19] and cellular RA-binding protein I [20] overexpression induced tumor formation in mice. Finally, the leukemogenic effect of the t(15;17) translocation that juxtaposes the promyelocytic leukemia and RARα genes, resulting in inhibition of the transactivation function of RARα [21], also encouraged us to pursue our hypothesis. Because both RARα1 and RARβ2 have been implicated in growth suppression [22,23], we considered studying RARα1/RARβ2 double null mice; however, these die shortly after birth [24], precluding this approach. We therefore addressed our hypothesis using a dominant negative RAR transgenic approach [25]. RARα bearing a G303E point mutation in the ligand binding domain (RARαG303E) binds ligand poorly but retains the ability to heterodimerize with RXR; when transfected into RAsensitive cells, RARαG303E inhibited the transcriptional activity of endogenous RAR by nearly 100% at physiological RA concentrations (1-10 nM) [26]. Moreover, 100fold greater concentrations of RA were required for RARα activation of a reporter gene when RARαG303E and RARα were expressed at a ~1:1 molar ratio [26]. These data and the potent in vivo biological effects of RARαG303E [27,28] led us to choose it as the dominant negative RAR for this study. RARαG303E transgenic mice We generated transgenic mice in which the mouse mammary tumor virus (MMTV) long terminal repeat (LTR) drives the expression of the N-terminally HA-tagged RARαG303E dominant negative mutant developed by Saitou et al. [26] (Fig. 1A). The MMTV-LTR has been widely used to target transgene expression to the mammary epithelium and was chosen in this study because of our particular interest in breast cancer. However, the MMTV-LTR is also known to target transgenes to lymphocytes [29][30][31]. Four female founder mice (#s 17,24,6,39) were obtained as determined by Southern blotting (Fig. 1B shows data for founder 17) or PCR. A hemizygous line was successfully established from founder 17 and maintained over eight generations. The data herein Tissue transgene expression Empty lanes correspond to mice that did not integrate the transgene. (C) RT-PCR analysis showing transgene expression in the spleen and mammary gland of a young, disease-free line 17 TG female (liver as negative control). Signal detected when the reversetranscription step was omitted (-RT) is due to contaminating DNA (see Fig. 2 for DNAse I treatment control). derived from the study of this line and to a lesser extent founder 24 (see Methods for information on founders 6 and 39). We evaluated transgene expression by RT-PCR in mammary tissue, spleen and liver from a transgenic female, taking advantage of the transgene's HA tag to distinguish it from endogenous RARα. Expression was readily detected in both the mammary gland and the spleen but not the liver (Fig. 1C), consistent with previous experience as related above. Tumor formation Based on the transgene expression pattern, we hypothesized that MMTV-RARαG303E mice might be susceptible to lymphoma and/or mammary tumor development. Adult transgenic (TG) female mice (line 17) and wild type (WT) female littermates were examined over time for evidence of tumor development. We found that TG mice developed node-based tumors at high penetrance, with a median latency of 40 weeks; no tumors arose among WT littermates ( Fig. 2A). Histological and molecular analyses established that the tumors were B cell lymphomas (below) and RT-PCR analysis demonstrated that they expressed the transgene (Fig. 2B). Founder 24 developed B cell lymphoma at 33 weeks of age. Thus, lymphoma development was associated with MMTV-RARαG303E expression and occurred in two independent TG lines against a tumor-free background in WT mice, suggesting that it is a product of transgene expression rather than insertional mutagenesis. Lymphomagenesis in line 17 was a reproducible feature over 8 generations. Notably, we did not detect mammary tumors, possibly because the latency period for their formation is longer (see below). Molecular analysis of tumors All lymphomas were high grade B cell lymphomas, predominantly node-based, and characterized by organ infiltration in liver, spinal cord, and muscle. The morphology ranged from immature blasts with small nucleolus, fine chromatin, mild or moderate nuclear pleomorphism, scanty cytoplasm, abundant mitotic activity (Fig. 3, panel A) to slightly larger cells with vesicular chromatin, prominent nucleolus, mild nuclear pleomorphism and often starry sky appearance, this latter morphology consistent with Burkitt's lymphoma [32]. Consistently, lymphomas were Pax5+, B220+, CD43+, AA41+, CD69+ and negative for surface heavy or light chain immunoglobulins, CD5, CD4, CD8, CD23, CD21, BCL6 (Fig. 3, panels B and C, and Fig. 4). In five typical cases, tumor cells were negative for c-MYC, BCL2, OCT-2, p53 and IRF4. In three typical cases tested, tumor cells were negative for MCL-1. The lymphomas that arose in founder 24 were histologically indistinguishable from typical line 17 lymphomas and like these were Pax5+ and BCL6-by immunostaining. One case had a biphasic cellular infiltrate in the spleen, composed of immature lymphoblasts with fine chromatin and monocytoid immature elements, in adjacent separate areas (Fig. 3, panel D). This tumor was composed of immature lymphoma/leukemia (Pax5+, TdT+, CD3-, Mac2-) and myelomonocytic leukemia (Pax5-, TdT-, CD3-, Mac2+) (Fig. 3, panels E and F) [33]. On secondary transplant, only the B cell component grew. Lymphoid system in young transgenic mice is normal Bone marrow, spleen, thymus and peritoneal cells from 3 TG and 3 WT littermates, age 17 weeks, were analyzed for Tumor incidence and transgene expression The relatively low -RT signal is due to contaminating DNA as demonstrated by loss of the signal after DNAse I treatment (last two lanes). Bottom, transgene expression in a tumor from another TG female, using RNA isolated with the less stringent Purescript protocol. The strong -RT signal is lost upon DNAse I treatment. Mammary tumor formation upon epithelial transplantation Whole mount analysis of mammary glands taken from young sexually mature females showed that ductal development was complete in both TG and WT animals, with the ductal tree reaching the confines of the fat pad (Fig. 5A). A similar finding was made by Costa et al. in a related dominant negative RAR transgenic model [34]. Therefore, the fact that we did not observed mammary tumor development within the time scale shown in Fig. 2A cannot be attributed to altered ductal development, nor to our choice of the FVB/N strain since MMTV-driven expression of "classical" oncogenes and other genes is tumorigenic in this strain. In several of these models, some or most tumors do not develop until late in the life of the animal [35][36][37]. We therefore wondered whether mammary tumors might arise in MMTV-RARαG303E mice were it not for the fact that they succumb to lymphomas first. To begin to test this idea, we performed an experiment designed to bypass lymphoma development. We transplanted TG mammary epithelium from young diseasefree females to a WT host whose abdominal mammary glands had been precleared of native epithelium (Fig. 5B). Sham controls confirmed that the endogenous epithelium was fully removed. A total of four transplant recipients were allowed to mate freely and followed over time. Two mice died disease-free and were lost to follow-up. One of the transplanted glands in the remaining two subjects developed a mammary adenocarcinoma with lung metastasis 17 months after transplantation (Fig. 5C). Neither mouse developed lymphoma. In total, a metastatic mammary carcinoma arose in one of four glands transplanted with TG epithelium but in none of sixteen host glands bearing intact WT epithelium. This result was not statistically significant (chi square), which is hardly surprising given the very preliminary nature of the study. However, we were able to document that the mammary carcinoma was transgenic (Fig. 5C), as expected from the fact that i) it arose in a gland that received a TG epithelium transplant and ii) sham controls invariably confirmed that clearing of endogenous WT epithelium was complete. Thus, there is virtually no doubt that the mammary epithelium is susceptible to transformation by RARαG303E. This and the success of our strategy in bypassing lymphoma development suggest that further studies of the susceptibility of MMTV-RARαG303E mice to mammary tumor formation, particularly in combination with other transforming events, is warranted. It will also be worthwhile to test for evidence of mammary hyperplasia in TG versus WT littermates at an early age, i.e., prior to the time when lymphomas begin to appear. As we pursue these studies, it will be of interest to test whether similar proproliferative and antiapoptotic strategies underlie dominant negative RAR transformation of B cells and mammary epithelial cells. For instance, decreased cyclin D1, cdk4 and cyclin E are associated with RA mediated G1 arrest [38]. Discussion Lymphomas in our model were transitional B cell type because of the presence of markers of immaturity (CD43, lack of surface Ig, occasional TdT positivity). They also lacked BCL6, which is otherwise typically present in human germinal center-derived lymphomas, including high grade human Burkitt's [39]. We did not find T cell lymphomas or frank myeloid leukemias. The interesting finding of a mixed B and myelomonocytic tumor in one mouse (Results section) suggests origination of the process in a progenitor retaining the plasticity to be programmed in two usually distinct lineages [40]. The growth of lymphomas is typically supported by the expression of an anti-apoptotic protein [39,41,42]; however, MMTV-RARαG303E lymphomas did not express BCL6, BCL2 or MCL1 (Results section), and preliminary data indicated that BCL-XL is weakly expressed. These findings suggest that a survival mechanism that is independent of these potent inhibitors

of apoptosis wards off cell death in our model. Despite the fact that they do not overexpress c-MYC protein, MMTV-RARαG303E lymphomas resembled, histologically and phenotypically, other types of high grade lymphomas with transitional phenotype, such as the ones driven by deregulated MYC expression [43][44][45][46]. However, in the latter models, most mice succumb to lymphomas at 14-30 weeks of life whereas in our model morbidity began later (27 weeks), suggesting a requirement for additional transforming events. Another difference was the absence of BM pro-B cell hyperplasia and activation, noticed in the Eµ-MYC mice [44]. In fact, healthy MMTV-RARαG303E mice had moderately reduced numbers of pro-B cells in the bone marrow and lacked CD69. Interestingly, Manshouri et al. generated RARα antisense transgenic mice and showed that they develop B and T cell lymphomas [19]. However, whereas in this model tumor development was associated with high BCL2 protein expression [47], our model generated exclusively BCL2 negative B cell lymphomas restricted to a specific window of differentiation, with a homogeneous transitional phenotype, all of high grade. The absence of immunologic abnormalities preceding the lymphoma onset is also a unique aspect of our model. These differences may be explained by differences in experimental design with respect both to the nature of the transgene (RARαG303E as opposed to RARα antisense) and the promoter driving its expression (MMTV-LTR as opposed to the pMAMneo expression vector which confers wider tissue expression). Though the evidence is very preliminary at this stage, our observation that RARαG303E has the potential to transform the mammary epithelium is intriguing. A large scale version of our pilot study is required to determine the magnitude of this potential. Wang et al. recently developed a transgenic model virtually identical to ours (below) and reported increased mammary branching in pubertal transgenic females [48], an observation that may or may not relate to our finding. It is important to note that our observation of mammary transformation by a dominant negative RAR is not without precedent. Thus, early studies by Berard et al. showed that lung and mammary carcinomas develop in MMTV-mRARβ4-like transgenic mice [49]. However, whether this can be attributed to a dominant negative-like activity of mRARβ4 [50] remains unclear since the transcriptional activity of this receptor is promoter context-and response elementdependent [51] and since the altered N-terminus of the mRARβ4-like transgene could conceivably have introduced an artifact. However, the use of RARαG303E is also not without caveat; thus, in K14-RARαG303E transgenic mice, which exhibit a skin phenotype, RARαG303E was shown to act via a pathway that is independent of DNA binding [52]. While this manuscript was in preparation, Wang et al. reported the generation of an apparently identical model to that we have described here, i.e., transgenic MMTV-RARαG303E mice in an FVB/N background [53]. Their mice developed B cell lymphomas with similar kinetics and the same phenotype as we have reported here. All seven transgenic founders they obtained developed B cell lymphoma, thus leaving little doubt that this phenotype is a result of transgene expression per se and not insertional mutagenesis. Our study and that of Wang et al. provide mutual and independent confirmation of the susceptibility of MMTV-RARαG303E mice to the formation of immature B cell lymphomas. Importantly, their study showed that RA treatment inhibited lymphoma growth, an observation that strongly suggests that RARαG303E acted as a bonafide dominant negative RAR. Our study extends theirs in several important regards: i) we demonstrate through extensive analysis that the transgene does not affect pre-tumor lymphoid differentiation (Results section); ii) we show by immunostaining that the tumors were positive for Pax-5, which is the most faithful and restricted (in the hematopoietic system) lymphoid transcription factor and remarkably evolutionarily conserved in B cells (Fig. 3); iii) we immunostained tumors Histological and immunostaining analysis for a number of relevant proteins including c-MYC, p53, BCL6, BCL2 and MCL1 (Fig. 3 and Results section); iv) we describe an unusual biphenotypic B-myelomonocytic tumor that implies origination of the transformation process in a plastic progenitor cell ( Fig. 3 and Discussion); and v) we provide very early but nevertheless intriguing evidence that RARαG303E can transform the mammary epithelium (Fig. 5). Conclusion Our data suggest that perturbation of physiological RAR activity has major health consequences. Thus, expression of a dominant negative RAR in mouse lymphoid tissue and mammary epithelium resulted in their malignant transformation. This is of potential clinical significance in view of the growing body of evidence showing that epigenetic changes underlie many human malignancies and specifically that epigenetic silencing of the cellular retinol binding protein I and RARβ2 are common events in human lymphoma and breast cancer. Generation and maintenance of transgenic mice All experiments were performed under Institutional Animal Care and Use Committee approval. RARαG303E was excised from pCMX-RAR-E [26] as a Hind III-Bam HI fragment and Klenow filled to generate blunt ends. This fragment was subcloned into MMTV-SV40-Bssk [54] digested with EcoRI and Klenow filled (blunt end ligation). Insert orientation was determined by DNA sequencing. The entire MMTV promoter-RARαG303E fragment was obtained by digestion with Sal I and Spe I (see Fig. 1A). This fragment was gel eluted and purified by ultracentrifugation through CsCl gradient and turned over to the Mouse Genetics Shared Research Facility for pronuclear microinjection (FVB/N background). The data herein was obtained from the study of a hemizygous line established from founder 17; founder 24 died of B cell lymphoma and its tumors were also characterized. Founder 6 transmitted the transgene to its offspring but a hemizygous line was not successfully maintained. Founder 39 died of unknown cause. Line 17 TG males were sterile as judged by their inability to sire litters after multiple attempts. Male sterility was previously described and studied in a related dominant negative RAR model [34] and was likewise observed by Wang et al. [48]. MMTV line 17 was maintained by interbreeding TG females to WT males; the transgene was transmitted in a Mendelian fashion. Genotyping and RT-PCR Tail DNA was digested with KpnI and SacI to release a 503 bp RARαG303E fragment and Southerns probed with RARαG303E cDNA. Endogenous RARα yielded a ~9 kb fragment. PCR genotyping as well as RT-PCR took advantage of the HA tag to distinguish the transgene from WT RARα. For PCR, amplification (25 cycles) was carried out with forward primer (5'-3') TACCCCTACGACGT-GCCCGACTATGCCAGC, reverse primer (5'-3') GTTTCT-CACAGACTCCTTGGACATGCCCAC, and 66°C Flow cytometric analysis annealing temperature. Total RNA was isolated using either the classical, more stringent Trizol method [55] or the Purescript kit from Gentra Systems (Minneapolis, MN). For RT-PCR, 2 µg of total RNA were reverse tran-scribed using random primers and 1/10 th of the cDNA used as PCR input. Amplification was carried out with the same primers and conditions as above. Amplification reactions in which the RT step was omitted were included to control for contaminating DNA. Histology, immunohistochemistry and immunofluorescence Four µm-thick formalin fixed, paraffin embedded sections were stained for H&E or immunostained as published previously [56,57]. Briefly, dewaxed, 0.01 mM EDTA antigenretrieved sections were blocked in 5% defatted milk powder in TBS-TritonX100, incubated overnight with primary antibodies against: rabbit anti BCL6, c-MYC, Bcl2, OCT-2, and IRF4 (Santa Cruz Biotechnology, CA), mouse anti Pax5, rat anti B220, CD138 (BD-Pharmingen), PNA lectin (Vector, Burlingame, CA) and rat anti Mac-2 (Cedarlane, Ontario, Canada), p53 CM5 (Novocastra, Newcastle-U-Tyne, UK). The specificity and reactivity of these reagents on mouse tissue was previously defined. Two rabbit antibodies against MCL-1 (S-19, sc-819 from Santa Cruz and 600-401-394 from Rockland, Gilbertsville, PA), generated against two unrelated epitopes of the molecule, were used to stain tumor tissue. These antibodies reacted with mouse normal germinal center and other selected tissues in an identical manner; the former was used at 0.2 µg/ml for further characterization of the lymphomas. The primary antibodies were counterstained with biotin-or Alkaline Phosphatase (AP) -conjugated secondary antibodies (Vector or Southern Biotechnology Associates, Birmingham, ALA). Biotin-conjugated antibodies received either Avidin-HRP (DAKOUSA, Carpintera, CA) or Avidin-AP. The enzymes were developed with AEC (AminoethylCarbazole, Sigma) or NBT-BCIP (Roche). Double immunohistochemical staining was performed as published previously [58] with non-crossreacting combinations of primary and secondary antibodies. Mammary gland procedures The technique of mammary epithelial transplantation has been described in detail [59]. Briefly, a 14-week old RARαG303E female (donor) and 3 to 4-week old virgin WT females (recipients) were anesthetized, the abdominal mammary glands of the donor surgically exposed, and 2 mm 3 gland fragments excised and placed on sterile PBS on ice. The abdominal mammary glands of the recipients were then surgically exposed and cleared of epithelium by cauterizing the fat pad at a point distal to the mammary lymph node (in 3 to 4-week old virgins the ducts lie proximal to the node, Fig. 5B); two donor fragments were implanted into a pocket prepared in the middle of each cleared fat pad. Finally, skin flaps were sutured with wound clips. Sham controls underwent gland clearing without receiving epithelial implants. Mammary whole mounts were prepared as described [60]. Competing interests The author(s) declare that they have no competing interests. Authors' contributions TSK synthesized and purified the MMTV-RARαG303E construct, designed the protocols for genotyping and RT-PCR, identified and bred the founder mice, generated the data in Figs. 1B and 5A, and contributed in numerous other ways. GC was responsible for all the expert and extensive molecular marker analyses of WT and TG lymphoid tissues (Figs. 3, 4, text and data not shown), and wrote the corresponding parts of the manuscript. CM generated the data in Figs. 1C, 2B and 5C. EFF performed the mammary gland transplantation experiments (Fig. 5B). RT contributed intellectually to all aspects of experimental design and interpretation. RML conceived and supervised the study, generated the data in Fig. 2A, and wrote the manuscript. All authors read and approved the final manuscript. Publish with Bio Med Central and every scientist can read your work free of charge Trends in out-of-hospital cardiac arrest incidence, patient characteristics and survival over 18 years in Perth, Western Australia Objectives To investigate trends in the incidence, characteristics, and survival of out-of-hospital cardiac arrests (OHCA) in the Perth metropolitan area between 2001 and 2018. Methods We calculated the crude incidence rate, age-standardised incidence rate (ASIR) and age- and sex-specific incidence rates (per 100,000 population) for OHCA of presumed cardiac aetiology. ASIRs were calculated using the direct method of standardisation using the 2001 Australian Population standard. Survival was assessed at return of spontaneous circulation at emergency department arrival and at 30 days. Temporal trends in patient and arrest characteristics were assessed with logistic regression, while trends in incidence were assessed using Joinpoint regression. Survival trends were assessed using binary logistic regression. Results A total of 18,417 OHCAs of presumed cardiac aetiology were attended by emergency medical services in Perth between 2001 and 2018. Overall, there were no significant changes in the crude or ASIR of OHCA over the study period, although OHCA incidence in 15–39 year-old males increased by 12.5% annually between 2011 and 2018. Both bystander cardiopulmonary resuscitation and bystander defibrillation increased over the study period, while the proportion of shockable arrests declined. Thirty-day OHCA survival improved significantly over time, with the odds of survival (in bystander-witnessed, initial shockable rhythm arrests) improving 12% (95% CI, 9.0% to 14.0%) annually, from 8.4% in 2001 to 44.0% in 2018. Conclusion Overall, there were no significant trends in OHCA incidence over the study period, although arrests in 15–39 year-old males increased significantly after 2011. There were significant improvements in 30-day survival between 2001 and 2018. Introduction Out-of-hospital cardiac arrest (OHCA) is a global health issue 1 that carries considerable societal and economic costs. 3 A recent study estimated that the annual economic loss to the Australian economy from sudden cardiac arrest was comparable to that from all cancers in Australia combined. 3 With an ageing population 4 it is likely these economic impacts will only be exacerbated in the future. Monitoring temporal trends in OHCA incidence enables the evaluation of the effectiveness of preventative public health strategies, while monitoring survival helps inform the clinical management of OHCAs. Importantly, understanding OHCA trends allows health authorities to better target future health spending. Globally, temporal trends in OHCA incidence have shown considerable variation. 2,5,6 This variation is likely to be a result of regional differences in socioeconomic status 7 and underlying population health characteristics. 8 Encouragingly, recent international studies have generally reported increasing survival over time. 2,[9][10][11] study

aimed to investigate temporal trends in the incidence, patient and arrest characteristics, and 30-day survival of OHCA of presumed cardiac aetiology in the Perth metropolitan area over an 18-year period. Study design We conducted a population-based retrospective cohort study of patients (of all ages) who experienced an OHCA of presumed cardiac aetiology in the Perth metropolitan area and were attended by St John Western Australia (SJ-WA) emergency medical services (EMS) between 1st January 2001 and 31st December 2018. 'Presumed cardiac aetiology' was defined based on the exclusion of other obvious non-cardiac causes e.g., trauma, poisoning, drowning, drug overdose or asphyxia. 12 This study was approved by the Human Research Ethics Committee of Curtin University as a substudy of the Western Australian Pre-hospital Care Record Linkage Project (HR128/2013). Study setting Perth is the capital, and largest city, in the state of WA, with a population of 1.39 million people in 2001 13 and 2.09 million in 2019. 14 The sole provider of road-based EMS in Perth is SJ-WA, which operates a single-tiered advanced life support service, staffed by nationally-registered paramedics. OHCA patients are currently transported to one of ten hospital emergency departments (ED) within the Perth area, 15 unless resuscitative efforts are ceased in the field by paramedics (or not commenced), due to futility or not-forresuscitation orders. Data sources We sourced data from the SJ-WA OHCA Database 16 which contains details for all metropolitan OHCAs attended by SJ-WA EMS since 1996. The database is maintained by the Prehospital, Resuscitation and Emergency Care Research Unit (PRECRU) at Curtin University, extracting data from patient care records (PCR) completed by SJ-WA paramedics, and linked to computer-aided dispatch data. The database contains data on patient demographics, arrest characteristics, EMS response intervals, pre-ambulance care by bystanders, and EMS interventions. The 'Utstein' prognostic variables 12 in the database include patient age and sex, witnessed status (bystander-witnessed; EMS-witnessed; unwitnessed), bystander cardiopulmonary resuscitation (CPR), EMS response time (call answer to arrival on scene), initial cardiac arrest rhythm (as recorded by paramedics), return of spontaneous circulation (ROSC) in the field, and ROSC on arrival at the Emergency Department (ED). Thirty-day survival following OHCA is determined by manual lookup in the WA Death Registry. 17 As a result of industrial action by SJ-WA paramedics in 2008, a number of OHCA cases for that year are thought to be missing/incomplete. We therefore excluded all cases from 2008. Statistical analysis Patient and arrest characteristics of the study cohort were summarised across 3-year time intervals (2001-2003, 2004-2006, 2007-2009 [excluding 2008], 2010-2012, 2013-2015, 2016-2018) as counts and percentages. Patient and arrest characteristics examined were patient age group (0-17, 18-39, 40-64, 65-79, and 80 + years), sex (male/female), arrest locations (public/other), EMS witnessed arrest (yes/no), bystander witnessed (yes/no), bystander CPR (yes/no), automated external defibrillator (AED) shock-given by bystander (yes/no), initial arrest rhythm (ventricular fibrillation/ pulseless ventricular tachycardia (VF/VT), Asystole/pulseless electrical activity (PEA), EMS-attempted resuscitation (yes/no), EMS response time (<10 mins, !10 mins), ROSC at ED (yes/no) and survived 30-days (yes/no). Annual crude, age-and sex-specific, and ASIR were calculated for the Perth metropolitan area and reported per 100,000 person-years. Additionally, ASIRs were determined for cases of EMS-attempted resuscitation, and the Utstein comparator group (bystander-witnessed arrests with initial shockable arrest rhythm). 12 Annual crude incidence rates were calculated as the number of OHCA events for each calendar year divided by the Perth population for that year (obtained from the Australian Bureau of Statistics (ABS)). Age-and sex-specific incidence rates were calculated by sex (male/female) and age group (0-14, 15-39, 40-64, 65-79, and 80 + years), using the number of OHCA cases for each group as the numerator and the Perth metropolitan population corresponding to each respective group as the denominator. ASIR were calculated using annual age-specific incidence rates across 5-year age groups, standardised using the direct method of standardisation to the 2001 Australian population standard. 18 Temporal trends in patient and arrest characteristics were assessed using logistic regression for dichotomous variables (e.g. sex), and ordinal logistic regression [after confirmation of proportional odds assumption] 19 for ordinal variables (e.g. patient age group). Temporal trends in OHCA incidence rates (e.g. crude, ageand sex-specific, and age-standardised) were examined using Joinpoint regression and reported as the Annual Percentage Change (APC) in incidence. Joinpoint regression applies a piecewise loglinear model using a permutation test to determine the optimal number of "joinpoints"; 20 allowing it to identify overall trend as well as shorter intra-study period trends (where present). Temporal trends in survival outcome (ROSC at ED, 30-day survival) were assessed using unadjusted and adjusted logistic regression for two patient groups: i) those who received EMS-attempted resuscitation and ii) the Utstein comparator group. 12 Variables included in the adjusted models were: calendar-year of arrest [continuous], patient age (0-17, 18-39, 40-64, 65-79, and 80+ years), sex (female/male), arrest location (public/other), arrest witness status (unwitnessed/bystander witnessed/EMS witnessed), bystander CPR (no/yes), bystander AED shock (no/yes), initial arrest rhythm (Asystole/PEA, VF/VT) and EMS response time (in minutes, from call to scene). The logistic regression model for the Utstein comparator group excluded 'bystander status' and 'initial arrest rhythm' as these are specified in the definition of the sub-group. Analysis was performed using SPSS v26 (IBM Inc., Armonk, NY, USA) and Joinpoint regression version 4.9.0.0 (https://surveillance.cancer.gov/joinpoint/). Results were considered statistically significant for P values < 0.05. Results Between 2001 and 2018 EMS attended 25,561 OHCA cases in the Perth metropolitan area, of which 19,188 (75.1%) were of a presumed cardiac aetiology. After excluding cases from 2008 (n = 755) and those with missing key patient demographic data (n = 16), 18,417 cases were included in our final cohort (Fig. 1). Overall, 1677 (9.1%) patients from our cohort achieved ROSC at ED and 920 (5.0%) survived to 30 days. Of the 8,340 (45.3%) cases who received a EMS-attempted resuscitation, 1599 (19.2%) attained ROSC at ED, and 841 (10.1%) survived to 30 days. Of the 1716 Utstein comparator group arrests (bystander-witnessed arrests with initial shockable rhythm), 585 (34.1%) achieved ROSC at ED and 435 (25.3%) survived to 30-days. Table 1 shows the characteristics of OHCA cases included over the study period, grouped by 3-year calendar intervals. Trends in ROSC at ED The overall proportion of OHCA patients with ROSC at ED increased over the study period, from 3.5% between 2001 and 2003 to 12.1% between 2016 and 2018 (Table 1). Fig. 2 shows the percentage of 30-day OHCA survivors by calendar year as a proportion of; i) all EMS-attempted resuscitations and ii) all Utstein comparator group resuscitations. In the EMS-attempted resuscitation group, the proportion of patients that attained ROSC at ED increased from 5.6% in 2001 to 26.5% in 2018 (p < 0.001), representing a crude improvement of 8.0% (OR 1.08; 95% CI, 1.07-1.10) per annum in the odds of ROSC at ED (Table 2a). In the Utstein comparator group, the propor-tion of patients that attained ROSC at ED increased from 9.5% in 2001 to 48.5% in 2018 (p < 0.001), representing a crude improvement of 12.0% (OR 1.12; 95% CI, 1.10-1.15) per annum in the odds of ROSC at ED. After adjustment for potential confounders (Table 2), the odds of ROSC at ED improved by 11.0% per annum for both the EMS-attempted resuscitation group (OR 1.11; 95% CI, 1.10-1.13) and the Utstein comparator group (OR 1.11; 95% CI, 1.09-1.14). Trends in 30-day survival There was a steady increase in 30-day survival over the study period, with overall 30-day survival in our study cohort increasing from 2.6% (between 2001 and 2003) to 6.7% (between 2016 and 2018). per annum after adjustment for potential confounders (Table 2a). Survival in the Utstein comparator group showed the greatest improvement over time, from 8.4% in 2001 to 44.0% in 2018 (p < 0.001). In this group the odds of surviving 30 days following arrest improved by 12% (OR 1.12; 95% CI, 1.09-1.14) per annum in the unadjusted logistic regression model and by 11% (OR 1.11; 95% CI 1.08-1.14) per annum after adjustment for potential confounders (Table 2b). Fig. 4 shows the overall crude, and age-standardised OHCA incidence rates along with the ASIR of the subgroup analyses (EMSattempted resuscitation, and Utstein comparator group). Supplementary Table 1 show the results of the trend analysis (i.e. Joinpoint Fig. 1). The improvement in the Utstein comparator group survival was particularly noteworthy, as survival in this sub-group of OHCAs is less susceptible to regional variations in EMS policy (with regards to selective resuscitation) or patient/arrest 'case-mix'; thereby providing a more comparable measure of survival trend. We speculate increasing rates of bystander CPR and AED use, coupled with improving EMS response times and provision of high-quality CPR over the study period have acted to increase survival. The benefits of high-quality CPR to OHCA survival has been well established. 21 Since 2000, resuscitation guidelines have progressively placed greater emphasis on optimizing chest compression (in relation to compression depth and rate) and limiting 'hands-off' periods. 22 Interestingly however, after adjusting our logistic regression models for pre/peri-arrest factors, the trends in 30-day survival and ROSC at ED for the Utstein group remained strong, with the odds of both outcomes increasing 11% annually (both were 12% in unadjusted models). This suggests the improvement in survival may have largely resulted from other, unmeasured factors. Ultimately, we were unable to identify the factors responsible for the large increase in 30-day survival. There were no significant trends in either crude or ASIRs between 2001 and 2018. A previous study 23 examining OHCA of presumed cardiac aetiology in Perth did not find a significant trend in crude incidence between 1997 and 2010, consistent with our study. This earlier study did however report a significant decrease in the ASIR, though this decrease appears to have exclusively occurred prior to 2002. A study 24 from Victoria, Australia, reported significant declines in the crude incidence and ASIR of OHCA of presumed cardiac aetiology between 2002 and 2012. However, this Victorian study only included adults (!18 years) and excluded all EMSwitnessed arrests, making direct comparisons with our present study challenging. Bystander CPR Concerningly, our study identified a growing prevalence of OHCA incidence in young males. Between 2011 and 2018, OHCAs in 15-39 year-old males increased at a rate of 12.5% per annum -the largest rate of annual increase of any demographic group. Although we were unable to identify the reason for this increase, a 2012 study 25 reported the most common underlying cause of OHCAs in those between 25 and 35 years of age was coronary artery disease (CAD). We suspect, as others [26][27][28] have suggested, that the rising prevalence of cardiovascular disease risk factors may be partly responsible for this trend. Despite encouraging OHCA survival trends, some trends are a cause for concern. Firstly, we found the proportion of OHCA patients presenting with initial shockable arrest rhythms has been steadily decreasing over time; consistent with findings from other Australian 5,29 and international 30,31 studies. Although the association between initial arrest rhythm and survival is well established, 32,33 the reason for this decreasing trend in shockable rhythms is poorly understood. A 2021 study 34 from the Netherlands suggested that comorbidity burden may be associated with lower odds of a shockable arrest, although this relationship was only found in males. Secondly, we found a decreasing trend in both the proportion of bystander-witnessed arrests and arrests occurring in public loca-tions. Both these factors result in a delay to resuscitative efforts, resulting in reduced survival. Lastly, given the generally negative effect comorbidity has on OHCA survival outcomes, 35 projected increases in population comorbidity 36 could result in lower OHCA survival in the future. Limitations Our study has several limitations. Firstly, our study excluded all data from 2008. However, this exclusion did not impact reported trends as our regression analyses modelled 'time' (i.e. calendar year) as a continues variable. Secondly, our study only included arrests of 'presumed cardiac aetiology'. We chose to focus on arrests of this aetiology as they make up the bulk of arrests in Western Australia (>70%) and therefore represent the greatest burden to healthcare. Thirdly, we defined the geographical boundary of metropolitan Perth according to the 2016 ABS definition. This may have resulted in the inclusion of some cases (prior to 2016) that were considered, at the time of arrest, to be rural. Lastly, our study did not examine survival outcomes beyond 30 days. However, in a prior study 9 we demonstrated that between

1998 and 2017 there was a significant improvement in 10-year survival, relative to the age-and sex-matched general population, of initial (30-day) OHCA survivors in Perth. Conclusion There were no significant trends in the overall incidence of OHCA of presumed cardiac origin in metropolitan Perth between 2001 and 2018, however incidence in 15-39 year-old males increased sharply after 2011. The rates of bystander CPR and the provision of bystander AED shocks increased over the study period, however the proportion of arrests presenting with initial VF/VT rhythms decreased. There was an overall improvement in OHCA survival, with the odds of 30-day survival in the Utstein comparator group (bystander wit- Risk factors for prolonged hypotension in patients with pheochromocytoma undergoing laparoscopic adrenalectomy: a single-center retrospective study Prolonged hypotension during pheochromocytoma resection is a significant complication. We sought to investigate the predictors of prolonged hypotension in patients with pheochromocytoma undergoing laparoscopic adrenalectomy (LA). Patients with pheochromocytoma who underwent LA between 2012 and 2015 were surveyed. Patients were considered to have prolonged hypotension if they had a mean arterial blood pressure <60 mmHg or required ≥30 consecutive minutes of catecholamine support intraoperatively. Among 123 patients, 54 (43.9%) developed prolonged hypotension requiring ≥30 consecutive minutes of catecholamine support. Compared with patients with nonprolonged hypotension, those with prolonged hypotension had higher levels of urinary norepinephrine (P = 0.011), epinephrine (P < 0.001), and dopamine (P = 0.019) preoperatively, and a higher incidence of vital organ injury postoperatively (P = 0.039). Multivariate logistic analysis showed that independent predictors for prolonged hypotension were multiples of the normal reference upper limit value of urinary epinephrine (odds ratio, 1.180; 95% confidence interval, 1.035–1.345) and dopamine (odds ratio, 4.375; 95% confidence interval, 1.207–15.855). The levels of preoperative urinary epinephrine and dopamine are clinical predictors for prolonged hypotension in patients with pheochromocytoma undergoing LA. Using these parameters, clinicians can assess and manage this patient population more effectively. Results Among 123 patients, 54 (43.9%) had intraoperative prolonged hypotension. These patients required continuous catecholamine support to maintain mean arterial blood pressure (MAP) ≥60 mmHg intraoperatively. The median duration of catecholamine infusion was 53 min (range, 32-217 min). The patients demographics and tumor characteristics are shown in Table 1. The fold changes of epinephrine, norepinephrine, and dopamine were greater in patients with prolonged hypotension than in those with nonprolonged hypotension. In addition, a greater proportion of patients with nonprolonged hypotension was asymptomatic, compared with those with prolonged hypotension. A lower proportion of patients with nonprolonged hypotension had diabetes than patients with prolonged hypotension. Table 2 shows that patients with prolonged hypotension required a longer duration of catecholamine support than those with nonprolonged hypotension. As shown in Table 3, patients with prolonged hypotension had a longer intubation time in the intensive care unit and length of hospitalization than those with nonprolonged hypotension. Compared with patients with nonprolonged hypotension, the incidence of postoperative vital organ injury was higher in patients with prolonged hypotension. In the multivariate analysis, multiples of the normal reference upper limit value of epinephrine and dopamine were independent predictors of prolonged hypotension during pheochromocytoma resection, with odds ratios of 1.180 (95% confidence interval, (Table 4). Receiver operating characteristic curve analysis was performed for each variable to identify the optimal cut-off point that correlated with prolonged hypotension during LA for patients with pheochromocytoma. The cut-off points for the multiples of the normal reference upper limit value of epinephrine and dopamine were 0.5 and 0.6, respectively (Fig. 1). Discussion In this study, we found that prolonged hypotension occurred in 54 (43.9%) patients and required continuous catecholamine support to maintain MAP ≥60 mmHg intraoperatively. Our multivariate analysis demonstrated that multiples of the normal reference upper limit value of epinephrine and dopamine were independent predictors of prolonged hypotension during LA. Intraoperative hypotension remains a common complication despite adequate preoperative preparation 6, 7, 12 . Previous reports have shown that intraoperative hypotension occurs in 39-48% of patients with pheochromocytoma 6,7 . In this study, 43.9% of patients needed catecholamine support after pheochromocytoma resection to maintain MAP ≥60 mmHg. This is consistent with the results of previous studies. Managing hypotension can represent a substantial challenge in a subset of cases. Management of hypotension begins with intravascular volume expansion. Continuous vasoconstrictors are also required in patients with severe hypotension during pheochromocytoma resection 6,7 . Generally, chronic hypovolemia and abrupt catecholamine withdrawal are thought to be the main causes of hypotension after pheochromocytoma resection 2, 4 . Furthermore, hypotension can be ascribed to down-regulation of αand β-adrenoceptors resulting from long-term plasma catecholamine elevation 13,14 . In the present study, we found that preoperative urinary epinephrine was associated with prolonged hypotension during LA in patients with pheochromocytoma. Catecholamines exceeding in majority of patients with pheochromocytoma exert their effects by adrenoceptors 15 . Epinephrine and norepinephrine have overlapping but distinct effects on αand β-adrenoceptors in the human body. In low doses, epinephrine acts predominantly on the peripheral β 1 -and β 2 -adrenoceptors. However, with increasing doses of epinephrine, the α 1 -adrenoceptormediated vasoconstrictor effect predominates. Norepinephrine acts predominantly on α-adrenoceptors to induce peripheral vasoconstriction, and has almost no effect on β-adrenoceptors. Although hypertension is the most common symptom in patients with pheochromocytoma, hypotension is frequently seen in patients with epinephrine-secreting tumors even before surgery 16,17 . Furthermore, excessive circulating epinephrine likely decreases cardiac contractility by down-regulating β-adrenoceptors in the heart 16,18 . In a previous study, investigators found that acute left cardiac dysfunction due to chronic elevated concentrations of epinephrine was the root cause of hypotension and circulatory collapse after pheochromocytoma resection 19 . Thus, the depression of cardiac contraction force may be induced, and aggressive volume expansion may increase the workload on the left heart 16,18,19 . According to the underlying actions of epinephrine and norepinephrine, catecholamine infusion improves low cardiac output, especially in patients with epinephrine-secreting pheochromocytoma, although both epinephrine and norepinephrine have been suggested to be related to postresection hypotension 15,16,18,20 . Hypotension after pheochromocytoma resection has been thought to be partially attributed to hypovolemia. However, a prospective study suggested that preoperative fluid therapy had no impact on postresection hypotension 21 that intraoperative hypotension may be largely due to down-regulation of β-adrenoceptors. We consider that the preoperative urinary levels of epinephrine are a robust risk factor for prolonged hypotension in patients with pheochromocytoma undergoing LA. This study also suggests that preoperative urinary dopamine is one of the predictors of prolonged hypotension during laparoscopic tumor resection. Dopamine-secreting tumors are rare compared to norepinephrine-secreting tumors, especially for dopamine-predominant pheochromocytoma. Patients with dopamine-secreting pheochromocytoma tend to be nonspecific and normotensive preoperatively 22 . This phenomenon can be explained by several mechanisms. In particular, dopamine offsets the vasoconstrictor effect of norepinephrine through D1 receptors in a dose-dependent manner. Dopamine suppresses norepinephrine release from neurons by acting on the presynaptic D2 receptors 23 . The same mechanisms may worsen the hypotensive scenario arising from catecholamine withdrawal during LA in patients with pheochromocytoma. It should also be noted that dopamine lacks the α-adrenoceptor affinity for norepinephrine and epinephrine 23 . In dopamine-producing tumors, there is little α-adrenoceptor stimulation excess. Researchers have speculated that the α blockade may lead to hypotension or even cardiovascular collapse 24 . Therefore, preoperative antihypertensive drugs are not recommended in patients with exclusively dopamine-secreting paraganglioma or pheochromocytoma. In this study, 14 patients had dopamine-producing pheochromocytomas, including three patients with exclusively dopamine-secreting tumors. Preoperative preparation was empirically used in all of these patients, and 11 of them, including two cases secreting only dopamine, developed prolonged hypotension. However, this result was not necessarily attributed to the preoperative α-blockade preparation, because dopamine itself was an independent predictor of prolonged hypotension. Furthermore, α and β blockades were empirically prescribed before surgery in one previously reported case with no adverse effects 25 . In another case where preoperative α blockade was not administered, the patient had drastic intraoperative hypertension and hypotensive episodes 26 . Thus, the relationship between preoperative antihypertension medication and intraoperative hypotension may need to be further evaluated in patients with dopamine-secreting tumors. Intraoperative hypotension is common during surgery and may be a significant determinant of postoperative vital organ injury-mainly acute kidney injury and cardiac injury-and other complications 8,9,27 . In this study, patients with prolonged hypotension experienced longer intubation time and length of hospitalization and had a higher incidence of postoperative vital organ injury than patients without prolonged hypotension. Our findings validate the previously published research by Sun et al. 8 and Walsh et al. 9 . Because blood pressure is one of very few adjustable intraoperative risk factors for postoperative adverse outcomes, it is important to prevent and correct hypotension in a timely manner during adrenalectomy in patients with pheochromocytoma, especially in those with epinephrine-and dopamine-secreting tumors. It is noteworthy that most of the patients in this study received phenoxybenzamine for preoperative medical preparation. Non-selective α blockade has been shown to relate with hypotensive episodes after tumor removal owing to the covalent binding of the drug to the α adrenoceptors. This situation may not improve until the α adrenoceptors are regenerated on the cell surface, which may not occur until several hours after tumor resection 28,29 . As a result, non-selective α blockade can lead to persistent intraoperative hypotension requiring fluid therapy and continuous catecholamine infusion, although it works better to control hypertensive crises 30,31 . So far, several studies have confirmed the relationship between phenoxybenzamine and perioperative hypotension in patients with pheochromocytoma [30][31][32] . Consequently, the effect of phenoxybenzamine on intraoperative hypotension in this study can not be ruled out, despite the fact that the ratio of phenoxybenzamine usage did not statistically significantly differ between the two groups. This study has several strengths. We surveyed patients who underwent LA very recently (2012-2015), which allowed for the homogeneous distribution of individual patients' physiology in response to the same anesthetic and surgical techniques. Furthermore, the invasive blood pressure was monitored electronically and recorded during pheochromocytoma resection in our center, which should decrease the potential bias associated with estimating these values. In addition, the relatively large sample size rendered the multivariable modelling stable, thus enabling the detection of associations between preoperative urinary catecholamines and intraoperative prolonged hypotension. However, we also acknowledge that our study has several limitations. First, the retrospective design of our study may affect the quality of the data. Second, intraoperative hemodynamic variables were collected from electronic recording updated every 5 min, which may have limited the accuracy of our counting parameters. Third, the urinary and plasma levels of metanephrine and normetanephrine were not routinely measured in our center. Future research is needed to evaluate the relationship between these two parameters and intraoperative hypotension. Finally, the results of the multivariate analysis pertaining to dopamine should be interpreted with caution, owing to the wide 95% confidence interval originated from the rarity of dopamine-secreting tumors. The relationship between preoperative medication and intraoperative hypotension also needs to be further elucidated, especially in patients with dopamine-secreting pheochromocytoma. Conclusion In conclusion, intraoperative prolonged hypotension is a relatively frequent complication (43.9%) in patients undergoing pheochromocytoma resection. Intraoperative prolonged hypotension is likely to be related to postoperative vital organ injury in pheochromocytoma patients. Furthermore, multiples of the normal reference upper limit value of urinary epinephrine (>0.5) and dopamine (>0.6) are independent predictors of prolonged hypotension in patients with pheochromocytoma undergoing LA. We retrospectively surveyed 127 patients with pheochromocytoma who underwent LA at our center between December 26, 2012, and October 9, 2015. Retroperitoneal laparoscopic surgery was preferred and performed for most patients. Four patients were excluded from the study because of missing data. All patients enrolled were identified as having pheochromocytoma from the postoperative pathology results. Methods In our institution, all patients diagnosed with pheochromocytoma via biochemical tests and imaging examinations were treated with phenoxybenzamine or doxazosin at least 2 weeks preoperatively. Metoprolol was added to control episodes of tachycardia if necessary. During the period of preoperative medication, oral hydration was encouraged, and patients were admitted for intravenous fluids 2 days before tumor resection. The criteria for efficacy include blood pressure <165/80 mmHg 2 days before surgery and standing blood pressure >80/45 mmHg. All patients received general anesthesia for the surgery. Intraoperatively, the invasive blood pressure and heart rate were automatically recorded. Intravenous vasodilators (sodium nitroprusside, phentolamine, and nitroglycerin) and esmolol were administered to control undesirable increases in blood pressure and heart rate, respectively. All patients were transferred to

the intensive care unit postoperatively. Prolonged hypotension during LA was managed using fluid replacement therapy and continuous catecholamine administration. Laboratory and clinical parameters and term definitions. Data on the patient demographics, clinical history, laboratory analysis, and intraoperative details were collected. Patients were considered to have prolonged hypotension if they had a MAP <60 mmHg or required ≥30 consecutive minutes of catecholamine support intraoperatively 33 . Vital organ injury was defined as acute kidney injury and cardiac injury. According to the Acute Kidney Injury Network threshold, patients were diagnosed with acute kidney injury if there was a 50% relative or 0.3 mg/dL absolute increase in creatinine from the preoperative value during the first two postoperative days 34 . Myocardial injury was defined as a postoperative cardiac enzyme concentration within 7 days of surgery that was greater than or equal to the suggested necrosis limit for troponin T and greater than the upper limit of normal for creatinine kinase MB 35 . Statistical analysis. The Mean (±standard deviation) and median (range) values were used to express normally and non-normally distributed continuous data, respectively. Comparisons between groups were conducted with the independent sample t-test for normally distributed variables, Mann-Whitney U test for non-normally distributed variables, and Chi square test for categorical variables. Multivariate logistic regression was used to determine factors associated with prolonged hypotension during LA in patients with pheochromocytoma. Using receiver characteristic operating curve, we calculated the cut-off points for the clinical risk factors associated with prolonged hypotension. Parameters with a P value < 0.10 in the univariate analysis were entered into a multivariate logistic regression analysis. Statistical significance was defined as P < 0.05. All statistical analyses were performed with SPSS version 19.0 (IBM, SPSS, Inc.). ArrayPitope: Automated Analysis of Amino Acid Substitutions for Peptide Microarray-Based Antibody Epitope Mapping Identification of epitopes targeted by antibodies (B cell epitopes) is of critical importance for the development of many diagnostic and therapeutic tools. For clinical usage, such epitopes must be extensively characterized in order to validate specificity and to document potential cross-reactivity. B cell epitopes are typically classified as either linear epitopes, i.e. short consecutive segments from the protein sequence or conformational epitopes adapted through native protein folding. Recent advances in high-density peptide microarrays enable high-throughput, high-resolution identification and characterization of linear B cell epitopes. Using exhaustive amino acid substitution analysis of peptides originating from target antigens, these microarrays can be used to address the specificity of polyclonal antibodies raised against such antigens containing hundreds of epitopes. However, the interpretation of the data provided in such large-scale screenings is far from trivial and in most cases it requires advanced computational and statistical skills. Here, we present an online application for automated identification of linear B cell epitopes, allowing the non-expert user to analyse peptide microarray data. The application takes as input quantitative peptide data of fully or partially substituted overlapping peptides from a given antigen sequence and identifies epitope residues (residues that are significantly affected by substitutions) and visualize the selectivity towards each residue by sequence logo plots. Demonstrating utility, the application was used to identify and address the antibody specificity of 18 linear epitope regions in Human Serum Albumin (HSA), using peptide microarray data consisting of fully substituted peptides spanning the entire sequence of HSA and incubated with polyclonal rabbit anti-HSA (and mouse anti-rabbit-Cy3). The application is made available at: www.cbs.dtu.dk/services/ArrayPitope. Identification of epitopes targeted by antibodies (B cell epitopes) is of critical importance for the development of many diagnostic and therapeutic tools. For clinical usage, such epitopes must be extensively characterized in order to validate specificity and to document potential cross-reactivity. B cell epitopes are typically classified as either linear epitopes, i.e. short consecutive segments from the protein sequence or conformational epitopes adapted through native protein folding. Recent advances in high-density peptide microarrays enable high-throughput, highresolution identification and characterization of linear B cell epitopes. Using exhaustive amino acid substitution analysis of peptides originating from target antigens, these microarrays can be used to address the specificity of polyclonal antibodies raised against such antigens containing hundreds of epitopes. However, the interpretation of the data provided in such largescale screenings is far from trivial and in most cases it requires advanced computational and statistical skills. Here, we present an online application for automated identification of linear B cell epitopes, allowing the non-expert user to analyse peptide microarray data. The application takes as input quantitative peptide data of fully or partially substituted overlapping peptides from a given antigen sequence and identifies epitope residues (residues that are significantly affected by substitutions) and visualize the selectivity towards each residue by sequence logo plots. Demonstrating utility, the application was used to identify and address the antibody specificity of 18 linear epitope regions in Human Serum Albumin (HSA), using peptide microarray data consisting of fully substituted peptides spanning the entire sequence of HSA and incubated with polyclonal rabbit anti-HSA (and mouse anti-rabbit-Cy3). The application is made available at: www.cbs.dtu.dk/services/ArrayPitope. a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 Introduction The highly diverse repertoire of antibodies constitutes a very important component of the immune-mediated protection against pathogens. The exquisite target specificity and high affinity of binding make antibodies attractive tools in scientific, diagnostic and therapeutic applications. Characterization of the specificity of antibodies towards their binding site (epitope) is important for their selection towards intended targets and preventing unintended cross-reactivity [1]. Protein epitopes are usually classified as linear or conformational, depending on whether the amino acids comprised are brought together by proximity in the peptide chain or by protein folding, respectively [2]. The majority of epitopes are thought to be conformational, but the distinction is not clear-cut since conformational epitopes often contain small segments of continuous residues able to bind the antibody on their own [3,4]. Since conformational epitopes rarely maintain readily detectable binding activity outside the context of the native protein structure, characterization of the conformational epitopes can be an extremely difficult task. On the other hand, linear epitopes (and linear segments of conformational epitopes) can be characterized by studying antibody binding to short peptide fragments of the protein. Characterization of the specificity of polyclonal antibodies toward any potential linear epitope within an antigen is challenging. Many different methods including solid phase peptide libraries and phage displayed peptide libraries [5,6] have been used to screen for linear epitopes. Although peptide display systems offer high-throughput identification of linear mimotopes [7] they have biases with regard to certain sequence populations and selection steps [6,8,9]. Using synthetic peptides to map target antigens, the mapping resolution depends on the length and overlap of the analysed peptides as well as subsequent truncations or substitutions, used to fine-map the location of the epitope and the contribution of the individual amino acids to the antibody binding [10]. Most studies based on synthetic peptides involve mapping of native-sequence proteins using overlapping peptides of length 15-30 amino acids [11,12] and some use alanine scans, in which alanine substitutions are introduced in the synthetic peptides to improve mapping resolution [13]. Recent advances in high-density peptide microarrays have enabled parallel synthesis of hundreds of thousands of peptides [10,14]. Two studies have used peptide microarrays to conduct full-resolution epitope mapping using exhaustive single-amino acid substitution analysis [10,15]. These studies use statistical methods to pinpoint which residues of a given peptide epitope are involved in antibody interaction, by identifying significant changes in signal intensity upon substitutions relative to the native sequences. By inspecting the individual amino acid substitutions at the different positions within a given peptide the antibody specificity can be characterized. However, microarray-driven amino acid substitution analysis can be quite cumbersome, e.g. when mapping epitopes in hundreds of proteins this way. Here, we present ArrayPitope, an automated analysis tool for characterizing and visualizing the specificity of the epitopes identified through large-scale single-amino acid substitution analysis of peptides. The tool identifies the contribution of each amino acid residue of the target protein for recognition of the corresponding antibodies and subsequently incorporates binding signals from overlapping peptides into one statistical analysis to precisely map the selectivity of each residue of the target protein involved in the recognition. As an illustration of the utility of this tool, we apply the method on a full-scale substitution analysis of the 69 kDa human serum albumin to address the specificities of polyclonal antibodies raised against the protein. An online implementation of the tool has been made freely available at www.cbs.dtu. dk/services/ArrayPitope. Methods The method takes quantitative peptide data and a set of protein sequences as input. All peptides are mapped back to the protein sequence including all single-amino acid derivatives of these peptides. The intensity values of peptides subjected to substitutions are rescaled relative to the intensity of corresponding native peptide (median is used if multiple copies of native peptide exist). As such, substitutions giving rise to a lower intensity relative to the native peptide result in substitution values less than 1, and substitutions with no effect on the intensity takes the substitution value of 1, whereas substitutions giving rise to a higher intensity relative to the native peptide (also known as heteroclitic responses) takes a substitution value more than 1. The signal variance is estimated from the pool of native peptides sampling N peptides using bootstrap method where N is the number of copies of the given native peptide. The algorithm next performs the statistical analyses in two steps: i) first, by calculating the statistical significance of the mean substitution effects of each position in individual native peptides to determine which peptide positions are part of the epitope, and next ii) the algorithm incorporates information from overlapping peptides containing a given position in the mapped protein and determines the position specific binding selectivity of the antibody. The two steps are outlined below. Epitope-calling in individual native peptides For each native peptide sequence, the substitution values are used to generate a position-specific scoring matrix (PSSM), in which columns represents positions in the peptide and rows represents amino acid substitutions. To determine if peptide positions undergoing substitution lead to a disruption of antibody binding, the importance of each peptide position is inferred by a Dunnett's test, i.e. comparing multiple sample means to a control population. Here the mean substitution value is compared to the theoretical value, μ 0 = 1, of no selectivity. When S 2 is the pooled variance of the PSSM, then SE i ¼ ffiffiffiffiffiffiffiffiffiffi S 2 =n i p is the pooled standard error of the i-th column of the PSSM, where n i is the number of substitutions in column i of the PSSM. In a complete amino acid substitution analysis n i (and thus SE i ) is the same for all positions. The leastsignificant-difference (LSD) is computed as: The critical value t d is computed as the quantile for the one-tailed noncentral Dunnett's test distribution corresponding to i) a given significance level, ii) the number of groups k, equal to the peptide length, minus 1 and iii) the number of degrees of freedoms equal to the number of substitution values minus the length of the peptide (∑n i − k). Peptide positions, where the relative change in signal exceeds the LSD value, i.e 1-μ i > LSD i , are characterized as being part of the epitope. Determining the selectivity of positions in overlapping peptides For each residue in the protein being mapped in overlapping peptides, the algorithm seeks to determine which amino acid substitutions leads to a significant change in signal intensity relative to the native amino acid. Here, the substitution values are used to generate a substitution matrix expressing substitution values of one protein residue being represented in different positions in the overlapping peptides (see Fig 1 for a schematic illustration of the procedure). Here, only peptides are included containing protein positions previously identified to be involved in the epitope. A Dunnett's multiple comparison procedure is used to test which mean, μ i (i.e. which replacing amino acid), that are significantly different from the value 1 of no selectivity. The LSD of each column of the substitution matrix is calculated as described above for each column of the PSSM. Protein positions will be reported and visualized, where the relative change in signal of one or more amino acid substitutions exceeds the LSD value. If SE i is the pooled standard error of the i-th column of the substitution matrix, then t i

= (μ i − μ g )/SE i is the t-statistic used to test the departure of the replacing amino acid i relative to the global mean, μ g , of the substitution matrix. Thus, mean substitution values above μ g yield positive t-statistics (amino acids favouring interactions), while substitution values below μ g yield negative t-statistics (amino acids disfavouring interactions). The p-value of the t-statistic is calculated from the cumulative distribution function for the noncentral Dunnett's test distribution with degrees of freedoms equal i) to the number of replacing amino acid (up to 19) and ii) the total number of substitution values minus the number of replacing amino acids. To visualize the selectivity profile, each protein residue is presented in a logo-plot with the corresponding amino acid and subsequently rescaled so that ∑|s i | = (1 − μ g ), consequently making the absolute sum of logoheights reflect the mean change caused by substitutions of the native amino acid. To illustrate this, a sample data is shown in Fig 2, exemplifying two positions with high and low selectivity, respectively. In Fig 2A, only the native amino acid E (highlighted in solid fill at μ = 1) retains complete antibody binding (μ E % 1). The majority of the remaining amino acid substitutions lead to a decrease in signal and thus lower substitution value. The p-value associated with the native amino acid E is hence low (p ( 1), since the departure from the global mean μ g is high (t > 0), leading to a high positive score, s E . The resulting logo-plot is shown in Fig 2C. The native amino acid employs the largest letter scale, but both the negatively charged amino acid D and the positively charged amino acids K, H and R employ larger letter scales due to their departure from μ g , in opposite directions. The absolute sum of the logo-plot column corresponds to the global effect of substitution (1 − μ g = 0.60). Fig 2B exemplifies positions with only two amino acid substitutions affecting the signal. Here, the native amino acid, which also happens to be E (highlighted in solid fill at μ = 1) will employ a high p-value (p % 1), since the substitution value is close to μ g , leading to a low score (s E % 0). The resulting logo-plot is shown in Fig 2D. The absolute sum of the logo-plot column is much smaller in this example (1 − μ g = 0.20) and only substitutions to K and H are affecting the signal, as seen by the relatively large negative scales. Results The ArrayPitope webserver was implemented to perform statistical analyses of the effects of single-amino acid substitution on a receptor-ligand interaction. Here, the webserver has been used to automatically map and characterize linear antibody epitopes in the human serum albumin (HSA) protein. Quantitative peptide microarray data was obtained from Hansen et al. [17] consisting of overlapping 15-mer peptides mapping the primary sequence of HSA (uniprot P02768) in 50 copies, including one copy of all possible single-amino acid substitutions. In total, the data consisted of 215,147 peptides. These peptides were measured for response to a commercially available polyclonal rabbit anti-HSA antibody using a secondary Cy3-conjugated goat anti-rabbit IgG, and fluorescence microscopy. Identifying peptide residues involved in antibody binding To identify which peptides and peptide residues are involved in epitopes, a complete singleamino acid substitution analysis was performed by first constructing a position specific scoring matrix (PSSM) for each overlapping native peptide in the microarray data (see materials and methods for details). Table 1, for the peptides spanning the region 56-88 of HSA. The LVNEVTEF epitope is seen sliding through the overlapping 15-mer peptides. As part of the epitope leaves the peptide window (at position 67), the median native signal drops to 255 and new epitope residues are shown to be part of the epitope for the 15-mer VNEVTEFAKTC-VADE. Some inconsistencies are found among the identified amino acids in overlapping peptides because these positions are only weakly affected by substitutions. Depending on the inclusion criterion the table shows a total of up to four overlapping epitopes. The results presented are nearly identical to previous studies made on the same data, using ANOVA-protected Tukey honest-significant-difference procedure to identify key residues (on the 0.01 significance level) involved in the epitopes of individual peptides [15]. Only the weakly affected residues fluctuate to a minor degree between the two approaches. A full table for the entire HSA protein can be found in S1 Table. Antibody selectivity toward individual residues The table output above presents a visual overview of the epitope mapping in overlapping peptides and enables the user to observe where one epitope ends and another begins. The output however, does not elucidate which positions are more selective than others for the antibody binding. To extract these differences, the algorithm incorporates the substitution values of a single protein residue from different peptides (native and 19 substitutions) overlapping this position. Such a substitution matrix is shown for position 518D in Fig 4, highlighting in the lower row the amino acid replacements with higher (green) or lower (red) substitution values relative to the global mean, μ g = 0.240 (white), of the substitution matrix. Blank rows depict the native residue being represented in peptides with residues found to be significantly affected by substitution. The matrix displays complete selectivity for the native amino acid D (μ D ) μ g ), with all 19 amino acid variations disrupting antibody binding in the majority of positions being represented. In order to visualize the selectivity, the algorithm constructs logo plots covering all positions in the target protein. As an example, the logo plot representation of three epitope regions of HSA (including residues 518D and 520T) is shown in Fig 5. A logo plot of the LEVDETY epitope is shown in Fig 5A. The figure shows amino acid letters in columns representing positions in the protein. Positions not previously found (using the single peptide analysis) to be significantly important for the epitope are shown as blank. The individual letters are log(p) scaled with positive letters denoting μ i > μ g and negative letters denoting μ i < μ g . The absolute height of each position reflects the mean change (1 − μ g ) caused by substitutions of the native amino acid. More details on the calculation of the logo plot and calculation of the position specific substitution matrices can be found in the methods section. The selectivity logo-plot shows a strong selectivity in positions 516E, 518D and 519E towards the native negatively charged amino acids while showing preference for non-polar residues in position 515L and 517V, small alcohol-containing residues (Serine and Threonine) in position 520T, and aromatic residues in position 521Y. Moreover, differences in the effect on substitutions of the native residues can be seen from the absolute height of letters in the logo-plot. Fig 5B and 5C shows examples of two other epitopes (ELFE-LGEYKFQ and DI-TLSEKERQI) found within peptides with relatively low binding signal (130 and 176 Au, respectively). A large number of singleamino acid derivatives of the ELFE-LGEYKFQ epitope share the binding signal of the native epitope, whereas only a few derivatives of the DI-TLSEKERQI epitope retain antibody binding. The results show that epitopes giving rise to similar binding signal may employ different effects upon substitutions (total height of columns in logo plot), and different representations of the amino acid selectivity. A full logo plot of the entire HSA protein can be found in S1 Fig. The algorithm identifies a total of 18 HSA epitope regions, as shown in Table 2. Each identified position has at least one amino acid substitution leading to a significant mean relative Table, and are not defined automatically by the algorithm. Two B cell epitopes for HSA have been mapped and are available within the IEDB [18]. Both of these are structural epitopes characterized with a large linear determinants; (E251, F252, A253, E254, S256, K257) and (S513, A514, L515, E516, E519, T520). Both of these are accurately captured by the ArrayPitope analysis of the peptide-chip data (see Table 2). The application was applied to a library of complete single-amino acid substitutions of a native protein antigen sequences. Amino acid substitutions are a critical requirement of this algorithm. However, a single-residue scan, such as an alanine scan is sufficient to produce meaningful sequence logos that can be used to pinpoint non-alanine positions involved in the epitope. In such case, the substitution matrix in Fig 4 will contain only two columns, and the Dunnett's statistical procedure will condense to a Student's t-test of comparing the mean substitution value of the replacing amino acid with that of the native amino acid. The sliding representation of the epitopes (Table 1) will not be produced for this type of data, since the PSSM of Fig 3 would only include one substitution value for every position, critically lowering the degrees of freedom used for the Dunnett's multiple comparison procedure. 13 epitope regions were found using a limited dataset including only substitutions with alanine (S2 Table). The alanine-only dataset was sufficient to identify 105 (55.9%) out of the 188 epitope residues found using the exhaustive substitution dataset (data not shown). 11 (10.5%) of the missed epitope positions can be ascribed to the native residue being alanine. The remaining missed epitope residues can be ascribed to a general limitation in epitope mapping when solely using alanine substitutions thus highlighting the value of a complete, exhaustive substitution strategy. Table showing 18 HSA epitope regions identified by the algorithm. Dashes mark gaps of residues with no selectivity. The regions were identified from the logo plots, and defined as having a minimum of 4 residues and a maximum gap-length of 2 residues. The algorithm has been made available as a webserver. It takes quantitative peptide data of fully or partially substituted overlapping peptides as well as protein sequence(s) in FASTA-format as input. Furthermore, options for specifying a custom significance level and length of peptides to be analyzed are available. The webserver outputs a table of overlapping peptides, each with their identified epitope residues highlighted, similar to Table 1, as well as sequence logos illustrating the specificity of the antibody-epitope complexes; see Fig 5 for details. Discussion Insight into antibody-specificity of the binding sites (epitopes) of target proteins is important for the identification and design of diagnostic targets as well as characterization of therapeutic antibodies. Through the ability to express large numbers of peptides, the recent advances in high-density peptide microarrays facilitate high-throughput discovery of linear antibody epitopes. The large number of results calls for an automated method for analysing antibody-specificity. Here, we present a statistical approach to analyze antibody-specificity of epitopes from peptide-based single amino acid substitution data. The pipeline is fully automated and consist of three main steps: i) mapping of peptides to the original target, ii) mapping of epitope positions of individual peptides by determining the statistical significance of substitutions and iii) determine the selectivity of each target residue involved in the epitope and visualize this by sequence logos. Using peptide microarray data containing a complete substitution analysis of HSA the tool was used to identify and characterize the specificity of 18 linear epitope regions of polyclonal rabbit anti-HSA antibodies. The high-resolution substitution analysis allows the user to distinguish close-proximity epitopes of polyclonal antibodies, by viewing the epitope mapping in overlapping peptides sliding through the protein sequence analyzed, as exemplified in Table 1. Here, we illustrate how the method can deconvolute the presence of four individual antibody epitopes; when part of the high-signal epitope LVNEVTEF, starting at position 66, leaves the query peptide, the weaker, but overlapping epitope TEF-KT-V-E is revealed. When the N-terminal of this epitope leaves the query peptide window another overlapping epitope C-ADESAE-C appears, and yet another after this one. Identifying these epitopes solely from a signal profile, where only the signals of entire 15-mers are known, would be a challenge. As thoroughly discussed by Buus et al. [10] the signal intensity is determined by a number of factors other than antibody affinity, such as peptide purity, solvation and antibody concentration. Moreover, it is often assumed that high-affinity antibodies are likely to be more specific, but affinity and specificity are not necessarily linked, as thoroughly discussed by Regenmortel

[1]. Selecting epitope candidates purely on the basis of microarray binding signal may lead to relevant candidates bring discarded. As a demonstration, the epitope LSEKERQI, spanning the 540-547 region of HSA are presented in multiple peptides with a relatively weak signal of (114-170 Au), but the antibody specificity towards this epitope is comparable the that of the epitope LEVDETY in the signal range of 500 Au, see Fig 5. When addressing specificity a word of caution is appropriate. The six CDRs of an antibody harbour multiple overlapping paratopes of 10-20 amino acids, and for an antibody to be monospecific to a single epitope it would require that the remaining part of the CDRs are unable to bind any other antigenic structure, which is unlikely [1]. The antibody may appear monospecific, however, only when tested for their capacity to bind one antigen and not another. Characterizing the specificity of antibody-epitope interaction by the degree of stereochemical complementarity upon single-amino acid substitutions may assist the selection of antibodies from polyclonal samples and understanding of the interaction of monoclonal antibodies. There is considerable difference in selectivity between individual residues of the epitopes. Most epitopes identified here, contain a core of 3-4 highly selective residues and few residues where a certain chemical property is preferred. This information may prove crucial when characterizing potential cross reactivity to related targets. The logo-plots from the algorithm serve as a visual representation of the selectivity of individual epitope residues. Here, they have been used to identify the boundaries of epitope regions in the protein, as seen in Table 2. One may find a few inconsistencies between the epitopes identified from individual peptides in Table 1 and the interpretation of selectivity by the logo-plots. A word of caution is hence appropriate, since the two analyses differ in representation of substitutions. While the PSSM of individual peptides is used to infer peptide positions significantly affected by substitutions, the substitutions of one position (column in the PSSM) with different amino acids may vary greatly, depending on the physiochemical properties of the replacing amino acid. As such, the different substitutions in the same peptide position may contribute a higher variance (and thus higher LSD) to the analysis, than multiple copies of the same replacing amino acid, consequently leading to false-negative epitope-positions. In the statistical analyses forming the basis of the logo-plots the pooled variance is calculated from copies of the same replacing amino acid, but represented in different peptide positions. Here, bias occurs when one protein position is represented in two overlapping epitopes, interacting with different antibodies. We strongly suggest that the two analyses outputs should be used in complement to each other. In conclusion, we have presented an online analysis tool for automated characterization and visualization of antibody selectivity toward specificity-determining epitope residues from peptide microarray-driven substitution analysis. Although the method was developed with the aim of characterizing antibody-peptide interactions, it is not restricted to such interactions, but can readily analyse quantitative peptide data from any receptor-ligand interaction, provided that single-amino acid derivative peptides of the ligand peptides exist. We expect this application to be useful in complete substitution analysis of pre-identified binding peptides from multiple proteins, such as complete linear antibody epitope mapping of entire proteomes. Table. HSA epitopes found through substitution with alanine only. Table showing 12 HSA epitope regions identified by the algorithm. The dataset was limited to only include alanine substitutions of native peptides of HSA. Dashes mark gaps of residues with no selectivity. The regions were identified from the logo plots (not shown), and defined as having a minimum of 4 residues and a maximum gap-length of 2 residues. (DOC) Author Contributions Conceptualization: CSH SB OL MN PM. Recompensation factors for patients with decompensated cirrhosis: a multicentre retrospective case–control study Objectives We aimed to evaluate recompensation factors among patients with decompensated cirrhosis. Design A multicentre retrospective case–control study was conducted. Data were collected from and compared between groups of patients with recompensated and acute decompensated cirrhosis. Univariable and multivariable logistic regressions were used to select indicators associated with recompensation among patients with decompensated cirrhosis with different complications. A decision tree with 10-fold cross-validation was used to develop the model to identify patients with recompensation. We followed the transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD) guideline for development and reporting of the new model. Setting The study was conducted in six tertiary public hospitals in Chongqing, China. Participants This study included 3953 patients with decompensated cirrhosis. Results In the total sample of included patients, there were 553 patients with recompensation and 3400 patients with acute decompensation, including 1158 patients with gastrointestinal bleeding, 1715 patients with a bacterial infection, 104 patients with hepatic encephalopathy and 423 patients with ascites. The most relevant indicator of recompensation selected by the decision tree model was albumin, with a threshold of 40 g/L. Total protein, haemoglobin, basophil percentage, alanine aminotransferase, neutrophil-to-lymphocyte ratio and diabetes were also selected to subsequently distinguish patients. The terminal nodes with a probability of recompensation was 0.89. The overall accuracy rate of the model was 0.92 (0.91–0.93), and it exhibited high specificity (86.9%) and sensitivity (92.6%). Conclusions The occurrence of recompensated cirrhosis could be identified by albumin, total protein, haemoglobin, basophil percentage, alanine aminotransferase, neutrophil-to-lymphocyte ratio and diabetes. These simple variables may help clinicians develop a treatment plan to encourage patients with decompensated cirrhosis to recompensate. INTRODUCTION Patients with decompensated cirrhosis have a poor prognosis, and are more likely to undergo hospital readmissions, liver transplantation, death or hepatocellular carcinoma. 1 In patients with acute decompensated cirrhosis without acute-on-chronic liver failure (ACLF), the 28-day mortality rate was 4.6%, which increased to 12.6% at 3 months, 18.3% at 6 months, and 27.6% at 1 year. In addition, acute decompensated cirrhosis occurs in up to 15% of cirrhotic patients each year. 2 Fortunately, due to aetiology control, effective treatment or prevention, some patients with decompensated cirrhosis may no longer have decompensation-related complications, for a long period of time, which is considered to be 'recompensation'. 3 4 Some studies have shown that patients with decompensated cirrhosis have improved transplantfree survival rates, Child-Turcotte-Pugh and model for end-stage liver disease scores after receiving antiviral treatment. [5][6][7][8][9][10] For patients with alcoholic decompensated cirrhosis listed for liver transplantation, the model for end-stage liver disease score less than 20 and serum albumin greater than or equal Strengths and limitations of this study ► The data contain more than 3000 patients with decompensated liver cirrhosis from six centres, making it the largest data sample for analysing recompensation indicators. ► The indicators included in the model are available in the information systems of hospitals at all levels, which makes our indicators easier to apply in clinical and even community hospitals. ► The knowledge of recompensated indicators may be useful for supporting different prevention strategies, so as to reduce the occurrence of acute decompensation. ► The cut-off values of the model in other regions need further external validation. Open access to 32 g/L at enrollment were independent predictors of recompensation/withdrawal from the transplant list. 11 Despite that, controversies remain regarding the evaluation time, evaluation indicators, and influencing factors of recompensation. Currently, research data on the recompensation markers of decompensated cirrhosis is scant. Better identification and understanding of recompensation in patients with decompensated liver cirrhosis is very important for the design of preventive interventions that reduce the overall burden. Hence, the purpose of this study was to describe the clinical characteristics of patients with recompensation and to determine the clinical variables relevant to recompensation. PATIENTS AND METHODS Patient and public involvement statement This was a retrospective study. Therefore, the patients and the public were not directly involved in the design and conception of this study. Patients and definitions This was a multicentre retrospective case-control study. Consecutive follow-up patients with decompensated cirrhosis came from six hospitals: the Second Affiliated Hospital of Chongqing Medical University, Yongchuan Hospital of Chongqing Medical University, the Third Affiliated Hospital of Chongqing Medical University, University-Town Hospital of Chongqing Medical University, the People's Hospital of Tongliang District and the Southeast Hospital of Chongqing. Clinical data of patients treated between January 2014 and October 2019 were collected using electronic medical record systems. The inclusion criteria were as follows: (1) diagnosis of decompensated liver cirrhosis based on clinical, biochemical, ultrasonographic and/or endoscopic findings and (2) age ≥18 years old. The exclusion criteria were as follows: (1) patients with liver cancer or other active malignancy; (2) ACLF, the diagnosis of which was based on the criteria from the consensus recommendation of the Asian Pacific Association for the Study of the Liver 12 ; (3) congestive heart failure, chronic kidney disease or other significant chronic extrahepatic disease; (4) selective admission, such as reasons for hospitalisation that were either to perform liver biopsy, endoscopy with potential band ligation or an evaluation for liver transplantation or (5) more than 20% of the data missing for the patients or indicators. Acute decompensated cirrhosis was defined as the rapid development of one or more major complications of liver disease, such as ascites, encephalopathy, gastrointestinal haemorrhage and bacterial infection, requiring hospitalisation. [13][14][15][16][17] Recompensated cirrhosis was defined as clinically stable outpatients with either controlled ascites or previously treated decompensation events who were in a stable clinical state for at least 1 year. 3 Ascites was recorded as the primary reason for admission if this was the sole criterion for admission and infection was absent. Hepatic encephalopathy as characterised by altered mental status or neuropsychiatric abnormalities in the presence of liver cirrhosis after exclusion of other causes. 18 Gastrointestinal bleeding was defined as the development of an upper and/or lower gastrointestinal haemorrhage of any aetiology. 15 Bacterial infection was defined in cases of spontaneous bacterial peritonitis, pneumonia, cellulitis, biliary tract infection, urinary system infection and spontaneous bacteraemia. 17 None of the included patients developed acute decompensated cirrhosis due to bacterial infection alone. In the presence of more than one contributory factor, the main cause of admission was defined as follows: (1) in patients admitted with gastrointestinal bleeding in the presence of ascites, bacterial infection or hepatic encephalopathy, gastrointestinal bleeding was considered the main cause because it frequently causes bacterial infection or hepatic encephalopathy; (2) in the absence of bleeding at admission, bacterial infection was the main cause of hospitalisation and (3) in patients with hepatic encephalopathy and ascites, the main cause was the former. 19 The principal cause of hospitalisation was subsequently assessed independently by two subspecialist physicians. Treatment Standard medical therapies were used for all patients after diagnosis, such as antiviral therapy and symptomatic and supportive therapies. Data collection Demographic, clinical and routine laboratory data were recorded during the first contact visit to the hospital. Demographic characteristics included age and sex. The aetiological characteristics, including hepatitis B virus (HBV)/hepatitis C virus (HCV) infection, autoimmunity and alcohol consumption, were assessed from the medical history. Clinical data included complications related to liver cirrhosis and comorbidities (such as hypertension and diabetes). Laboratory analyses included red cell counts, white cell counts, platelet counts, haemoglobin, basophil percentage, eosinophil percentage, total protein, albumin, direct bilirubin, indirect bilirubin, alanine aminotransferase (ALT), aspartate aminotransferase (AST), γ-glutamyl aminotransferase (γ-GT), alkaline phosphatase (ALP), neutrophil-to-lymphocyte ratio (NLR) and lymphocyte-to-monocyte ratio. Data from patients with recompensated cirrhosis were recorded during the first contact visit. For patients with acute decompensated cirrhosis, data were obtained within 24 hours of the initial diagnosis. Statistical analysis All statistical analyses were conducted using R software (V.4.0.2). All data were presented as counts with percentages for categorical variables and medians (IQRs) for continuous variables. We followed the transparent Open access reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD) guideline for development and reporting of the new model. 20 For variables with omission rates <20%, the mean imputation was used. χ 2 and Kruskal-Wallis tests were used when appropriate to evaluate the significance of differences in distributions between patients with recompensation, patients with gastrointestinal bleeding, patients with bacterial infection, patients with hepatic encephalopathy and patients with ascites. Indicators with p<0.05 were subsequently included in the univariate logistic regression to identify the factors associated with recompensation, followed by multivariate analysis using those factors with p<0.05 in the univariate logistic regression analysis. We combined the factors in each multivariate regression analysis and used decision trees to select recompensation indicators. The decision tree algorithm selected the most relevant of

these clinical variables, their position within the decision tree and their optimal cut-off values. The model was fitted using the R software (V.4.0.2) rpart package with the settings minsplit=20 and maxdepth=6. A 10-fold cross-validation was used to reduce overfitting and to assess the discrimination ability of the model by estimating the corresponding sensitivity and specificity of the model and computing the overall accuracy rate along with a 95% CI for this rate (using binom. test) and a one-sided test to see if the accuracy is better than the 'no information rate,' which is taken to be the largest class percentage in the data. 21 22 All tests were considered statistically significant when p<0.05. Baseline characteristic of patients In the total sample of included patients, there were 553 patients with recompensation and 3400 patients with acute decompensation, including 1158 patients with gastrointestinal bleeding, 1715 patients with a bacterial infection, 104 patients with hepatic encephalopathy and 423 patients with ascites. The aetiology of liver cirrhosis in these patients was 1955 (49.5%) positive for HBV, 190 (4.8%) positive for HCV, 573 (14.5%) for alcohol abuse, 245 (6.2%) for autoimmunity and 990 (25%) for other causes. The median age was highest among the patients with ascites, while the proportion of male patients was highest among the patients with hepatic encephalopathy. The median red cell counts, platelet counts, haemoglobin, basophil percentage, eosinophil percentage, total protein, albumin, lymphocyte-to-monocyte ratio and proportion of patients with other aetiologies were highest among the patients with recompensation, while the median white cell counts, direct bilirubin, ALT, AST, γ-GT, proportion of patients with HBV infection and proportion of patients with autoimmunity were highest among the patients with bacterial infection. The proportion of patients with alcoholic cirrhosis, hypertension, diabetes, the median indirect bilirubin and ALP levels were highest among the patients with hepatic encephalopathy, while the median NLR was highest among the patients with gastrointestinal bleeding (table 1). Tool to evaluate recompensation A decision tree was fitted to illustrate the observed associations and to detect other specific subgroups and relationships that may not be available through multivariate regression analysis. The variables which were statistically significant in the multivariate logistic regression from Open access gastrointestinal bleeding group, bacterial infection group, hepatic encephalopathy group and ascites group were combined, and a decision tree was employed to determine the index correlations with recompensation in patients with decompensated cirrhosis. The most relevant indicator of recompensation selected by the decision tree model was albumin, with a threshold of 40 g/L (figure 1). Total protein, haemoglobin, basophil percentage, ALT, NLR and diabetes were also selected to subsequently distinguish patients. The overall accuracy rate of the model was 0.92 (0.91-0.93), with high specificity (86.9%) and sensitivity (92.6%). If a patient with decompensated cirrhosis has serum albumin equal to or greater than 40 g/L, total protein equal to or greater than 72 g/L, basophil percentage equal to or greater than 0.07, NLR less than 5, haemoglobin equal to or greater than 104 g/L, and has no history of diabetes, then he/she has 89% probability to be discriminated as being recompensation. This is an example of an interpretation from figure 1. DISCUSSION Recompensation is a special phase of decompensated liver cirrhosis. After a period of effective treatment, the liver function can meet the patients' daily activities, and there will be no complications related to decompensated liver cirrhosis. 4 Until now, there has been a lack of a comprehensive evaluation index to identify patients with a 'recompensation advantage.' In this study, we analysed the recompensation-related factors of different complications of decompensated liver cirrhosis and combined these factors to establish a decision tree based on the presence of several routine laboratory indicators and comorbidities. It has demonstrated that albumin, total protein, haemoglobin, basophil percentage, ALT, NLR and Open access diabetes is associated with recompensation for patients with decompensated cirrhosis. The level of albumin was the most important indicator. The algorithm generally had good accuracy with high specificity and sensitivity. The proposed decision tree included three liver function indexes (albumin, total protein and ALT). Many studies have suggested a correlation between albumin and the prognosis of cirrhosis, with low serum albumin concentration being an important factor in the poor prognosis of cirrhosis. [23][24][25] For patients with alcohol-related liver disease on the liver transplantation waiting list, model for end-stage liver disease score <20 and albumin ≥32 g/L at entry were found to be the optimum cut-off points for predicting withdrawal from the transplantation list after recompensation. 11 Long-term use of albumin has a low hospital admission rate and mortality due to tension ascites or complications, such as hepatic encephalopathy, spontaneous peritonitis, bacterial infections other than spontaneous peritonitis, renal insufficiency, type 1 hepatorenal syndrome and side effects caused by potential diuretics, such as hyponatraemia and hyperkalaemia. 24 25 In our study, the cut-off value of albumin was 40 g/L, which was consistent with the results of the ANSWER study and indicated that only sufficient albumin concentrations can play a protective role. 24 ALT is generally considered to be an indicator of liver damage due to steatosis and inflammatory responses. 26 Severe impairment of liver function may lead to the risk of hepatic encephalopathy. 27 Our study found that regardless of whether there was gastrointestinal bleeding, the level of haemoglobin in patients with decompensated liver cirrhosis was a relevant factor for recompensation. Anaemia is another factor that has been recently characterised as a predictor of poor outcomes in patients with cirrhosis, including ACLF occurrence in outpatients with liver cirrhosis, and hepatocellular carcinoma induced death. [28][29][30][31][32] Bothou et al found that a low level of haemoglobin was a strongly Open access and independent predictor of hospital admission in outpatients with decompensated liver cirrhosis. 33 The mean value of haemoglobin reported by Bothou et al was 134 g/L. Anaemia can predict the development of ACLF in outpatients with cirrhosis, the median value was 108 g/L. 29 Thus, improving anaemia is a therapeutic target for maintaining the stability of decompensated liver cirrhosis, and it is very probable that for patients with cirrhosis even mild anaemia should be treated. The pathophysiology of patients with liver cirrhosis may be related to anaemia, which leads to arterial hypotension and tachycardia, resulting in circulatory dysfunction. However, the specific mechanism needs to be investigated in future studies. NLR is a novel indicator of systemic inflammation and has recently been reported to predict the outcome of patients with decompensated cirrhosis. [34][35][36] The PREDICT study indicated that patients with stable decompensated cirrhosis had less systemic inflammation compared with pre-ACLF patients who exhibited systemic inflammation with rapid progression (leading to the development of ACLF and death within 90 days) and unstable decompensated cirrhosis patients who were readmitted at least once during 90 days but did not progress to ACLF. 21 Our research indicated that the NLR was markedly higher in patients with acute decompensated cirrhosis with gastrointestinal bleeding, bacterial infection and ascites compared with that in patients with recompensation. An NLR equal to or greater than five is more likely to classified as acute decompensated liver cirrhosis. These findings clearly suggest that the systemic inflammatory response is predictive of a poor prognosis. Therefore, clinicians should pay significant attention to the prevention of infections, which could avoid secondary complications (further development of decompensation, recurrent infections, ACLF and death) of cirrhosis. 37 Basophils may induce and expand inflammation by producing specific cytokines and proteases and are Open access related to T helper 2 immune responses. 38 39 However, their role in decompensated liver cirrhosis has rarely been reported. Our study found that the increase in the basophil percentage is an evaluation index of recompensation, especially for the occurrence of gastrointestinal bleeding or bacterial infection. We speculate that basophils may regulate local and systemic inflammatory responses and shape innate and adaptive immune responses to prevent acute decompensation, but the specific mechanism needs further study. Diabetes is closely related to complications of liver cirrhosis. Diabetes may be related to an increased risk for the existence of covert hepatic encephalopathy and the development of overt hepatic encephalopathy in patients with liver cirrhosis. 40 Uncontrolled diabetes is associated with an increased risk of infection, an enhanced propensity for renal insufficiency, and a variety of other related complications. 41 Diabetes increases the risk of rehospitalisation within 30 or 90 days for patients with decompensated cirrhosis. 42 43 These results are in line with our study that diabetes is a risk indicator for acute decompensation, and patients with decompensated cirrhosis without diabetes are more likely to have recompensation. The strengths of our study As far as we know, this was the first study for including patients who did not experience acute decompensation within 1 year as recompensation. Compared with the previous study, the patient's condition was stable for a longer period of time. Furthermore, we combined logistic regression and decision tree to screen recompensation indexes, and proposed cut-off values of different indicators, which was more convenient for clinical application. LIMITATIONS Our study has several limitations. First, since our study of factors associated with recompensation in cirrhosis Open access is unique, there were no other cohorts that could available for external validation. We will conduct a prospective study to further evaluate the effectiveness of this new model. Second, we did not compare known models, such as the Child-Turcotte-Pugh, model for end-stage liver disease, or chronic liver failure-consortium acute decompensation scores. Because our model neither includes subjective clinical symptoms, such as hepatic encephalopathy and the severity of ascites, nor indicators that cannot be detected in community hospitals, such as international normalised ratio. We only included the most widely used laboratory indicators and comorbidities. The study from the real world makes our indicators easier to apply in clinical and even community hospitals. Besides, we performed 10-fold cross-validation to reduce overfitting of the model. Third, these thresholds have not yet been validated in the external cohort. Therefore, the difference between these cut-off values may be related to the difference in the prevalence of cirrhosis among the studied populations. CONCLUSIONS This study showed the level of albumin, total protein, haemoglobin, basophil percentage, ALT, NLR and the history of diabetes is related to recompensation in patients with decompensated cirrhosis. The decision tree algorithm identified albumin with a threshold of 40 g/L as the indicator that most influenced the occurrence of recompensation. The knowledge of recompensated indicators may help support the development of different prevention strategies to reduce the incidence of acute decompensation. Figure 1 Decision tree plot for identifying recompensated cirrhosis. Each node shows the percentage of patients classified and their probability of recompensation (also represented by the colours and colour intensity). The blue colour represents acute decompensated cirrhosis. The green colour represents recompensated cirrhosis. The intensity of the colour indicates the accuracy of the classification of each category, the more intense the colour, the higher the accuracy. Up-regulation of Prostaglandin E2 Synthesis by Interleukin-1β in Human Orbital Fibroblasts Involves Coordinate Induction of Prostaglandin-Endoperoxide H Synthase-2 and Glutathione-dependent Prostaglandin E2 Synthase Expression* Prostaglandin E2(PGE2) production involves the activity of a multistep biosynthetic pathway. The terminal components of this cascade, two PGE2 synthases (PGES), have very recently been identified as glutathione-dependent proteins. cPGES is cytoplasmic, apparently identical to the hsp90 chaperone, p23, and associates functionally with prostaglandin-endoperoxide H synthase-1 (PGHS-1), the constitutive cyclooxygenase. A second synthase, designated mPGES, is microsomal and can be regulated. Here we demonstrate that mPGES and PGHS-2 are expressed at very low levels in untreated human orbital fibroblasts. Interleukin (IL)-1β treatment elicits high levels of PGHS-2 and mPGES expression. The induction of both enzymes occurs at the pretranslational level, is the consequence of enhanced gene promoter activities, and can be blocked by dexamethasone (10 nm). SC58125, a PGHS-2-selective inhibitor, could attenuate the induction of mPGES, suggesting a dependence of this enzyme on PGHS-2 activity. IL-1β treatment activates p38 and ERK mitogen-activated protein kinases. Induction of both mPGES and PGHS-2 was susceptible to either chemical inhibition or molecular interruption of these pathways with dominant negative constructs. These results indicate that the induction of PGHS-2 and mPGES by IL-1β underlies robust PGE2 production in orbital fibroblasts. The past decade has witnessed dramatic advances in our understanding of the prostanoid biosynthetic pathways in mammalian cells. Of particular importance was the discovery and molecular characterization of two cyclooxygenase isoforms and the realization that each protein might possess distinct functions and patterns

of expression and regulation (1,2). Prostaglandinendoperoxide H synthase-1 and -2 (EC 1.14.99.1, PGHS) 1 are membrane-associated and contain a heme prosthetic group. The PGHS isoforms have received substantial attention, in large part because they are the targets of aspirin and a large series of nonsteroidal anti-inflammatory compounds (3). PGHS-1 is constitutively expressed in most tissues and is thought to generate prostaglandins involved in housekeeping activities (1,2,4,5). In contrast, PGHS-2 is expressed in most tissues only following cell activation by cytokines, growth factors, and mitogens (6 -10). Prostanoids generated through the activities of PGHS-2 are thought to represent those produced under circumstances of inflammation and tissue disruption. A very recent and significant advance has resulted from the identification of two prostaglandin E 2 (PGE 2 ) synthase enzyme isoforms (EC 5.3.99.3) (11)(12)(13). Both are glutathione-dependent and catalyze the terminal conversion reaction of PGH 2 to PGE 2 . One enzyme is a constitutively expressed cytosolic protein, designated cPGES (12). Evidence has been advanced indicating that cPGES is identical to p23, a putative chaperone of hsp90 that stabilizes steroid hormone receptor-hsp90 complexes (14). It exhibits substantial constitutive expression in many tissues. Moreover, this expression was reported to be invariant in several cell lines in vitro with regard to treatment with IL-1␤ or TNF-␣ (12). cPGES protein was found to be induced modestly in rat brain 48 h after injection of lipopolysaccharide (12). The other isoform, mPGES, is a 16-kDa microsomal protein that can be regulated (11,13). This enzyme may be expressed and regulated in a cell type-specific manner. mPGES is induced by lipopolysaccharide in rat macrophages, and this up-regulation can be blocked by the glucocorticoid, dexamethasone (13). Neither treatment altered the levels of cPGES in macrophages. Thus mPGES represents a regulated enzyme and a potentially important drug target. Very recent studies imply a preferential, functional association of each PGES isoform with a particular PGHS enzyme (12,13). The studies demonstrated that cPGES is linked to PGHS-1 and that mPGES utilizes PGH 2 generated by PGHS-2. Further studies have extended the notion that functional, stimuli-dependent linkage exists between specific PGHS isoforms and cell typespecific downstream enzymes (15). These observations were made in transfected cells where one or more of the relevant enzymes had been over-expressed. Importantly, nothing is known currently about whether the expression of endogenous PGHS isoforms might in some way exhibit physiological coordination with that of the PGES enzymes. Such a coupling FIG. 1. IL-1␤ up-regulation of PGE 2 synthesis in human fibroblasts is associated with the coordinate induction of PGHS-2 and mPGES proteins. Orbital fibroblasts, in this case from a patient with severe thyroid-associated ophthalmopathy, were allowed to proliferate to confluence in medium supplemented with 10% FBS. A, monolayers were shifted to medium without serum for 24 h and then treated with IL-1␤ (10 ng/ml) for the times indicated along the abscissa. For the PGE 2 determinations, medium was removed 30 min before the cultures were harvested and replaced with PBS with the respective additives. PBS was then collected and subjected to the assay described under "Experimental Procedures." For Western blot analysis of PGES and PGHS protein levels, monolayers were harvested and analyzed as described. B, cultures were treated as described for panel A, but IL-1␤ concentrations were graded as indicated along the abscissa, and the treatment time was uniformly 16 h. Western blots were scanned; the relative densities are represented as columns underneath the images. Data from the PGE 2 assay are expressed as the mean Ϯ S.D. of triplicate determinations. The data derive from a single representative experiment of the three performed. would be consistent with a model of PGE 2 production that exhibited functional cellular compartmentalization. Fibroblasts have been shown to express components of the prostanoid biosynthetic pathways, and cultures derived from particular anatomic regions and tissues can generate large amounts of PGE 2 when provoked (16). These sentinel cells are critical to the orchestration of inflammatory responses, tissue remodeling, and wound healing and are key participants in the evolution of fibrosis (17). Diversity among human fibroblasts has only recently been appreciated as being potentially important to normal tissue function and disease manifestation. It is now clear that fibroblast subsets represent highly specialized cells that participate in reactive processes in the context of both normal function and pathology (18,19). Fibroblasts function in various aspects of immunity, the recruitment of bone marrowderived cells, tissue repair, and remodeling and provide the molecular and structural infrastructure supporting cellular cross-talk (17,18,20). Human orbital fibroblasts exhibit a phenotype that sets them apart from fibroblasts derived from other anatomic regions. They are composed of discrete subsets (18), one of which represents pre-adipocytes (21); possess a characteristic morphology (22); and display a distinct pattern of gangliosides (23) and surface receptors (24). Of particular relevance to their proposed participation in orbital inflammation are the robust responses to a variety of disease mediators such as cytokines, growth factors, and bioactive lipids (25)(26)(27)(28)(29). These fibroblasts have been implicated in the pathogenesis of thyroid-associated ophthalmopathy (TAO), an autoimmune process. Two hallmarks of the tissue remodeling observed in TAO are the accumulation of hyaluronan and an often intense inflammatory reaction (30). We hypothesize that it is this set of attributes that renders connective tissues in the human orbit susceptible to the remodeling associated with TAO. Among the characteristics that set apart orbital fibroblasts is the dramatic up-regulation of PGE 2 synthesis observed following exposure of these cells in culture to proinflammatory cytokines (25,26). These increases in PGE 2 can be blocked by specific inhibitors such as SC58125 and NS-398, suggesting a dominant role for PGHS-2 in mediating the up-regulation (25). In fact, orbital fibroblasts when provoked by IL-1␤ or leukoregulin or through engagement of surface-displayed CD40 by CD154, exhibit a particularly robust induction of PGHS-2 (25,26,31). Here, we report that the treatment of orbital fibroblasts with IL-1␤ results in dramatic increases in PGE 2 production and is associated with coordinate induction of both PGHS-2 and mPGES expression. These responses are mediated through elevations in the steady-state levels of their respective mRNAs and involve the use of overlapping intracellular signaling pathways, including the p38 and ERK mitogen-activated protein (MAP) kinases. Inhibiting PGHS-2 activity results in the blockade of mPGES induction by IL-1␤, indicating some involvement of the products of the former enzyme in the expression of the latter. These latest findings provide insights into the complex interactions between PGHS-2 and mPGES in orbital fibroblasts that culminate in the generation of PGE 2 . Cell Culture-Orbital fibroblast cultures were initiated from tissue explants obtained as surgical waste during decompression surgery from severe TAO or were from normal appearing orbital tissues in patients undergoing surgery for non-inflammatory conditions. These activities have been approved by the Institutional Review Boards of Albany Medical College and Harbor-UCLA Medical Center. Some of the fibroblast strains were kindly provided by Dr. Rebecca Bahn (Mayo Clinic, Rochester, MN). Tissue fragments were generated by mechanical disruption of explants, and fibroblasts were then allowed to adhere to plastic culture plates. They were covered with Eagle's medium to which 10% fetal bovine serum (FBS), glutamine (435 g/ml), and penicillin/ streptomycin were added as described previously (33). Medium was changed every 3-4 days, and monolayers were maintained in a 5% CO 2 , humidified incubator at 37°C. Culture strains were utilized between the second and twelfth passage from initiation. All experimental manipulations were conducted after a state of confluence had been reached. We have already established the purity of these cultures and found them to be essentially free of contamination by endothelial and smooth muscle cells (18). RNA Isolation and Northern Hybridization-Fibroblasts were cultivated in 100-mm-diameter plates to a confluent state and were then treated with the test agents specified in the figure legends. Cellular RNA was extracted from rinsed monolayers by the method of Chomczynski and Sacchi (34) with an RNA isolating system purchased from Biotecx (Houston, TX). The nucleic acid was subjected to electrophoresis through denaturing 1% agarose, formaldehyde gels. Integrity of the RNA was established by determining the 260/280 spectroscopic ratios and by staining the electrophoresed samples with ethidium bromide and inspecting them under UV light. The RNA was transferred to Zeta-probe membrane (Bio-Rad), and immobilized samples were hybridized with [ 32 P]dCTP-labeled PGHS-1, PGHS-2, p23, and mPGES cDNA probes generated by the random primer method. Hybridization was conducted in a solution containing 5ϫ SSC, 50% formamide, 5ϫ Denhardt's solution, 50 mM phosphate buffer (pH 6.5), 1% SDS, and 0.25 mg/ml salmon sperm at 48°C overnight. Membranes were washed under high stringency conditions, and then the RNA/DNA hybrids were visualized by autoradiography on X-Omat film (Kodak, Rochester, NY) following exposure at Ϫ70°C. Bands resulting from radioactive hybrids were scanned by densitometry. Membranes were then stripped according to the instructions of the manufacturer and rehybridized with a GAPDH cDNA probe, and the band densities were normalized to this signal. For mPGES and PGHS-2 mRNA stability studies, cultures were treated with IL-1␤ for 3 h as a pretreatment. Cells were washed and incubated in growth medium for 4 h. At time 0, DRB (20 g/ml), an inhibitor of gene transcription, was added to the medium of all plates without or with IL-1␤ (10 ng/ml) for the intervals indicated in Fig. 4. Abundance of mRNAs for the two enzymes was quantified by Northern blot hybridization. mPGES and PGHS-2 mRNA signals were normalized to their respective GAPDH levels. Western Blot Analysis of Fibroblast Proteins-Cellular proteins were solubilized from rinsed fibroblast monolayers following the treatments indicated in the figure legends. The ice-cold harvest buffer contained 0.5% Nonidet P-40, 50 mM Tris-HCl (pH 8.0), and 10 M phenylmethylsulfonyl fluoride. Lysates were taken up in Laemmli buffer and subjected to SDS-PAGE, and the separated proteins were transferred to polyvinylidene difluoride membrane (Bio-Rad). Primary monoclonal antibodies directed against PGHS-1 and PGHS-2 (10 g/ml, Cayman) were incubated with the membranes for 2 h at room temperature. Following washes, membranes were reincubated with secondary peroxidase-labeled antibodies. In other experiments, primary antibodies directed against human mPGES and cPGES were utilized for the detec- FIG. 3. A, Northern blot analysis of the effects of IL-1␤ on the steadystate levels of mRNAs encoding mPGES, cPGES, PGHS-2, and PGHS-1 in orbital fibroblasts. Confluent fibroblast cultures were shifted to medium without FBS to which IL-1␤ (10 ng/ml) was added for the time intervals indicated along the abscissa. Monolayers were rinsed and cellular RNA harvested as indicated under "Experimental Procedures." RNA was then subjected to Northern blot hybridization with the respective 32 P-labeled cDNA probes. The radioactive RNA/DNA hybrids were visualized by exposing membranes to X-Omat film. The relative densities of the PGHS-2 and mPGES hybridization are displayed as columns underneath the images following their normalization with GAPDH signals. B, induction of mPGES mRNA by IL-1␤ in orbital fibroblasts is independent of intermediate protein synthesis. Cultures were shifted to medium without FBS to which nothing (control), IL-1␤ (10 ng/ml), cycloheximide (Cyclo) (10 g/ml), or a combination of the test compounds was added for 6 h. Cellular RNA was extracted and subjected to Northern blot analysis. (OD of GAPDH-corrected PGHS-2 signal in IL-1␤-treated sample is 0.98 and that in the IL-1␤ ϩ cycloheximide sample is 0.21). tion of these enzyme proteins. The ECL (Amersham Biosciences) chemiluminescence detection system was used to generate signals, and the resulting bands were analyzed with a densitometer. With regard to assessing the activation of p38 and ERK MAP kinases, cell lysates from untreated orbital fibroblast cultures and those treated with IL-1␤ (10 ng/ml) were subjected to SDS-PAGE; the proteins were transferred to membrane and then probed with phospho-specific antibodies against the two kinases. PGE 2 Assay-Fibroblasts were grown to confluence in 24-well plastic cluster plates in medium containing 10% FBS. Monolayers were shifted to serum-free medium for the final 24 h of incubation. IL-1␤ and the other test compounds were added at the times and concentrations indicated in the figure legends. Medium was removed from the cultures, and the monolayers were covered with phosphate-buffered saline (PBS) in the presence of the respective agents for the final 30 min of the treatment period. PBS was collected quantitatively, clarified by centrifugation, and subjected to PGE 2 radioimmunoassay using a commercially available kit (Amersham Biosciences). Transient Transfection of Orbital Fibroblasts with Plasmids Containing mPGES and PGHS-2 Promoters and DN Mutant p38 and ERK-For studies involving the transient transfection of human fibroblasts, cultures were

allowed to proliferate to 80 -90% confluence in medium containing 10% FBS. With regard to assessment of promoter activities, a 510-bp fragment spanning Ϫ538 to Ϫ28 of the putative mPGES promoter was cloned with the Human GenomeWalker kit (CLONTECH, Palo Alto, CA) according to the instructions of the supplier. Two reverse primers used for the PCR reactions included 5Ј-CGCAGCTCAACTGTGGGTGTGATC-3Ј and 5Ј-GTGATCAGCTCGA-CAGAGGAGCAG-3Ј. The amplified fragment was sequenced and subcloned from pCR2.1-TOPO vector (Invitrogen) into a promoter-less pGL2-luciferase vector (Promega, Madison, WI). With regard to the human PGHS-2 promoter, a plasmid designated Ϫ1800pGL2, containing Ϫ1840 to ϩ123 and thus located five base pairs upstream from the ATG of the human PGHS-2 promoter, was used. Promoter constructs were transiently transfected into fibroblasts using the LipofectAMINE PLUS system (Invitrogen). 0.75 g of pGL2 promoter DNA and 0.1 g of pRL-TK vector DNA (Promega), serving as a transfection efficiency control, were mixed with PLUS reagent for 15 min before being combined with LipofectAMINE PLUS for another 15 min. The DNA-lipid mixture was added to culture medium of 80% confluent cells for 3 h at 37°C. Dulbecco's modified Eagle's medium containing 10% FBS replaced the transfection mixture overnight. Transfected cultures were then serum-starved, and some received either IL-1␤ (10 ng/ml) for 2 h or nothing (control) as indicated in the figure legends. Cellular material was harvested in buffer provided by the manufacturer (Promega) and stored at Ϫ80°C until assayed. Luciferase activity was monitored with the Dual-Luciferase Reporter Assay System (Promega) in an FB12 tube luminometer (Zylux). Values were normalized to internal controls, and each experiment was performed at least three times. To interrupt the expression of potentially relevant signaling pathway components, DN constructs for p38 and ERK1 were ligated into pcDNA3.1 (Invitrogen). These were transiently transfected into cells as described above. Control cultures received a constant amount (2 g) of empty vector DNA. The diminished levels of the kinases were documented by Western blotting an aliquot of the lysate with relevant antibodies. IL-1␤ Up-regulates PGE 2 Production in Orbital Fibroblasts; This Is Associated with Induction of PGHS-2 and mPGES Proteins-Confluent orbital fibroblast monolayers were shifted to serumless medium without or with IL-1␤ (10 ng/ml) for up to 48 h. As the data in Fig. 1A suggest, the cytokine elicits a dramatic increase in PGE 2 levels, which are 90-fold above control levels at 6 h, remain near peak values for 24 h, and then begin to decline at 48 h. Western blot analysis of cellular proteins from fibroblasts treated under identical conditions reveal a large, time-dependent induction of mPGES and PGHS-2. Neither is expressed at detectable levels under basal conditions. With regard to PGHS-2, the protein is detectable at 4. A, effect of IL-1␤ on the activities of human mPGES and PGHS-2 gene promoters in orbital fibroblasts. Cells, in this case from an individual with severe TAO, were seeded in 6-well plastic culture plates and allowed to proliferate to 70% confluence. They were then shifted to a mixture of LipofectAMINE PLUS with empty vector or the PGHS-2 or mPGES promoter/reporter constructs described under "Experimental Procedures" for 3 h. Some wells received IL-1␤ (10 ng/ml) for 2 h. Cellular material was harvested and luciferase activity determined. Data are expressed as the mean Ϯ S.D. of three replicates. B, effect of IL-1␤ on the disappearance of PGHS-2 and mPGES mRNAs in orbital fibroblasts. Cultures were allowed to proliferate to confluence and then they were treated with IL-1␤ (10 ng/ml) for 3 h. Monolayers were washed and incubated in complete growth medium for 4 h, the cultures were then shifted to medium containing DRB (20 g/ml) without or with IL-1␤ for the duration of time indicated along the abscissas. Cultures were harvested and the RNA subjected to Northern blot hybridization with cDNA probes for PGHS-2 or mPGES. These signals were normalized with their respective GAPDH levels. and do not reach an apparent maximum until 24 h. Unlike PGHS-2, mPGES protein levels remain close to their peak at 48 h. In contrast, levels of cPGES and PGHS-1 are not altered as a consequence of IL-1␤ treatment. The effects of IL-1␤ on PGE 2 synthesis, the expression of PGHS-2 and mPGES are dose-dependent, as the data in Fig. 1B suggest. A near maximal effect on PGHS-2 expression and PGE 2 production is achieved at a concentration of 1 ng/ml. That concentration yields a suboptimal induction of mPGES. IL-1␤ at 10 ng/ml, the highest concentration of the cytokine used, results in a considerably higher level of mPGES. The induction of both mPGES and PGHS-2 by IL-1␤ was susceptible to blockade by the synthetic glucocorticoid, dexamethasone (10 nM, Fig. 2). That concentration of dexamethasone is associated with a high fractional occupancy of the nuclear glucocorticoid receptor and near maximal effects on human fibroblast metabolism (35). The steroid had no effect on basal enzyme expression. The blockade of IL-1␤-dependent enzyme expression was accompanied by a substantial inhibition of cytokine-provoked PGE 2 production. Induction of PGHS-2 and mPGES by IL-1␤ Is Mediated at the Pre-translational Level-Northern blot analysis was employed to determine the relationship in orbital fibroblasts between the induction of PGHS-2 and of mPGES mRNAs. Confluent cultures were shifted to medium without FBS overnight, and then IL-1␤ (10 ng/ml) was added at various times prior to monolayer harvest. As the Northern blot shown in Fig. 3A indicates, the transcript encoding PGHS-2 is not detectable under basal culture conditions but was induced as a single, 4.8-kb mRNA with the addition of IL-1␤ within 6 h. At 12 h, the mRNA levels were at least 100-fold above control levels. At 24 h, PGHS-2 mRNA levels had dropped substantially and were again undetectable at 48 h. When the membrane was rehybridized with an mPGES cDNA probe, a 2-kb transcript was apparent under control conditions and was strongly upregulated with IL-1␤, an effect that reached a maximum at 12 h, when it was ϳ7-fold above control levels. The induction was partially sustained for 48 h, the duration of the study, when it remained increased by 2.7-fold. Steady-state levels of cPGES and PGHS-1 mRNA were relatively constant following the addition of IL-1␤ to the culture medium (Fig. 3A). The former is expressed on Northern blot analysis as a single band. PGHS-1 mRNA migrates as an ϳ5.2-kb transcript, which is consistent with the pattern found previously in monocytes (36) and endothelial cells (37) and differs from the predominant 2.8-b mRNA found in some other human cells (5). Thus the relative levels of their respective mRNAs mirrored the pattern of protein induction provoked by cytokines. The Up-regulation of mPGES by IL-1␤ Represents a Primary Induction That Is Not Dependent upon Intermediate Protein Synthesis-We have reported previously that the induction of PGHS-2 mRNA by the proinflammatory cytokine leukoregulin is partially dependent on ongoing protein synthesis (25). We next determined whether the effects of IL-1␤ on mPGES and PGHS-2 mRNA levels were altered by an inhibition of protein synthesis created by cycloheximide (10 g/ml). This inhibitor concentration blocks Ͼ95% of protein synthesis in human fibroblasts (35). As the Northern blot in Fig. 3B clearly indicates, the inhibitor failed to attenuate the induction of mPGES after 6 h of IL-1␤ treatment. The inhibitor did down-regulate by ϳ78% the induction of PGHS-2 mRNA by IL-1␤. Thus, although the up-regulation by IL-1␤ of mPGES mRNA apparently represents a primary inductive event, the increase in PGHS-2 is dependent, at least in part, on the induction of an intermediate protein(s). genes were transiently transfected into orbital fibroblasts, and then cultures were incubated without or with the cytokine. As the data contained in Fig. 4A suggest, the PGHS-2 promoter exhibited considerable activity under basal conditions, at least 20-fold more luciferase activity than that found in controls transfected with the empty vector. IL-1␤ (10 ng/ml) increases the activity of PGHS-2 promoter by ϳ2-fold after 2 h of treatment. The up-regulation is consistent with our earlier finding that PGHS-2 gene transcription, as assessed by nuclear run-on assays, was enhanced 2-fold by leukoregulin (25). The mPGES promoter exhibits considerably less basal activity in orbital fibroblasts. IL-1␤ increases the activity of this promoter by ca 2.5-fold after 2 h of treatment. This response of the mPGES promoter is consistent with the modest effects of IL-1 observed in A549 human alveolar cells transfected with similar promoter constructs (38). Induction of mPGES and PGHS-2 by IL-1␤ Involves the Modest Up-regulation of Their Respective Gene Promoters-The The effect of IL-1␤ on the PGHS-2 and mPGES promoters was relatively modest. This fractionally small effect on PGHS-2 was unexpected because of the large induction of steady-state PGHS-2 mRNA levels observed (Fig. 3A). The 2-fold increase in mPGES promoter activity following IL-1␤ treatment was also less than anticipated, given the 7-fold increase in mPGES mRNA. We therefore next examined the impact of the cytokine on the decay of the respective hybridizable mRNAs. As the data from Northern blots in Fig. 4B indicate, under conditions without IL-1␤, PGHS-2 mRNA decays rapidly, and the cytokine dramatically retards the decay of the PGHS-2 transcript following transcriptional blockade with DRB (20 g/ml). The apparent t1 ⁄2 of PGHS-2 mRNA under culture conditions in the absence of IL-1␤ is around 1 h, and the mRNA levels have fallen 80% by 1.5 h. This instability is consistent with its behavior in a number of cell types (9). The disappearance of PGHS-2 mRNA was dramatically decreased following treatment with IL-1␤ (10 ng/ml), and levels remained constant for 5 h. mPGES mRNA was far more stable than that of PGHS-2 in control orbital fibroblasts not treated with the cytokine. The data in Fig. 4B demonstrate a t1 ⁄2 of ϳ6 h in untreated cultures. The addition of IL-1␤ failed to further enhance the survival of this relatively long-lived transcript. The Magnitude of mPGES and PGHS-2 Induction in Orbital Fibroblasts by IL-1␣, IL-1␤, and CD154 Is Considerably Greater than That Elicited by Other Cytokines-We next determined whether the dramatic induction of mPGES and PGHS-2 by IL-1␤ in orbital fibroblasts could be seen following exposure to other cytokines implicated in human autoimmune disease. Confluent cultures were treated with TNF-␣ (10 ng/ml), TGF-␤ (5 ng/ml), IL-4 (10 ng/ml), IL-1␣ (10 ng/ml) as well as IL-1␤ for 16 h. The effects of IL-1␣, IL-1␤, and CD154 on both mPGES and PGHS-2 expression are substantially greater than those of the other cytokines tested (Fig. 5). Especially surprising is the apparent lack of response to TNF-␣, considering the relatively large induction of PGHS-2 and PGE 2 found in other cell types. IL-4 exerts modest effects on PGE 2 production in orbital fibroblasts (25,39) but the cytokine failed to up-regulate either mPGES or PGHS-2 in the current studies. These fibroblasts display high levels of CD40, the receptor for CD154 (40). Cao et al. (31) have demonstrated that PGHS-2 expression can be induced by activation of the CD40/CD154 bridge. Cultures pretreated with interferon ␥ (100 units/ml) respond to recombinant human CD154 with a sizable induction of both mPGES FIG. 6. A, IL-1␤ activates p38 and ERK 1/2 MAP kinases in orbital fibroblast cultures in a time-dependent manner. Confluent cultures were shifted to medium without FBS for at least 16 h and were then treated with IL-1␤ (10 ng/ml) for the indicated times. Cell lysates were harvested and subjected to SDS-PAGE, transferred to membranes, and Western blotted with phospho-specific antibodies against p38 and ERK. Films were subjected to densitometric analysis, the results of which are shown. B, effect of specific inhibitors of MAP kinases on the induction by IL-1␤ of PGHS-2 and mPGES. Cultures were shifted to medium without FBS for 24 h and then incubated for 16 h without or with IL-1␤, alone or in combination with PD98059 (10 M) or/and SB203580 (10 M). Cell lysates were collected and subjected to Western blot analysis with specific antibodies against PGHS-2 and mPGES. The resulting bands were analyzed for relative densities. and PGHS-2 proteins after 12 h (Fig. 5). The action of CD154 on prostanoid generation in orbital fibroblasts has been shown to result from an intermediate induction of IL-1␣ induction (31). FBS was also found to induce these enzymes when fibroblasts were incubated under reduced serum conditions and were then shifted to medium with 10% serum. A 20-fold induction of PGHS-2 and 4-fold up-regulation of mPGES expression resulted from serum treatment (data not shown). Thus, it would appear that agents found to provoke the

expression of PGHS-2 also enhance the levels of mPGES in orbital fibroblasts. Moreover, mPGES, like PGHS-2, appears to represent a proximate target for proinflammatory signals derived from T cells and conveyed through the CD40/CD154 activational bridge. IL-1␤ Activates ERK and p38 MAP Kinases in Human Orbital Fibroblasts: The Induction of PGHS-2 and mPGES Is Dependent on the Activities of ERK and p38 MAP Kinases- Studies examining the signal transduction pathways utilized by proinflammatory cytokines in the induction of PGHS-2 suggest that multiple pathways are involved. Notable among the pathways thus far implicated are the MAP kinase pathways (41,42). Nothing has thus far been reported about the signaling involved in the activation of mPGES expression by cytokines. Thus, we next compared the signaling pathways utilized in the activation of PGHS-2 and mPGES by IL-1␤ in orbital fibroblasts. The cytokine rapidly activates both p38 and ERK 1/2 MAP kinases (Fig. 6A). These effects are time-dependent and are sustained for at least 48 h. We treated orbital fibroblasts with specific kinase inhibitors in combination with IL-1␤ to determine whether the MAP kinase pathway was utilized in the induction of mPGES. When the activity of the p38 MAP kinase pathway was interrupted with SB203580 (10 M), the induction of PGHS-2 and mPGES by IL-1␤ was attenuated by 55 and 80%, respectively (Fig. 6B). PD98059 (10 M), a specific inhibitor of MEK, which is immediately up-stream from ERK, also blocked the induction of PGHS-2 and mPGES by 33 and 65%, respectively. These inhibitor concentrations have been shown to specifically inhibit their respective target pathways (43,44). When the two inhibitors were added together, the induction of both PGHS-2 and mPGES was blocked further. The combination of compounds reduced the respective induction by 83 and 89%. An alternate, molecular approach was then employed for interrupting the p38 and ERK MAP kinase pathways by transiently transfecting DN constructs for these kinases into orbital fibroblasts. The induction of mPGES protein by IL-1␤ is attenuated partially following transfection with either the p38 or ERK kinase DN constructs (Fig. 7). Control cultures were transfected with an empty vector. When both DN constructs are co-transfected into the same cultures, a near complete blockade of the mPGES induction can be appreciated. Thus, it would appear that both p38 and ERK kinase signaling pathways are involved in the induction of mPGES by IL-1␤ in orbital fibroblasts. This pathway utilization overlaps that involved in the induction of PGHS-2. Congruent results were obtained with pharmacological and molecular strategies for pathway interruption. Inhibition of PGHS-2 Activity Is Associated with a Decrease in IL-1␤-provoked mPGES Expression-A functional link between PGHS-2 and mPGES was implied by the earlier results of studies in HEK293 cells, where cDNAs encoding the two enzymes were co-transfected (13). A marked increase in PGE 2 production occurred when a low concentration of exogenous arachidonate (2-5 M) was added to the culture medium. When PGHS-1 was co-transfected with mPGES, increased PGE 2 production occurred only at a relatively high concentration of arachidonate (10 M) (13). We thus determined whether a functional relationship between endogenous PGHS-2 and mPGES exists in orbital fibroblasts. To ascertain whether the activity of PGHS-2 was directly influencing the expression of mPGES, fibroblasts were treated with IL-1␤ (10 ng/ml) alone or in combination with SC58125 (5 M) for 16 h. As the data in Fig. 8A indicate, IL-1␤ increases the levels of mPGES protein by 13-fold. Addition of the PGHS-2 inhibitor resulted in a 92% attenuation of that induction. Moreover, SC58125 also partially blocked the induction by IL-1␤ of mPGES mRNA (Fig. 8B). The magnitude of this inhibition was 70%. We have reported previously that SC58125, at comparable concentrations to those used in the current studies, inhibits cytokine-dependent PGE 2 production in orbital fibroblasts (25,26). To determine whether SC58125 exerts its effect on mPGES expression directly through the inhibition of PGHS-2 activity, exogenous FIG. 7. The effect of transfected DN p38 and ERK1 expression vectors on the induction of mPGES by IL-1␤ in orbital fibroblasts. 80 -90% confluent cultures were shifted to serum-free medium containing LipofectAMINE plus and the plasmid DNA constructs (2 g each) for 3 h as described under "Experimental Procedures." Control cultures received empty vector DNA. Cultures were shifted to medium supplemented with 10% FBS for 24 h and then to serum-free medium for 16 h. Subsequently, some plates received IL-1␤ (10 ng/ml) for an additional 16 h. The resulting monolayers were solubilized and subjected to Western blot analysis with anti-mPGES primary antibody. The right panel demonstrates the analysis of cultures transfected with empty vector, DN p38, or DN ERK followed by Western blotting with anti-p38 or ERK. arachidonate (10 M) was added to IL-1␤-treated cultures also receiving SC58125. Exogenous arachidonate partially restored the cytokine-provoked mPGES expression (Fig. 8A). The downregulation of mPGES expression resulting from SC58125 treatment (92%) was reduced to a 36% inhibition. This result suggests strongly that SC58125 acts on mPGES expression through its effects on PGHS-2 activity. DISCUSSION The up-regulation of PGE 2 synthesis by IL-1␤ in human orbital fibroblasts involves the coordinate induction of multiple enzymes in the prostanoid biosynthetic pathway. The levels of PGHS-2, an enzyme that catalyzes two rate-limiting intermediate reactions that ultimately yield PGH 2 , and mPGES, which facilitates the terminal reaction converting PGH 2 to PGE 2 (1), are both increased. Induction of the two enzymes occurs sequentially, with PGHS-2 mRNA and protein becoming elevated above base line more rapidly and transiently than those of mPGES. The susceptibility of cytokine-provoked mPGES induction to a highly selective inhibitor of PGHS-2 activity suggests that up-regulation of the former enzyme by IL-1␤ is dependent, at least in part, on a product of PGHS-2. This is further supported by the partial restoration of IL-1␤-dependent mPGES expression with high concentrations of exogenous arachidonate. Thus, it would appear that mPGES expression is linked functionally in the orbital fibroblast with the activity of PGHS-2. The current findings regarding the functional interaction of endogenously expressed mPGES and PGHS-2 are entirely consistent with previous findings in transfected cells over-expressing both enzymes (13). Those earlier studies indicate a preferential coupling of mPGES to PGHS-2. In addition, the level of arachidonate generated and its distribution among cellular compartments might also influence the efficiency with which PGHS-2 and mPGES interact (15). The dramatic downregulation of PGHS-2 and mPGES in these fibroblasts by physiologically relevant concentrations of dexamethasone would suggest that both enzymes are targets of steroid regulation. Thus, the impact of glucocorticoids on prostanoid synthesis is complex and apparently involves multiple levels of control. The up-regulation by IL-1␤ of the steady-state levels of mPGES and PGHS-2 mRNAs are a consequence, at least in part, of modest increases in the activities of their respective promoters (Fig. 4A). With regard to mPGES, this small increase (2-3-fold) is consistent with previously reported experience in A549 cells (38), where the promoter activity was only minimally influenced. With regard to PGHS-2, Wang et al. (25) reported that the proinflammatory cytokine leukoregulin increases steady-state levels of PGHS-2 mRNA dramatically in orbital fibroblasts, but the increase in PGHS-2 gene transcription, as assessed by nuclear run-on assays, was modest. Of considerable potential importance to the induction of PGHS-2 by IL-1␤ in these cells is the impact that the cytokine appears to exert on the stability of the mRNA (Fig. 4B). Our data clearly demonstrate a dramatic enhancement of PGHS-2 mRNA stability and are consistent with the structure of the 3Ј-untranslated region of human PGHS-2. That sequence contains a number of AUUUA instability elements, and the transcript has been shown to exhibit rapid turnover in a number of cell types (9,45). mPGES mRNA exhibits considerably greater stability and treatment of orbital fibroblasts with IL-1␤ failed to alter the survival of this already long-lived mRNA. The sustained induction of mPGES mRNA following IL-1␤ treatment (Fig. 3) thus can be explained on the basis of a small increase in gene transcription in the setting of a relatively stable mRNA. During the course of these studies, a report appeared from Thorén and Jakobsson (46) demonstrating that PGES activity in A549 cells could be inhibited by sulindac sulfide and the PGHS-2 selective agent, NS-398, with IC 50 values of 80 and 20 M, respectively. In contrast, the IC 50 for NS-398 on PGHS-2 is ϳ3.8 M (47). These findings are of considerable interest because they suggest structural similarities in the drug-binding sites on PGHS-2 and PGES. Coupled with our current observations that selective inhibition of PGHS-2 activity also interferes with mPGES expression, it is possible that many partic- The monolayers were then harvested and subjected to Western blot analysis with primary anti-mPGES antibody. B, confluent cultures were shifted to serumless medium overnight. Some plates were treated with IL-1␤ (10 ng/ml) alone or in combination with SC58125 (5 M) for 16 h. RNA was harvested and subjected to Northern blot hybridization. The membrane was rehybridized with a GAPDH probe, and the signals were used to normalize the levels. ularly effective anti-inflammatory drugs may target multiple enzymes in the PGE 2 biosynthetic cascade. The precise roles for PGE 2 in the initiation and evolution of the inflammatory response and in tissue fibrosis are uncertain, and the topics are open to considerable debate. Clearly the insinuation of dramatic PGE 2 production into the repertoire of the orbital fibroblast phenotype suggests a great potential for this prostanoid in conditioning the immune reactivity of the orbit, both in healthy and pathological states, such as those occurring in TAO. The capacity of resident fibroblasts to generate high levels of PGE 2 following activation by recruited bone marrow-derived cells could impact substantially on immune responses occurring there and could direct the pattern of tissue remodeling. This is particularly relevant with regard to the potential for fibrotic reactions in orbital muscles in late stage TAO, because that process, associated with substantial morbidity, has been thought to be the consequence of T H 2 predominance (48). Thus, our current finding that mPGES is also highly inducible in orbital fibroblasts defines an additional component of the inflammatory machinery that might be exploited as a potential therapeutic target. From the current studies, it would appear that IL-1␤ utilizes overlapping signaling pathways with regard to the induction of PGHS-2 and mPGES expression in orbital fibroblasts. Although nothing had been reported previously concerning signal transduction related to the mPGES expression provoked by cytokines, an extensive literature has evolved suggesting a complex array of pathway utilization up-stream from PGHS-2 gene activation in several cell types. Prominent among these are the p38 and ERK MAP kinases (43,44,49). We have reported that CD40 engagement with CD154 leads to the induction of PGHS-2 and that interruption of the ERK pathway substantially attenuates this cellular response in orbital fibroblasts (31). In the current studies, the addition of specific inhibitors of both p38 and ERK could partially attenuate the induction of PGHS-2 by IL-1␤. This was also true with regard to the induction of mPGES. Similar results were obtained when each of the two kinase pathways was interrupted by the transfection of DN constructs. The co-transfection of both DN plasmids led to a further and additive blockade of the induction. Taken together, it would appear that signaling from IL-1␤ to mPGES in orbital fibroblasts involves multiple pathways. This insight defines potential therapeutic targets for the interruption of PGE 2 -related inflammation in the orbit. Moreover, an overlap in cell signaling that culminates in the activation of both PGHS-2 and mPGES expression suggests that potentially important coordination exists between the enzymes. This association is further supported by the finding that cytokines inducing PGHS-2 also up-regulate mPGES expression. These results demonstrate a substantial capacity of IL-1␤ to induce multiple enzymes in the synthetic cascade for PGE 2 in orbital fibroblasts. Why these cultures should exhibit greater responses to proinflammatory cytokines than do other types of fibroblasts (25,26) is not clear, but it could be related to the normal function of orbital connective tissue. What emerges from our current findings is a further clarification of the molecular basis underlying the particularly robust involvement of orbital connective tissue in autoimmune inflammation, such as that associated with TAO. Caspases disrupt the nuclear-cytoplasmic barrier. During apoptosis, caspases, a family of proteases, disassemble a cell by cleaving a set of proteins. Caspase-3 plays a major role in the disassembly of the nucleus by processing several nuclear substrates. The question

is how caspase-3, which is usually cytoplasmic, gains access to its nuclear targets. It was suggested that caspase-3 is actively transported to the nucleus through the nuclear pores. We found that caspase-9, which is activated earlier than caspase-3, directly or indirectly inactivates nuclear transport and increases the diffusion limit of the nuclear pores. This increase allows caspase-3 and other molecules that could not pass through the nuclear pores in living cells to enter or leave the nucleus during apoptosis by diffusion. Hence, caspase-9 contributes to cell disassembly by disrupting the nuclear-cytoplasmic barrier. Introduction Apoptosis is cell suicide that is induced by various stimuli. During apoptosis, the cell is disassembled by caspases, a family of highly specific proteases (reviewed in Cryns and Yuan, 1998;Thornberry and Lazebnik, 1998;Budihardjo et al., 1999). Caspases are expressed constitutively as precursors, which are activated by proteolytic processing. The current model is that caspases are activated in a proteolytic cascade, which includes initiator and effector caspases, their inhibitors, and activators. Initiator caspases are activated by autoproteolysis, which is induced by binding to specific activators. Each initiator caspase is activated in response to a subset of cytotoxic stimuli. For example, caspase-8 is activated by binding to DISC, a protein complex, which is formed after stimulation of the death receptors such as Fas and TNFR (reviewed in Ashkenazi and Dixit, 1998). Caspase-9 is activated by binding to a complex containing APAF-1 and cytochrome c, which is formed after a number of stimuli, including DNA damage. Importantly, cytochrome c has to be released from mitochondria to participate in caspase-9 activation. The activated initiator caspases activate effector caspases by proteolytic processing. While initiator caspases are specific for each pathway of apoptosis, the effector caspases are shared. Initiator and effector caspases disassemble a cell by cleaving a set of proteins. This processing results in dismantling of cellular structures, disrupting metabolism, inactivating antiapoptotic activity of some proteins, and promoting pro-apoptotic activity of others (reviewed in Thornberry and La-zebnik, 1998;Porter, 1999). Remarkably, only a small fraction of proteins is cleaved, indicating that cell disassembly is highly efficient. Although it is not clear how the processing of caspase substrates leads to cell death, it appears that caspases target key components of cell structures and signaling pathways. For example, during apoptosis, DNA is degraded by CAD, a DNAse, which is constitutively expressed but is bound to its inhibitor ICAD/DFF45 in live cells (Liu et al., 1997;Enari et al., 1998). Caspase-3 releases active CAD by cleaving ICAD/DFF45 at only two sites . In another example, caspases disassemble the nuclear lamina, which is formed by polymers of lamins, by cleaving these proteins at a single site (Lazebnik et al., 1995;Orth et al., 1996;Takahashi et al., 1996). Interestingly, caspase-3 is primarily cytoplasmic (Chandler et al., 1998;Mancini et al., 1998;Zhivotovsky et al., 1999;this study), while the complex of CAD with ICAD/ DFF45 is nuclear (Liu et al., 1998;Earnshaw, 1998, 2000). Therefore, the question is, how does caspase-3 reach ICAD/DFF45 and its other nuclear targets? This question is part of a larger problem, which is how the apoptotic machinery, which is initially activated in the cytoplasm, reaches the rest of the cell (Porter, 1999). It was suggested that caspase-3 and perhaps other caspases are translocated into the nucleus by active transport and that transport is required for apoptosis (Yasuhara et al., 1997;Kuwana et al., 1998;Kohler et al., 1999). Protein import and export is regulated by a complex machinery, which includes soluble components and nuclear pores (reviewed in Mattaj and Englmeier, 1998;Nakielny and Dreyfuss, 1999;Talcott and Moore, 1999). Nuclear pores are gated channels composed of at least 50 proteins, many of which have not been characterized. Particles of Ͻ 9 nM in diameter or globular proteins of less than 50-60 kD can enter the nucleus by diffusion, whereas larger objects should be actively transported. To be imported through the nuclear pores, a protein should have a nuclear localization signal (NLS) 1 or be bound to an NLS-containing protein. The NLS is recognized by importins that form a transport complex with the protein. The complex translocates through the pores by a multi-step process, which involves interactions between the complex and nuclear pore components. Once in the nucleus, the transport complex is dissociated upon binding of Ran-GTP to the importins which release the protein. The dimeric complex of importin and Ran-GTP is returned to the cytoplasm, where Ran-GTP is converted into Ran-GDP by the subsequent action of Ran-GAP and Ran-BPs. Ran-GDP is translocated into the nucleus as a complex with NTF2, where it is regenerated into Ran-GTP by RCC1. Nuclear export occurs similarly except that the exported protein forms a complex with exportins and Ran-GTP. According to this model, both import and export of proteins requires a gradient of Ran-GTP in which a majority of the protein is concentrated in the nucleus (Izaurralde et al., 1997). To understand how the cytoplasmic apoptotic machinery disassembles the nucleus, we investigated how caspase-3 reaches its nuclear substrates. We found that nuclear transport is inactivated after the activation of caspase-9. Furthermore, caspase-9 directly or indirectly increases the diffusion limit of nuclear pores, thereby allowing caspase-3 and other molecules to enter the nucleus by diffusion. Hence, caspase-9 promotes cell disassembly by disrupting the nuclear-cytoplasmic barrier. Materials and Methods Cell Culture MCF-7 cells (American Type Culture Collection) were maintained in MEM media supplemented with 10% fetal bovine serum and nonessential amino acids. Induction of Apoptosis Apoptosis in cells was induced with 50 M cisplatin (Sigma-Aldrich), 1 M staurosporine, or 35 g/ml TNF (Pharmingen) plus 10 g/ml cycloheximide (Sigma-Aldrich) as indicated. Apoptosis was monitored by release of cytochrome c and/or chromatin condensation. Immunofluorescence/Confocal Microscopy Cells were grown on poly-L -lysine (Sigma-Aldrich)-coated glass coverslips, fixed in PBS containing 4% paraformaldehyde, permeabilized in PBS containing 0.2% Triton and 0.5% BSA, and blocked with 2% BSA and 2% normal goat serum. Primary antibody incubation was performed according to the manufacturers' protocols (exceptions are described below) and was followed by secondary antibody conjugated to Alexa 594 or Alexa 488 (Molecular Probes). Cells were stained with either 4 Ј 6-diamidino-2-phenylindole (DAPI) or propidium iodide and coverslips were mounted using Prolong Antifade (Molecular Probes). Monoclonal antibody to caspase-3 (0.5 g/ml; Transduction Laboratory) was incubated with 1% acetone powder prepared from MCF-7 cells for 1 h at room temperature, and clarified by centrifugation (10,000 g , 30 min) before using for immunofluorescence. Images were acquired either on a Zeiss Axiophot microscope equipped with a Photometrics SenSys cooled CCD camera using Image 2.0.5 software (Oncor) or on a confocal laser scanning microscope (LSM410; Carl Zeiss, Inc.). Retroviral Gene Transduction cDNAs of interest were cloned into a MarX-IV-neo (caspase-3 and caspase-3-green fluorescent protein, GFP) or MarX-IV-puro (caspase-9) retroviral gene transfer vector (Hannon et al., 1999). Retrovirus was produced by transfection into LinX-A packaging cells (L.Y. Xie, D. Beach, and G.J. Hannon, unpublished observations). Media from transfected LinX cells were supplemented with 8 g/ml of polybrene, 10% FBS, and nonessential amino acids and added to plates of MCF-7 cells. Plates were centrifuged at 1,000 g for 1 h, and then incubated for 12-18 h at 32 Њ C. Media were then replaced after 2 d and infected cells were selected using 1 g/ml puromycin or 600 g/ml Geneticin for 4 or 7 d, respectively. Infected cell lines were maintained in MEM media supplemented with 0.5 g/ml puromycin or 300 g/ml Geneticin. Caspase-3 Translocates to the Nucleus during Apoptosis To investigate how caspase-3 reaches its nuclear substrates during apoptosis, we used a breast carcinoma cell line MCF-7, which does not express caspase-3 because of a mutation in the caspase-3 gene (Janicke et al., 1998). As a result of the mutation, MCF-7 cells undergo apoptosis without typical apoptotic nuclear changes, such as DNA cleavage and condensation of chromatin into distinct round particles, although some changes in chromatin structure are detectable (Woo et al., 1998). Expression of caspase-3 in MCF-7 cells restores both DNA fragmentation and chromatin condensation, providing a model system to study the function of caspase-3 and its derivatives (Janicke et al., 1998). To monitor caspase-3 localization in live and apoptotic cells, we decided to use a fusion of this protein with GFP, which would eliminate potential problems associated with the detection of proteins in apoptotic cells by immunofluorescence. In particular, we were concerned that chromatin condensation would make nuclear proteins less accessible to the antibodies. First, we investigated whether fusing caspase-3 to GFP affects the function of this protease. We used a retroviral gene transduction system (Hannon et al., 1999) to make MCF-7 cell lines that express caspase-3 (MCF-7/C3) or a fusion of caspase-3 to E-GFP (MCF-7/ C3-GFP). Both caspase-3 and the fusion protein were expressed at levels comparable with the levels of caspase-3 in other breast carcinoma cell lines ( Fig. 1 A). Consistent with the previous reports, we detected no caspase-3 in MCF-7 cells ( Fig. 1 A). Both MCF-7/C3 and MCF-7/C3-GFP cells, but not the parental cell line, underwent typical apoptotic chromatin condensation after treatment with cisplatin, an anticancer drug that induces DNA damage ( Fig. 1 B). This indicated that the fusion to GFP did not affect the ability of caspases-3 to induce nuclear changes of apoptosis. This conclusion was also supported by the finding that caspase-3 activity in extracts from apoptotic MCF-7/C3 and MCF-7/C3-GFP cells was similar when measured with DEVD-AFC, a peptide substrate for caspase-3 (data not shown). Consistent with the previous reports (Krajewska et al., 1997;Chandler et al., 1998;Mancini et al., 1998;Zhivotovsky et al., 1999), caspase-3 was found primarily in the cytoplasm in live cells, as was caspase-3-GFP (Fig. 1 C). Because caspase-3 and its fusion to GFP were similar in their activity and localization, and because the fusion protein could be visualized more reliably, we used caspase-3 GFP in this study. Caspase-3 Entry into the Nucleus Requires Active Caspase-9 Apoptosis can be considered as a two-step process in which the first step is a preparation for activating caspases and the second is the activation and its consequences. We asked whether caspase-3 translocation to the nucleus occurs before activation of caspases or requires their activity. Because we used a drug that damages DNA and because apoptosis induced by DNA damage is often mediated by caspase-9, we tested whether cisplatin kills MCF-7 cells through the caspase-9 pathway. Treatment with cisplatin released cytochrome c from mitochondria, indicating that the caspase-9 pathway is initiated. The release of cytochrome c correlated with translocation of caspase-3 into the nucleus and with typical chromatin condensation (Fig. 2, B and C). To test whether caspase-9 activity is required, we used retroviral gene transduction to make a cell line that expressed caspase-9 dominant-negative mutant (C9DN) (MCF-7/C3-GFP/C9DN). This mutant prevents the activation of caspase-9, but not the preceding steps of apoptosis by competing with the endogenous caspase-9 for binding to APAF-1. Consistent with previous reports (Fearnhead et al., 1998), expression of C9DN did not affect release of cytochrome c from mitochondria, but prevented chromatin condensation (Fig. 2, B and D) and detachment of the cells (data not shown). These observations were consistent with the notion that cisplatin induces apoptosis in MCF-7 by activating caspase-9. The entry of caspase-3-GFP into the nucleus during apoptosis was blocked in MCF-7/C3-GFP/C9DN cells (Fig. 2, B and C), indicating that caspase-9 activity is required for caspase-3 translocation. To determine whether caspase-3 activity or Total cell lysates prepared from MCF-7/C3, MCF-7/C3S (catalytically inactive C3 mutant), MCF-7/C3-GFP, and MCF-7/C3S-GFP, and the three indicated breast carcinoma cell lines were analyzed by immunoblotting using a monoclonal antibody to caspase-3. 50 g of lysate were loaded per lane. (B) MCF-7 cells that express either caspase-3 or caspase-3-GFP undergo typical apoptotic chromatin condensation. MCF-7, MCF-7/C3, and MCF-7/C3-GFP cells were treated with 50 M cisplatin for 12 h, fixed, stained with DAPI to visualize chromatin, and scored by fluorescence microscopy for typical apoptotic chromatin condensation (indicated with white arrows). 60% of cells expressing C3 and 64% of cells expressing C3-GFP had typical apoptotic nuclear morphology, which was not observed in MCF-7 cells. (C) Caspase-3 and caspase-3-GFP have similar, cytoplasmic localization. Caspase-3 in MCF-7, MCF-7/C3, and MCF-7/C3-GFP cells was visualized by fluorescence microscopy using a monoclonal antibody to caspase-3, as described in Materials and Methods. As expected, no caspase-3 was detected in MCF-7 cells. processing were required for translocating this caspase, we made cell lines that express various caspase-3 mutants, including those with mutated

catalytic cysteine, lacking some or all processing sites (D 9 , D 28 , D 175 ), or missing the prodomain (M 1 to D 28 ). All the mutants translocated into the nucleus similarly to caspase-3 (data not shown). Hence, caspase-9 activity is required for caspase-3 to enter the nucleus, whereas caspase-3 activity is not. The Nuclear Transport Is Altered during Apoptosis Previous reports suggested that caspase-3 is translocated into the nucleus by active transport (Yasuhara et al., 1997). Active transport typically concentrates a protein in a particular compartment. However, we observed that caspase-3 concentrations in the nucleus and cytoplasm equilibrated. Therefore, we tested whether active nuclear transport functions during apoptosis. A common assay to monitor nuclear transport is to use GFP fused to three copies of the SV40 T-antigen nuclear localization signal (GFP-NLS), a soluble molecule that accumulates in the nucleus of living cells (Roberts and Goldfarb, 1998). As expected, GFP-NLS transiently expressed in MCF-7 cells localized to the nucleus (Fig. 3, A and B). However, after treatment with cisplatin, GFP-NLS was distributed throughout the cell. The release of GFP-NLS from the nu- Caspase-3-GFP (GFP, green) and the nuclei (PI, red), were visualized by confocal microscopy. An arrow indicates a cell in which chromatin begins to condense. Cells at later stages of apoptosis were difficult to observe because they shrink and detach. (B-D) Translocation of caspase-3 into the nucleus during apoptosis requires caspase-9 activity. (B) MCF-7/C3-GFP and MCF-7/C3-GFP cells expressing a dominant-negative mutant of caspase-9 (MCF-7/C3-GFP/C9DN) were treated with 50 M cisplatin for 8 h. The cells were stained with DAPI to visualize the chromatin and with a cytochrome c antibody to detect the release of this protein from mitochondria. GFP, DAPI, and cytochrome c were visualized by fluorescence microscopy. White arrows indicate cells with released cytochrome c. (C) The rate of caspase-3 translocation into the nucleus is similar to that of chromatin condensation, and of cytochrome c release from mitochondria. MCF-7/C3-GFP cells were treated and stained as in B, except cells were fixed at the indicated time. The incidence of C3-GFP translocation to the nucleus, cytochrome c release, and change in chromatin structure was scored. (D) Dominant-negative mutant of caspase-9 (C9DN) prevents C3-GFP translocation to the nucleus and chromatin condensation. MCF-7/C3-GFP and MCF-7/C3-GFP/C9DN cells were treated and stained as in B. The incidence of C3-GFP translocation to the nucleus, cytochrome c release and change in chromatin structure was scored. cleus, like the equilibration of caspase-3 between the nucleus and cytoplasm, was prevented in MCF-7 cells that expressed caspase-9 dominant-negative mutant (Fig. 3, A and B). These observations were consistent with the notion that active transport is altered during apoptosis and that this alteration is caused by the activity of caspase-9. A prerequisite of active nuclear transport is the nuclearcytoplasmic gradient of Ran-GTP, in which a majority of the protein is concentrated in the nucleus (Izaurralde et al., 1997). We found that Ran, like GFP-NLS, was released from the nucleus during apoptosis and distributed throughout the cell (Fig. 3, C and D). As we observed with GFP-NLS, expression of the caspase-9 dominant-negative mutant prevented Ran release. The dissipation of the Ran gradient and the release of NLS-GFP were both consistent with the idea that nuclear transport is inactivated in apoptosis. Hence, caspase-3 would have to enter the nucleus by other means. Caspase-3 could be actively exported from the nucleus in living cells. In this case, inactivating nuclear transport would lead to diffusion of caspase-3 into the nucleus. We have not identified any obvious nuclear export signals (NES) in caspase-3. Deletion of one motif that might be an NES (I 20 to D 34 , L. Englmeier, unpublished observations) had no effect on caspase-3 localization (data not shown). Therefore, we considered an alternative possibility, that caspase-3 is excluded from the nucleus because of the diffusion limit of the nuclear pores. Although caspase-3 is a 32-kD protein, which could diffuse into the nucleus, it is conceivable that this caspase may form complexes with other proteins, such as caspase inhibitors, or it may oligomerize. We reasoned that if caspase-3 or caspase-3-containing complexes are excluded because of their size, then the diffusion limit of the nuclear pores would have to increase for caspase-3 to enter the nucleus during apoptosis. (C) Activation of caspase-9 leads to release of Ran from the nucleus. MCF-7 and MCF-7/C9DN cells were treated as in A, fixed, and stained with a monoclonal antibody to Ran, a polyclonal antibody to cytochrome c, and DAPI to visualize nuclei. The antibodies were visualized with Alexa488-conjugated anti-mouse and Alexa596-conjugated anti-rabbit secondary antibodies. White arrows indicate cells with released cytochrome c. Cells were scored for the release of cytochrome c, loss of nuclear Ran staining, and change in chromatin structure (D). Note that expression of C9DN causes the decrease in the ratio between the cells that release cytochrome c and Ran. 200-250 cells were counted for each measurement in all experiments. Shown is a representative of three experiments. Nuclear Diffusion Limit Is Increased during Apoptosis To determine whether the diffusion limit of nuclear pores is altered during apoptosis, we transiently expressed GFP oligomers ranging in size from a single GFP molecule (28 kD) to a GFP pentamer (140 kD). In addition, to test a large molecule, we used a fusion of GFP with bacterial ␤ -galactosidase, a soluble enzyme that is used as a tracer or reporter in mammalian cells (GFP-␤ -Gal), although we have no direct proof that GFP-␤ -Gal can freely diffuse in the cell. Considering that-␤ -Gal is a tetramer, the predicted size of GFP-␤ -Gal is ‫ف‬ 580 kD. As expected, in untreated cells, GFP was present throughout the cell, whereas the GFP dimer (56 kD), GFP pentamer, and GFP-␤ -Gal were excluded from the nucleus (Fig. 4, A and B). In apoptotic cells, the GFP dimer and pentamer were observed in both the nucleus and the cytoplasm, whereas GFP-␤ -Gal remained in the cytoplasm. We also found that GFP-␤ -Gal containing a nuclear localization signal remained in the nucleus of apoptotic cells (data not shown). Considering that the GFP oligomers are not substrates of the nuclear transport machinery, it is reasonable to conclude that they enter the nucleus during apoptosis by diffusion. Therefore, the permeability of the nuclear pores increases during apoptosis, allowing even large proteins and complexes to freely enter the nucleus during apoptosis. Hence, caspase-3 could enter the nucleus simply because the orifice of the nuclear pore increases. It is conceivable that the nuclear pores remain intact but the permeability is increased by perforation of the nuclear membranes. However, the observations that nuclear membranes remain intact during apoptosis argue against this model (Wyllie et al., 1980;Allen, 1987). Consistent with these reports, we found no evidence of membrane perforations in apoptotic MCF-7 cells that expressed C3-GFP by electron microscopy (data not shown). The observation that the nuclear-cytoplasmic barrier remains impermeable for GFP-␤ -Gal also argues against the perforation model. Another possibility is that the collapse of the nuclear lamina during apoptosis contributes to the increase in nuclear pore permeability. Lamins, which form the lamina, are thought to be cleaved by caspase-6, which is processed and activated by caspase-3 (Slee et al., 1999). Consistent with this, we found that lamin B1 remained intact in apoptotic MCF-7 cells, in which the nuclear permeability is increased, although it was processed in the cells that expressed caspase-3 or caspase-3 GFP (data not shown). Therefore, cleavage of lamins is not required for the increase in nuclear permeability during apoptosis. Nuclear Pores Are Altered during Apoptosis Because caspases disassemble a cell by processing their substrates, and because the increase in nuclear permeability required caspase activity, we reasoned that some components of the nuclear transport machinery, either soluble or insoluble, may be caspase targets. We found that none of the tested proteins were cleaved in MCF-7 cells (Table I), although two proteins, RanGap1 (a soluble component of the transport machinery) and p270/TPR (a component of the nuclear pore complex) were processed in MCF-7 cells that expressed caspase-3. Because nuclear transport is disrupted even without caspase-3, we concluded that processing of these two proteins is not required for the increase in nuclear permeability. We confirmed that nuclear pore protein NUP153 is cleaved in apoptotic HeLa cells (Buendia et al., 1999), although we failed to detect it in MCF-7 cells (data not shown), thus leaving open the question as to whether processing of this protein is required for the increase in nuclear permeability. It should be emphasized that our inspection of the transport machinery was limited by the available tools. We tested only 5 of the Ͼ 50 proteins of the vertebrate nuclear pore, many of which have not been characterized (Doye and Hurt, 1997;Talcott and Moore, 1999). This leaves open the possibility that other nuclear pore complex components are caspase targets and their processing causes an increase in nuclear permeability. For example, deficiency in yeast nup188 and nup170 results in an increase of the nuclear permeability (Shulga et al., 2000), suggesting vertebrate homologues of these proteins as potential targets of caspases. Considering that caspases usually target the key elements of a system or structure, the putative caspase substrates in the nuclear pore are likely to be critical to the pore function in living cells. Since the inspection of the nuclear pore components provided no clues for the increase in the permeability, we considered several speculative mechanisms. One speculation was that caspase activity leads to removal of the central plug of the nuclear pore, which would explain the increase in the nuclear permeability. To test this possibility, we used antibodies to p62, a component of the plug (Talcott and Moore, 1999). As expected, the antibody detected nuclear pores in untreated MCF-7 cells, which appeared as a typical nuclear rim, whereas hardly any nuclear rim was detected in apoptotic cells (Fig. 5 A). We found that p62 is intact in apoptotic cells as detected by immunoblotting and reasoned that this protein may become soluble if the plug is removed. However, we found that p62 detectable by immunoblotting remained with the nuclear fraction of both living and apoptotic cells during cell fractionation (data not shown). Furthermore, a monoclonal antibody 114 (mAb114) (Davis and Blobel, 1987), which detects several nuclear pore proteins in the plug and beyond, also failed to detect nuclear pores in apoptotic cells (Fig. 5 B), arguing against the plug hypothesis and suggesting that many pore proteins are affected. The failure to detect the proteins may be due to a modification of the epitopes in apoptotic cells or a failure of the antibodies to access them. Interestingly, a fusion of nuclear pore membrane protein POM121 with GFP was reported to be less detectable in apoptotic cells (Imreh et al., 1998), consistent with some changes in the nuclear pores other than a change in accessibility of the epitopes. Hence, it appears that some alterations in nuclear pores accompany the increase in their permeability during apoptosis, although what these changes are remains to be determined. A practical implication of the disappearance of nuclear pore staining during apoptosis may be in using this observation to identify apoptotic cells that, like MCF-7, do not have typical apoptotic morphology. We found that nuclear pores were hardly detectable by immunofluorescence during apoptosis not only in MCF-7 cells, but also all other cell lines that we tested (HeLa, Cos, U2OS, RKO, and U87) (data not shown), indicating that the apparent disappearance of the nuclear pore staining is a common phenomenon. Caspase-8 Requires Caspase-9 Activity to Disrupt the Nuclear-Cytoplasmic Boundary A current view is that common apoptotic changes can be induced by activating one of the several initiator caspases, such as caspase-8 or caspase-9. Therefore, we tested whether initiator caspases other than caspase-9 can inactivate nuclear transport. Because caspase-8 can indirectly activate caspase-9 Luo et al., 1998), we evaluated the effect of caspase-8 activity not only in MCF-7 cells, but also in the cells that express the dominant-negative mutant of caspase-9 (MCF-7/C9DN). Apoptosis in these cell lines was induced by TNF-␣ in the presence of cycloheximide, which, for unclear reasons, potentiates the NC C MCF-7 and MCF-7/C3 were either left untreated or were treated with 50 M cisplatin for 12 h, which resulted in Ͼ 75% apoptotic cells as scored by chromatin condensation. Cell lysates were analyzed by immunoblotting using monoclonal antibodies

to nuclear transport proteins. NC, not cleaved; C, cleaved; ND, not detected. Figure 5. Nuclear pores in apoptotic cells are not detectable by antibodies to nuclear pore proteins. MCF-7 were treated with 50 M cisplatin for 10 h, fixed, and stained with DAPI to visualize chromatin and with antibodies to nuclear pore protein np62 (A) or monoclonal antibody 414, which reacts with multiple nuclear pore proteins (B). White arrows indicate cells that have lost nuclear pore staining and have condensed chromatin. The Journal of Cell Biology, Volume 151, 2000 958 effect of TNF-␣ . As expected, treatment with TNF-␣ released cytochrome c (Fig. 6 A) and processed caspase-8 in both cell lines (Fig. 6 C). However, Ran was released only from the nucleus of MCF-7 cells, but not of the cells that expressed C9DN (Fig. 6, A and B), indicating that, in the absence of caspase-3, caspase-8 alone could not change the nuclear pore permeability, but required the activity of caspase-9 or its substrates. The finding that caspase-8 fails to process caspase-7, a substrate of caspase-9, in MCF-7/ C9DN cells (Fig. 6 C) leaves both possibilities open. In summary, activation of caspase-9 during apoptosis increases permeability of the nuclear pore, which allows cytoplasmic caspases to reach their nuclear substrates and lets soluble proteins that are normally restricted to the nucleus or cytoplasm to distribute throughout the cell. Figure 6. Activity of caspase-9 is required for the change in nuclear permeability even if apoptosis is induced by activating caspase-8. Caspase-9 activity is required for the release of Ran from the nucleus during apoptosis induced by TNF-␣. MCF-7 and MCF-7/C9DN cells were treated with 35 ng/ml TNF-␣ and 10 g/ml cycloheximide for 6 h, fixed, and stained with a monoclonal antibody to Ran and a polyclonal antibody to cytochrome c. White arrows indicate cells that have released cytochrome c. Cells were scored for the release of cytochrome c, loss of nuclear Ran staining, and change in chromatin structure (B). Note that expression of C9DN causes the decrease in the ratio between the cells that release cytochrome c and Ran. The processing of caspases-9, -8 and, -7 was confirmed by immunoblotting (C). Caspase precursors are indicated with black and the processed caspases with white arrows. Note that C9DN prevents processing of caspases-9 and -7, but not of caspase-8. Two experiments were performed, and Ͼ200 cells were scored for each experiment. Lymphopenia in hospitalized patients and its relationship with severity of illness and mortality Background Lymphopenia is associated with various pathologies such as sepsis, burns, trauma, general anesthesia and major surgeries. All these pathologies are clinically expressed by the so-called Systemic Inflammatory Response Syndrome which does not include lymphopenia into defining criteria. The main objective of this work was to analyze the diagnosis of patients admitted to a hospital related to lymphopenia during hospital stay. In addition, we investigated the relationship of lymphopenia with the four levels of the Severity of Illness (SOI) and the Risk of Mortality (ROM). Method and findings Lymphopenia was defined as Absolute Lymphocyte Count (ALC) <1.0 x109/L. ALC were analyzed every day since admission. The four levels (minor, moderate, major and extreme risk) of both SOI and ROM were assessed. A total of 58,260 hospital admissions were analyzed. More than 41% of the patients had lymphopenia during hospital stay. The mean time to death was shorter among patients with lymphopenia on admission 65.6 days (CI95%, 57.3–73.8) vs 89.9 (CI95%, 82.4–97.4), P<0.001. Also, patients with lymphopenia during hospital stay had a shorter time to the mortality, 67.5 (CI95%, 61.1–73.9) vs 96.9 (CI95%, 92.6–101.2), P<0.001. Conclusions Lymphopenia had a high prevalence in hospitalized patients with greater relevance in infectious pathologies. Lymphopenia was related and clearly predicts SOI and ROM at the time of admission, and should be considered as clinical diagnostic criteria to define SIRS. Introduction Lymphopenia is associated with various pathologies such as sepsis, burns, trauma, general anesthesia and mayor surgeries. This leads to a life-threatening period of immunosuppression [1][2][3][4][5]. This situation represents a serious public health problem, since both incidence and mortality from these processes are increasing [6][7][8]. All these pathologies have in common the clinical expression of the so-called Systemic Inflammatory Response Syndrome (SIRS), which includes a number of clinical and laboratory criteria such as leukocytosis and leukopenia, but does not include lymphopenia [9]. There is a need for a consensual definition of SIRS and sepsis, since there is sometimes difficulty, with the current criteria, to define pathological cases with SIRS caused by infection, since it is not easy to identify the infectious focus in some situations. Therefore, those criteria should be better adjusted to avoid erroneous diagnoses. Persistent lymphopenia is an independent predictor of increased mortality in critically ill emergency general surgical patients [10], septic patients [11], and outpatients [12,13]. However, to our knowledge, lymphopenia has not been studied in all the pathologies that make up the repertoire of diagnoses of patients admitted to hospital. The All Patients Refined Diagnosis Related Groups (APR-DRGs) are a patient classification scheme by means of several attributes which include severity of illness, risk of dying, prognosis, treatment difficulty, need for intervention, and resource intensity. Severity of Illness (SOI) refers to the extent of physiologic decompensation or organ system and Risk of Mortality (ROM) refers to the likelihood of dying. There are four SOI and ROM subclasses which are numbered sequentially from 1 to 4 indicating respectively, minor (1), moderate (2), major (3), or extreme (4) severity of illness or risk of mortality [14][15][16]. The main objective of this work was to analyze the diagnosis of patients admitted to hospital related to lymphopenia on admission and during hospital stay. In addition, we investigated the relationship of lymphopenia with the four levels of SOI and ROM according to DRGs. Study design and obtaining information This retrospective cohort study was conducted at the Arnau de Vilanova-Llíria Health Department in Valencia, Spain, including two hospitals: Arnau de Vilanova's Hospital and Llíria's Hospital, with a total of 450 beds, which cover an area of 300,000 inhabitants. Patients older than 14 years were admitted from January 1, 2016, to December 31, 2019, from Emergency Department (80%) and Scheduled Admission (20%). A file was generated with all analytical parameters of patients seen in the emergency room and hospitalization in the four years of the study. A cross was made with hospital episodes and emergency data. The resulting file was crossed, in turn, with the Database of DRGs from those same years. We obtained a Database of 72,984 records that combines data related to the hospitalization episode (date of admission, date of discharge, reason for admission, diagnoses, procedures); data related to the patient (age, sex); data related to DRGs (DRG, CMD, Severity, Risk of Mortality). Finally, patients who had undergone in-hospital laboratory tests were selected, ordered by evolution of the laboratory results in the admission episode, obtaining a sample of 58,860 patients. Clinical diagnoses were coded using International Classification of Diseases (ICD-10, 2016 version). The study protocol was approved by the Ethics and Investigation Committee of the Arnau de Vilanova-Llíria Hospital (approval number 2019-12), Valencia City, Spain. Study variables Demographic. Age and gender. Hospital stay, up to 15 diagnoses associated with each patient and at each admission, all blood counts performed in each episode, and in-hospital mortality. Lymphopenia was defined as Absolute Lymphocyte Count (ALC) <1.0 x10 9 /L. ALC were analyzed every day from admission. The mean of the analyses performed each day of hospitalization was calculated. We assessed the Severity of Illness (SOI) and the Risk of Mortality (ROM), with its four levels, minor (1), moderate (2), major (3), and extreme risk (4). Statistical analysis When normality was assumed (Kolmogorov-Smirnov test), Student t test was used to compare the means between two different groups of patients. When the hypothesis of normality was not accepted, the Mann-Whitney U test was used. Two-tailed Fischer's exact test was used to compare lymphopenia with diagnosis (Odds Ratio). Logistic regression (Exp B) and Cox regression (Hazard Ratios) were used to compare the two groups (with lymphopenia and without lymphopenia) taken into account analytical variables. Repeated Measures ANOVA were used to assess the differences of lymphocytes frequency between dead and alive patients over time. The Kaplan-Meier survival method was used to estimate the cumulative probability of mortality according to lymphopenia on admission and lymphopenia during hospital stay. Differences between curves were tested using the Log-Rank test (Statistical program R, version 3.3). A Cox proportional hazards regression model was used to assess whether lymphopenia variables were independently associated with the risk of mortality. P value <0.05 was considered statistically significant. Characteristics of admitted patients and diagnosis A total of 58,260 hospital admissions were analyzed. The mean age was 67.8 ± 18.4 years (66.4 ± 17.7 in males and 69.5 ± 19.0 in females, P<0.001). The mean age of patients with lymphopenia was 73.5 ± 15.6 years vs 63.9 ± 19.2 in patients without lymphopenia, p<0.001. The mean of hospital stay was 6.5 ± 6.6 days (6.8 ± 6.9 in males and 6.1 ± 6.1 in females, p <0.001). The mean of hospital stay in lymphopenic patients was 7.7 ± 7.6 days vs 5.6 ± 5.7 days in no lymphopenic patients (p<0.001). A total of 23,892 (41.0%) patients presented lymphopenia during hospital stay. Eighteen thousand three patients (30.9%) presented lymphopenia on admission. The characteristics of the admitted patients and the relationship of the variables studied according to lymphopenia are shown in Table 1. Table 2 shows the relationship of lymphopenia according to Major Diagnostic Categories (MDC). The patients with Septic Shock were 515/58.260 (0.9%), 446 (86.6%) of which had lymphopenia. Only 69 (13.4%) patients with septic shock did not have lymphopenia, OR = 9.5 (7.3-12.2), p<0.001. S1 Appendix shows the number of diagnoses in each patient according to disease groups. Each patient can have multiple diagnoses, so there are more diagnoses than patients. S1 Appendix shows the number of patients who have one or more diagnoses at the same time according to disease groups. Finally, it can be seen that the sum of the totals is equal to the sum of the diagnoses. The Kaplan-Meier curve for the risk of mortality is shown in Fig 2. Kaplan-Meier plots illustrating Survival probability following lymphopenia on admission (A) and lymphopenia (Fig 2C). Relationship of lymphopenia with severity and mortality risk has been described in Fig 2, panel D and E, respectively. Table 3 shows relationship between leukocyte and platelet counts with mortality rates. Discussion This is the first study to evaluate the association of lymphopenia with all hospital diagnoses in such a large number of patients. Highlights the high prevalence of lymphopenia in our study, since more than 30% of the patients admitted to the hospital had lymphopenia on admission. In addition, the incidence increased to 41% when including lymphopenia during hospital stay. Lymphopenia was more common in men than in women. On the other hand, lymphopenia was higher among patients admitted for surgical pathologies vs medical processes and it was also more frequent in Emergency compared to Scheduled Admission. The greatest decrease in lymphocytes was observed in pathological processes associated with infection. The lowest values were detected in sepsis and septic shock [17,18]. The most probable cause of this transient phenomenon of immunoparalysis with decreased of lymphocytes has been attributed to increased apoptosis of this cells [19,20]. Considering the importance of lymphocytes in the host's immune defense, it is not surprising the relationship of lymphopenia with severity and mortality. This fact was previously observed in other studies, although in a smaller number of patients. A prospective Danish population-based study, in subjects without acute pathology and underwent voluntary analysis, demonstrated an association between lymphopenia and increased risk of hospitalization and infection-related death [13]. We previously observed that lymphopenia was present in 75% of patients admitted to the hospital with sepsis. An inverse relationship of γδ T cells with disease severity and mortality was observed in septic patients [21]. In this work, a significant direct relationship between SOI and ROM with lymphopenia has been demonstrated. SOI and ROM variables are not available for the clinician at the time of admission. However, lymphocyte count and other objective parameters are available at the time of admission. The definition of SIRS is based on clinical and analytical criteria easily obtained and quickly accessible to clinicians. Leukocytes and especially total

neutrophils and "band" neutrophils are routinely performed in the emergency laboratory. The interpretation of leukocytosis and leukopenia is used as criteria for SIRS and infection. In this work we showed that lymphopenia has a higher predictive value for mortality than the measurement of leukocytes and total neutrophils. The inclusion criteria for SIRS did not allow distinguishing uninfected patients from septic patients. However, as we demonstrated in our study, lymphopenia is clearly related to infectious pathology, being much more marked in sepsis and in septic shock [9]. PLOS ONE A greater association was observed between mortality and lymphopenia during hospitalization than with lymphopenia on admission. This finding could be due to the fact that the percentage of patients who die increased progressively according to the number of analyzes performed during their hospitalization. In summary, the decrease in lymphocytes should be valued by physicians as a factor to consider in the prognosis of their patients. In conclusion, lymphopenia should be included among the criteria that evaluate the SOI and ROM at the time of admission. Furthermore, lymphopenia should be a criterion for evaluating SIRS patients with infectious and non-infectious causes at the time of admission. Limitations Our study has the limitations of retrospective studies. In addition, it lacks pediatric patients and patients with large burns since these pathologies are centralized in another hospital center whose database we have not been able to access. In any case, we think that the number of patients evaluated in the present study is large enough to be able to draw solid conclusions. Conclusion Lymphopenia has a high prevalence in patients admitted to a general hospital, with greater relevance in subjects with infectious pathology. Furthermore, lymphopenia is related and clearly predicts the Severity of Illness and the Risk of Mortality at the time of admission, and should be considered as a clinical diagnostic criteria to define SIRS patients with infectious and noninfectious causes at the time of admission. Supporting information S1 (DOCX) S1 Appendix. Number of diagnoses in each patient according to disease groups. Each patient can have multiple diagnoses, so there are more diagnoses than patients. S1 Appendix shows the number of patients who have one or more diagnoses at the same time according to disease groups. Finally, it can see that the sum of the totals is equal to the sum of the diagnoses. (DOCX) Chk1 Inhibition Restores Inotuzumab Ozogamicin Citotoxicity in CD22-Positive Cells Expressing Mutant p53 Inotuzumab ozogamicin (IO) is an anti-CD22 calicheamicin immunoconjugate that has been recently approved for the treatment of relapsed or refractory B-Acute Lymphoblastic Leukemia (r/r B-ALL). We employed both immortalized and primary cells derived from CD22-positive lymphoproliferative disorders to investigate the signaling pathways contributing to IO sensitivity or resistance. We found that the drug reduced the proliferation rate of CD22-positive cell lines expressing wild-type p53, but was remarkably less effective on cells exhibiting mutant p53. In addition, CD22-positive cells surviving IO were mostly blocked in the G2/M phase of the cell cycle because of Chk1 activation that, in the presence of a wild-type p53 background, led to p21 induction. When we combined IO with the Chk1 inhibitor UCN-01, we successfully abrogated IO-induced G2/M arrest regardless of the underlying p53 status, indicating that the DNA damage response triggered by IO is also modulated by p53-independent mechanisms. To establish a predictive value for p53 in determining IO responsiveness, we expressed mutant p53 in cell lines displaying the wild-type gene and observed an increase in IO IC50 values. Likewise, overexpression of an inducible wild-type p53 in cells natively presenting a mutant protein decreased their IC50 for IO. These results were also confirmed in primary CD22-positive cells derived from B-ALL patients at diagnosis and from patients with r/r B-ALL. Furthermore, co-treatment with IO and UCN-01 significantly increased cell death in primary cells expressing mutant p53. In summary, our findings suggest that p53 status may represent a biomarker predictive of IO efficacy in patients diagnosed with CD22-positive malignancies. INTRODUCTION In the adult population, Acute Lymphoblastic Leukemia (ALL) is an uncommon hematological disorder (0.4% of all new cancer cases in the US) characterized by highly proliferative immature lymphoid progenitors usually derived from the B-cell lineage (1,2). Human CD22 is a surface antigen expressed in pro-B and pre-B cells that heavily contributes to the regulation of B-cell function (8). After ligand binding, CD22 undergoes constitutive internalization followed by lysosomal degradation (9). Hence, this surface antigen represents an ideal target to kill leukemic Bcells using antibody-drug conjugates (ADCs) that combine the antibody specificity for a selected antigen with the cytotoxicity ability of different cell-killing agents (10). Inotuzumab ozogamicin (IO), also known as CMC-544, is a highly specific ADC targeting CD22-positive lymphoproliferative diseases. IO consists of a semi-synthetic derivative of N-acetil-γ calicheamicin 1,2-dimethyl hydrazine dichloride (CalichDMH) covalently linked-via an acid-labile 4-(4'-acetylphenoxy) butanoic acid-to a humanized monoclonal IgG4 anti-CD22 antibody (11,12). CalichDMH derives from the actinomyces Micromonospora echinospora, subspecies calichensis, and its cytotoxicity relies on the ability to bind the minor groove of the DNA helix producing double strand breaks (13,14). In turn this DNA damage arrests the cell cycle in G2/M, activating multiple apoptotic mechanisms (15,16). The Chk1/2 and the p53 signaling pathways have both been implicated in maintenance of the G2/M arrest triggered by DNA damage as the former proteins, upon induction by ATM (Ataxia Telangiectasia Mutated Kinase) and/or ATR (Ataxia Telangiectasia And Rad3-Related Protein) up-regulate 14-3-3 proteins or lead to p21 transactivation. In both circumstances, the net biological result of these events is prolongation of the G2/M arrest via 14-3-3-dependent cytoplasmic sequestering of Cdc25C or p21-dependent regulation of Retinoblastoma protein (17). The p53 protein-encoded by TP53 gene -plays a pivotal role in modulating DNA damage response, cell proliferation, differentiation, and death (18,19). Most p53 mutations result in protein loss of function and, if coupled with deleterious alterations involving the p53 region of the remaining allele, favor cellular oncogenic transformation. These non-synonymous p53 mutations usually occur in the DNA binding domain encoded by exons 5-8 of the TP53 gene. As a result, p53 protein structure is disrupted and p53 can no longer bind to its target genes and exert its transcriptional activity (20,21). In adult B-ALL, the most commonly reported TP53 alterations are missense mutations that, while infrequent, are usually associated with a poor outcome (22). Furthermore, the incidence of TP53 mutations increases at disease relapse and has been frequently reported in adult ALL that does not display recurrent fusion genes (23). IO has been recently approved for the treatment of adult patients with relapsed or refractory CD22-positive B-ALL (24) or adult patients with Ph+ ALL that have failed treatment with at least one TKI (25,26), showing significantly higher remission rates than standard therapy. In the present study we investigated the role of p53 in modulating the IO responsiveness of both immortalized and primary CD22-positive B-ALL cells. Immortalized Cells Burkitt lymphoma (BL-2, Namalwa, Raji, and Ramos), ALL (SUP-B15) and Acute Myeloid Leukemia (HL-60) cell lines were obtained from the German Collection of Microorganisms and Cell Cultures DSMZ and used for fewer than 6 months after receipt. Human bone marrow-derived mesenchymal stem cells (MSCs) immortalized by forcing the expression of telomerase reverse transcriptase (TERT) (donated by Dario Campana, Department of Pediatrics, Yong Loo Lin School of Medicine, National University of Singapore) were grown in RPMI-1640 medium supplemented with 10% FBS, 2 mmol/L glutamine, 10 −6 M hydrocortisone (Sigma-Aldrich), 100 U/mL penicillin and 50 µg/mL streptomycin as previously described (27). Immortalized MSCs were seeded in 96-well plates coated with 1% gelatin (Sigma-Aldrich) and grown until they reached confluence. Before seeding primary cells, the RPMI-1640 medium was removed from MSCs and cells were washed seven times with AIM-V medium (Thermo Fisher Scientific) to remove FBS and hydrocortisone. All cell lines were maintained in an incubator set at 37 • C with 5% CO 2 . Primary Cells Bone marrow (BM) samples were collected from six patients with newly diagnosed B-ALL and four refractory-relapsed B-ALL (r/r B-ALL) according to the 2008 WHO criteria. Patients were followed in the Division of Hematology of the A.O.U. Policlinico-Vittorio Emanuele and signed an informed consent releasing anonymously their samples for research purposes in accordance with the Declaration of Helsinki. Only subjects with neoplastic cells expressing >80% CD22-positive were eligible for this study. BM mononuclear cells were isolated by Ficoll-Paque Premium (GE Healthcare) density-gradient centrifugation according to the manufacturer's protocol. For apoptosis evaluation, BM mononuclear cells were maintained in AIM-V medium and seeded onto immortalized MSC cells. Primary cells were instead plated in AIM-V medium in stroma free wells to calculate the 50% inhibitory growth concentration (IC 50 ) for the drugs employed in the study. Immunophenotype Analysis of Immortalized Cell Lines The expression of surface markers of immortalized cell lines was determined by flow cytometry using the following monoclonal antibodies: anti-CD22 Fluorescein isothiocyanate (FITC) (Clone SJ10.1H11) and anti-CD33 Phycoerythrin (PE) (clone D3HL60.251) (both from Beckman Coulter). Cell lines were stained according to the manufacturer's instructions and analyzed by flow-cytometry using Cytomics FC500 (Beckman Coulter). For each condition, 10.000 events were acquired. Results were expressed as the percentage of CD22-or CD33positive cells over total number of analyzed events. Chk1 and p53 Constructs and Mutagenesis To achieve the inducible pTRIPZ short-hairpin RNA (shRNA) anti-Chk1 constructs we moved human shRNAs anti-Chk1 (cat. n • RHS4531, Dharmacon/Horizon Discovery Ltd) from the pGIPZ to the pTRIPZ vector according to the manufacturer's protocol. To obtain different p53 constructs we employed the following strategies. The human p53-EGFP sequence was first excised with NheI and NotI restriction enzymes from the pEGFP-N1-p53 plasmid (a gift of Prof. Francesco Frasca, Division of Endocrinology, Department of Clinical and Experimental Medicine of the University of Catania) and then cloned in the pcDNA3.1 expression vector (Thermo Fisher Scientific). To obtain the p53 R248Q mutant, the pcDNA3.1-p53-EGFP plasmid was subjected to a mutagenesis reaction using the Quick-Change II XL Site-Direct Mutagenesis Kit (Agilent Technologies) as elsewhere specified (28). Primers employed to generate the required mutations in the p53 wild-type sequence were 5 ′ -GGCGGCATGAACCAGAGGCCCATCCTC-3 ′ (forward) and 5 ′ -GAGGATGGGCCTCTGGTTCATGCCGCC-3 ′ (reverse). Codons indicated in bold mutagenize Arginine (R) in position 248 in Glutamine (Q). Proper incorporation of the desired mutation was verified by Sanger sequencing. Afterwards, the p53 R248Q -EGFP sequence was excised from pcDNA3.1 with XhoI and cloned in the pLEX lentiviral vector (Open Biosystem). Transfection and Lentiviral Transduction Production of lentiviral supernatant was performed as previously described (29). Briefly, pTRIPZ shRNA anti-Chk1 constructs, pTRIPZ Inducible Lentiviral non-silencing shRNA Control (Cat n • RHS4743), pLEX p53R248Q-EGFP, pTRIPz-p53-EGFP transfer vectors and their respective empty vector controls were co-transfected with the packaging plasmids in the HEK293T cell line, using the calcium phosphate transfection method according to the Open Biosystem protocol (Trans-Lentiviral Packaging Kit, Open Biosystem). Viral supernatants were harvested 48 h after transfection. The Namalwa cell line was then transduced with pTRIPz-EGFP or pTRIPz-p53-EGFP by a double round of spin-infection. Cells were centrifuged for 90 min at 32 • C and 1,200 x g in the presence of 1 mL of viral supernatant and 8 µg/mL Polybrene (Sigma-Aldrich). Forty-eight hours post-transduction, Namalwa cells were finally selected with 1 µg/mL puromycin (Sigma-Aldrich) for 72 h. BL-2 and SUP-B15 cell lines were transduced using 100-fold concentrated viral supernatants as previously described (30). Cells were subjected to two rounds of spinoculation as specified above and selected with 1 µg/mL puromycin for 72 h. Cell Proliferation and Cell Death Assay To measure the IC 50 of IO or Gemtuzumab Ozogamicin (GO), 1 × 10 4 BL-2, SUP-B15, Namalwa, Ramos, Raji, HL-60 and lentivirally transduced BL-2 and SUP-B15 cells or 2.5 × 10 4 primary leukemic cells were seeded in triplicate in 96-well plates and incubated for 48 h in the presence of increasing logarithmic doses of IO or GO (a gift of Pfizer) expressed as equivalents of N-acetyl-γ-calicheamicin dimethylhydrazide as previously reported (31). The doses employed ranged from 0.001 to 1000 ng/mL CalichDMH. Cell proliferation was determined using the luminescence ATP detection assay system ATPlite 1 step (Perkin-Elmer), following the manufacturer's instructions as previously reported (32). Subsequently, the IC 50 value was calculated by logistic non-linear regression using Prism 5.0 Software (GraphPad Software Inc) and was reported as the amount of CalichDMH equivalents (nM) from each treatment group that caused 50% loss of cell viability. To calculate IO IC 50 for Namalwa cells transduced with pTRIPz-EGFP empty vector or pTRIPz-p53-EGFP, 1

× 10 4 cells-either untreated or incubated with doxycycline (Sigma Aldrich) 1 µg/ml-were incubated with increasing logarithmic doses of IO (from 0.1 to 10.000 ng/mL CalichDMH) for 24 h. Cell viability and IC 50 values were determined as previously described. To analyze cell death, 1 × 10 4 BL-2, SUP-B15, Namalwa and Raji and 2 × 10 4 Ramos cells were treated with CalichDMH equivalents corresponding to their IO IC 50 , with 100 nM UCN-01 (Merk Biosciences) or p53 inhibitors, alone or in combination. Cells were then harvested and their apoptotic rate was determined using the Annexin V FITC/ 7AAD kit (Beckman Coulter) following the manufacturer's instructions. The Namalwa cell line transduced with the non-silencing shRNA or with anti-Chk1 shRNAs was induced with 1 µg/ml doxycycline for 24 h before performing death assays. Cells were kept in the presence of doxycicline for an additional 24 h to maintain shRNA silencing and were then treated with IO. They were then harvested and apoptosis was measured using the Annexin V FITC/ 7AAD kit following the manufacturer's instructions. For primary BM mononuclear samples, 20 × 10 4 leukemic cells were plated on a stromal feeder and then exposed to IO or UCN-01, alone or in combination. Cells were initially stained with monoclonal antibodies anti-CD45-Phycoerythrin Cyanin 7 (PC7) (clone J33) and anti human CD19-R-Phycoerythrin-Texas Red (ECD) (clone J3-119) (both Beckman Coulter) and incubated in the dark for 20 min at room temperature, then washed in PBS and finally labeled with the Annexin V FITC/7AAD kit according to the manufacturer's instructions. All data acquisition and analysis were performed using a Cytomics FC500 flow-cytometer (Beckman Coulter). Cell Cycle Distribution Analysis 1 × 10 4 BL-2, SUP-B15 and Namalwa cells were plated and treated with their respective IO IC 50 for 12, 24 and 48 h. Cells were then harvested, fixed in 70% of ethanol in Phosphate-Buffered Saline (PBS) for 24 h at −20 • C and incubated with 40 µg/mL RNase A and 20 µg/mL Propidium Iodide (both from Sigma-Aldrich) for 30 min. Cell cycle distribution was then evaluated employing the Cell Quest software (Becton-Dickinson) for acquisition and WinMDI 2.9 Software (Joseph Trotter, The Scripps Institute, La Jolla, CA) and Cylchred Software packages (Cell Cycle Analysis Software, Cardiff, UK) for analysis. Statistical Analysis Statistical analysis was performed using GraphPad Prism 5.0a (GraphPad Software Inc). Unpaired, single-tail t-tests with 95% confidence intervals were used to compare cell viability in different experimental conditions. The 1-way ANOVAs according to Bonferroni's post-test were used to compare the effect of IO on cell cycle phases at different time points. Inotuzumab Ozogamicin Shows Different Anti-proliferative Efficacy on CD22-Positive Leukemic Cell Lines With Different p53 Status To ascertain if the CD22-specific cytotoxicity of IO would be influenced by p53 expression, we employed three cell lines chosen for their diverse p53 profiles as BL-2 cells express wild-type p53, SUP-B15 present low levels of wild-type p53 (because of MDM2 gene amplification) and Namalwa cells display the p53 mutation R248Q (33). FACS analysis confirmed expression of the CD22 B-lymphoid antigen in >95% of BL-2, SUP-B15 and Namalwa cells while the myeloid-specific CD33 antigen was not detected in any of the above-mentioned cells but was expressed in 94% of the Acute Myeloid Leukemia cell line HL-60 employed as a negative control ( Figure 1A). We then incubated the above-mentioned cell lines with increasing doses of CalichDMH-conjugates (IO or the anti-CD33 ADC Gemtuzumab Ozogamicin-GO) and found that IO was much more effective than GO in reducing cell proliferation ( Figure 1B), generating IC 50 values that were consistently lower than those of their anti-CD33 counterpart ( Figure 1C). As expected, in CD33-positive HL-60 cells, GO was >400 fold more effective than IO. BL-2 and SUP-B15 were very sensitive to IO displaying IC 50 values (expressed as nM of CalichDMH equivalents) of 0.82 and 0.098, respectively. On the contrary, the Namalwa cell line was considerably less responsive to the drug, displaying an IC 50 of 23.70 nM that was comparable to that observed in the CD22-negative HL-60 cells (36.78 nM). Inotuzumab Ozogamicin Induces a G2/M Arrest That Is Associated With Chk1 and Chk2 Phosphorylation Next, we wanted to determine if the reduced proliferation rate that we observed in CD22-positive cells after IO treatment was due to induction of programmed cell death or to cellcycle arrest. To this end, we incubated BL-2, SUP-B15 and Namalwa cells for 48 h using IO concentrations reflecting their IC 50 values and then measured cell death by staining cells with Annexin V/7AAD (Figures 2A,B). We found that IO killed 60% of BL-2, 91% of SUP-B15 and 81% of Namalwa cell lines compared to 9, 23, and 8% of apoptotic rates in the untreated condition (Figures 2A,B). While these apoptotic rates may seem comparable, the fact that Namalwa cells required IO doses 29 fold (BL-2) and 241 fold (SUP-B15) higher than those employed for the other CD22-positive cell lines suggested a possible CD22independent cytotoxicity derived from the release of unbound CalichDMH as previously described (31). We then analyzed the cell cycle progression of each cell line after IO exposure for 12, 24, and 48 h. We found that cells surviving drug treatment exhibited a modified cell-cycle profile as both BL-2 and SUP-B15 displayed a progressive increase in the population blocked in the G2/M phase of the cell cycle ( Figure 2C and Table 1 Figure 2C). Our results confirm that IO arrests cells in the G2/M phase of the cell cycle (16), but indicate that this block is differently modulated in BL-2 and SUP-B15 cells as compared to the Namalwa cell line. Given these findings, we wanted to establish if the differing p53 status of our CD22-positive cell lines contributed to their different response to IO. Consolidated evidence has shown that induction of DNA damage can block cell cycle progression by triggering multiple signal transduction hubs (34,35) that converge on the Chk1/Chk2/p53/p21 pathway (36). To establish if this was the case for the G2/M arrest displayed by CD22-positive cells exposed to IO, we analyzed their protein lysates after drug incubation for 12, 24, and 48 h and found that all cells displayed a considerable increase in Chk1 phosphorylation (Figure 3). Moreover, BL-2 and Namalwa also exhibited increased Chk2 phosphorylation at each considered time point. As Chk1 and Chk2 induction by DNA damage results in p53 activation (37,38), we investigated p53 expression after IO exposure and detected its up-regulation in BL-2 and SUP-B15 cells. As expected, we failed to observe variations in the p53 levels of Namalwa cells as they overexpress mutant p53 at baseline (Figure 3). To confirm preservation of p53-dependent transcriptional activity, we analyzed p21 protein expression and detected p21 induction in both BL-2 and SUP-B15 but not in the Namalwa line (Figure 3). Taken together these findings indicate that IOdependent G2/M cell cycle arrest is associated with Chk1 phosphorylation. Our results also suggest that successful IO-mediated killing may require wild-type p53 as Namalwa cells were poorly responsive to IO despite their high CD22 expression. The Sequential Combination of Inotuzumab Ozogamicin and the Chk1 Inhibitor UCN-01 Increases the Apoptotic Rate of CD22-Positive Leukemic Cells Several reports have indicated that Chk1 pharmacological inhibition by UCN-01 increases the cytotoxic effect of different chemotherapeutic agents by abrogating Chk1-induced cell cycle arrest via both p53-dependent and -independent mechanisms (39)(40)(41). To establish if we could increase the IO sensitivity of Namalwa cells, we employed the sequential combination of IO and UCN-01 as depicted in Figure 4A. We found that, in BL-2 and SUP-B15 cells displaying wild-type p53, the two-drug combination increased apoptosis by 1.5 fold (Figures 4B,C). Strikingly, when we repeated this experiment on the Namalwa cell line employing an IO concentration calculated by averaging the IC 50 values of BL-2 and SUP-B15 responsive cells (0.459 nM), we observed a 2.6 fold increase in cell death as compared to IO alone (67.82% vs 26.43%) (Figures 4B,C). Furthermore, this result was achieved employing IO concentrations that were 50 fold lower than the previously calculated IC 50 for Namalwa cells ( Figure 1C). As UCN-01 is a staurosporine analog displaying multiple targets we could not exclude a non-specific effect attributable to the many substrates of this drug. We therefore silenced Chk1 expression in the Namalwa cell line employing a pool of inducible anti-Chk1 shRNAs. We initially performed an immunoblot to verify Chk1 silencing after 24 and 48 h of doxycycline induction (Supplemental Figure 1A). We next measured the apoptotic rate of cells transduced with the control non-silencing shRNA (shRNA NS) or with the anti-Chk1 specific shRNA and treated for 24 h with an IO concentration calculated by averaging the IC 50 values of BL-2 and SUP-B15 responsive cells. The silencing of Chk1 expression and the treatment with IO produced a 1.8 fold increase in cell death as compared to IO alone obtained after shRNA NS induction (29.83 vs 16.72%) (Supplemental Figures 1B,C). While these data implied the presence of a strong Chk1dependent increase in cell death, we still had not defined which protein downstream of Chk1 contributed to this event. It is well-established that-after DNA damage-activated Chk1 phosphorylates the cdc25C phosphatase, thereby favoring its interaction with 14-3-3 proteins that results in the nuclear export of cdc25C. Cytoplasmic retention by 14-3-3 prevents mitosis progression thus determining a G2/M arrest (42). To verify if UCN-01-dependent inhibition of Chk1 by-passed IO-induced cell cycle arrest, we performed an anti-phospho-cdc25C immunoblot. We found that UCN-01-alone or in combination with IO-reduced both cdc25C expression and phosphorylation ( Figure 4D). These results suggest that inhibition of Chk1 activity by UCN-01 may abolish IO-dependent G2/M arrest, inducing significant increases in cell death regardless of the p53 cellular background. Results are expressed as mean ± standard deviation values from three independent experiments. *p < 0.01; **p < 0.01; ***p < 0.001; ns, not significant. FIGURE 3 | Chk1 and Chk2 are involved in the G2/M arrest induced by Inotuzumab Ozogamicin. BL-2, SUP-B15 and Namalwa cell lines were exposed to IO according to their IC 50 values, for the indicated time points. Cells were then lysed and protein extracts were used to perform immunoblots employing the specified antibodies. Actin was used as a loading control. The depicted blots are representative of three separate experiments. p53 Status Determines Inotuzumab Ozogamicin Efficacy on Immortalized CD22-Positive Leukemic Cells To provide genetic confirmation that p53 status determines IO sensitivity in CD22-positive cells, we introduced the R248Q mutant p53 in cells harboring wild-type p53 and -converselyexpressed wild-type p53 in the Namalwa cell line. Specifically, BL-2 and SUP-B15 were stably transduced with lentiviral vectors expressing either GFP-tagged p53 R248Q or an empty vector used as a control. Anti-GFP immunoblots confirmed expression of the p53 R248Q mutant (Figure 5A). Transduced cells were subsequently exposed to different IO concentrations to determine their 48-h IC 50 . We observed that cells overexpressing p53 R248Q presented higher IC 50 values than the empty vector-transduced counterpart, requiring a 5-fold (BL-2: from 1.63 to 8 nM) or a 2.5-fold increase (SUP-B15: from 1.37 to 3.4 nM) to achieve 50% cell killing ( Figure 5B). Since expression of wild-type p53 induces a marked cytotoxic effect, we engineered a doxycycline-inducible lentiviral vector expressing the EGFP-tagged p53 and used it to transduce the Namalwa cell line. An immunoblot confirmed that doxycycline exposure for 24 h induced wild-type p53 expression ( Figure 5C). We then exposed the transduced Namalwa cells to increasing IO concentrations in the presence of doxycycline. As p53 induction per se determines an increase in cell death, we decreased the time of exposure to IO and used drug doses ranging from 0.1 to 10.000 ng/mL CalichDMH. When we determined the cells IC 50 , we observed different values as compared to those initially calculated for Namalwa cells and indicated in Figure 1. This finding was not surprising as the cells were continuously cultivated for 8 weeks (to allow lentiviral transduction and puromycin selection) and were incubated with IO for a different time (24 vs. 48 h) than the one employed in our previous experiments. Moreover, we found that forcing wild-type p53 expression in Namalwa cells increased anti-CD22 CalichDMH sensitivity, recording an IC 50 value 2.5 fold lower than that displayed by the empty vector-induced cells (86.27 vs. 216.64 nM). On the contrary, minimal differences were observed in the IC 50 values of cells that were not incubated with doxycycline (i.e., no

p53 induction) or were transduced with the empty vector control (Figure 5D). To further validate the role of mutant p53 in determining a reduced IO sensitivity in CD22-positive cells, we inhibited p53 expression in Namalwa cells and in two other lines expressing mutant p53 isoforms. To this end, we employed Raji cells that express the R273H gain of function mutation that involve an amino acidic residue directly involved in DNA binding and is therefore devoid of transcriptional activity (43) and the Ramos cell line displaying the I254D loss of function mutation (44). To confirm the importance of wild-type p53 in contributing to different responses to IO we calculated the drug's IC 50 in Raji and Ramos cell lines. As expected, after 48 h of treatment we found Raji and Ramos cells to be poorly responsive to IO displaying IC 50 values of 54.56 and 185.5 nM, respectively (data not shown). Several reports have indicated that chemical inhibition of p53 gain of function mutations by pifithrin-alpha (PFT-α) (45) or restoration of a p53 active conformation with APR-246 (also named PRIMA-1 Met ) (46) increases the apoptosis of cells expressing different mutant p53 isoforms. To establish if IO sensitivity is dependent on p53 status, we co-treated Namalwa and Raji cell lines with IO and PFT-α or alternatively exposed Ramos cells to the combination of IO and APR-246. When we assayed cell proliferation and survival in the presence of IO and PFT-α or APR-246 we found significant reduction on cell growth and viability of all p53 mutated cell lines. Specifically, we found that co-treatment with IO and PFT-α or APR-246 for 24 h caused a decrease in cell viability of about 1.7 fold compared to treatment with IO alone in all three cell lines ( Figure 6A). Furthermore, we observed an increase in the amount of apoptotic cells from 6.7 to 68.3% in Namalwa, from 8.2 to 40.5% in Raji and from 34.3 to 94.7% in Ramos cells (Figures 6B,C). Our data suggest that the anti-proliferative effect of IO is heavily influenced by p53 status, suggesting that p53 evaluation may predict the response of CD22-positive leukemic cells to a Calicheamicin derivative. p53 Modulates Inotuzumab Ozogamicin Efficacy on Primary CD22-Positive Leukemic Cells To validate the data generated in immortalized cell lines, we isolated mononuclear cells from six newly diagnosed B-ALL patients and four subjects presenting refractory or relapsed B-ALL (r/r B-ALL). These cells were incubated with logarithmic IO concentrations for 48 h and evaluated for cell viability (Figure 7A). Our experiment showed that cells derived from newly diagnosed B-ALL were more sensitive to IO than r/r B-ALL cells, with the former displaying average IC 50 values of 29.96 nM and the latter presenting an average IC 50 of 901.56 nM (Figure 7A). While alterations in p53 are considered an infrequent event in B-ALL, the disruption of both p53 alleles or the presence of missense mutations is associated with an adverse prognosis (47,48). To further confirm that the higher IC 50 values displayed by r/r B-ALL patients correlated with their p53 status, we initially analyzed their p53 expression levels after a 48 hour incubation with IO. As reported in Figure 7B, we found that newly diagnosed B-ALL patients only expressed p53 after IO treatment, while three (R1, R5, and R6) of the four r/r B-ALL subjects in our series displayed p53 overexpression in the untreated condition suggesting the presence of a mutant p53. Indeed, the r/r B-ALL patients who exhibited the highest IC 50 values also showed a mutant p53. We next wanted to establish if Chk1 inhibition would increase IO efficacy on primary B-lymphoid cells. Hence, these cells were exposed to the IO and UCN-01 combination previously described and scored for cell death. When we incubated cells derived from a B-ALL patient at diagnosis with 29.96 nM IO (mean of the IC 50 values of B-ALL patients at diagnosis) and UCN-01-alone or in combination-we detected no significant differences in the amount of cell death between IO plus UCN-01 (82.33%) and IO alone (74.48%; Figures 7C,D). However, when we repeated this experiment using primary cells derived from a r/r B-ALL patient expressing mutant p53, addition of UCN-01 to IO doubled the apoptotic rate compared to IO alone (from 53.9 to 96.84%) employing an IO dose 30-fold lower than the mean IC 50 calculated for r/r B-ALL patients (Figures 7C,D). Furthermore, we observed remarkable cell death rates after treatment with UCN-01 alone, confirming the importance of Chk1 inhibition in B-ALL patients (49). These findings strengthen our hypothesis that Chk1 phosphorylation triggers a G2/M arrest that will evolve in apoptosis in the presence of wild-type p53. Our data also suggest that p53 mutations may represent a limiting factor FIGURE 5 | p53 wild-type increases the sensitivity of CD22-positive cells to Inotuzumab Ozogamicin. (A) BL-2 and SUP-B15 cell lines were lentivirally transduced with an empty vector (EV) or the GFP-tagged p53 R248Q -pLEX construct. After puromycin selection, cells were lysed and protein extracts were blotted using an anti-GFP antibody to verify protein expression. Actin was employed as a loading control. (B) The modified cell lines, stably expressing either EV (•) or mutant p53 ( ) were then exposed to logarithmic doses of IO and their reduction in cell proliferation was evaluated employing the ATPLite luminescence assay. Results represent the average ± standard deviation of at least three different experiments performed in triplicates with relative luminescence of untreated cells arbitrarily set at 100%. IC 50 (Continued) FIGURE 5 | values were calculated by logistic non-linear regression and are presented as nM equivalents of CalichDMH. (C) Namalwa cells were transduced with an empty vector (EV) or GFP-tagged wild-type (WT) p53-pTRIPZ vector using an inducible lentiviral system. After puromycin selection, cells were treated for 24 h with doxycycline to induce wild-type p53 expression. Cells were harvested, lysed and protein extracts were simultaneously blotted for p53 and Actin, the latter employed as a loading control. (D) The reductions in proliferation of Namalwa cells expressing EV (•) or p53 WT ( ) were then evaluated after co-treatment for 24 h with doxycycline and logarithmic doses of CalichDMH. Results represent the average ± standard deviation of at least three different experiments performed in triplicates with relative luminescence of untreated cells set arbitrarily at 100%. IC 50 values are reported as nM equivalents of CalichDMH. negatively modulating IO efficacy in patients with CD22-positive lymphoproliferative disorders. DISCUSSION While antibody-drug conjugates represent one of the many successful therapeutic strategies recently introduced in the fight against cancer, selected neoplastic clones eventually escape the cell killing mechanisms induced by these drugs (50). To explain the resistance to IO we investigated the intracellular signaling triggered by the drug and report several findings with potential clinical consequences for patients diagnosed with B-cell derived disorders receiving IO. First of all, while it has been previously reported that CD22 levels play a minor role in determining IO efficacy, the general consensus is that high CD22 expression accelerates IO-induced cell death (51). Our data contradict this conclusion as we found that both immortalized cell lines (BL-2, SUP-B15, Namalwa) and primary cells exhibiting comparable CD22 levels displayed very different IO sensitivity. Hence, evaluation of CD22 expression cannot be considered a reliable biomarker to predict IO efficacy on CD22-positive cells. We have also actively investigated the signaling network elicited by IO. As calicheamicin is a well-established DNAdamaging agent (13,14), exposure to the compound would be expected to activate the ATM/ATR proteins (17) that, in turn, trigger the Chk1/Chk2/p38 pathway (52). This signaling network induces a reversible cell cycle arrest mediated by p53dependent induction of p21 (53). This complex response to DNA damage represents an evolutionary selected failsafe mechanism aimed at preserving the DNA integrity of healthy cells (54,55). Indeed, upon complete and accurate repair of the accumulated damage, cells may exit their cell cycle arrest and return to proliferate. However, in the presence of persisting (unrepaired or irreparable) DNA damage, cells will either senesce or undergo apoptosis (56). Our data confirm that IO blocks cells in the G2/M phase of the cell cycle as previously described (16,57). Moreover, the fact that this arrest was detected throughout our panel of immortalized cell lines and primary cells indicates that this event is independent of p53 status. However, we noticed that the initial G2/M block detected in all CD22-positive cells evolved in two alternative scenarios depending on their underlying p53. Indeed, while BL-2 and SUP-B15 (p53 wild-type) cells underwent apoptosis-possibly due to partial or inaccurate repair of IOinduced DNA-damage, p53-mutant Namalwa cells stabilized their G2/M arrest. Although we are still investigating this phenomenon, it is tempting to speculate that a prolonged G2/M arrest may enable cells to further repair their DNA damage, potentially favoring an escape from senescence by selected IOresistant clones. Our experiments also showed that, after IO exposure, the amount of death triggered by IO paralleled the p53 profile of the cells. Indeed, BL-2 and SUP-B15 expressing wild-type p53 both displayed high death rates (ranging between 60 and 90%) in the presence of <1 nM of CalichDMH equivalents. On the contrary, to achieve a 73% apoptotic rate, Namalwa cells with mutant p53 had to be incubated with large amounts of IO suggestive of cellular cytotoxicity derived from the release of antibodyfree calicheamicin. Moreover, the IO IC 50 for two further cell lines devoid a functional p53 confirmed that the CD22-specific cytotoxicity of IO is influenced by p53 expression. While these findings all point to p53 as a potential biomarker for IO efficacy, they are not in agreement with a previous report by Prokop and colleagues suggesting that p53 status bears no consequence on the apoptotic response triggered by calicheamicin (58). To explain these discrepancies, we engineered an inducible lentiviral vector expressing GFP-tagged p53 that was used to transduce Namalwa cells. In agreement with our previous observations, we found increased IO sensitivity (i.e., lower IC 50 ) in the presence of wildtype p53. Furthermore, when we inhibited p53 transactivation (with PFT-α) or reactivated p53 function (with APR-246) in cell lines (Raji and Ramos) expressing gain or loss of function p53 mutations we again observed an increased sensitivity to IO. While we are devoid of a specific explanation for the discordance between our results and those previously published by Prokop and colleagues, it is possible to assume that the different cellular background between the two studies (colon cancer cells in the paper by Prokop et al. vs Burkitt's lymphoma and ALL cells in our study) or the use of the unconjugated drug (i.e., free calicheamicin unbound to the anti-CD22 antibody epratuzumab in the Prokop study) may have significantly contributed to these diverging results. We should also point out that the IO IC 50 value calculated in primary cells derived from ALL patients at diagnosis or at the time of relapse also correlated with their p53 status, further strengthening the suggestion that p53 integrity predicts for IO efficacy. Hence, our study confirms the promising results of a CD22-specific immunoconjugate in patients affected by B-cell lymphoproliferative disorders expressing wild-type p53. We also demonstrated that the combination of IO and the Chk1 inhibitor UCN-01 as well as Chk1 silencing forced G2/M-arrested cells to progress along the cell cycle increasing their apoptotic rate. Unexpectedly, we found that-unlike what we observed in immortalized cell lines-in B-ALL patients at diagnosis the combination of IO and UCN-01 did not determine a significant increase in cell death compared to IO alone. As it has been The indicated primary cells derived from B-ALL patients at diagnosis (D) or from r/r B-ALL (R) were also treated for 48 h with calicheamicin equivalents corresponding to their IO IC 50 average values. Lysates from these cells were then blotted using the specified antibodies. Actin was used as a loading control. (C) Primary cells, derived from one B-ALL (D11) and one r/r B-ALL (R1) patient, were left untreated or exposed to the IO IC 50 average value for B-ALL patients at diagnosis (29.96 nM) and UCN-01 100 nM, alone or in combination, employing the scheme described in Figure 4A. Apoptosis was then evaluated after Annexin V-FITC/7AAD double staining. The indicated percentage values show the distribution of necrotic, early and late apoptotic cells. (D) Histograms representing the average percentage of Annexin V and 7 AAD positive cells in the untreated condition or after exposure to IO, UCN-01 or a combination of

the two drugs. Columns represent average ± standard deviation of two independent experiments. ***p < 0.001; ns, not significant. reported that the genomic profiling of B-ALL at diagnosis and relapse shows substantial changes in both the number and nature of the detected genetic alterations (59), it is possible to speculate that patients expressing wild-type p53 at diagnosis and lacking additional alterations in checkpoint pathways, remain sensitive to a DNA damaging agent but will not benefit from a checkpoint inhibitor (41). Our findings are in line with multiple studies suggesting the efficacy of Chk1 inhibitors as a new therapeutic strategy for B-and T-ALLs (49,60). Furthermore, as in primary cells UCN-01 sensitivity is strictly related to the integrity of the p53 sequence, we hypothesize that the combination of IO and a Chk1 inhibitor could greatly benefit patients affected by r/r B-ALL displaying mutant p53. Currently, several clinical trials are evaluating the use of Chk1 inhibitors for the treatment of patients with either solid tumors or hematological malignances (ClinicalTrials.gov Identifier: NCT02203513 and NCT03495323). While these drugs are still in the early phase of their clinical development, the initial use of the Chk1 antagonist prexasertib in ovarian cancer patients has been associated with good tolerability with the only grade 3-4 adverse events related to decreased white blood cell counts and neutropenia (61). An alternative approach to the use of Chk1 inhibitors is the combination of chemo-and immune-therapy that was recently described in two different phase 2 clinical trials coupling IO with mini-hyper-CVD in patients with relapsed or refractory ALL (62), or in older patients with newly diagnosed ALL (63). In both cases the trials generated excellent clinical outcomes, although patients experienced different grade 3-4 adverse events including veno-occlusive disease. In summary, our findings suggest that p53 mutational status may represent a predictive biomarker for IO efficacy and that an approach combining IO and Chk1 inhibition could represent a potential therapeutic approach for patients with r/r B-ALL expressing mutant p53 that are unlikely to benefit from IO monotherapy. AUTHOR CONTRIBUTIONS ET, MM, AF, PL, and NP performed the experiments. ET, MM, CR, MP, SV, and SS analyzed and interpreted the data. ET and MM designed the experiments and wrote the paper. FS, GP, and AR made a critical revision of paper. LM and FD helped supervise the project. PV conceived the original idea and supervised the project. Reasons, procedures, and outcomes in ventriculoatrial shunts: A single-center experience Background: Ventricular shunts are used to drain cerebrospinal fluid into extra-cranial spaces. Ventriculoatrial (VA) shunts are provided to transfer cerebrospinal fluid from the cerebral ventricle into the right atrium of the heart. A single center experience of indications, procedure, and clinical outcomes in VA shunt was presented in current study. Methods: VA shunts were applied in 10 patients who had repeated previous shunt dysfunction or infection. The reasons, clinical findings, replacement methods, and postoperative clinical follow-ups and outcomes were recorded retrospectively. Results: There were seven female (70%) and three (30%) male patients; their ages ranged from 5 to 13 years (mean ± SD; 8.5 ± 2.6 years). Shunt re-placement reasons were as follows: Shunt occlusion in five patients, intraperitoneal infection in four patients and a distal catheter was kinked and knotted in one patient. Postoperative early complications were seen in one patient as early catheter thrombosis and catheter revision were applied. Late complications were seen in two patients as follows: Catheter infection and infective endocarditis occurred in one patient and pulmonary thrombus occurred in one other patient. There was not any catheter-related mortality observed at the one year follow-up period. Conclusion: VA shunts may be an option for cerebrospinal fluid drainage at necessary conditions. However, sterilization and general training on asepsy and antisepsy are the most important determinants affecting the clinical outcome due to the cardio systemic relationship. INTRODUCTION There are limited options for continuous cerebrospinal fluid drainage. The ventriculoperitoneal (VP) option is more popular than ventriculoatrial (VA) shunts. However, shunt revisions may be required due to shunt infection, obstruction, and migration conditions in VP shunts. In such special events, VA shunts may be an appropriate option for continuous cerebrospinal fluid drainage. [4,5,11] The intraoperative appropriate vein selection and exact shunt placement is important to reduce complications such as obstruction. [4] Placement strategies and monitoring methods have been improved to achieve more success in VA shunt catheter replacement. [4,8] In the present study, reasons, protocols, and clinical outcomes of VA catheter placements were reported in 10 patients. MATERIALS AND METHODS A VA shunt operation was applied in 10 patients who suffered from hydrocephaly; this was done with the collaboration of the departments of neurosurgery and cardiovascular surgery. Age, gender, reasons, complications, additional applications, and clinical outcomes were evaluated retrospectively. Preoperative scopy scans were applied to monitor the venoatrial route in each patient. RESULTS There were seven female (70%) and three (30%) male patients (female to male ratio, 2.3:1). The ages ranged from 5 to 13 years (mean ± SD; 8.5 ± 2.6 years). All patients were diagnosed with hydrocephalus on neurological verification. A VP shunt was applied previously in all patients. Repetition of hydrocephalus complaints were detected in six patients. Repeated catheter obstruction was determined in five of the six patients. A kinked and knotted catheter was detected in one patient. Fever, abdominal tenderness, and elevated infection markers were detected in other four patients. These four patients were diagnosed with peritonitis. The ventricular side of catheters was placed by neurosurgeons. After that, neck dissection was performed with a parallel longitudinal incision to the sternocleidomastoid muscle by cardiovascular surgeons. Just behind the muscle, the internal jugular vein was explored and evaluated with its branches. If an appropriate branch was detected, a catheter was placed in the branch. If an appropriate branch was not detected (e.g., too narrow for the catheter's diameter), the catheter was placed in the main body of the internal jugular vein. Before replacement the purse sutures were added to prevent migration and bleeding into the vein. The placement of catheter was demonstrated in Figure 1. All patients were followed postoperatively for one year. Early complications of catheter thrombosis were seen in one patient on the 15 th day following operation. Catheter revision and anticoagulant therapy (0.5 mg/kg enoxaparine prophylaxis was administered) were applied in this patient and no secondary thrombosis was observed. Late complications were seen in two patients as follows: In the first patient, probable catheter-related methicillin-resistant Staphylococcus aureus bacteremia and infective endocarditis were detected in blood cultures the second month after the operation. The catheter was removed urgently and appropriate antibiotherapies (Vancomycin 1 gram IV + Gentamycin 1.5 mg/kg IV) were administered to the patient, according to current literature suggestions. [6] The removal of catheter cultures confirmed the diagnosis. The blood parameters and fever returned to normal ranges after one week of antibiotherapy. In the second patient, dyspnea and sweating occurred at the postoperative 162 nd day. A high resolution computed tomography scan revealed pulmonary thrombus on the small pulmonary artery branches. The catheter was removed immediately and rapid anticoagulation and antiagregant therapy (25000-50000 IU continuous intravenous unfractionated heparin infusion with a partial thromboplastin time (PTT) monitorization in the first two days, 1 mg/kg enoxaparine and 100 mg daily acetylsalicylic acid maintenance therapy was started after two days) was initiated in the patient presenting with pulmonary thromboemboly. The thrombosis treatment was applied with the guidance of the current literature. [10,15] The respiratory complaints were recovered after the first two days of treatment. There were no additional catheter-related events detected during the follow-up period. Additionally, catheter related mortalities were not observed during the one year follow-up period. The details of the cases are summarized in Table 1. DISCUSSION The VP shunts provides a prolonged relief of intracranial pressure, and it is quite easily to be performed. The major limitation is catheter dysfunction in this treatment method. Catheter dysfunction reasons can be listed as follows: Shunt infection, valve obstruction, catheter migration, shunt disconnection, kinked catheter (malposition), or any combination of these reasons. This problem is emergency-like situation that can be lead to mortalities. However, it can be treatable and also, the risk is reduced in shunt-related complications with recent advances in surgical technique and shunt design. [17] In contrast, VA shunts became the standard treatment for intracranial hypertension due to hydrocephalus since 1952. Furthermore, over the subsequent years, the favorable intervention led to notable concerns with the recognition of various range of severe and even life-threatening complications that closely related to the circulatory system. [16] Recent reports suggest that despite VP shunts being the most preferred method; there is a notable patient population that remains where VA shunt is needed. [16] VA shunts may be good option for the desired result cannot be achieved recurrent VP cases in experienced hands. Also, minimal invasive methods and radiographic guidance's are developed for VA shunts too. [12] Thus, VA shunts are important alternative to VP shunts in selected cases. [7] However, complications should not be neglected that range from something simple like an occlusion to other more significant problems like bacteremias and cardiac disorders due to cardiac placement and systemic drainage. [2,7,14] Therefore, new placement strategies and monitoring methods have been improved to achieve more success and reduced complications in VA shunt catheter replacement. [4,8,9] Chuang, et al. reported the use of percutaneous VA placement with real-time transesophageal echocardiogram monitoring. They claimed that this method can be used less invasively, as well as more accurately, quickly and safely. [4] Endovascular placement of a VA shunt was presented by Gonzales, et al., who reported three advantages of this technique: The venous system can be identified easily, the jugular vein patency can be demonstrated clearly, and the true placement or malposition of the catheter can be determined quickly. [8] In another study, a ventriculosubgaleal shunt was mentioned by Hansasuta, et al., who claimed that this approach is simple and inexpensive. [9] Metellus, et al. also reported that percutaneous placement of a VA shunt with radiographic guidance improves the effectivity and safety of the technique. [12] Ten patients who had previous shunt dysfunction were presented in our report. VA shunts were performed with the guidance of scopy in these patients. Only one catheter obstruction was detected. Catheter thrombosis was determined at the postoperative 15 th day in this patient. Additional occlusion and malposition were not observed. VA shunt approaches may cause potentially life threatening complications. Obstructions, malpositions, and shunt infections are the most frequent problems in VA shunts. [4,13,14] Numerous complications have been reported in the literature and most of them could have been treated or prevented. For example, Elhammady, et al. reported a VA shunt displacement in a case with a partial anomalous pulmonary venous return. They suggested that radiographic evidences are beneficial for the management and treatment of such variations and further scans such as computed tomography should be applied in suspected cases. [7] Arıbas, et al. reported pulmonary hypertension development in a case following a VA shunt implantation. Pulmonary perfusion scintigraphy revealed segmental and subsegmental perfusion defects in this case. They removed the VA shunt and it was replaced with a VP shunt. [1] Cardiac complications can occur during or after VA shunt implantation as Natarajan, et al. reported a case of a 57-year-old male patient who was treated for pneumonia and new-onset atrial fibrillation. According to their report, the patient had a VA shunt since childhood that was implanted for hydrocephalus and there he had no additional health problems. To sum up, they diagnosed catheter-related right heart complications, such as significant calcific tricuspid stenosis and a dilated right atrium that was revealed in transthoracic echocardiography. [14] Ben-Ami, et al. reported a catheter related Gram-positive bacteremia and nephritis in a 47-year-old woman who had a VA shunt for 10 years due to hydrocephalus. The VA shunt was removed; then, vancomycin therapy was started and a new VP shunt was inserted in this case. According to the report, the patient fully recovered. [2] Chaw, et al. reported infective endocartitis in one VA shunt case. They claimed that appropriate treatment should be applied with the removal of foreign material in such cases. [3] In our study, we implanted all of the VA shunt catheters under scopy scans to appropriately position them in the right atrium. An early catheter revision was made in one patient due to catheter occlusion 15

days following operation. Presumptive diagnoses of catheter-related methicillin-resistant Staphylococcus aureus bacteremia and infective endocarditis were detected in one patient at the second month of application. The catheter was removed urgently and appropriate antibiotherapies (vancomycin 1 g IV + Gentamycin 1.5 mg/kg IV) were administered. The patient fully recovered after therapy. Dyspnea and sweating occurred in another patient at the 162 nd day of implantation. A high resolution computed tomography scan revealed pulmonary thrombus on the small pulmonary artery branches. As a result, the catheter was removed immediately and rapid anticoagulation and antiagregant therapy was initiated, and the patient's respiratory complaints were recovered after two days of treatment. Additional catheter-related morbidity and mortality were not observed at the one-year follow-up period. As Celal Yavuz and collaborators acknowledge the VA shunt is a solution for a patient who has had already a set of complications. These complications are often secondary to the infectious etiology of the hydrocephalus or the surgical etiologies associated with a "cold" peritoneum. It is reasonable to forward the argument that the prevalence of VA shunt, particularly in the pediatric population, is a subtle indicator of the social reality of a country. The small but significant series of patients reported by Celal Yavuz and collaborators may not be that interesting for neurosurgeons that perform many of these procedures, or to those who don't have to do place any VA shunt. But its significance lays in that it is telling us all that neurosurgical pathologies have a geographical and social identity and that perhaps the standard textbooks and publications coming from the industrialized world needs to be modified to accommodate the particularities of every region and its neurosurgical patients. When with the best of the intentions we try to share our understanding of disease mechanism with our colleagues from different regions we naively assume that what we see is what they see and Celal Yavuz and collaborators with a series of just ten patients are proving us wrong. I am interested in Chiari type I malformation, and always wondered why my colleagues from Nicaragua operate on a handful a patients a year. Jorge A. Lazareff University of California at Los Angeles. E-mail: [email protected] Factor XII mutations, estrogen-dependent inherited angioedema, and related conditions The clinical, biochemical and genetic features of the conditions known as estrogen-dependent inherited angioedema, estrogen-associated angioedema, hereditary angioedema with normal C-1 inhibitor, type III angioedema, or factor XII angioedema are reviewed. Discussion emphasizes pathogenesis, diagnosis, and management. Novel forms of inherited angioedema, either completely dependent on, or associated with high estrogen levels, but otherwise clinically indistinguishable from classic forms of HAE, were independently reported by North American and European investigators in 2000 [1,2]. Cases are increasingly recognized around the world [3][4][5][6][7]. The nomenclature of these conditions is evolving as their underlying genetic abnormalities are elucidated. Originally referred to by clinical phenotype as estrogendependent (or estrogen-associated) inherited angioedema (EDIA, EAIA) [1], HAE with normal C1-INH activity [2]; or simply distinguished from classic forms as HAE type III [OMIM 610618] [2], the terms Factor XII-HAE or HAE-FXII have been used to identify the condition when associated with the recently identified gain-of-function mutation in the gene encoding factor XII (F 12) [11,12]. Some clinically indistinguishable cases do not carry this mutation [11], so underlying genetic diversity is apparent, and the nomenclature to describe these conditions will likely continue to evolve. Clinical heterogeneity is evident in described cases. In a large multigenerational family of Italian origin, affected individuals experienced angioedema only during pregnancy, use of oral contraceptives or hormone replacement therapy [1]. In contrast, in different European families, phenotypes were far more variable [2]. Some patients experienced angioedema prior to menarche, with exacerbations after puberty and/or with high estrogen states, but in many cases, angioedema occurred even in low or normal estrogen level states. Initial reports [1,2] described only affected female patients, with an unaffected obligate male carrier [1]. More recently, pedigrees with affected male members have been described [13][14][15]. In one of the original reports [1], ethical considerations precluded the study of biochemical features during symptomatic episodes, as the index patients presented in the postmenopausal period, and none of their daughters became pregnant during the period of observation. As multiple family members had experienced laryngeal edema during high estrogen states, investigators reasoned that administration of estrogen could have lifethreatening consequences, and affected individuals and individuals of unknown phenotype were advised to avoid estrogen. Indeed, death due to sudden airway obstruction was reported in some family members in the other originally reported pedigrees [2]. Thus, the only available biochemical analyses, performed when the affected individuals were asymptomatic, including normal C1-INH quantitative and functional assays, C3, C4,, and factor XII levels, at the time did not allow the investigators to preclude abnormalities in these parameters during symptomatic periods [1]. In the other initial report [2], biochemical analyses were reported in some symptomatic patients. C1 inhibitor level and activity, C3 and C4 were normal, even during acute attacks. These observations helped to distinguish EDIA and EAIA as being pathogenetically distinct from classic forms of HAE. Genetic features The mode of inheritance could not be determined precisely in either of the original reports. Autosomal dominant transmission was considered most likely in the pedigree with strict estrogen dependence, though other modes of transmission could not be excluded [1,2]. Detailed information was reported in two multigenerational European pedigrees [2], one of which showed transmission of disease to children from an unaffected female, a phenomenon not seen in other reported pedigrees. Investigators speculated that the restriction to women suggested an X-linked dominant mode of inheritance; autosomal dominant transmission with hormonal control of the expression of the trait (the favoured explanation for the pedigree in the strictly estrogendependent pedigree) was thought to be less likely due to onset of symptoms in childhood, prior to significant hormonal effects. Autosomal dominant transmission seemed likely in a French pedigree [3] The recent identification heterozygosity for a gain-of-function mutation in F12 in female subjects in patients with EAIA [5,11,12,15,16] and EDIA, including those from the originally reported pedigree of Italian origin [17] suggests that autosomal dominant transmission is likely. However, the involvement of other genetic polymorphisms likely contributes to the diversity of clinical phenotypes [17]. In the family of Italian origin, the coding sequences as well as the 5' untranslated region (UTR) of the gene encoding C1 INH (SERPING1) were determined to be normal, clearly establishing this condition as being separate from the classic forms of hereditary angioedema (characterized by mutational inactivation of the C1 inhibitor gene). The 5' UTR of F12 (known to contain an estrogen-response element, alteration of which might explain the clinical phenotype of estrogen dependence) was also determined to be normal [1]. The biochemical and genetic observations from these two studies indicated that abnormalities in C1 INH could be excluded, and efforts to find the underlying cause of EDIA/EAIA were redirected elsewhere. On the basis of co-segregation patterns, two different missense mutations in 6 index patients of 20 families (confirmed in 22 additional family members), mapping to 5 q 33-qter of F12 (Online Mendelian Inheritance in Man, [OMIM] 610619) were identified in European pedigrees of hereditary angioedema with normal C1-INH. Both in exon 9, one involved a threonine-to-lysine substitution (Thr309 Lys); the other a threonine-to arginine substitution (Thr309Arg) [11]. The presence of Thr328Lys in the family of Italian origin with estrogendependent angioedema was confirmed in affected family members living in Canada [17] and in Italy (R. Colombo, personal communication), In addition, affected family members living in Canada were found to have polymorphisms in the genes for angiotensin-converting enzyme (ACE) and aminopeptidase P (APP) that are associated with lower circulating levels of these enzymes that are responsible for the degradation of bradykinin and its active metabolite [17]. Insertion/deletion polymorphisms in the ACE gene (ACE) account for 50% of the variability in human serum levels of ACE [18], with the insertion (I) allele associated with lower expression of ACE mRNA, and decreased degradation of bradykinin [19]. All three index patients had at least one copy of the inserted allele (I) in intron 16 of the ACE gene that is associated with lower levels of ACE. Genetic variants in the gene encoding APP (XPNPEP2), resulting in reduced enzyme activity, higher bradykinin and des-Arg9-BK have been associated with angioedema induced by ACE inhibitors [20]. All three affected female subjects also had at least one copy of the A allele at the SNP rs3788853 locus, located 5' of XPNPEP2, which codes for membrane-bound APP, and is associated with decreased APP activity, decreased bradykinin and des-Arg9-BK degradation, and angioedema induced by ACE inhibitors [20,21] Additional families with HAE and normal C1 inhibitor have been identified as carrying the Thr328Lys mutation [5,12,15,16,22], while other factor XII mutations have been described in different pedigrees [23]. Bradykinin accumulation: the final common pathway A new picture is emerging of the hereditary angioedemas as a group of genetically heterogeneous disorders of bradykinin metabolism, leading to its periodic accumulation. Bradykinin and its active metabolite, des-Arg9-BK, are the key mediators of angioedema [9,10,24,25]. Not only can mutations in different components (C1 INH, factor XII, ACE, APP, and as yet other, unidentified, factors) of bradykinin-related pathways occur, multiple different mutations can occur in each factor, and it seems likely that different combinations of these mutations contribute to the observed clinical heterogeneity of the conditions. In addition, the unique sensitivity of many of these components in bradykinin-related pathways to androgens and estrogens further modifies the clinical presentations. An appreciation of the pathways that result in the formation and degradation of bradykinin, and its active metabolite, des-Arg9-BK, and their regulation by sex hormones, contributes to the rational treatment of both classical and estrogen-dependent/factor XII-associated forms of hereditary angioedema. Effects of sex hormones on bradykinin pathways, and contribution to clinical phenotype Before considering the influence of the sex hormones on key enzymes if bradykinin pathways, outlined below, it is helpful to review key aspects of the reciprocal regulation of bioavailable estrogen and testosterone through their effects sex hormones binding globulin (SHBG) (reviewed in [26]). The activity of estrogen and testosterone is determined by the free or bioavailable fraction. In males, approximately 65% of testosterone circulates bound to SHBG, 78% in females. This bound fraction is essentially a reservoir; only the remaining free testosterone is biologically active. The fraction of estrogen bound to SHBG is less; only 30% is bound in males, 58% in females. The clinical relevance of this differential binding is apparent as abnormal variants of SHBG that bind sex hormones less efficiently result in a preferential increase in bioavailable testosterone with resulting masculinization. By influencing the level of SHBG, each of the sex hormones enhances its own bioavailability, while decreasing the relative bioavailability of the other. For example, estrogen increases levels SHBG, this in turn binds more testosterone than estrogen, increasing the relative bioavailability of estrogen. Conversely, androgens decrease levels of SHBG, resulting in a preferential increase in the bioavailability of androgens. This type of negative reciprocal regulation of bioavailability can amplify the effects of small changes in the relative amounts of estrogen versus testosterone, and may in part explain the exquisite sensitivity of the clinical phenotype to relatively small changes in sex hormones levels. Danazol has been shown to suppress SHBG levels in classic HAE patients [27], although other observations suggest there may be additional effects of SHBG [28]. Estrogen: effect on bradykinin production Factor XII High levels of estrogen, such as occurr during pregnancy or oral contraceptive use [29,30], are associated with increased levels of factor XII, likely due to an estrogenresponse element in the promoter region of the gene [31,32]. When activated, factor XII converts pre-kallikrein to kallikrein, which produces bradykinin from high molecular weight kininogen. Under the conditions of high estrogen levels, the increased availability of factor XII for activation would favor increased bradykinin production. C-1 INH High levels in estrogen during pregnancy [33][34][35], or oral contraceptive use [36],are associated with reduced levels of C-1 INH. As C-1 INH normally inhibits activated factor XII and kallikrein; reduced inhibition of factor XII and kallikrein with high estrogen levels would favour increased bradykinin production. Estrogen: effect on bradykinin degradation ACE Estrogen suppresses ACE expression [37]. As ACE is important both for the degradation of bradykinin and its active metabolite, des-Arg9-BK, reduced levels of ACE under conditions of high estrogen levels

result in reduced degradation of bradykinin and its active metabolite, favouring their accumulation. APP The effect of estrogen on APP levels is not known. However, it has been reported that androgens increase APP levels [38], and, as estrogen increases SHBG, and reduces bioavailability of testosterone, it is reasonable to speculate that estrogen might reduce APP levels. As APP is particularly important in the degradation ofdes-Arg9-BK, and to a lesser extent bradykinin itself, reduced APP levels would favor the accumulation of bradykinin. Androgens: effect on bradykinin production C-1 INH Androgens increase the level of C-1 INH [39,40], which in turn inhibits activated factor XII and kallikrein, reducing bradykinin formation. Factor XII In rats, danazol was found to increase factor XII [41]. Specific studies in humans could not be located. Given the clinical efficacy of attenuated androgens in classic HAE, one might speculate that the clinically beneficial effects on other components of the bradykinin pathway (increased C-1 INH, increased APP, with secondary effects of the relative bioavailability of estrogen) outweigh the effect of increased factor XII. However, this observation has intriguing consequences for HAE-FXII. In this situation, androgen-induced increase in the over-activeThr328Lys factor XII could be deleterious. This has not been observed clinically [16], suggesting that, as in classic HAE, beneficial effects of androgens on other components of bradykinin metabolism overweight their effects on factor XII. Androgens: effect on bradykinin degradation APP Androgens increase APP levels [38]which would favor bradykinin degradation. ACE Animal studies suggest androgens are responsible for increased ACE levels [42,43]. Studies specifically addressing the influence of androgens on human ACE levels could not be located. In summary, androgens and estrogens have reciprocal, antagonistic effects on bradykinin metabolism through their effects on multiple components in these pathways relevant to the pathogenesis and treatment of classic and estrogen-related HAEs. Primary effects result in direct modification of the levels of key components in the pathways for bradykinin formation and degradation. Secondary effects, mediated through alterations in the level of SBHG, may amplify these primary effects by changing the relative bioavailability of the opposing sex hormone. High estrogen levels result in conditions favorable to increased bradykinin accumulation, whereas high androgen levels result in conditions that lead to low levels of bradykinin. The reciprocal antagonistic effects on multiple key components of bradykinin metabolism likely account for the sensitivity of disease expression to small changes in hormone levels. Exquisite sensitivity is most evident in patients with a strict estrogen-dependent phenotype [1]. For example, affected members in the family with the F 12 Thr328Lys mutation, the I allele of ACE, and the A allele of rs3788853 at the XPNPEP2 locus of the APP gene never experienced angioedema during normal menstrual cycles; however, angioedema occurred during pregnancy within days of the first missed menstrual period, a time when estrogen levels would be only marginally higher than the end of a normal cycle. Diagnosis The diagnosis of estrogen-related HAEs remains challenging as there is no specific, readily available assay. They should be suspected in the setting of otherwise unexplained episodes of angioedema, occurring in, or made worse, by high estrogen states, noting that strict estrogen dependence does not occur in every pedigree, even those with established factor XII Thr328Lys mutations [16]. Classic forms of HAE can also be exacerbated by high estrogen states, but these can be excluded if C-4, C-1 INH function and C-1 INH activity are normal [44]. Genetic analysis of suspected cases has been performed a research basis, however, the methodology required is likely within the capabilities of tertiary genetic referral centres. Identification of pre-symptomatic individuals in established pedigrees should be a priority so that exogenous estrogens (primarily oral contraceptives in young women) and the possibility of laryngeal edema can be avoided. Treatment: avoidance of aggravating medications Two distinct classes of medications contribute to bradykinin accumulation and should be avoided. Exogenous estrogens (oral contraceptives and hormone replacement therapy) have multiple effects that favour bradykinin accumulation, and have been associated with clinical exacerbations in both the estrogen-related [16] and classic forms of HAE [44]. Cardiovascular medications, ACE inhibitors, act at single point in bradykinin degradation. They have been associated with exacerbation of angioedema in both classic and estrogen-related HAEs. One patient experienced worsening of HAE-FXII with the angiotensin II receptor blocker losartan [16]; the mechanism for this effect is unclear. It would seem prudent to avoid angiotensin receptor blockers in patients with estrogen-associated HAE, if possible. Treatment: acute management Treatment experience of this newly recognized condition is limited; there are no well controlled trials. C1-INH concentrate was moderately or very effective in 6/7 patients experiencing 63 angioedema attacks [16]. Presumably, the additional C-1 INH achieved this clinical outcome by inhibiting activated factor XII and kallikrein, preventing the positive feedback loops that amplify their activity. The risks associated with this treatment would be those associated with the use of blood products. It is unclear if any of these reported uses occurred during pregnancy. Recombinant C-1 INH would be expected to have similar effects, but the potential for blood-borne infections would be eliminated. Fresh frozen plasma (FFP), is effective in classic forms of HAE [45]; its use is considered if C-1 INH concentrates are not readily available to treat an acute attack. Consideration of mechanisms responsible for bradykinin accumulation in estrogen-related angioedemas suggests FFP might be useful in these conditions. With respect to factor XII, transfusion of FFP (with normal factor XII activity) might be expected to dilute theThr328Lys factor XII with increased activity, helping to return overall factor XII activity towards normal, thereby reducing further bradykinin formation. With respect to C1-INH, transfusion of FFP would help replace any C-1 INH consumed by uncontrolled factor XII and kallikrein activation, helping to restore appropriate levels of inhibition of factor XII and kallikrein. With respect to the enzymes responsible bradykinin degradation, ACE and APP, transfusion of FFP would supplement levels in individuals having low levels of these enzymes due to genetic polymorphisms of their corresponding genes, as in individuals described [17]. Therefore, there is a theoretical basis for the use of FFP in estrogen-related angioedemas if C-1 INH concentrates are not readily available to treat an acute attack. Ecallantide, is a potent, selective, reversible inhibitor of kallikrein [46] that has recently become available for clinical use. This compound blocks the binding site of kallikrein, and reduces the conversion of high molecular weight kininogen (HMWK) to bradykinin. It also prevents the positive feedback loop in which kallikrein increases activation of factor XII, enhancing further kallikrein production. This compound has been shown to be effective in treating acute episodes of angioedema in classic HAE [47]. There are no published reports of its use in the estrogen-related angioedemas. No published data regarding use in pregnancy could be located. Icatibant, a bradykinin receptor-2 antagonist has been shown to be effective in ameliorating acute attacks of classic HAE [48]. It may be useful in the estrogenrelated angioedemas [49].Safety during pregnancy is not established. Ineffective treatments include corticosteroids, in 27 patients, and antihistamines in 15 patients, which were ineffective in controlling acute attacks [16], as is seen in patients with classic HAE. Treatment: prophylaxis Progesterone use has been reported. Eight women on various progesterone-only preparations were symptom free during progesterone treatment [16], but the frequency of previous attacks and whether these occurred only during high estrogen states is not reported, so it is difficult to evaluate whether the absence of symptoms was attributable to the use of progesterone, or the avoidance of estrogen. Further studies of the efficacy of progesterone seem warranted in patients who experience ongoing symptoms despite estrogen avoidance. However, caution is warranted as high progesterone levels have been associated with a higher number of episodes of angioedema in classic HAE [28]. Danazol use has been reported. Two patients experienced amelioration of symptoms with danazol [16]. Though not stated specifically, it seems likely that symptoms occurred during normal estrogen states. Attenuated androgens act at many points in bradykinin pathways to favour lower levels of bradykinin, thereby ameliorating symptoms. Androgens have been a cornerstone of treatment of classic HAEs for decades. However, they are contraindicated in pregnancy due to their masculinization of the fetus. The use of androgens would likely be limited to patients who experience ongoing symptoms despite estrogen avoidance, i.e., cases without strict estrogen dependence. For example, in the family with the strict EDIA phenotype [1] women of childbearing age were asymptomatic if they avoided oral contraceptives and used alternate methods of birth control, so androgens were not required. Postmenopausal individuals were asymptomatic if they avoided hormone replacement therapy (one affected individual with severe menopausal symptoms was successfully managed with very low dose transdermal estrogen without recurrence of angioedema, K. Binkley, unpublished observation), so androgens were not required. In this pedigree, identification of the phenotype allowed symptoms to be successfully managed by avoidance of triggers. Pregnancy was the only state during which treatment would be required, when androgens are contraindicated. Tranexamic acid is used in classic forms of HAE, but its efficacy is generally lower than that of the attenuated androgens. It is thought that this antifibrinolytic agent acts through the inhibition of plasmin. There is risk of thromboembolic events with its use. Tranexamic acid was used successfully in one patient with estrogenrelated angioedema [16]. It would seem the primary use of this agent would be in cases in which angioedema continued despite avoidance of estrogens. In summary, various treatment options are available for patients with estrogen-related angioedema that is not controlled despite avoidance of exogenous estrogens, though data is limited. The greatest need is for safe and effective treatments for patients who desire pregnancy. Currently, C-1 INH replacement with concentrates or recombinant C-1 INH seemed to be the best options. Conclusions In the decade since their original description, significant progress has been made in characterizing the underlying responsible genetic abnormalities in the estrogen-related HAEs. Significant clinical and genetic heterogeneity in these conditions is apparent, and it is likely that multiple genetic factors contribute to disease expression, even within the same pedigree. By extension, some of the more common genetic polymorphisms contributing to increased bradykinin accumulation, reported in patients with EDIA, might also contribute to the well-recognized phenotypic heterogeneity within individual pedigrees of classic HAEs. The emerging picture is that both classic and estrogen-related HAEs belong to a family of diverse genetic disorders of bradykinin metabolism that favour its periodic accumulation, resulting in angioedema. In both classic and estrogen-related HAEs, the profound effects of estrogens and androgens on multiple components in bradykinin metabolism pathways contribute to the expression of clinical phenotype, and have important implications for treatment. Limited data are encouraging that C-1 INH replacement is effective in treating acute attacks caused by mutations in F 12. Ecallantide and icatibant may also be useful, but further studies will be required. Optimal management of estrogenrelated angioedemas remains to be determined. Currently, definitive diagnosis remains challenging as genetic analysis is not immediately available to most clinicians. As these conditions are increasingly recognized, and the need for access to this analysis becomes apparent, specialized tertiary and quaternary genetic centres may be able to offer analysis in carefully selected patients. The most pressing needs relate to treatment during pregnancy, the one high-estrogen state that patients may be unwilling to avoid, and the one in which agents for long-term prophylaxis (androgens and tranexamic acid) are contraindicated, and safety data on agents used to treat acute attacks (C-1 INH replacement, kallikrein inhibitors, and bradykinin receptor antagonists) is almost nonexistent. Large controlled trials of treatment will be challenging due to the heterogeneity and rarity of these conditions. Genetic Characterization of Chikungunya Virus Among Febrile Dengue Fever–Like Patients in Xishuangbanna, Southwestern Part of China Co-infection of chikungunya virus (CHIKV) has been recently reported during dengue fever epidemics. However, the infection of CHIKV is often neglected due to its misdiagnosis as dengue virus (DENV) infection. In the summer of 2019 when dengue fever was epidemic, we collected 697 serum samples from febrile dengue fever–like patients in Xishuangbanna, southwestern part of China. DENV RNA was detectable in 99.42% of these patients. Notably, 88 patients (12.62%) showed the presence of CHIKV RNA, among which 86 patients were co-infected with DENV and CHIKV. We sequenced and analyzed the full genome of CHIKV virus in four out of 88 samples (two CHIKV infected and two co-infected). The results suggested that the four strains were

all Asian genotype and had the highest homology (99.4%) with the SZ1239 strain (accession number MG664851) isolated in 2012 and possibly introduced from Indonesia. Further comparison with the conserved sequences in the whole genome of 47 strains of CHIKV showed that there were 13 and 15 amino acid mutants in structural proteins and non-structural proteins, respectively. The previously reported adaptive mutations of E2-W64R, E2-I211T, E2-K233E, E1-A98T, and E1-K211E occurred in the four strains of this study. In conclusion, this study reports a co-infection of CHIKV during the DENV epidemic in the city Xishuangbanna, 2019. Molecular epidemiology revealed that CHIKV identified in this study was indigenous and belongs to Asian lineage with lineage-specific mutations and some reported adaptive mutations, which is distinct from the recently reported CHIKV (East/Central/South African) in Ruili, the city next to Xishuangbanna. INTRODUCTION Arboviral diseases, caused by arthropod-borne viruses, pose a global threat to global public health, which is recently called for priorities for research and development (Wilder-Smith et al., 2017). Arthropod vectors, essential for arbovirus transmission cycle, are mostly confined to specific ecological niches. Therefore, arboviral diseases are usually recognized as natural focal diseases. However, because of global changes of climate and socioeconomics and increase in international travel, viruses and vectors are spreading and emerging in new geographical areas where residents have no antibody or effective treatment against these pathogens. In addition, adaptive mutation of virus leads to transcontinental transmission and outbreak of disease in some unexpected regions (Tsetsarkin et al., 2007). Dengue and chikungunya diseases are mostly concerned arboviral diseases due to their morbidity, mortality, and socioeconomic burden. Dengue disease is caused by dengue virus (DENV), which consists of four serotypes (I, II, III, and IV), belonging to genus Flavivirus of the Flaviviridae family (Wen et al., 2018). Chikungunya disease is resulted due to infection of chikungunya virus (CHIKV) that is a member of the genus Alphavirus, Togaviridae family (Vasconcellos et al., 2019). Although DENV and CHIKV belong to two different viral families, they are both transmitted by the common vector Aedes mosquito. In fact, cocirculation of DENV and CHIKV has been documented in many locations (Raza et al., 2021). However, higher incidence of DENV and similar clinical manifestations contribute to neglect of CHIKV detection in patients with DENV or even misdiagnosis of CHIKV as DENV, which may lead to serious consequences. For this purpose, multiplex RT-PCR (Reverse Transcription-Polymerase Chain Reaction) has been developed to detect DENV, CHIKV, and ZIKV (Zika virus) in single tube/well, by which co-infections of DENV, CHIKV, and/or ZIKV are actually found (Sańchez-Arcila et al., 2020). Notably, during dengue fever outbreak in 2019, there are two reports of CHIKV infection in Ruili city, Yunnan Province China, where CHIKV belongs to East/Central/South African (ECSA) lineage and are imported from Southeast Asian countries Yin et al., 2021). Therefore, it is important to carry out surveillance of CHIKV. Here, we report a co-infection of CHIKV during dengue fever outbreak in Xishuangbanna and describe the viral origin of Asian lineage. Ethics Statement This study was approved by ethics committees of Xishuangbanna Dai Autonomous Prefecture People's Hospital, Yunnan, China. Informed consent was written and obtained from each participant. Sample Collection and Laboratory Diagnosis From September to November of 2019, we collected 697 serum samples from febrile dengue fever-like outpatients in Xishuangbanna Dai Autonomous Prefecture People's Hospital, Yunnan, China. Total RNA was extracted and purified from serum samples according to the manufacturer's instructions. Briefly, the serum samples were separated from the collected blood, followed by viral RNA extraction. Viral RNA was extracted from 200 µl of CHIKV-infected serum using the AxyPrep Body Fluid Viral DNA/RNA Miniprep Kit (Axygen, Inc., USA). The RNA was eluted in 50 µl of nuclease-free water and stored at −80°C. Multiplex one-step RT-PCR was performed to amplify gene fragments via specific primers for DENV (I-IV), CHIKV, and ZIKV (Calvo et al., 2016). The amplified products were visualized via agarose electrophoresis. The DNA band with the expected size was cut and purified for Sanger sequencing. Sequences of PCR products were aligned with published viral sequences to confirm the specific pathogens. Genome Sequencing Viral RNA was extracted with an AxyPrep Body Fluid Viral DNA/RNA Miniprep Kit (Axygen, Inc., USA) according to the manufacturer's instructions. Viral RNA was reversely transcribed into cDNA using the PrimeScript ™ II First Strand cDNA Synthesis Kit (Takara Bio, Shiga, Japan). The cDNA was amplified by PCR via 12 pairs of overlapping primers (Supplementary Table 2). PCR was performed with the following protocol (50-µl volume): denaturation at 94°C for 2 min, followed by 30 cycles of denaturation at 94°C for 30 s, annealing at 50°C for 30 s, and elongation at 72°C for 1 min; with a final elongation step at 72°C for 5 min. CHIKV's 5′UTR (5'untranslated region) and 3′UTR (3'-untranslated region) were amplified using SMARTer ® RACE 5′/3′ Kit (Takara Bio USA, Inc.). The PCR products were confirmed by agarose gel electrophoresis and sequenced by Sanger sequencing (Sangon Biotech, Shanghai). The full length of CHIKV genomic sequence was assembled on basis of 12 fragments, 5′UTR, and 3′UTR. Phylogenetic Analysis The full length or nearly full length of CHIKV genomic sequences were downloaded from the GenBank database (https://www.ncbi.nlm.nih.gov/genbank/) and processed via the software SnapGene (GSL Biotech, USA). The nucleotide sequences were aligned using MUSCLE. The phylogenetic analysis based on nucleotide sequences of full genomes, structure proteins, and non-structure proteins was carried out using the maximum likelihood method with the Tamura-Nei model in MEGA7. The tree with the highest log likelihood (−34,152.40) is constructed. Analysis of Characteristic Mutations Amino acid sequences of structural polyprotein (E1, E2, and E3) and non-structural polyprotein (nsP1, nsP2, nsP3, and nsP4) from 47 CHIKV strains, including four CHIKV strains of this study, were aligned with the prototypic CHIKV strain S27 via the software MEGA7. Furthermore, on basis of the prototypic CHIKV strain S27, SWISS-MODEL was used for comparative homology modeling of E1 and E2 of four strains identified in this study. Superimposed model was then visualized through PyMOL. The mutations identified in the study were highlighted as Stick format. Co-Infection of Chikungunya Virus with Dengue Virus in Febrile Patients From September to November of 2019 during the epidemic dengue fever, we collected 697 serum samples from febrile dengue feverlike outpatients in Xishuangbanna Dai Autonomous Prefecture People's Hospital, Yunnan, China. They were all local cases without any travel history. Multiplex one-step RT-PCR was performed to amplify gene fragments via specific primers for DENV (I-IV), CHIKV, and ZIKV. Sequences of PCR products were aligned with published viral sequences to confirm the specific pathogens. Among a total of 697 samples from febrile patients, 695 were positive for at least one of DENV, CHIKV, and ZIKV, and there were two samples negative for any of DENV, CHIKV, or ZIKV. Six hundred and ninety-three samples were DENV positive, accounting for 99.42% (693/697), among which DENV only was 87.09% (607/697), and coinfection of DENV and CHIKV was 12.33% (86/697). CHIKV only was 0.29% (2/697). Therefore, there were a total of 88 CHIKVpositive cases. Detailed results of PCR analysis were summarized in Table 1. Next, we further analyzed demographics and symptoms of 74 CHIKV-positive cases except 14 CHIKV-positive cases lack of symptom record. The top three symptoms of CHIKV infection were fever, myalgia, and headache, accounting for 90%. About 30% of patients showed nausea, and 26.03% were arthralgia. The remaining patients showed weakness, rash, fear of cold, and emesis with 15.07%, 10.96%, 4.11%, and 1.37%, respectively ( Figure 1A). In terms of gender and age, majority (66/88) of CHIKV-positive patients were at the age of 18-64 years, among which 31 cases were at the age of 18-40 years and 35 cases were at the age of 41-60 years. Male patients were more than female patients at the age of 18-64 years. There were 13 (seven females and six males) CHIKV-positive patients at the age of 1-12 years. Four (two females and two males) CHIKV-positive patients were at the age of 13-17 years. There was one patient with CHIKV older than 65 years ( Figure 1B). Furthermore, four CHIKV-positive serum samples, two from serum samples only with CHIKV and the other two from sera co-infected with DENV and CHIKV, were chosen for amplification of the whole genome via 12 pairs of primers, 5′ RNA was extracted and purified from serum samples. One-step RT-PCR was performed via primers specific for four genotypes of DENV, CHIKV, and ZIKV, respectively. The gene fragments obtained from RT-PCR were sequenced, which was blasted with viral sequences of DENV, CHIKV, and ZIKV on the NCBI server. Table 1). The full length of genomic sequences with 5′UTR and 3′UTR were eventually obtained from four CHIKV-positive samples and submitted to GenBank with accession numbers OK316990, OK316992, OK316993, and OK316995. Chikungunya Viruses Identified in Febrile Dengue Fever-Like Patients Belong to the Asian Lineage To trace the potential origin of CHIKV identified in this study, we firstly compared the full length of CHIKV genomic sequences with those from the NCBI GenBank database. We found that the four CHIKV strains in this study shared 99.4% homology of nucleotide with the strain SZ1239 (accession number MG664851) in the Asian lineage and 91.2% of nucleotide homology with the strains in the sublineage Indian Ocean Lineage (IOL) of ECSA recently reported in Ruili, southwestern part of China (Supplementary Table 2) Yin et al., 2021). The phylogenetic analysis of the full length of genomic sequences indicated that four CHIKV strains in this study were clustered to the Asian lineage with the CHIKV strain reported in 2012, China (Figure 2), distinct from the three strains documented in 2010, 2017, and 2019 in China that belong to the sublineage IOL of ESCA lineage. The gene fragments for structural proteins (C, E3, E1, and E2-6K) and non-structural proteins (nsP1, nsP2, nsP3, and nsP4) phylogenetically revealed four CHIKV strains in this study were also clustered in the Asian lineage (Supplementary Figure 1). To further predict the potential effects of mutations on the interaction between virus and host cell, three-dimensional structures of surface proteins E1 and E2 of four CHIKV strains in this study were modeled on basis of the prototypic strain S27 (accession number: AF369024) via the software PyMOL. Three mutations (E1-T65A, E1-V135I, and E1-V291A) in E1 protein of four CHIKV strains in this study did not make these residues protrude from the surface of E1 protein, but mutations of E1-A321V and E1-A329V resulted in obvious protrusion of amino acids (Supplementary Figure 2A). However, seven reported adaptive mutations in E1 protein led to their protruding from the surface of E1 protein, including E1-A98T and E1-K211E in CHIKV strains of this study. The mutation E2-N5R only observed in this study did not result in their protruding from the surface of E2 protein. The other two mutations (E2-S27P and E2-H123R/L) only in this study were revealed to be involved in interaction with the key residues Q63/Q64 and D68/R69 on MXRA8, one of important cellular receptor for CHIKV (Song et al., 2019), suggesting that MXRA8-binding property of CHIKV in this study may alter due to these two mutations. As expected, the nine reported adaptive mutations in E2 protein, including three mutations (E2-W64R, E2-I211T, and E2-K233E) observed in this study, resulted in their protruding from the surface of E2 protein (Supplementary Figure 2B). DISCUSSION Geographically, Yunnan province is located at the southwestern part of China and shares its border with several Southeast Asian countries (including Laos, Vietnam, and Myanmar) that are mostly affected by CHIKV. With the growth of tourism and trade with Southeast Asian countries, cases of imported CHIKV FIGURE 2 | Phylogenetic analysis of the full length of CHIKV genomic sequences. Nearly complete sequences of representative CHIKV strains from each genotype were downloaded from NCBI GenBank database. Alignment and comparison were conducted to screen the genomic sequences with more than 59 bp at 5'UTR and more than 407 bp at 3'UTR for further phylogenetic analysis using MEGA7 maximum likelihood method with 500 bootstrap values. Representative strains of each genotype in the phylogenetic tree were named by accession number, country of origin, and year of isolation. The numbers on the branches of phylogenetic tree represent the posterior probability values. The red circles represent CHIKV strains isolated in China, the red dots for CHIKV strains isolated in this study, the red triangles for the earliest ECSA isolate S27, and the red squares for the earliest

Asian isolate. The tree is drawn to scale, with branch lengths measured in the number of substitutions per site. CHIKV, chikungunya virus; WA, West African; ECSA, East/Central/South African; IOL, Indian Ocean Lineage. infection are constantly increasing, which may result in the reemergence and autochthonous transmission of CHIKV to Yunnan. The present study highlights the urgent need for continuous molecular screening and epidemic surveillance for CHIKV and its vectors to prevent future outbreaks of CHIKV infection among the human population of Yunnan. CHIKV was initially isolated from Tanzania (accession number AF369024, 1953), belonging to ECSA (Khan et al., 2002). The first CHIKV isolate in Asia can be traced to Thailand (accession number HM045810, 1958) (Volk et al., 2010). During 2018-2019, a number of imported and indigenous chikungunya cases were found in Yunnan Province, Southwest China (Xu et al., 2020;Liu et al., 2021;Yin et al., 2021). In addition molecular epidemiology revealed that CHIKV identified in this study was indigenous and belongs to Asian lineage, which is distinct from the reported CHIKV (ECSA lineage) in Ruili, the city next to Xishuangbanna, 2019 Yin et al., 2021). However, there may be other lineages of CHIKV including ESCA in this outbreak since we sequenced and analyzed only four strains of CHIKV in this study. In fact, CHIKV RNA in the other 84 samples is not sufficient for further analysis. Therefore, we should incubate serum samples with susceptible cells to isolate and amplify CHIKV in the future to get more viral RNA. It is also very important that further studies should be conducted to trace the origin of CHIKV identified in this study. Co-infection of DENV and CHIKV has been recently reported in several countries, which attracts great attentions due to its possibility of more severe diseases (Vogels et al., 2019). A multicenter study in Pakistan reveals the 12% of CHIKV coinfection in 590 DENV-positive patients at the average age of 28 (Raza et al., 2021). In India, a study in the population of 21-to 40year-old patients shows 9.54% of DENV and CHIKV co-infection (Kaur et al., 2018). The results in these two reports are consistent with this study (Table 1). Unxpectedly, 35.7% of co-infection DENV and CHIKV has been reported in children under 18 years old (Castellanos et al., 2021). However, genotypes of DENV in CHIKV samples and the lineage of co-infected CHIKV have not been analyzed in these reported studies of co-infection. Thrombocytopenia and lymphopenia are frequently documented in dengue, chikungunya, or co-infected patients (Kaur et al., 2018). Patients with dengue show more severe thrombocytopenia and lymphopenia than patients with chikungunya (Tomashek et al., 2017). Coinfection of CHIKV may increase the severity of thrombocytopenia and lymphopenia in dengue patients . To further understand the mutual effects between DENV and CHIKV during the coinfection in this study, it is necessary to perform the routine laboratory test. With the alignment of amino acid sequences, there were five of previously reported adaptive mutations also observed in the four CHIKV strains of this study, two in E1 protein (E1-A98T and E1-K211E), and three in E2 protein (E2-W64R, E2-I211T, and E2-K233E) ( Table 2). E1-A98T mutation located in the fusion loop of the E1 protein (E1: 83-100 aa) occurs only in the Asian lineage and inhibits E1-A226V mutation (Zhang et al., 2018). The mutation E1-K211E significantly enhances the adaptability of CHIKV to Aedes aegypti in the context of E1-A226 (Agarwal et al., 2016). The residue E2-W64 is located at the tip of the spike protein (52-82 aa) and interacts with cellular receptors such as MXRA8 (Song et al., 2019). The mutation E2-W64R, also observed in one CHIKV isolate of this study, has been reported to show the decreased dissemination in Ae. aegypti but not Ae. Albopictus (Tsetsarkin et al., 2014). E2-I211T interacts with E1's fusion ring and mediates virus entry, which is reported to significantly affect the infectivity of E1-A226V mutant in Ae. albopictus (Tsetsarkin et al., 2009). E2-K233E can directly enhance the infectivity of CHIKV in Ae. albopictus, mainly by alteration of viral infection and/or replication at the initial midgut (Berry et al., 2019). These reported adaptive mutations indicate that all of four CHIKV isolates in this study are potentially of high infectivity and adaption to both Ae. aegypti and Ae. albopictus, which probably increases the risk of arthropod-borne CHIKV infection in the population. In the nonstructure proteins, four CHIKV strains of this study contained two adaptive mutations (nsP1-G230R and nsP3-R524G) (Supplementary Table 3). NsP1-G230R is reported to work together with nsP3-*524 to enhance the replication of CHIKV in Ae. albopictus (Mounce et al., 2017). The mutation nsP3-524R can attenuate arthritis and pathology in a mouse model, including reducing swelling and inflammation in the feet and ankles of mice, and inhibiting recruitment of pathological immune mediators, compared with the parent virus containing nSP3-*524 (Jones et al., 2017). Through homology modeling, we hypothesize that mutations of E1-A321V and E1-A329V resulted in obvious protrusion of amino acids and MXRA8-binding property of CHIKV in this study may alter due to the two mutations (E2-S27P and E2-H123R/L). More experiments are necessary to test this hypothesis. In summary, this study reports a DENV epidemic accompanied by CHIKV infection in Xishuangbanna, southwest part of China, in 2019. CHIKV from four identified cases in this study were characterized to be distinct from CHIKV previously reported in China. CHIKV strains in this study show the mutations potentially associated with enhanced infectivity for Ae. aegypti or Ae. albopictus. These findings provide important reference and guidance for the prevention and control of arboviruses such as DENV and CHIKV. DATA AVAILABILITY STATEMENT The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material. ETHICS STATEMENT The studies involving human participants were reviewed and approved by Xishuangbanna Dai Autonomous Prefecture People's Hospital, Yunnan, China. Written informed consent to participate in this study was provided by the participants' legal guardian/next of kin. Written informed consent was obtained from the individual(s)' and minor(s)' legal guardian/next of kin, for the publication of any potentially identifiable images or data included in this article. AUTHOR CONTRIBUTIONS HL, XP, and QS contributed to the design of the study. HL and MZ were involved in writing this manuscript. MZ, CS, JL, and NL were involved in data acquisition and analysis. JZ, TL, DM, and DL contributed to collection of samples and clinical information. All authors contributed to the article and approved the submitted version. Public healthcare practitioners’ knowledge, attitudes and practices related to oral antibiotic prescriptions for dental use in Pietermaritzburg, KwaZulu-Natal Background There is limited published evidence on health workers’ perspectives on trends in oral antibiotic prescription for dental conditions in the public health sector. Aim This study set to determine healthcare practitioners’ knowledge, attitudes and practices related to oral antibiotic prescriptions for dental use. Setting This included two public hospitals in Pietermaritzburg. Methods This was a cross-sectional study using quantitative data. Purposive sampling was used to select medical and dental practitioners from Institution A and B (n = 122). A self-administered questionnaire was developed using open and close-ended questions. Data were collected and analysed using the Statistical Package for Social Sciences (IBM SPSS version 25R). Results The response rate for the study was 72.1%. The majority of study participants (n = 72, 81.8%) indicated awareness of an antibiotic stewardship programme in their respective institutions. However, a significant number (n = 42; 47.7%) of participants were unsure of whether this programme was active. Most participants (n = 80, 90.9%) indicated the need for improving oral antibiotic prescription for dental conditions. Participants indicated prescription of antibiotics for orofacial swellings (n = 52; 59.0%) and dental pain related to irreversible pulpitis (n = 29; 32.9%), reversible pulpitis (n = 33; 37.5%) and dental fillings (n = 15; 17.0%). Antibiotics were also prescribed for pericoronitis (n = 58; 65.9%), periodontitis (n = 57; 64.7%) and impacted teeth (n = 21; 23.8%). All dental practitioners (n = 14) supported the need for antibiotic cover for pericoronitis and periodontitis. Conclusion The results indicated inconsistencies in healthcare practitioners’ reported knowledge, attitudes and practices related to antibiotic prescription patterns. Contribution This study highlights the need for clear evidence-based guidelines for antibiotic prescription for dental conditions. Introduction Antibiotics have played a major role in the management of infectious diseases; however, limited access in the availability of affordable antibiotics, especially in developing countries, remains a challenge (Watkins et al. 2019). At the same time, there is a reported increase in the use of antibiotics for dental clinical management (Haliti et al. 2017). From a South African dental management perspective, antibiotics appear to be most commonly prescribed for orofacial and dental infections (Huang & Owen 2012). The overuse and misuse of antibiotics for health conditions, including dental conditions that do not require such interventions, place further strain on the already scarce resources, especially in the public health sector (Mthethwa & Matjila 2018). Studies further show that the overuse of antibiotics has collectively contributed to the emergence of resistant bacterial strains, and that this has necessitated a renewed focus on a post-antibiotic era (Boyles et al. 2017;Hoffman et al. 2015;Mendelson & Matsoso 2015). This post-antibiotic era highlights the need for managing the emergence of antimicrobial resistance through monitoring and a review of current trends and patterns in antibiotic prescriptions (Carter, Sun & Jump 2016;Crowther-Gibson et al. 2011;Truter 2015). A well-coordinated antibiotic stewardship programme could be one of the possible public health strategies to ease the burden on the public healthcare system. Likewise, dental health workers in South Africa need to be aware of the impact of antimicrobial resistance and how this could affect clinical management of dental conditions, as well as clinical outcomes (Mthethwa & Matjila 2018). Trends in antibiotic prescriptions for dental use Not much published evidence exists in South Africa on the related trends and patterns in antibiotic prescriptions for dental use; however, studies show that there are inconsistencies in how dental conditions are managed, specifically in relation to the indications for antibiotic use. The following studies highlight some of these inconsistent trends in antibiotic prescriptions. Sancho-Puchades et al. (2009) observed a mismatch between the practice of prescribing antibiotic prophylaxis for the prevention of local odontogenic infections and the recommended associated guidelines for dental clinical management. Similarly, other authors pointed out that overprescription of antibiotics occurred even when service providers such as dentists had adequate knowledge on the indications and contraindications of antibiotic use (Carter et al. 2016;Truter 2015). At the same time, Swiss dentists reported not being sure of when to prescribe antibiotics (Sweeney et al. 2004). This observation is not unique. While it might appear that dental practitioners are prescribing antibiotics judiciously in South Africa, according to Lalloo et al., there was no consensus amongst these practitioners on when and why to prescribe antibiotics for dental use (Lalloo et al. 2017a). The implications of these findings are that there could be inconsistencies in antibiotic prescription trends for dental use (Ncube et al. 2017). This study arose owing to the limited available data on the trends and practitioners' perceptions and attitudes towards antibiotic prescription in South Africa, specifically in KwaZulu-Natal (Ramnarain 2021). There are limited available data on the patterns of antibiotic prescription for dental conditions and of the treatment outcomes in terms of statistical records in the public oral health sector. This creates an unclear picture of prescription trends and patterns on oral antibiotics for dental conditions in KwaZulu-Natal (Ramnarain 2021). The resultant impact of the current trends in dental management in relation to antibiotic prescriptions and the possible antimicrobial resistance are largely unknown (Ramnarain 2021). To this end, the aim of this study was to determine healthcare practitioners' knowledge, attitudes and practices related to antibiotic prescription for dental purposes in the Pietermaritzburg region of KwaZulu-Natal. The Pietermaritzburg complex Pietermaritzburg, which is located in the uMgungundlovu district, is considered the second largest city in KwaZulu-Natal with a population of about 11 527 96 people (Statistics South Africa 2013. In addition, about 925 695 people in the uMgungundlovu district remain uninsured (Statistics South Africa 2013-2018), thus creating a huge burden on the public health system for health service provision (KwaZulu-Natal Department of Health 2015). Similarly, there is a huge dependency on the public oral health system to provide oral healthcare to the majority of the population;

yet, only a small fraction of the oral health workforce is located in this sector (Singh 2011;Singh, Myburgh & Lalloo 2010;Thema & Singh 2017). From a service delivery perspective, public oral healthcare can be accessed at the district, regional or provincial tertiary levels. The district-and regional-level facilities offer services for the management of oral conditions such as dental caries, dental impactions, periodontal disease, trauma and various types of oral pathology, while provincial tertiary-level hospital offers specialised maxillofacial services (KwaZulu-Natal Department of Health 2018). These include the management of maxillofacial injuries caused by motor vehicle accidents, assaults, sports injuries and gunshot wounds. Pathology related to the maxillofacial area is also managed at this level (KwaZulu-Natal Department of Health 2018). Methods This was a cross-sectional study using quantitative data to determine healthcare practitioners' knowledge, attitudes and practices related to oral antibiotic prescriptions for dental use (Ramnarain 2021). The research site was the Pietermaritzburg Complex in KwaZulu-Natal. Two public oral health facilities were purposively selected for the study (Institutions A and B). Purposive sampling was also used to select the study sample (n = 122). The sample comprised 108 healthcare workers and 14 oral healthcare workers (Table 1). The selection criteria included all healthcare professionals who prescribed and dispensed antibiotics for dental use. The study excluded all dental and medical managers and practitioners not involved in the clinical management and prescription of antibiotics for dental use (Ramnarain 2021). A written consent was first obtained from all participants, and ethical issues such as confidentiality and anonymity were upheld. A self-administered questionnaire with open-and closedended questions was developed for the study. The questionnaire comprised 15 questions that were designed to assess health workers' knowledge, perceptions, attitudes and practices related to antibiotic prescriptions for dental use. The first part of the questionnaire included sociodemographic data, such as gender and employment. Staff No. of participants Medical Surgeons 6 Dental therapists 6 Total 14 The questionnaire focused on participants' knowledge of guidelines developed by National Institute for Health care and Excellence and the guideline by the American Heart Association, the South African National Strategic Framework on Antimicrobial Resistance and antibiotic stewardship (Ramnarain 2021). The next part of the questionnaire focused on the health practitioners' practices related to antibiotic prescriptions for dental use. Participants were asked to indicate if they would prescribe oral antibiotics for specific dental conditions such as orofacial swellings, pulpitis and mandibular fractures or maxillary fractures. The last part of the questionnaire used open-ended questions to explore practitioners' perceptions and attitudes towards the current trends in antibiotic prescription. Participants were also asked to make recommendations on how to improve the trends in antibiotic prescription for dental use (Ramnarain 2021). Data were first cleaned and coded. A code book was prepared to ensure that codes were exclusive. Participants were also, provided code names such as P1, P2, etc. Data were then captured onto an Excel spread sheet and analysed using the Statistical Package for Social Sciences' (IBM SPSS Version 25 R ). Univariate descriptive statistics, such as frequency and mean distribution, were conducted for all variables. Bivariate statistics was also be used to assess the outcome, and the outcome was analysed by the explanatory variable (Bertani et al. 2018). Inferential techniques included Pearson's chi-square test to assess a possible relationship between the independent variable (employment or occupation) and dependent variable (antibiotic prescription). A p-value of 0.05 was established as being statistically significant. Thematic analysis was used for the open-ended questions. The responses were first read by the researcher several times to gain familiarity with the data (making connections in the data). For the subsequent steps, the data were coded and collated so that emergent themes could be further examined and compared for possible associations (through reflection, critical thinking and understanding). Results Only 88 participants returned the completed questionnaire, yielding a response rate of 72.1%. The majority of study participants were female (n = 59; 67.0%) and employed as medical practitioners (n = 74; 84.1%), while only 2.3% were community service dentists (n = 2). Knowledge of oral antibiotic prescriptions An overwhelming majority of participants (n = 85; 96.5%) were aware of the Standard Treatment Guidelines and the Essential Medicines List. Similarly, a large number of participants (n = 73; 82.9%) were aware of the American Heart Association guidelines, while more than half of the study sample (n = 53; 60.2%) was aware of the National Institute of Health Care and Excellence guidelines. More than two-thirds of the study sample (n = 72, 81.8%) indicated awareness of an antibiotic stewardship programme in their respective institutions. However, a significant number of participants (n = 42; 47.7%) were unsure of whether the institutional programme was active. Practitioner perceptions and attitudes towards oral antibiotic prescriptions The majority of participants (n = 80, 90.9%) indicated that there was a need for improving oral antibiotic prescription for dental conditions. Some of the emergent themes that arose from analysis of the open-ended questions included inappropriate prescription of antibiotics and limited availability of appropriate antibiotics. Respondents pointed out that overprescription of antibiotics occurred or that antibiotics were prescribed for conditions not requiring such intervention as reflected by the following quotes: 'Antibiotics are sometimes prescribed for viral illness.' (P108, female, aged 33, dentist) 'Over-prescription of antibiotics with insufficient diagnosis.' (P62, male, aged 52, medical practitioner) The qualitative data analysis further indicated that antibiotics are prescribed mainly for prophylaxis purposes. The limited availability of appropriate antibiotics could also influence the practitioner's pattern of antibiotic prescriptions, as indicated by the following quote: 'Limited spectrum of medication or antibiotics available after hours when the pharmacy is closed.' (P23, female, aged 42, medical practitioner) Other recommendations included proper use of prescription guidelines when prescribing antibiotics and proper monitoring and control of drug dosage. Antibiotic prescription practices More than half of the study sample (n = 60; 68.1%) indicated referring to the Standard Treatment guidelines 'some times' when prescribing antibiotics. At least half of the study sample (n = 44; 50%) followed the American Heart Association guidelines when prescribing antibiotics for the prevention of infective endocarditis, while 26 (29.5%) followed the National Institute for Health Care and Excellence guidelines for the same condition. More than half of the study sample (n = 52; 59.0%) indicated https://www.hsag.co.za Open Access that they would prescribe antibiotics for orofacial swellings, while 15 health practitioners (17.0%) were not sure (Table 2). Likewise, study participants indicated antibiotic cover for dental pain related to irreversible pulpitis (n = 29; 32.9%), reversible pulpitis (n = 33; 37.5%) and dental fillings (n = 15; 17.0%). With regard to the management of dental pain related to irreversible pulpitis, the following results were obtained from an institutional perspective. About 20 participants (57.1%) from Institution A and 16 (30.1%) from Institution B would prescribe antibiotics for this dental condition (p = 0.029). Interestingly, the majority of dental practitioners in the study sample indicated that antibiotic cover is not needed for irreversible pulpitis (n = 9), reversible pulpitis (n = 13) and dental fillings (n = 13). Only 32 health practitioners would prescribe antibiotics for maxillary and mandibular fractures, while all dental practitioners (n = 14) indicated the need to do so but 20 study participants were unsure (22.7%). A total of 34 health practitioners indicated the need for antibiotic cover for post-operative dental surgical treatment, while only 50% of dental practitioners (n = 7) supported this. It was noteworthy that more than half of the study sample (n = 50; 56.8%), including the majority of dental practitioners (n = 13) would not prescribe antibiotics for the management of aphthous ulcers. Less than half of the study participants (n = 43, 40%) indicated that sometimes they would review their clients in a followup visit after antibiotic prescription. Discussion The study findings suggested general inconsistencies in health practitioners' knowledge and practices related to antibiotic prescriptions for dental use. Participants in this study were aware of the Standard Treatment guidelines and the Essential Medicines List (96.5%), the American Heart Association guidelines (82.9%) and the National Institute of Health Care and Excellence guidelines (60.2%); yet, differences were observed in the prescription trends for the same dental condition. About 68.1% of the study sample indicated referring to the Standard Treatment guidelines 'sometimes' for the prescription of oral antibiotics. In contrast, a previous study reported that only 45% of practitioners adhered to the Standard Treatment guidelines and Essential Medicines List in the Primary Health Care in South Africa (Gasson 2018). The study findings are consistent with Marra et al. who also observed differences in antibiotic prescription trends in their study (Marra et al. 2016). Likewise, Mthethwa and Matjila reported that dentists in their study were aware of the treatment guidelines but few followed the recommendations for antibiotic prophylaxis (Mthethwa & Matjila 2018). According to Gajdács et al. (2020), healthcare practitioners had proper levels of knowledge on antibiotic therapy but were less familiar with the patho-mechanisms of infectious diseases and bacterial resistance. The implication of this finding is that adequate practitioner knowledge does not guarantee proper prescription practices because many reports have also suggested that practices are mainly influenced by the attitudes of the healthcare practitioners. At the same time, Lalloo et al. (2017b) postulated that antibiotic prescription patterns do not appear to follow a coherent set of guidelines or sound indications for antibiotic use. Inconsistencies in the recognised guidelines for antibiotic prescriptions have added to this confusion (Ramnarain 2021). The National Institute for Health Care and Excellence changed the routine prescription of antibiotics for the prevention of infective endocarditis in 2008; however, the American and European cardiologists still recommend the use of antibiotic prophylaxis for the prevention of same condition (Bunce & Hellyer 2018). These observations suggest that clarity and guidance are required at an antibiotic stewardship level. The inappropriate prescription or the excessive use of antibiotics in dental practice could occur when antibiotics are prescribed for conditions that are not indicated for antibiotic prophylaxis such as dry socket, irreversible pulpitis, acute periapical infection and pulpitis, non-clinical factors such as patient expectations, convenience or for viral infections such as herpes simplex virus-1 infections (Agnihotry et al. 2016;Mthethwa & Matjila 2018;Sancho-Puchades et al. 2009). Alarmingly, the study findings suggested that antibiotics were prescribed for dental conditions that did not require such cover. Some participants in this study indicated that they would prescribe antibiotics for pain related to reversible and irreversible pulpitis, dental fillings, pericoronitis, periodontitis, impacted teeth, trismus and aphthous ulcers. These findings are consistent with Perić et al. (2015:111) who reported that antibiotics were prescribed as a precaution because of 'uncertainty concerning the diagnosis, patient's expectations, unavailability of dental services and in short-term cases where there is insufficient time for doing any treatment'. It is also noteworthy that a significant number of study participants (especially medical practitioners) were unsure of when to prescribe antibiotics for dental use. At the same time, shortages in the availability of oral healthcare workers and inadequate infrastructure for facility-based care in the public sector (Singh et al. 2010) could be the possible reasons for the reported 'overprescription' of antibiotics in this study. This could explain why study participants opted to prescribe antibiotics for dental-related pain instead of offering or referring for treatment services. Interestingly, a pattern of antibiotic prescription also emerged in this study based on the clinical site where the respondent was located. These differences in prescription patterns for the same dental condition were surprising (Ramnarain 2021). More research is thus required to further understand these differences across clinical settings. One of the possible reasons for the reported difference in prescription patterns in this study could be professional territorialism where practitioners at one institution tend to follow a particular trend in prescribing antibiotics. Gutiérrez et al. (2006), therefore, highlight the need for professional agreement and consensus building with regard to the conditions for antibiotic prescriptions. Such efforts are also needed in a South African context to facilitate practitioner consensus building to ensure consistency in antibiotic prescription for dental use (Ramnarain 2021). This highlights the need for more practitioner support through the use of developed guidelines on antibiotic prescriptions and collaboration at a multidisciplinary team level (Lalloo et al. 2017b;Perumal-Pillay & Suleman 2017). It is concerning to note that while 81.8% of respondents reported the existence of a hospital-based antibiotic stewardship programme, yet less than half of the study sample (47.7%) were

aware of the activities related to this initiative. This suggests an urgent need to strengthen antibiotic stewardship at an institutional level (Ramnarain 2021). Antimicrobial stewardship and infection and prevention control teams could provide opportunities to augment prescribing practices and streamline this process (South African National Department of Health 2015a,b; Ntšekhe et al. 2013). Clearly, this study findings indicate serious shortcomings in the antibiotic prescription patterns for dental conditions and further highlight the obvious gaps to support medical and dental practitioners' decision-making in this regard. It is, therefore, incumbent upon the National Department of Health to review its National Strategic Framework on Antimicrobial Resistance to include the oral health sector in the planning and implementation of guidelines for antibiotic prescription for dental conditions, as well as in other related activities associated with creating awareness on the need for reducing antimicrobial resistance (Ramnarain 2021). Oral health promotion and the awareness of antimicrobial resistance should be an important part of the healthcare service provision in the public and private healthcare facilities (Ramnarain 2021). National and international recommendations and facility-specific recommendations are required to combat antimicrobial resistance (Elias et al. 2017). The hospital pharmaco-therapeutic committees should be able to provide support and mentorship to the prescribing healthcare practitioner (Ntšekhe et al. 2013). Limitations of the study This study provided valuable insight into the knowledge, practices and attitudes of healthcare practitioners regarding the trends in antibiotic prescription patterns for dental conditions. The generalisability of the study findings is limited to the study sites. There could have also been some over-reporting in respect of antibiotic prescription trends. More research is required to unpack the complexities associated with antibiotic prescription trends and patterns. The findings clearly emphasise an urgent need for clear evidence-based guidelines for healthcare practitioners that prescribe antibiotics for dental conditions (Ramnarain 2021). Conclusion The results reveal that healthcare practitioners at the identified research sites reported inconsistent knowledge, attitudes and practices related to antibiotic prescription patterns for dental use. The study highlights the need for clear evidence-based guidelines for antibiotic prescription for dental conditions. A Predominant Cariogenic Genotype of Streptococcus mutans in Schoolchildren of Mexico City: A Follow-Up Study Background: Streptococcus mutans is considered the primary bacterial species closely associated with the etiology of dental caries in humans. Recent studies suggest an association between caries and the genetic diversity of S. mutans . Objectives: The aim of this study was to evaluate the genotypic diversity of S. mutans in schoolchildren, its stability in a one-year follow-up study, and its association with dental caries. Methods: We studied 25 schoolchildren. Representative S. mutans colonies were isolated from the dental plaque of each child, grownonmitis-salivarius-bacitracinagar,andinoculatedintrypticasesoybroth. Weperformed16SrRNAgenePCR-RFLPanalysison S. mutans isolates. Dental caries in deciduous and permanent surfaces was scored according to the WHO criteria. After 12 months, the caries incidence, S. mutans count, and genotypic diversity were compared. We grouped samples to observe similarities using cluster analysis. A similarity coefficient of > 95% was considered for defining the genotypes. Results: At baseline, we proved the genotypic diversity of S. mutans with five different genotypes. Caries scores were higher in children with genotype A (dmfs 5.4 vs. 3.4). The genotypic diversity of S. mutans in the schoolchildren decreased after one year, with the predominance of genotype A (52% to 92%) that was also associated with high bacterial counts (P = 0.0063). Conclusions: This study supports that there are changes in the genotype of S. mutans over time, and the more cariogenic genotype is more stable than the others. Background Dental caries is a transmissible infectious and multifactorial disease associated with the mutans streptococci group. The group includes seven species, with Streptococcus mutans as the best-known species. The acquisition of mutans streptococci in children occurs during the first year of life, which is called the window of infectivity (1). The principal source from where infants acquire mutans streptococci is thought to be their main care provider, especially the mother, because the mothers and children present similar or identical bacterial profiles (2,3). Some authors have suggested that mutans streptococci acquisition in children increases with the erupted teeth (4). Streptococcus mutans is considered the primary species closely associated with the etiology of dental caries in humans, due to its ability to produce acid and extracellular polysaccharides (5,6). Recent studies suggest that there is a wide genetic diversity in S. mutans strains associated with caries in children (7)(8)(9)(10)(11). Many methods have been used to identify the mutans streptococci (MS) diversity, such as DNA genomic fingerprinting by pulse-field gel electrophoresis (PFGE) (12), arbitrary-primer polymerase chain reaction (AP-PCR) (11), repetitive extragenic palindromic PCR (rep-PCR), and multilocus sequence typing (MLST) (13). PCR based upon the dexA gene sequence has been successfully applied for the identification of S. mutans species. RFLP analysis of the PCRamplified 16S rRNA gene has also been used for the identification of MS species (14)(15)(16). However, the 16S rRNA gene PCR-RFLP analysis has not been applied yet to differentiate the genotypes of S. mutans species. On the other hand, strains of S. mutans that are exclusively cariogenic have not been genotyped, diversity has always been observed (7,11); the genetic stability of these strains has neither been stud-ied extensively. We only found a prospective study on the diversity and genetic stability of the strains of cariogenic S. mutans (8). Objectives The aim of this study was to identify the genotypic diversity of S. mutans types and their association with caries in schoolchildren through 16S rRNA gene PCR-RFLP analysis of S. mutans isolates in two episodes. Subjects In 2013, we initiated the study in 7 -8-year-old schoolchildren from a state-funded elementary school in Mexico City. Table salt available in Mexico City is all fluoridated. In this area, the fluoride content in water was < 0.247 ppm. We registered 36 children in the second school year. The subjects' families belonged to the mid socioeconomic status. All the children were from the same grade and staying at the school for eight hours per day. They had breakfast and lunch at the school canteen and performed unsupervised tooth brushing after every meal. These children had three holiday periods: 15 days of spring break, 10 days for year-end holidays, and 45 days between grades, during which their diet was not regulated. None of the schoolchildren had chronic diseases and received no antibiotic therapy and fluoride therapy during the last two weeks before biofilm sampling. None of them used orthodontic appliances. Oral samples were collected simultaneously with the clinical examination (baseline sample), and the procedure was repeated 12 months later. Clinical Examination Two calibrated examiners (Kappa = 0.95) performed the oral examination for caries prevalence at baseline using a natural light source, a mirror, and a periodontal probe. The caries prevalence was scored by the sum of decayed, missing, and filled deciduous surfaces (dmfs) and decayed, missing, and filled permanent surfaces (DMFS), according to the World Health Organization criteria (17). To determine the caries incidence, the oral examination was repeated 12 months later by the same examiners. To avoid bias, the schoolchildren were evaluated without access to their caries record. For each child, the caries increment was calculated by subtracting the baseline dmfs or DMFS score from the new dmfs or DMFS score. Biofilm Sampling Two hours after the most recent meal, the bacterial dental biofilm was obtained from the right lower first permanent molars by scraping 10 times from distal to mesial central fissures and pits using a short sterile hypodermic needle (18). The needle was then immersed in 2 mL of trypticase soy broth (TSB) (Difco/Becton Dickison, Sparks, MD, USA), supplemented with 200 U/L of bacitracin and 20% sucrose, using three sterile glass beads. Biofilm samples were stored on ice while transported to the laboratory and processed within two hours of collection. The samples were vortexed (Thermo Fisher Scientific, Inc., USA) for 30 s and used undiluted for further analysis. Streptococcus mutans Counts Biofilm samples were inoculated on selective mitis salivarius agar (Difco/Becton Dickison, Sparks, MD, USA) containing 200 U/L of bacitracin and 20% sucrose (19). Culture plates were incubated in 5% CO 2 at 37ºC for 72 h. After incubation, characteristic S. mutans colonies were counted and the number of colony-forming units per milliliter (CFU/mL) was determined. To present the data, the counts were transformed into an ordinal scale of < 10 4 to 10 6 and analyzed. Streptococcus mutans Identification One colony resembling S. mutans (rough appearance and adherent to the agar) from each sample was transferred to 8 mL of TSB containing 200 U/L of bacitracin and 20% sucrose, and incubated at 37°C for 48 h. The genomic DNA of the isolates cultured in TSB was extracted using the Wizard ® Genomic DNA Purification kit (Promega Corporation, Madison, WI, USA) according to the instructions provided by the manufacturer with a modified lytic enzymes mixture consisting of 10 mg/mL lysozyme and 10 mg/mL mutanolysin (20). Prior to genotyping by PCR-RFLP analysis, all isolates were confirmed as S. mutans using S. mutans-specific primers for the dexA gene to specifically amplify a 1272-bp fragment, based on the description provided by Igarashi et al. (14,21). The dexA gene was amplified with the primers, SD1, 5´-TAT GCT GCT ATT GGA GGT TC-3´and SD2, 5´-AAG GTT GAG CAA TTG AAT CG-3´. The PCR mixture consisted of 1X PCR buffer, 1.5 mM MgCl 2 , 60 ng/µL acety- extension at 72°C for 1 min, and the final extension at 72°C for 10 min. Streptococcus mutans Genotyping Streptococcus mutans genotyping was performed by 16S rRNA gene PCR-RFLP analysis using the universal primers 8F and 1492R. Amplification was performed according to the conditions previously described by Sato et al. (16). The primer sequences were 8F, 5´-AGA GTT TGA TCC TGG CTC AG-3´and 1492R, 5´-TAC GGG TAC CTT GTT ACG ACT T-3´. The PCR mixture consisted of 1X PCR buffer, 3.0 mM MgCl 2 , 200 µM each of dNTPs, 1.0 µM of each primer, 2.5 U of Taq DNA polymerase, and 1.0 µg of DNA. Amplification was performed by: initial denaturation at 94°C for 10 min, followed by 40 cycles of denaturation at 94°C for 30 s, annealing at 52°C for 40 s, an extension at 72°C for 40 s, and the final extension at 72°C for 7 min. The amplicons of the 16S rRNA gene were digested with restriction endonuclease HaeIII (Promega Co.). The digestion mixture consisted of 1X buffer C, 10 µg/µL acetylated bovine serum albumin, HaeIII 10 U/µL, and 1 µg of PCR product. Digests were incubated at 37°C for 90 min and separated on a 3.0% agarose gel in Tris-borate EDTA buffer at 80 V for 5 h. Images were captured with the Gel-Doc ™ XR System (Bio-Rad, Hercules, CA, USA). The products' digestion patterns were analyzed to create dendrograms with Gene Tools and Gene Directory software (Syngene, Cambridge, UK). Similarity percentages were obtained from the unweighted pair group with mathematical average (UP-GMA) based on Dice coefficients. Band position tolerance was set at 1.25%, which provided a hierarchical cluster representation of similarities between samples and indicated S. mutans strain level grouping. A similarity coefficient of > 95% was considered to define the genotype. Statistical Analysis Statistical analysis was performed using the JMP ® version 8.0.2 program (SAS Institute Inc., NC, USA). The chisquare test and Fisher's exact test were applied to determine differences in the distribution of S. mutans genotypes according to gender at baseline and after one year, respectively. The Student's t-test was used to evaluate differences in dmfs/DMFS scores and S. mutans counts of all the children between baseline examination and one-year examination. The analysis of variance (ANOVA) and Student's ttest were applied to evaluate differences in dmfs/DMFS values and S. mutans counts between genotypes at baseline and after one year, respectively. Values of P < 0.05 were considered significant. Regarding S. mutans genotyping at baseline, five genotypes were found among the 25 isolates ( Figure 1A). Genotype A (52%) and genotype B (36%) were more frequent. C, D, and E genotypes were found in one child only (Figure 2). After 12 months, the genotypic diversity was reduced from five to two (C, D, and E genotypes were not found) ( Figure 1B). In this episode, 23 children presented genotype A (92%) and genotype B was identified in only two children ( Figure 3). There was no difference between genders in the distribution of genotypes at baseline

(P = 0.5799) and after 12 months (P = 0.4867). Of the 14 boys (56%), seven presented genotype A, five genotype B, one genotype C, and another genotype E. Of the 11 girls, six presented genotype A, four genotype B, and one genotype D. No significant differences were found in relation to the distribution of genotypes by gender and age (P = 0.5799). When analyzing the genotype by age, giving weight to the gender variable, age did not influence the distribution of genotypes (P = 0.4177). Therefore, the data are presented for the whole study group. Regarding the caries prevalence at baseline, the 25 children had a mean dmfs of 3.84 ± 5.28 and a mean DMFS of 0.32 ± 1.06. No significant differences were found between dmfs/DMFS scores at baseline and after one year (P = 0.62). The dmfs/DMFS scores were not statistically associated with the genotypes at baseline and after 12 months. However, the children with genotype A presented more caries lesions than children with other genotypes ( Table 1). The analysis of variance revealed that those children harboring genotypes A and C had the highest S. mutans counts at baseline (P = 0.0063). In contrast, no associations were found between S. mutans counts and genotypes A and B after 12 months (Table 1). The Student's t-test revealed that baseline and final S. mutans counts did not differ among genotype A carriers (P = 0.1169). Discussion In this study, 16S rRNA gene PCR-RFLP analyses were used to evaluate the genotypic diversity and stability of S. mutans strains isolated from a group of schoolchildren. These analyses have been successfully applied for the differentiation of mutans streptococci species (16). However, this analysis had not been used to genotype S. mutans strains. For the genotyping of S. mutans, the techniques of AP-PCR (11), rep-PCR, and MLST (13) have been used so far. However, these techniques are usually more laborious and expensive, so the use of the 16S rRNA gene PCR-RFLP analysis offers a simpler and cheaper method to compare genotypic diversity and stability of S. mutans. The results of this study showed that S. mutans displayed a genetic diversity in children with caries, which is in line with previous reports (7,(9)(10)(11)22). With respect to the group of schoolchildren studied in this work, the genotypic diversity of S. mutans included five genotypes at baseline, two of which were more frequent: genotype A and genotype B (Table 1). Interestingly, we noticed that children who carried genotype A, which was more prevalent at baseline and 12month follow-up than the other found genotypes, showed higher values of CFU/mL, as well as the highest dmfs and DMFS scores (Table 1). When running a model giving weight to the number of CFU/mL, into which caries indices were entered as the dependent variables and persistence of genotype as the independent variable, we found that genotype A was influenced by the amount of CFU/mL. These results show that S. mutans counts and dmfs/DMFS indices were higher in those children harboring genotype A, except the child who presented genotype C. In this sense, a positive correlation has been reported between S. mutans counts in biofilm and caries indices, showing that children with high counts of S. mutans had high caries scores (23,24). After 12 months, a loss of genotypic diversity of S. mutans was observed, with genotype A still as the dominant genotype ( Figure 3). To the best of our knowledge, only one study has investigated. The stability of genotypes over time (8), reporting that genotypes decreased and only prevailed the dominant genotypes. We postulate that genotype A possibly colonized children more easily and displaced other genotypes after one year. Thus, the presence of cariogenic competing bacteria with probably different virulence abilities, under favorable environmental conditions, may facilitate the occurrence of microevolution and might influence the stability and adaptability of certain S. mutans genotypes in the oral cavity. These changes in the microenvironment and the stability of strains can also be observed in the findings of Dame-Teixeira et al. (22), in which, despite decreasing the number of genotypes after partial dentin caries removal, the virulence of the strains remained. The reduction in the number of genotypes observed after one year could be associated with the horizontal transmission of genotype A as a predominant genotype of S. mu-4 Jundishapur J Microbiol. 2019; 12(4):e82869. tans among the schoolchildren in this study. This assumption might be influenced by the fact that our study group had breakfast and lunch at school that is in the same place, which may promote horizontal transmission. Addition- Figure 2. Dendrogram of genotypic diversity among Streptococcus mutans strains isolated from schoolchildren at baseline ally, as mentioned above, children brushed their teeth after every meal without supervision. In this regard, it has been reported that cultural differences, such as eating with the same cutlery, sharing the same toothbrush, or defective washing up routines, increase the possibility of horizontal transmission (25)(26)(27). The variability in transmission can be associated with individual susceptibilities (1). Conclusions In conclusion, we found diversity in the S. mutans genotypes among the study group of schoolchildren. This diversity of genotypes decreased after 12 months and the more stable genotype associated with dental caries predominated. Significance Analysis of Microarrays (SAM) Offers Clues to Differences Between the Genomes of Adult Philadelphia Positive ALL and the Lymphoid Blast Transformation of CML Philadelphia positive malignant disorders are a clinically divergent group of leukemias. These include chronic myeloid leukemia (CML) and de novo acute Philadelphia positive (Ph(+)) leukemia of both myeloid, and lymphoid origin. Recent whole genome screening of Ph(+)ALL in both children and adults identified an almost obligatory cryptic loss of Ikaros, required for the normal B cell maturation. Although similar losses were found in lymphoid blast crisis the genetic background of the transformation in CML is still poorly defined. We used Significance Analysis of Microarrays (SAM) to analyze comparative genomic hybridization (aCGH) data from 30 CML (10 each of chronic phase, myeloid and lymphoid blast stage), 10 Ph(+)ALL adult patients and 10 disease free controls and were able to: (a) discriminate between the genomes of lymphoid and myeloid blast cells and (b) identify differences in the genome profile of de novo Ph(+)ALL and lymphoid blast transformation of CML (BC/L). Furthermore we were able to distinguish a sub group of Ph(+) ALL characterized by gains in chromosome 9 and recurrent losses at several other genome sites offering genetic evidence for the clinical heterogeneity. The significance of these results is that they not only offer clues regarding the pathogenesis of Ph(+) disorders and highlight the potential clinical implications of a set of probes but also demonstrates what SAM can offer for the analysis of genome data. Introduction Array CGH (aCGH) has shown itself to be a mature technology capable of detecting genomic gains and losses at a resolution of at least 250 base pairs. Clearly at this resolution there will be masses of data that will challenge the analyst, particularly in the light of the general variance of the genome from individual to individual-so called copy number variations (CNV) and knowledge that the function of a substantial part of the genome is still unknown hence referred to as 'orphan' or 'predicted'. It is also clear that genomic copy number aberrations (CNA) associated with diseased cells are likely to interfere with transcriptional pathways and affect gene function. Significance Analysis of Microarrays (SAM) was developed by Tusher 1 as a straightforward way of comparing data sets using an internally generated false discovery rate (FDR) as the criterion for the identification of probes that significantly differ between 2 or more classes. It has been successfully used to analyse gene expression data and is now routinely applied. 2,3 Philadelphia positive malignant disorders are a clinically divergent group of leukaemias with a unique identifying feature, the BCR/ABL1 fusion gene, usually resulting from the chromosome rearrangement t(9;22)(q34;q11) or its variants, that leads to constitutive expression of an aberrant tyrosine kinase. These include chronic myeloid leukaemia (CML) and de novo acute leukaemia of both myeloid Ph(+)AML and lymphoid origin Ph(+)ALL. The latter two disorders are clinically aggressive and therapy challenging even in the era of the powerful tyrosine kinase inhibitors. CML is a multistage progressive disorder which if untreated inevitably ends as fatal acute myeloid or lymphoid blast transformation. The latter, from which it has been reported to differ karyotypically, is usually clinically indistinguishable from Ph(+)ALL the most common type of ALL in adults. 4 Although non-random chromosome changes may accompany disease progression in CML, the genetic background of the malignant transformation from the benign chronic phase (CP) to acute leukaemia (blast crisis, BC) is poorly understood. Recent whole genome screening identified a spectrum of cryptic aberrations associated with disease progression 5,6 sometimes even present at the onset of CML. 7 Also similar investigations of Ph(+)ALL in both children and adults identified recurrent cryptic loss of Ikaros, required for the normal B cell maturation 8 in addition to the known deletions of the p16 (CDKN2A) gene. These findings led us to look in CML and Ph(+) ALL for imbalances in DNA sequences significantly associated with the disease stage and lineage origin. We used array CGH data obtained from 40 anonymous bone marrow samples comprising 10 CML chronic phase, 10 CML lymphoid blast phase, 10 CML myeloid blast phase, 10 Ph(+)ALL from the UKALLXII(R) trial [9] and 10 peripheral blood samples from disease free individuals. Of the ALL samples 5 had t(9;22)(q34;q11) as a sole cytogenetic abnormality, one was Ph negative but BCR/ABL positive, 3 showed hyperdiploid karyotype (HEH) but none showed aberrations detectable by G banding in the short arm regions of chromosome 7 and 9 ( Table 1). The presence of the BCR/ABL1 fusion gene was confirmed in all samples by qPCR and/or FISH (D-FISH probe, Vysis, USA) as reported previously. 9 All Ph(+)ALL and 5 out 10 CML blast crisis samples had established B cell immunophenotype. 6,10 Having identified genome regions of potential interest, ranked in order of significance, out of the thousands of array results, it is then a challenge to design further experiments to evaluate their contribution to the biology of the BCR/ABL positive disease. Materials and Methods Array CGH analysis was performed as described previously. 6 Briefly, Agilent (Wilmington, DE, USA) oligonucleotide arrays were hybridized following the manufacturer's protocol. 500 ng genomic test DNA was extracted from either peripheral blood or BM samples. Sex mismatched pooled DNA from peripheral blood mononuclear fraction of 6-8 disease free individuals (Promega, UK) was used as reference. Customized Agilent oligonucleotide arrays comprising 8 × 15 k probe sets per slide were designed from an analysis of active loci (hot spots) in the CML BC genome corresponding to pairs of probes that exceeded a 3SD threshold in 3 or more CML BC samples from a previous study. 5 Probes were selected to cover regions at ~1 k intervals except where the presence of repetitive sequences disallowed the inclusion of reliable probes. The arrays were scanned, features extracted and the data analyzed using an Agilent scanner and Mathematica software (http://www.wolfram.com). In addition, all samples had been subjected to whole genome screening using 105 K Agilent oligonucleotide arrays as part of published study. 6 The emergence of high throughput technology such as microarrays raises a fundamental statistical issue relating to testing hundreds of hypotheses thus rendering the standard P value meaningless. 11 The False Discovery Rate (FDR) concept is an alternative to the P value. Tusher 1 described such a method: Significance Analysis of Microarrays (SAM) and the implementation due to Chu et al has been incorporated by J Craig Venter Institute into their suite of 'MeV' routines. 12 SAM uses permutations of sample labels to estimate the FDR. We report the application of SAM for 5,000 permutations setting the median number of false significant probes to zero, for the supervised analysis of the myeloid and lymphoid blast crisis, chronic CML, Ph(+)ALL and control samples. Firstly, after removing data for the sex chromosomes, we constructed a table defining the log fluorescence ratio (FR) for each locus and assigned classes eg, Lymphoid blast phase CML (L) or Myeloid blast phase CML (M); Ph(+)ALL (ALL); Chronic phase CML (C); Control (Ctrl); Male (m) or Female (f). We chose a two class unpaired test and applied SAM to ask if there were

any probes that were uniquely associated with either classification. All genome addresses are derived from build 35 (March 2006) of the Human Genome. genomic difference between lymphoid and myeloid lineages We applied SAM to seek correlations between genome imbalances and clinical presentation. We asked which probes were significantly involved in the discrimination between lymphoid and myeloid lineages using the classes of myeloid and lymphoid CML BC as a model. Altogether we identified 489 significant probes, the top 100 of which were restricted to the TCR, IKZF1 and IgH genomic regions. Figure 1 shows cluster analysis Table 2). Lymphoid samples including Ph(+)ALL clustered together displaying losses ( Fig. 1 on the left), while the myeloid blast crisis and chronic CML samples formed a separate cluster with the control samples ( Fig. 1 on the right). We noted that 3 samples sat at the myeloid/lymphoid borderline and that some of the control samples showed losses in the TCR region. Comparison of CML lymphoid blast crisis and Ph(+)ALL 84% of the 155 probes differentiating lymphoid blast crisis CML and Ph(+)ALL map to one of two regions of the short arm of chromosome 9, namely 9p21.3-p21.2 and 9p24.1-p23, the latter housing genes PTPRD and MLLT3 among others. A hierarchical cluster analysis shows that five of the 10 Ph(+)ALL cases form a cluster of gains (Fig. 2, in red) although cytogenetic revealed no structural or numerical changes of 9p ( Table 1). In contrast, half of the 10 CML BCL cases formed a cluster with extensive genome loss (in green) that had been previously shown to be complex by G-banding and 105 K oligonucleotide array. 6 See Figure 2 and Table 3. Since many of the significant probes fell on chromosome 9p we repeated the analysis excluding all chromosome 9 loci. The top 10 of 80 probes meeting our significance threshold are revealed and GSTT1 in band 22q11.23, one of the most commonly reported polymorphic marker (CNV) in man. Genome loss (in green, Fig. 3) dominates the profile of 6 out 10 Ph(+)ALL samples. Surprisingly 5 of the latter cases (297, 299, 300, 301, 303) exhibit gains in the chromosome 9p21-p24 region (Fig. 3, heat map A). It is suggested from the heat maps in Figures 2 and 3 that the Ph(+)ALL samples split into two by cluster analysis (see Fig. 3 and Table 4). Associated with these loci are known genes such as PDEA4 (cAMP-diestarase) in band 19p13.2 groups, 5/10 cases showing dominant amplification of loci in the chromosome 9p region and losses elsewhere in the genome, while the remainder (5/10) lack recurrent genome imbalances. However, we were unable to detect any consistent differences in the two groups of Ph(+)ALL samples from an inspection of their chromosome status (see Table 1). In summary, SAM analysis revealed that while the lymphoid blast stage CML and Ph(+)ALL samples share common losses within the IGH, TCR, and Ikaros gene regions together with loci within the 9p21-p24 region, they form separate clusters at other sites on the genome thus suggesting that these acute malignant conditions may represent separate biological entities. Discussion Whilst huge progress has been made in the analysis of the genome and the identification of genes associated with malignant disease, there is still much work to be done evaluating the function of coding and noncoding regions 13 . We have identified numerous short 60 mer sequences that appear to play a significant role in the evolution of Philadelphia positive hematological malignancy. We offer no explanation of their function, but provide convincing evidence that their involvement is not a random event. SAM is used for the analysis of expression arrays to classify samples into groups according to phenotype using false discovery rate (FDR) as a test for significance. 1,14 Here we use SAM to study DNA from a cohort of CML and Ph(+)ALL patients to identify sequences that may help to distinguish between these Philadelphia positive diseases and enlighten their pathogenesis. Numerous software packages are available for the detection of genomic gains and losses across a range of array technologies, reviewed by Shah, 15 but highresolution array data presents special problems as typified by a wide variance making detection of small features complicated. Individual signals are rarely if ever considered to be significant on their own but only in the context of a contiguous collections of gains and losses. However if an individual locus is compared across a number of similarly processed arrays, the probability of a random single signal exceeding a 3SD threshold for n arrays is reduced to approximately (0.003)^n. FR signals meeting these criteria could be attributable to artifacts of the array and laboratory procedures or could be bona fide data with clinical significance. Since all the samples in our study were prepared under the same conditions and hybridized to the same batch of arrays, we believe that the results reflect recurrent genome copy number aberrations. Our data shows consistent, recurrent gains and losses although in many cases the FR data falls well short of the theoretical values suggesting the presence of clonal cell populations-a common phenomenon Case_284 Case_273 A B Figure 3. Ph positive all with gains at 9p21-p24 share common losses elsewhere in the genome. Notes: Cluster analysis of SAM data identified cryptic gains within the 9p21-p24 region in a sub-group of Ph(+)ALL samples (framed in heat map A, losses in green and gains in red). When cluster analysis was performed on SAM results of the genome excluding chromosome 9 data, some 115 probes were shown to discriminate between Ph(+)ALL and BCL CML, the top 10 of which are shown on the heat map B (losses in green and gains in red). All but one (arrowed) of these Ph(+)ALL samples with gains in the 9p21-p24 region share cryptic loss elsewhere in the genome (framed, heat map B), which involves relevant genes such as CYP11B2 (cytochrome P450), gCSTT1 (member of the glutathione transferase gene family with known role in carcinogenesis) and ChAF1A (chromatin assembly factor) among others. in haematological disease. SAM does not guarantee that the list of 'significant' loci are involved in the various classifications discussed here, but it does offer a list of candidates for cluster analysis or other investigative methods. Many gene copy number changes irrespective of the length of the affecting sequences such as those recorded here can contribute directly to monogenic diseases 16 . In recessive diseases, hemizygosity due to deletion of a gene, or part of a gene, could unmask a mutation on the other gene copy, while duplication of a healthy gene copy on one chromosome could theoretically mask the effects of a disease-causing mutation in the gene on the other chromosome, thus rescuing the phenotype. We designed a high resolution array (~1 kb intervals) designed to explore regions of the genome shown previously at low (33 kb) resolution to display gains or losses in a cohort of 35 samples from CML patients in blast phase 6 . This necessitated sacrificing large areas of the genome to concentrate on these areas for detailed inspection. Using this set of ~15,000 genetic loci enabled us to confirm that lymphoid phenotypes formed a single group characterized primarily by unique deletions within the IgH regions consistent with an early VDJ rearrangement as part of the B cell receptor formation occurring in per-B cells together with loss of the TCR gamma sequences also indicating gene rearrangement. Loss of whole or part of the IKZF1 (Ikaros) gene is the third most common feature in the genome profile of these cases. In contrast with a typical CNV that could affect any part of the IgH gene on 14q32.33 the deletions identified by us always involve the sequences 105.41-105.48 mbp and are almost universally accompanied by deletions in the TCR region of chromosome 7. Both IgH and TCR sequences are usually excluded from aCGH analysis as they are reported to be CNVs. We have demonstrated that these deletions are consistent throughout the sample set, suggesting that they are disease specific. These findings could be explained by a chain of events initiated by BCR/ABL1 that leads to compromised V(D)J recombinase machinery thus creating clonal populations of early B-cell progenitors with cross lineage rearrangements. 6 Mullighan et al in their poster presentation "Genome wide analysis of Genetic Aberrations in Chronic Myeloid Leukemia" (Mullighan et al, http://ash.confex.com/ash/2008/ webprogram/Paper5715.html) reported results from SNP analysis of 90 CML samples of which 9 were diagnosed as lymphoid blast crisis. This study could not find any genomic features that could differentiate between BC/L and Ph(+)ALL. In contrast we were able to reveal genomic differences in these clinically similar conditions. Many of the 'significant' probes that distinguish between Ph(+)ALL and BC/L cluster within chromosome 9p21 region, which harbours the CDKN2A/B gene, the loss of which has long been associated with both haematological and solid tumours and shown to result from RAG impairment. 8 to carry imbalances of the short arm of chromosome 9, it is possible that some probes in this location were lost 'by association' and not involved in discriminating between these two diseases. However, we found other loci that did discriminate between Ph(+)ALL and BC/L CML. For example, while half of the BC/L cases had deletions in chromosome 9p, half of the of Ph(+)ALL showed gains at these loci as shown in Figure 2. Omitting chromosome 9 records and reanalyzing the data, the same five Ph(+)ALL samples showed significant losses in 80 loci from other chromosomal locations (Fig. 3). Taken together the tandem CNA-gains at 9p with recurrent losses elsewhere in the genome offer a way to differentiate a Ph(+)ALL from CML lymphoid BC. Whilst we recognize that single aberrant 60 mer sequences could easily be dismissed as random events, the fact that there are more than 80 widely distributed probes not associated with morphological or cytogenetic anomalies but associated with a significant minority of the Ph(+)ALL samples is worth consideration. Further work is required to explore the possible role of these genome aberrations. In conclusion, SAM results offer clues regarding the pathogenesis of BCR/ ABL1 positive disorders and furthermore identifies a sets of probes with diagnostic potential. Authors' Contribution CG carried out the SAM analysis and co-wrote the manuscript; EPN designed the study and co-wrote the manuscript. Cytocell. Research funding was also received from operation of a clinic associated with UCL. CG discloses no conflict of interest. Disclosures and Ethics As a requirement of publication author(s) have provided to the publisher signed confirmation of compliance with legal and ethical obligations including but not limited to the following: authorship and contributorship, conflicts of interest, privacy and confidentiality and (where applicable) protection of human and animal research subjects. The authors have read and confirmed their agreement with the ICMJE authorship and conflict of interest criteria. Diagnosis of acute canine leptospirosis using multiple laboratory tests and characterization of the isolated strains Background Dogs presenting with acute leptospirosis may present non-specific clinical and laboratory findings, and the definitive diagnosis may require additional confirmatory tests, including bacterial culture, for the direct or indirect identification of the pathogen. The present study describes the diagnosis of leptospirosis in suspected dogs based on the use of multiple diagnostic tests, including serological, molecular and bacteriological tests, along with the characterization of the recovered leptospiral strains. Results Urine, serum and blood samples were collected from 33 dogs with suspected clinical leptospirosis treated at the University of São Paulo Veterinary Hospital Service (Hovet FMVZ-USP) between 2013 and 2016. Only dogs with high blood urea nitrogen and creatinine levels in association with multiple clinical manifestations of the disease were included. Leptospiral culture, PCR and serology (Microscopic agglutination test - MAT) were performed in blood and urine samples taken from all suspected dogs at clinical presentation, and an additional prospective MAT titration was performed in seven dogs. Infection could be identified exclusively by PCR in 10 dogs (30.3%), exclusively by MAT in four dogs (12.1%) and by both tests in four dogs, totaling 18 dogs (54.5–95%CI: 37.6–71.5). Six out of eight MAT-confirmed cases presented with the highest titers against the Icterohaemorrhagiae serogroup. Leptospires were recovered from urine samples from two PCR-positive dogs, and both strains could be characterized by Multilocus Sequence Analysis and serogrouping as L. interrogans serogroup Icterohaemorrhagiae. Both isolates were shown to be pathogenic in the hamster model. Conclusions The simultaneous use of MAT and PCR was able

to increase the diagnosis of leptospirosis in clinically suspected cases. Despite the increasing incidence of new serovars affecting dogs being reported in different locations, our results suggest that leptospiral strains belonging to the Icterohaemorrhagiae serogroup are still a major causative agent of canine leptospirosis in São Paulo, Brazil. Electronic supplementary material The online version of this article (10.1186/s12917-018-1547-4) contains supplementary material, which is available to authorized users. Background Leptospirosis is a bacterial disease caused by infection with pathogenic species of the genus Leptospira, which can affect virtually all mammals [1]. Dogs are considered highly susceptible to the infection because of a marked environmental exposure to leptospires, and canine leptospirosis has been largely described worldwide [1,2]. Recent reports have shown the reemergence of clinical illness in dogs and humans [1,3], highlighting the importance of improving current diagnostic approaches and prevention strategies. Infected dogs may manifest a broad spectrum of clinical symptoms, varying from hepatic and renal failure, often accompanied by hemorrhagic and pulmonary disorders, to mild, self-limiting febrile illness and asymptomatic infections [4]. Clinical and laboratory findings are usually non-specific, and a definitive diagnosis requires additional confirmatory tests for the direct or indirect identification of the pathogen, such as dark-field microscopy, Polymerase Chain Reaction (PCR), bacterial culture and Microscopic agglutination test (MAT) [1]. Since the exact time of infection at clinical presentation is typically unknown, and given that the early and accurate identification of infected dogs is crucial to alter the course of the disease with appropriate drug therapy [5], the use of multiple simultaneous tests may improve the chance for a correct diagnosis and consequent therapeutic success [6]. Despite the development of several serological tests for the detection of anti-Leptospira antibodies in recent years [7][8][9], the MAT is still widely employed for the serodiagnosis of acute leptospiral infection [10]. However, critical limitations may hamper its use in a clinical setting; the evaluation of one single MAT test may fail to detect antibodies at the early phase of the disease, and titration of convalescent serum samples is frequently required to reveal seroconversion, which may impose some diagnostic difficulties due to the high mortality of the disease. MAT also has a poor ability to predict the infecting serovar and may not distinguish between infection and vaccine-induced titers [10]. Conversely, PCR has been successfully used to confirm leptospiral infection at the early stages of infection [6,11], and sequencing PCR amplification products has enabled the identification of the different leptospiral species infecting dogs [12,13]. Nonetheless, culturing leptospires still stands as the gold standard reference test to unmistakably confirm leptospiral infection, and only serological characterization of the isolated strains may provide reliable information regarding serogroup or serovar identity [1]. The establishment of a panel of leptospiral strains circulating among dog populations remains the major strategy to support the development and commercialization of vaccines with more specific serovar composition, which would hypothetically increase immunization effectiveness for local canine populations. Unfortunately, culture is challenging due to frequent contamination and the fastidious growth of the pathogen [1]. Recovering leptospires from suspected dogs is also limited by the early institution of therapeutic intervention, which is usually required after the disease is suspected [14]. Few reports have studied canine leptospirosis in Brazil [15], and the characterization of leptospiral strains recovered from suspected clinical cases is poorly documented [16]. Given the benefits of using multiple laboratory tests to increase the likelihood of an accurate diagnosis, we hereby describe the diagnosis of leptospirosis in suspected cases treated at a University reference center, based on the association of multiple diagnostic strategies, such as serological, molecular and bacteriological tests, along with the characterization of the recovered leptospiral strains. Sample collection and inclusion criteria Urine and blood samples were collected from 33 dogs with suspected clinical leptospirosis presented to the University of São Paulo Veterinary Hospital Service (Hovet FMVZ-USP) between 2013 and 2016. The dogs were suspected of acute leptospirosis when presenting with high serum blood urea nitrogen (BUN) and creatinine levels of unknown origin (> 60 mg/dL and 1.4 mg/dL, respectively) in association with two or more clinical manifestations suggestive of leptospirosis (hemorrhagic disorders, fever, vomiting, jaundice, prostration, hyporexia/anorexia). Dogs presenting with polyuria, polydipsia and weight loss a month prior to the presentation at the veterinary ambulatory were not included in the study in order to rule out chronic kidney disease bias. Leptospiral culturing, PCR and MAT were performed in all suspected dogs at clinical presentation, and samples were collected prior to the institution of antimicrobial therapy. Biochemistry and hematological parameters were also determined and included blood urea nitrogen (BUN) and creatinine (CR) serum concentrations, alkaline phosphatase (ALP) and alanine aminotransferase activity (ALT), including also hematocrit (Ht) and white blood cell count (WBC). Reevaluations were performed in seven dogs and included exclusively serum antibody titration and evaluation of biochemistry and hematological parameters; no PCR or bacterial culture was performed in the prospective evaluation. Blood samples were collected from the jugular or cephalic vein and drawn into BD Vacutainer tubes (BD Diagnostics, New Jersey, USA) and Venosafe™ tubes containing K 3 EDTA (Terumo, Terumo Europe N.V, Leuven, Belgium) to obtain serum and whole-blood samples, respectively. All urine samples were taken aseptically by cystocentesis. Diagnostic criteria Acute leptospiral infection was confirmed by the detection of at least a four-fold increase in MAT titers between acute and convalescent serum samples [1] and/or the presence of MAT titers ≥800 in single serum samples from non-vaccinated dogs, as previously suggested [5]. Dogs with positive PCR results (urine and/or blood samples) were considered to be infected only after identification of Leptospira spp. was confirmed by DNA sequencing. Anti-Leptospira antibody detection MAT was performed to detect anti-Leptospira antibodies in patient's serum samples [17] using a panel of 22 serovars representing 18 serogroups, as previously described [12]. Endpoint titers were determined using two-fold dilutions until the last well showing 50% agglutination was recorded. The cut-off for a positive MAT reaction was defined as a titer ≥100; however, only non-vaccinated dogs with single MAT titers of 800 or above were considered to be acutely infected. Culturing of Leptospira For the recovery of leptospires, 0.5 mL aliquots from blood and urine samples were diluted in sterile physiological solution to a final concentration of 1:10 and 1:100, and 0.5 mL of each solution was further inoculated in semi-solid Fletcher and liquid EMJH medium (Difco Laboratories, Franklin Lakes, NJ, USA). The tubes were incubated at 28°C for 12 weeks and examined weekly by dark-field microscopy for the presence of spirochetes. Leptospiral DNA detection DNA was extracted from blood or urine samples using NucliSens® miniMAG™ (BioMérieux Inc., Durham, NC, USA) according to the manufacturer's instructions. Urine samples were centrifuged (6.500 G-force, 25°C, 25 min) and pellets were resuspended in 1 ml sterile phosphate-buffered saline (PBS -pH 7.2) prior to DNA extraction. Extracted DNA was subjected to PCR amplification using a previously reported Leptospira genus-specific protocol and primers targeting a 331 bp fragment of the 16S rRNA gene [18]. Cycling conditions were carried out as follows: 94°C for 5 min, 40 cycles at 94°C for 30 s, 60°C for 30 s, 72°C for 30 s and a final extension at 72°C for 5 min. Pure L. interrogans serovar Canicola (strain Hond Utrecht IV) genomic DNA was used as a positive control and DNase-free water as a negative control in all PCR runs. The amplified products were separated by electrophoresis on a 2% agarose gel stained with SYBR Safe DNA gel stain (Invitrogen, Thermo Fisher Scientific Inc., Carlsbad, CA, USA) and analyzed under UV transillumination. 16S rRNA sequencing The amplicons were sequenced on an ABI 7500 Genetic Analyzer (Life Technologies, Waltham, MA, USA). The sequences were edited using BIOEDIT Sequence Alignment Editor 7.0.9 (Hall, 1999 -Ibis Biosciences, Carlsbad, CA, USA) and compared to reference sequences deposited in GenBank using the BLAST tool (http://www.ncbi.nlm.nih. gov/BLAST/). Virulence characterization A pure culture of each of the isolated strains was counted in a Petroff-Hausser chamber and 0.5 mL containing 108 leptospires was inoculated intraperitoneally in thirty-day-old male hamsters (one hamster for each isolate) to determine if the isolates would produce infection [21]. The animals were purchased at Anilab Animais de Laboratório Criação e Comércio, Paulínia, SP, and were bred strictly for research purposes. The hamsters were kept in individual 30 X 20 X 12 cm polypropylene cages lined with wood shavings; free fresh water and food was daily supplied. The animals were daily monitored for signs of acute leptospiral infection (prostration, ruff hair coat, jaundice, external hemorrhage and dehydration) and were immediately euthanized after presenting two or more clinical signs. The hamster's kidneys were aseptically removed, macerated, resuspended and inoculated in liquid EMJH medium for reisolation. Ethical considerations The euthanasia procedures conducted in the animal models were in strict accordance with the recommendations in the CONCEA (National Council for Control of Animal Experimentation), and consisted of intraperitoneal administration of xylasine/ketamine and isofluran, followed by the use of a CO 2 chamber. Dogs presenting no response to treatment were euthanized by intravenous infusion of thiopental (75 mg/kg) and acepromazine (0.05 mg/kg), followed by the intravenous use of potassium chloride (dose-response curve -average of 10 ml). All euthanasia procedures were approved by the CEUAVET Committee, as stated in the Declarations section, and all efforts were made to minimize animal suffering. Statistical methods Univariate analysis was performed to describe average and standard deviation values for hematological and biochemical laboratory tests. These values were compared between leptospirosis confirmed cases (either through MAT or PCR) and non-confirmed cases using a T-test. The association of outcome with leptospirosis confirmed and non-confirmed groups was analyzed by a Chi-square test. All statistical analyses were performed in IBM SPSS Statistics 21. P-values of < 0.05 were considered statistically significant. Acute leptospirosis was diagnosed in eight dogs (24.2%) by MAT: six dogs without a history of recent vaccination had titers ≥800 in a single serum sample (dogs 3, 4, 11, 29, 31 and 32 - Table 1); additionally, prospective evaluation performed in seven dogs revealed seroconversion in two other dogs (dogs 15 and 30, see Additional file 1), which showed fourfold increase in MAT titers against Icterohaemorrhagiae serogroup (convalescent titers of 400 and 800, respectively). Overall, six dogs with MAT-confirmed leptospirosis had the highest titers against the Icterohaemorrhagiae serogroup; only one dog (dog 4) had the highest titers against the Sejroe serogroup. Leptospiral DNA was detected in 18 out of 33 animals; however, samples from four of these dogs (dogs 15, 17, 20 and 26) did not result in readable sequences after several sequencing attempts and therefore were not considered positive, leaving 14 PCR-positive dogs (42.4% - Table 1). Positive yields were obtained from urine (n = 8), blood (n = 2) or both (n = 4). The 16S rRNA phylogenetic analysis showed that the recovered sequences from all PCR-positive dogs were identified as L. interrogans (Table 1), with high similarity (> 99%) to L. interrogans representatives (AY996798, AY996800). The recovered sequences were submitted to GenBank under accession numbers KX891325 -KX891333, MG640117 and MG640121. The comparison between results from MAT-confirmed cases and PCR-positive dogs revealed that the infection could be identified exclusively by PCR in 10 dogs ( Leptospires were recovered from the urine samples of two dogs (dog 8, strain DU84; dog 24, strain DU100), which had PCR-positive results: only dog 8 had titers detectable by MAT, but serum titers were found to be below 800 (Table 1). Culturing as a single diagnostic strategy had a prevalence of 6.1% (0-14.2) and had no effect in the overall prevalence when combining all confirmatory tests (PCR, MAT and culturing) for the diagnosis of leptospirosis. The MLST analysis of both isolated strains (DU84 and DU100) revealed ST 17 (Fig. 1), which characterizes L. interrogans serogroup Icterohaemorrhagiae according to a previously described protocol [20]. Serogrouping of the DU84 strain showed a strong and specific reaction against serovars Copenhageni (6400) and Icterohaemorrhagiae (12800); the DU100 strain also had specific reactions against these serovars (Copenhageni 12,800; Icterohaemorrhagiae 51,200), revealing that both isolated strains belong to the Icterohaemorrhagiae serogroup. The inoculation of the isolates in the hamster model produced acute and intense clinical manifestations of leptospirosis. The hamsters developed acute lethal infection within six days post-inoculation, and macroscopic alterations included epistaxis, generalized petechial stains, pulmonary/liver congestion and pulmonary hemorrhage. Leptospires were successfully recovered from the kidney and liver tissues of both hamsters after euthanasia, thus indicating leptospiral

infection as the cause of disease. Discussion The use of MAT and PCR (with further DNA sequencing) as independent diagnostic strategies enabled the identification of eight (24.2%) and 14 (42.4%) dogs with leptospiral Although PCR was able to identify a higher proportion of infected dogs than MAT, the high number of dogs that follow-up care was discontinued and the high mortality/euthanasia rates found in the studied population has limited a more appropriate comparison between PCR and MAT results. It is usually recommended that the diagnosis of acute infection by MAT should be based on testing paired samples, collected 7 to 14 days apart, in order to detect at least a fourfold increase in agglutinating antibodies against Leptospira [1]. Only 7 out of 33 dogs could be prospectively evaluated, and serum samples with more than 7-day interval between acute and convalescent phases were obtained only from five dogs, including both dogs in which seroconversion was detected. Moreover, the cut off value of 800 adopted in the present study, which has been previously suggested to increase the specificity of MAT [5], and the natural delay to produce an adaptive serologic response after infection, may have led to false-negative results when testing a single serum sample by MAT. Even in face of such limitations, active leptospiral infection was determined in four dogs exclusively by MAT, while PCR was able to detect 10 infected dogs at clinical presentation that a single MAT sample evaluation could not, reinforcing that the use of multiple tests is beneficial for improving the diagnosis of canine leptospirosis in practical situations in which the clinical outcome and the possibility of follow-up care are uncertain [6]. These results are markedly different from those reported by Fraune et al. [5], who found no significant benefit in the use of PCR in relation to MAT for the early diagnosis of acute canine leptospirosis. On the other hand, our results are in agreement with the findings reported by Harkin et al. [6,11], who found a sensitivity of 100% and specificity of 88.3% when testing a leptospiral-specific PCR assay with urine samples from dogs with clinical suspicion of leptospirosis. It is important to emphasize that the application of PCR in a clinical setting, notably when using urine specimens, may provide false-positive results [22], and the association with DNA sequencing, as performed in the present study, is highly recommended to provide more reliable results for proper comparison with other diagnostic techniques. Most dogs diagnosed by PCR and DNA sequencing were identified using urine samples, and even though the detection of leptospiral DNA in urine specimens taken from dogs with multiple signs of leptospirosis is highly suggestive of acute leptospiral infection, it may not distinguish dogs with acute infection from those with chronic renal carriage of leptospires associated with other underlying diseases causing clinical manifestations similar to leptospirosis. Instead, the detection of leptospiral DNA from Fig. 1 MLST analysis of the DU84 and DU100 strains. The Maximum-likelihood tree was based on the concatenated sequences of the seven loci for the 229 available Leptospira STs blood and serum samples must be considered a more reliable strategy to diagnose acute leptospirosis in dogs by molecular methods. Our results have also shown no distinction in laboratory findings and clinical outcome between confirmed and non-confirmed cases of leptospirosis, highlighting the non-specificity of clinical and laboratory evaluations for the diagnosis of the disease. Both isolated strains could be characterized as L. interrogans serogroup Icterohaemorrhagiae and were shown to be pathogenic in a hamster model. The overall serological pattern found in the population studied indicates that the suspected dogs admitted to the veterinary hospital of the University of São Paulo, an ambulatory service that provides care for dogs from different regions of São Paulo city, are highly exposed to the Icterohaemorrhagiae serogroup. The isolation of serogroup Icterohaemorrhagiae from dogs with leptospirosis has been previously described in Brazil [21,23], and previous serological surveys have consistently shown Icterohaemorrhagiae as the most reactive serogroup in dogs suspected of leptospirosis [15,[24][25][26][27]. Interestingly, the serological profile of human subjects suspected of leptospiral infection also indicates Icterohaemorrhagiae/Copenhageni as the main causative agents attributed to human infection [28][29][30][31], and most strains recovered from human patients with leptospirosis in Brazil were characterized as Icterohaemorrhagiae [29,32]. Our findings corroborate the hypothesis that both dogs and humans are exposed to environmental contamination promoted by rodents [33], notably the brown rat (Rattus norvegicus), which was implicated as the main reservoir host of Icterohaemorrhagiae/ Copenhageni serovars in urban areas from Brazil and other locations [34,35]. Serological findings also revealed that leptospiral strains belonging to the Sejroe serogroup might be associated with acute leptospiral infection in dogs. This serogroup has been previously described as a causative agent of canine leptospirosis [36], and it was recently recovered from an asymptomatic dog in São Paulo city [12], thus indicating that Sejroe strains are actually circulating among dog populations in Brazil. In contrast, MAT titration against Canicola was only found in two dogs, which had low titers against this serogroup. Canicola strains are frequently isolated from dogs in Brazil and other locations [15,16,37], although serological evidence of Canicola infection is not often observed [38]. A previous study including dogs suspected of clinical leptospirosis conducted by our group between 2008 and 2012 found that even though most dogs reacted against serogroup Icterohaemorrhagiae, Canicola strains could still be recovered from three symptomatic dogs which had no MAT titers against this serovar [15]. All sequences from PCR-positive dogs could be identified as L. interrogans in this study, yet molecular characterization using single-gene sequencing methods has poor discriminatory power to distinguish among leptospires at a serovar/serogroup level [39], and infection caused by Canicola, although possible, could not be confirmed. Conclusions Our results suggest that the use of multiple tests is essential for a more accurate and sensitive diagnosis of acute leptospirosis in dogs in a clinical setting. The results also suggest that despite the rising incidence of new serovars affecting dogs in different locations [3], mostly attributed to the extensive use of multivalent vaccines containing few representative serovars, and the increased contact between dogs and wildlife from rural-urban interface regions [5,40,41], the Icterohaemorrhagiae serogroup is still a major causative agent of canine leptospirosis in São Paulo, Brazil. Proper characterization of leptospiral isolates remain a crucial bottleneck to access the role of particular Leptospira strains in the epidemiology of canine leptospirosis and may provide evidence-based knowledge to support the development and commercialization of multivalent vaccines containing serovars that are circulating among local populations. Availability of data and materials The datasets generated and/or analysed during the current study are not publicly available due to the confidentiality agreement with the participants (dog owners), but are available from the corresponding author on reasonable request. Cost-Effectiveness of Pneumococcal Vaccines for Adults in the United States Introduction In 2012, the Advisory Committee on Immunization Practices (ACIP) revised recommendations for adult pneumococcal vaccination to include a sequential regimen of 13-valent pneumococcal conjugate vaccine (PCV13) followed by 23-valent pneumococcal polysaccharide vaccine (PPSV23) for certain high-risk adults with immunocompromising conditions. This study, from a payer perspective, examined: (1) the cost-effectiveness of the new 2012 ACIP vaccine policy recommendation relative to the 1997 ACIP recommendation; (2) the cost-effectiveness of potential future pneumococcal vaccination policies; and (3) key assumptions that influence study results. Methods A static cohort model that incorporated costs, health outcomes, and quality-adjusted life-year (QALY) losses associated with invasive pneumococcal disease and non-bacteremic pneumococcal pneumonia (NBPP) was developed to evaluate seven pneumococcal vaccination strategies for a 50-year-old adult cohort over a 50-year period using incremental cost-effectiveness ratios (ICERs). Results For objective 1, the 2012 ACIP recommendation is the more economically efficient strategy (ICER was $25,841 per QALY gained vs. no vaccination). For objective 2, the most efficient vaccination policy would be to maintain the 2012 recommendation for PPSV23 for healthy and immunocompetent adults with comorbidities, and to modify the recommendation for adults with immunocompromising conditions by replacing PPSV23 with a sequential regimen of PCV13 and PPSV23 at age 65 (ICER was $23,416 per QALY gained vs. no vaccination). For objective 3, cost-effectiveness ratios for alternative pneumococcal vaccine policies were highly influenced by assumptions used for vaccine effectiveness against NBPP and accounting for the herd protection effects of pediatric PCV13 vaccination on adult pneumococcal disease. Conclusion Modifying the 2012 recommendation to include an additional dose of PCV13 at age 65, followed by PPSV23, for adults with immunocompromising conditions appears to be a cost-effective vaccine policy. Given the uncertainty in the available data and the absence of key influential data, comprehensive sensitivity analyses should be conducted by policy-makers when evaluating new adult pneumococcal vaccine strategies. Electronic supplementary material The online version of this article (doi:10.1007/s12325-014-0115-y) contains supplementary material, which is available to authorized users. other countries for all adults 65 years of age and older as well as younger adults with certain chronic conditions [1]. Like PPSV23, the 13-valent pneumococcal conjugate vaccine (PCV13) has been approved by the US Food and Drug Administration and selected other regulators for vaccination of adults aged 50 years and older. The basis for licensure for the adult indication for PCV13 relies on its noninferior immunogenicity for matched serotypes, when compared to adult responses to PPSV23 [2]. There is no known correlate of protection, or antibody level at which a clinical effect would be expected, for pneumococcal disease in adults, hence it is not possible to translate such immunogenicity measures into expected efficacy or effectiveness measure [1]. In February 2012, the US ACIP chose to continue its 1997 recommendation for routine adult use of PPSV23 rather than recommend routine use of PCV13, until more data become available related to clinical effectiveness of PCV13 and the magnitude of indirect reduction in adult disease due to routine vaccination of children with PCV13 [2,3]. In June 2012, the ACIP voted to recommend a sequential regimen of PCV13 and PPSV23 for certain high-risk adults with immunocompromising conditions, functional or anatomic asplenia, cerebrospinal leaks, or cochlear implants. These recommendations were published in October 2012 [1]. (Fig. 1). Annual age-specific transition probabilities among risk groups were derived from Weycker et al. [12] (see Table S9 in Appendix 1, Distribution of Risk Groups). Vaccination strategies were assumed to reduce the baseline risk of pneumococcal disease. Vaccination Strategies Given the need to assess two vaccines, three risk categories, and multiple vaccination age points in this economic analysis, many options were considered for defining a reasonable set of vaccination strategies that are both clinically relevant and practically implementable. For modeling purposes, this analysis is restricted to vaccination strategies for adults 50 years of age and older, for whom the 1997 ACIP Disease Estimates Estimates of annual rates of each disease outcome were based on data presented by Weycker et al. [12] (see Table S2 in Appendix 1). The rates of IPD, death and disability from IPD, hospitalized NBPP, death and disability from NBPP, and outpatient visits associated with NBPP were calculated, varying by age and risk of pneumococcal status (see Appendix 1 for description). Baseline rates of disease for the novaccination strategy were estimated using methods described previously [13]. Proportions of the population that fall into each risk category were based on published literature (see Table S9 in Appendix 1, Distribution of Risk Groups) [12] and individuals were assumed to move from lower risk categories to higher risk categories probabilistically as they aged in the model. Cost Estimates Medical-care costs related to hospitalizations, outpatient visits, and deaths were included in Table S2 in Appendix 1) [12]. Vaccine costs were based on Centers for Disease Control and Prevention reports of private-sector prices as of September 10, 2012 [14]. PPSV23 was assumed to be $57.70 per dose and PCV13 to be $120.95 per dose. Administration cost was assumed to be $15.00 per dose for each product. Indirect costs were not included in this analysis. Costs are expressed in 2012 US dollars using medical Consumer Price Index adjustor [15]. Utility Weights Utility weights associated with IPD and NBPP inpatient cases, as well as baseline utilities for average-and high-risk populations were based on values described by Smith et al. [9] (see Table S9 in Appendix 1). Utility weights associated with outpatient visits for NBPP were based on estimates related to influenza outpatient visits [10]. Duration of an event is incorporated in QALY calculation. Vaccine Uptake Vaccine uptake was assumed to

vary depending on age and risk category and was based on contemporary US data regarding PPSV23 (see Table S9 in Appendix 1). Vaccine uptake for both PCV13 and PPSV23 was assumed to be the same for each age/risk category described. For routine age-based recommendations (at age 50 years or age 65 years), vaccine uptake was assumed to be 12.0% for healthy, 33.9% for immunocompetent with comorbidities and immunocompromised populations at age 50 years, and 60.1% for all risk categories at age 65 years. Vaccine Effectiveness Due to the absence of data comparing PCV13 to PPSV23 efficacy in adults, assumptions around PCV13 effectiveness against IPD and NBPP as [17]. Sensitivity Analyses Given the uncertainty around several key model parameters, sensitivity analyses are critical to understanding the potential economic implications of the various vaccination strategies explored. Sensitivity analyses were conducted to examine the impact of varying assumptions related to vaccine effectiveness, impact of herd protection on disease incidence and vaccine serotype coverage. In addition, a set of scenario analyses was conducted to examine the potential cost-effectiveness of broad agebased vaccine recommendations starting at age 50. Projected Impact of Herd Protection To be conservative, the impact of herd protection was excluded in the base case. To assess model sensitivity to the impact of herd protection on disease incidence and serotype coverage, additional scenarios were evaluated, anticipating some future point when full reduction in adult disease due indirectly to pediatric PCV13 has been realized. Data related to reductions in IPD observed after routine pediatric use of 7-valent pneumococcal conjugate vaccine (PCV7) were applied to the modeled base-case rates of disease to generate projected disease-incidence rates after routine pediatric use of PCV13 [17]. In that future scenario, IPD and NBPP disease reductions post-PCV13 were assumed to be comparable to those seen for IPD reductions post-PCV7. There are sufficient surveillance data to indicate that a significant herd impact post-PCV7 introduction has already occurred in the US, however, there is no consensus in the published literature regarding the potential herd impact of use of PCV7 on NBPP in adults [18,19]. The projected herd-impact parameters varied based on certain serotype characteristics. Disease incidence associated with serotypes contained in PCV7 as well as serotypes 1 and 5 were assumed to remain at current low levels. Serotypes 1 and 5 are believed to behave differently than other serotypes in terms of carriage and transmission, so it is not clear how adult disease related to these two serotypes would be affected by pediatric use of PCV13. The incidence of disease associated with the four other new serotypes in PCV13 was assumed to decrease in a pattern similar to the observed reductions in PCV7 types after routine pediatric vaccination began. All other serotypes (PPSV23 unique types and non-vaccine types) were assumed to increase slightly, similar to increases observed post-PCV7 implementation. Serotype coverage for each vaccine was assumed to vary and three sets of assumptions were applied to three separate sensitivity analyses: 1. Full herd impact: PPSV23 was assumed to cover 68% of disease in adults aged 50-64 years and PCV13 20%; PPSV23 was assumed to cover 56% of disease in adults older than 65 years, and PCV13 18%; 2. Herd impact was assumed to be 50% of that projected for scenario 1; 3. Herd impact for IPD described as in scenario 1, but no herd impact assumed for NBPP. In this scenario NBPP rates of disease and serotype distributions were equivalent to base-case assumptions. Waning of Effectiveness Recent studies demonstrate that PPSV23 induces antibody responses that remain above the levels of unvaccinated adults for 5-10 years or more [20]. Projected Pneumococcal Disease Events Total projected IPD cases (survived), NBPP hospitalizations (survived), disabled cases (survived), outpatient visits, and deaths for the no-vaccination strategy, along with the projected events averted for the seven vaccination strategies were generated ( Table 4). The model projects that for a single cohort of 100,000 adults starting at age 50 years, in the absence of pneumococcal vaccination, there Sensitivity Analysis Results Sensitivity analyses related to impact of disease incidence and serotype changes and due to herd protection along with changes to vaccine effectiveness assumptions are reported in Table 6. Only non-dominated strategies are shown. Dominated strategies could have been eliminated from consideration due to strong dominance (i.e., a less expensive and more beneficial strategy available), or weak dominance (i.e., a more economically efficient strategy available). Incremental cost-effectiveness ratios are compared as follows: rank 1 vs. no vaccination, rank 2 vs. rank 1, and rank 3 vs. rank 2. Several sensitivity analyses had more than three non-dominated strategies, but only the top three strategies are shown. Full herd impact projections based on changes in adult IPD observed post-introduction of 7-valent pneumococcal conjugate vaccine [17] ICER, incremental cost-effectiveness ratio; IPD, invasive pneumococcal disease; NBPP, non-bacteremic pneumococcal pneumonia; PCV13, 13-valent pneumococcal conjugate vaccine; PPSV23, 23-valent pneumococcal polysaccharide vaccine; US$, United States dollars Herd Protection The potential impact on cost-effectiveness results due to projected changes in disease incidence for both IPD and NBPP along with serotype coverage for each vaccine was examined using three sensitivity analysis scenarios. For the first scenario, simulating the impact of full herd protection, the three most economically efficient strategies were S1 ($51,000 vs. no vaccination), followed by S2 ($77,000 vs. S1) and S5 ($103,000 vs. S2). If 50% less herd impact were to occur, S5 would be the most economically efficient ($36,000 vs. no vaccination), followed by S7. When it was assumed that herd protection did not impact incidence of NBPP, S1 was found to be the most economically efficient strategy followed by S5 and S3 (Table 6). no vaccination). Waning of Effectiveness Vaccine-waning patterns were found to impact cost-effectiveness results in the sensitivity analyses. When the two vaccines were assumed to have equivalent waning periods of 10 years, the most economically efficient strategies were S1 followed by S2 and S5, and all had ICERs below $50,000 per QALY. DISCUSSION The results of this analysis are intended to Flavonoids from Agrimonia pilosa Ledeb: Free Radical Scavenging and DNA Oxidative Damage Protection Activities and Analysis of Bioactivity-Structure Relationship Based on Molecular and Electronic Structures To clarify the substantial basis of the excellent antioxidant capacity of Agrimonia pilosa Ledeb. Fourteen flavonoids were isolated and identified from Agrimonia pilosa Ledeb, seven of which have notable DPPH radical scavenging activities, i.e., catechin, luteolin, quercetin, quercitrin, hyperoside, rutin, luteolin-7-O-β-glucoside with IC50 values of 5.06, 7.29, 4.36, 7.12, 6.34, 6.36 and 8.12 µM, respectively. The DNA nicking assay showed that five flavonoids from Agrimonia pilosa Ledeb—taxifolin, catechin, hyperoside, quercitrin and rutin—have good protective activity against DNA oxidative damage. Further, we analyzed the bioactivity-structure relationship of these 14 flavonoids by applying quantum theory. According to their O-H bond dissociation enthalpy (BDE), C ring’s spin density and stable molecular structure, the relationship between their structures and radical scavenging capacities was evaluated and clarified. We found that among flavonoid aglycones from Agrimonia pilosa Ledeb, the O-H BDE of quercetin is lowest with the values of 69.02 and the O-H BDE of apigenin is highest with the values of 79.77. It is interesting that the O-H BDE value of isovitexin (78.55) with glycoside at C-6 position is lower than that of its aglycone (79.77) and vitexin (99.20) with glycoside at C-8 position. Further analysis indicated that the glycosidation of flavonoids at C-6 in the A-ring makes a more uniform distribution of spin density and improves the stability of free radicals leading to the increase in antioxidant capacity. Flavonoids with good antioxidant capacity might contribute to the pharmacological effects of Agrimonia pilosa Ledeb. Introduction The normal metabolism of all oxygen and exogenous factors could generate reactive oxygen species (ROS), such as superoxide anion radicals (O 2 − ·), hydroxyl radical species (·OH), singlet oxygen ( 1 O 2 ), and hydrogen peroxide (H 2 O 2 ) [1,2]. ROS are a class of highly reactive molecules that can attack a wide range of molecules found in living cells, such as protein, lipids, and nucleic acids, leading to tissue oxidative damages [3,4]. This oxidative damage has been confirmed to be involved in occurrence and development of several chronic human diseases, namely cardiovascular diseases, neurodegenerative diseases, rheumatism, diabetes mellitus and cancer [5][6][7][8]. Based on the lack of effective therapies for most chronic diseases and the usefulness of foods rich in antioxidants in the prevention of the mentioned diseases, there is a growing interest in searching for natural antioxidants from vegetables, fruits, tea, spice, and medical herbs [9][10][11]. Agrimonia pilosa Ledeb, a traditional Chinese medicine and a wild vegetable for tonic function, belongs to Rosaceae. As the most common Agrimonia spp. in China, it is used to treat tumor, and blood, gastrointestinal, genitourinary, and gynecological diseases in Chinese traditional medicine [12,13]. Our previous study found that the aqueous extract from Agrimonia pilosa Ledeb, especially its AcOEt-soluble fraction and n-BuOH soluble fraction, was rich in phenolic compounds and exhibited high antioxidant activities [14]. In the present paper, we isolated several flavonoids from the Ethyl acetate-soluble fraction (ESF) and evaluated their radical scavenging activities by using DPPH scavenging assay and protective activities on DNA oxidative damage by using DNA nicking assay to elicit the active principles present in ESF from Agrimonia pilosa Ledeb. Further, we studied the bioactivity-structure relationship by applying quantum chemistry theory. To date, this is the first report on free radical-scavenging effects of pure compounds obtained from the title plant. Especially, we reported firstly that glycosylation at C-6 could enhance antioxidant activity compared to the corresponding aglycones, which would offer theoretical guidance for the design and the development of antioxidants. DPPH Radical Scavenging Activity of Flavonoids from Agrimonia pilosa Ledeb To clarify which compounds isolated are responsible for the excellent antioxidant activity of Agrimonia pilosa Ledeb, the antioxidant ability of each compound was evaluated by DPPH radical scavenging activity assay, and compared with the well-known reference antioxidants, vitamin C and BHT (butylated hydroxytoluene). As shown in Table 1, the active compounds 1, 5-9 and 12, which all possessed the catechol group in ring B, exhibit potent anti-oxidant activities against DPPH with IC 50 values of 5.06, 7.29, 4.36, 7.12, 6.34, 6.36 and 8.12 µM, respectively, whereas the positive control, vitamin C and BHT showed DPPH scavenging activities with IC 50 values of 14.62 and 17.67 µM. The DPPH radical scavenging activity of kaempferol (IC 50 = 16.09 µM) with the 3-hydroxyl group in the C-ring is equivalent to BHT. However, the compounds 4, 10, 11, 13 and 14 which all have not the catechol group in the B-ring and the 3-hydroxyl group in the C-ring show weak radical scavenging activity with IC 50 values larger than 100 µM. Additionally, it is implied that the antioxidant activity of flavonoid glycoside is correlate with the position of the flavonoids glycosylation. Glycosylation sharply brings down the antioxidant activity if -OH is the active group, e.g., kaempferol-3-O-glucoside. On the other hand, when -OH is not the active group, glycosylation hardly affects the antioxidant activity, e.g., luteolin-7-O-β-glucoside. The same phenomenon was found in trans-resveratrol and trans-piceid [21]. In DPPH radical scavenging activity assay, the antioxidants reduce the very stable DPPH radical to a yellow-colored compound, diphenylpicrylhydrazine, and the extent of the reaction depends on the hydrogen-donating capacity of the antioxidants. An antioxidant candidate that proves promising in the DPPH antioxidant assay would provide an optimistic scaffold for prospective in vivo studies [22]. Thus, this assay has been extensively used for screening antioxidants from fruit and vegetable juices or natural extracts [23][24][25]. Additionally, it is implied that the antioxidant activity of flavonoid glycoside is correlate with the position of the flavonoids glycosylation. Glycosylation sharply brings down the antioxidant activity if -OH is the active group, e.g., kaempferol-3-O-glucoside. On the other hand, when -OH is not the active group, glycosylation hardly affects the antioxidant activity, e.g., luteolin-7-O-β-glucoside. The same phenomenon was found in trans-resveratrol and trans-piceid [21]. In DPPH radical scavenging activity assay, the antioxidants reduce the very stable DPPH radical to a yellow-colored compound, diphenylpicrylhydrazine, and the extent of the reaction depends on the hydrogen-donating capacity of the antioxidants. An antioxidant candidate that proves promising in the DPPH antioxidant assay would provide an optimistic scaffold for prospective in vivo studies [22]. Thus, this assay has been extensively used for screening antioxidants from fruit and vegetable juices or natural extracts [23][24][25]. Protective Effect against DNA Oxidative

Damage Supercoiled plasmid pBR322 DNA, a closed double-stranded DNA molecule, has been widely used to screen the active substance protecting DNA from oxidative damage [26,27]. Plasmid pBR322 DNA exists three forms: complete pBR322 DNA, usually in supercoiled conformation (supercoiled conformation (SC)), open-loop notch type (nicked circular form (NC)), in which an incision occurs on a chain, and line form (linear form (LC)), in which the two chains break at the same position. Three forms of pBR322 DNA have completely different mobility in the gel electrophoresis: supercoiled DNA (SC) is the most tightly structured and migrates fastest, open-loop notch-type DNA (NC) migrates slowest due to its loose structure, and linear DNA (LC) is somewhere in between. Therefore, the protective effects of samples on DNA oxidative damage could be studied through observing migration position and the quantity of these three forms DNA on gel. The experimental results were shown in Figure 2. As shown in Figure 2, SC DNA was found as the main form making up 88.2% of pBR322 DNA in the control group without H 2 O 2 , while NC DNA (87.0%) was the main form existing in the model group in which ·OH was generated to fragment DNA by using UV irradiation on H 2 O 2 . Flavonoids from Agrimonia pilosa Ledeb all have protective activity on the oxidative fragment of DNA with a concentration-dependent manner. Among flavonoid aglycones from Agrimonia pilosa Ledeb, taxifolin exhibited the strongest efficiency, followed by catechin, luteolin, apigenin, quercetin, and kaempferol. At the presence of 1 mM taxifolin, SC percentage remains 48.8%. Three glycosides of quercetin, namely hyperoside, quercitrin and rutin, have higher protective effects than their flavonoid aglycone against the oxidative fragment of DNA with SC percentage of 49.5%, 46.7% and 38.5% at 1 mM, respectively. The protective effect of tiliroside on DNA damage is higher than that of kaempferol, while the protective effect of kaempferol-3-O-glu is similar to that of kaempferol. The protective effect of luteolin-7-O-glucoside is slightly lower than that of luteolin. Vitexin and isovitexin, as the C-glycosides of apigenin, have a higher protective effect on oxidative damage of DNA than their flavonoid aglycone. Analysis of Bioactivity-Structure Relationship Based on the Electronic Structure Using quantum chemistry calculation methods, the O-H bond dissociation enthalpy (BDE) and free radical spin density of the 14 flavonoids from Agrimonia pilosa Ledeb were studied. The relationship between the free radical scavenging activity of flavonoids and their electronic structure was illustrated from the perspective of quantum chemistry theory. Analysis of Bioactivity-Structure Relationship Based on the Electronic Structure Using quantum chemistry calculation methods, the O-H bond dissociation enthalpy (BDE) and free radical spin density of the 14 flavonoids from Agrimonia pilosa Ledeb were studied. The relationship between the free radical scavenging activity of flavonoids and their electronic structure was illustrated from the perspective of quantum chemistry theory. The O-H BDEs of flavonoids from Agrimonia pilosa Ledeb were calculated as listed in Table 2. In non-polar solvents, the antioxidants scavenge free radicals primarily by hydrogen abstraction reaction. The lower the BED of O-H is, the easier the hydrogen abstraction of the compound is, and the stronger the activity of the antioxidant is [28]. For the compound with several phenolic hydroxyl groups, its free radical scavenging activity mainly depends on the phenolic hydroxyl group with lowest BDE. From Table 2, we found that among flavonoid aglycones from Agrimonia pilosa Ledeb, the O-H BDE of quercetin is lowest and the O-H BDE of apigenin is highest. According to the values of BDEs, the order of free radical scavenging activities is quercetin > catechin > luteolin > taxifolin > kaempferol > apigenin. Among eight flavone glycosides, the O-H BDEs of quercitrin, hyperoside, rutin and luteolin-7-O-glu are lower than 73.30, which showed that they have higher free radical scavenging activities than isovitexin, vitexin, kaempferol-3-O-glu and tiliroside with the O-H BDEs higher than 76.32. It is interesting that the O-H BDE of isovitexin with glycoside at C-6 position is lower than that of its aglycone (apigenin) and vitexin with glycoside at C-8 position, indicating that glycoside at C-6 position is helpful for free radical scavenging activity. The evaluation results of free radical scavenging activities of flavonoids from Agrimonia pilosa Ledeb according to the O-H BDEs were consistent with those of the DPPH scavenging assay (see Table 1). Besides O-H BDE, the uniformity of the spin density of radicals generated is another factor relating to the antioxidant activity of compounds [29,30]. If the unpaired electrons of the radical are highly delocalized through conjugated system, the radical energy will significantly reduce and the radical becomes stable. Semi-quinone radical has strong antioxidant activity, in which uniform spin density distribution is easy to generate. The spin density sketch maps of six flavonoid aglycones from Agrimonia pilosa Ledeb were shown in Figure 3. The C-ring's spin density of compounds was listed in Tables 3-6. The spin density of 4 -O is closely related to antioxidant activity. We found that 4 -Os of luteolin, quercetin, quercitrin and rutin have low spin density, indicating these compounds have good radical scavenging activities, while 4 -Os of kaempferol-3-O-glu, tiliroside, apigenin, apigenin-6-glu, apigenin-8-glu have high spin density, implying that these compounds possess weak radical scavenging activities. The O-H BDEs of flavonoids from Agrimonia pilosa Ledeb were calculated as listed in Table 2. In non-polar solvents, the antioxidants scavenge free radicals primarily by hydrogen abstraction reaction. The lower the BED of O-H is, the easier the hydrogen abstraction of the compound is, and the stronger the activity of the antioxidant is [28]. For the compound with several phenolic hydroxyl groups, its free radical scavenging activity mainly depends on the phenolic hydroxyl group with lowest BDE. From Table 2, we found that among flavonoid aglycones from Agrimonia pilosa Ledeb, the O-H BDE of quercetin is lowest and the O-H BDE of apigenin is highest. According to the values of BDEs, the order of free radical scavenging activities is quercetin > catechin > luteolin > taxifolin > kaempferol > apigenin. Among eight flavone glycosides, the O-H BDEs of quercitrin, hyperoside, rutin and luteolin-7-O-glu are lower than 73.30, which showed that they have higher free radical scavenging activities than isovitexin, vitexin, kaempferol-3-O-glu and tiliroside with the O-H BDEs higher than 76.32. It is interesting that the O-H BDE of isovitexin with glycoside at C-6 position is lower than that of its aglycone (apigenin) and vitexin with glycoside at C-8 position, indicating that glycoside at C-6 position is helpful for free radical scavenging activity. The evaluation results of free radical scavenging activities of flavonoids from Agrimonia pilosa Ledeb according to the O-H BDEs were consistent with those of the DPPH scavenging assay (see Table 1). Besides O-H BDE, the uniformity of the spin density of radicals generated is another factor relating to the antioxidant activity of compounds [29,30]. If the unpaired electrons of the radical are highly delocalized through conjugated system, the radical energy will significantly reduce and the radical becomes stable. Semi-quinone radical has strong antioxidant activity, in which uniform spin density distribution is easy to generate. The spin density sketch maps of six flavonoid aglycones from Agrimonia pilosa Ledeb were shown in Figure 3. The C-ring's spin density of compounds was listed in Tables 3-6. The spin density of 4′-O is closely related to antioxidant activity. We found that 4′-Os of luteolin, quercetin, quercitrin and rutin have low spin density, indicating these compounds have good radical scavenging activities, while 4′-Os of kaempferol-3-O-glu, tiliroside, apigenin, apigenin-6-glu, apigenin-8-glu have high spin density, implying that these compounds possess weak radical scavenging activities. According to the BDE values, the C-ring's spin density and the stable molecular structure, the following conclusions could be deduced: (1) The presence of 3-OH, the ortho-hydroxy in the B-ring, 2,3-double bond and 4-carbonyl contribute to a more balanced distribution of spin density, reducing the BDE of O-H in the B-ring and increasing free radical scavenging activity of compounds; (2) Glycosylation of 3-OH in flavonol increases the dihedral angle between the B-ring and C-ring, resulting in an increase in spin density imbalance and the BDE of O-H in the B-ring, and thus reduces the free radical scavenging activity of compounds; (3) Glycosidation of flavonoids at C-6 in the A-ring makes a more uniform distribution of spin density and improves the stability of free radicals; whereas glycosidation at C-8 in the A-ring increases the dihedral angle between C-ring and B-ring, resulting in reduction of the delocalization degree and the stability of free radicals. The results of theoretical calculations are consistent with those of experimental studies. Plant Material The dried aerial parts of Agrimonia pilosa Ledeb were purchased from Western Medicine City in Chongqing, China. Structure Elucidation of the Isolated Compounds Examination of TLC plates under UV light and after visualization by spraying reagents revealed that these compounds possess a flavonoid skeleton. Molecular weights of isolated flavonoids were analyzed by using 5050E MS (VG, UK). Structure of isolated flavonoids were elucidated by using 1 H-NMR (500 MHz; with TMS as international standard) and 13 C-NMR (100.2 MHz) spectroscopy in AVANCE DRX-500 NMR Spectrometers (Bruker BioSpin, Fallanden, Switzerland). The NMR spectra of compounds were taken in DMSO. Then the compounds were identified by direct comparisons of their molecular weight and spectral data ( 1 H-NMR and 13 C-NMR) with those of literature data. DPPH Radical Scavenging Assay The ability of the flavonoids to scavenge DPPH radicals was determined according to the method of Blois [31] with some modifications. Briefly, a 0.1 mL sample was mixed with 1. The IC 50 (concentration for 50% DPPH radical-scavenging activity) values were determined by SPSS software. Ascorbic acid and butylated hydroxytoluene (BHT) were used as positive controls. DNA Nicking Assay DNA nicking assay was used to test the damage effect of ·OH on DNA and the protective ability of flavonoids from Agrimonia pilosa Ledeb on DNA oxidative damage. DNA damage was measured by the conversion of supercoiled pBR322 plasmid DNA to nicked circular DNA and linear DNA forms. Tris-HCl buffer (50 mmol/L pH 7.2), the sample solution, pBR322 DNA (200 ng/reaction), and H 2 O 2 were added into a 0.5 mL EP tube, diluted to a total volume of 10 µL with distilled water, and mixed. After the mixture was irradiated for 10 min by UV lamp, 2 µL of loading buffer was added to terminate the reaction. Then the mixture was subjected to electrophoresis in a horizontal slab gel apparatus and 1 × TAE buffer, which was performed at 75 V for 1.5 h. The gel was stained with a solution of 0.5 µg/mL ethidium bromide for 30 min, followed by destaining in water. Agarose gel electrophoresis of plasmid DNA was visualized by photographing the fluorescence of intercalated ethidium bromide under a UV illuminator. After electrophoresis, the proportion of DNA in each form was estimated quantitatively from the intensity of the bands using Glyko BandScan software. The control group did not contain H 2 O 2 and the compound tested, while the model group contained H 2 O 2 but not the compound tested. Quantum Theory Study of Bioactivity-Structure Relationship The O-H bond dissociation enthalpy (BDE) and free radical spin density were adapted to study and analyze the bioactivity-structure relationship of flavonoids from Agrimonia pilosa Ledeb. Using a combination method of density functional theory proposed by Wright et al. [32], the O-H bond dissociation enthalpy (BDE) was calculated. First, the optimization molecular structure of flavonoids was obtained by AMl approach. Then, the frequency of molecular vibration was calculated at B3LYP/6-31(d) level in order to check whether a stable configuration of each molecular was obtained. Meanwhile, zero point vibrational energy (ZPVE) was calculated and corrected by using the correction factor of 0.9806. Finally, single point electronic energy (SPE) was calculated at the B3LYP/6-31G (d) level. The total energy of molecules includes single-point energy and zero-point vibrational energy corrected. For flavonoid containing glycoside, layered Oniom (B3LYP/6-31d: UHH) was used for calculation. The following formula was used to give the value of BDE. Statistical Analysis Data were collected and expressed as the mean ± standard deviation of three independent experiments and analyzed for statistical significance from control, using the Dunnett test (SPSS 11.5 Statistics Software; SPSS, Chicago, IL, USA). The criterion for significance was set at p < 0.05. IC 50 values, from the in vitro data, were calculated by regression analysis.

Conclusions We isolated and identified 14 flavonoids from Agrimonia pilosa Ledeb. Catechin, taxifolin, luteolin, quercetin, quercitrin, hyperoside, rutin and luteolin-7-O-β-glucoside have notable radical scavenging activities and are responsible for the excellent antioxidant activity of Agrimonia pilosa Ledeb according to DPPH radical scavenging assay and quantum theory study. It is worth noting that glycosidation of flavonoids at C-6 in the A-ring makes a more uniform distribution of spin density and improves the stability of free radicals, leading to the increase of antioxidant capacity. In addition, catechin, taxifolin, quercitrin, hyperoside and rutin exhibit a good protective effect on DNA oxidative damage. It is implied that these flavonoids with good antioxidant capacity might contribute to some of the pharmacological and nutrient effects of Agrimonia pilosa Ledeb. Some researches should be launched to develop function foods or drugs for diseases relating to oxidative damage. A Case of Neurocysticercosis in Entire Spinal Level Cysticercosis is the most common parasitic infection affecting the central nervous system. Spinal neurocysticercosis (NCC) is very rare compared with intracranial NCC and requires more aggressive management because these lesions are poorly tolerated. The authors report a case of intradural extramedullary cysticercosis of the entire level of spine with review of the literature. INTRODUCTION Neurocysticercosis (NCC), one of the most prevalent forms of parasitosis of the central nervous system (CNS), is caused by the larvae of the tapeworm Taenia solium in the nervous system. It typically involves the brain parenchyma, intracranial subarachnoid space or ventricular system, mainly in endemic regions. In these areas, the incidence of NCC accounts for up to 4% of the general population 1) . The parasitic involvement of the spinal canal is a relatively rare event even in endemic areas with an estimated incidence of 1.5 to 3% among the total NCC 8) . The authors report a 42-year-old woman who had complaints of progressive bilateral leg motor weakness, finally diagnosed with spinal NCC along the entire level of spine (C2-L2). The report includes comments on the surgical experience and a review of the literature in terms of treatment modality. CASE REPORT A 42-year-old female with a history of hydrocephalus visited the institute complaining of unsteady gait, progressive lower-extremity weakness over several months. Muscle strength was Grade III/VI in lower extremities bilaterally with enhanced deep tendon reflexes. She had mental retar-dation and nearly total deafness for a long time, probably from childhood. Non-enhanced computed tomography (CT) of the head performed one year before had revealed hydrocephalus that was not investigated further as the headache and vomiting had subsided. Magnetic resonance imaging (MRI) of the brain was normal. MRI of the spine revealed multiple intradural, extramedullary cystic structures displacing the spinal cord along the C2-L2 levels. The mass lesion was distributed throughout the entire level of the spine. (Fig. 1). The NCC was considered on neuroimaging and serological studies preoperatively. We operated the level of T3-T5 and L1-L3 where the spinal cord was severely compressed. The operation included laminectomy and intradural exploration, during which we found thickened wall of the sac adhered to the spinal cord, longitudinally (Fig. 2). Complete resection of the lesion was impossible because of its huge size. We performed a limited resection of the thickened longitudinal sac. Sharp exision of the sac was required for contained cyst removal, ultimately allowing free flow of multiple translucent cysts from the excised sac. On gross examination, the each cyst was a thin-walled and contained whitish fluid which is consistent with neurocysticercosis without scolices. The final diagnosis was made as a cysticercosis after histopathological examination (Fig. 3). The patient was given a course of albendazole with steroids for 8 weeks after surgery. Her neurological status gradually improved over several months. After 6-months he was able to walk independently. Repeated serial MR imaging revealed successful decompression of the spinal cord, and the patient experienced no recurrence of symptoms during the 12-month follow-up period (Fig. 4). The enzyme-linked immunosorbent assay (ELISA) test showed no evidence of infection during followup period. DISCUSSION Despite the high incidence of NCC in endemic areas, spinal NCC is very rare compared with intracranial NCC. There have been fewer than 200 cases of spinal NCC reported in the world literature 2,11) . The low incidence of spinal involvement is believed to be related to the fact that the blood flow to the brain is approximately 100-fold greater than to the spine 7,9) . Intradural spinal cysticercosis can be subdivided in leptomeningeal (subarachnoid) or intramedullary forms (parenchymal), and the former is more prevalent than the later. Leptomeningeal form occur in approximately 80% of cases which is assumed to result from larval migration through the ventricular system into the spinal subarachnoid space, where their movement could be assisted by gravity 3) . Queiroz et al. 3,9) reported that spinal distribution of cysticerci occured as follows : 34% in the cervical; 44.5% in the thoracic; 15.5% in the lumbar; and 6% in the sacral region. The disease may have a pleomorphic quality that is related to individual differences in the number and location of the lesions within the CNS 4,12) . In an older study, the authors reported that 30% also had intracerebral involvement in patients afflicted with spinal cysticercosis, whereas in a more recent study the authors noted 100% concurrent intracere- bral involvement 10) . Some authors have reported that spinal involvement typically appears a few months or years after the development of hydrocephalus and/or episodes of cysticercosis-related meningitis 5) . Hydrocephalus may be explained by the presence of cysticerci within the ventricular system, arachnoiditis causing obstruction of the ventricles outlets (isolated ventricle), or subarachnoid natural CSF pathways. Spinal NCC-related symptoms are myelopathy and progressive weakness, induced by spinal cord and/or cauda equina compression. This case study has pleomorphic clinical entity in that the patient exhibited progressive spastic paraparesis due to spinal involvement with hydrocephalus and signs of dementia for more than 10 years. In this case, laboratory test including the ELISA showed no evidence of ongoing intracranial infection and neuroimaging showed arrested hydrocephalus, assumed to be from a previous infection. Magnetic resonance imaging is the diagnostic study of choice for evaluating spinal NCC because it provides noninvasive multiplanar images of any potential intraspinal pathological entity. Immunodiagnostic tests of serum samples have been widely used to confirm the diagnosis of neurocysticercosis. The ELISA performed to test CSF samples is more accurate than that of serum; its sensitivity is 87% and its specificity is 95% 1) . The treatment of spinal NCC includes a combination of specific cysticidal drugs, noncysticidal therapy and surgery, but there is a lack of definite guidelines due to the rarity of spinal involvement. The efficacy of medical therapy with praziquantel or albendazole is well described in a number of cases by various authors with the resolution of neurological deficits 12) . Albendazole may be more effective than praziquantel 6) . Surgical treatment is indicated in cases of spinal NCC when patients experience severe and progressive neurological dysfunction regardless of whether medical therapy has been attempted. Excision of extramedullary lesions is often difficult because of the arachnoidal scarring secondary to cyst degeneration, but sharp dissection, gentle irrigation, and Valsalva maneuvers may assist in extirpating adherent cysts 8) . All cases of spinal neurocysticercosis, surgery does not provide an decisive cure without morbidity because of the various mechanisms responsible for neurological symptoms in spinal neurocysticercosis. However, in patients with acute onset of symptoms and in those where the diagnosis is doubtful, surgical excision is necessary to eliminate the compressive element. This is both to confirm the diagnosis and to provide maximal chances of recovery before any irreversible cord changes occurs. In this case, due to the widespread nature of the cystic lesion, it was inevitable to do a limited excision rather than total resection. However, we did get an appropriate decompression of spinal cord without significant complications. CONCLUSION Spinal neurocysticercosis should be considered in the differential diagnosis in high-risk populations with symptoms suggestive of a spinal mass lesion. Current therapy for leptomeningeal spinal cysticercosis is laminectomy and resection followed by medical management. In the case that total resection is not feasible due to entire spinal lesion such as this case, the limited resection with cyst removal and chemotherapy for the treatment of spinal neurocysticercosis may be an option. Risk factors of infective endocarditis in persons who inject drugs Background The rising incidence of infective endocarditis (IE) among people who inject drugs (PWID) has been a major concern across North America. The coincident rise in IE and change of drug preference to hydromorphone controlled-release (CR) among our PWID population in London, Ontario intrigued us to study the details of injection practices leading to IE, which have not been well characterized in literature. Methods A case–control study, using one-on-one interviews to understand risk factors and injection practices associated with IE among PWID was conducted. Eligible participants included those who had injected drugs within the last 3 months, were > 18 years old and either never had or were currently admitted for an IE episode. Cases were recruited from the tertiary care centers and controls without IE were recruited from outpatient clinics and addiction clinics in London, Ontario. Results Thirty three cases (PWID IE+) and 102 controls (PWID but IE-) were interviewed. Multivariable logistic regressions showed that the odds of having IE were 4.65 times higher among females (95% CI 1.85, 12.28; p = 0.001) and 5.76 times higher among PWID who did not use clean injection equipment from the provincial distribution networks (95% CI 2.37, 14.91; p < 0.001). Injecting into multiple sites and heating hydromorphone-CR prior to injection were not found to be significantly associated with IE. Hydromorphone-CR was the most commonly injected drug in both groups (90.9% cases; 81.4% controls; p = 0.197). Discussion Our study highlights the importance of distributing clean injection materials for IE prevention. Furthermore, our study showcases that females are at higher risk of IE, which is contrary to the reported literature. Gender differences in injection techniques, which may place women at higher risk of IE, require further study. We suspect that the very high prevalence of hydromorphone-CR use made our sample size too small to identify a significant association between its use and IE, which has been established in the literature. In June of 2016, a public health emergency was declared in London, ON, Canada following an increase in IE [13] and injection-related HIV [14]. We identified an association between increasing IVDU-associated IE hospitalization and increasing prescriptions of hydromorphone across Ontario [4]. Prescription rates for hydromorphone controlledrelease (CR; Hydromorph Contin ®, Purdue Pharma, Pickering, Ontario) in London are in the top quartile in Ontario [4,13,15]. Furthermore, studies have shown that the misuse of prescribed opiates such as hydromorphone-CR amplifies the risk of infections due to the nature of how these substances are prepared for injection [11,[16][17][18]. Hydromorphone-CR is difficult to dissolve in solution, often requiring PWID to use unsterile methods to prepare injectates. The components of hydromorphone-CR capsules that provide the controlled release, increase survival of HIV [19] and Staphylococcus aureus [12], which cause the vast majority of IE cases [7]. Moreover, residual drug remains in cookers and filters after an initial injection, allowing PWID to resolubilize the remaining drug and conduct multiple injections (Fig. 1) [11,14,16,19,20]. This highrisk practice of multiple injections involves keeping, sharing, and reusing injection drug preparation equipment (IDPE), which increases infections among PWID [14,16,20]. However, studies have also found that heating the injectate before injection of hydromorphone-CR can significantly reduce the inoculum of bacteria causing IE [12]. With the coincident rise in IE and change of drug preference to hydromorphone-CR among PWID in London, Ontario, this study aimed to identify demographic variables and injection practices that pose a risk for IE. We hypothesized that using hydromorphone-CR would increase the risk of IE. Secondly, we hypothesized that being a male, injecting into multiple sites, using non-sterile equipment, and failing to heat hydromorphone-CR injectates would be further risk factors for IE. As a secondary goal, we also sought to explore demographic variables and injection practices of PWID in London to generate hypotheses for further studies. Design/setting/participants We conducted a case-control study where persons who inject drugs (PWID) ≥ 18 years were eligible for participation. To be classified as a PWID, participants had to self-report injection drug use within the last 3 months. Our cases were PWID with

"Definite IE", based on the Modified Duke Criteria [21]. Cases were recruited from the three tertiary care centers in London that provide all inpatient care for patients with endocarditis. Outpatients being followed up for recent IE (within 6 weeks) were also recruited as cases through the outpatient Infectious Disease clinics covering London. Our controls were PWID with no history of IE episodes; these participants were recruited from addiction and outpatient clinics commonly serving PWID in London, Ontario. Recruitment for controls was conducted at Addiction Services of Thames Valley, Regional HIV/AIDS Connection, St Joseph's Health Care Infectious Disease clinics and the London Intercommunity Health Center. This choice of sites allowed us to observe a diverse subset of participants with varied injection practices. Sampling from the Regional HIV/AIDS and Addiction Services of Thames Valley allowed us to observe participants who had access to counselling, whereas sampling from St Joseph's Healthcare and London Intercommunity Health Center allowed us to observe participants who had riskier injection practices complicated with infections other than IE. We recruited participants from August 11, 2016 to July 27, 2018. Fig. 1 Process of preparation and injection of hydromorphone controlled-release capsule for injection drug use. Storage of the used cooker and filter for use of residual hydromorphone is almost very commonly performed and leads to bacterial contamination [12,14]. Heating the cooker with a cigarette lighter prior to use reduces bacterial burden [12] Anonymous interviews were conducted with a questionnaire querying demographic data, medical history pertaining to current and previous IE episodes (if any) and other infectious complications, and history of intravenous drug use in the past 3 months and over one's lifetime. Each participant was then assigned an identification number. Patient medical records were consulted by Infectious Disease physicians to verify definite cases of IE. Study data were digitized and managed using REDCap electronic data capture tools hosted at Lawson Health Research Institute [22]. The study protocol was approved by an institutional board review (Health Sciences Research Ethics Board, Western University). Written informed consent was obtained from all participants. Participants were compensated for their time with a Canadian $10 Tim Horton's coffee shop gift card. Reporting of all aspects of this study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies [23]. Statistical analysis Analyses were completed using R 3.6.2 [24]. Demographic and clinical characteristics of cases (PWID IE+) and controls (PWID IE-) were evaluated through independent samples t-tests (for continuous variables) and chi-square or Fisher's exact test (for categorical variables). To evaluate our first hypothesis, we used chi-square analysis to compare the proportion of cases (IE+) and controls (IE−) who utilized hydromorphone-CR. To evaluate our second hypothesis, we compiled two multivariable logistic regression models to evaluate variables we hypothesized would increase the risk of IE. One model included gender, use of government-distributed IDPE (Stericup) and injection into multiple sites among all participants. The other model included the previous variables and additionally, whether PWID do not heat their hydromorphone-CR; this model was completed among hydromorphone-CR users because the additional variable only applied to these individuals. Although reuse of hydromorphone-CR for a second wash is a known risk factor of IE, it was not placed in these models given our small sample size and the ubiquity of this practice our cohort; similarly, use of hydromorphone-CR was also not entered into the multivariable model. Results One hundred and forty-one interviews were conducted, with 135 individuals being included in the final study. Of the 34 case interviews conducted, 1/34 (3%) was excluded from the final study as this individual did not have a definite episode of IE; thus, 33/34 (97%) were included in the final study. Our cases primarily had tricuspid-valve endocarditis and the most common organism causing IE was methicillin-sensitive Staphylococcus aureus (Table 1). Of the 107 control interviews conducted, 4/107 (4%) were excluded as they had stopped injecting drugs for more than 3 months prior to interview. Moreover, 1/107 (1%) was excluded due to inconsistent answers being provided between the different sections of the interview. In finality, 102/107 (95%) controls were included in the final study. The demographic characteristics of cases and controls are presented in Table 2 and discussed below. With respect to our first hypothesis, the most commonly injected drug in both groups was hydromorphone-CR; however, hydromorphone-CR use was not significantly different between cases (91%) and controls (81%; p = 0.20). Cases and controls injected a wide range of drugs (Table 3); hydromorphone-immediate release tablets was the second most commonly injected drug. Furthermore, there was a trend towards crystal methamphetamine being injected by control participants more than case participants (54.4% cases; 78.4% controls; p = 0.07). Injection of fentanyl tablets or patches was sparse among PWID in London at this time. With respect to our second hypothesis, we evaluated the association of IE and sex, government-dispensed IDPE, location of injection, and heating of hydromorphone-CR. Multivariable logistic regression analyses found that being female and not using government-dispensed IDPE (Stericups) were independently associated with having IE in unadjusted and adjusted models (Tables 4 and 5); injecting in multiple sites and always heating hydromorphone-CR were not significantly associated with IE. In the adjusted model, among all participants, females had 4.65 times higher odds of having IE (95%CI 1.85, 12.28; p = 0.001), and those who did not use government-distributed IDPE had 5.76 times higher odds of having IE (95% CI 2.37, 14.91; p < 0.001). Lastly, we also sought to explore the demographic variables and injection practices of PWID in London to generate hypotheses for further studies. The demographic characteristics of the two cohorts are shown in Table 2. Cases also more commonly had completed secondary or post-secondary education (did not have incomplete degrees) compared to controls (61.3% cases; 33.7% controls; p = 0.01). Controls and cases were similar in age, race, housing status, concurrent HIV and hepatitis C infections, and past complications. Injection location data (Table 3) showed that cases and controls differed in the location they frequently injected (Fishers' exact test, p = 0.012); this difference was driven by the higher frequency of multi-site (32% vs 13%) among cases compared to controls. The injection preparation techniques and behaviors of cases and controls are highlighted in Table 6. Both cases and controls used cookers to prepare drugs for injection at a rate of 50-60%. Controls were more likely to use a Stericup, which is distributed in the IDPE kits by the provincial government (42.4% cases; 84.2% controls; p < 0.001) to prepare drugs. Furthermore, PWID with IE were more likely to use cooker types that were not listed in our surveys (Stericup, spoon, glass bottle, or shot glass) than controls (48.5% cases; 11.8% controls; p < 0.001). A cellulose filter was commonly used by both groups (96%). Both groups indicated reusing their filters at 63.7% controls and 60.6% cases for multiple injections. Additionally, both groups stored their filters in cookers or pockets and/or at body temperature in similar proportions. Many participants (18% controls and 15.6% cases) shared their filters. We did not find that heating drugs were protective; however, we did find that controls used lighters as crushers for drug preparation significantly more than cases (58.8% vs 36.4%; p = 0.025), suggesting that controls may have greater accessibility to a heating source. While there were no significant differences between cohorts in the frequencies of needle and syringe sharing or reuse, the rates of reuse are high with 74.5% of controls and 69.7% of cases reporting reuse of needles. Furthermore, the rate of reuse for syringe barrels also remains high with 66.7% of controls and 60.6% of cases reporting reuse. Discussion Understanding the risk factors associated with IE in PWID is important in developing harm reduction strategies. We hypothesized that the use of hydromorphone-CR would be a risk for IE among PWID; however, we did not find a significant increase in hydromorphone-CR use in IE patients vs controls (91% vs 84%). In contrast, our previous work has demonstrated evidence of such a relationship. Our population-wide study in Ontario with over 60,000 PWID showed a 3.3-fold higher risk of acquiring IE within 120 days when prescribed hydromorphone-CR compared with other opioids (p < 0.0001) [25]. Moreover, we have also shown that drug excipients within hydromorphone-CR preserve S. aureus survival in vitro [12]. This was not the case for immediate release hydromorphone or controlled-release Oxycodone [12]. Furthermore, we found that the injectate obtained from aspirating from equipment previously used to inject hydromorphone-CR was contaminated with S. aureus in 14% of cases, and thus, injection of this drug would commonly be associated with bacteremia [12]. We suspect that the very high prevalence of hydromorphone-CR use in both cases and controls led to a lack of power to identify a difference in use between the two groups in this study. There has been very little data assessing the detailed injection practices associated with developing IE. The literature primarily studies the clinical and epidemiological characteristics of PWID developing IE [3,5,7,26,27]. Some studies assessing injection practices of PWID are in relation to the development of skin and soft tissue infections [28] or infections in general [14,20,29]. To our knowledge, this is the largest study (n = 33) showcasing detailed survey data regarding injection practices of PWID with IE. Understanding PWID-IE risk factors are of importance to inform public health authorities in the development of harm reduction strategies reducing infections in this at-risk population. Our one-onone surveys have allowed for the collection of comprehensive quantitative and qualitative data to thoroughly understand injection practices and behaviors of PWID in our region to elucidate the etiology of our high IE rates. Previous studies suggested that IE in PWID was more frequently seen in males, younger patients, and those with concurrent HIV infections [3,8,9]. However, our cases and controls had similar ages, concurrent HIV, and HCV infections. Nearly a quarter of our cases and controls had HIV; the high incidence of HIV in this population is likely related to our co-existent local HIV epidemic [19]. Hepatitis C rates were based on selfreport, and a lack of awareness of status may have led to lower than expected rates in both cases and controls. Unexpectedly, being a female PWID was a risk factor for IE in our population (OR 4.65; 95% 1.85-12.28). Wurcel et al. also showed a greater parity in PWID-IE distribution by sex (female = 53%) between the ages of 15-34 over a 13-year review of IE hospitalizations in the USA [5]. We suspect that gender differences may exist with regards to injection technique. Women are more likely to have sex partners that initiate them into injection practices and are more likely to share IPDE than men [30,31]. Women can be identified sub-populations for targeted harm reduction, and in particular, interventions should account for intimate partner dynamics concerning high-risk practices [32]. Furthermore, female anatomy increases the difficulty of IVDU. We hypothesize that women have smaller veins, which may be difficult to visualize, often requiring increased manipulation during injections. This inability to find an adequate injection site with smaller veins can promote the usage of larger, more accessible central veins like the internal jugular, which further increases risks of infection. Additionally, local surveys in our region from the Middlesex-London Health Unit found that female PWID in London were more likely to borrow and share their IDPE [33]. It was also anecdotally noted, through our other project in progress, that women were less likely to access supervised injection sites, leading to unsafe injection practices that place them at risk of IE. Usage of provincial-distributed IDPE, i.e., the Stericup, for mixing drugs was found to be protective against IE. PWID who are more likely to use equipment from needle exchange programs are also more likely to be exposed to education on safe injection practices and consistently use sterile equipment. PWID with IE were also more likely to use objects for mixing and heating drugs that were not distributed through IDPE kits or commonly listed in our interview questions. This suggests that cases might be injecting in severe withdrawal states, where concern for safe practices do not take precedence over the need to use. Additionally, the increased use of a lighter may be suggestive that controls are using drugs that require heating such as heroin, crystal methamphetamine, and cocaine, and these may reflect a lower risk

of using hydromorphone-CR. However, we did not see a significant protective effect of always heating hydromorphone-CR preparations (OR 1.57; 95% CI 0.57, 4.71). We feel this may be due to the variations in practice that PWID follow. It is likely that PWID heat their drugs depending on the circumstance (state of withdrawal, supervised site, environmental factors, etc.), and this information was not captured in our line of questioning. We asked participants to choose a definite answer of whether they heated all the time or never heated their hydromorphone-CR injectates, which does not reflect the reality of injection behaviors. In other local literature, heating drugs has been shown to reduce bacterial load within cookers which contain hydromorphone-CR [12]. These findings have been translated into public health campaigns promoting "cooking one's drugs" to reduce infectious complications of IVDU in London, Ontario [34]. It may be that our cases and our controls were both engaging in this behavior, but our sample size was insufficient to detect any difference. Another hypothesized risk factor for IE was the site used for injection. Entrenched drug users tend to have thickened scar tissue from chronic injections in the same location; in many cases, this will be near the veins of the arm [35]. Furthermore, difficulty in accessing common sites may lead participants to inject in multiple sites for their hit, which can further increase the risk of infections given the multiple entry points for bacteria. Alternative sites of injection and IE likely reflects a greater difficulty in accessing safer sites, with alternate sites having a greater likelihood of contamination. Alternative injection sites may also be a surrogate marker for more venous damage from previous injections and thus entrenched drug use [29]. In particular, one study of PWID in the UK found the high-risk practice of injecting into the jugular vein was associated with the female gender and multiple body-site injections [29]. We found a significant difference between cases and controls with respect to the site of injection, which was driven by a higher frequency of cases using multiple sites. This was also seen in unadjusted logistic regression, where injection into multiple sites was associated with higher odds of IE (2.29; 95% CI 0.83, 6.07), albeit not statistically significant (p = 0.10). This effect was somewhat diminished when adjusting for sex and government-dispensed IDPE (OR 1.67; 95% CI 0.53, 4.97). Our multivariable analyses did not include reuse of hydromorphone-CR for a second wash given the universality of this practise in our cohort and our sample size. A review of the literature had shown that conducting multiple washes of hydromorphone-CR injectates could also serve as a risk factor for IE. We did not find an increased likelihood of performing multiple washes with hydromorphone-CR (88% vs 74%). Our sample size was likely inadequate to identify these differences because the frequencies of both of these behaviours were much greater in both groups than expected. However, a companion study surveying PWID in London found that PWID with HIV were 22.12 (4.51 to 108.59) times more likely to share cookers, filters, or washes in their threemonth recall period [14]. The high-risk practice of injecting prescription opioids from equipment that is reused multiple times is prevalent in our region and appears to be related to a high incidence of infectious complications including Hepatitis C and a very high incidence of IE [4,36]. The trend towards increased use of crystal methamphetamine in controls may be suggestive that participants using less hydromorphone but who substitute with other agents, may be at lower risk of IE. Homelessness and unstable housing have been associated with injecting in public spaces and other high-risk injection practices [37,38]. However, in exploratory analysis, cases (PWID IE+) were not more likely to lack stable housing compared to controls. This is supported by Roy et al. [16], who found that unstable housing was not associated with conducting multiple washes (utilizing residual drug for multiple injections), which is often the preparatory method used to inject prescription opioids, such as hydromorphone-CR. We hypothesize that PWID using hydromorphone-CR, which is a more costly illicit substance, can be associated with stable housing, which is reflective of financial stability. Hydromorphone is one of the most expensive prescription opioids to purchase illicitly, costing Canadian $5.57/mg (US$4.28/mg) or Canadian $100.26 for an 18-mg capsule [39]. Interestingly, we found that our cases were more likely to have completed secondary or post-secondary education (61.3% cases vs 33.7% controls). In contrast, previous studies have linked the incompletion of education to illicit substance use [40]. Higher education likely is correlated with income and again may reflect greater accessibility to expensive prescription opiates such as hydromorphone-CR, which has properties that increase the risk of infections [12]. Our evaluation of the relationships between housing status and education with IE were exploratory and will require further studies to confirm. Limitations A major limitation of this paper is that social and relationship factors which may put women at increased risk of IE were not explored in this survey. Limitations on the number of questions which could be asked in a single sitting led to this limitation, but it is important that these factors be extensively explored in subsequent studies. Further research should question whether people inject alone, use with a steady partner, are able to self-inject or rely on someone else most of the time, and whether they use with someone who controls their injection practices, such that they have no choice when receiving a used needle (and thus have a greater risk of infection). Case-control study design may lead to recall biases with cases more likely to recall perceived hazardous behaviors. Our study reviewed cases admitted to or transferred (due to IE complexity) to the hospitals in the city of London. We did not capture cases admitted to regional or rural sites as other research has found poor harm reduction practices prevalent in rural settings [20]. Our controls were PWID accessing community resources as well as those presenting to outpatient clinics for other infectious complications. We did not capture PWID that do not seek medical or community supports, potentially missing controls who inject only at home, are not actively followed up for their addiction disorders or are of high socioeconomic status. These patients may therefore be less likely to perform high-risk behaviors. In particular, using the provincially distributed harm reduction materials may have been more common due to the controls having more access to these as many were attending addiction service centers. Providing a monetary compensation for participation could have compromised the integrity of participant responses received; although, the relatively modest funds given in this study make this less likely. Finally, the relatively small sample size limited our ability to identify several hypothesized relationships. Further larger studies would be helpful. Conclusion This study did not find an association between hydromorphone-CR use and infective endocarditis. The very high prevalence of hydromorphone-CR use in London possibly made our sample size too small to identify a significant association. We found being a female PWID and not using clean injection materials were risk factors for IE. Our work supports the necessity of harm reduction equipment distribution programs in reducing infectious complications among PWID. Further study of the potential social relationships as well as biological factors that may lead to a higher risk of IE in women are warranted. A cyclometalated iridium(III) complex used as a conductor for the electrochemical sensing of IFN-γ A novel iridium(III) complex was prepared and used as a conductor for sensitive and enzyme-free electrochemical detection of interferon gamma (IFN-γ). This assay is based on a dual signal amplification mechanism involving positively charged gold nanoparticles ((+)AuNPs) and hybridization chain reaction (HCR). To construct the sensor, nafion (Nf) and (+)AuNPs composite membrane was first immobilized onto the electrode surface. Subsequently, a loop-stem structured capture probe (CP) containing a special IFN-γ interact strand was modified onto the (+)AuNP surface via the formation of Au-S bonds. Upon addition of IFN-γ, the loop-stem structure of CP was opened, and the newly exposed “sticky” region of CP then hybridized with DNA hairpin-1 (H1), which in turn opened its hairpin structure for hybridizing with DNA hairpin-2 (H2). Happen of HCR between H1 and H2 thus generated a polymeric duplex DNA (dsDNA) chain. Meanwhile, the iridium(III) complex could interact with the grooves of the dsDNA polymer, producing a strong current signal that was proportional to IFN-γ concentration. Thus, sensitive detection of IFN-γ could be realized with a detection limit down to 16.3 fM. Moreover, satisfied results were achieved by using this method for the detection of IFN-γ in human serum samples. Interferon gamma (IFN-γ ) is a type of cytokine that can activate multiple signal transduction pathways through the transcriptional regulation of immunologically relevant genes 1 . IFN-γ is unrestricted by T-helper cells with stimulation by antigens, and is clinically employed to diagnose latent tuberculosis 2 . In addition, dysregulation of IFN-γ secretion is associated with various diseases, such as inflammatory bowel disease, genital herpes simplex virus (HSV) infections and Alzheimer's disease 3 . Thus, sensitive detection of IFN-γ can be used to investigate the vigor of the immune response as well as for the diagnosis of infectious diseases. Electrochemical sensing platforms have been extensively used for the detection of biomolecules due of their intrinsic advantages, including good portability, low cost, high sensitivity and simple operation 4 . Recently, numerous electrochemical methods have been developed for IFN-γ detection coupled with a number of signal amplification-based on strategies 1,[5][6][7] . For instance, Zhao et al. constructed an electrochemical aptasensor for IFN-γ sensing based on the hybridization chain reaction (HCR) coupled with enzyme-signal amplification 8 ; Luo and Pu groups realized IFN-γ detection using a dual signal amplification approach by exonuclease-mediated surface-initiated enzymatic polymerization 2 ; Wang and He groups developed a sensitive label-free electrochemical aptasensor for IFN-γ detection through nuclease cleavage-assisted target recycling amplification 9 . However, these methods have certain drawbacks, such as the use of enzymes and/or the labelling of probe molecules, which limit the utilization of them for the routine detection of IFN-γ . Thus, developing low-cost, enzyme-free, simple and sensitive strategies for IFN-γ detection continues to attract tremendous interest. Iridium(III) complexes have been widely used coupled with electrochemiluminescent 10-12 and fluorescent 13-17 detection techniques due to their high luminescence efficiency, variable photophysical properties and different excited-state characteristics. In addition, such metal complexes can also efficiently intercalate into the grooves of dsDNA strands 18 . Recently, a number of electrochemiluminenscent sensors have been constructed for the sensitive sensing of protein [19][20][21] and DNA 22,23 . However, to the best of our knowledge, no electrochemical platform has yet been developed by using the iridium(III) complex as an electron conductor for the detection of protein. In this work, we synthetized an iridium(III) complex that possesses good conductivity, using it as an electron mediator, a sensitive and enzyme-free electrochemical sensor for the determination of IFN-γ based on HCR signal amplification was developed. To construct the sensor, Nf/AuNPs composite membrane was first immobilized onto the electrode surface. After that, a loop-stem structured capture probe (CP) containing a special IFN-γ interact strand was modified onto the (+ )AuNP surface via the formation of S-Au bonds. Upon addition of IFN-γ , the loop-stem structure of CP was opened, and the newly exposed "sticky" region of CP then hybridized with H 1 to open the hairpin structure of it, accompanied with the happen of HCR between H 1 and H 2 for the formation of dsDNA polymer. Subsequently, iridium(III) complex could bind with the grooves of the dsDNA polymer, producing a strong current signal that was proportional to IFN-γ concentration (Fig. 1). Thus, sensitive IFN-γ detection could be achieved based on identifying the current signal change of differential pulse voltammetry (DPV) signal. Compared to enzyme-assisted signal amplification, HCR avoids the use of enzymes, accordingly improved the robustness and stability of the sensor. Notably, the iridium(III) complex used here could interact in a stable fashion with the unlabelled dsDNA polymers, making the detection of IFN-γ simple and low-cost 24,25 . Materials and Methods Materials. Gold Table S1. Apparatus. All electrochemical measurements were performed on a CV-3 electrochemical workstation (CH Instruments, Inc. U.S.A.). A conventional three-electrode system including a modified gold electrode, a platinum wire counter electrode and an Ag/AgCl reference electrode was utilized in all of the experiments. Gel electrophoresis

was carried out on a GT Mini-Gel system (Bio-Rad Laboratories, Inc., Itally). Size of gold nanoparticles (AuNPs) was detected by using high resolution transmission electron microscopy (HRTEM, Tecnai G2 F20, USA). Nanosurface of the electrode was characterized by scanning electron microscopy (SEM, Hitachi 2000, Japan). Synthesis of iridium(III) complex. The structure of the iridium(III) complex [Ir(ppy) 2 (phen-dione)]PF 6 was shown in Figure S2a, and it was synthetized according to the literature method 21 . A suspension of [Ir(ppy) 2 ] 2 Cl 2 (0.2 mmol) and phen-dione (0.42 mmol) in a solvent solution of mixed CH 2 Cl 2 :MeOH (1:1, 20 mL) was refluxed overnight under the protection of N 2 . The reaction solution was then allowed to cool down to 25 °C, and filtered to remove solid dimer that was not reacted. An aqueous solution of NH 4 PF 6 (excess) was added into the filtrate and the resulting solution was rotary evaporated until crude product was precipitated out. The precipitate was then filtered and washed with several portions of water (2 × 40 mL) followed by diethyl ether (2 × 40 mL). Afterwards, the product was recrystallized by acetone:diethyl ether vapor diffusion to yield the titled compound as a brown solid ( Figure S1). Preparation of (+)AuNPs. Positively charged gold nanoparticles ((+ )AuNPs) were obtained through the reported method 22 based on the reduction of 15 mL HAuCl 4 (1.0 mM) by using 2 mL of NaBH 4 (100 mM) in the presence of 2 mL cetyltrimethyl ammonium bromide (CTAB, 10 mM). The color of prepared (+ )AuNPs was orange red and the average sizes of them estimated from high resolution transmission electron microscopy (HRTEM) analysis were about 4 nm (Fig. 2d). method 28 . After the immobilization of Nf and (+ )AuNPs onto the surface electrodes, they were incubated in CP solution over night at room temperature (RT). Subsequently, the CP/(+ )AuNPs/Nf modified electrodes were immersed in 6-mercapto-1-hexanol (MCH, 2.0 mM) for 1.5 h at RT to decrease the nonspecific DNA adsorption and to optimize the orientation of the aptamer. Then, the CP/(+ )AuNPs/Nf modified electrodes were incubated in IFN-γ solution (1 h) to open the hairpin structure of CP for the following HCR between H 1 and H 2 . At last, the {H 2 /H 1 } n /IFN-γ /CP/(+ )AuNPs/Nf modified electrodes were immersed in PBS and CH 3 CN (8/2, v/v) buffer containing 20 μ M of iridium(III) complex to realize the interaction of iridium(III) complex with dsDNA polymers. Fabrication of iridium(III) complex/{H Gel electrophoresis. Before conducting the gel electrophoresis, H 1 and H 2 samples were heated to 95 °C for 5 min and then allowed to cool down to 25 °C before using. 1.0 wt% agarose gel was prepared for DNA samples loading. After the addition of different DNA samples (4.0 μ L, 1.0 μ M), gel electrophoresis was performed by using 1.0 × TAE as running buffer at a constant potential of 78 V for 40 min. At last, the gels were photographed by using the gel image system after Stains-All staining by ethidium bromide (EB) solution for 15 min. Electrochemical detection of IFN-γ. To construct the electrochemical IFN-γ detection, CP/(+ )AuNPs/ Nf modified gold electrodes were immersed into PBS (20 mM, pH 7.0) containing different concentration of IFN-γ for 60 min at 37 °C. Then, they were immersed into H 1 and H 2 solution (2.0 μ M) for 50 min for the happen of HCR at RT. After that, the {H 2 /H 1 } n /IFN-γ /CP/(+ )AuNPs/Nf modified gold electrodes were immersed into the solution that contained of iridium(III) complex for 90 min to realize the interaction between iridium(III) complex and dsDNA polymers. Finally, the electrochemical characteristics of the sensor were determined in 20 mM of PBS (pH 7.0) through differential pulse voltammetry (DPV) experiments from − 0.2 to 0.2 V. Results and Discussion Characteristics of the modified electrodes. The establishment of (+ )AuNPs on the electrode surface can greatly improve the amount of CP immobilization. Here, SEM images were used to monitor the formation of Nf/(+ )AuNPs, and an obvious thin film of Nf appeared after 5 μ L of 1.0 wt% Nf added onto the gold electrode surface (Fig. 2b) compared to that of the bare gold electrode (Fig. 2a). Then, after the adsorption of (+ )AuNPs, a homogeneous nano-surface was observed (Fig. 2c). Moreover, Nf and Nf/AuNPs were investigated using X-ray diffraction (XRD) (Fig. 2f). The XRD pattern of Nf is consistent with cubic phase, referring to the standard spectra (JCPDS 15-0770), which displays two characteristic peaks occurring at 2θ of 17°, 39°. Then, two obvious characteristic peaks of AuNPs at 39.5° and 49° for 111 and 200 appeared after the formation of Nf/AuNPs, and a shift in the diffraction peak of Nf was observed due to its combination with AuNPs. Conductivity of different iridium(III) complexes. To monitor the conductivity mechanism of the iridium(III) complex, different iridium(III) complexes were synthesized, and scanned in PBS buffer (pH 7.0). The results in Figure S3 showed that the iridium(III) complex used in our assay (a) has an good redox peak while that for others (b and c) exhibited no obvious peak current (curves b and c). The conductivity of the iridium(III) complex used in our assay could be attributed to the reversible oxidation-reduction reaction between the two ketonic groups and two oxhydryl groups involving the transfer of protons and electrons (Fig. 3). Characteristics of different electrodes. The modification process of the electrodes was characterized by using cyclic voltammetry (CV) experiments in the presence of 50 mM Fe(CN) 6 3−/4− . As shown in Fig. 4A, a stable and well-defined redox peak was obtained for the bare gold electrode (curve a), and the peak current decreased greatly after the immobilization of Nf due to the character of it that can hinder electron transfer (curve b). Subsequently, an obvious increase of the peak current appeared after the adsorption of (+ )AuNPs, because of the good electrical conductivity of (+ )AuNPs (curve c). Then, the immobilization of CP onto (+ )AuNPs surface based on Au-S bond greatly resulted in a decreased peak current (curve d). The reason for that mainly because MB is negatively charged, which can lead to the electrostatic repulsion with negatively charged [Fe(CN) 6 ] 3−/4− and hinder the electron transfer between the electrode and [Fe(CN) 6 ] 3−/4− . Moreover, after the happen of HCR between H 1 and H 2 , the current response of [Fe(CN) 6 ] 3−/4− redox peak decreased further due to the increase of negatively-charged DNA strands (curve e). Meanwhile, CV experiments were also conducted in PBS solution to demonstrate the interaction of iridium(III) complex with dsDNA polymers. It could be seen from Fig. 4B that a well-fined redox peak appeared after the incubation of {H 2 /H 1 } n /IFN-γ /CP/(+ )AuNPs/Nf modified gold electrodes with the iridium(III) complex (curve c), which could be attributed to the interaction of the iridium(III) complex with the grooves of the dsDNA polymers. On the contrary, no peaks appeared for the bare gold electrode (curve a) or {H 2 /H 1 } n /IFN-γ /CP/(+ )AuNPs/Nf modified gold electrodes (curve b) in the absence of iridium(III) complexes, indicating the outstanding conductive capability of the iridium(III) complex. In addition, electrochemical impedance spectroscopy (EIS) was also used to monitor the fabrication process of the modified electrodes. As shown in Fig. 4C, using 50 mM of Fe[(CN) 6 ] 3−/4− as the detection solution, a relatively small resistance was observed at the cleaned GCE (diagram a). Then, after the immobilization of the Nf membrane, the resistance gradually increased (diagram b), indicating that Nf inhibited the electron transfer between the solution and the base electrode. Subsequently, an obvious decrease of the resistance appeared after the adsorption of (+ )AuNPs, because of the good electrical conductivity of (+ )AuNPs (diagram c). However, after the immobilization of CP, IFN-γ , and the initiation of HCR between H 1 and H 2 , the resistance increased gradually (diagrams d and e). The reason for the increased resistance is mainly due to the fact that the DNA strands are negatively charged, which directly led to electrostatic repulsion with the negatively charged [Fe(CN) 6 ] 3−/4− , hindering the electron transfer between the electrode and [Fe(CN) 6 ] 3−/4− . Finally, after the interaction of the iridium(III) complex with dsDNA polymers, the resistance decreased significantly, indicating that the iridium(III) complex has an outstanding conductive capability (diagram f). Meanwhile, chronoamperometric measurements that constructed in 50 mM of [Fe(CN) 6 ] 3−/4− at a constant potential of -0.3 V further illustrated the modification process of the electrodes. It could be seen from Fig. 4E that a stable current appeared within 200 s, and a corresponding current change was obtained with the stepwise-modification of the electrode, which was consistent with the EIS results. Fourier transform-infrared spectroscopy (FTIR) is a useful tool for characterizing the characteristics of different materials. As shown in Fig. 4D, characteristic IR peaks of Nf appeared around 970 cm −1 (-COC-linkages), 1075 cm −1 (-SO3 H) and 1156 cm −1 (-CF 2 -) (curve a) 29 . After the adsorption of AuNPs, an obvious shift of the IR peaks of Nf could be found, indicating the effective adsorption of AuNPs onto the Nf membrane (curve b). Then, two obvious IR peaks appeared around 1543 cm −1 (amide II band, C-N stretch coupled with N-H bend) and 1638 cm −1 (amide I band, mainly C= O stretch), which displayed the FT-IR spectra of IFN-γ protein (curve "a") after the immobilization of CP and IFN-γ onto the electrode surface (curve c) 30 . After that, the interaction of the Optimization of the experimental conditions. The electrochemical signal of the iridium(III) complex mainly based on the amount of it that interacted into the grooves of dsDNA polymers. Thus, the specific bind of IFN-γ with CP can directly affect the happen of HCR, and accordingly affect the interaction of iridium(III) complex. As shown in Fig. 5A, the oxidation peak current of iridium(III) complex enhanced with the increase of the binding time between IFN-γ and CP in the range from 5 to 60 min and then reached a plateau, indicating 60 min could realize the effective binding of CP with IFN-γ . Thus, 60 min was selected as the optimum binding time between IFN-γ and CP. In addition, the hybridization time between CP and H 1 /H 2 mixture directly affected the happen of HCR. As displayed in Fig. 5B, the electrochemical signal of iridium(III) complex enhanced with increasing hybridization time of them in the range of 0-50 min. Such result indicated that HCR could effectively happen within 50 min. Therefore, 50 min was chosen as the optimum hybridization time. The interaction time of iridium(III) complex into the grooves of dsDNA polymers was another crucial parameter that could influence the sensitivity of the detection assay. As displayed in Fig. 5C, the current signal was proportional to the interaction time of iridium(III) complex over the range from 5 to 90 min and then reached a plateau. So 90 min was used for the interaction of iridium(III) complex with dsDNA polymers. The pH of the PBS also has an effect on the electrochemical behavior of the sensor because the activity of IFN-γ and the oxidation-reduction of iridium(III) complex may be affected by the acidity of the solution 33 . Just as shown in Fig. 5D, the peak current of iridium(III) complex decreased along with the increasing pH value from 5.0 to 9.0 and a plateau from 6.0 to 7.0. Considering that high acid solution may affect the activity of the protein, pH 7.0 was selected as the optimum pH of PBS. Sensitivity of the assay. Under optimal conditions, IFN-γ detection could be realized based on monitoring the current signal change. As shown in Fig. 6A, the DPV peak currents of the iridium(III) complex increased along with the increase of IFN-γ concentration, and a linear range from 50 fM to 3.0 pM was obtained for IFN-γ detection (Fig. 6B, linear a), with a detection limit of 16.3 fM (3σ/slope). The linear regression equation was I = 63.35 + 63.58c (c: pM, R 2 = 0.9971). The detection limit of the proposed method was lower

than the reported aptamer-based IFN-γ detection methods (Table S2) 5,8,9,[34][35][36] . Such high sensitivity mainly attributed to the dual signal amplification of HCR and (+ )AuNPs. Meantime, we compared the performance of this sensor with a similar sensor {H 2 /H 1 } n /IFN-γ /CP that lacked (+ )AuNPs. A linear relationship between peak current and IFN-γ concentration was obtained in the range of 200 fM-2.5 pM (Fig. 6B, linear b), illustrating the function of the (+ ) AuNPs as an electron mediator and signal amplifier. Selectivity and stability of the sensor. The selectivity of the sensor was investigated by monitoring the current response of the sensor to IFN-γ , bovine serum albumin (BSA), immunoglobulin G (IgG) and human serum albumin (HSA). As shown in Figure S4, the current signal of iridium(III) complex for 1.5 pM of IFN-γ was greater than 15 pM of various other substances. These results obviously illustrated the high selectivity of the sensor for IFN-γ detection, which was mainly due to the high specificity of CP to IFN-γ . In addition, the stability of the sensor was studied by storing the sensor at room temperature for 10 days, and the sensor kept 95.2% of its initial response, which indicated that the sensor has a good life time for IFN-γ detection. IFN-γ detection in human serum samples. Moreover, the application of the sensor for IFN-γ determination in serum samples was evaluated in a detection system containing 5.0% (v/v) human serum. As shown in Table S3, a good recovery in the range of 96.6-108.1% was obtained for IFN-γ detection with a relative standard deviation (RSD) between 1.87% and 4.06%, indicating the potential application of the assay for quantitative detection of IFN-γ in complex human samples. Conclusion In conclusion, an enzyme-free and highly sensitive electrochemical sensor for IFN-γ detection was developed by utilizing a novel iridium(III) complex as conductor coupled with a dual signal amplification mechanism involving HCR and (+ )AuNPs. Importantly, this amplification mechanism avoids the use of enzymes. Moreover, the iridium(III) complex used here could interact stably with formed dsDNA polymers without modification, making the IFN-γ detection assay simple and low-cost. Finally, the assay was highly selective for IFN-γ due to the specificity of CP for IFN-γ , and could also function in diluted human serum. Thus, this approach could possibly be applied for developing biosensors to monitor IFN-γ levels associated with various human diseases. A Comprehensive Evaluation of NIPAM Polymer Gel Dosimeters on Three Orthogonal Planes and Temporal Stability Analysis Polymer gel dosimeters have been proven useful for dose evaluation in radiotherapy treatments. Previous studies have demonstrated that using a polymer gel dosimeter requires a 24 h reaction time to stabilize and further evaluate the measured dose distribution in two-dimensional dosimetry. In this study, the short-term stability within 24 h and feasibility of N-isopropylacrylamide (NIPAM) polymer gel dosimeters for use in three-dimensional dosimetry were evaluated using magnetic resonance imaging (MRI). NIPAM gels were used to measure the dose volume in a clinical case of intensity-modulated radiation therapy (IMRT). For dose readouts, MR images of irradiated NIPAM gel phantoms were acquired at 2, 5, 12, and 24 h after dose delivery. The mean standard errors of dose conversion from using dose calibration curves (DRC) were calculated. The measured dose volumes at the four time points were compared with those calculated using a treatment planning system (TPS). The mean standard errors of the dose conversion from using the DRCs were lower than 1 Gy. Mean pass rates of 2, 5, 12, and 24 h axial dose maps calculated using gamma evaluation with 3% dose difference and 3 mm distance-to-agreement criteria were 83.5% ± 0.9%, 85.9% ± 0.6%, 98.7% ± 0.3%, and 98.5% ± 0.9%, respectively. Compared with the dose volume histogram of the TPS, the absolute mean relative volume differences of the 2, 5, 12, and 24 h measured dose volumes were lower than 1% for the irradiated region with an absorbed dose higher than 2.8 Gy. It was concluded that a 12 h reaction time was sufficient to acquire accurate dose volume using the NIPAM gels with MR readouts. Introduction Intensity-modulated radiation therapy (IMRT) has been widely applied in modern radiation therapy. Pretreatment verifications have become a crucial part of routine patient-specific quality control in IMRT [1]. Traditional measurement tools, such as ion chambers and films, have been used to verify the dose distribution of IMRT. However, these tools provide only point or planar dose measurements. To fully verify a three-dimensional (3D) dose distribution, Gore et al. [2] used a ferrous sulfate gel dosimeter (Fricke gel) to measure the dose distribution in three dimensions. However, the measured dose distribution is strongly influenced by readily dispersed ferric ions, resulting in low signals, blurred images, and, ultimately, errors in dose measurement. In 1958, Hoecker and Watkins reported that the critical diffusion of ferric ions in Fricke gel can be prevented using a radiation-induced polymerized monomer [3]. In the last decade, polymer gel dosimeters have become useful in measuring dose distribution. In contrast to traditional measurement tools, polymer gel dosimeters can capture the entire 3D dose distribution in a single measurement without signification diffusion. In addition, polymer gel dosimeters demonstrate the advantages of easy shaping and are equivalent to human tissues. The basic physical process of polymer gel dosimetry relies on water radiolysis, leading to radicals that interact with monomers, thereby initializing the polymerization reaction. When the polymerization reaction is completed, the chains become spatially trapped in the parts of the gel matrix affected by radiation. Therefore, the dose distribution can be obtained by measuring the changes of growing polymer chains. Several modalities have been used, including Xray computed tomography [4], optical computed tomography [5,6], ultrasound [7], and magnetic resonance imaging (MRI) [8,9]. In MRI, the spin-spin relaxation rate (R2) depends on the mobility of water molecules. The polymer chains formed in the gel matrix reduce the mobility of water molecules. MRI can therefore determine the degree of polymerization through T2-weighted imaging. Additionally, MRI demonstrates the advantages of high spatial resolution and no additional dose to gel. Senden et al. [10] proposed a new polymer gel, mainly composed of gelatin, N-isopropylacrylamide (NIPAM), Bis, and tetrakis (hydroxymethyl) phosphonium chloride (THPC), with a high radiation sensitivity that allows the reaction of monomers and free radicals in the irradiated region. An increasing number of reports [11][12][13] show that NIPAM gel dosimeters have potential for use in the verification of radiotherapy dose distributions. Previous work [11,12] has focused on the fundamental characteristics of NIPAM polymer gel dosimeters and the feasibility of an NIPAM/MRI system for clinical 3D dosimetry. Gel dosimeters are generally considered stabilized and readable at 24 h postirradiation [11,13]. In subsequent investigations, simple dose distributions and γ-index maps from gel measurements have been compared with those from treatment planning system (TPS) calculations regarding central axial planar dose distributions [14]. In this study, the short-term stability and feasibility of the NIPAM gel dosimeter with MR readouts were evaluated using a clinical case of eye tumor intensity-modulated radiation therapy (IMRT). Dose maps and dose volume measured from the NIPAM gel dosimeters at 2, 5, 12, and 24 h postirradiation were compared to that calculated from a treatment planning system. Dose maps in axial, coronal, and sagittal views were evaluated using the isodose maps, dose profiles and gamma evaluation. In addition, the entire dose volume was evaluated uising dose volume histogram (DVH). NIPAM gel dosimeter preparation The gel used in this study consisted of 5% gelatin (300 Bloom Tape A, Sigma-Aldrich), 3% NIPAM (97%, Sigma-Aldrich), 3% BIS (Merck), 10 mM THPC (80%, Sigma-Aldrich), and 87% deionized water. The NIPAM polymer gel was prepared according to the instructions of Senden et al. [10]. After manufacture, the gels were poured into Pyrex tubes for calibration and into three cylindrical gel phantoms for dose distribution measurement; the tubes were subsequently placed in cylindrical polymethylmethacrylate phantom containers (130 mm diameter, 130 mm height, and 5 mm wall thickness) covered with aluminum foil. The containers were then stored in a refrigerator at 4°C to prevent light-induced prepolymerization until complete solidification was achieved. After dose delivery and between MR scans, the gel phantoms and calibration tubes were placed in the scanning room at 23 ± 1°C to reduce the influence of temperature on the polymerization reaction. Dose delivery Calibration tubes. The NIPAM gels were irradiated using a 6 MV linear accelerator (Clinac 21EX LINAC, Varian Medical Systems, USA). Quality assurance of the linear accelerator was regularly performed, thereby passing the regulations of Taiwan. The photon output error of the medical accelerator was validated daily at lower than 3%. The irradiation was performed at the following settings: beam angle, 0°; dose rate, 4 Gy/min; and field size, 10 × 10 cm 2 . The dosimeters were placed in an acrylic phantom (length, 30 cm; width, 30 cm; thickness, 4 cm), placed between two 3 cm-thick solid water slabs. Seven calibration tubes were prepared to determine a dose response curve (DRC) for the dose conversion. The doses delivered to the tubes were 0, 1, 2, 5, 8, 10, and 12 Gy. Cylindrical gel phantoms. For treatment planning, CT images of a cylindrical phantom were acquired to obtain the geometry by using a simulation CT (CT Simulation, Marconi AcQ-Sim, Philips, UK). The cylindrical phantom was filled with gelatin to prevent unnecessary dose absorption in the gels and to mimic the gels' photon attenuation characteristics. The CT images were imported into the TPS, and an IMRT plan for eye tumor treatment was generated using the Eclipse TPS v10.0 (Varian Medical Systems, Palo Alto, CA). Three cylindrical gel phantoms filled with the NIPAM gel were irradiated using dose delivery techniques identical to those commonly used in patient treatments. The irradiation condition settings were as follows: prescribed dose at isocenter, 5 Gy; photon beam energy, 6 MV; dose rate, 400 cGy/min; number of fields, five; and source-to-axis distance (SAD), 100 cm. MRI scanning and data analysis A clinical 1.5 Tesla MRI scanner (MAGNETOM Aera MRI Scanner, Siemens, Germany) with a head coil was used to scan the gel phantoms. As shown in Fig 1, the calibration tubes and gel phantoms were inserted into a customized acrylic holder. The T2-weighted images of the tubes and phantoms were acquired using a multiple-spin echo sequence with the following parameters: TR, 3000 ms; echo spacing, 22 ms; number of echoes, 16; FOV, 240×240 mm 2 ; matrix size, 512×512; and slice thickness, 5 mm. Comprehensive Evaluation of NIPAM Dosimeters on Three Orthogonal Planes After scanning, the acquired MR images were analyzed using MATLAB (The MathWorks, Natick, MA, USA). The DRC was determined using the following three procedures. First, regions of interest were drawn in the tubes to measure the mean signal. Employing the many-points method [15], least-square fittings were performed to determine the T2 values of each tube, using the T2 relaxation model, mean signals, and echo times of 16 echoes. Second, R2 values were calculated by 1/T2 values. Finally, the DRC was determined by linear fitting the R2 values and the absorbed doses of the tubes by using R2 = a × Dose + b. Similarly, the MR images of the cylindrical gel phantom were converted to R2 maps by using the aforementioned procedures on a pixelby-pixel basis. Using the DRC, the R2 maps were converted into dose maps directly. In addition, to investigate the time polymerization reaction characteristics of NIPAM gel under MRI, the calibration tubes and three gel phantoms were scanned at four time points: 2, 5, 12, and 24 h after dose delivery. To accurately obtain T2-weighted images at the four time points, the calibration tubes and three gel phantoms were separately irradiated with a 40 min interval for MR imaging. On the basis of the irradiation time, the T2-weighted images of the calibration tubes and three gel phantoms were separately acquired at the four time points. The R2 maps were converted into dose maps by using the DRCs obtained from the same time points. Evaluation Uncertainty of dose conversion. To evaluate the errors from dose conversion, the uncertainty of the dose converted from the measured R2 maps by using the DRCs was calculated. Dose-response relationship, or calibration curve, is a widely recognized quantitative tool in science and technology. Typically, a single response measurement y is related to a single predictor x for each

observation using a simple linear regression as follows: where y i is the ith observation of the response to dose x i ; β is the intercept; α is the slope; and ε i is an unobservable error term with zero mean and constant variance σ 2 , i.e., {ε i : 1≦i≦n} is independent and identically normally distributed as N(0, σ 2 ). In our study, response y is the relaxation rate R2 and the predictor x is dose D. We seek to fit the n data points (D i , R2 i ) into the linear model given as follows: where R _ 2 be the prediction of R2, a is the least squares estimator of the slop, and b is the least squares estimator of the intercept. The standard error of R _ 2 is the appropriate quantity to use for error bars on R2 values obtained from the best fit in the least squares. The standard error of R _ 2 at D = D p can be computed using the following formula [16]: where s R2,D is the standard deviation of R2(D) and is defined by The inverse of this problem is a point of interest. In reverse regression, we regard D as the response and R2 as the predictor. In our study, the values (R2 i , D i ) comprise a set of n data pairs in which we wish to fit the reverse regression model. We express our estimated model [17] with This method disregards the simple linear regression assumption that the predictor is measured with a negligible error (Parker et al., 2010). For a future observed value of R2 0 , the standard error of D 0 using reverse regression is as follows: where s D,R2 is the standard deviation of D(R2) and is defined by In our study, the standard errors of the parameters a and b (Δa and Δb), and the mean standard errors of the dose conversion from using the DRCs (mean ΔD) were calculated for each measured time point. Evaluation of measured dose distribution. In this study, the gamma evaluation proposed by Low et al. [18] was used to evaluate the dose maps measured using NIPAM gel with MR readouts. The gamma evaluation comprises both dose difference and distance-to-agreement (DTA) comparison criteria for comparing two dose maps. The gamma index quantitatively represents the difference between dose maps and can be calculated using Eqs 1,2,3 and 4. where Γ is a gamma function and is described as follows: where rðr m ; r c Þ ¼ jr c ; r m j ð 11Þ and where r m and r c are the spatial locations of the dose pixels in the measured and calculated dose maps, respectively. The terms d and δ respectively denote the spatial distance and dose difference between pixels, and Δd M and ΔD M represent the DTA and dose difference comparison criteria, correspondingly. A pixel with a gamma value lower than 1 indicates that the combined errors of the spatial distance and dose difference of the pixel were lower than the predefined criteria, thus passing the evaluation. By contrast, a pixel with a gamma index greater than 1 corresponds to the locations where the measured value does not satisfy the criteria. Finally, the pass rate of a gamma map was calculated using the number of passed pixels divided by the pixel number of the gamma map and multiplied by 100%. The pass rate therefore represents the percentage of a dose map satisfying the evaluation criteria. In this study, the dose maps measured using NIPAM gels were compared with those from the TPS by using the criteria of a 3% dose difference (ΔD M ) and 3 mm DTA (Δd M ), which are the most frequently used criteria in published comparisons of treatment plans [19]. In addition to the gamma evaluation, DVH, which is generally used to analyze the 3D dosimetry and quality of treatment plans in clinical practice, was calculated to compare the dose volumes of NIPAM gel and TPS. The relative volume (RV) covered by an absorbed dose was calculated from 0% to 100% of the prescribed dose with an interval of 1%. The percent RV differences between the dose volumes from the TPS and the measurements across a dose range from an absorbed dose starting at 0 Gy to the prescribed dose were also calculated using the following equation: where d RV % is the percent RV differences, RV TPS and RV m represent the relative volumes calculated from the dose volumes of the TPS and the measurements, D s is the starting dose of the dose range, D p is the prescribed dose, and t denotes the measurement time. Results and Discussion Characteristics of gel dose response The values corresponding to short T2 represent a high absorbed dose. The DRCs of the NIPAM gel dosimeters from 2, 5, 12, and 24 h measurements are shown in Fig 3. The fitting parameters and linearity of the DRC fits are listed in Table 1. The linearity of four DRCs was higher than 0.99. The coefficient of variance (CoV) of the slope among four DRCs was lower than 8%, indicated no significant sensitivity difference among the four readout time points. The temporal instability is mainly caused by continuous slight polymerization in normoxic polymer dosimeters after irradiation. The 12 and 24 h DRCs were highly matched for absorbed doses higher than 2 Gy. Uncertainty of dose conversion from using dose response curves Table 2 lists the values and standard errors (Δa and Δb) of the fitting parameters a and b as well as the linearity of DRC (r 2 ). The mean standard errors of the dose conversion using the DRCs (mean ΔD) are also listed. For all the measured time points, Δa and Δb are lower than 0.002 s −1 Gy −1 and 0.015 s −1 , respectively, which indicates that the DRCs fit well. In addition, the mean ΔD of all the measured time points is lower than 1 Gy, which demonstrates that the dose conversion that uses the DRCs is accurate and reliable for all the measured time points. Temporal stability of dose distributions in the cylindrical gel phantom Fig 4 illustrates the DVH and the percent RV differences between the dose volumes from the TPS and the measurements. When the starting dose was below 2.1 Gy, the differences between the 2 h and 5 h dose volumes were higher than 8%. When the initial dose exceeded 2.6 Gy, the differences decreased and were lower than 5%. By contrast, the differences between the 12 h and 24 h dose volumes were lower than 5% for volumes with an absorbed dose lower than 4 Gy. When the starting dose exceeded 4 Gy, the RV was lower than 2.1%, and the percent RV differences increased to over 5% for all the measured time points. The absolute mean percent RV differences between the 2, 5, 12, and 24 h dose volumes and that of the TPS were 6.48%, 5.77%, 2.78%, and 1.98%, respectively. Table 2 lists the mean pass rates and standard deviations of the three orthogonal dose maps calculated from the three gel phantoms. In the 2 h and 5 h dose maps, the pass rates were lower than 95%. The poor performance of the 2 h and 5 h measurements was mainly attributed to a continuous polymerization reaction surrounding the irradiated region where the absorbed dose was lower than 50% of the prescribed dose, as observed in the DVHs of the measured dose volumes and TPS (Fig 4). The pass rates of the dose maps increased as time elapsed and became stable after 12 h. The pass rates of 12 and 24 h dose maps were higher than 0.97, which is a common acceptable criterion for dose delivery in clinical practice. The gamma maps of the 24 h dose maps with 3%/3 mm criteria are shown in Fig 5. The gamma values of most of the Table 1. Standard errors of the parameters a and b (Δa and Δb), linearity of the DRC (r 2 ), and the mean standard errors of the dose conversion from using the DRCs (mean ΔD). dose points were lower than 1, thus passing the evaluation. A few failed points appeared at the edges of the sagittal and coronal dose maps. These failed points may be attributable to the image noise. In the gamma evaluation, the mismatch between the dose maps from the NIPAM gel and TPS could also reduce the pass rates. The pass rates of a measured dose map converted from the R2 maps shown in Fig 2(A) were calculated using the original simulated map and the maps before and after the original map. On the basis of the 3%/3 mm criterion, the pass rates of the measured dose map to the previous, original, and subsequent simulated maps were 79.14%, 83.13%, and 81.47%, respectively. The results indicate that a poor mismatch among the dose maps strongly influences and degrades the pass rate of the dose maps. To ensure that the spatial location of the measured and simulated maps was as close as possible, the slice location and thickness of the MR scans were set to match the values used in the CT scans. On the basis of the MR-acquired dose map, the closest simulated dose map was used for gamma evaluation. The differences between the spatial location of the measured and simulated maps should be minimal. De Deene et al. showed that most of the polymerization reactions occurred within the first 24 h after irradiation, with some lasting up to 30 days [20]. The stabilization of polymerization could be observed in the slope difference per hour of the DRC with time. The slope differences per hour between the DRCs of 2 h to 5 h, 5 h to 12 h, and 12 h to 24 h were 3×10 −3 , 1×10 −3 , and 3.3×10 −4 s −1 Gy −1 h −1 , respectively. After 22 h, the slope difference was reduced tenfold, which indicated that the polymerization had stabilized. However, the results of the DVHs showed that the NIPAM gel dosimeters with an absorbed dose higher than 2.8 Gy could be stabilized after 2 h post-irradiation. This finding confirmed that the polymerization reaction of the NIPAM gel dosimeters could be completed within a short post-irradiation period, as previously reported in [21]. This result was probably attributed to the intensive radicals produced at this dose level, which rapidly depleted most of the monomers in the NIPAM gels to form stable polymer structures for accurate dose readouts. By contrast, the NIPAM gels with an absorbed dose lower than 2.8 Gy retained a sufficient number of monomers to form relatively larger polymer structures than the initial ones. In addition, the production rate of radicals under such dose level was low. Consequently, stable polymer structures could not be formed instantly, which caused the errors in the 2 h DVH and the failure in the low-dose region during the gamma evaluation. In this case, the continuous reactions that occur in the NIPAM gels until 12 h post-irradiation are crucial to determine the low-dose region accurately. In this case, continuous reactions occurred in the NIPAM gels until 12 h postirradiation are crucial for accurately determining the low-dose region. In addition, the relationship between stabilization time and absorbed dose may be changed with different gel formulas, thus indicating the need for further study. Conclusion In this study, the short-term stability of the NIPAM gels with MR readouts was evaluated using a clinical IMRT case. The results demonstrate that the NIPAM gel dosimeter with MR readouts could accurately provide the entire dose volume after a 12 h reaction time. In addition, volumes with an absorbed dose higher than 2.8 Gy can be rapidly obtained after a 2 h reaction time. It is therefore concluded that NIPAM gel dosimeters with MR readouts could be useful and convenient in dose verification of clinical IMRT. Effect of stabilization exercise combined with respiratory resistance and whole body vibration on patients with lumbar instability: A randomized controlled trial Background: Lumbar stability exercise promotes deep muscle functions, and it is an effective intervention method for increasing proprioceptive sensation. This study aims to explore and compare the

effects of lumbar stability exercise with respiratory resistance and whole body vibration on patients with lumbar instability. Methods: This study is a 3-group randomized control trial. Through screening tests, 48 patients with lumbar instability were selected and randomly assigned to SE group (n = 16), stabilization exercise program using respiratory resistance (SER) group (n = 16), and stabilization exercise program using respiratory resistance and whole body vibration (SERW) group (n = 16). In order to compare the effects depending on the intervention methods, quadruple visual analogue scale (QVAS), Functional Ability Roland-Morris low back pain and disability questionnaire ([RMDQ], center of pressure path length, velocity, and area), Korean version of fear-avoidance beliefs questionnaire, and Pulmonary Function were used for measurement. Results: All of the groups showed significant improvements in QVAS, RMDQ, Korean version of fear-avoidance beliefs questionnaire, and balance abilities before and after the interventions. The SER group and SERW group showed a significant difference in QVAS and RMDQ than the SE group (P < .05). In addition, balance ability showed a significant difference in SERW group (P < .05), where only the SER group showed a significant difference in pulmonary function indexes including forced vital capacity, forced expiratory volume in 1 second, maximum inspiratory pressure, and maximum expiratory pressure (P < .05). Conclusion: Stabilization exercise program using respiratory resistance and whole-body vibration administered according to the purpose of intervention methods may be effective exercise programs for people with lumbar instability. Introduction Low back pain (LBP) is a common muscular disorder in which approximately 60% to 80% of the population experience at least once in their lifetime. [1] LBP is a syndrome that occurs in the area between second lumbar spine to sacroiliac joint, where there is an abnormal overload in the area due to weakening of the soft tissues and muscles. [2] As a result, issues including physical dysfunctions such as activities of daily living (ADL) and psychological factors are reported due to LBP. [3] Sensitivity of body tissues increases from the area of LBP. [4] This makes hardening of muscles around the spine, decreased range of motion of spinal joints, and ultimately difficulties in the performance of ADLs. Patients with LBP usually show kinesiophobia due to the functional issues. [5] In addition, patients also develop negative psychosocial effects such as low self-confidence and apathy due to fear of pain, decreased quality of life, and depression. [6] Many intervention methods are introduced to address alleviation of LBP and functional recovery, but exercise therapy Medicine is recommended to improve muscle instability. Stabilization exercise allows prioritized collaborative contraction of deep muscles such as transverse abdominalis and multifidus, and greatly contributes to dynamic lumbar stabilization and static stabilization of spinal segments. [7] The stabilization exercise activates sensory-motor control to modify muscle asymmetry and decrease LBP relapse, and demonstrates positive effects in improving signs of LBP through increased motor performance. [8][9][10] In order to provide lumbar stabilization to patients with lumbar pain, deep muscle contractions are essential. [11] To enhance the effects during lumbar stabilization exercises, exercise interventions with respiratory methods that promote diaphragm contraction and stabilization exercises that come with respiratory resistance to induce strengthening of deep muscles in abdominals, spine, and pelvis by holding respiratory resistance apparatus with the mouth to facilitate resistance during respiration are introduced. Stabilization exercises with respiratory resistance demonstrates decreased pain, increased motor functions, and sense of psychosocial stability, and it is reported to be more effective in increasing pulmonary functions that the conventional stabilization exercises. [12] Vibration stimulations are reported to have positive effects by enhancing motor functions and increasing energy metabolization and blood circulation. [13] When vibration is applied to muscles, sensory stimuli are provided to activate muscle spindles, and thereby strengthening deep muscles that are critical for postrural stabilization. In addition, whole body vibration (WBV) exercise activates proprioceptors and Golgi tendons, inducing reflective contraction of muscles to affect kinaesthesis. Stabilization exercise with whole body vibration is effective for pain, balance ability, and deep muscle activation in LBP patients. [14] Additionally, stabilization exercise with respiratory resistance is reported to be an effective intervention for LBP, dysfunction, and psychosocial stability. [15] Respiratory resistance and whole body vibration given with stabilization exercises are both suggested as important factors for the recovery of LBP, but comparisons in exercise intensity or effects depending on the functional levels of patients with lumbar Tinstability are not yet recommended. Therefore, this study aims to investigate the effects of stabilization exercises with respiratory resistance and whole body vibration on pain, functional level, psychosocial level, and pulmonary functions and to suggest effective stabilization exercise method and clinical feasibility. Participants This study recruited 82 patients who were receiving therapy for lumbar pain as either outpatient or inpatient in P hospital in D city. The study took place between November 2020 and January 2021. Due to the COVID-19 pandemic, preventive measures against the pandemic such as taking body temperatures were taken into consideration prior to the recruitment. The inclusion criteria were: persons who experienced LBP within 6 weeks, mean score of 3 or higher in quadruple visual analogue scale (QVAS), more than, positives in lumbar instability test, [16] and persons who are able to stand on 1 leg for more than 30 seconds. The exclusion criteria were: persons who have difficulty in participating in the intervention due to vestibular disorders, persons who have respiratory disorders, persons with the history of spinal surgery, who are pregnant, and persons with the study participation rate less than 80%. All participants signed the agreement ensuring that they fully understood the purpose and process of the study and are voluntarily participating. Study design This study is a 3-group randomized control trial (3-group RCT study design). G*power program was used to determine the size of the participants. Medium effect size was set to .25 according to Cohen's f, significance level (α) as .05, and Power(1−β) = .8, 42 participants were needed for this study. However, 15% of drop out rate was considered, therefore 48 participants were ultimately selected. The selected 82 participants were given a lumbar instability test to screen appropriateness for inclusion in the study. The lumbar instability test considered the participant to have lumbar instability when 3 or more items out of the 5 items resulted in positive. [16] The items tested were: instability test prone (positive when pain disappears with manual pressure), passive extension test of lumbar area (positive when there is pain or not able to maintain position), lumbar segment posterior/anterior movement test (positive when there is abnormal movement), direct lift of lower extremity test (positive if greater than 90 degrees), and age (positive when over 41). Through the screening test 34 total participants were excluded for the study. There were participants who did not satisfy the lumbar instability test (n = 30) and who scored below 2 in pain level (QVAS) (n = 4). After conducting screening test for the 48 participants, a random number production program was used [17] to randomly assign the participants to stabilization exercise group (SE) (n = 16), stabilization exercise program using respiratory resistance group (SER) (n = 16), and stabilization exercise program using respiratory resistance and whole body vibration (SERW) (n = 16). The participants were not given information about the group to which they had been assigned. Pain level, functional ability, Korean version of fear-avoidance beliefs questionnaire (FABQ), and pulmonary functions were conducted before and after the interventions for all 3 groups. All participants received stabilization exercises, and therefore, the participants were not able to determine which group they were assigned to. All assessments were made by a physiotherapist with 8 years of clinical experience and specialization in musculoskeletal system. This study was approved by the Ethics Committee, and is registered in the WHO International Clinical Trials Registry Platform: KCT0005773. The study process is shown in Figure 1. Stabilization exercise program. The stabilization exercise that all 3 groups received were modified from the exercise program method suggested by Zheng, et al [18] This exercise program induces lumbar stability through contraction of the abdominal muscles and strengthening of lumbar area and lower extremity muscles. The exercise program consisted 6 exercises that include squat, lunge, flank, curl up, bridge, and bridge with knee extension. Stretching for 10 minutes is provided with the purpose of warm up and cool down before and after each interventions. Each exercises were performed for 5 repetitions in a set, 10 seconds for each set, and 5 sets total. Break time of 20 seconds were given between each sets. The intervention was provided for 60 minutes per session, 3 sessions per week, and a total period of 5 weeks. Stabilization exercise program using respiratory resistance. The SER group performed the exercise program using a respiratory resistance apparatus (Expand a Lung, Miami, USA). Ventilation during respiration may be controlled using the respiratory resistance apparatus, and it is also used to strengthen respiratory muscles. In addition, a reliable index Rating of Perceived Exertion (RPE) was used to measure the level of exercise fatigue to observe motion resistance during the exercise. This study controlled the RPE between 13 to 14 to maintain the level between " difficult" and "slightly difficult." [19] After attaching the respiratory resistance apparatus, explanation confirming that interventions may be discontinued in situations of abnormal situations, dizziness, and breathing difficulties. SERW. The SERW group performed the exercise program using a respiratory resistor apparatus and vibrator (SW-VH11, Wonju, Korea) (Fig. 2). Vibration stimulation is administered as an exercise method that contributes stabilization by delivering the stimulus to lumbopelvic area or proximal part of the lower extremities. In the intervention, the frequency followed the study methods suggested by Di Giminiani, et al [20] and the sound wave intensity was set to 30 and the frequency was set to 30 Hz, which is reported to be appropriate from muscle activation. Assessment methods 2.4..1. Pain level. In order to determine the pain level depending on the intervention programs, 4-item visual analogue scale (QVAS) was used. This scale consists 4 items that ask for: current pain level, average pain level experienced, pain level when it is least severe, and pain level when it is most severe. Each questions are scored from 0 to 10 where 0 means no pain at all and 10 with the most severe pain. This test has a high reliability of r = .76-.84. [21] 2.4.2. Functional ability. In order to determine the lumbar functional ability of the participants, Roland-Morris Low Back Pain and Disability Questionnaire (RMQD) was used. This assessment is a self reported questionnaire that has 24 items, where each items are scored wither 0 or 1, and the highest possible score is 24. Higher score means greater functional disability. RMQD is useful in explaining the levels of functional limitations. This assessment has a high reliability of r = .92. [22] Wii Balance Board (WBB) (Nintendo, Kyoto, Japan) was used to measure static balance ability of the participants of different intervention groups. On top of the platform shaped WBB, the participants' change in center of pressure is traced to calculate path length, velocity, and area. The data measured with WBB were collected using Balancia software (Balancia software, Mintosys, Korea). The interrater reliability of WBB is ICC = .92-.98 (Holmes et al, 2013), and Balancia program has high interrater reliability (r = .79-.96) and validity (r = .85-.96). [23] 2.4.3. Psychosocial factor. To determine the psychisocial factors of the participants, Korean Version of FABQ was used. This questionnaire has 16 items that are categorized into physical activity (PA) and work (W). In fear-avoidance response, FABQ-PA consists of 5 items with the score range from 0 to 24. FABQ-W has 11 items and the score ranges from 0 to 42. The reliability of this test is .95. [24] 2.4.4. Pulmonary function. Pulmonary function test was conducted using Microquark (COSMED, Roma, Italy) after entering sex, age, height, and weight of the participants. The pulmonary function test measured forced vital capacity (FVC), forced expiratory volume in 1 second (FEV1), forced expiratory volume in 1 second/forced vital capacity (FEV1/FVC), and maximum voluntary ventilation (MVV). In addition, Pony FX MIP/MEP (COSMED, Roma, Italy) was used to measure the changes in respiratory muscles. In a standing position, the participants positioned their legs shoulder width, holding the mouth piece with the mouth and nose closed with a clip. Maximum inspiratory pressure (MIP) and maximum

expiratory pressure (MEP) were measured in this position. After 3 light respirations, a total of 3 MIP and MEP were measured, and the highest values of each were used for analyzation. The measurement was performed 3 times by a skilled physiotherapists with many years of experience in pulmonary function test. Between each tests, 10-minute vreaks were given. Interrater reliability of pulmonary function test using Pony FX MIP/MEP is r = .99. [25] Statistical analysis All data collected through the interventions were analyzed using SPSS version 25.0 (IBM, Chicago, IL). General characteristics of the participants were determined with test of normality through Shapiro-Wilk test, and mean and standard deviations were analyzed through 1-way ANOVA and χ2 test. In order to investigate the amount of changes of the intervention effects before and after the intervention for each groups, paired t test was made and 1-way ANOVA was used to compared each groups. Statistical significance level (α) was set to .5. Results Among the participants, a total of 5 participants were excluded from the study where 2 participants dropped out due to pain aggrevation and 3 participants dropped out due to decondition. Ultimately, data were collected from 15 participants in the SE group, 14 from the SER group, and 14 from the SERW group. There was no statistically significant difference in the general characteristics of the 3 groups (Table 1), and the pretest outcome values of the 3 groups were homogeneous. QVAS Pain levels of the 3 groups showed significant decrease before and after the intervention, and SER and SERW groups showed significant decrease compared to the SE group (P < .05) ( Table 2). Functional ability Roland-Morris low back pain and disability questionnaire and balance ability of the participants showed significant increase before and after the intervention for all 3 groups (P < .05). When the 3 groups were compared, SERW and SER group showed a significant increase that the SE group in Roland-Morris low back pain and disability questionnaire (P < .05), and only the SERW group showed significant increase in balance ability(P < .05) ( Table 2). FABQ FABQ that shows the psychosocial level of the participants showed significant increase in all 3 groups (P < .05), but there was no significant increase in the amount of changes among the groups (Table 3). Pulmonary function FVC, FEV1, MVV, MIP, and MEP of the participants showed significant increase before and after the intervention in the SER and SERW group (P < .05), and in the differences among the groups, the SER group showed a significant increase in FVC, FEV1, MIP, and MEP (P < .05) ( Table 4). Discussion This study aimed to explore the effects of stabilization exercise programs depending on intervention methods on LBP patients with lumbar instability. The results demonstrate that all participants showed statistically significant differences (P < .05) in pain level, functional level, and psychosocial level. Additionally the SER group showed significant improvement in pain level, functional level and lung function (P < .05), and the SERW group showed improvement in pain level, functional level, and balance ability. Lumbar stabilization exercise controls the pressure applied to the lumbar area through dynamic movements, and ultimately improves pain and increases functional activities. [26] Park and Lee [11] studied the effects of stabilization exercise with respiratory resistance in patients with LBP, and demonstrated that there was a significant difference in pain level (P < .05, d = 2.74). When Zheng et al [18] provided stabilization exercise with whole body vibration, there was a significant difference in maintaining postures and pain level (P < .05, d = .52). This study also demonstrated significant increase (P < .05) in pain level before and after the intervention in all 3 groups (SE group (d = .33), SER group (d = .90), and SERW group (d = .91)) and it was consistent with the existing studies. There was no significant difference among the 3 groups, but SER and SERW groups showed a greater effect size compared to the SE group. Stabilization exercise with whole body vibration may have reduced pain from the increased control ability of the deep muscles by increasing proprioception during whole body vibration. When respiratory resistance was added to the stabilization exercise, increase in abdominal pressure from strong contraction in the abdominals may have contributed to pain reduction. As LBP is a general health issue, it limits many activity performances due to pain and dysfunction, thereby resulting in poor quality of life. [27] Yang et al [28] argued that pain and dysfunction Mean ± standard deviation. SE = stabilization exercise, SER = stabilization exercise program using respiratory resistance, SERW = stabilization exercise program using respiratory resistance and whole body vibration. Table 2 Comparison of pain level, functional ability before and after intervention between groups. This study assessed lumbar dysfunction along with pain level. Calculating the score percentage in relation to the total score of 24, the SE group showed a dysfunction decrease from 88% to 54%, the SER group from 89% to 42%, and the SERW group from 88% to 42%, suggesting that lumbar dysfunction level has decreased significantly (P < .05). In addition, the SER and SERW groups showed a statistically significant increase (P < .05) compared to the SE group. The SER group may have shown increased intramedullary and intraabdominal pressures from the strengthening of deep abdominal muscles and diaphragm through respiratory resistance training. Furthermore, it may have affected respiratory function, pain, and ADL performance, thereby affecting functional limitations in the lumbar area. In the SERW group, it may be suggested that pain and functional limitations have decreased through training on unstable surfaces with whole body vibration, where this method may have increased proprioception and neuromuscular control ability. Park and Lee [29] reported that with the onset of LBP, psychological changes occur and this change affects physical and work-related activities. Choi, et al [2] reported that pain level and FABQ have significant correlations. Therefore this study used Korean Version of FABQ to investigate the psychosocial factors of the participants before and after the intervention. The results showed that all participants have significantly decreased phychological anxiety (P < .05), but there was no significant difference among the 3 groups. All participants have performed the same Table 3 Comparison of psychosocial level before and after intervention between groups. Mean ± standard deviation FABQ = Korean version of fear-avoidance beliefs questionnaire, PA = physical activity, SE = stabilization exercise, SER = stabilization exercise program using respiratory resistance, SERW = stabilization exercise program using respiratory resistance and whole body vibration, W = work. *P < .05. Table 4 Comparison of pulmonary function before and after intervention between groups. stabilization exercises, and with the randomized assignment to different groups, not knowing which group they were assigned to, and stabilization exercise that avoided direct movement of the lumbar area may have affected psychological factors. In order to maintain balance stabilization exercise on unstable surface requires co-contraction of many muscles that go through body joints. [29] This study measured and compared the variables of center of pressure velocity, length and area to determine balance abilities of the participants before and after the interventions. As a result, all 3 groups showed significant increase (P < .05) in all variables before and after the interventions. In addition, there was difference between the 3 groups showed a significant difference only in the SERW group that combined respiratory resistance and whole-body vibration (P < .05). It is consistent with the results of previous studies that there was a significant improvement in balance ability by performing neuromuscular stabilization exercise using wholebody vibration for LBP patients (P < .05). [14] Rittweger, et al [30] reported that frequency lower than 20Hz may cause excessive relaxation of the muscles and frequency exceeding 50 Hz may cause muscle pain, thereby suggesting the frequency to be set between 20 and 50 Hz. This study used 30 Hz for the intervention as suggested by Di Giminiani, et al [20] in their study, where they reported that this frequency induces the greatest muscle activation. Since whole body vibration exercise implements a novice form of stimulation to the muscular system, it not only provides additional neural adaptation, but also improves balance abilities by giving positive effects in muscle activation when vibration is given. Many surrounding muscles are speculated to be activated in motor units rather than a single muscle activation to maintain balance against external resistance through whole body vibration. This study showed a significant increase in FVC, FEV1, MVV, MIP, and MEP before and after the intervention in the SER and SERW group (P < .05), and in the differences among the groups, the SER group showed a significant difference in FVC, FEV1, MIP, and MEP (P < .05). It is thought that, in the SER group, high-intensity respiratory resistance training with breathing resistance in addition to stabilization exercise led to strengthening of the diaphragm, and thus focused on training for breathing. However, in the SERW group, it is thought that the whole body vibration training in parallel with respiratory resistance applied various movements along with vibration stimulation of the lower extremities, so it is thought that only the breathing resistance could not be focused. Stabilization exercise with respiratory resistance may have applied pressure on the supplementary inspiration muscles and diaphragm, enhance internal pressure of the chest and trunk muscle activation during respiratory resistance, and increase spinal stabilization to ultimately enhance pulomary function, MIP, and MEP. While conducting this study, 1 to 2 days of preceding training was required for participants in stabilization exercise with whole body vibration group. In addition, the mouth piece of respiratory resistance apparatus is made with silicon material. The participants in SER and SERW group complained about the intervention when there was an exposure of saliva during training, creating nauseating sensations. Furethermore, this study has some limitations. First, the intervention period was not long enough to consistently compare and analyze the effects after the intervention. Secondly, the age range of the participants were limited to 30's, making it difficult to generalize to all LBP patients of all ages. These limitations must be addressed in studies determining the long term effects of lumbar stabilization exercise in the future. Conclusion This study aimed to explore the effects of stability exercise programs depending on intervention methods for patients with lumbar instability. The results showed that pain and dysfunction decreased with increased pulmonary functions in the SER before and after the interventions, and decreased pain and dysfunction with increased balance ability in the SERW group. In current clinical settings, various methods of stabilization exercises are facilitated to patients with lumbar instability. If stabilization exercise program using respiratory resistance and whole-body vibration administered according to the purpose of intervention methods may be effective exercise programs for people with lumbar instability. Engineering and Design of Chimeric Antigen Receptors T cells engineered with chimeric antigen receptors (CARs) have emerged as a potent new class of therapeutics for cancer, based on their remarkable potency in blood cancers. Since the first clinical reports of their efficacy emerged 7 years ago, investigators have focused on the mechanisms and properties that make CARs effective or toxic, and their effects on T cell biology. Novel CAR designs coupled with improvements in gene transfer technology, incorporating advances in gene editing, have the potential to increase access to engineered cell therapies, as well as improve their potency in solid tumors. T cells engineered with chimeric antigen receptors (CARs) have emerged as a potent new class of therapeutics for cancer, based on their remarkable potency in blood cancers. Since the first clinical reports of their efficacy emerged 7 years ago, investigators have focused on the mechanisms and properties that make CARs effective or toxic, and their effects on T cell biology. Novel CAR designs coupled with improvements in gene transfer technology, incorporating advances in gene editing, have the potential to increase access to engineered cell therapies, as well as improve their potency in solid tumors. Adoptive transfer of T cells collected from autologous peripheral blood and engineered to express chimeric antigen receptors (CARs) has produced impressive clinical responses in patients with hematologic malignancies. 1 Together with checkpoint blockade therapy, 2 CAR T cells are revolutionizing the field of cancer therapies, providing hope to patients with previously refractory cancers. [3][4][5][6][7][8][9] CAR T cells targeting CD19 were

recently approved by the US Food and Drug Administration (FDA) and the European Commission (EC) for the treatment of relapsed or refractory acute lymphoblastic leukemia (ALL) in pediatric and young adults (Kymriah), and for the treatment of relapsed or refractory diffuse large B cell lymphoma in adults (Kymriah and Yescarta). Emerging clinical data are proving the potential of CD22 and B cell maturation antigen (BCMA)-directed CAR T cells to eradicate leukemia and multiple myeloma, respectively, heralding a new era for the treatment of blood cancers. 10,11 Although the human T cell is a highly potent anti-tumor agent, several challenges remain for successfully applying CAR T cells to solid tumors. First, the potency of these "living drugs" can lead to lethal on-target toxicity if the target is expressed on a life-sustaining tissue. 12 Second, active proliferation of the CAR T cells in the body can itself lead to potentially serious side effects, such as cytokine release syndrome 13 and neurologic toxicity, 14 although these are typically transient and reversible . Third, although finding a single uniquely expressed tumor antigen is already a challenge, 15 targeting only one antigen may be insufficient to obtain sustained responses, because antigen loss has been identified as a frequent cause of tumor resistance to CAR T cell therapies. 3,10,16 Fourth, in order to be effective in solid tumors, CAR T cells will have to overcome multiple challenges, including an immunosuppressive tumor microenvironment. Recent clinical reports suggest that CAR T cells can engraft in the peripheral blood, traffic to solid tumors, and respond to antigen. [17][18][19][20] We define engraftment as the ability for CAR T cells to be detectable in peripheral blood at any time after infusion. Moreover, evidence of transient antitumor activity has been observed in patients with difficult-to-treat tumors. 21,22 However, with rare exceptions, clinical responses in patients with solid tumors have been minor and transient, highlighting the importance of optimizing T cell therapeutics to tackle the solid tumor challenge. Finally, simplified methods to culture and engineer T cells are required to lower the costs, accelerate development, and improve access to this novel treatment. In this review, we discuss the importance of fine-tuning each of the modules that form a CAR as an initial step to improve T cell specificity, antigen recognition, and T cell function. Second, we will summarize the current and novel tools that are being used to engineer CAR T cells, including recent developments in gene-editing technology that allow researchers to edit specific genes within the human genome, and even knock-in transgenes in specific integration sites. Finally, we will provide a comprehensive overview of exciting strategies that are being developed to generate new synthetic receptors with enhanced function and reduced toxicities. For simplicity, this review focuses on modification of ab-T cells with CARs. Substantial effort has been invested in identifying the best T cell subsets for adoptive cell transfer. Moreover, strategies to introduce CARs into other immune cell types, including gd-T cells, natural killer cells, natural killer T cells, and macrophages have garnered growing attention. CAR design will need to be tailored to each cell type. The pursuit of the ideal cell type for redirection with CARs is beyond the scope of this review and has been covered elsewhere recently. [23][24][25] domain that anchors the CAR to the cell membrane, and one or more intracellular domains that transmit activation signals. Depending on the number of costimulatory domains, CARs can be classified into first (CD3z only), second (one costimulatory domain + CD3z), or third generation CARs (more than one costimulatory domain + CD3z). Introduction of CAR molecules into a T cell successfully redirects the T cell with additional antigen specificity and provides the necessary signals to drive full T cell activation. Because antigen recognition by CAR T cells is based on the binding of the target-binding single-chain variable fragment (scFv) to intact surface antigens, targeting of tumor cells is not MHC restricted, co-receptor dependent, or dependent on processing and effective presentation of target epitopes. In this section, we describe the properties and function of each of the modules that compose CARs. Despite intensive research efforts to define optimal CAR design, a universally improved CAR structure has not yet been discovered. As of now, each CAR construct needs empirical testing for evaluation, and several studies indicate that small modifications can have major consequences on the therapeutic outcome. scFv The antigen-binding capability of the CAR is defined by the extracellular scFv, not the targeted antigen. For instance, the scFv of both of the approved anti-CD19-targeted CARs, Kymriah and Yescarta, is derived from the murine anti-human CD19 antibody FMC63, which targets a specific epitope of CD19 found in exon 4 of the CD19 gene. 26 Other anti-CD19 CARs may recognize different epitopes of CD19, or the same immunodominant epitope but with differential affinities, such as the scFv derived from SJ25C1, 14 which is in the JCAR015 product. Investigators targeting MUC1 with CARs have used scFvs derived from antibody clones HMFG2, SM3, and 5E5, among others. HMFG2 and SM3 both recognize the PDTR epitope of the MUC1 variable tandem repeat; however, MUC1 is a highly glycosylated protein, especially within these tandem repeats, and SM3 only recognizes the deglycosylated protein, whereas HMFG2 can recognize deglycosylated or fully glycosylated protein. 27 CAR T cells targeting MUC1, either utilizing SM3 scFv and co-expressing interleukin-12 (IL-12) or utilizing an HMFG2-based scFv, were employed in a phase I clinical trial to treat metastatic seminal vesicle cancer. HMFG2based CAR T cells caused tumor necrosis, whereas SM3-based CAR T cells did not seem to have efficacy. 28 In contrast, CARs using the 5E5 scFv recognize the GSTA epitope of the MUC1 repeat only when glycosylated with Tn antigen, an abnormal glycoform found in many epithelial cancers. 29 Because glycosylation of tumor-associated antigens can be modulated by cell type or oncogenic transformation, this difference in epitope binding can define the target of CAR T cells and potentially discriminate between normal tissue expressing a TAA and a tumor cell expressing the same TAA. In fact, changes to the tumor cell surface glycome can affect cell signaling, affect cellmatrix and cell-cell interactions, and influence immune modulation and the cellular metastatic potential. 30 Investigating the changes of the tumor membrane glycolipids and glycoproteins, which are encompassed in a broader view of the tumor "surfaceome," can provide additional targetable avenues for cancer immunotherapy. The format of an scFv is generally two variable domains linked by a flexible peptide sequence, either in the orientation VH-linker-VL or VL-linker-VH. The orientation of the variable domains within the scFv, depending on the structure of the scFv, may contribute to whether a CAR will be expressed on the T cell surface or whether the CAR T cells target the antigen and signal. 31 In addition, the length and/or composition of the variable domain linker may be an important contribution to the stability of the scFv; in studies generating scFvs from a TAG72 antibody (clone B72.3), linkers up to 6xGGGGS demonstrated higher molecular weight dimers and multimers, with clustering decreasing with increasing linker length. 32 If these scFvs are utilized in CAR molecules, this clustering may lead to antigen-independent signaling, or "tonic signaling." Another linker of a TAG72based scFv (clone CC49) designed to enhance proteolytic stability (Whitlow "218" linker: GSTGSGSKPGSGEGSTKG) had the effect of enhancing scFv affinity, 33 which may alter CAR T cell behavior as well. Two studies have demonstrated that variance of CAR scFv affinities, either through mutagenesis of complementary-determining regions while holding the epitope constant, or through CAR development with scFvs derived from therapeutic antibodies against the same target, but not the same epitope, can change the strength of the T cell signal and allow CAR T cells to differentiate overexpressed antigens from normally expressed antigens. 34,35 In fact, researchers demonstrated that trastuzumab-based CAR T cells with reduced scFv affinity could degranulate in response to HER2 + breast cancer cells without reactivity against normal primary human cells that express HER2. 34 These results suggest that the scFv, a critical component of a CAR molecule, can be carefully designed and manipulated to influence specificity and differential targeting of tumors versus normal tissues. Given that these differences may only be measurable with CAR T cells (as opposed to soluble antibodies), pre-clinical testing of normal tissues for expression of the target, and susceptibility to on-target toxicities, requires live-cell assays rather than immunohistochemistry on fixed tissues. Hinge The hinge, also referred to as a spacer, is the extracellular structural region of the CAR that separates the binding units from the transmembrane domain (Figure 1). With the exception of a few CARs based on the entire extracellular moiety of a receptor, such as NKG2D, 36 the majority of CAR T cells are designed with immunoglobulin (Ig)-like domain hinges. These spacers generally supply stability for efficient CAR expression and activity. The hinge also provides flexibility to access the targeted antigen. Several studies have demonstrated that the optimal spacer length of a given CAR depends on the position of the targeted epitope. 37 Long spacers provide extra flexibility to the CAR and allow for better access to membrane-proximal epitopes 37,38 or complex glycosylated antigens. 39 By contrast, CARs bearing short hinges are more effective at binding membrane-distal epitopes. 40,41 It has been proposed that the length of the spacer is crucial to provide adequate intercellular distance for immunological synapse formation, 42 highlighting the need to optimize spacers for individual epitopes accordingly. Hinges can affect the overall performance of CAR T cells. Human IgG-derived spacers consisting of two Ig-like domains, CH2 and CH3, have been used for efficient antigen recognition and are especially useful in assessing the level of CAR expression at the surface of the T cell. However, several studies indicate that IgG-based spacers retain their capacity to bind to Fc gamma receptor (FcgR) through their CH2 domain even after they are incorporated into CARs. This feature leads to off-target activation by myeloid and lymphoid cells expressing FcgRs, 43 and correlates with off-tumor localization and poor engraftment in animal models. 41,44,45 Point mutations to ablate the FcgR binding domain and spacers trimmed down to a single CH3 domain have improved CAR T cell engraftment and anti-tumor efficacy in in vivo experiments. 41,45,46 Alternatively, extracellular domains naturally lacking FcgR binding activity, such as CD28 and CD8a, have been extensively used as spacers. 47,48 In fact, these two spacers are found in Yescarta and Kymriah, respectively. Transmembrane Domain The transmembrane domain consists of a hydrophobic a helix that spans the cell membrane and is probably the least characterized region of the CAR. Although the main function of the transmembrane is to anchor the CAR in the T cell membrane, some evidence suggests that the transmembrane can be relevant for CAR T cell function. The nature of the CAR signaling is not well understood; however, if CARs behave similar to TCRs, it is likely that CARs need to dimerize or associate with other accessory molecules for proper signaling initiation. In that regard, one study demonstrated that first generation CARs need specific regions within the CD3 transmembrane for CAR dimerization or incorporation into endogenous TCR clusters. 49 Second and third generation CARs have been designed with numerous single-span transmembranes, mainly derived from CD4, CD8a, or CD28. 47,50,51 A recent study demonstrated that when using CARs containing the inducible T cell costimulator (ICOS) intracellular domain, incorporation of the ICOS transmembrane, instead of the extensively used CD8a transmembrane, was required for increased CAR T cell persistence and overall anti-tumor efficacy. 52 In another study, Morin et al. 53 analyzed the function of the extracellular and transmembrane domains of the endogenous CD28 molecule using a transgenic mouse that lacks the intracellular tail of CD28. Interestingly, the authors show that the extracellular and transmembrane domains of CD28 can partially induce T cell activation. 53 Altogether, these results suggest that the transmembrane domain of certain costimulatory molecules can be involved in synapse formation or T cell signaling. Linking the proximal intracellular domain to its corresponding transmembrane domain may enable proper CAR T cell signaling, while using the widely used CD8a or CD28 transmembrane domains may favor CAR expression or stability. Costimulatory Domains: The Art of Signaling Substantial effort at CAR T cell engineering has been invested in understanding the effects of CAR costimulation with the aim of identifying an optimal endodomain to include in CAR constructs.

CAR endodomains, or intracellular domains, are usually derived from costimulatory molecules from the CD28 family (including CD28 and ICOS) or the tumor necrosis factor receptor (TNFR) family of genes (including 4-1BB, OX40, or CD27). The CD28 family signals through the phosphatidylinositol 3-kinase (PI3K)-Akt pathway, with ICOS inducing a stronger PI3K activation than CD28. 52,54 Unique to CD28 is the recruitment of Grb2 and Lck to its intracellular tail, which results in high levels of IL-2 production, 55 as well as the recruitment of Itk. The TNFR family members signal through recruitment of TRAF proteins 56 and are implicated in modulating T cell proliferation, differentiation, and survival. [57][58][59] CD28 and 4-1BB are the most widely used costimulatory endodomains in CARs. Clinical trials with CARs incorporating CD28 or 4-1BB intracellular domains showed similar response rates in patients with hematologic malignancies. However, the persistence of T cells engineered with these two CAR designs is strikingly different. Preclinical studies identified these T cell persistence differences in head-to-head comparisons of CD28-and 4-1BB-based CAR T cells in animal models. 48,60 Clinical trials for B cell malignancies have shown that CD28-based CAR T cells are typically undetectable beyond 3 months, 6,61 whereas 4-1BB-based CAR T cells can persist . Two Families of Hinges IgG-based hinges, derived from IgG1, IgG2, or IgG4, provide flexibility to the scFv to target membraneproximal epitopes. However, binding of the CH2 domain to Fc receptors in myeloid cells can impair functionality. Modification or deletion of the CH2 domain can restore CAR T cell functionality. CARs with shorter hinges, including an IgG-derived hinge lacking the CH2-CH3 regions, or hinges derived from native CD28 and CD8 hinges, can be used to target membrane-distal epitopes. in patients for several years after treatment. 62 Exhaustive studies indicate that signaling through CD28-based CARs results in more rapid T cell activation, proliferation, cytolysis, and increased glycolysis, but shorter T cell persistence. By contrast, 4-1BB signaling induces a slower T cell effector response and promotes mitochondrial biogenesis, greater oxidative metabolism, and sustained T cell persistence. 52,63-65 The high effector function and self-limited expansion of CD28-based CARs may be ideal to transiently treat diseases with a rapid tumor elimination and short-term persistence of the CAR (i.e., as a bridge therapy for allogeneic hematopoietic stem cell transplantation). By contrast, 4-1BB-based CARs may be used to treat diseases in which complete responses require sustained T cell persistence. 62 Similar to 4-1BB, CD27 signaling has been shown to enhance CAR T cell survival when compared with CD28. 66 Incorporation of ICOS into a CAR drives CD4 + T cells toward a Th1/Th17 phenotype, enhancing T helper functions and increasing in vivo T cell persistence when compared with the widely used CD28 and 4-1BB intracellular domains. 67 Recent data suggest that various lymphocyte subsets require distinct costimulation signals for optimal function and persistence. For instance, whereas the ICOS intracellular domain enhances the persistence of CD4 + CAR T cells, the 4-1BB intracellular domain provides optimal persistence in CD8 + T cells. 52 One option to join the properties of different intracellular domains in one single T cell is to combine two intracellular domains in a third generation CAR. Typically, such combinations include one intracellular domain from the CD28 family and one intracellular domain from the TNFR family, resulting in the simultaneous activation of different signaling pathways. 52,60,68,69 Third generation CARs are consistently expressed at lower levels when compared with second generation CARs, but they usually show improved effector functions. Optimization of these more complex receptors, including the position of each intracellular domain in relation to the cell membrane, is required to attain complete synergy of the combination. 52 Altogether, these findings indicate that the nature of costimulation during CAR T cell activation critically regulates the metabolism, survival, and effector function of T cells. Each costimulatory domain has unique properties, and it is unlikely that an optimal costimulatory domain will serve all purposes. Differences in the affinity of the scFv, the intensity of antigen expression, the probability of off-tumor toxicity, or the disease to be treated may influence the selection of the intracellular domain. Further understanding of the molecular signaling effects of CARs may help in choosing the right intracellular domain or combination of intracellular domains for each condition. Ultimately, carefully designed clinical trials may be required to understand the effects of costimulation on the success of CAR T cell therapies. Avoiding Tonic Signaling Although the main purpose of engineering T cells with CARs is to drive activation upon antigen-specific recognition in tumors, some CAR constructs can also induce antigen-independent activation, usu-ally referred to as tonic signaling. High density of CARs at the T cell surface, usually driven by strong constitutive promoters, can result in clustering of CAR molecules that leads to chronic T cell activation. Tonic signaling can be detected during in vitro primary T cell expansions and is characterized by differences in the growth patterns, phenotype, and function of CAR T cells when compared with control T cells. The level and consequences of tonic signaling are highly dependent on the selection and configuration of CAR modules, but they usually include chronic T cell activation with accelerated differentiation and exhaustion, and impaired anti-tumor effects. 52,70 Moreover, tonic signaling in cMET-specific CARs containing the CD28 signaling domain can result in continuous in vitro proliferation, with constitutive secretion of diverse cytokines, including IL-2. 71 In a GD2-specific CAR, substitution of the CD28 intracellular domain for the 4-1BB intracellular domain ameliorated the T cell exhaustion caused by tonic signaling. Although low levels of tonic signaling in 4-1BB-based CARs may be beneficial because of enhanced T cell persistence, 48 recent reports indicate that in some cases, 4-1BB-driven tonic signaling may result in augmented T cell apoptosis, impaired in vitro expansion, and progressive downregulation of CAR expression. 52,72 The level of tonic signaling can also depend on the length of the hinge, 45 the expression system, 52,71 and the vector used. 72 In third generation CARs, tonic signaling may be mitigated by placing the 4-1BB intracellular domain distal to the cell membrane. 52 Tonic signaling is not usually assessed systematically in most CAR reports. However, emerging evidence underscores the importance of avoiding the use of CARs with configurations that may induce this chronic activation. Mitigating tonic signaling may require testing different combinations of scFvs, hinges, and intracellular domains. CAR aggregation can be attenuated by using alternative promoters to reduce the levels of CAR expression; however, this can also result in impaired efficacy. 52 A recent report suggests that knocking in the CAR into the TRAC locus through homology-directed recombination averts tonic CAR signaling by optimizing CAR internalization and re-expression kinetics. 73 Genetic Engineering Strategies Viral-and non-viral-based genetic engineering tools have been used to generate CAR T cells, resulting in permanent or transient expression of therapeutic genes (Table 1). However, most of the current clinical trials developing modified T cells employ the latest generations of retroviral (gamma retroviral and lentiviral) vectors. Retrovirus-based gene delivery is a mature, well-characterized technology, which permanently integrates the CAR into the host cell genome. Although serious adverse events were linked to retroviral gene transfer in CD34 + stem cells, 74,75 mature T cells have been safely modified with this framework now for decades with hundreds of treated patients, resulting in stable engraftments and without cell transformation. 76,77 The downsides of this approach are the cost associated with clinical-grade production and the limited amount of genetic material that can be included in the vector. A recent study using lentiviral vectors in patients with ALL reported that unintentional transduction of a single leukemic B cell during the CD19-CART manufacturing resulted in resistance to CAR T cell therapy through masking of the CD19 epitope. Although this is a rare event, this finding highlights the importance of properly isolating T cells from leukemic cells before redirection. 78 Non-viral DNA transfection transposition systems have also been used for permanent expression of therapeutic payload. Sleeping Beauty (SB) transposon is the most developed system of its kind to engineer CAR T cells, 79 complying with current good manufacturing practice (cGMP), 80 and is currently under investigation in the clinic (ClinicalTrials.gov: NCT00968760 and NCT01653717). Multiple SB enzymes 81,82 have been used to deliver more than one transgene from a multi-cistronic single plasmid 83,84 or multiple plasmids. 85 Others have developed CAR T cells using the piggyBac transposon system, [86][87][88] which can integrate larger transgenes. 89,90 In general, transposon technology might be easier and more economical to produce than viral vectors. However, the longer expansion protocols currently employed may result in T cell differentiation, as well as impaired activity and poor persistence of the infused cells. 91 In this regard, recent developments using minicircle vectors might help address these issues because of higher efficiency integrations. 92 Non-integrative gene transfer, such as RNA transfection, enables transient expression of the CAR for up to 1 week. This strategy has been used to evaluate potential toxicities or to limit the side effects of the therapy. 20,93 However, the weaknesses of a transient CAR expression are that the transgene is rapidly lost, and patients will require multiple infusions for clinical effect. Alternatively, non-integrative CAR delivery, such as episomal gene transfer, may result in longer-term, but not permanent, expression. 94 Finally, in order to simplify and reduce the cost of CAR T cell manufacturing, alternative approaches using in situ T cell modification are being explored. In situ T cell programming using DNA nanocarriers 95 or lentiviral vectors specifically targeting human CD8 + T cells 96 have been proved safe and efficacious in pre-clinical models of B cell malignancies. Gene Editing: The New Era Improvements to site-directed nuclease gene-editing tools are unveiling a new era for CAR T cell therapy. Most efforts at T cell gene editing thus far have been focused on developing off-the-shelf universal CAR T cells from allogeneic healthy donors. The majority of current clinical trials use autologous CAR T cells, because use of allogeneic CAR T cells could induce graft-versus-host disease (GVHD) due to TCR-mediated recognition of the recipient alloantigens. The immune system of the recipient could also attack the infused human leukocyte antigen (HLA)-disparate CAR T cells, causing rapid T cell rejection. In order to prevent GVHD, different groups have focused on removing the TCR from allogeneic CAR T cells using genome editing. Infusion of gene-edited universal CD19-CART cells in two pediatric patients with ALL induced complete responses in both patients, proving the feasibility of this approach. 97 Strategies to prevent CAR T cell rejection by the recipient are also being explored and include the elimination of HLA class I expression by editing the HLA-A or b 2 -microglobulin genes. Gene editing is also being explored to selectively remove the expression of T cell-suppressive genes. 98-100 Checkpoint inhibitor PD-1 is probably the most prominent gene currently targeted, with PD-1 knockout anti-MUC1 CAR T cells entering clinical trials this year (ClinicalTrials.gov: NCT03525782). Recently, Ren and colleagues 99 set up a versatile system for editing multiple genes simultaneously in CAR T cells through a single-vector lentiviral system. Target-specific endonucleases have also been exploited to insert transgenes inside loci relevant for T cell biology (Table 1). Unlike random insertion strategies, this method may result in finely regulated CAR expression independent of constitutive promoters. 73,101 In a landmark study, Eyquem et al. 73 demonstrated that targeting the CAR insertion into the TCR locus results in more uniform expression and increased potency over conventionally generated CAR T cells. Different endonucleases have been used to knock out genes in T cells, including zinc-finger nucleases (ZFNs), 102 TALENs, 97,103,104 and Cas9. 98,99 For knock-in approaches, the transgene can be supplied from donor DNA using non-integrative viral vectors, such as adeno-associated viral vector type 6 (AAV6), 73,101 or non-viral approaches, such as single-stranded DNA. 105 The enzymes are typically electroporated into T cells as mRNA or protein. 98,99 More recently, a study using the CRISPR/Cas9 system has demonstrated that donor DNA can be electroporated as endonuclease-complexed and long linear double-stranded DNA (dsDNA), which is otherwise toxic, yielding relatively high T cell viability and transgene insertion efficiencies. 105 This system avoids the use of viral vectors, which is expensive and time-consuming to generate and test, and could therefore substantially reduce the cost and time for gene editing in T cells. Overall, this report underscores the potential to develop next generation CAR T cell

therapies; however, further protocol optimization will be needed to achieve large-scale manufacturing of gene-edited CAR T cells. Finally, a better understanding of the potential deleterious effects of genome editing is required to ensure that benefits derived from further modifications in CAR T cells are greater than risks. In this regard, whole-genome sequencing of CRISPR/Cas9 and TALEN-targeted cell lines revealed a low incidence of off-target mutations. 106 Recent studies suggest that Cas9 can induce a p53-mediated toxicity in certain cell types, which can lead to a selection of p53 mutant cells more tolerant to DNA damages. 107,108 Careful monitoring of the p53 status in gene-edited cells together with efforts to enhance nuclease specificity may help reduce the potential risks of genome editing in CAR T cells. The Next Generation of CARs Several new approaches to enhance the therapeutic outcome of CAR T cells are being explored based on the lessons learned in recent clinical trials and fueled by rapid advances in synthetic biology. These approaches are mainly focused on improving T cell activation and persistence, overcoming tumor immunosuppression, addressing tumor escape, mitigating toxicities, and increasing access. CAR T cell therapies incorporating costimulatory molecules, inhibitory-to-costimulatory switch signals, cytokines, or safety switches have been developed to address these challenges and have been reviewed elsewhere. 23,[109][110][111] Here, we will highlight some of the innovative strategies based on new antigen receptors able to recognize a combination of antigens in an effort to avoid antigen escape or reduce toxicity ( Figure 2). Limiting Tumor Escape A major challenge for CAR T cell therapies is tumor escape due to antigen loss in the setting of inherent tumor heterogeneity and plasticity. Loss or decreased expression of the targeted antigen after CAR T cell therapy and outgrowth of antigen-negative cells after treatment has been observed in several clinical studies, 10,17,18,112,113 highlighting the importance of simultaneously targeting more than one antigen. One approach to overcome this obstacle is to express two different CARs in one single cell in order to target two different antigens, ideally in one single bicistronic vector to ensure dual-antigen targeting in all cells, an approach known as "dual CARs." 114 Expression of two CAR molecules in one viral plasmid may require codon optimization of duplicated DNA sequences (i.e., CD3z) to reduce the chances of DNA recombination. Another alternative is to design a receptor fusing two scFvs with different specificities with a single intracellular module in a tandem CAR. 10,115,116 Tandem CARs have the advantage of a smaller transgene size when compared with dual CARs, which is especially important when other transgenes (i.e., safety switches or cytokines) are also incorporated into the CAR plasmid. The main disadvantage is that in order to achieve optimal antigen recognition and full T cell activation, tandem CARs require optimization of the spacer length and linker sequence between the two scFvs, as well as the orientation of scFvs, which is more complex than with second generation CARs. 117 However, computational modeling methods can help predict the structure of tandem CARs, making design and development of these receptors more feasible. 115 Dual and tandem CARs are known as OR-gate CARs, as binding of either CAR to its cognate antigen is enough to drive full T cell activation. Interestingly, enhanced T cell function is observed when both targets are present, which makes them more efficient at inducing anti-tumor responses than a pooled combination of CAR T cells. 114,116,118 One of the main limitations of OR-gate CARs is that targeting two antigens may result in increased risk for toxicity, especially when treating solid tumors. Ongoing clinical trials are currently testing dual (ClinicalTrials.gov: NCT03330691) and tandem (ClinicalTrials.gov: NCT03185494, NCT03097770, and NCT03019055) CAR T cells simultaneously targeting CD19 and CD22 or CD20 for the treatment of leukemia and lymphoma. Minimizing Toxicities Lack of truly tumor-specific antigens is a major limitation for the development of CAR T cells. Most of the antigens currently targeted with CAR T cells are also expressed in normal cells, where even low levels of antigen expression can potentially result in severe on-target off-tumor toxicities. 12,119,120 Novel strategies in which engineered T cells require a specific combination of two antigens for activation are being explored to prevent toxicities. In a first approach in this direction, T cells were engineered to co-express two different CARs, one containing signal 1 (CD3z) and the other containing signal 2 (costimulation). 121,122 In this split-signal approach, recognition of a single antigen, which may happen in normal cells, leads to suboptimal T cell activation and limited killing of normal cells. In tumors, where both antigens may be expressed, full T cell activation is achieved after engagement of both scFvs to their cognate antigens. 122 In a different approach, the activation of CAR T cells can be controlled in situ by the addition of a small molecule. "ON-switch" CARs are designed as fragmented CAR receptors, where the extracellular antigenbinding module is dissociated from the intracellular signaling components. 123 Assembly of these two receptors in the presence of a heterodimerizing small molecule, such as tacrolimus-based analogs, leads to CAR T cell activation. The magnitude of responses depends on the dosage of the drug, which allows for titratable control. A more recent strategy involves an AND-gate recognition system in which activation of one receptor (synthetic Notch) induces the expression of a second receptor (CAR) that induces T cell activation following antigen recognition. 124 Full T cell activation and tumor elimination occur only when both antigens are expressed. However, tumor escape of CAR T cells through loss of the first antigen targeted by the synthetic Notch receptor is a major limitation of this strategy. A distinct class of strategies to limit toxicity includes the co-expression of a classic CAR with an antigen-specific inhibitory CAR (iCAR) bearing the signaling domain of an immunoinhibitory receptor (i.e., CTLA-4 or PD-1). 125 Engagement of iCARs to antigens in normal cells can constrain T cell functions, which can be resumed in the absence of the iCAR-targeted antigen and following antigen recognition by the activating CAR. One of the challenges of this approach is the identification of cell surface antigens that are expressed in normal cells but absent in tumor cells, a problem that is complementary to the obstacle of identifying tumor-specific antigens absent in normal tissues. Universal CARs: Increasing Versatility A different strategy that could simultaneously address antigen escape while mitigating toxicities is the utilization of so-called universal CARs. Instead of engineering T cells with fixed antigen specificities, these CAR therapies are composed of two parts: (1) an antibodybased molecule that recognizes a tumor antigen and is modified to express a "switch"; and (2) a universal CAR T cell without tumor specificity by itself, which contains a construct with an extracellular portion that binds to the switch and is linked to the intracellular signaling domains. In a first approach, universal CAR T cells were designed to bind to biotinylated antigen-specific molecules. 126 Other approaches incorporated the small molecule fluorescein isothiocyanate (FITC) or a short peptide in the antibody-based switch, allowing activation of the corresponding switchable CAR T cells after tumor recognition by the antibody-switch. 127,128 In a more recent approach, a split-CAR system named SUPRA CAR combines zipCAR T cell containing an extracellular leucine zipper with an scFv domain fused to a second leucine zipper (zipFv). 129 These versatile systems allow researchers to easily combine universal CAR T cells with zipFvs targeting different antigens to avoid tumor escape. At the same time, modulating the dose or affinity of the antibody-based switches may control toxicities. In the case of SUPRA CAR T cells, T cell activation can be prevented when necessary by addition of a competitive zipFv that can bind with high affinity to the zipFv with tumor specificity. 129 Conclusions The clinical development of CAR T cells has undergone tremendous growth over the past decade. However, this novel therapeutic modality is still in its infancy, and much remains to be learned and improved. Clinical trials evaluating CAR T cells in patients with both hematologic malignancies and solid tumors have revealed insights into the parameters that determine clinical outcome. Based on the lessons learned, a new arsenal of novel receptors and strategies has been developed and is poised to enter clinical trials. Preclinical studies have unveiled the importance of optimizing CAR constructs. The affinity and flexibility of the extracellular binding domain is key for improved tumor recognition, whereas the nature of costimulation during CAR T cell activation critically regulates the metabolism, survival, and effector function of T cells. A major challenge to the field is that the optimal CAR structure for each given antigen to be targeted remains empirical and involves the generation and testing of several constructs. As novel, more complex receptors emerge, optimizing each module may lead to an unrealistic number of constructs to test, both in animal models and in humans. In this regard, bridging the scientific divide between chimeric molecular engineering and protein structure identification and modeling may help predict which molecules would function optimally and reduce the number of constructs to evaluate. Identifying general rules to guide CAR design would also simplify this process. Also, a better understanding of the interactions of CAR proteins with endogenous T cell molecules and detailed mechanisms of CAR signaling would facilitate the design of more effective CAR T cell treatments. One major limitation, especially when treating solid tumors, is the absence of truly tumor-specific antigens that are expressed in tumor cells but absent in normal tumors. Expression of the targeted antigen in normal cells can result in severe toxicity, whereas absence or decreased antigen expression in tumor cells can lead to tumor escape. Thus, it is critical to explore the tumor surfaceome, including differential post-translational modifications of tumor-associated antigens, in order to identify new targets for antibody-based therapies, such as glycan-directed CAR T cells. 130 Strategies to optimize CAR specificity and reduce toxicity will be critical as we increase the number of tumor antigens to target. In this regard, recent technological advances in single-cell analysis tools and precision bioinformatics may help unveil additional differences between normal and tumor cells, 131 albeit with increasing tumor heterogeneity. Engineered T cells have shown their potential to dramatically change clinical outcomes in patients with B cell malignancies. Remarkable advances in the fields of molecular biology, immunology, gene editing, synthetic biology, and computational analysis provide new tools for CAR T cell development and are critical to achieving a greater therapeutic success when expanding this therapy to other cancer types. CONFLICTS OF INTEREST S.G., A.D.P., and M.V.M. are listed as inventors on various patents related to chimeric antigen receptors. Meta-analysis of incidence of early lung toxicity in 3-dimensional conformal irradiation of breast carcinomas Background This meta-analysis aims to ascertain the significance of early lung toxicity with 3-Dimensional (3D) conformal irradiation for breast carcinomas and identify the sub-groups of patients with increased risk. Methods Electronic databases, reference sections of major oncological textbooks and identified studies were searched for synonyms of breast radiotherapy and radiation pneumonitis (RP). Major studies in thoracic irradiation were reviewed to identify factors frequently associated with RP. Meta-analysis for RP incidence estimation and odds ratio calculation were carried out. Results The overall incidence of Clinical and Radiological RP is 14% and 42% respectively. Ten studies were identified. Dose-volume Histogram (DVH) related dosimetric factors (Volume of lung receiving certain dose, Vdose and Mean lung Dose, MLD), supraclavicular fossa (SCF) irradiation and age are significantly associated with RP, but not sequential chemotherapy and concomitant use of Tamoxifen. A poorly powered study in IMN group contributed to the negative finding. Smoking has a trend towards protective effect against RP. Conclusion Use of other modalities may be considered when Ipsilateral lung V20Gy > 30% or MLD > 15 Gy. Extra caution is needed in SCF and IMN irradiation as they are likely to influence these dosimetric parameters. Introduction Postoperative radiotherapy (RT) after the breast conservative surgery (BCT) or mastectomy, have been shown to reduce the rates of local recurrence and death in breast carcinomas (BC) [1][2][3][4][5]. It is the standard practice to offer patients adjuvant RT to whole breast or chest wall, with or without loco-regional RT (LRRT), depending on the stage of disease. The regional fields include the ipsilateral supraclavicular fossa (SCF). The Axilla (Ax) and internal mammary nodal region (IMN) RT is uncommon due to the toxicities associated with them [6][7][8][9][10]. Radiation pneumonitis

(RP) and lung fibrosis are two known toxicities that arise from incidental irradiation of adjacent ipsilateral lung in BC. Other toxicities include breast fibrosis, cardiac toxicity, skin toxicity and lymphoedema of the ipsilateral upper limb. The risk of cardiac toxicity in tangential radiotherapy treatment of left breast or chest wall is well studied in literature [6][7][8]. However most studies were pre-conformal CT-based planning and IMN irradiation was a common practice then. RP and lung fibrosis is believed to represent the different ends in the clinical course of the disease rather than separate entities altogether. The natural history of the radiation lung injury can be divided into 5 phases: immediate phase (hours to days); latent phase; acute exudative/clinical RP phase (4-12 weeks post-RT); intermediate phase with resolution of exudate and deposition of fibroblast and the final phase when fibrosis is established (usually 6-12 months post-RT) [11][12][13]. Type II pneumocytes which produce surfactant are the cells associated with RP [13,14]. Literature on RP in breast irradiation is very heterogeneous: different simulation techniques (conventional fluoroscopy-based versus CT-based), different treatment planning systems [2-Dimensional (2-D) versus highly conformal CT-based], different sites treated (chest wall/ breast +/− Ax/SCF/IMN) and also use of electrons to treat the chest wall +/− IMN after mastectomy. McDonald et al. [11] reported in 1995, the incidence of radiological and clinical RP to be in the range of 27 -40% and 0 − 10% respectively in patients with BC undergoing radiotherapy based on 2 studies in 1980. The details of the radiotherapy techniques and the grades of both radiological and clinical RP were not mentioned in that review. These figures are the most likely estimates of RP incidence in non-conformal planning days. Currently there is a trend towards minimizing toxicity to organs at risks in the adjuvant treatment of breast cancer. The newer RT modalities or techniques used include Intensity Modulated Radiotherapy (IMRT), Tomotherapy, Accelerated Partial Breast Irradiation (APBI) and Intraoperative Radiotherapy (IORT). IMRT and Tomotherapy produce more conformal radiation delivery at the expense of increased integral dose. These newer modalities are more expensive and needs longer planning time. Smith BD et al. [15] has recently shown that the adoption of IMRT for BC in the United States have increased the cost of breast irradiation significantly. In centers where there are limited resources or financial constraints, it is important to allocate the available resources accordingly without compromising the breast cancer patient's treatment outcome. Hence the important questions are: 1. Which patient subgroups will benefit from these newer modalities of treatment in regards to early lung toxicity? 2. What are the dosimetric parameters, treatment factors and patient factors which predict for RP or early lung toxicity in adjuvant 3-Dimensional Conformal RT (3D-CRT) for BC? Methodology Electronic databases were searched from 1995 till April 2011 (3D-CRT techniques would have been very unlikely before 1995) using the following inclusion and exclusion criteria for adjuvant RT studies in BC: Search was performed using PubMed search builder for Breast Radiotherapy AND Meta-analysis (or Systematic review); Breast Radiotherapy AND Toxicity; Breast AND Radiotherapy AND Pneumonitis. This produced 536 hits. Cochrane library, Google Scholar and reference sections of the major Radiation Oncology textbooks were searched to ensure no important studies were missed. Eight studies from these results were initially identified for the analysis and two further studies were identified by searching the reference sections of those 8 studies. Although four of the identified studies had overlapping data, different endpoints were used in the analysis [16][17][18][19]. Only one retrospective study was included [20]. Factors affecting the incidence and severity of RP were identified from the above studies and other thoracic irradiation studies (lung, oesophagus, lymphoma and others), which reported on RP. These were dosimetric parameters (V dose and MLD), patient factors (smoking and age) and treatment factors (SCF and IMN RT; concomitant Tamoxifen and sequential chemotherapy). Radiological RP grading criteria used include Arrigada, modified lung fibrosis CTC (common toxicity criteria) and Nishioka system [21]. The grouping of low grade radiological RP in this analysis are: Modified Arrigada grade 1 in Holli's study [19] and Arrigada scores 1-3 in Goldman's study [22]. Any grades or scores above these values are grouped as high grade radiological RP. The clinical RP is also divided into low and high grades with clinical RP grade 1 classified as low grade and clinical RP grade 2 or more grouped as high grade. Statistical methods Data was analysed using Stata® software, version 11.0 (Stata Corp College Station, TX, USA), and level of significance set at 5%. To estimate the incidences of clinical RP and radiological RP, we pooled the values from each study, using the random effects meta-analysis model. As for the association between treatment-related factors and RP, we combined the individual effect sizes (Odds Ratios) using the random-effects model, using the method of DerSimonian and Laird, with the estimate of heterogeneity being taken from the inverse-variance fixed-effect model. Heterogeneity between studies was assessed by the chi-square test for heterogeneity as well as examining the i-squared statistic, which quantifies the level of heterogeneity. In the event of heterogeneity, we used the random effect model instead of the fixedeffect model to analyse the data. Publication bias for the primary endpoint (incidence of RP) was assessed via the Egger's test. Radiological RP The overall incidence of radiological RP in our metaanalysis [ Figure 1] is 42% (95% CI = 22-62%) with large heterogeneity in the included studies (I-squared = 97.4%, p < 0.001) and Egger's test did not show significant publication bias (p = 0.151). In the study by Marco Krengli et al. [23], the high detection rate of RP (85%) could be attributed to the use of high resolution CT images for post radiotherapy lung assessment. The incidence of RP was low, as reported by Akiko Kubo et al. [24] because only tangential RT fields were used and the RP assessment was mainly done by Chest X-ray (CXR). Clinical radiation pneumonitis The overall incidence of clinical RP [ Figure 2] is 14% (95% CI = 8-21%) with large heterogeneity in the included studies (I-squared = 89.7%, p < 0.001). The Egger's test did not show significant publication bias (p = 0.376). Either CTC2.0 or CTC3.0 assessment was used in the studies. There are no reported cases of CTC grade 4 or 5 toxicity in any of the studies. The overall incidence of low grade and high grade clinical RP is 14% (95% CI = 9-18%) and 4% (95% CI = 2-7%) respectively. There is also large heterogeneity in the studies with I-squared = 79.3% (p = 0.002) and I-squared = 82.4% (p = 0.001) for low and high grade clinical RP respectively. The Egger's test for publication bias is not significant for either low or high grade clinical RP (p = 0.855 and 0.407). DVH-related parameters Volume of lung receiving a specified dose [V dose ] A systematic review by George Rodrigues et al. [25] in 2004 analysed 5 trials and reported a strong correlation between V dose and RP in lung tumour irradiations. In contast, V dose of the lung in breast or chest wall irradiation is likely to be smaller in value and hence we limit our analysis to studies reporting on V 20Gy to V 25Gy [ Table 1] though some studies do report on other V dose values. Berit Wennberg et al. [16] analyzed 121 patients who had node-positive stage II BC, which was a subgroup from the main study by Perh Lind et al. [17]. He presented the data in the form of mean cumulative ipsilateral lung DVHs for four different treatment techniques. Data extracted from this paper showed that treatments techniques with V 20Gy ≤ 20% had a lower incidence of RP compared to V 20Gy > 20% (12.5% vs 28.4% respectively). Perh Lind et al. [26] also looked at a subgroup of 128 patients from his earlier paper [17]. He analyzed the data using the ROC (receiver operating characteristics curves), which showed the significance of ipsilateral V 20Gy in clinical RP (p = 0.008) and radiological RP (p = 0.009). Mean lung dose (MLD) MLD is also significantly correlated with the incidence and grade of RP, mostly in studies involving lung cancer or other thoracic irradiations [27][28][29][30]. However the data on MLD in BC is very limited [ Table 2]. Though these studies used different end points, all of them showed significant impact of MLD on lung toxicity. As with the V dose , the study by Perh Lind et al. showed significant correlation between increasing MLD and the grade of lung toxicity using gamma statistics (p < 0.001) [17]. The MLD was 7.5 Gy, 13.5 Gy and 16.0-16.6 Gy for no RP, mild RP and moderate RP respectively. In Zsusana Kahan's study 20.5% of patients with radiological RP developed clinical symptoms. The MLD of patients with no RP versus RP in this study was 12.2 Gy vs 15.0 Gy respectively [31]. Treatment factors SCF irradiation Current meta-analysis shows a significant effect of SCF irradiation on the incidence of RP [ Figure 3]. The odds ratio (OR) of having RP in this group is 5.07 (95% CI = 1.95-13.22). The heterogeneity between the studies (I-squared) is 70.1% (p = 0.035). The subgroup analysis of 121 patients with node positive stage II BC by Berit Wennberg et al. also showed significant effect of SCF irradiation on RP [16]. Figure 4 shows that OR of having RP with IMN irradiation is 1.04 (95% CI = 0.43-2.54) and there is no statistically significant heterogeneity between the studies (I-squared = 66.3%, p = 0.052). The study by Goldman et al. had only 9 patients in the no IMN RT group [22]. The fields used in the IMN irradiation in the studies above vary significantly and they included oblique electron beam, anterior photon beam or deep oblique field. Sequential chemotherapy-RT Data on sequential chemotherapy and RP in breast irradiation is very scarce in the literature. Most studies looked at individual chemotherapeutic agents especially the taxane group [34,35]. The analysed studies are heterogeneous with multiple agents used, different doses and intensity of the chemotherapy. [20] did not mention the type of chemotherapy in their studies. The OR of having RP in this group is 1.40 (95% CI = 0.44-4.50) with I-squared = 88.5% (p < 0.001) for heterogeneity of the included studies. Patient factors Smoking The effect of smoking on RP has been studied in a few trials involving thoracic irradiation. In two lung cancer studies [36,37], smoking has been associated with lower incidence of lung toxicity. This positive effect of smoking is also seen in a trial by Leif Bjermer et al. [38] which looked at inflammatory response in breast cancer patients by bronchoalveolar lavage. Another large retrospective review by Silvia Johansson et al. [39] looked at 405 women who underwent radiotherapy for breast or oesophageal cancers also showed similar findings. Discussion The early radiation induced pulmonary toxicity has been well studied in other thoracic irradiation, mostly involving lung cancers. Caution is needed in extrapolating the data in lung tumors to breast cancer patients. The demographics of the patients may differ significantly as can be seen in the proportion of smokers, age group, effect of the tumor itself on the lung function and also the location of the lung that is irradiated. Furthermore, the lung in BC patients is a healthy organ in contrast to lung cancer patients. The major limitation that the authors faced in completing current analysis is the heterogeneity of the data in world literature, such as the clinical endpoints measurement, radiological and clinical RP grading system, different sensitivity in the detection tests of the radiological RP and the poor reporting of confounding factors in some studies. Besides the dosimetric, treatment and patient factors mentioned earlier, other reported factors are pre-irradiation lung function and performance status, pre-existing respiratory diseases (chronic obstructive airway disease, interstitial pneumonitis, etc) and genetic predisposition. There is very little or no data on contribution of this factors in 3D-CRT for BC towards early lung toxicity and hence no analysis can be done. Radiological and clinical RP The previous review by Mc Donald et al. [11] in 1995 have reported incidence of radiological RP in range of 27 -40% and clinical RP in to be in the range of 0-10%. The overall incidence of radiological RP in current

metaanalysis is 42% (95% CI = 22-62%) which is marginally higher compared to the previous review by Mc Donald et al. This may be due to the use of more CT scans in current studies. Clinical relevance low grade or high grade radiological RP leading to late lung toxicity is not well known currently and is an area for further research not only in breast cancer patients but also other thoracic irradiations. Five studies in current analysis have shown poor correlation between radiological RP and clinical RP [18,19,22,31,43]. Fourteen per cent incidence of clinical RP in current analysis (95% CI = 8-21%), is also higher than the previous estimate by McDonald.. This can be due to more vigilant assessment by the physicians in current clinical trial settings. Furthermore, McDonald and colleagues only reported approximate values and the precise information was not always reported in the quoted studies. The late sequelae of low or high grade clinical RP is also not well known currently, as with radiological RP. A subgroup analysis of 121 patients with stage II, node positive breast cancers by Berit Wennberg et al. [16] (from the large group of 475 patients in the study by Perh Lind et al. [17]) showed a slightly higher incidence of 23.1% clinical RP compared to 18.9% in the main study. This is probably due to the fact that the 475 patients recruited into the main study may include the mixture of very early stage disease who may not have had regional RT. The reported incidences of clinical and radiological RP vary significantly between studies as seen in the test for heterogeneity (I-squared). This can be attributed to the different techniques, fields used and measurement variations of RP. Though the assessment of clinical RP is quite standardized using either CTC2.0 or CTC3.0 toxicity criteria, the threshold to diagnose clinical RP may vary significantly between physicians and between centers. DVH related dosimetric factors The DVH related parameters (V dose and MLD) are strongly associated with the clinical RP, radiological RP and change in physiologic lung functioning in the current analysis. The study by Akiko Kubo et al. [24] was insignificant most likely because of low ipsilateral V 20Gy of 9.6% (mean) and 18.8% (maximum). In another study, Javier Jaen et al. [44] had shown that bilateral V 20 was closely correlated to the change in perfused volume. However it is beyond the scope of this paper to discuss the merits of physiological or functional lung damage assessment. It is difficult to deduce the exact threshold value of V 20Gy for clinical practice from this meta-analysis. However, other RT techniques or modalities can be considered, when ipsilateral V 20Gy > 30% [17,22,31]. Some clinical RP radiation pneumonitis, n number of patients, CI confidence interval, CTC/CTC-NCIC common toxicity criteria, RT radiotherapy, * year of data collection, NS statistically not significant. scenarios with high V 20Gy are very curved chest wall, bilateral breast radiotherapy and locally advanced BC. The ipsilateral V 20Gy should be kept below 24% if possible without compromising the required RT field coverage. The study by Akiko Kubo et al. [24] implies that further attempts at reducing ipsilateral V 20Gy below 20% may not be of any benefit. There are other models in radiation induced lung injury that are not analyzed in this paper such as the normal tissue complication probability (NTCP) model and more advanced biologically based models [25,45,46]. There is very limited data on the use of these models in BC. The large study of lung cancer patients by Kwa et al. [27] addressing RP have recommended the use of MLD or NTCP to predict the risk of RP. MLD is also found to be highly correlated to V 20Gy in lung irradiation studies [27,29]. The three studies quoted in this current analysis do show a strong relationship between MLD and early lung toxicity. Based on the studies by Perh Lind et al. [17] and Zsusana Kahan et al. [31], it should be safe if the MLD is limited below the range of 12-15 Gy to avoid serious lung toxicity in 3D-CRT irradiation for BC. Other techniques or modalities of radiotherapy may be considered if MLD exceeds 15 Gy. Treatment factors SCF irradiation shows a strong association with RP incidence (OR = 5.07). The addition of SCF irradiation increases the amount of lung tissue irradiated and hence MLD and V dose [47,48]. The reported incidence of IMN and medial SCF LN metastasis is between 4-9% in axillary node negative patients and 16-52% in axillary node positive patients [49][50][51][52]. Despite the reported toxicities, positive outcomes from 2 studies by Overgaard M et al. [2,3] have renewed the interest in IMN irradiation as part of LRRT. Furthermore, modern conformal techniques are likely to reduce the heart dose. The current view of IMN irradiation may change depending on the outcome of EORTC trial 22922/10925 which completed accrual in 2004 and results are expected in 2012 [49]. The three years toxicity profile update of 4004 patients with IMNmedial SCF field irradiation reported a 4.3% lung toxicity rate in the IMN arm versus 1.3% in no IMN arm (p < 0.0001). There were no statistically significant difference between the two groups with regards to cardiac toxicity. Poorly powered study by Goldman et al. (9 patients in no-IMN RT group) [22] caused the nonsignificant findings in current meta-analysis. Concomitant use of Tamoxifen has no statistically significant effect on RP in this meta-analysis, as in earlier studies [32,33]. Three large retrospective studies [53][54][55] examined the concomitant and sequential Tamoxifen in BC patients, and concluded that the use of both concomitant and sequential Tamoxifen was acceptable in terms of local relapse, distant metastasis and secondary malignancies. There was no statistically significant difference in lung toxicity between the groups. The influence of concurrent chemotherapy-RT on RP incidence has been well studied in thoracic irradiation [11,28,56] and also two BC studies [11,57]. Concurrent chemotherapy-RT is not analyzed in current paper as it is seldom practised nowadays. Data on sequential chemotherapy and RP in breast irradiation is very scarce in the literature [34,35]. Our result is not significant probably because of heterogeneous studies. Adjuvant Taxanes and Cyclophosphamide or biological agents (Bevacizumab) [58] have been reported to increase the risk of RP. Anthracyclines and Taxanes [59] are potent radiosensitizers. Cyclophosphamide have been reported to cause lung injury with or without the addition of RT in some case series [60]. Radiation recall can occur if Paclitaxel is used after RT and can cause pathological changes in the lungs at the previous RT field. Hence readers are encouraged to refer to data on individual drugs in the literature for clinical practice. Patient factors There is a trend towards protective effect for smokers in breast irradiation in our meta-analysis. This effect was also reported in two other studies of thoracic irradiation [36,37]. This finding is quite puzzling because we expect that smokers with already damaged lung would be more susceptible to radiation induced lung injury. Some possible explanations are: 1) smokers may have higher threshold to develop clinical symptoms due to their already damaged epithelium; 2) the local immune reaction may not be as strong as non-smokers due to the damage to immune cells such as tissue macrophages in the lung epithelium causing reduced antibody secretion and exudates; 3) tissue hypoxia in smokers may have radio protective effect; 4) lung scarring in smokers may mask the appearance of RP on CT scans. In current analysis, there is a conflicting effect of age on both radiological and clinical RP in 3D-CRT for BC. The studies by Akiko Kubo et al. [24] and Perh Lind et al. (Duke's) [20] did not show statistically significant effect of age. In the former study, only tangential whole breast irradiation was used and the volume of irradiated lung was very small. There could have been significant bias in retrospective data in the latter study. Based on the other three studies, age > 55 yrs is a risk factor for RP in 3D-CRT for BC. However, the long term effect of early lung toxicity is not known and may take many years before it becomes clinically significant, especially in the younger age group. Another type of early lung toxicity reported in the literature but not discussed here is the radiationinduced bronchiolitis obliterans organizing pneumonia (BOOP) syndrome in which the lung injury occurs outside the radiation field or even in the contra-lateral lung. Kubo et al. have reported 2.9% incidence of BOOP syndrome with all cases of radiological RP ≥ grade 2 developing BOOP syndrome. A survey in major hospitals in Japan from August 1999 to March 2000 showed an incidence of 1. 8-2.19% in patients radiated after BCT [24,61]. Conclusion The DVH-related parameters (V dose and MLD) and SCF irradiation are the strongest parameters associated with RP. Attempt should be made where possible to keep the ipsilateral lung V 20Gy < 24% and MLD < 15 Gy without compromising the required radiotherapy coverage. Other RT techniques can be considered when ipsilateral lung V 20Gy >30 Gy or MLD > 15 Gy. Other factors that increase the risk of RP are age >55 yrs and probably IMN irradiation. However, caution is needed in treating younger age group due to the possibilities of late sequealae. Concomitant Tamoxifen do not increase the risk of developing RP. Readers are encouraged to search the data on individual chemotherapeutic agents for clinical practice. Though smoking is noted to have protective effect, this can just be a masking effect rather than physiological protection against radiation induced damage. Biochemical markers can be another factor to predict for early lung toxicity in 3D-CRT for BC. The combination of elevated transforming growth factor (TGF)-beta1 levels during RT and MLD has been reported as having predictive value in non-small cell lung cancers [62]. This together with use of radioprotectors such as Pentoxifylline can be an area for further research in RP [63]. Liquid Biopsy-Guided Interventional Oncology: A Proof of Concept with a Special Focus on Radiotherapy and Radiology Simple Summary Over the last decade radiology interventions for tumor ablation have steadily increased and new procedures of interventional Oncology (IO) have been developed. Minimal residual disease (MRD) management and its quantification has a critical role to assess the efficacy or failure of the treatment adopted.The assessment of the MRD depends by multiple and repetitive measurements performed through a no invasive procedure like as Liquid biopsy (LB). LB is a rapidly evolving technology with potentially relevant consequences in the management of patients with cancer. Despite its high potential, the LB approach cannot give spatial information and still lacks of standardization and required specificity and sensitivity to be extensively implemented in the routine clinical practice. The review is aimed at investigating how LB can monitor the effects of locoregional treatment highliting the potential synergies between interventional techniques and LB to improve cancer patient’ management. The review suggests combining Interventional techniques and LB technology could improve LB overcoming its sensitivity limitations. Furthermore, LB techniques may enhance the outcomes of minimally invasive loco-regional treatments not only via ongoing periodic MRD monitoring of a patient during and subsequent to IO procedures. Abstract Although the role of liquid biopsy (LB) to measure minimal residual disease (MRD) in the treatment of epithelial cancer is well known, the biology of the change in the availability of circulating biomarkers arising throughout treatments such as radiotherapy and interventional radio-oncology is less explained. Deep knowledge of how therapeutic effects can influence the biology of the release mechanism at the base of the biomarkers available in the bloodstream is needed for selecting the appropriate treatment-induced tumor circulating biomarker. Combining existing progress in the LB and interventional oncology (IO) fields, a proof of concept is provided, discussing the advantages of the traditional risk assessment of relapsing lesions, limitations, and the timing of detection of the circulating biomarker. The current review aims to help both interventional radiologists and interventional radiation oncologists evaluate the possibility of drawing a tailor-made board of blood-based surveillance markers to reveal subclinical diseases and avoid overtreatment. Introduction Over the last decade, radiology interventions for tumor ablation have steadily increased, and new procedures have been developed. Radiofrequency, microwave, or cryotherapy are the most commonly used techniques supported by further progress in imaging tools such as ultrasound fusion and contrast, computed tomography (CT) fluoroscopy, magnetic resonance

(MR) thermometry, and positron emission tomography PET guidance [1]. An additional alternative to thermal ablation is irreversible electroporation, which transiently disrupts cell membranes through electric pulses, killing cancer cells, although this technology is still in its early stages [2]. An expanded role in the future will deliver novel therapeutics into tumors through either catheter-based approaches or direct intratumor injections of chemotherapeutic agents or oncolytic viral therapies [2] and bacterial therapies [3]. Moreover, several ongoing treatment trials combine immunotherapy with interventional oncology (IO) to improve antigen presentation, which can be accomplished with ablative or catheter-directed therapies [3]. For infiltrative tumors or to locally control nonresectable lesions, which cannot be effectively accomplished by either external beam radiation therapy or standard thermal ablative techniques, a new procedure in interventional radiotherapy, called brachytherapy, combining an advanced imaging technique and an image navigator system, has been implemented. The integrated approach has the potential to shorten the acquisition time of diagnostic-therapeutic information and radioactive seed or catheter placement, promoting tumor control. Another emerging therapeutic procedure is high-intensity focused ultrasound ablation, which allows intratumor drug delivery through either selective opening of the vascular barrier or nanoparticle disruption [1]. In cancer patients showing an oligometastatic setting, the primary endpoint of IO is to achieve complete remission of the treated lesion and to reduce the incidence of distant spreading cells by controlling the risk of local recurrence [4]. This risk, together with the incidence of distant spreading, is strictly dependent on the number of residual cancer cells, locally and/or circulating, in the peripheral blood after treatment [5]. Although the goal of cytoreductive therapy is to eradicate all malignant cells, a substantial portion of patients develop resistant cancer cell clones, so-called minimal residual disease (MRD), detectable after therapy, which ultimately leads to cancer recurrence [6]. A critical milestone in MRD management is its quantification by multiple and repetitive measurements to detect the precise cut-off corresponding to the minimal threshold of detectable cancer cells of patients responding well to therapy and, on the other hand, that can be used as an end-point of the therapy when the values of the markers fall below the minimum detectable [7]. The MRD level, at any time, reflects the anticancer treatment response, as well as intratumor and systemic immunologic effects, having the potential to counteract cancer relapse. The deep analysis of MRD needs to include characterization of the high-risk features that antagonize the response to treatment. On the other hand, the type of treatment and the anatomical accessibility of the tumor would conditionate the MRD monitoring methods. Because of the differences in clinical practice, it has not yet been identified as a gold-standard method for quantifying MRD [8]. Liquid biopsy (LB) can indirectly diagnose the presence of MRD by revealing both quantitative and qualitative dynamic modifications of the circulating tumor cells (CTCs), circulating tumor DNA (ct-DNA), tumor-specific microRNA, and circulating exosomes [8]. These biomarkers are shed by the tumor in the bloodstream, with their entity and availability reflecting different tumor tissue stages, i.e., vascularization, vessel permeabilization, inflammation, and apoptosis or necrosis cell death [9]. Additionally, release of CTCs could be influenced by high proliferation, spontaneous detachment from the matrix, and molecular equipment that enables cancer cells to cross the endothelial wall, leading to cancer relapse [10]. All circulating biomarkers are increasingly being investigated to improve their pertinence, standardization, and reproducibility. LB can be considered a precious parameter in the therapeutic decision-making process, to monitor and detect MRD in patients with solid cancers eligible for interventional therapy, and also to select patients needing treatment adjustments to improve their quality of survival [8]. The promising clinical impact of molecular and cellular circulating markers consists of a personalized approach based on the control and monitoring of specific tumor biological key markers anticipating the patients with solid cancers eligible for interventional therapy, and also to select patients needing treatment adjustments to improve their quality of survival [8]. The promising clinical impact of molecular and cellular circulating markers consists of a personalized approach based on the control and monitoring of specific tumor biological key markers anticipating the development of the secondary lesion through quantitative and qualitative estimation before their clinical image-guided evidence ( Figure 1). Figure 1. The liquid paradigm of the tumor biopsy. The liquid biopsy (LB) has the advantage of including the analysis of the circulating tumor products, representing the subclinical level of cancer disease systemic dissemination by unmasking emerging or expanding subclones involved in local recurrence or metastasis. Impact of Liquid Biopsy in Radiotherapy As per the up-to-date state of the art, cancer cell clonal monitoring within the tumor tissue under the selective pressure of the treatment is dictated by anatomic imaging, conducting scans every few months, or testing blood tumor-associated antigens (carcinoembryonic antigen (CEA), carbohydrate antigen 19-9 (Ca19.9), etc.) at the systemic level [11]. To start the process, the first step is to obtain an informative biopsy before and over therapy time. A biopsy is considered "informative" when it shows high diagnostic yields, is less invasive, is well tolerated by patients, can include information about more biomarkers, and can provide an evaluation of the tumor microenvironment adaptations to therapeutic pressures. Although the anatomic approach has generally worked well, currently, interest is focused on gaining tissue informative biopsies able to guarantee molecular readouts of the most active cancer areas rather than those of quiescent or necrotic lesions. Targets are primarily chosen based on accessibility, surrounding structures, obstacle avoidance, and patient conditions (e.g., sedation requirements and pain control). Furthermore, numerous reports evaluating a large number of tissue biopsies have defined that harvesting multiple tissue cores does not increase complication rates [12], tissue sampling procedures remain invasive, and the standardization of biopsy materials procedures remains problematic. In an ideal setting, future biopsies will be performed by targeting "hot spots" identified by molecular imaging within the tumor lesion [13]. In recent years, the development of liquid biopsy methodology has introduced an alternative point of view for serially monitoring treatment response, MRD, and relapse. The quantification of MRD can be based on combining information from two main sources, liquid and tumor biopsy, to mine more accurate and informative data points. Impact of Liquid Biopsy in Radiotherapy As per the up-to-date state of the art, cancer cell clonal monitoring within the tumor tissue under the selective pressure of the treatment is dictated by anatomic imaging, conducting scans every few months, or testing blood tumor-associated antigens (carcinoembryonic antigen (CEA), carbohydrate antigen 19-9 (Ca19.9), etc.) at the systemic level [11]. To start the process, the first step is to obtain an informative biopsy before and over therapy time. A biopsy is considered "informative" when it shows high diagnostic yields, is less invasive, is well tolerated by patients, can include information about more biomarkers, and can provide an evaluation of the tumor microenvironment adaptations to therapeutic pressures. Although the anatomic approach has generally worked well, currently, interest is focused on gaining tissue informative biopsies able to guarantee molecular readouts of the most active cancer areas rather than those of quiescent or necrotic lesions. Targets are primarily chosen based on accessibility, surrounding structures, obstacle avoidance, and patient conditions (e.g., sedation requirements and pain control). Furthermore, numerous reports evaluating a large number of tissue biopsies have defined that harvesting multiple tissue cores does not increase complication rates [12], tissue sampling procedures remain invasive, and the standardization of biopsy materials procedures remains problematic. In an ideal setting, future biopsies will be performed by targeting "hot spots" identified by molecular imaging within the tumor lesion [13]. In recent years, the development of liquid biopsy methodology has introduced an alternative point of view for serially monitoring treatment response, MRD, and relapse. The quantification of MRD can be based on combining information from two main sources, liquid and tumor biopsy, to mine more accurate and informative data points. However, to optimize the liquid biopsy approach to unmask MRD-positive patients, it is reasonable to advocate that MRD testing would be performed by circulating biomarkers. Compliance between the treatment and the choice of MRD biomarker is a crucial step in designing a personalized monitoring platform, for example, to discriminate the molecular products released in peripheral blood or vesicles (exosomes, microvesicles, etc.), circulating cells of epithelial and/or mesenchymal coming from tumor tissue, and endothelial cells with different differentiation grades (Table 1). Compliance between the treatment and the choice of MRD biomarker is a crucial step in designing a personalized monitoring platform, for example, to discriminate the molecular products released in peripheral blood or vesicles (exosomes, microvesicles, etc.), circulating cells of epithelial and/or mesenchymal coming from tumor tissue, and endothelial cells with different differentiation grades (Table 1). Circulating DNA (ct-DNA) is the most widely analyzed tumor-related element during anticancer treatment, as it keeps the same genomic signatures in the matching tumor, and the analysis performed by next-generation sequencing technologies improves the rates of MRD quantification. Ct-DNA provides data about the ability of some cancer cell subpopulations to fight elimination, revealing the competitive phenomena that undergo continuous remodeling at genetic levels during therapy. This approach allows the interrogation of clonally divergent, distant lesions without bias, allowing longitudinal monitoring of patients [14,15]. On the other side, ct-RNA in the bloodstream might be linked to epigenetic pathways of tumor plasticity, development, and progression. Both ct-DNA and ct-RNA are released in body fluid by tumor tissue as a consequence of cell death, necrosis, or apoptosis, or ct-RNA designing a personalized monitoring platform, for example, to discriminate the molecular products released in peripheral blood or vesicles (exosomes, microvesicles, etc.), circulating cells of epithelial and/or mesenchymal coming from tumor tissue, and endothelial cells with different differentiation grades (Table 1). Circulating DNA (ct-DNA) is the most widely analyzed tumor-related element during anticancer treatment, as it keeps the same genomic signatures in the matching tumor, and the analysis performed by next-generation sequencing technologies improves the rates of MRD quantification. Ct-DNA provides data about the ability of some cancer cell subpopulations to fight elimination, revealing the competitive phenomena that undergo continuous remodeling at genetic levels during therapy. This approach allows the interrogation of clonally divergent, distant lesions without bias, allowing longitudinal monitoring of patients [14,15]. On the other side, ct-RNA in the bloodstream might be linked to epigenetic pathways of tumor plasticity, development, and progression. Both ct-DNA and ct-RNA are released in body fluid by tumor tissue as a consequence of cell death, necrosis, or apoptosis, or EVs designing a personalized monitoring platform, for example, to discriminate the molecular products released in peripheral blood or vesicles (exosomes, microvesicles, etc.), circulating cells of epithelial and/or mesenchymal coming from tumor tissue, and endothelial cells with different differentiation grades (Table 1). Circulating DNA (ct-DNA) is the most widely analyzed tumor-related element during anticancer treatment, as it keeps the same genomic signatures in the matching tumor, and the analysis performed by next-generation sequencing technologies improves the rates of MRD quantification. Ct-DNA provides data about the ability of some cancer cell subpopulations to fight elimination, revealing the competitive phenomena that undergo continuous remodeling at genetic levels during therapy. This approach allows the interrogation of clonally divergent, distant lesions without bias, allowing longitudinal monitoring of patients [14,15]. On the other side, ct-RNA in the bloodstream might be linked to epigenetic pathways of tumor plasticity, development, and progression. Both ct-DNA and ct-RNA are released in body fluid by tumor tissue as a consequence of cell death, necrosis, or apoptosis, or CTCs products released in peripheral blood or vesicles (exosomes, microvesicles, etc.), circulating cells of epithelial and/or mesenchymal coming from tumor tissue, and endothelial cells with different differentiation grades (Table 1). Circulating DNA (ct-DNA) is the most widely analyzed tumor-related element during anticancer treatment, as it keeps the same genomic signatures in the matching tumor, and the analysis performed by next-generation sequencing technologies improves the rates of MRD quantification. Ct-DNA provides data about the ability of some cancer cell subpopulations to fight elimination, revealing the competitive phenomena that undergo continuous remodeling at genetic levels during therapy. This approach allows the interrogation of clonally divergent, distant lesions without bias, allowing longitudinal monitoring of patients [14,15]. On the other side, ct-RNA in the bloodstream might be linked to epigenetic pathways of tumor plasticity, development, and progression. Both ct-DNA and ct-RNA are released in body fluid by tumor tissue as a consequence of cell death, necrosis, or apoptosis, or CECs, EPCs ing cells

of epithelial and/or mesenchymal coming from tumor tissue, and endothelial cells with different differentiation grades (Table 1). Circulating DNA (ct-DNA) is the most widely analyzed tumor-related element during anticancer treatment, as it keeps the same genomic signatures in the matching tumor, and the analysis performed by next-generation sequencing technologies improves the rates of MRD quantification. Ct-DNA provides data about the ability of some cancer cell subpopulations to fight elimination, revealing the competitive phenomena that undergo continuous remodeling at genetic levels during therapy. This approach allows the interrogation of clonally divergent, distant lesions without bias, allowing longitudinal monitoring of patients [14,15]. On the other side, ct-RNA in the bloodstream might be linked to epigenetic pathways of tumor plasticity, development, and progression. Both ct-DNA and ct-RNA are released in body fluid by tumor tissue as a consequence of cell death, necrosis, or apoptosis, or Circulating DNA (ct-DNA) is the most widely analyzed tumor-related element during anticancer treatment, as it keeps the same genomic signatures in the matching tumor, and the analysis performed by next-generation sequencing technologies improves the rates of MRD quantification. Ct-DNA provides data about the ability of some cancer cell subpopulations to fight elimination, revealing the competitive phenomena that undergo continuous remodeling at genetic levels during therapy. This approach allows the interrogation of clonally divergent, distant lesions without bias, allowing longitudinal monitoring of patients [14,15]. On the other side, ct-RNA in the bloodstream might be linked to epigenetic pathways of tumor plasticity, development, and progression. Both ct-DNA and ct-RNA are released in body fluid by tumor tissue as a consequence of cell death, necrosis, or apoptosis, or encapsulated in vesicles as secretion products by living cells [15]. These encapsulated vesicles are distinguished by size in microvesicles, apoptotic bodies, circulating exosomes, and tumoreducated platelets (TEPs) and are actively released in body fluids containing different molecules, nucleic acid fragments, proteins such as chemokines, and cytokines [16]. Their analysis can provide information on the dynamic interplay between cancer and immune cells throughout the treatments [17]. Real-time quantification and characterization of those populations can be considered the best goal in order to identify what happens inside the treated tumor for the following: 1. CTCs correspond to the tumor progeny triggering, with a high probability, further lesions [19,20]. In oligometastatic patients, the presence of CTCs in peripheral blood reflects the persistence of tumor lesions and their high probable access to the systemic bloodstream [22]. 4. In patients with multimetastatic disease, CTCs can be identified as a longitudinal scan of the heterogeneous cell clonality [23]. 5. The endothelial cell subtype, circulating endothelial cells (CECs), and endothelial progenitor cells (EPCs) provide useful information about tumor angiogenesis and when associated with cardiovascular parameters, as well as side effects due to anticancer treatment on heart function [18]. Minimal Residual Disease Assessment in Radiotherapy by Circulating Tumor DNA Based on the biology of ct-DNA as free fragmented molecules in peripheral blood (cellfree-DNA or cf-DNA), its application in MRD surveillance during interventional radiology provides specific indications. For ct-DNA use in clinical practice, some limits must be considered. Very low levels must be detected within a total pool of cf-DNA (cell-free DNA), especially in pretreated cancer patients submitted to interventional oncology. Ct-DNA variations must be distinguished from those resulting from premalignant lesions such as age-related clonal hematopoiesis or due to the effect of chemotherapy on normal cells undergoing apoptosis. Moreover, it is necessary to consider the impact of biological activity (metabolic activity, cell death mechanisms, innate and adaptative immunity reaction) and anatomical factors (blood-tissue barrier permeabilization) on the dispersion of ct-DNA after local radiological treatment. The detection of epigenetic alterations is also applicable to the analysis of ct-DNA carried out by polymerase chain reaction (PCR) or sequencing [14]. With increased accuracy and reduced sequencing costs, routine clinical implementation of ct-DNA analysis can be considered now more feasible and may achieve a greater impact on patients with radiotherapy-treated cancer. Currently, the dose prescription or the decision between only radiotherapy or combined treatment with chemotherapy or surgery is determined by clinicopathological factors and is not influenced by biomarkers; therefore, some patients receive too much radiation or too many therapies, while others underestimate the risk of recurrence. The stratification of patients gained by LB allows to orient the therapy toward a more or less aggressive treatment. In the case of low levels of ct-DNA, the dose of radiotherapy can be reduced, or combined therapies can be avoided; instead, in patients with high levels of ct-DNA, it can lead to more aggressive therapeutic regimens ( Figure 2). Ct-DNA profiling at baseline and Days 91, 181, and 361 was performed [24,25], emphasizing that long timing of longitudinal observation for this biomarker is needed after therapy. Patients with high ct-DNA levels after neoadjuvant chemoradiotherapy and before surgery have very high relapse rates [24]. Furthermore, it must be considered that unchanged or increased ct-DNA levels during radiotherapy could indicate resistance to the treatment or progression of micrometastases, giving the chance to change the treatment strategy [25]. Emerging studies are showing promising performance of ct-DNA analysis for detecting MRD after radical radiotherapy [24][25][26][27][28]. In non-small-cell lung cancer (NSCLC). ct-DNA analysis of post-radiotherapy plasma samples achieved a 94% detection rate in patients who experienced recurrence [24]. Patients with high ct-DNA levels after neoadjuvant chemoradiotherapy and before surgery have very high relapse rates [24]. Furthermore, it must be considered that unchanged or increased ct-DNA levels during radiotherapy could indicate resistance to the treatment or progression of micrometastases, giving the chance to change the treatment strategy [25]. Few studies use LB analysis to detect minimal/molecular residual disease in radiotherapy-treated patients; one factor impeding the conduct of prospective clinical trials in this field is the rapidly evolving technological platforms for the detection of ct-DNA and CTCs [24][25][26][27][28]. Many studies evaluating novel LB platforms in the MRD setting have been retrospective. Numerous other retrospective studies have suggested the role of LB-based surveillance after radical or neoadjuvant treatment. LB is also used to identify predictive biomarkers, which are mainly based on qualitative properties reflecting the biological performance of the tumor. Currently approved ones are related to molecular target therapy such as epidermal growth factor receptor (EGFR) tyrosine kinase in NSCLC [27]. Response prediction to radiotherapy, by molecular biomarkers, remains a useful research area to correct the use of radiotherapy, select suitable and unsuitable patients, and can adjust the dose prescription. Patients with oligometastatic disease could be candidates for more aggressive treatment incorporating local therapies [29,30]. A recent study in oligometastatic prostate cancer has uncovered the potential of LB as a predictive biomarker for response to advanced treatment [25] (Table 2). Few studies use LB analysis to detect minimal/molecular residual disease in radiotherapytreated patients; one factor impeding the conduct of prospective clinical trials in this field is the rapidly evolving technological platforms for the detection of ct-DNA and CTCs [24][25][26][27][28]. Many studies evaluating novel LB platforms in the MRD setting have been retrospective. Numerous other retrospective studies have suggested the role of LB-based surveillance after radical or neoadjuvant treatment. LB is also used to identify predictive biomarkers, which are mainly based on qualitative properties reflecting the biological performance of the tumor. Currently approved ones are related to molecular target therapy such as epidermal growth factor receptor (EGFR) tyrosine kinase in NSCLC [27]. Response prediction to radiotherapy, by molecular biomarkers, remains a useful research area to correct the use of radiotherapy, select suitable and unsuitable patients, and can adjust the dose prescription. Patients with oligometastatic disease could be candidates for more aggressive treatment incorporating local therapies [29,30]. A recent study in oligometastatic prostate cancer has uncovered the potential of LB as a predictive biomarker for response to advanced treatment [25] (Table 2). There is also the possibility to combine ct-DNA with circulating proteins (prostatespecific antigen in prostate cancer) or with imaging (combination of ct-DNA and MRI imaging) [25][26][27] to improve the informative data setting linked to molecular biomarkers in MRD monitoring. Minimal Residual Disease Assessment in Radiotherapy by Circulating Tumor Cells Serial collection of CTCs during radiotherapy is useful to study the mechanism of treatment resistance in cancer types. CTCs can be grown in vitro [9,10,14] or subcutaneously engrafted into immunodeficient mice to provide a renewable source of patient tumor tissue for molecular profiling, genomic studies, and drug tests [18]. In patients with NSCLC treated with fractioned radiotherapy, it was hypothesized that the destruction of the tumor mass promotes the passage of tumor cells into the circulation, contributing to the formation of distant metastases [31]. The data do not support this hypothesis. The increased CTC positivity rate after radiotherapy could be due to an epiphenomenon depending on the actinic effects on the vessel's permeabilization [32]. Real-time monitoring data indicated a statistically significant (compared to pretherapy levels) increase in the number of CTCs starting soon (10-15 min) after therapy. The increased number of CTC lasts on average 20 min before returning to the pretherapy level [32]. No significant changes were observed in CTC count in a time range of 8 and 16 h (hrs) after therapy crosswise all treatments. Current studies revealed an acute increase in CTCs in a time window consistent with fluctuations in tumor vascular permeability. Moreover, the short-lived increase in CTC concentration was followed by a notable absence of CTCs 8 h post-treatment, consistent with physiological suppression of blood flow in tumor tissue. This is indirectly confirmed by the return of CTCs count back to pretherapy levels 16 h after therapy when vessel occlusion effects are minimized. Combination therapy also had a similar trend; however, these changes were not as appreciable [33]. Aggregates of tumor and healthy cells and clusters of tumor cells are released into the bloodstream; these aggregates are intercepted by the lung or the spleen and rapidly eliminated from circulation, while individual cancer cells can remain in circulation for much longer periods (up to 30-90 min). Long-term effects were analyzed using tumor growth delay analysis and histopathology data in addition to the last CTCs scan on Day 21. By Day 21, following the initial treatment, radiation significantly delayed tumor growth and reduced the number of tumor cells in circulation. The therapeutic effect on tumor volume correlated well with the total CTC count at the end of the evaluation period [33][34][35]. After high-dose radiotherapy, lack of metastatic tumor mass following CTC release indicates low viability, clonogenicity, or both; on the other hand, real-time monitoring of CTCs provides a window on acute therapeutic events in real time, different from traditional approaches which can take months to detect changes associated with therapy [35]. Recently, a work demonstrated that at doses of 10 gray (Gy) or above, radiation disrupts the tumor vasculature leading to irreparable damage to nearby well-oxygenated tumor cells. While hypoxic tumor tissues are known to be radio-and chemoresistant; these data highlight the importance and relevance of adjuvant chemotherapy to control systemic disease with the use of vascular disrupting drugs in combination with radiation or not [36]. The decrease in CTCs after radiotherapy treatment was demonstrated by different studies and independently by the technology of analysis used. The Cell Search system demonstrated that 15% of patients are a positive CTC count before and 7% after treatment with Fluoro-EPISPOT assay 49% versus 36% and with Cell Collector 48% versus 39% [26][27][28][29] (Table 2). CTC count is correlated with clinical course and response to treatment; patients who become CTC-negative after radiotherapy had a response rate of 86%, so CTC-positive after treatment is a poor outcome indicator. Similarly, a significant decrease in CTC number is found after radiotherapy in patients with localized NSCLC [27]. In prostate cancer (PC), prostate-specific antigen (PSA) serum level is not significantly different in patients with CTC-negative and positive samples, suggesting that the major source of PSA serum is not related to PSA-Secreting CTCs but rather to the tumor itself, independently from its transformation condition [34]. Several studies have reported significant correlations between the number of CTCs with patient survival and/or prostate cancer recurrence after treatment. These correlation studies have highlighted a clear trend towards a CTC value in correlation with the disease stage and towards a prognostic and predictive impact of CTC detection in prostate cancer [24][25][26][27][28][29]34]. Additionally, a significant correlation between CTC count and a patient's body mass index (BMI) was found. The possible cause of this correlation is the high concentration of leptin, insulin, insulin-like growth factor-1, and low

adiponectin, which together are likely to promote growth and progression. This observation underlies as in some types of tumors the metabolism induces a different behavior on CTCs availability [35]. Minimal Residual Disease Assessment in Radiotherapy by Circulating Exosomes The secretion of cellular products is one of the direct effects that can be observed during radiotherapy. Irradiated tumor cells produce exosomes, which are released into the tumor microenvironment. The DNA damage triggers the mechanism of increased exosome secretion by radiation cells, mainly by the activation of the P53/TSAP6 axis and the Wnt pathway. Through exosome secretion and exchange, irradiated cells target nonirradiated cells promoting migration, radioresistance, and DNA damage by promoting cytokines secretion by dendritic cells in the tumor microenvironment. The content of exosomes produced and released into body fluids significantly changes composition after tumor radiation therapy. Exosomal loads mainly include exosomal miRNAs, proteins, and other substances, such as circRNA and lipids regulating the aggressive cancer phenotype. Exosomes after radiotherapy suppress aggressive phenotypes of tumor cells, such as proliferation and migration. On the other hand, after irradiation, exosome-mediated signaling [16]. The "Liquid Redundancy" of Interventional Oncology As data reported in the literature suggest that LB can be used in monitoring radiotherapy outcomes, a few interventional oncology clinical trials use LB technologies with the same aim. Interventional radiology plays an integral role in delivering locoregional therapies for patients with both primary and metastatic disease. Both experimental treatment response monitoring and clinical follow-up may include tumor markers and CT, PET, and MRI. Microsatellite instability and mutational status, alpha-fetoprotein, and CA 19-9 are tumor markers that have been widely used for patients. Monitoring CTCs and ct-DNA may serve as a helpful additional tool for interventional radiologists to monitor patients before and after procedures to overcome the limited sensitivity and specificity of traditional markers. Some studies have observed that mutational status assessed in tumor tissue predicts response to some minimally invasive therapies, including ablation and embolization [36][37][38], and can extend this concept to the mutational state of ct-DNA to improve the management of cancer patients. Furthermore, the biological response of tumor cells at the tissue level to irradiation is the result of the coexistence and interaction of five radiobiology factors (5R), and it is possible to hypothesize that these factors may affect one or more circulating biomarkers. The five radiobiological determinants of treatment outcome in external radiotherapy are the same determinants conditioning the brachytherapy outcome. Repair of sublethal-induced cellular/DNA damage, redistribution of the cells within the cell cycle, revascularization/reoxygenation of the surviving cells, matrix remodeling, growth, apoptosis, cell death, and the reaction of the immune response reassume the principal biological consequences of the local treatment [37][38][39]. Local effects during brachytherapy could be tracked in the blood. Conceptually, the typology of tumor products released in the blood could be influenced by the prevalent biological effect within the 5R, as graphed in Figure 3. The five biological effects described as 5R could be considered to describe the local effects induced by hypoxia during chemoembolization or transarterial chemoembolization (TACE), which combines embolization with the local delivery of chemotherapy [40]. The liquid redundancy of TACE effects could reiterate the same points of view graphed in Figure 3 for brachytherapy, with a difference in terms of intensity, duration, and fluctuation of the levels of the circulating markers considered. On these bases, the applications of liquid biopsy could extend to additional opportunities for interventional radiologists to use this technology in treatment selection and postprocedural treatment of patients with cancer. chemoembolization (TACE), which combines embolization with the local delivery of chemotherapy [40]. The liquid redundancy of TACE effects could reiterate the same points of view graphed in Figure 3 for brachytherapy, with a difference in terms of intensity, duration, and fluctuation of the levels of the circulating markers considered. On these bases, the applications of liquid biopsy could extend to additional opportunities for interventional radiologists to use this technology in treatment selection and postprocedural treatment of patients with cancer. Figure 3. The "liquid redundancy" of the 5R biological factors. A proof of concept to individuate the circulating biomarkers linked to the prevalent biological factor, within the 5R, induced by interventional oncology treatment, influencing the relative effectiveness. Note: ct-RNA, circulating tumor RNA; ct-EVs, circulating vesicles TEPs, tumor-educated platelets. Discussion MRD has been identified as the single most robust prognosticator of overall and cancer-free survival in both cancer pediatric and adult patients. Serial testing of MRD is recommended, although robust trial evidences to validate the optimal timing and frequency are required. In the oncology of solid cancer, LB offers the opportunity to measure MRD in molecular and cellular terms, opening a new alternative era to traditional follow-up. On the other hand, the possibility to choose the different types of circulating biomarkers, and the relative methodology of analysis, suggest the necessity to develop a proof of concept on the utility of LB to personalize MRD monitoring [6,11]. Interventional oncology, focused on local control of the lesion, has the option of choosing a good MRD marker within the Discussion MRD has been identified as the single most robust prognosticator of overall and cancer-free survival in both cancer pediatric and adult patients. Serial testing of MRD is recommended, although robust trial evidences to validate the optimal timing and frequency are required. In the oncology of solid cancer, LB offers the opportunity to measure MRD in molecular and cellular terms, opening a new alternative era to traditional follow-up. On the other hand, the possibility to choose the different types of circulating biomarkers, and the relative methodology of analysis, suggest the necessity to develop a proof of concept on the utility of LB to personalize MRD monitoring [6,11]. Interventional oncology, focused on local control of the lesion, has the option of choosing a good MRD marker within the LB panel. LB markers can be expressed in real time, especially if analyzed in combination with all tissue modifications, before, during, and after treatment. Starting from the permeabilization of the vessel in the first step of irradiation or thermal ablation until tumor remodeling over time, the increased permeabilization of tumor vessels and necrosis induced by local treatment generate a storm of molecules that diffuse into body fluids following tumor remodeling and cell death. After a variable number of days from the local intervention, studies have demonstrated agreement with the observation of the decrease in circulating biomarkers associated with local tumor size reduction [24][25][26]. One of the question points raised here is after how many days does the decrease or increase in the selected biomarker reflect the tumor size modification? Results reported by clinical trials focusing on LB, reported in Table 2, monitoring MRD after radiotherapy, emphasize that there are different timings of longitudinal observation when assessing the selected biomarker. All studies highlighted that after therapy, accidental dissemination of molecules and cells occurs in the proximal fluids of the tumor [33,34]. Their systemic diffusion includes molecular products from healthy cells of the surrounding tissues and the tumor. Thus, unintentional release determines a deep interference that increases the general analytical noise or disturbance, favoring mistakes in the estimation phase of the LB biomarkers. This condition suggests the need to respect the time of latency or postpone the analysis and estimation phase after it. However, clinicians should be informed of this particular timing and that the results of the analysis can be affected by a high risk of false-positive or false-negative events. This period of latency, between the first treatment and the biomarker collection, could be called "molecular sedimentation time" (MST), because it represents the time needed for the interfering factors to sediment or reduce their presence in the blood (Figure 4). The time required for the catabolism of the interfering factors, occurring through the function of the hepatic, renal, and immune emunctories, is variable and individual. However, it is possible to reduce the latency time by using a strong tumor-specific circulating biomarker. The high specificity of a marker can guarantee a lower rate of false interpretation. Therefore, the MST is inversely proportional to the specificity of the circulating biomarker. In this scenario, the waiting time between the most used LB biomarkers, as shown in the clinical trials (reported in Table 2), is lower for CTCs rather than cf-DNA. This difference is potentially due to the high specificity and sensibility of CTCs compared to the other circulating biomarkers in the LB panel. in circulating biomarkers associated with local tumor size reduction [24][25][26]. One of the question points raised here is after how many days does the decrease or increase in the selected biomarker reflect the tumor size modification? Results reported by clinical trials focusing on LB, reported in Table 2, monitoring MRD after radiotherapy, emphasize that there are different timings of longitudinal observation when assessing the selected biomarker. All studies highlighted that after therapy, accidental dissemination of molecules and cells occurs in the proximal fluids of the tumor [33,34]. Their systemic diffusion includes molecular products from healthy cells of the surrounding tissues and the tumor. Thus, unintentional release determines a deep interference that increases the general analytical noise or disturbance, favoring mistakes in the estimation phase of the LB biomarkers. This condition suggests the need to respect the time of latency or postpone the analysis and estimation phase after it. However, clinicians should be informed of this particular timing and that the results of the analysis can be affected by a high risk of falsepositive or false-negative events. This period of latency, between the first treatment and the biomarker collection, could be called "molecular sedimentation time" (MST), because it represents the time needed for the interfering factors to sediment or reduce their presence in the blood (Figure 4). The time required for the catabolism of the interfering factors, occurring through the function of the hepatic, renal, and immune emunctories, is variable and individual. However, it is possible to reduce the latency time by using a strong tumorspecific circulating biomarker. The high specificity of a marker can guarantee a lower rate of false interpretation. Therefore, the MST is inversely proportional to the specificity of the circulating biomarker. In this scenario, the waiting time between the most used LB biomarkers, as shown in the clinical trials (reported in Table 2), is lower for CTCs rather than cf-DNA. This difference is potentially due to the high specificity and sensibility of CTCs compared to the other circulating biomarkers in the LB panel. is the period of latency between the first IO treatment and biomarker collection. During this period, interfering factors existing in the blood are produced prevalently by healthy tissue around the treated tumor lesion. Monitoring of minimal residual disease (MRD) begins with the first specific circulating tumor biomarker available. CTCs, being highly specific to tumor disease, are candidates to be a good marker for qualifying MRD by reducing the MST duration induced by IO. After MST, further combined analysis of CTCs and ct-DNA could represent a better surveillance approach for a longitudinal scan CTCs can inform about the persistence of MRD and, from their characterization, also about the type or area of the tumor that restarts its proliferative behavior. Ct-DNA can provide information on clonal resistance in terms of mutations. Moreover, isolation of CTCs and short-time culture offer the opportunity to understand the survival of the prevalent clonal subset or the cause of the growing relapse through their direct sequencing (in situ hybridization, single-cell sequencing) or their immunocytochemical characterization. Recently [17,18], it has been demonstrated that the molecules secreted by cultured cells during in vitro short culture, the so-called secretome, are enriched by cytokines and growth factors. The characterization of the secretome, after IO, could represent an interesting point of view in the characterization of MRD to identify an integrative therapeutic approach. Moreover, the enhanced effect of IO for immunotherapy [19] found in the characterization of the secretome of CTCs is a further point to monitor in real time the interplay between tumor and immune system [17]. Conclusions In the era of immunotherapy alternatives for MRD-directed therapy and treatment of frankly relapsed disease, it is yet imprecise which patients with MRD-positive disease will receive the most benefit from the chosen treatment or therapeutic consequences. It remains uncertain the importance of repeating MRD assessments in patients who contract MRDnegative disease after treatment. Therefore, before a patient relapses, a transition through an MRD-positive state is required, and

this may be an opportunity for intervention. It is uncertain at this point whether achieving deeper molecular responses is clinically significant or if additional therapy produces deeper molecular responses. In the future, we need to focus on these questions to minimize exposure to the adverse effects associated with IO therapy, including cytokine release syndrome, as well as for important health and economic reasons. In the future, it is important that the complicated instrumentation used to detect and characterize circulating biomarkers [41,42] becomes less complicated by using lab-onchip and point of care able to inform, in real time, clinicians about tumor dynamism [43,44] before and after an oncology intervention [45]. Moreover, the combination of LB and imaging can be addressed to improve the compliance of cancer patient management for reducing MRD and preventing tumor relapse. LB is a rapidly evolving technology with potentially relevant consequences in the management of patients with cancer, probably destined to modify oncological therapeutic strategies in the next decade. Despite its high potential, the LB approach cannot provide spatial information and still lacks the standardization, specificity, and sensitivity required to be extensively implemented in routine clinical practice. Interventional oncologists should accept this new technology to potentiate the synergies between interventional techniques and LB to improve cancer diagnosis and therapy. Interventional techniques may help LB technology overcome sensitivity limitations. Furthermore, LB techniques may improve the application and enhance the outcomes of minimally invasive locoregional treatments. Table 1. T.F. prepared Table 2. G.K., C.R., F.B., P.C., R.I. and L.T. reviewed the manuscript. All authors have read and agreed to the published version of the manuscript. Funding: This research received no external funding. Disease spread in age structured populations with maternal age effects Abstract Fundamental ecological processes, such as extrinsic mortality, determine population age structure. This influences disease spread when individuals of different ages differ in susceptibility or when maternal age determines offspring susceptibility. We show that Daphnia magna offspring born to young mothers are more susceptible than those born to older mothers, and consider this alongside previous observations that susceptibility declines with age in this system. We used a susceptible‐infected compartmental model to investigate how age‐specific susceptibility and maternal age effects on offspring susceptibility interact with demographic factors affecting disease spread. Our results show a scenario where an increase in extrinsic mortality drives an increase in transmission potential. Thus, we identify a realistic context in which age effects and maternal effects produce conditions favouring disease transmission. INTRODUCTION The ecological context of a host species will shape key demographic features, which in turn affect disease spread. Extrinsic mortality, for example, will regulate population density, a well-established demographic feature that modulates disease spread. However, extrinsic mortality will shape other population characteristics such as age structure. The age structure of a population could affect disease dynamics through a number of mechanisms, including (1) when individuals of different ages show different susceptibilities (i.e. age effects) or (2) when the offspring of mothers of different ages show different susceptibilities (i.e. maternal age effects). The age of a host at exposure to a parasite is well established to affect the probability and severity of infection, this is partly due to the ontogeny of the immune system (Lesser et al. 2006;Nussey et al. 2008;Hasselquist & Nilsson 2009). In vertebrates, this includes the development of adaptive immunity followed by immunosenescence, but invertebrates also appear to show a clear development of the immune system or pathogen resistance (Rheins & Karp 1985;Wilson-Rich et al. 2008;Piñera et al. 2013;Garbutt et al. 2014a;. The consideration of age-related effects on epidemiological processes has contributed greatly to disease prevention strategies (Anderson & May 1982;Katzmann & Dietz 1984;M€ uller 2000). By contrast, maternal age effects on offspring susceptibility to pathogens have received little attention. Generally, maternal effects have been shown to affect population dynamics and demography through alterations in offspring reproduction, maturation and growth rate (Gaillard et al. 2003;Benton et al. 2005;Beamonte-Barrientos et al. 2010), and maternal effects have been shown to alter population robustness in the face of ecological challenges (R€ as€ anen & Kruuk 2007;Kuijper & Hoyle 2015). The maternal condition, for example nutrition availability or disease status, is increasingly recognised to affect offspring susceptibility to infection (Huang & Song 1999;Little et al. 2003;Rahman et al. 2004;Mitchell & Read 2005;Gasparini et al. 2007;Lorenz & Koella 2011;Stjernman & Little 2011;Tidbury et al. 2011;Boots & Roberts 2012;Garbutt & Little 2017). This suggests variation in phenotypic traits, as a response to the maternal environment, can drive population and epidemiological dynamics (Beckerman et al. 2002;Mitchell & Read 2005;Garbutt et al. 2014b). Maternal age has been shown to affect offspring performance as offspring of older mothers are often born larger and mature at a greater size (Priest et al. 2002;Benton et al. 2008). However, although older mothers produce larger offspring, they tend to produce fewer of these, suggesting they are subject to a 'size vs. number' trade-off. These effects on performance measures indicate that maternal age effects may alter the competitive environment experienced by successive generations (Beckerman et al. 2006;Kindsvater et al. 2011;Prior et al. 2011). To understand how maternal age influences susceptibility, we studied four clutches of offspring from the parthenogenic crustacean Daphnia magna when exposed to a bacterial pathogen. In addition to pathogen resistance, we collected data on reproduction and body size. The experiment showed that older mothers give birth to more resistant offspring. Earlier work from our laboratory (Garbutt et al. 2014a) and others showed that older mothers themselves are more resistant to this pathogen. We used these two observationsthat older mothers are both more resistant and give birth to offspring that are more resistantto develop a compartmental model describing how population age structure, the effects of age and maternal effects combine to influence the spread of infection. STUDY ORGANISMS Daphnia magna (Crustacea: Cladocera) are filter feeding planktonic crustaceans found in small freshwater ponds. Pasteuria ramosa is a sterilising specialist bacterial pathogen of D. magna that is transmitted horizontally when spores released from infected cadavers are ingested by uninfected filter feeding hosts (Ebert et al. 1996(Ebert et al. , 2016. In this experiment, we used a Daphnia clone (Kc49a) from a population in a pond in Kaimes in the Scottish borders, and a P. ramosa strain isolated from the same population. This host clone was part of a study of genetic variation in maternal effects (Stjernman & Little 2011) and shows a response that is typical of its population (see Little & Colegrave 2016 for discussion of the merits of single vs. multigenotype studies). METHODS The basic design of our experiment was to measure the pathogen susceptibility of offspring from first, second, third and fifth clutches. Two newborn Daphnia were isolated from each clutch. One of these offspring was exposed to P. ramosa. The second offspring was used for measurement of reproduction. Each newborn also had their body size measured before being placed into their treatment groups. This experiment ran from July 31-September 24 2014. Acclimation Twenty-four replicates, each an individual Daphnia in a 60 mL glass jar, were acclimatised for three generations under standardised conditions. This process is designed to equilibrate uncontrolled maternal effects and ensure that each replicate is independent. During this time, Daphnia were kept in artificial pond medium (Kluttgen et al. 1994) in an incubator with a light:dark cycle of 12 : 12 L:D at 20°C. They were fed daily with 7 9 10 6 cells of green algae, Chlorella spp., and changed into fresh medium twice a week and when offspring were present in the jar. Each new generation was initiated with offspring from the second clutch. After the third homogenising generation, one individual was taken from each replicate and placed in fresh jars to be the mothers of the experimental animals. These individuals were kept in the same conditions as the homogenising generations. Experiment 1 Offspring from clutch 1, 2, 3 and 5 of the maternal generation were our experimental animals. Two offspring were taken from each mother and were randomly assigned to either pathogen exposure or controlused for the measurement of fecundity in the absence of infection. For pathogen exposure, 20 000 P. ramosa spores were added to each jar and media was not changed for 5 days. Spore mixes were prepared earlier by crushing infected Daphnia and counting spores with a haemocytometer. At the end of the 5-day exposure treatment, Daphnia were changed into clean jars with fresh media. Daphnia were maintained for 28 days after the exposure period under the conditions used for acclimation. During this period, the date of birth of each clutch and number of individuals born to each clutch was recorded. At the end of the 28-day period, infections were diagnosed (infections are easy to discern with the naked eye as Daphnia have a clear carapace and reddish-brown bacterial growth is visible in the haemolymph in addition to the lack of reproduction). The control Daphnia whose fecundity was measured were handled identically, except they were not exposed to parasite spores. All newborn Daphnia were photographed for later measurement of size at birth. Independent replication of experiment 1 The design and methodology of this experiment were identical to that detailed above: we measured the fitness of offspring from first, second, third and fifth clutches following identical protocol for pathogen exposure and control individuals. Analysis Body size at birth and total reproduction were analysed in a general linear model. Age at first reproduction was studied with a parametric survival analysis following a Weibull distribution. A generalised linear model (with binomial errors and logit link function) was used to study probability of becoming infected. In all cases, the explanatory variable was the clutch the Daphnia originated from. The different response variables were analysed with different subsets of the data: size at birth considered the entire data set, reproduction considered only Daphnia that were unexposed to P. ramosa, while the probability of infection included only Daphnia that were exposed to the pathogen. EXPERIMENTAL RESULTS The age of the mother had a significant effect on the probability of offspring infection (v 2 = 24.8, P < 0.0001) where offspring from young mothers were highly susceptible to infection, as shown in Fig. 1a. Maternal age (or variable 'Clutch' for clutch number) also had a significant effect on total reproduction (F 3,72 = 15.8, P < 0.0001) as seen in Fig. 1b and size at birth as seen in Fig. 1c (F 3,91 = 219, P < 0.0001). Age at first reproduction was not influenced by maternal age (v 2 = 5.07, P = 0.16). These patterns were confirmed through the independent replication of the experiment. Maternal age had a significant effect on the probability of becoming infected (v 2 = 20.6, P < 0.0001; Fig. 1a), total reproduction (F 3,76 = 3.91, P < 0.002; Fig. 1b) and size at birth (F 3,105 = 96, P < 0.0001; Fig. 1c). Our experimental results showed a strong effect of maternal age on offspring resistance, with older mothers producing more resistant offspring. Previous research showed that D. magna show age-specific susceptibility, where older individuals are more resistant to infection than young individuals (Garbutt et al. 2014a;. Host population demography and environmental factors, such as baseline mortality rates, predation and density, have been shown to influence transmission rates and pathogen virulence evolution. Pathogen transmission is a function of an infected individual's rate of contact with other potential hosts. As increased host population density should increase contact rates, it is generally expected that an increase in population density will result in an increase in transmission potential (Anderson & May 1979;Ebert & Mangin 1997;Choo et al. 2003). The theoretical consideration of age-specific effects on epidemiological dynamics tends to consider age-specific mortality and age-specific contact rates rather than the strength or weakness of an ageassociated immune response (Anderson & May 1982;Katzmann & Dietz 1984;Castillo-Chavez et al. 1989;M€ uller 2000). In addition to this, the presence of maternal age effects on offspring susceptibility has not been considered. We therefore developed a compartmental model incorporating these age effects on the expected number of secondary infections resulting from the introduction of an individual infected with a novel pathogen into a completely susceptible population (R 0 ). R 0 depends on the duration of infection, the probability of infecting a susceptible during one contact and the rate of contact per unit of time (Dietz 1993). The

output value can be used as a benchmark. If the value of this is > 1, then the disease will spread (Anderson & May 1979). We divide the population into four age classes, each of which could be infected (I) or uninfected (U), giving a total of eight classes of individuals in the model - (Fig. 2). Here, the first subscript indicates the individual's age (Y for young, O for old), and the second subscript indicates the individual's mother's age at that individual's birth. The parameters in the model are: b is the baseline transmission rate between infected and uninfected individuals; d is the baseline death rate; a is pathogen virulence as measured by disease induced death; p is the probability that non-pathogen induced death of an infected individual will also lead to transmission; m is the rate of maturation (i.e. the rate at which individuals move from the young age classes to the old); M describes the proportional reduction in an individual's susceptibility as associated by having a mother of the age class 'Old' applied to the baseline b; A describes the proportional reduction to an individual's susceptibility by themselves being of the age class 'Old' applied to the baseline b; r is the maximum per capita growth rate of hosts; K is the carrying capacity of the host population, controlling density dependent limitation on reproduction. The transmission terms, as shown in Fig. 2, contain the rates of death, due to P. ramosa being transmitted only upon host death. We make the assumptions that there is a constant rate of death across age groups that susceptibility varies between these classes, and that reproduction is resource limited with a constant maximum reproductive rate per day for uninfected individuals. We assume that infected individuals do not reproduce as P. ramosa castrates the host during infection (Ebert et al. 1996). For full details of the model, see the appendix. In the absence of the pathogen, the equilibrium densities of each age class are Here, we see that the equilibrium density of U O,O decrease with increasing mortality d (as r > d is necessary for a positive equilibrium). The equilibrium density of both U Y,O and U O,Y initially increases with increasing mortality before decreasing once d [ 2r þ m. Similarly, the density of the age class U Y,Y initially increases with increasing mortality before decreasing (Fig. 3). This bias of the age structure towards more susceptible younger individuals and individuals from younger mothers occurs as increased mortality frees up resources for reproduction, resulting in the production of more young individuals who will themselves reproduce leading to more individuals from young mothers. At this equilibrium, the total density of uninfected hosts, N U , is meaning that the total density of uninfected individuals monotonically declines with increasing mortality rate (Fig. 3). To calculate the R 0 of the pathogen, we consider the number of secondary infections caused by a rare pathogen at the pathogen-free equilibrium given in eqns 1a-d. This yields a pathogen R 0 of In the appendix, we show that the relationship between pathogen R 0 and the extrinsic death rate, d, must be negative whenever our age-related and maternal age-related effects on susceptibility are absent (i.e. A = 0, M = 0). However, this relationship is humped whenever with this condition satisfied for sufficiently large reductions in susceptibility with age and maternal age (high A and M). The relationship is illustrated with and without both the age-and maternal age-related reductions in susceptibility in Fig. 4. These qualitative results also hold when an environmental transmission stage of the pathogen is explicitly included in the model (see Appendix). The humped shape can be explained as follows. In the absence of reductions in susceptibility with age and maternal age, reductions in host population density with increasing mortality reduce pathogen R 0 owing to reduced transmission. However, increased mortality also shifts the age structure of the population towards younger individuals and individuals from younger mothers, which can increase in density even as the total population density declines. If these age classes of individual are sufficiently more susceptible to infection, this increase in their density can more than compensate for the total reduction in host population density, leading to an increase in R 0 with increasing non-pathogen induced mortality. However, at very high mortality rates the total host population density is sufficiently reduced that shifts in age structure no longer compensate, and R 0 declines with mortality, resulting in a humped relationship. Transmission potential begins to decline at lower mortality rates when, p (the probability that non-pathogen induced death of an infected The compartmental model with uninfected and infected groups of four age classes. UY;Yyoung individuals with a young mother. UY;Oyoung individuals with an old mother. UO;Y-old individuals with a young mother. UO;Oold individuals with an old mother. m is the rate at which an individual goes from being a young individual to an old individual. b is the rate of transmission. d is the baseline death rate, with d +a representing baseline death rate plus a measure of virulence. Other parameters are rmaximum reproduction; Kcarrying capacity; Ntotal population number; pthe probability that non-pathogen induced death of an infected individual will lead to transmission. As p. ramosa is only transmitted upon host death, death rates are included in the transmission terms. individual will lead to transmission) is low, as increasing mortality will result in loss of infections with little opportunity for transmission. Note, however that our qualitative results hold even in the case of p = 0. DISCUSSION Our experiment makes the novel observation that younger mothers give birth to offspring that are more pathogen susceptible. This data combined with recent observations that younger hosts are themselves more susceptible (Garbutt et al. 2014a; formed the basis of a compartmental model showing that increasing extrinsic mortality, and thus decreasing total host population density, can lead to increasing transmission. The relationship between mortality and transmission that is classically seen, where higher host densities generate greater transmission, is so pervasive it has become a central assumption of the epidemiological theory that contributes to our understanding of pathogen traits and virulence. (Anderson & May 1979;Read 1994;Levin 1996;Alizon et al. 2009;Schmid-Hempel 2011). Thus, we have identified a new, and biologically realistic context in which age effects and maternal effects combine to alter the ecological conditions favouring disease transmission. Our observed effect of maternal age on offspring performance is an example of maternal effect-driven, delayed life history effects (Beckerman et al. 2002) that lead to whole cohort effects (Ginzburg et al. 1994;Lindstr€ om 1999). The maternal condition has previously been shown to have prolonged effects on offspring fecundity, maturation and growth rate over successive generations, influencing population dynamics (Gaillard et al. 2003;Benton et al. 2005;Beamonte-Barrientos et al. 2010). We now add offspring susceptibility to this list, and can infer from our model that maternal age effects on susceptibility can then impact successive cohorts, by altering the condition of each cohort at reproduction. Considering P. ramosa specifically, the predictions of our model are biologically feasible. This pathogen sterilises hosts, but often one clutch is produced before reproduction stops. If no effect of maturation or maternal effects were present, the transmission potential would decrease as seen in Fig. 4. In the presence of these effects, however, those who are young when infected only reproduce once whilst young and then do not contribute to the population at a later age. Their offspring are therefore expected to be the most susceptible. As mortality increases, the density of young individuals contributing highly susceptible individuals to the population increases, resulting in the dynamic we show. The maternal effect of increased resistance in offspring born to older mothers could be explained by size at birth. Older mothers are known to produce larger offspring (Marshall et al. 2010;Kindsvater et al. 2011) and therefore the size at birth could be responsible to some degree for the resistance to disease. This correlation of body size to susceptibility has been made before in Daphnia, and it has been suggested this is due to increased resources availability for costly immune defences and somatic maintenance (Garbutt et al. 2014a). In Daphnia spp. and other systems, however, this correlation between increased size and reduced infection probability does not always hold true (Hall et al. 2007;Stjernman & Little 2011). Furthermore, the assumption that increased body size equates to better fitness and performance does not explain how smaller offspring of young mothers are capable of reproducing in higher numbers than their larger counterparts. The observed trade-off between offspring size and the decreasing number of offspring in each successive reproductive event could be adaptive if clutches of older mothers were to experience more challenging environments than young mothers. This is plausible given population size could increase, increasing competition, and potentially pathogen prevalence. It has been seen previously that Daphnia produce larger and more resilient offspring in tough conditions (Mitchell & Read 2005;Garbutt et al. 2014b) and larger individuals are better competitors for resources in a more dense populations (Brockelman 1975). In addition, individuals born to older mothers may need to be larger in order to compete with their older siblings (Plaistow et al. 2007). In summary, our experimental results add evidence to a growing body of work showing maternal age effects on offspring performance (Berkeley et al. 2004;Plaistow et al. 2007;Marshall et al. 2010). Furthermore, our model shows that where age-related and maternal age-related effects occur they can fundamentally transform the ecological conditions that favour disease transmission. Divergent host susceptibilities within a population could also have marked effects on pathogen evolution and virulence, and our model lends itself to an extension that includes studies of optimal virulence. Figure 4 Modelling outputs of the relationship between mortality and transmission potential (R 0 Þ with three different parameter values for p (extrinsic mortality of infected individuals contributing to transmission). p = 1 (yellow line), p = 0.5 (blue line) and p = 0 (red line). (a) No age effects or maternal age effects present, (b) Maternal age effects, (c) Age specific susceptibility, or (d) -Both maternal and age effects present. In the presence of no effects (a), the expected negative relationship between mortality and transmission potential is shown at all levels of extrinsic mortality. The presence of maternal age effects (b), age-specific susceptibility (c) or both effects (d) within a population results in a humped relationship, where an increase in mortality initially increases transmission, due to the shift in density of susceptible individuals. This positive relationship between mortality and transmission potential is most pronounced at all levels of extrinsic mortality when both effects are present. Other parameters are: r = 3; K = 1; a = 1/5; m = 1/10; b Y;Y = 3.5. Therapeutic Drug Monitoring Characteristics in a Tertiary University Hospital in 2019 Introduction Therapeutic drug monitoring (TDM) is defined as measuring drug concentration in a biological sample to optimize pharmacotherapy. This study aims to evaluate TDM requests in a tertiary university hospital retrospectively. Materials and methods TDM requests were evaluated retrospectively for lithium, valproic acid, carbamazepine, and digoxin in 2019. The age and gender of the patient, requesting department, and measurement results were evaluated. Lower levels than the reference values were considered as subtherapeutic, while levels higher than the reference were considered as toxic. Results A total of 415 drug level measurement records were found. The pediatric age sample ratio was 13.7%, and the elderly age sample ratio was 11.8%. When all samples were evaluated according to the relevant laboratory cut-off values, 72.8% of samples were within the therapeutic level range, 21.9% of samples were subtherapeutic, and 5.3% were toxic. The pediatric age group had a higher ratio of toxic levels for the four drugs studied (54.5%). Conclusions Tests for lithium, valproic acid, carbamazepine, and digoxin would not be considered sufficient for TDM. Multidisciplinary teamwork might be appropriate for further implementation and interpretation of TDM. Introduction Therapeutic drug monitoring (TDM) is the clinical practice of measuring the drug concentration in biological samples to optimize pharmacotherapy. It has been used since the 1970s to provide better drug response and avoid adverse effects. Identification of non-compliant patient, personalization of dosage, and investigation of non-response are some benefits of TDM. There are many prescribed drugs, but few have a clinical impact on TDM [1]. Drugs that

are candidates for TDM should satisfy pharmacological properties such as a narrow therapeutic index, significant pharmacokinetic variability, presence of a reasonable relationship between drug concentration and clinical effect, and an established target for concentration range [2]. There are forthcoming drugs, for which TDM is accepted as a standard of medical care. Examples are aminoglycoside antibiotics, cardiac glycosides, antiepileptic drugs, digoxin (DIGOX), theophylline, cyclosporin, tacrolimus, methotrexate, and salicylates, when used in high doses [2,3]. TDM requiring drugs might be indicated in epilepsy, mood disorders, heart failure, respiratory disease, and neoplastic or rheumatologic disease treatment. This study aims to evaluate TDM requests retrospectively in a tertiary university hospital. Materials And Methods Ethical approval was obtained from the ethical committee of Maltepe University Medical Faculty (issue number: 2019/900/08). TDM requests were evaluated retrospectively from electronic records in 2019. Both inpatient and outpatient data were included in the study. Patient age and gender were obtained from the records. The age data were stratified as 0 to 18 years as pediatric, 18 to 65 years as an adult, and > 65 years as elderly. The requesting department and measurement results were evaluated. Therapeutic ranges were assigned based on routine laboratory ranges. Lower levels than reference values were considered as subtherapeutic, and higher levels than reference values were considered toxic. All data were expressed as mean ± standard error of the means (SEM) [4]. The evaluation was performed by descriptive statistics and, to assess categorical data, a Pearson's chi-squared test was used. GraphPad Prism V.8.01 (San Diego, CA) and IBM SPSS Statistics for Windows, Version 25.0. (IBM Corp., Armonk, NY) were used for statistical analysis and graphing. Results A total of 415 drug level measurement records were found for 2019. The mean age was 40.4 years. The pediatric age sample ratio was 13.7%, and the elderly age sample ratio was 11.8%. Fully 74.5% of blood samples were drawn from patients in the adult age group. The gender ratio was 33% men to 67% women. The drugs monitored were lithium (LITHIUM), valproic acid (VALP), carbamazepine (CARB), and DIGOX. When all samples were evaluated according to the relevant laboratory cut-off, 72.8% of the samples were within the therapeutic level range, 21.9% of the samples were subtherapeutic, and 5.3% of the samples were toxic. The most frequent TDM request was LITHIUM, while the least frequent was DIGOX ( Table 1). The samples for LITHIUM and VALP TDM resulted 25.9% (52) and 19.9% (37) in the subtherapeutic levels, respectively. The highest number of samples was 144 and in LITHIUM therapeutic range group ( Figure 1). DIGOX was within the therapeutic range the least (50%). When age groups (pediatrics, adult, elderly) and drug level groups (subtherapeutic, therapeutic, toxic) were evaluated, the differences were found to be statistically significant. The pediatric age group had a higher ratio of toxic drug levels than the other age groups. The department requesting TDM differed based on the drug ( Figure 2). The three departments that most frequently requested drug level measurement were psychiatry (n=301), pediatrics (n=53), and neurology (n=20). Pediatric neurology requests were counted in pediatrics (n=35) and made up 66% of pediatrics requests. On the other hand, VALP and LITHIUM were most frequently requested TDMs by psychiatry specialists, CARB was from pediatrics, and DIGOX was from intensive care ( Figure 2). DIGOX level measurements were requested by intensive care specialists (37.5%) and by cardiology specialists (25%). Discussion The present retrospective study involved a one-year evaluation of 415 blood samples for TDM in approximately 200 beds and 15,000 outpatient/years in a tertiary university hospital setting. The drug requests were CARB, VALP, DIGOX, and LITHIUM. The most commonly monitored drugs mentioned were CARP, VALP, and DIGOX; these were among our monitored drugs [2]. The sample size was comparable with a four-year study from Turkey that evaluated a total of 7,759 measurements in a university hospital with more than 2,500 beds [5]. The study results indicated that the pediatric age group had a lower frequency of samples within the therapeutic range. This might be due to the fact that many of the drugs being monitored in the pediatric population were developed based on adult drug pharmacokinetics and disease states, which might be quite different from the pediatric population [6]. VALP was a frequently prescribed drug for TDM in neuropsychiatric disorders, with requests from neurology, pediatrics, and psychiatry departments. Age consideration was important for VALP as its use was not recommended in geriatric patients; however, 8.6% (n=16) of the patients in our study were elderly. The clinical need for VALP in this population should be evaluated as it is potentially inappropriate for older adults. In all age groups, subtherapeutic level results were found in 19.9% of samples. This might be interpreted as undertreatment. VALP metabolism was also altered with CYP2C9 and CYP2A6 polymorphisms [7]. Pharmacogenetics testing might contribute to a better understanding of the pharmacokinetic variability observed between individuals. LITHIUM was frequently found in our TDM samples. Nearly one-quarter (25.9%) of the results were for subtherapeutic levels, while 71.6% were within the therapeutic range. Even though it was frequent in our study, it was much more than expected. In Japan, regulatory warnings for LITHIUM have been shown to be effective in increasing TDM for the drug [8]. DIGOX, a cardiac glycoside, had the lowest ratio for being in the therapeutic range. Intensive care (37.5%) constituted the primary requesting department. This might indicate that DIGOX was monitored primarily for toxicity in intensive care. In this hospital setting, the results for DIGOX were returned in a few days and might be the reason it did not apply in outpatient settings. Even though TDM for DIGOX was highly recommended in the outpatient setting, a lower than the maximum therapeutic dose may have been believed to be safer for the treatment of chronic heart failure [9]. As an antiepileptic, CARB had the highest ratio of toxic levels and the least number of samples. There was a single CARB result for a sample below the lower limit; that particular sample might have been a request to monitor for adherence to therapy or a differential diagnosis for toxic clinical presentation. Our therapeutic range was only different for CARB in cases of concomitant antiepileptic use. This fact might support that CARB should be better disseminated for monitoring with TDM. All biological samples were from blood, though other biological fluids might be eligible for drug monitoring. The pediatric age group samples correlated with a lesser degree of being in the therapeutic range. For a better drug response, other biological fluids that can be obtained noninvasively, such as saliva, should be considered for TDM. Saliva samples may provide accurate levels for some drugs and might be chosen in the pediatric age group [10]. Limitations of the study include the lack of information for appropriateness of collecting time, the purpose for requesting TDM, and the presence of concomitant drugs. The time of blood collection, perhaps more than any other factor, can contribute to the misinterpretation of drug levels [11]. In the study, TDM was requested by the relevant departments. The blood collection time accuracy was considered appropriate as patients are informed about the accurate time. There are many indications for TDM, such as monitoring drug use compliance for individualizing pharmacotherapy, diagnosing undertreatment, or checking for drug toxicity [12]. Protein binding was another major factor that varies by drug, but that also can be altered by drug-drug and drug-disease interactions. [11]. Some of the patients were using more than one antiepileptic or other drugs. Time intervals after a drug were taken, drug-drug, drugdisease, and drug-nutrient interactions also should be evaluated for accurate TDM interpretation. TDM is a valuable tool to ameliorate pharmacotherapy to increase the efficacy and decrease the adverse drug reactions. Most patients had a therapeutic response within the reference range, but many patients might need target concentrations outside the reference range [13]. Interpretation of drug concentrations should involve clinical monitoring for efficacy and adverse drug reactions, and consider drug-drug and drug-disease interactions, as well as variability in pharmacogenetics. The present study in the university hospital setting prompted psychiatry, pediatrics, neurology, cardiology, and intensive care departments that were more relevant to TDM practice. In these departments, four drugs were requested for TDM. Interestingly, phenytoin, an antiepileptic drug, was not requested even though it was in the recommended drug list for TDM and has clinically important drug-drug interactions with newer antiepileptics [2,13]. TDM of aminoglycoside antibiotics, theophylline, cyclosporin, tacrolimus, methotrexate, and salicylate was not in the requested samples. The TDM reporting duration advised and indicated to be important was preferably within 24 hours of dosing, especially during dosage adjustments, and in diagnosing toxicity [12]. Our laboratory fulfills these criteria only for VALP and LITHIUM; these were found to be the most requested. Our study result yielded that the pediatric age group might be more vulnerable in maintaining a therapeutic range. Non-invasive monitoring with saliva samples for eligible drugs might be an option to maximize TDM. Conclusions In our study, a total of 415 drug level measurement records were found. The drugs monitored were LITHIUM, VALP, CARB, and DIGOX. When the samples were interpreted according to laboratory cut-off, 72.8% of the samples were within the therapeutic level range, 21.9% of the samples were subtherapeutic, and 5.3% of the samples were toxic. A multidisciplinary team approach is essential for appropriate TDM. LITHIUM, VALP, CARB, and DIGOX, as the requested tests in our hospital, would not be considered sufficient for TDM management. Pharmacologic aspects, laboratory analysis, and clinical monitoring are the three main components of the teamwork required. Biochemistry, pharmacology, and relevant departments should connect to interpret TDM for optimizing pharmacotherapy. Additional Information Disclosures Human subjects: Consent was obtained by all participants in this study. Ethical committee of Maltepe University Medical Faculty issued approval 2019/900/08. Ethical approval was obtained from the Ethical committee of Maltepe University Medical Faculty. Animal subjects: All authors have confirmed that this study did not involve animal subjects or tissue. Conflicts of interest: In compliance with the ICMJE uniform disclosure form, all authors declare the following: Payment/services info: All authors have declared that no financial support was received from any organization for the submitted work. Financial relationships: All authors have declared that they have no financial relationships at present or within the previous three years with any organizations that might have an interest in the submitted work. Other relationships: All authors have declared that there are no other relationships or activities that could appear to have influenced the submitted work. Discovery of a selective, state-independent inhibitor of NaV1.7 by modification of guanidinium toxins The voltage-gated sodium channel isoform NaV1.7 is highly expressed in dorsal root ganglion neurons and is obligatory for nociceptive signal transmission. Genetic gain-of-function and loss-of-function NaV1.7 mutations have been identified in select individuals, and are associated with episodic extreme pain disorders and insensitivity to pain, respectively. These findings implicate NaV1.7 as a key pharmacotherapeutic target for the treatment of pain. While several small molecules targeting NaV1.7 have been advanced to clinical development, no NaV1.7-selective compound has shown convincing efficacy in clinical pain applications. Here we describe the discovery and characterization of ST-2262, a NaV1.7 inhibitor that blocks the extracellular vestibule of the channel with an IC50 of 72 nM and greater than 200-fold selectivity over off-target sodium channel isoforms, NaV1.1–1.6 and NaV1.8. In contrast to other NaV1.7 inhibitors that preferentially inhibit the inactivated state of the channel, ST-2262 is equipotent in a protocol that favors the resting state of the channel, a protocol that favors the inactivated state, and a high frequency protocol. In a non-human primate study, animals treated with ST-2262 exhibited reduced sensitivity to noxious heat. These findings establish the extracellular vestibule of the sodium channel as a viable receptor site for the design of selective ligands targeting NaV1.7. www.nature.com/scientificreports/ recovery from inactivation ( Fig. 1B) [13][14][15][16] . By contrast, cationic guanidinium toxins and peptide cone snail toxins inhibit ion conduction by sterically occluding the extracellular vestibule of the channel pore (Site 1). The former are a unique collection of small molecule natural products exemplified by saxitoxin and tetrodotoxin-high affinity, state-independent blockers against six of nine Na V subtypes ( Fig. 1C) 17 . In the pursuit of isoform-selective inhibitors of Na V 1.7, two binding sites, the cystine knot toxin site at VSD II and the sulfonamide site at VSD

IV, have been heavily interrogated. Certain cystine knot toxins that engage VSD II such as HwTx-IV, Pn3a, and ProTx-II exhibit 6-1,000× selectivity for Na V 1.7 over other channel isoforms. Potency and selectivity for this target have been improved with synthetic peptide toxin derivatives [22][23][24][25][26] . Among small, Lipinski-type molecules, only the aryl and acyl sulfonamides pioneered by Icagen/Pfizer and subsequently investigated by Amgen, Chromocell, Genentech/Xenon, Merck, and others have shown evidence of significant Na V 1.7 isoform selectivity 7,16 . Within the sulfonamide series, selectivity levels are > 1,000× over certain off-target isoforms including the cardiac isoform, Na V 1.5, but generally 10-50× against Na V 1.2 and Na V 1.6. Many but not all sulfonamide Na V 1.7 inhibitors suffer from undesirable pharmaceutical properties, including high plasma protein binding (e.g. > 99.8%), cytochrome p450 inhibition, in vitro hepatotoxicity and high unbound clearance 27,28 , which have hindered clinical development. Although a number of candidates have been advanced to human testing, one compound has been discontinued after a Phase 2 study likely due to limited efficacy (PF-05089771); others have been discontinued after Phase 1 trials for reasons that may be related to safety liabilities such as elevated expression of liver transaminases and hypotension (GDC-0276) 29,30 . Electrophysiology studies with naturally occurring Site 1 ligands against different wild-type and mutant Na V isoforms have identified the extracellular vestibule of Na V 1.7 as a promising locus for selective inhibitor design [31][32][33] . The outer mouth of the channel is formed from residues that link the S5-S6 helices (referred to as pore loops) from each of the four domains. The domain III pore loop of human Na V 1.7 contains a T1398/I1399 sequence motif that is not present in other human Na V subtypes (which contain MD at equivalent positions, Suppl Table 1) 31 . Comparison of the amino acid sequence of the domain III pore loop across species indicates that the sequence motif in hNa V 1.7 is unique to primates. The half-maximal inhibitory concentration (IC 50 ) value for saxitoxin (STX) is markedly altered (250-fold change) depending on the presence or absence of the T1398 and I1399 residues. Against rNa V 1.4, the IC 50 of STX is 2.8 ± 0.1 nM compared to 702 ± 53 nM for hNa V 1.7 31 . Introduction of the alternative domain III pore loop sequence by mutagenesis restores potency (hNa V 1.7 T1398M/I1399D IC 50 = 2.3 ± 0.2 nM). These findings suggest that it may be possible to capitalize on structural differences in the extracellular vestibule between hNa V isoforms to design Na V 1.7-selective inhibitors. Recent advances in the de novo synthesis of guanidinium toxin analogues have enabled systemic examination of the structure-activity relationship (SAR) properties that govern hNa V 1.7 potency and isoform selectivity [34][35][36][37] . Prior to 2016, the binding orientation of STX proposed in the literature indicated that the C11 methylene carbon was positioned proximally to domain III pore loop residues [38][39][40] . SAR and mutant cycle analysis studies posited a revised pose in which the C13 carbamate moiety abuts DIII 32 . This revised binding pose was recently confirmed by cryoelectron microscopy (cryo-EM) structures of STX bound to Na V PaS and hNa V 1.7 18,41 . In the present study, analogues of STX substituted at both the C11 and C13 positions were investigated to understand the requirements for selective inhibition of hNa V 1.7. These efforts led to the discovery of ST-2262, a potent and selective inhibitor of Na V 1.7 that reduces sensitivity to noxious heat in a preliminary study in non-human primates (NHPs). Discovery of ST-2262. ST-2262 was discovered through a rational design strategy aimed at identifying derivatives of natural bis-guanidinium toxins that preferentially inhibit hNa V 1.7 over other off-target hNa V isoforms 31 . Mutagenesis, homology modeling, and docking studies conducted prior to 2016 suggested that bis- [38][39][40] . Exploration of substitution at C11 of decarbamoyl saxitoxin (dcSTX) led to the identification of a series of analogues bearing aryl amide groups at this site. Certain compounds, as exemplified by ST-282, show excellent potency against hNa V 1.7 but minimal selectivity (~ 1:1) over off-target isoforms such as hNa V 1.4 (Fig. 2B). The finding that hNa V 1.7 isoform selectivity could not be achieved by modification of the C11 substituent led us to investigate SAR at alternative positions. These studies followed evidence that the proper binding orientation of STX is rotated ~ 180° from earlier models, thus placing the C13 substituent in close proximity to domain III ( Fig. 2A, Revised pose) 32 . Derivatives of STX bearing amide, carbamate, ester, ether, and urethane substituents at the C13 position were prepared in an effort to identify compounds with improved selectivity for hNa V 1.7. Insight from studies of a naturally-occurring STX C13 acetate congener, STX-OAc, helped guide selection of compounds for synthesis (Suppl Figure 1) 32 . The difference in potencies between STX and STX-OAc is striking considering that these two structures vary at a single position (NH 2 → CH 3 ). Following this lead, we explored substituents at C13 that could replace the hydrolytically unstable acetate group. Ultimately, the C13 succinimide was discovered as a suitable acetate isostere, which was paired with a C11 tetrahydronaphthyl amide to generate ST-2262, the focus of the present study. Table 2). Potencies against off-target sodium channel isoforms (hNa V 1.1-1.6, hNa V 1.8) were determined following a similar protocol. Activity against hNa V 1.9 was not evaluated due to the difficulty of expressing this subtype heterologously 42 . ST-2262 was determined to be > 200-fold selective over hNa V 1.6 (IC 50 = 17.9 µM, 95% CI 14.8-22.1), > 900-fold selective over hNa V 1.3 (IC 50 = 65.3 µM, CI 62.7-68.1), and > 1,000-fold selective over all other Na V isoforms tested. Similar IC 50 values against the eight hNa V subtypes were obtained in an independent study using the PatchXpress automated electrophysiology platform (Suppl Table 3). ST-2262 is a potent and selective inhibitor of hNa Exposure of hNa V 1.1 and hNa V 1.2 to high concentrations of ST-2262 (10-100 µM) resulted in a reduction of the rate of fast inactivation; a similar effect was noted, albeit to a lesser degree, with hNa V 1.3 and hNa V 1.4 (Suppl Figure 2). Lower concentrations of ST-2262 (1-3 µM), which remain sufficiently high to achieve > 90% inhibition of hNa V 1.7, had no measurable effect on fast inactivation of hNa V 1.1 and hNa V 1.2. It is possible that elevated concentrations of ST-2262 result in a secondary mode of binding against these Na V subtypes, however, efforts have not been made to examine such a mechanism at this time. To our knowledge, changes in the rate of fast inactivation have not been observed with STX. To investigate whether the potency of ST-2262 was dependent on the membrane holding potential or frequency of stimulus, an IC 50 value was measured against hNa V 1.7 using a two-pulse protocol with a pre-pulse to the voltage at half-inactivation (8 s step) and with a protocol that depolarizes the cell at high frequency (30 Hz stimulus). The potency of ST-2262 was not appreciably altered using either stimulation protocol (IC 50 Table 4). These results indicate that ST-2262 is a selective, use-independent inhibitor of hNa V 1.7. Species variation in potency and mutagenesis. The potency of ST-2262 was assessed against a panel of species variants of Na V 1.7, including mouse, rat, and cynomolgus monkey (Suppl Table 5). Consistent with the hypothesis that Na V 1.7 potency is affected by the presence of the T1398/I1399 sequence motif in the DIII pore loop, the IC 50 of ST-2262 against cynoNa V 1.7 (0.101 µM, 0.073-0.140) was similar to human. In contrast, ST-2262 was > 50 × less potent against mouse (IC 50 = 3.78 µM, 3.23-4.43) and rat Na V 1.7 (IC 50 = 4.95 µM, 4.17-5.87) than the human ortholog. Affinity was restored within twofold of the hNa V 1.7 potency by introduction of domain III MD-TI mutations to mouse Na V 1.7 (IC 50 = 0.130 µM, 0.055-0.307; Fig. 3C, Suppl Table 6). Multiple lines of evidence suggest that ST-2262 binds to the extracellular vestibule of the sodium channel (i.e., Site 1) including: (i) the structural similarity of ST-2262 to natural bis-guanidinium toxin ligands, (ii) the Fig. 3D, Suppl Table 6). Collectively, these results indicate that ST-2262 binds to the extracellular vestibule of Na V 1.7, displaying significant species variation in potency and isoform selectivity in large part due to molecular interactions with residues T1398 and I1399, which are unique to human and nonhuman primate Na V 1.7 orthologs 31,32 . www.nature.com/scientificreports/ Na V 1.7 in species that lack the T1398/I1399 sequence motif (Suppl Table 5). A NHP model of acute thermal pain was identified that uses a heat lamp to deliver a stimulus to the dorsal surface of the hand of lightly anesthetized cynomolgus macaques and measures the time to withdrawal 47 . Prior to advancing ST-2262 into the NHP acute thermal pain model, a standard battery of preclinical assays was completed to evaluate ADME and pharmacokinetic properties of this compound in cynomolgus macaques (Suppl Table 7). Off-target activity of ST-2262 using a commercially available radioligand binding assay panel against 68 different targets was also measured (LeadProfilingScreen, Eurofins, Taipei, Taiwan). No hits were identified on the off-target panel, defined as > 50% inhibition with 10 µM ST-2262 (Suppl Table 8). ST-2262 increases withdrawal latency in a nonhuman Male cynomolgus monkeys were anesthetized with propofol to a level in which the withdrawal reflex of the hand occurred at a consistent latency of approximately 3 s, a response time that was comparable to the detection of sharp pain from Aδ fibers when tested in prior studies on human volunteers 48,49 . The dorsal surface of the hand was exposed to a thermal stimulus that selectively activates Aδ-fiber nociceptors (Fig. 4A-C) 47,50 . The thermal stimulus was turned off at 5 s to prevent tissue damage. Heart rate was monitored throughout the study, and presentation of the noxious thermal stimuli consistently led to a transient increase in heart rate that peaked seconds after the stimulus and then returned to baseline (ΔHR). Acute noxious thermal stimuli transiently increase heart rate in human subjects; the percent change in heart rate correlates with subjective pain score 51 . ST-2262 hydrochloride administered IV increased the withdrawal latency to noxious thermal stimuli (Fig. 4A). Efficacy was assessed in one subject at four dose levels (0.01, 0.05, 0.25, 1.25 mg/kg), in two subjects at the three higher dose levels (0.05, 0.25, 1.25 mg/kg), and in one additional subject at the highest dose level only (1.25 mg/kg). At the highest dose of 1.25 mg/kg, all four animals showed no hand withdrawal prior to the 5 s cut-off latency (Fig. 4A) Plasma samples were obtained from animals to assess the PK/PD relationship between drug exposure and thermal withdrawal latency. We found that 0.25 mg/kg ST-2262 resulted in ~ 1,400 ng/ml in plasma at the 5 min time point (n = 2), which corresponds to 7× the IC 50 value of ST-2262 against cynoNa V 1.7, corrected for plasma protein binding (cyno PPB = 73.5%). The unbound exposure of drug was reduced to 3.4 × cynoNa V 1.7 IC 50 at the 30 min time point. At a dose of 1.25 mg/kg, the total plasma concentration was ~ 7,000 ng/ml at 5 min (n = 2), which corresponds to an unbound exposure of 32 × cynoNa V 1.7 IC 50 , and was maintained above 15× cynoNa V 1.7 IC 50 for over 100 min (Fig. 4C). Lumbar CSF samples collected from two animals receiving the 1.25 mg/kg dose indicated that ST-2262 was peripherally restricted, with CSF:plasma ratios < 10 -3 (n = 2; [ST-2262] 0.8, < 0.5 ng/ ml in CSF). By adjusting radiant heat parameters, the noxious heat model can be used to selectively assess responses to cutaneous C-fiber nociceptor activation, which produces a burning pain in volunteers 48,49 . The effect of ST-2262 on C-fiber induced hand withdrawal and heart rate change was investigated on two cynomolgus subjects 47 . As with the Aδ nociceptive response, 1.25 mg/kg ST-2262 completely abolished the C-fiber-mediated hand withdrawal and ΔHR (Fig. 4D,E).

Collectively, these results are consistent with the hypothesis that pharmacological block of Na V 1.7 reduces sensitivity to noxious heat, phenotypically analogous to studies of Na V 1.7 lossof-function in CIP patients 2 . In addition, analysis of the PK/PD relationship of ST-2262 in this model provides insight into the level of Na V 1.7 target occupancy that may be necessary to achieve a pharmacodynamic effect. Recognizing the limited number of animals tested due to the challenge of working with non-human primates, additional work is warranted to further define the relationship between pharmacological inhibition of Na V 1.7 and sensitivity to noxious thermal stimuli. Discussion The finding that humans lacking functional Na V 1.7 exhibit an inability to experience pain raises the intriguing possibility that selective inhibitors of Na V 1.7 may be potent analgesics [1][2][3] . In the present study, we describe the discovery and characterization of ST-2262, a selective pore blocker of hNa V 1.7 advanced through rational modification of a natural small molecule toxin lead, STX. In whole cell voltage clamp recordings, ST-2262 exhibited > 200-fold selectivity for hNa V 1.7 over hNa V 1.1-1.6 and hNa V 1.8. The selectivity of ST-2262 was not examined against hNa V 1.9, a channel subtype that is difficult to express in heterologously. hNa V 1.9 contains a residue in the domain I p-loop, S360, that confers resistance to STX and lacks the domain III threonine/isoleucine sequence motif that is essential for high potency of ST-2262 against hNa V 1.7. Thus, inhibition of Na V 1.9 by ST-2262 is unlikely 42 . The properties of ST-2262 are in contrast to other preclinical and clinical inhibitors of Na V 1.7, which preferentially bind to an inactivated conformation(s) of the channel 52 . Mutagenesis experiments indicate that specific residues in the extracellular pore of Na V 1.7, including a two amino acid sequence variation in the domain III pore loop that is unique to primates, are required for ST-2262 binding to cyno-and human Na V 1.7 31,39 . These findings establish the extracellular vestibule of Na V 1.7 as a viable receptor site for the design of potent and selective channel inhibitors. Whereas congenital insensitivity to pain in humans is the result of complete and permanent Na V 1.7 loss-offunction, inhibition by small molecule agents is incomplete and transient. This difference raises several important questions regarding the pharmacology of Na V 1.7: (1) is transient inhibition sufficient for analgesia, (2) what level of target engagement is required for efficacy, and (3) www.nature.com/scientificreports/ is sufficient to produce analgesia 45 . Furthermore, recognizing that ST-2262 is a polar small molecule with low membrane permeability and therefore unlikely to reach efficacious concentrations in the CNS (analysis of CSF samples obtained during NHP experiments gave a CSF:plasma ratio of < 10 -3 ), the observed effects on thermal withdrawal latency and ΔHR are likely the result of peripheral inhibition. Our findings, however, do not rule out an additional role for Na V 1.7 at the central terminals of primary afferents or in dorsal horn neurons, as has been suggested 53 . In the present study, the effect of ST-2262 on withdrawal latency to noxious heat was measured in NHPs at doses of 0.01, 0.05, 0.25 and 1.25 mg/kg IV. Doses of 0.05, 0.25 and 1.25 mg/kg resulted in unbound plasma concentrations of ST-2262 of 0.7×, 3.4× and 16× the IC 50 value against cynoNa V 1.7 at a time point 30 min following drug administration. Assuming a unitary Hill coefficient, which is consistent with the dose-response relationship for ST-2262 in whole cell recordings against cyno-and human Na V 1.7, these unbound exposures correspond to 41%, 78% and 94% inhibition of Na V 1.7, respectively. Further work to understand whether a similar relationship exists between Na V 1.7 target occupancy and analgesic pharmacodynamic effects in other preclinical pain models is ongoing. 47 . In two subjects, the C-fiber-induced hand withdrawal response was replicable for testing. The efficacy endpoints measured were withdrawal latency (A) and heart rate change (B). Symbiotic Bacterium-Derived Organic Acids Protect Delia antiqua Larvae from Entomopathogenic Fungal Infection The protection of associated microbiota for their animal hosts against pathogen infection has been studied widely over the last 100 years. However, how those microbes protect the animal host remains unclear. In former studies, body surface microbes of one insect, Delia antiqua, protected the insect larvae from infection with the entomopathogen Beauveria bassiana. By comparing the metabolites produced by microbes that protect the insect and microbes that cannot protect the insect, the question of how the microbes protect the insect is answered. It turns out that body surface bacteria produce a metabolite cocktail that inhibits colonization of B. bassiana and consequently protects the insect. This work reveals novel molecules with antifungal activity, which may aid in discovery and expansion of new prophylactic and therapeutic natural chemicals for treating infectious diseases. indigenous microbiota suppresses resident pathogenic species such as Clostridium difficile to low levels within the intestine (6). Most studies have focused on symbioses of mammals and microbes, revealing the interaction between indigenous microbiota and migratory pathogens during the process of colonization resistance and even identifying various metabolites that mediate the process. These findings likely represent the tip of the iceberg, and such symbioses thus need further investigation. Colonization resistance in symbioses formed by microbes and insects such as wood wasps (7), bumble bees (8), and beetles (9) has also attracted the attention of ecologists and entomologists. For example, a mixture of four Lactobacillus kunkeei strains isolated from the gut microbial community of bees decreased honeybee infections by the pathogens Paenibacillus sp. and Nosema ceranae (10). Microbial probiotic treatment can rescue honeybees from adverse effects due to N. ceranae by stimulating immunity in the honeybee (11). Due to insects' worldwide distribution, rich diversity, the possibility of extrapolating research studies to vertebrates, and the relatively low cost of rearing, insect-microbe symbiotic systems have potential as alternative models to investigate and develop colonization resistance theory in broader taxonomic clades. Delia antiqua (Meigen) (Diptera: Anthomyiidae) is a devastating pest that feeds on liliaceous crops (12) and can cause a reduction of crop yields by more than half of the total yield (13,14). Previously, we found that six most frequently isolated bacterial symbionts including Citrobacter freundii, Enterobacter ludwigii, Pseudomonas protegens, Serratia plymuthica, Sphingobacterium faecium, and Stenotrophomonas maltophilia protected D. antiqua larvae from Beauveria bassiana infection by inhibiting conidial germination and mycelial growth (15), while the bacterium Klebsiella oxytoca showed no antifungal activity against B. bassiana. Collectively, the D. antiqua-associated bacteriumentomopathogen tripartite interaction system provides an ideal model to investigate whether and which metabolites produced by the bacterial associates prevent fungal infection of D. antiqua larvae. In this study, we first verified and compared the effects of six bacterial strains including C. freundii B505, E. ludwigii B424, P. protegens B108, S. plymuthica B585, S. maltophilia B263, and S. faecium B253 and one control strain, K. oxytoca B313, on B. bassiana and its infection of D. antiqua larvae individually. In addition, the effects of the selected bacterial strains on the conidial germination and mycelial growth of B. bassiana were also determined. Second, the metabolomic profiles of the six bacterial strains and K. oxytoca B313 were compared to identify candidate metabolites that may protect D. antiqua larvae from B. bassiana infection. Third, candidate in situ metabolites of nonaxenic and axenic larvae were quantified and compared. Fourth, effects of artificial metabolite cocktails (prepared according to in situ concentrations of each metabolite) on B. bassiana and its infection of D. antiqua larvae were determined. This work promotes the understanding of metabolic interactions between entomopathogens and the symbiotic system formed by D. antiqua larvae and its associated bacteria. In addition, this work may also provide novel strategies for discovering new natural antifungal metabolites. Experiment II: metabolomic analysis revealed 10 candidate metabolites from the six bacterial strains that may protect D. antiqua larvae. In total, 17,983 and 12,275 metabolites were detected in all samples under positive and negative ion modes, respectively. Annotations were made for 6,389 metabolites among 14,522 metabolites with high-quality features under positive mode and 3,688 metabolites among 10,476 metabolites with high-quality features under negative mode. The unsupervised model principal-component analysis (PCA) showed that the samples from each group including quality control (QC) showed a trend of shifting away from each other under both positive (Fig. S3a) and negative (Fig. S3b) ion modes. To compare the differences in metabolites between lysogeny broth (LB) medium and each of the bacterial strains, individual PCAs and partial least-squares discriminant analyses (PLS-DAs) (Fig. S4) were conducted. PCA results showed that samples from selected bacterial strains and LB media were clearly separated from each other under both positive and negative ion modes. Similar separating trends were also observed in PLS-DA (Fig. S4). For all permutation tests of PLS-DA models of selected bacterial strains and LB medium, all blue Q2 values to the left are lower than the original points to the right, and the regression line of the blue Q2 points intersects the vertical axis below zero, which indicated a low risk of model overfitting (data not shown). Organic Acids Protect Delia antiqua surface at a dosage equal to in situ conditions, conidial germination of B. bassiana BB1101 was slightly reduced to 90.9% relative to control (Fig. 4a), while conidial germination of B. bassiana BB1101 was sharply reduced to 15.2% relative to control under the effect of metabolite cocktail for nonaxenic larval body surface at a dosage equal to in situ conditions (Fig. 4b). Similar inhibitory effects of the two metabolite cocktails on the mycelial growth of B. bassiana BB1101 were detected. Results showed that metabolite cocktails for body surfaces of axenic larvae (Fig. 4c, Welch's ANOVA, F 8,18.605 ϭ 32.641, P Ͻ 0.001) and nonaxenic larvae (Fig. 4d, one-way ANOVA, F 8,45 ϭ 7,772.135, P Ͻ 0.001) each significantly inhibited mycelial growth of B. bassiana BB1101 individually. As the concentrations of metabolite cocktails for body surface of nonaxenic larvae increased, the mycelial growth decreased. Under the effect of a metabolite cocktail for axenic larval body surface at a dosage equal to the in situ conditions, mycelial growth of B. bassiana BB1101 was slightly reduced to 96.4% relative to control (Fig. 4c), while for nonaxenic larval body surface mycelial growth of B. bassiana BB1101 was significantly reduced to 16.3% relative to control under the effect of a metabolite cocktail at a dosage equal to in situ conditions (Fig. 4d). Experiment V: artificial metabolite cocktail for nonaxenic larval body surface inhibited fungal infection of D. antiqua larvae with B. bassiana BB1101. Kaplan-Meier analysis showed that the B. bassiana BB1101 treatment significantly decreased the survival of axenic larvae (Fig. 5, log rank test, 2 ϭ 52.755, df ϭ 1, P Ͻ 0.001), while the metabolite cocktail showed no effects on the survival of axenic D. antiqua larvae (Fig. 5, log rank test, 2 ϭ 0.225, df ϭ 1, P ϭ 0.636). For the B. bassiana-infected axenic larvae, the final survival of larvae that were preliminarily treated with the metabolite cocktail was significantly higher than that of the larvae without the metabolite cocktail treatment (Fig. 5, 2 ϭ 19.428, df ϭ 1, P Ͻ 0.001), though the cocktail cannot exclusively prevent mortality. Specifically, under B. bassiana BB1101 treatment, the survival rate of axenic larvae treated with the metabolite cocktail was 62.2%, while the survival rate of axenic larvae that were not treated with the candidate metabolite cocktail was 15.6%. DISCUSSION The symbiotic system formed by bacteria and D. antiqua larvae provides an ideal model for studying the process of colonization resistance and identifying metabolites that mediate the process. In this study, six most frequently isolated species (C. freundii, E. ludwigii, P. protegens, S. plymuthica, S. maltophilia, and S. faecium) and K. oxytoca B313 showed different effects on the infection of D. antiqua larvae with B. bassiana. Specifically, neither the six bacterial strains nor K. oxytoca B313 affected the survival of axenic D. antiqua larvae (see Fig. S1 in the supplemental material). However, the six bacterial strains inhibited B. bassiana infection of axenic D. antiqua larvae, while K. oxytoca B313 did not (Fig. 1). We confirmed that the six bacterial strains protect D. antiqua larvae by inhibiting the conidial germination (Fig. S2a) and mycelial growth (Fig. S2b) of B. bassiana, which is consistent with

previous reports (15). Comparison of metabolomic data from the six bacterial strains and K. oxytoca B313 (Fig. 2) indicated that 10 types of metabolites, including glutaric acid, thymine, ethylmalonic acid, hypoxanthine, kynurenic acid, picolinic acid, ketoisocaproic acid, phenyllactic acid, adipic acid, and indoleacetic acid, were produced at different levels among the six bacteria compared to B313. These metabolites may be the bioactive chemicals that protect D. antiqua larvae from entomopathogenic infection. Subsequent experiments quantified these 10 chemicals on the body surface of both axenic and nonaxenic D. antiqua larvae (Fig. 3). According to the in situ concentration of each chemical from samples of axenic and nonaxenic larval body surfaces, two kinds of metabolite cocktails were prepared in a series of concentrations. Further experiments revealed that the metabolite cocktail of the axenic larval body surface showed a slightly inhibitory effect on both conidial germination (Fig. 4a) and mycelial growth (Fig. 4c) of B. bassiana BB1101 at a dosage equal to in situ conditions, while the metabolite cocktail of the nonaxenic larval body surface showed a significantly inhibitory effect on both conidial germination (Fig. 4b) and mycelial growth (Fig. 4d) of B. bassiana BB1101 at a dosage equal to in situ conditions. Furthermore, B. bassiana BB1101's successful infection of axenic larvae may result from the concentration and chemical composition of the metabolite cocktail, which was confirmed in Fig. 5. Colonization resistance involves direct interactions between the associated microbiota and the pathogens (16). The six bacteria associated with D. antiqua selected in this study directly inhibited conidial germination and mycelial growth of B. bassiana by producing a metabolite cocktail including glutaric acid, ethylmalonic acid, picolinic acid, ketoisocaproic acid, adipic acid, and indoleacetic acid. Indoleacetic acid has been reported to mediate colonization resistance by inhibiting biofilm in mammal guts (17), Organic Acids Protect Delia antiqua while the essential roles of glutaric acid, ethylmalonic acid, picolinic acid, ketoisocaproic acid, and adipic acid in mediating colonization resistance were reported for the first time in this study. Antimicrobial activity of these organic acids has been reported occasionally. For example, glutaric acid was found in metabolites of an antifungal Bacillus strain (18), and direct evidence of virucidal activity of glutaric acid has also been reported (19). Similarly, antimicrobial activity of picolinic acid (20,21), adipic acid (22), and indoleacetic acid (23) has also been reported occasionally. It seems that microbes associated with D. antiqua mediate colonization resistance against B. bassiana by producing anti-B. bassiana metabolites. Associated microbiota mediate colonization resistance by producing antimicrobial metabolites (24), which indicates that the active metabolites should be accumulated in situ. Correspondingly, concentrations of glutaric acid, adipic acid, indoleacetic acid, and ketoisocaproic acid from nonaxenic larval body surfaces were much higher than those from axenic ones (Fig. 3). Higher concentrations of certain organic acids on body surfaces of nonaxenic larvae may result from the microbial production of certain organic acids from plant amino acids such as tryptophan, leucine, or lysine as well as several other amino acids that have been detected in garlic and onion (25,26). For example, glutaric acid is naturally produced during the metabolism of lysine and tryptophan by bacteria (27). Indoleacetic acid is produced by various bacteria through metabolism of tryptophan (28). Ketoisocaproic acid is an intermediate in the metabolism of leucine (29). Taking phylogenetic relationships of the six bacterial species into consideration (15), it seems that these bacteria produce those organic acids mainly through primary metabolization. Those metabolites may be essential in colonization resistance. In contrast, concentrations of phenyllactic acid and kynurenic acid from nonaxenic larval body surfaces were much lower than those from axenic ones (Fig. 3). This may result from the microbial assimilation of certain organic acids from plant tissue leading to reduction of precursors for phenyllactic acid and kynurenic acid on the nonaxenic insects. Even though concentrations of phenyllactic acid and kynurenic acid from axenic larval body surfaces were significantly higher than those from nonaxenic ones, this difference may not influence the whole scenery of colonization resistance as concentrations of phenyllactic acid (Fig. 3e, 6.3 to 14.0 mg/liter) and kynurenic acid (Fig. 3f, 0.1 to 0.5 mg/liter) were much lower than those of other organic acids from nonaxenic larval body surfaces (Fig. 3a, adipic acid, 587.5 mg/liter; Fig. 3b, indoleacetic acid, 191.2 mg/liter; Fig. 3c, glutaric acid, 63.7 mg/liter; Fig. 3d, ketoisocaproic acid, 66.1 mg/liter). Besides, the in vivo abundance of the microbes tested in this study in D. antiqua remains unclear although they were most frequently isolated species in former studies (15). Dominant species, especially those unculturable ones, in the symbiotic system formed by D. antiqua larvae and bacteria may also play a significant role in metabolite transformations, which may explain why the in vivo metabolite abundances do not match those predicted from the in vitro experiments using individual strains. Further investigation combining microbiome analysis and metabolomic analysis in vivo needs to be conducted. Several reasons may explain the anti-B. bassiana activity of the metabolite cocktail. First, the organic acid may disrupt primary metabolomic activity of B. bassiana during conidial germination and mycelial growth as most organic acids can cross cell membranes (30,31). For example, the antimicrobial activity of picolinic acid (a metalchelating agent) may result from the deprivation of free nutrient iron essential for the conidial germination and mycelial growth of B. bassiana, which has been reported previously (21,32). As an intermediate of the metabolism of leucine (29), ketoisocaproic acid has been reported to suppress insulin-stimulated glucose transport in skeletal muscle cells (33). This metabolite might function in glucose transport in B. bassiana, which leads to repressed conidial germination and mycelial growth of the fungus. Second, the accumulation of adipic acid, indoleacetic acid, glutaric acid, and ketoisocaproic acid on nonaxenic D. antiqua larval body surface (Fig. 3a to d) may lead to a low-pH condition for the larval body surface as all these compounds belong to organic acids. Although the concentration of phenyllactic acid and kynurenic acid decreased on body surfaces of nonaxenic D. antiqua larvae compared to axenic ones, this may scarcely contribute to increase of the lowered-pH condition as the concentrations of phenyllactic acid and kynurenic acid were much lower than those of adipic acid, indoleacetic acid, glutaric acid, and ketoisocaproic acid (Fig. 3). Combined with previous studies showing that low pH (below 6.3) leads to mycelial growth inhibition for B. bassiana (34), it seems that the candidate metabolites decreased the pH of the larval habitat and thus protected the larvae from B. bassiana infection. Third, low pH of the D. antiqua larval habitat may in turn enhance the toxicity of organic acids to the fungus. For example, adipic acid will become undissociated and enter cells via passive diffusion over the plasma membrane. Once in the cytosol, the carboxylic groups of this acid become deprotonated due to the almost neutral pH in the cytosol and cause acid stress in the cell (35). Further experiments such as testing the B. bassiana infection of D. antiqua after neutralizing the low pH, the transcriptomic response of B. bassiana against the organic acid cocktail, and the synergistic effects of individual organic acids need to be conducted to illustrate the molecular mechanism of the protective effect of the organic acid cocktail. As no standard protocols are available to identify differential metabolites from different bacterial strains that mediate colonization resistance, especially considering the background noise from microbial culture media, we took a relatively stricter cutoff strategy to detect candidate metabolites that mediate colonization resistance in the D. antiqua-microbe symbiosis. Based on the hypothesis that the most discrepant metabolites between the six bacterial strains and K. oxytoca B313 may be the chemicals with the highest potential to protect D. antiqua larvae from fungal infection, candidate metabolite chemicals were quantified in situ and their effects on the fungal infection of D. antiqua larvae were determined. Although this design may lead to artificial deviations in detecting acting members among the metabolite pool, especially when detecting potential metabolites with minor differences between the two groups of bacterial strains, several metabolites that mediate colonization resistance were successfully detected. Besides, only a very small portion of metabolites were annotated and identified as shown in Fig. 2. There should be a much greater metabolic diversity of unknown composition and function for the natural situation in the D. antiqua-microbe symbiosis. This work focused only on the small subset of metabolites that could be identified by database annotation, for which there is no technological alternative at present. Moreover, as shown in Fig. 2o to t, the majority of differential metabolites were not annotated or identified. There are likely more metabolites than those selected in this study that provide protection for the insect host. Further experiments to isolate metabolites with chromatographic column separation and identify the bioactivity of these metabolites need to be conducted. Bacterial strains selected in this work are the most frequently isolated bacterial strains among all strains obtained in prior research by a culture-dependent method, which indicates that the selected bacteria in this work may not be the dominant species among the associated microbiota of D. antiqua. Thus, our work identified a limited number of metabolites that function in colonization resistance in D. antiqua-bacterium symbiosis, consistent with the assumption that identified metabolites that mediate colonization resistance are likely a small proportion of the total (17). Further analysis combining data of the host microbiome and metabolomic data sets may lead to deep understanding of the process of colonization resistance in D. antiqua-bacterium symbiosis. During the process of colonization resistance, associated microbiota can kill other microbes by secreting small molecules such as short-chain fatty acids and small carboxylic acids as well as some tryptophan metabolites (17) in mammal guts. For insects, some metabolites mediating colonization resistance are some antibiotics. For example, the Penicillium symbionts associated with a leaf-rolling weevil defend the weevil's offspring from mold fungi and antagonistic bacteria by producing the antibiotic (ϩ)-scleroderolide (36). Streptomyces microbial symbionts associated with beewolves can produce antibiotics including piericidin derivatives, streptochlorin deriva-tives, and nigericin to protect the insect host against mold fungi (7). However, in the symbiotic system formed by the D. antiqua larvae and their associated bacteria, the metabolite cocktail of organic acids derived from amino acids (excluding tryptophan) has been proven to protect the insect host from B. bassiana infection. Their functions in mediating colonization resistance were reported for the first time. Given the essential roles of those novel metabolites that mediate colonization resistance, this work may aid in discovery and expansion of the list of new bioactive antibiotics, promoting development of prophylactic and therapeutic approaches for treating many infectious diseases for both humans and other animals. MATERIALS AND METHODS Insects, microbial strains, and materials. Adults of D. antiqua were originally collected from garlic fields in Taian City, China (N36°14=, E117°25=). As this insect is not in the list of wild animals under state priority conservation in China, permissions to collect it were not needed. Eggs laid by these fieldcollected adult females were randomly selected for further experiments. Axenic larvae were obtained by rearing larvae from hatched surface-sterilized eggs (75% ethanol twice for 30 s) with axenic artificial diets containing antibiotics (12,15). Detailed information on the preparation of artificial diets is in Table S1 in the supplemental material. Axenic second-instar larvae (hatched larvae reared with artificial diets at 24°C for 7 days in darkness) were used in this study as their body size is suitable for experimental procedures and there was enough time for experimental observation before pupation. Additional bacterial and fungal isolation experiments from axenic larvae (15) were conducted to ensure that all culturable microbes were eliminated in the axenic larvae. Nonaxenic second-instar larvae were collected from garlic fields. The fungus B. bassiana BB1101 was originally isolated from infected D. antiqua adults with the single-spore method (37), and its conidia were preserved in 25% glycerol in our laboratory at Ϫ80°C. The effects of the six bacterial strains and K. oxytoca B313 on the survival of axenic larvae were determined. For each bacterial strain, 4 ml of overnight lysogeny broth (LB) cultures (28°C and rotary shaking at 150 rpm for 12 h) was centrifuged at 3,000 ϫ g for 5 min to collect bacterial cells, and then the bacterial cells were washed three times with phosphate-buffered saline (PBS) and resuspended in PBS (10 6 CFU/ml). Subsequently, 200 l of the suspension was added to sterilized filter paper

in a 30-mm petri dish, and three surface-sterilized D. antiqua eggs were placed on the paper. After hatching, artificial diets (containing no antibiotics) were put inside the petri dish to feed the hatched larvae at 24°C in darkness. For the control group, 200 l of PBS was used in place of the bacterial cell suspension. Subsequently, one hatched larva from each petri dish was picked out and reared with artificial diets containing no antibiotics for further observation. Ninety petri dishes were set up for each bacterial strain (45 for bacterial suspension and 45 for PBS). Those larvae were incubated at 24°C in darkness, and larval survival was recorded every 2 days until pupation. The effects of the six bacterial strains and K. oxytoca B313 on entomopathogenic infection of D. antiqua larvae with B. bassiana BB1101 were compared. For the comparison between each bacterial strain and K. oxytoca B313, overnight LB cultures (28°C and rotary shaking at 150 rpm for 12 h) for each mentioned bacterial strain or K. oxytoca B313 were centrifuged at 3,000 ϫ g for 5 min, and then the bacterial cells were collected, washed, and resuspended in PBS (10 6 CFU/ml). Moreover, a conidial suspension of B. bassiana BB1101 was obtained by washing the fungal plate with sterilized water containing 0.05% Tween 80 and filtered through a sterilized multilayer gauze, and the suspension was adjusted to 10 6 CFU/ml with sterilized water containing 0.05% Tween 80. A volume of 200 l bacterial suspension of each bacterial strain or B313 was added into a 30-mm petri dish containing three surface-sterilized D. antiqua eggs placed on sterilized filter paper, and the hatched larvae were reared with artificial diets (containing no antibiotics) for 1 week (second instar). Subsequently, one of the three larvae was randomly selected, taken out, and sprayed with the above conidial suspension and air dried on a piece of sterilized filter paper. Then, the larva was reared with an artificial diet containing no antibiotics for further observation (Bbϩbacterial strain). Forty-five larvae (one hatched larva from each petri dish containing three larvae) treated with one of the six bacterial strains or K. oxytoca B313 were individually collected and sprayed with B. bassiana BB1101 conidial suspension. In addition, another group of axenic larvae (45 larvae) were treated with the same conidial suspension one by one without inoculation of bacteria (Bb). Then, for each comparison, each larva from these three groups (Bb, selected bacterial strainϩBb, and B313ϩBb) was separately fed with artificial diets (containing no antibiotics). Those larvae were incubated at 24°C in darkness, and larval survival was recorded every 2 days until pupation as described above. A preliminary test showed that B. bassiana BB1101 has excellent insecticide activity against axenic larvae (Fig. S6), so the comparison between axenic larvae treated by B. bassiana BB1101 conidial suspension and those without treatment was not set up in each comparison. The effect of the six bacterial strains and K. oxytoca B313 on the conidial germination and mycelial growth of B. bassiana BB1101 was determined using a previously described method with minor modifications (38). A conidial suspension of B. bassiana BB1101 (10 6 CFU/ml) was obtained as described in the previous study. LB cultures of each bacterial strain (incubated for 72 h at 28°C with rotary shaking at 150 rpm) were centrifuged at 3,000 ϫ g for 5 min, and the supernatant was filtered with a 0.20-m filter {polytetrafluoroethylene (PTFE) syringe filter [catalog no. SCAA-1114; ANPEL Laboratory Technologies (Shanghai) Inc.]} and diluted by 1ϫ, 5ϫ, and 25ϫ using LB medium. For each strain, 100 l of the conidial suspension was combined with 3.9 ml of diluted LB culture supernatant and then incubated at 25°C with rotary shaking at 180 rpm for 24 h in darkness. LB medium instead of the diluted supernatant was used in the control group. Conidia were defined as germinated when the length of the germ tube was greater than or equal to the conidia under a microscope (39). Each test for one specific bacterial strain on conidial germination was repeated six times. The conidial germination rate was presented as the percent value relative to that of the control. To determine the effects of each individual strain on the mycelial growth of B. bassiana BB1101, each bacterial strain was cultured in LB medium for 12 h at 28°C with rotary shaking at 150 rpm, and then the culture was centrifuged at 3,000 ϫ g for 5 min to collect bacterial cells. Subsequently, these cells were washed three times with PBS and resuspended to 100, 20, and 4 CFU/ml. Five milliliters of the bacterial cell suspension at different concentrations was added to 15 ml of melted potato dextrose agar (PDA; stored at 50°C in a water bath after autoclaving) and then mixed and poured into a 90-mm petri dish (approximately 500, 100, and 20 CFU/petri dish). These plates were used in the treatment group. For the control group, 5 ml of PBS was used in place of the bacterial cell suspensions. Agar plugs (3 mm, without conidia) taken from the leading edge of the B. bassiana BB1101 culture growing on 1/4 PDA plates were inoculated in the center of the PDA plates containing bacterial cells (the treatment groups) or without bacterial cells (the control group), and agar plugs from the same plate were randomly assigned to different treatment groups. Each petri dish was regarded as one replicate, and each test for one specific bacterial strain on mycelial growth was repeated six times. The plates were subsequently incubated at 25°C for 10 days in darkness. Starting the next day, the diameter of fungal mycelia was measured each day in two directions at right angles to each other during the experimental period. The mycelial growth rate was calculated as the percent value relative to that of the control. Experiment II: metabolomic analysis of the six bacterial strains and K. oxytoca B313 metabolites. Individual bacterial LB cultures (rotary shaking at 150 rpm and 28°C for 72 h) for each strain including K. oxytoca B313 were centrifuged at 3,000 ϫ g for 5 min. Subsequently, 20 l of supernatant filtered with a 0.22-m PTFE syringe filter [catalog no. SCAA-1114; ANPEL Laboratory Technologies (Shanghai) Inc.] was mixed with 120 l of precooled 50% methanol (4°C for 24 h), vortexed for 1 min, and incubated at room temperature for 10 min. The mixture was then stored overnight at Ϫ20°C and centrifuged at 4,000 ϫ g for 20 min, and the supernatants were transferred separately into new glass vials for the liquid chromatography-mass spectrometry (LC-MS) analysis. LB medium was also extracted as described above and analyzed with LC-MS to subtract the background noise of LB. In addition, pooled quality control (QC) samples were also prepared by combining 10 l of each extraction sample. Six replicates were used for each group (B585, B263, B108, B505, B253, B424, B313, LB, and QC). Each sample from all nine groups was analyzed using an LC-MS system according to a previously published method with minor modifications (40). An Acquity ultraperformance liquid chromatography (UPLC) ethylene-bridged hybrid (BEH) amide column (100 mm ϫ 2.1 mm, 1.7 m; Waters, United Kingdom) was used, with a mobile phase of solvent A (25 mM ammonium acetate) and solvent B (isopropanol:acetonitrile ϭ 9:1 ϩ 0.1% formic acid). Gradient elution conditions were as follows: 0 to 0.5 min, 95% B; 0.5 to 9.5 min, 95% to 65% B; 9.5 to 10.5 min, 65% to 40% B; 10.5 to 12 min, 40% B; 12 to 12.2 min, 40% to 95% B; and 12.2 to 15 min, 95% B. The instrumentation conditions of a high-resolution tandem mass spectrometer, TripleTOF 5600 Plus (Sciex, United Kingdom), were set up according to previously published methods (41). Metabolomic data were analyzed using the XCMS software package (https:// sciex.com/products/software/xcms-plus-software) according to methods published previously (42). Metabolites were annotated if the mass difference between the observed metabolite and the one from databases (Kyoto Encyclopedia of Genes and Genomes Database and the Human Metabolome Database) was less than 10 ppm. Besides, an in-house fragment spectrum library of metabolites was used to identify the metabolite. Principal-component analysis (PCA) was performed for outlier detection and batch effects evaluation with the preprocessed data set. Supervised partial least-squares discriminant analysis (PLS-DA) was conducted to discriminate the different variables between groups with calculated variable importance of projection (VIP) values. A VIP cutoff value of 2 was used to select important features. To subtract the background noise from LB medium, additional data analysis procedures were conducted (Fig. S7). A data set union was created by merging metabolites from each bacterial sample group which were significantly upregulated at least 2-fold compared with the LB medium (step I in Fig. S7). Subsequently, a more critical parameter (metabolites produced by the selected bacterial strains upregulated at least 10 times compared with the B313 group) was used to ensure that the obtained metabolites were produced by the six strains and of high antifungal activity (step II in Fig. S7). Metabolomics data were deposited in the EMBL-EBI MetaboLights database (43). Experiment III: quantification of candidate metabolites on the body surface of axenic and nonaxenic D. antiqua larvae. The difference between laboratory standard conditions and in situ field environments with bacteria may lead to variation in concentrations and compositions of bioactive metabolites that protect D. antiqua larvae from B. bassiana infection. Moreover, metabolomic analysis cannot provide in situ information, particularly concentrations of candidate metabolites. Thus, a quantification experiment of candidate metabolites, including ketoisocaproic acid, glutaric acid, adipic acid, phenyllactic acid, indoleacetic acid, kynurenic acid, picolinic acid, ethylmalonic acid, hypoxanthine, and thymine, on the body surface of D. antiqua second-instar larvae was conducted. In addition, the concentrations of the candidate metabolites on the body surface of axenic and nonaxenic larvae were compared. To quantify the candidate metabolites on the body surface of nonaxenic D. antiqua larvae, 20 field-collected second-instar larvae from five randomly selected garlic plants (four larvae from each plant) were collected, and the frass on the body surface of the larvae was cleaned with a little brush. Then all 20 larvae were put into one centrifuge tube (2 ml), mixed with 1.5 ml of double-distilled water (ddH 2 O), and sonicated for 30 s at 50°C. Subsequently, the sample was vortexed for 60 s, and the liquid inside the tube was collected and centrifuged at 3,000 ϫ g for 10 min in another centrifuge tube. The supernatant was collected and filtered with a 0.22-m PTFE syringe filter [catalog no. SCAA-1114; ANPEL Laboratory Technologies (Shanghai) Inc.], followed by quantification of the candidate metabolites. Detailed information on instrumentation and detection conditions is in Text S1 in the supplemental material. Another group of 20 axenic larvae at the second instar (surface-sterilized eggs reared with artificial diets containing no antibiotics for 7 days) was also collected. The body surface samples for quantification of candidate metabolites of axenic D. antiqua larvae were prepared as described above. According to the quantity of the selected metabolites in the supernatant from 20 D. antiqua larvae, the quantity of selected metabolites for each larva was calculated. Previous preliminary experiments showed that there was about 0.0920 Ϯ 0.047 mg H 2 O on the body surface of second-instar D. antiqua larva (unpublished data; average Ϯ standard deviation). Thus, the actual concentration for each selected metabolite on the body surface of D. antiqua larva was estimated. Each group of 20 larvae was regarded as one replicate, and the above two tests were repeated five times. Experiment IV: effects of bacterial metabolite cocktail on conidial germination and mycelial growth of B. bassiana BB1101. In experiment III, the concentrations of selected metabolites (ketoisocaproic acid, glutaric acid, adipic acid, phenyllactic acid, indoleacetic acid, kynurenic acid, picolinic acid, ethylmalonic acid, hypoxanthine, and thymine) on the larval body surface of both axenic and nonaxenic larvae were determined. Based on these results, two metabolite cocktails for body surface of axenic and nonaxenic larvae were prepared separately to perform in situ simulation. For each of the above cocktails, a serial concentration of 2 1 , 2 0 , 2 Ϫ1 , 2 Ϫ2 , 2 Ϫ3 , 2 Ϫ4 , 2 Ϫ5 , and 2 Ϫ6 times the actual concentration for each metabolite was set. Details of the dosage setup of selected metabolites are in Table S2. Effects of the above bacterial metabolite cocktails on conidial

germination and mycelial growth of B. bassiana BB1101 were determined as follows. To determine the effects of bacterial metabolite cocktails on conidial germination of B. bassiana BB1101, a series of concentrations of solutions (Table S2) for each cocktail was prepared with sterilized 1/4 potato dextrose broth (PDB) and filtered with 0.22-m PTFE syringe filters [catalog no. SCAA-1114; ANPEL Laboratory Technologies (Shanghai) Inc.]. According to previous reports (44,45), 1/4 PDB or PDA was used. Subsequently, 10 l of B. bassiana BB1101 conidial suspension (10 6 CFU/ml) obtained as described above was added into a test tube containing 4 ml of the above PDB-metabolite cocktail. Test tubes containing the above mixture were incubated at 25°C with rotary shaking at 180 rpm for 24 h in darkness. Conidial germination was determined via microscopy as described for experiment I. For each metabolite cocktail, the test for each dose was repeated six times, and 1/4 PDB medium without metabolites was used as a control. The conidial germination rate was presented as the percent value relative to that of the control. To test the effect of bacterial metabolite cocktails on mycelial growth of B. bassiana BB1101, an aliquot of individual metabolite cocktail stock solution {filtered using a 0.22-m PTFE syringe filter [catalog no. SCAA-1114; ANPEL Laboratory Technologies (Shanghai) Inc.]} was added into 10 ml 1/4 soft PDB agar inside a petri dish to final concentrations as described in Table S2. Agar plugs (3 mm) taken from the leading edge of the B. bassiana BB1101 culture growing on 1/4 PDA plates (without conidia) were inoculated into the center of the PDA plates that contained different doses of metabolite cocktails (for the treatment groups). For the control group, 1/4 PDB was used instead of the metabolite cocktail solution. For each metabolite cocktail, the test for each dose was repeated six times, and agar plugs from the same B. bassiana BB1101 culture PDA plate were randomly assigned to different treatment groups. Subsequently, these plates were incubated at 25°C in darkness for 10 days. The diameter of fungal mycelia was measured every 2 days in two directions at right angles to each other during the experiment. The mycelial growth rate was calculated as the percent value relative to that of the control group. Experiment V: effects of candidate metabolite cocktail on fungal infection of D. antiqua larvae with B. bassiana BB1101. A bacterial metabolite cocktail containing ketoisocaproic acid (66.1 mg/liter), glutaric acid (63.7 mg/liter), adipic acid (587.5 mg/liter), phenyllactic acid (6.3 mg/liter), indoleacetic acid (191.2 mg/liter), kynurenic acid (0.1 mg/liter), and picolinic acid (0.2 mg/liter) was prepared according to the concentration of each metabolite from the nonaxenic larval body surface. One artificial diet containing the above organic mixture was prepared by adding organic acid solutions to sterilized diets at 55°C; antibiotics were not added into the diets. Final concentrations for each organic acid in the diet were the same as that in the above metabolite cocktail. To detect the effect of the cocktail on B. bassiana infection of axenic D. antiqua larvae, 2 ml of the cocktail was added to sterilized filter paper in a 90-mm petri dish. An axenic second-instar D. antiqua larva was sprayed with the above cocktail on sterilized filter paper and air dried, and the larva was sprayed with a B. bassiana BB1101 conidial suspension (10 6 conidia/ml, cocktailϩBb) as described above. Then the larva was put inside the petri dish containing a piece of cocktail-soaked filter paper and fed with the artificial diets containing the seven organic acids. To detect the effect of the cocktail on survival of axenic D. antiqua larvae, one second-instar axenic larva was treated with the cocktail combined with sterilized water instead of the B. bassiana BB1101 conidial suspension, put inside a petri dish containing a piece of cocktail-soaked filter paper, and fed with artificial diets containing the seven organic acids (cocktail). To detect the effect of B. bassiana on survival of axenic D. antiqua larvae, one second-instar axenic larva was treated with sterilized water instead of the cocktail, further treated with the B. bassiana BB1101 conidial suspension, put inside a petri dish containing a piece of water-soaked filter paper, and fed with artificial diets containing neither organic acids nor antibiotics (Bb). For the control (Blank), one axenic larva without treatment was incubated inside a petri dish containing water-soaked filter paper and fed with artificial diets containing neither organic acids nor antibiotics. All larvae were incubated in darkness at 24°C. Larval survival was recorded every 2 days until pupation. Data analysis. Data analysis was conducted with IBM SPSS 20.0 (International Business Machines Corp., Armonk, NY, USA). The normality and homogeneity of observed variances were tested with the Kolmogorov-Smirnov test and Levene's test, respectively. Larval survival rates under various treatments were compared with Kaplan-Meier analysis (log rank test) in experiments I and IV. Conidial germination or mycelial growth data in experiments I and IV were compared with one-way ANOVA followed by Tukey's multiple comparisons, Welch's ANOVA followed by Dunnett's T3 test, or the Kruskal-Wallis test depending on the normality and homogeneity of the variances. In experiment III, quantifications of candidate metabolites on the body surface of axenic and nonaxenic larvae were compared using the independent t test or Mann-Whitney U test. The figures for the above-described experiments were produced using SigmaPlot 12.5 (Systat Software Inc., San Jose, CA, USA). Data availability. All data during the study are available from the corresponding author by request. Metabolomics data have been deposited in the EMBL-EBI MetaboLights database (https://doi.org/10 .1093/nar/gkz1019; PMID: 31691833) with the identifier MTBLS2074. SUPPLEMENTAL MATERIAL Supplemental material is available online only. TEXT S1, PDF file, 0.2 MB. Changes in Gaze Behavior during the Learning of the Epidural Technique with a Simulator in Anesthesia Novices Background: Current literature demonstrates the ability of eye tracking to provide reliable quantitative data as an objective assessment tool, with potential applications to medical and surgical training to improve performance. Objective: The aim of this study was to evaluate the changes in gaze behavior in anesthesia novice trainees when performing a simulated epidural technique before and after a hands-on training on the epidural simulator. Methods: We enrolled 48 novice trainees who had never previously performed an epidural block. After a standardized learning module, each trainee practiced the epidural procedure on the epidural simulator while wearing a pair of eye tracking glasses (Tobii Pro Glasses 50 Hz wearable wireless eye tracker). After this baseline recording, each trainee spent two hours practicing with the epidural simulator and afterwards once again performed the eye tracking epidural procedure. Eye tracking metrics and epidural learning (duration of the procedure and number of attempts) before and after the simulated practice were recorded. Results: The duration of the epidural procedure and of the epidural needle advancement phase (P < 0.05) and the number of epidural attempts (P < 0.001) were reduced after the tutorial. When considering the eye tracking metrics, after the tutorial, the number of visit counts decreased and their duration increased (P < 0.05). The number of epidural needle insertions (additional attempts) showed a significant positive correlation with the visits number (aOR = 2.02 (95% CI = 1.26 3.55; P = 0.008)) and a significant negative correlation with the visit duration (aOR = 0.65 (95% CI = 0.39 0.99; P = 0.05)). Conclusion: We observed significant changes in gaze behavior associated with increased performance during the epidural technique learning with a simulator in anesthesia trainees who had never previously performed an epidural block. These results may create a prototype for future studies on How to cite this paper: Capogna, E., Salvi, F., Del Vecchio, A., Velardo, M. and Capogna, G. (2020) Changes in Gaze Behavior during the Learning of the Epidural Technique with a Simulator in Anesthesia Novices. Open Journal of Anesthesiology, 10, 361-370. https://doi.org/10.4236/ojanes.2020.1011032 Received: September 16, 2020 Accepted: November 6, 2020 Published: November 9, 2020 Copyright © 2020 by author(s) and Scientific Research Publishing Inc. This work is licensed under the Creative Commons Attribution International License (CC BY 4.0). http://creativecommons.org/licenses/by/4.0/ Background: Current literature demonstrates the ability of eye tracking to provide reliable quantitative data as an objective assessment tool, with potential applications to medical and surgical training to improve performance. Objective: The aim of this study was to evaluate the changes in gaze behavior in anesthesia novice trainees when performing a simulated epidural technique before and after a hands-on training on the epidural simulator. Methods: We enrolled 48 novice trainees who had never previously performed an epidural block. After a standardized learning module, each trainee practiced the epidural procedure on the epidural simulator while wearing a pair of eye tracking glasses (Tobii Pro Glasses 50 Hz wearable wireless eye tracker). After this baseline recording, each trainee spent two hours practicing with the epidural simulator and afterwards once again performed the eye tracking epidural procedure. Eye tracking metrics and epidural learning (duration of the procedure and number of attempts) before and after the simulated practice were recorded. Results: The duration of the epidural procedure and of the epidural needle advancement phase (P < 0.05) and the number of epidural attempts (P < 0.001) were reduced after the tutorial. When considering the eye tracking metrics, after the tutorial, the number of visit counts decreased and their duration increased (P < 0.05). The number of epidural needle insertions (additional attempts) showed a significant positive correlation with the visits number (aOR = 2.02 (95% CI = 1.26 -3.55; P = 0.008)) and a significant negative correlation with the visit duration (aOR = 0.65 (95% CI = 0.39 -0.99; P = 0.05)). Conclusion: We observed significant changes in gaze behavior associated with increased performance during the epidural technique learning with a simulator in anesthesia trainees who had never previously performed an epidural block. These results may create a prototype for future studies on Introduction Epidural block is a complex procedure and requires cognitive skills such as the knowledge of the anatomy along with psychomotor skills such as the skills required to perform the technique. The learning curve is one of the most common tools to assess how the physician in training is progressing at a skill, but the achievement of competence is highly dependent on the subjective and qualitative nature of the definition of learning. Eye tracking is the process of measuring an individual's eye movements, using a device called an eye-tracker, to register his attentional behavior while performing a task. The principle underlying the use of eye-tracking technology is the "eye-mind" theory that suggests there is a relationship between where an individual looks (fixates) and the insight into the cognitive processes that guide this scanning, essentially what he is paying attention to, or thinking about at that point in time. Although cognitive processes are complex, and it is possible that an individual may be simultaneously fixating on one thing, thinking about something else, studies have demonstrated that an individual's attention and thoughts at a given point in time most likely correspond to their point of fixation [1] [2]. In the medical field, perception and visual expertise have a high impact on work efficiency and effectiveness, as well as on the correctness of analysis and diagnosis. Eye tracking has been successfully used for quantitative proficiency assessment and may be a new promising tool to evaluate competency. Current literature demonstrates the ability of eye tracking to provide reliable quantitative data as an objective assessment tool, with potential applications to medical and surgical training to improve performance [3]. Eye metrics can consistently distinguish non-experts from experts, and many studies report that as proficiency in a task improves, there is an increase in focused attention and visual representation during its performance [4] Using the eye-tracking methodology in medical students, Zumwalt et al. [9] demonstrated that education may quite quickly affect visual behavior as new knowledge is gained. Soh et al. [10] examined trainee radiologist readers while they attempted to detect breast lesions on radiologic images before and after an e-learning tutorial, observing an improvement in lesion detection as demonstrated by improved eye-tracking metrics. We performed this study to evaluate the changes in gaze behavior in anesthe-sia novice trainees when performing a simulated epidural technique before and after a hands-on training using an epidural simulator. Methods Forty-eight anesthesia trainees volunteered from the University of Catania Medical School to be enrolled in this prospective, observational study. Each participant gave written

informed consent and privacy, confidentiality and anonymity were fully guaranteed by the EESOA Research Board. In our Region simulation centers do not have access to a formal ethical approval process. Our study was eligible for Ethical Committee's exemption, in accordance with US Federal Human Subject Regulations-Protection of Human Subjects, due to the nature of the study itself, as no patients were involved, the trainees participating were volunteers, researchers ensured that those taking part in the research would not be caused distress, all participants' personal and other data were completely anonymized, and all the investigators had no conflict of interest and were not involved in any of the participants' university teaching programs. In addition, the study was in line with the Healthcare Simulationist Code of Ethics supported by the Society for Simulation in Healthcare [11]. All the novice trainees included in this study had never previously performed an epidural block. All trainees received a standardized three-hour learning module on epidural anatomy and technique as described elsewhere [12] and subsequently practiced the epidural procedure on the epidural simulator while wearing a pair of eye tracking glasses. Each trainee then spent two hours practicing the epidural technique with the epidural simulator, under the supervision of a trained instructor. At the end of this individual hands-on tutorial, every trainee was asked to perform once again the epidural procedure, wearing the eye tracking glasses. For this study we used a commercially available Tobii Pro Glasses 50 Hz wearable wireless eye tracker. This system can measure eye movements using cameras integrated into the eyeglasses which record the corneal reflection of infrared lighting to track pupil position, mapping the subject's focus of attention on video recordings of the subject's field of vision (gaze). In addition to tracking gaze, this system enables the measurement of various eye metrics including fixation frequency and dwell time, used as a surrogate measure of perceived stimulus importance [13]. All the eye-tracked epidural procedures were recorded immediately after accurate individual calibration, during which the participant, after wearing the glasses unit, focused on the center of the calibration target. All the eye tracking video-recordings were stored and analyzed by using a Tobii Pro Lab Software. We selected five areas of interest (AOI), to define regions of a displayed stimulus, and to extract metrics specifically for those regions. The areas were the following: 1) point of the epidural needle at its insertion into the skin; 2) shaft of the epidural needle; 2) hub of the needle; 3) barrel of the syringe; 4) plunger of the syringe; 5) other fields of view. We also divided the epidural procedure into three phases: 1) epidural needle insertion and advancement; 2) loss of resistance procedure (LOR); 3) duration of the whole epidural simulated procedure. Number and duration of fixations, visit counts and their duration for each area of interest and for each phase before and after the hands-on tutorial, were examined. For the purpose of the study we used a validated, modified version of a commercially available epidural simulator (P61, 3B Scientific) which can realistically reproduce the anatomical features of the layers the needle must pass through to reach the epidural space [14]. Number of attempts and total duration of the epidural procedure before and after the hands-on tutorial were recorded. An attempt was defined as a complete withdrawal of the needle from the skin and its reinsertion at the same or at a different interspace. The total duration of the procedure was defined as the length of time between the insertion of the epidural needle into the skin of the simulator and the completion of the loss of resistance technique (finding of the epidural space). Statistical Analysis The power analysis was set according to a parametric model (paired t-test) in order to set 90% test power and a 95% significance level and resulted in a required sample size of 44 observations. A paired t-test was used to evaluate the differences between the variables observed before and after the hands-on tutorial. A logistic regression model was used to evaluate the correlation between the dependent variable "number of epidural needle insertions" and the control variable "eye tracking metrics". It was not possible to make a priori power analysis because the impossibility to make assumptions on number of insertions before the experiment so a post hoc power analysis was carried out for a logistic regression model. Considering the enrolled sample size and setting the significance level of 95% and the effect size of 0.5, the power of test for the logistic regression model resulted to be 80%. The adjusted odds ratio (aOR) was calculated from the logistic regression parameters. Results All the trainees completed the task. Two eye tracking recordings were excluded for technical reasons, leaving 46 for the analysis. The duration of the epidural procedure and of the epidural needle advancement phase (P < 0.05) and the number of epidural attempts (P < 0.001) were reduced after the tutorial (Table 1). When considering the eye tracking metrics, no differences in the number and duration of fixations were noted. After the tutorial, the number of visit counts decreased and their duration increased (P < 0.05) ( Table 2). No differences were noted between the groups during the LOR procedure phase. The radar chart of mean visit count and mean visit duration (seconds) in the assessed AOI during the epidural procedure is reported in Figure 1. Examples of typical gaze plot and heat map before and after the tutorial are shown in Figure 2. Discussion In this study we have examined changes in gaze behavior during the initial training with an epidural simulator in novice anesthesia trainees who had never previously performed an epidural block. The assessment of gaze and attention and their relationship to expertise is the most common use of eye tracking [3]. Significantly greater fixation of relevant anatomic targets in laparoscopic procedures was demonstrated by expert surgeons and similar findings were reported for studies in other non-surgical specialties [4] [5] [6] [7] [8]. Scan path analysis has shown that novices had a much wider search strategy, compared to their more experienced colleagues [4]. In one preliminary study on epidural simulator [15], expert anesthesiologists had an overall shorter duration of fixations during the epidural procedure, and spent more time fixating a more specific target location such as the point of the epidural needle rather than the syringe's barrel, whilst trainees split their attention between tracking their tools and the target location. Interestingly, in this study we have observed significant changes in visit counts and duration, but not in number and duration of fixations, which are the eye tracking metrics associated to experience and expert behavior [4] [5] [6] [7] [8]. A visit is the period of time when a participant first focuses on a region until the person looks away from that region and may consist of at least one fixation but could include dozens depending on the size and content in the region. Visit count and total visit duration can be informative when examining individual participant's interest or ease of understanding. Our study has evidenced more visit counts of shorter duration before the practical tutorial and less visit counts of longer duration after it, which may suggest the acquisition of an initial competence and a more focused attention after the tutorial. These eye tracking pattern changes might represent the early stages of a rudimentary transition from novice to a more capable (but not still expert) behavior. In addition, we noted that changes in visit count and duration were associated with increased performance in carrying out the epidural practical task (number of attempts), and therefore it could be assumed that visit patterns may reflect the very initial process of learning, whereas the fixation patterns may mirror the proficient behavior. The changes we have observed may reflect the perceptual learning (the improvement of perceptual abilities with practice) which, in turn, reflects plasticity in the adult visual system [16]. We hypothesize that exposure to the tutorial might have triggered some learning-related neural changes at an early level of visual processing, representing stimuli relating to task performance [17]. The results of our study are not only purely speculative, but pave the way for future application developments of teaching with simulator task trainers. The changes in eye tracking metrics correlated well with the number of epidural attempts which may suggest the adequacy of eye tracking as a tool to evaluate the advancements of the trainee's learning. In anesthesia, a huge amount of information is presented visually. However, little is known about the anesthetists' distribution of visual attention to gather the relevant information to correctly perform a practical technique. The attention of the operator selects the relevant sources of information from the sensory input and determines which information will be perceived and processed to accomplish the skill, in our case, the epidural technique. This process is performed by the working memory that takes into account the knowledge, expectations and mental models of the epidural anatomy and technique. Eye-tracking provides an informative metric and may be used to develop an objective assessment tool, with which trainees may be objectively monitored during their training period. Using eye tracking to help understand the learning progress of the trainee when performing a practical activity could also supplement the methodology of utilizing gaze behavior to aid in teaching. Chetwood et al. [18] have shown that using eye-tracking technology to provide visual instruction improves completion times and reduces errors in a simulated laparoscopic surgery environment. Although this reported technology requires significant development, the future potential applications are wide-ranging. Conclusions In this study we were able to observe subtle, but significant, changes in gaze behavior associated with increased performance during learning of epidural technique with a simulator in anesthesia novices who had never previously performed an epidural block. We believe that our paper may create a paradigm for future studies on the evolution of eye tracking techniques as a teaching tool and to estimate the progress of learning. MORTALITY AND ITS ASSOCIATED FACTORS IN PATIENTS WITH SPINAL CORD INJURIES AT PARAPLEGIC urvival rates amongst patients with Sspinal cord injury (SCI) have improved in the last few decades, yet mortality of these patients due to certain complications still remains a 1,2 challenge to health care professionals. SCI is a life lasting disability which affects almost every organ of body due to which they are prone to developing 3 complications throughout life. Minor preventable complications such as pressure ulcer (PU), if not prevented and treated timely, can lead to life 4 threatening conditions in such patients. INTRODUCTION urvival rates amongst patients with Sspi nal cord injury (SCI) have improved in the last few decades, yet mortality of these patients due to certain complications still remains a 1,2 challenge to health care professionals. SCI is a life lasting disability which affects almost every organ of body due to which they are prone to developing R e s p i r a t o r y c o m p l i c a t i o n s a n d cardiovascular diseases are the most commonly reported causes of death among SCI patients in developed countries as per recent data. 7 -9 Nevertheless, studies from developing countries reported that infections and septicemia due to UTI and/or PU are still the leading causes of death of such patients in developing countries. [10][11][12][13] The higher incidence of untimely deaths in SCI patients is reported in developing countries as compared to developed countries. In Pakistan, most of the [10][11][12][13] data available about SCI patients is limited to demographic information [14][15][16][17][18] and only a single study has published data on mortality among SCI patients. 19 Therefore, the current study was designed to determine mortality and its METHODS This retrospective cross sectional study was conducted in Paraplegic Center Peshawar, Pakistan, in which clinical records of SCI patients admitted to Paraplegic Center from January 2011 to March 2017 were evaluated. All those SCI patients were included in Peshawar. All patients reported in current study were admitted to Paraplegic Center for rehabilitation and/ or pressure ulcer management after t h e i r i n i t i a l t r e a t m e n t / a c u t e management in acute care hospitals. (Table I). Out The most common cause of SCI

in these patients was fall from height (n=23; 37.1%), followed by road traffic accident (n=17; 27.4%) and then fire arm injury (n=11; 17.7%) while remaining (n=11; 17.7%) patients was having other causes. Majority of patients (n=32; 51.6%) were having complete thoracic paraplegia (Table II). Analysis of data regarding complications and co-morbid conditions showed that 85.5% (n=53) patients were having pressure ulcers while 11.3% (n=7) patients were having limb fractures (Table III). during the 1st year post injury. In accordance with results of previous studies, current study also reported that majority of patients died (82.3%) within the first year of SCI. The most common complication among patients who died in the rehabilitation center was PU. Previous study in same rehabilitation center reported that prevalence of PU in SCI patients to be 63.8% while current study reported that 85.5% patients were having PU at time of death. Two major reasons for high prevalence of PU in SCI patients reported in previous studies can be that, in most hospitals of Pakistan facilities regarding PU prevention are scarce and that's why majority of SCI patients develop PU in hospital. Secondly due to lack of awareness in Pakistani society about rehabilitation most patients visit rehabilitation center after developing 16,21,22 PU mostly they stay at home. Lack of awareness among health care professionals and general population r e g a r d i n g P U p r e v e n t i o n a n d management, inadequate access to health care services, poor hygienic conditions and warm weather are b e l i e v e d t o b e t h e a d d i t i o n a l c o n t r i b u t i n g f a c t o r s f o r t h e development of PU in SCI patients of the country. Due to advances in management and skin care, death rates due to PU have been decreased in last 10 two decades, however, studies from developing countries still has same old high figure of PU as leading cause of [10][11][12][13] death in SCI patients. In current study, 4.8% patients were having DVT at the time of death, while previously 2.7% prevalence of DVT was 1 5 reported in SCI. DVT and its rd complications are reported to be the 3 major cause of death in SCI survivors. Previous studies reported suicide to be one of the leading cause of death among 2,8,13 SCI patients. Though our study as well as other previously conducted studies regarding SCI failed to report suicide among SCI patients. Strong religious beliefs, joint family system and support of local community are believed to be the contributing factors for the low rate of suicides among SCI patients in Pakistan. The data presented in current study was limited to only those patients who died in rehabilitation center that is why all those patients who died in other hospitals or in homes are missed. Secondly, due to retrospective nature of study and incomplete documentation, exact cause of death was not reported. Regardless of these limitations, current study was unique which reported d e m o g r a p h i c i n f o r m a t i o n a n d complications of patients with SCI who died in rehabilitation center. CONCLUSION In our study, relatively younger patients were predominant and complete thoracic paraplegia was the commonest SCI level. Presence of pressure ulcers, limb fractures and deep venous thrombosis in patients with SCI were major contributing factors to morbidity leading to mortality in our patients. mean age of SCI patients, which shows that majority of SCI patients who died in rehabilitation center tended to be older as compared to SCI patients admitted to rehabilitation center in Pakistan. In current study, more than half of SCI patients (56.5%) who died in rehabilitation center were uneducated while previous study from same rehabilitation center reported that 43.0% of the SCI patients were 16 uneducated, which shows that deaths are more common among uneducated SCI patients. Educated SCI patients may take extra care to prevent secondary complications, which can be reason for low mortality among educated SCI patients. The majority of patients in current study were having complete thoracic paraplegia (51.6%) followed by complete cervical tetraplegia (30.6%). Previous studies reported that high neurological level and complete SCI are associated with increased risk of 2,4, 6-8 12 mortality. Similarly, Levy L, et al. reported that 2/3 of SCI patients who st died within 1 year post injury were tetraplegics while 1/3 were paraplegics. However, in current study paraplegic patients were more as compared to t e t r a p l e g i c p a t i e n t s . T h e o n l y explanation for this is less admission of tetraplegics to Paraplegic Center, Peshawar that reported to be only Proximate Chemical Analysis of Luncheon and Burger at Egyptian Markets Department of Food Hygiene, Faculty of Veterinary Medicine, Suez Canal University, Ismailia, Egypt Department of Food Hygiene, Animal Health Research Institute, Ismailia, Egypt Key Laboratory of Animal Epidemiology and Zoonosis, Sharkia Vet. Directorate, General Org. Vet. Services, Ministry of Agriculture, Egypt Department of Avian and Rabbit Medicine, Faculty of Veterinary Medicine, Suez Canal University, Ismailia, Egypt Introduction Adulteration is a serious meat safety and quality issue which becomes the focus of attention for the food industry and consumers in the last few decades (Ahmed et al., 2016). The high price of meat and passiveness of consumer safety warranty further encourage the sellers to substitute components with other replacers in the manufacturing of meat products (Roostita et al., 2014). The meat products adulteration can take many forms such as complete or partial omission or substitution of valuable constituents with undeclared alternatives to increase product bulk or weight or to make the product appears of better value than it is (Hargin, 1996). The modern technology in different fields gives chance for the meat processors to produce new products in different shapes, easily handled, stored and rapidly used. Meat products are highly demanded due to high biological value, reasonable price and agreeable taste and easy during serving (Edris et al., 2012). Prime importance in the meat industry is assuring a good quality control of the products on the market, especially of meat products to avoid the problem of adulterated food. The qualities of raw materials as well as the additives used in the final products are very important for public health. Therefore, the use of low quality ingredients in the processing yields low quality meat products (Pearson and Gillette, 1996). Meat products adulteration can be done by many ways such as complete or partial absence of valuable constituents, whole or partial substitution of components with an undeclared alternative to increase bulk or weight or to improve value (Hargin, 1996). Increases in profitability may be achieved by adulteration to enhance the perceived quality of products, reduce manufacturing costs or for product extension purposes (Ahmed et al., 2016). The detailed information on the chemical and nutritional contents is essential for consumers in choosing meat products (Erwanto et al., 2012). Consequently, consumers have become more selective and more considered about the quality of the product (brand quality, consumer reviewer comments, authority frequent reports….etc.), which became a more significant factor in marketing meat products (Eman, 2009). The widespread of species adulteration in retail markets may be attributed to the inadequate meat inspection and the lack of suitable and affordable analytical methods. However, the ability to detect less desirable or objectionable species in meat products is important for people whose religious practices limit the types of meat they eat as Muslim communities who are particularly concerned about the meat they eat and the accurate labeling which is critical to detect the species that are considered not permissible to eat (Abdelmoety et al., 2019;Abouelmaatti et al., 2013;Farouk, 2013;Rady et al., 2020;Sultan et al., 2020). In addition to increased rate of adulteration with chicken mean, this may be associated with the poultry pathogen epidemiological situation in Egypt in the recent five years and emerging of new pathogens or new variant from old endemic pathogens which associated with increase mortalities in poultry farms either due to viral pathogen like infectious bronchitis, avian influenza, Newcastle virus (Ayoub et al., 2019;Diab et al., 2019;Elhady et al., 2018;Fawzy et al., 2019;Sedeik et al., 2018;Sultan et al., 2019a;2019b) or bacterial pathogen like Pasteurella, e-coli and clostridium or parasite like coccidia (Eid et al., 2016;Elfeil et al., 2012;Enany et al., 2018). Thus, identifying the species of meat in the finished meat products is the main target to fulfil Halal requirement and Islamic regulations (Hamzah et al., 2014). Therefore; the chemical analysis is applied to ensure compliance with legal and compositional standards of some meat products. Materials and Methods Meat products quality is a critical subject and the need of further information about the composition of meat products is increased. Collection of Samples A total of 100 different commercial meat product samples (25 each of beef luncheon, beef burger, chicken luncheon and chicken burger) were randomly collected from local and high different retail markets located in Ismailia city and transferred under refrigeration to the central lab, faculty of veterinary medicine to determine their chemical profiles. Preparation of Samples All samples were prepared and examined according to the technique recommended by AOAC (2003). Determination of Moisture Content Moisture Content was carried using the drying hot air oven. Determination of Crude Protein Content Crude protein content was measured using the Kjeldahl digestion block KJELDATHERM and VAPODEST 50s distillation systems (C. Gerhardt GmbH and Co. KG Cäsariusstraße 97, 53639 Königswinter, Germany). Determination of Total Ash Ash content was measurement using a muffle furnace at 550C. Statistical Analysis Data were statistical analysis using SPSS, Version 16.0 computer program (SPSS, 2007). Least Significant Differences (LSD) at 0.05% level between means were calculated. Correlations between protein and meat content, fat and moisture were performed. Results and Discussion The chemical analysis indicates the nutritive criteria of the examined meat products which is important for consumer health and acceptability. Moisture Content Data in Table 1 showed that the mean values of moisture were 62.66, 61.02, 66.87 and 63.72% for beef luncheon, beef burger, chicken luncheon and chicken burger, respectively. No significant differences between beef luncheon and beef burger and between beef luncheon and chicken burger. But there are significant differences between chicken luncheon and all the other products. These results are lower than those had been recorded by (Ali, 2011) in beef burger and by (Al-Bahouh et al., 2012) in chicken burger and similar to those obtained by (Edris et al., 2012) in beef burger and higher than those been recorded by (Edris et al., 2012) in beef luncheon and by (Prayson et al., 2008) in beef burger as well as by (Jecan et al., 2013), (Ahmad et al., 2015) and (Al-Dughaym and Altabari, 2010) in chicken burger. As Egypt has its own legislative nutritive criteria for meat products, the samples were compared with the Egyptian standards (EOS, 2005) to estimate their acceptability. The percentages of the non-accepted samples for moisture were 52, 44, 100 and 0.0 for beef luncheon, beef burger, chicken luncheon and chicken burger, respectively (Table 1). Multiple factors can affect the moisture content include fat percentage (inverse relationship), addition of water and non-meat ingredients and degree and type of processing and cooking. The moisture content of a meat product is important because of the properties of water, its interaction with other components in the product and its contribution to the chemical, biological and physical properties of foods (Cornejo and Chinachoti, 2003). Protein Content The mean values of protein % in the examined meat products were 8.50, 11.54, 12.45 and 11.34 for beef luncheon, beef burger, chicken luncheon and chicken burger, respectively. There are significant differences between beef luncheon and the other products, while no significant differences between them (Table 2). These results were consistent with previous reports of El-Tahan et al. (2006) in chicken luncheon (El-Tahan et al., 2006). These results were much higher percentages had been reported by (Edris et al., 2012) in beef luncheon and beef burger and (El-Tahan et al., 2006) in chicken burger. On the other hand, these results were

lower the results obtained by (Ali, 2011) in beef burger. The percentage of the non-accepted samples for protein % according to the Egyptian standards (EOS, 2005) which were 96, 84, 48 and 64 for beef luncheon, beef burger, chicken luncheon and chicken burger, respectively ( Table 2). The low protein content of some meat products may be due to addition of improper meat cuts or the use of meat trimmings in preparation or substitution with non-meat components, since meat proteins are relatively more expensive than non-meat components (Lawrie, 1998). Many of the protein sources used in preparing commercial products used presently are partially replaced by non-meat protein source. In addition to lowering the cost, non-meat protein sources such as egg, whey protein and soy protein, are able to improve the flavour and texture of the product by increasing the fat and moisture binding ability (Kassem and Emara, 2010). Definitely, these ingredients decrease the production cost and also enhance sensory quality but could not fulfill the prescribed limit for proteins (Turhan et al., 2007). Fat Content The mean values of fat % in the examined meat product samples were 5.25, 17.13, 4.65 and 8.26 for beef luncheon, beef burger, chicken luncheon and chicken burger, respectively. There was a significant difference between beef burger and all the other products and between chicken burger and all the other products, while there were no significant differences between beef luncheon and chicken luncheon (Table 3). These results were relatively lower than found by (Edris et al., 2012) in beef luncheon and beef burger as well as (Al-Dughaym and Altabari, 2010) and (El-Tahan et al., 2006) in chicken burger and by (Ali, 2011) in beef burger. Similar results to that were obtained by (Jecan et al., 2013) in chicken burger and by (El-Tahan et al., 2006) in chicken luncheon. On the other hand, These results were lower results to that obtained by (Ahmad et al., 2015) and (Al-Bahouh et al., 2012) in chicken burger. By comparing the results with the Egyptian Standards (EOS, 2005), 36% of beef burger only were unaccepted based on their fat content as revealed in Table 3. The function of fat is mainly influencing the sensory quality of burgers, particularly its flavor (Suman and Sharma, 2003). Fat as a major food component is used for its sensory and physiological benefits that contribute to the flavor, taste and aroma/odor of the final products (Moghazy, 1999). From the present result, it was clear that the examined products contained low cost fat substitute that resulted in the production of low-fat products. The production of low fat meat product can be achieved by increase in the carbohydrates and added water content that does not affect the traditional full-fat flavor, taste and texture but reduce the formulation cost (Ibrahim et al., 2011). This trend toward production of low-fat meat product might be a result of adulteration and the trials of the industry to reduce the cost, which make their products non-conferment with the standard. Ash Content The mean values of ash % in the examined meat products were 3.07, 4.01, 3.40 and 3.17 for beef luncheon, beef burger, chicken luncheon and chicken burger, respectively. There were significant differences between beef burger and all the other products, no significant differences between beef luncheon and chicken burger and no significant differences between chicken luncheon and chicken burger. The current results were lower the results obtained by (Edris et al., 2012) in beef burger as well as by (Al-Dughaym and Altabari, 2010), (Ramadhan et al., 2011) and (Al-Bahouh et al., 2012) in chicken burger. The current results were higher finding were reported by (Edris et al., 2012) in beef luncheon. The percentages of the non-accepted samples for ash % according to the Egyptian standards (EOS, 2005) were 12, 52 and 84 for beef luncheon, chicken luncheon and chicken burger, respectively as shown in Table 4. Ashes represent the total minerals found in food such as sodium, phosphorus and iron, that can be contributed by the meat as raw material, salt and spices added (Fernández-López et al., 2006). The ash content in meat products not only depend on muscle minerals but also on the curing salt added (Kirk and Sawyer, 1991). The high ash content could be achieved by the addition of spices for seasoning, high fiber carbohydrate, starches, cereals, soya-protein and salt. Incorporation of mechanically deboned chicken meat might be another reason for increasing the ash content (Babji et al., 2000). Carbohydrate Content The mean values of carbohydrate % in the examined samples were 20.49, 6.29, 12.63 and 13.49 for beef luncheon, beef burger, chicken luncheon and chicken burger, respectively. There were significant differences between beef luncheon and all the other products and between beef burger and all the other products. But there were no significant differences between chicken luncheon and chicken burger (Table 5). These results were nearly consistent with previous reports of ( Ramadhan et al., 2011) in chicken burger. While current results were higher the results had been reported by (Abd-Elhak et al., 2014) in beef burger. However, these results were lower the results were obtained by (Babji et al., 2000) and (Al-Bahouh et al., 2012) in chicken burger. The Egyptian Standards (EOS, 2005) of beef burger stated that carbohydrate should not exceed 10%. The results showed that 20% of the samples were not in compliance with the Egyptian standards. Carbohydrates in burgers are mainly from the use of starches as ingredients. Starches, such as maize, tapioca, rice, potato and wheat, have been used as meat filler and water binder and this result could be due to the use of cheap ingredients like rusk, bread crumbs, cereal and soya protein (Joly and Anderstein, 2009). Meat Content The mean values of meat content % in the examined meat products were 26. 05, 47.16, 49.67 and 44.14 for beef luncheon, beef burger, chicken luncheon and chicken burger, respectively. There were significant differences between beef luncheon and the other products which had no significant differences between them (Table 6). These results are lower than those been recorded by (Ramadhan et al., 2011) and(Al-Bahouh et al., 2012) in chicken burger and higher than those been recorded by (Prayson et al., 2008) in beef burger. (EOS, 2005) Means with the same letter are not significant difference (P< 0.05). By comparing the results with the Egyptian Standards (EOS, 2005), it is evident that 100%, 80%, 100% and 88% of beef luncheon, beef burger, chicken luncheon and chicken burger, respectively were unaccepted based on their meat content which were revealed in Table 6. The low meat content of the examined products could be contributed to the presence of fat and the replacement of expensive meat protein by cheap binders and fillers (Babji et al., 2000). The statistical results showed that the Pearson's correlation between protein and meat content was positive and significant at the 0.01 level (R 2 = 0.975) as shown in Table 7. While Pearson's correlation between moisture and fat is negative and significant at the 0.01 level (R 2 = -0.759) as shown in Table 8. Conclusion The present study showed a significant difference in quality attributes between examined meat products. Most of them were lower in fat, protein and meat contents and higher in carbohydrate and moisture contents compared to the Egyptian Standards. It is quite evident from our study that the food industry is non-conferment with the regulatory requirements for meat quality standards. Characteristics of carcass and non-carcass in F1 population crossbred brown and black Japanese quails The aim of this study was to evaluate characteristics of carcass and non-carcass in F1 population crossbred brown and black Japanese quails. A total of 80 birds have been raised under the same feeding and management in battery cage system. They were divided into four groups based on type of crossing. Brown male quail and Black female quail crossing produced Brown layer lines (BL12 and BL21 lines), on the other hand Black male quail and Brown female quail crossing produced layer populations called LB12 and LB21 lines. The data observed was live body weight, carcass and non-carcass percentages, and abdominal fat percentage. The data was analysed using analysis of variance and if there were pairwise differences among lines Duncan’s multiple range test (DMRT) was performed. Significant differences among lines have been found for live body weight, carcass and non-carcass percentages. Live body weight in LB12 and LB21 populations were significantly higher than others (P<0.001). In addition, highest carcass percentage and lowest non-carcass percentage were found in LB12 population (P=0.0175 and P=0.0159, respectively). Abdominal fat percentage was not different among lines. In conclusion, F1 population of LB line had overall better live body weight, carcass and non-carcass percentages than BL lines. Introduction Japanese quail (Coturnix japonica) is potential small poultry species to be developed as an animal protein producer due to easy to be raised and it does not need large area in its farming practice. Japanese quail is commonly raised not only to produce egg but also to produce meat in some countries [1]. It has most carcass percentage to be utilized for human consumption among domesticated poultry species [2]. In Indonesia, most of unproductive female quails will be culled out to be slaughtered due to economically unprofitable, on the other hand male quail is intentionally raised to produce meat [3]. Dressed poultry carcass, dead poultry body after being partially butchered without internal organs, head, and legs, is a kind of poultry performances which is affected by both genetic and environment factors such as age of animal, sex, line, and livestock management [4]. Meat quail with excellent dressed carcass characteristics could be produced by crossbred quails genetically superior for meat producer. Genetic potential of quails can be evaluated if they are maintained under the identical condition and management [5]. There are two lines of Japanese quail commonly found in small-holder farmers in Indonesia, i.e. Black plumage and brown plumage lines [6]. Different quail lines developed from different regions may perform different carcass and non-carcass traits. To date, report regarding carcass and non-carcass characteristics of different Japanese quail lines in Indonesia is not previously found. Farmers raises their quails without scientific knowledge and only based on their field experiences, therefore developing quail with superior traits of interest including carcass and non-carcass characteristics becomes very important. A way to create superior quail lines for meat and carcass traits is by crossing among lines. The objective of present study was to evaluate live body weight, dressed carcass percentage, non-carcass percentage, and abdominal fat percentage of four F1 population crossbred brown and black Japanese quails. Quail population and management A total of 80 heads of Japanese quail have been used in this study. They were constructed by four different crossings. They were 20 heads of crossbred Brown male quail 1 and Black female quail 2 (BL12), 20 heads of crossbred Black male quail 1 and Brown female quail 2 (LB12), 20 heads of crossbred Brown male quail 2 and Black female quail 1 (BL21), and 20 heads of crossbred Black male quail 2 and Brown female quail 1 (LB21). Day old quails (DOQs) were weighed after their plumages were dried, and they were placed into brooder cage. The DOQ was fed by commercial BR1 containing 22% crude protein and 3100 kcal energy metabolism (PT. Wonokoyo Jaya Corporindo, Indonesia) for 30 days of age. Quails were transferred to colony cage (50 x 30 x 20 cm) and each colony cage was occupied by five quails. Furthermore, they were fed by commercial feed for laying quail containing 20% crude protein and 2900 kcal energy metabolism (PT. Japfa Comfeed Indonesia, Indonesia). The quantity and nutrient content of feed given to quail were following Shanaway [7]. All quails were free to access water (ad libitum). In addition, they were vaccinated with ND La Sota at 30 days of age and Vitastress was given to quails every week where the dose was given according to company procedure (PT. Medion Farma Jaya, Indonesia). They were exposed by sunlight for about 12 hours and dim light was applied at night to prevent rodent and other animal attacks using bulbs. Slaughtering and carcass traits of F1 population Crossbred Brown and Black Japanese quail All quails have been fasted for 3.5 to 4 hours and they were weighed using digital weighing scale (Notebook

Digital Scale, Indonesia) before slaughtered. Moreover, they were slaughtered by following Islamic law and carcass parts were processed according to procedure developed by Genchev and Mihaylov [8]. Whole dressed carcass and carcass parts have been weighed with Notebook Digital Scale. Data collected in this study were live body weight, carcass traits, non-carcass traits, abdominal fat percentage. Carcass, non-carcass traits and their parts have been measured by following formulas [9,10] Statistical analysis The data observed in this study was analyzed using analysis of variance (ANOVA) and the alpha was set at 5%. In addition, Duncan's Multiple Range Test (DMRT) was performed to distinguish among quail populations [11]. Mathematical model used in this study is as follows: Yij is the observed value at i th crossing and j th replicate, μ is overall mean, Ʈi is the effect of i th crossing, ɛij is random error. Live body weight Live body weights of each Japanese quail line in this study is presented in Table 1. Live body weight of crossbred black male quail and brown female quail (LB lines) was significantly higher than crossbred brown male quail and black female quail (BL lines) (P<0.01). The LB quail lines were 8 to 14 grams higher than BL quail populations. This result indicated that LB lines is genetically more potential to be developed as meat-type quail than BL lines. Crossing between quail lines generates genetically different quail that may affect quantitative traits [2,12]. Previous study also explained that live body weight significantly affected by genetic factor [4]. Lotfi et al. [13] approved that genetic factor contributed greatly to live body weight in quail which is indicated by high heritability value (h 2 ) for live body weight. The h 2 value of live body weight is 0.45 that means 45% of body weight traits in quail inherited by parents to their offspring. In addition, live body weight of quail population in this study is quite similar with previous study reported by Al-Kafajy et al. [14]. Carcass traits Carcass percentage and its parts are presented in Table 1. Crossing affected carcass parts of Japanese quail significantly (P<0.05). Average values of carcass percentage, breast, wings, back, and thighs in LB12 line population were found higher than both LB21 and BL12 lines, however there was no differences with BL21 line. Comparing with previous study, these results were relatively lower where carcass percentage of mature quail can reach up to 67.19% [2]. Genetic and environment condition may greatly contribute to this difference. Moreover, live body weight may directly affect carcass percentage. Higher body weight of quail produces higher dressed carcass weight and carcass percentage [2,15]. Non-carcass traits Profile of non-carcass traits produced by four lines of Japanese quail is presented in Table 1. Non carcass percentages of BL12 and LB21 lines were significantly higher than LB12 line, and yet they were statistically not different with BL21 line. Non-carcass components were also significantly different among quail lines. This result was much higher than previous study where non-carcass percentage of Japanese quail was about 26 to 32% [2]. Higher non-carcass percentage is tightly correlated with lower carcass percentage and vice versa since they are negatively correlated [15]. The LB12 population have overall better carcass and non-carcass traits compared to other quail lines. Abdominal fat percentage The difference of abdominal fat percentage among quail lines was not found in this present study (Table 1). It may be due to quails raised under similar environment conditions that may contribute to similar abdominal fat percentage. Tumuva and Teimouri [16] explained that abdominal fat percentage in poultry was affected by slaughtering age, sex, nutrient contents and feed consumption. In addition, sex and the amount of feed given to quail greatly affect abdominal fat deposition [17,18]. Conclusion In conclusion, crossbred Black male and Brown female quails was better in live body weight, carcass and non-carcass traits than crossbred Brown male and Black female quails. It could be considered for quail breeding program and strategy to create meat-type Japanese quail. Adaptation reveals multi-stage coding of visual duration In conflict with historically dominant models of time perception, recent evidence suggests that the encoding of our environment’s temporal properties may not require a separate class of neurons whose raison d'être is the dedicated processing of temporal information. If true, it follows that temporal processing should be imbued with the known selectivity found within non-temporal neurons. In the current study, we tested this hypothesis for the processing of a poorly understood stimulus parameter: visual event duration. We used sensory adaptation techniques to generate duration aftereffects: bidirectional distortions of perceived duration. Presenting adapting and test durations to the same vs different eyes utilises the visual system’s anatomical progression from monocular, pre-cortical neurons to their binocular, cortical counterparts. Duration aftereffects exhibited robust inter-ocular transfer alongside a small but significant contribution from monocular mechanisms. We then used novel stimuli which provided duration information that was invisible to monocular neurons. These stimuli generated robust duration aftereffects which showed partial selectivity for adapt-test changes in retinal disparity. Our findings reveal distinct duration encoding mechanisms at monocular, depth-selective and depth-invariant stages of the visual hierarchy. When interacting with our environment we continually monitor event duration. For example, when watching streamed video content, the quality of our viewing experience is greatly degraded by 'packet loss' leading to the stream's 'stalling' or 'freezing' 1 . The extent of this perceptual degradation increases sharply as the stalling duration increases between 0-1000 ms 2 . Duration-dependency of this kind is just one example of perception underpinned by the rapid, accurate encoding of sub-second visual durations. Although this encoding often yields duration estimates that are both noisy 3,4 and distorted 5,6 , the lack of any duration-specific sensory receptor surfaces arguably makes it surprising that it can be completed at all. Uncertainty about how the nervous system computes duration has led to an assortment of candidate mechanisms whose proposed neural loci range from the pre-cortical to the association cortices. The most peripheral examples are founded in the properties of pre-cortical neurons with circular, small-diameter spatial receptive fields. These models provide explanations for visual temporal distortions with very narrow (~1°) spatial tuning and no specificity for visual orientation [7][8][9] . A pre-cortical locus for duration processing is given further credence by studies where temporal perception has been measured in patients whose inter-hemispheric communication is prevented via commissurotomy 10 . These patients can show normal cross-hemispheric duration discrimination performance alongside marked deficits in cross-hemispheric spatial discrimination 11 . This strongly suggests pre-cortical processing of duration which then projects bilaterally to both cortical hemispheres 12 . Alternatively, several lines of evidence point towards duration encoding being a cortical phenomenon. For example, training-induced improvement of duration discrimination sensitivity transfers fully between locations spanning trained and untrained visual hemispheres 13 . Moreover, some perceptual distortions of perceived duration show broad 14 or non-existent 15,16 spatial tuning. Where spatial tuning has been reported, its characteristics have variously included dependency on stimulus size 14 or spatiotopic position 17,18 (but see 19 ). In some cases, orientation specificity has also been reported, consistent with the orientation tuning profiles of striate cortical neurons 20,21 . Although these findings are suggestive of a cortical basis for duration processing, uncertainty persists about whether they are examples of specificity for duration or other perceptual parameters. For example, the cross-hemisphere transfer of training-induced sensitivity improvements could simply reflect non-specific reductions in high-level decisional noise 22 , rather than improvements in duration encoding accuracy per se. Similarly, whilst a lack of orientation specificity could reflect a (pre-cortical) processing stage that precedes orientation encoding it is also consistent with downstream stages where orientations are pooled (e.g. 23,24 ) to provide high-level specificity for object categories 25 . Finally, experimental manipulations designed to test the spatial specificity of duration distortions have the unintended consequence of manipulating the locus of spatial attention (e.g., adapting to the temporal properties of a stimulus presented at one location then testing at a different location). Given that spatial attention itself strongly influences perceived duration [26][27][28] , it likely contributes to the presence or absence of spatial tuning via the attentional modulation of (a) adapting and/or test stimulus' duration, or (b) gain adaptation amplitude itself. In the current study, we address the question of whether duration encoding has a localised position within the visual processing hierarchy. To do so, we combine the technique of sensory adaptation with methodology that allows us to utilise the anatomical progression from monocular, binocular, depth-selective and depth-invariant processing. Observers adapted to repeated presentations of relatively long or short sub-second visual durations. In keeping with earlier reports 29,30 adaptation generated bidirectional duration aftereffects: subsequently viewed test durations were perceptually expanded/contracted in the opposite direction to the adapting stimulus. Using dichoptic presentation, adapting and test stimuli were isolated within monocular channels. This allowed us to test for duration aftereffect selectivity within processing stages at or above those underpinned by neurons in V1: the major recipient of geniculate input 31 and the first site of anatomical convergence between the eyes 32 . We found robust interocular transfer: substantial duration aftereffects persisted when adapted or non-adapted eyes viewed a luminance-defined test stimulus. Nevertheless, aftereffects failed to transfer completely between the eyes, revealing a small but significant contribution from monocular channels. We then systematically tested the relationship between duration selectivity and more sophisticated levels of binocular processing. To do so, we deployed a novel stimulus which allowed the presentation of durations defined solely by inter-ocular retinal disparity. Despite being completely invisible to monocular mechanisms, these durations generated robust duration aftereffects. Finally, we demonstrate that disparity-defined duration aftereffects show limited transfer across large adapt-test differences in disparity-defined depth plane. This transfer exposes a downstream contribution from a mechanism which encodes duration after pooling across its disparity-selective afferents. These findings indicate that visual duration processing cannot be explained by sub-cortical mechanisms alone. Instead, it receives input from multiple processing stages, consistent with duration being a generic stimulus property whose encoding is shared across mechanisms located both upstream and downstream from the site of binocular integration. Results We began by assessing the extent to which the mechanisms contributing to duration aftereffects exhibit elementary binocularity. During the adaptation phase (see Fig. 1A and Methods for details), observers viewed repeated presentations of isotropic, luminance-defined Gaussian blobs. All stimuli were presented via a two-mirror stereoscope which allowed adapting and test stimuli to be presented to the same or different eyes. The duration of these visual stimuli was fixed at relatively long (666 ms) or short (166 ms) durations which were held constant within an experimental session. Following adaptation, the test phase comprised a duration discrimination judgment between a fixed duration 333 ms reference stimulus (auditory tone) and test stimulus (Gaussian blob) whose duration varied around 333 ms. When these judgments were plotted as a function of test duration, the resultant psychometric functions allowed extraction of the Point of Subjective Equality (PSE): the physical test duration perceptually equivalent to reference stimulus ( Fig. 1B and C). Longer/shorter PSE values reflect adaptation-induced perceptual contraction/ expansion of test duration, respectively. Differences between PSE values corresponding to the 'adapt 166 ms' and 'adapt 666 ms' conditions reflect the magnitude and direction of any duration aftereffect. When the adapting and test stimuli were both presented to the same eye, adaptation to relatively long durations induced perceptual contraction of the moderate duration test stimulus. This can be seen in Fig. 1B (yellow bars): post-adaptation, longer duration test stimuli (and thus longer PSE values) were required to maintain perceptual equivalence with the 333 ms reference stimulus. The reverse is evident following adaptation to relatively short durations (blue bars). This pattern of perceptual distortion represents a repulsion-type after effect: adaptation distorts perceived duration in the opposite direction to the adapting stimulus. Consistent with earlier reports 29,33,34 , this effect is asymmetrical around the 333 ms reference stimulus value ( Fig. 1B and C, horizontal red line), reflecting the fact that auditory durations are typically perceived as being longer than their (physically identical) visual counterparts, irrespective of adaptation. Importantly, a similar pattern of duration aftereffects can also be observed when adapting and testing durations are presented to opposite eyes (Fig. 1C). The arithmetic difference between PSEs for each adapting duration provides a measure of duration aftereffect magnitude. Plotting these values for each observer's 'same' and

'different' eye conditions ( Fig. 2 -green circles) allows visualisation of any systematic trends towards greater duration aftereffect magnitude values in either condition. Averaged across observers, duration aftereffect magnitude values were 69 ms in the "same eye" condition and 52 ms for the "different eye" condition ( Fig. 2, black data point). Permutation testing showed that both values were significantly different from zero (p < 0.001) and that the difference between them (Fig. 2 -black circle's departure from the diagonal) was also highly significant (p < 0.001). These data therefore represent duration aftereffects mediated by mechanisms that are predominantly binocular, alongside a small but significant monocular component. Despite some evidence for modest inhibitory interactions between eye inputs at the level of the LGN 35,36 , the first examples of excitatory convergence upon binocular neurons are found in the superficial and deep layers of V1 37 . The aftereffect's small monocular component could therefore reflect pre-cortical adaptation 38 or adaptation within monocular V1 (i.e. layer 4 C 39 ) neurons. The much larger binocular component could be mediated within cortical neurons showing rudimentary binocularity (i.e. responsivity to binocular stimulation but limited integration of monocular inputs) or more sophisticated forms of binocular processing such as the extraction of retinal www.nature.com/scientificreports www.nature.com/scientificreports/ disparity. In primate V1 for example, approximately 50% of binocularly responsive neurons show selectivity for retinal disparity information that can only be extracted by comparing the spatial properties of their monocular inputs 40,41 . To test whether the binocularity shown in Fig. 2 allows the extraction of duration information that is invisible to monocular channels we measured the effects of adapting to durations defined solely by the presence or absence of crossed retinal disparity. By presenting the disparity only during the periods corresponding to adapt/test stimulus durations, we excluded monocular contributions to perceived duration. Observers viewed left and right stimuli simultaneously via a mirrored stereoscope. These images were identical except for a lateral shift within a subset of pixels (see Methods) which defined a disc shaped region of +6′ crossed retinal disparity (Fig. 3A). To prevent monocular artefacts (e.g., positional shift cues), we embedded the disparity within dynamic luminance noise (see Methods) which precluded monocular tracking of pixel positions around the time of the disparity's introduction/removal (i.e. the disparity-defined duration's onset/offset). The disparity was presented for intervals matching the required adapting or testing duration and therefore disappeared during the (zero disparity) inter-stimulus intervals. The disparity-defined adapting and test stimuli were identical except for their differences in duration. Adapting and test durations matched those used in the interocular transfer experiment. The resultant PSE values for each adapting duration are shown in Fig. 3B. Despite the invisibility of adapting and test stimuli to monocular mechanisms, all observers show a robust pattern of duration aftereffects. Specifically, adaptation to relatively long/short disparity-defined durations causes significant (p < 0.001) perceptual contraction/expansion of subsequently view disparity-defined test durations. Taken together with the results of the inter-ocular transfer experiment, this finding suggests that duration encoding must occur within at least two processing stages: one monocular and a second with selectivity for duration, and retinal disparity. In the primate visual system, the proportion of disparity-tuned neurons increases progressively from V1 through V2-V3/V3A 42 , MT 43 and V4 44 . However, higher-order visual processing in V4 45 , inferotemporal 46 , intraparietal 47,48 and lateral occipital 49 cortices shows neuronal selectivity for object shape that is invariant across changes in retinal disparity. In many cases, these object shapes are defined by their three-dimensional structure, suggesting that this structure is constructed via pooling across upstream disparity-selective inputs 50 . It therefore follows that if duration aftereffects fail to transfer across adapt-test changes in stimulus depth plane the locus of the underlying mechanism would be constrained to processing stages between the encoding of, and pooling across, retinal disparity. We tested this hypothesis by measuring duration aftereffects across a 48′ difference in the retinal disparity defining the adapting and test stimuli (Fig. 4A). Test stimuli were always presented with −24′ retinal disparity www.nature.com/scientificreports www.nature.com/scientificreports/ whereas adapting stimuli were presented with either the same (−24′ (Fig. 3C)) or different (+24′ (Fig. 4A)) retinal disparity (see Methods for details). We selected these values to minimise the possibility that adapting and test stimuli might stimulate common populations of neurons with broad disparity tuning (Parker, 2007). Figure 3D shows PSE values for the 'same disparity' conditions. Following adaptation to durations defined by uncrossed retinal disparity, all observers show duration aftereffects comparable with those observed following adaptation to luminance defined (Fig. 1B) and crossed retinal disparity defined (Fig. 3B) durations. This equivalence suggests that duration adaptation may be insensitive to the nature of the disparity defining the adapting and/or test stimulus' duration. When this possibility is tested by the introduction of a large adapt-test retinal disparity difference (Fig. 4A) significant duration aftereffects persist (Fig. 4B), albeit of a substantially reduced magnitude relative the 'same disparity' condition (Fig. 3D). This reduction represents a degree of selectivity for retinal disparity. This selectivity is highlighted when Figs 3D and 4B's PSE data is expressed as duration aftereffect magnitude (Fig. 5). For example, all observers show significant duration aftereffects when adapting and test stimuli's durations are defined by the same retinal disparity (horizontal distance between Fig. 5's green data points and the y-axis) but for 6 out of 7 observers these aftereffects are of reduced magnitude when adapting and test stimuli are defined by different retinal disparities. Averaged across observers, duration aftereffect magnitude values were 71 ms in the "same disparity" condition and 39 ms for the "different disparity" condition ( Fig. 5, black data point). Permutation testing showed that both values were significantly different from zero (p < 0.001) and that the difference between them (Fig. 5 -black circle's departure from the diagonal) was also highly significant (p < 0.001). Discussion We sought to investigate the binocularity of human duration perception. We used sensory adaptation techniques to generate duration aftereffects. We tested the specificity of these aftereffects via adapt-test stimulus manipulations that allowed the quantification of input from monocular, disparity-selective and disparity-invariant processing stages. Duration aftereffects showed extensive interocular transfer and are therefore generated by duration-selective mechanisms that are primarily binocular in nature. Nevertheless, a small but significant component of these aftereffects showed selectivity for eye of input. We then explored the extent to which the binocular component is mediated by neurons able to extract duration information from retinal disparity-defined stimuli which were invisible to monocular mechanisms. Disparity information alone was sufficient to generate robust duration aftereffects, indicating a duration encoding stage that is distinct from the most basic forms of binocularity. Finally, we found partial transfer across changes in the disparity which defined adapting and test durations, indicative of a third processing stage which pools duration information across disparity-selective inputs. www.nature.com/scientificreports www.nature.com/scientificreports/ Bidirectional, repulsion-type duration aftereffects are consistent with the 'channel-based' duration encoding predicted by the response properties of duration-tuned neurons (DTNs). DTNs have been documented at multiple neural sites within a range of species 51 . Specifically, duration aftereffect magnitude varies with the similarity between adapting and test duration 30 , in keeping with model predictions rooted in activity within bandwidth-limited DTNs that varies around the neuron's 'preferred' duration 30,51,52 . Although auditory DTNs have been reported in both cortical and subcortical structures 53 reports of DTNs within the visual system havethus far -implicated cortical sites alone 54,55 . The small but significant degree of eye input selectivity reported here (Fig. 2) is not consistent with selectivity for duration in extrastriate areas 55 where virtually all neurons are binocular [56][57][58] . It is, however compatible with duration aftereffects coded via the activity of monocular neurons in V1. Although duration-tuned activity has been reported within the simple cells of area 17 of cat visual cortex 54 , the degree of binocularity exhibited by these cells remains unclear. Alternatively, monocularity could be explained by a pre-cortical contribution to the mechanism generating the duration aftereffects themselves or that mechanism's upstream temporal inputs. Duration-dependent firing patterns have been documented within retinal ganglion cells (RGCs) 59 and cells within the lateral genicular nucleus (LGN) 60 . Both cell types show inhibition during stimulus presentation followed by rebound spiking response at stimulus offset. The strength of this offset response increases monotonically with increases in stimulus duration across the millisecond range. However, unlike their cortical analogue 54,61 , this duration-dependency is 'long-pass' rather than the bandpass type of tuning needed to produce duration aftereffects tuned around the adapting duration 30 . Thus, any subcortical contribution to duration adaptation mechanisms is more likely to occur via adaptation-induced distortion of afferent temporal information prior to its arrival at cortical bandpass DTNs. In addition to the duration-dependent RGC and LGN cell characteristics described above, several alternative candidate mechanisms could also distort the signal within monocular channels. As discussed earlier (see Introduction), adaptation to drifting or flickering gratings induces compressive duration distortions that display multiple characteristics consistent with a neural locus within magnocellular layers of the LGN 7,8,62 . Equally, a monocular contribution could arise via duration encoding mechanisms which monitor the spatiotemporal evolution of neural activity within 'state dependent networks' (SDN) 63,64 or duration-dependent changes in response magnitude within any neuron directly activated by the stimulus (so called 'Energy Readout' models [65][66][67] ). Both SDN and Energy Readout are examples of 'distributed' duration processing that could -theoretically, at leastoperate at any neural scale of stimulus-driven activity and therefore provide distorted monocular input to downstream duration aftereffect generating mechanisms. The extensive degree of interocular transfer (Fig. 2) requires the input of binocular neurons in striate or extrastriate cortices. The fact that duration aftereffects can be generated by durations defined solely by retinal disparity (Figs 3-5) suggests that adaptation to such stimuli is generated via at least two possible routes. The first would be via a mechanism that is selective for both duration and retinal disparity. Figure 3's data shows that precisely this mechanism operates during adaptation: when adapting and test durations are defined by retinal disparities that stimulate the same populations of disparity selective neurons, aftereffect strength is maximal. When adapting and www.nature.com/scientificreports www.nature.com/scientificreports/ test disparities stimulate distinct neural populations, duration aftereffect magnitude is markedly reduced (Figs 4 and 5). This implicates duration selectivity within disparity-selective striate or extrastriate cortices, with the later arguably more likely given the increasing prevalence of disparity selectivity (and absence of monocular neurons) in extrastriate areas [42][43][44]68 . Although the DTNs reported in area 17 and 18 of cat cortex could also be tuned for retinal disparity, this association has yet to be tested directly 54 . The second route would be via a duration-selective mechanism downstream from the initial coding of retinal disparity. For example, higher-level DTNs could inherit temporal information via their disparity-selective inputs. These neurons could therefore show selectivity for duration without having any intrinsic selectivity for disparity. Figures 4 and 5's data show that duration aftereffects are reduced but not abolished by adapt-test changes in retinal disparity, revealing significant contributions to duration encoding from both disparity-selective and disparity-invariant mechanisms. The latter's contribution to duration aftereffects could, in fact be a signature of a mechanism that pools across all visual stimulus features. However, such an extreme form of stimulus invariance is not compatible with the aftereffect's size-dependent spatial specificity 14 . More plausibly, disparity invariance may be a signature of duration encoding that is selective for stimulus duration and higher-order stimulus features that are re-constructed after pooling across disparity-selective inputs. For example, human fMRI evidence shows neural selectivity within lateral occipital complex (an object selective mid-level visual area bordering V5) for depth-defined object shape but not the depth plane occupied by the object 49 . Likewise, primate V4 neurons show selectivity for 3D slant (i.e. linear disparity gradient) across changes in the shape's distance from fixation 45 . More abstract still are objects defined by 3D curvature in depth 50 . In inferotemporal and intraparietal cortex, neurons maintain selectivity for depth-defined curvature and orientation across changes in retinal disparity and the cues defining 3D shape 46,47,69 . Could these same disparity-invariant neurons also encode stimulus duration and therefore mediate the duration aftereffect's partial transfer across disparity? Although the answer remains unclear, duration-dependent firing patterns have been observed in monkey intraparietal cortex 70 and recent human fMRI adaptation experiments report

duration-tuned modulation of neural activity in the same region 55 . Human psychophysical evidence also supports the concept of processing stages which pool across retinal disparity. Specifically, higher-level examples of shape, curvature, numerosity and motion perception are insensitive to changes in disparity-defined depth [71][72][73] . The binding of duration to stimulus features of this type would allow perceived duration to retain object specificity across changes in a range of horizontal positions 14 , depth planes and viewing angles. In summary, we have demonstrated that duration encoding receives input from three different processing mechanisms at distinct strata within the visual processing hierarchy. An important question is the extent to which these processing stages operate in serial or parallel. The latter risks the prospect of multiple, possibly independent, duration estimates co-existing within the nervous system. This could be problematic unless they are integrated by a central mechanism which then outputs a single duration estimate. That degree of centralisation is difficult to reconcile with the stimulus specificity demonstrated in the current study and elsewhere. A perhaps more credible scenario would be for duration processing to form a serial 'cascade' where downstream duration encoding mechanisms apply cumulative adaptation to effects inherited from their upstream counterparts. Cascading adaptation through retinal to extrastriate sites has been reported for encoding of contrast and motion 38,74,75 . Specifically, the same spatial adaptation changes response gain in primate LGN neurons and stimulus selectivity in V1 38 , whilst motion adaptation changes responsivity in extrastriate area MT with a spatial specificity predicted by adaptation inherited from V1 74 . Although our knowledge of hierarchical interaction in the temporal domain remains primitive 76 , a picture is beginning to emerge of temporal perception underpinned by neural mechanisms whose primary function has hitherto appeared to be ostensibly spatial. Ongoing experiments will address the degree to which temporal perception can be explained by the characteristics of spatial perception, and vice versa. Methods observers. Ten observers (seven naïve) took part in the IOT task, seven observers (five naive) in the disparity tasks. All observers gave their informed, written consent to participate, and had normal or corrected to normal vision, stereoacuity and hearing at the time of the experiment. All experiments were conducted in accordance with the relevant guidelines and regulations, the informed written consent of each observer and received prior approval from the Research Ethics Committee at the School of Optometry and Vision Sciences, University of Bradford. stimuli and Apparatus. Visual stimuli were presented on two gamma-corrected Eizo FG2421 LCD monitors with a refresh rate of 120 Hz and a resolution of 1920 × 1080. For the IOT task these were connected to a 2 × 2.8 GHz Quad-core Apple Mac Pro desktop computer running Mac OS 10.5.8, and for the disparity task they were connected to a 3 GHz E5-1660v3 8-Core HP Z440 desktop computer running Windows 8.1 Pro. All stimuli were generated using Matlab (version 7.7.0.471 or 8.4.0, Mathworks, USA) running the Psychtoolbox Extension 77 version 3.0.8, (www.psychtoolbox.org). The physical durations and simultaneity of all auditory and visual stimuli were verified using a dual-channel oscilloscope. Visual stimuli were viewed dichoptically through a two mirror stereoscope which allowed the left and right monitors to be monocularly viewed by left and right eyes respectively. A forehead and chin rest were used to maintain a viewing distance of 2.3 meters (one pixel subtend 0.4 arc minutes), and a head position allowing right and left eyes to remain centred in front of the stereoscope's right and left mirrors. Prior to each experimental session, observers viewed monocular nonious lines and adjusted the tip and tilt of each mirror until these were horizontally and vertically aligned. This procedure neutralised any individual oculomotor anomalies and thus aided stable fusion through the experiment. www.nature.com/scientificreports www.nature.com/scientificreports/ The auditory reference stimulus was a 500 Hz tone presented through Sennheiser HD 280 headphones. The specific visual stimuli pertaining to each task are detailed below. Inter-ocular transfer (Iot) task. Visual stimuli were isotropic, luminance-defined Gaussian blobs (mean luminance 74 cd/m 2 ) presented at fixation against a uniform grey background of 37 cd/m 2 , whose luminance (L) profile was defined as follows: where L mean is the mean luminance of the Gaussian and σ Stim is the standard deviation. Throughout the IOT experiment σ Stim (the size of the stimulus), was set to 1°. Fixation was maintained on a white 0.07° circular fixation marker presented at the centre of left and right screens. Disparity task. Visual stimuli were constructed from dynamic luminance noise stereograms viewed through the dual-mirror stereoscope. They consisted of two dichoptically presented left and right eye circular regions in which each pixel was randomly assigned a luminance value (either black or white), which was randomly updated every five frames (24 Hz) to create dynamic luminance noise. These images were identical except for a disk-shaped sub-region referred to as the 'target zone' (e.g. Fig. 3A). This sub-region had a diameter of 500 pixels (subtending 3.33° of visual angle when viewed at 2.3 m) within which one eye's noise patch could be shifted laterally relative to the opposite eye's corresponding region. The introduction/removal of this shift allowed the presentation of crossed (+ve) or uncrossed (−ve) disparity within the target zone. The surrounding dynamic noise annulus was always presented with zero disparity, as was the entire display during interstimulus intervals. To ensure that there were no monocular artefacts at either the onset of offset of the disparity-defined target, we (i) maintained constant density by 'wrapping' noise pixels shifted beyond the target region to the opposite side; (ii) avoided motion cues by synching changes in disparity to the noise regeneration cycle and (iii) monitored for dropped frames during stimulus presentation. The surround was bordered by a static checkered annulus which was presented to both eyes, and thus aided stable binocular fusion. Beyond this border, the luminance of the background was set to mid-grey, which was equal to the average luminance of the dynamic noise. Observers maintained fixation at the centre of the screen throughout. procedure. For all experiments, a block of trials began with an initial adaptation phase consisting of 100 serially presented visual stimuli. Within a block, the duration of these stimuli was fixed at either 166 ms or 666 ms. Interstimulus interval (ISI) was randomly jittered between 500-1000 ms. The adaptation phase was a followed by a further four 'top up' adapting stimuli and a subsequent test phase (e.g. Fig. 1A) consisting of a fixed (333 ms) duration auditory reference stimulus and a variable duration visual test stimulus. Observers then made a two alternative forced choice (2AFC) duration discrimination judgment as to "which was longer, the visual test or auditory reference stimulus?" Visual test stimuli varied logarithmical in seven steps which were appropriately spaced relative to individual observer's duration discrimination threshold. The order of test stimuli presentation was randomly interleaved within a method of constant stimuli. Observers responded via key press which triggered the next top-up and test cycle, until all test durations had been presented five times per block of trials. In the IOT task, adapting and test stimuli were presented monocularly. In the "same" condition, the adapting stimulus was presented to the same eye as the visual test stimulus, and in the "different" condition (e.g. Fig. 1A) adapting and test stimuli were presented to opposite eyes. The choice of test eye was randomly assigned to each observer at the start of the experiment so that half of the observers viewed the test stimuli with their right eye and the other half their left eye. Half of the observers also performed additional conditions to ensure that the choice of test eye did not affect the magnitude of the duration aftereffect. These five observers performed all four permutations of adapt/test eye. A paired samples t-test revealed that there was no significant difference in the size of the aftereffect generated in either of the "same" test eye conditions (i.e. adapt and test right eye versus adapt and test left eye, p = 0.13), or the "different" conditions (i.e. adapt right, test left versus adapt left, test right, p = 0.54) For the remaining analysis, data was combined across equivalent 'same' and 'different' conditions. In the retinal disparity experiments the procedure was identical to that described above with the following exceptions: all stimuli were presented binocularly and the 'same' vs 'different' manipulations pertained to situations where observers adapted and tested with stimuli defined by the same retinal disparity (adapt and test +6′ or adapt and test −24′ (Fig. 3A,C, respectively)) vs different retinal disparity (adapt +24′, test −24′ (Fig. 4A)). Each observer completed multiple blocks for each of the two adapting durations, and for each of the 'same' vs 'different' adapting conditions, giving grand totals of 380 (IOT task) 200 (adapt and test +6′ disparity) and 210 (adapt and test −24′, adapt +24′, test −24′, disparity) repetitions per test duration, per condition. For all experiments, the proportion of 'test longer than reference' responses were plotted against the physical visual test durations for each adapting duration (166 ms and 666 ms), and each experimental condition. The psychometric functions were then fitted with a logistic function of the form: where μ is the Point of Subjective Equality (PSE): the test duration that is perceptually equivalent to the 333 ms auditory reference duration, and θ is an estimate of the discrimination threshold. For each condition, PSE values were extracted and an estimate of duration aftereffect magnitude was obtained by subtracting the PSE for the Primary tumor side is associated with prognosis of colorectal cancer patients with brain metastases Background Brain metastases (BM) are a rare complication in colorectal cancer (CRC) patients and associated with an unfavorable survival prognosis. Primary tumor side (PTS) was shown to act as a prognostic and predictive biomarker in several trials including metastatic CRC (mCRC) patients. Here, we aim to investigate whether PTS is also associated with the outcome of CRC patients with BM. Methods Patients treated for CRC BM between 1988 and 2017 at an academic care center were included. Right-sided CRC was defined as located in the appendix, cecum and ascending colon and left-sided CRC was defined as located in the descending colon, sigma and rectum. Results Two hundred and eighty-one CRC BM patients were available for this analysis with 239/281 patients (85.1%) presenting with a left-sided and 42/281 patients (14.9%) with a right-sided primary CRC. BM-free survival (BMFS) was significantly longer in left-sided compared with right-sided CRC patients (33 versus 20 months, P = 0.009). Overall survival from CRC diagnosis as well as from diagnosis of BM was significantly longer in patients with a left-sided primary (42 versus 25 months, P = 0.002 and 5 versus 4 months, P = 0.005, respectively). In a multivariate analysis including graded prognostic assessment, PTS remained significantly associated with prognosis after BM (hazard ratio 0.65; 95% confidence interval: 0.46-0.92 months, P = 0.0016). Conclusions PTS was associated with survival times after the rare event of BM development in CRC patients. Therefore, its prognostic value remains significant even thereafter. INTRODUCTION Recent data strongly support the biological heterogeneity of colorectal cancer (CRC), arguing that CRCs are actually several different diseases originating at the same location. In addition to molecular biomarkers such as KRAS, NRAS, BRAF, mismatch repair (MMR) deficiency and HER2, primary tumor side (PTS) of CRC has been recently described to act as a prognostic and predictive surrogate parameter in metastatic CRC (mCRC) patients. One-third of colorectal tumors are right-sided and originate from the embryonic midgut, whereas two-thirds are left-sided and derive from the embryonic hindgut. 1,2 Right-sided tumors are associated with a generally worse prognosis compared with left-sided colorectal tumors, as reflected by a higher incidence of mucinous, undifferentiated and signet-ring cell tumors and a usually more advanced stage of disease at initial diagnosis. [3][4][5][6] Significant underlying molecular differences could be identified, since right-sided tumors are highly immunogenic characterized by higher rates of MMR deficiency as well as BRAF mutations and exhibit a higher incidence of activated RAS and PIK3CA mutations. 7,8 Moreover, the microbial richness was shown to increase from the proximal to the distal colon. 9 In line with these differences in biological behavior, PTS was only recently incorporated in treatment guidelines as a predictive surrogate parameter for the selection of targeted therapies in the metastatic setting. 10 As observed

in several retrospective analyses of phase II and III randomized trials, overall survival (OS) benefit with antiepidermal growth factor receptor (EGFR) antibodies such as cetuximab and panitumumab was only evident in patients with left-sided RAS wild-type mCRC, whereas patients with right-sided RAS wild-type mCRC may rather benefit from anti-vascular endothelial growth factor receptor (VEGFR) antibodies such as bevacizumab. [11][12][13][14][15][16] So far, this impact might be especially relevant in metastatic disease, since studies of CRC patients with early stages suggested no significant outcome differences with regards to PTS. 17 Only little is known about differences in metastatic behavior between patients with left-sided and right-sided CRC. Whereas liver and lung metastases are more often observed in left-sided CRC patients, peritoneal metastases may be more common in right-sided CRC. However, the incidence of brain metastases (BM) seems to be comparable, although evidence in this distinct patient population remains scarce due to the rare occurrence of BM in CRC patients. 17 Only 6% of BM patients present with gastrointestinal primaries most frequently located in the rectum and esophagus. 18 Small series so far suggested that BM from CRC are associated with a particularly poor prognosis between 3 and 11 months. Performance status was thereby shown to be significantly worse compared with other entities of primary tumors at BM diagnosis. 19,20 Within this study, we aim to investigate the influence of PTS on the clinical course and prognosis in a uniquely large cohort of CRC BM patients. Patients Overall, 323 patients treated between 1988 and 2017 for CRC BM at the Medical University of Vienna were identified from the Vienna Brain Metastasis Registry. Seven patients had to be excluded due to incomplete information regarding the clinical course of disease, 10 patients due to incomplete information regarding PTS and 12 patients due to diagnosis of a second primary tumor. Furthermore, 13 patients had to be excluded due to non-exact localization of the primary tumor in the transverse colon and the splenic flexure, respectively, to avoid a potential classification bias in terms of sidedness. Therefore, 281 patients were available for this retrospective analysis (Supplementary Figure S1, available at https://doi.org/10. 1016/j.esmoop.2021.100168). If leptomeningeal carcinomatosis (LC) was present concomitantly to diagnosis of parenchymal BM, patients were also eligible for inclusion. Information relating to patient demographics, case history and survival were collected by retrospective chart review. This study was conducted in accordance with the Declaration of Helsinki and approval by the institutional review board (IRB) was obtained (ethics committee of the Medical University of Vienna, 1167/2019). All authors had access to the study data and reviewed and approved the final manuscript. All patients were managed by a dedicated team of CRC BM specialists. Treatment decisions were taken in an interdisciplinary tumor conference. Treatment was carried out according to best clinical evidence and according to current standard of care. Localization of primary tumor and classification of sidedness Information about PTS was retrieved according to surgery protocols and histology reports. Patients with primaries in the transverse colon were excluded to avoid a potential classification bias in terms of tumor side allocation. Sidedness of the primary tumor was categorized according to recent international standards 11 : tumors of the appendix, cecum and ascending colon were categorized as right-sided and tumors of the descending colon, sigma and rectum as left-sided tumors ( Figure 1). Statistical analysis For comparisons patients were grouped in two groups based on the PTS: left-sided and right-sided CRC. OS was defined as the interval from first diagnosis of CRC, diagnosis of mCRC and diagnosis of BM, respectively, until death or last date of follow-up and estimated with the KaplaneMeier product limit method. To test for differences between two parameters, the chi-square test was used for binary variables and the ManneWhitney U test for differences in mean ranks between two variables. To test for differences between OS curves, the log-rank test was used. BM-free survival (BMFS) was defined as the interval from diagnosis of CRC until diagnosis of BM. Two-tailed P values <0.05 were considered to indicate statistical significance. The association of PTS with OS from diagnosis of CRC BM was the main point of interest of the present study. The graded prognostic assessment (GPA) including Karnofsky performance status (KPS) (<70, 70-89, 90-100), age (<50, 50-59, 60 years), extracranial metastases (present, absent) and number of BM (1, 2-3, >3) and the recently updated GPA for gastrointestinal cancers (GI-GPA), respectively, including KPS (<80, 80, 90-100), age (<60, 60 years), extracranial metastases (present, absent) and number of BM (1, 2-3, >3) are the best established prognosticators of outcome in CRC BM patients. 19 Therefore, we predefined a priori the inclusion of the PTS together with either the GPA or the GI-GPA into the multivariate model, depending on their significance in the univariate analysis. A multivariate analysis was carried out using the Cox regression model. Due to the exploratory and hypothesisgenerating design of the present study, no adjustment for multiple testing was applied and no formal sample-size calculation was conducted. 21 All statistics were calculated using statistical package for the social sciences (SPSS®) 26.0 software (SPSS Inc., Chicago, IL). Patient characteristics A total of 281 patients with CRC BM were available for this analysis. Median age at initial diagnosis of CRC was 61 years Association of PTS with clinical characteristics of CRC BM patients Baseline characteristics were well balanced between leftand right-sided tumors. Median age at diagnosis of BM and median KPS were not statistically different between rightsided and left-sided CRC patients (64 versus 69 years and 70%; P > 0.05; ManneWhitney U test). At diagnosis of BM, 55.2% with left-sided CRC and 56.8% patients with rightsided CRC presented with progressive extracranial disease (P > 0.05; chi-square test). Median number of BM at initial diagnosis of BM was one in left-sided as well as right-sided CRC patients (P > 0.05; ManneWhitney U test). Incidence of concomitant LC diagnosis to solid BM was not significantly different between left-and right-sided CRC patients (2.9% versus 0%, P > 0.05; chi-square test) as well as incidence of intracranial recurrence after initial BM therapy (26.5% versus 42.3%, P > 0.05; chi-square test). Detailed patient characteristics according to PTS are listed in Table 1. Association of PTS with survival times in CRC BM patients Median OS from diagnosis of CRC BM according to the time period of initial CRC diagnosis was not significantly different between patients diagnosed before the year 2000, between 2000 and 2010 and after the year 2010 (5 versus 4 versus 4 months, P > 0.05; log-rank test). Median OS from BM diagnosis was numerically, but not significantly different between GI-GPA classes (4 months with class 1 versus 4 months with class 2 versus 5 months with class 3 versus 5 months with class 4, P > 0.05; log-rank test) ( Figure 3A). Therefore, we also carried out survival analysis according to GPA classes, which was significantly associated with OS from diagnosis of BM (13 months with class 1 versus 13 months with class 2 versus 5 months with class 3 versus 4 months with class 4, P ¼ 0.004; log-rank test) ( Figure 3B). Patients with left-sided tumors had a significantly longer BMFS compared with patients with right-sided tumors (33 versus 20 months, P ¼ 0.009; log-rank test) ( Figure 3C). Median OS from first diagnosis of CRC was significantly longer in patients with left-sided tumors compared with right-sided tumors (42 versus 25 months, P ¼ 0.002, logrank test) ( Figure 3D). Median OS from diagnosis of mCRC was not statistically different between left-and right-sided tumors (23 versus 21 months, P > 0.05; log-rank test) ( Figure 3E). Median OS from diagnosis of BM was significantly longer in patients with left-sided tumors compared with right-sided tumors (5 versus 4 months, P ¼ 0.005, logrank test) ( Figure 3F). To evaluate the independent association of sidedness of the primary tumor on prognosis of CRC BM patients, we (Table 2). DISCUSSION Sidedness of the primary tumor was significantly associated with the clinical course in the present series of CRC BM patients. Time from diagnosis of the primary tumor to BM development, as well as survival from BM diagnosis was significantly shorter in patients with a right-sided primary tumor than in patients with a left-sided primary tumor. The present observation suggests that the biological metastatic drivers differing between right-and left-sided CRC might even impact the disease course in the rare event of BM. Therefore, our data further support the theory that CRCs comprise several molecular diverse diseases with differing metastatic behavior originating in the same organ. In the present cohort of CRC BM patients, left-sided primary was with 85.1% more frequently observed than right-sided primary tumor. This 4 : 1 side distribution is well in line with the one previously observed for mCRC without BM. 12,22,23 Therefore, as previously postulated in rather small series, PTS per se might not influence the development of BM. 24 Survival prognosis of CRC patients in our study was more than 1.5-fold better with a left-sided compared with a right-sided primary. An underlying reason, therefore, might display profound differences in molecular and biological characteristics as well as resulting targeted treatment approaches. Differences in embryological origins lead to distinct gene expression patterns with different methylation and mutation profiles, as well as distinctions in the microbiome of patients. 6,9,25 Recent next generation sequencing studies revealed higher rates of KRAS, NRAS, BRAF, PIK3CA, CTNNBI and SMAD mutations as well as CpG island methylator phenotype (CIMP) and MMR defects in right-sided CRC, whereas left-sided tumors presented more TP53 mutations. 26 Based on that, the consensus molecular subtypes (CMS) of CRC originally described in 2015 by Guinney et al. 27 and defined by gene-expression arrays had been analyzed with regards to prognostic relevance of PTS. Here, CMS2 indicating a rather favorable prognosis was more common in left-sided and CMS1 indicating a rather poor prognosis in right-sided tumors. 27,28 These differences in molecular profiles might impact the brain-specific metastatic behavior, resulting in an easier colonization of right-sided CRC cells in the brain parenchyma. Indeed, RAS mutant CRC was previously shown to present with a significantly higher cumulative incidence of lung, bone and brain metastasis. 29 Further, PIK3CA mutations were postulated to increase the brain metastatic behavior in breast cancer. 30 Preclinical and clinical data further support that PIK3CA inhibitors have clinical efficacy in BM. [31][32][33] Further, molecular research focusing specifically on molecular drivers could reveal targets for targeted treatment approaches. Here, we were able to report a unique large cohort of CRC BM patients to gain further insight into the correlation of PTS and prognosis in the specific setting of BM. Clearly, our study comprises some limitations, which have to be considered. First, due to the retrospective nature, our results need to be interpreted with caution. Second, unfortunately only limited information on the molecular profile of tumors was available, which would have been indeed of great interest to further investigate We applied the standard prognostic assessment scores in our population including GPA as well as the disease-specific form of the GI-GPA. Only the GPA and not the later updated GI-GPA remained significantly associated with OS after BM diagnosis. A potential reason could be that only half of the patients of the validation study for the GI-GPA had a primary within the colon, while the rest presented mainly upper and other gastrointestinal primaries. CRC might therefore display a distinct subgroup of gastrointestinal malignancies. Moreover, the GPA distinguishes more precisely with regards to age and KPS compared with the GI-GPA, which may have allowed for a better discrimination of our patient population. Conclusion To our best knowledge, our study represents the largest single-center analysis of CRC BM patients to date. We could determine a clear association between PTS and BMFS as well as OS, since patients with right-sided CRC develop BM significantly earlier and exhibit a significantly impaired prognosis compared with left-sided CR CBM patients. Further investigation of the underlying molecular drivers is warranted to identify potential future treatment targets. FUNDING None declared. The Effect of Non-immersive Virtual Reality Exergames Versus Band Stretching on Cardiovascular and Cerebral Hemodynamic Response: A Functional Near-Infrared Spectroscopy Study Background Exercise is one of the effective ways to improve cognition. Different forms of exercises, such as aerobic exercise, resistance exercise, and coordination exercise, have different effects on the improvement of cognitive impairment. In recent years,

exergames based on Non-Immersive Virtual Reality (NIVR-Exergames) have been widely used in entertainment and have gradually been applied to clinical rehabilitation. However, the mechanism of NIVR-Exergames on improving motor cognition has not been clarified. Therefore, the aim of this study is to find whether NIVR-Exergames result in a better neural response mechanism to improve the area of the cerebral cortex related to motor cognition under functional near-infrared spectroscopy (fNIRS) dynamic monitoring in comparison with resistance exercise (resistance band stretching). Methods A cross-over study design was adopted in this study, and 15 healthy young subjects (18–24 years old) were randomly divided into group A (n = 8) and group B (n = 7) according to a computerized digital table method. Task 1 was an NIVR-Exergame task, and Task 2 was resistance band stretching. Group A first performed Task 1, rested for 30 min (i.e., a washout period), and then performed Task 2. Group B had the reverse order. The fNIRS test was synchronized in real time during exercise tasks, and heart rate measurements, blood pressure measurements, and 2-back task synchronization fNIRS tests were performed at baseline, Post-task 1, and Post-task 2. The primary outcomes were beta values from the general linear model (GLM) in different regions of interest (ROIs), and the secondary outcomes were heart rate, blood pressure, reaction time of 2-back, and accuracy rate of 2-back. Results The activation differences of Task 1 and Task 2 in the right premotor cortex (PMC) (P = 0.025) and the left PMC (P = 0.011) were statistically significant. There were statistically significant differences in the activation of the right supplementary motor area (SMA) (P = 0.007), left dorsolateral prefrontal cortex (DLPFC) (P = 0.031), left and right PMC (P = 0.005; P = 0.002) between baseline and Post-task 1. The differences in systolic pressure (SBP) between the two groups at three time points among women were statistically significant (P1 = 0.009, P2 < 0.001, P3 = 0.044). Conclusion In this study, we found that NIVR-Exergames combined with motor and challenging cognitive tasks can promote the activation of SMA, PMC and DLPFC in healthy young people compared with resistance exercise alone, providing compelling preliminary evidence of the power for the rehabilitation of motor and cognitive function in patients with central nervous system diseases. INTRODUCTION The incidence of cognitive dysfunction gradually increases with age (Feigin et al., 2009;Reitz et al., 2011Reitz et al., , 2020. The prevalence of mild cognitive dysfunction for those over 65 years old is 10-20% (Langa and Levine, 2014), and it significantly decreases the ability to live independently, which increases burdens on families and society. Therefore, it is particularly important to find how to effectively prevent dysfunction. In a review of cognitive-enhancing technologies, certain drugs (nootropic, methylphenidate, and modafinil) can enhance cognitive abilities, including attention, concentration, and vigilance. But it can also be dangerous, causing headaches, diarrhea, insomnia, fatigue, tremors and nausea. Among nonpharmaceutical interventions, both exercise and music are one of the effective methods to improve cognition (Al-Shargie et al., 2019;Wu et al., 2019). Different types of music have different effects on cognition and perception. For example, fast-paced music can improve a person's concentration and cognitive function, while maintaining vigilance. Moreover, many studies have confirmed that exercise can promote neuroplasticity of the brain, thereby improving cognitive function. The potential related mechanisms include the improvement of oxygen consumption and increased cerebral blood flow, which promote brain-cell regeneration in brain regions related to cognitive function (Han et al., 2017). Exercise could also up-regulate brainderived neurotrophic factor (BDNF) and downstream redox signaling pathways, enhance the repair process of antioxidant and oxidative damage, and increase brain metabolism and neuronal activation (Quan et al., 2020). In addition, exercise could also regulate the release of neurotransmitters (Cassilhas et al., 2016), affect the neurotrophic effect of BDNF (Vaynman and Gomez-Pinilla, 2005), and promote the generation of new neurons (van Praag et al., 2005), thereby promoting neural plasticity. Different forms of exercise, such as aerobic exercise, resistance exercise, and coordination exercise, have different effects on cognitive dysfunction (Hötting and Röder, 2013). There are relevant behavioral data suggesting that coordinated exercise has greater benefits over pure aerobic exercise in different exercise types. Voelcker-Rehage et al. (2011) conducted a 12month longitudinal study and showed that cardiorespiratory training such as aerobic exercise was associated with increased sensorimotor network activation and had a positive impact on control tasks (Voelcker-Rehage et al., 2011). Furthermore, coordination exercise was associated with increased visualspatial network activation and had a positive effect on cognitive processing speed, but the specific improvement mechanism remained unclear. Meanwhile, the neural regulation mechanism of resistance exercise in related brain regions remains to be studied. Therefore, it is important to clarify the influence of different forms of exercise on brain function, which could be helpful to guide clinical cognitive training and improve treatment efficiency. Related research indicates that traditional action video games (such as "Jumping Square") usually involve visual spatial perception, working memory, executive function, and information processing speed. Repeated game training could shorten reaction time without reducing accuracy or continuing attention (Dye et al., 2009;Bediou et al., 2018). A meta-analysis by Bediou et al. (2018) demonstrated that traditional action video games could significantly enhance top-down attention and spatial perception. A review by Al-Shargie et al. (2019) has found particularly positive effects of Action Video Game on perception, sustained attention, memory, cognitive skills, and decision-making. Such skills include sustained attention and preservation of multitasking tendencies. In addition, video games are often accompanied by rhythmic music, and may have superimposed effects on cognitive improvement. Nevertheless, there are still some drawbacks in traditional action video games. For example, most traditional video games, such as action, adventure, imitation, and strategy games, only involve using the fingers to control the game interface (Straker et al., 2014). Thus, they mainly exercise the abilities of reaction and handeye coordination to improve cognitive function, but they rarely involve large-scale activities of the whole body. Overuse of a mouse or keyboard to play games could lead to health problems such as tenosynovitis and benign joint hypermobility syndrome (Queiroz et al., 2018). In addition, prolonged sitting may lead to adverse health consequences such as musculoskeletal pain or syndromes (de Rezende et al., 2014;Straker et al., 2014;Queiroz et al., 2018). Therefore, a healthier, efficient, and feasible way of games to improve cognitive function is needed. In recent years, with the continuous development of video games, exergames based on Non-Immersive Virtual Reality (NIVR-Exergames) have been widely used in entertainment and have gradually been applied to clinical rehabilitation. Clinical studies have shown that sports games may be beneficial to the improvement of motor function and cognitive function (Swinnen et al., 2020). One example is the game "Just Dance" on the Nintendo Switch. The player mainly uses a visual display screen and a controller to complete dance movements through body movements that follow the rhythm of music (Adcock et al., 2022), and the game provides real-time feedback. In clinical rehabilitation, the advantages of such games are embodied in various aspects. Patients perform active and systematic repetitive training through specific game tasks involving different cognitive abilities, which realize the reorganization of neural circuits and improve the treatment effect (Gamito et al., 2017). Furthermore, through human machine interaction technology, dynamic feedback and incentive stimulation can be provided, and the actions of patients can be corrected immediately, thereby improving the enthusiasm for training (Gamito et al., 2017). Another benefit is that extensive visual and auditory stimulation as an external factor in training (Pompeu et al., 2012) provides a more attractive, novel, and rich training environment than conventional rehabilitation therapy (Mirelman et al., 2016), thereby increasing patient compliance during long-term rehabilitation (Pompeu et al., 2012). Non-Immersive Virtual Reality-Exergames combine the cognitive training provided by traditional video games with a form of exercise training (Pichierri et al., 2011). However, further research is needed to determine whether they can integrate the advantages of direct cognitive improvement (game elements improving cognition through cognitive training) and indirect cognitive improvement (exercise elements improving cognition). The results could lead to the realization of a double superposition effect of NIVR-Exergames on cognitive function improvement. Functional near-infrared spectroscopy (fNIRS) is an optical and non-invasive neuroimaging technology that has been widely used in the assessment of cognitive-related brain functions. Examples include a study on the effect of caffeine on cognitive function using the Stroop task (Yuan et al., 2020) and a study on the change of fall-risk-related brain functional connectivity when using smart phones while walking (Takeuchi et al., 2016). fNIRS is sensitive to mental task load and practice level. An article provide evidence of the fNIRS deployment in the field for its ability to monitor hemodynamic changes that are associated with relative cognitive workload changes of operators. fNIRS allows realtime assessment and has the advantages of portability, economy, and non-invasiveness compared with other brain functional imaging technologies, such as electroencephalography (EEG), magnetoencephalography (MEG), and functional magnetic resonance imaging (fMRI) (McKendrick et al., 2016;Curtin et al., 2019;Pinti et al., 2020). These advantages are especially apparent for subjects with high exercise tolerance (Liu et al., 2016). Furthermore, the method is suitable for subjects performing physical exercise tasks. Studies have shown that simple exercise and NIVR-Exergames can improve cognitive function (Sakudo, 2016). However, most of the related studies have focused on the function and neuroimaging effects before and after treatment (Saposnik et al., 2016;Bonilauri et al., 2020), and there has been a lack of real-time dynamic research on neural regulation mechanisms.Therefore, the aim of this study is to explore whether NIVR-Exergame tasks lead to a better neural response mechanism to improve cerebral cortex areas related to motor cognition. This was accomplished based on fNIRS for real-time monitoring of brain activation while playing NIVR-Exergames (Nintendo's "Just Dance"). The results were compared to those obtained with stretching using a resistance band. The findings could guide clinical treatment and home rehabilitation for cognitive dysfunction in the future. Subjects This study recruited 15 healthy young subjects (8 males and 7 females) from the Fifth Affiliated Hospital of Guangzhou Medical University. The inclusion criteria were the following: (1) healthy young people between 18 and 24 years old; (2) right handedness according to the Edinburgh Handedness Inventory;voluntarily signing of an informed consent form. The exclusion criteria were: (1) a history of diseases with motor dysfunction, such as fracture, limb pain, etc.; (2) cardiopulmonary diseases resulting in inability to complete the exercise; (3) sudden diseases occurring during the task affecting the experimental results; (4) history of brain-related diseases with cognitive dysfunction or medications that improve cognitive function being taken; (5) severe mental disorders; and (6) no relevant professional training. Statements of Ethical Consideration This study was performed in accordance with the principles of the Declaration of Helsinki and was approved by the Ethics Committee of the Fifth Affiliated Hospital of Guangzhou Medical University (Ethics No.: GYWY-L2021 -70). The intervention program has also been registered with the International Clinical Laboratory Registry (registration number: ChiCTR2100054358). All subjects signed an informed consent form before participating. Study Design A crossover study design was adopted in this study (Figure 1), and the subjects were randomly divided into group A (n = 8) and group B (n = 7) according to the computerized digital table method with a ratio of 1:1. The experimental scheme consists of two tasks and three measurement time points (baseline, Posttask 1, and Post-task 2). An NIVR-Exergame, "Just Dance" by Nintendo, was used in Task 1, and resistance band stretching was used in Task 2. Group A first performed Task 1, rested for 30 min (i.e., a washout period), and then did Task 2. Group B had the opposite order. Heart rate measurements, blood pressure measurements, and the 2-back test were performed at baseline, Post-task 1, and Post-task 2. The fNIRS measurement was carried out simultaneously during the two exercise tasks and 2-back task. Resistance Band Stretching Task The subjects used a conventional red resistance band (Thera-Band, United States) to perform Task 2. The subjects were FIGURE 1 | Cross-over design of the study. Task 1: NIVR-Exergame task; Task 2: resistance band stretching task; Post-task 1: after the NIVR-Exergame task; Post-task 2: after the resistance band stretching task. min, minute. required to complete the stretching of the upper limbs as much as possible within 3 min and to keep the position of the feet unchanged. NIVR-Exergame Task Participants

used the game "Just Dance" (Soy Yo-Snake Version) based on the Nintendo Switch system (Nintendo Co., Ltd., Kyoto, Japan) for Task 1. Subjects were asked to wear a controller on their right hand and to play the "snake" role. The NIVR-Exergame task lasted for 3 min. During the task, only the upper limbs of subjects were required to complete the action with the requirements of the game and try to keep the position of the feet unchanged (this was done to take into account the stretching task of the upper limb with the resistance band). Clinical Assessment Heart rate and blood pressure were measured at baseline, Post-task 1, and Post-task 2 using an upper-arm electronic sphygmomanometer (Model: U10L, OMRON Healthcare Co., Ltd., China), which met the standards of the American Medical Device Promotion Association (AAMI). Heart rate was recorded in beats per minute, and blood pressure was recorded in mmHg. Blood pressure and heart rate were recorded immediately after the exercise stopped. Subjects were asked to keep the body straight in a sitting position, and the center of the cuff was kept at the same height as the heart. Blood pressure and heart rate were measured three times in total, and all measurements were collected by a well-trained researcher. Cognitive Assessment This study used the N-back task (2-back) based on E-prime (Psychological Software Tools Company) to evaluate the cognitive function (working memory) of subjects at three time points (baseline, Post-task 1, and Post-task 2) (Figure 2). Before the 2-back task, the subjects were asked to practice for 2 min first to eliminate the influence of the familiarity effect on the subsequent evaluation. The 2-back task paradigm continuously presents a series of random letters (A-H) in the center of the display on a laptop running Windows. Subjects have to remember the letters that appear and respond as quickly and accurately as possible by clicking a response button corresponding to the computer keyboard. In the 2-back task, the subjects need to compare the letters that they see at present with the letters at an interval ahead of them. The subjects are asked to press "F" (if the two letters are the same), or "J" (otherwise). This task consists of three continuous 2-back blocks. A separate block consists of 60 s of trial and 60 s of rest. The trial presentation time is 1000 ms, and the reaction time is 2000 ms, which makes a total time of 3000 ms. The subjects received no feedback on their scores during the test. During the evaluation task, fNIRS was used to synchronously collect hemodynamic signals in the cortex. The main measurement indexes are the number of correct answers, the number of wrong answers, the number of missing answers, the accuracy rate (AR) {(the number of correct answers/total answers) × 100%}, and the average reaction time (RT). Functional Near-Infrared Spectroscopy Hemodynamic changes were monitored in real time during the exercise tasks and 2-back task using a two wavelengths (730 and 850 nm) CW-fNIRS equipment (Nirsmart, Danyang Huichuang Medical Equipment Co., Ltd., China). 12 emitters and 10 detectors constituting 24 channels were placed in accordance with the international 10-20 system (Figure 3), using a fixed 3cm source-detector spacing. The cerebral cortex hemodynamic response was collected and recorded at a sampling rate of 10 Hz. Regions of Interest The statistical parameters of the Montreal Neurology Institute (MNI) were mapped to NIRS_SPM software to calculate the coordinates of each signal source and detector position. The software provides the Brodmann region corresponding to each fNIRS channel. Based on the Brodmann partition, six regions of interest (ROIs) were identified in the bilateral dorsolateral prefrontal cortex (DLPFC), supplementary motor areas (SMA), and premotor cortex (PMC) (Figure 3). The channel distribution of the selected ROI is such that the left DLPFC corresponds to channels 5 and 6, and the right DLPFC corresponds to channels 1 and 2. The left and right SMAs correspond to channels 17 and channel 12, respectively. The left PMC corresponds to channels 21, 22, and 23, and the right PMC corresponds to channels 14, 15, and 16, respectively. The changes in oxyhemoglobin concentrations in the ROIs were collected during the 2 motor tasks and at the 3 time points, and the beta values were calculated using the general linear model (GLM) to quantify the correlation between brain regions and tasks. Functional Near-Infrared Spectroscopy Data Processing and Analysis All the original optical intensity time series were derived from NIRSMART. Firstly, the channels with low-quality signals were removed to reduce the error according to the received optical signal quality of each channel. Secondly, the modified Beer-Lambert law (Herold et al., 2017) is used to convert the optical intensity into relative change of oxyhemoglobin (HbO2) and deoxyhemoglobin (HbR) concentrations. Then, the NIRS-SPM toolbox 1 (Dye et al., 2009) was used to process the HbO2 signal of the target channel, since HbO2 generally has better signal-noise ratio than the HbR signal (Strangman et al., 2002). A statistical parameter mapping method based on the GLM is used to fit the fNIRS data using NIRS-SPM. For each separate HbO2 time series during three different task (NIVR-Exergame Task, Resistance Band Stretching Task, 2-back Task), 1 https://www.nitrc.org/projects/nirs_spm/ the model parameters for GLM could be described as y ∼ 1 + x i in Wilkinson-notation, where y represents the hemodynamic response, x i denotes the corresponding regressor for each task (de Winkel et al., 2017). Because several physiological processes (e.g., respiration, blood-pressure changes, heartbeat, etc.) are known to produce structured "noise" within the data (i.e., autocorrelation), a precoloring treatment is performed, through which such temporal correlations are "swamped" by an imposed 4s Gaussian temporal correlation structure and are thus effectively reduced. Wavelet minimum description length detrending algorithm is applied to the both model and data to count for the global trends. Then model estimation was done and β value was used as the indicator of brain activation. The largest beta outcome for all channels within a single ROI is selected and entered into the group-level analyses reported below (Ashlesh et al., 2020;Baker et al., 2020). Finally, the activation beta value (the weight coefficient in the GLM) in each task is obtained to quantify the correlation between the brain region and the task. Outcomes The primary outcome indicators are the GLM beta values in different ROIs. The secondary outcome indicators are the heart rate, blood pressure, reaction time of 2-back, and accuracy rate of 2-back. Sample Size The sample size for the comparison of the two exercise interventions (using a paired t-test) was estimated using G * Power 3.0 (Faul et al., 2007). The sample size was calculated using data regarding cortical activation of the premotor cortex from five subjects in a preliminary experiment. The analysis indicated that the total sample size should be 12 based on a level of significance of 0.05 and statistical power of 0.8. Statistical Analysis SPSS 25.0 was used for statistical analysis of the data. When processing the clinical evaluation data, the influence of gender on the results was tested by comparison between male and female groups using an independent-samples t-test. Multicomparison correction procedure was ued to avoid the type I error. An analysis of variance (ANOVA) was used to compare the effects of three different time points on the results in the male and female groups. When processing cognitive assessment data, a paired t-test was used to compare the effects of the two different tasks on the results. A repeated measure ANOVA was used to analyze the effect of multiple factors. P < 0.05 indicated that the difference was statistically significant. RESULTS The study recruited 15 healthy subjects (8 males and 7 females) between 18 and 24 years old. Baseline characteristics show the data of male and female group, including age, height, weight, body mass index (BMI), resting blood pressure, and heart rate (HR) ( Table 1). Compared with the female group, the male group had greater height (171.938 ± 5.710 cm in the male group and 160.143 ± 4.811 cm in the female group, P = 0.001), weight (60.913 ± 6.132 kg in the male group and 46.686 ± 3.231 kg in the female group, P < 0.001), BMI (20.633 ± 2.158 in the male group and 18.210 ± 1.069 in the female group, P = 0.019), baseline systolic pressure (SBP) (124.625 ± 9.039 mmHg in the male group and 111.571 ± 5.884 mmHg in the female group, P = 0.006). These differences between groups were statistically significant. The results of the ANOVA with a 2 × 3 factorial design shows the effects of gender and time on cardiovascular response and whether there is interaction between them ( Table 2). The Significance level set at P < 0.05. Significant correlations are marked with *P < 0.05, **P < 0.01, and ***P < 0.001. BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; cm, centimeter; kg, kilogram; mmHg, millimeters of mercury; min, minute; P, P value. Significance level set at P < 0.05. Significant correlations are marked with *P < 0.05, **P < 0.01, and ***P < 0.001. HR, heart rate; SBP, systolic blood pressure; DBP, diastolic blood pressure; min, minute; mmHg, millimeters of mercury; P, P value. Significant correlations are marked with ***P < 0.001. results show that the differences in SBP measured for different genders and at different time points were statistically significant (gender factor: F = 31.137, P < 0.001; time factor: F = 11.745, P < 0.001). However, the interaction between gender and time was not statistically significant (F = 3.665, P = 0.581). The activation of the ROIs during Task 1 and Task 2 is also shown according to the GLM (Table 3). In Task 1, the beta values of the right PMC and left PMC were higher than those in Task 2 (right PMC: Task 1: 0.229 ± 0.186 vs. Task 2: 0.096 ± 0.156, P = 0.025; left PMC: Task 1: 0.206 ± 0.188 vs. Task 2: 0.043 ± 0.144, P = 0.011) (Figure 5). The activation of the right SMA, right PMC, left PMC, right DLPFC and left DLPFC regions of the brain was statistically significant when time was the main effect (F = 5.870, P = 0.007, F = 7.894, P = 0.002; F = 7.550, P = 0.002; F = 5.305, P = 0.021; F = 3.199, P = 0.056) ( Table 4). The results of post hoc tests for comparison of the GLM beta values in different ROIs in the left and right brain regions and at different time points are shown (Figure 6). Compared with baseline, Right SMA (P = 0.007), Right and Left PMC (P = 0.002; P = 0.005) and Left DLPFC (P = 0.031) were significantly activated at Posttask 1; Right SMA and Right DLPFC (P = 0.037; P = 0.013) were significantly activated at Post-task 2. Compared with Post-task 1, Right and Left PMC (P = 0.025; P = 0.011) were significantly activated at Post-task 2. DISCUSSION In recent years, NIVR-Exergames have effectively integrated the advantages of sports and traditional games. As a commercial mass-entertainment product, NIVR-Exergames have also been gradually applied in the field of clinical rehabilitation (Swinnen et al., 2020). However, there are few studies on NIVR-Exergames to improve motor cognition, and the relevant mechanisms are not yet clear. In this study, NIVR-Exergames and a resistanceband stretching exercise were compared, and fNIRS was used for the first time to monitor the activation of ROIs during and after exercise. Furthermore, the neural regulation mechanism of NIVR-Exergames in improving motor cognition was explored. The results show that the NIVR-Exergame task was more helpful for the immediate response of ROIs than the resistance-band stretching task and was not affected by gender. However, the SBP after two exercise tasks was affected by gender factors. Compared with the resistance-band stretching task, the NIVR-Exergame task significantly activated the bilateral PMC. Its mechanism may be related to the PMC's participation in motion control, motor planning, and learning (Kantak et al., 2012). The PMC is located in the frontal lobe and the front end of the anterior central gyrus adjacent to the posterior primary motor cortex and is a functional area related to movement in the frontal lobe of the brain (Nakajima et al., 2019;Fornia et al., 2020). Studies have shown that brain activities in motor planning and

various cognitive tasks are jointly located in the PMC (Hanakawa, 2011). This study used NIVR-Exergames combined with dance elements for training. Studies have shown that dance movement therapy (DMT) has obvious benefits in improving cardiovascular health (Li et al., 2015), musculoskeletal health (Matthews et al., 2006), dexterity, balance, and overall endurance. The training process involves the practice of sensory motor skills, which requires observation of specific movements and imitation, learning, and motor planning (Mura et al., 2018), which are closely related to PMC function. At the same time, when subjects participated in interactive virtual tasks and imitate the interactive process, the PMC performs predictive decoding of virtual tasks, action observation, and prediction, and then promotes action imitation (Sacheli et al., 2019). The resistance-band stretching task involves a relatively simple, repetitive, mechanized upper-limb movement that is less affected by rhythm perception, motor planning, and learning factors. This explains why the NIVR-Exergames significantly activated the bilateral PMC compared with the resistance-band stretching task in this study. Studies have shown that the execution and imagination of motion require activation of the connectivity of the brain region, and the common coupling nodes include DLPFC-PMC and PMC-SMA (Kim et al., 2018). We speculate that the activation of PMC in the NIVR-Exergame task may be the key connection point of brain activation during cognitive-motor tasks and that it plays an important role in improving motor and cognitive function. The results of this experiment showed that the NIVR-Exergame task and resistance-band stretching task could significantly activate the right SMA compared with the baseline. The potential mechanism may be related to the specialization in the right hemisphere region The SMA is located in the precentral gyrus of the frontal lobe and plays a key role in the execution of FIGURE 4 | Comparison of HR, SBP and DBP between male group and female group at 3 time points between groups and within groups. (A) Comparison of HR at Baseline, Post-task 1 and Post-task 2 between male and female group. (B) Comparison of HR at Baseline, Post-task 1 and Post-task 2 in male group and female group. (C) Comparison of SBP at Baseline, Post-task 1 and Post-task 2 between male and female group. (D) Comparison of SBP at Baseline, Post-task 1 and Post-task 2 in male group and female group. (E) Comparison of DBP at Baseline, Post-task 1 and Post-task 2 between male and female group. (F) Comparison of DBP at Baseline, Post-task 1 and Post-task 2 in male group and female group. Post-task 1: after the NIVR-Exergame task; Post-task 2: after the resistance band stretching task. mmHg, millimeters of mercury; min, minute; HR, heart rate; SBP, systolic blood pressure; DBP, diastolic blood pressure. autonomous motion. It is not only involved in motor execution, but also movement sequence organization (Gerloff et al., 1998). Studies have shown that performing dexterous exercises involves spatial coordination of multi-joint, proximal, and distal muscle activity. This requires complex movement sequences that can be acquired and refined through extensive exercises, thus allowing engagement in skilled activities such as typing, dancing, musical performance, and exercise (Strick et al., 2021). Therefore, improving the dexterity of movement is potentially associated with SMA activation. In addition, previous studies have found significant functional specialization in music perception, and the central areas of perception of pitch, melody, and harmony are located in the right hemisphere (Tramo, 2001;Tramo et al., 2002). The right hemisphere is the center of visualization, imagination, and conceptualization, with advantages in metaphorical thinking, Significance level set at P < 0.05. Significant correlations are marked with *P < 0.05, **P < 0.01, and ***P < 0.001. ROIs, regions of interest; SMA, supplementary motor cortex; PMC, premotor cortex; DLPFC, dorsolateral prefrontal cortex; Task 1, NIVR-Exergame task; Task 2, resistance band stretching task; GLM, general linear model; T, T value; P, P value. gameplay, and finding solutions. The NIVR-Exergame task in this study requires (Hoppe, 1988) subjects to complete a series of links such as music rhythm perception, game video recognition, motion imagination, and imitation. The resistanceband stretching task in this study also requires subjects to exercise in rhythm. The functions of these links are closely related to the specialization of music perception areas. Therefore, the significant activation of the right SMA may be dominated by music-related right-hemisphere specialization. The results show that NIVR-Exergames have more development potential than resistance exercise and could be preferable as a method of clinical and home rehabilitation training. In clinical rehabilitation, targeted training such as music and games is of great significance for patients with brain injury, especially right hemisphere injury. Compared with traditional exercise rehabilitation, it has obvious advantages in restoring and improving the dysfunction of patients. Moreover, the results of this experiment also show the activation of DLPFC by the NIVR-Exergame task. DLPFC is related to cognitive, emotional, and sensory processing and is a key node in multiple brain networks (Seminowicz and Moayedi, 2017). Studies have shown that the DLPFC, especially the tail, combines spatial information with object identity, behavior rules, reward mechanisms, and other information and plays a key role in the cognitive control of motor behavior (Hoshi, 2006). The cognitive function assessment results of the 2-back task synchronous fNIRS test showed that there was significant difference in the Left DLPFC activation between Post-task 1 and Post-task 2. The results suggest that NIVR-Exergames may play an important role in maintaining and improving cognitive function. However, the 2-back results show that there was no significant difference in RT and AR between Baseline and Post-task 1. Previous studies have found that the improvement of working memory function after a single aerobic exercise depends on the baseline working memory function but is irrelevant to exercise intensity (Yamazaki et al., 2018). That is, subjects with better performance in baseline assessment are less affected by exercise factors, while subjects with poor performance are greatly affected by exercise factors. Considering that the subjects in this study were healthy and young, their baseline performance was good, so the cognitive improvement effect after exercise was not obvious. In addition, a number of studies have shown that a single short-term exercise has no effect on working memory. When aerobic exercise is performed three times a week for 12 weeks, Significance level set at P < 0.05. Significant correlations are marked with *P < 0.05, **P < 0.01, and ***P < 0.001. ROIs, regions of interest; SMA, supplementary motor cortex; PMC, premotor cortex; DLPFC, dorsolateral prefrontal cortex; Post-task 1, after the NIVR-Exergame task; Post-task 2, after the resistance band stretching task; F, F value; P, P value. subjects show benefits in executive function, memory, and complex attention as early as week 6, and continuous benefits are shown in week 12 (Cabral et al., 2019). Combined with the cognitive function evaluation results of the 2-back task synchronous fNIRS test in this study, the 2-back AT and RT Post-task 1 showed no obvious improvement effect. This may have been because the 2-back task designed in this study was less difficult for healthy young people. At the same time, it may be due to the short time of the single exercise task as well. Moreover, it has been demonstrated in the relevant literature that elevated oxygenated hemoglobin levels of PFC (prefrontal cortex, PFC) in physical exercise and cognitive tasks may be associated with an increase in neurometabolic activation, which is one of the candidate mechanisms for cognitive function improvement (Bediz et al., 2016). In a follow-up study, we could extend the treatment cycle and improve the difficulty of cognitive assessment tasks to further examine the effect of two kinds of exercise tasks on improving motor cognition at the functional level. In terms of cardiovascular response, previous studies have shown that gender is one of the factors affecting blood pressure (Cheng et al., 2012). The baseline results of this experiment also showed that the mean resting systolic blood pressure of the male group was higher than that of the female group, which was consistent with previous results, although the systolic blood pressures of the two groups were healthy. This study also found that compared with the baseline, the systolic blood pressure of the female group Post-task 1 was significantly increased, while that Post-task 2 was significantly decreased. The potential mechanism of elevation may be related to the heart and nerve response. Immediately after exercise, the sympathetic nervous system is still in an excited state, and the blood vessels in the skin and internal organs contract, while the cardiac contractility is strengthened, resulting in an increase in systolic blood pressure (Teng et al., 2005). Studies have shown that a potential mechanism of blood pressure reduction after resistance training may be related to the decrease of peripheral vascular resistance (Lopes et al., 2021) or the improvement of autonomic nerve function (Oliveira-Dantas et al., 2020). At the same time, resistance exercise reduces the inflammatory factors in patients with hypertension, thereby reducing the inflammatory response to damage vascular endothelial cells and slowing the proliferation and migration of endothelial cells, thus lowering blood pressure. In clinical practice, resistance training can increase the effect of antihypertensive drugs on individual muscle strength and reduce resting blood pressure (Polito et al., 2021). Therefore, when guiding hypertensive patients, blood pressure should be closely monitored during exercise training, especially systolic blood pressure changes. In addition, resistance exercise may be an optimal option for exercise therapy for patients with hypertension. It is also worth noting that the decrease in blood pressure in seconds or minutes after resistance exercise can be attributed to sudden perfusion and transient pressure drop in previously occluded muscle blocks and may result in post-exercise hypotension (PEH) (MacDonald et al., 1999). It is suggested that there is a certain risk of resistance exercise, and blood pressure should be closely monitored during exercise training in FIGURE 6 | Comparison between the left and right brain of bate value in different regions of interest during Baseline, Post-task 1 and Post-task 2. (A) The left brain; (B) the right brain. * P < 0.05 and * * P < 0.01. SMA, supplementary motor cortex; PMC, premotor cortex; DLPFC, dorsolateral prefrontal cortex; Post-task 1: after the NIVR-Exergame task; Post-task 2, after the resistance band stretching task. healthy and low-blood-pressure groups. However, there was no significant change in blood pressure in male groups after exercise in this study, which might be related to the fact that the exercise load of this study was small for them and that they had strong exercise tolerance. Limitation This study was a cross-sectional study that only discussed the effect of single exercise. A follow-up longitudinal study could be done to examine a long-term intervention to explore the potential mechanism of exercise in improving cognition. Moreover, this study can not completely eliminate the interference factors to the signal, and the short channel separation can be used to further reduce the interference to the signal. In addition, this study examined healthy young people, so future studies could look at more different types of exercises among young people, the elderly, and even stroke and hypertension patients, which could shed more light on the neural regulation mechanisms and effects of exercise in improving cognition. CONCLUSION NIVR-Exergames use a combination of motor and challenging cognitive tasks based on virtual reality technology, and more and more evidence has shown its prospects for clinical application. In this study, synchronous fNIRS was used to monitor two groups of motor tasks in real time. The results showed that NIVR-Exergames might be conducive to the realization of the motor-cognitive dual function superposition effect, which could have significance for guiding future clinical research. DATA AVAILABILITY STATEMENT The data and trial protocol are available from the corresponding author upon reasonable request (QL; [email protected]). ETHICS STATEMENT The studies involving human participants were reviewed and approved by the Ethics Committee of the Fifth Affiliated Hospital of Guangzhou Medical University. The patients/participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article. AUTHOR CONTRIBUTIONS QL, AY, and HF designed the study. QL, YZ, and TY drafted the manuscript. YZ, TY, and RD performed data analysis. JL, JZ, TP, BZ, HO, and YJ collected the data. RD, JL, JZ, and TP wrote sections of the manuscript. QL, AY, and HF approved the final version of

the manuscript. All authors contributed to manuscript revision, read, and approved the submitted version. Presence of DQ2.2 Associated with DQ2.5 Increases the Risk for Celiac Disease Background. Celiac disease (CD) is a genetically determined immune-mediated disorder in which gluten immunogenic peptides are presented to CD4 T cells by HLA-DQ2.5, DQ8, DQ2.2, and their combinations. Our aim is to establish a risk gradient for celiac disease based on HLA-DQ profile in a brazilian representative population and the relevance of DQ2.2 in celiac disease development. Materials and Methods. 237 celiac patients and 237 controls (both groups with 164 females and 73 males) were included. All samples were tested for the presence of predisposing HLA-DQ alleles using the PCR-SSP method. Results were considered significant when p < 0.05. Disease risk was expressed as 1 : N for each HLA-DQ category described at this study. Results. DQ2.5 and/or DQ8 were detected in 224 celiac patients (94.5%) and 84 controls (35.4%). Eight celiac patients (3.4%) and 38 controls (16%) disclosed only DQ2.2. Even though DQ2.2 (β2/β2 or β2/x) showed a low CD risk of 1 : 251 and 1 : 550, respectively, the genotype DQ2.5/DQ2.2 (β2/β2) showed high CD risk of 1 : 10 (p < 0.0001). The disease risk gradient ranged from 1 : 3014 to 1 : 7. Conclusion. Our study allowed the determination of a risk gradient for celiac disease development in at-risk population, showing that DQ2.2 variant was relevant when associated with DQ2.5. Introduction Celiac disease (CD) is a genetically determined immunemediated disorder, in which individuals carrying specific HLA haplotypes (DQ2 and/or DQ8) mount an immunologic response to the ingestion of gluten that leads to a broad range of clinical signs and symptoms. Gastrointestinal disorders are the most common manifestations and include chronic diarrhea, abdominal distention, and nutrients malabsorption. However, extraintestinal manifestations are also frequent and include numerous conditions such as dermatitis herpetiformis, anemia, dental enamel hypoplasia, osteoporosis, and neurologic problems [1]. CD activity is characterized by the production of IgA antiendomysial antibody (IgA-EmA) and IgA anti-transglutaminase antibody (IgA-tTG), which are good markers of the active phase of the disease and are usually used as a first step in its diagnosis. In most cases, a definitive diagnosis requires a jejunal biopsy showing typical histologic abnormalities such as villous atrophy, crypt hyperplasia, and lymphocytic infiltration [2]. In the general population of Europe, United States, and countries predominantly populated by individuals of European origin, the prevalence of CD is approximately 1%. In Brazil, several prevalence studies performed to date revealed significant differences around the country, probably consequent to genetic and environmental factors and possibly also due to interlaboratory assay variability [3][4][5][6][7]. Although multicenter epidemiological studies that could yield reliable information on CD prevalence in Brazil are still lacking, the existing reports suggest that the disease prevalence in this country is similar to the general prevalence found in other areas of the world [2,4,5,7]. Virtually all CD patients carry the alleles that code for DQ2 and/or DQ8 molecules or at least for one chain of the DQ2 heterodimer, normally chain, encoded by DQB1*02 allele. The occurrence of CD in the absence of these atrisk DQ factors is extremely rare [8]. The presence of these molecules does not predict with accuracy that CD will develop, since they are present in 25 to 50% of the general population, although the vast majority of these individuals will never develop the disease [9]. Consequently, in view of the nearly 100% negative predictive value, the HLA typing has been used as a screening tool in high-risk population such as carriers of type 1 diabetes, Down syndrome, or Turner syndrome [10]. HLA typing has also been used as a prognostic factor of the disease severity [11,12] and sex distribution [13] and as an accessory element in the diagnosis of difficult cases [14]. Finally, the HLA-DQ typing to determine the future risk of CD has been extensively discussed, although its practical usage remains not clinically defined. Genetic testing of individuals could eliminate more than 60% of the population considered to have a low CDrisk (DQ2 or DQ8 negative) from future antibody testing, and the identification of high-risk individuals would allow a prospective screening, enabling an early therapeutic intervention [15]. As far as we know, no previous study has focused on the frequency of CD predisposing HLA genotypes in affected and nonaffected individuals in a Brazilian population. Consequently, our aim in the present study is to determine the frequency of CD predisposing DQ genotypes in celiac and nonceliac subjects and establish a CD-risk gradient focusing on the prevalence of DQ2.2, in Brazil. Celiac Patients and Ethics Committee. This retrospective study included celiac patients followed during the period of 2006 to 2014 at the Celiac Disease Outpatient Clinic of the Brasilia University Hospital. The diagnosis was achieved according to the criteria of the European Society of Pediatric Gastroenterology and Nutrition (ESPGAN) [16] and the revised guidelines of the European Society of Pediatric Gastroenterology, Hepatology and Nutrition (ESPGHAN) [17]. The study was approved by the School of Health Sciences Ethic Committee in Human Research (project number 070/06) and is in accordance with the latest Declaration of Helsinki. Written informed consent for participation was obtained from all participants prior to their enrollment in the study, including consent to publish. Patients and Controls . 237 serologically and biopsy confirmed celiac patients were included in the study (164 females and 73 males, age range: 1 to 75, mean age 21.5 ± 16.2 years at sample collection). Control group included 237 unrelated, gender and age matched, healthy individuals (164 women and 73 males, age range: 1 to 75 years, mean age 18.5 ± 16.14 years at sample collection), without history of autoimmune diseases and from the same geographic area and socioeconomic stratum as the celiac patients. All controls underwent serologic tests: IgA anti-transglutaminase antibodies (QUANTA Lite6 h-tTG IgA ELISA, INOVA Diagnostic, Inc., San Diego, CA, USA) and IgA anti-endomysium antibodies (NOVA Lite5 Monkey Oesophagus IFA Kit, INOVA Diagnostic, Inc., San Diego, CA, USA) and disclosed negative results. HLA Alleles Genotyping. Whole blood was obtained from celiac patients and controls in EDTA containing tubes following the H3-A6 criteria of Clinical and Laboratory Standards Institute (CLSI). DNA extraction was performed using commercial kit (Illustrat6 Blood Genomic Prep Mini Kit, GE Healthcare, Buckinghamshire, UK), and quantified with a NanoVue spectrophotometer (GE Healthcare, Buckinghamshire, UK). The final concentration of DNA was adjusted to 20 ng/ L. Statistical Analysis. Celiac patients and controls were categorized according to their HLA-DQ profile. Fisher's exact and 2 tests were applied to compare the frequencies of predisposing HLA-DQ alleles between celiac patients and controls. Results were considered significant when < 0.05. The risk assessment calculations was based on the study by Megiorni et al. [18], in which the degree of risk was represented by 1 : , where was the number of healthy individuals among which one patient is present. The analysis was performed considering a presumed prevalence of 1 : 100 in the general Brazilian population. For each HLA-DQ category, was calculated as a percentage of controls with a distinct HLA-DQ multiplied by 100 and divided by the percentage of patients with the same DQ typing. Considering that previous studies have shown that the dose-dependent effect of the allele DQB1*02 significantly increases the degree of risk, we classified as 2/x those individuals with only one copy of the DQB1*02 allele (heterozygosis) and as 2/ 2 those with two copies of the same allele (homozygosis). Individuals classified as 2/x may show either DR3/X, DR5/DR7, X/DR7, or DR7/DR4 haplotypes. Individuals pertaining to the group 2/ 2 may show haplotypes DR3/DR3 or DR3/DR7 and DR7/DR7. The HLA-DQ8 molecule is encoded by allelic variants DQA1*03 and DQB1*03:02 which are in linkage disequilibrium and consequently are commonly present in cisconformation in the DR4 haplotype. This haplotype can be found either in association with DQ2.5 (DR3/DR4), DQ2.2 (DR7/DR4) or with DQA1*05 ( 5/DR4). The 5 individuals carry the allele DQA1*05 and 8 individuals carry the allele DQA1*03, without any other CD predisposing allele. Individuals without any CD predisposing alleles were designed as absent. "X" refers to undetermined haplotypes and "x" to undetermined alleles. Celiac patients and controls, who did not disclose any of the risk variants (DQ2.5, DQ2.2, or DQ8), were explored for the presence of alleles that could confer low risk for the development of CD. The frequency and absence of these alleles, in celiac patients and controls, are shown in Table S1 (see Supple- The comparison between the frequency of CD HLA predisposing alleles in celiac patients and controls was used to establish a disease risk gradient that, in our sample, ranged from 1 : 7 to 1 : 3014 ( Figure 1). Figure 1: Risk gradient according to HLA haplotype combinations, considering a disease prevalence of 1 : 100. 2 refers to DQB1*02; 5 refers to DQA1*05; "x" denotes corresponding allele not determined; individuals with only one copy of the DQB1*02 allele were classified as 2/x (heterozygosis); individuals with two copies of the DQB1*02 allele were classified as 2/ 2 (homozygosis); trans and cis indicate, respectively, 5 and 2 position in different or in the same chromosome 6 pair. The association of DQ8 and 5 disclosed a significant decrease in disease risk of 1 : 1005 ( = 0.0108; OR 0.096 95% 0.0122-0.758), when compared to the presence of DQ8 alone. Finally, the exclusive presence of 5 showed a risk of 1 : 1594 ( < 0.0001; OR 0.050 95% CI 0.015-0.165), while the complete absence of any CD-HLA-DQ predisposing alleles was associated with a risk of 1 : 3014 ( < 0.0001; OR 0.024 95% CI 0.006-0.101). The risk for the presence of 8 was undetermined. Discussion The onset of CD is strongly associated with the presence of HLA-DQ2 and DQ8, which are considered to account for up to 40% of the genetic risk for the disease development. Other non-HLA genes are involved in CD susceptibility, although their genetic contribution to CD is weak and is not generally considered in the calculation of the disease risk [19,20]. Consequently, CD risk calculation, based on the presence of these HLA alleles, will allow a practical evaluation of the need for subsequent periodic tests on subjects included in the atrisk group. On the other hand, in view of the high negative predictive value of their absence, the need of further serologic testing is practically excluded in negative subjects. Data obtained in the present study regarding the distribution of haplotypes containing DQ2.5/DQ8/DQ2.2 and their corresponding alleles, in a representative sample of Brazilian celiac patients and of presumably healthy controls, were used to calculate the risk gradient for future development of celiac disease in our population. The Brazilian population has a high degree of genetic heterogeneity resulting from more than 500 years of interbreeding among three main ethnicities: Europeans, Amerindians, and Africans. In addition, during the last two centuries, successive migratory waves of Italians, Spaniards, Germans, Japanese, Lebanese, and Syrian further increased our population racial miscegenation. The analysis of ancestry informative markers of a representative sample of the Brazilian population disclosed a major contribution of Europeans (0.771%), followed by Africans (0.143%), and Amerindians (0.085%) [21]. The current population of Brasilia can be considered representative of the Brazilian population, since, during more than fifty years from its foundation, this city, presently with more than 2,500,000 inhabitants [22], has hosted people from all over the country. Despite the great ethnical diversity of the distinct macroregions of Brazil, the frequency of different haplotypes in Brazilian celiac patients is relatively stable. Previous studies performed in small series of celiac patients revealed that HLA-DQ2.5 and/or DQ8 were present in 93.1% of celiac patients in the Northeastern region of Brazil, where the miscegenation with Africans and Amerindians is more intense [23]. In the Southern region, where Caucasian ancestry predominates, the frequency of 91.1% was found [24]. In this study, conducted in the Midwestern region of the country, the frequency of these genotypes was 94.5%. It is of interest that these frequencies are very similar to those found in Europe, where over 90% of the celiac patients carry the DQ2 heterodimer, the remaining being mostly characterized by the presence of DQ8 [8]. Only a small number of celiacs carry neither the DQ2 nor the DQ8 genotypes, although several have been reported to have just one chain of the

DQ2 heterodimer [8]. Additionally, in our study, when considering the presence of the DQ2.5 variant without association with DQ8, the frequency found in celiac patients was 75.1%, while in previous studies performed in the Southern and Northeastern regions of Brazil was 65.3% [24] and 68.5%, respectively [23]. The presence of DQ8 alone (i.e., without the concomitant presence of DQ2.5) among our celiac patients was 10.5%. Comparatively, the frequency in the previously cited regions was, respectively, 11.9% and 17.8% [23,24]. Interestingly, when comparing celiac patients and controls for the presence of DQ8, alone or in combination with other genotypes or alleles (Table 1), the same frequency of 19.4% was detected in both groups. The only differences between groups were the DQ8 allelic combinations and their related risks. The association of DQ8 with DQB1*02 (DQ2.5/ DQ8 or DQ2.2/DQ8) is predominantly observed in celiac patients and in a minor degree in controls. The presence of this association increases 7 to 15 times the risk of developing the disease compared to DQ8 alone. These data are in agreement with the fact that the presence of DQB1*02 is associated with an increased risk of CD due to its higher antigenic repertoire, a fact which does not occur with an uncombined DQ8 [8,18,25,26]. The low frequency of DQ2.2 ( 2/x and 2/ 2) ( Table 1) found among celiac patients ( = 8, 3.4%) when compared to controls ( = 38, 16%) shows that although this variant is a predisposing factor for CD, it has a minor capacity to trigger an autoimmune process. This fact is in accordance with low binding stability of DQ2.2 heterodimer with gluten antigens decreasing the inflammatory response [27,28]. However, its genotyping is recommended in at-risk groups; since the genotype DQ2.2 is present in association with DQ2.5 and DQ8, it becomes a major risk factor significantly increasing the risk for CD onset [29]. In our study, the DQ2.5 heterodimer disclosed a risk of 1 : 20 (in trans) and 1 : 30 (in cis), showing statistically significant differences. However, when cisand trans-conformations Autoimmune Diseases 5 of DQ2.5 were considered together, the resulting risk was 1 : 30 (data not shown), similar to the risk found in the literature [18,30]. The combination DQ8/ 5 showed 3.5 times decreased risk of developing CD than that reported for isolated DQ8 (Figure 1). Similar relationship was found in Southern Spain [30]. Both patients and controls with this combination disclosed a DR5/DR4 haplotype configuration. It could be possible that the trans-conformations (DQA1*05/DQB1*03:02 and DQA1*03/DQB1*03:01) lead to the production of not very efficient heterodimers regarding to the presentation of some major immunogenic gluten peptides, since there is a lack of evidence for these events in the current study [26]. Consequently, we suggest that the association of DQA1*05/ DQB1*03:02 and DQA1*03/DQB1*03:01 reduces the binding of immunogenic epitopes of gluten and transglutaminase 2, therefore not triggering an effective immune response mediated by lymphocytes B and T. Individuals belonging to at-risk groups that only show the 5 allele or do not disclose any CD predisposing HLA alleles could be excluded from periodical follow-ups in view of their extremely low risk of developing CD [8,18] (Figure 1). Conclusion The present study showed, in a representative population of celiac patients and controls, the frequency of HLA-DQ variants DQ2.5/DQ2.2/DQ8, determining the probability of developing CD and establishing a risk gradient for the future onset of the disease. These data are useful for the screening of subjects included in at-risk groups for CD, discriminating subjects that are unlikely to develop the disease from those who must be monitored for a possible future development of CD. Additionally, this study demonstrated that although the HLA-DQ2.2 allele has a negligible risk factor when alone, its importance for the onset of CD will significantly increase when associated with DQ2.5. Metabolomics study of fibroblasts damaged by UVB and BaP We have recently shown that both UVB and BaP can induce the production of ROS, apoptosis and even cancer. However, the differences in the metabolic profiles of skin damaged by UVB, BaP or UVB combined with BaP have not been studied. Therefore, we examined the metabolic changes in the human foreskin fibroblast injured by UVB or BaP or the combination of the two, using ultra performance liquid chromatography (UPLC) coupled with quadrupole time-of-flight mass spectrometry (qTOF-MS). 24 metabolites were altered in the UVB damage group, 25 in the BaP damage group, and 33 in the UVB combined with BaP group. These alterations indicated that the metabolic mechanisms of HFF-1 cells treated with UVB or BaP are related to multiple main metabolites including glycerophosphocholine (PC), lactosylceramide (LacCer), guanidinosuccinic acid (GSA), glutathione(GSH), and lysophosphatidylcholine (LysoPC) and the main mechanisms involved glycerophospholipid and glutathione metabolism. Thus, our report provided useful insight into the underlying mechanisms of UVB and BaP damage to skin cells. UVB exposure and Cell viability assay. Prior to the experiment, Human foreskin fibroblast (HFF-1) cells were seeded at 50,000 cells/mL in 96-well plates incubated for 24 h. The cells were divided into two groups: (1) UVB blank control group: kept in a thin layer of PBS without UVB exposure; (2) UVB groups: cells were placed inside UVB in a thin layer of PBS and serially irradiated in doses of 5, 10, 15, 20 mJ/cm 2 . To evaluate the toxicity after acute UVB irradiation and determine the appropriate dose of UVB, cell viability was determined using the CCK-8 kit and measuring optical density at 450 nm. BaP exposure and ROS assay. HFF-1 cells were divided into the blank control group, negative control group (0.1% DMSO), positive control group and BaP damage group to select the concentration of BaP for further use. The cells were grown in 24-well plates for 24 h, followed by exposure to the predetermined doses of UVB. Then the BaP damage group was treated with culture media supplemented with 10 nM to 100 μM BaP, the negative control group was treated with 0.1% DMSO/culture media, and the other groups were only incubated with culture media. After 24 h BaP treatment, intracellular ROS generation of HFF-1 cells were measured using the oxidationsensitive dye 2′,7′-dichlorofluorescin diacetate (DCFH-DA). All groups were then incubated for 20 min with DCFH-DA. All cells were centrifuged and rinsed twice with phenol red-free DMEM (PRD) to clear the DCFH-DA. Cells were detected at an excitation wavelength of 485 nm and an emission wavelength of 535 nm using a microplate reader for ROS detection. Sample preparation. HFF-1 cells were divided into four groups: control group (C), UVB exposure group (U), BaP exposure group (B), UVB and BaP group (D). C and D were exposed to UVB in a thin layer of PBS which was then replaced with the fresh culture medium, and in groups B and D, 10 μM BaP was then added. Afterward, cells were incubated for another 24 h before the culture medium was discarded and cells were washed twice at 37 °C with PBS in order to reduce temperature shock to cells. Cells from each group were flash-frozen in liquid nitrogen and immediately stored in a − 80 °C freezer 5until further analysis. Prior to metabolomics analysis, taking out the cell samples in a − 80 °C freezer, and briefly thawed on ice ,then vortexed with 1 mL cold methanol/chloroform/H 2 O (8:1:1, v/v/v) for 1 min. After that scraped off the cells with a cell scraper, Cells of each sample were detached and collected into a centrifugation tube and sonicated at 4 °C for about 10 min. Cells lysates were then incubated for 30 min on ice and centrifuged at 4 °C for 10 www.nature.com/scientificreports/ 4 °C for UPLC-QTOF-MS analysis. The QC samples were inserted 6 samples at a time throughout the run to ensure the stability of the instrument. UHPLC-QTOF-MS analysis. All samples were analysed with an UPLC-QTOF-MS system. Briefly, full scan spectra between 50 and 1200 Da were acquired, and injected into an Acquity UPLC BEH C18 column (2.1 mm × 100 mm, 1.7 μm, Waters, USA). The injection volume was 5 μL and the column temperature was maintained at 50 °C. The phases were (A) water with 0.1% formic acid and (B) acetonitrile with 0.1 formic acid. The UPLC separations were 20 min/sample at a flow rate of 0.40 mL/min using the following scheme: 5% B, increased linearly to 38% B at 10 min, then to 100% B at 15 min, held for 3 min, and decreased linearly to 5% B at 19 min, and held for 1 min. The mass spectrum was used in the positive and negative electrospray ionisation mode. In both the positive and negative mode, capillary voltage was set at 3200 (+)/2500 (−) V and the sampling cone voltage was 35 V. Nitrogen was used as desolvation gas at a flow rate of 800 L/h at 400 °C. The ionisation source temperature was set at 120 °C. Data were collected in centroid mode. Samples were run in a random sequence. Data processing. Metabolomics data were processed using Markerlynx XS and EZ info software (Waters) combined with the Human Metabolome Database (HMDB). The (O)PLS-DA or PLS-DA methods were used to further analyse the data provided by the HMDB. Then, biomarker identification was conducted according to variable importance in projection values (VIP > 1), using an independent t-test (P < 0.05), accurate molecular ions were obtained by MS (mass error < 5 ppm), and fragment ions were obtained by MS/MS. The respective metabolic pathways of marker metabolites were mapped using Metaboanalyst 4.0 (https:// www. metab oanal yst. ca/). Statistical significance was calculated by Student's t-test or analysis of variance (ANOVA) using IBM SPSS Statistics 21.0.0.0. False discovery rate (FDR) correction was calculated to reduce the risk of a false-positive by the adjusted p values (< 0.05) based on Holm-Bonferroni method. For all analyses, ***P < 0.001; **P < 0.01; *P < 0.05. Results Cytotoxicity after UV irradiation. To determine the cytotoxicity of UVB irradiation, cells were irradiated by UVB ranging from 5 to 40 mJ/cm 2 after incubation for 24 h (Fig. 1a). The survival rates of HFF-1 cells were 93%, 81%, 70% and 65%, respectively. When the radiation dose of UVB reached 40 mJ/cm 2 , the cell viability showed obvious concentration gradient dependence (p < 0.01). It is generally believed that the 70% survival rate indicates that the injury effect can be determined, and the dose of 20 mJ/cm 2 is selected for further use. Intracellular ROS generation. BaP (10 NM-100 μ M) treatment had no effect on the viability of HFF-1 cells in the short term, but UVB and BaP could affect the ROS content of cells 3,4,11,14 , so we chose ROS as the modeling parameter. Followed irradiation by UVB with BaP (10 nM-100 μM) didn't affect the viability of HFF-1 cells (data not shown). UVB irradiation and BaP both lead to increased oxidation in the human skin cells.Therefore, the cytotoxicity of BaP against HFF-1 cells was determined by measuring the level of intracellular ROS generation. The results showed that ROS levels in HFF-1 cells showed dose-dependent on BaP and reached the highest value at a dose of 10 μM (Fig. 1b). Thus, 10 µM BaP was selected for further use. Multivariate statistical analysis. The preprocessed data were subjected to multivariate analysis for sample classification. PCA-class analysis was first performed to evaluate the similarity of the samples within each class (Fig. 1c). To further investigate changes in metabolites among the C, B, D and U groups, we applied the OPLS-DA method to identify differences between groups, as shown in the figure (Fig. 1d-f). The OPLS-DA was further validated using permutation test, the regression line of the Y-permuted Q2 was intercepted at − 0.172 (C versus B), − 0.379 (C versus U) and − 0.0284 (C versus D) respectively, indicating a reliable predictive power of the model. Subsequent processing with VIP analysis, VIP value larger than 1.0 were selected as putative biomarkers and are summarised in Fig. 2. After statistical analysis, Zero or missing values were replaced by 1/5 of the min positive value for each variable, and no data filtering was performed. C versus U and C versus B were found to have 18 altered metabolites in common, 12 of which also appeared in C versus D. Then we use the ANOVA and find that 11 substances have P ≤ 0.05, except